Professional Documents
Culture Documents
(Download pdf) Foundations Of Quantitative Finance Book V General Measure And Integration Theory 1St Edition Robert R Reitano full chapter pdf docx
(Download pdf) Foundations Of Quantitative Finance Book V General Measure And Integration Theory 1St Edition Robert R Reitano full chapter pdf docx
https://ebookmass.com/product/foundations-of-quantitative-
finance-book-iv-distribution-functions-and-expectations-1st-
edition-robert-r-reitano/
https://ebookmass.com/product/foundations-of-finance-pearson-
series-in-finance/
https://ebookmass.com/product/foundations-of-finance-john-d-
martin/
https://ebookmass.com/product/foundations-of-antenna-radiation-
theory-eigenmode-analysis-wen-geyi/
Foundations of Airport Economics and Finance Hans-
Arthur Vogel
https://ebookmass.com/product/foundations-of-airport-economics-
and-finance-hans-arthur-vogel/
https://ebookmass.com/product/foundations-of-finance-global-
edition-9th-edition-ebook-pdf/
https://ebookmass.com/product/general-surgery-examination-and-
board-review-1st-edition-robert-b-lim/
https://ebookmass.com/product/chemistry-in-quantitative-language-
fundamentals-of-general-chemistry-calculations-2nd-edition-
oriakhi/
https://ebookmass.com/product/integration-of-mechanical-and-
manufacturing-engineering-with-iot-r-rajasekar/
Foundations of Quantitative
Finance
Chapman & Hall/CRC Financial Mathematics Series
Series Editors
M.A.H. Dempster
Centre for Financial Research
Department of Pure Mathematics and Statistics
University of Cambridge, UK
Dilip B. Madan
Robert H. Smith School of Business
University of Maryland, USA
Rama Cont
Mathematical Institute
University of Oxford, UK
Robert A. Jarrow
Ronald P. & Susan E. Lynch Professor of Investment ManagementSamuel Curtis Johnson Graduate School of
Management Cornell University
Recently Published Titles
Machine Learning for Factor Investing: Python Version
Guillaume Coqueret ad Tony Guida
Introduction to Stochastic Finance with Market Examples, Second Edition
Nicolas Privault
Commodities: Fundamental Theory of Futures, Forwards, and Derivatives Pricing, Second Edition
Edited by M.A.H. Dempster, Ke Tang
Introducing Financial Mathematics: Theory, Binomial Models, and Applications
Mladen Victor Wickerhauser
Financial Mathematics: From Discrete to Continuous Time
Kevin J. Hastings
Robert R. Reitano
Brandeis International Business School
Waltham, MA
First edition published 2024
by CRC Press
2385 Executive Center Drive, Suite 320, Boca Raton, FL 33431
Reasonable efforts have been made to publish reliable data and information, but the author and publisher cannot as-
sume responsibility for the validity of all materials or the consequences of their use. The authors and publishers have
attempted to trace the copyright holders of all material reproduced in this publication and apologize to copyright holders
if permission to publish in this form has not been obtained. If any copyright material has not been acknowledged please
write and let us know so we may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or
utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including pho-
tocopying, microfilming, and recording, or in any information storage or retrieval system, without written permission
from the publishers.
For permission to photocopy or use material electronically from this work, access www.copyright.com or contact the
Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. For works that are
not available on CCC please contact mpkbookspermissions@tandf.co.uk
Trademark notice: Product or corporate names may be trademarks or registered trademarks and are used only for iden-
tification and explanation without intent to infringe.
DOI: 10.1201/9781003264576
Preface xi
Author xiii
Introduction xv
1 Measure Spaces 1
1.1 Lebesgue and Borel Spaces on R . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.1 Starting Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.2 Lebesgue Measure Space . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.1.3 Borel Measure Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2 General Extension Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 Measure Space Constructions . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.1 Finite Products of Measure Spaces . . . . . . . . . . . . . . . . . . . 11
1.3.2 Borel Measures on Rn . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3.3 Infinite Products of Probability Spaces . . . . . . . . . . . . . . . . . 14
1.4 Continuity of Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2 Measurable Functions 17
2.1 Properties of Measurable Functions . . . . . . . . . . . . . . . . . . . . . . 19
2.2 Limits of Measurable Functions . . . . . . . . . . . . . . . . . . . . . . . . 22
2.3 Results on Function Sequences . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.4 Approximating σ(X)-Measurable Functions . . . . . . . . . . . . . . . . . . 26
2.5 Monotone Class Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.5.1 Monotone Class Theorem . . . . . . . . . . . . . . . . . . . . . . . . 32
2.5.2 Functional Monotone Class Theorem . . . . . . . . . . . . . . . . . . 34
vii
viii Contents
4 Change of Variables 84
4.1 Change of Measure: A Special Case . . . . . . . . . . . . . . . . . . . . . . 85
4.1.1 Measures Defined by Integrals . . . . . . . . . . . . . . . . . . . . . . 85
4.1.2 Integrals and Change of Measure . . . . . . . . . . . . . . . . . . . . 89
4.2 Transformations and Change of Measure . . . . . . . . . . . . . . . . . . . 92
4.2.1 Measures Induced by Transformations . . . . . . . . . . . . . . . . . 92
4.2.2 Change of Variables under Transformations . . . . . . . . . . . . . . 95
4.3 Special Cases of Change of Variables . . . . . . . . . . . . . . . . . . . . . 99
4.3.1 Lebesgue Integrals on R . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.3.2 Linear Transformations on Rn . . . . . . . . . . . . . . . . . . . . . 102
4.3.3 Differentiable Transformations on Rn . . . . . . . . . . . . . . . . . . 106
References 231
Index 235
Taylor & Francis
Taylor & Francis Group
http://taylorandfrancis.com
Preface
The idea for a reference book on the mathematical foundations of quantitative finance has
been with me throughout my professional and academic careers in this field, but the com-
mitment to finally write it didn’t materialize until completing my introductory quantitative
finance book in 2010.
My original academic studies were in pure mathematics in a field of mathematical anal-
ysis, and neither applications generally nor finance in particular were then even on my
mind. But on completion of my degree, I decided to temporarily investigate a career in ap-
plied math, becoming an actuary, and in short order became enamored with mathematical
applications in finance.
One of my first inquiries was into better understanding yield curve risk management, ulti-
mately introducing the notion of partial durations and related immunization strategies. This
experience led me to recognize the power of greater precision in the mathematical specifica-
tion and solution of even an age-old problem. From there my commitment to mathematical
finance was complete, and my temporary investigation into this field became permanent.
In my personal studies, I found that there were a great many books in finance that
focused on markets, instruments, models, and strategies, which typically provided an in-
formal acknowledgment of the background mathematics. There were also many books on
mathematical finance focusing on more advanced mathematical models and methods, and
typically written at a level of mathematical sophistication requiring a reader to have signif-
icant formal training and the time and motivation to derive omitted details.
The challenge of acquiring expertise is compounded by the fact that the field of quanti-
tative finance utilizes advanced mathematical theories and models from a number of fields.
While there are many good references on any of these topics, most are again written at
a level beyond many students, practitioners and even researchers of quantitative finance.
Such books develop materials with an eye to comprehensiveness in the given subject matter,
rather than with an eye toward efficiently curating and developing the theories needed for
applications in quantitative finance.
Thus the overriding goal I have for this collection of books is to provide a complete and
detailed development of the many foundational mathematical theories and results one finds
referenced in popular resources in finance and quantitative finance. The included topics
have been curated from a vast mathematics and finance literature for the express purpose
of supporting applications in quantitative finance.
I originally budgeted 700 pages per book, in two volumes. It soon became obvious
this was too limiting, and two volumes ultimately turned into ten. In the end, each book
was dedicated to a specific area of mathematics or probability theory, with a variety of
applications to finance that are relevant to the needs of financial mathematicians.
My target readers are students, practitioners, and researchers in finance who are quanti-
tatively literate and recognize the need for the materials and formal developments presented.
My hope is that the approach taken in these books will motivate readers to navigate these
details and master these materials.
Most importantly for a reference work, all 10 volumes are extensively self-referenced.
The reader can enter the collection at any point of interest, and then using the references
xi
xii Preface
cited, work backward to prior books to fill in needed details. This approach also works for
a course on a given volume’s subject matter, with earlier books used for reference, and for
both course-based and self-study approaches to sequential studies.
The reader will find that the developments herein are presented at a much greater level
of detail than most advanced quantitative finance books. Such developments are of necessity
typically longer, more meticulously reasoned, and therefore can be more demanding on the
reader. Thus before committing to a detailed line-by-line study of a given result, it can be
more efficient to first scan the derivation once or twice to better understand the overall logic
flow.
I hope the scope of the materials, and the additional details presented, will support your
journey to better understanding.
