0% found this document useful (0 votes)
11 views149 pages

Measure Theory

The document is an introduction to measure and integration, covering fundamental concepts such as algebras, measures, and the construction of measures. It includes detailed chapters on the properties of measures, the integration of functions, product measures, and the relationship between integration and differentiation. The text serves as a comprehensive guide for understanding the mathematical foundations of measure theory and its applications.

Uploaded by

uxuehernani
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
11 views149 pages

Measure Theory

The document is an introduction to measure and integration, covering fundamental concepts such as algebras, measures, and the construction of measures. It includes detailed chapters on the properties of measures, the integration of functions, product measures, and the relationship between integration and differentiation. The text serves as a comprehensive guide for understanding the mathematical foundations of measure theory and its applications.

Uploaded by

uxuehernani
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 149

Measure and integration,

an introduction

Henk de Snoo (Rijksuniversiteit Groningen)


Henrik Winkler (TU Ilmenau)
Contents

Chapter 1. Algebras and measures 1


1. Finitely additive algebras and σ-algebras 1
2. Continuity properties of measures 4
3. Generators of σ-algebras 6
4. Dynkin systems 7
5. Completion of σ-algebras 10
Chapter 2. Construction of measures 15
1. σ-Algebras via outer measures 15
2. Lebesgue measure on Rd 17
3. Regularity properties of Lebesgue measurable sets 25
4. Lebesgue-Stieltjes measure on R 28
5. On the construction of measures 31
Chapter 3. Measurability of functions 35
1. Measurable mappings 35
2. Measurable functions 37
3. Approximation by simple functions 40
4. Properties which are valid almost everywhere 41
Chapter 4. Integrability of functions 45
1. Integrals of nonnegative measurable functions 45
2. Integrable functions 50
3. Integration with respect to Lebesgue measure 53
4. Convergence theorems 55
5. Parameter dependent integrals 61
6. Densities and transformations of measures 62
Chapter 5. Product measures 65
1. Construction of product measures 65
2. Fubini-Tonelli theorems 70
3. Product measures on Rp × Rq 73
4. Completion of product measures 76
5. Convolutions 79

Chapter 6. Integration and differentiation on R 83


1. Vitali covers 83
2. Differentiability of monotone functions 84
3. Termwise differentiation of series of monotone functions 89
4. Bounded variation and absolute continuity 93
5. Differentiation and integration 95
iii
iv CONTENTS

Chapter 7. Integration in Rd 101


1. Transformations in measure spaces 101
2. Non-linear transformations in Rd 103
3. Some coordinate systems 108
4. Lebesgue integration on manifolds in Rd 112

Chapter 8. Spaces of integrable functions 117


1. Semi-normed linear spaces 117
2. Minkowski and Hölder inequalities 119
3. Completeness 122
4. Dense linear subsets 124

Chapter 9. Decompositions of measures 127


1. Bounded linear functionals on Hilbert spaces 127
2. Majorization 128
3. Absolutely continuous measures 129
4. Lebesgue decomposition 132
5. Monotonicity, absolute continuity, and measures 134

Chapter 10. Some useful facts 137


1. Manipulations with sets 137
2. The extended real line 139
3. Open sets in Rd 140
4. Der große Umordnungssatz 141

Bibliography 145
CHAPTER 1

Algebras and measures

The notion of measure is a generalization of the notion of volume. A measure


on a subset A of a set Ω is a nonnegative number µ(A) (possibly ∞). Some sets
in Ω are too wild to have a measure in a consistent way. The subsets of Ω which
have a well-defined measure are the measurable sets. It will be required that the
measurable sets form a σ-algebra and that the measure is countably additive: the
measure of a countable union of disjoint measurable sets is the sum of the measures
of the individual sets. This chapter gives a brief introduction to measures on σ-
algebras.

1. Finitely additive algebras and σ-algebras


In order to introduce σ-algebras and measures it is helpful to understand the
situation for finitely additive algebras and measures.
Definition 1.1. A collection A of subsets of a set Ω is an algebra (or field) if
(1) Ω ∈ A;
(2) A ∈ A ⇒ Ac ∈ A;
(3) A, B ∈ A ⇒ A ∪ B ∈ A.
Note that in an algebra one has
(1) ∅ = Ωc ∈ A;
(2) A, B ∈ A ⇒ A ∩ B = (Ac ∪ B c )c ∈ A;
(3) A, B ∈ A ⇒ A \ B = A ∩ B c ∈ A;
(4) A, B ∈ A ⇒ A∆B = (A \ B) ∪ (B \ A) ∈ A.
Definition 1.2. A finitely additive measure (or content) µ on an algebra A is
an extended real-valued function µ : A → [0, ∞] which satisfies
(1) µ(∅) = 0;
(2) A, B ∈ A pairwise disjoint ⇒ µ(A ∪ B) = µ(A) + µ(B).
Lemma 1.3. Let µ be a finitely additive measure on an algebra A. Then
(1) µ(A ∪ B) + µ(A ∩ B) = µ(A) + µ(B);
(2) A ⊂ B ⇒ µ(B) = µ(A) + µ(B \ A);
(3) A ⊂ B, µ(A) < ∞ ⇒ µ(B) − µ(A) = µ(B \ A).
Proof. Note that the right-hand sides of the identities
A ∪ B = A ∪ (B \ A), B = (A ∩ B) ∪ (B \ A)
are disjoint unions. Hence by definition
(1.1) µ(A ∪ B) = µ(A) + µ(B \ A), µ(B) = µ(A ∩ B) + µ(B \ A).
1
2 1. ALGEBRAS AND MEASURES

Therefore one obtains


µ(A ∪ B) + µ(A ∩ B) = µ(A) + µ(B \ A) + µ(A ∩ B) = µ(A) + µ(B),
and item (1) follows. Furthermore, item (2) follows from the last of the identities
in (1.1). Item (3) is clear from (2). 
Let µ be a finitely additive measure on an algebra. It follows from Lemma 1.3
that µ is monotone
A ⊂ B ⇒ µ(A) ≤ µ(B),
and that µ is subadditive:
µ(A ∪ B) ≤ µ(A) + µ(B).
By induction, it is clear that for any n ∈ N
n
! n
[ X
µ Ai = µ(Ai ), A1 , . . . , An pairwise disjoint,
i=1 i=1
and that, in general, !
n
[ n
X
µ Ai ≤ µ(Ai ).
i=1 i=1
Algebras are also known as finitely additive algebras. The following simple obser-
vations concerning sets will be useful in the further development of algebras and
measures.
Lemma 1.4. Let An ⊂ Ω, n ∈ N, and define the subsets
(1.2) A01 = A1 , A0n = An \ (A1 ∪ · · · ∪ An−1 ), n ≥ 2.
Then the sets A0n are pairwise disjoint and
[n [n ∞
[ ∞
[
0 0
(1.3) Ak = Ak , An = An .
k=1 k=1 n=1 n=1

In particular, if the sequence An is monotonically nondecreasing then (1.2) reads


as
(1.4) A01 = A1 , A0n = An \ An−1 , n ≥ 2,
and (1.3) reads as
n
[ ∞
[ ∞
[
(1.5) A0k = An , A0n = An .
k=1 n=1 n=1

Definition 1.5. A collection A of subsets of a set Ω is a σ-algebra (or σ-field)


if
(1) Ω ∈ A;
(2) A ∈ A ⇒ Ac ∈ A; S

(3) An ∈ A, n ∈ N ⇒ n=1 An ∈ A.
A σ-algebra is a finitely additive algebra. Note that in a σ-algebra one has
∞ ∞
!c
\ [
An ∈ A, n ∈ N ⇒ An = Acn ∈ A.
n=1 n=1
A measurable space (Ω, A) is a set Ω provided with a σ-algebra A on Ω. The
elements of A are called measurable sets.
1. FINITELY ADDITIVE ALGEBRAS AND σ-ALGEBRAS 3

Example 1.6. Let Ω be a set. Then the collection A = {∅, Ω} is a σ-algebra,


and the collection A = P (Ω) (the power set of Ω) is a σ-algebra. These collections
are the smallest and largest σ-algebra that can be associated with Ω.
Definition 1.7. A measure µ on a σ-algebra A is an extended real-valued
function µ : A → [0, ∞] which satisfies
(1) µ(∅) = 0; S∞ P∞
(2) An ∈ A, n ∈ N, pairwise disjoint ⇒ µ ( n=1 An ) = n=1 µ(An ).
The last property in the definition is called the countable additivity of the
measure µ. Here the natural convention is that the sum of a divergent series of
nonnegative terms is ∞. In particular, a measure is finitely additive.
Proposition 1.8. A measure µ on a σ-algebra is countably subadditive:
∞ ∞
!
[ X
An ∈ A, n ∈ N ⇒ µ An ≤ µ(An ).
n=1 n=1
Proof. Introduce the pairwise disjoint sets A0n as in (1.2). Then A0n ∈ A,
0
µ(An ) ≤ µ(An ), and it follows from (1.3) that
∞ ∞ ∞ ∞
! !
[ [ X X
0
µ An = µ An = µ(A0n ) ≤ µ(An ),
n=1 n=1 n=1 n=1
which completes the proof. 
A set A ⊂ Ω is said to be a set of measure 0 if A ∈ A and µ(A) = 0.
Corollary 1.9. Let An ∈ A, n ∈ N, be such that µ(An ) = 0, then
∞ ∞
!
[ [
An ∈ A and µ An = 0,
n=1 n=1
i.e. the countable union of sets of measure 0 is a set of measure 0.
Definition 1.10. A measure space is a triple (Ω, A, µ) consisting of a set Ω,
a σ-algebra A on Ω, and a measure µ : A → [0, ∞]. A measure space is finite if the
measure is finite: µ(Ω) < ∞. A measure space is a probability space if the measure
is a probability measure: µ(Ω) = 1.
The notions in this definition form the heart and soul of measure, integration,
and probability theory.
Example 1.11 (Dirac measure). Let Ω be a set and let A be a σ-algebra of
subsets of Ω. For ω ∈ Ω define the extended real-valued function δω : A → [0, ∞]
by (
1, if ω ∈ A,
δω (A) =
0, if ω 6∈ A.
Then δω is a measure on A, the so-called Dirac measure. A set A ∈ A has measure
0 precisely when ω 6∈ A. The Dirac measure is finite as δω (Ω) = 1.
Example 1.12 (Counting measure). Let Ω be a set and let A = P (Ω). Let
A ∈ A, i.e. A ⊂ Ω. If A is a finite set, then define µ(A) as the number of elements
of A; if A is an infinite set, then define µ(A) = ∞. This definition impies that µ is
a measure on A, the so-called counting measure. The only set of measure 0 is the
set ∅. The measure µ is finite if and only if Ω is a finite set.
4 1. ALGEBRAS AND MEASURES

2. Continuity properties of measures


Measure spaces have an important continuity property with respect to mono-
tone sequences of measurable sets. For sets An , n ∈ N, and A in Ω the notations
An ↑ A or An ↓ A,
will denote that An is monotonically increasing or An is monotonically decreasing
in the sense that An ⊂ An+1 or An+1 ⊂ An for n ∈ N, and that

[ ∞
\
An = A or An = A,
n=1 n=1
respectively.
Theorem 1.13. Let (Ω, A, µ) be a measure space. Then:
(1) An ∈ A and An ↑ A ⇒ A ∈ A and µ(An ) ↑ µ(A);
(2) An ∈ A and An ↓ A, while µ(A1 ) < ∞ ⇒ A ∈ A and µ(An ) ↓ µ(A).
Proof. (1) Define A0n as in (1.4). Then A0n ∈ A and it follows from (1.5) that
∞ ∞ ∞
! !
[ [ X
0
µ(A) = µ An = µ An = µ(A0k )
n=1 n=1 k=1
n n
!
X [
= lim µ(A0k ) = lim µ A0k = lim µ(An ).
n→∞ n→∞ n→∞
k=1 k=1

(2) From An ↓ A it follows that (A1 \ An ) ↑ (A1 \ A), and hence by (1)
µ(A1 \ An ) → µ(A1 \ A).
Since An ⊂ A1 , A ⊂ A1 , and µ(A1 ) < ∞ this means
µ(A1 ) − µ(An ) → µ(A1 ) − µ(A),
(see Lemma 1.3), from which the claim follows. 
In particular, if (Ω, A, µ) is a finite measure space, then it follows from Theorem
1.13 that
(1.6) An ↓ ∅ ⇒ µ(An ) ↓ 0.
There is a partial converse.
Theorem 1.14. Let A be a σ-algebra and let µ be a finitely additive measure
on Ω such that
(1) µ(Ω) < ∞;
(2) µ satisfies (1.6).
Then µ is a measure.

S∞ Let (An ) be a sequence of pairwise disjoint


Proof. P∞ subsets of A and define A
by A = n=1 An . One has to show that µ(A) = n=1 µ(An ). In order to do this,
introduce !
[n
Bn = A \ Ak .
k=1
Then Bn ∈ A and the sequence (Bn ) is monotonically nonincreasing with Bn ↓ ∅.
According to the assumption, limn→∞ µ(Bn ) = 0.
2. CONTINUITY PROPERTIES OF MEASURES 5

Now observe that the definition of Bn implies


n
! n
!
[ [
A = Bn ∪ Ak with Bn ∩ Ak = ∅.
k=1 k=1
Since µ is finitely additive, this leads to the identity
Xn
µ(A) = µ(Bn ) + µ(Ak ).
k=1
P∞
For n → ∞, it follows that µ(A) = n=1 µ(An ). 
The following trivial observation is a direct consequence of the definitions. Let
(Ω, A, µ) be a measure space and let Ω0 ⊂ Ω belong to A . Then µ and Ω0 give rise
to the so-called induced measure µ0 on (Ω, A) by
µ0 (A) = µ(A ∩ Ω0 ), A ∈ A.
To see this, note that µ is well defined and that µ0 (∅) = 0. Furthermore, let An ∈ A
0

be disjoint sets, then also An ∩ Ω0 are disjoint sets in A, and


∞ ∞ ∞
! ! ! !
[ [ [
0 0 0
µ An = µ An ∩ Ω = µ (An ∩ Ω )
n=1 n=1 n=1

X ∞
X
= µ(An ∩ Ω0 ) = µ0 (An ).
n=1 n=1

Hence µ is a measure on (Ω, A). Clearly, µ is a finite measure on (Ω, A) when


0 0

µ(Ω0 ) < ∞.
Lemma 1.15. Let (Ω, A, µ) be a measure space. The following statements are
equivalent:
(1) there is a sequence Ωn ∈ A such that

[
(1.7) µ(Ωn ) < ∞, Ω= Ωn ;
n=1

(2) there is a mutually disjoint sequence Ωn ∈ A such that (1.7) holds;


(3) there is a sequence Ωn ∈ A such that
(1.8) µ(Ωn ) < ∞ and Ωn ↑ Ω.
Proof. (1) ⇒ (2) This follows from Lemma 1.4.
(2) ⇒ (3) By definition there exist Ω0n ∈ A which are mutually disjoint such
that µ(Ω0n ) < ∞ and

[
Ω= Ω0n
n=1
Define Ωn = Ω01 ∪ · · · ∪ Ω0n , so that µ(Ωn ) < ∞ and Ωn ↑ Ω.
(3) ⇒ (1) This is trivial. 
Definition 1.16. A measure space (Ω, A, µ) is σ-finite if any of the equivalent
conditions in Lemma 1.15 hold.
Example 1.17. Let Ω be a set and let µ be the counting measure. Then µ is
σ-finite if Ω is countable.
6 1. ALGEBRAS AND MEASURES

Proposition 1.18. Let (Ω, A, µ) be a σ-finite measure space and let the se-
quence Ωn ∈ A be such that (1.8) holds. Define µn : A → [0, ∞] by
µn (A) = µ(A ∩ Ωn ), A ∈ A.
Then µn defines a finite measure on Ω and µn (A) ↑ µ(A) for all A ∈ A.
Proof. It is clear by the text preceding Lemme 1.15 that µn is a measure
since Ωn ∈ A, and that this measure µn is finite since µ(Ωn ) < ∞. Now observe
(A ∩ Ωn ) ↑ A, which implies that
µn (A) = µ(A ∩ Ωn ) ↑ µ(A), A ∈ A,
due to Theorem 1.13. 

3. Generators of σ-algebras
There are various ways to construct σ-algebras on a set Ω. A typical situation
is: given a collection E of subsets of Ω find the smallest σ-algebra which contains
the collection E.
Lemma 1.19. The intersection of a nonempty family of σ-algebras on a set Ω
is a σ-algebra.
Proof. Let Aα with α ∈ I, some index set,Tbe a collection of σ-algebras.
T Since
Ω in all Aα , it also belongs to the intersection α∈I Aα . If A belongs to α∈I Aα ,
Ac belong to all Aα and hence Ac belongs to α∈I Aα . Now let
T
then A and also T
A , n ∈ N, be in α∈I Aα . Then for each α ∈ I all sets An belong to Aα and hence
Sn∞
n=1 An belongs to Aα and thus to α∈I Aα .
T

Proposition 1.20. Let E be a collection of subsets of a set Ω. Then there is
precisely one σ-algebra A such that
(1) E ⊂ A;
(2) if B is a σ-algebra with E ⊂ B, then A ⊂ B.
Proof. The family of σ-algebras which contain E is nonempty as the power
set P (Ω) is such a σ-algebra. By Lemma 1.19 the intersection of this family is a
σ-algebra and it contains E, hence (1) has been shown. Item (2) is trivial.
Let A0 be another σ-algebra with the properties (1) and (2). Then clearly
A ⊂ A0 and, likewise, A0 ⊂ A. Hence a σ-algebra with the properties (1) and (2)
is uniquely determined. 
Definition 1.21. Let E be a collection of subsets of a set Ω. The unique σ-
algebra in Proposition 1.20 is said to be generated by E, denoted by σ(E), and E is
said to be the generator of this σ-algebra.
Let E and F be collections of subsets of a set Ω. If E ⊂ F then it is clear that
σ(E) ⊂ σ(F). A σ-algebra A may be generated by different collections of subsets
of Ω.

The definition of the σ-algebra σ(E) generated by a collection E of subsets of


a set Ω is greatly non-explicit. Fortunately, in analysis it is rarely necessary to
know the set-theoretic description of σ(E): most sets that come up can be proved
to be measurable by expressing them as countable unions, intersections, and com-
plements of sets that are already known to be measurable.
4. DYNKIN SYSTEMS 7

By adding more structure to the set Ω one can obtain “natural” generators.
Definition 1.22. Let Ω be a topological space. The σ-algebra generated by all
open sets of Ω is the Borel σ-algebra B(Ω); its elements are called Borel measurable
subsets.
Note that countable intersections of open sets and countable unions of closed
sets are Borel sets. Such sets are sometimes called Gδ and Fσ sets, respectively.
In the following the only topological spaces that will be considered are the spaces
Ω = Rd with d ≥ 1. For d = 1 the σ-algebra B(R) will be denoted by B when no
confusion arises; likewise for d > 1 the σ-algebra B(Rd ) will be denoted by Bd .
Proposition 1.23. The Borel σ-algebra B on R is generated by
(1) the collection of closed subsets in R;
(2) the collection of intervals (−∞, b], b ∈ R;
(3) the collection of intervals (a, b], a, b ∈ R, a < b.
Moreover, the Borel σ-algebra Bd on Rd is generated by
(4) the collection of closed subsets;
(5) the collection of half-spaces {(x1 , . . . , xd ) : xi ≤ b} for some index i and
some b ∈ R;
(6) the collection of rectangles (a1 , b1 ] × · · · × (ad , bd ] where ai , bi ∈ R, ai < bi ,
and 1 ≤ i ≤ d.
Proof. Let B1 , B2 , and B3 be the σ-algebras, generated by the sets in (1),
(2), and (3), respectively. The Borel σ-algebra B includes the family of open sets
of R and, since it is closed under complementation, it also includes the family of
closed subsets of R. But then it also contains the σ-algebra B1 generated by the
closed subsets of R: B ⊃ B1 . The sets (−∞, b], b ∈ R, are closed and belong to B1 ;
hence B1 ⊃ B2 . Observe that for all a < b ∈ R:
(a, b] = (−∞, b] ∩ (−∞, a]c .
Since B2 is closed under complementation, each interval (a, b] belongs to B2 , and
therefore B2 ⊃ B3 . Each open subset of R belongs to B3 , due to its dyadic decom-
position; hence B3 ⊃ B. The first part of the proposition now follows from
B ⊃ B1 ⊃ B2 ⊃ B3 ⊃ B.
The second part is proved in the same way. 
In a similar way one can show the following useful observation.
Proposition 1.24. The Borel σ-algebra Bd is generated by
(1) the collection of compact subsets in Rd ;
(2) the collection of rectangles [a1 , b1 ] × · · · × [ad , bd ] where ai , bi ∈ R, ai < bi ,
and 1 ≤ i ≤ d.

4. Dynkin systems
The study of generators of a σ-algebra is facilitated by the notion of a Dynkin
system. A Dynkin system is weaker than a σ-algebra: it does not require closedness
under countable unions, but only closedness under countable unions of pairwise
disjoint sets.
Definition 1.25. A collection D of subsets of Ω is a Dynkin system if
8 1. ALGEBRAS AND MEASURES

(1) Ω ∈ D;
(2) A ∈ D ⇒ Ac ∈ D; S∞
(3) An ∈ D, n ∈ N, pairwise disjoint ⇒ n=1 An ∈ D.
It is obvious that a σ-algebra is a Dynkin system. However, not every Dynkin
system is a σ-algebra.
The following lemma and proposition can be proved in the same way as for the
corresponding results for σ-algebras.
Lemma 1.26. The intersection of a nonempty family of Dynkin systems is a
Dynkin system.
Proposition 1.27. Let E be a collection of subsets of Ω. Then there is precisely
one Dynkin system D such that
(1) E ⊂ D;
(2) if F is a Dynkin system with E ⊂ F, then D ⊂ F.
Definition 1.28. Let E be a collection of subsets of a set Ω. The unique Dynkin
system in Proposition 1.27 is said to be generated by E, denoted by d(E), and E is
said to be the generator of this Dynkin system.
Let E be a collection of subsets of Ω. The σ-algebra σ(E) contains E. Since
σ(E) is a Dynkin system, which contains E, it follows that
(1.9) d(E) ⊂ σ(E).
It is of great practical interest to know when equality holds in (1.9). The following
lemmas will lead to the main theorem concerning this question.
Lemma 1.29. The following statements are equivalent:
(1) D is a σ-algebra;
(2) D is a Dynkin system which is closed under intersections.
Proof. (1) ⇒ (2) This implication is clear.
(2) ⇒ (1) Let D be a Dynkin system which is closed under intersections. First
note that for A, B ∈ D one has
A ∪ B = (Ac ∩ B c )c ∈ D,
as Ac , B c ∈ D and Ac ∩ B c ∈ D by the extra assumption. Hence D is an algebra
and thus D is closed under finite unions. It will now be shown that D is closed
under countable unions. Let An ∈ D and introduce the pairwise disjoint sets A0n
as in (1.2). Then A0n = An \ (A1 ∪ · · · ∪ An−1 ) ∈ D, and

[ ∞
[
An = A0n ∈ D,
n=1 n=1
since D is a Dynkin system. Therefore D is a σ-algebra. 
Lemma 1.30. Let E be a collection of subsets of Ω. Then the collection
(1.10) DD = {A ∈ P (Ω) : A ∩ D ∈ d(E)} when D ∈ d(E),
is a Dynkin system. Moreover, if E is closed under intersections, then
(1.11) d(E) ⊂ DD when D ∈ d(E),
and, consequently, d(E) is closed under intersections.
4. DYNKIN SYSTEMS 9

Proof. First it will be shown that DD in (1.10) is a Dynkin system. It is clear


that Ω ∈ DD . Now let A ∈ DD , i.e., A ∩ D ∈ d(E). Note that
Ac ∩ D = ((A ∩ D) ∪ Dc )c and (A ∩ D) ∩ Dc = ∅.
Since A ∩ D and Dc are disjoint sets from the Dynkin system d(E), it follows that
Ac ∩ D ∈ d(E). Thus Ac ∈ DD . Finally, let AnS∈ DD be mutually disjoint. Note

that An ∩D ∈ d(E) and they are disjoint. Hence n=1 (An ∩D) ∈ d(E) and therefore

!
[
An ∩ D ∈ d(E),
n=1
S∞
so that n=1 An ∈ DD .
In order to prove the statement concerning (1.11), observe that if E ∈ E, then
E ∈ d(E) and hence the collection
DE = {A ∈ P (Ω) : A ∩ E ∈ d(E)}
is a Dynkin system. Now assume that E is closed under intersections. Then
(1.12) E ⊂ DE .
To see this let A ∈ E, then A ∩ E ∈ E ⊂ d(E), so that A ∈ DE . Since DE is a
Dynkin system it follows from (1.12) that
(1.13) d(E) ⊂ DE .
Let D ∈ d(E) and E ∈ E, then (1.13) implies
D ∩ E ∈ d(E).
Hence it follows that
E ⊂ DD when D ∈ d(E).
Since DD is a Dynkin system one sees that d(E) ⊂ DD . Hence (1.11) follows.
Finally it is shown that d(E) is closed under intersections, when E is closed
under intersections. For this purpose let A, D ∈ d(E). Since D ∈ d(E) one has
d(E) ⊂ DD by (1.11). Moreover for A ∈ d(E) one has A ∈ DD , in other words
A ∩ D ∈ d(E). Hence d(E) is closed under intersections. 
By combining Lemma 1.29 and Lemma 1.30 one obtains the main result: a
collection of sets which is closed under intersections generates a σ-algebra and a
Dynkin system which are equal.
Theorem 1.31. Let E be a collection of subsets of Ω. If E is closed under
intersections, then d(E) = σ(E).
Proof. The assumption that E is closed under intersections implies that d(E)
is closed under intersections by Lemma 1.30. By Lemma 1.29 this implies that
the Dynkin system d(E) is a σ-algebra and it contains E. Hence it follows that
σ(E) ⊂ d(E). Together with the inclusion (1.9) this shows that d(E) = σ(E). 
A result similar to Theorem 1.31 may be obtained when the notion of Dynkin
system is replaced by that of a monotone system, which can be found in many
textbooks.

The following corollaries are typical applications of the main result, Theorem
1.31. It suffices to prove a statement concerning the σ-algebra generated by a set
10 1. ALGEBRAS AND MEASURES

which is closed under intersections, by checking it for the corresponding Dynkin


system.
Corollary 1.32. Let (Ω, A) be a measurable space and let E be a generator of
A, which is closed under intersections. Let µ and ν be finite measures on A, such
that
(1) µ(A) = ν(A) for all A ∈ E;
(2) µ(Ω) = ν(Ω).
Then µ = ν.
Proof. Define the collection
M = {A ∈ A : µ(A) = ν(A)}.
Observe that Ω ∈ M and that for A ∈ M one has, since µ and ν are finite,
µ(Ac ) = µ(Ω) − µ(A) = ν(Ω) − ν(A) = ν(Ac ),
so that Ac ∈ M. Now let An ∈ M be disjoint, then, since µ and ν are measures,
∞ ∞ ∞ ∞
! !
[ X X [
µ An = µ(An ) = ν(An ) = ν An .
n=1 n=1 n=1 n=1
S∞
It follows that n=1 An ∈ M. Therefore M is a Dynkin system and it contains E,
so that d(E) ⊂ M. It follows from Theorem 1.31 that d(E) = σ(E) since E is closed
under intersections. By assumption σ(E) = A, and it follows that A ⊂ M. 

Corollary 1.33. Let (Ω, A) be a measurable space and let E be a generator


of A, which is closed under intersections. Let µ and ν be σ-finite measures on A,
such that
(1) µ(A) = ν(A) for all A ∈ E;
(2) µ(Ωn ) = ν(Ωn ) < ∞ for some sequence Ωn ∈ E such that Ωn ↑ Ω.
Then µ = ν.
Proof. Define the finite measures µn and νn by
µn (A) = µ(A ∩ Ωn ), νn (A) = ν(A ∩ Ωn ), A ∈ A,
cf. Proposition 1.18. Note that for A ∈ E, also A ∩Ωn ∈ E. Hence µn and νn satisfy
the conditions of Corollary 1.32. Therefore for all n ∈ N it follows that µn = νn ,
which leads to µ = ν by Proposition 1.18. 

5. Completion of σ-algebras
Let (Ω, A, µ) be a measure space and let A ∈ A have measure 0. Since the
measure µ is monotone it follows that µ(B) = 0 when B ⊂ A and B belongs to A.
However, it is not necessarily true that subsets of a set of measure 0 are measurable.
Definition 1.34. A measure space (Ω, A, µ) is complete if every subset of a
set of measure 0 is measurable (and, hence, is itself a set of measure 0).
Definition 1.35. Let (Ω, A, µ) be a measure space. Let Z be the class of all sets
N such that to each N there corresponds some F ∈ A with N ⊂ F and µ(F ) = 0.
5. COMPLETION OF σ-ALGEBRAS 11

Note that the class Z in Definition 1.35 is closed under countable unions. To
see this, let Nn ∈ Z. Then there exist Fn ∈ A with Nn ⊂ Fn and µ(Fn ) = 0. Hence
∞ ∞ ∞ ∞
!
[ [ [ [
Nn ⊂ Fn , Fn ∈ A, µ Fn = 0,
n=1 n=1 n=1 n=1
S∞
which shows that n=1 Nn ∈ Z.

Definition 1.36. Let (Ω, A, µ) be a measure space and let Z be the class in-
troduced in Definition 1.35. Let Ā be the class of all sets of the form E ∪ N with
E ∈ A and N ∈ Z. The completion µ̄ of µ is defined on Ā by

µ̄(E ∪ N ) = µ(E), E ∈ A, N ∈ Z.

A measure space (Ω, B, ν) is said to extend a measure space (Ω, A, µ) if A ⊂ B


and ν(E) = µ(E) for all E ∈ A. It will be verified that Ā is a σ-algebra, that µ̄ is
well-defined measure on A, and that (Ω, Ā, µ̄) extends (Ω, A, µ).

Theorem 1.37. Let (Ω, A, µ) be a measure space and let A and µ̄ be as in


Definition 1.36. Then the triple (Ω, Ā, µ̄) is a complete measure space extending
(Ω, A, µ). Moreover, this extension is minimal: if (Ω, A0 , µ0 ) is another complete
measure space which extends (Ω, A, µ), then it also extends (Ω, Ā, µ̄).

Proof. First it will be shown that Ā is a σ-algebra. It is clear that Ω ∈ Ā,


since Ω ∈ A. To show that Ā is closed under complements, consider E ∪ N with
E ∈ A and N ∈ Z. Then N ⊂ F for some F ∈ A and µ(F ) = 0, and thus

E ∪ N = E ∪ [F ∩ (F \ N )c ] = (E ∪ F ) ∩ (E ∪ (F \ N )c ),

so that
(E ∪ N )c = (E ∪ F )c ∪ ((F \ N ) ∩ E c ).
Now E ∪ F ∈ A implies that (E ∪ F )c ∈ A, while (F \ N ) ∩ E c ⊂ F . Hence
(E ∪ N )c ∈ Ā. Thus Ā is closed under complements. To see that Ā is closed under
countable unions, let En ∈ A and Nn ∈ Z. Then
∞ ∞ ∞
! !
[ [ [
(En ∪ Nn ) = En ∪ Nn ∈ Ā,
n=1 n=1 n=1

since Z is closed under countable unions.


Next it will be shown that (Ω, Ā, µ̄) is a measure space. In order to show that
µ̄ is well-defined it will be verified that E1 ∪ N1 = E2 ∪ N2 , with E1 , E2 ∈ A
and N1 , N2 ∈ Z, implies that µ(E1 ) = µ(E2 ) and which leads to µ̄(E1 ∪ N1 ) =
µ̄(E2 ∪ N2 ). It is clear that
E1 \ E2 ⊂ N2 ,
and N2 ⊂ F2 where F2 ∈ A is a set of measure 0, so that µ(E1 \ E2 ) = 0. It
follows from the identity E1 = (E1 ∩ E2 ) ∪ (E1 \ E2 ) that µ(E1 ) = µ(E1 ∩ E2 ). By
symmetry µ(E1 ) = µ(E2 ). It remains to demonstrate that µ̄ is a measure. It is
clear that µ̄(∅) = 0. Now let the sets An ∈ Ā be pairwise disjoint. Recall that

An = En ∪ Nn , En ∈ A, Nn ∈ Z, µ̄(An ) = µ(En ).
12 1. ALGEBRAS AND MEASURES

Hence it follows that


∞ ∞ ∞ ∞
! ! !
[ [ [ [
µ̄ An = µ̄ En ∪ Nn =µ En
n=1 n=1 n=1 n=1

X X∞ ∞
X
= µ(En ) = µ̄(En ∪ Nn ) = µ̄(An ).
n=1 n=1 n=1

Therefore µ̄ is a measure and (Ω, Ā, µ̄) is a measure space.


Now it will be shown that the measure space (Ω, Ā, µ̄) is complete. To see this
let A ∈ Ā be a set of measure 0, i.e., µ̄(A) = 0, and let M ⊂ A. By definition
A = E ∪ N with E ∈ A, µ(E) = 0, and N ∈ Z. There exists F ∈ A with N ⊂ F
and µ(F ) = 0. Hence M ⊂ A = E ∪ N ⊂ E ∪ F with E ∪ F ∈ A and µ(E ∪ F ) = 0.
Therefore M ∈ Ā. This shows that (Ω, Ā, µ̄) is complete.
Finally it will be shown that (Ω, Ā, µ̄) is a minimal extension of (Ω, A, µ).
First of all, it is clear that A ⊂ Ā, since ∅ ∈ Z. In order to verify that the
extension (Ω, Ā, µ̄) is minimal, let (Ω, A0 , µ0 ) be another complete measure space
which extends (Ω, A, µ). It will be shown that (Ω, A0 , µ0 ) extends (Ω, Ā, µ̄). To see
this, let E ∪ N with E ∈ A and N ∈ Z be an arbitrary element of Ā. Then N ⊂ F
for some F ∈ A with µ(F ) = 0. Since (Ω, A0 , µ0 ) extends (Ω, A, µ) it follows that
E ∈ A0 , F ∈ A0 , µ0 (F ) = 0.
However (Ω, A0 , µ0 ) is complete, so that it follows that N ∈ A0 and µ0 (N ) = 0.
Observe that E ∪ N ∈ A0 . Furthermore, the following inequalities
µ0 (E) ≤ µ0 (E ∪ N ) ≤ µ0 (E) + µ0 (N ) = µ0 (E),
lead to
µ0 (E ∪ N ) = µ0 (E) = µ(E) = µ̄(E ∪ N ).
Thus (Ω, A0 , µ0 ) extends (Ω, Ā, µ̄). 

Corollary 1.38. Let (Ω, A, µ) be a measure space and assume that it is σ-


finite. Then the completion (Ω, Ā, µ̄) is also σ-finite.
Sometimes it is useful to know that a measurable space (Ω, A) gives rise to a
measurable space on a subset Ω0 ⊂ Ω. If (Ω, A, µ) be a (complete) measure space
then Ω0 even inherits a (complete) measure from µ.
Proposition 1.39. Let (Ω, A, µ) be a measure space and let Ω0 ⊂ Ω. Then the
so-called trace, defined by
A0 = { A ∩ Ω0 : A ∈ A },
is a σ-algebra on Ω0 . Moreover, if Ω0 ∈ A, then
µ0 (B) = µ(B), B ∈ A0 ,
defines a measure on Ω0 . If the measure space (Ω, A, µ) is complete, then the mea-
sure space (Ω0 , A0 , µ0 ) is complete.
Proof. It is clear that Ω0 ∈ A0 , as Ω0 = Ω ∩ Ω0 . If A ∈ A, then
Ω0 \ (A ∩ Ω0 ) = Ac ∩ Ω0 ∈ A0 .
5. COMPLETION OF σ-ALGEBRAS 13

S∞
Now let An ∈ A, then n=1 An ∈ A, so that
∞ ∞
!
[ [
0
(An ∩ Ω ) = An ∩ Ω0 ∈ A 0 .
n=1 n=1
Hence A is a σ-algebra on Ω .
0 0

If Ω0 ∈ A, then A ∩ Ω0 ∈ A for any A ∈ A, and it follows that A0 ⊂ A. Hence,


µ is a well defined measure on the σ-algebra A0 .
0

To show that (Ω0 , A0 , µ0 ) is complete, let N ∈ A0 with µ0 (N ) = 0. Let E ⊂ N .


Then E ∈ A since (Ω, A, µ) is complete and N ∈ A, and since E = E ∩ Ω0 it follows
that E ∈ A0 . 
CHAPTER 2

Construction of measures

How to provide an arbitrary set Ω with measurable sets and a measure? One
way to do this is by means of the notion of outer measure. Outer measures are
defined on all subsets of Ω and they are required to be monotone and countably
subadditive. In this context a subset A ⊂ Ω is defined to be measurable if for
all subsets Z ⊂ Ω the outer measure is additive on the disjoint union of Z ∩ A
and Z ∩ Ac . The restriction of the outer measure to the measurable sets gives a
measure space which is complete (subsets of sets of measure 0 are measurable).
In this chapter measurable sets and measures are constructed via outer measures.
In particular the Lebesgue measure and the Lebesgue-Stieltjes measure will be
introduced.

1. σ-Algebras via outer measures


An important construction of measures makes use of the notion of outer mea-
sure. It will be sketched how outer measures give rise to complete σ-algebras.
Definition 2.1. Let Ω be a set and let P (Ω) be its power set. An extended
real-valued function µ∗ : P (Ω) → [0, ∞] is an outer measure on Ω if
(1) µ∗ (∅) = 0;
(2) A ⊂ B ⇒ µ∗ (A) ≤ µ∗ (B);
S∞ P∞
(3) An ⊂ Ω, n ∈ N ⇒ µ∗ ( n=1 An ) ≤ n=1 µ∗ (An ).
Definition 2.2. Let µ∗ be an outer measure on Ω. A set A ⊂ Ω is measurable
with respect to µ∗ if for any set Z ⊂ Ω
(2.1) µ∗ (Z) = µ∗ (Z ∩ A) + µ∗ (Z ∩ Ac ).
Lemma 2.3. Let µ∗ be an outer measure on Ω. A set A ⊂ Ω is measurable if
and only if for any set Z ⊂ Ω with µ∗ (Z) < ∞
(2.2) µ∗ (Z) ≥ µ∗ (Z ∩ A) + µ∗ (Z ∩ Ac ).
Proof. Note that Z = (Z ∩ A) ∪ (Z ∩ Ac ), so that automatically
µ∗ (Z) ≤ µ∗ (Z ∩ A) + µ∗ (Z ∩ Ac )
by the definition of outer measure. Hence A ⊂ Ω is measurable if and only if the
reverse inequality holds. Clearly, it suffices to check the reverse inequality only for
those sets Z which have finite outer measure. 
Lemma 2.4. Let A ⊂ Ω satisfy µ∗ (A) = 0. Then A is measurable with respect
to µ∗ .
Proof. Let µ∗ (A) = 0. For any Z ⊂ Ω one has (2.2), since the monotonicity
of µ implies that µ∗ (Z ∩ A) ≤ µ∗ (A) = 0 and µ∗ (Z ∩ Ac ) ≤ µ∗ (Z). Hence A is

measurable. 
15
16 2. CONSTRUCTION OF MEASURES

Theorem 2.5. Let µ∗ be an outer measure on Ω. The set A∗ of all µ∗ -


measurable sets A ⊂ Ω is a σ-algebra on Ω. Moreover, the restriction of µ∗ to
A∗ is a measure, which is complete.
Proof. It is clear from Definition 2.1 that Ω ∈ A∗ and that A∗ is closed under
complements. Next it will be shown that A∗ is an algebra (hence closed under
intersections) and that A∗ is a Dynkin system. Then by Theorem 1.29 the Dynkin
system A∗ , being closed under intersections, is a σ-algebra.
In order to show that A∗ is an algebra, let A, B ∈ A∗ , so that both (2.2) and
(2.3) µ∗ (Z) = µ∗ (Z ∩ B) + µ∗ (Z ∩ B c )
are valid for any Z ⊂ Ω. Replace in (2.3) Z by Z ∩ A and Z ∩ Ac , which gives
µ∗ (Z ∩ A) = µ∗ (Z ∩ A ∩ B) + µ∗ (Z ∩ A ∩ B c )
and
µ∗ (Z ∩ Ac ) = µ∗ (Z ∩ Ac ∩ B) + µ∗ (Z ∩ Ac ∩ B c ).
Together with (2.2) the last two identities give
(2.4) µ∗ (Z) = µ∗ (Z ∩ A ∩ B) + µ∗ (Z ∩ A ∩ B c ) + µ∗ (Z ∩ Ac ∩ B) + µ∗ (Z ∩ Ac ∩ B c ).
Replace in (2.4) Z by Z ∩ (A ∪ B), so that
µ∗ (Z ∩ (A ∪ B)) = µ∗ (Z ∩ (A ∪ B) ∩ A ∩ B) + µ∗ (Z ∩ (A ∪ B) ∩ A ∩ B c )
+ µ∗ (Z ∩ (A ∪ B) ∩ Ac ∩ B) + µ∗ (Z ∩ (A ∪ B) ∩ Ac ∩ B c ),
or, equivalently,
(2.5) µ∗ (Z ∩ (A ∪ B)) = µ∗ (Z ∩ A ∩ B) + µ∗ (Z ∩ A ∩ B c ) + µ∗ (Z ∩ Ac ∩ B).
Combine (2.4) and (2.5) to obtain
µ∗ (Z) = µ∗ (Z ∩ (A ∪ B)) + µ∗ (Z ∩ (A ∪ B)c ).
This implies that with A, B ∈ A∗ also A ∪ B ∈ A∗ . Hence A∗ is an algebra, so that
A∗ is closed under intersections. Note that it also follows from (2.5) that
(2.6) A∩B =∅ ⇒ µ∗ (Z ∩ (A ∪ B)) = µ∗ (Z ∩ A) + µ∗ (Z ∩ B).
In order to show that A∗ is a Dynkin system, let An ∈ A∗ , n ∈ N, be pairwise
disjoint. Define the sets

[ n
[
(2.7) A= Ai and Bn = Ai .
i=1 i=1

It will be shown that A ∈ A∗ . Since A∗ is an algebra one has Bn ∈ A∗ . Hence for


any Z ⊂ Ω
(2.8) µ∗ (Z) = µ∗ (Z ∩ Bn ) + µ∗ (Z ∩ Bnc ),
and, in addition, it follows from (2.6) that
n
X
(2.9) µ∗ (Z ∩ Bn ) = µ∗ (Z ∩ Ai ).
i=1

Clearly, Bn ⊂ A, which implies that Z ∩ Ac ⊂ Z ∩ Bnc ; and, hence


(2.10) µ∗ (Z ∩ Ac ) ≤ µ∗ (Z ∩ Bnc ).
2. LEBESGUE MEASURE ON Rd 17

Combine (2.8), (2.9), and (2.10), to obtain


n
X

µ (Z) ≥ µ∗ (Z ∩ Ai ) + µ∗ (Z ∩ Ac ),
i=1

valid for any n ∈ N. Hence, it follows that



X

(2.11) µ (Z) ≥ µ∗ (Z ∩ Ai ) + µ∗ (Z ∩ Ac ) ≥ µ∗ (Z ∩ A) + µ∗ (Z ∩ Ac ),
i=1
S∞
where the last inequality is based on Z ∩ A = n=1 (Z ∩ An ) and the countable
subadditivity of µ∗ . Due to µ∗ (Z ∩ A) + µ∗ (Z ∩ Ac ) ≥ µ∗ (Z) it follows from (2.11)
that

X
(2.12) µ∗ (Z) = µ∗ (Z ∩ Ai ) + µ∗ (Z ∩ Ac ) = µ∗ (Z ∩ A) + µ∗ (Z ∩ Ac ).
i=1

Clearly (2.12) shows that A is measurable with respect to µ∗ . Hence A∗ is a Dynkin


system. Since A∗ is closed under intersections it is a σ-algebra.
Note that the choice of Z = A in (2.12) with A as in (2.7) leads to the identity

X
µ∗ (A) = µ∗ (Ai ),
i=1

for any set of pairwise disjoint Ai ∈ A∗ , i ∈ N. Hence µ∗ restricted to the σ-algebra


A∗ is a measure.
In order to show that µ∗ on A∗ is complete, let A be a set in A∗ with measure 0
and let N ⊂ A. Since µ∗ (A) = 0, the monotonicity property of µ∗ shows µ∗ (N ) = 0,
so that by Lemma 2.4 also N is measurable. Hence A∗ is complete. 

