Fa e PDF
Fa e PDF
Fa e PDF
of
Func€ionel
Arnalysis
Kutateladze
Fundamentals of Functional Analysis
Kluwer Texts in the Mathematical Sciences
VOLUME 12
The titles published in this series are listed at the end of this volume.
Fundamentals of
Functional Analysis
by
S. S. Kutateladze
Sobolev Institute of Mathematics,
Siberian Branch of the Russian Academy of Sciences,
Novosibirsk, Russia
This is a concise guide to basic sections of modern functional analysis. Included are
such topics as the principles of Banach and Hilbert spaces, the theory of multinormed
and uniform spaces, the Riesz-Dunford holomorphic functional calculus, the Fredholm
index theory, convex analysis and duality theory for locally convex spaces.
With standard provisos the presentation is self-contained, exposing about a hun-
dred famous "named" theorems furnished with complete proofs and culminating in
the Gelfand-Nalmark-Segal construction for C*-algebras.
The first Russian edition was printed by the Siberian Division of "Nauka" Pub-
lishers in 1983. Since then the monograph has served as the standard textbook on
functional analysis at the University of Novosibirsk.
This volume is translated from the second Russian edition printed by the Sobolev
Institute of Mathematics of the Siberian Division of the Russian Academy of Sciences·
in 1995. It incorporates new sections on Radon measures, the Schwartz spaces of
distributions, and a supplementary list of theoretical exercises and problems.
This edition was typeset using AMS-'lEX, the American Mathematical Society's
'lEX system.
To clear my conscience completely, I also confess that := stands for the definor,
the assignment operator, <J marks the beginning of a (possibly empty) proof, and [>
signifies the end of the proof.
s. K utateladze
Preface
to the First Russian Edition
As the title implies, this book treats functional analysis. At the turn of the century
the term "functional analysis" was coined by J. Hadamard who is famous among
mathematicians for the formula of the radius of convergence of a power series. The
term "functional analysis" was universally accepted then as related to the calculus of
variations, standing for a new direction of analysis which was intensively developed by
V. Volterra, C. Arzela, S. Pincherle, P. Levy, and other representatives of the French
and Italian mathematical schools. J. Hadamard's contribution to the present discipline
should not be reduced to the invention of the word "functional" (or more precisely to
the transformation of the adjective into a proper noun). J. Hadamard was fully aware
of the relevance of the rising subject. Working hard, he constantly advertised problems,
ideas, and methods just evolved. In particular, to one of his students, M. Frechet, he
suggested the problem of inventing something that is now generally acclaimed as the
theory of metric spaces. In this connection it is worth indicating that neighborhoods
pertinent to functional analysis in the sense of Hadamard and Volterra served as
precursors to Hausdorff's famous research, heralding the birth of general topology.
Further, it is essential to emphasize that one of the most attractive, difficult, and
important sections of classical analysis, the calculus of variations, became the first
source of functional analysis.
The second source of functional analysis was provided by the study directed to
creating some algebraic theory for functional equations or, stated strictly, to simplify-
ing and formalizing the manipulations of "equations in functions" and, in particular,
linear integral equations. Ascending to H. Abel and J. Liouville, the theory of the
latter was considerably expanded by works of I. Fredholm, K. Neumann, F. Noether,
xii Preface
A. Poincare, et al. The efforts of these mathematicians fertilized soil for D. Hilbert's
celebrated research into quadratic forms in infinitely many variables. His ideas, devel-
oped further by F. Riesz, E. Schmidt, et al., were the immediate predecessors of the
axiomatic presentation of Hilbert space theory which was undertaken and implemented
by J. von Neumann and M. Stone. The resulting section of mathematics h.as vigor-
ously influenced theoretical physics, first of all, quantum mechanics. In this regard
it is instructive as well as entertaining to mention that both terms, "quantum" and
"functional," originated in the same year, 1900.
The third major source of functional analysis encompassed Minkowski's geometric
ideas. His invention, the apparatus for the finite-dimensional geometry of convex bod-
ies, prepared the bulk of spatial notions ensuring the modern development of analysis.
Elaborated by E. Helly, H. Hahn, C. Caratheodory, I. Radon, et al., the idea of convex-
ity has eventually shaped the fundamentals of the theory of locally convex spaces. In
turn, the latter has facilitated the spread of distributions and weak derivatives which
were recognized by S. L. Sobolev as drastically changing all tools of mathematical
physics. In the postwar years the geometric notion of convexity has conquered a new
sphere of application for mathematics, viz., social sciences and especially economics.
An exceptional role in this process was performed by linear programming discovered
by L. V. Kantorovich.
The above synopsis of the history of functional analysis is schematic, incomplete,
and arbitrary (for instance, it casts aside the line of D. Bernoulli's superposition prin-
ciple, the line of set functions and integration theory, the line of operational calculus,
the line of finite differences and fractional derivation, the line of general analysis, and
many others). These demerits notwithstanding, the three sources listed above reflect
the main, and most principal, regularity: functional analysis has synthesized and pro-
moted ideas, concepts, and methods from classical sections of mathematics: algebra,
geometry, and analysis. Therefore, although functional analysis verbatim means anal-
ysis of functions and functionals, even a superficial glance at its history gives grounds
to claim that functional analysis is algebra, geometry, and analysis of functions and
functionals.
A more viable and penetrating explanation for the notion of functional analy-
sis is given by the Soviet Encyclopedic Dictionary: "Functional analysis is one of the
principal branches of modern mathematics. It resulted from mutual interaction, unifi-
cation, and generalization of the ideas and methods stemming from all parts of classical
mathematical analysis. It is characterized by the use of concepts pertaining to various
Preface xiii
abstract spaces such as vector spaces, Hilbert spaces, etc. It finds diverse applications
in modern physics, especially in quantum mechanics."
The S. Banach treatise Theorie des Operationes Lineares, printed half a century
ago, inaugurated functional analysis as an essential activity in mathematics. Its influ-
ence on the development of mathematics is seminal: Omnipresent, Banach's ideas,
propounded in the book, captivate the realm of modern mathematics.
An outstanding contribution toward progress in functional analysis was made by
the renowned Soviet scientists: I. M. Gelfand, L. V. Kantorovich, M. V. Keldysh, A.
N. Kolmogorov, M. G. KreIn, L. A. Lyusternik, and S. L. Sobolev. The characteristic
feature of the Soviet school is that its research on functional analysis is always con-
ducted in connection with profound applied problems. The research has expanded the
scope of functional analysis which becomes the prevailing language of the applications
of mathematics.
The next fact is demonstrative: In 1948 even the title of Kantorovich's insightful
article Functional Analysis and Applied Mathematics was considered paradoxical, but
it provided a basis for the numerical mathematics of today. And in 1974 S. L. Sobolev
stated that "to conceive the theory of calculations without Banach spaces is just as
impossible as trying to conceive of it without the use of computers".
The exponential accumulation of knowledge within functional analysis is now
observed alongside a sharp rise in demand for the tools and concepts of the discipline.
The resulting conspicuous gap widens permanently between the current level of anal-
ysis and the level fixed in the literature accessible to the reading community. To alter
this ominous trend is the purpose of the present book.
Preface
to the Second Russian Edition
For more than a decade the monograph has served as a reference book for compul-
sory and optional courses in functional analysis at Novosibirsk State University. This
time span proves that the principles of compiling the book are legitimate. The present
edition is enlarged with sections addressing the fundamentals of distribution theory.
Theoretical exercises are supplemented and the list of references is updated. Also,
inaccuracies, mostly indicated by my colleagues, have been corrected.
I seize the opportunity to express my gratitude to all those who helped me in the
preparation of the book. My pleasant debt is to acknowledge the financial support of
xiv Preface
the Sobolev Institute of Mathematics of the Siberian Division of the Russian Acade-
my of Sciences, the Russian Foundation for Fundamental Research, the International
Science Foundation and the American Mathematical Society during the compilation
of the second edition.
March, 1995 S. Kutateladze
Chapter 1
An Excursion into Set Theory
1.1. Correspondences
1.1.1. DEFINITION. Let A and B be sets and let F be a subset of the product
A X B := {(a, b): a E A, b E B}. Then F is a correspondence with the set
of departure A and the set of arrival B or just a correspondence from A (in )to B.
1.1.2. DEFINITION. For a correspondence F C A x B the set
F- I := {(b, a) E B x A: (a, b) E F}
These are the partial correspondences of F. In this regard F itself is often sym-
bolized as F( . , . ) and referred to as a correspondence in two arguments. This
beneficial agreement is effective in similar events.
1.1.4. DEFINITION. The composite correspondence or composition of corre-
spondences F C A x B and Gee x D is the set
HoGoF= U F-\b)xH(e).
(b,c)EG
Go F = U F-l(b) x G(b).
bEimF
<] Insert H:= G, G:= limP and F:= F into 1.1.10. I>
set, or even a poset. It is said that "x is less than y," or "y is greater than x,"
or "x ~ y," or "y 2: x," or "x is in relation (1 to y," or "x and y belong to (1,"
etc. Analogous agreements apply customarily to a preordered set, i.e. to a set
furnished with a preorder. The convention is very propitious and extends often
to an arbitrary relation. However, an equivalence is usually denoted by the signs
like "'.
1.2.3. EXAMPLES.
(1) The identity relation; each subset Xo of a set X bearing a relation (1
is endowed with the (induced) relation (10 := (1 n Xo x Xo.
(2) If (1 is an order (preorder) in X, then (1-1 is also an order (preorder)
which is called reverse to (1.
(3) Let I : X -+ Y be a mapping and let 7 be a relation on Y. Consider
the relation 1-1 070 I appearing on X. By 1.1.10,
Hence it follows that (Xl, XZ) E 1-1 0 7 0 I {:} (f(X1), I(xz» E 7. Thus, if 7
is a preorder then 1-1 0 7 0 I itself is a preorder called the preimage or inverse
image of 7 under I. It is clear that the inverse image of an equivalence is also
an equivalence. Whereas the preimage of an order is not always antisymmetric.
In particular, this relates to the equivalence 1-1 0 f (= 1-1 0 ly 0 f).
( 4) Let X be an arbitrary set and let w be an equivalence on X. Define
a mapping rp : X -+ &,(X) by rp(x) := w(x). Recall that &,(X) stands for
the powerset of X comprising all subsets of X and also denoted by 2x. Let
X := X/w := im rp be the quotient set or factor set of X by w or modulo w.
A member of X/w is usually referred to as a coset or equivalence class. The
mapping rp is the coset mapping (canonical projection, quotient mapping, etc.).
Note that rp is treated as acting onto X. Observe that
1.2.8. REMARK. The claim of 1.2.7 gives rise to two generalizations of the
concept of greatest element.
1.2.9. DEFINITION. Let X be a (preordered) set and let U eX. A supremum
of U in X is a least upper bound of U, i.e. a least element of the set of all upper
bounds of U. This element is denoted by supx U or in short sup U. Certainly,
in a poset an existing supremum of U is unique and so it is in fact the supremum
of U. An infimum, a greatest lower bound inf U or infx U is defined by reversal.
1.2.10. DEFINITION. Let U be a subset of an ordered set (X, (1). A member x
of X is a maximal element of U if (1( x) n U = {x}. A minimal element is again
defined by reversal.
1.2.11. REMARK. It is important to make clear the common properties and
distinctions of the concepts of greatest element, maximal element, and supremum.
In particular, it is worth demonstrating that a "typical" set has no greatest element
while possibly possessing a maximal element.
1.2.12. DEFINITION. A lattice is an ordered set with the following property:
each pair (Xl, X2) of elements of X has a least upper bound, Xl V X2 := sup{ Xl, X2},
the join of Xl and X2, and a greatest lower bound, Xl 1\ X2 := inf {Xl, X2}, the meet
of Xl and X2.
1.2.13. DEFINITION. A lattice X is complete if each subset of X has a supre-
mum and an infimum in X.
