Professional Documents
Culture Documents
Fundamentals of Functional Analysis 1St Edition Ammar Khanfer Online Ebook Texxtbook Full Chapter PDF
Fundamentals of Functional Analysis 1St Edition Ammar Khanfer Online Ebook Texxtbook Full Chapter PDF
https://ebookmeta.com/product/measure-theory-and-integration-1st-
edition-ammar-khanfer/
https://ebookmeta.com/product/functional-analysis-fundamentals-
and-applications-2nd-edition-michel-willem/
https://ebookmeta.com/product/functional-analysis-1st-edition-
jan-van-neerven/
https://ebookmeta.com/product/functional-analysis-2nd-edition-s-
kesavan/
Basic Analysis V: Functional Analysis and Topology 1st
Edition James K. Peterson
https://ebookmeta.com/product/basic-analysis-v-functional-
analysis-and-topology-1st-edition-james-k-peterson/
https://ebookmeta.com/product/functional-interfaces-in-java-
fundamentals-and-examples-1st-edition-ralph-lecessi/
https://ebookmeta.com/product/fundamentals-of-analysis-with-
applications-atul-kumar-razdan/
https://ebookmeta.com/product/fundamentals-of-spatial-analysis-
and-modelling-1st-edition-jay-gao/
https://ebookmeta.com/product/an-introduction-to-functional-
analysis-1st-edition-james-c-robinson/
Ammar Khanfer
Fundamentals
of Functional
Analysis
Fundamentals of Functional Analysis
Ammar Khanfer
Fundamentals of Functional
Analysis
Ammar Khanfer
Department of Mathematics and Sciences
Prince Sultan University
Riyadh, Saudi Arabia
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Singapore Pte Ltd. 2023
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
Preface
The present book provides a thorough treatment and a careful introduction to the field
of functional analysis with a clear exposition to the basic concepts and fundamentals
of functional analysis, and it covers the necessary topics required to master this
area of mathematics. Efforts have been made to make the book a self-study and
reader-friendly text to allow the student to get acquainted with the fundamentals of
advanced analysis and provide the basis for further studies that can be carried more
deeply. Because of these reasons, the book serves as a typical introductory book of
functional analysis for the students. The book contains material suitable for a two-
semester graduate course in analysis, so it is suitable for all graduate mathematics
students. The reader is expected to be familiar with linear algebra, real analysis, in
addition to basics of measure theory and Lebesgue’s integration.
The book is divided into eight chapters, and each chapter contains eight sections.
Chapter 1 introduces normed space and examples of some normed function spaces
and normed sequence spaces. Topology of normed spaces is discussed, with sepa-
rability and compactness properties. The chapter is concluded by equivalence of
norms.
Chapter 2 is, historically, the beginning of the area of functional analysis. The
chapter deals with linear operators and functionals between normed spaces rather
than finite-dimensional vector spaces, and we discuss the topologies of these spaces.
The main result is the famous Hahn–Banach theorem and its consequences. Most
books discuss this theorem in the treatment of Banach spaces. However, the present
book discusses the theorem before introducing Banach spaces to emphasize that the
Hahn–Banach theorem is applicable to general normed space.
Chapter 3 introduces the locally convex spaces and discusses the geometric
sequences of the Hahn–Banach theorem. The separation results and the main sepa-
ration theorems are discussed thoroughly. Then we study extreme points in normed
spaces with some examples, followed by the Krein–Milman theorem.
Chapter 4 introduces the Banach spaces and their geometric and topological prop-
erties. This is the place we introduce completeness for the first time in this text
to emphasize on the fact that the main results of Chap. 3 don’t require complete-
ness. Some great theorems that establish the cornerstones of functional analysis are
v
vi Preface
discussed in detail, such as the Baire category theorem, open mapping theorem,
closed graph theorem, uniform bounded principle, and the Banach–Steinhaus
theorem.
Chapter 5 deals with a particularly important class of Banach spaces in which an
inner product can be endowed. These are called Hilbert spaces. These spaces can
be viewed as the natural extension of the finite-dimensional Euclidean spaces. We
study these spaces in detail and establish essential and fundamental results based on
their rich geometry and orthogonality.
Chapter 6 presents the topology of Banach spaces in comparison to the topology
produced by their norms. These topologies are called the weak and weak-star topolo-
gies. We study how the space structure will be affected when these topologies are
adopted instead of the norm topology. Convergence will be a crucial factor in deter-
mining these topologies’ efficiency, giving rise to more flexible criteria in obtaining
compactness and convergence under these topologies. Some fundamental theorems
will be discussed, such as the Banach–Alaoglu theorem, Goldstine’s theorem, Krein–
Milman theorem, Eberlein–Smulian theorem, James theorem, and more. The chapter
concludes with a discussion of Schauder’s basis and weak basis.