2. Lebesgue measure on Rd
The notion of an outer measure for Rd is introduced via the volume of a (closed)
d-dimensional rectangle R of the form
R = [a1 , b1 ] × · · · × [ad , bd ], ai , bi ∈ R, ai ≤ bi , i = 1, . . . , d,
whose volume is defined by `(R) = (b1 − a1 ) · · · (bd − ad ).
Definition 2.6. Let A ⊂ Rd . Then m∗ (A) ∈ [0, ∞] is defined by
( ∞ ∞
)
X [
∗ d
(2.13) m (A) = inf `(Rn ) : Rn ⊂ R closed rectangle, A ⊂ Rn .
n=1 n=1

P∞ Rn , n ∈ N, form a cover of A and m (A) is the infimum of the
The rectangles
total volumes n=1 `(Rn ) of all possible covers Rn .
Theorem 2.7. The extended real-valued function m∗ : P (Rd ) → [0, ∞] in
(2.13) is an outer measure on Rd , called the Lebesgue outer measure.
Proof. It will be shown that Definition 2.2 applies. It is clear that m∗ (∅) = 0.
Moreover, if A ⊂ B ⊂ R, then a covering of B is also a covering of A. Hence it
follows that
A ⊂ B ⇒ m∗ (A) ≤ m∗ (B).
18 2. CONSTRUCTION OF MEASURES

Next it is shown that m∗ is countably subadditive, i.e.,


∞ ∞
!
[ X

m An ≤ m∗ (An ),
n=1 n=1
P∞
for any sequence An ⊂ Ω. It suffices to show this for n=1 m∗ (An ) < ∞. Assume
that this sum is finite and let ε > 0. Then for each n there exists a sequence of
rectangles Rkn , k ∈ N, such that
∞ ∞
[ X ε
An ⊂ Rkn , `(Rkn ) < m∗ (An ) + .
2n
k=1 k=1

Hence it follows that



[ ∞ [
[ ∞
An ⊂ Rkn ,
n=1 n=1 k=1
1
and observe that
∞ X∞ ∞  ∞
X X ε  X ∗

`(Rkn )
≤ m (An ) + n = m (An ) + ε,
n=1 k=1 n=1
2 n=1
S∞ P∞
which leads to m∗ ( n=1 An ) ≤ n=1 m∗ (An ) + ε. The result follows since ε > 0
is arbitrary. 

Next it is verified that the definition of Lebesgue outer measure is reasonable,


i.e., that the Lebesgue outer measure of a rectangle coincides with its volume.
Lemma 2.8. Let the Lebesgue outer measure m∗ be as in (2.13) and let R ⊂ Rd
be a closed rectangle with volume `(R). Then
m∗ (R) = `(R).
Proof. It is clear that m∗ (R) ≤ `(R), as {R} covers R.
For the converse
S∞ inequality choose ε > 0. Let Rn , n ∈ N, be rectangles which
cover R: R ⊂ n=1 Rn . By enlarging Rn one obtains a rectangle Sn whose interior
Sn0 contains Rn and
ε
`(Sn ) ≤ `(Rn ) + n .
2
Then the interiors Sn0 form an open cover of the compact set R. Hence there exists
N ∈ N so that {S10 , . . . , SN
0
} and, hence, also {S1 , . . . , SN } cover the rectangle R.
Note that
N N h ∞
X X ε i X
`(R) ≤ `(Sn ) ≤ `(Rn ) + n ≤ `(Rn ) + ε.
n=1 n=1
2 n=1

Since ε > 0 is arbitrary, one concludes



X
`(R) ≤ `(Rn ),
n=1

and it follows that `(R) ≤ m∗ (R). 

1Here the so-called ”grosse Umordnung” is used: give an enumeration of the cover and take
the sum via this enumeration.
2. LEBESGUE MEASURE ON Rd 19

Definition 2.9. The Lebesgue measurable sets of Rd are the measurable sets
defined by the Lebesgue outer measure m∗ in (2.13), and Lebesgue measure is the
restriction of m∗ to the Lebesgue measurable sets. The corresponding complete
measure space is denoted by (Rd , Md , m).
Let a ∈ Rd , then m∗ ({a}) = 0. Hence the subset {a} ⊂ Rd is Lebesgue
measurable with Lebegue measure 0. As a consequence any subset of the form
{an ∈ Rd : n ∈ N} is Lebesgue measurable with Lebesgue measure 0.
Let A = {x ∈ Rd : xi = 0} for a fixed 1 ≤ i ≤ d; i.e., A ⊂ Rd is a d − 1
dimensional hyperspace. Then A is Lebesgue measurable and m(A) = 0. To show
this, it suffices to treat the subspace A = Rd−1 × {0}. For any ε > 0 one has
∞  h ε
[ ε i
A⊂ [−n, n]d−1 × − · (2n)−(d−1) · 2−n , · (2n)−(d−1) · 2−n ,
n=1
2 2

while the corresponding sum of volumes satisfies


∞ h ε
X  ε i
` [−n, n]d−1 × − · (2n)−(d−1) · 2−n , · (2n)−(d−1) · 2−n = ε.
n=1
2 2

Hence m∗ (A) = 0, so that A is Lebesgue measurable and m(A) = 0.


It can be shown that there exist subsets of Rd which are not Lebesgue measur-
able. However most useful subsets of Rd are Lebesgue measurable.
Proposition 2.10. Every closed rectangle in Rd is Lebesgue measurable.
Proof. Let R be a closed rectangle in Rd and let Z ⊂ Rd be an arbitrary
subset with m∗ (Z) < ∞. For the Lebesgue measurability of R it suffices to show
that
(2.14) m∗ (Z ∩ R) + m∗ (Z ∩ Rc ) ≤ m∗ (Z).
For this purpose, let ε > 0 and choose a cover of closed rectangles Rn , n ∈ N,
of the subset Z with
X∞
m∗ (Z) ≤ `(Rn ) < m∗ (Z) + ε.
n=1

Decompose each Rn with respect to the rectangle R into an almost disjoint finite
union of rectangles (i.e., these rectangles have disjoint interiors):
 
N
[n
Rn = Ren ∪  Tn,j  , Ren = Rn ∩ R ⊂ R, Tn,j ⊂ Rc .
j=1

Then it follows that


Nn
X
`(Rn ) = `(R
en ) + `(Tn,j ).
j=1

In other words one has


 

X ∞
X XNn
m∗ (Z) ≤ `(R
en ) +  `(Tn,j ) < m∗ (Z) + ε.
n=1 n=1 j=1
20 2. CONSTRUCTION OF MEASURES

en cover Z ∩ R and the rectangles Tnj cover Z ∩ Rc so that


The rectangles R
 
X∞ ∞
X Nn
X
m∗ (Z ∩ R) ≤ en ), m∗ (Z ∩ Rc ) ≤
`(R  `(Tn,j ) .
n=1 n=1 j=1

Therefore
 

X ∞
X XNn
m∗ (Z ∩ R) + m∗ (Z ∩ Rc ) ≤ `(R
en ) +  `(Tn,j ) < m∗ (Z) + ε.
n=1 n=1 j=1

Since ε > 0 is arbitrary, this implies (2.14). 


It has been shown that the closed rectangle R = [a1 , b1 ] × · · · × [ad , bd ] is
Lebesgue measurable and thus it follows from Lemma 2.8 that m(R) = `(R). The
open rectangle R0 = (a1 , b1 ) × · · · × (ad , bd ) is also Lebesgue measurable as it can
be approximated by the increasing sequence
   
1 1 1 1
a1 − , b1 − × · · · × ad − , bd − ,
n n n n
and it follows from Theorem 1.13 that m(R0 ) = `(R). Note that ∂R = R \ R0
implies that ∂R is Lebesgue measurable and that
m(∂R) = m(R) − m(R0 ) = 0,
cf. Lemma 1.3. In particular, the faces of a rectangle are Lebesgue measurable
sets with measure 0. Similar statements as above hold for rectangles of the form
[a1 , b1 ) × · · · × [ad , bd ). In particular dyadic rectangles are measurable.
Corollary 2.11. Every open set in Rd is Lebesgue measurable.
Proof. Every open set in Rd can be expressed as the union of a countable
number of dyadic rectangles. 
Hence it is clear that Bd , the σ-algebra generated by the open sets in Rd , is
contained in the σ-algebra Md ; cf. Definition 1.22. It can be shown that there exist
Lebesgue measurable sets which are not Borel measurable.
Corollary 2.12. The Lebesgue measure on Rd is σ-finite.
Proof. Define An = (−n, n)d , so that it is clear that
S∞An is measurable. Fur-
thermore it is clear that m(An ) = (2n)d < ∞ and Rd = n=1 An . 
Example 2.13 (Cantor set). Let C0 = [0, 1] and remove the open middle third
and define    
1 2
C1 = 0, ∪ ,1 .
3 3
Again remove the open middle thirds of the intervals of C1 and define
       
1 2 1 2 7 8
C2 = 0, ∪ , ∪ , ∪ ,1 .
9 9 3 3 9 9
Continuation of this process of deleting the open middle thirds of the intervals leads
to a nested sequence Cn such that every Cn is the union of 2n intervals, each of
length 1/3n . Define the Cantor set to be
\∞
C= Cn .
n=1
2. LEBESGUE MEASURE ON Rd 21

Since each Cn is closed, it follows that C is closed. There is a one-to-one correspon-


dence between C and the interval [0, 1] (via ternary expansions for instance), so that
C is not countable. By construction each Cn , being a finite union of 2n disjoint
closed intervals of length (1/3)n , is Lebesgue measurable and its Lebesgue measure
equals (2/3)n . Hence C is Lebesgue measurable and, according to Theorem 1.13,
the set C has measure 0.
Recall that the outer measure in Rd is defined in terms of volumes of rectangles,
whose sides are parallel to the coordinate axis of Rd . So it comes as no surprise
that the Lebesgue outer measure is invariant under translations.
Proposition 2.14. Let A ⊂ Rd and let x ∈ Rd . Then m∗ is invariant under
translation:
(2.15) m∗ (A + x) = m∗ (A).
Hence A + x is Lebesgue measurable if and only if A is Lebesgue measurable and
(2.16) m(A + x) = m(A).
Proof. Let R1 , R2 , . . . be a cover of closed rectangles of A + x, then
−x + R1 , −x + R2 , · · ·
is a cover of closed rectangles of A with `(Rn ) = `(−x + Rn ) for all n. Hence it
follows that
X∞ ∞
X
`(Rn ) = `(−x + Rn ) ≥ m∗ (A).
n=1 n=1
Taking the infimum on the left-hand side over all covers of closed rectangles of A+x
leads to
m∗ (A + x) ≥ m∗ (A).
Since this inequality holds for arbitrary x ∈ Rd , one also has
m∗ (A) = m∗ ((A + x) − x) ≥ m∗ (A + x),
proving (2.15).
Now assume that A is Lebesgue measurable and let x ∈ Rd . For any Z ⊂ Rd
one obtains
m∗ (Z ∩ (A + x)) + m∗ (Z ∩ (A + x)c ) = m∗ (Z ∩ (A + x)) + m∗ (Z ∩ (Ac + x))
= m∗ ((Z − x) ∩ A) + m∗ ((Z − x) ∩ Ac )
= m∗ (Z − x) = m∗ (Z),
where the second and fourth equalities follow from (2.15) and the third equality
follows from the measurability of A. This shows that A + x is Lebesgue measurable,
and m(A + x) = m(A) is then an immediate consequence of (2.15). 
The space Rd may be considered as an inner product space with a basis parallel
to the natural coordinate axes. An orthogonal transformation in Rd leads to a new
orthogonal basis with rectangles which are oblique in the original coordinate system.
What happens when a set in Rd is measured in terms of these new rectangles?
Lemma 2.15. Let Q be an orthogonal transformation in Rd and let A ⊂ Rd .
Then m∗ is invariant under orthogonal transformations:
(2.17) m∗ (Q(A)) = m∗ (A).
22 2. CONSTRUCTION OF MEASURES

Hence Q(A) is Lebesgue measurable if and only if A is Lebesgue measurable, and


in this case
(2.18) m(Q(A)) = m(A).
Proof. The lemma will be first proved for rectangles of the following form
R0 = [0, a1 ) × · · · × [0, ad ), a1 , . . . , ad > 0.
It is clear that R0 is Lebesgue measurable, see discussion following Proposition
2.10. Note that the image Q(R0 ) may be written as
∞        
[ 1 1
Q(R0 ) = Q 0, 1 − a1 × · · · × 0, 1 − ad .
n=2
n n
The sets in the above union are compact (since Q is continuous), hence closed and
thus measurable. Therefore Q(R0 ) is Lebesgue measurable. It will now be shown
that the finite numbers m(Q(R0 )) and `(R0 ) are equal.
For each n ∈ N partition the rectangle R0 into 2dn rectangles of the same form
(by partitioning each [0, ai ) into 2n equidistant intervals). By translation of R0 the
whole space Rd can be seen as the disjoint union Λn of such rectangles, all with the
same area. Since R0 is the union of 2dn rectangles R in Λn , each with the same
area, it follows for each rectangle R ∈ Λn that
`(R) = 2−dn `(R0 ).
Let B the unit ball in Rd and define for each n ∈ N
[
Cn = {R ∈ Λn : R ∩ B 6= ∅}, Cn = R.
R∈Cn

Then it is clear that C1 ⊃ C2 ⊃ · · · ⊃ B; cf. Theorem 1.13. Since B is closed, it is


Lebesgue measurable, so m(B) is well-defined and finite. Moreover one sees that
X
(2.19) m(B) = lim m(Cn ) = lim `(R) = lim 2−dn · `(R0 ) · #Cn ,
n→∞ n→∞ n→∞
R∈Cn

where #Cn denotes the number of elements of Cn . Since Q is orthogonal one has
that Q(B) = B. In addition, Q(C1 ) ⊃ Q(C2 ) ⊃ · · · ⊃ Q(B) = B, cf. Theorem 1.13,
so that
X
(2.20) m(B) = lim m(Q(Cn )) = lim m(Q(R)).
n→∞ n→∞
R∈Cn
dn
Note that Q(R0 ) is the union of 2 oblique rectangles Q(R) with R ∈ Λn . All
these oblique rectangles Q(R), R ∈ Λn , are translates of each other. According to
Proposition 2.14 Lebesgue measure is translation invariant and thus
m(Q(R)) = 2−dn m(Q(R0 )).
Hence it follows from the last term in (2.20) that
(2.21) m(B) = lim 2−dn · m(Q(R0 )) · #Cn .
n→∞

Comparing (2.19) and (2.21), one concludes that m(Q(R0 )) = `(R0 ) as desired.
Now assume R0 is a rectangle of the form R0 = [a1 , b1 ] × · · · × [ad , bd ], with
ai < bi for each i. Let x = (a1 , . . . , an ) and, for each n ∈ N, let
       
1 1
Sn = a1 , 1 + b1 × · · · × ad , 1 + bd ,
n n
2. LEBESGUE MEASURE ON Rd 23

so that S1 ⊃ S2 ⊃ · · · R0 and ∩Sn = R0 , and also Q(S1 ) ⊃ Q(S2 ) ⊃ · · · Q(R0 ) and


∩Q(Sn ) = Q(R0 ). We then have
m(Q(R0 )) = lim m(Q(Sn )) = lim m(Q(Sn ) − x)
n→∞ n→∞
= lim m(Q(Sn − x)) = lim m(Sn − x) = lim m(Sn ) = lim m(R0 ),
n→∞ n→∞ n→∞ n→∞

where the first and sixth equalities follow from Theorem 1.13, the second and fifth
equalities follow from Proposition 2.14, the third equality follows from the linearity
of Q and the fourth equality follows from the fact that Sn − x is a rectangle of the
form [0, c1 ) × · · · [0, cd ) (with ci = −ai + (1 + n1 )bi ), which we have already treated.
Now fix A ⊂ Rd . For any ε > 0, there exist rectangles R1 , R2 , . . . such that
A ⊂ ∪∞n=1 Rn and
X∞
`(Rn ) ≤ m∗ (A) + ε.
n=1
Then, since Q(A) ⊂ ∪∞
n=1 Q(Rn ), we have

X ∞
X
m∗ (Q(A)) ≤ m(Q(Rn )) = `(Rn ) ≤ m∗ (A) + ε.
n=1 n=1
∗ ∗
This shows that m (Q(A)) ≤ m (A), and the reverse inequality follows by symme-
try and the fact that Q−1 is also orthogonal.
Now assume that A is Lebesgue measurable, then for any set Z ⊂ Ω
m∗ (Z ∩ Q(A)) + m∗ (Z ∩ (QA)c )
= m∗ (Z ∩ Q(A)) + m∗ (Z ∩ Q(Ac ))
(2.22)
= m∗ (Q−1 Z ∩ A) + m∗ (Q−1 Z ∩ Ac )
= m∗ (Q−1 (Z)) = m∗ (Z),
which implies that Q(A) is Lebesgue measurable. The converse follows by symme-
try. 

Proposition 2.16. Let T be an invertible linear transformation in Rd and let


A ⊂ Rd . Then
(2.23) m∗ (T (A)) = | det T | m∗ (A).
Hence T (A) is Lebesgue measurable if and only if A is Lebesgue measurable, and in
this case
(2.24) m(T (A)) = | det T | m(A).
Proof. The idea of the proof of (2.24) is to decompose T in transformations
which are easier to handle. For this reason the proof is broken up in several steps.
Step 1. The transformation T can be decomposed as
(2.25) T = U DV,
where U and V are orthogonal (that is U > U = V > V = I), and D is a dilation
(diagonal with positive entries). To see this note that T > T is strictly positive.
Hence there exists an orthogonal transformation V such that
T > T = V > ∆V, ∆ = diag (λ1 , . . . , λd ),
24 2. CONSTRUCTION OF MEASURES

where λ1 , . . . , λd are the necessarily positive eigenvalues of T > T . Define the linear
transformation S by
p p
S = V > DV where D = diag ( λ1 , . . . , λd ).
Then S is symmetric, invertible, and S 2 = V > D2 V = V > ∆V = T > T . Define
W = T S −1 . Since S −1 is symmetric, it follows that
W > W = S −1 T > T S −1 = S −1 S 2 S −1 = I,
and therefore W is orthogonal. Hence T can be written as
T = W S = W V > DV = U DV with U = W V >.
Note that U is orthogonal, while D is diagonal with positive entries.
Step 2. The dilation of a cover of A is a cover of the dilation of A. A dilated
rectangle has an area that is obtained by multiplying the area of the rectangle by
det D. Hence one obtains
m∗ (D(A)) = (det D) m∗ (A).
Step 3. The decomposition of T as in (2.25) and a successive argument involving
Lemma 2.15 and Step 2 shows

m∗ (T (A)) = m∗ (U DV (A)) = m∗ (DV (A))


= (det D) m∗ (V (A)) = (det D) m∗ (A),
where clearly det D = | det T |. Thus (2.23) has been shown. Now assume that A
is Lebesgue measurable, then for any set Z ⊂ Ω
m∗ (Z ∩ T (A)) + m∗ (Z ∩ (T A)c )
= m∗ (Z ∩ T (A)) + m∗ (Z ∩ T (Ac ))
(2.26) = | det T | (m∗ (T −1 Z ∩ A) + m∗ (T −1 Z ∩ Ac ))
= | det T | m∗ (T −1 (Z))
= m∗ (Z),
which implies that T (A) is Lebesgue measurable. Since T is invertible, it also
follows that if T (A) is Lebesgue measurable, then A is Lebesgue measurable. The
identity (2.24) is now clear. 

Corollary 2.17. Let T be a non-invertible linear transformation in Rd and


let A ⊂ Rd . Then T (A) is Lebesgue measurable and
m(T (A)) = | det T | m(A) = 0.
Proof. Let A ⊂ Rd , then T (A) is contained in a linear subspace of Rd of
dimenion at most d − 1. Hence it suffices to show that any linear subspace of
dimension d − 1 is Lebesgue measurable and has measure 0.
Let B ⊂ Rd be a linear subspace of dimension d − 1 After an appropriate
orthogonal transformation B takes the form Rd−1 × {0}. Since by Lemma 2.15
the outer measure remains invariant it follows that m∗ (B) = m∗ (Rd−1 × {0}) = 0.
Thus m∗ (B) = 0 and B is Lebesgue measurable with Lebesgue measure 0. 
3. REGULARITY PROPERTIES OF LEBESGUE MEASURABLE SETS 25

3. Regularity properties of Lebesgue measurable sets


In the construction of the σ-algebra (Rd , Md ), see Definition 2.9, it has been
shown that the open sets of Rd belong to Md ; cf. Corollary 2.11. Since the Borel
σ-algebra (Rd , Bd ) is generated by all open sets in Rd , cf. Definition 1.22, it follows
that (Rd , Bd ) ⊂ (Rd , Md ). There exist Lebesgue measurable subsets of Rd which
are not Borel measurable, so the inclusion is strict. The measurable space (Rd , Bd )
becomes a measure space when the Lebesgue measure is restricted to the Borel
measurable sets: (Rd , Bd , m) ⊂ (Rd , Md , m). Recall from the construction via outer
measures that the measure space (Rd , Md , m) is complete. According to Theorem
1.37, the completion of (Rd , Bd , m) is contained in (Rd , Md , m). It will be shown
that in fact the completion of (Rd , Bd , m) equals (Rd , Md , m). Furthermore it will
be shown that Lebesgue measurable subsets of Rd are regular in the sense that they
are “approximately open” and “approximately closed”.
Lemma 2.18. Let A ⊂ Rd be any set with m∗ (A) < ∞. For every ε > 0 there
exists an open set O ⊂ Rd such that
(2.27) A ⊂ O, m∗ (A) ≤ m(O) < m∗ (A) + ε.
Hence, for any A ⊂ Rd there exists a sequence of open sets On ⊂ Rd such that
∞ ∞
!
\ \

(2.28) A⊂ On , m (A) = m On .
n=1 n=1

Proof. For ε > 0 there exists a cover of closed rectangles Rn , n ∈ N, of A


with

X ε
`(Rn ) < m∗ (A) + .
n=1
2
Let Sn be a closed rectangle whose interior Sn0 contains Rn , and such that
ε
`(Sn ) ≤ `(Rn ) + n+1 .
2
S∞ S∞
Let O = n=1 Sn0 , so that O ⊂ Rd is open. Clearly, A ⊂ O and, since O ⊂ n=1 Sn ,
one obtains via the monotonicity and countable subadditivity of outer measures
∞ ∞
X X ε
m∗ (A) ≤ m(O) ≤ `(Sn ) ≤ `(Rn ) + < m∗ (A) + ε.
n=1 n=1
2
Hence (2.27) follows.
In order to show (2.28), apply (2.27) with ε = 1/n, so that there exist open
sets On ⊂ Rd such that
1
(2.29) A ⊂ On , m∗ (A) ≤ m(On ) < m∗ (A) + .
n
Then clearly the inclusion in (2.29) implies
∞ ∞ ∞
! !
\ \ \
∗ ∗
A⊂ On and m (A) ≤ m On = m On ;
n=1 n=1 n=1
T∞
recall that n=1 On is Borel measurable. Hence it follows from the inequality in
(2.29) that

!

\ 1
m (A) ≤ m On ≤ m(On ) < m∗ (A) + .
n=1
n
26 2. CONSTRUCTION OF MEASURES

Therefore (2.28) is obtained when n → ∞. 

Theorem 2.19. The σ-algebra (Rd , Md , m) of the Lebesgue measurable sets is


the completion of the σ-algebra (Rd , Bd , m) of the Borel measurable sets.
Proof. Denote the measure spaces (Rd , Bd , m) and (Rd , Md , m) by the abbre-
viations B and M, respectively, and denote the completion of (Rd , Bd , m) by C. It
is clear that C ⊂ M; cf. Theorem 1.37. Hence it suffices to show that M ⊂ C. For
this purpose let A ∈ M.
First assume that m(A) < ∞. By (2.28) in Lemma 2.18 there exists a Borel
measurable set B such that
B ∈ B, A ⊂ B, m(B) = m(A).
Denote N = B \ A. Then N ∈ M and B = A ∪ N while A ∩ N = ∅. Hence
m(B) = m(A) + m(N ). The assumption m(A) < ∞ leads to m(N ) = 0. Again, by
(2.28) in Lemma 2.18 there exists a Borel measurable set C ⊂ Rd such that
C ∈ B, N ⊂ C, 0 = m(C) (= m(N )),
so that N ∈ C. Since B ∈ B ⊂ C and C is a σ-algebra, this implies A = B \ N ∈ C.
Now assume m(A) = ∞. Define An = A ∩ (−n, n)d , then An ∈ M and
m(An ) < ∞. Hence by the previous reasoning it follows that An ∈ C and

[
A= An ∈ C,
n=1

since C is a σ-algebra.
Thus in both cases it follows that A ∈ C. Therefore M ⊂ C. 

Proposition 2.20. Let A ⊂ Rd be a set. Then the following statements are


equivalent:
(1) A is Lebesgue measurable;
(2) for every ε > 0 there exists an open set O ⊂ Rd such that A ⊂ O and
m∗ (O \ A) < ε;
(3) for every ε > 0 there exists a closed set F ⊂ Rd such that F ⊂ A and
m∗ (A \ F ) < ε.
Proof. (1) ⇒ (2) Let A be Lebesgue measurable and let ε > 0. There are
two cases to consider.
If m(A) < ∞, then by Lemma 2.18 there exists an open set O ⊂ Rd
A ⊂ O, m(A) ≤ m(O) < m(A) + ε.
This leads to m∗ (O \ A) = m(O \ A) = m(O) − m(A) < ε.
If m(A) = ∞, defineS An = A ∩ (−n, n)d . Then An is Lebesgue measurable and

m(An ) < ∞, while A = n=1 An . By the previous reasoning there exist open sets
On ⊂ Rd such that
ε
An ⊂ On , m∗ (On \ An ) < n+1 .
2
S∞
Define O = n=1 On , then O is open and A ⊂ O. Observe that

[ ∞
[ ∞
[
O\A= On \ An ⊂ (On \ An ).
n=1 n=1 n=1
3. REGULARITY PROPERTIES OF LEBESGUE MEASURABLE SETS 27

The monotonicity and countable subadditivity of m∗ lead to


∞ ∞
!
[ X
m∗ (O \ A) ≤ m∗ (On \ An ) ≤ m∗ (On \ An ) < ε.
n=1 n=1

Hence in each of the above cases there exists an open set O ⊂ Rd such that
A ⊂ O and m∗ (O \ A) < ε, so that (2) is satisfied.
(2) ⇒ (1) For every n ∈ N there
T∞exists an open set On ⊂ Rd such that A ⊂ On

and m (On \ A) < 1/n. With O = n=1 On one has O Lebesgue measurable, while
A ⊂ O. It follows for every n ∈ N that

!
\ 1
m∗ (O \ A) = m∗ (On \ A) ≤ m∗ (On \ A) ≤ .
n=1
n

Therefore m∗ (O \ A) = 0, so that O \ A is Lebesgue measurable. Hence it follows


from A = O \ (O \ A) that A is Lebesgue measurable, and (1) is satisfied.
(1) ⇒ (3) Let A be Lebesgue measurable, then also Ac is Lebesgue measurable.
Thus by (2) for every ε > 0 there exists an open set O ⊂ Rd such that Ac ⊂ O and
m∗ (O \ Ac ) < ε. This implies Oc ⊂ A and m∗ (A \ Oc ) < ε (as A \ Oc = O ∩ A =
O \ Ac ). Let F = Oc , then F is closed, F ⊂ A, and m∗ (A \ F ) < ε, so that (3) is
satisfied.
(3) ⇒ (1) For every ε > 0 there exists a closed set F ⊂ Rd such that F ⊂ A
and m∗ (A \ F ) < ε, or in other words, Ac ⊂ F c and m∗ (F c \ Ac ) < ε. Let
O = F c , then O is open, Ac ⊂ O, and m∗ (O \ Ac ) < ε. Hence the conditions in
(2) are satisfied for Ac . Thus by (1) Ac is Lebesgue measurable, and therefore A is
Lebesgue measurable. 

Corollary 2.21. Let A ⊂ Rd be a set. Then A is Lebesgue measurable if and


only if for every ε > 0 there exists an open set O ⊂ Rd and a closed set F ⊂ Rd
such that F ⊂ A ⊂ O and m(O \ F ) < ε.
Proof. (⇒) Let A ⊂ Rd be Lebesgue measurable and let ε > 0. Then by
Theorem 2.20 there exist an open set O ⊂ Rd and a closed set F ⊂ Rd such that
F ⊂ A ⊂ O and
m∗ (O \ A) < ε/2, m∗ (A \ F ) < ε/2.
Then the inclusion
O \ F ⊂ (O \ A) ∪ (A \ F ),
and the monotonicity and subadditivity of m∗ lead to m(O \ F ) = m∗ (O \ F ) < ε.
(⇐) Let ε > 0 so that there are an open set O ⊂ Rd and a closed set F ⊂ Rd
with F ⊂ A ⊂ O and m(O \ F ) < ε. Note that O \ A ⊂ O \ F , so that the
monotonicity of m∗ implies
m∗ (O \ A) ≤ m∗ (O \ F ) = m(O \ F ) < ε.
Hence the open set O satisfies A ⊂ O and m∗ (O \ A) < ε. The measurability of A
follows from Theorem 2.20. 

Corollary 2.22. Let A ⊂ Rd be a set. Then A is Lebesgue measurable and


bounded if and only if for every ε > 0 there exists a bounded open set O ⊂ Rd and
a compact set F ⊂ Rd such that F ⊂ A ⊂ O and m(O \ F ) < ε.
28 2. CONSTRUCTION OF MEASURES

Proof. Only the implication (⇒) needs a proof. If A measurable then there
exists an open set O ⊂ Rd and a closed set F ⊂ Rd such that F ⊂ A ⊂ O and
m(O \ F ) < ε. Since A is bounded also F is bounded and hence compact; moreover
there exists an open ball B around 0 such that A ⊂ B. Then O1 = O ∩ B is a
bounded open set and A ⊂ O1 , while O1 \ F ⊂ O \ F , so that m(O1 \ F ) < ε. 

4. Lebesgue-Stieltjes measure on R
The construction of the Lebesgue measure on Rd has an analog on R when the
outer measure is defined in terms of a non-decreasing function.
Proposition 2.23. Let F : (a, b) → R be a non-decreasing real-valued function.
Then for x ∈ (a, b) the limits F (x−) and F (x+) exist and
sup F (t) = F (x−) ≤ F (x) ≤ F (x+) = inf F (t).
a<t<x x<t<b

Moreover, the function F has at most countably many discontinuities x ∈ (a, b),
which are jump discontinuities: F (x−) < F (x+). Furthermore
a<x<y<b ⇒ F (x+) ≤ F (y−).
Proof. Let x ∈ (a, b), then the set {F (t) : a < t < x} is bounded above by
F (x). Hence
A = sup F (t) ≤ F (x).
a<t<x
Now choose ε > 0. Then by the supremum property there exists δ > 0 such that
a<x−δ <x and A − ε < F (x − δ) ≤ A.
The monotonicity property of F shows that for all t with x − δ < t < x
A − ε ≤ F (x − δ) ≤ F (t) ≤ A.
Hence it has been shown that A = F (x−). A similar reasoning holds for F (x+).
Now let a < x < y < b, then the monotonicity of F in fact gives
F (x+) = inf F (t) = inf F (t), F (y−) = sup F (t) = sup F (t).
x<t<b x<t<y a<t<y x<t<y

Hence one obtains


F (x+) ≤ F (y−).
Let E be the set of points x ∈ (a, b) where F is not continuous. Clearly, for
every x ∈ E there is r(x) ∈ Q such that
F (x−) < r(x) < F (x+).
Now let x1 , x2 ∈ E with x1 < x2 . Then F (x1+ ) ≤ F (x2− ),
F (x1 −) < r(x1 ) < F (x1 +) ≤ F (x2− ) < r(x2 ) < F (x2 +),
so that r(x1 ) < r(x2 ). Hence there is a one-to-one correspondence between E and
a subset of Q. 
In the following it will be assumed that F is continuous from the right, i.e.,
F (x) = F (x+); in this case the jumps of F have the form F (x) − F (x−). Define
the length of a left-open, right-closed interval (a, b] relative to F by
`F (a, b] = F (b) − F (a).
Note that for a < c < b one has `F (a, b] = `F (a, c] + `F (c, b].
4. LEBESGUE-STIELTJES MEASURE ON R 29

Definition 2.24. Let A ⊂ R. Then (mF )∗ (A) is an element of [0, ∞] defined


by
∞ ∞
( )
X [

(2.30) (mF ) (A) = inf `F (In ) : In = (an , bn ], A ⊂ In .
n=1 n=1

Theorem 2.25. The extended real-valued function (mF )∗ : P (R) → [0, ∞] in


(2.30) is an outer measure on R, called the Lebesgue-Stieltjes outer measure.
The proof of this theorem and the proof of the following lemma run very much
as in the case of Lebesgue outer measure.
Lemma 2.26. Let the Lebesgue-Stieltjes outer measure (mF )∗ be as in (2.30).
Then
(mF )∗ ((a, b]) = `F (a, b].
Definition 2.27. The Lebesgue-Stieltjes measurable sets of R are the mea-
surable sets defined by the Lebesgue-Stieltjes outer measure (mF )∗ in (2.30), and
Lebesgue-Stieltjes measure mF is the restriction of (mF )∗ to the Lebesgue-Stieltjes
measurable sets.
Proposition 2.28. Every interval (a, b] in R is Lebesgue-Stieltjes measurable.
Proof. The proof is very similar to that of Proposition 2.10. Fix an interval
(a, b] and Z ⊂ R with (mF )∗ (Z) < ∞; it suffices to prove that
(mF )∗ (Z ∩ (a, b]) + (mF )∗ (Z ∩ (a, b]c ) ≤ (mF )∗ (Z).

Pε∞> 0, there exist intervals In = (an , bn ], with n ∈ N, such that Z ⊂ ∪n In


Given
and n=1 `F (In ) < (mF )∗ (Z) + ε. For each n, let
In0 = In ∩ (−∞, a], In00 = In ∩ (a, b], In000 = In ∩ (b, ∞)
note that these three sets are disjoint intervals, all of which are open on the left
and closed on the right. In particular, `F (In ) = `F (In0 ) + `F (In00 ) + `F (In000 ). Since
(a, b] ∩ Z ⊂ ∪n In00 and (a, b]c ∩ Z ⊂ (∪n In0 ) ∪ (∪n In000 ), we obtain
X X
(mF )∗ (Z ∩ (a, b]) + (mF )∗ (Z ∩ (a, b]c ) ≤ `F (In00 ) + (`F (In0 ) + `F (In000 ))
n n
X
= `F (In ) < (mF )∗ (Z) + ε.
n

Corollary 2.29. Every open set in R is Lebesgue-Stieltjes measurable.
Proof. Every open set in R can be expressed as the union of a countable
number of dyadic intervals. 
Corollary 2.30. The Lebesgue-Stieltjes measure on R is σ-finite.
Proof. For each n ∈ N, let An = (−n, n]. Each An is then Lebesgue-Stieltjes
measurable. Moreover, R = ∪∞
n=1 An and, for each n, mF (An ) = `F (An ) = F (n) −
F (−n) < ∞. 
Of course now the measure of a set consisting of one point need not be 0, due
to the possible jumps of the function F . In fact the following lemma describes the
situation.
30 2. CONSTRUCTION OF MEASURES

Lemma 2.31. The singletons {a}, a ∈ R, are Lebesgue-Stieltjes measurable and


mF ({a}) = F (a) − F (a−), a ∈ R.
Proof. The first statement follows from writing {a} = ∩∞n=1 (a − 1/n, a]. For
the second statement, from the Theorem 1.13 we have
 
1
mF ({a}) = lim mF a − ,a
n→∞ n
 
1
= F (a) − lim F a − = F (a) − F (a−), 
n→∞ n
Lemma 2.32. Let A ⊂ R be any set with (mF )∗ (A) < ∞. For every ε > 0 there
exists an open set O ⊂ R such that
(2.31) A ⊂ O, (mF )∗ (A) ≤ m(O) < (mF )∗ (A) + ε.
Hence, for any A ⊂ R there exists a sequence of open sets On ⊂ R such that
∞ ∞
!
\ \

(2.32) A⊂ On , (mF ) (A) = m On .
n=1 n=1

Proof. Given A ⊂ R with (mF ) P (A) < ∞ and ε > 0, there exist intervals

In = (an , bn ] such that A ⊂ ∪n In and n=1 `F (In ) < (mF )∗ (A) + ε/2. Then, for
each n, using the fact that F is continuous from the right, there exists δn > 0 such
that F (bn + δn ) − F (bn ) < ε2−n−1 . Hence, the open set O = ∪∞ n=1 (an , bn + δn ) is
such that A ⊂ O and
X∞ ∞
X
mF (O) ≤ mF ((an , bn + δn )) = (mF ((an , bn ]) + mF ((bn , bn + δ]))
n=1 n=1

ε X
≤ + (F (bn + δn ) − F (bn )) < ε.
2 n=1
This proves (2.31). The second statement can be proved in the same way as the
second statement of Lemma 2.18 was proved from (2.27). 
The following two propositions can now be proved from Lemma 2.32 in the
exact same way as Theorem 2.19 and Proposition 2.20 are proved from Lemma
2.18.
Proposition 2.33. The σ-algebra of Lebesgue-Stieltjes measurable sets is the
completion of the Borel σ-algebra.
Proposition 2.34. Let A ⊂ R be a set. Then the following statements are
equivalent:
(1) A is Lebesgue-Stieltjes measurable;
(2) for every ε > 0 there exists an open set O ⊂ R such that A ⊂ O and
m∗F (O \ A) < ε;
(3) for every ε > 0 there exists a closed set F ⊂ R such that F ⊂ A and
m∗F (A \ F ) < ε.
Example 2.35. Define the function F : R → R by
F (x) = x, x ∈ R,
then mF is Lebesgue measure with mF ((a, b]) = b − a.
5. ON THE CONSTRUCTION OF MEASURES 31

Example 2.36. Define the function F : R → R by


(
1, if x ≥ a,
F (x) =
0, if x < a,
where a ∈ R. Then mF is the Dirac measure supported at a:
(
1, if a ∈ A,
mF (A) =
0, if a 6∈ A,
for any Borel set A ⊂ R.

5. On the construction of measures


We first recall some function terminology. Let A and B be two sets and A0 ⊂ A.
Given functions f : A → B and g : A0 → B such that f (x) = g(x) for all x ∈ A0 ,
we say that f is an extension of g to A, and also that g is the restriction of f to
A0 . It is common that one abuses notation and denotes both functions by the same
symbol; the notation g = f |A0 is also common.
Let us briefly recapitulate the construction of the Lebesgue measure on Rd
that we have seen in this chapter. Our starting point was a set function `, which
was defined on a simple class of subsets of Rd , namely the set of rectangles. We
then followed a two-step procedure. First, we extended the function ` to an outer
measure m∗ , whose domain was the set of all subsets of Rd . Second, we showed that
m∗ has a restriction m, a measure defined in a domain Md which is a σ-algebra and
contains the initial class of rectangles of Rd (in particular, m is also an extension
of `). The procedure relied crucially on important properties of the initial function
` and of its domain.
The purpose of this section is to explain that, although we have worked on
Rd , a similar construction can be carried out in a much more general and abstract
setting. Such is the content of Carathéodory’s extension theorem, which we will
state shortly. The starting point of the extension described in this theorem is again
a set function (call it µ) defined on a class of subsets (call it S) of a set Ω; µ and S
will need to satisfy certain properties (µ has to be a pre-measure and S a semi-ring;
definitions are given below). We then again follow a two-procedure. First, µ is
extended to an outer measure µ∗ on the power set of Ω (in fact, this extension is
somewhat involved, so this step is split into two subparts, as we will detail). Second,
it is shown that µ∗ has a restriction which is a measure defined on a σ-algebra that
contains S.
We now turn to the formal definitions.
Definition 2.37. Let A be a class of subsets of Ω with ∅ ∈ A. Then the
extended real-valued set function µ : A → [0, ∞] is a pre-measure when:
(1) µ(∅) = 0; S∞
(2) An ∈ A pairwise disjoint and n=1 An ∈ A imply that
∞ ∞
!
[ X
µ An = µ(An ) (∈ [0, ∞]).
n=1 n=1
The pre-measure
S∞ on A is σ-finite when there exist Ωn ∈ A with µ(Ωn ) < ∞ such
that Ω = n=1 Ωn .
Definition 2.38. A collection S of subsets of a set Ω is a semi-ring if
32 2. CONSTRUCTION OF MEASURES

(1) ∅ ∈ S;
(2) A, B ∈ S ⇒ A ∩ B ∈SS;
n
(3) A, B ∈ S ⇒ A \ B = k=1 Ck , where C1 , . . . , Cn ∈ S are pairwise disjoint.
Theorem 2.39 (Carathéodory extension theorem). Let S be a semi-ring on Ω
and let µ : S → [0, ∞] be a pre-measure. Then µ has an extension to a measure
on σ(S), the σ-algebra generated by S. If the pre-measure on S is σ-finite, then its
extension to σ(S) is unique.
Let us now show how both the Lebesgue measure and the Lebesgue-Stieltjes
measure can be obtained from this theorem.
Example 2.40 (Lebesgue measure). The collection of rectangles of Rd of the
form (a1 , b1 ] × · · · × (ad , bd ] forms a semi-ring S as is easily verified (we need to take
semi-open intervals here, in contrast to what was done in the course of the chapter,
in order to guarantee that the difference of two members of S produces a disjoint
union of members of S). Denote the function which assigns the volume `(R) to each
block R ∈ S by µ. Then the function µ is a pre-measure.
Example 2.41 (Lebesgue-Stieltjes measure). Let the collection of subsets S of
R be defined by
S = {(a, b] : −∞ < a ≤ b < ∞}.
Then S is a semi-ring which generates the Borel σ-algebra on R. Let F : R → R be
a non-decreasing and right-continuous function and define µ : S → [0, ∞] by
µ((a, b]) = F (b) − F (a).
Then µ is a pre-measure on S etc.
We will not give the full proof of Theorem 2.39, as much of it would be a rep-
etition of the work we have done in constructing the Lebesgue measure on Rd . We
will simply indicate the steps involved, stating the corresponding results without
proof.

Step 1a: From a pre-measure on a semi-ring to a pre-measure on a ring


Definition 2.42. A collection R of subsets of a set Ω is ring if
(1) ∅ ∈ R;
(2) A, B ∈ R ⇒ A \ B ∈ R;
(3) A, B ∈ R ⇒ A ∪ B ∈ R.
Note that a ring is also a semi-ring. Due to A ∩ B = A \ (A \ B) a ring is
closed under intersections. If R is a ring with Ω ∈ R, then R is an algebra. The
smallest ring R(S) which contains the semi-ring S (i.e., the intersection of all rings
containing S) is given by
( n
)
[
R(S) = A ∈ P (Ω) : A = Ak : A1 , . . . , An ∈ S for some n ∈ N ,
k=1
where these unions may also be taken over mutually disjoint sets from S.
Lemma 2.43. If S is a semi-ring, then σ(S) = σ(R(S)).
Proposition 2.44. A pre-measure on a semi-ring S has a unique extension to
a pre-measure on R(S).
5. ON THE CONSTRUCTION OF MEASURES 33

Step 1b: From a pre-measure on a ring to an outer measure on the power set
The results in this step are sometimes stated only for the case that R is an
algebra.
Definition 2.45. Let R be a ring on Ω and let µ : R → [0, ∞] be a pre-measure
on R. Define for any set E ⊂ Ω
( ∞ ∞
)
X [
(2.33) µ∗ (E) = inf µ(An ) : E ⊂ An , A n ∈ R .
n=1 n=1

Proposition 2.46. The set function µ : P (Ω) → [0, ∞] in (2.33) is an outer
measure on Ω with the property
µ∗ (A) = µ(A), A ∈ R.