1.2.14. An ordered set X is a complete lattice if and only if each subset of X
has a least upper bound. <ll>
1.2.15. DEFINITION. An ordered set (X, (1) is filtered upward provided that
X 2 = (1-1 0 (1. A downward-filtered set is defined by reversal. A nonempty poset
is a directed set or simply a direction if it is filtered upward.
6 Chapter 1. An Excursion into Set Theory
1.3. Filters
1.3.1. DEFINITION. Let X be a set and let fJI, a nonempty subset of 9'(X),
consist of nonempty elements. Such fJI is said to be a filterbase (on X) if fJI
is filtered downward. Recall that 9'(X) is ordered by inclusion. It means that
a greater subset includes a smaller subset by definition; this order is always pre-
sumed in 9'(X).
1.3.2. A subset fJI of 9'(X) is a filterbase if and only if
(1) fJI i= 0 and 0 rt fJI;
(2) B l , B2 E fJI =} (3B E fJI) Be Bl n B 2 •
1.3.3. DEFINITION. A subset $ of 9'(X) is a filter (on X) if there is a fil-
terbase fJI such that $ is the set of all supersets of fJI; i.e.,
In this case fJI is said to be a base for the filter $ (so each filterbase is a filter
base).
1.3.4. A subset $ in 9'(X) is a filter if and only if
(1) $ i= 0 and 0 rt $;
(2) (A E $ & A C B C X) =} B E $;
(3) AI, A2 E $ =} Al nA2 E $. <It>
1.3. Filters 7
1.3.5. EXAMPLES.
(1) Let F C X x Y be a correspondence and let .9B be a downward-filtered
subset of &,(X). Put F(.9B):= {F(B): B E .9B}. It is easy to see that F(.9B) is
filtered downward. The notation is alleviated by putting F(.9B):= fil F(.9B). If §
is a filter on X and B n dom F i- 0 for all B E §, then F( §) is a filter (on Y)
called the image of § under F. In particular, if F is a mapping then the image
of a filter on X is a filter on Y.
(2) Let (X, a) be a direction. Clearly, .9B: = {a( x): x EX} is a filter-
base. For a net F : X - t Y, the filter fil F(.9B) is the tail filter of F. Let (X, a)
and F : X - t Y be another direction and another net in Y. If the tail filter of F
includes the tail filter of F then F is a subnet (in a broad sense) of the net F.
If there is a subnet (in a broad sense) G : X - t X of the identity net «X)XEX
in the direction (X, a)) such that F = FoG, then F is a subnet of F (sometimes
F is addressed as a Moore subnet or a strict subnet of F). Every subnet is a sub net
in a broad sense. It is customary to speak of a net having or lacking a subnet with
some property.
1.3.6. DEFINITION. Let §(X) be the collection of all filters on X. Take
§l, §2 E §(X). Say that §l is finer than §2 or §2 refines §l (in other words,
§l is coarser than §2 or §l coarsens §2) whenever §l ::::> §2.
1.3.7. The set §(X) with the relation "to be finer" is a poset. <ll>
<I It is obvious that $0 is the greatest element of ~ and {X} is the least
element of ~. Therefore, the empty set has a supremum and an infimum in ~:
in fact, sup0 = inf ~ = {X} and inf0 = sup~ = $0. By 1.2.17 and 1.3.8,
it suffices to show that the join $1 V$2 is available for all $1, $2 E ~. Consider
$:= {AI n A2 : Al E $1, A2 E $2}. Clearly, $ C $0 while $ :::) $1 and
$ :::) $2. Thus to verify the equality $ = $1 V $2, it is necessary to establish
that $ is a filter.
Plainly, $ i- 0 and 0 ¢ $. It is also immediate that (B 1 , B2 E $ =?
B1 n B2 E $). Moreover, if C :::) Al n A2 where Al E $1 and A2 E $2, then
C = {AI n Ad u C = (AI U C) n (A2 U C). Since Al U C E $1 and A2 U C E $2,
conclude that C E $. Appealing to 1.3.4, complete the proof. I>
Exercises
1.1. Give examples of sets and nonsets as well as set-theoretic properties and non-set-
theoretic properties.
1.2. Is it possible for the interval [0, 1] to be a member of the interval [0, I]? For the
interval [0, 2]?
1.3. Find compositions of the simplest correspondences and relations: squares, disks
and circles with coincident or distinct centers in jRM X ]RN for all feasible values of M and N.
1.4. Given correspondences R, S, and T, demonstrate that
(R U S)-1 =
R- 1 U S-1; (R n S)-1 =
R- 1 n S-1;
(R U S) 0 T =(R 0 T) U (S 0 T); R 0 (S U T) =(R 0 S) U (R 0 T);
(RnS) 0 T C (Ro T) n (S oT); Ro (S nT) C (Ro S) n (R 0 T).
1.12. Let X and Y be ordered sets and M(X, Y), the set of increasing mappings from
X to Y with the natural order (specify the latter). Prove that
(1) (M(X, Y) is a lattice) {:} (Y is a lattice);
(2) (M(X, Y) is a complete lattice) {:} (Y is a complete lattice).
1.13. Given ordered sets X, Y, and Z, demonstrate that
(1) M(X, Y X Z) is isomorphic with M(X, Y) x M(Y, Z);
(2) M(X x Y, Z) is isomorphic with M(X, M(Y, Z)).
1.14. How many filters are there on a finite set?
1.15. How do the least upper and greatest lower bounds of a set of filters look like?
1.16. Let 1 be a mapping from X onto Y. Prove that each ultrafilter on Y is the image
of some ultrafilter on X under I.
1.17. Prove that an ultrafilter refining the intersection of two filters is finer than either
of them.
1.18. Prove that each filter is the intersection of all ultrafilters finer than it.
1.19. Let d be an ultrafilter on N containing cofinite subsets (a cofinite subset is
a subset with finite complement). Given :z:, yEs:= lW.N, put:z: "'.<If y:= (3A E d) :z:IA = yiA'
Denote *lW.:= lW.N /"'.<If. For t E lW. the notation *t symbolizes the coset with the constant sequence t
defined as t(n):= t (n EN). Prove that *lW. \ {*t : t E lW.} :f: 0. Furnish *lW. with algebraic and
order structures. How are the properties of lW. and *lW. related to each other?
Chapter 2
Vector Spaces
Thus :!£ is furnished with the structure of a vector space. This space, denoted
by X/Xo, is the quotient (space) of X by Xo or the factor space of X modulo Xo.
2.1.5. Let X be a vector space and let Lat (X) stand for the collection of all
subspaces of X. Ordered by inclusion, Lat (X) presents a complete lattice.
<l It is clear that infLat(X) = 0 and supLat(X) = X. Further, the inter-
section of a nonempty set of subspaces is also a subspace. By 1.2.17, the proof is
complete. t>
2.1.6. REMARK. With Xl, X 2 E Lat (X), the equality Xl V X 2 = Xl + X 2
holds. It is evident that inf g = n{Xo: Xo E g} for a nonempty subset g
of Lat (X). Provided that g is filtered upward, sup g = U {Xo : Xo E g}. <It>
2.1.7. DEFINITION. Subspaces Xl and X 2 of a vector space X split X into
(algebraic) direct sum decomposition (in symbols, X = Xl EEl X 2), if Xl /\ X 2 = 0
and Xl V X 2 = X. In this case X 2 is an (algebraic) complement of Xl to X, and
Xl is an (algebraic) complement of X 2 to X. It is also said that Xl and X 2 are
(algebraically) complementary to one another.
2.1.8. Each subspace of a vector space has an algebraic complement.
<l Take a subspace Xl of X. Put
isomorphism between ~ and X~, where ~:= TI"ES X., with X.,:= 0 if." =1= e
and X~:= X~.
( 4) Let X := Xl EB X 2. Since + -1 is an isomorphism between X and
Xl x X 2, we may define PI, P2 E 2(X) as PI := P X dlX 2 := PrIO(+-I) and
P2 := P X2 11x1 := Pr2 0(+ -1). The operator PI is the projection of X onto Xl
along X 2 and P2 is the complementary projection to PI or the complement of PI
(in symbols, P2 = pt). In turn, PI is complementary to P2, and P 2 projects X
onto X 2 along Xl. Observe also that PI + P2 = Ix. Moreover, Pl:= PIPI = PI,
and so a projection is an idempotent operator. Conversely, every idempotent P
belonging to 2(X) projects X onto P(X) along P-I(O).
For T E 2(X) and P a projection, the equality PTP = T P holds if and only
if T(Xo) C Xo with Xo = im P (read: Xo is invariant under T). <II>
The equality TPX1 ll x2 = PXdlx2T holds whenever both Xl and X 2 are in-
variant under T, and in this case the direct sum decomposition X = Xl EB X 2
reduces T. The restriction of T to Xl is acknowledged as an element TI of 2(Xd
which is called the part of T in Xl. If T2 E 2(X2) is the part of Tin X 2, then T
is expressible in matrix form
Q5 = Q2 Q 4·
o -+ Xo •
----7 X 'P
----7 X / Xo -+ 0
16 Chapter 2. Vector Spaces
is exact. (On similar occasions the letters t and c.p are usually omitted.) In a sense,
this sequence is unique. Namely, consider a so-called short sequence
O-X~Y~Z-o
and assume that it is exact. Putting Yo:= im T, arrange the following commuta-
tive diagram
T S
O-X-Y-Z-O
o- Yo - Y - Y j Yo- 0
where a, /3, and 'Yare some isomorphisms. In other words, a short exact sequence
actually presents a subspace and the corresponding quotient space. <It>
(6) Every T in 2'(X, Y) generates the exact sequence
o- ker T - X ~ Y _ coker T - 0
which is called the canonical exact sequence for T.
2.3.6. DEFINITION. Given To E 2'(Xo, Y), with Xo a subspace of X, call
an operator T from X to Y an extension of To (onto X, in symbols, T :::> To),
provided that To = Tt, where t : Xo - X is the identical embedding of Xo
into X.
2.3.7. Let X and Y be vector spaces and let Xo be a subspace of X. Then
each To in .2'(Xo, Y) has an extension T in 2'(X, Y).
Putting T:= ToPxo, where PX o is a projection onto X o, settles the claim. t>
<l
~l~
Z
is commutative for some ~ in 2'(Y, Z) if and only if ker ACker B.
It is evident that ker ACker B in case B = ~ A.
<l :::}:
~: Set ~:= BoA-I. Clearly, ~ 0 A(x) = B 0 (A- I 0 A)x = B(x + ker A) =
Bx. Show that .:to:= ~hm A is a linear operator. It suffices to check that ~ is
single-valued. Suppose that y E im A and ZI, Z2 E ~(y). Then ZI = BXI, Z2 =
BX2 and AXI = AX2 = y. By hypothesis B(XI - X2) = 0; therefore, ZI = Z2.
Applying 2.3.7, find an extension ~ of .:to to Y. t>
2.3. Equations in Operators 17
T
x .y
2.3.12. Let X be a vector space and let fo, !I, ... ,IN belong to X#. The
functional fo is a linear combination of !I, ... , fN (i.e., fo = ~f=1 Adj with Aj
in IF) if and only if ker fo ::) nf=l
ker fJ.
<l Define the linear operator (!I, ... ,IN) : X --+ IFN by (!I, ... ,fN)X:=
(!I (x), ... , f N (x». Obviously, ker (!I , ... ,IN) = nf=l ker fJ. Now apply 2.3.8 to
the problem
~
on recalling what IFN# is. D>
Z
is commutative for some !r in ~(Z, Y) if and only if im A ::) im B.
18 Chapter 2. Vector Spaces
im B = B(Z) = A('?£(Z)) C A(Y) = im A.