Chapter 7 introduces operators on Banach spaces. The chapter discusses adjoint
of operators and self-adjoint operators in addition to other classes of operators, with
a focus on compact operators, and their approximation by finite-rank operators.
Chapter 8 presents the basics of the geometry of Banach spaces; strict convexity,
uniform convexity, smoothness, and differentiability of functionals are introduced.
Then the idea of renorming Banach spaces is discussed in brief.
The textbook contains more than 150 examples and 350 problems, in addition to
extensive historical and bibliographical remarks. The problems are distributed at the
end of each chapter. Hints and answers are given at the end of the book. The questions
aim to test whether the reader has absorbed the topic’s basics and gained a complete
understanding. Therefore, the levels of the questions range from straightforward to
challenging, and some of them are standard in their subjects that can be found in
many other books. Moreover, these questions help students prepare for the Ph.D.
comprehensive exams. It should be noted that some questions are answered in detail,
while others are left with some hints or key answers. Similarly, some details in the
proofs throughout the book are left to the reader to complete independently.
It is important to know that mastering this book’s topics is not possible without
the time and effort needed to overcome all the difficulties that may arise in terms of
grasping the material and mastering the subject. This will prepare students to think
independently and test their maturity and background to accomplish the task by filling
in the gaps and unveiling all the hidden details whenever needed—a common practice
when one enrolls in a graduate program and inevitable when one starts independent
research.
I have to start by thanking the God most of all for giving me the strength, health,
knowledge, and patience to endure and complete this work successfully.
It is a pleasure to offer my grateful appreciation to everyone who helped me in
completing this book. Writing a book is more exhausting than one can imagine and
more gratifying than one might expect. Before being rewarded, it is important in this
regard to recall our math teachers. I am immensely thankful to everyone who taught
me mathematics at various stages of my life from childhood to the Ph.D. level. Words
cannot express gratitude to them. I sincerely thank all of you.
It is a pleasure to express my sincere thanks to Prince Sultan University for its
continuing support. I also wish to express my deep thanks and gratitude to Prof.
Mahmoud Al Mahmoud, the Dean of the college (CHS), and Prof. Wasfi Shatanawi,
the Chair of the Department of Mathematics (MSD), for their support and recogni-
tion of my work. Many thanks extend to my colleagues of our department and the
colleagues at the Research and Initiative Center (RIC). I am grateful to all of them
for their warm encouragement.
I am profoundly indebted to the anonymous reviewers for their valuable comments
and helpful suggestions which improved the book significantly.
I would also like to thank the editorial board and the staff at Springer who were
involved in the review and publication of the book and bringing my work to fruition. I
would like to single out some individuals for special thanks: Shamim Ahmad, Iydah
Grace, and Ramya Somasundaram for their cooperation and being tremendously
helpful and patient in dealing with this project professionally from the review stage
to the production stage.
A special debt of gratitude is owed to my wife and family for their emotional
support. I deeply acknowledge the love and encouragement that I received from
them during the preparation of this book.
Lastly, I want to thank everyone who supported me with a positive word or feeling.
I heard it all .. and it meant quite something!
vii
Contents
1 Normed Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Gate to Infinite-dimensional Spaces . . . . . . . . . . . . . . . . . . . 1
1.1.2 Why We Need Norms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Function Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 The Space of Continuous Functions . . . . . . . . . . . . . . . . . . . 3
1.2.2 The Space of n–Continuously Differentiable
Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.3 The Space of All Continuous Functions
with Compact Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.4 The Space of Continuous Functions Vanishing
at Infinity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.5 The Space of Functions of Bounded Variation . . . . . . . . . . . 4
1.2.6 The Space of Absolutely Continuous Functions . . . . . . . . . 5
1.2.7 The Space of Riemann Integrable Functions . . . . . . . . . . . . 6
1.2.8 The Space of Riemann–Stieltjes Integrable Functions . . . . 6
1.3 The Notion of Norm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.1 Definition and Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.2 Continuity of Norms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.3 Norms Versus Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4 Definition and Examples of Normed Spaces . . . . . . . . . . . . . . . . . . . . 10
1.4.1 Definition of Normed Spaces . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4.2 Convergence in Normed Spaces . . . . . . . . . . . . . . . . . . . . . . . 11
1.4.3 Examples of Normed Spaces . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5 L p Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.5.1 L p Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.5.2 L ∞ Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.5.3 p and ∞ Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.5.4 Basic Inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.5.5 Measure Theory and Integration: Quick Review . . . . . . . . . 19
ix
x Contents
Ammar Khanfer earned his Ph.D. from Wichita State University, USA. His area of
interest is analysis and partial differential equations (PDEs), focusing on the interface
and links between elliptic PDEs and hypergeometry. He has notably contributed to
the field by providing prototypes studying the behavior of generalized solutions of
elliptic PDEs in higher dimensions in connection to the behavior of hypersurfaces
near nonsmooth boundaries. He also works on the qualitative theory of differential
equations, and in the area of inverse problems of mathematical physics. He has
published articles of high quality in reputable journals.