Step 2: From an outer measure on the power set to a measure on a σ-algebra


Let µ∗ be the outer measure on Ω as given in (2.33). A set A ⊂ Ω is measurable
with respect to µ∗ if for any set Z ⊂ Ω
µ∗ (Z) = µ∗ (Z ∩ A) + µ∗ (Z ∩ Ac ).
The set of all µ∗ -measurable sets in Ω is denoted by A∗ . Recall the following result.
Proposition 2.47. The collection A∗ is a σ-algebra on Ω which contains S
(and hence also σ(S) = σ(R(S)). Moreover, the restriction of µ∗ to A∗ is a complete
measure.
Note that, except for the statement that σ(S) ⊂ A∗ , this proposition is a con-
sequence of Theorem 2.5. Theorem 2.39 follows from putting together Proposition
2.47 and
Lemma 2.48. If µ is a σ-finite pre-measure on the semi-ring S and C is a σ-
algebra with S ⊂ C ⊂ A∗ , then there exists a unique measure on C that extends
µ.
CHAPTER 3

Measurability of functions

Measurable mappings are structure-preserving mappings between measurable


spaces; they form a natural context for the theory of integration. Specifically,
a mapping between measurable spaces is said to be measurable if the pre-image
of each measurable set is measurable, analogous to the situation of continuous
functions between topological spaces. Special care must be taken regarding the
σ-algebras involved. In particular, for a real-valued function the range space R is
provided with the σ-algebra of Borel sets. The main emphasis in this chapter is on
properties of measurable functions with values in the extended real line.

1. Measurable mappings
Definition 3.1. Let (Ω, A) and (Ω0 , A0 ) be measurable spaces. The mapping
f : Ω → Ω0 is measurable, or, more precisely, (A, A0 )-measurable, if f −1 (A0 ) ∈ A
for any A0 ∈ A0 .
Let the mappings f : (Ω, A) → (Ω0 , A0 ) and g : (Ω0 , A0 ) → (Ω00 , A00 ) be measur-
able. Then it is clear from the definition that also the composition
g ◦ f : (Ω, A) → (Ω00 , A00 )
is measurable.
Theorem 3.2. Let (Ω, A) and (Ω0 , A0 ) be measurable spaces and let E0 be a
generator for A0 . Then the following statements are equivalent:
(1) the mapping f : Ω → Ω0 is measurable;
(2) f −1 (E 0 ) ∈ A for any E 0 ∈ E0 .
Proof. (1) ⇒ (2) This is clear.
(2) ⇒ (1) Define the collection F0 ⊂ A0 by
F0 = { A0 ∈ A0 : f −1 (A0 ) ∈ A }.
Clearly Ω0 ∈ F0 , since f −1 (Ω0 ) = Ω. If A0 ∈ F0 , then
f −1 (Ω0 \ A0 ) = Ω \ f −1 (A0 ) ∈ A.
If A0n ∈ F0 , then
∞ ∞
!
[ [
f −1
A0n = f −1 (A0n ) ∈ A.
n=1 n=1
Hence F0 is a σ-algebra in Ω0 . Due to the assumption E0 ⊂ F0 , then also
A0 = σ(E0 ) ⊂ F0 ;
so that A0 = F0 . Hence f is measurable. 
35
36 3. MEASURABILITY OF FUNCTIONS

On the basis of the above definition it is reasonable to call a mapping f from a


measurable space (Ω, A) to a topological space Y measurable if Y is provided with
the σ-algebra B(Y ) generated by the open sets in Y and f −1 (B) ∈ A for every
B ∈ B(Y ). The following result is then clear due to Theorem 3.2.
Proposition 3.3. Let Y be a topological space and let B(Y ) be the σ-algebra
generated by the open sets in Y . Let f be a mapping from the measurable space
(Ω, A) to the measurable space (Y, B(Y )). Then the following statements are equiv-
alent:
(1) f is measurable;
(2) the set f −1 (O) belongs to A for each open set O ⊂ Y .
It will be useful to consider the cases where Y = R or Y = R2 in more detail.
Recall the notation B = B(R) and B2 = B(R2 ).
Definition 3.4. Let f be a function from the measurable space (Ω, A) to R.
Then f is called measurable if f is measurable as a function from (Ω, A) to (R, B).
The Borel σ-algebra B = B(R) is also generated by half-open and half-closed
intervals; cf. Proposition 1.23. In the present context the following result is equiv-
alent to Proposition 3.3.
Corollary 3.5. The real-valued function f : (Ω, A) → (R, B) is measurable
if and only if for each c ∈ R one (and hence all) of the following assertions are
satisfied:
(1) {ω ∈ Ω : f (ω) < c} ∈ A;
(2) {ω ∈ Ω : f (ω) ≤ c} ∈ A;
(3) {ω ∈ Ω : f (ω) > c} ∈ A;
(4) {ω ∈ Ω : f (ω) ≥ c} ∈ A.
Definition 3.6. Let f be a mapping from the measurable space (Ω, A) to R2 .
Then f is called measurable if f is measurable as a mapping from (Ω, A) to (R2 , B2 ).
Proposition 3.7. Let f = (f1 , f2 ) be a mapping from the measurable space
(Ω, A) to R2 . Then f is measurable if and only if the functions f1 and f2 are
measurable.
Proof. (⇒) Assume that f is measurable. Let A ⊂ R be open, then A × R is
open in R2 and therefore
(f1 )−1 (A) = {ω ∈ Ω : f1 (ω) ∈ A}
= {ω ∈ Ω : f1 (ω) ∈ A, f2 (ω) ∈ R} = f −1 (A × R)
is measurable. Hence f1 is measurable. A similar argument applies to f2 .
(⇐) Each open set O ⊂ R2 is the countable union of dyadic intervals in O.
Hence it suffices to show that the inverse image of a rectangle (a1 , b1 ] × (a2 , b2 ] is
measurable. Note that
f −1 ((a1 , b1 ] × (a2 , b2 ]) = (f1 )−1 ((a1 , b1 ]) ∩ (f2 )−1 ((a2 , b2 ]),
and (f1 )−1 ((a1 , b1 ]) and (f2 )−1 ((a2 , b2 ]) are measurable. 
Complex-valued functions will be met frequently in applications. According to
Proposition 3.7, with B(C) and B(R2 ) identified, such functions are measurable if
and only if their real and imaginary parts are measurable.
2. MEASURABLE FUNCTIONS 37

2. Measurable functions
A real-valued function f : Ω → R has been called measurable in the context
of a σ-algebra A on Ω and the Borel σ-algebra B on R; cf. Definition 3.4. It
will be necessary to widen this definition to include extended real-valued functions
f : Ω → R. Here R denotes the extended real line which is provided with the
usual ordering and topology. The Borel σ-algebra B(R) is then defined as the
smallest σ-algebra containing all open sets in R. The following notation B = B(R)
will be used throughout. If (Ω, A) is a measurable space, the extended real-valued
function f on Ω is measurable if it is measurable as an extended real-valued function
f : (Ω, A) → (R, B).
Proposition 3.8. Let f be an extended real-valued function from the measur-
able space (Ω, A) to (R, B). Then the following statements are equivalent:
(1) f is measurable;
(2) the set f −1 (O) belongs to A for each open set O ⊂ R.
The Borel σ-algebra B is generated by half-open and half-closed intervals in R.
In the present context the following result is equivalent to Proposition 3.8.
Corollary 3.9. The extended real-valued function f : (Ω, A) → (R, B) is mea-
surable if and only if for each c ∈ R one, and hence all, of the following assertions
are satisfied:
(1) {ω ∈ Ω : f (ω) < c} ∈ A;
(2) {ω ∈ Ω : f (ω) ≤ c} ∈ A;
(3) {ω ∈ Ω : f (ω) > c} ∈ A;
(4) {ω ∈ Ω : f (ω) ≥ c} ∈ A.
Proposition 3.10. Let f, g : (Ω, A) → (R, B) be measurable extended real-
valued functions and let A ∈ A. Then the sets
{ω ∈ Ω : f (ω) < g(ω)}, {ω ∈ Ω : f (ω) ≤ g(ω)}, {ω ∈ Ω : f (ω) = g(ω)},
are measurable.
Proof. Note that f (ω) < g(ω) if and only if there is a number r ∈ Q with
f (ω) < r < g(ω). This observation shows the identity
[
{ω ∈ Ω : f (ω) < g(ω)} = {ω ∈ Ω : f (ω) < r} ∩ {ω ∈ Ω : r < g(ω)},
r∈Q

from which measurability follows. Now consider the identity


{ω ∈ Ω : f (ω) ≤ g(ω)} = Ω \ {ω ∈ Ω : f (ω) > g(ω)}.
This shows the measurability of the left-hand side. Finally the identity
{ω ∈ Ω : f (ω) = g(ω)} = {ω ∈ Ω : f (ω) ≤ g(ω)} \ {ω ∈ Ω : f (ω) < g(ω)}
shows the measurability of the left-hand side. 
Proposition 3.11. Let f, g : (Ω, A) → (R, B) be measurable extended real-
valued functions. Then the maximum and the minimum of these functions, given
by
max (f, g), min (f, g)
are measurable.
38 3. MEASURABILITY OF FUNCTIONS

Proof. The measurability of the maximum follows from


{ω ∈ Ω : max (f, g)(ω) ≤ c} = {ω ∈ Ω : f (ω) ≤ c} ∩ {ω ∈ Ω : g(ω) ≤ c},
and the measurability of the minimum follows from
{ω ∈ Ω : min (f, g)(ω) ≥ c} = {ω ∈ Ω : f (ω) ≥ c} ∪ {ω ∈ Ω : g(ω) ≥ c}.
This completes the proof. 

Proposition 3.12. Let fn : (Ω, A) → (R, B) be a sequence of measurable


extended real-valued functions. Then also
sup fn , inf fn , lim sup fn , lim inf fn ,
n∈N n∈N n∈N n∈N

are measurable extended real-valued functions from (Ω, A) to (R, B).


Proof. The measurability of supn∈N fn follows from the identity

\
{ω ∈ Ω : sup fn (ω) ≤ c} = {ω ∈ Ω : fn (ω) ≤ c}.
n∈N n=1

Since each function fn is measurable, each set {ω ∈ Ω : fn (ω) ≤ c} is measurable.


Since countable unions of measurable sets are measurable, it follows that supn∈N fn
is measurable. Similarly, it is shown that inf n∈N fn is measurable.
Recall the definition of lim supn∈N fn :
lim sup fn = inf {g1 , g2 , . . . } where gk = sup {fk , fk+1 , fk+2 , . . . }.
n∈N

By the previous reasoning concerning the supremum the functions gk are mea-
surable, and then the same reasoning for the infimum show that lim supn∈N fn is
measurable. The result for lim inf n∈N fn follows in an analogous way. 

Definition 3.13. Let Ω be a set. A sequence of extended real-valued functions


fn : Ω → R converges pointwise to an extended real-valued function f : Ω → R if
fn (ω) → f (ω) for all ω ∈ Ω.
Proposition 3.14. Let (Ω, A) be a measurable space. Let the sequence of
measurable extended real-valued functions fn : Ω → R converge pointwise to an
extended real-valued function f : Ω → R. If all fn are measurable, then f is
measurable.
Proof. Use the identity
lim inf fn = lim fn = lim sup fn ,
n∈N n→∞ n∈N

and apply Proposition 3.12. 

Proposition 3.15. Let f, g : (Ω, A) → (R, B) be measurable extended real-


valued functions. Then the following extended real-valued functions are also mea-
surable:
f + g, f g, f /g
on their domains of definition.
2. MEASURABLE FUNCTIONS 39

Proof. Observe that the sets in Ω where f or g take on the values ∞ or


−∞ are measurable. Hence the domain of definition is measurable. Note that
f (ω) + g(ω) < c if and only if there is r ∈ Q with f (ω) < r and g(ω) < c − r. This
observation shows the identity
[
{ω ∈ Ω : f (ω) + g(ω) < c} = {ω ∈ Ω : f (ω) < c − r} ∩ {ω ∈ Ω : g(ω) < r}.
r∈Q

Hence f + g is measurable.
If h is measurable, then also h2 is measurable. To see this observe that for
c≤0
{ω ∈ Ω : h(ω)2 < c} = ∅,
while for c > 0
√ √
{ω ∈ Ω : h(ω)2 < c} = {ω ∈ Ω : − c < h(ω) < c}.

Now the measurability of the product f g follows from


1
(f + g)2 − f 2 − g 2 .

fg =
2
Observe that the domain of definition of f /g is given by {ω ∈ Ω : g(ω) 6= 0}
which is a measurable set. For the quotient f /g the set

{ω ∈ Ω : f /g(ω) < c}

is the union of

{ω ∈ Ω : g(ω) > 0} ∩ {ω ∈ Ω : f (ω) < cg(ω)}

and
{ω ∈ Ω : g(ω) < 0} ∩ {ω ∈ Ω : f (ω) > cg(ω)}.
Hence f /g is measurable on its domain of definition. 

Remark 3.16. Let the extended real-valued function f : (Ω, A) → (R, B) be


measurable. Then the positive and negative parts of f , defined by

f + = max (f, 0), f − = max (−f, 0),

are measurable, since constant functions are measurable. Hence with f also |f | =
f + + f − is measurable. Conversely, if f + and f − are measurable, then also f =
f + − f − is measurable.

Remark 3.17. Many of the notions introduced so far play a role in probability
theory. Here is a very brief outline. Let (Ω, A, µ) be a probability space, i.e., (Ω, A, µ)
is a σ-algebra and µ(Ω) = 1. The measurable sets A ∈ A are called events in the
context of probability. The probability of an event A ∈ A is defined by µ(A) and
the conditional probability of A ∈ A given B ∈ A is defined by µ(A ∩ B)/µ(B). The
events A, B ∈ A are said to be independent if µ(A ∩ B) = µ(A)µ(B). A random
variable is just a function f : (Ω, A) → (R, B) which is measurable.
40 3. MEASURABILITY OF FUNCTIONS

3. Approximation by simple functions


Simple functions are the building blocks for the theory of measurable functions:
every nonnegative measurable function is the limit of simple functions.
Definition 3.18. A real-valued function f : (Ω, A) → (R, B) is called simple
if f is measurable and takes on only a finite number of values.
Thus if f : (Ω, A) → (R, B) is simple and α1 , . . . , αr ∈ R are its values then
the sets
Ak = f −1 (αk ), k = 1, . . . , r,
are measurable and pairwise disjoint. Therefore the function f can be written as
Xr
f= αk 1Ak .
k=1

Theorem 3.19. Let the extended real-valued function f : (Ω, A) → (R, B) be


measurable. Then there exists a sequence of simple functions fn on Ω such that
fn → f pointwise, i.e., for all ω ∈ Ω one has fn (ω) → f (ω). Moreover:
(1) if f is bounded, then the convergence is uniform;
(2) if f ≥ 0, then the sequence fn may be chosen such that
fn ≥ 0, fn ≤ fn+1 .
Proof. First consider the case of a nonnegative extended real-valued function
f . For each n ∈ N define the sets
 
i−1 i
Ein = ω ∈ Ω : ≤ f (ω) < n , 1 ≤ i ≤ n2n ,
2n 2
and the set
Fn = {ω ∈ Ω : n ≤ f (ω)}.
Clearly for each n ∈ N all the sets Ein and Fn are measurable and disjoint, while
their union equals Ω. Define the function fn : Ω → R by
(
i−1
fn (ω) = 2n , if ω ∈ Ein ,
n, if ω ∈ Fn .
Then fn is simple and fn ≥ 0. It remains to show that fn → f pointwise. Assume
that f (ω) < ∞. Then f (ω) < m for some m ∈ N. This implies that f (ω) < n for
all n ≥ m. Hence for each n ≥ m one has ω ∈ Ein for some i and it follows that
1
fn (ω) ≤ f (ω) < fn (ω) + n , n ≥ m,
2
so that
1
0 ≤ f (ω) − fn (ω) < n , n ≥ m.
2
Therefore fn (ω) → f (ω). Now assume that f (ω) = ∞. Then ω ∈ Fn for all n ∈ N
and it follows that fn (ω) = n for each n ∈ N. Thus fn (ω) → f (ω). In each case it
has been shown that fn → f .
If f is nonnegative and in addition bounded, then there exists m ∈ N such that
for all ω ∈ Ω one has 0 ≤ f (ω) ≤ m. For n ≥ m it then follows for all ω ∈ Ω that
0 ≤ f (ω) ≤ n and
1
0 ≤ f (ω) − fn (ω) < n , n ≥ m.
2
4. PROPERTIES WHICH ARE VALID ALMOST EVERYWHERE 41

Therefore the convergence is uniform when f is bounded.


Now consider the case of a not necessarily nonnegative extended real-valued
function f . Then write f as f = f + − f − . It has been just shown that there exist
simple nonnegative functions gn and hn such that
gn → f + , hn → f − .
It follows that the function fn = gn − hn is simple and fn → f + − f − = f . Note
that f is bounded if and only if f + and f − are bounded simultaneously.
Finally it will be shown that if f ≥ 0 the constructed sequence fn is monoton-
ically increasing. In order to do this let ω ∈ Ω and choose n ∈ N. It will be shown
that fn (ω) ≤ fn+1 (ω) by considering several cases.
(a) First assume that f (ω) < n, so that for some 1 ≤ i ≤ n2n one has
i−1 i i−1
≤ f (ω) < n and fn (ω) = n .
2n 2 2
Going from n to n + 1 amounts to adding an intermediate point:
i−1 2i − 1 i
n
< n+1 < n .
2 2 2
Hence it follows that either
i−1 2i − 1 i−1
n
≤ f (ω) < n+1 and fn+1 (ω) = n = fn (ω),
2 2 2
or
2i − 1 i 2i − 1 i−1
≤ f (ω) < n and fn+1 (ω) = n+1 > n = fn (ω).
2n+1 2 2 2
(b) Now assume that n ≤ f (ω) < n + 1. Then it follows that for some 1 ≤ i ≤ 2n+1
one has
i−1 i i−1
n + n+1 ≤ f (ω) < n + n+1 and fn+1 (ω) = n + n+1 ≥ n = fn (ω).
2 2 2
(c) Assume that f (ω) ≥ n + 1. Then
fn+1 (ω) = n + 1 > n = fn (ω).
In each of these cases it has thus been shown that fn+1 (ω) ≥ fn (ω). 

Example 3.20. Let A ⊂ Ω be a measurable set and define the extended real-
valued function ∞A from Ω to R by
(
∞, ω ∈ A,
(3.1) ∞A (ω) =
0, ω 6∈ A.
Then clearly ∞A is measurable and the sequence fn = n1A is a monotone sequence
of simple functions such that fn → ∞A pointwise.

4. Properties which are valid almost everywhere


Let (Ω, A, µ) be a measure space. Recall that a set N ⊂ Ω has measure 0 if
N ∈ A and µ(N ) = 0.
Definition 3.21. Let (Ω, A, µ) be a measure space. A property is said to
hold almost everywhere (a.e.) on Ω if it holds on a measurable set A ∈ A whose
complement Ac has measure 0.
42 3. MEASURABILITY OF FUNCTIONS

Example 3.22. Let (Ω, A, µ) be a measure space and let f and g be extended
real-valued functions from Ω to R. Then f = g a.e. means that there is a set A ∈ A
such that
(3.2) f (ω) = g(ω) for all ω ∈ A,
while µ(Ac ) = 0.
Almost everywhere equality of functions gives rise to an equivalence relation.
Lemma 3.23. Let (Ω, A, µ) be a measure space. Let f , g, and h be extended
real-valued functions from Ω to R, such that f = g a.e. and g = h a.e. Then f = h
a.e.
Proof. Since f = g a.e. and g = h a.e., there are sets A, B ∈ A such that
f (ω) = g(ω) for ω ∈ A, and g(ω) = h(ω) for ω ∈ B,
while µ(Ac ) = µ(B c ) = 0. Define C = A ∩ B, so that it is clear that C ∈ A and
µ(C c ) = µ((A ∩ B)c ) = 0. Observe that
f (ω) = g(ω) = h(ω) for ω ∈ C,
so that f = h a.e. 
Note that the set A in (3.2) need not be the maximal subset of Ω on which
the functions f and g have equal values. However, if the functions f and g are
measurable one has the following result.
Lemma 3.24. Let (Ω, A, µ) be a measure space and let f and g be measurable
extended real-valued functions from Ω to R. Then
f = g a.e. ⇔ µ({ω ∈ Ω : f (ω) 6= g(ω)}) = 0.
Proof. Define the set N ⊂ Ω by
N = {ω ∈ Ω : f (ω) 6= g(ω)}.
It is clear that the sets N and N c belong to A.
(⇒) Assume that f = g a.e. Hence there exists A ∈ A such that f (ω) = g(ω)
for all ω ∈ A, while µ(Ac ) = 0. Note that N ⊂ Ac and hence µ(N ) = 0.
(⇐) For all ω ∈ N c one has f (ω) = g(ω). Since (N c )c = N and µ(N ) = 0 it
therefore follows that f = g a.e. 
Let (Ω, A, µ) be a measure space. A sequence of extended real-valued functions
fn from Ω to R converges pointwise almost everywhere to an extended real-valued
function f from Ω to R, i.e.,
lim fn = f a.e.,
n→∞
if there exists a set A ∈ A such that
fn (ω) → f (ω) for all ω ∈ A,
c
while µ(A ) = 0.
Lemma 3.25. Let (Ω, A, µ) be a measure space and let fn and f be measurable
extended real-valued functions from Ω to R, Then
lim fn = f a.e. ⇔ µ({ω ∈ Ω : lim fn (ω) 6= f (ω)}) = 0.
n→∞ n→∞
4. PROPERTIES WHICH ARE VALID ALMOST EVERYWHERE 43

Proof. Define the set N ⊂ Ω by


N = {ω ∈ Ω : lim fn (ω) 6= f (ω)}.
n→∞

It is clear that N is measurable.


(⇐) For all ω ∈ N c one has limn→∞ fn (ω) = f (ω). Since (N c )c = N and
µ(N ) = 0 it therefore follows that limn→∞ fn (ω) = f (ω) a.e.
(⇒) Assume that limn→∞ fn (ω) = f (ω) a.e. Hence there exists A ∈ A such
that limn→∞ fn (ω) = f (ω) for all ω ∈ A, while µ(Ac ) = 0. Note that N ⊂ Ac and
hence µ(N ) = 0. 

Theorem 3.26. Let (Ω, A, µ) be a complete measure space. Let f and g be


extended real-valued functions from Ω to R, such that f = g a.e. If f is measurable,
then g is measurable.
Proof. Let O be an open set in R. In order to show that g −1 (O) is measurable,
let A ∈ A be a set such that
f (ω) = g(ω) for all ω ∈ A,
while µ(Ac ) = 0. Clearly the identity
g −1 (O) = (g −1 (O) ∩ A) ∪ (g −1 (O) \ A) = (f −1 (O) ∩ A) ∪ (g −1 (O) \ A)
holds. The set f −1 (O) ∩ A is measurable since the sets f −1 (O) and A are measur-
able. Moreover,
g −1 (O) \ A ⊂ Ac and µ(Ac ) = 0,
which implies that the set g −1 (O) \ A is measurable, as (Ω, A, µ) is assumed to be
complete. Hence g −1 (O) is measurable. It follows that g is measurable. 

Proposition 3.27. Let (Ω, A, µ) be a complete measure space. Let fn and f be


extended real-valued functions from Ω to R. Assume that fn is measurable. Then
lim fn = f a.e. ⇒ f is measurable.
n→∞

Proof. Let A ⊂ Ω be a set such that fn (ω) → f (ω) for all ω ∈ A, while
µ(Ac ) = 0. Then fn 1A converges to f 1A . Note that fn 1A is measurable so that
the pointwise limit f 1A is measurable. Note that f 1A (ω) = f (ω) for ω ∈ A while
µ(Ac ) = 0. Hence f is measurable by Theorem 3.26. 

Let (Ω, A, µ) be a measure space and let (Ω, Ā, µ̄) be its completion. Extended
real-valued functions on Ω which are measurable with respect to A are also mea-
surable with respect to Ā. If an extended real-valued function on Ω is measurable
with respect to Ā, then there is an extended real-valued function measurable with
respect to A which comes very close to the original function. The closeness is mea-
sured in terms of the complete measure. This is where the notion “equality almost
everywhere” comes up naturally.
Theorem 3.28. Let (Ω, A, µ) be a measure space and let (Ω, Ā, µ̄) be its com-
pletion.
(1) Let f : (Ω, A) → (R, B) be a measurable extended real-valued function.
Then f is also measurable with respect to (Ω, Ā).
44 3. MEASURABILITY OF FUNCTIONS

(2) Let f : (Ω, Ā) → (R, B) be a measurable extended real-valued function.


Then there is an extended real-valued function g : (Ω, A) → (R, B), mea-
surable with respect to A, such that
(3.3) µ̄({ω ∈ Ω : g(ω) 6= f (ω)}) = 0,
i.e. g = f almost everywhere with respect to µ̄. Furthermore, g may be
constructed such that g(ω) = 0 if g(ω) 6= f (ω).
Proof. (1) This is clear, since A ⊂ Ā.
(2) Let the extended real-valued function f : (Ω, Ā) → (R, B) be measurable.
Then by Theorem 3.19 there exists a sequence of functions fn on Ω which are simple
with respect to Ā such that fn → f pointwise. Each function fn is of the form
Xr
αk 1Ak with Ak ∈ Ā, αk ∈ R,
k=1
and recall that for k = 1, . . . , r one has
Ak = B k ∪ C k , Bk ∈ A, Ck ∈ Ā, µ̄(Ck ) = 0.
Therefore the function gn defined by
r
X
gn = αk 1Bk
k=1
Sr
is measurable with respect to A, while gn and fn differ
S∞only on Nn = k=1 Ck , a
set which belongs to Ā and µ̄(Nn ) = 0. Now let N = n=1 Nn , so that N ∈ Ā and
µ̄(N ) = 0. Choose M ∈ A with N ⊂ M and µ(M ) = 0 and define the extended
real-valued function g : Ω → R by
g = lim gn 1Ω\M ,
n→∞
so that g is measurable with respect to A with the property that g(ω) = f (ω),
ω ∈ Ω \ M . Hence
{ω ∈ Ω : g(ω) 6= f (ω)} ⊂ M,
which confirms (3.3), since the set in the left-hand side belongs to Ā. 
CHAPTER 4

Integrability of functions

The integrability of a function will be defined when the function is measur-


able and when a measure is available. In fact, if a function f is measurable and
nonnegative on a measure space (Ω, A, µ), then f can be approximated pointwise
by a sequence
R of so-called simple functions whose integral is easy to define. The
integral Ω f dµ is then the limit of the integrals over the approximating simple
functions. This basic idea provides an integration theory for so-called integrable
functions (which are not necessarily nonnegative).

1. Integrals of nonnegative measurable functions


Definition 4.1. Let (Ω, A, µ) be a measure space and let f : (Ω, A) → (R, B)
be a nonnegative simple function with values α1 , . . . , αr ∈ R. Define
Ak = f −1 (αk ), k = 1, . . . , r,
so that the sets Ak are measurable, pairwise disjoint, and the function f can be
written as
Xr
f= αk 1Ak .
k=1
R
The integral Ω
f dµ of f over Ω with respect to µ, is defined by
Z Xr
f dµ = αk µ(Ak ).
Ω k=1

R instance, if A ∈ A, then 1A is a nonnegative simple measurable function


For
and 1 dµ = µ(A).
Ω A

Example 4.2. Let f : R → R be the characteristic function Rof the set Q and
let m be the Lebesgue measure. Then f is a simple function and R f dm = 0.
Proposition 4.3. Let (Ω, A, µ) be a measure space. Let f, g : (Ω, A) → (R, B)
be nonnegative simple functions and α ≥ 0. Then αf and f +g are also nonnegative
simple functions and
R R
(1) RΩ αf dµ = α ΩRf dµ, α ≥R 0;
(2) Ω (f + g)Rdµ = Ω f Rdµ + Ω g dµ;
(3) f ≤ g ⇒ Ω f dµ ≤ Ω g dµ.
Proof. Assume in general that
r
X
f= αk 1Ak with Ak = f −1 (αk ), k = 1, . . . , r,
k=1

45
46 4. INTEGRABILITY OF FUNCTIONS

and
s
X
g= βl 1 B l with Bl = g −1 (βl ), l = 1, . . . , s.
l=1
(1) If α = 0 the statement is trivial. If α > 0 note that αf is simple with values
αα1 , . . . , ααr ∈ R, and
Ak = (αf )−1 (ααk ), k = 1, . . . , r,
hence Z r
X r
X Z
αf dµ = ααk µ(Ak ) = α αk µ(Ak ) = α f dµ.
Ω k=1 k=1 Ω

(2) It is clear that f + g is simple and nonnegative. Let


Ck,l := Ak ∩ Bl , k = 1, . . . , r, l = 1, . . . , s,
and note that the sets Ck,l are disjoint (possibly empty) with
s
[ r
[
Ck,l = Ak and Ck,l = Bl .
l=1 k=1

Let γ be a value of f + g and let Jγ be the set of all pairs (k, l) with γ = αk + βl .
Then
[ X
Dγ := (f + g)−1 (γ) = Ck,l with µ(Dγ ) = µ(Ck,l ).
(k,l)∈Jγ (k,l)∈Jγ

Then it follows that


Z X r X
X s
(f + g) dµ = γµ(Dγ ) = (αk + βl )µ(Ck,l )
Ω γ∈ran (f +g) k=1 l=1
r s
! s r
!
X X X X
= αk µ(Ck,l ) + βl µ(Ck,l )
k=1 l=1 l=1 k=1
Xr s
X
= αk µ(Ak ) + βl µ(Bl )
k=1 l=1
Z Z
= f dµ + g dµ.
Ω Ω
(3) With the notations in the proof of (2), note that f ≤ g implies that αk ≤ βl on
Ck,l (which is trivial if Ck,l is empty). It follows that
Z Xr X s r X
X s Z
f dµ = αk µ(Ck,l ) ≤ βl µ(Ck,l ) = g dµ. 
Ω k=1 l=1 k=1 l=1 Ω

The functions in the following corollary are linear combinations of characteristic


functions of measurable sets which are not necessarily pairwise disjoint. Yet one
does not need to have inhibitions to integrate in the obvious way, due to Proposition
4.3.
Corollary 4.4. Let γk ≥ 0 and let Ck be measurable sets, k = 1, . . . , n. Then
the function f defined by
X n
f= γk 1Ck
k=1
1. INTEGRALS OF NONNEGATIVE MEASURABLE FUNCTIONS 47

is simple function and


Z n
X
f dµ = γk µ(Ck ).
Ω k=1

Proof. Note that fk = γk 1Ck is a nonnegative simple measurable function for


k = 1, . . . , n, and that f = f1 + . . . + fn . The statement follows from an induction
argument using (2) of Proposition 4.3. 

Definition 4.5. Let (Ω, A, µ) be a measure space and let the extended real-
valued function
R f : (Ω, A) → (R, B) be measurable and nonnegative. Then the
integral Ω f dµ is defined by
Z Z 
f dµ = sup g dµ : 0 ≤ g ≤ f, g simple ∈ [0, ∞].
Ω Ω

Note that if f in this definition is a simple function, then the definition agrees
with Definition 4.1. The following result is preliminary version of what will be
called the monotone convergence theorem.
Lemma 4.6. Let (Ω, A, µ) be a measure space and let the extended real-valued
function f : (Ω, A) → (R, B) be measurable and nonnegative. Let fn : (Ω, A) →
(R, B) be a sequence of nonnegative simple functions such that fn ↑ f . Then
Z Z
fn dµ ↑ f dµ.
Ω Ω

Proof. For n ≤ m it follows from Proposition 4.3 that


Z Z
fn dµ ≤ fm dµ.
Ω Ω
R
Hence limn→∞ Ω
fn dµ exists in [0, ∞]. It follows from Definition 4.5 that
Z Z
fn dµ ≤ f dµ,
Ω Ω
R R
which leads to limn→∞ Ω fn dµ ≤ Ω f dµ.
For the reverse inequality, it suffices to show that for any nonnegative simple
function g with g ≤ f one has
Z Z
g dµ ≤ lim fn dµ,
Ω n→∞ Ω

as this, by Definition 4.5, implies


Z Z
f dµ ≤ lim fn dµ.
Ω n→∞ Ω

Therefore, let the nonnegative simple function g with g ≤ f have the representation
r
X
g= γk 1Ck , Ck = g −1 (γk ),
k=1

where the sets Ck are disjoint and measurable. Choose 0 < ε < 1 and define
Ωn = {ω ∈ Ω : fn (ω) ≥ (1 − ε)g(ω)}.
48 4. INTEGRABILITY OF FUNCTIONS

Then clearly the S sets Ωn are measurable, and fn ≤ fn+1 implies Ωn ⊂ Ωn+1 .

Furthermore Ω = n=1 Ωn . To see this let ω ∈ Ω and assume that for all n one has
ω 6∈ Ωn , so that for all n ∈ N one has
fn (ω) < (1 − ε)g(ω).
Then it follows that fn (ω) < (1 − ε)f (ω), which contradicts fn (ω) ↑ f (ω). Thus
one may conclude Ωn ↑ Ω. Now observe that
Z Z Z
fn dµ ≥ fn 1Ωn dµ ≥ (1 − ε) g1Ωn dµ
Ω Ω Ω
Z r
!
X
= (1 − ε) γk 1Ck ∩Ωn dµ
Ω k=1
r
X
= (1 − ε) γk µ(Ck ∩ Ωn ),
k=1

by Corollary 4.4. Therefore it follows that


Z Xr r
X
lim fn dµ ≥ lim (1 − ε) γk µ(Ck ∩ Ωn ) = (1 − ε) γk µ(Ck ),
n→∞ Ω n→∞
k=1 k=1

since Ωn ↑ Ω implies that µ(Ck ∩ Ωn ) → µ(Ck ) by Theorem 1.13. Hence one


concludes Z Z
lim fn dµ ≥ (1 − ε) g dµ.
n→∞ Ω Ω
R R
Since 0 < ε < 1 is arbitrary, it follows that limn→∞ Ω fn dµ ≥ Ω g dµ, and the
proof is complete. 

Recall that any nonnegative measurable function can be approximated by a


monotonically increasing sequence of simple functions; cf. Theorem 3.19. This
leads to a useful direct corollary concerning the completion of a measure space; cf.
Theorem 3.28.
Corollary 4.7. Let (Ω, A, µ) be a measure space and let (Ω, Ā, µ̄) be its com-
pletion. Let the extended real-valued function f : (Ω, A) → (R, B) be measurable
and nonnegative. Then, in addition to f being measurable with respect to (Ω, Ā),
one has
Z Z
(4.1) f dµ = f dµ̄,
Ω Ω

and the equality is understood in [0, ∞].


Proof. Let fn : (Ω, A) → (R, B) be a sequence of nonnegative simple functions
such that fn ↑ f ; cf. Theorem 3.19. Then it follows from Lemma 4.6 that
Z Z
fn dµ ↑ f dµ.
Ω Ω

However, f and fn are also measurable with respect to Ā; cf. Theorem 3.28.
Therefore an application of Lemma 4.6 in the context of (Ω, Ā, µ̄) shows that
Z Z
fn dµ̄ ↑ f dµ̄.
Ω Ω
1. INTEGRALS OF NONNEGATIVE MEASURABLE FUNCTIONS 49

It follows from Definition 4.1 and Definition 1.36 that


Z Z
fn dµ̄ = fn dµ,
Ω Ω
which leads to (4.1). 
The approximation of a nonnegative measurable extended real-valued function
by simple functions and Lemma 4.6 provide a natural way to extend the results of
Proposition 4.3.
Proposition 4.8. Let (Ω, A, µ) be a measure space. Let the extended real-
valued functions f, g : (Ω, A) → (R, B) be measurable and nonnegative and let
α ≥ 0.
R R
(1) RΩ αf dµ = α ΩRf dµ, α ≥R 0;
(2) Ω (f + g)Rdµ = Ω f Rdµ + Ω g dµ;
(3) f ≤ g ⇒ Ω f dµ ≤ Ω g dµ.
Proof. Approximate the functions f and g by simple functions fn and gn as
in Theorem 3.19. Apply Lemma 4.6 and Proposition 4.3. 
In particular, it is of interest to consider the approximation of the extended
real-valued function ∞A in Example 3.20 by a sequence of simple functions, as it
has a useful consequence.
Example 4.9. Let A ⊂ Ω be a measurable set. Then according to Example
3.20 the sequence fn = n1A satisfies fn ↑ ∞A . Hence by Lemma 4.6
Z Z
nµ(A) = fn dµ ↑ ∞A dµ.
Ω Ω
Therefore the integral of the function ∞A in (3.1) is given by:
Z (
0, if µ(A) = 0,
(4.2) ∞A dµ =
Ω ∞, if µ(A) > 0.
Corollary 4.10. Let (Ω, A, µ) be a measure space and let the extended real-
valued function f : (Ω, A) → (R, B) be measurable and nonnegative. Then
Z
f = 0 a.e. ⇔ f dµ = 0.

Proof. Define the subset N ⊂ Ω by


N = {ω ∈ Ω : f (ω) > 0}.
Since f is measurable, it follows that N is measurable. The statement f = 0 a.e.
is equivalent to µ(N ) = 0; cf. Lemma 3.24.
(⇒) Assume that µ(N ) = 0. It is clear that 0 ≤ f ≤ ∞N and it follows from
Proposition 4.8 and (4.2) that
Z Z
0≤ f dµ ≤ ∞N dµ = 0.
Ω Ω
R
This gives Ω f dµ = 0.
(⇐) Define the measurable sets
An = {ω ∈ Ω : f (ω) > 1/n}.
50 4. INTEGRABILITY OF FUNCTIONS

Note that An ↑ N . Hence, by Proposition 1.13, one obtains µ(An ) ↑ µ(N ). It


follows from Z Z
1 1
µ(An ) = 1An dµ ≤ f dµ = 0,
n Ω n Ω
that µ(An ) = 0, and hence, that µ(N ) = 0. 
Corollary 4.11. Let (Ω, A, µ) be a measure space and let the extended real-
valued functions f, g : (Ω, A) → (R, B) be measurable and nonnegative. Then
Z Z
f = g a.e. ⇒ f dµ = g dµ.
Ω Ω
Proof. Define the set N by
N = {ω ∈ Ω : f (ω) 6= g(ω)}.
Since f and g are measurable it follows that N is measurable. The statement f = g
a.e. is equivalent to µ(N ) = 0; cf. Lemma
R 3.24. R
Assume that µ(N ) = 0. Hence Ω f 1N dµ = 0 and Ω g1N dµ = 0 by Corollary
4.10. Note that Z Z
f 1Ω\N dµ = g1Ω\N dµ,
R R Ω Ω
which leads to Ω f dµ = Ω g dµ. 
Corollary 4.12. Let (Ω, A, µ) be a measure space and let the extended real-
valued function f : (Ω, A) → (R, B) be measurable and nonnegative. Then
Z
f dµ < ∞ ⇒ µ({ω ∈ Ω : f (ω) = ∞}) = 0.

Proof. The proof is complete by showing the equivalent statement
Z
µ({ω ∈ Ω : f (ω) = ∞}) > 0 ⇒ f dµ = ∞.

Let A = {ω ∈ Ω : f (ω) = ∞}, then 0 ≤ ∞A ≤ f . It follows from Proposition 4.8
and (4.2) that Z Z
∞= ∞A dµ ≤ f dµ ≤ ∞.
Ω Ω
This gives the statement. 

2. Integrable functions
Let the extended real-valued function f : (Ω, A) → (R, B) be measurable. Then
the functions f + and f − are nonnegative measurable functions and
(4.3) f = f + − f −, |f | = f + + f − .
Definition 4.13. Let (Ω, A, µ) be a measure space and let the extended real-
valued function f : (Ω, A) → (R, B) be measurable. Then the integral Ω f dµ is
R

defined by Z Z Z
f dµ = f + dµ − f − dµ,
Ω Ω Ω
if at least one of the integrals in the right-hand side is finite. The function f is said
to be integrable if both integrals in the right-hand side are finite, or, equivalently, if
Z
|f | dµ < ∞.

2. INTEGRABLE FUNCTIONS 51

Proposition 4.14. Let (Ω, A, µ) be a measure space, let the extended real-
valued functions f, g : (Ω, A) → (R, B) be integrable, and let α ∈ R. Then αf and
f + g (when defined) are also integrable and
R R
(1) RΩ αf dµ = α ΩRf dµ, α ∈RR;
(2) Ω (f + g)Rdµ = Ω f Rdµ + Ω f dµ;
(3) fR≤ g ⇒ ΩRf dµ ≤ Ω g dµ;
(4) | Ω f dµ| ≤ Ω |f | dµ.
Proof. Items (1), (2), and (3) follow from Definition 4.13 and Proposition 4.8
as applied to f + , g + , f − , and g − . For (4) observe that
Z Z Z Z Z Z
+ − + −
f dµ = f dµ − f dµ ≤ f dµ + f dµ = |f | dµ,
Ω Ω Ω Ω Ω Ω
due to (4.3). 
Proposition 4.15. Let (Ω, A, µ) be a measure space and let the extended real-
valued functions f, g : (Ω, A) → (R, B) beRmeasurable.
R Assume that f integrable
and that g = f a.e. Then g integrable and Ω g dµ = Ω f dµ.
Proof. Assume that f is integrable. It follows from g = f a.e. that
g + = f + a.e. and g − = f − a.e.
By Corollary 4.11 it follows that
Z Z Z Z
g + dµ = f + dµ and g − dµ = f − dµ,
Ω Ω Ω Ω
and each of the integrals is finite. Hence g is integrable and the result follows. 
Corollary 4.16. Let (Ω, A, µ) be a measure space and let the extended real-
valued function f : (Ω, A) → (R, B) be integrable. Then
(4.4) µ({ω ∈ Ω : |f (ω)| = ∞}) = 0,
and there exists an integrable real-valued function g : (Ω, A) → (R, B) such that
Z Z
(4.5) g = f a.e. and g dµ = f dµ.
Ω Ω
Proof. Let N = {ω ∈ Ω : |f (ω)| = ∞}, then it follows from Corollary 4.12
that µ(N ) = 0, so that (4.4) holds. Now define the function g by
g = f 1Ω\N .
Then g is measurable
R and gR= f a.e. Hence it follows from Proposition 4.15 that g
is integrable and Ω g dµ = Ω g dµ. 
In Theorem 3.28 the measurability of a function was studied in terms of the
completion of a measure space. A similar question is now treated for the integra-
bility of a function. Corollary 4.7 will play an important role.
Theorem 4.17. Let (Ω, A, µ) be a measure space and let (Ω, Ā, µ̄) be its com-
pletion.
(1) Let the extended real-valued function f : (Ω, A, µ) → (R, B) be integrable.
Then f is also integrable with respect to (Ω, Ā, µ̄), and
Z Z
(4.6) f dµ = f dµ̄.
Ω Ω
52 4. INTEGRABILITY OF FUNCTIONS

(2) Let the extended real-valued function f : (Ω, Ā) → (R, B) be integrable
with respect to µ̄. Then the the extended real-valued function g : (Ω, A) →
(R, B) in Theorem 3.28 is integrable with respect to µ, and
Z Z
(4.7) g dµ = f dµ̄.
Ω Ω

Proof. (1) Let f : (Ω, A, µ) → (R, B) be integrable. Then by definition f +


and f − are integrable with respect to µ. According to Corollary 4.7
Z Z Z Z

+
f dµ = +
f dµ̄, f dµ = f − dµ̄.
Ω Ω Ω Ω
+ −
so that f and f are also integrable with respect to µ̄. Hence f is integrable with
respect to µ̄ and (4.6) holds.
(2) According to Theorem 3.28 the function g is measurable with respect to A
and g = f almost everywhere with respect to µ̄. By Proposition 4.15 it follows that
g is integrable with respect to µ̄, and that
Z Z
g dµ̄ = f dµ̄.
Ω Ω
Since g is measurable with respect to A, it follows from Corollary 4.7 that
Z Z Z Z

+
g dµ = +
g dµ̄, g dµ = g − dµ̄.
Ω Ω Ω Ω

The integrability of g with respect to µ̄ implies that g + and g + are integrable with
respect to µ̄. Then the above identies show that g + and g + are integrable with
respect to µ. Therefore it follows that g is integrable with respect to µ and that
Z Z
g dµ = g dµ̄.
Ω Ω

The identity (4.7) is now clear. 