<1 =}:
Let Yo be an algebraic complement of ker A to Y and Ao := Alyo. Then Ao
-¢=:
is an isomorphism between Yo and im A. The operator .?£:= lA01 B is obviously
a sought solution, with l the identical embedding of Yo into Y. I>
2.3.14. REMARK. Provided that the operator A is a monomorphism, there
is a unique operator .?£ in 2.3.13. <11>
2.3.15. REMARK. Theorems 2.3.8 and 2.3.13 are in "formal duality." One
results from the other by "reversing arrows," "interchanging kernels and images,"
and "passing to reverse inclusion."
2.3.16. Snowflake Lemma. Let S E 2'(Y, Z) and T E 2'(X, Y). There
are unique operators 0:1, ... ,0:6 such that the following diagram commutes:
0 0
\ 0:2 I
ker ST • ker S
0:/ \
T
/ ,,\3
. y - - - coker T - O
0 -+-kerT 'X
S~ Is Z
/\
coker S - coker ST
is exact. <11>
Exercises
2.1. Give examples of vector spaces and nonvector spaces. Which constructions lead to
vector spaces?
2.2. Study vector spaces over the two-element field 1::2.
2.3. Describe vector spaces with a countable Hamel basis.
2.4. Prove that there is a discontinuous solution f : JR -+ JR to the function equation
2.5. Prove that the algebraic dual to the direct sum of (Xe) is presentable as the product
of the algebraic duals of (Xe).
2.6. Let X :J Xo :J Xoo. Prove that the spaces X/Xoo and (X/Xo)/(Xoo/Xo) are
isomorphic.
2.7. Define the "double sharp" mapping by the rule
Show that this mapping embeds a vector space X into the second dual X## .
2.8. Prove that finite-dimensional spaces and only such spaces are algebraically reJlezivej
i.e.,
##(X) =
X## <=> dimX < +00.
2.9. Are there any analogs for a Hamel basis in general modules?
2.10. When does a sum of projections present a projection itself?
2.11. Let T be an endomorphism of some vector space which satisfies the conditions
Tn-l I- 0 and Tn = 0 for a natural n. Prove that the operators TO, T, ... ,Tn-l are linearly
independent.
2.12. Describe the structure of a linear operator defined on the direct sum of spaces
and acting into the product of spaces.
2.13. Find conditions for unique solvability of the following equations in operators
~A = B and A~= B (here the operator ~ is unknown).
2.14. What is the structure of the spaces of bilinear operators?
2.15. Characterize the real vector space that results from neglecting multiplication
by imaginary scalars in a complex vector space (cf. 3.7.1).
2.16. Given a family of linearly independent vectors (z.).E£', find a family offunctionals
(Z~).E£' satisfying the following conditions:
2.17. Given a family of linearly independent functionals (Z~).E£" find a family of vec-
tors (z.).E£' satisfying the following conditions:
2.18. Find compatibility conditions for simultaneous linear equations and linear inequal-
ities in real vector spaces.
2.19. Consider the commutative diagram
with exact rows and such that 0/ is an epimorphism, and cS is a monomorphism. Prove that
ker i= =
T(ker,8) and 'f-1(im i) im,8.
Chapter 3
Convex Analysis
(10) Given r c IF2, observe that X E (r) for every vector space X
over IF. Note also that in 3.1.2 (1)-3.1.2 (9) the set r is itself a r-set.
3.1.3. Let X be a vector space and let tf: be a family of r-sets in X. Then
n{u: U E im tf:} E (r). Provided that im tf: is filtered upward (by inclusion),
U{U: U E im tf:} E (r). <It>
3.1.4. REMARK. The claim of 3.1.3 means in particular that the collection
of r-sets of a vector space, ordered by inclusion, presents a complete lattice.
3.1.5. Let X and Y be vector spaces and let U and V be r-sets, with U c X
and V c Y. Then U x V E (r).
<l If U or V is nonempty then U x V = 0, and there is nothing to prove.
Now take UI, U2 E U, VI, V2 E V, and (AI, A2) E r. Find AIUl + A2U2 E U and
AIV2 + A2Vl E V. Hence, (AIUl + A2U2, AIVI + A2V2) E U x V. t>
3.1.6. DEFINITION. Let X and Y be vector spaces and r c IF2. A correspond-
ence T C X x Y is a r-correspondence provided that T E (r).
3.1.7. REMARK. When a r-set bears a specific attribute, the latter is pre-
served for naming a r-correspondence. With this in mind, we speak about linear
and convex correspondences, affine mappings, etc. The next particularity of the
nomenclature is worth memorizing: a convex function of one variable is not a con-
vex correspondence, save trivial cases (cf. 3.4.2).
3.1.8. Let T C X x Y be a rl-correspondence and let U c X be a r 2 -set.
Ifr 2 c r l then T(U) E (r 2 ).
<l Take Yl, Y2 E T(U). Then (Xl, yd E T and (X2' Y2) E T with some
Xl, X2 E U. Given (AI, A2) E r 2 , observe that Al(Xl, yd + A2(X2, Y2) E T,
because by hypothesis (AI, A2) E rl. Finally, infer that AIYl + A2Y2 E T(U). t>
3.1.9. The composition ofr-correspondences is also a r-correspondence.
<J Let Fe X x V and G C W x Y, with F, G E (r). Note that
<l Denote the right side of the Motzkin formula by V. Since Uo C U; applying
3.1.12 (3), deduce that Hr(Uo ) C Hr(U); and, hence, Hr(U) :::::> V. By 3.1.12 (2),
it is necessary (and, surely, sufficient) to verify that V E (r). The last follows
from 3.1.3 and the inclusion Hr(Uo) U Hr(Ut) C Hr(Uo U Ut ). t>
3.1.14. REMARK. The Motzkin formula reduces the problem of describing
arbitrary r-hulls to calculating r-hulls of finite sets. Observe also that r-hulls
for concrete rs have special (but natural) designations. For instance, if r :=
{(.At, .Az) E ~: .At +.A2 = I}, then the term "convex hull" is used and co(U)
stands for Hr(U). If U 1= 0, the notation for HlF2(U) is lin (U); moreover, it is
natural and convenient to put lin (0):= O. The subspace lin (U) is called the linear
span of U. The concepts of affine hull, conical hull, etc. are introduced in a similar
fashion. Note also that the convex hull of a finite set of points comprises their
convex combinations; i.e.,
Then a is a preorder compatible with vector structure and K coincides with the
cone a(O) of positive elements of X. The relation a is an order if and only
if K n (-K) = O.
It is clear that 0 E K:::} Ix C a and K +K C K :::} aoa C a. Furthermore,
<I
a-I = {(x, y) E X2: x - Y E K}. Therefore, a n a-I C Ix {:} K n (-K) = O.
To show that a is a cone, take (Xl, yd, (X2' Y2) E a and al, a2 E lR+. Find
alYl + a2Y2 - (alxl + a2x2) = al(Yl - xd + a2(Y2 - X2) E alK + a2K C K. I>
3.2.4. DEFINITION. A cone K is an ordering cone or a salieni cone provided
that K n ( - K) = O.
3.2.5. REMARK. By virtue of 3.2.2 and 3.2.3, assigning a preorder to X is
equivalent to distinguishing some cone of positive elements in it which is the pos-
itive cone of X. The structure of an ordered vector space arises from selecting
an ordering cone. Keeping this in mind, we customarily call a pair (X, X+) with
positive cone X+, as well as X itself, a (pre)ordered vector space.
3.2.6. EXAMPLES.
(1) The space of real-valued functions JR.=: with the positive cone JR.~:=
(JR.+)=: constituted by the functions assuming only positive values has the "natural
order."
(2) Let X be an ordered vector space with positive cone X+. If Xo is
a subspace of X then the order induced in Xo is defined by the cone Xo n X+.
In this way Xo is considered as an ordered vector space, a subspace of X.
(3) If X and Yare preordered vector spaces then T E 2'(X, Y) is
a positive operator (in symbols, T 2: 0) whenever T(X+) C Y+. The set 2'+(X, Y)
of all positive operators is a cone. The linear span of 2'+(X, Y) is denoted
by 2'r(X, Y), and a member of 2'r(X, Y) is called a regular operator.
3.2.7. DEFINITION. An ordered vector space is a vector lattice or a Riesz
space if the ordered set of its vectors presents a lattice.
3.2.8. DEFINITION. A vector lattice X is called a Kantorovich space or briefly
a K -space if X is boundedly order complete or Dedekind complete, which means
that each nonempty bounded above subset of X has a least upper bound.
3.2.9. Each nonempty bounded below subset of a Kantorovich space has
a greatest lower bound.
<I Let U be bounded below: x ::; U for some x. So, -x 2: -U. Applying 3.2.8,
find sup( -U). Obviously, - sup( -U) = inf U. I>
3.2.10. If U and V are nonempty bounded above subsets of a K -space then
<I If U or V is a singleton then the equality is plain. The general case follows
from the associativity of least upper bounds. Namely,
x + y = x V y + x 1\ y.
°
<I The first equality follows from 3.2.13 on letting y:= O. Furthermore, Ixl =
x V (-x) = -x + (2x) V = -x + 2x+ = -x(x+ - x_) + 2x+ = x+ + x_. I>
3.2.15. Interval Addition Lemma. If x and yare positive elements of a vec-
tor lattice X then
[0, x + y) = [0, x) + [0, y).
(As usual, [u, v):= O'(u) n O'-l(v) is the (order) interval with endpoints u and v.)
°: ;
<I The inclusion [0, x)+[O, y) C [0, x+y) is obvious. Assume that Z ::; x+y
and put Zl := Z 1\ x. It is easy that Zl E [0, x). Now if Z2 := Z - Zl then Z2 ~
and Z2 = z-zl\x = z+(-z)V(-x) = OV(z-x)::; OV(x+y-x) = OVy = y. I>
°
3.2.16. REMARK. The conclusion of the Interval Addition Lemma is often
referred to as the Riesz Decomposition Property.
3.2.17. Riesz-Kantorovich Theorem. Let X be a vector space and let Y
be a K-space. The space of regular operators 2'r(X, Y), ordered by the cone
of positive operators 2'+(X, Y), is a K -space. <II>
3.3. Extension of Positive Functionals and Operators 25
3.3. Extension of Positive Functionals and Operators
3.3.1. COUNTEREXAMPLES.
(1) Let X be B([O, 1], JR.), the space ofthe bounded real-valued functions
given on [0, 1]; and let Xo be C([O, 1], JR.), the subspace of X comprising all
continuous functions. Put Y := Xo and equip X o, X, and Y with the natural
order (cf. 3.2.6 (1) and 3.2.6 (2)). Consider the problem of extending the identical
embedding To : Xo ---T Y to a positive operator T in .2"+ (X, Y). If the problem had
a solution T, then each nonempty bounded set ~ in Xo would have a least upper
bound sUPxo ~ calculated in Xo· Namely, sUPxo ~ = Tsupx~, where sup X ~ is
the least upper bound of ~ in X; whereas, undoubtedly, Y fails to be a K -space.
(2) Denote by s:= JR.N the sequence space and furnish s with the nat-
ural order. Let c be the subspace of s comprising all convergent sequences, the
convergent sequence space. Demonstrate that the positive functional fo : c ---T JR.
defined by fo(x):= limx(n) has no positive extension to s. Indeed, assume that
fES#, f2.0andf"Jfo. Putxo(n):=nandxk(n):=kAnfork, nEN. Plainly,
fO(Xk) = k; moreover, f(xo) 2. f(Xk) 2. 0, since Xo 2. Xk 2. 0, a contradiction.
3.3.2. DEFINITION. A subspace Xo of an ordered vector space X with pos-
itive cone X+ is massive (in X) if Xo + X+ = X. The terms "coinitial" or
"minorizing" are also in common parlance.
3.3.3. A subspace Xo is massive in X if and only if for all x E X there are
elements Xo and XO in Xo such that Xo ~ x ~ xo. <II>
3.3.4. Kantorovich Theorem. H X is an ordered vector space, Xo is mas-
sive in X, and Y is a K -space; then each positive operator To, a member of
.2"+(Xo, Y), has a positive extension T, a member of .2"+ (X, Y).