Ammar taught at several universities in the USA: Western Michigan University,
Wichita State University, and Southwestern College in Winfield. He was a member
of the Academy of Inquiry Based Learning (AIBL) in the USA. During the period
2008–2014, he participated in AIBL workshops and conferences on effective teaching
methodologies and strategies of creative thinking, which made an impact on his
engaging and motivational writing style. He then moved to Saudi Arabia to teach
at Imam Mohammad Ibn Saud Islamic University, where he taught and supervised
undergraduate and graduate students of mathematics. Furthermore, he was appointed
as coordinator of the PhD program establishment committee in the department of
mathematics. In 2020, he moved to Prince Sultan University in Riyadh, and has been
teaching there since then.
xix
Chapter 1
Normed Spaces
1.1 Motivation
and the goal is to find an appropriate function y that would maximize or minimize
I (y). We have the following observations here:
(1) I (y) is a real-valued function of functions, i.e., the domain of this function is
the set of all continuous or measurable functions on [a, b], and the range of the
function is R.
(2) Because of the integral, the function is linear, i.e., I (c1 y1 + c2 y2 ) = c1 I (y1 ) +
c2 I (y2 ), so we require the function c1 y1 + c2 y2 to be a candidate function that
compete with other functions. Hence this function must also belong to the same
set. But a set with such property is nothing but a vector space. So the set of
candidate functions must be a vector space.
(3) The elements of this vector space are of the form of functions, so we can call it
a “function space”. If we consider a vector v ∈ Rn , then v = (x1 , x2 , . . . , xn ),
so v has n components in the space. In order to write the function f (x) in the
same representation, we see that x varies from a to b, and f will depend on
each value x from a to b, so writing f in the same representation will result in
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 1
A. Khanfer, Fundamentals of Functional Analysis,
https://doi.org/10.1007/978-981-99-3029-6_1
2 1 Normed Spaces
an array of the form ( f (x))x∈[a,b] , and this means that f (x) will have infinitely
many components.
These spaces containing the candidate functions are, in fact, spaces of infinite dimen-
sions, and consequently many rules and properties regarding the finite-dimensional
spaces fail to satisfy.
In order to obtain a useful tool, these new spaces must be equipped with a geo-
metric structure. Convergence is one of the most important and basic concepts in
mathematical analysis through which we can characterize continuity of functions,
compactness of sets, convergence of sequences, and many other useful notions and
concepts. The convergence is based on the notion of distance between numbers which
is given by the absolute value function. Now, the elements in our new spaces are no
longer numbers but functions, so how to define the “length” of a function and the
“distance” between functions? What is the appropriate geometric structure that can
be provided with this feature? What is the generalized version of the absolute value
function? These questions can be answered by the notion of “norm” , which will be
essential for any space of functions to behave well and turn the space into a useful
mathematical tool.
Keeping in mind that the original problem that motivated us was the problem of
optimizing an integral of the form
b
I (y) = y(x)d x,
a
the only tool at our disposal to treat this problem is the derivative, but I (y) is a function
of functions, and we do not know how to give a derivative for such functions because
the limit needs the concept of convergence and distance between functions to control
these distances and make them as small as desired. Hence, if the space of functions
can be endowed with a norm, the space will have a good geometric structure that
allows measuring the length of a function and the distance between two functions,
in addition to algebraic structure that allows the functions to interact between each
other via addition and other operations.
1.2 Function Spaces 3
Review of Vector Spaces. Recall that a vector (or linear) spaceV is a set of elements
(called vectors) in which two operations are defined; the addition of two vectors and
the scalar multiplication of a vector with a constant in a field F. These operations
obey the following axioms:
(1) If v1 , v2 ∈ V and c1 , c2 ∈ R then c1 v1 + c2 v2 ∈ V.
(2) Commutative and associative laws of addition hold.
(3) Distributivity of scalar multiplication with respect to field and vector additions.
So there exist v1 , v2 ∈ V and a, b ∈ F such that a(v1 + v2 ) = av1 + av2 , and
(a + b)v = av + bv.
(4) Additive and multiplicative identities exist. There exists 0 ∈ V such that v + 0 =
v for all v ∈ V, and there is 1 ∈ V such that a.v = v for all v ∈ V .
(5) Scalar multiplication is associative, that is, a(bv) = (ab)v for all v ∈ V.
(6) There exists an additive inverse −v such that v + (−v) = 0.
Taking into account the following remarks:
(1) It is because of axiom 1 these spaces are also called “linear spaces” since all
elements obey linearity.
(2) We are interested only in real-valued functions; hence we will always let F = R.
Any space obeying these axioms is called “vector space”, or sometimes “linear
space”. The two names are used interchangeably, though the former is rather more
common.