I am grateful for the support of my family: Lisa, Michael, David, and Jeffrey, as well as
the support of friends and colleagues at Brandeis International Business School.
Robert R. Reitano
Brandeis International Business School
Author
xiii
Taylor & Francis
Taylor & Francis Group
http://taylorandfrancis.com
Introduction
xv
xvi Introduction
the integrals of nonnegative functions, including Fatou’s lemma and Lebesgue’s monotone
convergence theorem, which are then instrumental in the development of the properties of
such integrals. As an application of this theory, a Chapter I.7 result is generalized, address-
ing countable additivity of the set function defined on measurable rectangles of product
measure spaces.
This integration theory is then applied to derive integrals of general measurable func-
tions and their properties, along with the integration to the limit results of Beppo-Levi’s
theorem and Lebesgue’s dominated convergence theorem. The bounded convergence theo-
rem is then derived, as is an integration to the limit result not seen in Book III, called the
uniform integrability convergence theorem. The Leibniz integral rule on the derivatives of
parametrically defined integrals is then studied, generalizing the result of the Riemann the-
ory of Book III. The chapter ends with a discussion of when the Lebesgue-Stieltjes integrals
of this chapter agree with the Riemann-Stieltjes integrals of Book III.
Chapter 4 develops various results on change of variables in Lebesgue-Stieltjes integrals.
The first investigation is fundamental in probability theory and addresses integrals whose
measures are defined by integrals of measurable functions with respect to other measures.
In probability theory, for example, this measure would be the Borel measure on R induced
by the density function of a given random variable.
A more general realization of this idea is studied next, related to measurable transfor-
mations and their induced measures on range spaces. Integrals on the domain and range
spaces can then be related using this transformation and related measures. Special cases
of transformations are then studied in detail, beginning with linear transformations on Rn ,
and then turning to continuously differentiable transformations.
Integrals on product spaces are studied in Chapter 5. The key question is, echoing a Rie-
mann result of Book III, when can a multiple integral be evaluated using so-called iterated
integrals, which integrate one variable at a time? Such results prove to be related to the
product space sigma-algebra used, whereby the choices contrasted are the complete sigma
algebra and the smallest sigma algebra that contains the defining algebra. In large measure,
the results look quite similar, but there are details related to the need to qualify certain
statements with “almost everywhere.” The fundamental results here are then Fubini’s the-
orem, applicable to integrable functions, and Tonelli’s theorem, applicable to nonnegative
measurable functions. Examples from earlier books are then used to illustrate the theory,
though the next two chapters find deeper applications.
In Chapter 6, the first application of the Chapter 5 theory is to Lebesgue-Stieltjes
integration by parts. For this, the notion of a “signed measure” is introduced and seen to
possess all the properties of a measure other than nonnegativity. The integration theory is
then developed for such measures when induced by functions of bounded variation, echoing
the Riemann-Stieltjes work of Book III. The second application of the Chapter V theory
is to the study of the integrability of the convolution of integrable functions, proving that
such functions are indeed integrable.
The theory of Fourier transforms is developed in Chapter 7, a theory that will have
its most important applications in Book VI in the guise of characteristic functions. After
introducing the notion of the integral of complex-valued functions, the Fourier transform of
integrable functions and finite Borel measures is defined and various properties developed. In
particular, there are key connections between the decay at infinity of a function (or measure)
and the differentiability of its Fourier transform. Conversely, there are connections between
the existence and integrability of the derivatives of a function and the rate of decay at
infinity of its Fourier transform.
Fourier inversion is then studied, whereby one recovers the measure from its Fourier
transform, or, recovers the integrable function in the case where this transform is integrable.
Finally, a continuity theory is studied, which addresses the connection between the weak
Introduction xvii
convergence of probability measures and the pointwise convergence of their Fourier trans-
forms. An application to Poisson’s limit theorem is made, while the more powerful applica-
tions are deferred to Book VI.
Chapter 8 investigates general measure relationships and various decompositions of mea-
sures vis-a-vis other measures. The chapter begins with an example of the decomposition of
a Borel measure on R into measures that have characteristic properties relative to Lebesgue
measure. These characteristic properties are then generalized, whereby given a measure µ
we identify what it means for other measures to be absolutely continuous, or, mutually sin-
gular, relative to µ. To generalize the Borel decomposition, the signed measures introduced
in Chapter 6 are now studied in some detail. Various results are developed, such as the
Hahn decomposition, which address the positive and negative sets underlying a signed mea-
sure, and the Jordan decomposition, which decomposes a signed measure into a difference
of measures.
Perhaps the most famous and useful result on relationships between measures is the
Radon-Nikodým theorem. It states that if a measure υ is absolutely continuous with re-
spect to a measure µ, then υ is given by the µ-integral of a measurable function. This key
result will have an important application in Chapter 9, and then in the stochastic pro-
cess studies of Books VII–IX. The chapter ends with the Lebesgue decomposition theorem,
which generalizes the chapter’s Borel example to σ-finite measures.
The final Chapter 9 investigates Banach spaces, adding to their introduction in Book
III by studying the so-called Lp spaces of variously integrable or bounded measurable func-
tions, and their properties. Bounded linear functionals on Lp spaces are investigated and
characterized by the Riesz representation theorem, which is proved with the aid of the
Radon-Nikodým theorem. The special space of L2 , which is a Hilbert space, is addressed.
I hope this book and the other books in the collection serve you well.
Notation 0.1 (Referencing within FQF Series) To simplify the referencing of results
from other books in this series, we use the following convention.
A reference to “Proposition I.3.33” is a reference to Proposition 3.33 of Book I, while
“Chapter III.4” is a reference to Chapter 4 of Book III, and “II.(8.5)” is a reference to
formula (8.5) of Book II, and so forth.
Taylor & Francis
Taylor & Francis Group
http://taylorandfrancis.com
1
Measure Spaces
This chapter summarizes some of the more important results from Book I on the various
constructions of a measure space (X, σ(X), µ). The importance of such constructions is
twofold:
• The requirements for a measure space (X, σ(X), µ) are quite demanding and, in fact,
so demanding that it is not immediately clear that such objects even exist.
• Given general existence, a measure space with particular properties will often be re-
quired. What if any restrictions are needed on these properties to ensure existence? For
example, can we create a measure on R so that the measure of any interval [a, b] is given
by f (b) − f (a) for a given function f ?
For existence, the space X is generally a collection of points, sometimes with other
special properties, but any set X can in theory be used. By Definitions I.2.1 and I.2.5, the
sigma algebra σ(X) is a collection of subsets of X with the following properties:
Definition 1.1 (Algebra of sets, sigma algebra) A collection A of sets from a space
X is called a algebra of sets on X:
S
1. If A ∈ A and B ∈ A, then the union A B ∈ A where:
[
A B ≡ {x|x ∈ A or x ∈ B}.
2. If A ∈ A, then A
e ∈ A, where the complement of A is defined:
e ≡ {x ∈ X|x ∈
A / A}.
The complement of A is also denoted Ac .
A collection σ(X) of sets from a space X is called a sigma algebra on X if σ(X) is
an algebra of sets, and:
S∞
3. If {Ai }∞
i=1 ⊂ σ(X), then i=1 Ai ∈ σ(X).
Thus algebras are closed under finite unions and complementation, and also finite inter-
sections by De Morgan’s laws of Exercise I.2.2. A sigma algebra is closed under countable
unions, and thus again countable intersections. In addition, algebras, and hence also sigma
algebras,
Smust contain X and the empty set ∅. For example, A ∈ A implies A e ∈ A and thus
both A A e = X ∈ A and A T A e = ∅ ∈ A.
The existence of algebras and sigma algebras can be demonstrated starting with any
collection of sets. Such a collection then obtains both an algebra and a sigma algebra by
defining A as the smallest algebra that contains this collection, and σ(X) as the smallest
such sigma algebra. This construction works because by Proposition I.2.8, the intersection
of any collection of algebras is an algebra, and similarly for sigma algebras. To ensure that
we do not have a vacuous intersection in this construction, note that the power set P(X),
which contains all subsets of X, is one such algebra and sigma algebra.
An important example of such constructions are the Borel sigma algebras.
DOI: 10.1201/9781003264576-1 1
2 Measure Spaces
Example 1.2 (Borel sigma algebra) By Definition I.2.13, the Borel sigma algebra
n
B(R ) on Rn is the smallest sigma algebra that contains the open sets of Rn . More generally,
if X is a topological space (Definition 1.5), the Borel sigma algebra B(X) is defined as the
smallest sigma algebra on X that contains all the open sets as defined by the topology T
on X.
By De Morgan’s laws, Borel sigma algebras can also be defined as the smallest sigma
algebras that contain the closed sets, but the above formulation is conventional.
Recall the definition of open and closed sets. For more background on open and closed
sets in various contexts, see Chapter 4 of Reitano (2010) or Section III.4.3.2, and also
Dugundji (1970) or Gemignani (1967).
