Instant Ebook Access, One Click Away – Begin at ebookgate.

com

Variational Methods in Nonlinear Analysis With

Applications in Optimization and Partial
Differential Equations Dimitrios C Kravvaritis

https://ebookgate.com/product/variational-methods-in-
nonlinear-analysis-with-applications-in-optimization-and-
partial-differential-equations-dimitrios-c-kravvaritis/

OR CLICK BUTTON

DOWLOAD EBOOK

Get Instant Ebook Downloads – Browse at https://ebookgate.com

Click here to visit ebookgate.com and download ebook now
Instant digital products (PDF, ePub, MOBI) available
Download now and explore formats that suit you...

Advanced Numerical Methods with Matlab 2 Resolution of

Nonlinear Differential and Partial Differential Equations
1st Edition Bouchaib Radi
https://ebookgate.com/product/advanced-numerical-methods-with-
matlab-2-resolution-of-nonlinear-differential-and-partial-
differential-equations-1st-edition-bouchaib-radi/
ebookgate.com

Methods of Nonlinear Analysis Applications to Differential

Equations 2ed. Edition Pavel Dra■Bek

https://ebookgate.com/product/methods-of-nonlinear-analysis-
applications-to-differential-equations-2ed-edition-pavel-drabek/

ebookgate.com

Partial Differential Equations Theory Analysis and

Applications Theory Analysis and Applications 1st Edition
Christopher L. Jang
https://ebookgate.com/product/partial-differential-equations-theory-
analysis-and-applications-theory-analysis-and-applications-1st-
edition-christopher-l-jang/
ebookgate.com

Wavelet methods for solving partial differential equations

and fractional differential equations 1st Edition Gupta

https://ebookgate.com/product/wavelet-methods-for-solving-partial-
differential-equations-and-fractional-differential-equations-1st-
edition-gupta/
ebookgate.com
Fourier Series in Several Variables with Applications to
Partial Differential Equations 1st Edition Victor Shapiro

https://ebookgate.com/product/fourier-series-in-several-variables-
with-applications-to-partial-differential-equations-1st-edition-
victor-shapiro/
ebookgate.com

Handbook of Nonlinear Partial Differential Equations

Second Edition Andrei D. Polyanin

https://ebookgate.com/product/handbook-of-nonlinear-partial-
differential-equations-second-edition-andrei-d-polyanin/

ebookgate.com

An introduction to nonlinear partial differential

equations 2nd ed Edition J. David Logan

https://ebookgate.com/product/an-introduction-to-nonlinear-partial-
differential-equations-2nd-ed-edition-j-david-logan/

ebookgate.com

Nonlinear Hyperbolic Equations Spectral Theory and Wavelet

Transformations A Volume of Advances in Partial
Differential Equations 1st Edition P. R. Popivanov (Auth.)
https://ebookgate.com/product/nonlinear-hyperbolic-equations-spectral-
theory-and-wavelet-transformations-a-volume-of-advances-in-partial-
differential-equations-1st-edition-p-r-popivanov-auth/
ebookgate.com

Introduction to Partial Differential Equations and Hilbert

Space Methods Karl E. Gustafson

https://ebookgate.com/product/introduction-to-partial-differential-
equations-and-hilbert-space-methods-karl-e-gustafson/

ebookgate.com
Dimitrios C. Kravvaritis, Athanasios N. Yannacopoulos
Variational Methods in Nonlinear Analysis
Also of Interest
Elementary Functional Analysis
Marat V. Markin, 2018
ISBN 978-3-11-061391-9, e-ISBN (PDF) 978-3-11-061403-9,
e-ISBN (EPUB) 978-3-11-061409-1

Real Analysis
Marat V. Markin, 2019
ISBN 978-3-11-060097-1, e-ISBN (PDF) 978-3-11-060099-5,
e-ISBN (EPUB) 978-3-11-059882-7

Applied Nonlinear Functional Analysis. An Introduction

Nikolaos S. Papageorgiou, Patrick Winkert, 2018
ISBN 978-3-11-051622-7, e-ISBN (PDF) 978-3-11-053298-2,
e-ISBN (EPUB) 978-3-11-053183-1

Functional Analysis. A Terse Introduction

Gerardo Chacón, Humberto Rafeiro, Juan Camilo Vallejo, 2016
ISBN 978-3-11-044191-8, e-ISBN (PDF) 978-3-11-044192-5,
e-ISBN (EPUB) 978-3-11-043364-7

Dynamics of Solid Structures. Methods using Integrodifferential

Relations
Georgy Viktorovich Kostin, Vasily V. Saurin
ISBN 978-3-11-051623-4, e-ISBN (PDF) 978-3-11-051644-9,
e-ISBN (EPUB) 978-3-11-051625-8
Dimitrios C. Kravvaritis,
Athanasios N. Yannacopoulos

Variational
Methods in
Nonlinear Analysis
|
With Applications in Optimization and Partial
Differential Equations
Mathematics Subject Classification 2010
35-02, 65-02, 65C30, 65C05, 65N35, 65N75, 65N80

Authors
Prof. Dr. Dimitrios C. Kravvaritis Prof. Dr. Athanasios N. Yannacopoulos
National Technical University of Athens Athens University of
School of Applied Mathematical and Economics and Business
Physical Sciences Department of Statistics
Heroon Polytechniou 9 Patission 76 104 34 Athens, Greece
Zografou Campus [email protected]
157 80 Athens, Greece
[email protected]

ISBN 978-3-11-064736-5
e-ISBN (PDF) 978-3-11-064738-9
e-ISBN (EPUB) 978-3-11-064745-7

Library of Congress Control Number: 2020934276

Bibliographic information published by the Deutsche Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie;
detailed bibliographic data are available on the Internet at http://dnb.dnb.de.

© 2020 Walter de Gruyter GmbH, Berlin/Boston

Cover image: your_photo / iStock / Getty Images
Typesetting: VTeX UAB, Lithuania
Printing and binding: CPI books GmbH, Leck

www.degruyter.com
|
DCK: To my wife Katerina and my two sons Christos and Nikos.

ANY: To my fixed-point Electra and to Nikos, Jenny and Helen

|
Spent the morning putting in a comma and the afternoon removing it.
G. Flaumbert or O. Wilde

If you can’t prove your theorem, keep shifting parts of the conclusion to the assump-
tions, until you can.
E. De Giorgi (according to Wikipedia)
Preface
Nonlinear analysis is a vast field, which originated from the need of extending clas-
sical linear structures in analysis in order to treat nonlinear problems. It can be con-
sidered as general term for describing a wide spectrum of abstract mathematical tech-
niques, ranging from geometry, topology and analysis, which nevertheless find im-
portant applications in a variety of fields, such as mathematical modeling (e. g., in
economics, engineering, biology, decision science, etc.), optimization or in applied
mathematics (e. g., integral, differential equations and partial differential equations).
It is the aim of this book to select out of this wide spectrum, certain aspects that the
authors consider as most useful in a variety of applications, and in particular aspects
of the theory related with variational methods, which are most closely related to ap-
plications in optimization and PDEs.
The book is intended to graduate students, lecturers or independent researchers
who wish a concise introduction to nonlinear analysis, in order to learn the material
independently, use it as a guideline for the organization of a course, or use it as a ref-
erence when trying to master certain techniques to use in their own research. We hope
that we have achieved a good balance between theory and applications. All the theo-
retical results are proved in full detail, emphasizing certain delicate points, and many
examples are provided guiding the reader either towards extensions of the fundamen-
tal results or towards important applications in a variety of fields. Large sections are
devoted to the applications of the theoretical results in optimization, PDEs and varia-
tional inequalities, including hints toward a more algorithmic approach. Keeping the
balance right, the volume manageable but at the same time striving to provide as wide
a spectrum of techniques and subjects as possible has been a very tricky business.
We hope that we have managed to present a good variety of these parts of nonlin-
ear analysis which find the majority of applications in the fields which are the within
the core interest of this book, i. e., optimization and PDEs. Naturally, we had to draw
lines in our indulgence to the beautiful subjects we have decided to present (other-
wise we would have results in a nine volume treatise rather than a nine chapter book)
but at the same time we have to completely omit important fields such as, e. g., degree
theory, and we apologize for this, hoping to make up for this omission in the future!
At any rate, we hope that our effort (which by the way was extremely enjoyable) will
make itself useful to those wishing a concise introduction to nonlinear analysis with
an emphasis toward applications, and will help newcomers of any mathematical age
to the field, colleagues in academia to organize their courses, and practitioners or re-
searchers toward obtaining the methodological framework that will help them toward
reaching their goal.
This book consists of nine chapters, which we think cover a wide spectrum of con-
cepts and techniques of nonlinear analysis and their applications.

https://doi.org/10.1515/9783110647389-201
X | Preface

Chapter 1 collects, mostly without proofs, some preliminary material from topol-
ogy, linear functional analysis, Sobolev spaces and multivalued maps, which are con-
sidered as necessary in order to proceed further to developing the material covered in
this book. Even though proofs are not provided for this introductory material, we try
to illustrate and clarify the concepts using numerous examples.
Chapter 2 develops calculus in Banach spaces and focuses on convexity as a fun-
damental property of subsets of Banach spaces as well as on the properties of con-
vex functions on Banach spaces. This material is developed in detail, starting from
topological properties of convex sets in Banach space, in particular, their remarkable
properties with respect to the weak topology, and then focuses on properties of con-
vex functions with respect to continuity and semicontinuity or with respect to their
Gâteaux or Fréchet derivatives. Extended comments on the deep connections of con-
vexity with optimization are made.
Chapter 3 is devoted to an important tool of nonlinear analysis, which is used
throughout this book, fixed-point theory. One could devote a whole book to the sub-
ject, so here we take a minimal approach, of presenting a selection of fixed-point
theorems which we find indispensable. This list consists of the Banach contraction
theorem, Brouwer’s fixed-point theorem which, although finite dimensional in na-
ture, forms the basis for developing a large number of useful fixed-point theorems
in infinite dimensional spaces, Schauder’s fixed-point theorem and Browder’s fixed-
point theorem. Special attention is paid to the construction of iterative schemes for
the approximation of fixed points, and in particular the Krasnoselskii–Mann iterative
scheme which finds interesting applications in numerical analysis. We also introduce
the Caristi fixed-point theorem and its deep connection with a very useful tool, the
Ekeland variational principle. The chapter closes with a fixed-point theorem for mul-
tivalued maps.
Chapter 4 provides an introduction to nonsmooth convex functions and the con-
cept of the subdifferential. The theory of the subdifferential as well as subdifferen-
tial calculus is developed gradually, leading to its applications to optimization. The
Moreau proximity operator in Hilbert spaces is introduced and its use in a class of
popular optimization algorithms, which go under the general name of proximal algo-
rithms is presented.
Chapter 5 is devoted to the theory of duality in convex optimization. The chap-
ter begins with an indispensable tool, a version of the minimax theorem which al-
lows us to characterize saddle points for functionals of specific structure in Banach
spaces. We then define and provide the properties of the Fenchel–Legendre conjugate
as well as the biconjugate, and illustrate them with numerous examples. To further il-
lustrate the properties of these important transforms, we introduce the concept of the
inf-convolution, a concept of interest in its own right in optimization. Having devel-
oped the necessary machinery, we present the important contribution of the minimax
approach in optimization starting with a detailed presentation of Fenchel’s duality
 Preface | XI

theory and then its generalization to a more general scheme that encompasses La-
grangian duality as well or the augmented Lagrangian method. These techniques are
illustrated in terms of various examples and applications. The chapter closes by pre-
senting an important class of numerical methods based on the duality approach and
discuss their properties.
Chapter 6 is an introduction to the calculus of variations and its applications. Af-
ter a short motivation, we start with a warm up problem, and in particular the vari-
ational theory of the Laplacian. This is a nice opportunity to either introduce or re-
mind (depending on the reader) some important properties of the Laplacian, such as
the solvability and the properties (regularity, etc.) of Poisson’s equation or the spec-
tral theory related to this operator. Having spent some time on this important case,
we return to the general case, presenting results on the lower semicontinuity of inte-
gral functionals in Sobolev spaces, the existence of minimizers and their connection
with the Euler–Lagrange equation. We then consider the problem of possible extra
regularity of the minmizer beyond that guaranteed by their membership in a Sobolev
space (already established by existence). To this end, we provide an introduction to
the theory of regularity of solutions variational problems, initiated by De Giorgi, and
in particular present some fundamental results concerning Hölder continuity of min-
imizers or higher weak differentiability. The chapter closes with two important exam-
ples, the application of the calculus of variations to semilinear elliptic problems and
to quasilinear elliptic problems and in particular problems related to the p-Laplace
operator.
Chapter 7 considers variational inequalities, a subject of great interest in various
fields ranging from mechanics to finance or decision science. As a warm-up, we re-
visit the problem of the Laplacian but now related to the minimization of the Dirichlet
functional on convex closed subsets of a Sobolev space, which naturally lead to a vari-
ational inequality for the Laplace operator. Then the Lions–Stampacchia theory for ex-
istence of solutions to variational inequalities (not necessarily related to minimization
problems) is developed, initially in a Hilbert space setting and with generalizations to
variational inequalities of second kind. We then proceed to study the problem of ap-
proximation of variational inequalities, presenting the penalization method, as well
as the internal approximation method. The chapter closes with an important class of
applications, the study of variational inequalities for second order elliptic operators.
As the Lions–Stampacchia theory covers equations as a special case, we take up the
opportunity to start our presentation with second-order elliptic equations and in this
way generalize certain properties that we have seen for the Laplacian in Chapter 6, to
more general equations. We then move to variational inequalities and discuss various
issues such as solvability, comparison principles, etc., and close the chapter with a
study of variational inequalities involving the p-Laplace operator.
Chapter 8 introduces critical point theory for functionals in Banach spaces. Hav-
ing understood quite well the properties of minimizers of functionals and their con-
XII | Preface

nection with PDEs, it is interesting to see whether extending this analysis to critical
points in general may provide some useful insights. The chapter starts with the presen-
tation of an important class of theorems related to the Ambrosseti–Rabinowitz moun-
tain pass theorem and it generalizations (e. g., the saddle point theorem). This the-
ory provides general criteria under which critical points of certain functionals exist.
Through the Euler–Lagrange equation such existence results may prove useful to the
study of certain PDEs or variational inequalities. Our applications focus on semilinear
and quasilinear PDEs, related to the p-Laplacian.
The book finishes with Chapter 9 which deals with a very important part of nonlin-
ear analysis, the study of monotone type operators. During the development of the ma-
terial up to now, we have seen that monotonicity in some form or another has played
a key role to our arguments (such as, e. g., the monotonicity of the Gâteaux deriva-
tive of convex functions, or the coercivity of bilinear forms in the Lions–Stampacchia
theory, etc.). In this chapter, motivated by these observations, we study the mono-
tonicity properties of operators as a primary issue and not as a derivative of other
properties, e. g., convexity. As we shall see, this allows us to extend certain interesting
results that we have arrived to via minimization or in general variational arguments
for operators that may not necessarily enjoy such a structure. We start our study with
monotone operators, then move to maximal monotone operators and then introduce
pseudomonotone operators. Our primary interest is in obtaining surjectivity results
for monotone- type operators, which will subsequently used in the study of certain
PDEs or variational inequalities, but along the way encounter certain interesting re-
sults such as, e. g., the Yosida approximation, etc. The theory is illustrated with appli-
cations in quasilinear PDEs, differential inclusions, variational inequalities and evo-
lution problems.
We wish to acknowledge the invaluable support of various people for various rea-
sons. We must start with acknowledging the support of Dr. Apostolos Damialis, at the
time the Mathematics Editor at De Gruyter whose contribution made this book a re-
ality. Then we are indebted to Nadja Schedensack at De Gruyter for her superb han-
dling of all the details and her understanding. We thank Ina Talandienė and Monika
Pfleghar for their superb technical assistance. As this material has been tested on au-
diences at our respective institutions over the years, we wish to thank our students
who have helped us shape and organize this material through the interaction with
them. We both feel compelled to mention specifically George Domazakis, whose help
in proof reading and commenting some of this material as well as with technical issues
related to the De Gruyter latex style file has been invaluable, and we deeply thank him
for that. We are also thankful to Kyriakos Georgiou for his kind offer for proofreading
the manuscript and his useful comments.
There are many colleagues and friends the interaction and collaboration with
which throughout these years have shaped our interests and encouraged us. We need
to mention, Professors N. Alikakos, I. Baltas, D. Drivaliaris, N. Papageorgiou, G. Papa-
giannis, G. Smyrlis, I. G. Stratis, A. Tsekrekos, S. Xanthopoulos, N. A. Yannacopoulos
 Preface | XIII

and A. Xepapadeas. At a personal level, ANY is indebted to Electra Petracou for mak-
ing all this feasible through her constant support and devotion and helpful advise.

Athens, 2019 D. C. Kravvaritis, NTUA,

A. N. Yannacopoulos, AUEB
1 Preliminaries
In this introductory chapter, we collect (mostly without proofs but with a number of
illustrative examples) the fundamental notions from linear functional analysis, con-
vexity, Lebesgue, Sobolev, Sobolev–Bochner spaces and multivalued mappings, that
are essential in proceeding to the main topic of this book which is nonlinear analysis
and its various applications. For detailed presentations of the material covered in this
chapter, we refer to, e. g., [1, 5, 4, 28, 48, 109].

1.1 Fundamentals in the theory of Banach spaces

Most of this book will concern Banach spaces, i. e., complete normed spaces. In this
section, we recall some of their fundamental properties.

1.1.1 Linear operators and functionals

Linear mappings between normed spaces and functionals play an important role in
(linear) functional analysis.

Definition 1.1.1 (Operators). Let X, Y be two normed spaces.