A vector space is called “function space” if all of its elements are functions. Some
examples of function spaces are
Example 1.2.1 C[a, b]: the space of continuous functions on [a, b]. This is clearly a
function vector space since the scalar multiple of a continuous function is continuous,
and the summation of two continuous functions is continuous; hence this defines a
vector space.
Example 1.2.2 C n [a, b]: the space of n−continuously differentiable functions, that
is, f, f , . . . , f (n) are all continuous on [a, b]. It is clear that for every n ≤ m we
have
4 1 Normed Spaces
where it is understood that C 0 [a, b] = C[a, b]. Moreover, the space C ∞ [a, b] defined
by
∞
∞
C [a, b] = C n [a, b]
n=0
consists of all continuous functions that are infinitely differentiable, i.e., smooth
functions that have derivatives of all orders.
Example 1.2.3 Cc (R): the space of all continuous functions f on R with compact
support. A compact suppor t of a function, denoted by supp( f ), is defined as
So, for every f ∈ Cc (R) there exists a compact interval, say [a, b] such that f (x) = 0
for all x ∈
/ [a, b].
Example 1.2.4 C0 (R): the space of continuous functions on R s.t. lim f (x) = 0,
x→±∞
f ∈ C0 (R).
Cc (X ) ⊆ C0 (X ) ⊆ C(X ).
Example 1.2.5 BV [a, b]: the space of functions of bounded variation on [a, b].
Let f be a real-valued function on [a, b] and P is a partition of [a, b], and let the
variation of f with respect to P be defined by
1.2 Function Spaces 5
n
vab ( f ; P) = | f (xi ) − f (xi−1 )| .
i=1
V ( f ) = supvab ( f ; P) < ∞,
P
where the supremum is taken over all possible partitions of [a, b]. This is a function
vector space. It can be shown1 that the function
⎧
⎪
⎨x sin( x ) 0 < x ≤ 1
1
f (x) =
⎪
⎩
0 x = 0.
The stronger absolute continuity condition suffices to give the inclusion above.
Example 1.2.6 AC[a, b]: the space of absolutely continuous functions on [a,b].
Recall that a function is called absolutely continuous on [a, b] if given > 0, there
exists δ > 0 such that if {[xi , yi ] : i = 1, 2, . . . , n} is a finite collection of pairwise
disjoint subintervals in [a, b] and
n
|yi − xi | < δ
i=1
then
n
| f (yi ) − f (xi )| < .
i=1
Example 1.2.7 R[a, b]: the space of Riemann integrable functions over [a, b].
It is well known that the Riemann integral is linear, so the space R[a, b] is a
function vector space.
Example 1.2.8 R(g)[a, b]: the space of Riemann–Stieltjes integrable functions. Let
b
f and g be bounded functions on [a, b], and consider the integral a f dg. This
integral is defined as follows:
Let
Pn = a = t0 < t1 < · · · < tn = b
n
S( f, g; P) = f (τ j )[g(t j ) − g(t j−1 )],
j=1
We denote the space of all Riemann–Stieltjes integrable functions over [a, b] with
respect to the function g by R(g)[a, b]. It is easy to see that the Riemann–Stieltjes
integral is linear with respect to f and to g, so the space is a function vector space.
1.3 The Notion of Norm 7
• The 2-norm,
1
x 2 = |x1 |2 + |x2 |2 + · · · + |xn |2 2
.
• The 2-norm,
∞ 1/2
x 2 = |xi |2 ,
i=1
These are the most common norms although we have many more norms to deal with
other than these. The 2-norm can be generalized to any number p, to become
∞ 1/ p
x p = |xi | p
i=1
for functions.
The absolute value is given to ensure the nonnegativity of the integral value and
fulfill axiom 1, and the power 1/ p in the 2-norm is taken to ensure homogeneity and
fulfil axiom 3. Surprisingly, axiom 2 is the most problematic axiom, and it fails to
hold in many cases. The proposed function can still give nice geometric properties
for the space, and thus it has been considered and given the name seminorm.
It is well known from elementary analysis courses that the absolute value function is
continuous throughout its domain. Since the norm function is a generalization of the
absolute function, we expect this would also be the case for norms, which is indeed
a fundamental property of norms.
Proposition 1.3.1 The norm function is continuous.
x = x−y+y ≤ x−y + y ,
| x − y |≤ x−y .
Another important notion that is associated with norms is the metric. Recall a metric
is a function d : X × X → [0, ∞) satisfying the following:
(1) Positive Definiteness: d(x, y) ≥ 0 and d(x, y) = 0 iff x = y.
(2) Symmetry: d x, y) = d(y, x).
(3) Triangle Inequality: d(x, y) ≤ d(x, z) + d(z, y).
The distance function is the metric on Rn , for if x, y ∈ Rn then
d(x, y) = (x1 − y1 )2 + · · · + (xn − yn )2 .