Definition 1.4 (Open sets in R, Rn , and metric X) A set E ⊂ R is called open if for
any x ∈ E there is an open interval containing x that is also contained in E. In other words,
there exists 1 , 2 > 0 so that (x − 1 , x + 2 ) ⊂ E. There is no loss of generality by requiring
1 = 2 .
A set E ⊂ Rn is called open if for any x ∈ E there is an open ball Br (x) about x of
radius r > 0:
Br (x) ≡ {y| |x − y| < r},
so that Br (x) ⊂ E. Here |x − y| denotes the standard metric on Rn :
hXn i1/2
2
|x − y| ≡ (xi − yi ) .
i=1
More generally, if X is a metric space with metric d, a set E ⊂ X is called open if for
any x ∈ E there is an open ball Br (x) about x of radius r > 0:
so that Br (x) ⊂ E.
In all cases, a set F is called closed if Fe, the complement of F, is open.
The above notions of an open set reflect the natural metric |·| on R and Rn , where
d(x, y) ≡ |x − y| , or more generally a metric d on a space X. For more on metrics, see the
above references and Section III.4.3.1.
Open sets can also be defined without metrics, and this notion allows one to then define
a continuous function. See Proposition 2.10.
1. ∅, X ∈ T ;
2. If {Aα }α∈I S ⊂ T , where the index set I is arbitrary (finite, countably infinite, uncount-
able), then α∈I Aα ∈ T ;
Tn
3. If {Ai }ni=1 ⊂ T , then i=1 Ai ∈ T .
Measure Spaces 3
Exercise 1.6 (Natural topology on Rn ) Prove that the collection of open sets on R, as
defined in Definition 1.4, is a topology on R, and then generalize to Rn . These topologies are
often referred to as the natural topologies on R and Rn , and also the Euclidean topolo-
gies. The same result is true for the metric space (X, d), and this is called the topology
induced by the metric d. Hint: Recall Definition III.4.39 for definitional properties of
metrics.
Returning to the existence question for a measure space (X, σ(X), µ), since it is relatively
easy to construct various (X, σ(X)) pairs, the real question of existence relates to the
existence of a measure µ on σ(X). Recalling Definition I.2.23:
Definition 1.7 (Measure space) Let X be a set and σ(X) a sigma algebra of subsets of
X. A measure µ on X is a nonnegative set function defined on σ(X), taking values in
+
the nonnegative extended real numbers R ≡ R+ ∪{∞}, and which satisfies the following
properties:
1. µ(∅) = 0;
2. Countable Additivity: If {Ai }∞i=1 is aT
countable collection of pairwise disjoint sets
in the sigma algebra σ(X), meaning Aj Ak = ∅ if j 6= k, then:
[∞ X∞
µ Ai = µ (Ai ) . (1.1)
i=1 i=1
In such a case, the sets in σ(X) are said to be measurable, and sometimes µ-
measurable, and the triplet (X, σ(X), µ) is called a measure space.
Definition 1.8 (Lebesgue, Borel measure spaces) For these special measure spaces, a
third requirement is added:
m(I) = |I| = b − a,
the length of I.
By the construction in Section I.7.6, this criterion
Qn generalizes for Lebesgue measure on
Rn , that for any measurable rectangle R ≡ i=1 hai , bi i , whether such intervals are
open, closed, or semi-closed:
Yn
m(R) = |R| = (bi − ai ) ,
i=1
the volume of R.
n
• For a Borel measure µ, X = Rn and σ(X) = B(R ), and Definitions I.5.1 and
I.8.4 add the requirement:
n
3’. For any compact set A ∈ B(R ), µ(A) < ∞.
As noted above, while it is relatively easy to construct a sigma algebra on given set, it
is by no means apparent how one then defines a measure on this sigma algebra.
4 Measure Spaces
Remark 1.9 (Additional properties of measures) A few comments on the above def-
initions:
• Lebesgue translation invariance: For Lebesgue measure, it follows from item 3 that
m is translation invariant on intervals:
where I + x ≡ {x + y|y ∈ I} for any x. This property is then satisfied for all Lebesgue
measurable sets on R by Propositions I.2.28 and I.2.35, and analogously generalizes to
Rn by the construction of Chapter I.7.
• Borel measures: The notion of a Borel measure is not standardized, and it is not
uncommon to see other restrictions added in place of item 3’, such as the inner and
outer regularity properties of Proposition I.5.29. Measure spaces that also satisfy item
3’ are often called Radon measures, after Johann Radon (1887–1956).
• Finite additivity: Setting Ai = ∅ for i > n in (1.1), it follows from this definition that
all measures also satisfy finite additivity for pairwise disjoint sets:
[n Xn
µ Ai = µ (Ai ) . (1.2)
i=1 i=1
• Monotonicity: Measures are monotonic, which means that if A ⊂ B with both mea-
surable, then:
µ (A) ≤ µ (B) . (1.3)
S
This follows from finite additivity since B = A (B − A) as a disjoint union.
The existence of measures is more readily justified when X has a finite or countable
number of points.
For uncountable sets X, the construction is of necessity more subtle and sometimes ob-
tains surprising results. In the next section, we review the special constructions of Lebesgue
and Borel measure spaces of Chapters I.2 and I.5. We then summarize the general construc-
tion theory of Chapter I.6, and in the following section, recall Book I applications of this
general framework to various special constructions.
Lebesgue and Borel Spaces on R 5
• Lebesgue: The collection of sets is the open intervals G ≡ {(a, b)}, and the measure of
I = (a, b), denoted |I| , is defined as interval length:
|I| = b − a. (1.5)
where each In is an open interval, and |In | denotes its interval length.
• Borel: The collection of sets is the right semi-closed intervals A0 ≡ {(a, b]}, and the
measure of I = (a, b], denoted |I|F , is defined as the F -length of the interval:
and then by translation invariance derived that all such sets must have the same outer
measure.
This construction completely crushes any hope that m∗ is countably additive. Indeed,
it follows that depending on whether m∗ (A0 ) = 0 or m∗ (A0 ) > 0:
X∞
m∗ (Aj ) ∈ {0, ∞}.
j=1
∗
But in no case can it be that m ([0, 1]) = 1.
∗
P∞As m∗ is countably subadditive by Proposition I.2.29, it follows that m∗ (A0 ) > 0 and
∗
j=1 m (Aj ) = ∞. It then follows from this that m is not even finitely additive. So, the
problem here is even more serious than it first appears.
We are left with two possibilities:
There is no hope for item 1, since any proposal will encounter the Vitali construction.
To pursue item 2, there are two conventional approaches to identify the sets in P (R) on
which m∗ is indeed a measure, and these are discussed in Section I.2.4. The approach used
there was one that generalizes well to other constructions.
In Definition I.2.33, the approach of Constantin Carathéodory (1873–1950) was used,
an approach he developed for the general theory of outer measures.
It was then proved in a series of results culminating in Proposition I.2.39 that ML (R)
is a sigma algebra that contains the intervals and hence the Borel sigma algebra B(R), and
that m∗ is a Lebesgue measure on this sigma algebra. Lebesgue measure m is then defined
on ML (R) by:
m ≡ m∗ ,
and (R, ML (R), m) is a Lebesgue measure space.
It was also proved that ML (R) is a complete sigma algebra, and thus (R, ML (R), m)
is a complete measure space. See Definition 1.16. As noted above, ML (R) contains the
intervals, and m agrees with the specification in (1.5) on this collection.
An important corollary of the definition of the outer measure m∗ in (1.6) is that Lebesgue
measurable sets can be approximated well with open sets, and various classes of sets defined
using open sets. See Propositions I.2.42 for approximations, and I.2.43 for regularity of
Lebesgue measure.
To create a Borel measure space, it thus makes sense to investigate if any such increasing
and right continuous function F (x) can be used to induce a Borel measure µF . Indeed, the
answer is in the affirmative. Initially defining the set function F -length on right semi-closed
intervals (a, b] by (1.7), which is now compelled by (1.10), a series of results extends this
set function.
The first extension is to a set function µA on the algebra A, constructed as the collection
of all finite disjoint unions of right semi-closed intervals. This extension is defined additively
by disjointness. In Proposition I.5.13, µA proves to be a measure on this algebra.
1. µ(∅) = 0,
2. Countable Additivity: If {Aj }∞
j=1 is a countable collection of pairwise disjoint sets
S∞
in A with j=1 Aj ∈ A, then:
[ ∞ X∞
µ Aj = µ (Aj ) . (1.11)
j=1 j=1
An outer measure µ∗A is introduced as in (1.8), and as in the Lebesgue case, we utilize
the Carathéodory criterion in Definition I.5.16:
It was proved in a series of results culminating in Proposition I.5.23 that MµF (R) is a
sigma algebra that contains the algebra A and hence the Borel sigma algebra B(R), and
that µ∗A is a Borel measure on this sigma algebra. The Borel measure µF is then defined on
MµF (R) by:
µF ≡ µ∗A ,
and R, MµF (R), µF is a Borel measure space.