(i) A map A : D(A) ⊂ X → Y (not necessarily linear) is called an operator (or func-
tional if Y = ℝ). We further define the sets

D(A) := {x ∈ X : A(x) ∈ Y},

Gr(A) := {(x, A(x)) : x ∈ D(A)} ⊂ X × Y,
N(A) := {x ∈ X : A(x) = 0},
R(A) := {y ∈ Y : ∃ x ∈ X, for which A(x) = y} = A(X),

called the domain, graph, kernel and range of A, respectively.

(ii) If A is such that A(λ1 x1 + λ2 x2 ) = λ1 A(x1 ) + λ2 A(x2 ), it is called a linear operator (or
linear functional, denoted by f if Y = ℝ).
(iii) The linear operator A : X → Y is called bounded if sup{‖A(x)‖Y : x ∈ X, ‖x‖X ≤
1} < ∞. The set of bounded linear operators between X and Y is denoted by
ℬ(X, Y).
(iv) The set of continuous linear operators between X and Y is denoted by ℒ(X, Y). In
the special case where Y = X, we will use the simplified notation ℒ(X) instead of
ℒ(X, X).

The following theorem collects some useful properties of linear operators and
functionals over normed spaces.

https://doi.org/10.1515/9783110647389-001
2 | 1 Preliminaries

Theorem 1.1.2 (Elementary properties of linear operators and functionals). Let A :

X → Y be a linear mapping.
(i) The following are equivalent: (a) A is continuous at 0, (b) A is continuous, (c) A is
uniformly continuous, (d) A is Lipschitz, (e) A is bounded. It therefore holds that
ℬ(X, Y) = ℒ(X, Y).
(ii) It holds that

‖A‖ℒ(X,Y) := sup{󵄩󵄩󵄩A(x)󵄩󵄩󵄩Y : x ∈ X, ‖x‖X ≤ 1}

󵄩 󵄩

= sup{󵄩󵄩󵄩A(x)󵄩󵄩󵄩Y : x ∈ X, ‖x‖X = 1}
󵄩 󵄩

‖A(x)‖Y
= sup{ : x ∈ X, x ≠ 0}
‖x‖X
= inf{c > 0 : 󵄩󵄩󵄩A(x)󵄩󵄩󵄩Y ≤ c‖x‖X , x ∈ X}.
󵄩 󵄩

In fact, ℒ(X, Y) is a normed space with norm ‖ ⋅ ‖ℒ(X,Y) , and if Y is a Banach space,
ℒ(X, Y) is a Banach space also.

An important special class of continuous linear functionals are isomorphisms and

isometries.

Definition 1.1.3 (Isomorphisms and isometries). Let X, Y be normed spaces.

(i) A linear operator A : X → Y which is 1–1 and onto (bijection), and such that A and
A−1 are bounded is called an isomorphism. If an isomorphism A exists between
the normed spaces X and Y, the spaces are called isomorphic, and this is denoted
by X ≃ Y.
(ii) A linear operator A : X → Y which is 1–1 and onto (bijection), and such that
‖A(x)‖Y = ‖x‖X , for every x ∈ X, is called an isometry. If an isometry A exists be-
tween the normed spaces X and Y, the spaces are called isometric. Clearly, an
isometry is an isomorphism.

An isomorphism leaves topological properties between the two spaces invariant,

while an isometry leaves distances invariant. One easily sees that if X1 , X2 , X3 are
normed spaces and X1 ≃ X2 , and X2 ≃ X3 , then X1 ≃ X3 , with a similar result holding
for isometric spaces. This shows that the relation X ≃ Y is an equivalence relation
in the class of normed spaces. Furthermore, one can show that a linear 1–1 and onto
(bijective) operator between two normed spaces is an isomorphism if and only if there
exists c1 , c2 > 0 such that c1 ‖x‖X ≤ ‖A(x)‖Y ≤ c2 ‖x‖X , for every x ∈ X, while an isome-
try maps the unit ball of X to the unit ball of Y, thus preserving the geometry of the
spaces. As we will see quite often, when two normed spaces are isometric we will
often “identify” them in terms of this isometry.
One of the fundamental results concerning linear spaces and linear functionals
is the celebrated Hahn–Banach theorem. We will return to this theorem very often,
in any of its multiple versions, especially its versions related to separation of convex
 1.1 Fundamentals in the theory of Banach spaces | 3

subsets of normed spaces. Here, we present its analytic form and will return to its
equivalent geometric forms (in terms of separation theorems) in Section 1.2, where we
study convexity and its properties in detail.
In its analytic form, the Hahn–Banach theorem deals with the extension of a lin-
ear functional defined on a subspace X0 of a (real) linear space X to the whole space.

Theorem 1.1.4 (Hahn–Banach I). Let X0 be a subspace of a (real) linear space X and let
p : X → ℝ be a positively homogeneous subadditive functional on X, i. e., a functional
with the properties (a) p(λ x) = λ p(x), for every λ > 0 and x ∈ X, and (b) p(x1 + x2 ) ≤
p(x1 )+p(x2 ), for every x1 , x2 ∈ X. If f0 : X0 → ℝ is a linear functional such that f0 (x) ≤ p(x)
for every x ∈ X0 , then, there exists a linear functional f : X → ℝ such that f (x) = f0 (x)
on X0 and f (x) ≤ p(x) for every x ∈ X.

As a corollary of the above, we may obtain the following version of the Hahn–
Banach theorem for normed spaces (simply apply the above for the choice p(x) =
‖f0 ‖ℒ(X0 ,ℝ) ‖x‖X ).

Theorem 1.1.5 (Hahn–Banach II). If X is a (real) normed linear space, X0 is a subspace

and f0 : X0 → ℝ is a continuous linear functional on X0 , then there exists a continuous
linear extension f : X → ℝ, of f0 such that ‖f0 ‖ℒ(X0 ,ℝ) = ‖f ‖ℒ(X,ℝ) .

A second fundamental theorem concerning bounded linear operators in Banach

spaces is the Banach–Steinhaus theorem or principle of uniform boundedness, which
provides a connection between the following two different types of boundedness and
convergence in the normed space ℒ(X, Y).

Definition 1.1.6 (Norm and pointwise bounded families). Let X, Y be normed spaces.
(i) A family of operators {Aα : α ∈ ℐ } ⊂ ℒ(X, Y) is called norm bounded if ‖Aα ‖ℒ(X,Y) <
∞ for every α ∈ ℐ , and pointwise bounded if for every x ∈ X (fixed) ‖Aα (x)‖Y < ∞
for every α ∈ ℐ .
(ii) A sequence {An : n ∈ ℕ} ⊂ ℒ(X, Y) is called norm convergent to A ∈ ℒ(X, Y) if
‖An − A‖ℒ(X,Y) → 0 as n → ∞, and pointwise convergent to A ∈ ℒ(X, Y) if An (x) →
A(x) for every x ∈ X.

The Banach–Steinhaus theorem states the following.

Theorem 1.1.7 ( Banach–Steinhaus). Let X, Y be Banach spaces.1 Consider a family of

continuous linear mappings 𝒜 := {Aα : X → Y : α ∈ ℐ } ⊂ ℒ(X, Y), (not necessarily
countable). 𝒜 is norm bounded if and only if it is pointwise bounded. The principle of
course remains valid for linear functionals, upon setting Y = ℝ.

1 Y may in fact simply be a normed space; see, e. g., the principle of uniform boundedness Theo-
rem 14.1 [48].
4 | 1 Preliminaries

In other words, if we have a family of linear operators {Aα : X → Y : α ∈ ℐ } ⊂

ℒ(X, Y) such that supα∈ℐ ‖Aα x‖Y < ∞ for every x ∈ X, then supα∈ℐ ‖Aα ‖ℒ(X,Y) < ∞,
i. e., there exists a constant c > 0 such that ‖Aα x‖Y ≤ c‖x‖X for every x ∈ X and α ∈ ℐ ,
with the constant c being independent of both x and α. The Banach–Steinhaus theo-
rem has important implications for operator theory. The following example shows us
how to use it to define linear bounded operators as the pointwise limit of a sequence
of bounded operators, a task often fundamental in numerical analysis and approxi-
mation theory.

Example 1.1.8 (Definition of operators as pointwise limits). Consider a sequence of

linear bounded operators {An : n ∈ ℕ} ⊂ ℒ(X, Y) with the property: For every x ∈ X,
there exists yx ∈ Y such that An (x) → yx in Y. Then these pointwise limits define an
operator A : X → Y by A(x) := yx = limn An (x) for every x ∈ X, which is linear and
bounded, and
‖A‖ℒ(X,Y) ≤ lim inf ‖An ‖ℒ(X,Y) . (1.1)
n→∞

By the Banach–Steinhaus theorem there exists a c > 0 such that ‖An x‖ ≤ c ‖x‖, for
all x ∈ X, n ∈ ℕ. We pass to the limit as n → ∞ and since An (x) → A(x) in Y for every
x ∈ X we have that ‖An (x)‖Y → ‖A(x)‖Y as n → ∞, therefore, ‖A(x)‖Y ≤ c ‖x‖X , for every
x ∈ X from which follows that A ∈ ℒ(X, Y). Furthermore, taking the limit inferior on
the inequality ‖An (x)‖Y ≤ ‖An ‖ℒ(X,Y) ‖x‖X for every x, we have that

󵄩󵄩A(x)󵄩󵄩󵄩Y = lim 󵄩A (x)󵄩󵄩󵄩Y = limninf󵄩󵄩󵄩An (x)󵄩󵄩󵄩Y ≤ limninf ‖An ‖ℒ(X,Y) ‖x‖X , ∀ x ∈ X,

󵄩󵄩 󵄩 󵄩󵄩 󵄩 󵄩 󵄩
n 󵄩 n

where upon taking the supremum over all ‖x‖ ≤ 1 we obtain (1.1). ◁

The other fundamental result concerning bounded linear operators between Ba-
nach spaces is the open mapping theorem, which is again a consequence of the Baire
category theorem (for a proof see, e. g., [28]).

Theorem 1.1.9 (Open mapping). Let X, Y be Banach spaces and A : X → Y a linear

bounded operator, which is surjective (onto). Then A is also open, i. e., the image of every
open set in X under A is also open in Y.

Example 1.1.10. Let A : X → Y be a linear bounded operator which is onto and one to
one (bijective). Then A−1 is a bounded operator.
By the open mapping theorem if 𝒪 ⊂ X is open, then A(𝒪) ⊂ Y is open. Since
A(𝒪) = (A−1 )−1 (𝒪), it follows that A−1 maps open sets to open sets hence, it is continu-
ous, and being linear it is bounded. ◁

An important consequence of the open mapping theorem is the closed graph the-
orem.

Theorem 1.1.11 (Closed graph). Let X, Y be Banach spaces and A : X → Y be a lin-

ear operator (not necessarily bounded). A is continuous (hence, bounded) if and only if
Gr(A) is closed.
 1.1 Fundamentals in the theory of Banach spaces | 5

Using the open mapping theorem, we can show that if Gr(A) is closed in X × Y, the
operator A is continuous, a fact which is not true for nonlinear operators.

1.1.2 The dual X ⋆

An important concept is that of the dual space, which can be defined in terms of con-
tinuous linear mappings from X to ℝ (functionals).

Definition 1.1.12 (Dual space). Let X be a normed space. The space ℒ(X, ℝ) of all con-
tinuous linear functionals from X to ℝ is called the (topological) dual space of X, and
is denoted by X ⋆ , its elements by x⋆ and their action on elements x ∈ X, defines a du-
ality pairing between X and X ⋆ by x⋆ (x) =: ⟨x⋆ , x⟩X ⋆ ,X , often denoted for simplicity as
⟨⋅, ⋅⟩.

Proposition 1.1.13 (The dual as a normed space). Let X be a normed space. The dual
space X ⋆ is a Banach space2 when equipped with the norm3 ‖x⋆ ‖X ⋆ = sup{|⟨x⋆ , x⟩| :
‖x‖X ≤ 1}, where ⟨⋅, ⋅⟩ is the duality pairing between the two spaces, ⟨x⋆ , x⟩ = x⋆ (x), for
every x ∈ X, x⋆ ∈ X.

Often it is convenient to understand the dual space of a normed linear space in

terms of its isometry to a different normed linear space. One of the most important
results of this type holds in the special case where X is a Hilbert space H. In this case,
one has the celebrated Riesz representation theorem.

Theorem 1.1.14 (Riesz). Let H be a Hilbert space with inner product ⟨⋅, ⋅⟩H . For every
x⋆ ∈ H ⋆ , there exists a unique xo ∈ H such that x⋆ (x) = ⟨x⋆ , x⟩ = ⟨xo , x⟩H for every x ∈ X.
Moreover, ‖x⋆ ‖H ⋆ = ‖xo ‖H .

For more general Banach spaces, one may still identify the dual space with a con-
crete vector space.

Example 1.1.15 (The dual space of sequence spaces). Let X = ℓp = {x = {xn : n ∈ ℕ} :

xn ∈ ℝ, ∑n∈ℕ |xn |p < ∞}, p ∈ (1, ∞), equipped with the norm ‖x‖ℓp = {∑n∈ℕ |xn |p }1/p .
Its dual space is identified with the sequence space ℓp with p1 + p1⋆ = 1.
⋆
◁

Similar constructions can be done for other spaces; see, e. g., Sections 1.4 and 1.5.
Since the dual space X ⋆ is the space of continuous linear functionals on X, the
Hahn–Banach theorem plays an important role in the study of the dual space. One
of the many instances where this shows up is in the construction of the duality map
from X to X ⋆ , the existence of which follows by an application of the Hahn–Banach
theorem.

2 even if X is not.
|⟨x⋆ ,x⟩|
3 or equivalently, ‖x⋆ ‖X ⋆ = supx∈X\{0} ‖x‖X
= supx∈X, ‖x‖X =1 |⟨x⋆ , x⟩|.
6 | 1 Preliminaries

Definition 1.1.16 (The duality map). The duality map J : X → X ⋆ is the set valued non-
linear map x 󳨃→ J(x), where

󵄩 󵄩2
J(x) := {x⋆ ∈ X ⋆ : ⟨x⋆ , x⟩ = ‖x‖2X = 󵄩󵄩󵄩x⋆ 󵄩󵄩󵄩X ⋆ }.

Example 1.1.17. For every x ∈ X, it holds that

‖x‖X = sup{󵄨󵄨󵄨⟨x⋆ , x⟩󵄨󵄨󵄨 : x⋆ ∈ X ⋆ , 󵄩󵄩󵄩x⋆ 󵄩󵄩󵄩X ⋆ ≤ 1}

󵄨 󵄨 󵄩 󵄩

= max{󵄨󵄨󵄨⟨x⋆ , x⟩󵄨󵄨󵄨 : x⋆ ∈ X ⋆ , 󵄩󵄩󵄩x⋆ 󵄩󵄩󵄩X ⋆ ≤ 1}.

󵄨 󵄨 󵄩 󵄩

The fact that the supremum is attained is very important, since in general the supre-
mum in the definition of the norm ‖⋅‖X ⋆ may not be attained (unless certain conditions
are satisfied for X, e. g., if X is a reflexive Banach space, see [28]). ◁

At this point, we recall the notion of the adjoint operator.

Let X, Y be Banach spaces, with duals X ⋆ , Y ⋆ , respectively. We will first define the
concept of the adjoint or dual operator.

Definition 1.1.18 (Adjoint of an operator). Let A : X → Y be a linear operator between

two Banach spaces. The adjoint operator A⋆ : Y ⋆ → X ⋆ is defined by

⟨A⋆ (y⋆ ), x⟩X ⋆ ,X = ⟨y⋆ , A(x)⟩Y ⋆ ,Y ,

for every x ∈ X, y⋆ ∈ Y ⋆ .

The adjoint or dual operator generalizes in infinite dimensional Banach spaces

the concept of the transpose matrix. If X, Y are Hilbert spaces, the term adjoint oper-
ator is used, and upon the identifications X ⋆ ≃ X and Y ⋆ ≃ Y the definition simplifies
to ⟨A⋆ y, x⟩X = ⟨y, Ax⟩Y for every x ∈ X, y ∈ Y. An operator A : X → X, where X = H is
Hilbert space, such that A⋆ = A is called a self-adjoint operator. For such operators, one
may show that the operator norm admits the representation4 ‖A‖ = sup‖x‖=1 |⟨Ax, x⟩|.

1.1.3 The bidual X ⋆⋆

Definition 1.1.19 (Bidual of a normed space). The dual space of X ⋆ is called the bidual
and is denoted by X ⋆⋆ := (X ⋆ )⋆ .

Let us reflect a little bit on the nature of elements of the bidual space X ⋆⋆ . They
are functionals from X ⋆ → ℝ, i. e., continuous linear mappings x⋆⋆ : X ⋆ → ℝ that

4 If m = sup‖x‖=1 |⟨Ax, x⟩|, it is straightforward to show that m ≤ ‖A‖. For the reverse inequality, express
⟨Ax, z⟩ = 41 (⟨A(x+z), x+z⟩−⟨A(x−z), x−z⟩) ≤ 41 (|⟨A(x+z), x+z⟩|+|⟨A(x−z), x−z⟩|) ≤ m4 (‖x+z‖2 +‖x−z‖2 ) =
m
2
(‖x‖2 + ‖z‖2 ) and choose z = ‖x‖
‖Ax‖
Ax.
 1.1 Fundamentals in the theory of Banach spaces | 7

take a functional x⋆ : X → ℝ and map it into a real number x⋆⋆ (x⋆ ). This allows us
to define a duality pairing between X ⋆ and X ⋆⋆ by ⟨x⋆⋆ , x⋆ ⟩X ⋆⋆ ,X ⋆ = x⋆⋆ (x⋆ ) for every
x⋆⋆ ∈ X ⋆⋆ and every x⋆ ∈ X ⋆ . Clearly, this is a different duality pairing than the one
defined between X and X ⋆ . If we were too strict on notation, we should use ⟨⋅, ⋅⟩X ⋆ ,X for
the latter and ⟨⋅, ⋅⟩X ⋆⋆ ,X ⋆ for the former, but this would make notation unnecessarily
heavy and we will often just use ⟨⋅, ⋅⟩ for both when there is no risk of confusion. For
instance, as x, x⋆ , x⋆⋆ will be used for denoting elements of X, X ⋆ , X ⋆⋆ , respectively, it
will be clear that ⟨x⋆ , x⟩ corresponds to the duality pairing between X and X ⋆ , whereas
⟨x⋆⋆ , x⋆ ⟩ corresponds to the duality pairing between X ⋆ and X ⋆⋆ .