Note here that a metric in general doesn’t satisfy homogeneity condition (axiom 3
of norms). So every norm induces a metric by the following definition of metric
d(x, y) = x − y (1.3.1)
Indeed,
d(x, y) = x − y ≥ 0,
and
d(x, y) = 0 ⇒ x − y = 0 ⇒ x = y.
d(x, y) = x − y
≤ x−z + z−y
= d(x, z) + d(z, y).
On the other hand, not every metric can induce a norm. Consider, for example,
the following metric:
1 x=y
d(x, y) = .
0 x = y
This metric obviously does not satisfy axiom 3 of norms, so it cannot induce a norm.
The following proposition provides sufficient conditions under which a metric can
induce a norm.
Proposition 1.3.2 If d is a metric and satisfies the following:
Consider the set containing all functions defined on the interval on [a, b]. Checking
the axioms of the norm above, we can conclude that this set forms a vector space,
with elements being of the form of functions. As said above, this space is of infinite
dimensions. By adding some specific conditions on the functions, the space gets
smaller because these conditions work as constraints that may rule out some functions
that do not obey the conditions. Therefore, more the conditions we impose, more
restrictive the space becomes. We begin by the definition.
1.4 Definition and Examples of Normed Spaces 11
Definition 1.4.2 (Convergent Sequence) Let (X, · ) be a normed space, and (xn )
be a sequence in X. Then we say that xn converges to x if for every > 0 there
exists N ∈ N such that xn − x < for all n ≥ N . This convergence is denoted by
xn − x → 0, and x is called the limit of (xn ).
Definition 1.4.3 (Bounded Sequence) Let (X, · ) be a normed space, and (xn ) be
a sequence in X. Then we say that xn is bounded on X if sup xn ≤ M < ∞.
n
Now, let
M = max{ x1 , x2 , . . . , x N −1 , x + 1}.
x − y = x − xn + xn − y
≤ x − xn + xn − y −→ 0.
The converse of (1) is not necessarily true as can be seen clearly from the sequence
xn = (−1)n for example.
12 1 Normed Spaces
Example 1.4.5 (Euclidean Spaces) The space Rn with the Euclidean norm:
x 2 = |x1 |2 + |x2 |2 + · · · + |xn |2
for x = (x1 , x2 , . . . , xn ) is a normed space, and the norm · 2 represents the distance
between a point and the origin. We usually denote it by (Rn , · 2 ). Using Cauchy–
Schwarz inequality, we obtain the triangle inequality. The spaces (Rn , · 1 ) and
(Rn , · ∞ ) are other normed spaces. It’s the same vector space but with different
norms. These norms affect the geometric and topological properties of the space.
Example 1.4.6 (Space of Continuous Functions) Consider the space C([a, b])
defined in Example 1.2.1. Define the following norm:
f ∞ = sup | f (x)| .
x∈[a,b]
This is a generalization to the sup-norm · ∞ with the point x ∈ Rn replaced with the
function f (x). The norm is well defined on C[a, b] since every continuous function
on a closed interval has a maximum. If f ∞ = 0 then f ≡ 0 on [a, b]. Moreover,
but with uncountably infinite components. The norm is well defined since any con-
tinuous function on [a, b] is integrable on [a, b]. If f, g ∈ C[a, b] then
b b b
f +g 1 = | f + g| ≤ |f|+ |g| = f 1 + g 1 .
a a a
Also,
1.4 Definition and Examples of Normed Spaces 13
b b
cf 1 = |c f (x)| d x = |c| | f (x)| d x. = |c| . f 1 .
a a
Let f 1 = 0. Then
b
| f (x)| d x = 0,
a
and since f is continuous on [a, b], this cannot occur without f ≡ 0 on [a, b].
Indeed, suppose on the contrary that f = 0. Then there exist > 0 and an interval
I ⊂ [a, b] such that | f x)| > on I. Then
b
f 1= | f (x)| d x > | f (x)| d x > (I ) > 0,
a I
where (I ) is the length of the interval. But this contradicts the assumption that
f 1 = 0. This proves that f 1 is a norm on C[a, b].
Example 1.4.8 (Space of Polynomials) Consider the space P[a, b] be the space of
polynomials on [a, b]. It is clear that
Then the space (P[a, b], · ∞ ) is a normed space. All axioms of the norm are clearly
satisfied.
f s = | f (a)| .
It is clear that f satisfies axioms 1, 3, and 4. Indeed, f ≥ 0 for all f ∈ BV [a, b],
and
c f = |c f (a)| = |c| | f (a)| = |c| f .
Moreover, the triangle inequality gives axiom 4. It remains axiom (2) which is obvi-
ously not correct since we may have a nonzero function f with f (a) = 0. Hence
(BV [a, b], · s ) is not a normed space, but rather a seminormed space.