<I STEP I. First, let X:= Xo EB XI, where Xl is a one-dimensional subspace,
Xl := {ax: 0' E JR.}. Since Xo is massive and To is positive, the set U:= {ToxO :
x O E X o , X O 2. x} is nonempty and bounded below. Consequently, there is some
11 such that 11:= inf U. Assign Tx:= {Toxo + 0'11: x = Xo + ax, Xo E X o, 0' E JR.}.
It is clear that T is a single-valued correspondence. Further, T "J To and dom T
= X. So, only the positivity of T needs checking. If x = Xo + ax and x 2. 0, then
°
the case of 0' equal to 0 is trivial. If 0' > then x 2. -xo/O'. This implies that
-Toxo/O' ~ 11, i.e., Tx E Y+. In a similar way for 0' < 0 observe that x ~ -xo/O'.
Thus, 11 ~ -Toxo/O' and, finally, Tx = Toxo + 0'11 E Y+.
STEP II. Now let ~ be the collection of single-valued correspondences S c
X x Y such that S "J To and S(X+) c Y+. By 3.1.3, ~ is inductive in order
by inclusion and so, by the Kuratowski-Zorn Lemma, ~ has a maximal element T.
If x E X \ dom T, apply the result of Step I with X := dom T EB Xl, Xo :=
dom T, To:= T and Xl := {ax: 0' E JR.} to obtain an extension of T. But T is
maximal, a contradiction. Thus, T is a sought operator. I>
26 Chapter 3. Convex Analysis
°
(2) p is a convex function that is positively homogeneous:
p(Qx) = Qp(x) for all 0. 2:: and X E X;
(3) ifQ}, 0.2 E JR.+ and Xl, X2 E X, then
P(QlXl + Q2 X2) :::; QlP(Xt} + Q2P(X2);
(4) p is a positively homogeneous functional that is subadditive:
P(XI + X2) :::; p(xt} + p(X2) (Xl, X2 EX). <ll>
3.4.8. EXAMPLES.
(1) A linear functional is sublinear; an affine functional is a convex func-
tion.
(2) Let U be a convex set in X. Define the indicator function 8(U)
X ~ JR.. of U as the mapping
0, if X EU
8(U)(x):= {
+=, if X 1. U.
28 Chapter 3. Convex Analysis
e
"'!O a
. f h(x + aht) + 1 h(x + ah 2 )) - f(x)
= 2 hm ---'--'----'----'-----'-'----'--'-
"'!O a
< 1· f(x+aht)-f(x)
1m
l' f(x+ah 2 )-f(x)
+ 1m ::...."..--~--':.....:....--'-
- "'!O a "'!O a
= f'(x)(ht) + f'(x)(h 2 ).
Appealing to 3.4.7, end the proof. I>
3.5. The Hahn-Banach Theorem 29
3.5. The Hahn-Banach Theorem
3.5.1. DEFINITION. Let X be a real vector space and let f : X -+ lR: be
a convex function. The subdifferential of f at a point x in dom f is the set
3.5.2. EXAMPLES.
(1) Let p: X -+ JR' be a sublinearfunctional. Put 8(p):= 8o(p). Then
8(b'(Xo)) = {I E X# : ker I J X o }.
8 x(f) = 8(f'(x))
8 x (f 0 T) = 8Tx(f) 0 T.
<l By 3.4.10 it follows that x E core dam f. From 3.5.2 (4) obtain 8x (f 0 T) =
8((f 0 T)'(x)). Moreover,
<I Let PI := fUx) and P2 := f~(x). Given Xl, X2 E X, define p(Xl' X2):=
Pl(xd + P2(X2) and t(xt):= (Xl, xI). Using 3.5.2 (4) and 3.5.3, infer that
define the correspondence seg C X 2 x X that assigns to each pair of points the open
segment joining them. If U is a convex set in X and segu is the restriction of seg
to U 2, then a convex subset V of U is extreme in U if segijl (V) C V2. An extreme
subset of U is sometimes called a face in U. A member X of U is an extreme point
in U if {x} is an extreme subset of U. The set of extreme points of U is usually
denoted by ext U.
3.6.2. A convex subset V is extreme in U if and only if for all Ul, U2 E U and
01, 02 > 0, 01 + 02 = 1, the containment 0lUl + 02U2 E V implies that Ul E V
and U2 E V. <It>
32 Chapter 3. Convex Analysis
3.6.3. EXAMPLES.
(1) Let p : X
-7 lR' be a sublinear functional and let a point x of X
°
<I For, if alit +a212 E 8 x (p) and 11, 12 E 8(p), where aI, a2 > 0, al +a2 = 1;
then = p(x) - (alit(x) + a2h(x)) = al(p(x) - Il(x)) + a2(p(x) - 12(X)) 2: 0.
°
Moreover,p(x)-lt(x) 2: andp(x)-12(x) 2: 0. Hence, 11 E 8 x (p) and 12 E 8 x (p).1>
(2) Let W be an extreme set in V and let V be in turn an extreme set
in U. Then W is an extreme set in U. <II>
(3) If X is an ordered vector space then a positive element x of X is
discrete if and only if the cone {ax: a E R+} is an extreme subset of X+.
<I ¢::: If ° : :;
Y :::; x then x = 1 h(2y) + 1 h(2(x - y)). Therefore, by 3.6.2,
2y = ax and 2(x - y) = f3x for some a, f3 E R+. Thus, 2x = (a + f3)x. The case
°
x = is trivial. Now if x#-o then 0< h E [0, 1] and, consequently, [0, x] C [0, l]x.
The reverse inclusion is evident.
=}: Let [0, x] = [0, l]x, and suppose that ax = alYl + a2Y2 for a 2: 0;
Now take 9 E (XJR)# and put f(x):= g(x)-ig(ix). Evidently, f E 2'(XJR, CJR)
and Ref(x) = g(x) for x E X. It is sufficient to show that f(ix) = if(x),
because this equality implies f E X#. Straightforward calculation shows f( ix) =
g(ix) + ig(x) = i(g(x) - ig(ix)) = if(x), which enables us to conclude that lRe is
also an epimorphism. t>
3.7.4. DEFINITION. The lear trap map or complexifier is the inverse lRe- 1 :
(XJR)# -+ (X#)JR of the real part map.
3.7.5. REMARK. By 3.7.3, for the complex scalar field
In the case of the reals, f:= lR, the complexifier lRe- 1 is the identity operator.
3.7.6. DEFINITION. Let (X, f, +, .) be a vector space over f. A seminorm
on X is a function p : X -+ lR· such that dom p =1= 0 and
181(p 0 T) = 181(p) 0 T.
<I Applying 2.3.8 and 3.7.10, successively infer that
{f < t} C Ut C {f ~ t} (t E T)
if and only if the mapping t 1-+ Ut increases.
36 Chapter 3. Convex Analysis
<J ::::}: Suppose that T contains at least two elements sand t. (Otherwise there
is nothing left to proof.) If s < t then
Us C {f :::; s} C {f < t} CUt.
¢::: Define a mapping f : X ---7 iii by setting f(x):= inf{t E T: x E Ud. If
{f < t} is empty for t E T then all is clear. If x E {f < t} then f(x) < +00, and
so there is some sET such that x E Us and s < t. Thus, {f < t} C Us CUt.
Continuing, for x E Ut find f( x) :::; t; i.e., Ut C {f :::; t}. I>
3.8.3. Function Comparison Lemma. Let f, g : X ---7 iii be functions
defined by (Ut)tET and (Vi)tET as follows:
{f < t} C Ut C {f :::; t};
{g < t} C vt C {g :::; t} (t E T).
IfT is dense in iii (i.e., (V r, t E iii, r < t) (3 SET) (r < s < t)) then the inequality
f :::; g (in iiix; i.e., f(x) :::; g(x) for x E X) holds if and only if
(tl' t2 E T & tl < t2) ::::} vtl C Ut2 •
<J ::::}: This is immediate from the inclusions
vtl C {g :::; tI} C {f :::; tI} C {f < t2} C Ut2 •
¢::: Assume that g(x) i= +00 (if not, f(x) :::; g(x) for obvious reasons). Given
t E lR. such that g( x) < t < +00, choose t l , t2 E T so as to satisfy the conditions
g(x) < h < t2 < t. Now
x E {g < tI} C vtl C Ut2 C {f :::; t2} c {f < t}.
Thus, f(x) < t. Since t is arbitrary, obtain f(x) :::; g(x). I>
3.8.4. Corollary. If T is dense in iii and the mapping t f-+ Ut (t E T) in-
creases then there is a unique function f : X ---7 iii such that
{f < t} C Ut C {f :::; t} (t E T).
Let p : X --t JR' be an arbitrary sublinear functional with positive values and
let {p < I} eSc {p ~ I}. Given t E JR+, put Vi:= {p < t} and Ut := tS; given
t < 0, put Vi:= Ut := 0. Plainly,
for t E IR. If 0 ::; tl < t2 then vtl = {p < ttl = tdp < I} C tl S = Uh C Ut2 .
Moreover, Ut1 C tdp::; I} C {p ::; ttl C {p < t2} C vt 2 • Therefore, by 3.8.3 and
3.8.4, p = PS. t>
In particular, this entails the inequality p(<p(x)) 2:: 1, since <p(x) t/. <p(U).
Using 3.5.6, find a functional j in ox(p 0 <p); now the Hahn-Banach Theorem
implies
Put H:= {7 = po <p( x)}. It is clear that H is a real hyperplane and, undoubtedly,
H :::> L. Appealing to 3.5.2 (1), conclude that Hncore U = 0. Now let f:= IRe-Ij
and H:= {f = f(x)}. There is no denying that L C He H. Thus, the hyperplane
H provides us with what was required. t>
3.8.12. REMARK. Under the hypotheses of the Separation Theorem, it may
be assumed that core Un L = 0. Note also that Theorem 3.8.11 is often referred
to as the Hahn-Banach Theorem in geometric form or as the Minkowski-Ascoli-
Mazur Theorem.
Exercises 39
4.1.15. REMARK. In connection with 4.1.14, the set 0p(T) is also referred
to as the topology of U, since T is uniquely determined from Op (T). The same
relates also to CI ( T ), the collection of all closed sets in X.
4.1.16. DEFINITION. A filterbase fJ6 on X converges to x in X or x is a limit
of fJ6 (in symbols, fJ6 ~ x) if fil fJ6 is finer than the neighborhood filter of Xj i.e.,
fil fJ6 J T(X).
4.1.17. DEFINITION. A net or (generalized) sequence (Xe)eES converges to x
(in symbols, xe ~ x) if the tail filter of (xe) converges to x. Other familiar terms
and designations are freely employed. For instance, x = lime xe and x is a limit
e
of (xe) as ranges over 3.
4.1.18. REMARK. A limit of a filter, as well as a limit of a net, is unique
in a metric space X. This is another way of expressing the fact that the topology
of X is separated. <II>
4.1.19. For a nonempty set U and a point x the following statements are
equivalent:
(1) x is an adherent point ofUj
(2) there is a filter $ such that $ ~ x and U E $j
(3) there is a sequence (Xe)eES whose entries are in U and which con-
verges to x.
<I (1) =} (2): Since x is not exterior to U, the join $ := T(X) V fil {U} is
available of the pair of the filters T( x) and fil {U}.
(2) =} (3): Let $ ~ x and U E $. Direct $ by reverse inclusion. Take
Xv E V n U for V E $. It is clear that Xv ~ x.
(3) =} (1): Let V be a closed set. Take a sequence (Xe)eE3 in V such that
xe ~ x. In this case it suffices to show that x E V, which is happily evident.
e
Indeed, were x in X \ V we would find at least one E 3 such that xe EX \ V. I>
4.1.20. REMARK. It may be assumed that $ has a countable base in 4.1.19
(2), and 3:= N in 4.1.19 (3). This property is sometimes formulated as follows:
"In metric spaces the first axiom of count ability is fulfilled."