Definition 4.18. Let (Ω, A, µ) be a measure space and let the extended real-
valued function f : (Ω, A) → (C, B) be measurable. Then the function f is said to
be integrableR if the real part f1 and and the imaginary part f2 are integrable and
the integral Ω f dµ is defined as
Z Z Z
f dµ = f1 dµ + i f2 dµ, f = f1 + if2 .
Ω Ω Ω

It follows from the following trivial inequalities


max(|f1 |, |f2 |) ≤ |f | ≤ |f1 | + |f2 |,
that the integrability of f is equivalent to the condition
Z
|f | dµ < ∞.

Note that the properties (1), (2), and (4) in Proposition 4.14 remain valid with the
understanding that now in (1) α ∈ C. To see that the inequality in (4) remains
valid, let
Z Z

f dµ = e f dµ,
Ω Ω
3. INTEGRATION WITH RESPECT TO LEBESGUE MEASURE 53

with a suitable θ ∈ R. Then


Z Z Z Z Z
f dµ = eiθ f dµ = Re eiθ f dµ = Re [eiθ f ] dµ ≤ |eiθ f | dµ.
Ω Ω Ω Ω Ω

It will be explicitly mentioned whenever complex-valued functions are integrated.


R
The integral Ω f dµ is often denoted in different ways, showing the “integration
variable” explicitly: Z Z
f (ω) dµ(ω), f (ω) µ(dω).
Ω Ω
This notation will be wuite convenient when several measures are considered simul-
taneously.
For a measurable set A ⊂ Ω one defines
Z Z
(4.8) f dµ = 1A f dµ,
A Ω
so that with disjoint measurable sets A and B one has
Z Z Z Z
f dµ + f dµ = [1A + 1B ]f dµ = f dµ.
A B Ω A∪B

In (4.8) it is assumed that the function f is defined on Ω and is then restricted to


A ⊂ Ω. Of course if f is only given on A then it can be trivially extended to Ω.

3. Integration with respect to Lebesgue measure


In order to apply the previous definitions to integrate a function f : Rd → R
over Rd one needs to provide Rd with a σ-algebra of measurable sets. Assume
that the space Rd is provided with the Lebesgue σ-algebra Md . Then a Lebesgue
measurable function in Md is integrable when
Z
|f | dm < ∞,
Rd

where m is Lebesgue measure on Md . If the space Rd is provided with the Borel


σ-algebra Bd , then a Borel measurable function in Bd is integrable when
Z
|f | dmB < ∞,
Rd

where mB denotes the restriction of the Lebesgue measure to Bd ⊂ Md . Recall


that Md is the completion of Bd ; cf. Theorem 2.19. By Theorem 4.17 a function f
which is (Rd , Bd , mB ) integrable is also (Rd , Md , m) integrable and
Z Z
f dmB = f dm.
Rd Rd

Conversely, if a function f is (R , M , m) integrable, then there exists a function g


d d

which is (Rd , Bd , mB ) integrable, such that


m({x ∈ Rd : g(x) 6= f (x)}) = 0,
i.e. g = f almost everywhere with respect to m, and
Z Z
g dmB = f dm,
Rd Rd
54 4. INTEGRABILITY OF FUNCTIONS

see Theorem 4.17 and Theorem 3.28. In the following a function f : Rd → R will
be called Lebesgue integrable over Rd if it is integrable with respect to (Rd , Md , m)
or with respect to (Rd , Bd , mB ). Except for sets of Lebesgue measure 0 there is no
difference in the two approaches and no distinction will be made between m and
mB when there is no need. Note that for functions f : Rd → R which are defined
only on a subset of Rd similar considerations are valid when the notion of trace
σ-algebra is introduced; cf. Proposition 1.39.
Any continuous function f : Rd → R is Borel measurable (and hence Lebesgue
measurable). If the continuous function f : Rd → R has compact support, then f
is clearly Borel measurable. Furthermore the function f is integrable. To see this
let the support K ⊂ Rd of f be compact, then
Z Z Z
|f | dm = |f | dm ≤ M dm = M m(K) < ∞,
Rd K K

where M = supx∈K |f (x)|. The actual calculation of an integral involving Lebesgue


measure on Rd needs some explanation. Later in Chapter 5 and in Chapter 7 the
higher dimensional cases will be treated.
Here it suffices to consider the case d = 1. Let f : [a, b] → R be continuous
where a ≤ b. The set [a, b] is provided
R with the trace Lebesgue σ-algebra of R; cf.
Proposition 1.39. The integral [a,b] f dm is usually denoted by
Z b Z b
f (t) dm(t) or f (t) dt,
a a
or variations thereof. If b < a, then one defines
Z b Z a
f (t) dt = − f (t) dt.
a b
It is also customary to use the notation
Z x Z b
f (t) dt = f (t)1[a,x] (t) dt, x ∈ [a, b],
a a
so that Z x Z y Z x
f (t) dt − f (t) dt = f (t) dt.
a a y

Lemma 4.19. Let the real-valued function f : [a, b] → R be continuous, then f


is Lebesgue integrable.
(1) If the function F : [a, b] → R is defined by
Z x
F (x) = f (t) dt, x ∈ [a, b],
a

then F is continuously differentiable on [a, b] and F 0 = f .


(2) If the function F : [a, b] → R is continuously differentiable on [a, b], such
that F 0 = f , then
Z b
f (t) dt = F (b) − F (a).
a

Proof. Since the function f is continuous, it is bounded on [a, b]. Hence the
Lebesgue integrability is clear.
4. CONVERGENCE THEOREMS 55

(1) Let x ∈ (a, b). Then for every ε > 0 there exists δ > 0 so that |t − x| < δ
implies that |f (t) − f (x)| < ε. For 0 < h < δ consider
Z x+h Z x !
F (x + h) − F (x) 1
− f (x) = f (t) dt − f (t) dt − f (x)
h h a a

1 x+h
Z
= [f (t) − f (x)] dt,
h x
since [a, x] and (x, x + h] are disjoint measurable sets. If −δ < h < 0, the last
relation holds analogously. Note that
Z x+h Z x+h
[f (t) − f (x)] dt ≤ |f (t) − f (x)| dt ≤ |h|ε.
x x

Consequently,
F (x + h) − F (x)
− f (x) ≤ ε.
h
Hence F 0 (x) = f (x). A similar argument holds for the endpoints.
R x (2) Let F be continuously differentiable on [a, b] with F 0 = f . Define G(x) =
0
a
f (t) dt. Then by (1) G = f , so that G − F is constant on [a, b]. In particular,
Rb
F (b) − F (a) = G(b) − G(a) = a f (t) dt. 
The lemma shows that as far as the Lebesgue integral of a continuous function
is concerned one obtains the value via the difference of a primitive at the end points.
Some more rules of calculation are as before.
Corollary 4.20. Let the real-valued functions h, k : [a, b] → R be continuously
differentiable, then
Z b Z b
0
h(t)k (t) dt = h(b)k(b) − h(a)k(a) − h0 (t)k(t) dt.
a a

Proof. Apply Lemma 4.19 to the function F = hk. Then F is continuously


differentiable and F 0 = f = hk 0 + h0 k. 
Corollary 4.21. Let the real-valued function ϕ : [a, b] → R be continuously
differentiable with ϕ0 > 0, and let the real-valued function f : [ϕ(a), ϕ(b)] → R be
continuous. Then (f ◦ ϕ) ϕ0 is Lebesgue integrable on [a, b] and
Z b Z ϕ(b)
(f ◦ ϕ)(x)ϕ0 (x) dx = f (y) dy.
a ϕ(a)

Proof. Note that F ◦ ϕ is continuously differentiable with derivative (f ◦ ϕ)ϕ0


and apply Lemma 4.19. 
The main theorem of calculus is contained in Lemma 4.19. In Chapter 6 one
returns to this topic by replacing continuity with Lebesgue integrability.

4. Convergence theorems
Let (Ω, A, µ) be a measure space. Let fn be a sequence of integrable functions
which
R converges
R pointwise to a measurable function f . In order to conclude that

fn dµ → Ω
f dµ, i.e., to interchange limit and integration, one needs conditions
to ensure the convergence of the integrals and the limiting identity.
56 4. INTEGRABILITY OF FUNCTIONS

Example 4.22. Let [a, b] ⊂ R be a compact interval and let fn : [a, b] → R


be a sequence of continuous functions, which converge uniformly to a (necessarily
continuous) function f : [a, b] → R. Since all functions involved are Lebesgue
measurable and integrable, one has
Z b Z b
lim fn dm = lim fn dm,
a n→∞ n→∞ a
which follows from
Z b Z b Z b
f dm − fn dm ≤ |f − fn | dm.
a a a

Hence limit and integration may be interchanged.


Example 4.23. Let the sequence of functions fn : R → R be defined by
1
fn = 1[−n,n] .
n
Then the measurable fn converges uniformly on R to the function f , defined by
f (x) = 0, x ∈ R. Note that
Z Z
lim fn dm = 0, fn dm = 2,
R n→∞ R
so that limit and integration may not be interchanged.
Example 4.24. Let the sequence of functions fn : R → R be defined by
fn = 1[−n,n] .
Then the measurable functions fn converge monotonically on R to the function f ,
defined by f (x) = 1, x ∈ R. Note that
Z Z
lim fn dm = ∞, fn dm = 2n → ∞,
R n→∞ R
so that limit and integration may be interchanged (when allowing ∞ = ∞).
Example 4.25. Let the sequence of functions fn : [0, 1] → R be defined by
fn (x) = xn , x ∈ [0, 1].
Then the contimuous functions fn converge pointwise, but not uniformly, on [0, 1]
to the measurable function f , defined by f (x) = 0, 0 ≤ x < 1, and f (1) = 1. Note
that Z 1 Z 1
lim fn dm = lim fn dm,
0 n→∞ n→∞ 0
so that limit and integration may be interchanged.
Theorem 4.26 (Monotone convergence theorem). Let (Ω, A, µ) be a measure
space. Let fn , n ∈ N, and f be nonnegative measurable extended real-valued func-
tions on Ω, such that
fn (ω) → f (ω), fn (ω) ≤ fn+1 (ω), ω ∈ Ω.
Then Z Z
lim fn dµ = f dµ.
n→∞ Ω Ω
4. CONVERGENCE THEOREMS 57

Proof. For n ≤ m it follows from Proposition 4.8 that


Z Z Z
fn dµ ≤ fm dµ ≤ f dµ.
Ω Ω Ω
R
Hence limn→∞ f dµ exists in [0, ∞] and it follows from that
Ω n
Z Z
lim fn dµ ≤ f dµ.
n→∞ Ω Ω
For the reverse inequality, it suffices to show that for any nonnegative simple
function g with g ≤ f one has
Z Z
g dµ ≤ lim fn dµ,
Ω n→∞ Ω
as this, by Definition 4.5, implies
Z Z
f dµ ≤ lim fn dµ.
Ω n→∞ Ω
In fact, the same arguments as in the proof of Lemma 4.6 can be used, although
now the functions fn need not be simple. 

By means of the monotone convergence theorem, the result in Lemma 4.19 can
be extended to the situation where I ⊂ R is a half-open and half-closed interval
and f : I → R is nonnegative and continuous.
Lemma 4.27 (Lebesgue integration on the real line). Let f : (a, b] → R or
f : [a, b) → R be nonnegative and continuous with primitive F . Then
Z Z
f dm = lim [F (b) − F (α)] or f dm = lim [F (β) − F (a)],
(a,b] α&a [a,b) β%b

respectively.
Proof. It has been shown that for a ≤ β < b one has
Z
f dm = F (β) − F (a).
[a,β]

According to the monotone convergence theorem one has


Z Z Z
f dm = f 1[a,β] dm → f dm as β → b.
[a,β] [a,b) [a,b)

The other case can be shown similarly. 

Example 4.28. The following nonnegative continuous real-valued functions f


are Lebesgue integrable on the indicated interval:
(1) [1, ∞), f (x) = xα , α < −1;
(2) (0, 1], f (x) = xα , α > −1;
(3) [1, ∞), f (x) = e−αx , α > 0.
Theorem 4.29 (Fatou’s lemma). Let (Ω, A, µ) be a measure space. Let fn ,
n ∈ N, be nonnegative measurable extended real-valued functions on Ω, then
Z Z
(lim inf fn ) dµ ≤ lim inf fn dµ.
Ω n→∞ n→∞ Ω
58 4. INTEGRABILITY OF FUNCTIONS

Proof. The functions


gn = inf {fn , fn+1 , fn+2 , . . . }
are measurable, gn ≤ fn , gn ≤ gn+1 , and
gn ↑ lim inf fm as n → ∞.
m→∞
Now apply the monotone convergence theorem:
Z Z Z
(lim inf fm ) dµ = ( lim gn ) dµ = lim gn dµ
Ω m→∞ Ω n→∞ n→∞ Ω
Z Z
= lim inf gn dµ ≤ lim inf fn dµ,
n→∞ Ω m→∞ Ω
which completes the proof. 
The following result is the dominated convergence theorem. With obvious
modifications it remains valid for complex-valued functions.
Theorem 4.30 (Dominated convergence theorem). Let (Ω, A, µ) be a measure
space. Let fn , n ∈ N, and f be measurable extended real-valued functions on Ω,
and let g : Ω → [0, ∞] be an integrable extended real-valued function, such that
(4.9) fn (ω) → f (ω), |fn (ω)| ≤ g(ω), for almost all ω ∈ Ω.
Then fn and f are integrable and
Z Z Z
(4.10) lim |fn − f | dµ = 0, lim fn dµ = f dµ.
n→∞ Ω n→∞ Ω Ω

Proof. It follows from |f | ≤ g that f is integrable. Note that |fn − f | ≤ 2g.


The functions 2g − |fn − f | are measurable and nonnegative, and it follows from
Fatou’s lemma that
Z Z
lim inf (2g − |fn − f |) dµ ≤ lim inf (2g − |fn − f |) dµ.
Ω n→∞ n→∞ Ω
R
Since fn → f , the left-hand equals Ω 2g dµ, so that
Z Z
2g dµ ≤ lim inf (2g − |fn − f |) dµ
Ω n→∞ Ω
Z Z
= 2g dµ + lim inf (−|fn − f |) dµ
Ω n→∞ Ω
Z Z
= 2g dµ − lim sup |fn − f | dµ.
Ω n→∞ Ω
R
Since Ω
g dµ is finite, it follows that
Z
lim sup |fn − f | dµ ≤ 0.
n→∞ Ω
This gives the first equality in (4.10); the second equality in (4.10) follows from the
first one. 
If the convergence in (4.9) is pointwise everywhere, then it follows from the
measurability of fn that f is measurable. If the convergence in (4.9) is pointwise
almost everywhere and the measure space (Ω, A, µ) is complete, then the measur-
ability of fn implies that f is measurable; cf. Proposition 3.27.
4. CONVERGENCE THEOREMS 59

Example 4.31. As an application of the dominated convergence theorem con-


sider the sequence of functions fn : [0, 1] → R given by
n sin x
fn (x) = √ , x ∈ [0, 1].
1 + n2 x
These functions are continuous and hence measurable. Further note that
lim fn (x) = 0, x ∈ [0, 1],
n→∞
so that the pointwise limit is also measurable. It is clear that for all n ∈ N
n 1
|fn (x)| ≤ √ ≤ √ , 0 < x ≤ 1,
1 + n2 x x
which shows the existence of an integrable majorant; cf. Example 4.28. Hence by
Theorem 4.30 Z 1
n sin x
lim √ dx = 0.
n→∞ 0 1 + n2 x

It is useful to reformulate the results concerning limits and integration in the


context of series and integration.
Proposition 4.32. Let (Ω, A, µ) be a measure space. Let fn be nonnegative
measurable extended real-valued functions on Ω, then
∞ ∞ Z
Z !
X X
fn dµ = fn dµ.
Ω n=1 n=1 Ω

Proof. For every N ∈ N one has by Proposition 4.8


Z N
! N Z
X X
fn dµ = fn dµ.
Ω n=1 n=1 Ω

Now apply the monotone convergence theorem. 

S∞ Corollary 4.33. Let (Ω, A, µ) be a measure space and assume that Ω =


n=1 Ωn with disjoint sets Ωn ∈ A. Let f be a nonnegative measurable extended
real-valued function on Ω. Then
Z ∞ Z
X
f dµ = f dµ.
Ω n=1 Ωn

Proof. Observe that f can be written as a series:



X
f= fn , fn = f 1Ωn .
n=1
Now apply Proposition 4.32. 
The following result is the series version of the dominated convergence theorem.
With obvious modifications it remains valid for complex-valued functions.
Theorem 4.34. Let (Ω, A, µ) be a measure space. Let fn , n ∈ N, be measurable
extended real-valued functions on Ω, such that
∞ Z
X
(4.11) |fn | dµ < ∞.
n=1 Ω
60 4. INTEGRABILITY OF FUNCTIONS

P∞
Then the series n=1 fn converges almost everyhere to an integrable extended real-
valued function and
∞ ∞ Z
Z !
X X
(4.12) fn dµ = fn dµ.
Ω n=1 n=1 Ω

Proof. In order to apply the dominated convergence theorem, define the fol-
lowing functions
N
X ∞
X
(4.13) gN = fn , h= |fn |.
n=1 n=1

Then clearly |gN | ≤ h. By Proposition 4.32 and the assumption (4.11) one has
Z ∞ Z
X
h dµ = |fn | dµ < ∞,
Ω n=1 Ω

and it follows from Corollary


P∞ 4.16 that (0 ≤) h < ∞ almost everywhere. Hence,
by (4.13), gN → f = n=1 fn almost everywhere (absolute convergence of a series
implying convergence of that series). By the dominated convergence theorem,
∞ Z
X XN Z Z Z
fn dµ = lim fn dµ = lim gN dµ = lim gN dµ,
Ω N →∞ Ω N →∞ Ω Ω N →∞
n=1 n=1

which gives the desired result (4.12). 

Example 4.35. Define the continuous functions fn : (0, 1) → R by


fn (x) = nxn−1 − (n + 1)xn , x ∈ (0, 1).
Then it is easy to see that
∞ ∞ Z
Z 1 X !
X 1
fn dm = 1, fn dm = 0.
0 n=1 n=1 0

Furthermore note by direct calculation that


∞ Z 1 ∞  n
X X n 1
|fn | dm = 2 = ∞.
n=1 0 n=1
n + 1 n + 1

Example 4.36. The function f : [0, ∞) → R defined by


sin t
f (t) = , t > 0, f (0) = 1,
t
is continuous on [0, ∞) and, hence, measurable on [0, ∞). Therefore it is clear that
f is integrable on each interval [0, R], R > 0. However, f is not Lebesgue integrable
on [0, ∞), as can been from the estimates
Z (2k+1)π Z 2kπ
sin t 2 sin t 2
dt ≥ , dt ≥ , k ∈ N,
2kπ t (2k + 1)π (2k−1)π t 2kπ
so that
Z ∞ ∞ Z kπ ∞
sin t X sin t 2X1
dt = dt ≥ = ∞,
0 t k t π n=1 n
k=0
5. PARAMETER DEPENDENT INTEGRALS 61

cf. Proposition 4.32. However one can see, for instance via the calculus of residues,
that Z R
sin t π
lim dt = .
R→∞ 0 t 2
Note that the mere existence of the limit is easily deduced by integration by parts
on the interval [1, R] and the dominated convergence theorem.

5. Parameter dependent integrals


The dominated convergence theorem gives some useful results about integrals
which depend on a parameter. With obvious modifications they remain valid for
complex-valued functions.
Theorem 4.37. Let (Ω, A, µ) be a complete measure space and let I ⊂ R be an
open interval with x0 ∈ I. Let f : I × Ω → R be a real-valued function, such that
(i) for every x ∈ I the function ω 7→ f (x, ω) is integrable;
(ii) for almost all ω ∈ Ω the function x 7→ f (x, ω) is continuous at x0 ∈ I;
(iii) there exists an integrable real-valued function g : Ω → R such that for
every x ∈ I
|f (x, ω)| ≤ |g(ω)|
holds almost everywhere on Ω.
Then the real-valued function F : I → R defined by
Z
F (x) = f (x, ω) dµ(ω), x ∈ I,

is continuous at x0 ∈ I.
Proof. Note that by (i) the function F is well-defined. In order to show that
F is continuous at x0 , it suffices to demonstrate that for any sequence (xn ) in I
with xn → x0 one has
F (xn ) → F (x0 ) as n → ∞.
For this purpose define
fn (ω) = f (xn , ω), f0 (ω) = f (x0 , ω), ω ∈ Ω.
Then (ii) implies that for almost all ω ∈ Ω
fn (ω) → f0 (ω),
while (i) and (iii) imply that the integrable functions fn are dominated by the
integrable function g. Hence by the dominated convergence theorem
Z Z
lim F (xn ) = lim fn dµ = f0 dµ = F (x0 ).
n→∞ n→∞ Ω Ω
Thus the function F is continuous at x0 . 
The next result is about “differentiation under the integral sign”, which is often
a very useful tool.
Theorem 4.38. Let (Ω, A, µ) be a complete measure space and let I ⊂ R be an
open interval with x0 ∈ I. Let f : I × Ω → R be a real-valued function, such that
(i) for every x ∈ I the function ω 7→ f (x, ω) is integrable;
(ii) for almost all ω ∈ Ω the function x 7→ f (x, ω) is differentiable at x0 ∈ I;
62 4. INTEGRABILITY OF FUNCTIONS

(iii) there exists an integrable function g : Ω → R such that for every x ∈ I



f (x, ω) ≤ |g(ω)|
∂x
holds almost everywhere on Ω.
Then the real-valued function F : I → R defined by
Z
F (x) = f (x, ω) dµ(ω), x ∈ I,

is differentiable at x0 ∈ I and
Z
F 0 (x0 ) = (∂/∂x)|x0 f (x, ω) dµ(ω).

Proof. From (i) and (ii) one sees that the function F is continuous at x0 ;
cf. Theorem 4.37. In order to show that F is differentiable at x0 , let (hn ) be any
sequence in R \ {0} with hn → 0, and define
f (x0 + hn , ω) − f (x0 , ω)
fn (ω) = , f0 (ω) = (∂/∂x)|x0 f (x, ω), ω ∈ Ω.
hn
Then (ii) implies with the mean value theorem that for almost all ω ∈ Ω there
exists a number θn (ω) with |θn (ω)| ≤ |hn | such that
fn (ω) = (∂/∂x)|x0 +θn (ω) f (x, ω).
By (i) the functions fn are integrable, and by (iii) the functions fn are dominated by
the integrable function g for almost all ω ∈ Ω. Hence by the dominated convergence
theorem Z Z
0
F (x0 ) = lim fn dµ = f0 (ω) dµ.
n→∞ Ω Ω R
Hence F is differentiable at x0 and its derivative is given by Ω f0 dµ. 

6. Densities and transformations of measures


Let (Ω, A, µ) be a measure space. There are various ways to construct a new
measure by means of µ. For instance, by multiplication with a nonnegative measur-
able function or by composition with the inverse of a measurable mapping (pulling
back). The question is how integrals over the new measure can be expressed in
terms of integrals over the original measure.
Theorem 4.39. Let (Ω, A, µ) be a measure space and let the extended real-
valued function h : (Ω, A) → (R, B) be nonnegative and measurable. Then λ defined
by
Z
(4.14) λ(E) = h dµ, E ∈ A,
E

is a measure on A. Let f : (Ω, A) → (R, B) be a measurable extended real-valued


function. Then the following statements hold.
(1) If the function f is nonnegative, then
Z Z
(4.15) f dλ = f h dµ.
Ω Ω
(2) The function f is integrable with respect to λ if and only if the function
f h is integrable with respect to µ, in which case (4.15) holds.
6. DENSITIES AND TRANSFORMATIONS OF MEASURES 63

Proof. It is clear that λ is well defined, that λ(E) ≥ 0, E ∈ A, and that


λ(∅) = 0.S To see that λ is a measure, let An ∈ A be a sequence of disjoint sets and

let A = n=1 An . Then by Proposition 4.32:

Z Z Z !
X
λ(A) = h dµ = 1A h dµ = 1An h dµ
A Ω Ω n=1
∞ Z
X ∞ Z
X ∞
X
= 1An h dµ = h dµ = λ(An ),
n=1 Ω n=1 An n=1

which shows that λ is a measure.


(1) For f = 1A , A ∈ A, the statement coincides with (4.14). By linearity
the result follows for simple functions. A limiting procedure extends the result to
nonnegative measurable functions.
(2) For every measurable function f it follows from (1) that
Z Z
|f | dλ = |f |h dµ ∈ [0, ∞].
Ω Ω

Hence f is integrable with respect to λ if and only if f h is integrable with respect


to µ. In this case the equality in (4.15) is obtained via the components f + and f − ,
separately. 

The function h in Theorem 4.39 is sometimes called the density of λ with


respect to µ. The new measure λ has the property that
A ∈ A, µ(A) = 0 ⇒ λ(A) = 0,
R
since 1A h = 0 almost everywhere with respect to µ and, thus, λ(A) = Ω 1A h dµ =
0. Moreover, if 0 ≤ h ≤ 1 µ-a.e., then λ(E) ≤ µ(E), E ∈ A. Further properties
will be studied in the context of absolute continuity.

A measurable mapping between measurable spaces will induce a new measure


on the range space, as shown in the following result.
Theorem 4.40. Let (Ω, A) and (Ω0 , A0 ) be measurable spaces and assume that
ϕ : (Ω, A) → (Ω0 , A0 ) is a measurable mapping. If µ is a measure on (Ω, A), then ν
defined by
(4.16) ν(B) = µ(ϕ−1 (B)), B ∈ A0 ,

is a measure on (Ω0 , A0 ). Let the extended real-valued function f : (Ω0 , A0 ) → (R, B)


be measurable. Then the following statements hold.
(1) If the function f is nonnegative, then
Z Z
(4.17) f dν = (f ◦ ϕ) dµ.
Ω0 Ω

(2) The function f is integrable with respect to ν if and only if the function
f ◦ ϕ is integrable with respect to µ, in which case (4.17) holds.
Proof. Since ϕ is measurable the expression in (4.16) is well-defined. Clearly
ν(B) ≥ 0 for all B ∈ A0 and ν(∅) = 0. So it remains to check the countable
64 4. INTEGRABILITY OF FUNCTIONS

additivity of ν. Let Bn be a sequence of pairwise disjoint sets in A0 , then ϕ−1 (Bn )


is a sequence of pairwise disjoint sets in A, and
∞ ∞ ∞
! !! !
[ [ [
−1 −1
ν Bn = µ ϕ Bn =µ ϕ (Bn )
n=1 n=1 n=1

X ∞
X
= µ(ϕ−1 (Bn )) = ν(Bn ).
n=1 n=1
Hence ν is a measure on (Ω , A ).
0 0

(1) For f = 1B , B ∈ A0 , the statement coincides with (4.16). By linearity


the result follows for simple functions. A limiting procedure extends the result to
nonnegative measurable functions.
(2) For every measurable function f it follows from (1) that
Z Z
|f | dν = |f ◦ ϕ| dµ.
Ω0 Ω
Hence f is integrable with respect to λ if and only if f h is integrable with respect
to µ. In this case the equality in (4.17) is obtained via the components f + and f − ,
separately. 
The measure ν on (Ω0 , A0 ) as given in (4.16) is called the induced measure. It
is sometimes denoted by ν = µ ◦ ϕ−1 . In particular if ϕ : (Ω, A) → (Ω, A), then ϕ
is said to be measure preserving if
µ(B) = µ(ϕ−1 (B)), B ∈ A.
Measure preserving mappings arise naturally in physics and other applications.
CHAPTER 5

Product measures

The area of a rectangle in the plane is the product of the length of the sides.
This is the basis for the introduction of two-dimensional Lebesgue measure. Like-
wise the double integral of a continuous function over a rectangle can be computed
as two iterated one-dimensional integrals. The question is now if two-dimensional
Lebesgue measure can be seen as the product of one-dimensional Lebesgue mea-
sures. In a more general context, it can be shown that the product of two measure
spaces provides a product measure, so that an integral of a function on the product
space can be computed as iterated integrals over the individual components of the
product space. However it turns out that the product measure is in general not
complete. This chapter gives a concise introduction to product measures and their
completions.

1. Construction of product measures


The product Ω of sets Ω1 and Ω2 is defined by
Ω = Ω1 × Ω2 = {(ω1 , ω2 ) : ω1 ∈ Ω1 , ω2 ∈ Ω2 }.
Let A1 be a σ-algebra on Ω1 and let A2 be a σ-algebra on Ω2 . Define the collection
G of products of measurable sets by
(5.1) G = {A1 × A2 : A1 ∈ A1 , A2 ∈ A2 }.
Note that for all A1 , B1 ∈ A1 and A2 , B2 ∈ A2 :
(A1 × A2 ) ∩ (B1 × B2 ) = (A1 ∩ B1 ) × (A2 ∩ B2 ),
and that A1 ∩ B1 ∈ A1 and A2 ∩ B2 ∈ A2 . Hence G is closed under intersections.
Definition 5.1. Let A1 be a σ-algebra on Ω1 and let A2 be a σ-algebra on
Ω2 . The σ-algebra on Ω generated by G in (5.1) is called the product σ-algebra
A = A1 ⊗ A2 on Ω1 × Ω2 , so that A1 ⊗ A2 = σ(G).
Since the generating set G is closed under intersections it follows that σ(G) =
d(G), i.e. the σ-algebra and the Dynkin system generated by G coincide, cf. Theo-
rem 1.31.
Definition 5.2. Let A be a subset of Ω = Ω1 × Ω2 . For ω1 ∈ Ω1 and ω2 ∈ Ω2
the sections Aω1 and Aω2 are defined as subsets of Ω2 and Ω1 , respectively, by
Aω1 = {ω2 ∈ Ω2 : (ω1 , ω2 ) ∈ A},
and
Aω2 = {ω1 ∈ Ω1 : (ω1 , ω2 ) ∈ A}.

65
66 5. PRODUCT MEASURES

In particular, note that for A = A1 × A2 ∈ G the corresponding sections are


given by
( (
A2 , ω1 ∈ A1 , ω2 A1 , ω 2 ∈ A2 ,
(5.2) Aω1 = A =
∅, ω1 6∈ A1 , ∅, ω2 6∈ A2 ,
which are measurable in A2 and A1 , respectively. In fact, for any measurable set
in A1 ⊗ A2 its sections are measurable.
Proposition 5.3. Let A ∈ A1 ⊗ A2 . Then
(1) ω1 ∈ Ω1 ⇒ Aω1 ∈ A2 ;
(2) ω2 ∈ Ω2 ⇒ Aω2 ∈ A1 .
Proof. It suffices to show (1). For ω1 ∈ Ω1 define
F = {A ∈ A1 ⊗ A2 : Aω1 ∈ A2 }.
From (5.2) it is clear that G ⊂ F, which implies
A1 ⊗ A2 = σ(G) ⊂ σ(F).
Hence, in order to prove (1), it suffices to show that σ(F) = F or, equivalently, that
F is a σ-algebra.
It will now be shown that F is a σ-algebra. It is clear that Ω ∈ F. If A ∈ F,
then Aω1 ∈ A2 , which gives
(Ac )ω1 = (Aω1 )c ∈ A2 ,

so that Ac ∈ F. Now let A = n=1 An with An ∈ F. Then (An )ω1 ∈ A2 and hence
S


[
Aω1 = (An )ω1 ∈ A2 . 
n=1

Definition 5.4. Let f : Ω1 × Ω2 → R be an extended real-valued function. For


ω1 ∈ Ω1 and ω2 ∈ Ω2 the sections fω1 : Ω2 → R and f ω2 : Ω1 → R are defined by
fω1 (ω2 ) = f (ω1 , ω2 ), ω2 ∈ Ω2 ,
and
f ω2 (ω1 ) = f (ω1 , ω2 ), ω1 ∈ Ω1 .
Note the following connection between sections of subsets in Ω1 × Ω2 and sec-
tions of characteristic functions:
(5.3) (1A )ω1 = 1Aω1 , (1A )ω2 = 1Aω2 .
Thus the sections of the characteristic function of a measurable set are measurable
by Proposition 5.3. In fact, for any A1 ⊗ A2 -measurable function its sections are
measurable.
Proposition 5.5. Let f : (Ω1 × Ω2 , A1 ⊗ A2 ) → (R, B) be an extended real-
valued measurable function. Then
(1) for ω1 ∈ Ω1 the function fω1 : (Ω2 , A2 ) → (R, B) is measurable;
(2) for ω2 ∈ Ω1 the function f ω2 : (Ω1 , A1 ) → (R, B) is measurable.
Proof. The statement is clear for characteristic functions and, by linearity,
for simple functions. By a limiting procedure the case of nonnegative measurable
functions follows, which gives the result for all measurable functions. 
1. CONSTRUCTION OF PRODUCT MEASURES 67

Let f : Ω1 × Ω2 → R be an extended real-valued function. It is straightforward


to see the useful identities
(5.4) (f + )ω1 = (fω1 )+ and (f − )ω1 = (fω1 )− .
In particular, if f is measurable, then also f + and f − are measurable. By Propo-
sition 5.5 it follows that all functions in (5.4) are measurable.

Let (Ω1 , A1 , µ1 ) and (Ω2 , A2 , µ2 ) be measure spaces. It will now be shown


how in certain cases a product measure on A1 ⊗ A2 can be constructed. The
construction of the product measure rests on the following simple observation. Let
A = A1 × A2 ∈ G, then it follows from (5.2) that
µ2 (Aω1 ) = µ2 (A2 )1A1 (ω1 ), ω1 ∈ Ω1 ,
(5.5) ω2
µ1 (A ) = µ1 (A1 )1A2 (ω2 ), ω2 ∈ Ω2 .
These formulas show that the functions ω1 7→ µ2 (Aω1 ) and ω2 7→ µ1 (Aω2 ) are
measurable with respect to A1 and A2 , respectively.
Now let A ∈ A1 ⊗ A2 , then by Proposition 5.3 Aω1 ∈ A2 and Aω2 ∈ A1 , so that
µ2 (Aω1 ) and µ1 (Aω2 ) are well-defined. In order to show that these expressions are
measurable functions of ω1 and ω2 , respectively, it will be assumed that µ1 and µ2
are σ-finite.
Proposition 5.6. Let (Ω1 , A1 , µ1 ) and (Ω2 , A2 , µ2 ) be σ-finite measure spaces
and let A ∈ A1 ⊗ A2 . Then
(1) the function ω1 7→ µ2 (Aω1 ) is measurable with respect to A1 ;
(2) the function ω2 7→ µ1 (Aω2 ) is measurable with respect to A2 .
Proof. It suffices to show (1). Note that (1) is concerned with the σ-algebra
A1 ⊗ A2 and only the measure µ2 on the measurable space (Ω2 , A2 ). Define the set
F = {A ∈ A1 ⊗ A2 : ω1 7→ µ2 (Aω1 ) measurable}.
From (5.5) it is clear that G ⊂ F, which implies
A1 ⊗ A2 = σ(G) = d(G) ⊂ d(F).
Hence, in order to prove (1), it suffices to show that d(F) = F or, equivalently, that
F is is a Dynkin system. It will now be shown that F is a Dynkin system for the
case that µ2 is finite and for the case that µ2 is σ-finite.
First assume that µ2 is finite. Then it is clear from (5.5) that Ω ∈ F. If A ∈ F,
then
µ2 ((Ac )ω1 ) = µ2 ((Aω1 )c ) = µ2 (Ω2 ) − µ2 (Aω1 ),
S∞
which is A1 -measurable in ω1 . Now let A = n=1 An with An ∈ F pairwise disjoint.
Then (An )ω1 ∈ A2 and
[∞
Aω1 = (An )ω1 ∈ A2 with (An )ω1 pairwise disjoint.
n=1
Hence
∞ ∞
!
[ X
µ2 (Aω1 ) = µ2 (An )ω1 = µ2 ((An )ω1 ),
n=1 n=1
and thus ω1 7→ µ2 (Aω1 ) is measurable with respect to A1 (being the sum of non-
negative measurable functions). This implies that A ∈ F and F is a Dynkin system.
Hence if µ2 is finite, the function ω1 7→ µ2 (Aω1 ) is measurable with respect to A1 .
68 5. PRODUCT MEASURES

Next assume that µ2 is σ-finite. Then there exists a sequence Ω2,n ∈ A2 such
that
µ2 (Ω2,n ) < ∞, Ω2,n ↑ Ω2 .
Recall from Proposition 1.18 that
µ2,n (A2 ) = µ2 (A2 ∩ Ω2,n ), A2 ∈ A2 ,
defines a finite measure µ2,n : A2 → [0, ∞] on Ω2 with the property
µ2,n (A2 ) ↑ µ2 (A2 ), A2 ∈ A2 .
By what has just been shown it follows that for every A ∈ A1 ⊗ A2 the function
ω1 7→ µ2,n (Aω1 ),
is A1 -measurable. Now observe that
µ2,n (Aω1 ) ↑ µ2 (Aω1 ),
so that the function in the right-hand side, as the pointwise limit of A1 -measurable
functions, is A1 -measurable itself. 
The following main result gives the definition of the product measure and shows
in what sense it is unique. In fact the theorem itself is the first form of the Fubini-
Tonelli theorems.
Theorem 5.7 (Product measure). Let (Ω1 , A1 , µ1 ) and (Ω2 , A2 , µ2 ) be σ-finite
measure spaces and let (Ω1 × Ω2 , A1 ⊗ A2 ) be the corresponding measurable product
space. Then
Z Z
(5.6) µ(A) = µ2 (Aω1 )µ1 (dω1 ) = µ1 (Aω2 )µ2 (dω2 ), A ∈ A1 ⊗ A1 ,
Ω1 Ω2

defines a σ-finite measure µ = µ1 ⊗ µ2 on A1 ⊗ A1 . Moreover µ = µ1 ⊗ µ2 is the


only measure on A1 ⊗ A1 which satisfies
(5.7) µ(A1 × A2 ) = µ1 (A1 )µ2 (A2 ), A1 ∈ A1 , A2 ∈ A2 .
Proof. By Proposition 5.6 the nonnegative integrands in (5.6) are measurable,
so both integrals are well-defined. It will be shown that µ = µ1 ⊗ µ2 defined by
the first integral in (5.6) is a σ-finite measure on A1 ⊗ A2 and that it is the only
measure A1 ⊗ A2 which satisfies (5.7). By symmetry it is clear that the second
integral in (5.6) also defines a σ-finite measure on A1 ⊗ A2 which satisfies (5.7). By
uniqueness it then follows that the two integrals in (5.6) are equal.
Define the function µ : A1 ⊗ A2 → [0, ∞] by
Z
(5.8) µ(A) = µ2 (Aω1 )µ1 (dω1 ), A ∈ A1 ⊗ A2 .
Ω1
S∞
Clearly µ(∅) = 0. Now let An ∈ A1 ⊗ A2 be pairwise disjoint and let A = n=1 An .
Then

[
Aω1 = (An )ω1 with (An )ω1 pairwise disjoint.
n=1
Therefore

X
µ2 (Aω1 ) = µ2 ((An )ω1 ).
n=1
1. CONSTRUCTION OF PRODUCT MEASURES 69

It follows from (5.8) and Proposition 4.32 that


Z Z X∞
µ(A) = µ2 (Aω1 )µ1 (dω1 ) = [µ2 ((An )ω1 )] µ1 (dω1 )
Ω1 Ω1 n=1
∞ Z
X ∞
X
= [µ2 ((An )ω1 )] µ1 (dω1 ) = µ(An ).
n=1 Ω1 n=1
Thus µ is σ-additive and, hence, µ is a measure on A1 ⊗ A2 . Note that for A =
A1 × A2 ∈ G it follows from (5.8) and (5.5) that
Z
µ(A1 × A2 ) = µ2 (A2 )1A1 (ω1 )µ1 (dω1 ) = µ1 (A1 )µ2 (A2 ),
Ω1
so that µ satisfies (5.7).
Since µ1 and µ2 are σ-finite, there are sequences Ω1,n ∈ A1 and Ω2,n ∈ A2 such
that
µ1 (Ω1,n ) < ∞, Ω1,n ↑ Ω1 and µ2 (Ω2,n ) < ∞, Ω2,n ↑ Ω2 .
Note that by definition Ω1,n × Ω2,n ∈ G and that Ω1,n × Ω2,n ↑ Ω, while
µ(Ω1,n × Ω2,n ) = µ1 (Ω1,n )µ2 (Ω2,n ) < ∞.
Hence also the product measure µ = µ1 ⊗ µ2 is σ-finite. Therefore, by Corollary
1.33, it follows that the product measure on A1 ⊗ A2 with the property (5.7) is
uniquely determined. 
Corollary 5.8. Let (Ω1 , A1 , µ1 ) and (Ω2 , A2 , µ2 ) be σ-finite measure spaces
with product measure µ1 ⊗ µ2 , and let A ∈ A1 ⊗ A2 . Then the following statements
are equivalent:
(1) (µ1 ⊗ µ2 )(A) = 0;
(2) µ2 (Aω1 ) = 0 for µ1 -almost all ω1 ∈ Ω1 ;
(3) µ1 (Aω2 ) = 0 for µ2 -almost all ω2 ∈ Ω2 .
Proof. By symmetry it suffices to show (1) ⇔ (2). Recall that the product
measure µ = µ1 ⊗ µ2 is given by
Z
µ(A) = µ2 (Aω1 )µ1 (dω1 ), A ∈ A1 ⊗ A2 .
Ω1
Hence (2) ⇒ (1) is clear. To show (1) ⇒ (2) assume µ(A) = 0. Since the integrand
is nonnegative (and measurable), it follows from Corollary 4.9 that µ2 (Aω1 ) = 0 for
µ1 -almost all ω1 ∈ Ω1 . 
As a final remark one notes that the present construction of a product measure
is based on the product σ-algebra A1 ⊗ A2 ; the product measure µ1 ⊗ µ2 , defined
on the σ-algebra A1 ⊗ A2 , is the only measure on A1 ⊗ A2 which satisfies (5.7).
Even if µ1 and µ2 are complete measures the product measure µ1 ⊗ µ2 is in general
not complete.
Example 5.9. Assume that A1 ⊂ Ω1 is a non-measurable set and let A2 ∈ A2
with µ(A2 ) = 0. Then clearly
A1 × A2 ⊂ Ω 1 × A2 ,
where Ω1 × A2 ∈ A1 ⊗ A2 and (µ1 ⊗ µ2 )(Ω1 × A2 ) = 0. Note that A1 × A2 is not
A1 ⊗A2 measurable (otherwise also its sections would be measurable by Proposition
5.3). This implies that A1 ⊗ A2 is not complete.
70 5. PRODUCT MEASURES

2. Fubini-Tonelli theorems
The unique product measure as given in Theorem 5.7 is the basis for the Fubini-
Tonelli theorems. The key ingredient is the formula in (5.6) for a measurable set
A ∈ A1 ⊗ A1 , which can be rewritten as an identity with iterated integrals:
Z Z Z 
1A d(µ1 ⊗ µ2 ) = 1Aω1 d µ2 µ1 (dω1 )
Ω1 ×Ω2 Ω Ω
Z 1 Z 2 
= 1A 2 d µ1 µ2 (dω2 ),
ω
Ω2 Ω1

involving sections of sets. From now on these identities will be used in the following
equivalent form involving sections of functions:
Z Z Z 
1A d(µ1 ⊗ µ2 ) = (1A )ω1 dµ2 dµ1 (ω1 )
Ω1 ×Ω2 Ω Ω
(5.9) Z 1 Z 2 
ω2
= (1A ) dµ1 dµ2 (ω2 ),
Ω2 Ω1

cf. (5.3). Recall that any measurable function f : (Ω1 × Ω2 , A1 ⊗ A2 ) → (R, B) has
measurable sections; cf. Proposition 5.5.
Theorem 5.10 (Fubini-Tonelli). Let (Ω1 , A1 , µ1 ) and (Ω2 , A2 , µ2 ) be σ-finite
measure spaces. Assume that f : (Ω1 × Ω2 , A1 ⊗ A2 ) → (R, B) is a measurable
nonnegative extended real-valued function. Then
(1) the function ω1 7→ Ω2 fω1 dµ2 is nonnegative and A1 -measurable;
R

(2) the function ω2 7→ Ω1 f ω2 dµ1 is nonnegative and A2 -measurable.