It was also proved that MµF (R) is a complete sigma algebra, and thus
R, MµF (R), µF is a complete measure space. See Definition 1.16. Further, A ⊂ MµF (R)
as noted above, and µF agrees with the specification in (1.7) on A.
An important corollary of the definition of the outer measure µ∗A in (1.6) is that sets
in MµF (R) can be approximated well with various classes of sets defined using right semi-
closed intervals. See Proposition I.5.26 and Remark I.5.27 for approximations and why these
appear to differ structurally from the Lebesgue approximations, and Proposition I.5.29 for
regularity of Borel measures.
1. If A01 , A02 ∈ A0 , then A01 A02 ∈ A0 , and thus this holds by induction for all finite
T
intersections.
f0 = Sn A0 .
2. If A0 ∈ A0 , then there exists disjoint {A0j }nj=1 ⊂ A0 so that A j=1 j
2’. If A ∈ A then A
e ∈ A.
The collection {(a, b]} is a semi-algebra, and by Exercise I.6.10, every semi-algebra A0
generates an algebra A defined as the collection of all finite disjoint unions of A0 -sets. This
General Extension Theory 9
implies that if we can define a set function on such a semi-algebra A0 , that the extension to
a measure on A will follow additively as in the Borel case.
Next, the notion of an outer measure is introduced. This definition captured the key
properties of the above special definitions that supported the conclusion that the collection
of Carathéodory measurable sets is a sigma algebra, and that µ∗ restricted to this
collection is a measure.
Definition 1.15 (Outer measure) Given a set X, a set function µ∗ defined on the power
sigma algebra σ(P (X)) of all subsets of X is an outer measure if:
1. µ∗ (∅) = 0.
2. Monotonicity: For sets A ⊂ B :
µ∗ (A) ≤ µ∗ (B).
This definition looks nothing like the outer measure definitions in (1.6) and (1.8). But
it can be noted that the first results developed after the Book I introductions of m∗ and
µ∗A were to establish that these outer measures had the above properties. Further, these
properties were essential in the subsequent developments.
The results seen in the Borel development are then completely generalized in a series of
Chapter I.6 results, which worked backward starting with the Carathéodory extension
theorem 1 of Proposition I.6.2 and named for Constantin Carathéodory (1873–1950). It
asserts that outer measures always obtain complete measure spaces by the Carathéodory
criterion.
Definition 1.16 (Complete measure space) A measure space (X, σ(X), µ) is com-
plete if for any A ∈ σ(X) with µ(A) = 0, then B ∈ σ(X) for all B ⊂ A. It then follows by
monotonicity of measures that µ(B) = 0 for all such B.
Equivalently, if A, C ∈ σ(X) with µ(A) = µ(C), then B ∈ σ(X) for all C ⊂ B ⊂ A, and
then µ(B) = µ(A) for all such B.
µ∗ (E) = µ∗ (A ∩ E) + µ∗ (A
e ∩ E). (1.13)
and:
2. Countable additivity: If {An }∞ 0
j=1 ⊂ A is a disjoint countable collection of sets and
S∞ 0
j=1 An ∈ A , then: [ ∞ X∞
µ0 Aj = µ0 (Aj ).
j=1 j=1
The algebra A in 1.b can be the algebra generated by a given semi-algebra A0 or defined
independently of a semi-algebra.
2. “Free” Steps:
This framework was applied in Chapters I.7–I.9 to obtain measure spaces in three im-
portant contexts. While still requiring a certain amount of effort to obtain the required step,
both the reader and the author were likely equally pleased to not have to then explicitly
derive all the results obtained in the free steps.
Definition 1.23 (Product space and set function) Given measure spaces {(Xi , σ(Xi ),
µi )}ni=1 , the product space:
Yn
X= Xi ,
i=1
is defined as:
X = {x ≡ (x1 , x2 , ..., xn )|xi ∈ Xi }. (1.15)
A measurable rectangle in X is a set A:
Yn
A= Ai = {x ∈ X|xi ∈ Ai }, (1.16)
i=1
0
where Ai ∈ σ(Xi ). We denote the collection of measurable
Qn rectangles in X by A .
The product set function µ0 is defined on A = i=1 Ai ∈ A0 by:
Yn
µ0 (A) = µi (Ai ), (1.17)
i=1
Unsurprisingly given the notation, the collection A0 proves to be a semi-algebra, and the
set function µ0 can be extended additively to a set function µA on the associated algebra
A of finite disjoint unions of A0 -sets. The derivation that µA so defined is a measure on
this algebra is subtle. This is in part due to the complexity of A-sets, and also that the
necessary results must be derived with (1.17) and properties of the measures {µi }ni=1 .
While generalized in Section 3.2.4, the Book I proof of countable additivity of µA required
that {µi }ni=1 be σ-finite measures. This obtained σ-finiteness of µA on A and of the resulting
measure µX on the complete sigma algebra, there denoted σ(X). The product measure space
of Proposition I.7.20 is then denoted (X, σ(X), µX ) . Qn
As noted in Notation I.7.21, it is common to express µX = i=1 µi . This notation also
reflects the fact that µX Q is an extension of µA from A to σ(X) and thus extends µ0 from
n
A0 to σ(X). So, for A = i=1 Ai ∈ A0 , (1.17) can be expressed:
Yn
µX (A) = µi (Ai ). (1.18)
i=1
When {(Xi , σ(Xi ), µi )}ni=1 = {(R, MFi (R), µFi )}ni=1 are Borel measure spaces, which
n Qn
are sigma finite by item 30 of Definition 1.8,Q
the final measure space (Rn , M(R ), i=1 µFi )
n n n n
contains a Borel measure space (Rn , B(R ), i=1 µFi ) since B(R ) ⊂ M(R ), and µX proves
to be a Borel measure.
However, there are Borel measures on Rn other than these product measures, and this
is the subject we discuss next.
We say that Fµ is continuous from above if the above property is true for all x.
2. Fµ satisfies
Qnthe n-increasing condition if given any bounded right semi-closed rect-
angle A = i=1 (ai , bi ] : X
sgn(x)Fµ (x) ≥ 0. (1.20)
x
Each x = (x1 , ..., xn ) in the summation is one of the 2n vertices of A, so xi = ai or
xi = bi , and sgn(x) equals −1 if the number of ai -components of x is odd, and equals
+1 otherwise.
Given the insights of this study of general Borel measures, the Chapter I.6 extension
n
theory is then applied to investigate if a Borel measure space (Rn , B(R ), µF ) can be con-
structed from a continuous from above and n-increasing function F : Rn → R. Given such
F (x), we begin by defining the class of bounded right semi-closed rectangles:
n
n
Yn o
A0B ≡ A ∈ B(R )|A = (ai , bi ], with − ∞ < ai ≤ bi < ∞ ,
i=1
and on A0B define the set function µ0 as in (1.21). It can be checked that A0B is not a
semi-algebra (Hint: consider A),
e so this will need to be addressed later in the development.
It is then proved in Proposition I.8.13 that µ0 is finitely additive and countably subad-
ditive on A0B . Since A0B is not a semi-algebra, Carathéodory’s extension theorem 2 cannot
be directly applied. Instead, the set function µ∗F is defined on A ⊂ Rn by:
nX [ o
µ∗F (A) = inf µ0 (An ) | A ⊂ An , An ∈ A0B , (1.22)
n
Definition 1.27 (Infinite product space and set function) Given probability spaces:
where
Qn Aj(i) ∈ σ(Xj(i) ). The cylinder set in (1.24) is said to be defined by J and
0
i=1 j(i) , and the collection of cylinder sets in X is denoted by A .
A
0 0
Qn product set function µ0 is defined on A as follows. If A ∈ A is defined by J
The
and i=1 Aj(i) , then:
Yn
µ0 (A) = µj(i) (Aj(i) ). (1.25)
i=1
The above restriction to probability spaces stems from the need to have µ0 (A) in (1.25)
well defined. For example, if J = (1, 2, ..., n) then:
and so µn+1 (Xn+1 ) = 1 is derived unless one of these sets has infinite or zero measure.
As the notation suggests, A0 so defined is a semi-algebra. Further, µ0 extends additively
to µA on the associated algebra A, and µA proves to be finitely additive and countably
subadditive. For countable additivity, the algebra A needed to be enlarged to A+ , and all
Xi were then restricted to R.
Definition 1.28 (Product space; general cylinder sets: A+ ) Given probability spaces
∞
{(R,
Q∞ B(R), µi )}i=1 , where B(R) denotes the Borel sigma algebra, the product space R =
N
The cylinder set in (1.26) will be said to be defined by H and J, and the collection of
general cylinder sets in RN is denoted by A+ .