Another candidate of a norm is
f BV = | f (a)| + V ( f ),
14 1 Normed Spaces
where
n
V ( f ) = sup | f (xi ) − f (xi−1 )|
P i=1
is the total variation of f , which is finite since f ∈ BV [a, b]. It is clear that
f BV ≥0
for all
f ∈ BV [a, b].
n
| f (xi ) − f (xi−1 )| = 0
i=1
for all possible partitions P of [a, b]. Finally, using properties of supremum and
definition of total variation, it is easy to show that
V ( f + g) ≤ V ( f ) + V (g),
f +g BV = | f (a) + g(a)| + V ( f + g)
≤ | f (a)| + |g(a)| + V ( f ) + V (g)
= f BV + g BV .
(1) The space of all bounded convergent sequences with the supremum norm,
denoted by (c, · ∞ ).
(2) The space of all sequences converging to zero, (c0 , · ∞ ), where
c00 = {(xn ) ∈ c0 : x j = 0
c00 ⊂ c0 ⊂ c
and that each one of these spaces endowed with · ∞ is a normed space.
1.5 L p Spaces
1.5.1 L p Spaces
We devote this section to one of the most important normed spaces. These are the
spaces of Lebesgue integrable functions which have numerous applications in func-
tional analysis.2 These spaces consist of measurable functions that are Lebesgue
integrable, so prior knowledge of measure theory and Lebesgue integration is impor-
tant to understand the basic theory of the spaces.3 The goal of this section is to remind
the reader about the important concepts and results regarding L p theory that may
help us in this text.
Recall the space L[a, b] is the space consisting of all Lebesgue integrable func-
tions on [a, b], that is, those functions f : [a, b] → R such that
b
| f (x)| d x < ∞.
a
This is a normed space, where it is understood that the elements of the space are
viewed as equivalence classes of functions rather than functions, which is possible
due to the fact that in measure theory, measurable functions that are equal except at
a set of measure zero are considered equal. In this case the norm defined in (1.5.1)
is a norm.
The space L[a, b] can also be generalized to L p [a, b], the space of all functions
such that | f | p is Lebesgue integrable on [a, b] for every f ∈ L p [a, b], where 1 ≤
p < ∞, together with the norm
b 1/ p
f p = | f (x)| p d x .
a
1.5.2 L ∞ Spaces
The space L ∞ () is defined as the space of all essentially bounded measurable
functions, with the norm
f ∞ = ess sup | f | ,
So the function f is bounded on [a, b], except maybe on a set of measure zero. Thus,
(L ∞ (), · ∞ ) is a normed space.
The following well-known inclusion is important and is proved in most texts on
real analysis.
Proposition 1.5.3 Let 1 ≤ r < s < ∞, and let μ() < ∞. Then,
Proof Let f ∈ L ∞ (). Then | f | < M a.e. for some M > 0. Choose s ≥ 1. Then
| f |s dμ ≤ M s · μ() < ∞,
|f| =
r
|f| +
r
| f |r
Ec
E
≤ | f |s + μ(E)
E
≤ ∞.
1.5 L p Spaces 17
So f ∈ L r ().
Another important class of spaces that can be derived from L p spaces is the class
of p spaces. The space p can be constructed as follows: In the space L p , consider
X = N, the set of natural numbers, and Σ is a σ−algebra of subsets of N, and ν is
the counting measure. Then L p = L p (N, Σ, υ), and f ∈ L p (N) will be of the form
f = ( f 1 , f 2 , . . . , ) such that
⎛ ⎞1/ p
∞
p
f p =⎝ f j ⎠ < ∞.
j=1
Then
1/ p
f p = cn χ[n,n+1] (x)d x =( cnp )1/ p = cn p .
Hence f p < ∞ if and only if cn p < ∞. So, the p space for 1 ≤ p < ∞ is
defined as 1/ p
p = {(xn ), such that xn p = |xn | p < ∞}.
The spaces ( p , · p) and (∞ , · ∞) are linear normed spaces. It is easy to see that
p ⊆ ∞
1 ⊂ p ⊂ q ⊂⊂ ∞ .
18 1 Normed Spaces
ap bq
ab ≤ + .
p q
Proof Since f (x) = − ln x is concave up, the line tln x + (1 − t) ln y must be above
the graph of f between x and y for all t ∈ (0, 1). Therefore, we have
ln(t x + (1 − t)y) ≥ t ln x + (1 − t) ln y.
Now substitute with x = a p and y = bq and take the exponential of both sides of the
equation.
2. Holder’s Inequality: Let f ∈ L p () and g ∈ L q (), for 1 ≤ p, q ≤ ∞, p and q
are conjugates in the sense that
1 1
+ = 1.
p q
f +g p ≤ f p + g p .