<I The equivalence (1) ¢} (2) follows from 4.1.11. It remains to demonstrate
that (1) '* (3) '* (4)'* (5) '*
(2).
(1) '* (3): If V E ry(l(x)) then W := int V E Op(ry) and f(x) E W.
Hence, f-l(W) E Op(rx) and x E f-l(W). In other words, f-l(W) E rx(x)
(see 4.1.14). Moreover, f-l(V) :J f-l(W) and, consequently, f-l(V) E rx(x).
Finally, V :J f(l-l (V)).
(3) '* (4): Given $ --+ x, by Definition 4.1.16 fil $ :J rx(x). From the
hypothesis infer that f($) :J f(rx(x)) :J ry(l(x)). Appealing to 4.1.16, again
reveal the sought instance of convergence, f( $) --+ f( x).
(4) '* (5): The image of the tail filter of (Xe)eE2 under f is coarser than the
tail filter of (I( xe) )eE2.
(5) '* (2): Take a closed set Fin Y. If F = 0 then f-l(F) is also empty and,
hence, closed. Assume F nonempty and let x be an adherent point of f-l(F).
Consider a sequence (Xe)eE2 converging to x and consisting of points in f-l(F)
(the claim of 4.1.19 yields existence). Then f(xe) E F and f(xe) --+ f(x). Another
citation of 4.1.19 guarantees that f(x) E F and, consequently, x E f-l(F). [>
4.2.2. DEFINITION. A mapping satisfying one (and hence all) of the equiva-
lent conditions 4.2.1 (1)-4.2.1 (5) is continuous. If 4.2.1 (5) holds at a fixed point x
then f is said to be continuous at x. Thus, f is continuous on X whenever f is
continuous at every point of X.
4.2.3. Every composition of continuous mappings is continuous.
<I Apply 4.2.1 (5) thrice. [>
4.2.4. Let f: X --+ Y and let 'Wx and 'Wy be uniformities on X and Y. The
following statements are equivalent:
(1) (VV E 'Wy ) (3U E 'Wx ) ((Vx, y)(x, y) E U '*
(I(x), fey)) E V);
(2) (VV E o//y) f- l 0 V 0 f E o//x;
(3) r('Wx) :J 'Wy, with r : x 2 --+ Y defined as fX : (x, y) I-t
(I(x), fey));
(4) (VV E 'Wy ) r-l(V) E 'Wx; i.e., r-l('Wy) c 'Wx .
<I By 1.1.10, given U C X 2 and V C y2, observe that
4.2.8. DEFINITION. Let g be a set of mappings from X into Y and let Cil/x
and Cil/y be the uniformities in X and Y. The set g is equicontinuous if
(\IV E Cil/Y) n
lEg
f- l 0 Vo f E Cil/x .
4.3. Semicontinuity
4.3.1. Let (Xl, dd and (X2' d 2) be metric spaces, and.¥:= Xl x X 2 • Given
x:= (Xl, X2) and y:= (Yl, Y2), put
Tben d is a metric on :Z". Moreover, for every X:= (Xl, X2) in :Z" tbe presentation
bolds:
TX(X) = fil{Ul X U2 : Ul E TXl(Xl), U2 E TX 2 (X2)}. <]1>
4.3.2. DEFINITION. The topology TX is called the product of TX l and TX 2 or
the product topology of Xl x X 2. This topology on Xl x X 2 is denoted by TX l x TX 2 •
4.3.3. DEFINITION. A function f : X -+ lR" is lower semicontinuous if its
epigraph epi f is closed in the product topology of X x JR.
4.3.4. EXAMPLES.
(1) A continuous real-valued function f: X -+ JR is lower semicontinuous.
(2) If fe : X -+ lR" is lower semi continuous for all E 3, then the e
upper envelope f( x):= sup {fe (x): e
E 3} (x E X) is also lower semicontinuous.
A simple reason behind this is the equality epi f = neE2 epi fe.
4.3. Semicontinuity 45
Here, as usual,
lim inf J(y):= lim J(y):= sup inf J(U)
y~x y~x UEr(x)
Therefore, inf J(U) > t for some neighborhood U of x. It follows that the comple-
ment of epi J to X x JR is open. I>
4.3.6. REMARK. The property, stated in 4.3.5, may be accepted as an initial
definition of lower semi continuity at a point.
4.3.7. A function J : X - t JR is continuous if and only if both J and - J are
lower semicontinuous. <11>
4.3.8. A function J : X - t JR. is lower semicontinuous if and only if for every
t E JR the level set {f :::; t} is closed.
<1 ::}: If x ¢ {f :::; t} then t < J(x). By 4.3.5, t < inf J(U) in a suitable
neighborhood U about x. In other words, the complement of {f :::; t} to X is
open.
'¢=:: Given limy-+x inf J(y) :::; t < J( x) for some x E X and t E JR, choose c > 0
such that t + c < J(x). Repeating the argument of 4.3.5, given U E r(x) take
a point Xu in Un {f :::; inf J(U) + c}. Undoubtedly Xu E {f :::; t + c} and Xu - t x,
a contradiction. I>
46 Chapter 4. An Excursion into Metric Spaces
4.4. Compactness
4.4.1. DEFINITION. A subset C of X is called a compact set whenever for
every subset rB' of Op (TX) with the property C C U{ G: G E rB'} there is a finite
subset rB'0 in rB' such that still C C U {G: G E 6O}.
4.4.2. REMARK. Definition 4.4.1 is verbalized as follows: "A set is compact
if its every open cover has a finite subcover." The terms "covering" and "subcov-
ering" are also in current usage.
4.4.3. Every closed subset of a compact set is also compact. Every compact
set is closed. <ll>
4.4.4. REMARK. With regard to 4.4.3, it stands to reason to use the term
"relatively compact set" for a set whose closure is compact.
4.4.5. Weierstrass Theorem. The image of a compact set under a contin-
uous mapping is compact.
<l The inverse images of sets in an open cover of the image compose an open
cover of the original set. I>
4.4.6. Each lower semicontinuous real-valued function, defined on a nonempty
compact set, assumes its least value; i.e., the image of a compact domain has a least
element.
<l Suppose that f : X R and X is compact. Let to := inf f(X). In the case
-+
to = +00 there is nothing left to proof.
If to < +00 then put T:= {t E lR.: t > to}.
The set Ut := {f ::::: t} with t E T is nonempty and closed. Check that n{Ut :
t E T} is not empty (then every element x of the intersection meets the claim:
f(x) = inf f(X)). Suppose the contrary. Then the sets {X \ Ut : t E T} compose
an open cover of X. Refining a finite subcover, deduce n{Ut : t E To} = 0.
However, this equality is false, since UtI n U t2 = U tI 1\t2 for t l , t2 E T. I>
4.4.7. Bourbaki Criterion. A space is compact if and only if every ultra-
filter on it converges (cf. 9.4.4).
4.4.8. The product of compact sets is compact.
<l It suffices to apply the Bourbaki Criterion twice. [>
4.5. Completeness
4.5.1. If ~ is a filterbase on X then {B2: B E ~} is a filterbase (and a base
for the filter ~X) on X2.
<l (BI x Bd n (B2 x B 2) J (BI n B 2) x (BI n B 2) [>
4.5. Completeness 47
4.5.2. DEFINITION. Let %'X be a uniformity on X. A filter ~ is called
a Cauchy filter or even Cauchy (the latter might seem preposterous) if ~x J %'x.
A net in X is a Cauchy net or a fundamental net if its tail filter is a Cauchy filter.
The term "fundamental sequence" is treated in a similar fashion.
4.5.3. REMARK. If V is an entourage on X and U is a subset of X then U is
V -small whenever U 2 C V. In particular, U is Be-small (or simply 6-small) if and
only if the diameter of U, diam U:= sup(U 2 ), is less than or equal to 6. In this
connection, the definition of a Cauchy filter is expressed as follows: "A filter is
a Cauchy filter if and only if it contains arbitrarily small sets."
4.5.4. In a metric space the following conditions are equivalent:
(1) every Cauchy filter converges;
(2) each Cauchy net has a limit;
(3) every fundamental sequence converges.
<I The implications (1) =} (2) =} (3) are obvious; therefore, we are left with
establishing (3) =} (1).
Given a Cauchy filter~, let Un in ~ be a BI/n-small set. Put Vn:= UI n ... n
Un and take Xn E Vn . Observe that VI J V2 J ... and diam Vn ::; 1 I n. Hence,
(x n ) is a fundamental sequence. By hypothesis it has a limit, x:= limx n . Check
that ~ -+ x. To this end, choose no E N such that d(xm, x) ::; 1 hn as m ~ no.
Then for all n E N deduce that d(xp, y) ::; diam Vp ::; 1 hn and d(xp, x) ::; 1 hn
whenever p:= no V 2n and y E Vp. It follows that y E Vp =} d(x, y) ::; 1 In; i.e.,
Vp C B1/n(x). In conclusion, ~ J r(x). I>
4.5.5. DEFINITION. A metric space satisfying one (and hence all) of the
equivalent conditions 4.5.4 (1 )-4.5.4 (3) is called complete.
4.5.6. Cantor Criterion. A metric space X is complete if and only if every
nonempty downward-filtered family of nonempty closed subsets of X whose diam-
eters tend to zero has a point of intersection.
<I =}: If !JIJ is such a family then by Definition 1.3.1 !JIJ is a filterbase. By hy-
pothesis, !JIJ is a base for a Cauchy filter. Therefore, there is a limit: !JIJ -+ x. The
point x meets the claim.
'¢::: Let ~ be a Cauchy filter. Put !JIJ:= {cl V: V E ~}. The diameters ofthe
sets in !JIJ tend to zero. Hence, there is a point x such that x E cl V for all V E ~.
Plainly, ~ -+ x. Indeed, let V be an e h-small member of ~ and y E V. Some
y' in V is such that d(x, y') ::; e h. Therefore, d(x, y) ::; d(x, y') + d(y', y) ::; 6.
Consequently, V C Be(x) and so Be(x) E ~. I>
4.5.7. Nested Ball Theorem. A metric space is complete if and only if
every nested ( = decreasing by inclusion) sequence of balls whose radii tend to zero
has a unique point of intersection. <II>
48 Chapter 4. An Excursion into Metric Spaces
u
(Ul,U2)EU 2
(X,d) - (X,d)
~ lw
(Xt,dt)
4.6. Compactness and Completeness 49
d(f, g):= sup dJF(f(x), g(x)):= sup II(x) - g(x)1 (f, 9 E C(Q, IF));
xEQ xEQ
U:= n
JE£'
1-1 0 ()' 0 I
(cf. 4.2.9). Given 9 E 0' and lEg' such that go 1-1 C ()', observe that
()' = ()'-1 J (g 0 1-1 )-1 = (f-l )-1 0 g-1 = I 0 g-l.
U c ng-I
gEe
0 ()' 0 g
lQ C U U(xo) x U(xo)
xoEQo
for every g in g and some I in g'. Using the estimates, successively infer that
. ff(x+h)-f(x).
) ·-1·Imm
D + f( x.- h '
hto
D +f() 1· f(x+h)-f(x)
X := Imsup h .
hto
The elements D+f(x) and D+ f(x) of the extended axis R are the lower right and
upper right Dini derivatives of f at x. Further, let D stand for the set of functions
fin C([O, 1], lR) such that for some x E [0, 1) the elements D+f(x) and D+ f(x)
belong to lR; i.e., they are finite. Then D is a meager set. Hence, the functions
lacking derivatives at every point of (0, 1) are everywhere dense in C([O, 1], lR).
However, the explicit examples of such functions are not at all easy to find and
grasp. Below a few of the most popular are listed:
Lb
00
D\ Uint D"
10=1
is a holey disk or, more formally, a perforated disk. A subset of the plane which is
diffeomorphic (= "smoothly homeomorphic") to a holey disk is called a connected
elementary compactum. The union of a finite family of pairwise disjoint connected
elementary compacta is an elementary compactum.