Proposition 1.1.20 (The bidual as a normed space). Let X be a normed space. The bid-
ual space X ⋆⋆ of X, is a Banach space endowed with the norm

󵄩󵄩x 󵄩󵄩X ⋆⋆ = sup{⟨x , x ⟩X ⋆⋆ ,X ⋆ : 󵄩󵄩󵄩x 󵄩󵄩󵄩X ⋆ ≤ 1}.

󵄩󵄩 ⋆⋆ 󵄩󵄩 ⋆⋆ ⋆ 󵄩 ⋆󵄩

In the following example, we construct an important linear mapping from X to

X ⋆⋆
.

Example 1.1.21. Consider any x ∈ X. For any x⋆ ∈ X ⋆ , define the mapping x⋆ 󳨃→

⟨x⋆ , x⟩X ⋆ ,X , which is a mapping from X ⋆ to ℝ. Call this mapping fx : X ⋆ → ℝ. Since
|fx (x⋆ )| = |⟨x⋆ , x⟩| ≤ ‖x⋆ ‖X ⋆ ‖x‖X it is easy to see that fx is a bounded mapping, and
‖fx ‖ℒ(X ⋆ ,ℝ) ≤ ‖x‖X . In fact, ‖fx ‖ℒ(X ⋆ ,ℝ) = ‖x‖X . ◁

The above example introduces the following important concept. For any x ∈ X, the
mapping x⋆ 󳨃→ ⟨x⋆ , x⟩X ⋆ ,X , can be considered as a linear mapping fx : X ⋆ → ℝ, which
by the fundamental estimate |⟨x⋆ , x⟩X ⋆ ,X | ≤ ‖x⋆ ‖X ⋆ ‖x‖X is continuous, therefore, it is
a linear functional on X ⋆ hence, an element of (X ⋆ )⋆ = X ⋆⋆ . Note that by definition
fx (x⋆ ) = ⟨x⋆ , x⟩, and we may interpret fx (x⋆ ) as the duality pairing between X ⋆ and
X ⋆⋆ , so that fx (x⋆ ) = ⟨fx , x⋆ ⟩X ⋆⋆ ,X ⋆ ; hence, fx may be interpreted as this element of X ⋆⋆
for which it holds that

⟨fx , x⋆ ⟩X ⋆⋆ ,X ⋆ = ⟨x⋆ , x⟩X ⋆ ,X , ∀ x⋆ ∈ X ⋆ . (1.2)

For this mapping, it holds that ‖fx ‖X ⋆⋆ = ‖x‖X . Now define the map j : X → X ⋆⋆ by j(x) =
fx for any x ∈ X. Note that by (1.2) for the map j : X → X ⋆⋆ it holds that ⟨j(x), x⋆ ⟩X ⋆⋆ ,X ⋆ =
⟨x⋆ , x⟩X ⋆ ,X , for every x⋆ ∈ X ⋆ . Since ‖fx ‖X ⋆⋆ = ‖j(x)‖X ⋆⋆ = ‖x‖X , the mapping j is one to
one and conserves the norm. If we define Y := R(j) = {x⋆⋆ ∈ X ⋆⋆ : ∃ x ∈ X, j(x) = x⋆⋆ },
then obviously j : X → Y is onto and since it is an isometry we may identify X with
Y = R(j). Note that in general j : X → X ⋆⋆ is not onto. In terms of the mapping j : X →
X ⋆⋆ , the vector space X can be considered as a subspace of its bidual space X ⋆⋆ , and
for this reason the mapping is called the canonical embedding of X into X ⋆⋆ .
The above discussion leads to the following fundamental theorem.

Theorem 1.1.22 (Canonical embedding of X into X ⋆⋆ ). Let X be a normed space. Then

there exists a linear operator j : X → X ⋆⋆ , with ⟨j(x), x⋆ ⟩X ⋆⋆ ,X ⋆ = ⟨x⋆ , x⟩X ⋆ ,X , for all
8 | 1 Preliminaries

x⋆ ∈ X ⋆ , such that ‖j(x)‖X ⋆⋆ = ‖x‖X for all x ∈ X. This operator is called the canonical
embedding of X into X ⋆⋆ and is not in general surjective (onto), but allows the identifi-
cation of X with a closed subspace of X ⋆⋆ , in terms of an isomorphic isometry.

For certain types of normed spaces, the canonical embedding may be surjective
(onto).

Definition 1.1.23 (Reflexive space). If j : X → X ⋆⋆ is surjective, i. e., j(X) = X ⋆⋆ , then

the space X is called reflexive.

For a reflexive space, j is an isometry between X and X ⋆⋆ ; hence, X and X ⋆⋆ are

isomorphic isometric and can be identified in terms of the equivalent relation X ≃ X ⋆⋆ .
By the Riesz isometry, all Hilbert spaces are reflexive.

Example 1.1.24 (Biduals of sequence spaces). We may illustrate the above concepts
using the example of sequence spaces (see Example 1.1.15). Since in the case p ∈ (1, ∞),
it holds (ℓp )⋆ ≃ ℓp , for p1 + p1⋆ = 1, following the same arguments as in Example 1.1.15
⋆

for ℓp we see that (ℓp )⋆⋆ = (ℓp )⋆ ≃ ℓp with p1⋆ + p⋆⋆ 1

⋆ ⋆ ⋆⋆
= 1. One immediately sees that
in the case there p ∈ (1, ∞) we have that p = p, so that (ℓp )⋆⋆ ≃ ℓp and the spaces
⋆⋆

are reflexive. The situation is different in the cases where p = 1 and p = ∞. ◁

Reflexive Banach spaces satisfy the following fundamental properties.

Proposition 1.1.25 (Properties of reflexive spaces). Let X be a Banach space.

(i) If X is reflexive, every closed subspace is reflexive.
(ii) X is reflexive if and only if X ⋆ is reflexive.

1.1.4 Different choices of topology on a Banach space: strong and weak topologies

Often the topology induced by the norm on a Banach space X, the strong topology is
too strong. Let us clarify what we imply by that. Compactness, openess, closedness of
sets as well as continuity and semicontinuity of mappings are topological properties,
and whether they hold or not depends on the topology with which a set is endowed. As
the choice of topology changes the choice of open and closed sets, it is clear that a set
A ⊂ X may be open (closed) for a particular choice of topology on X and not enjoying
these properties for another choice. Similarly, whether or not a set A ⊂ X is compact
depends on whether any open covering of the set admits a finite subcover, and in turn
whether a cover is open or not depends on the choice of topology. Hence, a subset
A ⊂ X may be compact for a particular choice of topology while not being compact for
some other choice. Similarly, a mapping f : X → Y is called continuous is f −1 maps
open sets of Y to open sets of X and again whether a set is open or not depends on the
choice of topology; hence, a map may be continuous for some choice of topology but
not for some other choice. Finally, a mapping F : X → ℝ is lower semicontinuous if
 1.1 Fundamentals in the theory of Banach spaces | 9

the set {x ∈ X : F(x) ≤ c} is closed for any c ∈ ℝ, and clearly again this depends upon
the choice of topology.
As a rule of thumb, one can intuitively figure out that as we weaken a topology
(i. e., if we choose a topology whose collection of open sets is a subset of the collection
of open sets of the original one) then it is easier to make a given set compact; however,
at the same time it makes it more difficult for a mapping to be continuous. As an effect
of that, if we choose a topology on X which is weaker than the strong topology we
may manage to turn certain important subsets of X, which are not compact when the
strong topology is used (such as for instance the closed unit ball BX (0, 1) = BX (0, 1) =
{x ∈ X : ‖x‖X ≤ 1}), into compact sets. This need arises when we need to guarantee
some limiting behavior of sequences in such sets, as is often the case in optimization
problems.
Since one tries to juggle between compactness and continuity one way to define a
weaker topology is by choosing the coarser (weaker) topology that makes an important
family of mappings continuous. If this family of mappings is a family of seminorms
on X then this topology enjoys some nice properties, e. g., it is a locally convex topol-
ogy which is Hausdorfff (i. e., enjoys some nice separation properties). When trying
this approach with X, for the choice of the family of mappings defined by the duality
pairing we see that we may obtain a weak topology, the weak topology on X, which
provides important compactness properties, and furthermore complies extremely well
with the property of convexity. This approach can be further applied on the Banach
space X ⋆ , again for a family of mappings related to the duality pairing, and obtain a
weak topology on X ⋆ , called the weak⋆ topology, which again enjoys some very con-
venient properties with respect to compactness.

1.1.5 The strong topology on X

We start by recalling the strong (norm) topology on a Banach space X.

1.1.5.1 Definitions and properties

Definition 1.1.26 (Strong topology). Let X be a Banach space. The strong topology on
X is the topology generated by the norm.5

The strong topology on X has the following important property, which will lead
us to the necessity of endowing X with a different (weaker) topology.

Theorem 1.1.27. The closed unit ball BX (0, 1) := {x ∈ X : ‖x‖X ≤ 1} is compact if and
only if X is of finite dimension.

5 i. e., the topology of the metric generated by the norm.

10 | 1 Preliminaries

1.1.5.2 Convergence in the strong topology

Definition 1.1.28 (Strong convergence in X). A sequence {xn : n ∈ ℕ} ⊂ X converges
strongly to x, denoted by xn → x, if and only if limn→∞ ‖xn − x‖X = 0.

By the reverse triangle inequality for the norm, according to which | ‖xn ‖X −‖x‖X | ≤
‖xn −x‖X we see that the norm is continuous with respect to strong convergence, in the
sense that if xn → x then ‖xn ‖X → ‖x‖X as n → ∞.
The next result, connects completeness with the convergence of series in X.

Theorem 1.1.29. Let (X, ‖ ⋅ ‖X ) be a normed linear space. The space (X, ‖ ⋅ ‖X ) is complete
if and only if it has the following property: For every sequence {xn : n ∈ ℕ} ⊂ X such
that ∑n ‖xn ‖X < ∞ we have that ∑n xn converges strongly.6

1.1.5.3 The strong topology on X ⋆

Since X ⋆ can be considered as a Banach space when endowed with the norm ‖ ⋅ ‖X ⋆ ,
we can consider X ⋆ as a topological space endowed with the strong topology defined
by the norm ‖ ⋅ ‖X ⋆ . This topology enjoys the same properties as the strong topology on
X, so in complete analogy with Theorem 1.1.27 we have the following.

Theorem 1.1.30. The closed unit ball BX ⋆ (0, 1) = {x⋆ ∈ X ⋆ : ‖x⋆ ‖X ⋆ ≤ 1} is compact if
and only if X ⋆ is of finite dimension.

We may further consider the strong convergence in X ⋆ .

Definition 1.1.31 (Strong convergence in X ⋆ ). A sequence {x⋆n : n ∈ ℕ} ⊂ X ⋆ con-

verges strongly to x⋆ , denoted by x⋆n → x⋆ , if and only if limn→∞ ‖x⋆n − x⋆ ‖X ⋆ = 0.

Clearly, the analogue of Theorem 1.1.29 holds for the strong topology on X ⋆ .

1.1.6 The weak topology on X ⋆

⋆

The strong topology on X ⋆ is perhaps too strong as Theorem 1.1.30 shows. To regain
the compactness of BX ⋆ (0, 1) in infinite dimensional spaces, we need to endow X ⋆ with
a weaker topology, called the weak⋆ topology.

1.1.6.1 Definition and properties

Let X be a normed space, X ⋆ its dual, and ⟨⋅, ⋅⟩ the duality pairing between them.

Definition 1.1.32 (Weak⋆ topology on X ⋆ ). The topology generated on X ⋆ by the fam-

ily of mappings M ⋆ := {fx : X ⋆ → ℝ : x ∈ X} defined for each x ∈ X by fx (x⋆ ) = ⟨x⋆ , x⟩

6 i. e., the sequence {sn : n ∈ ℕ} ⊂ X defined by sn = ∑ni=1 xi converges strongly to some limit x ∈ X.
 1.1 Fundamentals in the theory of Banach spaces | 11

for every x⋆ ∈ X ⋆ is called the weak⋆ topology on X ⋆ and is the weaker topology on
X ⋆ for which all the mappings in the collection M ⋆ are continuous. The weak⋆ topol-
ogy will be denoted by σ(X ⋆ , X) and the space X ⋆ endowed with the weak⋆ topology
is denoted by Xw⋆⋆ .

For the construction of topologies by a set of mappings, see [28].

An alternative (and equivalent) way of defining the weak⋆ topology on X ⋆ is by
considering the collection of seminorms on X ⋆ defined by the collection of mappings
{px : X ⋆ → ℝ+ : x ∈ X} where px (x⋆ ) = |⟨x⋆ , x⟩|, and then defining the weak topology
on X as the topology generated by this family of seminorms, in the sense that it is the
weakest topology making these seminorms continuous. Recall that such topologies
are called locally convex topologies. The construction of a local basis for this topology
is presented in the next example.

Example 1.1.33 (A local basis for the weak⋆ topology). For any x⋆0 ∈ X ⋆ , fix {x1 , . . . ,
xn } ⊂ X, finite, ϵ > 0 and consider the sets of the form

n
𝒪 (x0 ; ϵ, x1 , . . . , xn ) := ⋂{x ∈ X : 󵄨󵄨󵄨⟨x − x0 , xi ⟩󵄨󵄨󵄨 < ϵ}, xi ∈ X,
⋆ ⋆ ⋆ 󵄨 ⋆ ⋆ 󵄨
i=1

called (x1 , . . . , xn ) semiball of radius ϵ centered at x⋆0 . The collection of sets

B⋆x⋆ := {𝒪⋆ (x⋆0 ; ϵ, x1 , . . . , xn ) : ∀ ϵ > 0, ∀ {x1 , . . . xn } ⊂ X, finite, n ∈ ℕ},

0

forms a basis of neighborhoods of x⋆0 for the weak⋆ topology σ(X ⋆ , X) and this helps
us to characterize weak⋆ open sets in X ⋆ . A subset 𝒪⋆ ⊂ X ⋆ is weak⋆ open if and
only if for every x⋆0 ∈ 𝒪⋆ there exists ϵ > 0 and x1 , . . . , xn ∈ X such that the semiball
𝒪⋆ (x⋆0 ; ϵ, x1 , . . . , xn ) ⊆ 𝒪⋆ . ◁

The weak⋆ topology on X ⋆ is a weaker topology than the strong topology on X ⋆ ,

so the same subsets of X ⋆ may have different topological properties depending on the
chosen topology. However, we have the following.

Example 1.1.34. A subset A ⊂ X ⋆ is bounded if and only if it is weak⋆ bounded.

If A is bounded (in the strong topology on X ⋆ ), then for every strong neighborhood
N0 there exists λ > 0 such that A ⊂ λN0 . To show that it is weak⋆ bounded, we need
to show the above for every weak⋆ neighborhood N0w , for a possibly different λ󸀠 > 0.
⋆

Consider any weak⋆ neighborhood of the origin N0w which by the fact that the weak⋆
⋆

topology on X ⋆ is weaker than the strong topology on X ⋆ is also a strong neighbor-

hood, so the result follows by the fact that A is bounded in the strong topology.
For the converse assume that A is weak⋆ bounded, i. e., for every weak⋆ neighbor-
hood N0w of the origin there exists λ󸀠 > 0 such that A ⊂ λ󸀠 N0w . Then, it holds that
⋆ ⋆

sup{|⟨x⋆ , x⟩| : x⋆ ∈ A} < ∞, for every x ∈ X, so by the Banach–Steinhaus theorem

sup{‖x⋆ ‖X ⋆ : x⋆ ∈ A} < ∞, therefore, A is bounded in the strong topology. ◁
12 | 1 Preliminaries

The weak⋆ topology has the following important properties (see, e. g., [28]).

Theorem 1.1.35. The weak⋆ topology, σ(X ⋆ , X), has the following properties:
(i) The space X ⋆ endowed with the weak⋆ topology is a Hausdorff 7 locally convex
space.
(ii) The weak⋆ topology is metrizable if and only if X is finite dimensional.
(iii) A normed space X is separable if and only if the weak⋆ topology restricted to the
closed unit ball BX ⋆ (0, 1) = {x⋆ ∈ X ⋆ : ‖x⋆ ‖X ⋆ ≤ 1} is metrizable.8
(iv) Every nonempty weak⋆ open subset A ⊂ X ⋆ is unbounded, in infinite dimensional
spaces.

Note that even though X ⋆ can be turned into a normed space using ‖ ⋅ ‖X ⋆ , when en-
dowed with the weak⋆ (σ(X ⋆ , X)) topology, it is no longer in general a metric space
on account of Theorem 1.1.35(ii). In this sense, assertion (iii) is very important since
metrizable topologies enjoy some convenient special properties resulting from the fact
that they admit a countable basis and are first countable. For example, it allows us to
use weak⋆ sequential compactness as equivalent to weak⋆ compactness for (norm)
bounded subsets of X ⋆ . However, caution is needed as the metrizability does not hold
over the whole of X ⋆ , thus requiring in general the use of nets to replace sequences
when dealing with the weak⋆ topology in general.
On the other hand, the weak⋆ topology displays some remarkable compactness
properties, known as the Alaoglu (or Alaoglu–Banach–Bourbaki) theorem, a fact that
makes the use of this topology indispensable.

Theorem 1.1.36 (Alaoglu). The closed unit ball in X ⋆ , BX ⋆ (0, 1) := {x⋆ ∈ X ⋆ : ‖x⋆ ‖X ⋆ ≤
1} is weak⋆ compact (i. e., compact for the σ(X ⋆ , X) topology).9 If furthermore, X is sep-
arable, then every bounded sequence in X ⋆ has a weak⋆ convergent subsequence.