R

Moreover,
Z Z Z 
f d(µ1 ⊗ µ2 ) = fω1 dµ2 dµ1 (ω1 )
Ω1 ×Ω2 Ω Ω
(5.10) Z 1 Z 2 
= f ω2 dµ1 dµ2 (ω2 ),
Ω2 Ω1

and the equalities are understood in [0, ∞].


Proof. Due to (5.9) the statements in (5.10) are clear for characteristic func-
tions of measurable sets. By linearity these statements remain valid for simple
functions. Now let f : (Ω1 × Ω2 , A1 ⊗ A2 ) → (R, B) be a nonnegative measurable
function and let fn be a sequence of nonnegative simple functions with fn ↑ f .
Then by Theorem 4.26
Z Z
(5.11) fn d(µ1 ⊗ µ2 ) ↑ f d(µ1 ⊗ µ2 ).
Ω1 ×Ω2 Ω1 ×Ω2

It will now be shown that this limit is also equal to


Z Z 
fω1 dµ2 dµ1 (ω1 ),
Ω1 Ω2

which proves the first equality in (5.10); the other equality is clear by symmetry.
It is clear from the pointwise limiting relation fn ↑ f that also (fn )ω1 ↑ fω1 for
any ω1 ∈ Ω1 . By Proposition 5.5 sections of measurable functions are measurable,
2. FUBINI-TONELLI THEOREMS 71

and it follows for each ω1 ∈ Ω1 that


Z Z
(5.12) (fn )ω1 dµ2 ↑ fω1 dµ2 ,
Ω2 Ω2

again by Theorem 4.26. The final step is integration of the last result (5.12) over
Ω1 , which is allowed when the functions in the left-hand side and in the right-hand
side are measurable in ω1 . To see that these functions are measurable observe that
the identity
Z Z
(1A )ω1 dµ2 = 1Aω1 dµ2 = µ2 (Aω1 )
Ω2 Ω2

defines an A1 -measurable function of ω1 ; cf. Proposition 5.6. Hence, as a finite


sum of such functions, also the function
Z
ω1 7→ (fn )ω1 dµ2
Ω2

is A1 -measurable. This implies that also the function


Z
ω1 7→ fω1 dµ2
Ω2

in (1) is A1 -measurable (being a monotone limit of A1 -measurable functions).


Therefore, Theorem 4.26 may be applied to (5.12), so that
Z Z  Z Z 
(5.13) (fn )ω1 dµ2 dµ1 (ω1 ) ↑ fω1 dµ2 dµ1 (ω1 ). 
Ω1 Ω2 Ω1 Ω2

Corollary 5.11 (Fubini-Tonelli). Let (Ω1 , A1 , µ1 ) and (Ω2 , A2 , µ2 ) be σ-finite


measure spaces. Assume that f : (Ω1 × Ω2 , A1 ⊗ A2 ) → (R, B) is a measurable
extended real-valued function. The following integrals are all equal:
Z Z Z 
|f | d(µ1 ⊗ µ2 ), |fω1 | dµ2 dµ1 (ω1 ),
Ω1 ×Ω2 Ω1 Ω2

and
Z Z 
ω2
|f | dµ1 dµ2 (ω2 ).
Ω2 Ω1

In particular, they are simultaneously finite.

Proof. Assume that f : (Ω1 × Ω2 , A1 ⊗ A2 ) → (R, B) is measurable. Then


in the decomposition f = f + − f − the nonnegative functions f + and f − are
measurable. By Theorem 5.10 one has for these components
Z Z Z 
+ +
(5.14) f d(µ1 ⊗ µ2 ) = (f )ω1 dµ2 dµ1 (ω1 ),
Ω1 ×Ω2 Ω1 Ω2

and
Z Z Z 
(5.15) f − d(µ1 ⊗ µ2 ) = (f − )ω1 dµ2 dµ1 (ω1 ).
Ω1 ×Ω2 Ω1 Ω2
72 5. PRODUCT MEASURES

Hence (5.14), (5.15), and a repeated application of Proposition 4.8 lead to


Z
|f | d(µ1 ⊗ µ2 )
Ω1 ×Ω2
Z Z
= f + d(µ1 ⊗ µ2 ) + f − d(µ1 ⊗ µ2 )
Ω ×Ω Ω1 ×Ω2
Z 1 Z2  Z Z 
+ −
= (f )ω1 dµ2 dµ1 (ω1 ) + (f )ω1 dµ2 dµ1 (ω1 )
Ω Ω Ω Ω
Z 1 Z 2  Z 1 Z 2 
+ −
= (fω1 ) dµ2 dµ1 (ω1 ) + (fω1 ) dµ2 dµ1 (ω1 )
Ω Ω Ω1 Ω2
Z 1 Z 2 
= |fω1 )| dµ2 dµ1 (ω1 ),
Ω1 Ω2

where in the penultimate step use has been made of (5.4). Likewise one finds
Z Z Z 
|f | d(µ1 ⊗ µ2 ) = |f ω2 | dµ1 dµ2 (ω1 ).
Ω1 ×Ω2 Ω2 Ω1
This completes the proof. 
Theorem 5.12 (Fubini-Tonelli). Let (Ω1 , A1 , µ1 ) and (Ω2 , A2 , µ2 ) be σ-finite
measure spaces. Assume that f : (Ω1 × Ω2 , A1 ⊗ A2 ) → (R, B) is an integrable
extended real-valued function. Then the set
(5.16) F1 = { ω1 ∈ Ω1 : fω1 not integrable with respect to µ2 }
has µ1 -measure 0, and the set
(5.17) F2 = { ω2 ∈ Ω2 : f ω2 not integrable with respect to µ1 }
has µ2 -measure 0. Furthermore
Z Z Z 
f d(µ1 ⊗ µ2 ) = fω1 dµ2 dµ1 (ω1 )
Ω1 ×Ω2 Ω1 \F1 Ω2
Z Z 
= f ω2 dµ1 dµ2 (ω2 ).
Ω2 \F2 Ω1

Proof. Assume that f : (Ω1 ×Ω2 , A1 ⊗A2 ) → (R, B) is an integrable function.


Then in the decomposition f = f + − f − the nonnegative functions f + and f −
are integrable. Hence it follows from (5.14), (5.15), and Corollary 4.16 that the
following set
 Z   Z 
+ −
(5.18) ω1 ∈ Ω1 : (f )ω1 dµ2 = ∞ ∪ ω1 ∈ Ω1 : (f )ω1 dµ2 = ∞
Ω2 Ω2

has µ1 -measure 0. Due to (5.4) it is clear that the set in (5.18) is equal to the set
F1 in (5.16).
By definition the integral of the integrable function f is given by
Z Z Z
f d(µ1 ⊗ µ2 ) = f + d(µ1 ⊗ µ2 ) − f − d(µ1 ⊗ µ2 ),
Ω1 ×Ω2 Ω1 ×Ω2 Ω1 ×Ω2

and by (5.14) and (5.15) the right-hand side is equal to


Z Z  Z Z 
(f + )ω1 dµ2 dµ1 (ω1 ) − (f − )ω1 dµ2 dµ1 (ω1 ).
Ω1 Ω2 Ω1 Ω2
3. PRODUCT MEASURES ON Rp × Rq 73

As the set F1 ⊂ Ω1 has µ1 -measure 0 this is also equal to


Z Z  Z Z 
+ −
(f )ω1 dµ2 dµ1 (ω1 ) − (f )ω1 dµ2 dµ1 (ω1 ),
Ω1 \F1 Ω2 Ω1 \F1 Ω2

or, equivalently, to
Z Z Z 
(f + )ω1 dµ2 − (f − )ω1 dµ2 dµ1 (ω1 ),
Ω1 \F1 Ω2 Ω2

as the difference of the integrands is well-defined. Due to (5.4) one has


Z Z Z Z
(f + )ω1 dµ2 − (f − )ω1 dµ2 = (fω1 )+ dµ2 − (fω1 )− dµ2 .
Ω2 Ω2 Ω2 Ω2

Since each of the integrals in the right-hand side is finite for ω1 ∈ Ω1 \ F1 , the
right-hand side is equal by definition to
Z
fω1 dµ2 .
Ω2

This proves the first identity in the theorem. The proof of the second identity is
similar. 

3. Product measures on Rp × Rq
The notions of Borel measurable sets and Lebesgue measurable sets with their
corresponding measures have been developed for the space Rd . Again for conve-
nience the notations Md and Bd will be used for the corresponding σ-algebras; the
Lebesgue measure in these cases is denoted by md . Now the products of these
notions will be considered.
Proposition 5.13 (Product of Borel algebras). The product Bp ⊗ Bq of the
σ-algebras Bp and Bq satisfies
(5.19) Bp ⊗ Bq = Bp+q .
Moreover, the product measure mp ⊗ mq on Bp ⊗ Bq is well-defined and satisfies
(5.20) mp ⊗ mq = mp+q ,
where mp+q is Lebesgue measure on Bp+q .
Proof. First it will be shown that Bp ⊗ Bq ⊂ Bp+q . Let A1 ∈ Bp and
A2 ∈ Bq , and let π1 and π2 be the canonical projections from Rp+q to Rp and
to Rq , respectively. These projections are continuous and thus Borel measurable.
Therefore it follows from
A1 × A2 = (A1 × Rq ) ∩ (Rp × A2 ) = π1−1 (A1 ) ∩ π2−1 (A2 ),
that A1 × A2 is a Borel set in Rp+q . Therefore
Bp ⊗ Bq ⊂ Bp+q .
For the reverse inclusion, recall that Bp+q is generated by intervals of the form
A1 × A2 with A1 ⊂ Rp and A2 ∈ Rq , and note that A1 × A2 ∈ Bp ⊗ Bq . Since
Bp ⊗ Bq is a σ-algebra, one obtains Bp+q ⊂ Bp ⊗ Bq . This establishes (5.19).
To see that the product measure mp ⊗ mq is well-defined on Bp ⊗ Bq it suf-
fices to note that mp and mq are σ-finite measures on the σ-algebras Ap and
74 5. PRODUCT MEASURES

Aq , respectively; cf. Theorem 5.7. Note that on sets of the form A × B where
A = [a1 , b1 ] × · · · × [ap , bp ] and B = [c1 , d1 ] × · · · × [cq , dq ] the measures coincide:
(mp ⊗ mq )(A × B) = mp (A)mq (B) = mp+q (A × B).
Now apply Proposition 1.24 with Corollary 1.33. 

Let Kp ⊂ Rp and Kq ⊂ Rq be compact and let f : Kp ×Kq → R be continuous.


Since Kp × Kq ⊂ Rp+q is compact it follows that f is integrable. Hence the
interchange of the order of integration is always allowed, so that
Z Z Z !
f dmp+q = f (x, y) dmq (y) dmp (x)
Kp ×Kq Kp Kq
Z Z !
= f (x, y) dmp (x) dmq (y).
Kq Kp

Note that with abuse of language these identities are frequently encountered in the
following form:
Z Z Z ! Z Z !
f dxdy = f (x, y) dy dx = f (x, y) dx dy,
Kp ×Kq Kp Kq Kq Kp

which is alright as long as one knows what one is doing. The following result is a
useful variation as one often integrates over a subset A ⊂ Kp × Kq which is Borel
measurable. Recall that
Z Z
f dmp+q = (f 1A ) dmp+q .
A Kp ×Kq

Corollary 5.14 (Elementary version of the Fubini-Tonelli theorem). Let K ⊂


Rp × Rq be compact and let f : K → R be a continuous real-valued function. Let
A ⊂ K be Borel measurable then
Z Z Z !
f dmp+q = (f 1A )(x, y) dmq (y) dmp (x)
A Kp Kq
Z Z !
= (f 1A )(x, y) dmp (x) dmq (y).
Kq Kp

The version of the Fubini-Tonelli theorem in Corollary 5.14 is the one that is
often presented in calculus texts.

In general, even if the repeated integrals of a function f with respect to the mea-
sures mp and mq exist, the integrability of f with respect to the product measure
mp+q is not guaranteed. The following simple example shows this.
Example 5.15. On the square (0, 1) × (0, 1) consider the function
x2 − y 2
f (x, y) = .
(x2 + y 2 )2
Clearly the function f is continuous and hence it is Borel measurable. Provide the
horizontal and vertical axis with the Borel σ-algebra and consider the square with
3. PRODUCT MEASURES ON Rp × Rq 75

the product measure. Observe that the sections of f are also Borel measurable and
that Z 1 Z 1
1 1
f (x, y) dm(x) = − 2 , f (x, y) dm(y) = 2 .
0 y + 1 0 x +1
Hence repeated integration gives
Z 1 Z 1  Z 1 Z 1 
π π
f (x, y) dm(y) dm(x) = , f (x, y) dm(x) dm(y) = − ,
0 0 4 0 0 4
so that the order of integration matters. Clearly the above function f is not inte-
grable on the square, i.e.,
Z
|f (x, y)| dm2 (x, y) = ∞,
(0,1)×(0,1)

which follows from Theorem 5.12.


Note that none of the measure spaces in the identity (5.19) is complete. The
situation is different for Lebesgue measure spaces. Both Mp and Mq are complete
measure spaces but the corresponding product measure space Mp ⊗Mq is not, while
the measure space Mp+q is complete.
Proposition 5.16 (Product of Lebesgue algebras). The product Mp ⊗ Mq of
the σ-algebras Mp and Mq satisfies
(5.21) Mp ⊗ Mq = Mp+q .
Moreover, the product measure mp ⊗ mq on Mp ⊗ Mq is well-defined and satisfies
mp ⊗ mq = mp+q ,
where mp+q is Lebesgue measure on Mp+q .
Proof. First it will be shown that Mp ⊗ Mq ⊂ Mp+q . For this, it suffices to
demonstrate that
A1 ∈ Mp , A2 ∈ Mq ⇒ A1 × A2 ∈ Mp+q .
To see this, let A1 ∈ Mp and A2 ∈ Mq .
First assume that A1 and A2 are bounded. Choose 0 < ε < 1. Then there exist
compact sets K1 , K2 and bounded open sets O1 , O2 , such that
K1 ⊂ A1 ⊂ O1 , K2 ⊂ A2 ⊂ O2 ,
cf. Corollary 2.22, and
ε ε
mp (O1 \ K1 ) ≤ , mq (O2 \ K2 ) ≤ .
2(mq (A2 ) + 1) 2(mp (A1 ) + 1)
Note that O1 \ A1 ⊂ O1 \ K1 so that mp (O1 \ A1 ) ≤ mp (O1 \ K1 ), which implies
ε
mp (O1 ) ≤ mp (A1 ) + ≤ mp (A1 ) + 1.
2(mq (A2 ) + 1)
Likewise it follows that
ε
mq (O2 ) ≤ mq (A2 ) + ≤ mq (A2 ) + 1.
2(mp (A1 ) + 1)
Now observe that K1 × K2 ⊂ A1 × A2 ⊂ O1 × O2 , with K1 × K2 compact and
O1 × O2 open, and that
(O1 × O2 ) \ (K1 × K2 ) ⊂ (O1 \ K1 ) × O2 ) ∪ (O1 × (O2 \ K2 )).
76 5. PRODUCT MEASURES

Since all these sets are open, hence Lebesgue measurable in Rp+q , one obtains
mp+q ((O1 × O2 ) \ (K1 × K2 ))
≤ mp+q ((O1 \ K1 ) × O2 )) + mp+q ((O1 × (O2 \ K2 )))
= mp (O1 \ K1 )mq (O2 ) + mp (O1 )mq (O2 \ K2 )
≤ ε/2 + ε/2 = ε.

Hence A1 × A2 is Lebesgue measurable in Rp+q .


Now assume that A1 or A2 is is not bounded. Then

[ ∞
[
A1 = A1,n , A2 = A2,n ,
n=1 n=1

where each A1,n and A2,m is bounded and measurable, so that by the earlier argu-
ment

[
A1 × A2 = A1,n × A2,m
n,m=1
p+q
is Lebesgue measurable in R .
Finally, observe that
Bp+q = Bp ⊗ Bq ⊂ Mp ⊗ Mq ⊂ Mp+q .
According to Theorem 2.19, Mp+q is the completion of Bp+q .
To see that the product measure mp ⊗ mq is well-defined on Mp ⊗ Mq it suffices
to note that mp and mq are σ-finite measures on the σ-algebras Mp and Mq ,
respectively; cf. Theorem 5.7. The restriction of mp ⊗ mq to Bp+q = Bp ⊗ Bq
coincides with mp+q . Hence the completion of mp ⊗ mq is mp+q . 

4. Completion of product measures


The product measure µ1 ⊗µ2 of two σ-finite measures µ1 and µ2 is not complete,
even if µ1 and µ2 are complete measures; cf. Example 5.9. Now a version of
the Tonelli-Fubini theorems will be given involving the completion of the product
measure. The new theorem is in essence an interpretation of Theorem 5.12, based
on the following two technical lemmas.
Lemma 5.17. Let (Ω1 , A1 , µ1 ) and (Ω2 , A2 , µ2 ) be σ-finite measure spaces, and
let
(Ω1 × Ω2 , L, λ)
be the completion of the product (Ω1 × Ω2 , A1 ⊗ A2 , µ1 ⊗ µ2 ). Assume that the
extended real-valued function f : (Ω1 × Ω2 , L) → (R, B) is measurable with respect
to λ. Then there is a measurable extended real-valued function g : (Ω, A1 ⊗ A2 ) →
(R, B) such that
(5.22) N = {ω ∈ Ω : f (ω) 6= g(ω)} ∈ L with λ(N ) = 0,
i.e., g = f almost everywhere with respect to λ. Moreover,
Z Z
(5.23) |g| d(µ1 ⊗ µ2 ) = |f | dλ (∈ [0, ∞]).
Ω1 ×Ω2 Ω1 ×Ω2
4. COMPLETION OF PRODUCT MEASURES 77

In particular, if the function f is integrable with respect to λ, then the function g


is integrable with respect to µ1 × µ2 and
Z Z
(5.24) g d(µ1 ⊗ µ2 ) = f dλ.
Ω1 ×Ω2 Ω1 ×Ω2

Proof. Let f : (Ω1 × Ω2 , L) → (R, B) be measurable with respect to λ.


The existence of the function g : (Ω, A1 ⊗ A2 ) → (R, B) such that (5.22) holds is
a consequence of Theorem 3.28. Moreover, the identity (5.23) then follows from
Corollary 4.7. If f is integrable with respect to λ then g is integrable with respect
to µ1 × µ2 due to Theorem 4.17, which also implies that (5.24) holds. 
Lemma 5.18. For the set N ∈ L in (5.22) there exists a set M , measurable
with respect to µ1 ⊗ µ2 , such that
(5.25) N ⊂M and (µ1 ⊗ µ2 )(M ) = 0.
Hence there exist an A1 -measurable set H1 ⊂ Ω1 with µ1 (H1 ) = 0 such that
(5.26) ω 1 ∈ Ω1 \ H 1 ⇒ µ2 (Mω1 ) = µ2 ({ω2 ∈ Ω2 : (ω1 , ω2 ) ∈ M }) = 0,
and an A2 -measurable set H2 ⊂ Ω2 with µ2 (H2 ) = 0 such that
(5.27) ω 2 ∈ Ω 2 \ H2 ⇒ µ1 (M ω2 ) = µ1 ({ω1 ∈ Ω1 : (ω1 , ω2 ) ∈ M }) = 0.
Proof. The measure space (Ω1 × Ω2 , L, λ) is the completion of the product
(Ω1 × Ω2 , A1 ⊗ A2 , µ1 ⊗ µ2 ). The set N ∈ L in (5.22) belongs to L and λ(N ) = 0.
Hence the existence of M , measurable with respect to µ1 ⊗ µ2 , such that (5.25)
holds follows from the definition of completion. As (µ1 ⊗ µ2 )(M ) = 0 the existence
of the sets H1 and H2 is a direct consequence of Corollary 5.8. 
Theorem 5.19. Let (Ω1 , A1 , µ1 ) and (Ω2 , A2 , µ2 ) be σ-finite measure spaces,
and let
(Ω1 × Ω2 , L, λ)
be the completion of the product (Ω1 × Ω2 , A1 ⊗ A2 , µ1 ⊗ µ2 ). Assume that the
extended real-valued function f : (Ω1 × Ω2 , L) → (R, B) is measurable with respect
to λ. Then there exist an A1 -measurable set H1 ⊂ Ω1 with µ1 (H1 ) = 0 and an
A2 -measurable set H2 ⊂ Ω2 with µ2 (H2 ) = 0 such that the following integrals are
all equal: Z Z Z 
|f | dλ, |fω1 | dµ2 dµ1 (ω1 ),
Ω1 ×Ω2 Ω1 \H1 Ω2
and Z Z 
ω2
|f | dµ1 dµ2 (ω2 ).
Ω2 \H2 Ω1
In particular, they are simultaneously finite.
Proof. Let the function f : (Ω1 ×Ω2 , L) → (R, B) be measurable with respect
to L and let the function g : (Ω, A1 ⊗ A2 ) → (R, B) be measurable with respect to
A1 ⊗ A2 as in Lemma 5.17. By Corollary 5.11 the following integrals are all equal:
Z Z Z 
|g| d(µ1 ⊗ µ2 ), |gω1 | dµ2 dµ1 (ω1 ),
Ω1 ×Ω2 Ω1 Ω2
and Z Z 
|g ω2 | dµ1 dµ2 (ω2 ).
Ω2 Ω1
78 5. PRODUCT MEASURES

In particular, they are simultaneously finite. Now each of these integrals will be
written in terms of the function f . It is clear from (5.23) that
Z Z
|g| d(µ1 ⊗ µ2 ) = |f | dλ.
Ω1 ×Ω2 Ω1 ×Ω2

To treat the remaining integrals observe that the function g satisfies (5.22). Asso-
ciated with the set N ∈ L in (5.22) is the set M ∈ A1 ⊗ A2 in (5.25). By Lemma
5.18 there exist an A1 -measurable set H1 ⊂ Ω1 with µ1 (H1 ) = 0 such that (5.26)
holds, and an A2 -measurable set H2 ⊂ Ω2 with µ2 (H2 ) = 0 such that (5.27) holds.
Hence for each ω1 ∈ Ω1 \ H1 one has the identity
Z Z Z Z
|gω1 | dµ2 = |gω1 | dµ2 + |gω1 | dµ2 = |gω1 | dµ2 ,
Ω2 Mω1 Ω2 \Mω1 Ω2 \Mω1

where the last step follows from (5.26). Note that in the last integral (ω1 , ω2 ) 6∈ M
(otherwise it would follow that ω2 ∈ Mω1 ), so that certainly (ω1 , ω2 ) 6∈ N , see
(5.25). Therefore in the last integral one has g(ω) = f (ω) and gω1 = fω1 . Hence
for each ω1 ∈ Ω1 \ H1
Z Z Z
|gω1 | dµ2 = |fω1 | dµ2 = |fω1 | dµ2 .
Ω2 Ω2 \Mω1 Ω2

Observe that
Z Z  Z Z 
|gω1 | dµ2 dµ1 (ω1 ) = |gω1 | dµ2 dµ1 (ω1 )
Ω1 Ω2 Ω1 \H1 Ω2
Z Z 
= |fω1 | dµ2 dµ1 (ω1 ).
Ω1 \H1 Ω2

The remaining integral is treated similarly. 

Theorem 5.20. Let (Ω1 , A1 , µ1 ) and (Ω2 , A2 , µ2 ) be σ-finite measure spaces,


and let
(Ω1 × Ω2 , L, λ)
be the completion of the product (Ω1 × Ω2 , A1 ⊗ A2 , µ1 ⊗ µ2 ). Assume that the
extended real-valued function f : (Ω1 × Ω2 , L) → (R, B) is integrable with respect
to λ. Then there exist subsets L1 ⊂ Ω1 and L2 ⊂ Ω2 with
(1) L1 is A1 -measurable and µ1 (L1 ) = 0;
(2) L2 is A2 -measurable and µ2 (L2 ) = 0,
such that
(3) for all ω1 ∈ Ω1 \ L1 the function fω1 is integrable with respect to µ2 ;
(4) for all ω2 ∈ Ω2 \ L2 the function f ω2 is integrable with respect to µ1 .
Furthermore
Z Z Z 
f dλ = fω1 dµ2 dµ1 (ω1 )
Ω1 ×Ω2 Ω1 \L1 Ω2
Z Z 
= f ω2 dµ1 dµ2 (ω2 ).
Ω2 \L2 Ω1
5. CONVOLUTIONS 79

Proof. Let f : (Ω, L) → (R, B) be integrable with respect to λ. According to


Lemma 5.17 the function g is integrable with respect to µ1 × µ2 and
Z Z
(5.28) f dλ = g d(µ1 ⊗ µ2 ).
Ω1 ×Ω2 Ω1 ×Ω2
Applying Theorem 5.12 to the integral involving the function g one sees that the
set
(5.29) G1 = { ω1 ∈ Ω1 : gω1 not integrable with respect to µ2 }
is A1 -measurable with µ1 -measure 0, and that
Z Z Z 
(5.30) g d(µ1 ⊗ µ2 ) = gω1 dµ2 dµ1 (ω1 ).
Ω1 ×Ω2 Ω1 \G1 Ω2

As in the proof of Theorem 5.19 for each ω1 ∈ Ω1 \ H1 one has the identity
Z Z Z Z
gω1 dµ2 = gω1 dµ2 + gω1 dµ2 = gω1 dµ2 .
Ω2 Mω1 Ω2 \Mω1 Ω2 \Mω1

Observe that in the last integral (ω1 , ω2 ) 6∈ M (otherwise it would follow that
ω2 ∈ Mω1 ), so that certainly (ω1 , ω2 ) 6∈ N , i.e., f (ω) = g(ω) and gω1 = fω1 .
Now note that the union G1 ∪H1 is A1 -measurable and µ1 (G1 ∪H1 ) = 0. Hence
one obtains from (5.28) and (5.30)
Z Z
f dλ = g d(µ1 ⊗ µ2 )
Ω1 ×Ω2 Ω1 ×Ω2
Z Z 
= gω1 dµ2 dµ1 (ω1 )
Ω1 \G1 Ω2
Z Z 
= gω1 dµ2 dµ1 (ω1 )
Ω1 \(G1 ∪H1 ) Ω2
Z Z !
= gω1 dµ2 dµ1 (ω1 )
Ω1 \(G1 ∪H1 ) Ω2 \Mω1
Z Z !
= fω1 dµ2 dµ1 (ω1 )
Ω1 \(G1 ∪H1 ) Ω2 \Mω1
Z Z 
= fω1 dµ2 dµ1 (ω1 ),
Ω1 \(G1 ∪H1 ) Ω2

where in the last step Proposition 4.15 has been used, so that for ω1 ∈ Ω1 \(G1 ∪H1 )
it follows that fω1 is integrable with respect to µ2 . By taking L1 = G1 ∪ H1 the
proof is complete for one case. The other case follows similarly. 

5. Convolutions
The Fubini-Tonelli theorems will be used to derive a few facts concerning con-
volutions. The following general observation is useful.
Lemma 5.21. Let the real-valued functions f, g : Rd → R be Lebesgue measur-
able. Then the real-valued function
(x, y) → f (x)g(y)
from Rd × Rd to R is Lebesgue measurable.
80 5. PRODUCT MEASURES

Proof. Note that the function (x, y) 7→ f (x) is Lebesgue measurable. To see
this observe that for every c ∈ R the set
{(x, y) ∈ Rd × Rd : f (x) ≤ c} = {x ∈ Rd : f (x) ≤ c} × R
is measurable in the product space. Likewise the function (x, y) 7→ g(y) is Lebesgue
measurable. In addition, products of measurable functions are measurable. 
Corollary 5.22. Let the real-valued functions f, g : Rd → R be Lebesgue
measurable. Then the real-valued function
(x, y) → f (x − y)g(y)
is Lebesgue measurable on R × Rd .
d

Proof. According to Lemma 5.21 the function (x, y) 7→ H(x, y) = f (x)g(y) is


Lebesgue measurable. Define T : Rd × Rd → Rd × Rd by
T (x, y) = (x − y, y).
Then T is an invertible linear transformation in R2 , mapping measurable sets onto
measurable sets, cf. Proposition 2.16. Therefore the composition H ◦ T is measur-
able. 
Definition 5.23. Let the real-valued functions f, g : Rd → R be Lebesgue
measurable. The convolution product of f and g is defined by
Z
(5.31) h(x) = f (x − y)g(y) dm(y),
Rd
for all x ∈ R for which the integrand is Lebesgue integrable. The function h is called
the convolution of f and g, denoted by h = f ∗ g.
Note that the integrand is measurable in (x, t) due to Corollary 5.22. The ques-
tion is to determine for which x ∈ Rd the convolution is defined. As an application
of the Fubini-Tonelli theorems one sees the following result.
Theorem 5.24. Let the functions f, g ∈ L(Rd ). Then
Z
|f (x − y)g(y)| dm(y) < ∞,
Rd

for almost all x ∈ Rd . For such x ∈ Rd define the function h = f ∗ g by (5.31).


Then h ∈ L(Rd ) and
Z Z  Z 
|h(x)| dm(x) ≤ |f (x)| dm(x) |g(x)| dm(x) .
Rd Rd Rd

Proof. Observe that


Z Z 
|f (x − y)g(y)| dm(y) dm(x)
Rd Rd
Z Z 
= |g(y)| |f (x − y)| dm(x) dm(y),
Rd Rd
and Z Z
|f (x − y)| dm(x) = |f (x)| dm(x).
Rd Rd
The rest follows from the Fubini-Tonelli theorems. 
5. CONVOLUTIONS 81

Remark 5.25. Often the reasoning in the treatment of convolutions is to re-


place Lebesgue measurable functions by Borel measurable functions almost every-
where. This makes it possible to make use of the Fubini-Tonelli theorems involving
the product measure rather than the version involving the completion of the product
measure.
CHAPTER 6

Integration and differentiation on R

The connection between differentiation and Lebesgue integration in R is an


interesting subject: in general the main theorem of calculus
Z x
f (x) = f (a) + f 0 (t) dt, x ∈ [a, b],
a

for the case that f is differentiable almost everywhere and f 0 is Lebesgue integrable,
is not valid; but what comes in its place is very rewarding not only from a theoretical
point of view but also from an applied point of view. The results in this chapter are
based on the fact that every monotone function is differentiable almost everywhere
with respect to Lebesgue measure and the derivative is Lebesgue integrable on any
compact interval. The main new notions here are functions of bounded variation
and absolutely continuous functions; later they will find an natural place in the
abstract decomposition theory of ordinary and signed or complex measures. There
are also higher-dimensional versions of the topics in this chapter, but they lead too
far away from the present purposes.

1. Vitali covers
Definition 6.1. Let E ⊂ R be a set. A Vitali cover of E is a collection V of
closed intervals with positive length with the property that for every x ∈ E and for
every η > 0 there is an interval I ∈ V with x ∈ I and m(I) < η.
Lemma 6.2. Let E ⊂ R be a set with m∗ (E) < ∞ and let V be a Vitali cover
of E. For each ε > 0 there are disjoint intervals I1 , . . . , IN ∈ V such that
m∗ (E \ (I1 ∪ · · · ∪ IN )) < ε.
Proof. Choose an open set O ⊂ R such that E ⊂ O and m(O) < ∞; cf. ??.
Observe that the collection W defined by
W = {I ∈ V : I ⊂ O}
is a Vitali cover of E. Choose an interval I1 ∈ W and continue by induction in the
following way. Assume that the disjoint intervals I1 , . . . , In have been chosen, then
either E ⊂ I1 ∪ · · · ∪ In and the selection process stops, or E \ (I1 ∪ . . . In ) 6= ∅.
In the last case let kn be the supremum of the lengths of the intervals in W that
have an empty intersection with I1 ∪ · · · ∪ In . Then choose such an interval In+1
with length greater than kn /2. If this process continues one finds mutually disjoint
closed intervals In , n ∈ N, from W such that
∞ ∞
!
X [
(6.1) m(Ik ) = m Ik ≤ m(O) < ∞,
k=1 k=1

83
84 6. INTEGRATION AND DIFFERENTIATION ON R

so that the sum converges. Now let ε > 0, then there exists N ∈ N such that
X∞
m(Ik ) < ε,
k=N +1

and let Jk be a closed interval with the same center as Ik but with five times its
length. It is now claimed that

[
(6.2) E \ (I1 ∪ · · · ∪ IN ) ⊂ Jk .
k=N +1

The claim in (6.2) clearly implies that



!
[
∗ ∗
m (E \ (I1 ∪ · · · ∪ IN )) ≤ m Jk
k=N +1

X ∞
X
≤ m(Jk ) = 5 m(Ik ) < 5ε,
k=N +1 k=N +1

which completes the proof of the lemma.


In order to substantiate the claim, let x ∈ E \ (I1 ∪ · · · ∪ IN ). Since W is a Vitali
cover for E there exists an interval I ∈ W with x ∈ I and such that I has empty
intersection with I1 ∪ · · · ∪ IN . Assume that this interval I has empty intersection
with I1 ∪ · · · ∪ In for all n > N . Then for each n > N the interval I is among all
the intervals in W that could have been picked for In+1 , which means that
m(I) ≤ kn < 2m(In+1 ), n > N,
which implies via (6.1) that m(I) = 0, a contradiction. Hence there is a smallest
integer with I ∩ Im 6= ∅. But note that the above reasoning also gives that
m(I) ≤ km−1 < 2m(Im ),
as I is among all the intervals in W that could have been picked for Im .
Now Im ⊂ Jm , I ∩ Im 6= ∅, and m(I) < 2m(Im ) imply that I ⊂ Jm , which
shows that (6.2) holds. 

2. Differentiability of monotone functions


Let the function f : (a, b) → R be monotone. It is known that f is continuous
except for a countable number of points where f has a jump; cf. Proposition 2.23.
It will now be shown that f is differentiable almost everywhere with respect to
Lebesgue measure.
Assume that the function f : (a, b) → R is nondecreasing. Define the so-called
Dini derivatives of f at the point x ∈ (a, b) by
f (y) − f (x) f (y) − f (x)
D+ f (x) = lim sup , D+ f (x) = lim inf ,
y↓x y−x y↓x y−x
f (y) − f (x) f (y) − f (x)
D− f (x) = lim sup , D− f (x) = lim inf .
y↑x y−x y↑x y−x
It is clear from the definitions that for all x ∈ (a, b) the following inequalities hold
(6.3) 0 ≤ D− f (x) ≤ D− f (x) ≤ ∞, 0 ≤ D+ f (x) ≤ D+ f (x) ≤ ∞.
Moreover, it is clear that the Dini derivatives of f are all Lebesgue measurable.
2. DIFFERENTIABILITY OF MONOTONE FUNCTIONS 85

Theorem 6.3. Let the real-valued function f : (a, b) → R be nondecreasing.


Then one has with respect to Lebesgue measure:
(1) D+ f (x) < ∞ and D− f (x) < ∞ almost everywhere;
(2) D+ f (x) = D+ f (x) and D− f (x) = D− f (x) almost everywhere;
(3) D− f (x) ≤ D+ f (x) and D− f (x) ≥ D+ f (x) except for an at most count-
able number of points.
In particular, f is differentiable almost everywhere.
Proof. Only the first assertion in each item will be proved; the other assertion
can be proved by similar arguments or by a reduction to the first assertion.
(1) It suffices to prove this on intervals of the form (a + 1/n, b − 1/n), since a
countable union of sets of measure 0 has measure zero. Hence it may be assumed
that f is bounded on (a, b), say |f | ≤ C. Let the set E be defined by
E = {x ∈ (a, b) : D+ f (x) = ∞}.
For any s > 0 the collection V of intervals [x, y] with
f (y) − f (x)
x ∈ E, y ∈ (x, b), > s,
y−x
is a Vitali cover for E. Choose ε > 0, then by Lemma 6.2 there exist pairwise
disjoint intervals [x1 , y1 ], . . . , [xN , yN ] in V such that
m∗ (E \ ([x1 , y1 ] ∪ · · · ∪ [xN , yN ])) < ε,
which leads to
m∗ (E) ≤ m∗ (E \ ([x1 , y1 ] ∪ · · · ∪ [xN , yN ])) + m∗ ([x1 , y1 ] ∪ · · · ∪ [xN , yN ])
< ε + (y1 − x1 ) + · · · + (yN − xN ).
Hence it follows that
s(m∗ (E) − ε) < s(y1 − x1 ) + · · · + k(yN − xN )
< f (y1 ) − f (x1 ) + · · · + f (yN ) − f (xN )
≤ f (b) − f (a) ≤ 2C.
If m (E) > 0 one may choose ε > 0 such that m∗ (E) − ε > 0. For s so large

that s(m∗ (E) − ε) > 2C, the above estimate leads to a contradiction. Therefore it
follows that m∗ (E) = 0, which implies that the set E is Lebesgue integrable with
measure 0.
(2) Due to (6.3) it suffices to show that the set F defined by
F = {x ∈ (a, b) : D+ f (x) < D+ f (x)}
is Lebesgue measurable with measure 0. For each x ∈ F there exist rx , sx ∈ Q with
D+ f (x) < rx < sx < D+ f (x).
S
Thus it is clear that F ⊂ r,s∈Q Ers , where
Ers = {x ∈ (a, b) : D+ f (x) < r < s < D+ f (x)}.
In order to show that F has measure 0 it is therefore sufficient to show that for
fixed r, s ∈ Q the set E = Ers has measure 0. Choose ε > 0 and let O ⊂ R be an
open set with
E ⊂ O, m(O) < m∗ (E) + ε;
86 6. INTEGRATION AND DIFFERENTIATION ON R

cf. Lemma 2.18. The collection Vr of intervals [x, y] with


f (y) − f (x)
x ∈ E, y ∈ O, y > x, < r,
y−x
is a Vitali cover for E. By Lemma 6.2 there exist pairwise disjoint intervals
[x1 , y1 ], . . . , [xN , yN ] in Vr such that
m∗ (E \ ([x1 , y1 ] ∪ · · · ∪ [xN , yN ])) < ε.
Note that with these intervals [x1 , y1 ], . . . , [xN , yN ] which nearly cover E one has
N
X N
X
(f (yk ) − f (xk )) < r (yk − xk ).
k=1 k=1

Define U = (x1 , y1 ) ∪ · · · ∪ (xN , yN ) and consider the set E ∩ U . The collection Vs


of intervals [u, v] with
f (v) − f (u)
u ∈ E ∩ U, v > u, [u, v] ⊂ (xk , yk ) for some k, > s,
v−u
is a Vitali cover for E ∩ U . By Lemma 6.2 there exist pairwise disjoint intervals
[u1 , v1 ], . . . , [uM , vM ] in Vs such that
m∗ ((E ∩ U ) \ ([u1 , v1 ] ∪ · · · ∪ [uM , vM ])) < ε.
Note that with these intervals [u1 , v1 ], . . . , [uM , vM ] which nearly cover E ∩ U one
has
M
X XM
(f (vj ) − f (uj )) > s (vj − uj ).
j=1 j=1
Due to the construction it is clear that
XM N
X
(f (vj ) − f (uj )) ≤ (f (yk ) − f (xk )),
j=1 k=1

which leads to
M
X N
X
(6.4) s (vj − uj ) < r (yk − xk ).
j=1 k=1

To majorize the sum in the right-hand side of (6.4) observe that by construction
N
X
(6.5) (yk − xk ) ≤ m(O) < m∗ (E) + ε.
k=1

To minorize the sum in the left-hand side of (6.4) first observe


m∗ (E \ U ) = m∗ (E \ ([x1 , y1 ] ∪ · · · ∪ [xN , yN ])) < ε,
since the sets involved differ only by a finite number of points. Therefore it follows
that
(6.6) m∗ (E) ≤ m∗ (E ∩ U ) + m∗ (E \ U ) < m∗ (E ∩ U ) + ε.
To estimate m∗ (E ∩ U ) observe that
m∗ (E ∩ U ) ≤ m∗ ((E ∩ U ) \ ([u1 , v1 ] ∪ · · · ∪ [uM , vM ]))
+ m∗ ([u1 , v1 ] ∪ · · · ∪ [uM , vM ]),
2. DIFFERENTIABILITY OF MONOTONE FUNCTIONS 87

which shows that


M
X
(6.7) m∗ (E ∩ U ) < ε + (vj − uj ).
j=1

The inequalities in (6.6) and (6.7) lead to


M
X
(6.8) m∗ (E) − 2ε < m∗ (E ∩ U ) − ε < (vj − uj ).
j=1

Hence combining (6.4), (6.5), and (6.8) one obtains


s(m∗ (E) − 2ε) < r(m∗ (E) + ε) or (s − r)m∗ (E) < (2s + r)ε
for any ε > 0. Since s > r the assumption that m∗ (E) > 0 leads to a contradiction.
Therefore it follows that m∗ (E) = 0, which implies that the set E is Lebesgue
measurable with measure 0.
(3) It suffices to show that the set F defined by
F = {x ∈ (a, b) : D− f (x) > D+ f (x)}
is at most countable. For each x ∈ F there exist qx ∈ Q with
D− f (x) > qx > D+ f (x).
S
Thus it is clear that F ⊂ q∈Q Eq , where
Eq = {x ∈ (a, b) : D− f (x) > q > D+ f (x)}.
In order to show that F is countable it is therefore sufficient to show that for
fixed q ∈ Q the set E = Eq is countable. For q ∈ Q define the function g by
g(x) = f (x) − qx. Then for x ∈ Eq it follows that
D− g(x) > 0 > D+ g(x).
Thus the function g has a strict local maximum at x ∈ Eq . The number of points
where g has a strict local maximum is at most countable1. Hence the number of
points in Eq is at most countable. Therefore it follows that the set F is countable.
As a consequence of (1), (2), and (3) the function f is differentiable almost
everywhere, as will be shown now. By (1) and (2) the function f has finite one-
sided derivatives almost everywhere. If these one-sided derivatives are not equal
then either
D− f (x) = D− f (x) > D+ f (x) = D+ f (x),
or
D− f (x) = D− f (x) < D+ f (x) = D+ f (x).
By (3) these inequalities can happen only on a set of measure 0. This completes
the proof. 

1A function h : (a, b) → R is said to have a strict local maximum at c ∈ (a, b) if there exist
pc , qc ∈ R such that for all x with pc < x < qc and x 6= c one has h(x) < h(c). The mapping
ι : c → (pc , qc ) from the points where h has a strict local maximum to their corresponding
intervals is injective. To see this, let h have local maxima at c and d, c 6= d, and assume that
(pc , qc ) = (pd , qd ); this implies that g(d) < g(c) and g(c) < g(d), a contradiction. Without loss
of generality one may take pc , qc ∈ Q, so that ι is an injective mapping into the countable set
Q × Q. Hence it follows that the number of points where h has a strict local maximum is at most
countable.
88 6. INTEGRATION AND DIFFERENTIATION ON R

The following result shows a remarkable feature of Lebesgue integration on R


in connection with differentiation.
Theorem 6.4. Let the real-valued function f : [a, b] → R be nondecreasing.
Then f 0 is Lebesgue integrable and
Z b
f 0 (x) dx ≤ f (b) − f (a).
a

Proof. Define the functions fn : [a, b] → R by


f (x + 1/n) − f (x)
fn (x) = ,
1/n
so that fn ≥ 0 and fn → f 0 almost everywhere. Here one defines f (x) = f (b) for
x > b. In particular, f 0 ≥ 0 is Lebesgue measurable. Now apply Fatou’s lemma
Z b Z b
f 0 (x) dx ≤ lim inf fn (x) dx
a n→∞ a
Z b
= lim inf n [f (x + 1/n) − f (x)] dx
n→∞
"aZ #
b+1/n Z b
= lim inf n f (x) dx − f (x) dx
n→∞ a+1/n a
" Z #
b+1/n Z a+1/n
= lim inf n f (x) dx − n f (x) dx
n→∞ b a

≤ f (b) − f (a),
where the last inequality follows from the monotonicity of f . Hence the Lebesgue
integrability of f 0 follows. 

