Fundamentals of Functional Analysis

1st Edition Ammar Khanfer

Visit to download the full and correct content document:
https://ebookmeta.com/product/fundamentals-of-functional-analysis-1st-edition-amma
r-khanfer/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Measure Theory and Integration 1st Edition Ammar

Khanfer

https://ebookmeta.com/product/measure-theory-and-integration-1st-
edition-ammar-khanfer/

Functional Analysis Fundamentals and Applications 2nd

Edition Michel Willem

https://ebookmeta.com/product/functional-analysis-fundamentals-
and-applications-2nd-edition-michel-willem/

Functional Analysis 1st Edition Jan Van Neerven

https://ebookmeta.com/product/functional-analysis-1st-edition-
jan-van-neerven/

Functional Analysis 2nd Edition S. Kesavan

https://ebookmeta.com/product/functional-analysis-2nd-edition-s-
kesavan/
Basic Analysis V: Functional Analysis and Topology 1st
Edition James K. Peterson

https://ebookmeta.com/product/basic-analysis-v-functional-
analysis-and-topology-1st-edition-james-k-peterson/

Functional Interfaces in Java Fundamentals and Examples

1st Edition Ralph Lecessi

https://ebookmeta.com/product/functional-interfaces-in-java-
fundamentals-and-examples-1st-edition-ralph-lecessi/

Fundamentals of Analysis with Applications Atul Kumar

Razdan

https://ebookmeta.com/product/fundamentals-of-analysis-with-
applications-atul-kumar-razdan/

Fundamentals of Spatial Analysis and Modelling 1st

Edition Jay Gao

https://ebookmeta.com/product/fundamentals-of-spatial-analysis-
and-modelling-1st-edition-jay-gao/

An Introduction to Functional Analysis 1st Edition

James C. Robinson

https://ebookmeta.com/product/an-introduction-to-functional-
analysis-1st-edition-james-c-robinson/
Ammar Khanfer

Fundamentals
of Functional
Analysis
Fundamentals of Functional Analysis
Ammar Khanfer

Fundamentals of Functional
Analysis
Ammar Khanfer
Department of Mathematics and Sciences
Prince Sultan University
Riyadh, Saudi Arabia

ISBN 978-981-99-3028-9 ISBN 978-981-99-3029-6 (eBook)

https://doi.org/10.1007/978-981-99-3029-6

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Singapore Pte Ltd. 2023

This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
Preface

The present book provides a thorough treatment and a careful introduction to the field
of functional analysis with a clear exposition to the basic concepts and fundamentals
of functional analysis, and it covers the necessary topics required to master this
area of mathematics. Efforts have been made to make the book a self-study and
reader-friendly text to allow the student to get acquainted with the fundamentals of
advanced analysis and provide the basis for further studies that can be carried more
deeply. Because of these reasons, the book serves as a typical introductory book of
functional analysis for the students. The book contains material suitable for a two-
semester graduate course in analysis, so it is suitable for all graduate mathematics
students. The reader is expected to be familiar with linear algebra, real analysis, in
addition to basics of measure theory and Lebesgue’s integration.
The book is divided into eight chapters, and each chapter contains eight sections.
Chapter 1 introduces normed space and examples of some normed function spaces
and normed sequence spaces. Topology of normed spaces is discussed, with sepa-
rability and compactness properties. The chapter is concluded by equivalence of
norms.
Chapter 2 is, historically, the beginning of the area of functional analysis. The
chapter deals with linear operators and functionals between normed spaces rather
than finite-dimensional vector spaces, and we discuss the topologies of these spaces.
The main result is the famous Hahn–Banach theorem and its consequences. Most
books discuss this theorem in the treatment of Banach spaces. However, the present
book discusses the theorem before introducing Banach spaces to emphasize that the
Hahn–Banach theorem is applicable to general normed space.
Chapter 3 introduces the locally convex spaces and discusses the geometric
sequences of the Hahn–Banach theorem. The separation results and the main sepa-
ration theorems are discussed thoroughly. Then we study extreme points in normed
spaces with some examples, followed by the Krein–Milman theorem.
Chapter 4 introduces the Banach spaces and their geometric and topological prop-
erties. This is the place we introduce completeness for the first time in this text
to emphasize on the fact that the main results of Chap. 3 don’t require complete-
ness. Some great theorems that establish the cornerstones of functional analysis are

v
vi Preface

discussed in detail, such as the Baire category theorem, open mapping theorem,
closed graph theorem, uniform bounded principle, and the Banach–Steinhaus
theorem.
Chapter 5 deals with a particularly important class of Banach spaces in which an
inner product can be endowed. These are called Hilbert spaces. These spaces can
be viewed as the natural extension of the finite-dimensional Euclidean spaces. We
study these spaces in detail and establish essential and fundamental results based on
their rich geometry and orthogonality.
Chapter 6 presents the topology of Banach spaces in comparison to the topology
produced by their norms. These topologies are called the weak and weak-star topolo-
gies. We study how the space structure will be affected when these topologies are
adopted instead of the norm topology. Convergence will be a crucial factor in deter-
mining these topologies’ efficiency, giving rise to more flexible criteria in obtaining
compactness and convergence under these topologies. Some fundamental theorems
will be discussed, such as the Banach–Alaoglu theorem, Goldstine’s theorem, Krein–
Milman theorem, Eberlein–Smulian theorem, James theorem, and more. The chapter
concludes with a discussion of Schauder’s basis and weak basis.
Chapter 7 introduces operators on Banach spaces. The chapter discusses adjoint
of operators and self-adjoint operators in addition to other classes of operators, with
a focus on compact operators, and their approximation by finite-rank operators.
Chapter 8 presents the basics of the geometry of Banach spaces; strict convexity,
uniform convexity, smoothness, and differentiability of functionals are introduced.
Then the idea of renorming Banach spaces is discussed in brief.
The textbook contains more than 150 examples and 350 problems, in addition to
extensive historical and bibliographical remarks. The problems are distributed at the
end of each chapter. Hints and answers are given at the end of the book. The questions
aim to test whether the reader has absorbed the topic’s basics and gained a complete
understanding. Therefore, the levels of the questions range from straightforward to
challenging, and some of them are standard in their subjects that can be found in
many other books. Moreover, these questions help students prepare for the Ph.D.
comprehensive exams. It should be noted that some questions are answered in detail,
while others are left with some hints or key answers. Similarly, some details in the
proofs throughout the book are left to the reader to complete independently.
It is important to know that mastering this book’s topics is not possible without
the time and effort needed to overcome all the difficulties that may arise in terms of
grasping the material and mastering the subject. This will prepare students to think
independently and test their maturity and background to accomplish the task by filling
in the gaps and unveiling all the hidden details whenever needed—a common practice
when one enrolls in a graduate program and inevitable when one starts independent
research.

Riyadh, Saudi Arabia Ammar Khanfer

2022
Acknowledgments

I have to start by thanking the God most of all for giving me the strength, health,
knowledge, and patience to endure and complete this work successfully.
It is a pleasure to offer my grateful appreciation to everyone who helped me in
completing this book. Writing a book is more exhausting than one can imagine and
more gratifying than one might expect. Before being rewarded, it is important in this
regard to recall our math teachers. I am immensely thankful to everyone who taught
me mathematics at various stages of my life from childhood to the Ph.D. level. Words
cannot express gratitude to them. I sincerely thank all of you.
It is a pleasure to express my sincere thanks to Prince Sultan University for its
continuing support. I also wish to express my deep thanks and gratitude to Prof.
Mahmoud Al Mahmoud, the Dean of the college (CHS), and Prof. Wasfi Shatanawi,
the Chair of the Department of Mathematics (MSD), for their support and recogni-
tion of my work. Many thanks extend to my colleagues of our department and the
colleagues at the Research and Initiative Center (RIC). I am grateful to all of them
for their warm encouragement.
I am profoundly indebted to the anonymous reviewers for their valuable comments
and helpful suggestions which improved the book significantly.
I would also like to thank the editorial board and the staff at Springer who were
involved in the review and publication of the book and bringing my work to fruition. I
would like to single out some individuals for special thanks: Shamim Ahmad, Iydah
Grace, and Ramya Somasundaram for their cooperation and being tremendously
helpful and patient in dealing with this project professionally from the review stage
to the production stage.
A special debt of gratitude is owed to my wife and family for their emotional
support. I deeply acknowledge the love and encouragement that I received from
them during the preparation of this book.
Lastly, I want to thank everyone who supported me with a positive word or feeling.
I heard it all .. and it meant quite something!

vii
Contents

1 Normed Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Gate to Infinite-dimensional Spaces . . . . . . . . . . . . . . . . . . . 1
1.1.2 Why We Need Norms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Function Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 The Space of Continuous Functions . . . . . . . . . . . . . . . . . . . 3
1.2.2 The Space of n–Continuously Differentiable
Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.3 The Space of All Continuous Functions
with Compact Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.4 The Space of Continuous Functions Vanishing
at Infinity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.5 The Space of Functions of Bounded Variation . . . . . . . . . . . 4
1.2.6 The Space of Absolutely Continuous Functions . . . . . . . . . 5
1.2.7 The Space of Riemann Integrable Functions . . . . . . . . . . . . 6
1.2.8 The Space of Riemann–Stieltjes Integrable Functions . . . . 6
1.3 The Notion of Norm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.1 Definition and Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.2 Continuity of Norms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.3 Norms Versus Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4 Definition and Examples of Normed Spaces . . . . . . . . . . . . . . . . . . . . 10
1.4.1 Definition of Normed Spaces . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4.2 Convergence in Normed Spaces . . . . . . . . . . . . . . . . . . . . . . . 11
1.4.3 Examples of Normed Spaces . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5 L p Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.5.1 L p Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.5.2 L ∞ Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.5.3  p and ∞ Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.5.4 Basic Inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.5.5 Measure Theory and Integration: Quick Review . . . . . . . . . 19

ix
x Contents

1.6 Topology of Normed Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

1.6.1 Balls and Sphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.6.2 Open and Closed Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.6.3 Quotient Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.6.4 Distance to Closed Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.6.5 Riesz’s Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.6.6 Compactness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.6.7 Separability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.7 Equivalent Norms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.7.1 Equivalence of Norms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.7.2 Topological Interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.7.3 Norms in Finite-Dimensional Spaces . . . . . . . . . . . . . . . . . . 32
1.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2 Linear Functionals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.1 Linear Mappings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.1.1 Definitions: Operators and Functionals . . . . . . . . . . . . . . . . . 37
2.1.2 Continuous Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.1.3 Identity and Shift Operators . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.1.4 Differential and Integral Operators . . . . . . . . . . . . . . . . . . . . 43
2.1.5 Integral Functional . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.2 Inverse of Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.2.1 Injective and Surjective Operators . . . . . . . . . . . . . . . . . . . . . 45
2.2.2 Inverses and Kernels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.2.3 Isometry and Isomorphism . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.3 Dual Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.3.1 The Notion of Dual Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.3.2 Dual Space of a Finite-dimensional Spaces . . . . . . . . . . . . . 50
2.3.3 Annihilators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
2.3.4 Dual Space of Euclidean Space . . . . . . . . . . . . . . . . . . . . . . . 52
2.3.5 Dual Space of  p . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.3.6 Dual Space of c0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.4 Riesz Representation Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.4.2 Riesz Representation Theorem for L p Spaces . . . . . . . . . . . 58
2.4.3 Dual Space of L p . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.4.4 Riesz Representation Theorem for C[a, b] . . . . . . . . . . . . . . 60
2.4.5 Dual Space of C[a, b] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.5 Extension Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.5.2 Extensions and Seminorms . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.5.3 Fundamental Extension Theorem . . . . . . . . . . . . . . . . . . . . . . 63
2.5.4 The Extension Theorem in Finite-Dimensional
Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Contents xi

2.6 Hahn–Banach Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

2.6.1 Historical Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
2.6.2 Hausdorff Maximal Principle . . . . . . . . . . . . . . . . . . . . . . . . . 68
2.6.3 HBT for Vector Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
2.6.4 HBT with Seminorms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.6.5 Hahn–Banach Theorem for Normed Spaces (HBT) . . . . . . 72
2.7 Consequences of Hahn–Banach Theorem . . . . . . . . . . . . . . . . . . . . . . 74
2.7.1 The Existence of Bounded Functional . . . . . . . . . . . . . . . . . . 74
2.7.2 Richness of Dual Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.7.3 Banach Separation Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.7.4 Separability of Normed Spaces . . . . . . . . . . . . . . . . . . . . . . . 76
2.7.5 Proof of Riesz Representation Theorem for C[a, b] . . . . . . 77
2.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3 Locally Convex Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.1 Hyperplanes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.1.1 Introduction and Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.1.2 Separating Hyperplanes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.1.3 Convex Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.2 Minkowski Functional . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.2.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.2.2 Properties of Minkowski Functional . . . . . . . . . . . . . . . . . . . 89
3.3 Topological Vector Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.4 Constructing Locally Convex Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.4.1 TVS with Seminorms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.4.2 Locally Convex Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3.4.3 Metrizability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
3.5 Separation Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
3.5.1 Mazur Separation Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
3.5.2 Bourgin Separation Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 102
3.5.3 Generalized Banach Separation Theorem . . . . . . . . . . . . . . . 103
3.5.4 Eidelheit Separation Theorem . . . . . . . . . . . . . . . . . . . . . . . . 104
3.5.5 Tuckey—Klee Separation Theorem . . . . . . . . . . . . . . . . . . . . 105
3.5.6 Historical Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
3.6 Extreme Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
3.6.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
3.6.2 Examples of Extreme Points . . . . . . . . . . . . . . . . . . . . . . . . . . 107
3.6.3 Faces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
3.7 Krein-Milman Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
3.7.1 Convex Hull . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
3.7.2 Statement and Proof of KMT . . . . . . . . . . . . . . . . . . . . . . . . . 112
3.7.3 Applications of KMT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
3.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
xii Contents