Example 1.1.37 (Weak compactness of the duality map). For any x ∈ X, the duality
map J(x) (recall Definition 1.1.16) is a bounded and weak⋆ compact set in X ⋆ .
This follows since for any fixed x⋆ ∈ J(x), it holds that ‖x⋆ ‖X ⋆ ≤ ‖x‖X , so that J(x) is
a bounded set. The weak⋆ compactness then follows by Alaoglu’s theorem (see Theo-
rem 1.1.36). ◁

Even though, nets provide a complete characterization of the weak⋆ topology,

there are still important properties that may be captured using sequences.

7 i. e., a topological space such that any two distinct points have distinct neighborhoods.
8 Clearly, this result extends to any norm bounded and closed (with respect to the strong topology on
X ⋆ ) set of the form A = {x⋆ ∈ X ⋆ : ‖x⋆ ‖X ⋆ ≤ c}, for some c ∈ (0, ∞). Note that the weak⋆ topology is
never metrizable on the whole of X ⋆ .
9 Hence, also the set cBX ⋆ (0, 1) = {x⋆ ∈ X ⋆ : ‖x⋆ ‖X ⋆ ≤ c}. This result also implies that any bounded
subset of X ⋆ is relatively weak⋆ compact in X ⋆ , so that bounded and weak⋆ closed subsets of X ⋆ are
weak⋆ compact.
 1.1 Fundamentals in the theory of Banach spaces | 13

1.1.6.2 Convergence in the weak topology

⋆

Definition 1.1.38 (Weak⋆ convergence). A sequence {x⋆n : n ∈ ℕ} ⊂ X ⋆ converges

weak⋆ to x⋆ , denoted by x⋆n ⇀ x⋆ , if it converges with respect to the weak⋆ topology
⋆

σ(X ⋆ , X), i. e., if and only if limn→∞ ⟨x⋆n , x⟩ = ⟨x⋆ , x⟩ for every x ∈ X.

One may similarly define weak⋆ convergence for a net and show that a net {x⋆α :
α ∈ ℐ } ⊂ X ⋆ converges weak⋆ to x⋆ if and only if ⟨x⋆α , x⟩ → ⟨x⋆ , x⟩ for every x ∈ X.
The weak⋆ convergence has the following useful properties.

Proposition 1.1.39. Consider the sequence {x⋆n : n ∈ ℕ} ⊂ X ⋆ . Then:

(i) If x⋆n → x⋆ then x⋆n ⇀ x⋆ .
⋆

(ii) If x⋆n ⇀ x⋆ , then the sequence {‖x⋆n ‖X ⋆ : n ∈ ℕ} ⊂ ℝ is bounded and ‖x⋆ ‖X ⋆ ≤

⋆

lim infn ‖x⋆n ‖X ⋆ , i. e., the norm ‖ ⋅ ‖X ⋆ is sequentially lower semicontinuous with re-
spect to weak⋆ convergence.
(iii) If x⋆n ⇀ x⋆ and xn → x, then ⟨x⋆n , xn ⟩ → ⟨x⋆ , x⟩.
⋆

Property (i) follows easily from the fact that |⟨x⋆n − x⋆ , x⟩| ≤ ‖x⋆n − x⋆ ‖X ⋆ ‖x‖X . Property
(iii) follows by a rearrangement of the difference |⟨x⋆n , xn ⟩ − ⟨x⋆ , x⟩| = |⟨x⋆n , xn − x⟩ +
⟨x⋆n − x⋆ , x⟩| and fact that the weak⋆ convergent sequence {x⋆n : n ∈ ℕ} ⊂ X ⋆ is norm
bounded, which in turn comes as a consequence of the Banach–Steinhaus theorem
(see Theorem 1.1.7). Property (ii) can be seen in a number of ways, we provide one of
them in the following example.

Example 1.1.40 (Weak⋆ (sequential) lower semicontinuity of the norm ‖ ⋅ ‖X ⋆ ).

The function x⋆ 󳨃→ ‖x⋆ ‖X ⋆ is weak⋆ lower semicontinuous and weak⋆ sequentially
lower semicontinuous.
The weak⋆ lower semicontinuity of the norm ‖ ⋅ ‖X ⋆ follows from the fact that the
functions fx : X ⋆ → ℝ, defined by fx (x⋆ ) = ⟨x⋆ , x⟩ for every x⋆ ∈ X ⋆ are (by defini-
tion) continuous with respect to the weak⋆ topology, and since ‖x⋆ ‖X ⋆ = sup{|⟨x⋆ , x⟩| :
‖x‖X ≤ 1} = sup{|fx (x⋆ )| : ‖x‖X ≤ 1}, we see that the function ‖ ⋅ ‖X ⋆ is the (pointwise)
supremum of a family of continuous functions; hence, it is semicontinuous. The lat-
ter can be easily seen as follows: if f (x) = supα∈ℐ fα (x), with fα continuous (or lower
semicontinuous) then {x ∈ X : f (x) ≤ c} = ⋂α∈ℐ {x ∈ X : fα (x) ≤ c} which is a closed
set as the intersection of the closed sets {x ∈ X : f (x) ≤ c}.
The weak⋆ sequential lower semicontinuity can follow by another application of
the Banach–Steinhaus Theorem 1.1.7, working as in Example 1.1.8, for the sequence
of operators {An : n ∈ ℕ} ⊂ ℒ(X, ℝ) defined by An (x) = ⟨x⋆n , x⟩ for any x ∈ X, which
converges pointwise to the operator A ∈ ℒ(X, ℝ) defined by A(x) = ⟨x⋆ , x⟩ for every
x ∈ X. ◁

Example 1.1.41 (Does every norm bounded sequence in X ⋆ have a weak⋆ convergent
subsequence?). Can we claim that if {x⋆n : n ∈ ℕ} ⊂ X ⋆ satisfies supn∈ℕ ‖x⋆n ‖X ⋆ < c,
14 | 1 Preliminaries

for some c > 0 independent of n, there exists x⋆o ∈ X ⋆ and a subsequence {x⋆nk : k ∈ ℕ}
such that x⋆nk ⇀ x⋆o as k → ∞? The answer is no, unless X is separable.
⋆

Note that since in general (see Theorem 1.1.35) the weak⋆ topology is not metriz-
able, Alaoglu’s theorem which guarantees weak⋆ compactness of (norm) bounded
subsets of X ⋆ , does not guarantee weak⋆ sequential compactness, i. e., that any se-
quence {x⋆n : n ∈ ℕ} ⊂ X ⋆ , such that there exists a constant c > 0 (independent of
n) with the property ‖x⋆n ‖X ⋆ ≤ c, for every n ∈ ℕ, has a convergent subsequence in
the sense of convergence in the weak⋆ topology on X ⋆ . This will only be true when X
is separable, in which case the weak⋆ topology restricted on this bounded set will be
metrizable. ◁

1.1.7 The weak topology on X

The strong topology on X is perhaps too strong as Theorem 1.1.27 shows. To regain the
compactness of the closed unit ball BX (0, 1) in infinite dimensional spaces, we need
to endow X with a weaker topology, called the weak topology.

1.1.7.1 Definitions and properties

Let X be a normed space, and X ⋆ its dual.

Definition 1.1.42 (Weak topology on X). The topology generated on X by the family of
mappings M := {fx⋆ : X → ℝ : x⋆ ∈ X ⋆ } = {⟨x⋆ , ⋅⟩ : X → ℝ : x⋆ ∈ X ⋆ } is called the
weak topology on X and is the weaker topology on X for which all the mappings in the
collection M are continuous. The weak topology will be denoted by σ(X, X ⋆ ), and the
space X endowed with the weak topology will be denoted by Xw .

For the construction of topologies by a set of mappings, see [28].

An alternative (and equivalent) way to define the weak topology on X is by con-
sidering the collection of seminorms on X defined by the collection of mappings
{px⋆ : X → ℝ+ : x⋆ ∈ X ⋆ } where px⋆ (x) := |⟨x⋆ , x⟩|, and then defining the weak
topology on X as the topology generated by this family of seminorms, in the sense
that it is the weakest topology making these seminorms continuous. Recall that such
topologies are called locally convex topologies. The construction of a local basis for
this topology is presented in the next example.

Example 1.1.43 (A local basis for the weak topology). For any x0 ∈ X, fix
{x⋆1 , . . . , x⋆n } ⊂ X ⋆ , finite, and ϵ > 0 and consider the sets of the form

n
𝒪(x0 ; ϵ, x1 , . . . , xn ) := ⋂{x ∈ X : 󵄨󵄨󵄨⟨xi , x − x0 ⟩X ⋆ ,X 󵄨󵄨󵄨 < ϵ, xi ∈ X },
⋆ ⋆ 󵄨 ⋆ 󵄨 ⋆ ⋆

i=1
 1.1 Fundamentals in the theory of Banach spaces | 15

called (x⋆1 , . . . , x⋆n ) semiball of radius ϵ centered at x0 . The collection of semiballs

Bx0 := {𝒪(x0 ; ϵ, x⋆1 , . . . , x⋆n ) : ∀ ϵ > 0, ∀ {x⋆1 , . . . x⋆n } ⊂ X ⋆ , finite, n ∈ ℕ},

forms a basis of neighborhoods of x0 for the weak topology σ(X, X ⋆ ), and this helps us
characterize the open sets for the weak topology on X (or weakly open sets). A subset
𝒪 ⊂ X is weakly open if and only if for every x0 ∈ 𝒪 there exists ϵ > 0 and x⋆1 , . . . , x⋆n ∈
X ⋆ such that the semiball 𝒪(x0 ; ϵ, x⋆1 , . . . , x⋆n ) ⊂ 𝒪. ◁

Example 1.1.44. As a simple geometrical illustration of how the semiballs may look
like, consider the case where X = ℝ2 , x0 = (0, 0) and x⋆1 = (1, 1), x⋆2 = (1, −1). Then

𝒪(x0 ; ϵ, x1 ) = {(x1 , x2 ) : |x1 + x2 | ≤ ϵ}, and

⋆

𝒪(x0 ; ϵ, x1 , x2 ) = {(x1 , x2 ) : |x1 + x2 | ≤ ϵ, |x1 − x2 | ≤ ϵ},

⋆ ⋆

which look like strips and intersection of strips, respectively. However, this geometri-
cal picture which is valid in finite dimensional spaces may be misleading in infinite
dimensions (see, e. g., Theorem 1.1.45(v)). ◁

The following theorem (see, e. g., [28]) collects some fundamental results for the
weak topology.

Theorem 1.1.45. The following are true for the weak topology σ(X, X ⋆ ):
(i) The space X endowed with the weak topology σ(X, X ⋆ ) is a Hausdorff locally convex
space.
(ii) The weak topology is weaker than the norm topology, i. e., every weakly open
(closed) set is strongly open (closed). The weak and the strong topology coincide if
and only if X is finite dimensional.
(iii) The weak topology is metrizable10 if and only if X is finite dimensional.
(iv) X ⋆ is separable if and only if the weak topology σ(X, X ⋆ ) on X restricted on the closed
unit ball BX (0, 1) = {x ∈ X : ‖x‖X ≤ 1} is metrizable. Clearly, the result extends to
any closed norm bounded A ⊂ X.
(v) If X is infinite dimensional then for any xo ∈ X, every weak neighborhood N w (xo )
contains an affine space passing through xo .

Assertion (iii) indicates that we have to be cautious with the weak topology since it
will not enjoy some of the convenient special properties valid for metrizable or first
countable topologies. For instance, it is not necessarily true that compactness coin-
cides with sequential compactness, and in general the weak closure of a set is ex-
pected to be a larger set than the set of limits of all weakly converging sequences in
the set. In this respect, Assertion (iv) is very important since it allows us to reinstate

10 That is, there exists a metric that induces that weak topology.
16 | 1 Preliminaries

these convenient characterizations in terms of sequences for the weak topology in spe-
cial cases. For example, it allows us to use sequential compactness as equivalent to
compactness for bounded subsets of X in the weak topology, or it allows the charac-
terization of weak closures of sets in terms of limits of sequences. However, caution
is needed as the metrizability does not hold over the whole of X, thus requiring in
general the use of nets to replace sequences when dealing with the weak topology in
general. An interesting question is what kind of Banach spaces X satisfy the condition
that X ⋆ is separable? This would always hold true for instance if X is separable and
reflexive, as in such a case X ⋆ is separable. A further example of nonreflexive X with
separable X ⋆ is the case where X = c0 and X ⋆ = ℓ1 (see, e. g., [48]).
The weak topology on X is weaker than the strong topology on X so concepts like
closed and open sets, compact sets, etc. vary with respect the topology chosen on X.
The following example shows that as far as the concept of boundedness is concerned
the choice of topology does not make a difference.

Example 1.1.46 (Weakly bounded set vs. bounded set). A subset A ⊂ X is weakly
bounded if and only if it is bounded.11
It is easy to see that if A is bounded then it is also weakly bounded. Since A is
bounded for every (strong) neighborhood N0 there exists λ > 0 such that A ⊂ λN0 . To
show that A is weakly bounded, we must show that for every weak neighborhood N0w
there exists λ󸀠 > 0 such that A ⊂ λ󸀠 N0w . However, any weak neighborhood N0w is also a
strong neighborhood so the required property holds for λ󸀠 = λ.
For the converse, the argument is slightly more involved. Assume that A is weakly
bounded so that for every weak neighborhood N0w there exists λ󸀠 > 0 such that
A ⊂ λ󸀠 N0w . This implies that the family of linear maps 𝒜 := {fx : X ⋆ → ℝ : x ∈ A},
defined by fx (x⋆ ) := ⟨x⋆ , x⟩ for every x⋆ ∈ X ⋆ , is pointwise bounded. By the Banach–
Steinhaus Theorem 1.1.7 this is also norm bounded, i. e., the elements of 𝒜 are
bounded when considered as elements of ℒ(X ⋆ , ℝ) which coincides with X ⋆⋆ . But
‖fx ‖ℒ(X ⋆ ,ℝ) = ‖x‖X (see Example 1.1.21) which means that A is norm bounded; hence, it
is strongly bounded. ◁

However, certain topological properties are significantly different if we change

topologies on X. Clearly, since the weak topology is weaker than the strong (norm)
topology, any weakly open set is open, however, as the following example shows the
converse is not true. Similarly, for closed sets (unless these are convex where in this
case the concept of weak and strong closedness coincide, as we will prove in Proposi-
tion 1.2.12).

Example 1.1.47. In infinite dimensions, the open unit ball BX (0, 1) = {x ∈ X : ‖x‖X < 1}
has empty weak interior; hence, is not open in the weak topology. In fact (see Theo-

11 In fact, this result holds in a more general setting than that of Banach spaces and is known as the
Mackey theorem (see, e. g., Theorem 6.24 in [5]).
 1.1 Fundamentals in the theory of Banach spaces | 17

rem 1.1.45(v)), every nonempty weakly open subset is unbounded in infinite dimen-
sions. Note, however, that the closed unit ball BX (0, 1) = {x ∈ X : ‖x‖X ≤ 1} is also
weakly closed (as a result of convexity).
Suppose not, and consider a point xo in the weak interior of BX (0, 1). Then there
exist ϵ > 0, n ∈ ℕ and x⋆1 , . . . , x⋆n ∈ X ⋆ such that 𝒪(xo ; ϵ, x⋆1 , . . . , x⋆n ) ⊂ BX (0, 1). We
claim that we may obtain a point zo ∈ X \ {0}, such that ⟨x⋆i , zo ⟩ = 0 for all i = 1, . . . , n.
Indeed, if not ⋂ni=1 N(x⋆i ) = {0}, where by N(x⋆i ) we denote the kernel of the functional
x⋆i considered as a linear mapping x⋆i : X → ℝ, and since for any x⋆ ∈ X ⋆ , {0} ⊂ N(x⋆ )
we conclude that ⋂ni=1 N(x⋆i ) ⊂ N(x⋆ ), therefore using a well-known result from linear
algebra12 x⋆ ∈ span{x⋆1 , . . . , x⋆n } which contradicts the assumption that dim(X) = ∞.
Then, for every λ ∈ ℝ we have that xo + λzo ∈ 𝒪(x0 ; ϵ, x⋆1 , . . . , x⋆n ) ⊂ BX (0, 1), which
implies that ‖xo + λzo ‖X ≤ 1 which is clearly a contradiction. ◁

Example 1.1.48 (The closure of the unit sphere). Assume that dim(X) = ∞ and let
w
SX (0, 1) = {x ∈ X : ‖x‖X = 1} and BX (0, 1) = {x ∈ X : ‖x‖X ≤ 1}. Then SX (0, 1) =
BX (0, 1). In general, not every element of BX (0, 1) can be expressed as the limit of a
weakly convergent sequence in SX (0, 1), except in special cases. Such is, e. g., the case
where X = ℓ1 on account of the Schur property that states the equivalence of the weak
and the strong convergence.
To check this, if we take for the time being for granted that BX (0, 1) is a weakly
w
closed set (for a proof see Proposition 1.2.12) so that SX (0, 1) ⊂ BX (0, 1), since SX (0, 1) ⊂
w
BX (0, 1) and SX (0, 1) is the smallest weakly closed set containing SX (0, 1). To prove the
reverse inclusion, consider any xo ∈ BX (0, 1) and take a weak (open) neighborhood
N w (xo ). Then, by the construction of the local basis for the weak topology (see Exam-
ple 1.1.43) there exists x⋆1 , . . . , x⋆n ∈ X ⋆ and ϵ > 0 such that 𝒪(xo ; ϵ, x⋆1 , . . . , x⋆n ) := {x ∈ X :
|⟨x⋆i , x − xo ⟩| < ϵ, i = 1, . . . , n} ⊂ N w (xo ). With the same arguments as in Example 1.1.47,
there exists nonzero zo ∈ X such that ⟨x⋆i , zo ⟩ = 0 for all i = 1, . . . , n, so that for any
λ ∈ ℝ, xo + λzo ∈ 𝒪(xo ; ϵ, x⋆1 , . . . , x⋆n ) ⊂ N w (xo ). Since ‖xo ‖X ≤ 1, there exists λo ∈ ℝ such
that ‖xo +λo zo ‖X = 1, therefore, xo +λo zo ∈ 𝒪(xo ; ϵ, x⋆1 , . . . , x⋆n )∩SX (0, 1) ⊂ N w (xo )∩SX (0, 1)
so that N w (xo ) ∩ SX (0, 1) ≠ 0. Since for any xo ∈ BX (0, 1) and any weak neighborhood
w
N w (xo ) of xo , we have that N w (xo ) ∩ SX (0, 1) ≠ 0 we see that xo ∈ SX (0, 1) , so that since
w
xo was arbitrary we conclude that BX (0, 1) ⊂ SX (0, 1) , and the proof is complete. ◁

On the other hand, when it comes to the question of continuity of linear operators
the choice of topology does not make too much difference ([28]).