4.8.6. REMARK. The boundary 8F of an elementary compact urn F com-
prises finitely many disjoint smooth loops. Furthermore, the embedding of F into
the (oriented) plane R2 induces on F the structure of an (oriented) manifold with
(oriented) boundary 8F. Observe also that, by 4.8.3, it makes sense to specify the
positive orientation of a smooth loop. This is done by indicating the orientation
induced on the boundary of the compact set surrounded by the loop.
4.8.7. If K is a compact subset of the plane and G is a nonempty open set
that includes K then there is a nonempty elementary compactum F such that
K C int Fe F c G. <It>
Exercises
4.1. Give examples of metric spaces. Find methods for producing new metric spaces.
4.2. Which filter on X2 coincides with some metric uniformity on X?
4.3. Let S be the space of measurable functions on [0, 1) endowed with the metric
1
d( )._/ I/(t)-g(t)1 dt (I, g E S)
I, g.- 1 + I/(t) - g(t)1
o
with some natural identification implied (specify it!). Find the meaning of convergence in the
space.
4.4. Given a, (3 E WN , put
dCa, (3):= 1/ min {k E W : ak :f. (3kl·
Check that d is a metric and the space WN is homeomorphic with the set of irrational numbers.
4.5. Is it possible to metrize pointwise convergence in the sequence space? In the func-
tion space IF[O, 1) ?
4.6. How to introduce a reasonable metric into the countable product of metric spaces?
Into an arbitrary product of metric spaces?
4.7. Describe the function classes distinguishable by erroneous definitions of continuity
and uniform continuity.
4.8. Given nonempty compact subsets A and B of the spaces ]W.N, define
Show that d is a metric. The metric is called the Hausdorff metric. What is the meaning
of convergence in this metric?
4.9. Prove that nonempty compact convex subsets of a compact convex set in ]W.N con-
stitute a compact set with respect to the Hausdorff metric. How does this claim relate to the
Ascoli-Arzela Theorem?
4.10. Prove that each lower semicontinuous function on ]W.N is the upper envelope
of some family of continuous functions.
4.11. Explicate the interplay between continuous and closed mappings (in the product
topology) of metric spaces.
4.12. Find out when a continuous mapping of a metric space into a complete metric
spaces is extendible onto a completion of the initial space.
4.13. Describe compact sets in the product of metric spaces.
4.14. Let (Y, d) be a complete metric spaces. A mapping F : Y --+ Y is called expanding
whenever d(F(x), F(y» ~ (3d(x, y) for some (3 > 1 and all x, y E Y. Assume that an expanding
mapping F : Y --+ Y acts onto Y. Prove that F is one-to-one and possesses a sole fixed point.
4.15. Prove that no compact set can be mapped isometrically onto a proper part of it.
4.16. Show normality of an arbitrary metric space (see 9.3.11).
4.17. Under what conditions a countable subset of a complete metric spaces is nonmea-
ger?
4.18. Is it possible to characterize uniform continuity in terms of convergent sequences?
4.19. In which metric spaces does each continuous real-valued function attain the supre-
mum and infimum of its range? When is it bounded?
Chapter 5
Multinormed and Banach Spaces
o
(5) X = dom p if and only if B p is absorbing.
<I If Xl, X2 E dom p and aI, a2 ElF then by 3.7.6
Hence, (1) holds. Suppose to the contrary that (2) is falsej i.e., p(x) < 0 for some
X EX. Observe that 0 ~ p( x) + p( -x) < p( -x) = p( x) < 0, a contradiction. The
claim of (3) is immediate from (2) and the subadditivity of p. The validity of (4)
and (5) has been examined in part (cf. 3.8.8). What was left unproven follows
from the Gauge Theorem. t>
5.1.2. Ifp, q: X -+ JR. are twoseminorms then theinequalityp::; q (in (JR.)X)
holds if and only if Bp J B q.
<I '*: Evidently, {q ::; I} C {p ::; I}.
'*=: In view of 5.1.1 (4), observe that p = PBp and q = PB q . Take tl, t2 E JR
such that tl < t 2. If tl < 0 then {q ~ ttl = 0, and so {q ::; ttl C {p ~ t 2}.
If tl 20 then tlBq C tlBp C t2Bp. Thus, by 3.8.3, p ::; q. t>
5.1.3. Let X and Y be vector spaces, let T C X x Y be a linear correspond-
ence, and let p : Y -+ JR. be a seminorm. If PT( x ) := inf p 0 T( x) for x E X then
5.1. Seminorms and Multinorms 57
PT : X -+ R· is a seminorm and the set BT := T-l(Bp) is absolutely convex.
In addition, PT = PBT.
<l Given Xl, X2 E X and al, a2 E Jr, infer that
IITxll
IITII:= sup{IITxll: x E X, IIxll :s:; I} = :~~ Txf'
(From now on, in analogous situations we presume 0 10 := 0.)
5.1. Seminorms and Multinorms 59
I/(x)1
IIIII = sup{l/(x)l: Ilxll ~ I} = :~~ W'
Consider X":= (X')':= B(X', IF), the second dual of X. Given x E X and
f E X', put x" := t(x) : f f-+ f(x). Undoubtedly, t(x) E (X')# = £l(X', IF).
In addition,
5.1.11. REMARK. The constructions, carried out in 5.1.10 (8), show some
symmetry or duality between X and X'. In this regard, the notation (x, f):=
(x If):= f(x) symbolizes the action of x E X on f E X, (or the action of f on x).
To achieve and ensure the utmost conformity, it is a common practice to denote
elements of X' by symbols like x'; for instance, (x I x') = (x, x') = x'(x).
<I =>: Let (x, y) ¢ Ix; i.e., x i= y. Then p(x - y) > 0 for some semi norm p
in rot Hence, (x, y) ¢ {d p :::; l/2p(x - y)}. But the last set is included in %'p,
and thus in %'!JJl. Consequently, X 2 \ Ix C X 2 \ n{v: V E %'rod. Furthermore,
Ix c n{v: V E %'!JJl}.
'¢::: If p( x) = 0 for all p E VR then (x, 0) E V for every V in %'!JJl. Hence,
(x, 0) E Ix by hypothesis. Therefore, x = O. t>
5.2.7. Given a space X witb uniformity %'x, define
TX(X)=SUp{Tp(X): pEVRx}
<l =>: Let Xl i- X2 and let c:= p(Xl-X2) > 0 for P E 9J1x. Put Ul := Xl +e hBp
and U2 := X2 + e h Bp. By 5.2.10, Uk E 7X(Xk). Verify that Ul n U2 = 0.
Indeed, if y E Ul n U2 then P(XI - y) ::; e hand p(X2 - y) ::; e h. Therefore,
P(XI - X2) ::; 2 h C < C = p(Xl - X2), which is impossible.
{=:: If (Xl, X2) E n{v: V E %'x} then X2 E n{V(xt): V E %'x}. Thus,
Xl = X2 and, consequently, 9J1x is separated by 5.2.6. I>
5.2.13. REMARK. The presence of a uniformity and the corresponding topol-
ogy in a multinormed space readily justifies using uniform and topological concepts
such as uniform continuity, completeness, continuity, openness, closure, etc.
5.2.14. Let (X, p) be a seminormed space and let Xo be a subspace of X.
The quotient space (X/Xo, px/xo) is separated if and only if Xo is closed.
<l =>: If X rf. Xo then cp(x) i- 0 where, as usual, cp : X -+ X/X o is the coset
mapping. By hypothesis, 0 i- c:= px/xo(cp(x)) = p",,-l(cp(X)) = inf{p(x + xo) :
Xo E X o}. Hence, the ball X + e h Bp does not meet Xo and X is an exterior point
of Xo. Thus, Xo is closed.
{=:: Suppose that x is a nonzero point of X/X o and x = cp(x) for some X
with suitable PI, ... ,Pn E ml and t > O. The right side of this inclusion is an ele-
ment of ~Pl V ... V ~Pn = ~{Pl'''' ,Pn} C ~!m. Hence, V is also a member of ~!m. t>
5.3.3. DEFINITION. Let p, q : X -+ lR. be two seminorms on X. Say that P
is finer or stronger than q and write P )- q whenever {p} )- {q}. The equivalence
of seminorms P '" q is understood in a routine fashion.
5.3.4. p)- q {:} (3 t > 0) q ::; tp {:} (3 t 2: 0) Bq J tBp;
(the series on the right side of the above formula is dominated by the convergent
series I:~l 112k, and so d is defined soundly).
5.4. Metrizable and Normable Spaces 65
1 1
L L
00 00
k=l
It remains to established that %'d and %'m coincide.
First, show that %'d C %'m. Take a cylinder, say, {d ::; c}; and let (x, y) E
{d p1 ::; t} n ... n {d pn ::; t}. Since a is an increasing function, deduce that
d( ~ 1 Pk(X-y) ~ 1 Pk(x-y)
x, y) = ~ 2k 1 + Pk(X - y) + k~l 2k -l..:..+-p-'-k...,..(x----'--'--y...,..)
t l 1 t 1
L 2k + L 2k ::; 1 + t +
n 00
::; 1 + t 2n .
k=l k=n+l
Since t(l + t)-l + 2- n tends to zero as n - t 00 and t - t 0, for appropriate t and n
observe that (x, y) E {d ::; c}. Hence, {d ::; c} E %'m, which is required.
Now establish that 'Wm C 'Wd • To demonstrate the inclusion, given Pn E 9J1
and t > 0, find c > 0 such that {d pn ::; t} J {d ::; c}. For this purpose, take
1 t
c·----
.- 2n 1 + t'
it may be assumed that V = Bp for some seminorm p in the given multinorm !.m.
Undoubtedly, p -<!.m. Now if U E T!JJl(O) then nU ::l V for some n E N. Hence,
U E Tp(O). Using Theorem 5.3.2, observe that p >-!.m. Thus, p rv !.m; and,
therefore, p is also separated by 5.2.12. This means that p is a norm. I>
5.4.6. REMARK. Incidentally, 5.4.5 shows that the presence of a bounded
neighborhood of zero in a multinormed space X amounts to the "seminormability"
ofX.
<I ::::}: Let 2::::11Ixnll < +00 for some (countable) sequence (xn). Then the
sequence of partial sums Sn := Xl + ... + Xn is fundamental because
m m
for m > k.
~: Given a fundamental sequence (x n ), choose an increasing sequence (nk)kEN
such that IIxn - xmll::; 2- k as n, m 2: nk. Then the series x n1 +(xn2 -xnJ+(xna-
x n2 ) + ... converges in norm to some x; i.e., x nk -+ x. Observe simultaneously
that Xn -+ x. t>
5.5.4. If X is a Banach space and Xo is a closed subspace of X then the
quotient space XI XO is also a Banach space.
<I Let r.p : X -+ !Z:= XIX o be the coset mapping. Undoubtedly, for every
x E !Z there is some x E r.p-1(x) such that 211xll 2: IIxll 2: IIxli. Hence, given
2:::'=1 x n , a norm convergent series in !Z, it is possible to choose Xn E r.p-1(xn )
so that the norm series 2:::'=1 IIxnll be convergent. According to 5.5.3, the sum
X:= 2:::'=1 xn is available. Ifx:= r.p(x) then
n n
extension of the initial integral f. (The traditional liberty is taken of using the
same symbol for the original and its successor.)
An element of dom .Ai is an integrable or summable function. The integrabil-
ity of f E IF,c is equivalent to the integrability of its real part Re f and imaginary
part 1m f, both members of lR.,c. For the sake of completeness, recall the definition
N(g):= inf {sup J Xn : (xn) eX, Xn ::; Xn+l, (\:f e E 6") Ig(e)1 = li~ xn(e)}
for an arbitrary g in IF,c. If IF = lR. then dom.Ai obviously presents the closure
of X in the normed space (dom N, N).