4 Banach Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

4.1 Completeness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
4.1.1 The Idea of Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
4.1.2 Cauchy Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
4.1.3 Complete Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.1.4 Absolutely Summable Series . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.1.5 Examples of Complete Spaces . . . . . . . . . . . . . . . . . . . . . . . . 123
4.2 Definition and Examples of Banach Spaces . . . . . . . . . . . . . . . . . . . . . 125
4.2.1 Notion of Banach Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
4.2.2 Continuous Functions Over Compact Sets . . . . . . . . . . . . . . 125
4.2.3 Continuous Functions Over σ -Compact Sets . . . . . . . . . . . . 127
4.2.4 C n (K ) & C ∞ (K ) Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
4.2.5 Completeness of  p Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
4.2.6 Completeness of L p Space . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
4.2.7 Incomplete Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
4.3 Properties of Banach Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
4.3.1 Closedness and Completeness . . . . . . . . . . . . . . . . . . . . . . . . 133
4.3.2 Operators Between Banach Spaces . . . . . . . . . . . . . . . . . . . . 134
4.3.3 Isomorphisms of Banach Spaces . . . . . . . . . . . . . . . . . . . . . . 135
4.3.4 Direct Sum of Banach Spaces . . . . . . . . . . . . . . . . . . . . . . . . . 138
4.3.5 Completion of Incomplete Spaces . . . . . . . . . . . . . . . . . . . . . 140
4.4 Baire Category Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
4.4.1 First and Second Category . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
4.4.2 Weak Version of the Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 142
4.4.3 Strong Version of the Theorem . . . . . . . . . . . . . . . . . . . . . . . . 143
4.4.4 Consequences of BCT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
4.5 Open Mapping Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
4.5.1 Open Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
4.5.2 The Open Mapping Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 146
4.5.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
4.5.4 Bounded Inverse Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
4.6 Closed Graph Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
4.6.1 Closed Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
4.6.2 The Closed Graph Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 152
4.6.3 Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
4.7 Uniform Bounded Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
4.7.1 Pointwise and Uniformly Boundedness . . . . . . . . . . . . . . . . 154
4.7.2 Uniform Bounded Principle (Weak Version) . . . . . . . . . . . . 155
4.7.3 Historical Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
4.7.4 Uniform Bounded Principle (Strong Version) . . . . . . . . . . . 157
4.7.5 Banach–Steinhaus Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 158
4.7.6 Totally Boundedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
4.7.7 Applications of UBP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
4.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
Contents xiii

5 Hilbert Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169

5.1 Inner Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
5.1.1 Inner Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
5.1.2 Cauchy–Schwartz Inequality . . . . . . . . . . . . . . . . . . . . . . . . . 171
5.1.3 Parallelogram and Polarization Laws . . . . . . . . . . . . . . . . . . 172
5.2 Inner Product Spaces and Hilbert Spaces . . . . . . . . . . . . . . . . . . . . . . . 173
5.2.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
5.2.2 Euclidean Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
5.2.3 L 2 and 2 Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
5.2.4 Counterexamples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
5.3 Functionals on Hilbert Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
5.3.1 Riesz Representation Theorem for Hilbert Spaces . . . . . . . 176
5.3.2 Hahn–Banach Theorem for Hilbert Spaces . . . . . . . . . . . . . . 178
5.4 Orthogonality in Hilbert Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
5.4.1 Orthogonality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
5.4.2 Best Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
5.4.3 Projections in Hilbert Spaces . . . . . . . . . . . . . . . . . . . . . . . . . 181
5.4.4 Orthonormal Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
5.5 Orthonormal Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
5.5.1 Total Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
5.5.2 Orthonormal Bases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
5.5.3 Parseval Identity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
5.5.4 Fourier Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
5.6 Nonseparable Hilbert Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
5.6.1 Uncountable Orthonormal Bases . . . . . . . . . . . . . . . . . . . . . . 194
5.6.2 Hilbert Dimension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
5.7 Isomorphisms Between Hilbert Spaces . . . . . . . . . . . . . . . . . . . . . . . . 197
5.7.1 Hilbert Isomorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
5.7.2 Classification of Hilbert Spaces . . . . . . . . . . . . . . . . . . . . . . . 200
5.7.3 Self Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
5.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
6 Topology on Banach Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
6.1 Weak Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
6.1.1 The Weak Limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
6.1.2 Weak Convergence in L p . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
6.1.3 Radon–Riesz Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
6.1.4 Radon–Riesz Property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
6.1.5 Schur’s Property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
6.2 Weak Topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
6.2.1 The Idea of Weak Topology . . . . . . . . . . . . . . . . . . . . . . . . . . 218
6.2.2 Weak Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
6.2.3 Weakly Closed Sets and Mazur’s Theorem . . . . . . . . . . . . . . 222
6.2.4 Weak Compactness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
xiv Contents

6.2.5 Link to Hausdorff Topology . . . . . . . . . . . . . . . . . . . . . . . . . . 226

6.2.6 Notion of Nets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
6.3 Weak* Topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
6.3.1 The Emebedding Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
6.3.2 The Bidual . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
6.3.3 Weak* Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
6.3.4 Properties of Weak* Topology . . . . . . . . . . . . . . . . . . . . . . . . 233
6.3.5 Link to Hausdorff Topology . . . . . . . . . . . . . . . . . . . . . . . . . . 234
6.3.6 Link to Locally Convex Topology . . . . . . . . . . . . . . . . . . . . . 235
6.3.7 Weak* Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
6.3.8 Compactness in the Product Spaces . . . . . . . . . . . . . . . . . . . . 238
6.3.9 Banach–Alaoglu Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
6.3.10 Goldstine Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
6.3.11 Weak* Heine–Borel Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 243
6.3.12 Krein–Milman Theorem on Weak Topology . . . . . . . . . . . . 244
6.3.13 Existence of Completion Space . . . . . . . . . . . . . . . . . . . . . . . 245
6.4 Reflexive Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
6.4.1 Definition and Basic Properties . . . . . . . . . . . . . . . . . . . . . . . 245
6.4.2 Reflexivity and Weak Compactness . . . . . . . . . . . . . . . . . . . . 247
6.4.3 Properties of Reflexive Spaces . . . . . . . . . . . . . . . . . . . . . . . . 249
6.4.4 Helly’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
6.4.5 Consequences of Reflexivity . . . . . . . . . . . . . . . . . . . . . . . . . . 251
6.5 Weakly Compactness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
6.5.1 The Sequentially Compactness Problem . . . . . . . . . . . . . . . . 253
6.5.2 Weak Compactness and Metrizibility . . . . . . . . . . . . . . . . . . 254
6.5.3 Eberlein–Smulian Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 257
6.5.4 James Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
6.6 Bases in Banach Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
6.6.1 Hamel Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
6.6.2 Schauder Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
6.6.3 Basic Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
6.6.4 Basis and Separability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
6.6.5 Bases in Banach Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
6.6.6 Biorthogonal System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
6.6.7 Applications to Banach Spaces . . . . . . . . . . . . . . . . . . . . . . . . 275
6.7 Weak Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
6.7.1 Notion of Weak Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
6.7.2 Weak Bases in Banach Spaces . . . . . . . . . . . . . . . . . . . . . . . . 278
6.7.3 Weak* Basis for Dual Spaces . . . . . . . . . . . . . . . . . . . . . . . . . 279
6.7.4 Dual Projection Maps on Dual Spaces . . . . . . . . . . . . . . . . . 280
6.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
Contents xv

7 Operators on Banach Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289

7.1 Adjoint of Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
7.1.1 The Idea of Adjoint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
7.1.2 The Notion of Adjoint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
7.1.3 Definition of Adjoint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
7.1.4 Basic Properties of Adjoints . . . . . . . . . . . . . . . . . . . . . . . . . . 291
7.2 Adjoint Operators on Hilbert Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . 293
7.2.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
7.2.2 Existence Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
7.2.3 Adjoint of Unbounded Operators . . . . . . . . . . . . . . . . . . . . . . 294
7.2.4 Basic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
7.2.5 Range and Null Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
7.2.6 Examples of Adjoint Operators . . . . . . . . . . . . . . . . . . . . . . . 297
7.3 Classes of Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
7.3.1 Self-adjoint Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
7.3.2 Characterization of Self-adjoint Operators . . . . . . . . . . . . . . 300
7.3.3 Fredholm Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
7.3.4 Decomposition of Bounded Operator . . . . . . . . . . . . . . . . . . 301
7.3.5 Normal Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
7.3.6 Unitary and Positive Operators . . . . . . . . . . . . . . . . . . . . . . . . 304
7.4 Compact Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
7.4.1 Characterization of Compact Operators . . . . . . . . . . . . . . . . 307
7.4.2 Compactness and Adjointness . . . . . . . . . . . . . . . . . . . . . . . . 308
7.4.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
7.4.4 Compactness on Finite Dimensions . . . . . . . . . . . . . . . . . . . . 311
7.5 Completely Continuous Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
7.5.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
7.5.2 Completely Continuity and Compactness . . . . . . . . . . . . . . . 312
7.5.3 Example of A Sequence That is Not Completely
Continuous . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
7.5.4 Example of A Completely Continuous But Not
Compact Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
7.6 Operators of Finite Rank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
7.6.1 The Notion of Finite Rank . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
7.6.2 Example of a Compact But Not Finite-Rank Operator . . . . 316
7.6.3 The Rank of the Adjoint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
7.7 Approximation of Compact Operators . . . . . . . . . . . . . . . . . . . . . . . . . 319
7.7.1 Sequence of Compact Operators . . . . . . . . . . . . . . . . . . . . . . 319
7.7.2 Approximation By Finite-Rank Operators . . . . . . . . . . . . . . 320
7.7.3 Constructing a Sequence of Finite-Rank Operators . . . . . . . 321
7.7.4 Compactness of the Adjoint . . . . . . . . . . . . . . . . . . . . . . . . . . 322
7.7.5 Historical Remark: The Approximation Problem . . . . . . . . 323
7.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
xvi Contents

8 Geometry of Banach Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329

8.1 Convexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
8.1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
8.1.2 Strictly Convexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330
8.2 Strict Convexity in p-Norm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
8.2.1 Convex Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
8.2.2 p-Characterization of Stricly Convex Spaces . . . . . . . . . . . . 332
8.2.3 Best Approximation in Stricly Convex Spaces . . . . . . . . . . . 334
8.2.4 Taylor–Foguel Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336
8.3 Uniform Convexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
8.3.1 Motivation for Uniform Convexity . . . . . . . . . . . . . . . . . . . . 337
8.3.2 Necessary and Sufficient Conditions . . . . . . . . . . . . . . . . . . . 339
8.3.3 Milman–Pettis Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
8.3.4 Glimpse of Super-Reflexivity . . . . . . . . . . . . . . . . . . . . . . . . . 342
8.4 Uniform Convexity in p-Norm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
8.4.1 A Different Look at Uniform Convexity . . . . . . . . . . . . . . . . 343
8.4.2 Best Approximation in Uniformly Convex Spaces . . . . . . . 345
8.4.3 Clarkson Inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346
8.4.4 Uniform Convexity of L p . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
8.4.5 Radon–Riesz Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
8.4.6 Historical Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
8.5 Smoothness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
8.5.1 Supporting Hyperplanes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
8.5.2 Duality Between Strict Convexity and Smoothness . . . . . . . 353
8.5.3 Uniformly Smooth Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
8.5.4 Modulus of Smoothness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
8.5.5 Lindenstrauss Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
8.5.6 Duality Between Uniform Convexity and Uniform
Smoothness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
8.6 Differentiability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
8.6.1 Gateaux Differentiability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
8.6.2 Basic Properties of Gateaux Derivative . . . . . . . . . . . . . . . . . 362
8.6.3 Frechet Derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365
8.7 Renorming Banach Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367
8.7.1 The Idea of Renorming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367
8.7.2 Dual Norm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 368
8.7.3 Smulian Criterions for Differentiability . . . . . . . . . . . . . . . . 369
8.7.4 Kadets Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
8.7.5 Locally Uniformly Conves Spaces . . . . . . . . . . . . . . . . . . . . . 372
8.7.6 LUC Norm Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
8.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374
Contents xvii

Answer Key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 441
About the Author

Ammar Khanfer earned his Ph.D. from Wichita State University, USA. His area of
interest is analysis and partial differential equations (PDEs), focusing on the interface
and links between elliptic PDEs and hypergeometry. He has notably contributed to
the field by providing prototypes studying the behavior of generalized solutions of
elliptic PDEs in higher dimensions in connection to the behavior of hypersurfaces
near nonsmooth boundaries. He also works on the qualitative theory of differential
equations, and in the area of inverse problems of mathematical physics. He has
published articles of high quality in reputable journals.
Ammar taught at several universities in the USA: Western Michigan University,
Wichita State University, and Southwestern College in Winfield. He was a member
of the Academy of Inquiry Based Learning (AIBL) in the USA. During the period
2008–2014, he participated in AIBL workshops and conferences on effective teaching
methodologies and strategies of creative thinking, which made an impact on his
engaging and motivational writing style. He then moved to Saudi Arabia to teach
at Imam Mohammad Ibn Saud Islamic University, where he taught and supervised
undergraduate and graduate students of mathematics. Furthermore, he was appointed
as coordinator of the PhD program establishment committee in the department of
mathematics. In 2020, he moved to Prince Sultan University in Riyadh, and has been
teaching there since then.

xix
Chapter 1
Normed Spaces

1.1 Motivation

1.1.1 Gate to Infinite-dimensional Spaces

b
In order to maximize (or minimize) the integral a y(x)d x for all possible values
of y(x), we see that the candidates competing with each other are in the form of
functions, not numbers or points. So we can rewrite the integral as
 b
I (y) = y(x)d x, (1.1.1)
a

and the goal is to find an appropriate function y that would maximize or minimize
I (y). We have the following observations here:
(1) I (y) is a real-valued function of functions, i.e., the domain of this function is
the set of all continuous or measurable functions on [a, b], and the range of the
function is R.
(2) Because of the integral, the function is linear, i.e., I (c1 y1 + c2 y2 ) = c1 I (y1 ) +
c2 I (y2 ), so we require the function c1 y1 + c2 y2 to be a candidate function that
compete with other functions. Hence this function must also belong to the same
set. But a set with such property is nothing but a vector space. So the set of
candidate functions must be a vector space.
(3) The elements of this vector space are of the form of functions, so we can call it
a “function space”. If we consider a vector v ∈ Rn , then v = (x1 , x2 , . . . , xn ),
so v has n components in the space. In order to write the function f (x) in the
same representation, we see that x varies from a to b, and f will depend on
each value x from a to b, so writing f in the same representation will result in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 1
A. Khanfer, Fundamentals of Functional Analysis,
https://doi.org/10.1007/978-981-99-3029-6_1
2 1 Normed Spaces

an array of the form ( f (x))x∈[a,b] , and this means that f (x) will have infinitely
many components.
These spaces containing the candidate functions are, in fact, spaces of infinite dimen-
sions, and consequently many rules and properties regarding the finite-dimensional
spaces fail to satisfy.