1.5 L p Spaces 19
Proof The inequality is clear for p = 1, and the case p = ∞ follows from the
properties of the supremum. Let 1 < p < ∞. Then
f +g p
p ≤ f +g p/q
p ( f p + g p ).
xn yn 1 ≤ xn p · yn q
for p > 1.
(2) Minkowski’s Inequality for Sequences: If xn , yn ∈ p then
xn + yn p ≤ xn p + yn p
for p ≥ 1.
Now, we state (without proof)4 some fundamental results from Lebesgue’s theory of
integration.
Definition 1.5.4 (Simple Function) A function n ϕ is called simple function if it is
defined on a set A, and is written as ϕ(x) = i=1 ci χ Ai (x), where Ai = A, and
χ Ai (x) is the characteristic function (or indicator function), and is defined as
1 x ∈ Ai
χ Ai (x) = .
0 x∈ / Ai
4 The interested reader is referred to Ref. [51], or any other graduate textbook on real analysis.
20 1 Normed Spaces
where the supremum is taken over all possible simple functions ϕ with 0 ≤ ϕ ≤
f. The function f is said to be integrable over E if
f < ∞.
E
and b
f (x)d x = F(b) − F(a).
a
In general normed spaces, the starting key in determining the topology of the space
is the norm, and the topology generated by the norm is called the norm topology. The
norm determines the topology of the space, and the shape of the unit ball determines
the geometric properties of the space. Determining the norm induced on the unit ball
is the cornerstone that builds the entire structure of the space, so the role of the unit
ball as the building block will continue to work in general normed spaces of infinite
dimensions.
The intervals in R are generalized in Rn to what are known as “balls”. The unit
ball in Rn , which is denoted by B1 (0), is a ball of radius one and centered at the
origin, and is given by
The closure of this ball is the closed unit ball consisting of all points satisfying
22 1 Normed Spaces
The boundary of the closed unit ball is known as the unit sphere, and is given by
|x1 | + |x2 | = 1.
x12 + x22 = 1
gives the shape of unit circle for S1 . Applying the norm · ∞ gives the shape of a
square. The norm · 2 gives the smoothest shape for the sphere, and as p exceeds
2, the circle gets flattened until it reaches the shape of a square when p = ∞. It is,
therefore, obvious that
Definition 1.6.1 (Open Ball) Let X be a normed space with a norm · . Then the
open ball of X of radius r and centered at x0 is given by
Br (x0 ) = {x ∈ X : x − x0 < r }.
Br (x0 ) = {x ∈ X : x − x0 ≤ r }.
When r = 1 and x0 = 0, then the ball is called “unit ball”. The unit sphere is given
by
S1 (x0 ) = {x ∈ X : x − x0 = 1}.
It suffices to study the structure of the unit balls and unit spheres in many cases
since all other balls and spheres can be obtained using shifting and scaling operations.
In fact, one advantage, among many others, of unit spheres is that one can always
reflect elements of the space to elements on the unit sphere by normalizing the
element. It should be noted that the term ball here does not reflect the actual shape
of the ball. The only space where we can use the terms “ball” and “sphere” in their
1.6 Topology of Normed Spaces 23
actual meanings is R3 . In n > 3, these building units are called “hyperballs” and
“hyperspheres”, or n-balls and n-spheres. We, however, can simply call them balls
and spheres if the context is clear from confusion.
In R, open and closed sets are defined based on intervals. Similarly, the balls play
the same role in normed spaces.
Definition 1.6.2 (Open Set) Let X be a normed space and O ⊆ X. Then, O is called
open set if for every x ∈ O, there exists r > 0 such that Br (x) ⊂ O. The set F ⊆ X
is called closed if F c (the complement of F) is open.
It is well known in elementary analysis that the arbitrary union of open sets is
open, and the finite intersection of open sets is open. The same property holds for
general normed space. The following fact provides a sequential characterization of
closed sets. The proof is straightforward and has been discussed in undergraduate
analysis courses, so it will simply be presented as a definition.
Definition 1.6.3 (Closed Set) A set F is called closed set in a normed space X if
for every sequence {xn } ⊆ F and xn −→ x, we have x ∈ F.
One immediate consequence is that the real line R is closed. One can extend this
claim by induction to say that the vector space Rn is closed, and it is well known
from linear algebra that vector spaces of the same dimension are isomorphic, which
means there exists a linear bijection mapping between them. We infer the following:
The concept of closedness helps us in many applications, but we will only mention
two examples: quotient normed spaces and distance to sets. For the first application,
we would like to study the quotient space of a normed space. Let M be a subspace
of a normed space X . Then, we define the quotient X/M as
where the elements of this quotient are cosets [x] = x + M, formed by equivalence
classes. Here, we define the equivalence relation ∼ as x1 ∼ x2 iff x1 − x2 ∈ M. It’s
clear that
24 1 Normed Spaces
[x] + [y] = [x + y]
and
[cx] = c[x].