12 If on a linear space for the linear maps f , f1 , . . . , fn : X → ℝ, it holds that ⋂ni=1 N(fi ) ⊂ N(f ) then f is
a linear combination of the f1 , . . . , fn . This can be proved by induction. It is easy to see that the claim is
true for n = 1. Assuming it holds for n, if ⋂n+1 n
i=1 N(fi ) ⊂ N(f ) then it holds that ⋂i=1 N(ϕi ) ⊂ N(ϕ) where
ϕi = fi |N(fn+1 ) , ϕ = f |N(fn+1 ) , i = 1, . . . , n so applying the result at level n to these functions we have that
ϕ ∈ span{ϕ1 , . . . , ϕn }, i. e., for some λ1 , . . . , λn ∈ ℝ we have that N(fn+1 ) ⊂ N(f − ∑ni=1 λi fi ) and applying
the result for n = 1 we conclude the existence of λn+1 such that f = ∑n+1 i=1 λi fi , concluding the proof.
18 | 1 Preliminaries

Theorem 1.1.49. A linear operator A : X → Y is continuous with respect to the weak

topologies on X and Y (weakly continuous) if and only if it is continuous with respect
to the strong topologies on X and Y (strong or norm continuous), the same holding of
course for linear functionals (Y = ℝ).

Even though, nets provide a complete characterization of the weak topology, there
are still important properties that may be captured using sequences.

1.1.7.2 Convergence in the weak topology

Definition 1.1.50 (Weak convergence). A sequence {xn : n ∈ ℕ} ⊂ X converges weakly
to x, denoted by xn ⇀ x, if it converges with respect to the weak topology σ(X, X ⋆ ),
i. e., if and only if

lim ⟨x⋆ , xn ⟩X ⋆ ,X = ⟨x⋆ , x⟩X ⋆ ,X , ∀ x⋆ ∈ X ⋆ .

n→∞

One may similarly define weak convergence for a net {xα : α ∈ ℐ } ⊂ X and show
that the net converges weakly to x if and only if ⟨x⋆ , xα ⟩X ⋆ ,X → ⟨x⋆ , x⟩X ⋆ ,X for every
x⋆ ∈ X ⋆ .

Remark 1.1.51 (The Urysohn property). Since X w := (X, σ(X, X ⋆ )) is a Hausdorff topo-
logical space, we have uniqueness of weak limits. A sequence weakly converges to
a point if each subsequence contains a further subsequence which converges to this
point (Urysohn property). The same property naturally holds for the strong topology.

Example 1.1.52. Show that the weak convergence enjoys the Urysohn property. In-
deed assume by contradiction that every subsequence of {xn : n ∈ ℕ} ⊂ X has a
further subsequence that converges weakly to x but xn ⇀̸ x in X. Then there exists
ϵ > 0, some x⋆i ∈ X ⋆ , i = 1, . . . , m, and a subsequence {xnk : k ∈ ℕ} with the property
|⟨x⋆i , xnk ⟩ − ⟨xi ⋆ , x⟩| ≥ ϵ, i = 1, . . . , m, for every k. However, by assumption this subse-
quence has a further subsequence {xnk : ℓ ∈ ℕ} such that xnk ⇀ x, which clearly is a
ℓ ℓ
contradiction. ◁

Proposition 1.1.53 (Properties of weak convergence). Let {xn : n ∈ ℕ} be a sequence

in X. Then:
(i) If xn → x, then xn ⇀ x.
(ii) If xn ⇀ x, then the sequence xn is bounded (i. e., supn∈ℕ ‖xn ‖X < ∞), and
‖x‖X ≤ lim infn ‖xn ‖X , i. e., the norm is sequentially weakly lower semicontinuous.
(iii) If xn ⇀ x and x⋆n → x⋆ in X ⋆ , then ⟨x⋆n , xn ⟩ → ⟨x⋆ , x⟩.

The first claim is immediate from the fact that the weak topology is weaker (coarser)
than the strong topology. The second claim may be proved in various ways; we present
one possible proof in Example 1.1.54 below. The third claim follows easily by a rear-
rangement of ⟨x⋆n , xn ⟩−⟨x⋆ , x⟩ = ⟨x⋆n −x⋆ , xn ⟩+⟨x⋆ , xn −x⟩, using the norm boundedness
of the sequence {‖xn ‖X : n ∈ ℕ} ⊂ ℝ.
 1.1 Fundamentals in the theory of Banach spaces | 19

Example 1.1.54 (Weak (sequential) lower semicontinuity of the norm). The function
x 󳨃→ ‖x‖X is (sequentially) lower semicontinuous for the weak topology.
The proof of this claim (stated in Proposition 1.1.53(ii)) follows by the observation
that the norm can be represented as ‖x‖X = sup{|⟨x⋆ , x⟩| : x⋆ ∈ X ⋆ , ‖x⋆ ‖X ⋆ ≤ 1} (see
Example 1.1.17) and then using the same reasoning as in Example 1.1.40, based on the
remark that in the weak topology, for any x⋆ ∈ X ⋆ fixed, the functions fx⋆ : X → ℝ,
defined by fx⋆ (x) = ⟨x⋆ , x⟩ are continuous and the supremum of a family of continuous
functions is lower semicontinuous (see the argument in Example 1.1.40), we obtain the
weak lower semicontinuity of the norm. For the sequential weak lower semicontinuity,
we may reason once more as in Example 1.1.40, reversing the roles of x and x⋆ . ◁

Example 1.1.55 (If xn ⇀ x and x⋆n ⇀ x⋆ , it may be that ⟨x⋆n , xn ⟩ ↛ ⟨x⋆ , x⟩). Consider the
⋆

space X = ℓp , p ∈ (0, ∞), with dual space X ⋆ = ℓp , with p1 + p1⋆ = 1, and bidual
⋆

X ⋆⋆ = (ℓp )⋆ = ℓp . Let en = {δin : i ∈ ℕ}, which may be considered either as an

⋆

element of X, X ⋆ or X ⋆⋆ . Consider the sequences {xn : n ∈ ℕ} = {en : n ∈ ℕ} ⊂ X and

{x⋆n : n ∈ ℕ} = {en : n ∈ ℕ} ⊂ X ⋆ . One may easily confirm that xn ⇀ 0, and x⋆n ⇀ 0.
⋆

p⋆
Indeed, considering any x ∈ X , of the form x = {xi : i ∈ ℕ, ∑i∈ℕ |xi | < ∞}, so
⋆ ⋆

that ⟨x⋆ , xn ⟩ = xn → 0 as n → ∞, since ∑i∈ℕ |xi |p < ∞, with a similar treatment for
⋆

x⋆n ⇀ 0. However, ⟨x⋆n , xn ⟩ = 1 for every n ∈ ℕ, so that ⟨x⋆n , xn ⟩ ↛ 0. ◁

Even though, the weak topology is not metrizable in general, the celebrated
Eberlein–Smulian theorem (see, e. g., [4] or [82]) shows that for the weak topology the
concepts of sequential compactness and compactness are equivalent as for metrizable
topologies.

Theorem 1.1.56 (Eberlein–Smulian). Let X be a Banach space and consider A ⊂ X. The

following are equivalent:
(i) A is compact in the weak topology σ(X, X ⋆ ) (weakly compact).
(ii) A is sequentially compact in the weak topology σ(X, X ⋆ ) (weakly sequentially com-
pact).

1.1.7.3 The weak topology for reflexive spaces

For the particular case of reflexive Banach spaces X endowed with the weak topology
one may obtain some important information, which essentially follows by the impor-
tant compactness properties of the weak⋆ topology on X ⋆ .
Before proceeding, we present in the next example the possible connections of
the weak⋆ topology on X ⋆ with the weak topology on X.

Example 1.1.57 (The weak topology on X ⋆ ). Since X ⋆ is a normed space, with dual
X ⋆⋆ , what would happen if we tried to repeat the construction of Definition 1.1.42 for
X ⋆ ? Replacing X by X ⋆ and X ⋆ by X ⋆⋆ in this construction we create thus the topology
σ(X ⋆ , X ⋆⋆ ). This will be another topology on X ⋆ , called the weak topology on X ⋆ . How
does this topology compare with the weak⋆ topology σ(X ⋆ , X) on X ⋆ ?
20 | 1 Preliminaries

In general, we may consider that X ⊂ X ⋆⋆ (in the sense of the canonical embedding
of Theorem 1.1.22) so that the weak⋆ topology σ(X ⋆ , X) is weaker than the weak topol-
ogy on X ⋆ constructed here σ(X ⋆ , X ⋆⋆ ). However, if X is reflexive, then the canonical
embedding is onto and X ≃ X ⋆⋆ , so that the topologies σ(X ⋆ , X) and σ(X ⋆ , X ⋆⋆ ) coin-
cide. This means that in reflexive spaces, we may consider the weak topology on X ⋆
as equivalent to the weak⋆ topology on X ⋆ , thus enjoying the important compactness
properties of the weak⋆ topology ensured by Alaoglu’s theorem. ◁

After the above preparation we may state the following result (see, e. g., [28]) or
[82]) which provides some important information concerning compact sets in this
topology, called weakly compact sets.

Theorem 1.1.58. Let X be a Banach space. The following statements are equivalent:
(i) X is reflexive.
(ii) The closed unit ball of X, BX (0, 1) = {x ∈ X : ‖x‖X ≤ 1} is weakly compact.
(ii󸀠 ) The closed unit ball of X ⋆ , BX ⋆ (0, 1) = {x⋆ ∈ X ⋆ : ‖x⋆ ‖X ⋆ ≤ 1} is compact when X ⋆ is
endowed with the weak topology.
(iii) Every bounded sequence {xn : n ∈ ℕ} ⊂ X has a subsequence which converges for
the topology σ(X, X ⋆ ) (a weakly converging subsequence), i. e., there exists a subse-
quence {xnk : k ∈ ℕ} and an element xo ∈ X such that xnk ⇀ xo in X.
(iv) For every x⋆ ∈ X ⋆ , the supremum in the definition of the dual norm is attained

󵄩󵄩x 󵄩󵄩X ⋆ = max{⟨x , x⟩ : ‖x‖X ≤ 1}.

󵄩󵄩 ⋆ 󵄩󵄩 ⋆

This theorem is very useful as it allows us to extract a weakly convergent subsequence

out of every bounded sequence in a reflexive Banach space, and so it will be used
very often throughout this book. It plays in reflexive Banach spaces the role that the
Bolzano–Weierstrass theorem plays in ℝ. To see how reflexivity leads to weak com-
pactness of the closed unit ball, we may follow the reasoning of the next example.

Example 1.1.59 (Weak compactness of the closed unit ball in reflexive spaces). For a
reflexive space, we know that the canonical embedding j : X → X ⋆⋆ is onto, so that
X ≃ X ⋆⋆ , and we can identify X = (X ⋆ )⋆ . By Proposition 1.1.25, since X is reflexive so
is X ⋆ ; hence, by the reasoning of Example 1.1.57, the weak⋆ topology on (X ⋆ )⋆ = X
coincides with the weak topology on (X ⋆ )⋆ = X. By Alaoglu’s theorem (see Theo-
rem 1.1.36), the (norm) closed unit ball BX (0, 1) of (X ⋆ )⋆ = X is weak⋆ compact, but
since (by reflexivity) the weak⋆ and the weak topology on (X ⋆ )⋆ = X coincide, BX (0, 1)
is also weakly compact. But then the Eberlein–Smulian theorem (see Theorem 1.1.56)
guarantees weak sequential compactness from which we can conclude that every se-
quence in BX (0, 1); hence, any (norm) bounded sequence admits a weakly convergent
subsequence. ◁
 1.1 Fundamentals in the theory of Banach spaces | 21

Example 1.1.60 (The supremum in the definition of the norm is attained for reflexive
spaces). We will show that if X is reflexive then for any x⋆ ∈ X ⋆ fixed, there exists
xo ∈ X such that ‖x⋆ ‖X ⋆ = ⟨x⋆ , xo ⟩ with ‖xo ‖X ≤ 1.
By the reflexivity of X, it holds that the (norm) closed unit ball BX (0, 1) is weakly
compact, hence from Eberlein–Smulian theorem (see Theorem 1.1.56) weakly sequen-
tially compact. By definition, ‖x⋆ ‖X ⋆ = sup{⟨x⋆ , x⟩ : x ∈ X, ‖x‖X ≤ 1}, so we construct
a sequence {xn : n ∈ X} ⊂ B, with the property13 ⟨x⋆ , xn ⟩ → ‖x⋆ ‖X ⋆ . The weak sequen-
tial compactness of BX (0, 1) implies the existence of a subsequence {xnk : k ∈ ℕ} of
the above sequence and a xo ∈ BX (0, 1) such that xnk ⇀ xo as k → ∞. This implies that
⟨x⋆ , xnk ⟩X ⋆ ,X → ⟨x⋆ , xo ⟩. On the other hand, since the whole sequence {xn : n ∈ ℕ}
satisfies ⟨x⋆ , xn ⟩ → ‖x⋆ ‖X ⋆ , the same holds for the weakly converging subsequence
{xnk : k ∈ ℕ}, and we easily conclude that ‖x⋆ ‖X ⋆ = ⟨x⋆ , xo ⟩. The converse statement
is known as James theorem and its proof is more subtle (see, e. g., [82]). The reasoning
used here will be used repeatedly in this book for providing results concerning the
maximization of functionals in reflexive Banach spaces. ◁

The following result (see, e. g., [51] or [108]) will be used frequently throughout
this book.
w
Proposition 1.1.61. Let X be a reflexive Banach space, A ⊂ X, bounded and xo ∈ A ,
w
where by A we denote the weak closure of A. Then there exists a sequence {xn : n ∈
ℕ} ⊂ A which is weakly convergent to xo in X.
w
Proof. Let us fix the point xo ∈ A and let us denote by BX ⋆ (0, 1) = {x⋆ ∈ X ⋆ : ‖x⋆ ‖X ⋆ ≤
1} the closed unit ball of X ⋆ .
w
First, we will show the existence of a countable subset A0 ⊂ A such that xo ∈ A0 .
w
To this end, we claim that for the fixed xo ∈ A , and for any choice of integers n, m,
and every (x⋆1 , . . . , x⋆m ) ∈ (BX ⋆ (0, 1))m = BX ⋆ (0, 1) × ⋅ ⋅ ⋅ × BX ⋆ (0, 1), we can find x ∈ A
w
such that |⟨x⋆k , x − xo ⟩| < n1 for all k = 1, . . . , m. Indeed, since xo ∈ A , by definition,
for any weak neighborhood N w (xo ) it holds that N w (xo ) ∩ A ≠ 0; hence, recalling the
construction of weak neighborhoods the claim holds. This can be rephrased as that
for any (x⋆1 , . . . , x⋆m ) ∈ (BX ⋆ (0, 1))m it holds that (x⋆1 , . . . , x⋆m ) ∈ ⋃x∈A Vn,m (x), where

m 󵄨 1
Vn,m (x) = {(x⋆1 , x⋆2 , . . . , x⋆m ) ∈ (BX ⋆ (0, 1)⋆ ) : 󵄨󵄨󵄨⟨x⋆k , x − xo ⟩󵄨󵄨󵄨 < , k = 1, . . . , m}
󵄨
n
m
⊂ (BX ⋆ (0, 1)) ,

which are clearly open sets (for the weak topology). By this rephrasal, we conclude
that {Vn,m (x) : x ∈ A} constitutes an open cover for (BX ⋆ (0, 1))⋆ , which is weakly com-

13 The construction of this sequence follows by elementary means from the definition of the supre-
mum as the least upper bound; for any n ∈ ℕ, ‖x⋆ ‖X − n1 is not an upper bound of the quantity ⟨x⋆ , x⟩
for x ∈ B, which implies the existence of an xn ∈ B such that ‖x⋆ ‖X − n1 < ⟨x⋆ , xn ⟩ ≤ ‖x⋆ ‖X which has
the required properties.
22 | 1 Preliminaries

pact by Alaoglu’s theorem; hence, it must admit a finite subcover. Therefore, there
exists a finite subset of A, which we denote by Sn,m ⊂ A, such that (BX ⋆ (0, 1)⋆ )m ⊂
⋃x∈Smn Vm,n (x). As this holds for every n, m, we may define A0 = ⋃n,m∈ℕ Sn,m which is a
w
countable subset of A, and has the sought for property that xo ∈ A0 . In order to show
w
that xo ∈ A0 , we must show that for every weak neighborhood N w (xo ) of xo , it holds
that N w (xo ) ∩ A0 ≠ 0. Indeed, let N w (xo ) be any weak neighborhood of xo . By the con-
struction of weak neighborhoods on X, there exists n, m and (x⋆1 , . . . , x⋆m ) ∈ (BX ⋆ (0, 1))m
such that ⋂m 1 w m
i=1 {x ∈ X : |⟨xi , x − xo ⟩ < n } ⊂ N (xo ). Since (x1 , . . . , xm ) ∈ (BX ⋆ (0, 1)) ⊂
⋆ ⋆ ⋆

⋃x∈Sn,m Vn,m (x), there exists x ∈ Sn,m for which (x⋆1 , . . . , x⋆m ) ∈ Vn,m (x), i. e., by the defini-
tion of Vn,m (x), there exists x ∈ Sn,m ⊂ A0 = ⋃n,m Sn,m , with the property |⟨x⋆i , x−xo ⟩| < n1
for all i = 1, . . . , m, which means that x ∈ ⋂m 1 w
i=1 {x ∈ X : |⟨xi , x − xo ⟩| < n } ⊂ N (xo ),
⋆

therefore x ∈ 𝒩xo . In summary, we have shown that for every weak neighborhood
󸀠

N w (xo ) of xo , there exists a x ∈ A0 such that x ∈ N w (xo ), so that N w (xo ) ∩ A0 ≠ 0. Since

N w (xo ) is arbitrary, we have that for every open neighborhood N w (xo ) of xo it holds
w
that N w (xo ) ∩ A0 ≠ 0, which implies that xo ∈ A0 .
Now, let X0 be the smallest closed linear subspace of X which contains A0 and
xo . Then X0 is a separable and reflexive Banach space. By the Hahn–Banach theorem,
each functional x⋆0 ∈ X0⋆ can be extended to a functional x⋆ ∈ X ⋆ . Therefore, xo lies in
the weak closure of A0 with respect to the weak topology on X0 . Since A0 is bounded
and X0 is separable and reflexive, the weak X0 topology on A0 is metrizable. Thus,
there exists a sequence {xn : n ∈ ℕ} ⊂ A0 such that xn ⇀ xo in X0 . Since X ⋆ ⊂ X0⋆ , we
conclude that xn ⇀ x in X.