1.1.2 Why We Need Norms

In order to obtain a useful tool, these new spaces must be equipped with a geo-
metric structure. Convergence is one of the most important and basic concepts in
mathematical analysis through which we can characterize continuity of functions,
compactness of sets, convergence of sequences, and many other useful notions and
concepts. The convergence is based on the notion of distance between numbers which
is given by the absolute value function. Now, the elements in our new spaces are no
longer numbers but functions, so how to define the “length” of a function and the
“distance” between functions? What is the appropriate geometric structure that can
be provided with this feature? What is the generalized version of the absolute value
function? These questions can be answered by the notion of “norm” , which will be
essential for any space of functions to behave well and turn the space into a useful
mathematical tool.
Keeping in mind that the original problem that motivated us was the problem of
optimizing an integral of the form
 b
I (y) = y(x)d x,
a

the only tool at our disposal to treat this problem is the derivative, but I (y) is a function
of functions, and we do not know how to give a derivative for such functions because
the limit needs the concept of convergence and distance between functions to control
these distances and make them as small as desired. Hence, if the space of functions
can be endowed with a norm, the space will have a good geometric structure that
allows measuring the length of a function and the distance between two functions,
in addition to algebraic structure that allows the functions to interact between each
other via addition and other operations.
1.2 Function Spaces 3

1.2 Function Spaces

Review of Vector Spaces. Recall that a vector (or linear) spaceV is a set of elements
(called vectors) in which two operations are defined; the addition of two vectors and
the scalar multiplication of a vector with a constant in a field F. These operations
obey the following axioms:
(1) If v1 , v2 ∈ V and c1 , c2 ∈ R then c1 v1 + c2 v2 ∈ V.
(2) Commutative and associative laws of addition hold.
(3) Distributivity of scalar multiplication with respect to field and vector additions.
So there exist v1 , v2 ∈ V and a, b ∈ F such that a(v1 + v2 ) = av1 + av2 , and
(a + b)v = av + bv.
(4) Additive and multiplicative identities exist. There exists 0 ∈ V such that v + 0 =
v for all v ∈ V, and there is 1 ∈ V such that a.v = v for all v ∈ V .
(5) Scalar multiplication is associative, that is, a(bv) = (ab)v for all v ∈ V.
(6) There exists an additive inverse −v such that v + (−v) = 0.
Taking into account the following remarks:
(1) It is because of axiom 1 these spaces are also called “linear spaces” since all
elements obey linearity.
(2) We are interested only in real-valued functions; hence we will always let F = R.
Any space obeying these axioms is called “vector space”, or sometimes “linear
space”. The two names are used interchangeably, though the former is rather more
common.
A vector space is called “function space” if all of its elements are functions. Some
examples of function spaces are

1.2.1 The Space of Continuous Functions

Example 1.2.1 C[a, b]: the space of continuous functions on [a, b]. This is clearly a
function vector space since the scalar multiple of a continuous function is continuous,
and the summation of two continuous functions is continuous; hence this defines a
vector space.

1.2.2 The Space of n–Continuously Differentiable Functions

Example 1.2.2 C n [a, b]: the space of n−continuously differentiable functions, that
is, f, f  , . . . , f (n) are all continuous on [a, b]. It is clear that for every n ≤ m we
have
4 1 Normed Spaces

C m [a, b] ⊆ C n [a, b] ⊆ C[a, b],

where it is understood that C 0 [a, b] = C[a, b]. Moreover, the space C ∞ [a, b] defined
by
∞

∞
C [a, b] = C n [a, b]
n=0

consists of all continuous functions that are infinitely differentiable, i.e., smooth
functions that have derivatives of all orders.

1.2.3 The Space of All Continuous Functions with Compact

Support

Example 1.2.3 Cc (R): the space of all continuous functions f on R with compact
support. A compact suppor t of a function, denoted by supp( f ), is defined as

supp( f ) = {x ∈ Dom( f ) : f (x) = 0}.

So, for every f ∈ Cc (R) there exists a compact interval, say [a, b] such that f (x) = 0
for all x ∈
/ [a, b].

1.2.4 The Space of Continuous Functions Vanishing

at Infinity

Example 1.2.4 C0 (R): the space of continuous functions on R s.t. lim f (x) = 0,
x→±∞
f ∈ C0 (R).

It is readily seen that Cc (X ) = C0 (X ). In general, we have

Cc (X ) ⊆ C0 (X ) ⊆ C(X ).

1.2.5 The Space of Functions of Bounded Variation

Example 1.2.5 BV [a, b]: the space of functions of bounded variation on [a, b].
Let f be a real-valued function on [a, b] and P is a partition of [a, b], and let the
variation of f with respect to P be defined by
1.2 Function Spaces 5


n
vab ( f ; P) = | f (xi ) − f (xi−1 )| .
i=1

A function f is of bounded variation on [a, b] if it has a finite total variation V ( f )

over [a, b], that is,

V ( f ) = supvab ( f ; P) < ∞,
P

where the supremum is taken over all possible partitions of [a, b]. This is a function
vector space. It can be shown1 that the function
⎧
⎪
⎨x sin( x ) 0 < x ≤ 1
1

f (x) =
⎪
⎩
0 x = 0.

is not of bounded variation, but it is continuous on [0, 1]. Consequently, we have

C[a, b]  BV [a, b].

1.2.6 The Space of Absolutely Continuous Functions

The stronger absolute continuity condition suffices to give the inclusion above.
Example 1.2.6 AC[a, b]: the space of absolutely continuous functions on [a,b].
Recall that a function is called absolutely continuous on [a, b] if given  > 0, there
exists δ > 0 such that if {[xi , yi ] : i = 1, 2, . . . , n} is a finite collection of pairwise
disjoint subintervals in [a, b] and


n
|yi − xi | < δ
i=1

then


n
| f (yi ) − f (xi )| < .
i=1

It is well known that a function that is absolutely continuous on [a, b] is of bounded

variation on [a, b], and we can also find functions of bounded variations on [a, b]
that are not absolutely continuous on [a, b], therefore we have

1 See Example 2.6.4, vol. 1 [51].

6 1 Normed Spaces

AC[a, b]  BV [a, b].

1.2.7 The Space of Riemann Integrable Functions

Example 1.2.7 R[a, b]: the space of Riemann integrable functions over [a, b].
It is well known that the Riemann integral is linear, so the space R[a, b] is a
function vector space.

1.2.8 The Space of Riemann–Stieltjes Integrable Functions

Example 1.2.8 R(g)[a, b]: the space of Riemann–Stieltjes integrable functions. Let
b
f and g be bounded functions on [a, b], and consider the integral a f dg. This
integral is defined as follows:
Let
Pn = a = t0 < t1 < · · · < tn = b

be any partition of [a, b], and consider the sum


n
S( f, g; P) = f (τ j )[g(t j ) − g(t j−1 )],
j=1

for t j−1 ≤ τ j ≤ t j . Then we say that f is Riemann–Stieltjes integrable with respect

to g if there exists a number I such that for every  > 0 there exists δ > 0 such that
|S( f, g; P) − I | <  whenever

max {t j − t j−1 } < δ.

1≤ j≤n

The number I is the Riemann–Stieltjes integral of f with respect to g and is written

as  b n
I = f dg = lim f (τ j )[g(t j ) − g(t j−1 )].
a j=1

We denote the space of all Riemann–Stieltjes integrable functions over [a, b] with
respect to the function g by R(g)[a, b]. It is easy to see that the Riemann–Stieltjes
integral is linear with respect to f and to g, so the space is a function vector space.
1.3 The Notion of Norm 7

1.3 The Notion of Norm

1.3.1 Definition and Examples

We need to impose some geometric structure on a vector space X. In particular, we

need to define a metric on that space, and since the space is linear, we can begin with
a norm, then we define a metric induced by the norm. Recall a norm on a space X is
a function · : X → [0, ∞) such that the following conditions hold:
(1) x ≥ 0 for every x ∈ X.
(2) x = 0 if and only if x = 0.
(3) ax = |a| . x for every x ∈ X and a ∈ R.
(4) x + y ≤ x + y , for every x, y ∈ X.
Any function that satisfies the four axioms above is called a norm. If the function ·
satisfies all conditions but (2), then we call · a seminorm. As can be seen from the
conditions above, the norm generalizes the idea of absolute value for numbers.
Some examples of norm of vectors (x1 , x2 , . . . , xn ) ∈ Rn are:
• The 1-norm
x 1 = |x1 | + |x2 | + · · · + |xn | .

• The 2-norm,
1
x 2 = |x1 |2 + |x2 |2 + · · · + |xn |2 2
.

This is also known as the Euclidean norm.

• The infinity norm:
x ∞ = max{|x1 | , |x2 | , . . . , |xn |}.

This is also known as the supremum norm.

All these norms can be extended to infinitely many coordinates when dealing with
infinite-dimensional spaces. Indeed, if n → ∞, then the vectors convert to sequences
(xn ) = x1 , x2 , . . . , and the norms become
• The 1-norm
∞

x 1 = |xi |
i=1

• The 2-norm,
∞ 1/2

x 2 = |xi |2 ,
i=1

• The supremum norm

x ∞ = sup{|xn |}.
n
8 1 Normed Spaces

If there are uncountably infinite number of coordinates, the sequences convert

to functions f (x), is not is the case the 1-norm will convert to an integral, and the
norms become
• The 1-norm 
f 1 = | f|,

• The 2-norm,  1/2

f 2 = | f |2 ,

• The supremum norm

f ∞ = sup{| f (x)|}.
n

These are the most common norms although we have many more norms to deal with
other than these. The 2-norm can be generalized to any number p, to become
∞ 1/ p

x p = |xi | p

i=1

for sequences, and

 1/ p
f p = | f |2

for functions.
The absolute value is given to ensure the nonnegativity of the integral value and
fulfill axiom 1, and the power 1/ p in the 2-norm is taken to ensure homogeneity and
fulfil axiom 3. Surprisingly, axiom 2 is the most problematic axiom, and it fails to
hold in many cases. The proposed function can still give nice geometric properties
for the space, and thus it has been considered and given the name seminorm.

1.3.2 Continuity of Norms

It is well known from elementary analysis courses that the absolute value function is
continuous throughout its domain. Since the norm function is a generalization of the
absolute function, we expect this would also be the case for norms, which is indeed
a fundamental property of norms.
Proposition 1.3.1 The norm function is continuous.

Proof Notice that the norm is a real-valued function. Therefore we write x as

1.3 The Notion of Norm 9

x = x−y+y ≤ x−y + y ,

from which we have

| x − y |≤ x−y . 

1.3.3 Norms Versus Metrics

Another important notion that is associated with norms is the metric. Recall a metric
is a function d : X × X → [0, ∞) satisfying the following:
(1) Positive Definiteness: d(x, y) ≥ 0 and d(x, y) = 0 iff x = y.
(2) Symmetry: d x, y) = d(y, x).
(3) Triangle Inequality: d(x, y) ≤ d(x, z) + d(z, y).
The distance function is the metric on Rn , for if x, y ∈ Rn then

d(x, y) = (x1 − y1 )2 + · · · + (xn − yn )2 .

Note here that a metric in general doesn’t satisfy homogeneity condition (axiom 3
of norms). So every norm induces a metric by the following definition of metric

d(x, y) = x − y (1.3.1)

Indeed,

d(x, y) = x − y ≥ 0,

and

d(x, y) = 0 ⇒ x − y = 0 ⇒ x = y.

Symmetry holds since

d(x, y) = x − y = y − x = d(y, x).