Example 6.5. Recall the construction of the Cantor set C in Example 2.13.
The Cantor function will be defined as follows. The set C1 is [0, 1] with (1/3, 2/3)
removed. For x ∈ (1/3, 2/3) define f (x) = 1/2. The set C2 is C1 with (1/9, 2/9)
and (7/9, 8, 9) removed. For x ∈ (1/9, 2/9) define f (x) = 1/4 and for x ∈ (7/9, 8/9)
define f (x) = 3/4. Continue by induction. The set Cn−1 is made up of 2n−1
intervals, each of length 1/3n−1 . The set Cn is obtained from Cn−1 by removing
the middle third of each of these intervals. On these deleted intervals define f by
1 3 5 2n − 1
, , , . . . , ,
2n 2n 2n 2n
respectively. Finally the function f is defined on all of [0, 1] by setting f (0) =
0, f (1) = 1, and for x ∈ C \ {0, 1} by setting f (x) = limt↑x f (t). Then f is
nondecreasing and continuous. It follows from Theorem 6.3 that f is differentiable
almost everywhere.
Actually, the construction of f gives more. Let x ∈ [0, 1] \ C, then x belongs to
an open interval where f is constant. Hence f is differentiable at x and f 0 (x) = 0.
Thus f 0 = 0 almost everywhere, so that
Z 1
f 0 (t) dt = 0 < 1 = f (1) − f (0),
0
3. TERMWISE DIFFERENTIATION OF SERIES OF MONOTONE FUNCTIONS 89

i.e., strict inequality may happen in Theorem 6.4. Finally it is claimed that f is
not differentiable on C. Let x ∈ C and assume that f is differentiable at x. Then
for any sequences αn , βn with
αn ↑ x and βn ↓ x,
one must have that the limit
f (βn ) − f (αn )
lim
n→∞ βn − αn
exists in R, as one can see from
 
f (βn ) − f (αn ) 0 f (βn ) − f (x) 0
− f (x) = λn − f (x)
βn − α n βn − x
 
f (αn ) − f (x) 0
+ (1 − λn ) − f (x) ,
αn − x
where λn = (βn − x)/(βn − αn ) is a bounded sequence. However one sees that
for x ∈ C it follows that x ∈ Cn for each n ∈ N. In other words x ∈ [an , bn ]
for some interval [an , bn ] in Cn (which is made up of 2n such intervals), so that
bn − an = 1/3n . Note that by definition
1
f (bn +) − f (an −) = ,
2n
which shows that
f (bn ) − f (an ) 3n
∼ n.
bn − an 2

3. Termwise differentiation of series of monotone functions


The results concerning the differentiability of monotone functions have a very
useful consequence, namely Fubini’s theorem on the termwise differentiation of
series of monotone functions.
Theorem 6.6 (Fubini). Let fk : (a, b) → R be real-valued nondecreasing func-
tions and assume that the series
X∞
f= fk
k=1

converges everywhere on (a, b). Then f is nondecreasing and f 0 is given by



X
f0 = fk0 a.e.
k=1

Proof. It is clear that the sum f is a nondecreasing function and thus f is


differentiable almost everywhere. Write the sum as
n
X ∞
X
f= fk + Rn , Rn = fk ,
k=1 k=n+1

and observe that also Rn is a nondecreasing function which is differentiable almost


everywhere. It clearly suffices to prove the result of a compact interval [a, b]. Let
90 6. INTEGRATION AND DIFFERENTIATION ON R

E ⊂ [a, b] the set of points where f and all fn are differentiable. Due to Theorem
6.3 the set [a, b] \ E has measure 0. For x ∈ E also Rn is differentiable so that
n
X
f 0 (x) = fk0 (x) + Rn0 (x) x ∈ E,
k=1

and since Rn0 (x) ≥ 0 for all x ∈ E, this leads to


n
X
(6.9) fk0 (x) ≤ f 0 (x), x ∈ E.
k=1

It follows from Theorem 6.4 applied to the function Rn that


Z b
Rn0 (t) dt ≤ Rn (b) − Rn (a),
a

P∞ Rn (b) − Rn (a) → 0 as n → ∞, due to the pointwise convergence


and observe that
of the series n=1 fk . Hence it follows that
Z b
(6.10) Rn0 (t) dt → 0 as n → ∞.
a
Now observe
Z b Z n
bX Z b
0
f (t) dt = fk0 (t) dt + Rn0 (t) dt
a a k=1 a
Z bX∞ Z b
≤ fk0 (t) dt + Rn0 (t) dt.
a k=1 a

By means of (6.10) this gives for n → ∞ that


Z b ∞
Z bX
f 0 (t) dt ≤ fk0 (t) dt.
a a k=1

Compare this with (6.9) and the proof is complete. 


Example 6.7. Let f be the Cantor function from Example 6.5 and extend it
as follows 
0,
 x < 0,
g(x) = f (x), 0 ≤ x ≤ 1,

1, x > 1.

Then g is a continuous nonnegative nondecreasing function and g 0 = 0 a.e. Let (rk )


be an enumeration of Q and define the function F : R → R by

X 1
F (x) = g(x − rk ), x ∈ R.
2k
k=1

Then F is clearly nondecreasing, F is continuous by the Weierstrass theorem, and F


is differentiable almost everywhere by Theorem 6.6. In fact F is strictly increasing.
To see this, let x 6= y and assume F (x) = F (y); then it follows that
g(x − rk ) = g(y − rk ) for all rk ∈ Q,
which, by the continuity of g, leads to a contradiction.
3. TERMWISE DIFFERENTIATION OF SERIES OF MONOTONE FUNCTIONS 91

Definition 6.8. Let (rk ), k ∈ N, be a countable set in an interval (a, b), and
let hk > 0 satisfy
X∞
hk < ∞.
k=1

The jump function f : (a, b) → R is defined by


X ∞
X
(6.11) f (x) = hk = hk H(x − rk ),
k: rk ≤x k=1

where the Heaviside function H defined by


(
1 if x ≥ 0,
H(x) =
0 if x < 0.

Proposition 6.9. The jump function f : (a, b) → R in Definition 6.8 is well-


defined and it has the following properties:
(1) f is nondecreasing; S∞
(2) f is continuous on (a, b) \ k=1 rk ;
(3) at a point r` one has
f (r` +) = f (r` ), f (r` −) = f (r` ) − h` ,
so that f is right continuous with a jump of size h` ;
(4) f is differentiable almost everywhere and f 0 = 0 a.e.
Proof. The jump function is well-defined since the series has nonnegative
terms and the result is independent of the ordering.
(1) For y < x one has by the definition (6.11) of f
X
f (x) − f (y) = hn ≥ 0,
n: y<rn ≤x

and it follows that the S


function f is nondecreasing.

(2) Let c ∈ (a, b) \ k=1 rk . It will be shown that f is continuous at c ∈ (a, b).
Choose ε > 0 and let N ∈ N be so large that

X
hk < ε.
k=N +1

Define δ by
δ = min{ |c − rk | : 1 ≤ k ≤ N },
so that δ > 0. Let c < x < c + δ. If for some k one has c < rk ≤ x, then clearly
k ≥ N + 1, which leads to
X X
f (x) − f (c) = hk ≤ hk < ε.
k: c<rk ≤x k=N +1

Now let c − δ < x < c. If for some k one has x ≤ rk ≤ c, then clearly k ≥ N + 1,
which leads to
X X
f (c) − f (x) = hn ≤ hk < ε.
n: x<rn ≤c k=N +1
92 6. INTEGRATION AND DIFFERENTIATION ON R

(3) It will be shown that at r` the function f is right continuous and that f
has a jump h` . Choose ε > 0. Let N ∈ N be so large that N > r` and

X
hk < ε.
k=N +1

Define δ by
δ = min{ |r` − rk | : 1 ≤ k ≤ N, k 6= ` },
so that δ > 0. Let r` < x < r` + δ, then as in the proof of (2)
f (x) − f (r` ) < ε.
Now let r` − δ < x < r` then
X
f (r` ) − f (x) = hk
k: x<rk ≤r`

so that X
f (r` ) − f (x) − h` = hk < ε,
k: x<rk <r`

as in the proof of (2).


(4) This result follows from Theorem 6.6, since the Heaviside function H in
(6.11) is differentiable except at 0. 

Proposition 6.10. Let the real-valued function f : [a, b] → R be nondecreasing


and right continuous. Then there is a unique decomposition of f into f = g + h,
where g is a continuous nondecreasing function and h is a right continuous jump
function.
Proof. Since f is nondecreasing there are at most countably points rk where
f is not continuous. Define at these points
hk = f (rk ) − f (rk − 0) > 0.
P∞
It is clear that k=1 hk ≤ f (b) − f (a). Hence the corresponding jump function is
given by
X
h(x) = hk .
k: rk ≤x

Define the function g by g = f − h. Then it is clear that g is nondecreasing since


for x < y
X
g(y) − g(x) = f (y) − f (x) − hk ≥ 0.
k: x<rk ≤y
S∞
Moreover, g is continuous at points x 6∈ k=1 rk . To see that g is continuous at
some point r` note that
g(r` ) = f (r` ) − h(r` ), g(r` −) = f (r` −) − h(r` −).
Hence by Proposition 6.9 one obtains
g(r` ) − g(r` −) = f (r` ) − f (r` −) − [h(r` ) − h(r` −)] = h` − h` = 0.
The uniqueness of the decomposition is straightforward to see. 
4. BOUNDED VARIATION AND ABSOLUTE CONTINUITY 93

4. Bounded variation and absolute continuity


Let f : [a, b] → R be a real-valued function. For each x ∈ [a, b] the variation
Vax (f ) of f over the interval [a, x] is defined by
( r )
X
Vax (f ) = sup |f (tk ) − f (tk−1 )| : a ≤ t0 < t1 < . . . tr ≤ x ,
k=1
where the supremum is taken over all finite partitions of [a, x]. Note that in general
0 ≤ Vax (f ) ≤ ∞ and that
|f (x) − f (a)| ≤ Vax (f ).
The variation function is additive in the following sense.
Proposition 6.11. Let f : [a, b] → R be a real-valued function and let c ∈ [a, b].
Then
Vab (f ) = Vac (f ) + Vcb (f ).
In particular, the function x 7→ Vax (f ) is nondecreasing.
Proof. Let P be a partition of [a, b] and let ΣP [a, b] be the corresponding
sum. Choose c ∈ [a, b] and let P 0 = P ∪ {c} a refinement of P . Then clearly
ΣP [a, b] ≤ ΣP 0 [a, c] + ΣP 0 [c, b].
Observe that ΣP 0 [a, c] ≤ Vac (f ) and ΣP 0 [c, b] ≤ Vcb (f ), which leads to
Vab (f ) ≤ Vac (f ) + Vcb (f ).
On the other hand, if P1 is a partition of [a, c] and P2 is a partition of [c, b], then
P1 ∪ P2 is a partition of [a, b] and
ΣP1 [a, c] + ΣP2 [c, b] = ΣP1 ∪P2 [a, b].
Clearly ΣP1 ∪P2 [a, b] ≤ Vab (f ), so that
Vac (f ) + Vcb (f ) ≤ Vab (f ).
This completes the proof. 
Definition 6.12. A real-valued function f : [a, b] → R is said to be of bounded
variation on [a, b] if
Vab (f ) < ∞.
If f : [a, b] → R is a nondecreasing function, then f is of bounded variation
and, in fact, one has Vab (f ) = f (b) − f (a). It is clear that the sum of functions of
bounded variation is of bounded variation.
Proposition 6.13. A real-valued function f : [a, b] → R is of bounded varia-
tion if and only if f is the difference of two nondecreasing functions on [a, b].
Proof. (⇐) Let f = g − h, where g and h are nondecreasing functions on
[a, b]. Then it is clear that
Vab (f ) ≤ g(b) − g(a) + h(b) − h(a).
(⇒) Define the function h by h(x) = Vax (f ) − f (x). Let x ≤ y, then it follows
from
Vax (f ) + f (y) − f (x) ≤ Vax (f ) + |f (y) − f (x)| ≤ Vax (f ) + Vxy (f ) = Vay (f ),
cf. Proposition 6.11, that h(x) ≤ h(y). Hence f (x) = Vax (f ) − h(x), the difference
of two nondecreasing functions. 
94 6. INTEGRATION AND DIFFERENTIATION ON R

Now Proposition 6.13 and Theorem 6.4 have the following direct consequence.
Corollary 6.14. Let f : [a, b] → R be a real-valued function of bounded
variation. Then f is differentiable almost everywhere and f 0 is Lebesgue integrable.
Definition 6.15. A real-valued function f : [a, b] → R is absolutely continuous
if for every ε > 0 there exists δ > 0 such that for each finite sequence of pairwise
disjoint subintervals [ak , bk ] ⊂ [a, b], 1 ≤ k ≤ n,
Xn Xn
(6.12) (bk − ak ) < δ ⇒ |f (bk ) − f (ak )| < ε.
k=1 k=1

By definition an absolutely continuous function is (uniformly) continuous. Note


that if f : [a, b] → R is Lipschitz continuous then f is absolutely continuous.
In particular, if f is differentiable with bounded derivative, then f is absolutely
continuous. It is clear that the sum of absolutely continuous functions is absolutely
continuous.
Proposition 6.16. Let the real-valued function f : [a, b] → R be absolutely
continuous. Then f is of bounded variation on [a, b].
Proof. Choose ε = 1 in Definition 6.15, then there exists δ > 0 such that the
total variation over any interval length δ is at most 1. Since [a, b] is the union of
at most N = [(b − a)/δ] + 1 such intervals, the total variation of f over [a, b] is at
most N ; cf. Proposition 6.11. 
Proposition 6.17. Let the real-valued function f : [a, b] → R be absolutely
continuous. Then also the function x 7→ Vax (f ) is absolutely continuous.
Proof. Choose ε > 0 and let δ > 0 be as in Definition 6.15. Let [ak , bk ] ⊂ [a, b]
be a sequence of pairwise disjoint subintervals with
Xn
(bk − ak ) < δ.
k=1
Choose for any such interval [ak , bk ] an arbitrary finite partition Pk as follows:
s(k)
X
ak ≤ xk0 ≤ xk1 ≤ ··· ≤ xks(k) ≤ bk , so that (xkj − xkj−1 ) ≤ bk − ak .
j=1

Then it is clear that


n s(k)
X X n
X
(xkj − xkj−1 ) ≤ (bk − ak ) < δ,
k=1 j=1 k=1

and hence, by definition it follows that


n s(k)
X X
|(f (xkj ) − f (xkj−1 )| < ε.
k=1 j=1

Now observe that this estimate holds for any choice of finite partition Pk of each
interval [ak , bk ]. This leads to
 
n n s(k)
X X X
Vabkk (f ) = sup  |(f (xkj ) − f (xkj−1 )| ≤ ε.
Pk j=1
k=1 k=1
5. DIFFERENTIATION AND INTEGRATION 95

By Proposition 6.11 one therefore obtains


Xn
[ (Vabk (f ) − Vaak (f ) ] ≤ ε,
k=1

which shows that the function x 7→ Vax (f ) is absolutely continuous. 

5. Differentiation and integration


By Proposition 6.16 absolutely continuous functions are of bounded variation;
hence they are differentiable almost everywhere. They satisfy the following impor-
tant property.
Lemma 6.18. Let the real-valued function f : [a, b] → R be absolutely continu-
ous. If f 0 = 0 a.e., then f constant.
Proof. The proof will be given by means of Vitali covers. Define the set
E = {x ∈ [a, b] : f 0 (x) = 0},
so that E is measurable and choose ε > 0.
By the definition of the set E for every x ∈ E there exists δx > 0 such that for
all 0 < h < δx one has
(6.13) |f (x + h) − f (x)| < hε.
The collection V of such closed intervals [x, x + h] forms a Vitali cover for E.
By Definition 6.15 there exists δ > 0 such that (6.12) holds. By Lemma 6.2
there exists a collection J1 of pairwise disjoint intervals of the form
[x1 , x1 + h1 ], . . . , [xN , xN + hN ],
such that they nearly cover E:
m∗ (E \ ([x1 , x1 + h1 ] ∪ · · · ∪ [xN , xN + hN ])) < δ.
The set [a, b]\([x1 , x1 +h1 ]∪· · ·∪[xN , xN +hN ]) determines another finite collection
J2 of intervals
[a, x1 ], [x1 + h1 , x2 ], . . . , [xN + hN , b],
and clearly the length of each individual interval in J2 is at most δ. Together J1
and J2 cover the interval [a, b] and clearly
!
X X
|f (b) − f (a)| ≤ + |f (x + h) − f (x)|.
J1 J2

Note that by (6.13) and by (6.12), respectively, one obtains


X X
|f (x + h) − f (x)| ≤ (b − a)ε, |f (x + h) − f (x)| < ε.
J1 J2

Therefore one gets


|f (b) − f (a)| ≤ (b − a)ε + ε,
where ε > 0 is arbitrary. Thus it follows that f (b) = f (a). This completes the
proof. 
Although every absolutely continuous function is of bounded variation, not
every function of bounded variation is absolutely continuous. There are nonconstant
functions of bounded variation whose derivative is 0 almost everywhere.
96 6. INTEGRATION AND DIFFERENTIATION ON R

Example 6.19. The Cantor function f in Example 6.5 is of bounded variation,


since it is increasing. However the function f is not absolutely continuous.
The following theorem is the foundation on which the relation between differ-
entiation and integration is built.
Theorem 6.20. Let g : [a, b] → R be a Lebesgue integrable real-valued function
and let the real-valued function f : [a, b] → R be defined by
Z x
f (x) = γ + g(t) dt, x ∈ [a, b],
a
where γ ∈ R. Then f (a) = γ, f is absolutely continuous, and f 0 = g a.e.
Proof. In order to check that f is absolutely continuous, define the functions
hn : [a, b] → R by
(
|gn (x)|, when |gn (x)| ≤ n,
hn (x) =
n, when |gn (x)| > n.
Then hn % |f | and by the monotone convergence theorem one has
Z b Z b
lim hn (t) dt = |g(t)| dt.
n→∞ a a
Choose ε > 0 and let n ∈ N be so large that
Z b
[ |g(t)| − hn (t) ] dt < ε.
a
Let δ be defined by δ = ε/n so that δ > 0 and choose
n
X
a ≤ a1 ≤ b1 ≤ · · · ≤ an ≤ bn ≤ b with (bk − ak ) ≤ δ.
k=1
Then it follows that
X n n Z
X bk n Z
X bk
|f (bk ) − f (ak )| = g(t) dt ≤ |g(t)| dt
k=1 k=1 ak k=1 ak
n
X bkZ n Z
X bk
= |g(t) − hn (t)| dt + |hn (t)| dt
k=1 ak k=1 ak

< ε + nδ = 2ε.
Therefore f is absolutely continuous. Hence f is of bounded variation, which implies
that f is differentiable a.e. and that f 0 is integrable.
It remains to show that f 0 = g a.e. This will be done in two steps: first the
case that g is bounded will be treated and then the general case will be reduced to
the bounded case.
First assume that g is integrable and bounded on [a, b]: |g(u)| ≤ K. Then
f (t + h) − f (t) 1 t+h
Z
= g(u) du ≤ K
h h t
and
f (t + h) − f (t)
→ f 0 (t) a.e.
h
5. DIFFERENTIATION AND INTEGRATION 97

For every x ∈ [a, b] it follows from the dominated convergence theorem that
Z x
1 x
Z
0
f (t) dt = lim [f (t + h) − f (t)] dt
a h→0 h a
" Z #
1 x+h 1 a+h
Z
= lim f (t) dt − f (t) dt
h→0 h x h a
= f (x) − f (a),
where in the last step the continuity of f has been used. Hence for every x ∈ [a, b]
it follows that Z x
[f 0 (t) − g(t)] dt = 0,
a
so that g = f 0 a.e.
Next consider the general case where g is integrable. From the definition of the
integral it follows that one may assume that g is nonnegative and integrable. In
this case define the functions gn by
gn (x) = min (n, g(x)).
Then g − gn ≥ 0 and observe that
Z x
fn (x) = [g(t) − gn (t)] dt
a
is a nondecreasing function on [a, b]. Therefore fn is differentiable almost every-
where with nonnegative integrable derivative fn0 . Furthermore observe that the
function gn is bounded so that it follows from the previous step in the proof that
Z x
d
gn (t) dt = gn (x) a.e.
dx a
By rewriting f as follows
Z x
f (x) = γ + fn (x) + gn (t) dt,
a
one sees that for all n ∈ N
f 0 (x) ≥ gn (x) a.e.
which gives
f 0 (x) ≥ g(x) a.e.
Integration of this inequality gives
Z b Z b
f 0 (t) dt ≥ g(t) dt = f (b) − f (a).
a a
Rb
Since by Theorem 6.4 f (b) − f (a) ≥ a f 0 (t) dt one concludes
Z b
[f 0 (t) − g(t)] dt = 0.
a
However f 0 ≥ g a.e. so that one sees that f 0 = g a.e. 
The next result is the main theorem of the calculus in the context of Lebesgue
integration: it shows precisely the connection between differentiation and integra-
tion; cf. Lemma 4.19.
98 6. INTEGRATION AND DIFFERENTIATION ON R

Theorem 6.21. Let f : [a, b] → R be a real-valued function on a compact


interval. Then the following statements are equivalent:
(1) f is absolutely continuous;
(2) f is differentiable almost everywhere, the derivative f 0 is Lebesgue inte-
grable, and
Z x
(6.14) f (x) = f (a) + f 0 (t) dt, x ∈ [a, b].
a

Proof. (1) ⇒ (2) Let f be absolutely continuous, then f is of bounded varia-


tion, see Proposition 6.16. According to Corollary 6.14 f is differentiable a.e. and
f 0 is Lebesgue integrable. Define the function ϕ by
Z x
ϕ(x) = f (x) − f 0 (t) dt, x ∈ [a, b].
a
Since the sum of absolutely continuous functions is absolutely continuous it follows
from Theorem 6.20 that the function ϕ is absolutely continuous. Moreover it follows
from Theorem 6.20 that ϕ0 = 0 a.e. Hence ϕ is constant by Lemma 6.18; i.e.
ϕ(x) = ϕ(a) = f (a). Therefore the identity (6.14) has been shown.
(2) ⇒ (1) This is clear from Theorem 6.20. 
Theorem 6.21 has an important consequence about the structure of monotone
functions or of functions of bounded variation. The decomposition below will be
further studied later in the context of ordinary measures when f is nondecreasing
and in the context of signed and complex measures when f is of bounded variation.
Theorem 6.22 (Lebesgue decomposition). Let f : [a, b] → R be a real-valued
function.
(1) If f is nondecreasing, then there exists a unique decomposition
(6.15) f = ϕ + ψ,
where ϕ is an absolutely continuous nondecreasing function with ϕ(a) =
f (a) and ψ is a nondecreasing function with ψ 0 = 0.
(2) If f is of bounded variation, then there exists a unique decomposition as
in (6.15), where ϕ is an absolutely continuous function with ϕ(a) = f (a)
and ψ is a function of bounded variation with ψ 0 = 0.
Proof. (1) Define the function ϕ : [a, b] → R by
Z x
ϕ(x) = f 0 (t) dt + f (a), x ∈ [a, b].
a
Then ϕ is absolutely continuous, ϕ0 = f 0 , and ϕ(a) = f (a). Define the function
ψ : [a, b] → R by ψ = f − ϕ. Then by Theorem 6.4
Z x
ψ(x) = f (x) − f 0 (t) dt − f (a) ≥ 0,
a
0 0 0
while ψ = f − f = 0 a.e.
For the uniqueness assume that f = ϕ1 + ψ1 where ϕ1 is an absolutely contin-
uous nondecreasing function with ϕ1 (a) = f (a) and ψ1 is a nondecreasing function
with ψ10 = 0. Then ϕ − ϕ1 = ψ1 − ψ, so that
ϕ0 − ϕ01 = ψ10 − ψ 0 = 0,
5. DIFFERENTIATION AND INTEGRATION 99

so that ϕ1 − ϕ = c, where c is constant. Due to ϕ(a) = f (a) and ϕ1 (a) = f (a) it


follows that ϕ1 = ϕ and, hence, ψ1 = ψ.
(2) This case is reduced to the case in (1) due to Proposition 6.13. 
Remark 6.23. The unique decomposition (6.15) of f is in terms of the abso-
lutely continuous part ϕ and of the so-called singular part ψ. The singular part ψ
itself can be uniquely decomposed in
ψ = g + h,
where g is a continuous nondecreasing function with g 0 = 0 a.e., and h is a jump
function which is continuous from the right; cf. Proposition 6.10. The function g
is called the singular continuous part of f and the function h is sometimes called
the discrete part of f . The case for functions of bounded variation is completely
analogous.
Proposition 6.24. Let f, g : [a, b] → R be real-valued Lebesgue integrable
function and define
Z x Z x
F (x) = α + f (t) dt, G(x) = β + g(t) dt.
a a
Then Z b Z b
G(t)f (t) dt + g(t)F (t) dt = F (b)G(b) − G(a)F (a).
a a

Proof. The functions F and G are absolutely continuous on [a, b] and hence
they are bounded: |F (x)| ≤ A and |G(x)| ≤ B. Thus it follows from
|F (x)G(x) − F (y)G(y)| = |F (x)G(x) − F (x)G(y) + F (x)G(y) − F (y)G(y)|
≤ |F (x)||G(x) − G(y)| + |G(y)||F (x) − F (y)|
≤ A|G(x) − G(y)| + B|F (x) − F (y)|,
that the product F G is absolutely continuous. Hence F G is differentiable a.e. and
(F G)0 = F 0 G + F G0 a.e.
0 0
Since F = f and G = g a.e. the statement now follows from Theorem 6.21. 
Note that by Theorem 6.21 the previous result could also be stated in the
following more familiar equivalent form.
Corollary 6.25. Let the real-valued functions F, G : [a, b] → R be absolutely
continuous, then
Z b Z b
G(t)F 0 (t) dt + G0 (t)F (t) dt = F (b)G(b) − G(a)F (a).
a a
CHAPTER 7

Integration in Rd

For an extended Rreal-valued integrable function f : Rd → R the actual calcu-


lation of the integral Rd f dm where m stands for Lebesgue measure depends very
often on Fubini-Tonelli type theorems. Another tool which plays an important role
takes care of the geometry of the situation: it is the transformation formula
Z Z
f dm = (f ◦ ϕ) | det ϕ0 | dm.
Y X
d
Here X and Y are open sets in R and ϕ is a diffeomorphism from X onto Y .
Usually the occurrence of the function ϕ is interpreted as a change of variables. In
the present context the expression | det ϕ0 | dm involving the Jacobian determinant
of ϕ0 can be interpreted as another measure on Rd . A rigorous argument to justify
the formula will be presented. First the general measure theoretic argument behind
the form of the formula will be given and it will be applied to linear transformations
in Rd . The general nonlinear case will be treated by applying local linearizations.
The transformation formula will be applied to some coordinate systems which play
a role in mathematical physics. The chapter end with a very brief introduction to
Lebesgue integration on manifolds in Rd .

1. Transformations in measure spaces


Let (Ω , A , ν) be a measure space and let f Rbe a measurable nonnegative ex-
0 0

tended real-valued function on (Ω0 , A0 ), so that Ω0 f dν ∈ [0, ∞]. Assume that a


change of variables in (Ω0 , A0 ) is due to a bijective mapping ϕ from a σ-algebra
(Ω, A) onto (Ω0 , A0 ), such that ϕ andR ϕ−1 are measurable with respect to (Ω, A)
R (Ω , A ), Rrespectively. Then also Ω f ◦ ϕ dµ ∈ [0, ∞]. The question is to relate
0 0
and
Ω0
f dν and Ω f ◦ ϕ dµ when the measures ν and µ are related. The main idea can
be found in Theorem 4.39.
Lemma 7.1. Let (Ω, A, µ) and (Ω0 , A0 , ν) be measure spaces. Assume that ϕ is
a bijective mapping from (Ω, A) onto (Ω0 , A0 ), such that ϕ and ϕ−1 are measurable
with respect to (Ω, A) and (Ω0 , A0 ), respectively. Assume that the measures µ and
ν satisfy
(7.1) ν ◦ ϕ ≤ µ.
Let the nonnegative extended real-valued function f : (Ω0 , A0 ) → (R, B) be measur-
able. Then
Z Z
(7.2) f dν ≤ (f ◦ ϕ) dµ.
Ω0 Ω

Proof. Let A ∈ A and let g be its indicator function. Then it follows that
g ◦ ϕ−1 = 1ϕ(A) .
101
102 7. INTEGRATION IN Rd

since g ◦ ϕ−1 (ω 0 ) = 1 if and only if ϕ−1 (ω 0 ) ∈ A if and only if ω 0 ∈ ϕ(A). The


inequality (7.1) can be rewritten as
Z Z
−1
(7.3) (g ◦ ϕ ) dν = ν(ϕ(A)) ≤ µ(A) = g dµ.
Ω0 Ω
By linearity it follows from (7.3) that
Z Z
−1
(7.4) (g ◦ ϕ ) dν ≤ g dµ,
Ω0 Ω
holds for all simple functions g. By monotonicity it is seen that (7.4) remains valid
for nonnegative measurable functions g. Now let g = f ◦ ϕ in (7.4) then (7.2)
follows. 
Theorem 7.2. Let (Ω, A, µ) and (Ω0 , A0 , ν) be measure spaces. Assume that ϕ
is a bijective mapping from (Ω, A) onto (Ω0 , A0 ), such that ϕ and ϕ−1 are measurable
with respect to (Ω, A) and (Ω0 , A0 ), respectively. Assume that the measures µ and
ν are connected by
(7.5) ν ◦ ϕ = µ.
Let the extended real-valued function f : (Ω0 , A0 ) → (R, B) be measurable. Then the
following statements hold.
(1) If the function f is nonnegative, then
Z Z
(7.6) f dν = (f ◦ ϕ) dµ.
Ω0 Ω
(2) The function f is integrable with respect to ν if and only if the function
f ◦ ϕ is integrable with respect to µ, in which case (7.6) holds.
Moreover, if there exists a nonnegative measurable function h and a measure ζ on
(Ω, A) such that
Z
(7.7) µ(A) = h dζ, A ∈ A,
A
then (7.6) can be written as
Z Z
(7.8) f dν = (f ◦ ϕ) h dζ.
Ω0 Ω

Proof. (1) This follows as in the proof of Lemma 7.1.


(2) For every measurable function f it follows from (1) and |f | ◦ ϕ = |f ◦ ϕ|
that Z Z
|f | dν = |f ◦ ϕ| dµ ∈ [0, ∞].
Ω0 Ω
Hence f is integrable with respect to ν if and only if f ◦ ϕ is integrable with respect
to µ. In this case the equality in (7.6) is obtained via the components f + and f − .
The last statement is clear. 
A combination of Theorem 7.2 with Proposition 2.14 leads to the following
invariance result for translations. Let a ∈ Rd , then the translation fa of a function
f : Rd → Rd is defined by fa (x) = f (x + a), x ∈ Rd .
Theorem 7.3. Let the extended real-valued function f : Rd → R be Lebesgue
measurable and let a ∈ Rd . Then the following statements hold.
2. NON-LINEAR TRANSFORMATIONS IN Rd 103

(1) If the function f is nonnegative, then


Z Z
(7.9) f dm = fa dm.
Rd Rd
(2) The function f is Lebesgue integrable if and only if the function fa is
Lebesgue integrable, in which case (7.9) holds.
Proof. Let (Ω, A) = (Ω0 , A0 ) be equal to (Rd , Md ). Define the translation
ϕ : Rd → Rd by ϕx = x + a, x ∈ Rd . Then ϕ and its inverse are Lebesgue
measurable; cf. Proposition 2.14. Apply Theorem 7.2 with ν = m and the present
choice of ϕ. Then by Proposition 2.14
µ(A) = m(ϕ(A)) = m(A), A ∈ Md .
R R
Hence it is is clear that Ω (f ◦ ϕ) dµ = Ω fa dm. Thus (7.6) implies (7.9). 
A combination of Theorem 7.2 with Proposition 2.16 leads to a similar trans-
formation result for invertible linear transformations.
Theorem 7.4. Let the extended real-valued function f : Rd → R be Lebesgue
measurable and let T be an invertible linear transformation in Rd .
(1) If the function f is nonnegative, then
Z Z
(7.10) f dm = | det T | (f ◦ T ) dm.
Rd Rd
(2) The function f is Lebesgue integrable if and only if the function f ◦ T is
Lebesgue integrable, in which case (7.10) holds.
Proof. Let (Ω, A) = (Ω0 , A0 ) be equal to (Rd , Md ). The invertible transforma-
tion T in Rd and its inverse are Lebesgue measurable; cf. Proposition 2.16. Apply
Theorem 7.2 with ν = m and ϕ = T . Then by Proposition 2.16
µ(A) = m(T (A)) = | det T | m(A), A ∈ Md .
Hence it is is clear that
Z Z Z
(f ◦ ϕ) dµ = (f ◦ T ) dµ = | det T | (f ◦ T ) dm.
Ω Ω Ω
Thus (7.6) implies (7.10). 
Corollary 7.5. The Lebesgue integral on Rd is invariant under unimodular
transformations T (i.e., whose determinant satisfies det T = ±1).

2. Non-linear transformations in Rd
Let X ⊂ Rd and Y ⊂ Rd be open subsets. A bijection ϕ from X onto Y is
called a diffeomorphism if ϕ and its inverse ϕ−1 are both C r , r ≥ 1. It follows from
(ϕ ◦ ϕ−1 )(y) = y, y ∈ Y , and the chain rule that
(7.11) ϕ0 (ϕ−1 (y))(ϕ−1 )0 (y) = I, y ∈ Y,
0 −1 −1 0
where ϕ (ϕ (y)) and (ϕ ) (y) are linear transformations in Rd . A direct conse-
quence of (7.11) is
(7.12) det(ϕ0 ◦ ϕ−1 )(y) det(ϕ−1 )0 (y) = 1, y ∈ Y.
In particular the linear transformations ϕ (ϕ (y)) and (ϕ−1 )0 (y) are invertible for
0 −1

all y ∈ Rd . Let ϕ be a homeomorphism from an open set X ⊂ Rd onto an open


104 7. INTEGRATION IN Rd

subset Y ⊂ Rd . Observe that if ϕ is C r , r ≥ 1, and det ϕ0 does not vanish on X,


then ϕ is in fact a diffeomorphism between X and Y , as follows from the inverse
mapping theorem.
Theorem 7.4 has been a first step towards a general form of the abstract trans-
formation formula mentioned in Theorem 7.2 in the context of Rd . In order to
obtain a nonlinear version of Theorem 7.4 involving a diffeomorphism the following
lemmas are needed.
Lemma 7.6. Let ϕ be a diffeomorphism from an open set X ⊂ Rd onto an open
set Y ⊂ Rd . If Q ⊂ X is a closed cube, then ϕ(Q) is Lebesgue measurable, and
Z
(7.13) m(ϕ(Q)) ≤ | det ϕ0 | dm.
Q

Proof. Let Q be a closed cube in X ⊂ Rd . It will be shown that ϕ(Q) (being


Lebesgue measurable) satisfies
Z
(7.14) m(ϕ(Q)) ≤ | det j | dm.
Q

Assume that a is the center of Q and that the length of its sides is r, so that
m(Q) = rd . Since Q is a closed cube in Rd it is compact and since ϕ is continuous,
the image ϕ(Q) is compact and thus Lebesgue measurable. It can be seen from the
generalized mean-value theorem that ϕ(Q) ⊂ Y lies in a cube with center ϕ(a) and
length of its sides at most r maxu∈Q kj(u)k. Hence it follows that
 d  d
(7.15) m(ϕ(Q)) ≤ r max kj(u)k = max kj(u)k m(Q).
u∈Q u∈Q

For any invertible linear transformation T in Rd apply the result in (7.15) to the
diffeomorphism T −1 ◦ ϕ. By the chain rule the derivative of T −1 ◦ ϕ is given by the
matrix T −1 ϕ0 . Hence one obtains
 d
−1 −1
(7.16) m(T (ϕ(Q))) ≤ max kT j(u)k m(Q).
u∈Q

Now apply Theorem 2.17 to the left-hand side of (7.16), then


m(T −1 (ϕ(Q))) = (det T −1 ) m(ϕ(Q)),
and so one obtains from (7.16)
 d
(7.17) m(ϕ(Q)) ≤ | det T | max kT −1 j(u)k m(Q).
u∈Q

Next choose a decomposition of Q as a finite union of almost nonoverlapping


closed cubes Qk , (that is, the intersection of two cubes is of lower dimension than d)
with center ak and length of sides rk . Apply the estimate (7.17) to each individual
Qk with the choice T = j(ak ). Then one obtains from (7.17)
n
X
m(ϕ(Q)) ≤ m(ϕ(Qk ))
k=1
(7.18) n  d
X
−1
≤ | det j(ak ) | max kj(ak ) j(u)k m(Qk ).
u∈Qk
k=1
2. NON-LINEAR TRANSFORMATIONS IN Rd 105

In order to treat the terms involving the maximum, consider for x, u ∈ Q


kj(x)−1 j(u)k = kj(x)−1 (j(u) − j(x)) + Ik
(7.19)
≤ kj(x)−1 kkj(u) − j(x)k + kIk ≤ 1 + M kj(u) − j(x)k,
where M is a global bound on Q as j −1 is continuous on Q. Due to uniform
continuity for every ε > 0 there exists δ > 0 such that kx − uk < δ implies that
kj(u) − j(x)k < ε/M , and thus the estimate in (7.19) gives
(7.20) kj(x)−1 j(u)k ≤ 1 + ε.
Additionally, let L be the global minimum of | det j| on Q. Then L > 0 since Q is
compact and det j is continuous and nonzero on Q. In (7.18) the decomposition of
Q may be chosen by taking rk sufficiently small such that each individual cube Qk
is contained in an open ball Uδ with the radius δ such that
| det j(ak ) − det j(u) | ≤ εL, u ∈ Qk .
The last relation implies
| det j(ak ) − det j(u) | ≤ ε| det j(u)|, u ∈ Qk ,
which leads with the triangle inequality to
(7.21) | det j(ak ) | ≤ (1 + ε)| det j(u)|, u ∈ Qk .
Hence it follows from (7.18) and (7.20) that for every ε > 0 there exists a finite
partition Q1 , . . . , Qn of the cube Q such that
n
X
(7.22) m(ϕ(Q)) ≤ (1 + ε)d | det j(ak ) | m(Qk ).
k=1

Clearly the sum in the right-hand side can be written as an integral


n Z n
!
X X
| det j(ak )| m(Qk ) = | det j(ak ) | 1Qk dm,
k=1 Q k=1

where the integrand is the nonnegative simple function


n
X
| det j(ak ) | 1Qk ,
k=1

which lies almost everywhere below the function (1 + ε)| det j | by relation (7.21).
Hence it follows from (7.22) that
Z
d+1
(7.23) m(ϕ(Q)) ≤ (1 + ε) | det j | dm.
Q

Since the inequality in (7.23) is true for every ε > 0 the claim (7.14) has been
shown. 

Lemma 7.7. Let ϕ be a C r -diffeomorphism, r ≥ 1, from an open set X ⊂ Rd


onto an open set Y ⊂ Rd . If A ⊂ X is Lebesgue measurable, then ϕ(A) is Lebesgue
measurable, and
Z
(7.24) m(ϕ(A)) ≤ | det ϕ0 | dm.
A
106 7. INTEGRATION IN Rd

Proof. The result has been shown for closed cubes in X in Lemma 7.6. The
proof of the general case will be given in a number of steps. In Step 1 the result
for closed is extended (via dyadic cubes) to open sets in X. Steps 2 and 3 then
succesively treat the case of measurable sets which are conveniently bounded by
compact sets and the case of general measurable sets in X. The notation j = ϕ0
will still be used as in the proof of Lemma 7.6.
Step 1. Let O be an open set in X ⊂ Rd . It will be shown that ϕ(O) (being
Lebesgue measurable) satisfies
Z
(7.25) m(ϕ(O)) ≤ | det j | dm.
O
Recall that any open subsetSO ⊂ X can be written asS∞a countable union of pairwise

disjoint dyadic cubes O = n=1 Qn . Then ϕ(O) = n=1 ϕ(Qn ), a pairwise disjoint
union of measurable sets. Since the complement of each dyadic cube Qn in its
closure Qn is of lower dimension than d, it follows that m(Qn \ Qn ) = 0 and thus
the relation (7.14) holds also for dyadic cubes. Therefore
X∞ X∞ Z Z
m(ϕ(O)) = m(ϕ(Qn )) ≤ | det j | dm = | det j | dm,
n=1 n=1 Qn O

which proves the claim (7.25).


Step 2. Let A be a Lebesgue measurable set in X, such that A ⊂ H 0 ⊂ H for
some compact set H ⊂ X. It will be shown that ϕ(A) is Lebesgue measurable and
that
Z
(7.26) m(ϕ(A)) ≤ | det j | dm.
A

Let M = maxu∈H | det j(u) |. Since A is measurable there exists compact sets Kn
and open sets On such that
1
Kn ⊂ A ⊂ On ⊂ H, m(On \ Kn ) <
Mn
(if necessary intersect the open set from Corollary 2.22 with H 0 to satisfy the
inclusion On ⊂ H). Since ϕ is a diffeomorphism, clearly ϕ(Kn ) is compact and
ϕ(On ) is open. Moreover by Step 1
m(ϕ(On )) − m(ϕ(Kn )) = m(ϕ(On \ Kn ))
Z
1
≤ | det j | dm ≤ M m(ϕ(On \ Kn )) ≤ ,
On \Kn n
which confirms that ϕ(A) is measurable. Furthermore
1
m(ϕ(A)) ≤ m(ϕ(On )) ≤ m(ϕ(Kn )) +
Z Z n
1 1
≤ | det j | dm + ≤ | det j | dm + ,
Kn n A n
which proves the claim (7.26).
Step 3. Let A be a Lebesgue measurable set in X ⊂ Rd . It will be shown that
ϕ(A) is Lebesgue measurable and that
Z
m(ϕ(A)) ≤ | det j | dm.
A
2. NON-LINEAR TRANSFORMATIONS IN Rd 107

Define the set Bn , n ∈ N, by


 
c 1
Bn = x ∈ A : kxk ≤ n, d(x, X ) ≥ ,
n
then clearly Bn is closed and bounded, so compact. Note that Bn ⊂ Bn+1 . Define
0
An = A ∩ Bn , then An is measurable and An ⊂ Bn+1 ⊂ Bn+1 , so that An satisfies
the condition in Step 2. Furthermore,
S∞ since A n ⊂ A n+1 , S
the sets (An ) form a

nondecreasing sequence with A = n=1 An , so that ϕ(A) = n=1 ϕ(An ). Hence
Z Z
m(ϕ(A)) = lim m(ϕ(An )) ≤ lim | det j | dm = | det j | dm
n→∞ n→∞ An A

This completes the proof of the lemma. 