As is natural certain functions which are continuous with respect to the strong
topology may no longer be continuous in the weak topology, as for example the func-
tion x 󳨃→ ‖x‖X , but this important function remains sequentially lower semicontinuous
with respect to the weak topology. Furthermore, the weak topology has some rather
convenient properties with respect to convexity, such as for instance that convex sets
which are closed under the strong topology are also closed under the weak topology,
but those will be treated in detail in Section 1.2.2.

1.2 Convex subsets of Banach spaces and their properties

Convexity plays a fundamental role in analysis. Here, we review some of the funda-
mental concepts that will be very frequently used in this book. For detailed accounts
of convexity, we refer to, e. g., [5, 19] or [25].

1.2.1 Definitions and elementary properties

We now introduce the concept of a convex subset of X.

 1.2 Convex subsets of Banach spaces and their properties | 23

Definition 1.2.1. Let X be a Banach space and C ⊂ X. The set C is called convex if it
has the property

tx1 + (1 − t)x2 ∈ C, ∀ x1 , x2 ∈ C, ∀ t ∈ [0, 1].

Notation 1.2.2. We will also use the notation [x1 , x2 ] = {tx1 + (1 − t)x2 : t ∈ [0, 1]},
(x1 , x2 ] = {tx1 + (1 − t)x2 : t ∈ (0, 1]} and [x1 , x2 ) = {tx1 + (1 − t)x2 : t ∈ [0, 1)} to
simplify the exposition. With this notation, C is convex if for any x1 , x2 ∈ C it holds
that [x1 , x2 ] ⊂ C.

Example 1.2.3. Let 𝒟 ⊂ ℝd and take X = Lp (𝒟) for p ∈ [1, ∞]. Then each element x ∈ X
is identified with a function u : 𝒟 → ℝ such that ‖u‖Lp (𝒟) < ∞. If ψ is any element of
X = Lp (𝒟), the set C := {u ∈ Lp (𝒟) : u(x) ≥ ψ(x) a.e.} is a convex subset of X. ◁

The following topological properties of convex sets (see, e. g., [5]) are useful.

Proposition 1.2.4 (Intersections, interior and closure of convex sets). In what follows,
C ⊂ X will be a convex set.
(i) Let {Cα : α ∈ ℐ } be an arbitrary family of convex sets. Then, ⋂α∈ℐ Cα is also convex.
(ii) If x ∈ int( C) and z ∈ C, then [x, z) ⊂ int( C).
(iii) If x1 ∈ int( C) and x2 ∈ C then [x1 , x2 ) ∈ int( C).
(iv) The interior int( C) and closure C are convex sets.
(v) If int( C) ≠ 0, then int( C) = int( C) and int( C) = C.

A useful concept is that of the algebraic interior.

Definition 1.2.5 (Algebraic interior). The algebraic interior or core of a set is defined
as

core(C) = {x ∈ C : ∀ z ∈ SX , ∃ϵ > 0 such that x + δz ∈ C, ∀ δ ∈ [0, ϵ]}.

For a convex set C, the core can equivalently be defined as

core(C) = {x ∈ C : X = ⋃ λ(C − x)}.

λ>0

Clearly, int( C) ⊂ core(C). However, there are important situations in which

int( C) = core(C). As we shall see when studying convex functions, it is important
to identify the interior of the domain of certain convex sets (as, e. g., the effective
domain of a convex function which has important effects on the continuity of the
function in question). On the other hand, it is much easier to check if a point is in
the core of a convex set. Therefore, knowing when the two sets coincide may often be
useful. However, before doing that we will need another definition.

Definition 1.2.6 (Convex series and convex series closed (CS) sets). An infinite sum
of the form ∑∞n=1 λn xn , where λn ∈ [0, 1] with ∑n=1 λn = 1 is called a convex series. A set
∞
24 | 1 Preliminaries

C is called convex series closed (or CS- closed) if for every convex series ∑∞
n=1 λn xn = x,
with xn ∈ C, it holds that x ∈ C.

It is straightforward for the reader to check that since X is a Banach space any con-
vex series of elements on a bounded subset A ⊂ X (which is by construction Cauchy)
is convergent. CS-closedness of C, guarantees that the limit is in C. Clearly, every CS-
closed set is convex. Every closed convex set C is CS-closed, but the converse does not
always hold true as CS-closed sets may not be closed. Therefore, CS-closedness is of-
ten a useful generalization of closedness. The following proposition (see [25]) is very
useful.

Proposition 1.2.7. Let C ⊂ X be a convex set. Then:

(i) If int( C) ≠ 0 then int( C) = core(C).
(ii) If C can be expressed as the countable union of closed sets C = ⋃n An , then
int( C) = core(C).
(iii) If C is CS-closed (see Definition 1.2.6) then core(C) = int( C) = int( C).

1.2.2 Separation of convex sets and its consequences

Convex sets enjoy separation properties related to the various forms of the Hahn–
Banach theorem (see, e. g., [28]). We state a form of this theorem which will be used
very often in this book.

Definition 1.2.8 (Closed hyperplanes and half spaces). Let X be a normed linear
space.
(i) The set Hx⋆ ,α ⊂ X defined by Hx⋆ ,α := {x ∈ X : ⟨x⋆ , x⟩ = α} for some x⋆ ∈ X ⋆ \ {0}
and α ∈ ℝ, is called a closed hyperplane.
(ii) The set Hx⋆ ≤α := {x ∈ X : ⟨x⋆ , x⟩ ≤ α} is called a closed half-space.

Using the above notions, we may present the geometric version of the Hahn–Banach
theorem, which essentially offers information concerning the containment of convex
sets in half-spaces.

Theorem 1.2.9 (Strict separation). Suppose C1 and C2 are two disjoint, convex subsets
of a normed space X, where C1 is compact and C2 is closed. Then there exists a closed
hyperplane that strictly separates C1 and C2 , i. e., there exists a functional x⋆ ∈ X ⋆ \ {0}
and an α ∈ ℝ such that

⟨x⋆ , x1 ⟩ < α, ∀ x1 ∈ C1 , and ⟨x⋆ , x2 ⟩ > α ∀ x2 ∈ C2 .

A convenient corollary of the separation theorem arises when one of the two con-
vex sets degenerates to a point.
 1.2 Convex subsets of Banach spaces and their properties | 25

Proposition 1.2.10. Let C ⊂ X be a nonempty closed convex subset of a normed linear

space. Then each xo ∈ ̸ C can be strictly separated from C by a closed hyperplane, i. e.,
there exists (x⋆ , α) ∈ (X ⋆ \ {0}) × ℝ, such that

⟨x⋆ , xo ⟩ > α, and ⟨x⋆ , x⟩ < α, ∀ x ∈ C.

This proposition14 essentially states that a nonempty closed convex set C is con-
tained in a (suitably chosen) closed half-space, whereas any point xo ∈ ̸ C in its com-
plement. In fact, this result can be rephrased as that a nonempty closed convex set
can be expressed in terms of the intersection of all the closed half-spaces Hx⋆ ≤α that
contain it, C = ⋂C⊂Hx⋆ ≤α Hx⋆ ≤α , an observation which is important in its own right.
A final reformulation of the separation theorem which is often useful is the fol-
lowing version.

Proposition 1.2.11. Let C1 , C2 be two nonempty disjoint convex subsets of a normed lin-
ear space, such that C1 has an interior point. Then C1 and C2 can be separated by a closed
linear hyperplane, i. e., there exists (x⋆ , α) ∈ (X ⋆ \ {0}) × ℝ, such that

⟨x⋆ , x1 ⟩ ≤ α, ∀ x1 ∈ C1 , and ⟨x⋆ , x2 ⟩ ≥ α ∀ x2 ∈ C2 .

As a result of the separation theorem, convex sets have a number of remarkable

and very useful properties. For example, for convex sets, the concepts of weak and
strong closedness coincide, meaning that if a convex set C ⊂ X is closed in the strong
topology of X it is also closed in the weak topology of X. Clearly, the same is true for
closed linear subspaces.

Proposition 1.2.12. A closed convex set C is also weakly closed.

Proof. Assume that C is strongly closed. We will show that it is also weakly closed. This
is equivalent to showing that its complement C c = X \ C is open in the weak topology.
Consider any xo ∈ C c . Since C is closed and xo ∈ ̸ C, by an application of the Hahn–
Banach theorem (in the form of Proposition 1.2.10) there exists x⋆o ∈ X ⋆ and α ∈ ℝ
such that ⟨x⋆o , xo ⟩ < α < ⟨x⋆o , z⟩, for any z ∈ C. Define the set V = {x ∈ X : ⟨x⋆o , x⟩ < α},
which is open in the weak topology. Observe that xo ∈ V so N w (xo ) := V is a (weak)
neighborhood of xo . By the strict separation V ∩ C = N w (xo ) ∩ C = 0, which implies
V = N w (xo ) ⊂ C c . So for every xo ∈ C c we may find a (weak) neighborhood N w (xo ) such
that N w (xo ) ⊂ C c which means that C c is weakly open.15

Remark 1.2.13. An alternative way to phrase the above statement is that for a convex
w
set C in a Banach space X, the strong and the weak closure of C, denoted by C and C ,

14 whose proof follows directly from Theorem 1.2.9 by considering the compact set C1 = {xo } and the
closed set C2 = C.
15 Since C c is a weak neighborhood of each of its points, C c = ⋃xo ∈Cc N w (xo ) so C c is weakly open as
the arbitrary union of weakly open sets.
26 | 1 Preliminaries

respectively, coincide. Since the weak topology is coarser than the strong it is clear that
w
C ⊂ C . It remains to show the opposite inclusion. Since C is (by definition) strongly
w
closed and convex, by Proposition 1.2.12 it is also weakly closed. Therefore, C ⊂ C
w
leading to the stated result C = C .

Remark 1.2.14. A weakly closed set is also weakly sequentially closed.16 By that ob-
servation, the result stated in Proposition 1.2.12 is very useful and will be employed
quite often in the following setting: If a convex set C ⊂ X has the property that for any
sequence {xn : n ∈ ℕ} ⊂ C such that xn → x in X it holds that x ∈ C then it also has the
property that for any sequence {x̄n : n ∈ ℕ} ⊂ C such that x̄n ⇀ x̄ it holds that x̄ ∈ C.

An important side result of Proposition 1.2.12 is the following weak compactness

result (the reader may also wish to revisit Example 1.1.47).

Proposition 1.2.15. Let X be a reflexive Banach space and C ⊂ X a convex, closed and
bounded set. Then C is weakly compact.

Proof. By Proposition 1.2.12, since C is convex and closed it is also weakly closed. Since
C is bounded, there exists λ > 0 such that C ⊂ λBX (0, 1) where BX (0, 1) is the closed
unit ball in X. This implies that C = C ∩ λBX (0, 1). Since X is reflexive by the Eberlein–
Šmulian theorem, BX (0, 1) is weakly compact; hence, λBX (0, 1) enjoys the same prop-
erty. Therefore, since C = C ∩ λBX (0, 1), it is expressed as the intersection of the weakly
closed set C and the weakly compact set λBX (0, 1) it is weakly compact.

An important corollary of the above proposition is Mazur’s lemma. In order to

prove it, we need the concept of the convex hull of a set.

Definition 1.2.16 (Convex hull). Let A ⊂ X. The convex hull of A denoted by convA is
the set of all convex combinations from A,

n n
convA = {x : ∃xi ∈ A, ti ∈ [0, 1], i = 1, . . . , n, ∑ ti = 1, and x = ∑ ti xi }.
i=1 i=1

The convex hull of A, convA, is the smallest convex set including A. The smallest
closed convex set including A is called the closed convex hull of A and is denoted by
convA. It can be easily seen that convA = convA.

Proposition 1.2.17 (Mazur). Let X be a Banach space and consider a sequence {xn :
n ∈ ℕ} ⊂ X such that xn ⇀ x in X. Then there exists a sequence {zn : n ∈ ℕ} ⊂ X
consisting of (finite) convex combinations of terms of the original sequence such that
zn → x (strongly). The sequence {zn : n ∈ ℕ} can be constructed so that for each n ∈ ℕ,
the term zn is either (i) a (finite) convex combination of the terms {xn , xn+1 , . . .} or (ii) a
(finite) convex combination of the terms {x1 , x2 , . . . , xn }.

16 In fact, a closed set is sequentially closed for any topology.

 1.2 Convex subsets of Banach spaces and their properties | 27

Proof. The proof is based on Proposition 1.2.12. Since for every n ∈ ℕ, we have that
w
xn ∈ C := conv{x1 , x2 , . . .} and because xn ⇀ x we have that x ∈ C . But C is convex,
w
so that by Proposition 1.2.12 it holds that C = C; hence, x ∈ C, so that there exists a
sequence {zn : n ∈ ℕ} ⊂ C, strongly converging to x. To show claim (i), fix any n ∈ ℕ.
Clearly, xk ∈ Cn := conv{xn , xn+1 , . . .} for any k ≥ n. Since xk ⇀ x as k → ∞, it holds
w w
that x ∈ Cn . But Cn is a convex set so that by Proposition 1.2.12, Cn = Cn for every
n, therefore, x ∈ Cn . The strong closure is characterized as the set of limits of strongly
convergent sequences, so there exists a sequence {zk : k ∈ ℕ} ⊂ Cn such that zk → x
as k → ∞. Since our result holds for any n, for each k pick n = n(k) = k which leads to
the desired result. Claim (ii) follows similarly.

Proposition 1.2.18 (Mazur). If X is a Banach space, A ⊂ X compact, then convA is com-

pact.

Proof. We briefly sketch the proof.17 If A = {x1 , . . . , xn } is finite, then we may consider
convA as the image of the map (t1 , . . . , tn ) 󳨃→ ∑ni=1 ti xi where (t1 , . . . , tn ) takes values in
the compact set {(t1 , . . . , tn ) : ti ∈ [0, 1], ∑ni=1 ti = 1, i = 1, . . . , n}. Since this map is
continuous, it follows that convA is compact, therefore, convA is also compact.
If A is not finite but compact, for any ϵ > 0, there exists a finite set Aϵ =
{x1 , . . . , xn } ⊂ A such that A ⊂ ⋃ni=1 B(xi , ϵ). Note that ⋃ni=1 B(xi , ϵ) ⊂ convAϵ + B(0, ϵ),
which by the above argument is closed and convex. Therefore, A ⊂ convAϵ + B(0, ϵ)
and since convAϵ +B(0, ϵ) is a closed convex set containing A and convA is the smallest
closed convex set containing A it must hold that

convA ⊂ convAϵ + B(0, ϵ). (1.3)

Since Aϵ is a discrete set, we know from the first part of the proof that convAϵ is
compact so there exists a finite set A1 := {z1 , . . . , zm } ⊂ convAϵ such that convAϵ ⊂
⋃m m
i=1 B(zi , ϵ). Noting that ⋃i=1 B(zi , ϵ) = A1 + B(0, ϵ) and combining the previous inclu-
sion with (1.3) we see that

convA ⊂ A1 + B(0, 2ϵ), (1.4)

which together with the observation that A1 ⊂ convAϵ ⊂ convA allows us to conclude
that convA is totally bounded; hence, compact.

Remark 1.2.19. A similar result, known as the Krein–Smulian weak compactness the-
orem, holds for weak compactness. According to that, if A ⊂ X be a weakly compact
w
set, then convA is also weakly compact. For a proof see, e. g., [82], page 254 or [48],
page 164.

17 The reader can find more details in [108], or in a slightly more general framework in [5] or [75].
28 | 1 Preliminaries

A natural question arising at this point is concerning the behavior of convex sets
of X ⋆ in the weak⋆ topology. In general, Proposition 1.2.12 is not true, i. e., a convex A ⊂
X ⋆ which is closed in the strong topology of X ⋆ is not necessarily closed in the weak⋆
topology of X ⋆ , unless of course X is reflexive. One may construct counterexamples
for the general case where X is a nonreflexive Banach space. A general criterion for
weak⋆ closedness of convex subsets A ⊂ X ⋆ , is the Krein–Smulian theorem according
to which A is weak⋆ closed if and only if A ∩ rBX ⋆ (0, 1) is weak⋆ closed for every r > 0
(such sets are often called bounded weak⋆ closed). If X is a separable Banach space,
then one may show that a convex set A ⊂ X ⋆ is bounded weak⋆ closed if and only if it
is weak⋆ sequentially, therefore, for separable X, a convex set A ⊂ X ⋆ is weak⋆ closed
if and only if it is weak⋆ sequentially closed. This characterization can be quite useful
in a number of cases.