Triangle Inequality holds since

d(x, y) = x − y
≤ x−z + z−y
= d(x, z) + d(z, y).

Hence, the function defined in (1.3.1) is a metric.

10 1 Normed Spaces

On the other hand, not every metric can induce a norm. Consider, for example,
the following metric: 
1 x=y
d(x, y) = .
0 x = y

This metric obviously does not satisfy axiom 3 of norms, so it cannot induce a norm.
The following proposition provides sufficient conditions under which a metric can
induce a norm.
Proposition 1.3.2 If d is a metric and satisfies the following:

(1) Translation Invariant: d(x, y) = d(x + z, y + z).

(2) Homogeneity: d(cx, cy) = |c| d(x, y).

Then the metric d induces a norm x = d(x, 0).

Proof Left to the reader. 
When defined on a set or space, the metric can give some geometric features because
it measures the distance between elements of the set. It, however, does not preserve
scale properties (which is assured by homogeneity condition) and does not preserve
translations and shifts. Norms preserve both conditions, and not only does it provide
distances between elements of the space but also gives rise to the notion of the length
of every element, which explains why norms provide rich information regarding the
structure of the space. This is a crucial difference between the two notions, and it
allows metrics to be defined on sets and spaces without any algebraic structure on
these spaces. On the other hand, norms require some algebraic structure to perform
the algebraic operations such as summation and multiplying with a scalar. Notice
that axioms 3 and 4 of norms justify the space’s necessity to be a vector space since
the values x + y and cx will not necessarily exist in X if it is not a vector space. If a
norm is defined on a vector space, it provides rich geometric and algebraic properties
to further investigate the space and introduce other notions.

1.4 Definition and Examples of Normed Spaces

1.4.1 Definition of Normed Spaces

Consider the set containing all functions defined on the interval on [a, b]. Checking
the axioms of the norm above, we can conclude that this set forms a vector space,
with elements being of the form of functions. As said above, this space is of infinite
dimensions. By adding some specific conditions on the functions, the space gets
smaller because these conditions work as constraints that may rule out some functions
that do not obey the conditions. Therefore, more the conditions we impose, more
restrictive the space becomes. We begin by the definition.
1.4 Definition and Examples of Normed Spaces 11

Definition 1.4.1 (Normed Space) Let X be a vector space. If X is endowed by a norm

· , then the space (X, · ) is called: “Normed vector space” or simply: “normed
space”.
The advantage of having a normed space is that it combines the geometric structure
provided by metrics in metric spaces with the algebraic structure provided by vectors
spaces, which enables us to handle many topological properties and explore further
results about the space.

1.4.2 Convergence in Normed Spaces

Definition 1.4.2 (Convergent Sequence) Let (X, · ) be a normed space, and (xn )
be a sequence in X. Then we say that xn converges to x if for every  > 0 there
exists N ∈ N such that xn − x <  for all n ≥ N . This convergence is denoted by
xn − x → 0, and x is called the limit of (xn ).

Definition 1.4.3 (Bounded Sequence) Let (X, · ) be a normed space, and (xn ) be
a sequence in X. Then we say that xn is bounded on X if sup xn ≤ M < ∞.
n

Proposition 1.4.4 Let (X, · ) be a normed space, and (xn ) be a convergent

sequence in X. Then:

(1) (xn ) is bounded.

(2) (xn ) has a unique limit.

Proof (1): Let  = 1. Since (xn ) is convergent, there is N ∈ N such that xn − x ≤ 1

for all n ≥ N , so
xn ≤ x + 1.

Now, let
M = max{ x1 , x2 , . . . , x N −1 , x + 1}.

Then xn ≤ M for all n.

(2): Let x, y be two limits of (xn ). Then

x − y = x − xn + xn − y
≤ x − xn + xn − y −→ 0. 

The converse of (1) is not necessarily true as can be seen clearly from the sequence
xn = (−1)n for example.
12 1 Normed Spaces

1.4.3 Examples of Normed Spaces

The following are examples of normed spaces.

Example 1.4.5 (Euclidean Spaces) The space Rn with the Euclidean norm:

x 2 = |x1 |2 + |x2 |2 + · · · + |xn |2

for x = (x1 , x2 , . . . , xn ) is a normed space, and the norm · 2 represents the distance
between a point and the origin. We usually denote it by (Rn , · 2 ). Using Cauchy–
Schwarz inequality, we obtain the triangle inequality. The spaces (Rn , · 1 ) and
(Rn , · ∞ ) are other normed spaces. It’s the same vector space but with different
norms. These norms affect the geometric and topological properties of the space.

Example 1.4.6 (Space of Continuous Functions) Consider the space C([a, b])
defined in Example 1.2.1. Define the following norm:

f ∞ = sup | f (x)| .
x∈[a,b]

This is a generalization to the sup-norm · ∞ with the point x ∈ Rn replaced with the
function f (x). The norm is well defined on C[a, b] since every continuous function
on a closed interval has a maximum. If f ∞ = 0 then f ≡ 0 on [a, b]. Moreover,

sup |c f (x)| = |c| sup | f (x)| ,

and from the triangle inequality we conclude that it is a norm.

Example 1.4.7 Consider again C[a, b] with the 1-norm defined by

 b
f 1= | f (x)| d x.
a

This norm can be viewed as a generalization of the 1-norm

x 1 = |x1 | + |x2 | + · · · + |xn | ,

but with uncountably infinite components. The norm is well defined since any con-
tinuous function on [a, b] is integrable on [a, b]. If f, g ∈ C[a, b] then
 b  b  b
f +g 1 = | f + g| ≤ |f|+ |g| = f 1 + g 1 .
a a a

Also,
1.4 Definition and Examples of Normed Spaces 13
 b  b
cf 1 = |c f (x)| d x = |c| | f (x)| d x. = |c| . f 1 .
a a

Let f 1 = 0. Then
 b
| f (x)| d x = 0,
a

and since f is continuous on [a, b], this cannot occur without f ≡ 0 on [a, b].
Indeed, suppose on the contrary that f = 0. Then there exist  > 0 and an interval
I ⊂ [a, b] such that | f x)| >  on I. Then
 b 
f 1= | f (x)| d x > | f (x)| d x > (I ) > 0,
a I

where (I ) is the length of the interval. But this contradicts the assumption that
f 1 = 0. This proves that f 1 is a norm on C[a, b].

Example 1.4.8 (Space of Polynomials) Consider the space P[a, b] be the space of
polynomials on [a, b]. It is clear that

P[a, b] ⊂ C[a, b].

Define the norm

f ∞ = max | f (x)| .
x∈[a,b]

Then the space (P[a, b], · ∞ ) is a normed space. All axioms of the norm are clearly
satisfied.

Example 1.4.9 (Space of Functions of Bounded Variation) Consider the space

BV [a, b] of all functions of bounded variations on [a, b]. Define

f s = | f (a)| .

It is clear that f satisfies axioms 1, 3, and 4. Indeed, f ≥ 0 for all f ∈ BV [a, b],
and
c f = |c f (a)| = |c| | f (a)| = |c| f .

Moreover, the triangle inequality gives axiom 4. It remains axiom (2) which is obvi-
ously not correct since we may have a nonzero function f with f (a) = 0. Hence
(BV [a, b], · s ) is not a normed space, but rather a seminormed space.
Another candidate of a norm is

f BV = | f (a)| + V ( f ),
14 1 Normed Spaces

where


n
V ( f ) = sup | f (xi ) − f (xi−1 )|
P i=1

is the total variation of f , which is finite since f ∈ BV [a, b]. It is clear that

f BV ≥0

for all
f ∈ BV [a, b].

Moreover, let f BV = 0. Then, V ( f ) = 0, which implies that


n
| f (xi ) − f (xi−1 )| = 0
i=1

for all possible partitions P of [a, b]. Finally, using properties of supremum and
definition of total variation, it is easy to show that

V ( f + g) ≤ V ( f ) + V (g),

hence by the triangle inequality

f +g BV = | f (a) + g(a)| + V ( f + g)
≤ | f (a)| + |g(a)| + V ( f ) + V (g)
= f BV + g BV .

Therefore the space (BV [a, b], · BV ) is a linear normed space.

Example 1.4.10 (Sequence Spaces) We introduce the following spaces of sequences

and endowed with the supremum norm:

(1) The space of all bounded convergent sequences with the supremum norm,
denoted by (c, · ∞ ).
(2) The space of all sequences converging to zero, (c0 , · ∞ ), where

c0 = {(xn ) ⊂ c : lim xn = 0}.

(3) The space of eventually zero sequences, (c00 , · ∞ ), where

c00 = {(xn ) ∈ c0 : x j = 0

for all j ≥ N for some N ∈ N}.

1.5 L p Spaces 15

It is obvious from the definitions that

c00 ⊂ c0 ⊂ c

and that each one of these spaces endowed with · ∞ is a normed space.

1.5 L p Spaces

1.5.1 L p Spaces

We devote this section to one of the most important normed spaces. These are the
spaces of Lebesgue integrable functions which have numerous applications in func-
tional analysis.2 These spaces consist of measurable functions that are Lebesgue
integrable, so prior knowledge of measure theory and Lebesgue integration is impor-
tant to understand the basic theory of the spaces.3 The goal of this section is to remind
the reader about the important concepts and results regarding L p theory that may
help us in this text.
Recall the space L[a, b] is the space consisting of all Lebesgue integrable func-
tions on [a, b], that is, those functions f : [a, b] → R such that
 b
| f (x)| d x < ∞.
a

Define the norm in L[a, b] as follows:

 b
f = | f (x)| d x. (1.5.1)
a

This is a normed space, where it is understood that the elements of the space are
viewed as equivalence classes of functions rather than functions, which is possible
due to the fact that in measure theory, measurable functions that are equal except at
a set of measure zero are considered equal. In this case the norm defined in (1.5.1)
is a norm.
The space L[a, b] can also be generalized to L p [a, b], the space of all functions
such that | f | p is Lebesgue integrable on [a, b] for every f ∈ L p [a, b], where 1 ≤
p < ∞, together with the norm
 b 1/ p
f p = | f (x)| p d x .
a

2 These function spaces were introduced by F. Riesz in 1910 [176].

3 For a detailed and careful treatment of the theory, consult the first volume of this series [51] or
alternatively any graduate text on real analysis.
16 1 Normed Spaces

The convergence in L p is defined as follows.

Definition 1.5.1 (Convergence in p−Norm) Let ( f n ) ∈ L p (). Then f n converges
to
 f in p−norm (or in the mean of order p) if f n − f p → 0, or equivalently,
|
 fn − f
| p −→ 0.
A fundamental result about the convergence in L p spaces is the following.
Proposition 1.5.2 Let f n → f a.e., and f n , f ∈ L p , for 1 ≤ p < ∞. Then f n −→
f in p−norm iff f n p −→ f p a.e.

1.5.2 L ∞ Spaces

The space L ∞ () is defined as the space of all essentially bounded measurable
functions, with the norm

f ∞ = ess sup | f | ,

where the essential supremum is defined as

ess sup( f ) = inf{M : m(x ∈  : | f (x)| > M}) = 0}.

So the function f is bounded on [a, b], except maybe on a set of measure zero. Thus,
(L ∞ (), · ∞ ) is a normed space.
The following well-known inclusion is important and is proved in most texts on
real analysis.
Proposition 1.5.3 Let 1 ≤ r < s < ∞, and let μ() < ∞. Then,

L ∞ () ⊆ L s (E) ⊆ L r ().

Proof Let f ∈ L ∞ (). Then | f | < M a.e. for some M > 0. Choose s ≥ 1. Then

| f |s dμ ≤ M s · μ() < ∞,


so f ∈ L s (). Now, let f ∈ L s () and E ⊆  defined by E = {x ∈  : | f | > 1}.

Then

  
|f| =
r
|f| +
r
| f |r
 Ec
 E

≤ | f |s + μ(E)
E
≤ ∞.
1.5 L p Spaces 17

So f ∈ L r (). 

1.5.3 p and ∞ Spaces

Another important class of spaces that can be derived from L p spaces is the class
of  p spaces. The space  p can be constructed as follows: In the space L p , consider
X = N, the set of natural numbers, and Σ is a σ−algebra of subsets of N, and ν is
the counting measure. Then L p = L p (N, Σ, υ), and f ∈ L p (N) will be of the form
f = ( f 1 , f 2 , . . . , ) such that
⎛ ⎞1/ p
∞
 p
f p =⎝  f j  ⎠ < ∞.
j=1

So we obtain sequences instead of functions, and these sequences converge in p-

norm. Let {cn } be a sequence, and define

f (x) = cn χ[n,n+1] (x).

Then
  1/ p 
f p = cn χ[n,n+1] (x)d x =( cnp )1/ p = cn p .

Hence f p < ∞ if and only if cn p < ∞. So, the  p space for 1 ≤ p < ∞ is
defined as  1/ p
 p = {(xn ), such that xn p = |xn | p < ∞}.

If p = ∞, then the space

∞ = {(xn ) : such that xn ∞ = sup |xn | < ∞}.

n

The spaces ( p , · p) and (∞ , · ∞) are linear normed spaces. It is easy to see that

 p ⊆ ∞

since every xn ∈  p is necessarily bounded. In general, for 1 ≤ p < q < ∞, we have

the following inclusion relations:

1 ⊂  p ⊂ q ⊂⊂ ∞ .
18 1 Normed Spaces

1.5.4 Basic Inequalities

We review some important inequalities:

1. Young’s Inequality: If a, b ≥ 0, and 1 < p, then

ap bq
ab ≤ + .
p q

Proof Since f (x) = − ln x is concave up, the line tln x + (1 − t) ln y must be above
the graph of f between x and y for all t ∈ (0, 1). Therefore, we have

ln(t x + (1 − t)y) ≥ t ln x + (1 − t) ln y.