The cylinder set A can also be characterized in terms of the projection mapping:
Yn Yn
πJ ≡ π j(i) : RN → Rj(i) ,
i=1 i=1
by:
A = π −1
J (H). (1.27)
+
For A ∈ A defined by H and J, the product set function µ0 is defined by:
where µJ denotes the finite dimensional product space probability measure associated with
{(R, B(R), µj(i) )}ni=1 .
or
Ai+1 ⊂ Ai , for all i.
In the former case, we are interested in the measure of the union, and in the latter, the
measure of the intersection.
The properties identified in this proposition are often referred to in terms of the “conti-
nuity” of measures and understood in the following sense. Given a collection of measurable
sets {Bj }∞
j=1 , define Ai by:
Si
1. Ai = j=1 Bj , or,
Ti
2. Ai = j=1 Bj , where it is assumed that µ(A1 ) < ∞.
16 Measure Spaces
Proposition 1.29 (Continuity of all measures) Given the measure space (X, σ (X) , µ)
and {Ai }∞
i=1 ⊂ σ (X):
Proof. To prove item 1, note that Ai ⊂ Ai+1 implies that µ(Ai ) ≤ µ(Ai+1 ) by monotonicity
of µ. Define B1 = A1 , and for i ≥ 2, let Bi = Ai − Ai−1 . Then {Bi }∞
i=1 ⊂ σ (X) are disjoint
sets, and: [∞ [∞
Ai = Bi .
i=1 i=1
By countable additivity of µ:
[∞ X∞
µ Ai = µ(Bi )
i=1 i=1
Xi
= µ(A1 ) + lim µ(Aj − Aj−1 ).
i→∞ j=2
Since Aj−1 and Aj − Aj−1 are disjoint with union Aj , finite additivity assures that:
Ti
For item 2, j=1 Aj = Ai by the nesting property, while monotonicity and the assump-
tion that µ(A1 ) < ∞ yields for all i :
\
i
µ Aj = µ(Ai ) < µ(A1 ).
j=1
Again by monotonicity, {µ(Ai )}∞i=1 is a bounded, nonincreasing sequence, and thus has a
well-defined limit as i → ∞, which proves (1.30).
2
Measurable Functions
f −1 (A) ∈ σ(X).
n
An extended real-valued transformation f : X → R defined on the measure space
(X, σ(X), µ) is said to be measurable, etc., if f −1 (A) ∈ σ(X) for every Borel set A ∈
n
B(R ).
More generally, a mapping f (x) between measure spaces (X, σ(X), µX ) and (Y, σ(Y ), µY )
is measurable or σ(X)/σ(Y )-measurable, if for all A ∈ σ(Y ):
f −1 (A) ∈ σ(X).
If D ⊂ X and f (x) is defined on D, then the criterion for measurability is the same as
above, and thus of necessity, D = f −1 (R) ∈ σ(X).
n
Remark 2.2 (On A ∈ B(R) or A ∈ B(R )) With f (x) an extended real-valued function,
n
f : X → R or f : X → R , it may seem odd that the measurability criterion only addresses
Borel sets in R and Rn . But note that if f is measurable by the above definition, then:
f −1 (∞) = X − f −1 (R/Rn ),
Notation 2.3 (σ(X)-measurable) Given the variety of labels used above to declare mea-
surability, σ(X)-measurable is the most accurate for extended real-valued functions and
transformations because the criterion f −1 (A) ∈ σ(X) for all Borel A is a sigma algebra
restriction. Measurability has nothing to do with the measure µ since there can be many
DOI: 10.1201/9781003264576-2 17
18 Measurable Functions
measures defined on a sigma algebra, and these do not affect which functions are measur-
able and which are not. Nonetheless, the use of µ-measurable is fairly common and rarely
causes confusion when there is one sigma algebra on the space.
However, there will be instances in coming studies where we will encounter measure
spaces (X, σ i (X), µ) with various sigma-algebras σ i (X). In other words, the space X is
fixed as is the measure µ, but there can be various sigma algebras on which µ satisfies
the definition of measure. A simple example but a common one is when (X, σ(X), µ) is a
measure space and σ i (X) ⊂ σ(X) is a sigma subalgebra, then (X, σ i (X), µ) is again a
measure space. But there can also be multiple sigma algebras with no such inclusions.
In these situations, the notion of a measurable function can become ambiguous, as
can the notion of a µ-measurable function. Thus when there is more than one sigma
algebra on X, it is necessary to say that f is σ(X)-measurable, identifying the defining
sigma algebra.
When f (x) is a mapping between general measure spaces (X, σ(X), µX ) and
(Y, σ(Y ), µY ), we will always say that f is σ(X)/σ(Y )-measurable as noted above.
Although a σ(X)-measurable function or transformation could be called σ(X)/B(R)-
n
measurable, or σ(X)/B(R )-measurable, this level of formality is rarely needed.
Although measurability of f −1 (A) for all A in the range space sigma algebra is the
requirement, it is not necessary to verify this condition for all such A to establish measura-
bility. This was seen in Proposition I.3.4 for Lebesgue or Borel measurability, meaning where
the respective domain space was (R, ML (R) , m) or (R, B (R) , m). Then, measurability for
all y of f −1 ((−∞, y)) is equivalent to this statement for all f −1 ([y, ∞)), all f −1 ((y, ∞)), or
all f −1 ((−∞, y]). Using the same ideas, this is equivalent to measurability of all f −1 ((a, b)).
In any of these cases, Proposition I.3.26 obtains that this is equivalent to measurability of
f −1 (A) for all A ∈ B (R) .
Exercise 2.5 Generalize the prior paragraph to Lebesgue or Borel measurability Qn of transfor-
mations defined on (Rn , ML (Rn ) , m) or (Rn , B (Rn ) , m). Show that if f −1 ( i=1 (−∞, yi ))
is measurable forQall y = (y1 , ...,
Qnyn ), then this
Qnis equivalent toQmeasurability of all f −1 (A)
n n
for A defined as i=1 [yi , ∞), i=1 (yi , ∞), i=1 (−∞, yi ] or i=1 (ai , bi ).
What is clear from the Book I development and the results of Exercise 2.5 is that these
collections of sets, and there are many others, have the property that they generate the Borel
sigma algebras B (R) or B (Rn ) , the sigma algebras of the range spaces. This generalizes as
might be expected.
In the following result, we specify this special collection as A0 , which is our standard
notation for a semi-algebra. This result is true for any collection of sets that generate σ(Y ),
not just semi-algebras. But we use this notation because it will often be the case that there
is an apparent semi-algebra A0 that generates an algebra A, which in turn generates the
range space sigma algebra σ(Y ). For example, this applies when (Y, σ(Y ), µY ) is a measure
space created by the extension theory of the prior chapter.
The following result is also true for f : (X1 , d1 ) → (X2 , d2 ) where, recalling Exercise 1.6,
open sets in X1 and X2 are those induced by these metrics. Details are left as an exercise
in changing notation.
Conversely, assume f −1 (G) is open for all open G ⊂ R. Let x0 ∈ Rn be given and
y0 = f (x0 ) ∈ R. Choose any open set G ⊂ R that contains y0 , for example, we could
choose G = R. By definition of open, there exists > 0 so that B (y0 ) ⊂ G. By assump-
tion f −1 (B (y0 )) is open in Rn and contains x0 . Again by definition of open, there exists
Bδ (x0 ) ⊂ f −1 (B (y0 )) and thus f (Bδ (x0 )) ⊂ B (y0 ). This now translates to the − δ
definition for continuity and the proof is complete.
This result now provides an immediate extension of the notion of continuity to functions
on topological spaces.
It now follows from this characterization of continuity that continuous functions are
σ(X)-measurable.
Proposition 2.12 (Continuous ⇒ µ-measurable) Given a measure space (X, σ(X), µ),
assume that X is also a topological space and that σ(X) contains the open sets of X, and
hence contains the Borel sigma algebra B(X). If f : X → R is continuous, then f is σ(X)-
measurable.
Proof. Consider the open set G ≡ (a, b). Since continuous, f −1 ((a, b)) is open in X, and
thus by definition:
f −1 ((a, b)) ∈ B(X) ⊂ σ(X).
The proof is complete by Proposition 2.7.
The next result summarizes that measurability is preserved under simple arithmetic
operations. We restrict to real-valued functions with range in R. The reader is referred to
Remark I.3.34 for a discussion on generalizing to extended real-valued functions with range
in R. Furthermore, items 1 and 2 remain true for measurable real-valued transformations
and this is left as an exercise.