The Holder inequality is valid:
.Ai (J g) ::; .%p (J).%pl (g),
with p' the conjugate exponent of p, i.e. 1/ p + 1/ p' = 1 for p > l.
<l This is a consequence of the Young inequality, xy - x P/ p ::; ypl / p' for all
x, y E lR+, applied to Ifl/.%p(J) and Igl/.%pl(g) when ./ij,(J) and ./ij,1(g) are both
nonzero. If ./ij,(J)./ij,1 (g) = 0 then the Holder inequality is beyond a doubt. !>
The set 2'p:= dom .%p is a vector space.
<l If + glP ::; (If I + Igl)P ::; 2P(lfl V Igl)P = 2P(lf1 P V IgIP) :::; 2P(lf1 P + IgI P) !>
The function .%p is a seminorm, satisfying the Minkowski inequality:
./ij,(J + g) ::; ./ij,(J) + ./ij,(g).
<l For p = 1, this is trivial. For p > 1 the Minkowski inequality follows from
the presentation
whose right side is the upper envelope of a family of seminorms. To prove the
above presentation, using the Holder inequality, note that g:= If( /pl lies in 2'q
when .%p(J) > 0; furthermore, .%p(J) = .Ai (Jg)/.%pl (g). Indeed, .Ai(Jg) =
f Ifl PIpl+l = .%p(J)P, because P/ pI + 1 = P (1 - 1/ p) + 1 = p. Continue arguing
to find .%pI (g)p l = f IgIP ' = f Ifl P = .%p(J)P, and so .%pl(g) = .%p(J/lpl. Finally,
.Ai(Jg)/.%pI(g) = .%p(J)P/.%p(J) pIpI = .%p(J) p-Plpl = .%p(J) p(l-ll pI) = .%p(J). !>
The quotient space 2'p/ ker.%p, together with the corresponding quotient
norm II . lip, is called the space of p-summable functions or Lp space with more
complete designations Lp(S), Lp(6", X, j), etc.
Finally, if a system with integration S arises from inspection of measurable
step functions on a measure space (n, d, /l) then it is customary to write
Lp(n, tzI, /l), Lp(n, /l) and even Lp(/l), with the unspecified parameters clear
from the context.
70 Chapter 5. Multinormed and Banach Spaces
It is in common parlance to call the elements of Leo (like the elements of 2'eo)
essentially bounded functions. The space Leo presents a Banach space. <ll>
The space Leo, as well as the spaces C(Q, IF), lp(g), co(g), c, lp, and
Lp (p ;::: 1), also bears the unifying title "classical Banach space." Nowadays
a Lindenstrauss space which is a space whose dual is isometric to Ll (with respect
to some system with integration) is also regarded as classical. It can be shown
that a Banach space X is classical if and only if the dual X, is isomorphic to one
of the Lp spaces with p ;::: l.
(6) Consider a system with integration S:= (g, X, J) and let p ;::: l.
Suppose that for every e in g there is a Banach space (Ye , II ·lIyJ. Given an arbi-
trary element! in I1eE&' Ye, define III!III : e 1-+ 11!(e)lly•. Put Np(f):= inf{JY;,(g) :
9 E 2'p, 9 ;::: III!III}. It is clear that dom Np is a vector space equipped with the
semi norm Np. The sum of the family in the sense of Lp or simply the p-sum
of (Ye)eE&' (with respect to the system with integration S) is the quotient space
dom Np/ker Np under the corresponding (quotient) norm 111·lll p '
The p-sum of a family of Banach spaces is a Banach space.
<l If L:~l Np(/k) < +00 then the sequence (Sn:= L:~=l 111!klll) tends to some
almost everywhere finite positive function 9 and Np(g) < +00. It follows that for
almost all e E g the sequence (sn(e)) (i.e., the series L:~lllfk(e)IIYJ converges.
By the completeness of Y e , the seriesaL.~l !k(e) converges to some sum !o(e)
in Y e with II!o(e)lly. ~ gee) for almost every e E g. Therefore, it may be assumed
that !o E dom Np. Finally, Np O=~=l!k - !o) ~ L:~n+l Np(fk) --+ O. I>
In the case when g:= N with conventional summation, for the sum!D of a se-
quence of Banach spaces (Yn)nEN (in the sense of Lp) the following notation is
often employed:
In the case Ye:= X for all e E g, where X is some Banach space over IF, put §p:=
dom Np and Fp:= §p/ ker N p. An element of the so-constructed space is a vector
field or a X-valued function on g (having a p-summable norm). Undoubtedly, Fp
is a Banach space. At the same time, if the initial system with integration contains
a nonmeasurable set then extraordinary elements are plentiful in Fp (in particular,
for the usual Lebesgue system with integration Fp f- Lp). In this connection the
functions with finite range, assuming each value on a measurable set, are selected
72 Chapter 5. Multinormed and Banach Spaces
in §p. Such a function, as well as the corresponding coset in Fp, is a simple, finite-
valued or step function. The closure in Fp of the set of simple functions is denoted
by Lp (or more completely Lp(X), Lp(S, X), Lp(n, PI, Il), Lp(n, Il), etc.) and is
the space of X-valued p-summable functions. Evidently, Lp(X) is a Banach space.
It is in order to illustrate one of the advantages of these spaces for p = 1. First,
notice that a simple function f can be written as a finite combination of charac-
teristic functions:
f = L Xf-l(X)X,
xEimf
JI I fill J L
= IIXf-l(x)XII
JL J
xEimf
=
xEim f
Xf-l(x) Ilxll = L
xEimf
Ilxll Xf-l(X) < +00.
Next, associate with each simple function f some element in X by the rule
J f:= L
xEimf
J Xf-l(x)X.
= JL xEimf
Ilxll Xf-l(X) = JIllflll·
By virtue of 4.5.10 and 5.3.8, the operator J has a unique extension to an element
of B(Ll(X), X). This element is denoted by the same symbol, f (or fg, etc.), and
is referred to as the Bochner integral.
(7) In the case of conventional summation, the usage of the scalar theory
is preserved for the Bochner integral. Namely, the common parlance favours the
term "sum of a family" rather than "integral of a summable function," and the
symbols pertinent to summation are perfectly welcome. What is more important,
infinite dimensions bring about significant complications.
5.6. The Algebra of Bounded Operators 73
Let (xn) be a family of elements of a Banach space. Its summability (in the
sense of the Bochner integral) means the summability of the numeric family (lIx n 11),
i.e. the norm convergence of (xn) as a series. Consequently, (xn) has at most
count ably many nonzero elements and may thus be treated as a (countable) se-
quence. Moreover, 2:::'=lllxnll < +00; i.e. the series Xl + X2 + ... converges in
norm. By 5.5.3, for the series sum x = 2::::"=1 Xn observe that x = limo 80, where
80:= 2::nEO Xn is a partial sum and () ranges over the direction of all finite sub-
sets of N. In this case, the resulting x is sometimes called the unordered sum of
(xn), whereas the sequence (xn) is called unconditionally or unorderly summable
to x (in symbols, x = 2::nEN xn). Using these terms, observe that summability in
norm implies unconditional summability (to the same sum). If dimX < +00 then
the converse holds which is the Riemann Theorem on Series. The general case is
explained by the following deep and profound assertion:
Dvoretzky-Rogers Theorem. In an arbitrary infinite-dimensional Banach
space X, for every sequence of positive numbers (t n ) such that 2::::"=1 t; < +00
there is an unconditionally summable sequence of elements (xn) with Ilxnll = tn
for all n E N.
In this regard for a family of elements (X e)eE6' of an arbitrary multinormed
space (X, 9)1) the following terminology is accepted: Say that (X e)eE6' is summable
or unconditionally summable (to a sum x) and write x:= 2::eE6' Xe whenever x is
the limit in (X, 9)1) of the corresponding net of partial sums 80, with () a finite
subset of tf:; i.e., 80 - t X in (X, 9)1). If for every p there is a sum 2::eE,cP(Xe)
then the family (X e)eE6' is said to be multinorm summable (or, what is more exact,
fundamentally summable, or even absolutely fundamental).
In conclusion, consider a Banach space!D and T E B(X, !D). The operator T
can be uniquely extended to an operator from L 1 (X) into L 1 (!D) by putting Tf :
e f-.-+ Tf(e) (e E tf:) for an arbitrary simple X-valued function f. Then, given
f E L 1 (X), observe that Tf E L 1 (!D) and f6' Tf = T Ie
f. This fact is verbalized
as follows: "The Bochner integral commutes with every bounded operator." <It>
Consequently,
1/ 1/ k(n)/
reT) :::; IITnl1 n:::; M n IIT"II n
1 k(n)./ 1/ (n-I(n))/
:::;M / n(r(T)+c:) n =M n(r(T)+c:) n.
Since M1/n -+ 1 and (n-I(n»/n -+ 1, find r(T):::; limsupllT n l1 1/n :::; reT) +c:.
The inequality liminf/lTnll1/n ;::: reT) is evident. Recall that c: is arbitrary, thus
completing the proof. I>
5.6.9. Neumann Series Expansion Theorem. With X a Banach space
and T E B(X), the following statements are equivalent:
(1) the Neumann series 1 +T + T2 + ... converges in the operator norm
of B(X);
(2) IITkl1 < 1 for some k and N;
(3) reT) < 1.
If one of the conditions (1)-(3) holds then 2:~o Tk = (1 - T)-I.
<l (1) => (2): With the Neumann series convergent, the general term (Tk) tends
to zero.
(2) => (3): This is evident.
(3) => (1): According to 5.6.8, given a suitable c: > 0 and a sufficiently large
kEN, observe that reT) :::; IITk 111 /k :::; reT) + c: < 1. In other words, some tail
of the series 2:~o IITkl1 is dominated by a convergent series. The completeness
of B(X) and 5.5.3 imply that 2:~o Tk converges in B(X).
Now let S:= 2:~o Tk and Sn:= 2:~=o Tk. Then
k=l k=O
Hence,
Hence, by 5.6.11, (T- 1S)-l belongs to B(X). Put R:= (T- 1S)-lT- 1. Clearly,
R E B(Y, X) and, moreover, R = S-1(T- 1)-IT- 1 = S-l. Further,
2) -l)k(A -
<Xl
1 Tk
= >: L
<Xl
R(T, A) Ak
k=O
holds. Moreover,IIR(T, A)II-+ 0 as IAI-+ +00.
<3 Since I (A - T) - (AO - T) I = IA - AO I, the openness property of res (T)
follows from 5.6.12. Proceed along the lines
R(T, A) = (A - T)-l
= L( -l)k(A -
<Xl
Tk
fA )-1
<Xl
R(T, A) = >:1 ( 1 - T
= >:1 "L.t If·
k=O
It is clear that
1 1
IIR(T, A)II S 1>:1. 1- ~TII/IAI· I>
5.6.17. REMARK. It is worth keeping in mind that the inequality IAI > reT)
is a necessary and sufficient condition for the convergence of the Laurent series,
R(T, A) = L:~o Tk / Ak+1, which expands the resolvent of T at infinity.
5.6.18. An operator S commutes with an operator T if and only if S com-
mutes with the resolvent of T.
78 Chapter 5. Multinormed and Banach Spaces
Exercises
5.1. Prove that a normed space is finite-dimensional if and only if every linear functional
on the space is bounded.
5.2. Demonstrate that it is possible to introduce a norm into each vector space.
5.3. Show that a vector space X is finite-dimensional if and only if all norms on X are
equivalent to each other.
5.4. Demonstrate that all separated multinorms introduce the same topology in a finite-
dimensional space.
5.5. Each norm on ]R.N is appropriate for norming the product of finitely many normed
spaces, isn't it?
5.6. Find conditions for continuity of an operator acting between multinormed spaces
and having finite-dimensional range.