A combination of Lemma 7.1 and Theorem 7.2 with Lemma 7.7 leads to a
nonlinear version of Theorem 7.4.
Theorem 7.8. Let ϕ be a C r -diffeomorphism, r ≥ 1, from an open set X ⊂ Rd
onto an open set Y ⊂ Rd and let the extended real-valued function f : Y → R be
Lebesgue measurable. Then the following statements hold.
(1) If the function f is nonnegative, then
Z Z
(7.27) f dm = (f ◦ ϕ) | det ϕ0 | dm.
Y X
(2) The function f is Lebesgue integrable if and only if the function
(f ◦ ϕ) | det ϕ0 |
is Lebesgue integrable, in which case (7.27) holds.
Proof. (1) Let A ⊂ Rd be a Lebesgue measurable set and rewrite the inequal-
ity (7.24) in Lemma 7.7 as follows
ν(ϕ(A)) ≤ µ(A),
where the measures ν and µ are defined by
Z
(7.28) ν(A) = m(A), µ(A) = | det ϕ0 | dm, A ∈ A.
A

Hence if f : Y → R is Lebesgue measurable and nonnegative, it follows from Lemma


7.1 that
Z Z
(7.29) f dm ≤ (f ◦ ϕ) | det ϕ0 | dm,
Y X
for any diffeomorphism ϕ taking X bijectively onto Y .
Since ϕ−1 is a diffeomorphism taking Y bijectively onto X, it follows by sym-
metry from (7.29) that for any g : X → R which is Lebesgue measurable and
nonnegative one has
Z Z
(7.30) g dm ≤ (g ◦ ϕ−1 ) | det(ϕ−1 )0 | dm.
X Y
Next replace in the inequality (7.30) the function g by the function (f ◦ ϕ)| det ϕ0 |,
so that in particular the term g ◦ ϕ−1 in the right-hand side of (7.30) becomes
((f ◦ ϕ)| det ϕ0 |) ◦ ϕ−1 = (f ◦ ϕ ◦ ϕ−1 )(| det ϕ0 | ◦ ϕ−1 ) = f | det(ϕ0 ◦ ϕ−1 )|.
108 7. INTEGRATION IN Rd

Therefore it follows from (7.30) that


Z Z
0
(f ◦ ϕ)| det ϕ | dm ≤ f | det(ϕ0 ◦ ϕ−1 ) || det(ϕ−1 )0 | dm
X
(7.31) ZY
= f dm,
Y
where in the last step the identity (7.12) has been used. The identity (7.27) now
follows from the inequalities in (7.29) and (7.31).
(2) This follows from Theorem 7.2 with ν and µ as in (7.28). 
Corollary 7.9. Let ϕ be a C r -diffeomorphism, r ≥ 1, from an open set
X ⊂ Rd onto an open set Y ⊂ Rd . Then ϕ(A) ⊂ Y is Lebesgue measurable if and
only if A ⊂ X is Lebesgue measurable, and in this case
Z
(7.32) m(ϕ(A)) = | det ϕ0 | dm.
A

For completeness the one-dimensional result is mentioned separately. Now ϕ is


a diffeomorphism from an interval (a, b) onto an interval (c, d). The absolute value
of the derivative takes care of the orientation: the derivative is either positive on
(a, b) or negative on (a, b).
Corollary 7.10. Let ϕ be a C r -diffeomorphism, r ≥ 1, from (a, b) onto (c, d)
and let f : [c, d] → R be a Lebesgue measurable real-valued function.
(1) If the function f is nonnegative, then
Z d Z b
(7.33) f (t) dt = (f ◦ ϕ)(t)|ϕ0 (t)| dt.
c a
(2) The function f is Lebesgue integrable if and only if the function (f ◦ ϕ)ϕ0
is Lebesgue integrable, in which case (7.33) holds.

3. Some coordinate systems


It is important to observe that the main transformation theorem is concerned
with functions defined on open sets. However in many applications this is not a
severe restriction as then the boundaries have measure 0. This will be illustrated
for several coordinate systems.

Cylindrical polar coordinates for R3 . First consider polar coordinates in the


space R2 ; they are based on the formulas
(
x = r cos ϕ,
(7.34) (r, ϕ) ∈ R2 .
y = r sin ϕ,
Then one sees that the following identities hold:
x2 + y 2 = r 2 , (sin ϕ) x = (cos ϕ) y.
Hence, it is clear that (7.34) produces circles around the origin and straight lines
emanating from the origin when (r, ϕ) varies in R2 with the parameters r and ϕ
fixed, respectively. If restricted to the open set (0, ∞) × (−π, π) the corresponding
mapping in (7.34) is injective, and hence a diffeomorphism, onto the slit plane
R2 \ {(x, y) ∈ R2 : x ≤ 0}.
3. SOME COORDINATE SYSTEMS 109

Note that {(x, y) ∈ R2 : x ≤ 0} has measure 0 in R2 . The corresponding transfor-


mation formula in (7.32) is expressed as
dxdy = r drdϕ.
Cylindrical polar coordinates are introduced by the formulas

 x = r cos ϕ,

(7.35) y = r sin ϕ, (r, ϕ, z) ∈ R3 .

z = z.

When restricted to (0, ∞) × (−π, π) × R the corresponding mapping in (7.35) is


injective, and hence a diffeomorphism, onto the slit space
R3 \ {(x, y, z) ∈ R2 : x1 ≤ 0}.
The corresponding transformation formula in (7.32) is expressed as
dxdydz = r drdϕdz.

Cylindrical elliptic coordinates for R3 . First consider elliptic coordinates in


the space R2 ; they are based on the formulas
(
x = cosh ρ cos ϕ,
(7.36) (ρ, ϕ) ∈ R2 .
y = sinh ρ sin ϕ,
Then one sees, when the relevant divisions are allowed, that the following identities
hold:  2  2  2  2
x y x y
+ = 1, − = 1.
cosh ρ sinh ρ cos ϕ sin ϕ
Hence, it is clear that (7.36) produces ellipses and hyperboles when (ρ, ϕ) varies in
R2 with the parameter ρ or ϕ fixed, respectively. When restricted to the open set
(0, ∞)×(−π, π) the corresponding mapping in (7.34) is injective. The corresponding
transformation formula in (7.32) is expressed as
1
dxdy = (cosh 2ρ − cos 2ϕ) dρdϕ.
2
Cylindrical elliptic coordinates are introduced by the formulas

 x = cosh ρ cos ϕ,

(7.37) y = sinh ρ sin ϕ, (ρ, ϕ, z) ∈ R3 .

z = z,

When restricted to (0, ∞) × (−π, π) × R the corresponding mapping in (7.37) is


injective. The corresponding transformation formula in (7.32) is expressed as
1
dxdydz = (cosh 2ρ − cos 2ϕ) dρdϕdz.
2
3
Spherical coordinates for R . The spherical coordinates are introduced by the
formulas

 x = r sin ϕ sin θ,

(7.38) y = r cos ϕ sin θ, (r, ϕ, θ) ∈ R3 .

z = r cos θ,

110 7. INTEGRATION IN Rd

Note that here and in the following examples the choice of the angles is as follows:
for a vector (x, y, z) ∈ R3 first the angle θ is chosen with respect to the z-axis;
for the corresponding projection in the xy-plane the angle ϕ is then chosen with
respect to the y-axis. Note that in the literature different choices may occur.
Clearly the following identities hold:
p
x2 + y 2 + z 2 = r2 , (cos ϕ) x = (sin ϕ) y, (sin θ) z = (cos θ) x2 + y 2 .
Hence (7.38) produce spheres, planes, and cones when the point (r, ϕ, θ) varies in
R3 with the parameter r, ϕ, or θ fixed, respectively. When restricted to the open
set (0, ∞) × (−π, π) × (0, π) the corresponding mapping in (7.38) is injective. The
corresponding transformation formula in (7.32) is expressed as
dxdydz = r2 sin θ drdϕdθ.
Prolate spheroidal coordinates for R3 . The prolate spheroidal coordinates are
based on the formulas:

 x = sinh ρ sin ϕ sin θ,

(7.39) y = sinh ρ cos ϕ sin θ, (ρ, ϕ, θ) ∈ R3 .

z = cosh ρ cos θ,

Then one sees that the following identities hold:


z2 x2 + y 2 z2 x2 + y 2
+ = 1, (cos ϕ) x = (sin ϕ) y, − = 1.
cosh2 ρ sinh2 ρ cos2 θ sin2 θ
Hence, it is clear that (7.39) produces ellipsoids, half-planes, and two-sheeted hyper-
boloids, when (ρ, ϕ, θ) varies in R3 with the parameter ρ, ϕ, or θ fixed, respectively.
When restricted to the open set (0, ∞) × (−π, π) × (0, π) the corresponding map-
ping in (7.39) is injective. The corresponding transformation formula in (7.32) is
expressed as
dxdydz = sinh ρ sin θ (sinh2 ρ + sin2 θ) dρdϕdθ.
Oblate spheroidal coordinates for R3 The oblate spheroidal coordinates are
based on the formulas:

 x = cosh ρ sin ϕ sin θ,

(7.40) y = cosh ρ cos ϕ sin θ, (ρ, ϕ, θ) ∈ R3 .

z = sinh ρ cos θ,

Then one sees that the following identities hold:


z2 x2 + y 2 z2 x2 + y 2
+ = 1, (cos ϕ) x = (sin ϕ) y, − = −1.
sinh2 ρ cosh2 ρ cos2 θ sin2 θ
Hence, it is clear that (7.40) produces ellipsoids, half-planes, and one-sheeted hyper-
boloids, when (ρ, ϕ, θ) varies in R3 with the parameter ρ, ϕ, or θ fixed, respectively.
When restricted to the open set (0, ∞) × (−π, π) × (0, π) the corresponding map-
ping in (7.40) is injective. The corresponding transformation formula in (7.32) is
expressed as
dxdydz = cosh ρ sin θ (sinh2 ρ + cos2 θ) dρdϕdθ.
Spherical coordinates for Rd . Spherical coordinates can be introduced in higher
dimensional Euclidean spaces Rd with d ≥ 3. To make the angles unique the
coordinates will be given for the proper subspace
e d = Rd \ {x ∈ Rd : x1 = 0, x2 ≥ 0}.
R
3. SOME COORDINATE SYSTEMS 111

Let x ∈ Rd , x 6= 0, be of the form x = (x1 , . . . , xd ) and let


x21 + · · · + x2d = r2 , r > 0.
Since x ∈ R e d , x 6= 0, it follows that x2 + x2 > 0. Hence there is precisely one
1 2
angle θd−2 with 0 < θd−2 < π, such that xd = r cos θd−2 . This leads then to
x21 + · · · + x2d−1 = r12 , where r1 = r sin θd−2 . In fact this is the first step in a process
of successively splitting off angles: for k = 1, . . . , d − 3 one obtains with r0 = r
(
xd−k+1 = rk−1 cos θd−k−1 , 0 < θd−k−1 < π,
x21 + · · · + x2d−k = rk2 , rk = rk−1 sin θd−k−1 ,
and for k = d − 2 one obtains xd = r cos θd−2 :
(
x3 = rd−3 cos θ1 , 0 < θ1 < π,
x21 + x22 = rd−2
2
, rd−2 = rd−3 sin θ1 .
The last step shows that there is precisely one angle θ0 with, this time, −π < θ0 < π,
such that 1
x2 = rd−2 cos θ0 , x1 = rd−2 sin θ0 , −π < θ0 < π.
The introduction of the various angles above leads to the following mapping Φm
(7.41) (r, θ0 , θ1 , · · · , θm−2 ) → (x1 , x2 , · · · , xm ),
given by
x1 = sin θ0 sin θ1 sin θ2 sin θ3 · · · · · · · · · sin θm−3 sin θm−2 r
x2 = cos θ0 sin θ1 sin θ2 sin θ3 · · · · · · · · · sin θm−3 sin θm−2 r
x3 = cos θ1 sin θ2 sin θ3 · · · · · · · · · sin θm−3 sin θm−2 r
(7.42) x4 = cos θ2 sin θ3 · · · · · · · · · sin θm−3 sin θm−2 r
························
xm−1 = cos θm−3 sin θm−2 r
xm = cos θm−2 r
Proposition 7.11. The mapping Φm given by (7.41) and (7.42) is a diffeomor-
phism from (0, ∞) × (−π, π) × (0, π)m−2 onto R
e d , and the corresponding Jacobian
determinant is given by
(7.43) rm−1 sin θ1 (sin θ2 )2 · · · (sin θm−2 )m−2 .
Proof. It follows from the above construction that the function is C 1 and that
it gives a bijection between (0, ∞) × (−π, π) × (0, π)m−2 and R e m . To determine
the Jacobian determinant, use induction. It is clear that (7.43) is valid for m = 2.
Now assume that (7.43) is valid for some m − 1 (≥ 2).
Denote the diffeomorphism in (7.41) by Φm . Then clearly the mapping Λm
defined by
(7.44) (ρ, θ0 , θ1 , · · · , θm−3 , xm ) → (Φm−1 (ρ, θ0 , θ1 , · · · , θm−3 ), xm ),
defines a diffeomorphism from
(0, ∞) × (−π, π) × (0, π)m−3 × R → R
e d−1 × R = R
e d,

1The angle θ is counted counterclockwise from the positive x -axis.


0 2
112 7. INTEGRATION IN Rd

whose Jacobian determinant (by induction hypothesis) is given by


(7.45) ρm−2 sin θ1 (sin θ2 )2 · · · (sin θm−3 )m−3 .
Furthermore, define the mapping Ψm by
(7.46) (r, θ0 , θ1 , · · · , θm−3 , θm−2 ) → (ρ, θ0 , θ1 , · · · , θm−3 , xm ),
with ρ and xm given by
(7.47) ρ = r sin θm−2 , xm = r cos θm−2 .
Then Ψm is a diffeomorphism from
(0, ∞) × (−π, π) × (0, π)m−2 → (0, ∞) × (−π, π) × (0, π)m−3 × R,
whose Jacobian determinant is r. It is straightforward to see that
Φm = Λm ◦ Ψm ,
so that Φ0m = (Λ0m ◦ Ψm )Ψ0m . Therefore det Φ0m = det(Λ0m ◦ Ψm ) det Ψ0m can be
determined by (7.45):
[ρm−2 sin θ1 (sin θ2 )2 · · · (sin θm−3 )m−3 ]r,
where ρ is given by (7.47). This shows that (7.43) is valid for m. 

Let f : Rm → R be a measurable function and define the function g by g =


f ◦ Φm . Then
Z Z ∞ Z 2π Z π Z π
f (x) dx = ... g(r, θ)J(r, θ) drdθ,
Rm 0 0 −π −π

where θ = (θ0 , . . . , θm ), dθ = dθ0 · · · dθm , and


J(r, θ) = rm−1 sin θ1 (sin θ2 )2 · · · (sin θm−2 )m−2 .
The function f is integrable if and only the function (r, θ) 7→ g(r, θ)J(r, θ) is inte-
grable.

4. Lebesgue integration on manifolds in Rd


This section gives a brief overview of Lebesgue integration for extended real-
valued functions defined on a k-dimensional differentiable manifold M which is
embedded in Rd . In order to integrate extended real-valued functions on a differen-
tiable manifold M one has to define a σ-algebra on M and then to define a measure
on that σ-algebra. Since a differentiable manifold is made up of differentiable co-
ordinate patches - which overlap differentiably - the essential step is to define a
σ-algebra and a measure on a coordinate patch.
Definition 7.12. Let X ⊂ Rk , 1 ≤ k ≤ d, be an open set and let α : X → Rd
be an injective C r -mapping, r ≥ 1. Assume that the derivative Dα, given by the
d × k matrix-valued function
 
dα dα
(7.48) Dα = ······ ,
dx1 dxk
has maximal rank. Then the set M = α(X) is called a C r -patch in Rd and the
mapping α is called its parametrization.
4. LEBESGUE INTEGRATION ON MANIFOLDS IN Rd 113

Associated with the parametrization α is the volume-function V (Dα) from X


to [0, ∞), defined by
(V (Dα))(x) = V ((Dα)(x)), x ∈ X,
where the function V : Rd × · · · × Rd → [0, ∞) given by
q
V (v1 , . . . , vk ) = det (vi , vj )i,j=1,...,k ,

determines the volume of the parallelepiped spanned by the vectors v1 , . . . , vk ∈ Rd .


It is clear that the function V (Dα) is of class C r−1 , so that it is continous and hence
Lebesgue measurable.
Definition 7.13. Let M be a C r -patch in Rd as in Definition 7.12. The set
A is defined as the collection of all subsets A ⊂ M for which α−1 (A) is Lebesgue
measurable. The mapping µ : A → [0, ∞] is defined by
Z
(7.49) µ(A) = V (Dα) dmk , A ∈ A.
α−1 (A)

The definition of µ : A → [0, ∞] can be geometrically motivated. For instance,


let A ⊂ M be the image by α of a rectangle in Rk . A partition of the rectangle
leads to a collection of parallelepipeds in Rd whose total volume has a limit, namely
µ(A), when the refinement of the partition goes to 0.
Proposition 7.14. The triple (M, A, µ) as defined in Definition 7.13 is a σ-
algebra. In particular, the measure µ ◦ α is absolutely continuous with respect to the
Lebesgue measure m with density V (Dα).
Remark 7.15. Let (Ω, A) be a measurable space and let ϕ be a one-to-one
correspondence between Ω and Ω0 . Define A0 as the collection of all ϕ(A) ⊂ Ω0 ,
A ∈ A. Then the mappings ϕ and ϕ−1 are measurable between (Ω, A) and (Ω0 , A0 ).
Proof. The space (X, M(X)) is measurable and the mapping ϕ is injective.
Due to Definition 7.13 and Remark 7.15 it follows that (M, A) is a measure space
and that ϕ and ϕ−1 are measurable.
To see that µ is a measure, let B be a measurable set in X. Then according to
(7.49) one has Z
µ(α(B)) = V (Dα) dmk .
B
Since V (Dα) is a nonnegative measurable function, it follows from Theorem 4.39
that µ ◦ α is a measure on (X, M(X)), which is absolutely continuous with respect
to the Lebesgue measure mk with density V (Dα). Observe that
µ = (µ ◦ α) ◦ α−1 .
Since the mapping α is measurable, it follows from Theorem 4.40 that µ is a mea-
sure. 
Now Definition 7.12, Definition 7.13, and Proposition 7.14 will be combined.
Proposition 7.16. Let M be a C r -patch in Rd and let (M, A, µ) be the corre-
sponding measure space. Let f : M → R be an extended real-valued function. Then
f is measurable with respect to A if and only if f ◦ α is Lebesgue measurable with
respect to M(X).
114 7. INTEGRATION IN Rd

Proposition 7.17. Let M be a C r -patch in Rd and let (M, A, µ) be the corre-


sponding measure space. Let f : M → R be an extended real-valued function. Then
the following statements hold.
(1) If the function f is nonnegative, then
Z Z
(7.50) f dµ = (f ◦ α) V (Dα) dmk .
M X
(2) The function f is integrable with respect to µ if and only if the function
f ◦ α is integrable with respect to V (Dα) dmk , in which case (7.50) holds.
Proof. (1) For f = 1A , A ∈ A, the statement coincides with (7.49). By
linearity the result follows for simple functions. A limiting procedure extends the
result to nonnegative measurable functions.
(2) For every measurable function f it follows from (1) that
Z Z
|f | dµ = |f ◦ α| V (Dα) dmk .
Ω0 Ω
Hence f is integrable with respect to µ if and only if (f ◦α)V (Dα) is integrable with
respect to mk . In this case the equality in (7.50) is obtained via the components
f + and f − , separately. 
Example 7.18. Let X ⊂ R be an open set and let α : X → R3 be an injective
mapping which belongs to C r , r ≥ 1, and whose derivative does not vanish. The
image α(X) is a one-dimensional C r -manifold in R3 , consisting of a single patch,
which is parametrized by α with
α(x) = ( α1 (x), α2 (x), α3 (x) )> , x ∈ X.
It is clear that the derivative Dα of α in (7.48) is given by the 3 × 1 matrix function
 >
∂α1 ∂α2 ∂α3
Dα = , , ,
∂x ∂x ∂x
which has maximal rank. The volume function V (Dα) is given by
s 2  2  2
∂α1 ∂α2 ∂α3
V (Dα) = + + .
∂x ∂x ∂x
Therefore the measure in Definition 7.13 has the following arc-length appearance,
well-known from calculus,
s 2  2  2
Z
∂α1 ∂α2 ∂α3
µ(A) = + + dx, A ∈ A,
α−1 (A) ∂x ∂x ∂x
where dx stands for the one-dimensional Lebesgue measure on X ⊂ R.
Example 7.19. Let X ⊂ R2 be an open set and let h : X → R be a function
which belongs to C r , r ≥ 1. The graph of this function
(x, y, h(x, y)), (x, y) ∈ X,
is a two-dimensional C r -manifold in R3 , consisting of a single patch. This patch is
parametrized by the mapping α : X → R3 when
α(x, y) = (x, y, h(x, y)), (x, y) ∈ X.
4. LEBESGUE INTEGRATION ON MANIFOLDS IN Rd 115

It is clear that the derivative Dα of α in (7.48) is given by the 3 × 2 matrix function


 
1 0
Dα =  0 1 ,
∂h ∂h
∂x ∂y

which has maximal rank. The volume function V (Dα) is given by


s  2  2
∂h ∂h
V (Dα) = 1 + + .
∂x ∂y
Therefore the measure in Definition 7.13 has the following surface area appearance,
well-known from calculus,
s  2  2
Z
∂h ∂h
µ(A) = 1+ + dxdy, A ∈ A,
−1
α (A) ∂x ∂y
where dxdy stands for the 2-dimensional Lebesgue measure on X ⊂ R2 .
It may look from Definition 7.13 that the definition of the σ-algebra (M, A) and
the definition of the measure µ on (M, A) are dependent on the parametrization
α : X → Rk . However it follows from the transformation formula in Theorem 7.8
and Corollary 7.9 that this is not the case.
Proposition 7.20. The σ-algebra (M, A) and the measure µ in Definition 7.13
are independent of the parametrization of M .
Proof. Assume that there is another parametrization for the patch M . In
other words, let Y ⊂ Rk be open and let β : Y → Rd be an injective C r -mapping
such that M = β(Y ). Then it can be shown that
ϕ = β −1 ◦ α
is a C r -diffeomorphism from X onto Y . Therefore the mappings ϕ and ϕ−1 are
measurable.
Let f : M → R be an extended real-valued function. Then α = β ◦ ϕ shows
that
f ◦ α = (f ◦ β) ◦ ϕ.
Hence measurability of functions is independent of the parametrization. Further-
more, observe that α = β ◦ ϕ implies via the chain rule that
Dα = (Dβ ◦ ϕ) ϕ0 ,
cf. (7.48), which leads to
(7.51) V (Dα) = (V (Dβ) ◦ ϕ) | det ϕ0 |.
Now let A ∈ A and observe that β −1 (A) = ϕ(α−1 (A)). Hence, according to
Theorem 7.8 and Corollary 7.9 one obtains
Z Z
V (Dβ) dmk = V (Dβ) dmk
β −1 (A) ϕ(α−1 (A))
Z
= (V (Dβ) ◦ ϕ) | det ϕ0 | dmk
α−1 (A)
Z
= V (Dα) dmk ,
α−1 (A)
116 7. INTEGRATION IN Rd

due to (7.51). This shows that the definition of the measure µ in (7.49) is indepen-
dent of the representation. 
Recall that a manifold M in Rd is a subset of Rd which can be viewed as a col-
lection of patches, which “overlap differentiably”. The totality of the corresponding
coordinate charts forms a countable atlas for the manifold. A set A ⊂ M is called
measurable if each point of M is contained in a patch whose intersection with A
is measurable in that patch. In fact, the measurable sets thus defined form a σ-
algebra on the manifold M , independent of the particular choice of atlas. In order
to obtain a measure on M let the atlas be given by patches Mi , parametrized by
αi : Xi ⊂ Rk → Rd with Mi = αi (Xi ), and denote the corresponding measure by
µi , i ∈ N. Let A ⊂ M be measurable, then the inductively defined sets
n
!
[
A1 = A ∩ M1 , An+1 = (A ∩ Mn+1 ) \ Ai
i=1
S∞
are measurable (in each patch) and disjoint, such that A = i=1 Ai . With this
decomposition one defines the function µ : M → [0, ∞] by
X∞
µ(A) = µi (Ai ).
i=1
The definition of the function µ is independent of the particular choice of atlas.
Moreover, it turns out that µ is a measure on the measurable sets of M .
CHAPTER 8

Spaces of integrable functions

Let (Ω, A, µ) be a measure space.R The integrable functions on Ω form a linear


space L1 (Ω, A, µ) and clearly kf k = Ω |f | dµ (with f integrable) provides a semi-
norm on that space. By making equivalence classes of integrable functions which
differ on a set measure 0, one then obtains a norm with which the space L1 (Ω, A, µ)
(of equivalence classes) is complete. Likewise, the space Lp (Ω, A, µ) of functions f
for which |f |p is integrable, is complete. These spaces are widely used in various
branches of mathematics. This chapter provides a short introduction and presents
the most commonly used properties of these spaces.

1. Semi-normed linear spaces


A linear space X over C is called semi-normed if the exists a semi-norm on X,
i.e. a function k · k : X → R, such that
(1) khk ≥ 0, h ∈ X;
(2) kλhk = |λ|khk, h ∈ X, λ ∈ C;
(3) kh + kk ≤ khk + kkk, h, k ∈ X.
With a semi-normed space one may speak about the convergence of sequences. A
sequence (hn ) in X is said to converge to h ∈ X if for every ε > 0 there exists N ∈ N
for which khn − hk < ε when n ≥ N , i.e., if khn − hk → 0 as n → ∞. A Cauchy
sequence in X is a sequence hn ∈ X such that for every ε > 0 there exists N ∈ N for
which khn − hm k < ε when m, n ≥ N . The semi-normed linear space X is called
complete if every Cauchy sequence in X converges in X. The following observation
is sometimes useful.
Lemma 8.1. A Cauchy sequence converges if it has a convergent subsequence.
Proof. For every ε > 0 there exists N ∈ N such that khn − hm k < ε/2 for
m, n ≥ N . Let hnk be the subsequence of hn which converges to h ∈ X. Hence by
definition of convergence there exists an index nk > N such that khnk − hk < ε/2.
With n ≥ N it now follows that
kh − hn k ≤ kh − hnk k + khnk − hn k < ε.
In other words the Cauchy sequence hn converges to h in X. 
Note that in semi-normed linear spaces limits are not uniquely determined; in
fact they are determined by the collection of all neutral elements in X:
N = {h ∈ X : khk = 0}.
Note that if hn → h in X and k ∈ N then also hn → h + k, as can be seen from
khn − (h + k)k ≤ khn − hk + kkk = khn − hk.
In fact this reasoning leads to the following result.
117
118 8. SPACES OF INTEGRABLE FUNCTIONS

Lemma 8.2. Let X be a semi-normed linear space. Then N is a closed linear


subspace of X.
Proof. It follows from the definition that N is a linear subspace of X. Let
hn ∈ N converge in X to h ∈ X, so that khn − hk → 0 as n → ∞. Then h =
h − hn + hn implies that
khk ≤ kh − hn k + khn k = kh − hn k,
so that khk = 0 and h ∈ N. 
It is sometimes useful to express completeness of a semi-normed
P∞ space in terms
of series rather than in terms of Cauchy sequences. A seriesP n=1 fn with terms

in X is convergent
P∞ if the partial sums converge in X. A series n=1 fn is absolutely
convergent if n=1 kfn k < ∞. The connection with Cauchy sequences is shown in
the following construction.

Let (hn ) be a Cauchy sequence in the semi-normed space X. Hence for εk = 2−k
there exists Nk so that
(8.1) khn − hm k ≤ 2−k , n, m ≥ Nk .
Thus there is a subsequence (hnk ), such that
(8.2) khnk+1 − hnk k ≤ 2−k , k ∈ N.
Now define the elements fk = hnk+1 − hnk , so that by construction

X
(8.3) kfk k < ∞,
k=1
P∞
i.e., the series k=1 fk converges absolutely. The connection between the Cauchy
sequence and the elements fn can be expressed by the identity
N
X
(8.4) hnN +1 − hn1 = fk .
k=1

P∞from (8.4) that the subsequence hnk converges in X if and only if the
It follows
series k=1 fk converges in X.
Proposition 8.3. Let X be a semi-normed linear space. Then the following
statements are equivalent:
(1) X is complete;
(2) every absolutely convergent series in X converges in X.
Proof. (1) ⇒ (2) The partial sums of an absolutely convergent series form a
Cauchy sequence.
(2)P⇒ (1) Let (hn ) be a Cauchy sequence in X and construct the associated

series k=1 fk as above. Since the series converges absolutely, cf. (8.3), it follows
from the assumption that the series converges. Hence (8.4) implies that the subse-
quence hnk converges in X, which by Lemma 8.1 implies that the Cauchy sequence
converges in X. 
The semi-normed linear space X can be turned into a normed linear space
by forming appropriate equivalence classes relative to the closed subspace N. For
h, k ∈ X define h ∼ k if h − k ∈ N. It is easily seen that this gives an equivalence
2. MINKOWSKI AND HÖLDER INEQUALITIES 119

relation on X. Denote by X/N the set of the corresponding equivalence classes.


Clearly X/N is a linear space with the following operations
α[h] = [αh], α ∈ C, h ∈ X,
and
[h] + [k] = [h + k], h, k ∈ X.
The linear space X/N provided with
k[h]k = khk, h ∈ X,
is a normed linear space. A complete noremd linear space is called a Banach space.
Lemma 8.4. Let X be a semi-normed linear space and let N be the linear space
of all neutral elements. If X is complete, then X/N is a Banach space.
Proof. For any elements h, k ∈ X one has according to the above definition
k[h] − [k]k = k[h − k]k = kh − kk.
Hence [hn ] is a Cauchy sequence in X/N if and only if hn is a Cauchy sequence in
X. Assume that X is complete and let [hn ] be a Cauchy sequence in X/N. Then
hn is a Cauchy sequence in X and, due to the completeness of X, there exists an
element h ∈ X such that hn → h in X. In other words
k[hn ] − [h]k = k[hn − h]k = khn − hk → 0, n → ∞.
Hence [hn ] → [h] in X/N as n → ∞. Therefore the normed linear space X/N is
complete. 

2. Minkowski and Hölder inequalities


Lemma 8.5. Let α, β ∈ R be nonnegative and let 1 ≤ p < ∞ and 1/p + 1/q = 1.
Then
α β
(8.5) α1/p β 1/q ≤ + ,
p q
and
 p
α+β 1
(8.6) ≤ (αp + β p ).
2 2
Proof. It is clear that (8.5) and (8.6) hold for p = 1. For the rest assume
therefore that p > 1. Observe that for t ≥ 1 one has the inequalities
 p
t 1 t+1 1
t1/p ≤ + , ≤ (tp + 1).
p q 2 2
To see this, observe that in each case there is equality for t = 1 and in each case
the derivative of the left-hand side is smaller than the derivative of the right-hand
side for t > 1. Next substitute α/β or β/α depending on the size to obtain (8.5)
and (8.6), respectively. 
Definition 8.6. Let (Ω, A, µ) be a measure space and let 1 ≤ p ≤ ∞. For
p < ∞ the space Lp (Ω) is the set of all measurable complex-valued functions f for
which
Z
(8.7) |f |p dµ < ∞.

120 8. SPACES OF INTEGRABLE FUNCTIONS

Furthermore, one defines for f ∈ Lp (Ω)


Z 1/p
p
(8.8) kf kp = |f | dµ .

For p = ∞ the space L (Ω) is the set of all measurable complex-valued functions

f for which there exists c ≥ 0 such that


(8.9) |f (ω)| ≤ c almost everywhere.
Furthermore, one defines for f ∈ L (Ω) ∞

(8.10) kf k∞ = inf { c ≥ 0 : |f (ω)| ≤ c almost everywhere }.

Lemma 8.7. The space Lp (Ω), 1 ≤ p ≤ ∞, is a linear space.


Proof. To see this, let f, g in Lp (Ω). Since f and g are measurable, also the
sum f + g is measurable.
First let 1 ≤ p < ∞. Observe that the inequality (8.6) implies that
|f (ω) + g(ω)|p ≤ 2p−1 (|f (ω)|p + |g(ω)|p ).
Hence, by (8.7) it follows that f + g ∈ Lp (Ω).
Now let p = ∞. Then |f (ω)| ≤ c almost everywhere and |g(ω)| ≤ d almost
everywhere. Hence it follows that |f (ω) + g(ω)| ≤ c + d almost everywhere. Hence
by (8.9) it follows that f + g ∈ L∞ (Ω).
Therefore Lp (Ω) is a linear space. 

Remark 8.8. If f ∈ L∞ (Ω) then the set {ω ∈ Ω : |f (ω)| ≤ c} with c in (8.9)


is measurable and its complement has measure 0. The quantity kf k∞ in (8.9) itself
satisfies
(8.11) |f (ω)| ≤ kf k∞ almost everywhere.
For this reason kf k∞ is called the essential supremum of |f | and is sometimes
denoted by
kf k∞ = ess supω∈Ω |f (ω)|.
To prove (8.11) introduce the measurable sets A and An by
A = {ω ∈ Ω : kf k∞ < |f (ω)|}, An = {ω ∈ Ω : kf k∞ < |f (ω)| − 1/n},
S∞
and note that A = n=1 An . Then each An has measure 0, so that µ(A) = 0,
which shows the claim. To see that each An has measure 0, assume that there
exists N ∈ N such that
(8.12) µ{ω ∈ Ω : |f (ω)| > kf k∞ + 1/N } > 0.
Since kf k∞ is the infimum of all c ∈ R for which (8.9) holds there exists c0 with
kf k∞ ≤ c0 < kf k∞ + 1/N , such that
µ{ω ∈ Ω : |f (ω)| > c0 } = 0.
However, it is clear that
{ω ∈ Ω : |f (ω)| > kf k∞ + 1/N } ⊂ {ω ∈ Ω : |f (ω)| > c0 },
which contradicts (8.12).
2. MINKOWSKI AND HÖLDER INEQUALITIES 121

Theorem 8.9 (Hölder’s inequality). Let 1 ≤ p ≤ ∞ and 1/p + 1/q = 1. If


f ∈ Lp (Ω) and g ∈ Lq (Ω), then f g ∈ L1 (Ω) and
kf gk1 ≤ kf kp kgkq .
In particular, for p = 2 and q = 2 this is the Cauchy-Schwarz inequality.
Proof. First consider the case 1 < p < ∞. If either kf kp = 0 of kgkq = 0 the
result is straightforward. Thus it suffices to assume that kf kp 6= 0 and kgkq 6= 0.
In this case apply the inequality (8.5) with
 p  q
|f (ω)| |g(ω)|
α= , β= .
kf kp kgkq
This results in  p  q
|f (ω)g(ω)| 1 |f (ω)| 1 |g(ω)|
≤ + .
kf kp kgkq p kf kp q kgkq
Integration of this inequality leads to
|f (ω)g(ω)|
Z
1 1
dµ ≤ + = 1,
Ω kf kp kgkq p q
which gives the desired result.
Now consider p = 1 so that q = ∞. Then by Remark 8.8
|f (ω)g(ω)| ≤ |f (ω)|kgk∞ almost everywhere.
Integration gives the desired result.
The case p = ∞ and q = 1 is treated in the same way. 
Corollary 8.10. Let 1 ≤ p ≤ ∞ and assume that µ(Ω) = 1. If f ∈ Lp (Ω)
then f ∈ L1 (Ω) and
kf k1 ≤ kf kp .
Theorem 8.11 (Minkowski’s inequality). Let 1 ≤ p ≤ ∞. The space Lp (Ω)
with k · kp defined by (8.8) is a semi-normed linear space. In particular for f, g ∈
Lp (Ω)
(8.13) kf + gkp ≤ kf kp + kgkp .
A function f ∈ L (Ω) has kf kp = 0 if and only if f = 0 almost everywhere.
p

Proof. First consider the case 1 < p < ∞. Observe that


|f (ω) + g(ω)|p = |f (ω) + g(ω)||f (ω) + g(ω)|p−1 ,
so that by the triangle inequality
(8.14) |f (ω) + g(ω)|p ≤ |f (ω)||f (ω) + g(ω)|p−1 + |g(ω)||f (ω) + g(ω)|p−1 .
The terms in the right-hand side of (8.14) contain the function |f + g|p−1 . With
1/p + 1/q = 1 observe that |f + g|p−1 ∈ Lq (Ω) and that, in fact,
Z 1/q
p−1 (p−1)q
k|f + g| kq = |f + g| dµ = kf + gkp−1
p .

Now integrate (8.14) and apply Hölder’s inequality, so that
kf + gkpp ≤ (kf kp + kgkp ) kf + gkp−1
p ,
which shows (8.13).
122 8. SPACES OF INTEGRABLE FUNCTIONS

Now consider the case p = 1. Then integration of


(8.15) |f (ω) + g(ω)| ≤ |f (ω)| + |g(ω)|,
leads to the desired result.
Finally consider the case p = ∞. It follows from (8.15) and Remark 8.8 that
|f (ω) + g(ω)| ≤ kf k∞ + kgk∞ almost everywhere.
This leads to the desired result. 

3. Completeness
The completeness of the spaces Lp (Ω), 1 ≤ p ≤ ∞, will now be considered. For
the case 1 ≤ p < ∞ the equivalence in Proposition 8.3 will be used. As a bonus
there is a statement about pointwise convergence.
Theorem 8.12. Let (Ω, A, µ) be a measure P∞space and let 1 ≤ p < ∞. Let
fn ∈ Lp (Ω),
P∞ n ∈ N, and assume that the series k=1 fk converges absolutely. Then
the series k=1 fk converges in Lp (Ω), i.e., there exists a function f ∈ Lp (Ω), such
that
X∞
(8.16) fk = f,
k=1

where the series converges in L (Ω). In addition, the series in (8.16) converges
p

pointwise almost everywhere.


P∞
Proof. Assume that the series k=1 fk converges absolutely in Lp (Ω), so that

X
(8.17) kfk kp = M < ∞.
k=1

Define the functions hn , h : Ω → [0, ∞] by


n
X ∞
X
(8.18) hn (ω) = |fk (ω)|, h(ω) = |fk (ω)|, ω ∈ Ω.
k=1 k=1

Then hn and h are measurable, and hn is monotonically increasing and converging


pointwise to h. By the monotone convergence theorem one obtains that
Z Z
p
(8.19) h dµ = lim hpn dµ.
Ω n→∞ Ω
By Minkowski’s inequality (for the sum of the functions |f1 |, . . . , |fn |) and (8.17) it
follows that
Xn
(8.20) khn kp ≤ kfk kp ≤ M,
k=1

and hn ∈ L (Ω). It follows from (8.19) and (8.20) that h ∈ Lp (Ω) with
p

(8.21) khkp ≤ M,
and, in particular, h is finite almost everywhere. Hence, according to (8.18), the
series
X∞
|fk (ω)|
k=1
3. COMPLETENESS 123

converges almost everywhere. Define the function f on Ω by


(P
∞ P∞
k=1 fk (ω) if k=1 |fk (ω)| < ∞,
f (ω) =
0 otherwise.
Then f is a measurable function with |f | ≤ h, so that f ∈ Lp (Ω). Due to this
estimate and (8.18) one sees
n p
X p
f− fk ≤ (|f | + hn ) ≤ (2h)p ∈ L1 (Ω).
k=1

Hence the dominated convergence theorem implies that


Z n p
X
f− fk dµ → 0,
Ω k=1
Pn
meaning that k=1 fk converges to f in the sense of Lp (Ω). 
Theorem 8.13. Let 1 ≤ p ≤ ∞. The semi-normed linear space Lp (Ω) is
complete.
Proof. First consider the case 1 ≤ p < ∞. Then the statement is a direct
consequence of Proposition 8.3 and Theorem 8.12.
Next consider the case p = ∞. Let fn be a Cauchy sequence in L∞ (Ω). Let
Ak and Bm,n be the sets where
|fk (ω)| > kfk k∞ , |fn (ω) − fm (ω)| > kfn − fm k∞ ,
and let E be the union of all these sets. Then µ(E) = 0 and on Ω \ E the sequence
fn converges uniformly to a bounded function f . Define f (ω) = 0 for ω ∈ E. Then
f ∈ L∞ (Ω) and kfn − f k∞ → 0 as n → ∞. 
Note that a Cauchy sequence in L∞ (Ω) converges pointwise almost everywhere.
A similar statement is not valid in Lp (Ω), but there is a weaker statement.
Corollary 8.14. Let 1 ≤ p < ∞. Each Cauchy sequence in Lp (Ω) contains
a subsequence which converges pointwise almost everywhere.
Proof. Choose a Cauchy sequence hn in Lp (Ω) and choose a subsequenceP∞ hnk
as in (8.2). Define the elements fk = hnk+1 − hnk Due to (8.4) the series
P∞ k=1 fk
converges absolutely. It follows from Theorem 8.12 that the series k=1 fk not only
converges in Lp (Ω), but it also converges pointwise almost everywhere. In other
words the Cauchy sequence hn has a subsequence hnk which converges pointwise
almost everywhere. 
For the following discussion it is important to indicate the appropriate mea-
sures, thus the above notions will be written as
Lp (Ω, µ), N(µ), Lp (Ω, µ) = Lp (Ω, µ)/N(µ).
By Theorem 8.13 the semi-normed linear space Lp (Ω, µ) is complete. The space
N(µ) is the closed linear subspace of all elements in Lp (Ω, µ) whose seminorm is
zero. By Lemma 8.4 the completeness of Lp (Ω, µ) guarantees that the quotient
space
Lp (Ω, µ) = Lp (Ω, µ)/N(µ)
124 8. SPACES OF INTEGRABLE FUNCTIONS

is a Banach space (a complete normed linear space). In practice it will always be


the spaces of equivalence classes Lp (Ω, µ) which will be used. As the measure µ in
the above discussion has not been assumed to be complete one might also look at
the completion µ̄ and the corresponding objects
Lp (Ω, µ̄), N(µ̄), Lp (Ω, µ̄) = Lp (Ω, µ̄)/N(µ̄),
and to consider their relationship with the earlier objects. Note that N(µ) ⊂ N(µ̄).
Now let f ∈ Lp (Ω, µ), so that f is measurable and |f |p is integrable with respect
to µ. Then it is clear that f is measurable and |f |p is integrable with respect to µ̄,
so that f ∈ Lp (Ω, µ̄) and, moreover,
Z Z
|f |p dµ = |f |p dµ̄.
Ω Ω
Thus it follows that
[f ]µ → [f ]µ̄
is a well defined linear mapping from Lp (Ω, µ)/N(µ) to Lp (Ω, µ̄)/N(µ̄), which pre-
serves the norm:
k[f ]µ kp = k[f ]µ̄ kp .
It is called the natural embedding from Lp (Ω, µ)/N(µ) into Lp (Ω, µ̄)/N(µ̄).
Theorem 8.15. The natural embedding from Lp (Ω, µ)/N(µ) into Lp (Ω, µ̄)/N(µ̄)
is surjective.
Proof. Let f ∈ Lp (Ω, µ̄). Then f is measurable and |f |p is integrable with
respect to µ̄. According to Theorem 3.28 there exists a function g : (Ω, A) → (R, B)
such that
(8.22) µ̄({ω ∈ Ω : f (ω) 6= g(ω)}) = 0,
i.e. g = f almost everywhere with respect to µ̄. Furthermore, g may be constructed
such that g(ω) = 0 if f (ω) 6= g(ω). According to Proposition 4.15 one has
Z Z
p
(8.23) |f | dµ̄ = |g|p dµ̄,
Ω Ω
while by Corollary 4.7 one sees that
Z Z
(8.24) |g|p dµ̄ = |g|p dµ,
Ω Ω
which completes the proof. 
The basic message of Theorem 8.15 is that there is no difference between the
spaces Lp (Ω, µ) and Lp (Ω, µ̄).

4. Dense linear subsets


Theorem 8.16. Let (Ω, A, µ) be a measure space and let 1 ≤ p ≤ ∞. The
simple functions which belong to Lp (Ω) are dense in Lp (Ω).
Proof. It suffices to consider the case of nonnegative functions f ∈ Lp (Ω).
The general result then follows by decomposition in positive and negative parts.
First consider the case 1 ≤ p < ∞. Let f ∈ Lp (Ω) be nonnegative. Then by
Theorem 3.19 there is a nondecrasing sequence of simple functions fn such that
fn ↑ f . Hence fn ∈ Lp (Ω) and
|f − fn |p ≤ f p ∈ L1 (Ω),
4. DENSE LINEAR SUBSETS 125

which shows dominated convergence. Therefore Theorem 4.30 implies that


Z
|f − fn |p dµ → 0 as n → ∞.