Now substitute with x = a p and y = bq and take the exponential of both sides of the
equation. 
2. Holder’s Inequality: Let f ∈ L p () and g ∈ L q (), for 1 ≤ p, q ≤ ∞, p and q
are conjugates in the sense that
1 1
+ = 1.
p q

Then f g ∈ L 1 (), i.e., Lebesgue integrable, and


| f g| ≤ f p . g q .


Proof For the case p = 1 and q = ∞, we have


| f g| ≤ g ∞ f 1 .


Consider p > 1. Let

f g
F= &G = .
f p g q

Clearly F p = G q = 1. Using Young’s inequality and then integrating over ,

we obtain
  
Fp Gq 1 1
|F G| ≤ + = + = 1.
  p  q p q

Writing F and G in terms of f and g. 

3. Minkowski’s Inequality: Let f, g ∈ L p for 1 ≤ p ≤ ∞. Then

f +g p ≤ f p + g p .
1.5 L p Spaces 19

Proof The inequality is clear for p = 1, and the case p = ∞ follows from the
properties of the supremum. Let 1 < p < ∞. Then

| f + g| p = | f + g| p−1 | f + g|1 ≤ | f | | f + g| p−1 + |g| | f + g| p−1 .

Since p = ( p − 1)q, we have ( f + g) p−1 ∈ L q . Now, apply Holder Inequality to

 1/q
both terms in the RHS of the inequality above, then taking | f + g| p as a
common factor gives
  1/q  1/ p  1/ p 
| f + g| ≤
p
| f + g| p
|f| p
+ |g| p
.

This can be written as

f +g p
p ≤ f +g p/q
p ( f p + g p ).

Now we divide both sides by f + g p/q

p . 
The inequalities also hold for the  p spaces. Strictly speaking, we have
(1) Holder Inequality for Sequences: If xn ∈  p and yn ∈ q then xn · yn ∈ 1 and

xn yn 1 ≤ xn p · yn q

for p > 1.
(2) Minkowski’s Inequality for Sequences: If xn , yn ∈  p then

xn + yn p ≤ xn p + yn p

for p ≥ 1.

1.5.5 Measure Theory and Integration: Quick Review

Now, we state (without proof)4 some fundamental results from Lebesgue’s theory of
integration.
Definition 1.5.4 (Simple Function) A function n ϕ is called simple  function if it is
defined on a set A, and is written as ϕ(x) = i=1 ci χ Ai (x), where Ai = A, and
χ Ai (x) is the characteristic function (or indicator function), and is defined as

1 x ∈ Ai
χ Ai (x) = .
0 x∈ / Ai

4 The interested reader is referred to Ref. [51], or any other graduate textbook on real analysis.
20 1 Normed Spaces

Definition 1.5.5 (Measurable Function) Let (X, Σ) be a measurable space and  ∈

Σ. A function f :  ⊆ X −→ R is called measurable if given any open set O ⊆ R,
f −1 (O) is a measurable set.
Theorem 1.5.6 (Simple Approximation Theorem) Let f be a measurable function
on a set E ∈ Σ in a measurable space (X, Σ). Then there exists a sequence of simple
functions {ϕn } such that |ϕn | ≤ | f | for all n and ϕn −→ f p.w. on E. If f is bounded
on E then ϕn −→ f uniformly.
Proposition 1.5.7 (Approximation of L ∞ ) The class of simple functions in L ∞ ()
are dense in L ∞ (), for μ() < ∞. The class of step functions in L ∞ () are dense
in L ∞ (), for μ() < ∞.
Theorem 1.5.8 (Egoroff Theorem) Let { f n } be a sequence of measurable functions
and f n → f a.e. on a set E, μ(E) < ∞ in a measure space (X, Σ, μ). Then for
every  > 0, there exists a set A, μ(A) <  such that f n −→ f uniformly on E \ A.
Definition 1.5.9 (Integral on Measure Spaces) Let (X, Σ, μ) be a measure space.
Then
(1) Integral of Nonnegative Function: Let f : E −→ [0, ∞] be a nonnegative mea-
surable function defined on a measurable set E ∈ Σ in a measure space
(X, Σ, μ). Then the integral of f over E is defined as
 
f = sup ϕ,
E E

where the supremum is taken over all possible simple functions ϕ with 0 ≤ ϕ ≤
f. The function f is said to be integrable over E if

f < ∞.
E

(2) Lebesgue Integrable Function: Let f be a measurable function defined on a

measurable set E. If f + and f − both integrable over E, then f is integrable.
Theorem 1.5.10 (Fatou’s Lemma) Let { f n } be a sequence of measurable functions
in a measure space (X, Σ, μ), f n ≥ 0, and f n −→ f a.e. on a set E ∈ Σ. Then
 
f dμ ≤ lim f n dμ.
E E

Theorem 1.5.11 (Monotone Convergence Theorem) Let { f n } be a sequence of

nonnegative and increasing measurable functions on E ∈ Σ in a measure space
(X, Σ, μ). If lim f n = f a.e., then
 
lim f n dμ = f dμ.
E E
1.6 Topology of Normed Spaces 21

Theorem 1.5.12 (Dominated Convergence Theorem) Let { f n } be a sequence of

measurable functions on E ∈ Σ in a measure space (X, Σ, μ). If f n → f p.w. a.e.,
and is dominated by some μ−integrable function g over E, that is, | f n | ≤ g for all
n and for almost all x in E, then f is μ−integrable and
 
lim f n dμ = f dμ.
E

Theorem 1.5.13 (Lebesgue Differentiation Theorem) Let f : (a, b) → R be mono-

tonic function. Then f is differentiable almost everywhere on (a, b).

Theorem 1.5.14 (Fundamental Theorem of Calculus x for Lebesgue Integration)

(a) Let f be integrable on [a, b] and F(x) = a f (t)dt. Then F  (x) = f (x) a.e.
(b) Let F ∈ AC[a, b], and F  (x) = f (x) a.e. on [a, b]. Then
 x
F(x) = f (t)dt + F(a),
a

and  b
f (x)d x = F(b) − F(a).
a

(c) Corollary: A function F is absolutely continuous on [a, b] if and only if it is

the indefinite integral of its derivative.

1.6 Topology of Normed Spaces

1.6.1 Balls and Sphere

In general normed spaces, the starting key in determining the topology of the space
is the norm, and the topology generated by the norm is called the norm topology. The
norm determines the topology of the space, and the shape of the unit ball determines
the geometric properties of the space. Determining the norm induced on the unit ball
is the cornerstone that builds the entire structure of the space, so the role of the unit
ball as the building block will continue to work in general normed spaces of infinite
dimensions.
The intervals in R are generalized in Rn to what are known as “balls”. The unit
ball in Rn , which is denoted by B1 (0), is a ball of radius one and centered at the
origin, and is given by

B1 (0) = {(x1 , . . . , xn ) ∈ Rn : x12 + x22 + · · · + xn2 < 1}.

The closure of this ball is the closed unit ball consisting of all points satisfying
22 1 Normed Spaces

B1 (0) = {(x1 , . . . , xn ) ∈ Rn : x12 + x22 + · · · + xn2 ≤ 1}.

The boundary of the closed unit ball is known as the unit sphere, and is given by

S1 (0) = {(x1 , . . . , xn ) ∈ Rn : x12 + x22 + · · · + xn2 = 1}.

The class of L p spaces is a fundamental class of spaces that is extremely helpful

in applications and is an indispensable tool in the theory and applications of partial
differential equations. It is well known that the shape of the sphere is completely
determined by the norm of the space, and any change to the norm in the space will
result in changing the shape of these units. For example, consider X = R2 , then
applying the norm · 1 gives the following equation for the unit sphere:

|x1 | + |x2 | = 1.

This gives the shape of diamond for S1 . Applying the norm · 2

x12 + x22 = 1

gives the shape of unit circle for S1 . Applying the norm · ∞ gives the shape of a
square. The norm · 2 gives the smoothest shape for the sphere, and as p exceeds
2, the circle gets flattened until it reaches the shape of a square when p = ∞. It is,
therefore, obvious that

(S1 , · 1 ) ⊆ (S1 , · 2 ) ⊆ (S1 , · ∞ ).

Definition 1.6.1 (Open Ball) Let X be a normed space with a norm · . Then the
open ball of X of radius r and centered at x0 is given by

Br (x0 ) = {x ∈ X : x − x0 < r }.

The closure of that ball is the closed ball given by

Br (x0 ) = {x ∈ X : x − x0 ≤ r }.

When r = 1 and x0 = 0, then the ball is called “unit ball”. The unit sphere is given
by
S1 (x0 ) = {x ∈ X : x − x0 = 1}.

It suffices to study the structure of the unit balls and unit spheres in many cases
since all other balls and spheres can be obtained using shifting and scaling operations.
In fact, one advantage, among many others, of unit spheres is that one can always
reflect elements of the space to elements on the unit sphere by normalizing the
element. It should be noted that the term ball here does not reflect the actual shape
of the ball. The only space where we can use the terms “ball” and “sphere” in their
1.6 Topology of Normed Spaces 23

actual meanings is R3 . In n > 3, these building units are called “hyperballs” and
“hyperspheres”, or n-balls and n-spheres. We, however, can simply call them balls
and spheres if the context is clear from confusion.

1.6.2 Open and Closed Sets

In R, open and closed sets are defined based on intervals. Similarly, the balls play
the same role in normed spaces.
Definition 1.6.2 (Open Set) Let X be a normed space and O ⊆ X. Then, O is called
open set if for every x ∈ O, there exists r > 0 such that Br (x) ⊂ O. The set F ⊆ X
is called closed if F c (the complement of F) is open.
It is well known in elementary analysis that the arbitrary union of open sets is
open, and the finite intersection of open sets is open. The same property holds for
general normed space. The following fact provides a sequential characterization of
closed sets. The proof is straightforward and has been discussed in undergraduate
analysis courses, so it will simply be presented as a definition.
Definition 1.6.3 (Closed Set) A set F is called closed set in a normed space X if
for every sequence {xn } ⊆ F and xn −→ x, we have x ∈ F.
One immediate consequence is that the real line R is closed. One can extend this
claim by induction to say that the vector space Rn is closed, and it is well known
from linear algebra that vector spaces of the same dimension are isomorphic, which
means there exists a linear bijection mapping between them. We infer the following:

Proposition 1.6.4 Every finite-dimensional space is closed.

We will give rigorous proof of this proposition later.

1.6.3 Quotient Spaces

The concept of closedness helps us in many applications, but we will only mention
two examples: quotient normed spaces and distance to sets. For the first application,
we would like to study the quotient space of a normed space. Let M be a subspace
of a normed space X . Then, we define the quotient X/M as

X/M = {[x] = x + M : x ∈ X } = {x + z : z ∈ M},

where the elements of this quotient are cosets [x] = x + M, formed by equivalence
classes. Here, we define the equivalence relation ∼ as x1 ∼ x2 iff x1 − x2 ∈ M. It’s
clear that
24 1 Normed Spaces

[x] + [y] = [x + y]

and

[cx] = c[x].

Hence, the quotient X/M forms a subspace of X. The space (X \ M, · Q ) is called:

“quotient space”. The subspace M is the zero coset of this quotient space. Let · X
be the norm in X. Define the following.

[x] Q = inf{ x + z : z ∈ M}. (1.6.1)

This is called the quotient seminorm. The reason is that if

[x] Q = x+M Q = 0,

then x may not be in M, but the infimum will take a zero value. We will see later
that the space (X \ M, · Q ) is normed if and only if the set M is closed. We need
to define the closedness of a set in a general normed space. If M is closed, then

x+M Q = 0 iff inf { x + z : z ∈ M} = 0,

z

and since the closed set M contains its boundary and limits points, this means that
if x0 ∈
/ M then x0 ∈ M c , which is an open set, so we can find a ball U = B (x0 )
around x0 such that x0 ∈ U ⊂ M c and so U ∩ M = Ø. This implies that

inf { x0 + z : z ∈ M} > 0.
z

Therefore, if

inf { x0 + z : z ∈ M} = 0
z

then x0 ∈ M. The other conditions can easily be verified. Let c ∈ R and x ∈ X , then
it is easy to show that · Q is a norm. Indeed,

c[x] = inf{ cx + z = inf{|c| x + z/c } = |c| inf{ x + z } = |c| [x] Q .

For the triangle condition,

[x] + [y] Q = x + z1 + y + z2 Q ≤ x − z1 Q + y − z2 Q .

Taking the infimum over z 1 and z 2 gives

[x] + [y] Q ≤ [x] Q + [y] Q .

1.6 Topology of Normed Spaces 25

We therefore have the following result.

Proposition 1.6.5 Let X be a normed space. Then the quotient space (X \ M, · Q)
is a normed space if and only if M is closed in X.