Proposition 2.13 Let f (x) and g(x) be real-valued σ(X)-measurable functions defined on
a measure space (X, σ(X), µ), and let a, b ∈ R. Then, the following are σ(X)-measurable:
1. af (x) + b,
2. f (x) ± g(x),
3. f (x)g(x),
4. f (x)/g(x) on {x|g(x) 6= 0}.
Proof. To simplify notation, the set {x|f (x) < r} is denoted by {f (x) < r}, and so forth.
Also, by Proposition 2.7, it is sufficient to prove that h−1 (A) ∈ σ(X) for any collection of
sets that generates B (R) , where h(x) denotes any function under consideration.
1. If a = 0, the function g(x) = b is σ(X)-measurable since g −1 (A) ∈ {∅, X} for all A. For
a > 0,
{af (x) + b < y} = {f (x) < (y − b)/a},
which is measurable since f (x) is σ(X)-measurable. A similar result applies to a < 0.
2. Consider the sum since then by part 1, −g(x) is measurable and this implies the result
for f (x) − g(x). For rational r, if f (x) < r and g(x) < y − r, then f (x) + g(x) < y.
Taking a union over all rational r:
[ h \ i
{f (x) < r} {g(x) < y − r} ⊂ {f (x) + g(x) < y}.
r
On the other hand, if f (x) + g(x) < y then f (x) < y − g(x), and by density of the
rationals, there exists rational r so that f (x) < r < y − g(x). This implies f (x) < r and
g(x) < y − r.
Hence, [ h \ i
{f (x) + g(x) < y} = {f (x) < r} {g(x) < y − r} ,
r
3. First note that both f 2 (x) and g 2 (x) are measurable. For f 2 (x), for example:
√ T √
{f (x) < y} {f (x) > − y}, y ≥ 0,
{f 2 (x) < y} =
∅, y < 0.
• y > 0: [
{1/g(x) < y} = {g(x) > 1/y} {g(x) < 0}.
• y = 0:
{1/g(x) < 0} = {g(x) < 0}.
• y < 0: \
{1/g(x) < y} = {g(x) > 1/y} {g(x) < 0}.
Thus 1/g(x) is measurable as is f (x)/g(x) by item 3.
The next result is a good example of when a measurability conclusion requires complete-
ness of the measure space (X, σ(X), µ). Recall that when f (x) = g(x), except on a set of
µ-measure 0, this is often written as f (x) = g(x) µ-a.e., and read, “µ almost everywhere.”
The first set is a subset of a set of µ-measure zero and is hence σ(X)-measurable by com-
pleteness, wheras the second set is the intersection of measurable E,
e the complement of E,
and σ(X)-measurable {x|f (x) < y}.
{fn (x)}∞
−∞, n=1 unbounded below,
inf n fn (x) = (2.4)
max{y|y ≤ fn (x) all n}, {fn (x)}∞
n=1 bounded below.
{fn (x)}∞
∞, n=1 unbounded above,
supn fn (x) = (2.5)
min{y|y ≥ fn (x) all n}, {fn (x)}∞
n=1 bounded above.
Definition 2.16 (Limits inferior/superior) Given a sequence of functions {fn (x)}∞ n=1 ,
the limit inferior and
T∞limit superior of the sequence are defined pointwise as follows.
For each x ∈ D ≡ n=1 Dmn{fn }:
When clear from the context, the subscript n → ∞ is often dropped from the lim inf and
lim sup notation.
Notation 2.17 The limit superior of a function sequence is alternatively denoted limfn (x),
and the limit inferior denoted limfn (x), but we will use the above notation throughout these
books.
Exercise 2.18 (The lim in lim inf and lim sup) From Definition 2.16, it may not be ap-
parent where the notion of limit appears. Prove that:
In other words, the limit inferior is the limit of infima, wheras the limit superior is the limit
of suprema. Hint: Consider how inf k≥n fk (x) and supk≥n fk (x) vary with n.
As anticipated, these limiting functions are σ(X)-measurable when the functions in the
sequence are σ(X)-measurable. For item 7, we recall Corollary I.3.46 that lim fn (x) exists
at x if and only if:
−∞ < lim inf fn (x) = lim sup fn (x) < ∞. (2.8)
Proposition 2.19 (Measurability of functions derived from {fn (x)}∞ n=1 ) Given a se-
quence of σ(X)-measurable functions {fn (x)}∞ n=1 defined on measurable domains
{Dn }∞
n=1 of the measure space
T∞ (X, σ(X), µ), the following functions are also σ(X)-
measurable as defined on D ≡ n=1 Dn :
24 Measurable Functions
Proof. By Proposition 2.7, it is sufficient to prove that h−1 (A) ∈ σ(X) for any collection
of sets that generate B (R) where h(x) denotes any function under consideration.
Item 1 follows from item 3, and 2 from 4, by defining fn (x) = fN (x) for n ≥ N.
For item 3, if h(x) is defined by h(x) = inf fn (x), then by (2.4):
Thus: \∞
{x|h(x) > y} = {x|fn (x) > y},
n=1
and is measurable as the intersection of measurable sets.
Similarly, with g(x) = sup fn (x):
\∞
{x|g(x) < y} = {x|fn (x) < y},
n=1
Corollary 2.20 (Measurability on complete (X, σ(X), µ)) Given a sequence of σ(X)-
measurable functions {fn (x)}∞ ∞
n=1 defined on σ(X)-measurable domains {Dn }n=1 of the com-
plete measure space (X, σ(X), µ), if h(x) denotes any of the functions identified in Propo-
sition 2.19, and g(x) = h(x) µ-a.e., then g(x) is σ(X)-measurable.
Proof. This is Proposition 2.14.
Proposition 2.21 Given (X, σ(X), µ), let {fn (x)}∞ n=1 be a sequence of real-valued σ(X)-
measurable functions defined on a measurable set D with µ(D) < ∞, and let f (x) be a
real-valued function so that fn (x) → f (x) pointwise for x ∈ D. Then given > 0 and δ > 0,
there is a measurable set A ⊂ D and an integer N , so that µ(A) < δ, and for all x ∈ D − A
and all n ≥ N :
|fn (x) − f (x)| < .
Proof. Given > 0, define:
Corollary 2.22 If (X, σ(X), µ) is complete, the conclusion of the above proposition re-
mains valid if fn (x) → f (x) for each x ∈ D outside a set of µ-measure 0.
Proof. Everything in the above proof remains the same, except that we can now only con-
clude that for every xT∈ D outside an exceptional set of measure 0, that there exists DN with
x∈ / DN , and hence N DN equals this set of measure 0. But then, limN →∞ µ[ DN ] → 0
again by Proposition 1.29, and the proof follows as above, with a final application of Corol-
lary 2.20.
This proposition does not imply that fn (x) converges uniformly to f (x) on D − A
because the set A depends on the given and δ. This result is close to but not equivalent to
Littlewood’s third principle of Chapter I.4, named for J. E. Littlewood (1885–1977).
To improve this result to Littlewood’s conclusion of “nearly uniform convergence,” it
must be shown that A can be chosen so that fn (x) → f (x) uniformly on D − A. That is,
we need to find a fixed set A with µ(A) < δ, so that for any > 0, there is an N, such that
|fn (x) − f (x)| < for all x ∈ D − A and all n ≥ N. See the introduction to Chapter I.4 for
more on Littlewood’s principles.
This next result formalizes Littlewood’s third principle and is known as Egorov’s the-
orem, named for Dmitri Fyodorovich Egorov (1869–1931), and sometimes phonetically
translated to Egoroff. It is also known as the Severini–Egorov theorem in recognition
of the somewhat earlier and independent proof by Carlo Severini (1872–1951).
Proposition 2.23 (Severini-Egorov theorem ) Given (X, σ(X), µ), let {fn (x)}∞ n=1 be
a sequence of σ(X)-measurable functions defined on a measurable set D with µ(D) < ∞,
and let f (x) be a σ(X)-measurable function so that fn (x) → f (x) pointwise for x ∈ D.
Then given δ > 0, there is a measurable set A ⊂ D with µ(A) < δ, so that fn (x) → f (x)
uniformly on D − A.
That is, for > 0, there is an N , so that |fn (x) − f (x)| < for all x ∈ D − A and
n ≥ N.
26 Measurable Functions
Proof. Given δ > 0, for each m define m = 1/m and δ m = δ/2m and apply Proposition
2.21. The result is a set Am with µ(Am ) < δ m , and
S∞an integer Nm , so that |fn (x) − f (x)| <
m for nP≥ Nm and all x ∈ D − Am . With A ≡ m=1 Am , countable subadditivity obtains
∞
µ(A) ≤ m=1 µ(Am ) = δ. We now claim that fn (x) → f (x) uniformly on D − A.
Given there is an m so that m < , and hence an Nm so that |fn (x) − f (x)| < m <
for n ≥ Nm and all x ∈ D − Am . But then this statement is also true for x ∈ D − A since
Am ⊂ A.