Now consider the case p = ∞. Let f ∈ L∞ (Ω) be nonnegative. It is possible
to change the function f on a set of measure 0 to make it bounded. Then by
Theorem 3.19 there is a nondecrasing sequence of simple functions fn such that
fn ↑ f uniformly. Hence fn → f in L∞ (Ω). 
The conditions in the theorem need some explanation. Note that a simple
function of the form
Xn
ci 1Ai
i=1
belongs to Lp (Ω) for 1 ≤ p < ∞ if and only if µ(Ai ) < ∞ for every Ai for which
ci 6= 0. However, every simple function belongs to L∞ (Ω).
In the context of Rd one may use Theorem 8.16 to obtain a very useful approx-
imation result in terms of continuous functions with compact support. A complex-
valued function f on Rd is said to have compact support if the support, i.e. the
closure of the set {x ∈ Rd : f (x) 6= 0} is bounded. Define Cc (Rd ) as the collection
of all complex-valued functions on Rd which are continuous and which have com-
pact support. It is clear that Cc (Rd ) is a linear space and that it is contained in
Lp (Rd ), 1 ≤ p ≤ ∞.
Theorem 8.17. The space Cc (Rd ) is dense in Lp (Rd ), 1 ≤ p < ∞.
Proof. Let f ∈ Lp (Rd ) and let ε > 0. Then by Theorem 4.30 there exists an
N such that R = [−N, N ]d is compact in Rd and such that
Z
|f − f 1R |p dm < ε.
Rd

Hence it suffices to assume that f ∈ Lp (Rd ) has compact support. Furthermore,


it suffices to consider the case of a nonnegative function, as the general result then
follows by decomposition in positive and negative parts.
So let f be a nonnegative measurable function in Lp (Ω) with compact support.
By Theorem 3.19 there is a nondecreasing sequence of compactly supported simple
functions fn such that fn ↑ f . Hence fn ∈ Lp (Rd ) and
|f − fn |p ≤ f p ∈ L1 (Rd ),
so that Z
|f − fn |p dm → 0 as n → ∞.
Rd
Recall that every simple function is a finite linear combination of characteristic
functions.
Hence it suffices to show that the characteristic function 1A where A ⊂ Rd is
compact and measurable can be approximated in the above sense. For ε > 0 there
exists a bounded open set O ∈ Rd and a compact set F such that F ⊂ A ⊂ O with
m(O \ F ) < ε; cf. Corollary 2.22. Recall that for a closed set C ⊂ Rd the distance
function d(x, C) = inf{|x − y| : y ∈ C} is continuous. Hence the function
d(x, Oc )
g(x) = , x ∈ Rd ,
d(x, F ) + d(x, Oc )
126 8. SPACES OF INTEGRABLE FUNCTIONS

is continous on Rd and has the properties 0 ≤ g ≤ 1 and


(
1, x ∈ F,
(8.25) g(x) =
0, x ∈ Oc .
Hence g ∈ Cc (Rd ) and it follows that
Z
k1A − gkp = |1A − g|p dm ≤ m(O \ F ) < ε.
O\F
Thus the characteristic function 1A can be approximated arbitrarily closely by a
function from Cc (R). 
Remark 8.18. For p = ∞ the situation is slightly more complicated. Denote
the class of all continuous functions on Rd which vanish at ∞ by C0 (Rd ) and provide
this linear space with the sup-norm. It is clear that Cc (Rd ) ⊂ C0 (Rd ) and, in fact,
C0 (Rd ) is the completion of Cc (Rd ) in L∞ (Rd ).
Corollary 8.19. Let f ∈ Lp (Rd ), 1 ≤ p < ∞, then
Z
lim |f (x + t) − f (t)|p dm(t) = 0.
x→0 Rd

Proof. Choose ε > 0 then there exists a function ϕ ∈ Cc (Rd ) such that
kf − ϕkp < ε. Then
Z Z
|f (x + t) − f (t)|p dm(t) ≤ |f (x + t) − ϕ(x + t)|p dm(t)
R d R d
Z Z
+ |ϕ(x + t) − ϕ(t)|p dm(t) + |ϕ(t) − f (t)|p dm(t).
Rd Rd
The last and the first integral (translation invariance) in the righthand side are at
most εp . Let K be the support of ϕ and define K1 by
K1 = {x ∈ Rd : d(x, K) ≤ 1}
Then K1 is a compact set and for |x| ≤ 1 the support of the integrand in the middle
integral is contained in K1 . On this compact set ϕ uniformly continuous so there
exists 0 < δ < 1 such that |ϕ(x + t) − ϕ(t)| ≤ ε for |x| < δ. 
CHAPTER 9

Decompositions of measures

Let λ and µ be σ-finite measures on a measurable space. The measure λ is


absolutely continuous with respect to the measure µ if there is a so-called Radon-
Nikodym derivative dλ/dµ. The measures λ and µ are mutually singular if they
are supported in some sense by disjoint measurable sets. It will be shown that
λ can be uniquely decomposed into the sum of a relatively well behaved measure
which is absolutely continuous with respect to µ and a rather wild measure which
is mutually singular with µ.

1. Bounded linear functionals on Hilbert spaces


Let X be a normed linear space over C. A linear function F : X → C is called
a bounded linear functional on X if F is linear and if F is bounded, i.e., if
 
|F h|
(9.1) kF k = sup : h ∈ X, h 6= 0 < ∞.
khk
The finite quantity kF k is called the norm of the functional F . The dual space X0
of X is the linear space of bounded linear functionals on X provided with the norm
as defined in (9.1).
Theorem 9.1. Let X be a normed linear space. The dual space X0 is a Banach
space.
Proof. Let Fn be a Cauchy sequence in X0 . Then for every ε > 0 there exists
N ∈ N such that for all m, n ≥ N
kFn − Fm k ≤ ε.
For each h ∈ X one has for m, n ≥ N
|Fm (h) − Fn (h)| ≤ kFm − Fn kkhk ≤ khkε,
so that Fn (h) is a Cauchy sequence in C. Now C is complete so that there exists
an element λh ∈ C such that Fn (h) → λh . Define F : X → C by F (h) = λh , then
clearly F is a linear functional on X. It follows that for n ≥ N
|F (h) − Fn (h)| ≤ khkε,
so that for h ∈ X
|F (h)| ≤ |FN (h)| + khkε ≤ (kFN k + ε)khk.
Hence F is a bounded linear functional and for n ≥ N
|F (h) − Fn (h)|
kF − Fn k = sup ≤ ε.
h∈X,h6=0 khk
Therefore Fn → F in the norm of X0 . 
127
128 9. DECOMPOSITIONS OF MEASURES

Let X be an inner product space over C. Then khk =


p
(h, h) defines a norm
on X and one has the Cauchy-Schwarz inequality
|(h, k)| ≤ khkkkk, h, k ∈ X.
The inner product space X is a Hilbert space if the space with the induced norm
is a Banach space. Note that if X is an inner product space and if k ∈ X, then
F : X → C defined by
(9.2) F (h) = (h, k), h ∈ X,
belongs to X0 and that kF k = kkk. This is clear for k = 0. To see this for k 6= 0
observe that F in (9.2) is a linear function and that |F (h)| = |(h, k)| ≤ khkkkk, so
that kF k ≤ kkk. Furthermore note that
kkk2 = (k, k) = F (k) = |F (k)| ≤ kF kkkk,
shows kkk ≤ kF k. Hence clearly kF k = kkk.
Theorem 9.2 (Riesz representation theorem). Let X be a Hilbert space and
assume that F ∈ X0 . Then there exists a unique element k ∈ X such that F is given
by (9.2).
Proof. If F = 0 then take k = 0. Now assume that F 6= 0. Hence ker F is a
closed subspace of X and
X = ker F ⊕ (ker F )⊥ , (ker F )⊥ 6= {0}.
Choose a non-trivial element ϕ ∈ (ker F )⊥ and normalize ϕ so that F (ϕ) = 1.
Then it follows that
F (h − F (h)ϕ) = F (h) − F (h)F (ϕ) = F (h) − F (h) = 0.
Hence h − F (h)ϕ ∈ ker F while ϕ ∈ (ker F )⊥ , so that
0 = (h − F (h)ϕ, ϕ) = (h, ϕ) − F (h)(ϕ, ϕ).
Define the element k ∈ X by
ϕ
k= .
kϕk2
Then F is given by (9.2). 

2. Majorization
Definition 9.3. Let (Ω, A) be a measurable space and let ν and µ be measures
on (Ω, A). Then ν is majorized by µ if
ν(A) ≤ µ(A), A ∈ A.
Let (Ω, A, µ) be a measure space and let h be a measurable nonnegative function
on Ω with 0 ≤ h ≤ 1 almost everywhere with respect to µ. Then
Z
ν(A) = h dµ, A ∈ A,
A

defines a measure ν which is majorized by µ; cf. Theorem 4.39 and the text following
the theorem.
3. ABSOLUTELY CONTINUOUS MEASURES 129

Theorem 9.4 (Radon-Nikodym derivative in majorization case). Let µ be a


finite measure on a measurable space (Ω, A) and let ν be a measure on (Ω, A) which
is majorized by µ. Then there exists a µ-measurable real-valued function h with
0 ≤ h ≤ 1 almost everywhere with respect to µ, such that
Z
(9.3) ν(A) = h dµ, A ∈ A.
A

The function h is uniquely determined up to sets of µ-measure 0.


Proof. The uniqueness of the function h is easily checked. The proof of the
existence is based on the Cauchy-Schwarz inequality and the Riesz representation
theorem for Hilbert spaces. Note that for f ∈ L2 (Ω, A, µ) one has by the Cauchy-
Schwarz inequality
Z Z Z Z  12
1
2
f dν ≤ |f | dν ≤ |f | dµ ≤ |f | dµ (µ(Ω)) 2 ,
Ω Ω Ω Ω

Hence for f ∈ L (Ω, A, µ) the function


2
Z
f 7→ f dν

is a bounded linear functional on the Hilbert space L2 (Ω, A, µ). Hence by the Riesz
representation theorem there exists a function h, square integrable with respect to
µ, such that Z Z
f dν = f h dµ, f ∈ L2 (Ω, A, µ).
Ω Ω
Note that h is uniquely defined up to a set of µ-measure 0. With f = 1A and A ∈ A
this representation reduces to
Z
(9.4) ν(A) = h dµ, A ∈ A.
A

It is clear that 0 ≤ h ≤ 1 almost everywhere with respect to µ. 

The function h in Theorem 9.4 is called the Radon-Nikodym derivative and it


is usually denoted by

h= .

Note that if ν is dominated by µ then
Z Z
(9.5) g dν = gh dµ,
Ω Ω

for all measurable functions g, cf. Theorem 4.39.

3. Absolutely continuous measures


It will be convenient to form the sum of two measures.
Definition 9.5. Let (Ω, A) be a measurable space and let λ1 and λ2 be measures
on (Ω, A). The sum λ1 + λ2 is defined by
(λ1 + λ2 )(A) = λ1 (A) + λ2 (A), A ∈ A.
130 9. DECOMPOSITIONS OF MEASURES

Clearly, the sum λ1 + λ2 is a measure on (Ω, A) and one has


Z Z Z
f d(λ1 + λ2 ) = f dλ1 + f dλ2 ,
Ω Ω Ω

for all measurable numerical functions f .


Definition 9.6. Let (Ω, A) be a measurable space and let λ and µ be measures.
Then λ is absolutely continuous with respect to µ, denoted by λ  µ, if
A∈A and µ(A) = 0 ⇒ λ(A) = 0.
Note that if λ is majorized by µ, then λ is absolutely continuous with respect to
µ. Let (Ω, A, µ) be a measure space and let h be a measurable nonnegative function
on Ω. Then Z
λ(A) = h dµ, A ∈ A,
A
defines a measure λ which is absolutely continuous with respect to µ; cf. Theorem
??. Clearly λ is majorized by µ when 0 ≤ h ≤ 1 almost everywhere with respect
to µ.
Lemma 9.7. Let (Ω, A) be a measurable space and let λ1 , λ2 , and µ be measures
on (Ω, A). Then
λ1  µ, λ2  µ ⇒ λ1 + λ2  µ.
Proof. Let A ∈ A and µ(A) = 0. Then λ1 (A) = 0 and λ2 (A) = 0. Hence
(λ1 + λ2 )(A) = 0. 

Theorem 9.8 (Radon-Nikodym derivative). Let λ and µ be σ-finite measures


on a measurable space (Ω, A) with λ  µ. Then there exists a nonnegative µ-
measurable extended real-valued function h such that
Z
(9.6) λ(A) = h dµ, A ∈ A.
A

The function h is uniquely determined up to sets of µ-measure 0.


Proof. The uniqueness of the function h is easily checked. The proof of the
existence is based on a reduction to Theorem 9.4. The measures λ and µ give rise
to a new measure α, defined as follows
(9.7) α(A) = λ(A) + µ(A), A ∈ A.
Then it is easily checked that, indeed, α is a measure on (Ω, A) and, moreover, that
both λ and µ are majorized by α.
First consider the case where the measures λ and µ are finite, in which case also
the measure α is finite. By Theorem 9.4 there exist two nonnegative measurable
functions hλ and hµ , such that 0 ≤ hλ ≤ 1 and 0 ≤ hµ ≤ 1, almost everywhere
with respect to µ, and
Z Z
(9.8) λ(A) = hλ dα, µ(A) = hµ dα, A ∈ A.
A A

The function hµ gives rise to the following subsets of Ω:


(9.9) F = {ω ∈ Ω : hµ (ω) > 0}, G = {ω ∈ Ω : hµ (ω) = 0},
3. ABSOLUTELY CONTINUOUS MEASURES 131

and it is clear that these sets are measurable, disjoint, and that Ω = F ∪ G.
Furthermore, define the function h by

(9.10) h= 1F ,

then clearly h is measurable and nonnegative.
The behavior of the measure λ relative to the sets F and G will be considered
separately. By (9.8) and (9.10) one obtains
Z Z Z
(9.11) λ(A) = hλ dα = hµ h dα = h dµ, A ∈ A, A ⊂ F,
A A A
where in the last equality the identity (9.5) has been used. Next observe that
Z
(9.12) µ(G) = hµ dα = 0, λ(G) = 0,
G
where the first identity is due to (9.8) and (9.9), and the second identity follows
from the first identity and the fact that λ  µ.
Now let A ∈ A, and decompose A according to Ω = F ∪ G, so that
A = (A ∩ F ) ∪ (A ∩ G).
Hence, by A ∩ G ⊂ G and by the second identity of (9.12),
λ(A) = λ(A ∩ F ) + λ(A ∩ G) = λ(A ∩ F ).
By A ∩ F ⊂ F and (9.11) it follows then that
Z
λ(A) = λ(A ∩ F ) = h dµ.
A∩F
Since A ∩ G ⊂ G it then follows from the first identity in (9.12) that
Z Z Z
λ(A) = h dµ + h dµ = h dµ,
A∩F A∩G A
which gives (9.6).
Now assume that the measures λ and µ are σ-finite. Then there exist measur-
able sequences Ln , Mn of subsets of Ω with λ(Ln ) and µ(Mn ) finite for all n ∈ N,
and
[∞ [∞
Ω= Ln , Ω = Mn .
n=1 n=1
Without loss of generality assume that the sets Ln are pairwise disjoint and that the
sets Mn are pairwise disjoint; cf Lemma
S∞ 1.4. Then clearly both λ and µ are finite
on Ln ∩ Mm , n, m ∈ N, and Ω = n,m=1 (Ln ∩ Mm ). Rearrange these intersections
into a sequence of pairwise disjoint measurable sets Ωn on which λ and µ are finite,
and note that

[
(9.13) Ω= Ωn .
n=1

Relative to this decomposition of the set, define for each component the collection
An = { A ∩ Ωn : A ∈ A },
which is a σ-algebra on Ωn ; cf. Theorem 1.39. As the restrictions of λ and µ to Ωn
are finite, the Radon-Nikodym theorem for finite measures may be applied. Hence
132 9. DECOMPOSITIONS OF MEASURES

there is a nonnegative function hn defined on Ωn and measurable with respect to


An , such that Z
λ(E) = hn dµ, E ∈ An .
E
Since (9.13) gives a decomposition of Ω into countably many pairwise disjoint sets
Ωn , the function h on Ω given by
h(ω) = hn (ω), ω ∈ Ωn ,
is well defined. It is straightforward to see that h is measurable. Now observe that
any A ∈ A has the form

[
A= An , An = A ∩ Ωn ∈ An .
n=1
It follows that from Corollary 4.33 that

X ∞ Z
X Z
λ(A) = λ(An ) = hn dµ = h dµ,
n=1 n=1 An A

which shows (9.6). 

4. Lebesgue decomposition
In order to state the Lebesgue decomposition theorem one needs the following
rather geometric notions of measures being concentrated on a set and of mutually
singular measures.
Definition 9.9. Let (Ω, A) be a measurable space and let λ be a measure. The
measure λ is said to be concentrated on a set S ∈ A if λ(S c ) = 0.
Lemma 9.10. Let (Ω, A) be a measurable space, let λ be a measure, and let S
be a set in A. The following statements are equivalent:
(1) The measure λ is concentrated on S.
(2) For all A ∈ A the inclusion A ⊂ S c implies λ(A) = 0.
(3) For all A ∈ A the identity λ(A) = λ(A ∩ S) holds.
Proof. Let S be a set in A. It is clear for all A ∈ A that
A = (A ∩ S) ∪ (A \ S) where (A ∩ S) ∩ (A \ S) = ∅.
Hence λ(A) = λ(A ∩ S) + λ(A \ S).
(1) ⇒ (2) This is trivial.
(2) ⇒ (3) Since A \ S ⊂ S c the assumption implies that λ(A \ S) = 0, which
leads to λ(A) = λ(A ∩ S).
(3) ⇒ (1) The assumption λ(A) = λ(A ∩ S) for all A ∈ A implies that λ(S c ) =
λ(S ∩ S c ) = 0. Hence λ is concentrated on S. 
Definition 9.11. Let (Ω, A) be a measurable space and let λ1 and λ2 be mea-
sures. Then λ1 and λ2 are mutually singular, denoted by λ1 ⊥ λ2 , if λ1 is concen-
trated on S1 ∈ A, λ2 is concentrated on S2 ∈ A, and S1 ∩ S2 = ∅.
Lemma 9.12. Let (Ω, A) be a measurable space and let λ1 and λ2 be measures.
Then the following statements are equivalent:
(1) λ1 and λ2 are mutually singular.
4. LEBESGUE DECOMPOSITION 133

(2) There exists a set N ∈ A such that λ1 is concentrated on N and λ2 is


concentrated on N c .
Proof. (1) ⇒ (2) Assume λ1 and λ2 are mutually singular. Then there exist
S1 , S2 ∈ A with S1 ∩ S2 = ∅, such that λ1 (S1c ) = 0 and λ2 (S2c ) = 0. Now take
N = S2c , then N ∈ A and λ2 (N ) = 0. Furthermore note that S1 ⊂ S2c = N , so that
N c ⊂ S1c . Therefore λ(N c ) ≤ λ1 (S1c ) = 0.
(2) ⇒ (1) This is trivial. 
Lemma 9.13. Let (Ω, A) be a measurable space and let λ, λ1 , λ2 , and µ be
measures on (Ω, A). Then
(1) λ1 ⊥ µ, λ2 ⊥ µ ⇒ λ1 + λ2 ⊥ µ;
(2) λ1  µ, λ2 ⊥ µ ⇒ λ1 ⊥ λ2 ;
(3) λ  µ, λ ⊥ µ ⇒ λ = 0.
Proof. (1) There is a set N ∈ A such that λ1 (N c ) = 0 and µ(N ) = 0; and
there is a set M ∈ A such that λ2 (M c ) = 0 and µ(M ) = 0. Then clearly
(λ1 + λ2 )(N c ∩ M c ) = 0, µ(N ∪ M ) = 0.
Hence λ1 + λ2 is concentrated on N ∪ M and µ is concentrated on (N ∪ M )c , which
shows that λ1 + λ2 and µ are mutually singular.
(2) Since λ2 ⊥ µ there is a set N ∈ A with λ2 (N c ) = 0 and µ(N ) = 0. Since
λ1  µ it follows that λ1 (N ) = 0. Hence λ1 and λ2 are mutually singular.
(3) By (2) it follows that λ ⊥ λ. Hence there exists N ∈ A such that λ(N ) = 0
and λ(N c ) = 0, which leads to λ = 0. 
Theorem 9.14 (Lebesgue decomposition). Let (Ω, A) be a measurable space
and let λ and µ be σ-finite measures. There is a unique pair of measures λa and
λs such that
(9.14) λ = λa + λs , λa  µ, λs ⊥ µ,
and, in addition, λa ⊥ λs .
Proof. The σ-finite measures λ and µ give rise to a new measure α on (Ω, A),
by the sum
(9.15) α(A) = λ(A) + µ(A), A ∈ A,
and α is also σ-finite. Moreover, both λ and µ are absolutely continuous with
respect to α. By Theorem 9.8 there exists a nonnegative measurable function hµ ,
such that
Z
(9.16) µ(A) = hµ dα, A ∈ A.
A
The function hµ gives rise to the following subsets of Ω:
(9.17) F = {ω ∈ Ω : hµ (ω) > 0}, G = {ω ∈ Ω : hµ (ω) = 0},
and it is clear that these sets are measurable, disjoint, and that Ω = F ∪ G. Clearly
Z
(9.18) µ(G) = hµ dα = 0.
G
Now define, in terms of the sets in (9.17), the mappings λa and λs by
λa (A) = λ(A ∩ F ), λs (A) = λ(A ∩ G).
134 9. DECOMPOSITIONS OF MEASURES

It is clear that λa and λs are measures, cf. Proposition 1.39, and that λ = λa + λs .
First observe that the definition implies that
(9.19) λs (F ) = λ(F ∩ G) = 0.
Hence from (9.18) and (9.19) it is clear that λs is concentrated on F c = G and
that µ is concentrated on Gc = F . Therefore λs and µ are mutually singular, i.e.,
λs ⊥ µ. Next let A ∈ A and assume that µ(A) = 0, which implies that
Z
hµ dα = µ(A) = 0.
A

Since hµ is positive on F ∩ A, this identity implies α(A ∩ F ) = 0, cf. Corollary 4.10.


Therefore one obtains
λa (A) = λ(A ∩ F ) ≤ α(A ∩ F ) = 0.
Thus A ∈ A and µ(A) = 0 imply that λa (A) = 0, i.e., λa  µ. The above
conclusions λa  µ and λs ⊥ µ lead to λa ⊥ λs via Lemma 9.13.
It remains to show the uniqueness of the Lebesgue decomposition (9.14). For
this purpose assume that there is another pair of measures λ0a and λ0s for which
(9.20) λ = λ0a + λ0s , λ0a  µ, λ0s ⊥ µ.
Since λs ⊥ µ and λ0s ⊥ µ, it follows from Lemma 9.12 that there exist subsets
N, N 0 ∈ A with
λs (N c ) = 0, µ(N ) = 0 and λ0s ((N 0 )c ) = 0, µ(N 0 ) = 0.
In particular, it follows from Lemma 9.10 that
(9.21) λs (A) = λs (A ∩ N ), λ0s (A) = λ0s (A ∩ N 0 ), A ∈ A.
Now define N0 = N ∪ N , then clearly N0 ∈ A and µ(N0 ) = 0. In particular, for
0

any A ∈ A it follows from A ∩ N0 ⊂ N0 that µ(A ∩ N0 ) = 0. Since λa  µ and


λ0a  µ, this leads to
λa (A ∩ N0 ) = 0, λ0a (A ∩ N0 ) = 0, A ∈ A.
Hence for all A ∈ A the decomposition (9.14) shows that
λ(A ∩ N0 ) = λa (A ∩ N0 ) + λs (A ∩ N0 ) = λs (A ∩ N0 ).
Apply the first identity in (9.21) (with A replaced by A ∩ N0 ) and use that N ⊂ N0 :
λs (A ∩ N0 ) = λs (A ∩ N0 ∩ N ) = λs (A ∩ N ) = λs (A).
Hence it follows that λ(A ∩ N0 ) = λs (A). Likewise, for all A ∈ A the decomposition
(9.20) and the second identity in (9.21) show λ(A ∩ N0 ) = λ0s (A). This leads to
λs = λ0s and, consequently, to λa = λ0a . 

5. Monotonicity, absolute continuity, and measures


Let f : R → R be a non-decreasing real-valued function. The function f
generates the so-called Lebesgue-Stieltjes measure mf ; cf. Theorem 2.25. The
connection between the function f and the measure mf is given by
(9.22) f (x) = mf ((−∞, x]), x ∈ R.
5. MONOTONICITY, ABSOLUTE CONTINUITY, AND MEASURES 135

Lemma 9.15. Let λ be a measure on the measurable space (R, B). Then the
real-valued function f : R → R defined by
(9.23) f (x) = λ((−∞, x]), x ∈ R.
is non-decreasing and λ = mf .
Definition 9.16. Let λ be a measure on R and let x ∈ R. Then λ has a
derivative L ∈ R with respect to the Lebegue measure m if for every ε > 0 there
exists δ > 0 such that
λ(I)
− L < ε,
m(I)
for every open interval I with x ∈ I and m(I) < δ. If the derivative exists at x ∈ R,
it will be denoted by (Dm λ)(x).
Lemma 9.17. Let f : R → R be a non-decreasing real-valued function, let λ
be the corresponding measure, and let x ∈ R. Then the following statements are
equivalent:
(1) f is differentiable at x and f 0 (x) = L;
(2) λ is differentiable with respect to m at x and (Dm λ)(x) = L.
Proof. (1) ⇒ (2) Assume that f is differentiable at x and f 0 (x) = L. Let
s < x < t and let I = (s, t). Then
λ(s, t] f (t) − f (s)
−L= − L,
m(s, t] t−s
which shows (2).
(2) ⇒ (1) Assume that λ is differentiable with respect to m at x and that
(Dm λ)(x) = L. Then for each ε > 0 there exists δ > 0 such that for any s < x < t
with t − s < δ one has
f (t) − f (s) λ(s, t]
−L = − L < ε.
t−s m(s, t]
Furthermore the assumption implies that λ({x}) = 0, so that f is continuous at x.
Hence f is differentiable at x and f 0 (x) = L. 
Let f : R → R be a non-decreasing real-valued function with corresponding
measure λ on the measurable space (R, B) as in (9.22). Then according to Theorem
9.14 the measure λ has the Lebesgue decomposition
λ = λa + λs , λa  m, λs ⊥ m,
and, in addition, λa ⊥ λs . Here λa has the representation
Z
λa (A) = h dm, A ∈ B.
A
where h is the Radon-Nikodym derivative of λ with respect to m. There is an
intimate connection between the Radon-Nikodym derivative h and the derivative
Dm λ.
Theorem 9.18. Let f : R → R be a non-decreasing real-valued function with
corresponding measure λ on the measurable space (R, B) as in (9.22). Then for
m-almost all x ∈ R the derivative (Dm λ)(x) exists. In fact, for m-almost all x ∈ R
one has (Dm λ)(x) = h(x).
136 9. DECOMPOSITIONS OF MEASURES

Corollary 9.19. Let f : R → R be a non-decreasing real-valued function and


let mf be the corresponding measure. Then
(1) λ  m if and only if λ(A) = A (Dm λ) dm, A ∈ B;
R

(2) λ ⊥ m if and only if (Dm λ)(x) = 0 for m-almost all x ∈ R.


Let f : [a, b] → R be a non-decreasing real-valued function. Recall that there
exists a unique decomposition
(9.24) f = ϕ + ψ,
where ϕ is an absolutely continuous nondecreasing function with ϕ(a) = f (a) and
ψ is a nondecreasing function with ψ 0 = 0. The decomposition (9.24) corresponds
to the Lebesgue decomposition of the measure mf .
CHAPTER 10

Some useful facts

1. Manipulations with sets


This section contains a number of results about abstract sets, which will be
very useful. For sets A and B one defines
A ∪ B = {x : x ∈ A or x ∈ B}, union,
A ∩ B = {x : x ∈ A and x ∈ B}, intersection,
A \ B = {x : x ∈ A and x 6∈ B}, difference,
A∆B = (A \ B) ∪ (B \ A), symmetric difference.
For the union and intersection one has A ∪ B = B ∪ A and A ∩ B = B ∩ A
(commutative laws) and the following properties:
A ∪ (B ∪ C) = (A ∪ B) ∪ C (associative law),
A ∩ (B ∩ C) = (A ∩ B) ∩ C (associative law),
A ∩ (B ∪ C) = (A ∩ B) ∪ (A ∩ C) (distributive law),
A ∪ (B ∩ C) = (A ∪ B) ∩ (A ∪ C) (distributive law),
C \ (A ∪ B) = (C \ A) ∩ (C \ B) (de Morgan’s law),
C \ (A ∩ B) = (C \ A) ∪ (C \ B) (de Morgan’s law).
For the difference one observes the following properties:
(A \ B) \ C = (A \ C) \ (B \ C),
A \ (B \ C) = (A \ B) ∪ (A ∩ C),
A ∩ (B \ C) = (A ∩ B) \ C,
while for the symmetric difference one observes A∆B = B∆A, and
A∆B = (A ∪ B) \ (A ∩ B),
A∆(B∆C) = (A∆B)∆C,
A ∩ (B∆C) = (A ∩ B)∆(A ∩ C).
If Ω is a universe and A ⊂ Ω, then Ω \ A will be denoted by Ac . In this case
∅ = Ω, Ωc = ∅, (Ac )c = A, and A \ B = A ∩ B c . If E is a collection of subsets of a
c

universe Ω, then one defines


[
{E : E ∈ E} = { ω ∈ Ω : ω ∈ E for some E ∈ E },
\
{E : E ∈ E} = { ω ∈ Ω : ω ∈ E for all E ∈ E }.
Clearly, one has for every A ∈ E:
\ [
{E : E ∈ E} ⊂ A ⊂ {E : E ∈ E},
137
138 10. SOME USEFUL FACTS

and [ [
A∩ {E : E ∈ E} = {A ∩ E : E ∈ E} (distributive law),
\ \
A∪ {E : E ∈ E} = {A ∪ E : E ∈ E} (distributive law).
Furthermore, one has
[ c \
{E : E ∈ E} = {E c : E ∈ E} (de Morgan’s law),
\ c [
{E : E ∈ E} = {E c : E ∈ E} (de Morgan’s law).
It is customary to denote unions and intersections of a collection of subsets of a
universe via an index set as follows
[ \
Eα , Eα ,
α∈I α∈I
and, in particular, when the collection is finite, one uses
[n \n
Ei , Ei .
i=1 i=1
Let En , n ∈ N, and E be subsets of a universe Ω. The notations En ↑ E or En ↓ E
will denote that En is monotonically increasing or En is monotonically decreasing
in the sense that En ⊂ En+1 or En+1 ⊂ En for n ∈ N, and that

[ ∞
\
En = E or En = E,
n=1 n=1
respectively.
Let Ω and Ω0 be sets and define the product Ω × Ω0 by
Ω × Ω0 = { (ω, ω 0 ) : ω ∈ Ω, ω 0 ∈ Ω0 }.
For A, B ∈ Ω and C, D ∈ Ω0 one has
(A ∪ B) × (C ∪ D) = (A × C) ∪ (A × D) ∪ (B × C) × (B × D),
(A ∩ B) × (C ∩ D) = (A × C) ∩ (B × D),
(A × C)c = (Ac × Ω0 ) ∪ (Ω × C c ),
(A \ B) × C = (A × C) \ (B × C),
(A × C) \ (B × D) = [(A \ B) × C] ∪ [A × (C \ D)].
Now let f be a function from Ω to Ω0 . For E ∈ Ω and G ∈ Ω0 one uses the
following notations:
f (E) = { ω 0 ∈ Ω0 : ω 0 = f (ω) for some ω ∈ E },
and
f −1 (G) = { ω ∈ Ω : f (ω) ∈ G }.
If E is a collection of subsets of Ω, then
[  [
f {E : E ∈ E} = {f (E) : E ∈ E},
\  \
f {E : E ∈ E} ⊂ {f (E) : E ∈ E},
and if G is a collection of subsets of Ω0 , then
[  [
f −1 {G : G ∈ G} = {f −1 (G) : G ∈ G},
2. THE EXTENDED REAL LINE 139

\  \
f −1 {G : G ∈ G} = {f −1 (G) : G ∈ G}.

2. The extended real line


It turns out to be useful to extend the real line by adjoining the two symbols
−∞ and ∞ with the obvious ordering. The extended real line will be denoted by
R = [−∞, ∞] and the nonnegative extended real numbers will be denoted by [0, ∞].
For x ∈ R define the following addition and multiplication by
x + ∞ = ∞ + x = ∞, x + (−∞) = (−∞) + x = −∞,
x · ∞ = ∞.x = ∞, x · (−∞) = (−∞) · x = −∞, if x > 0,
x · ∞ = ∞ · x = −∞, x · (−∞) = (−∞) · x = ∞, if x < 0.
Furthermore,
∞ + ∞ = ∞, −∞ + (−∞) = −∞,
∞ · ∞ = (−∞) · (−∞) = ∞, ∞ · (−∞) = (−∞) · ∞ = −∞,
while
0 · ∞ = ∞ · 0 = 0 · (−∞) = (−∞) · 0 = 0.
The sums ∞ + (−∞) and (−∞) + ∞ are left undefined. The products 0 · ∞, ∞ · 0,
0 · (−∞), (−∞) · 0 are often left undefined, but the present definitions are very
convenient in measure theory. Finally, define
|∞| = | − ∞| = ∞.
Every subset of R has a least upper bound, or supremum, and a greatest lower
bound, or infimum, in R. For a subset A ⊂ R they are denoted by sup A and inf A,
respectively. Note that sup A, inf A ∈ R. These facts need some explanation.
(1) When A ⊂ R is nonempty and bounded above by an element in R, then
the least upper bound belongs to R, except when A = {−∞} in which
case sup A = −∞.
(2) When A ⊂ R is nonempty and bounded below by an element in R, then
the greatest lower bound belongs to R, except when A = {∞} in which
case inf A = ∞.
(3) When A ⊂ R is nonempty and not bounded above by an element in R,
then sup A = ∞.
(4) When A ⊂ R is nonempty and not bounded below by an element in R,
then inf A = −∞.
(5) When A ⊂ R is empty, then sup A = −∞ and inf A = ∞.
The extended real line R is provided by a topology by means of the open
intervals of R and the half-intervals (x, ∞] and [−∞, x). Let (xn ) be a sequence in
R and define
ak = sup{xk , xk+1 , xk+2 , . . . }, a = inf{a1 , a2 , . . . },
bk = inf{xk , xk+1 , xk+2 , . . . }, b = sup{b1 , b2 , . . . }.
Then clearly
b1 ≤ b2 ≤ · · · ≤ bn ≤ bn+1 ≤ · · · ≤ an+1 ≤ an ≤ · · · ≤ a2 ≤ a1 .
It follows that ak ↓ a and bk ↑ b, and a ≤ b. Moreover, there is a subsequence xni of
xn such that xni → a and a is the largest number in R with this property. Likewise,
140 10. SOME USEFUL FACTS

there is a subsequence xni of xn such that xni → b and b is the smallest number
with this property. The numbers a and b are called the upper limits of (xn ):
a = lim sup xn , b = lim inf xn .
Then lim inf xn ≤ lim sup xn , and lim inf xn = − lim sup(−xn ). If lim xn exists in
R, then
lim xn = lim sup xn = lim inf xn .
Furthermore
xn ≤ yn ⇒ lim inf xn ≤ lim inf yn and lim sup xn ≤ lim sup yn .

3. Open sets in Rd
The real line R is provided in the usual way with a topology via open sets or via
open intervals. Recall that a subset of R is an interval if and only if it is non-empty
and contains each point between
T any two ofSits points. Hence if Iγ is a non-empty
class of intervals of R and γ Iγ 6= ∅, then γ Iγ is an interval. The following fact
plays a fundamental role.
Theorem 10.1. Every non-empty open set of R is the union of countably many
disjoint open intervals.
Proof. Let O be a non-empty open subset of R and let x ∈ O. Since O is
open there is an open interval with midpoint x which is completely contained in O.
Let Ix be the union of all open intervals which contain x and which are contained
in O. Then
(1) Ix is an open interval which contains x and which is contained in O;
(2) Ix contains every open interval which contains x and which is contained
in O;
(3) if y ∈ Ix , then Iy = Ix .
Let x, y ∈ O be distinct points. Then either Ix and Iy are disjoint or they are
identical; for if they are not disjoint then there exists z ∈ Ix ∩ Iy , so that Iz = Ix
and Iz = Iy , which gives Ix = Iy .
Now let I be the class of all disjoint open intervals of the form Ix for points in
x ∈ O. For each such interval Ix choose a rational point r ∈ Ix such that Ir = Ix .
Hence I is countable. 
A dyadic interval of R of length 1/2m is an interval of the form
1
(10.1) (k, k + 1], k ∈ Z, m ∈ N ∪ {0}.
2m
Dyadic intervals of a given length are pairwise disjoint and cover R: every point of
R belongs to a unique dyadic interval of given length. Dyadic intervals of different
length intersect precisely when one is contained in the other. Parallel to Theorem
10.1 one has a result for dyadic intervals.
Theorem 10.2. Every non-empty open set O of R is the union of countably
many pairwise disjoint dyadic intervals with closure in O.
This statement allows a generalization to Rd . For each m ∈ N ∪ {0} let Fm be
the family of d-dimensional rectangles
   
k1 k1 + 1 kd kd + 1
(10.2) , × ··· × , , k1 , . . . , kd ∈ Z.
2m 2m 2m 2m
4. DER GROSSE UMORDNUNGSSATZ 141

The rectangles in Fm are pairwise disjoint and cover Rd . If Qm is a rectangle (as


in (10.2)) in Fm and Qm+1 is a rectangle in Fm+1 , then either Qm and Qm+1 are
disjoint or Qm+1 ⊂ Qm .
Theorem 10.3. Every non-empty open set O of Rd is the union of countably
many pairwise disjoint dyadic rectangles with closure contained in O.
Proof. Let O ⊂ Rd be open. For m = 0 let G0 be the collection of all
rectangles in F1 with closure contained in O. For m ≥ 1 let Gm be the collection
of all rectangles in Fm with closure contained in O and not contained in any of the
rectangles in Gr for r < m. Once it has been shown that
∞ [
[ 
(10.3) O= {Q : Q ∈ Gm } ,
m=1
the theorem follows.
In order to see (10.3), let x ∈ O. Then there is an open rectangle P ⊂ O with
x ∈ P . Assume that P is of the form
P = (a1 , b1 ) × · · · × (ad , bd ).
Then there is an integer m ∈ N with
1 1
m
< xi − ai , < bi − xi , 1 ≤ i ≤ d.
2 2m
For each i with 1 ≤ i ≤ d there is an integer ki with
ki ki + 1
ai < m < xi ≤ < bi .
2 2m
Then    
k1 k1 + 1 kd kd + 1
Q= , × · · · × ,
2m 2m 2m 2m
is a rectangle in F with closure contained in O and x ∈ Q. Let m be the smallest
m

integer with this property, then Q ∈ Gm , so that x belongs to the right-hand side
of (10.3). This completes of the identity (10.3). 

4. Der große Umordnungssatz


P∞
P∞Let (an ), n ∈ N, be a sequence of real numbers. If n=1 |an | < ∞, then
n=1 an converges and, in fact, for any enumeration of (an ) the corresponding
series converges to the same limit.
Theorem 10.4. Let (ank ), n, k ∈ N, be a doubly indexed sequence of real
numbers. The following statements are equivalent:
P∞ P∞ P∞
(1) for each n ∈ N P k=1 |ank | < ∞ and Pn=1 ( P k=1 |ank |) < ∞;
∞ ∞ ∞
(2) for each k ∈ N n=1 |ank | < ∞ and k=1 ( n=1 |ank |)P< ∞;

(3) for any enumeration am , m ∈ N, of the sequence (ank ) m=1 |am | < ∞.
In this case
∞ ∞ ∞ ∞ ∞
! !
X X X X X
|ank | = |ank | = |am |,
n=1 k=1 k=1 n=1 m=1
and, moreover, the following series converge and are equal:
∞ ∞ ∞ ∞ ∞
! !
X X X X X
ank , ank , am .
n=1 k=1 k=1 n=1 m=1
142 10. SOME USEFUL FACTS

Proof. It suffices to show (1) ⇔ (3) and the associated statements, as the
rest follows by symmetry.
(1) ⇒ (3) Let M ∈ N, and choose N and K so large that a1 , . . . , aM occur
among
{ank : n ≤ N, k ≤ K}.
Then it follows that
∞ ∞
M N K
! !
X X X X X
|am | ≤ |ank | ≤ |ank | ,
m=1 n=1 k=1 n=1 k=1

which implies
∞ ∞ ∞
!
X X X
|am | ≤ |ank | .
m=1 n=1 k=1
(3) ⇒ (1) Choose K and fix n. Choose M so large that an1 , . . . , anK occur
among a1 , . . . , aM . Then it follows
K
X M
X ∞
X
|ank | ≤ |am | ≤ |am |,
k=1 m=1 m=1

which implies

X ∞
X
|ank | ≤ |am |.
k=1 m=1
Let N ≥ 1 and choose K so that for all n = 1, . . . , N
∞ K ∞ K
X X X X 1
|ank | − 1/N < |ank | or |ank | − |ank | < .
N
k=1 k=1 k=1 k=1

Next choose M so large that {ank : n ≤ N, k ≤ K} are among a1 , . . . , aM . Then



N
! N K
! N K
!
X X X X 1 X X
|ank | ≤ |ank | + ≤ |ank | + 1
n=1 n=1
N n=1
k=1 k=1 k=1
M
X ∞
X
≤ |am | + 1 ≤ |am | + 1.
m=1 m=1

This implies
∞ ∞ ∞
!
X X X
|ank | ≤ |am | + 1.
n=1 k=1 m=1

Now assume the equivalent conditions (1) and (3), so that by what has been
shown above:
∞ ∞ ∞ ∞
!
X X X X
|am | ≤ |ank | ≤ |am | + 1.
m=1 n=1 k=1 m=1
Replace in these inequalities ank and am by Cank and Cam with C > 0. Hence it
follows that
∞ ∞ ∞ ∞
!
X X X X 1
|am | ≤ |ank | ≤ |am | + .
m=1 n=1 m=1
C
k=1
4. DER GROSSE UMORDNUNGSSATZ 143

Since C > 0 is completely arbitrary, one obtains


∞ ∞ ∞
!
X X X
|am | = |ank | .
m=1 n=1 k=1
In order to show the last statements about the convergence of the series (with-
out absolute values), introduce the nonnegative numbers
bnk = |ank | − ank and bm = |am | − am .
The equivalent assumptions (1) or (3) imply that

X ∞
X
bm ≤ |am | < ∞,
m=1 m=1
so that by these same equivalences it follows that also
∞ ∞ ∞
!
X X X
bnk < ∞ and bnk < ∞,
k=1 n=1 k=1
while
∞ ∞ ∞
!
X X X
bm = bnk .
m=1 n=1 k=1
P∞
According to the above definitions one sees that k=1 ank exists and

X ∞
X ∞
X
ank = |ank | − bnk .
k=1 k=1 k=1
P∞ P∞
As a consequence, n=1 ( k=1 ank ) exists and
∞ ∞ ∞ ∞ ∞ ∞
! ! !
X X X X X X
ank = |ank | − bnk
n=1 k=1 n=1 k=1 n=1 k=1
X∞ ∞
X
= |am | − bm
m=1 m=1
X∞
= am ,
m=1
which completes the proof. 
Bibliography

[1] S. Abbott, Understanding Analysis, Springer Verlag, 2002.


[2] E. Asplund and L. Bungart, A first course in integration, Holt, Rinehart and Winston, Inc.,
New York, 1966.
[3] H. Bauer, Mass- und Integrationstheorie, de Gruyter, Berlin, 1990.
[4] J.C. Burkill, The Lebesgue integral, Cambridge University Press, 1963.
[5] M. Capiński and E. Kopp, Measure, Integral and Probability, Springer Verlag, 2004.
[6] D.L. Cohn, Measure Theory, Birkhäuser, Boston, 1980.
[7] J. Elsstrodt, Mass- und Integrationstheorie, Springer, Berlin, 2002.
[8] G.B. Folland, Real Analysis, 2nd ed., Wiley, New York, 1999.
[9] J. Gärtner, Mass- und Integrationstheorie, Lecture notes, TU Berlin, 2008.
[10] J.K. Hunter, Measure Theory, Lecture notes, University of California at Davis, 2011.
[11] G. Klambauer, Real Analysis, Dover, 2005.
[12] S. Lang, Real Analysis, Addison Wesley, 1969.
[13] I.P. Natanson, Theorie der Funktionen einer reellen Veränderlichen, Akademie Verlag, 1981.
[14] W. Rudin, Principles of Mathematical Analysis, McGraw-Hill, New York, 1964.
[15] W. Rudin, Real and Complex Analysis, McGraw-Hill, New York, 1966.
[16] G.E.F. Thomas, Integraalrekening, Lecture notes, University of Groningen, 1990.
[17] R.L. Wheeler and Z. Zygmund, Measure and Integral, Marcel Dekker, 1977.
[18] J. Yeh, Real Analysis: Theory of Measure and Integration, 3rd ed., World Scientific, 2014.

145

You might also like