1.6.4 Distance to Closed Sets

The second application is the distance to sets. Recall that the distance between a
point x and a set Y is given by

dist(x, Y ) = inf x − y .
y∈Y

Note also that if Y is a subspace of a normed space, then it is a vector space. So we

have

inf x − y = inf x − (−y) = inf x + y .

y∈Y y∈Y y∈Y

If Y is closed, then the distance is strictly greater than zero because Y contains its
boundary. Indeed, if dist(x, Y ) = 0, then there exists a sequence yn ∈ Y such that
yn −→ x. But this means x ∈ Y = Y since Y is closed, which is a contradiction.
Hence, we have
Proposition 1.6.6 If Y is closed subset of a normed space X and x ∈
/ Y , then
dist(x, Y ) > 0.

1.6.5 Riesz’s Lemma

The preceding proposition predicts that the distance between a point and a closed
subset of a normed space is strictly greater than zero. The following important result
by Riesz in 1918 gives a good lower bound when the set is a subspace.
Lemma 1.6.7 (Riesz’s Lemma) Let Y be a closed subspace of a normed space
X. Then, for every δ, 0 < δ < 1, there exists x0 ∈ X such that x0 = 1 and
dist(x0 , Y ) ≥ δ.

Proof Since Y is closed, dist(x, Y ) = d > 0. Let 0 < δ < 1. Then, by definition of
infimum, there exists y0 ∈ Y such that
d
x − y0 ≤ . (1.6.2)
δ
Note that
26 1 Normed Spaces

y0 + x − y0 y ∈ Y. (1.6.3)

Let
x − y0
x0 = .
x − y0

Then, by (1.6.2) and (1.6.3) and the fact that x0 = 1, we obtain

x − y0 − x − y0 y d
x0 − y = ≥ ≥ r. 
x − y0 x − y0

1.6.6 Compactness

Another fundamental set is the compact set. Recall a set K is compact if every open
cover of K has a finite subcover. By an open cover, we mean a collection of open
sets. The following definition plays a dominant role in analysis.
Definition 1.6.8 (Compact Set) A set K is said to be compact if every sequence in
K has a subsequence that converges in K .
The characterization in the previous definition is known as the sequentially com-
pact definition of compact sets. In an introductory course of analysis, it was shown
that a set is compact if and only it is sequentially compact in a vector space.5 We have
used this characterization to define compact sets, which seems important and inter-
esting since it connects the compactness property with the completeness property.6
These concepts are cornerstones of analysis and can be used in establishing many
important results. We will state two great theorems related to this notion that have
numerous applications and consequences. The first is considered one of the most
important theorems in analysis, and its proof is available in most undergraduate
textbooks in analysis; thus, it will not be mentioned here.
Theorem 1.6.9 (Heine–Borel Theorem) In a finite-dimensional space, a set is com-
pact if and only if it is closed and bounded.
The second is a fundamental result that holds in any topological space, not nec-
essarily normed.

5 The reader has already studied compactness in an introductory course of analysis, so we shall not
repeat the discussion here.
6 The completeness property will be discussed in detail in Sect. 4.1.
1.6 Topology of Normed Spaces 27

Theorem 1.6.10 A closed subset of a compact set is compact.

The proof of Theorem 1.6.10 can be easily obtained by using definitions of closed-
ness and sequential compactness (Definition 1.6.8), and the reader should be able to
obtain it without any difficulties. The difference between the two theorems is that
Theorem 1.6.10 holds for all normed spaces, finite dimensional or infinite dimen-
sional, whereas Theorem 1.6.9 holds only for finite-dimensional spaces. We always
have the direction: compact ⇒closed + bounded in all normed spaces, but the con-
verse holds only in finite-dimensional spaces. The following result is fundamental
in the topology of general normed spaces.
Theorem 1.6.11 Let X be a normed space. Then, X is finite dimensional if and only
if its closed unit ball is compact.

Proof The (⇒) direction is immediate by Heine–Borel Theorem. Conversely, let

dim(X ) = ∞ and assume the closed unit ball is compact. Let x1 ∈ S X , i.e., x1 = 1.
Define the space

Y1 = span{x1 }.

By Riesz Lemma, there exists x2 , x2 = 1 such that

1
dist(Y1 , x2 ) ≥ .
2
Let

Y2 = span{Y1 , x2 }.

Then,
1
x1 − x2 ≥ .
2
Since X has infinite dimensional, we can continue inductively to construct a sequence
{xn } such that

Yn = span{x1 , x2 , . . . , xn−1 }

and
1
dist(Yn , xn+1 ) ≥ .
2
Hence
1
xn+1 − xi ≥
2
28 1 Normed Spaces

for all i ≤ n. So we conclude that the sequence {xn } does not have any convergent
subsequence, which implies the unit sphere S is not compact. But the unit sphere is
closed in the closed unit ball, the result then follows from Theorem 1.6.10. 
The previous result can be used to identify infinite-dimensional spaces by their non-
compact closed unit balls.
Example 1.6.12 (Standard Sequence) In ∞ and 2 , consider the sequence {δn } ∈ S,
defined as

δi = (0, 0, . . . , 1 , 0, . . .), i = 1, 2, . . . .
ithcoordinate

That is, δi = 1 on the ith coordinate and zero elsewhere. This is called the
standard sequence, and it could also be represented as

δn = (δnm )∞
m=1 = (0, . . . , 1, 0, . . .), n ∈ N},
nth

where 
1 n=m
δnm =
0 n = m

is the Kronecker delta function. It is clear that

    √
δi − δ j  = 1 & δi − δ j  = 2.
∞ 2

In either case, we cannot find a convergent subsequence. Hence, 2 and ∞ are

infinite-dimensional spaces. One can easily extend the result to all  p spaces.

1.6.7 Separability

Definition 1.6.13 (Separable Space) A linear space X is said to be separable if there