1.3 Compact operators and completely continuous operators

Compact operators is an important class of operators that plays a fundamental role in
both linear and nonlinear analysis. We review here some of the fundamental proper-
ties of compact operators (see, e. g., [48]).

Definition 1.3.1 (Compact and completely continuous operators).

(i) Let X and Y be Banach spaces. An operator A : X → Y (not necessarily linear) is
called compact if it maps bounded sets of X into relatively compact sets of Y (i. e.,
into sets whose closure is a compact set).
(ii) Let X and Y be Banach spaces. An operator A : X → Y (not necessarily linear)
is called completely continuous (or strongly continuous) if the image of every
weakly convergent sequence in X under A converges in the strong (norm) topology
of Y, i. e., if xn ⇀ x in X implies A(xn ) → A(x) in Y, (i. e., ‖A(xn ) − A(x)‖Y → 0).

Example 1.3.2 (Finite rank operators are compact). Consider a linear operator A ∈
ℒ(X, Y) such that dim(R(A)) < ∞. Such an operator, called a finite rank operator, is
compact.
This follows easily from the definition since closed and bounded sets in finite di-
mensional spaces are compact. ◁

In general, the classes of compact operators and completely continuous operators

are not comparable. However, the following general results hold.

Theorem 1.3.3. Let X, Y, Z be Banach spaces and A : X → Y, B : Y → Z be linear

operators.
(i) If A is a compact operator, then it is completely continuous.
(ii) If X is reflexive, and A is completely continuous, then A is compact.
(iii) (Schauder) A bounded operator A is compact if and only if A⋆ is compact.
 1.3 Compact operators and completely continuous operators | 29

(iv) The product BA : X → Z is compact if one of the two operators is compact and the
other is bounded.
(v) Consider a sequence of compact operators {An : n ∈ ℕ} ⊂ ℒ(X, Y) that converges in
the strong operator topology to an operator A : X → Y, such that A ∈ ℒ(X, Y). Then
A is compact.

An important class of compact operators are integral operators.

Example 1.3.4 (Certain integral operators are compact). Consider a compact interval
[a, b] ⊂ ℝ and the Banach space

X = C([a, b]) = {ϕ : [a, b] → ℝ, ϕ continuous},

endowed with the norm ‖ϕ‖ = supt∈[a,b] |ϕ(t)|. An element x ∈ X is identified with
a continuous function ϕ : [a, b] → ℝ. We now consider the integral operator A :
C([a, b]) → C([a, b]) defined by
ˆ b
Aϕ(t) = K(t, τ)ϕ(τ)dτ,
a

where K : [a, b]×[a, b] → ℝ is continuous, and we use the simpler notation Aϕ instead
of A(ϕ) to denote the element of C([a, b]) which is the image of ϕ ∈ C([a, b]) under
A. Then A is a linear and compact operator. The result can be extended for integral
operators on X = C(𝒟) in the case where 𝒟 ⊂ ℝd is a suitable domain, defined by its
´
action on any ϕ ∈ C(𝒟) by Aϕ(x) = 𝒟 K(x, z)ϕ(z)dz for any z ∈ 𝒟.
The linearity of the operator follows by standard properties of the integral. In or-
der to show compactness, consider the bounded set U ⊂ X, of functions ϕ ∈ X such
that supt∈[a,b] |ϕ(t)| < c for some constant c. Then the set A(U) is bounded, as can be
seen by the estimate

󵄨󵄨ˆ b 󵄨󵄨
󵄨󵄨 󵄨󵄨
󵄨󵄨
󵄨󵄨 K(t, τ)ϕ(τ)dτ 󵄨󵄨 ≤ (b − a) max 󵄨󵄨󵄨K(t1 , t2 )󵄨󵄨󵄨‖ϕ‖ < c󸀠
󵄨󵄨 󵄨 󵄨
󵄨󵄨 a 󵄨󵄨 t1 ,t2 ∈[a,b]

for a suitable constant c󸀠 . We will show that A(U) is equicontinuous; hence, relatively
compact in X := C([a, b]) by the Arzelá–Ascoli theorem (see Theorem 1.8.3). Indeed,
since K is continuous on the compact set [a, b] × [a, b] it is also uniformly continuous;
hence, for every ϵ󸀠 > 0 there exists δ > 0 such that

󵄨󵄨K(t1 , τ) − K(t2 , τ)󵄨󵄨󵄨 < ϵ

󵄨󵄨 󵄨 󸀠
for all t1 , t2 , τ ∈ [a, b] with |t1 − t2 | < δ. (1.5)

Then,
ˆ b
󵄨󵄨Aϕ(t1 ) − Aϕ(t2 )󵄨󵄨󵄨 ≤
󵄨󵄨 󵄨
󵄨󵄨K(t1 , τ) − K(t2 , τ)󵄨󵄨󵄨󵄨󵄨󵄨ϕ(τ)󵄨󵄨󵄨dτ ≤ (b − a)ϵ c,
󵄨󵄨 󵄨󵄨 󵄨 󸀠 (1.6)
a
30 | 1 Preliminaries

ϵ
for any ϕ ∈ U. Consider any ϵ > 0. Let ϵ󸀠 = (b−a)c and choose δ > 0 such that (1.5)
holds. Then from (1.6) we can conclude that for all t1 , t2 ∈ [a, b] with |t1 − t2 | < δ
and all ϕ ∈ U it holds that |Aϕ(t1 ) − Aϕ(t2 )| < ϵ, hence A(U) is equicontinuous and
by the Arzelá–Ascoli theorem (see Theorem 1.8.3) we have the compactness of the
operator A. ◁

Definition 1.3.5 (Fredholm operators). A bounded linear operator A : X → Y is called

a Fredholm operator if (a) R(A) is a closed subset of Y, (b) dim(N(A)) < ∞, and
(c) dim(Y/R(A)) < ∞. The quotient space Y/R(A) is often called the cokernel of
the operator A, and is denoted by coker(A). The difference index(A) := dim(N(A)) −
dim(Y/R(A)) is called the Fredholm index of A.

Example 1.3.6 (Finite dimensional operators are Fredholm). If dim(X) < ∞ and
dim(Y) < ∞, then any A ∈ ℒ(X, Y) is Fredholm with index(A) = dim(X) − dim(Y).
This is a well-known result from linear algebra.

Example 1.3.7 (Fredholm operators related to integral operators). In the framework

of Example 1.3.4, consider the operator B : X → X defined by B = I − A, which acts
´
on any function ϕ ∈ C(𝒟) by Bϕ(x) = ϕ(x) − 𝒟 K(x, z)ϕ(z)dz. This is a Fredholm
operator.

Fredholm operators have some interesting properties with respect to duality, as

this is reflected in the properties of the adjoint (dual) operator, as well as some useful
connections with compact operators.

Theorem 1.3.8 (Properties of Fredholm operators). Let A : X → Y be a bounded linear

operator between two Banach spaces X, Y. Then:
(i) A is Fredholm if and only if A⋆ is Fredholm.
(ii) If A is Fredholm then dim(N(A⋆ )) = dim(Y/R(A)) and dim(X ⋆ /R(A⋆ )) = dim(N(A)).
(iii) A is a Fredholm operator if and only if there exists a bounded linear operator B :
Y → X such that the operators IX − BA : X → X and IY − AB : Y → Y are compact.
(iv) If X, Y, Z are Banach spaces and A : X → Y, B : Y → Z are Fredholm operators
then their composition BA : X → Z is also a Fredholm operator with index(BA) =
index(A) + index(B).
(v) If A, B : X → Y are a Fredholm and a compact operator, respectively, then A + B is
Fredholm and index(A + B) = index(A).
(vi) If A, B : X → Y are a Fredholm and a bounded operator, respectively, then there
exists ϵ0 > 0 such that A + ϵB is Fredholm for ϵ < ϵ0 and index(A + B) = index(A).

Theorem 1.3.8 in the special case where X = Y can lead to a very useful result, known
as the Fredholm alternative, concerning the solvability of linear operator equations,
of the form (I − A)x = z, where A : X → X is a linear compact operator and z ∈ X is
given.
 1.3 Compact operators and completely continuous operators | 31

Theorem 1.3.9 (Fredholm alternative). Let A ∈ ℒ(X) be a compact operator. Then, upon
defining the annihilator of A ⊂ X as

A⊥ := {x⋆ ∈ X ⋆ : ⟨x⋆ , x⟩ = 0, ∀ x ∈ A},

we have that: dim(N(I − A)) = dim(N(I − A⋆ )) < ∞, and R(I − A) is closed with R(I − A) =
N(I − A⋆ )⊥ . In particular N(I − A) = {0} if and only if R(I − A) = X.

In particular, the Fredholm alternative states that either the operator equation
(I − A)x = z admits a unique solution for any z ∈ X, or the homogeneous equation
(I − A)x = 0 admits n linearly independent solutions and then the inhomogeneous
equation admits solutions only as long as z ∈ N(I −A⋆ )⊥ . Then the solvability condition
of the operator equation (I − A)x = z is related to the existence of nontrivial solutions
to problem (I − A)x = 0 or the invertibility of the operator I − A. The latter problem is
called an eigenvalue problem.

Example 1.3.10 (Solvability of integral equations). In the multidimensional frame-

work of Example 1.3.4, consider the integral equation
ˆ
ϕ(x) − λ K(x, z)ϕ(z)dz = ψ(x),
𝒟

for any x ∈ 𝒟, where ψ ∈ C(𝒟) is a known function and λ ∈ ℝ is a known constant.

Then, on account of the results of Example 1.3.4 we may apply the Fredholm alterna-
tive and conclude that this problem admits a unique solution for any ψ ∈ C(𝒟) as long
´
as λ ∈ ℝ is such that the homogeneous problem ϕ(x) − λ 𝒟 K(x, z)ϕ(z)dz = 0 admits
only the trivial solution ϕ(x) = 0 for all x ∈ 𝒟. This tells us that the solvability of the
´
problem occurs only for special types of ψ, if λ is such that ϕ(x)−λ 𝒟 K(x, z)ϕ(z)dz = 0
admits nontrivial solutions, i. e., if λ coincides with eigenvalues of the integral opera-
tor. A characterization of such ψ requires the study of the dual operator. ◁

As problems of the form (I − A)x = 0 play an important role in linear functional

analysis we close this section by recalling some fundamental facts concerning the
spectrum of linear operators in Banach spaces. Since it is well known from linear
algebra that real matrices may well have complex eigenfunctions, the same is antic-
ipated to hold in infinite dimensions and we therefore in general need to consider
complex valued Banach spaces, in which the concept of the norm is generalized so
that ‖λx‖ = |λ| ‖x‖ for all x ∈ X and λ ∈ ℂ, so as to treat this problem in its full gener-
ality. However, here this is not required for the needs of the material covered, and we
will (following [28]) present a selected view of these concepts restricting ourselves to
real valued Banach spaces. For a more general viewpoint, we refer, e. g., to [82].

Definition 1.3.11 (The resolvent set and the spectrum). Let A : X → X be a bounded
linear operator.
(i) The resolvent set of A is the set ρ(A) := {λ ∈ ℝ : (A − λI) : X → X is bijective}.
Another Random Scribd Document
with Unrelated Content
"I'm taking a powder, as they used to say back on good old Terra.
I'm gambling. Gambling that I can get back to the one cruiser that
hasn't burnt up all its wiring, and call the ship." He slapped the leg-
pocket of his breeches. Kemp nodded. The pocket contained a
ready-packed emergency chute.

Cornelia shook her head slowly. "You'll never make it—"

"He might," said Kemp.
"I'll bet I can," said Pritchard. "I've got to."
"The One will get you," she said. "It can get into one of those things
easily. It'll take only a little of Its digestive juice...." Her face
puckered and those emerald eyes shone brighter, but she fought for
and regained control.
"So?" Pritchard smiled. "You'll be well rid, then, of that notorious big-
game hunter, Elmer Pritchard."
"I don't want to be rid of him," she said softly. "I want him with me
—at the end."
He bowed. "Thank you, Miss Boyce."
"Call me Cornelia, please."
"And you," he said, "may call me Elmer—a name I permit no one to
use—" he bent forward "—but you, now."
Their lips brushed and clung.
"Fine time for love-making," muttered Kemp.
Below them, a flamer squealed suddenly. Sturgis, unknown to them,
had lingered behind. Now, a hundred feet down the ridge, he fired a
burst at the worm—a warning burst, for the dread feelers hung high
above his head on a long, curving tendon of red wetness.
The flamer had an effective range of only thirty feet, but the slimy
scarlet rope curved away, dropping off to one side and extending out
into the air. The feelers contracted to mere knobs and the end
thickened into a club.
A haymaker, drawing back, poised and cocked. Pritchard saw it and
howled, "Sturgis! Duck!"
But there was to be no ducking that swing. Sturgis hugged the thin
spine of crag and threw up a blossom of fire. But the rope came
flailing about, slashing through the flame, and neatly flicked him off.
They watched the body arcing out over the meadow, the spare
flamer of Pritchard twisting after it, and saw it sink on down, to stop
suddenly against the turf.
Kemp began to curse. Pritchard pulled the emergency chute pack
from his leg-pocket and began snapping the light harness about his
long frame.
"Cut that out," he said coldly. "Just hang on, Kemp, and watch. If
I've got this baby figured out right, It's going to lose interest in you
two in about as many seconds."
"Good-bye, Elmer," came Cornelia's voice forlornly.
The worm's first half was recovering from the follow-through of that
swing, draping itself back along the ridge yard by relentless yard.
Pritchard turned, holding the chute cord in his fist. He forced a grin
that he was afraid looked more like a grimace. "So long, kiddies," he
said, and jumped.
At this point the leaning peak overhung the ground and he flung
himself as far out as possible, trying for distance. The smooth,
almost polished wall slanted away from him and the meadow swung
upward.
He pulled the cord at the last minute. As the filmy neosilk billowed
above him, and the harness seemed to jerk him back up from the
onrushing ground, he managed to twist a glance back up at the
ridge.
The One was motionless. That was good.
It had seen him.
Then he drew up his knees. The ground slammed into him and he
lay there, stunned, letting the filmy folds flutter down over him.
Then he was up, bruised but whole, on his knees and scrabbling out
from under the light gray stuff. By crawling under every line he
avoided entanglement and in a minute was clear and running,
unsnapping the harness as he went.
Not until he was well away from the mountain did he dare a glance
over his shoulder. Then he almost stumbled, at the chill terror
gusting through him, freezing every muscle.
The worm was a red festoon, drooping from the ridge. Even as he
looked, Its whole length came off, to fall writhing out of sight
momentarily at the base of the mountain.
He hadn't expected that. He had planned for It to back laboriously
down the way It came, giving him a decent margin of time. But it
had crossed him up. Now he had seconds instead of minutes.
He put his head down and dug in, pumping his tired, aching legs
furiously. This was the worst gamble of his career, against the
longest odds. He had no idea how fast the worm could go on level
ground.
Suddenly, he was racing a shadow. In the slanting light of Piramus,
setting through the afternoon, something like an elongated
caricature of a snail's head crept across the grass beside him—two
long slivers of tapering purple shadow.
Then he saw his flamer, lying almost dead ahead where it had
landed after being catapulted off the ridge. Sobs rasping his throat,
he slanted toward it, dove and rolled, to come up clutching it.
There was a spattering sound close by, a spatter that changed to an
angry fizzing. Pritchard swung the nozzle up in the very face of the
glistening red column swaying toward him. He squeezed the handle-
grip.
Through the booming flame, he saw the shape twisting aside and
followed it with fire. It went down to the ground, backing away into
a swelling body. The worm writhed desperately away from that
searing plume of licking flames.
Pritchard wheeled and ran toward the cruiser that had not burned.
Evil-smelling juice slashed across the upturned belly of the ship as
he savagely wrenched open the buckled door and tumbled in,
dragging the flamer in after him. He stumbled across the roof-struts
and lunged for the upside-down radio panel.
The cruisers' radios were on their own battery-powered circuits. He
snapped the power on and heard the slow hum and sputter of the
warming tubes. He poked in the button labeled AUT. EM. SIG. a
standard repeating distress call on a tight beam.
Then he was flung against the opposite wall. As he struggled back to
his feet, pressure against them told him the cruiser was rising, and
he knew very well it was not doing so under its own power.
A glistening red wall bulged against the door-frame through which
he had come. Pritchard realized that once again the cruiser was
being hoisted aloft in the worm's coil. It was going to drop him, to
kill him quickly, rather than poke inside and face his flamer.
Pritchard snatched the flamer and staggered toward the opening.
Jabbing the nozzle into that scarlet slime, he gripped the handles.