[x] Q = x+M Q = 0,
then x may not be in M, but the infimum will take a zero value. We will see later
that the space (X \ M, · Q ) is normed if and only if the set M is closed. We need
to define the closedness of a set in a general normed space. If M is closed, then
and since the closed set M contains its boundary and limits points, this means that
if x0 ∈
/ M then x0 ∈ M c , which is an open set, so we can find a ball U = B (x0 )
around x0 such that x0 ∈ U ⊂ M c and so U ∩ M = Ø. This implies that
inf { x0 + z : z ∈ M} > 0.
z
Therefore, if
inf { x0 + z : z ∈ M} = 0
z
then x0 ∈ M. The other conditions can easily be verified. Let c ∈ R and x ∈ X , then
it is easy to show that · Q is a norm. Indeed,
[x] + [y] Q = x + z1 + y + z2 Q ≤ x − z1 Q + y − z2 Q .
The second application is the distance to sets. Recall that the distance between a
point x and a set Y is given by
dist(x, Y ) = inf x − y .
y∈Y
If Y is closed, then the distance is strictly greater than zero because Y contains its
boundary. Indeed, if dist(x, Y ) = 0, then there exists a sequence yn ∈ Y such that
yn −→ x. But this means x ∈ Y = Y since Y is closed, which is a contradiction.
Hence, we have
Proposition 1.6.6 If Y is closed subset of a normed space X and x ∈
/ Y , then
dist(x, Y ) > 0.
The preceding proposition predicts that the distance between a point and a closed
subset of a normed space is strictly greater than zero. The following important result
by Riesz in 1918 gives a good lower bound when the set is a subspace.
Lemma 1.6.7 (Riesz’s Lemma) Let Y be a closed subspace of a normed space
X. Then, for every δ, 0 < δ < 1, there exists x0 ∈ X such that x0 = 1 and
dist(x0 , Y ) ≥ δ.
Proof Since Y is closed, dist(x, Y ) = d > 0. Let 0 < δ < 1. Then, by definition of
infimum, there exists y0 ∈ Y such that
d
x − y0 ≤ . (1.6.2)
δ
Note that
26 1 Normed Spaces
y0 + x − y0 y ∈ Y. (1.6.3)
Let
x − y0
x0 = .
x − y0
1.6.6 Compactness
Another fundamental set is the compact set. Recall a set K is compact if every open
cover of K has a finite subcover. By an open cover, we mean a collection of open
sets. The following definition plays a dominant role in analysis.
Definition 1.6.8 (Compact Set) A set K is said to be compact if every sequence in
K has a subsequence that converges in K .
The characterization in the previous definition is known as the sequentially com-
pact definition of compact sets. In an introductory course of analysis, it was shown
that a set is compact if and only it is sequentially compact in a vector space.5 We have
used this characterization to define compact sets, which seems important and inter-
esting since it connects the compactness property with the completeness property.6
These concepts are cornerstones of analysis and can be used in establishing many
important results. We will state two great theorems related to this notion that have
numerous applications and consequences. The first is considered one of the most
important theorems in analysis, and its proof is available in most undergraduate
textbooks in analysis; thus, it will not be mentioned here.
Theorem 1.6.9 (Heine–Borel Theorem) In a finite-dimensional space, a set is com-
pact if and only if it is closed and bounded.
The second is a fundamental result that holds in any topological space, not nec-
essarily normed.
5 The reader has already studied compactness in an introductory course of analysis, so we shall not
repeat the discussion here.
6 The completeness property will be discussed in detail in Sect. 4.1.
1.6 Topology of Normed Spaces 27
Y1 = span{x1 }.
Y2 = span{Y1 , x2 }.
Then,
1
x1 − x2 ≥ .
2
Since X has infinite dimensional, we can continue inductively to construct a sequence
{xn } such that
Yn = span{x1 , x2 , . . . , xn−1 }
and
1
dist(Yn , xn+1 ) ≥ .
2
Hence
1
xn+1 − xi ≥
2
28 1 Normed Spaces
for all i ≤ n. So we conclude that the sequence {xn } does not have any convergent
subsequence, which implies the unit sphere S is not compact. But the unit sphere is
closed in the closed unit ball, the result then follows from Theorem 1.6.10.
The previous result can be used to identify infinite-dimensional spaces by their non-
compact closed unit balls.
Example 1.6.12 (Standard Sequence) In ∞ and 2 , consider the sequence {δn } ∈ S,
defined as
δi = (0, 0, . . . , 1 , 0, . . .), i = 1, 2, . . . .
ithcoordinate
That is, δi = 1 on the ith coordinate and zero elsewhere. This is called the
standard sequence, and it could also be represented as
δn = (δnm )∞
m=1 = (0, . . . , 1, 0, . . .), n ∈ N},
nth
where
1 n=m
δnm =
0 n = m
1.6.7 Separability