exists a countable subset that is dense in X .
The simplest and most famous example is the set of rational numbers, which is
dense in R. So the space Rn is separable for any finite n. We will use this information
to build up some dense subsets that are countable in their spaces.
Theorem 1.6.14 The space L p [a, b] is separable for all 1 ≤ p < ∞.
Proof According to Proposition 1.5.7, the collection S[a, b] of step functions is
dense in L p [a, b]. We form the subcollection SQ [a, b] of all step functions tak-
ing rational values on [a, b] with a partition P = {a = x0 , x1 , . . . , xn = b} which
consists of rational numbers, and each function in SQ [a, b] takes a constant value
between xi and xi+1 . Then the collection SQ [a, b] is clearly dense in S[a, b], which
Another random document with
no related content on Scribd:
 Chapter Eight: A Fresh Start
It is wonderful what an effect a good night’s sleep and a bright
morning will have on the mind of a healthy man. French had gone to
bed tired and worried about this case. He woke cheery and
optimistic, philosophic as to his reverses, and hopeful for the future.
On such a morning, indeed, it was impossible that anyone could
be despondent. Though October had begun, the sun shone with a
thin brilliancy reminiscent of early summer. The air, floating up gently
from the garden in the rear of the hotel, was surprisingly warm and
aromatic for the time of year. Birds were singing in the trees and
there was a faint hum of insects from below. As he looked out of his
window French felt that life was good and that to squander it in sleep
was little better than a sin.
He breakfasted at his leisure, then lighting his pipe, he sauntered
out into the little town to take what he called “a turn” before settling
down to the serious work of the day.
Though his conclusions of the previous evening still seemed
incontrovertible, he was surprised to find that his sense of
disappointment had vanished. At first he thought this was due simply
to his night’s rest, then gradually he realised the reason.
In his heart of hearts he distrusted these conclusions. In spite of
the difficulties involved, he was not satisfied that the Berlyn-Pyke
affair should be eliminated from the case.
The murderer had shown himself an extremely ingenious man.
Could it not be that these seeming impossibilities were really
intentionally designed to throw investigating detectives off the scent?
French reconsidered the strength of the coincidences otherwise
involved.
A disappearance at a certain time and place was required to
account for the body in the crate. At that very time and place, and
there only, a disappearance was known to have occurred. French
could not bring himself to dismiss the possibility of a connection
between the two facts.
He decided that he had not exhausted the possibilities. He must
learn more about Berlyn and Pyke.
For preliminary enquiries Sergeant Daw seemed the most
hopeful source of information, and he lost no time in walking down to
the police station and asking his help.
“I want to know who everybody is, Sergeant. You know the local
people and you might tell me something which would give me the
hint I am looking for.”
The sergeant did not think this likely, but he was willing to do
anything to oblige.
“Very good. Then I’ll ask questions. First of all, will you tell me
what you can about Mr. Berlyn?”
Daw put on his best police-court manner and proceeded to
deliver himself.
“Mr. Berlyn was junior partner at the works. I understand that
some eight or nine years ago he and Colonel Domlio bought up
nearly the whole of the stock between them. Mr. Berlyn dealt with the
commercial side and attended the office every day as if he was an
official, but the colonel looked on the business as a hobby. He acted
as a sort of consulting engineer and only went to the works when it
pleased him. I believe there are other directors, but in practice they
don’t amount to anything.”
“Was Mr. Berlyn liked?”
“As a matter of fact, sir, he wasn’t altogether popular among the
work people. From what I’ve heard, he wanted too much and he
wouldn’t make allowances for people making mistakes. It was get on
or get out with him, and you know yourself, Mr. French, that if that’s
pushed too far it doesn’t always work. But he was straight enough
and what he said he stuck to.”
“A man like that would make enemies. Do you know of anyone he
was on bad terms with?”
“No, sir. No one.”
“He hadn’t his knife in Mr. Pyke, for instance?”
“Not to my knowledge.”
“Was Mr. Berlyn married?”
 “Yes, four or five years ago. Very pleasant lady, Mrs. Berlyn.”
“Any children?”
“No, sir.”
“Where did they live?”
“Out along the Buckland road about ten minutes’ walk from the
works. Place called Soller. They say it’s the name of some foreign
town where he’d met the lady and popped the question, but of
course I don’t know anything about that.”
“Then he was a traveller, Mr. Berlyn?”
“Yes; used to go away to France and such places when he had
holidays.”
“Wise man,” French commented. “And how did the match turn
out?”
For the first time the sergeant hesitated.
“There, Mr. French, you have me. I couldn’t really tell you. From
all accounts they got on as well as most people whose tastes differ.
He was quiet and liked sitting at home in the evenings, and she
wanted a bit of life. There’s not much of what you might call gaiety in
this town, as you may guess, but whatever there was Mrs. Berlyn
was in the centre of it. At first he used to go out with her to Torquay
and so on, but he gradually gave that up and she had to find some
one else to go with or stay at home.”
“And she found some one?”
“Any number. The gentlemen up at the works, mostly. They were
all glad to go with her! Colonel Domlio had been taking her about
lately—I mean before Mr. Berlyn’s death—and before that it was Mr.
Pyke and sometimes Mr. Cowls, the engineer. She was friends, too,
with Dr. and Mrs. Lancaster, and I’ve often seen her out with people
called Tucker that live close by.”
All this seemed suggestive to French and his facile brain was
already building up tentative theories.
“Was there ever any suggestion of anything between Mrs. Berlyn
and any of those men?”
“There was a bit of talk at one time, but I don’t believe there was
anything in it.”
“But there was talk. Just tell me what was said.”
 “She was talked about with Mr. Pyke. They certainly saw a deal
of each other at one time. He was constantly at the house and they
went out motoring together. She was a top-hole driver.”
“You say they saw a deal of each other at one time. Did that not
continue?”
“It was supposed to come to an end about four months before the
tragedy. But that’s only local gossip and I can’t vouch for it. All the
same, I don’t remember seeing them motoring since, except once
when Mr. Pyke’s cousin came for three or four days.”
“And you have no idea what happened?”
“No, sir. Some said the lady heard of the talk and thought she
had gone far enough; others, that Mr. Berlyn got wise to it; and
others again, that they got tired of each other. I don’t know. Whatever
happened, it was all quite amicable, for I’ve seen them together
different times since.”
“And was that the only time there was talk?”
“After that there was talk about her and Colonel Domlio. But you
know, Mr. French, in a place this size they’re hard up for something
to talk about. I don’t believe there was anything in either story.”
“Tell me what was said anyway.”
“Well, that she used to go out to see him in the afternoons. The
colonel was believed to be very fond of her, but she was only
supposed to be amusing herself with him.”
“You say this took place recently?”
“That was the rumour.”
French shrugged.
“Safety in numbers, Sergeant. I agree it doesn’t sound hopeful.
Did Mr. Berlyn seem upset about it?”
“Not that I ever heard of.”
“No good for us, Sergeant. Now about these others. Mr. Pyke
was not married, was he?”
“No, Mr. Pyke was not married, nor were Mr. Cowls nor Mr.
Samuel nor Mr. Leacock, other young men about the works with
whom Mrs. Berlyn had seemed on good terms. Mr. Pyke had been
with the firm for several years and was said to be highly thought of.
He was pleasant-mannered and jolly and a general favourite. He
lodged in the town; in fact, his rooms were nearly opposite the hotel.”
 “Is Mrs. Berlyn still here?”
“She left three or four days ago. There was an auction and she
waited till it was over. I heard she had gone to London.”
“Will she be well off?”
“I believe so. They say Mr. Berlyn left her everything.”
“You spoke of Mr. Pyke’s cousin. Who is he?”
“A Mr. Jefferson Pyke, a farmer in the Argentine. Rather like the
late Mr. Pyke, that’s Mr. Stanley, in appearance, but a bit taller and
broader. He was on a visit to England and was down here twice.
First he came and stayed with Mr. Stanley for three or four days
about a couple of months before the tragedy; that was when Mrs.
Berlyn took them both out motoring. I wasn’t speaking to him then,
but I saw him with Mr. and Mrs. Berlyn and Mr. Stanley in the car.
Then the morning after the tragedy Mrs. Berlyn gave me his London
address and told me to wire for him. I did so and he came down that
evening. He stayed for three or four days in Torquay and came over
to make enquiries and to look after Mr. Stanley’s affairs. A very nice
gentleman I found him, and a good business man, too.”
French noted the London address and then asked what servants
the Berlyns had.
“They had three—two house servants and the gardener.”
“Any of them available?”
“One of the girls, Lizzie Johnston, lives not far away. The others
were strangers.”
French continued his inquisitions in his slow, painstaking way,
making notes about everyone connected with the Berlyns and Pyke.
But he learned nothing that confirmed his suspicions or suggested a
line of research. It was true that in Mrs. Berlyn he had glimpsed a
possible source of trouble between her husband and Pyke. All the
essentials of a triangle drama were there—except the drama itself.
Mrs. Berlyn might easily have hated her husband and loved one of
these other men, but, unfortunately for theorising detectives, if not for
moralists, there was no evidence that she had done so. However, it
was a suggestive idea and one which could not be lost sight of.
As these thoughts passed through French’s mind a further
consideration struck him, a consideration which he saw might not
only prove a fifth test of the case he was trying to make, but which, if
so, would undoubtedly be the most conclusive of them all. He turned
once more to Daw.
“There’s a point which is worrying me rather, Sergeant,” he
declared. “Suppose one of these two men murdered the other on
that night. Now why would the murderer go to the trouble of getting
the body into the works and sending it off in the crate? Could he not
simply have thrown it into one of these mires?”
Daw nodded.
“I thought of that when you suggested your idea, but I don’t
believe there’s anything in it. It wouldn’t be so easy as it sounds. In
fact, I couldn’t see any way it could be done.”
“I’m glad to hear you say so, Sergeant. Explain, please.”
“Well, if you go into one of those places and begin to sink you
throw yourself on your back. As long as your weight is on the small
area of your feet you go down, but if you increase your area by lying
on your back you reduce the weight per unit of area and you float—
because it really is a kind of floating. You follow me, sir?”
“Quite. Go ahead.”
“Now if you walk to a soft place carrying a body you have
doubled the weight on your feet. You will go down quickly. But the
body won’t go down. A man who tried to get rid of his victim that way
would fail, and lose his own life into the bargain.”
“That sounds conclusive. But I didn’t know you could save
yourself by throwing yourself down. If that is so wouldn’t Berlyn and
Pyke have escaped that way? Why did you then accept the idea that
they had been lost?”
“There were two reasons. First there was nothing to make me
doubt it, such as knowing about the crate, and secondly, though the
accident was not exactly likely, it was possible. This is the way I
figured it out. Suppose one of these mists had come on. They do
come on unexpectedly. One of the men gets into a soft place. Mists
are confusing, and in trying to get out, he mistakes his position and
flounders in further. That’s all perfectly possible. Then he calls to the
other one, and in going to the first one’s help the other gets in also—
both too far to get out again.”
“But you said it was a clear night?”
 “So it was when I got there. But three or four hours earlier it might
have been thick.”
“Now, Sergeant, there’s another thing. Could the murderer not
have used some sort of apparatus, a ladder or plank to lay on the
soft ground, over which he could have carried the body and escaped
himself? Same as you do on ice.”
“I thought of that too, but I don’t believe it would be possible. A
ladder wouldn’t do at all. With its sharp edges it would go down
under the weight. And I don’t think a man could handle a big enough
plank. It would have to be pretty wide to support the weight of two
men and it would have to be long to get beyond the edge of the mire.
You see, Mr. French, it’s only well out into the big mires that a flat
body will sink. Near the edges it would have to be kept upright with
the weight on the feet. That couldn’t be done off the end of a plank
which would itself be sinking; in fact, I don’t think it could be done at
all.”
French nodded. This was certainly very satisfactory.
“Besides, sir,” Daw went on, “think of a plank laid as you’ve
suggested and with the end of it partly sunk. It’ll not be easy to pull
out, particularly when the ground you’re pulling from is not very firm.
You won’t do it without leaving pretty deep footmarks, and the plank
will leave a sort of trough where it was slid out. If that had been done
that night the marks would have been there next morning, and if they
had been there I should have seen them. No, sir, I think you may
give up that idea. You couldn’t get rid of a body by hiding it in a
mire.”
“I’m uncommonly glad to hear you say so,” French repeated. “If
the thing had been possible it would have knocked my case into a
cocked hat. Well, Sergeant, I’ve bothered you enough for one
morning. I’ll go along and have a word with Mrs. Berlyn’s maid.”
Lizzie Johnston lived with her mother in a little cottage on the hill
behind the railway station. She proved to be a dark, good-looking girl
of about five and twenty, and when French talked with her he soon
discovered she was observant and intelligent also.
She had lived, she said, with Mrs. Berlyn for about two years, and
French, in his skilful, pleasant way drew her out on the subject of the
household. It consisted of the two Berlyns, herself, and cook, unless
Peter Swann, the gardener, might be included.
Mr. Berlyn she had not greatly liked. He was quiet in the house,
but was rather exacting. He was not socially inclined and preferred
an evening’s reading over the fire to any dinner party or dance. He
had been civil enough to her, though she had really come very little
in contact with him.
About Mrs. Berlyn the girl was not enthusiastic, either, though she
said nothing directly against her. Mrs. Berlyn, it appeared, was also
hard to please, and no matter what was done for her, she always
wanted something more. She was never content to be alone and
was continually running over to Torquay to amusements. After their
marriage Mr. Berlyn had gone with her, but he had gradually given
up doing so and had allowed her to find some other escort. This she
had had no difficulty in doing, and Mr. Pyke, Mr. Cowls and others
were constantly in attendance.
No, the girl did not think there had been anything between Mrs.
Berlyn and any of these men, though for a time Mr. Pyke’s attentions
had been rather pronounced. But some four months before the
tragedy they appeared to have had a disagreement, for his visits had
suddenly fallen off. But it could not have been very serious, for he
still had occasionally come to dinner and to play bridge. She
remembered one time in particular when Mr. Pyke had brought a
relative; she heard it was a cousin. There were just the four, the two
Pykes and the two Berlyns, and they all seemed very friendly. But
there was a coolness all the same, and since it had developed,
Colonel Domlio had to some extent taken Mr. Pyke’s place.
About the Berlyns’ history she could not tell much. Mr. Berlyn had
lived in the town for several years before his marriage. He seemed to
have plenty of money. He had bought the house on the Buckland
road just before the wedding and had had it done up from top to
bottom. It was not a large house, but beautifully fitted up. At the
same time he had bought the car. Peter Swann, the gardener,
washed the car, but he did not drive it. Both Mr. and Mrs. Berlyn
were expert drivers and good mechanics. Mrs. Berlyn also used her
push bicycle a good deal.
 French then came to the evening of the tragedy. On that evening
dinner had been early to allow Mr. Berlyn to get away in the car at
seven o’clock. It had been her, Lizzie Johnston’s, evening out, but
Mrs. Berlyn had told her she would have to take the next evening
instead, as some friends were coming in and she would be wanted
to bring up supper. About eight o’clock Mr. Fogden, Mr. Cowls, a Dr.
and Mrs. Lancaster, and three or four other people had arrived. She
had brought them up coffee and sandwiches about half past ten.
They had left about eleven. She had got to bed almost at once, and
a few minutes later she had heard Mrs. Berlyn go up to her room.
The next thing she remembered was being wakened in the
middle of the night by Mrs. Berlyn. The lady was partly dressed and
seemed agitated. “Lizzie,” she had said, “it’s nearly three o’clock and
there’s no sign of Mr. Berlyn. I’m frightened. I’ve just been out to the
garage to see if the car has come back, but it’s not there. What do
you think can be wrong?”
They hurriedly discussed the matter. Mr. Berlyn was the last man
to alter his plans, and both were afraid of an accident on that
dangerous Tavistock road.
In the end they decided that Mrs. Berlyn should knock up
Sergeant Daw, who lived near. This she did, while Lizzie dressed.
Presently Mrs. Berlyn came back to say that the sergeant was going
out to investigate. They had some tea and lay down without taking
off their clothes. In the early morning a policeman brought the news
of the tragedy.
Mrs. Berlyn was terribly upset. But she grew calmer in time, and
the arrangements for the auction and for her removal to London
taking her out of herself, in a week she was almost normal.
She had been very nice to Lizzie at the last, giving her an
excellent testimonial and an extra month’s wages.
French thanked the girl for her information and rose as if to take
his leave.
“I suppose Mrs. Berlyn was something of a needlewoman?” he
said, carelessly. “Some one told me she made her own dresses.”
Lizzie laughed contemptuously.
“Made her dresses, did she?” she repeated. “I don’t think. She
didn’t hardly know how to wear a thimble, she didn’t. She wouldn’t
have sat down to a job of sewing, not for no person on earth she
wouldn’t.”
“Then who did the household mending?”
“Yours truly. Anything that was done I had to do.”
“But not the clothes, surely? Who darned Mr. Berlyn’s socks, for
instance?”
“Yours truly. I tell you Mrs. Berlyn wouldn’t have touched a sock
or a bit of wool not to save her life.”
This was a piece of unexpected luck. French turned away.
“You are a good girl,” he declared. “Would half a sovereign be of
any use to you?”
Miss Johnston left him in no doubt on the point.
“Very well,” he went on. “You come down to the hotel after dinner
to-night and ask for me. I want you to mend some clothes and socks
for me. Or rather,” he paused, “I have to come up in this direction
after lunch to-day in any case, and I’ll bring them.”