Corollary 2.24 (Severini–Egorov theorem) If (X, σ(X), µ) is complete, the above re-
sult remains valid if fn (x) → f (x) µ-a.e. for x ∈ D.
Proof. Left as an exercise.
Remark 2.25 (On µ(D) < ∞) To perhaps state the obvious, the above results apply with-
out the explicit need for the restriction of µ(D) < ∞ in finite measure spaces, and in par-
ticular, in probability spaces. In such a space, we can conclude that pointwise convergence
on any measurable set assures nearly uniform convergence.
'I know! I know!' she riposted, with the volubility of intense excitement.
'I know your proud device: "Roy ne suys, ne Prince, ne Duc, ne Comte. Je
suys Sire de Froide Monte." Ah, Messire Gilles! you were fated to belie that
device! Prince de Froidmont—'tis no mean title.'
'You are that indeed, Messire, and more,' she rejoined solemnly. 'Ah! if
my brother were only like you, what glorious destiny would have been his!'
'You have started to carve them out for yourself now, Messire Gilles, on
the tablets of my memory.'
'The Fates?' she cried gaily. 'Why, you and I have conquered the Fates,
Messire. Will you deny that they are our handmaidens now?'
CHAPTER VI
I
And three days later, an' it please you, Messire Gilles presented himself,
his safe-conduct and his faithful Jehan at the Porte de Cantimpré.
'My equerry hath fallen sick on the way,' explained Gilles airily. 'He lies
at a village inn close by and will come as soon as may be.'
It was at once arranged that whenever the equerry did present himself at
the gate, Monseigneur le Prince de Froidmont was immediately to be
apprised of his arrival so that he might at once stand guarantee for the man's
identity. Needless to say that no such equerry existed, nor does the Captain
of the Guard appear to have worried his head over so small a matter. But,
anyway, Gilles now was inside Cambray, the scene of his coming
adventure, and I can assure you that on this first occasion—it was late
evening then and a cold, drizzling rain was blurring every outline of the
picturesque city—Gilles did not stride about the streets with that careless
jauntiness which characterized his usual demeanour.
'My mother,' stammered the youth, after he had recovered from his
primary emotion, 'is bedridden now, alas! but I will do my best to serve you,
Messire, and your henchman, to the best of my ability. I will tend you and
wait on you, and whatever this humble abode hath to offer is entirely at
your disposal. My liege lady commands,' he added, drawing up his spare
frame with the air of a devotee in the presence of his hero. 'I will obey her
in all things!'
We will not say that Gilles was exactly gratified to hear that the hostess
of 'Les Trois Rois' was bedridden and would be unable to attend on him, but
it is certain that he was not grieved. With this young enthusiast alone to
attend on him and to share the secret of his adventure, he was as secure
from untimely discovery as it was possible under the circumstances to be.
II
'Yes,' he said. 'He, at any rate, seems determined to carry the adventure
through.'
'I hope to God that we have done right,' rejoined d'Inchy. 'The whole
thing, now that it is upon us, appears to me more foolish than ever it did
before.'
'Then you must blame yourself too, my friend,' retorted d'Inchy dryly.
'You, too, consented, and so did I....'
'I know that well enough! Like yourself, de Montigny and I acted for the
best, though I for one could even now with zest strike that Valois Prince in
the face for this insult upon our ward.'
But d'Inchy apparently was all for a conciliating attitude and a cheerful
view of the situation.
'And all our friends think the same, as you well know, my good de
Lalain,' continued d'Inchy firmly. 'An alliance with a prince of the House of
France is safer than a submission to the leadership of Orange. We want the
help of France; we want her well-trained armies, her capable generals, the
weight of her wealth and influence to drive the Spaniards out of our
provinces. Elizabeth of England promises much but holds little. She is on
the side of Orange. I am on the side of France.'
'So am I, my good d'Inchy,' rejoined de Lalain; 'else I had never
consented to the Queen of Navarre's madcap scheme.'
'You wrong Jacqueline by such doubts, my friend. She is not a child nor
yet an irresponsible girl. She knows that her person and her fortune are
powerful assets in the future of her country. She is a patriot, and will never
allow sentiment to overrule her duty.'
And reason gained the victory. D'Inchy already had gone a few steps
forward in order to greet his exalted visitor. De Lalain composed himself
too, even paid an involuntary tribute of admiration to that tall and martial-
looking figure which enshrined, so rumour had it, a soul that was both weak
and false.
III
The two Flemish lords had also waited until the usher had disappeared;
then only did they make obeisance, with all the ceremonious empressment
which the presence of a Royal personage demanded.
'Monseigneur understood, I hope, how it was that we did not present our
respects to you in person. Such a ceremony would have set the tongues of
our town gossips wagging more furiously than before.'
Already, it seemed that the presence of the stranger inside Cambray had
created some comment. In these days, when the Spanish armies swarmed all
over the province, when plots and counter-plots were being constantly
hatched in favour of one political side or another, strangers were none too
welcome inside the city. There was the constant fear of spies or of traitors,
of emissaries from Spain or France or England, of treason brewed or
brewing, which might end in greater miseries yet for any unfortunate
province which was striving for its own independence and the overthrow of
Spanish tyranny. Gilles, listening with half an ear to Monseigneur d'Inchy's
elaborate compliments, was inwardly marvelling whether spies had not
already come upon his track and would upset the Queen of Navarre's plans
even before they had come to maturity. He had a curious and exceedingly
uncomfortable sensation of unreality, as if these two stern-looking Flemings
were not actual personages but puppets moved by an unseen hand for the
peopling of his dreams. He answered the elaborate flummeries of the
governor with a vague: 'I thank you, Messire.' Then he added a little more
coherently: 'I understood everything, believe me, and must again thank you
for acceding to my wishes and to those of my sister, the Queen of Navarre.'
Gilles roused himself. It was no use and ill policy to boot to allow that
feeling of unreality to dominate his mood so utterly. If he let himself drift
upon these waves of somnolence he might, with one unguarded word,
betray the grave interests which had been committed to his care.
Half an hour drew its weary length along while the discussion
proceeded, and it was at the very end of that time that M. le Baron d'Inchy
said quite casually:
'Of course, you, Monseigneur, will understand that since you choose to
do your wooing under a mask, our ward, Madame Jacqueline de Broyart,
Dame de Morchipont, Duchesse et Princesse de Ramèse, will not appear in
public either, save also with a mask covering her face.'
''Tis hard for a man to woo a maid whose face he is not allowed to see,'
he said, by way of protest.
'Oh, Monseigneur is pleased to jest!' was d'Inchy's calm rejoinder. 'It was
agreed that you should come to Cambray and see the noble lady who holds
in her dainty hand the sovereignty of the Netherlands for her future lord;
but, as Messire de Montigny had the honour to tell you, Madame Jacqueline
de Broyart is not going to be trotted out for any man's inspection—be he
King or Emperor, or Prince—like a filly that is put up for sale.'
'A just arrogation, Messire,' riposted Gilles. 'But why the mask?' he
added blandly.
'If Monseigneur will woo Madame definitely and openly,' replied d'Inchy
firmly, 'she will not wear a mask either. But then there can be no question of
withdrawal if she consents.'
Now, to woo Madame Jacqueline definitely and openly was just the one
thing Gilles could not do. So there was the difficulty and there the cunning
and subtlety of these Flemish lords, who had very cleverly succeeded in
getting Monsieur into a corner and in safeguarding at the same time the
pride and dignity of the greatest heiress in Flanders. Gilles would have
given all the worlds which he did not possess for the power to consult with
Madame la Reyne de Navarre over this new move on the part of the
Flemings. But, alas! she was far away now, flying across France after her
faithless brother, hoping soon to catch him by the tails of his satin doublet
and to drag him back to the feet of the rich heiress whom that unfortunate
Gilles was deputed to woo and win for him. And Gilles was left to decide
for himself, which he did with a 'Very well, Messire, it shall be as you
wish!' and as gracious a nod and bow to these two obstinate men as he
could bring himself to perform; for, of a truth, he would gladly have given
each a broken head.
Thus the actual discussion of the affair was ended. After that, there were
only a few minor details to talk over.
'You two gentlemen,' Gilles said after a slight pause, during which he
had been wondering whether it were a princely thing to do to rise and take
his leave. 'You two gentlemen are alone in the secret of this enterprise?'
'For the moment, yes,' replied d'Inchy guardedly. 'But others will have to
know ... some might even guess. I shall have to explain the matter to my
private secretary, and one or two members of my Privy Council have certain
rights which we could not disregard.'
'To every one else, then, Messire,' he continued more firmly, 'I shall pass
as the Prince de Froidmont.'
'Certainly, Monseigneur.'
'None.'
'Doth she know that it is your desire she should become the wife of the
Duc d'Anjou ... that she should become my wife, I mean?'