Roaring heat beat back at him. He braced himself, ignoring his own
singeing flesh and crisping hair.
The cruiser struck ground with a crash. He was flung sideways,
threw up an arm, and heard it snap. He dragged himself to the door
which now was turned to the ground. Gritting his teeth against the
pain, Pritchard hung his head through the opening and peered out.
It was a crazy nightmare. The meadow was a ceiling, to his inverted
eyes, against which a giant red riband rolled and writhed in fantastic
configurations. Every melting convolution, every arching loop,
expressed pain and wrath. And, now and again, a livid blotch
appeared along its length, alternately turning purple and yellow, and
dripping streamers of drool.
Then came a sound, a great tearing sound in the sky. Pritchard
hauled himself back into the ship and crawled to the radio. He
switched off the automatic signal and cut in the transmission band.
"—the hell you got down there?" came Captain Savage's rasp. "Is
that you up on the rock, Mr. Pritchard? Mr. Pritchard—"
"Captain!" yelled Pritchard. "Step on it! Come down on that monster.
I'm all right. Come ahead!"
Then he snatched up a pair of solar goggles and worked his way to
a viewport, in time to see the Apollo, a magnificent column of metal
in the sky, descend on a pillar of incandescence—at the bottom of
which lay something that bubbled and cooked, rising in a last great
arch of simmering agony.
The snaggle-toothed horizon of Thisbe II was rising across the dull
indigo disk of setting Piramus. Pritchard and Savage sat in their
gimbal chairs in the Forward Lounge. The old man's wispy white
hairs stirred in the evening breeze sucked in by the blowers.
"And every time I wonder if my hunting days aren't over," sighed
Pritchard. Experimentally, he worked on the flexicast on his right
arm.
"Huh," grunted the captain. "Not you. One week on Terra and you'll
be telling yourself the next time it just can't be as bad. Or that this
wasn't as bad as it seemed. Anything, you'll tell yourself. Anything to
start—"
Cornelia appeared in the doorway. "Good evening, gentlemen," she
said coolly. She was wearing cordron slacks and a soft neosilk
blouse, that seemed to enjoy clinging to her contours.
"Good evening," croaked Captain Savage. He stood up, and
stretched restlessly.
"Oh, don't go," said Cornelia.
"Well, if we're blasting off in the morning, I've got things to do.
These days it's the old men who do all the work." He chuckled as he
eased past her through the door, and gave her shoulder a little pat.
"Good hunting."
The girl watched him go down the passage. "Whatever did he mean
by that?" she inquired. "'Good hunting'."
"Oh, it's just an expression," said Pritchard vaguely.
She came over to him and turned about on her bare feet. "No shoes
that fit," she said. "How do you like what I managed to scrounge
from the men?"
He pulled her down to him with one lazy reach of his good arm. "I'm
afraid," he murmured, "that I liked you better the way you were."
"You know," she spoke muffledly against his shoulder, "you're
something of a beast."
 *** END OF THE PROJECT GUTENBERG EBOOK AS IT WAS ***

Updated editions will replace the previous one—the old editions will
be renamed.

Creating the works from print editions not protected by U.S.

copyright law means that no one owns a United States copyright in
these works, so the Foundation (and you!) can copy and distribute it
in the United States without permission and without paying
copyright royalties. Special rules, set forth in the General Terms of
Use part of this license, apply to copying and distributing Project
Gutenberg™ electronic works to protect the PROJECT GUTENBERG™
concept and trademark. Project Gutenberg is a registered trademark,
and may not be used if you charge for an eBook, except by following
the terms of the trademark license, including paying royalties for use
of the Project Gutenberg trademark. If you do not charge anything
for copies of this eBook, complying with the trademark license is
very easy. You may use this eBook for nearly any purpose such as
creation of derivative works, reports, performances and research.
Project Gutenberg eBooks may be modified and printed and given
away—you may do practically ANYTHING in the United States with
eBooks not protected by U.S. copyright law. Redistribution is subject
to the trademark license, especially commercial redistribution.

START: FULL LICENSE

THE FULL PROJECT GUTENBERG LICENSE
 PLEASE READ THIS BEFORE YOU DISTRIBUTE OR USE THIS WORK

To protect the Project Gutenberg™ mission of promoting the free

distribution of electronic works, by using or distributing this work (or
any other work associated in any way with the phrase “Project
Gutenberg”), you agree to comply with all the terms of the Full
Project Gutenberg™ License available with this file or online at
www.gutenberg.org/license.

Section 1. General Terms of Use and

Redistributing Project Gutenberg™
electronic works
1.A. By reading or using any part of this Project Gutenberg™
electronic work, you indicate that you have read, understand, agree
to and accept all the terms of this license and intellectual property
(trademark/copyright) agreement. If you do not agree to abide by all
the terms of this agreement, you must cease using and return or
destroy all copies of Project Gutenberg™ electronic works in your
possession. If you paid a fee for obtaining a copy of or access to a
Project Gutenberg™ electronic work and you do not agree to be
bound by the terms of this agreement, you may obtain a refund
from the person or entity to whom you paid the fee as set forth in
paragraph 1.E.8.

1.B. “Project Gutenberg” is a registered trademark. It may only be

used on or associated in any way with an electronic work by people
who agree to be bound by the terms of this agreement. There are a
few things that you can do with most Project Gutenberg™ electronic
works even without complying with the full terms of this agreement.
See paragraph 1.C below. There are a lot of things you can do with
Project Gutenberg™ electronic works if you follow the terms of this
agreement and help preserve free future access to Project
Gutenberg™ electronic works. See paragraph 1.E below.
1.C. The Project Gutenberg Literary Archive Foundation (“the
Foundation” or PGLAF), owns a compilation copyright in the
collection of Project Gutenberg™ electronic works. Nearly all the
individual works in the collection are in the public domain in the
United States. If an individual work is unprotected by copyright law
in the United States and you are located in the United States, we do
not claim a right to prevent you from copying, distributing,
performing, displaying or creating derivative works based on the
work as long as all references to Project Gutenberg are removed. Of
course, we hope that you will support the Project Gutenberg™
mission of promoting free access to electronic works by freely
sharing Project Gutenberg™ works in compliance with the terms of
this agreement for keeping the Project Gutenberg™ name associated
with the work. You can easily comply with the terms of this
agreement by keeping this work in the same format with its attached
full Project Gutenberg™ License when you share it without charge
with others.

1.D. The copyright laws of the place where you are located also
govern what you can do with this work. Copyright laws in most
countries are in a constant state of change. If you are outside the
United States, check the laws of your country in addition to the
terms of this agreement before downloading, copying, displaying,
performing, distributing or creating derivative works based on this
work or any other Project Gutenberg™ work. The Foundation makes
no representations concerning the copyright status of any work in
any country other than the United States.

1.E. Unless you have removed all references to Project Gutenberg:

1.E.1. The following sentence, with active links to, or other

immediate access to, the full Project Gutenberg™ License must
appear prominently whenever any copy of a Project Gutenberg™
work (any work on which the phrase “Project Gutenberg” appears,
or with which the phrase “Project Gutenberg” is associated) is
accessed, displayed, performed, viewed, copied or distributed:
 This eBook is for the use of anyone anywhere in the United
States and most other parts of the world at no cost and with
almost no restrictions whatsoever. You may copy it, give it away
or re-use it under the terms of the Project Gutenberg License
included with this eBook or online at www.gutenberg.org. If you
are not located in the United States, you will have to check the
laws of the country where you are located before using this
eBook.

1.E.2. If an individual Project Gutenberg™ electronic work is derived

from texts not protected by U.S. copyright law (does not contain a
notice indicating that it is posted with permission of the copyright
holder), the work can be copied and distributed to anyone in the
United States without paying any fees or charges. If you are
redistributing or providing access to a work with the phrase “Project
Gutenberg” associated with or appearing on the work, you must
comply either with the requirements of paragraphs 1.E.1 through
1.E.7 or obtain permission for the use of the work and the Project
Gutenberg™ trademark as set forth in paragraphs 1.E.8 or 1.E.9.

1.E.3. If an individual Project Gutenberg™ electronic work is posted

with the permission of the copyright holder, your use and distribution
must comply with both paragraphs 1.E.1 through 1.E.7 and any
additional terms imposed by the copyright holder. Additional terms
will be linked to the Project Gutenberg™ License for all works posted
with the permission of the copyright holder found at the beginning
of this work.

1.E.4. Do not unlink or detach or remove the full Project

Gutenberg™ License terms from this work, or any files containing a
part of this work or any other work associated with Project
Gutenberg™.

1.E.5. Do not copy, display, perform, distribute or redistribute this

electronic work, or any part of this electronic work, without
prominently displaying the sentence set forth in paragraph 1.E.1
with active links or immediate access to the full terms of the Project
Gutenberg™ License.

1.E.6. You may convert to and distribute this work in any binary,
compressed, marked up, nonproprietary or proprietary form,
including any word processing or hypertext form. However, if you
provide access to or distribute copies of a Project Gutenberg™ work
in a format other than “Plain Vanilla ASCII” or other format used in
the official version posted on the official Project Gutenberg™ website
(www.gutenberg.org), you must, at no additional cost, fee or
expense to the user, provide a copy, a means of exporting a copy, or
a means of obtaining a copy upon request, of the work in its original
“Plain Vanilla ASCII” or other form. Any alternate format must
include the full Project Gutenberg™ License as specified in
paragraph 1.E.1.

1.E.7. Do not charge a fee for access to, viewing, displaying,

performing, copying or distributing any Project Gutenberg™ works
unless you comply with paragraph 1.E.8 or 1.E.9.

1.E.8. You may charge a reasonable fee for copies of or providing

access to or distributing Project Gutenberg™ electronic works
provided that:

• You pay a royalty fee of 20% of the gross profits you derive
from the use of Project Gutenberg™ works calculated using the
method you already use to calculate your applicable taxes. The
fee is owed to the owner of the Project Gutenberg™ trademark,
but he has agreed to donate royalties under this paragraph to
the Project Gutenberg Literary Archive Foundation. Royalty
payments must be paid within 60 days following each date on
which you prepare (or are legally required to prepare) your
periodic tax returns. Royalty payments should be clearly marked
as such and sent to the Project Gutenberg Literary Archive
Foundation at the address specified in Section 4, “Information
 about donations to the Project Gutenberg Literary Archive
Foundation.”

• You provide a full refund of any money paid by a user who

notifies you in writing (or by e-mail) within 30 days of receipt
that s/he does not agree to the terms of the full Project
Gutenberg™ License. You must require such a user to return or
destroy all copies of the works possessed in a physical medium
and discontinue all use of and all access to other copies of
Project Gutenberg™ works.

• You provide, in accordance with paragraph 1.F.3, a full refund of

any money paid for a work or a replacement copy, if a defect in
the electronic work is discovered and reported to you within 90
days of receipt of the work.

• You comply with all other terms of this agreement for free
distribution of Project Gutenberg™ works.

1.E.9. If you wish to charge a fee or distribute a Project Gutenberg™

electronic work or group of works on different terms than are set
forth in this agreement, you must obtain permission in writing from
the Project Gutenberg Literary Archive Foundation, the manager of
the Project Gutenberg™ trademark. Contact the Foundation as set
forth in Section 3 below.

1.F.

1.F.1. Project Gutenberg volunteers and employees expend

considerable effort to identify, do copyright research on, transcribe
and proofread works not protected by U.S. copyright law in creating
the Project Gutenberg™ collection. Despite these efforts, Project
Gutenberg™ electronic works, and the medium on which they may
be stored, may contain “Defects,” such as, but not limited to,
incomplete, inaccurate or corrupt data, transcription errors, a
copyright or other intellectual property infringement, a defective or
damaged disk or other medium, a computer virus, or computer
codes that damage or cannot be read by your equipment.

1.F.2. LIMITED WARRANTY, DISCLAIMER OF DAMAGES - Except for

the “Right of Replacement or Refund” described in paragraph 1.F.3,
the Project Gutenberg Literary Archive Foundation, the owner of the
Project Gutenberg™ trademark, and any other party distributing a
Project Gutenberg™ electronic work under this agreement, disclaim
all liability to you for damages, costs and expenses, including legal
fees. YOU AGREE THAT YOU HAVE NO REMEDIES FOR
NEGLIGENCE, STRICT LIABILITY, BREACH OF WARRANTY OR
BREACH OF CONTRACT EXCEPT THOSE PROVIDED IN PARAGRAPH
1.F.3. YOU AGREE THAT THE FOUNDATION, THE TRADEMARK
OWNER, AND ANY DISTRIBUTOR UNDER THIS AGREEMENT WILL
NOT BE LIABLE TO YOU FOR ACTUAL, DIRECT, INDIRECT,
CONSEQUENTIAL, PUNITIVE OR INCIDENTAL DAMAGES EVEN IF
YOU GIVE NOTICE OF THE POSSIBILITY OF SUCH DAMAGE.

1.F.3. LIMITED RIGHT OF REPLACEMENT OR REFUND - If you

discover a defect in this electronic work within 90 days of receiving
it, you can receive a refund of the money (if any) you paid for it by
sending a written explanation to the person you received the work
from. If you received the work on a physical medium, you must
return the medium with your written explanation. The person or
entity that provided you with the defective work may elect to provide
a replacement copy in lieu of a refund. If you received the work
electronically, the person or entity providing it to you may choose to
give you a second opportunity to receive the work electronically in
lieu of a refund. If the second copy is also defective, you may
demand a refund in writing without further opportunities to fix the
problem.

1.F.4. Except for the limited right of replacement or refund set forth
in paragraph 1.F.3, this work is provided to you ‘AS-IS’, WITH NO
OTHER WARRANTIES OF ANY KIND, EXPRESS OR IMPLIED,
INCLUDING BUT NOT LIMITED TO WARRANTIES OF
MERCHANTABILITY OR FITNESS FOR ANY PURPOSE.

1.F.5. Some states do not allow disclaimers of certain implied

warranties or the exclusion or limitation of certain types of damages.
If any disclaimer or limitation set forth in this agreement violates the
law of the state applicable to this agreement, the agreement shall be
interpreted to make the maximum disclaimer or limitation permitted
by the applicable state law. The invalidity or unenforceability of any
provision of this agreement shall not void the remaining provisions.

1.F.6. INDEMNITY - You agree to indemnify and hold the Foundation,

the trademark owner, any agent or employee of the Foundation,
anyone providing copies of Project Gutenberg™ electronic works in
accordance with this agreement, and any volunteers associated with
the production, promotion and distribution of Project Gutenberg™
electronic works, harmless from all liability, costs and expenses,
including legal fees, that arise directly or indirectly from any of the
following which you do or cause to occur: (a) distribution of this or
any Project Gutenberg™ work, (b) alteration, modification, or
additions or deletions to any Project Gutenberg™ work, and (c) any
Defect you cause.

Section 2. Information about the Mission

of Project Gutenberg™
Project Gutenberg™ is synonymous with the free distribution of
electronic works in formats readable by the widest variety of
computers including obsolete, old, middle-aged and new computers.
It exists because of the efforts of hundreds of volunteers and
donations from people in all walks of life.

Volunteers and financial support to provide volunteers with the

assistance they need are critical to reaching Project Gutenberg™’s
goals and ensuring that the Project Gutenberg™ collection will
remain freely available for generations to come. In 2001, the Project
Gutenberg Literary Archive Foundation was created to provide a
secure and permanent future for Project Gutenberg™ and future
generations. To learn more about the Project Gutenberg Literary
Archive Foundation and how your efforts and donations can help,
see Sections 3 and 4 and the Foundation information page at
www.gutenberg.org.

Section 3. Information about the Project

Gutenberg Literary Archive Foundation
The Project Gutenberg Literary Archive Foundation is a non-profit
501(c)(3) educational corporation organized under the laws of the
state of Mississippi and granted tax exempt status by the Internal
Revenue Service. The Foundation’s EIN or federal tax identification
number is 64-6221541. Contributions to the Project Gutenberg
Literary Archive Foundation are tax deductible to the full extent
permitted by U.S. federal laws and your state’s laws.

The Foundation’s business office is located at 809 North 1500 West,

Salt Lake City, UT 84116, (801) 596-1887. Email contact links and up
to date contact information can be found at the Foundation’s website
and official page at www.gutenberg.org/contact

Section 4. Information about Donations to

the Project Gutenberg Literary Archive
Foundation
Project Gutenberg™ depends upon and cannot survive without
widespread public support and donations to carry out its mission of
increasing the number of public domain and licensed works that can
be freely distributed in machine-readable form accessible by the
widest array of equipment including outdated equipment. Many
small donations ($1 to $5,000) are particularly important to
maintaining tax exempt status with the IRS.

The Foundation is committed to complying with the laws regulating

charities and charitable donations in all 50 states of the United
States. Compliance requirements are not uniform and it takes a
considerable effort, much paperwork and many fees to meet and
keep up with these requirements. We do not solicit donations in
locations where we have not received written confirmation of
compliance. To SEND DONATIONS or determine the status of
compliance for any particular state visit www.gutenberg.org/donate.

While we cannot and do not solicit contributions from states where

we have not met the solicitation requirements, we know of no
prohibition against accepting unsolicited donations from donors in
such states who approach us with offers to donate.

International donations are gratefully accepted, but we cannot make

any statements concerning tax treatment of donations received from
outside the United States. U.S. laws alone swamp our small staff.

Please check the Project Gutenberg web pages for current donation
methods and addresses. Donations are accepted in a number of
other ways including checks, online payments and credit card
donations. To donate, please visit: www.gutenberg.org/donate.

Section 5. General Information About

Project Gutenberg™ electronic works
Professor Michael S. Hart was the originator of the Project
Gutenberg™ concept of a library of electronic works that could be
freely shared with anyone. For forty years, he produced and
distributed Project Gutenberg™ eBooks with only a loose network of
volunteer support.
Project Gutenberg™ eBooks are often created from several printed
editions, all of which are confirmed as not protected by copyright in
the U.S. unless a copyright notice is included. Thus, we do not
necessarily keep eBooks in compliance with any particular paper
edition.

Most people start at our website which has the main PG search
facility: www.gutenberg.org.

This website includes information about Project Gutenberg™,

including how to make donations to the Project Gutenberg Literary
Archive Foundation, how to help produce our new eBooks, and how
to subscribe to our email newsletter to hear about new eBooks.
Welcome to Our Bookstore - The Ultimate Destination for Book Lovers
Are you passionate about books and eager to explore new worlds of
knowledge? At our website, we offer a vast collection of books that
cater to every interest and age group. From classic literature to
specialized publications, self-help books, and children’s stories, we
have it all! Each book is a gateway to new adventures, helping you
expand your knowledge and nourish your soul
Experience Convenient and Enjoyable Book Shopping Our website is more
than just an online bookstore—it’s a bridge connecting readers to the
timeless values of culture and wisdom. With a sleek and user-friendly
interface and a smart search system, you can find your favorite books
quickly and easily. Enjoy special promotions, fast home delivery, and
a seamless shopping experience that saves you time and enhances your
love for reading.
Let us accompany you on the journey of exploring knowledge and
personal growth!

ebookgate.com