No object in advertising the lines on which he was working, he
thought. The less that was known of his researches, the more hope
there was of their proving fruitful.
A couple of hours later he returned with a small suitcase.
“Here are the clothes,” he said. “I wish you’d see what they want,
so that I’ll know when I’m likely to get them.”
He laid four pairs of socks on the table—three brown pairs of his
own and the grey pair found in the crate. The girl looked them over
one by one. French watched her in silence. He was anxious, if
possible, to give her no lead.
“There isn’t much wrong with these,” she said, presently. “They
don’t want no darning.”
“Oh, but they have been badly mended. You see these grey ones
have been done with a different-coloured wool. I thought perhaps
you could put that right.”
Miss Johnston laughed scornfully.
“You’re mighty particular, mister, if that darning ain’t good enough
for you. I’d just like to know what’s wrong with it.”
“You think it’s all right?” French returned. “If so, I’m satisfied. But
what about these underclothes?”
 The girl examined the clothes. They were almost new and neatly
folded, just as they had come back from the laundry, so that her
contemptuous reply was not inexcusable. At all events, it was
evident that no suspicion that they were other than her visitor’s had
crossed her mind.
French, with his half formed theory of Berlyn’s guilt, would have
been surprised if she had answered otherwise. The test, however,
had been necessary, and he felt he had not lost time. Mollifying her
with a tip, he returned to the hotel.
 Chapter Nine: A Step Forward
French believed that he had obtained all the available information
about the Berlyns from his interview with the sergeant and Lizzie
Johnston. Pyke was the next name on his list and he now crossed
East Street to the house in which the travelling representative had
lodged. The door was opened by a bright-eyed, bustling little woman
at sight of whom French’s emotional apparatus registered
satisfaction. He knew the type. The woman was a talker.
But when for the best part of an hour he had listened to her,
satisfaction was no longer the word with which to express his state of
mind. He had no difficulty in getting her to talk. His trouble was to
direct the flood of her conversation along the channel in which he
wished it to flow.
He began by explaining that he was staying at the hotel, but that
as he liked the district and might want to remain for some time, he
was looking about for rooms. He had heard she had some to let.
Was this so, and if it was, could he see them?
It was so and he could see them. She had had a lodger, a very
nice gentleman and a very good payer, but she had lost him recently.
Mr. ——? She had heard his name: was it not Mr. French? Mr.
French must have heard about his dreadful death? His name was
Mr. Pyke. Had Mr. French not heard?
Mr. French had heard something about it. It seemed a very sad
affair.
It was a very sad affair. Mr. Pyke had gone out that evening as
well as Mr. French or herself, and he had never come back, had
never been seen again. Terrible, wasn’t it? And a terrible shock to
her. Indeed, she didn’t feel the same even yet. She didn’t believe she
ever would. Between that and the loss of the letting. . . .
What had he said before he started? Why, he hadn’t said
anything! At least he had said he wouldn’t be home until about
midnight, and for her not to forget to leave the hall door on the latch
and to put some supper on the table in his room. And she had done.
She had left everything right for him, and then she had gone to bed.
And she had slept. She was a good sleeper, except that one time
after she had had scarlet fever, when the doctor said . . .
Yes, the rooms were ready at any time. She believed in keeping
her house clean and tidy at all times, so that everything was always
ready when it was wanted. She had once been in service with Mrs.
Lloyd-Hurley in Chagford and she had learnt that lesson there. Mrs.
Lloyd-Hurley was very particular. She . . .
Mr. Pyke’s things? Oh yes, they were gone. She thought that
would be understood when she said the rooms were ready. She . . .
It was his cousin. His cousin had come down from London and
taken everything there was. That was Mr. Jefferson Pyke. Her Mr.
Pyke was Mr. Stanley. Mr. Jefferson was the only remaining relative,
at least so she understood. He packed up everything and took it
away. Except a few things that he said he didn’t want. These she
had kept. Not that she wanted them, but if they were going begging,
as Mr. French might say, why, then. . . .
No, she had only seen Mr. Jefferson once before. He lived in the
Argentine, or was it Australia? She wasn’t rightly sure—she had no
memory for places—but he lived away in some strange foreign
country, anyhow. He happened to be over on a visit and was going
back again shortly. Her brother James lived in Australia and she had
asked Mr. Jefferson . . .
So French sat and listened while the unending stream poured
about his devoted head. At times by summoning up all his resolution
he interposed a remark which diverted the current in a new direction.
But his perseverance was rewarded as from nearly all of these
mutations he learned at least one fact. When at last, exhausted but
triumphant, he rose to take his leave, he had gained the following
information:
Mr. Stanley Pyke was a jolly, pleasant-mannered man of about
five-and-thirty, who had lodged with the talkative landlady for the
past four years. He had been connected with the works for much
longer than that, but at first had had other rooms farther down the
street. Hers, the landlady modestly explained, were the best in the
town, and Mr. Pyke’s removal was an outward and visible sign of his
prosperity. For the rest, he was satisfactory as lodgers go, easy to
please, not stingy about money, and always with a pleasant word for
her when they met.
On the evening of the tragedy he had dined at six-fifteen instead
of seven, his usual hour. He had gone out immediately after, giving
the instruction about the door and his supper. The landlady had gone
to bed as usual, and the first intimation she had had that anything
was wrong was the visit of the police on the following morning.
Some one, she did not know who, must have informed the
cousin, Mr. Jefferson Pyke, for that evening he turned up. He had
stayed at Torquay for three or four days, coming over to Ashburton to
see the police and make enquiries. On one of these visits he had
called on her and stated that, as he was the only surviving relative of
his cousin, he would take charge of his personal effects. He had
packed up and removed a good many of the dead man’s things,
saying he did not want the remainder and asking her to dispose of
them.
It had not occurred to her to question Mr. Jefferson Pyke’s right to
take her lodger’s property. She had seen him once before, in Mr.
Stanley’s lifetime. Some two months before the tragedy Mr. Stanley
had told her that his cousin was home on a visit from the Argentine—
she believed it was the Argentine and not Australia—and that he was
coming down to see him. He asked her could she put him up. Mr.
Jefferson had arrived a day or two later and she had given him her
spare bedroom. He stayed for four days and the cousins had
explored the moor together. Mrs. Berlyn, she had heard, had driven
them about in her car. The landlady had found Mr. Jefferson very
pleasant; indeed, when the two men were together they had nearly
made her die laughing with their jokes and nonsense. Mr. Jefferson
had told her that he owned a ranch in the Argentine and that he was
thinking of starting flower gardens from which to supply the cities. He
was then on his way back from the Scillys, where he had gone to
investigate the industry. A week after Mr. Jefferson left, Mr. Stanley
took his holidays, and he had told her he was going with his cousin
to the south of France to a place called Grasse, where there were
more gardens. He was only back some three weeks when he met his
death.
All this was given to French with a wealth of detail which, had it
been material to his investigation, he would have welcomed, but by
which, as it was, he was frankly bored. However, he could do nothing
to stop the stream and he simulated interest as best he could.
“By the way, Mrs. Billing,” he said, pausing on his way out, “if I
take these rooms could you look after the mending of my clothes?
Who did it for Mr. Pyke?”
Mrs. Billing had, and she would be delighted to do the same for
Mr. French.
“Well, I have some that want it at the present time. Suppose I
bring them over now. Could you look at them?” Five minutes later he
returned with his suitcase and spread out the clothes as he had done
for Lizzie Johnston an hour or two before. Like the maid, Mrs. Billing
glanced over them and remarked that there didn’t seem to be much
wrong.
French picked up the grey sock.
“But you see they have not been very neatly darned. This grey
one has been done with a different coloured wool. I thought perhaps
you could put that right.”
Mrs. Billing took the grey sock and stared at it for some time,
while a puzzled expression grew on her face. French, suddenly
keenly excited, watched her almost breathlessly. But after turning it
over she put it down, though the slightly mystified look remained.
“Here are some underclothes,” French went on. “Do these want
any mending?”
Slowly the landlady turned over the bundle. As she did so
incredulity and amazement showed on her bird-like features. Then
swiftly she turned to the neck of the vest and the shirt cuffs and
scrutinised the buttons and links.
“My Gawd!” she whispered, hoarsely, and French saw that her
face had paled and her hands were trembling.
“You recognise them?”
She nodded, her flood of speech for once paralysed.
“Where did you get them?” she asked, still in a whisper.
 French was quite as excited as she, but he controlled himself and
spoke easily.
“Tell me first whose they are and how you are so sure of them.”
“They’re Mr. Pyke’s, what he was wearing the night he was lost. I
couldn’t but be sure of them. See here. There’s the wool first. I
darned that and I remember I hadn’t the right colour. Then these
buttons.” She picked up the vest. “I put that one on. See, it’s not the
same as the rest; it was the only one I could get. And then if that
wasn’t enough, these are the cuff-links. I’ve seen them hundreds of
times and I’d know them anywhere. Where did you get them?”
French dropped his suave, kindly manner and suddenly became
official and, for him, unusually harsh.
“Now, Mrs. Billing,” he said, sharply, “I’d better tell you exactly
who I am and warn you that you’ve got to keep it to yourself. I am
Inspector French of New Scotland Yard; you understand, a police
officer. I have discovered that Mr. Pyke was murdered and I am on
the track of the murderer.”
The landlady gave a little scream. She was evidently profoundly
moved, not only by surprise and excitement, but by horror at her late
lodger’s fate. She began to speak, but French cut her short.
“I want you to understand,” he said, threateningly, “that you must
keep silence on this matter. If any hint of it gets about, it will be a
very serious thing for you. I take it you don’t want to be mixed up in a
murder trial. Very well, then; keep your mouth shut.”
Mrs. Billing was terrified and eagerly promised discretion. French
questioned her further, but without result. She did not believe her late
lodger was on bad terms with anyone, nor did she know if he had a
birthmark on his upper arm.
French’s delight at his discovery was unbounded. The
identification of the dead man represented the greatest step towards
the completion of his case that he had yet made. He chuckled to
himself in pure joy.
But his brain reeled when he thought of his four test points. If this
news were true, he had made some pretty bad mistakes! Each one
of his four conclusions must be false. As he remembered the facts
on which they were based, he had to admit himself completely
baffled.
 Presently his mood changed and a wave of pessimism swept
over him. The identification of the underclothes was not, after all, the
identification of the body. Such an astute criminal as he was dealing
with might have changed the dead man’s clothes. But when he
reminded himself that the man who called for the crate resembled
Berlyn, the thing became more convincing. However, it had not been
proved, and he wanted certainty.
Fortunately there was the birthmark. French had examined it
carefully and was satisfied that it was genuine. Who, he wondered,
could identify it?
The most likely person, he thought, was Jefferson Pyke. It would
be worth a journey to London to have the point settled. That night,
therefore, he took the sleeping-car express to Paddington.
Daw had given him the address—17b, Kepple Street, off Russell
Square, and before ten next morning he was there.
Jefferson Pyke was a clean-shaven man of about forty, of rather
more than medium height and stoutly built. He was a study in
browns: brown eyes, a dusky complexion, hair nearly black, brown
clothes and shoes, and a dark-brown tie. He looked keenly at his
visitor, then pointed to a chair.
“Mr. French?” he said, speaking deliberately. “What can I do for
you, sir?”
“I’ll tell you, Mr. Pyke,” French answered. “First of all, here is my
professional card. I want some help from you in an investigation I am
making.”
Pyke glanced at the card and nodded.
“A case on which I was engaged took me recently to Ashburton,
and while there I heard of the tragic death of Mr. Pyke and Mr. Berlyn
of the Veda Works staff. I understand that Mr. Pyke was a relative of
yours?”
“That is so. My first cousin.”
“Well, Mr. Pyke,” French said, gravely, “I have to inform you that a
discovery had been made which may or may not have a bearing on
your cousin’s fate. A body has been found—the body of a murdered
man. That body has not been identified, but there is a suggestion
that it may be your cousin’s. I want to know if you can identify it?”
Mr. Pyke stared incredulously.
 “Good Heavens, Inspector! That’s an astonishing suggestion. You
must surely be mistaken. I went down to Ashburton directly I heard
of the accident, and there seemed no doubt then about what had
happened. Tell me the particulars.”
“About a fortnight ago, as you may have noticed in the papers, a
crate was picked up in the sea off Burry Port in South Wales, which
was found to contain the body of a murdered man. The face had
been disfigured and there was no means of identification. However, I
traced the crate and I learned that it was sent out from the Veda
Works on the morning after your cousin and Mr. Berlyn
disappeared.”
“Good Heavens!” Mr. Pyke exclaimed again. “Go on.”
“I made enquiries and the only persons known to have
disappeared were those two men. You see the suggestion? I am
sorry to have to ask you, but can you help me to identify the
remains?”
Mr. Pyke’s face showed both amazement and horror.
“This is terrible news, Inspector. I need hardly say I hope you are
mistaken. Of course you may count on me to do all I can.”
“You think you can identify the body, then?”
“Surely I ought to recognise my own cousin?”
“Otherwise than by the face? Remember the face has been
disfigured. I might say, indeed, it is nonexistent, it has been so
savagely battered.”
“By Heaven! I hope you will get the man who did it!” Pyke said,
hotly. “But that does not answer your question.” He hesitated. “If it is
not possible to recognise the features, I’m not so sure. How do you
suggest it might be done?”
French shrugged.
“Identification otherwise than by the features is usually possible.
It is a matter of observation. Some small physical defect, a crooked
finger, the scar of an old cut, or mole on the neck—there are scores
of indications to the observant man.”
Mr. Pyke sat in silence for a few minutes.
“Then I’m afraid I’m not very observant,” he said at last. “I can’t
remember any such peculiarity in poor Stanley’s case.”
 “Nothing in the shape of the finger nails,” French prompted. “No
birthmark, no local roughness or discoloration of the skin?”
“By Jove!” Mr. Pyke exclaimed with a sudden gesture. “There is
something. My cousin had a birthmark, a small red mark on his left
arm, here. I remembered it directly you mentioned the word.”
“Then you have been fairly intimate with your cousin? Have you
often seen this mark?”
“Seen it? Scores of times. We were boys together and I have
noticed it again and again. Why, now I come to think of it, I saw it on
these last holidays I spent with Stanley. We went to the south of
France and shared a cabin in the steamer to Marseilles.”
“Could you describe it?”
“No, but I could sketch it.” He seized a piece of paper and drew a
rough triangle.
French laid his photograph beside the sketch. There could be no
doubt that they represented the same object. Pyke seized the
photograph.
“That’s it. I could swear to it anywhere. You’ve found Stanley’s
body right enough. Good Heavens! Inspector, it’s incredible! I could
have sworn he hadn’t an enemy in the world. Have you any clue to
the murderer?”
Natural caution and official training made French hedge.
“Not as yet,” he answered, assuring himself that his ideas about
Berlyn were hypothetical. “I was hoping that you could give me a
lead.”
“I?” Jefferson Pyke shook his head. “Far from it. Even now I can
scarcely credit the affair.”
“Well, I should like you to run over his associates and see if you
can’t think of any who might have hated him. Now to start with the
senior partner: What about Colonel Domlio?”
Mr. Pyke had never met him and knew nothing about him, though
he had heard his cousin mention his name. French went on through
the list he had made at Ashburton till in the natural sequence he
came to Berlyn.
“Now Mr. Berlyn. Could he have had a down on your cousin?”
“But he was lost, too,” Pyke rejoined, then stopped and looked
keenly at French. “By Jove! Inspector, I get your idea! You think
Berlyn may have murdered him and cleared out?” He shook his
head. “No, no. You are wrong. It is impossible. Berlyn wasn’t that
sort. I knew him slightly and I confess I didn’t care for him, but he
was not a murderer.”
“Why did you not like him, Mr. Pyke?”
Pyke shrugged.
“Hard to say. Not my style, perhaps. A good man, you know, and
efficient and all that, but—too efficient, shall I say? He expected too
much from others; didn’t make allowances for human errors and
frailties. Poor Mrs. Berlyn had rather a time with him.”
“How so?”
“Well, an example will explain what I mean. On this last holiday
after Stanley and I got back to London we met Berlyn and his wife,
who were in town. The four of us dined together and went to a
theatre. We were to meet at the restaurant at seven. Well, Mrs.
Berlyn had been off somewhere on her own and she was five
minutes late. What was that for a woman? But Berlyn was so ratty
about it that I felt quite embarrassed. You see, he wouldn’t have
been late himself. If he had said seven, he would have been there—
on the tick. He couldn’t see that other people were not made the
same way.”
“I follow you. You say that Mrs. Berlyn had rather a time with him.
Did they not get on?”
“Oh, they got on—as well as fifty per cent of the married people
get on. Berlyn did his duty to her strictly, even lavishly, but he
expected the same in return. I don’t know that you could blame him.
Strictly speaking, of course he was right. It was his instinct for
scrupulously fair play.”
“Your late cousin and Mrs. Berlyn were very good friends, were
they not?”
“We were both good friends with Mrs. Berlyn. Stanley and I knew
her as children. In fact, it was through Stanley that Berlyn met her. I
was in the Argentine at the time, but he told me about it. Berlyn was
going for a holiday—one of those cruises round the western
Mediterranean. Stanley happened to have met Phyllis Considine, as
she was then, in London, and she had mentioned she was going on