'A clear field, Monseigneur,' broke in de Lalain firmly, 'for two weeks.'
'Two weeks?' retorted Gilles with a quick frown. 'Why only two weeks?'
'Because,' said the other with solemn earnestness, 'because the Duke of
Parma's armies are already swarming over our province. If they should
invest Cambray we could not hold out alone. Monseigneur must be ready
by then to support us with influence, with men and with money. If you
turned your back on us and on the proposed alliance with a Flemish heiress,
we should have to look once more to Orange as our future Lord.'
'I understand,' rejoined Gilles dryly. ''Tis an "either—or" that you place
before me.'
Then, as d'Inchy remained respectfully silent, M. de Lalain broke in
abruptly:
'Think you, Monseigneur, that the people of the Netherlands, after all
that they have suffered in intolerance and religious persecution, would
accept a Catholic sovereign unless his wife, at least, were of their nation
and of their faith?'
He threw back his head and laughed in hearty echo. The whole humour
of the situation suddenly struck him with the full force of its irresistible
appeal. Life had been so dull, so drab, so uneventful of late! Here was
romance and excitement and gaiety; a beautiful maid—Gilles had become
suddenly convinced that she was beautiful—some blows; some knocks; a
master to serve; a beautiful, sorrowing Queen to console; spurs to be won
and a fortune to be made!
And here the matter ended for the nonce. Gilles took leave of his stiff-
necked hosts and returned to 'Les Trois Rois,' having declared most
solemnly that he must have time to prepare himself for so strange a wooing.
A masked wench; think on it! It changed the whole aspect of the situation!
A respite of four days was, however, all that was respectfully but firmly
granted to him for this preparation, and Messire Gilles spent the next few
hours in trying to devise some means whereby he could outwit the Flemish
lords and catch sight of Madame Jacqueline ere he formally set out to woo
her. Of a truth, the dull-witted and stodgy Flemings whom Monsieur
affected to despise, had not much to learn in the matter of finesse and
diplomacy from the wily Valois! This counter-stroke on their part was a real
slap-in-the-face to the arrogant prince who was condescending to an
alliance, of which every other reigning house in Europe would have been
proud.
CHAPTER VII
Old Nicolle, restless and cross, was fidgeting about the room, fingering
with fussy inconsequence the beautiful clothes which her mistress had taken
off half an hour ago preparatory to going to bed—clothes of great value and
of vast beauty, which had cost more money to acquire than good Nicolle
had ever handled in all her life. There was the beautiful gown which
Madame had worn this evening at supper, fashioned of black satin and all
slashed with white and embroidered with pearls. There was the underdress
of rich crimson silk, worked with gold and silver braid; there were the
stockings of crimson silk, the high-pattened shoes of velvet, the delicately
wrought fan, the gloves of fine chamois skin, the wide collarette edged with
priceless lace. There was also the hideous monstrosity called the farthingale
—huge hoops constructed of whalebone and of iron which, with the no less
abominable corset of wood and steel, was intended to beautify and to refine
the outline of the female figure and only succeeded in making it look
ludicrous and ungainly. There were, in fact, the numberless and costly
accessories which go to the completion of a wealthy lady's toilet.
Madame had divested herself of them all and had allowed Nicolle to
wrap a woollen petticoat round her slender hips and to throw a shawl over
her shoulders. Then, with her fair hair hanging in heavy masses down her
back, she had curled herself up in the high-backed chair beside the open
window—the open window, an it please you! and the evening, though mild,
still one of early March! Old Nicolle had mumbled and grumbled. It was
ten o' the clock and long past bedtime. For awhile she had idled away the
hour by fingering the exquisite satin of the gown which lay in all its rich
glory upon the carved dowry chest. Nicolle loved all these things. She loved
to see her young mistress decked out in all the finery which could possibly
be heaped up on a girlish and slender body. She never thought the silks and
satins heavy when Jacqueline wore them; she never thought the farthingale
unsightly when Jacqueline's dainty bust and shoulders emerged above it like
the handle of a huge bell.
But gradually her patience wore out. She was sleepy, was poor old
Nicolle! And Madame still sat squatting in the tall chair by the open
window, doing nothing apparently save to gaze over the courtyard wall to
the distance beyond, where the graceful steeple of St. Géry stood outlined
like delicate lace-work against the evening sky.
''Tis time Madame got to bed,' reiterated the old woman for the twentieth
time. 'The cathedral tower hath chimed the quarter now. Whoever heard of
young people not being abed at this hour! And Madame sitting there,' she
added, muttering to herself, 'not clothed enough to look decent!'
'Not decent?' she exclaimed with a laugh. 'Why, my dear Colle, nobody
sees me but you!'
'People passing across the courtyard might catch sight of Madame,' said
Nicolle crossly.
'E'en so,' rejoined Jacqueline lightly, 'my attire, meseems, is not lacking
in modesty. I am muffled up to my nose in a shawl and—— Oh!' she added
with a quick sigh of impatience, 'I am so comfortable in this soft woollen
petticoat. I feel like a human being in it and not like a cathedral bell. How I
wish my guardian would not insist on my wearing all these modish clothes
from Paris! I was so much more comfortable when I could don what I most
fancied.'
'Oh! a murrain upon my rank and upon my wealth!' cried the young girl
hotly. 'My dear mother rendered me a great disservice when she bare me to
this world. She should have deputed some simple, comfortable soul for the
work, who could have let me roam freely about the town when I liked, run
about the streets barefooted, with a short woollen kirtle tied round my waist
and my hair flying loose about my shoulders. I could have been so happy as
a humble burgher's daughter or a peasant wench. I do so loathe all the
stiffness and the ceremony and the starched ruffles and high-heeled shoes.
What I want is to be free—free!—Oh!——'
And Jacqueline de Broyart stretched out her arms and sighed again, half-
longingly, half-impatiently.
'Don't scold, old Colle!' pleaded the girl softly. And now her arms were
stretched towards the old waiting-woman.
Nicolle resisted the blandishment. She was really cross just now. She
turned her back resolutely upon the lovely pleader, avoiding to look into
those luminous blue eyes, which had so oft been compared by amorous
swains to the wild hyacinths that grow in the woods above Marcoing.
'Come and kiss me, Colle,' whispered the young charmer, 'I feel so
lonely somehow to-night. I feel as if—as if——'
And the young voice broke in a quaint little gasp which was almost like
a sob.
Whereupon Jacqueline broke into a sudden, gay and rippling laugh, even
though the tears still glistened on her lashes.
'I shouldn't at all enjoy,' she said lightly, 'seeing my dear old Colle cut
into a thousand pieces.'
Jacqueline made no reply. For a few seconds she remained quite silent,
her eyes fixed into nothingness above old Colle's head. One would almost
have thought that she was listening to something which the old woman
could not hear, for the expression on her face was curiously tense, with eyes
glowing and lips parted, while the poise of her girlish figure was almost
rigidly still. The flame of the wax candles in the tall sconces flickered
gently in the draught, for the casement-window was wide open and a soft
breeze blew in from the west.
But old Colle evidently thought that she knew better than that. 'Heu!
heu!' she muttered with a shrug of the shoulders, accompanied by a
knowing wink. 'What chance wouldst thou have then of seeing M. de
Landas?'
'I hardly can speak with M. de Landas during those interminable
banquets,' rejoined Jacqueline with a sigh. 'My guardian or else M. de
Lalain always seem in the way now whenever he tries to come nigh me.'
'Even so,' said Jacqueline with sudden irrelevance, 'there are moments
when one likes to be alone. There is so much to think about—to dream of
——'
'I know, I know,' murmured the old woman crossly. 'Thy desire is to sit
here half the evening now by the open window, and catch a deathly ague
while listening to that impudent minstrel who dares to serenade so great a
lady.'
She went on muttering and grumbling and fidgeting about the room,
unmindful of the fact that at her words Jacqueline had suddenly jumped to
her feet; eyes blazing, small fists clenched, cheeks crimson, she suddenly
faced the garrulous old woman.
'Silent? Silent?' grumbled the woman. 'I have been silent quite long
enough, and if Monseigneur were to hear of these doings 'tis old Nicolle
who would get the blame. As for M. de Landas, I do verily believe that he
would run his sword right through the body of the rogue for his impudence!
I know.... I know,' she added, with a tone of spite in her gruff voice. 'But let
me tell thee that if that rascally singer dares to raise his voice again to-night
——'
She paused, a little frightened at the fierce wrath which literally blazed
out of her mistress's eyes.
'Pierre will make it unpleasant for him, that's all!' replied the old woman
curtly.
'Pierre?'
In a moment Jacqueline was by Nicolle's side. She seized the old woman
by the wrist so that poor Colle cried out with pain, and it was as the very
living image of a goddess of wrath that the young girl now confronted her
terrified serving-maid.