5.7. Describe the operator norms in the space of square matrices. When are such norms
comparable?
S.S. Calculate the distance between hyperplanes in a normed space.
5.9. Find the general form of a continuous linear functional on a classical Banach space.
5.10. Study the question of reflexivity for classical Banach spaces.
5.11. Find the mutual disposition of the spaces Ip and Ipl as well as Lp and Lpl. When
is the complement of one element of every pair is dense in the other?
5.12. Find the spectrum and resolvent of the Volterra operator (the taking of a primi-
tive), a projection, and a rank-one operator.
5.13. Construct an operator whose spectrum is a prescribed nonempty compact set in C.
5.14. Prove that the identity operator (in a nonzero space) is never the commutator
of any pair of operators.
5.15. Is it possible to define some reasonable spectrum for an operator in a multinormed
space?
5.16. Does every Banach space over IF admit an isometric embedding into the space
C(Q, IF), with Q a compact space?
5.17. Find out when Lp(X)' = Lpl(X'), with X a Banach space.
S.lS. Let (Xn) be a sequence of normed spaces and let
be their co-sum (with the norm Ilxll = sup{ IIxnll : n E N} induced from the loo-sum). Prove that
Xo is separable if and only if so is each of the spaces X n .
5.19. Prove that the space C(p)[O, 1] presents the sum of a finite-dimensional subspace
and a space isomorphic to C[O, 1].
Chapter 6
Hilbert Spaces
<l If (x, x) = (y, y) = 0 then 0:::; (x+ty, x+ty) = t(x, y)* +t*(x, y). Letting
t:= -(x, y), find that -21(x, y)12 ~ 0; i.e., in this case the claim is established.
If, for definiteness, (y, y) i= 0; then in view of the estimate
which reads: the sum of the squares of the lengths of the diagonals equals the sum
of the squares of the lengths of all sides.
Given the mapping ( " Y)JR, from the Parallelogram Law successively derive
Since the mapping ( " Y)IR is continuous for obvious reasons, conclude that
(', Y)JR E (HIR)#' Put
Furthermore,
6.1.10. EXAMPLES.
(1) A Hilbert space is exemplified by the L2 space (over some system with
integration), the inner product introduced as follows (j, g):= fg* for f, g E L2. J
In particular, (x, y):= L:eE8' xeY: for x, y E l2(6").
(2) Assume that H is a pre-Hilbert space and ( " .): H2 --t IF is a semi-
inner product on H. It is clear that the real carrier HJR with the semi-inner
product (', ')JR : (x, y) 1-+ Re(x, y) presents a pre-Hilbert space with the norm
of an element of H independent of whether it is calculated in H or in HJR. The pre-
Hilbert space (HJR, (', . )JR) is the realification or decomplexification of (H, (', . )).
In turn, if the real carrier of a seminormed space is a pre-Hilbert space then
the process of complexification leads to some natural pre-Hilbert structure of the
original space.
(3) Assume that H is a pre-Hilbert space and H* is the twin vector
space of H. Given x, y E H*, put (x, y)*:= (x, y)*. Clearly, (', .)* is a semi-
inner product on H *. The resulting pre-Hilbert space is the twin of H, with the
denotation H* preserved.
(4) Let H be a pre-Hilbert space and let Ho := ker II . II be the kernel
of the semi norm II . lion H. Using the Cauchy-Bunyakovskil'-Schwarz inequality,
Theorem 2.3.8 and 6.1.10 (3), observe that there is a natural inner product on
the quotient space H/Ho: If XI:= <p(xI) and X2:= <p(X2), with Xl, x2 E Hand
<p: H --t H/Ho the coset mapping, then (Xl, X2):= (Xl, X2). Moreover, the pre-
Hilbert space H / Ho may be considered as the quotient space of the seminormed
space (H, 11·11) by the kernel of the semi norm 11·11. Thus, H / Ho is a Hausdorff space
referred to as the Hausdorff pre-Hilbert space associated with H. Completing the
normed space H/Ho, obtain a Hilbert space (for instance, by the von Neumann-
Jordan Theorem). The so-constructed Hilbert space is called associated with the
original pre-Hilbert space.
(5) Assume that (He)eE8' is a family of Hilbert spaces and H is the 2-sum
ofthe family; i.e., h E H if and only if h:= (he)eE8', where he E He for e E 6" and
IIhll:= (2:
eE8'
Il h el1 2 )
'/
2 < +00.
84 Chapter 6. Hilbert Spaces
= 1 Iz (L
eE,c
life + gel1 2 + + L
eE,c
life - ge11 2)
6.2. Orthoprojections
6.2.1. Let Ue be a convex subset oftbe spbericallayer (r +E)BH \ rBH witb
r, E > 0 in a Hilbert space H. Tben tbe diameter of Ue vanisbes as E tends to o.
<l Given x, y E Ue , on considering that 1 Iz(x + y) E Ue and applying the
Parallelogram Law, for E ~ r infer that
<l Put Ue := {u E U: Ilx - ull ~ inf IIU - xii +E}. By 6.2.1, the family (Ue)e>O
constitutes a base for a Cauchy filter in U. t>
6.2.3. DEFINITION. The element Uo appearing in 6.2.2 is the best approxima-
tion to x in U or the projection of x to U.
6.2.4. Let Ho be a closed subspace of a Hilbert space H and x E H \ Ho.
An element Xo of Ho is tbe projection of x to Ho if and only if (x - Xo, h o ) = 0
for every ho in H o.
6.2. Orthoprojections 85
<I It suffices to consider the real carrier (Ho)JR of Ho. The convex function
f(h o ) := (ho - x, ho - x) is defined on (Ho)JR. Further, Xo in Ho serves as the
projection of x to Ho if and only if 0 E oxo (1). In view of 3.5.2 (4) this containment
means that (x - Xo, h o) = 0 for every ho in H o, because f'(xo) = 2(xo - x, .). I>
6.2.5. DEFINITION. Elements x and y of H are orthogonal, in symbols x 1.. y,
if (x, y) = o. By U.l.. we denote the subset of H that comprises all elements
orthogonal to every point of a given subset U; i.e.,
n n
k=l k=l
for x E Hand l #- m.
In particular, putting x:= P,X, observe that
Furthermore, in virtue of6.2.10 (4), (Pkx, y) = (x, PkY) and so (Px, y) = (x, Py).
It suffices to appeal to 6.2.10 (4) once again. t>
6.2.14. REMARK. Theorem 6.2.13 is usually referred to as the pairwise or-
thogonality criterion for finitely many orthoprojections.
6.3. A Hilbert Basis 87
<l Let S(J := L:eE,g' Xe , where () is a finite subset of G'. By 6.2.8, IIs(J11 2 =
L:eE(J Ilxe 112. Given a finite subset ()f of G' which includes (), thus observe that
<l Given a finite subset () of G', put S(J := L:eE(J Pe. By Theorem 6.2.13, S(J is
an orthoprojection. Hence, in view of 6.2.8, IIs(Jx11 2 = L:eE(J IIPexI1 2 ::; IIxII 2
for every x in H. Consequently, the family (IiPexII 2)eE,g' is summable (the net
of partial sums is increasing and bounded). By the Pythagoras Theorem, there
is a sum Px := L:eE,g' Pex; i.e., Px = lim(J S(JX. Whence p 2 x = lim(J s(JPx =
lim(J S(J lim(J1 S(JIX = lim(J lim(J1 S(JS(JIX = lim(J lim(J1 S(Jn(J'X = lim(J S(JX = Px. Finally,
IIPxll = Illim(J s(Jxll = lim(J Ils(Jxll ::; Ilxll and, moreover, p 2 = P. Appealing to
6.2.10, conclude that P is the orthoprojection onto im P.
If x E im P, i.e., Px = x; then x = L:eE,g' Pex and by the Pythagoras Theorem
L:eE,g' IIPexll 2 = IIxII 2 = IIPxI1 2 < +00. Since Pex E He (e E G'); therefore, x E .Ye.
If Xe E He and L:eE,g' IIxe 112 < +00 then for X:= L:eE,g' Xe (existence follows from
the same Pythagoras Theorem) observe that x = L:eE,g' Xe = L:eE,g' Pexe = Px;
i.e., x E im P. Thus, im P = .Ye. I>
88 Chapter 6. Hilbert Spaces
6.3.4. REMARK. This theorem may be treated as asserting that the space.Yt'
and the Hilbert sum of the family (He)eE<C are isomorphic. The identification is
clearly accomplished by the Bochner integral presenting the process of summation
in this case.
6.3.5. REMARK. Let h in H be a normalized or unit or norm-one element;
i.e., IIhll = 1. Assume further that Ho := lFh is a one-dimensional subspace of H
spanned over h o. For every element x of H and every scalar .x, a member of IF,
observe that
n-l
Yn
Yn:= Xn - L(ek)x n , (n EN).
k=O
e n := IIYnll
(x, y) = L xefJ.*·
eE&'
<l By the Pythagoras Theorem, the Fourier transform acts in 12( I!). By Theo-
----
rem 6.3.3, ~ is an epimorphism. By the Steklov Theorem, ~ is a monomorphism.
It is beyond a doubt that §-I X = x for x E Hand §-I(x) = x for x E 12(1!).
The equality
IIxII 2 = L IIxII 2 = IIxell~ (x E H)
eE&'
90 Chapter 6. Hilbert Spaces
6.3.17. REMARK. The Parseval identity shows that the Fourier transform
preserves inner products. Therefore, the Fourier transform is a unitary operator
or a Hilbert-space isomorphism; i.e., an isomorphism preserving inner products.
This is why the Riesz-Fisher Theorem is sometimes referred to as the theorem
on Hilbert isomorphy between Hilbert spaces (of the same Hilbert dimension).
Therefore, x t-t x' is an isometry of H * into H'. Check that this mapping is
an epimorphism.
Let 1 E H' and Ho := ker I i= H (if there no such I then nothing is to be
proven). Choose a norm-one element e in Ht and put grad 1:= I(e)*e. If x E Ho
then
(grad I)'(x) = (x, grad I) = (x, I(e)*e) = I(e)**(x, e) = o.
Consequently, for some a in IF and all x E H in virtue of2.3.12 (grad I)'(x) = al(x).
In particular, letting x:= e, find
for all x E H t and y E H2. Moreover, T* E B(H2' Ht) and IIT*II = IITII.
<l Let y E H 2. The mapping x 1-+ (Tx, y) is the composition y' 0 T; i.e.,
it presents a continuous linear functional over H t . By the Riesz Prime Theorem
there is precisely one element x of H t for which x' = y' 0 T. Put T*y := x.
It is clear that T* E !L'(H2' Ht). Furthermore, using the Cauchy-Bunyakovskil-
Schwarz inequality and the normative inequality, infer that
IITxl1 2= (Tx, Tx) = I(Tx, Tx)1 = I(T*Tx, x)1 ~ IIT*Txll Ilxll ~ IIT*TII.
Furthermore, using the submultiplicativity of the operator norm and 6.4.4, infer
IIT*TII ~ IIT*IIIITIl = IITII2, which proves (4). t>
92 Chapter 6. Hilbert Spaces
6.4.7. Let H l , Hz, and H3 be three Hilbert spaces. Assume further that
T E B(Hl' Hz) and S E B(Hz, H3). Then (ST)* = T*S*.
<l (STx, z) = (Tx, S*z) = (x, T*S*z) (x E H l , Z E H 3) I>
<l Put t:= sup{I(Tx, x)l: Ilxll ~ I}. It is clear that I(Tx, x)1 ~ IITxllllxll ~
IITII provided IIxll ~ 1. Thus, t ~ IITII·
Since T = T*; therefore, (Tx, y) = (x, Ty) = (Ty, x)* = (y, Tx)*; i.e.,
(x, y) 1-+ (Tx, y) is a hermitian form. Consequently, in virtue of 6.1.3 and 6.1.8
ii>.x - Txll ~ I(>'x - Tx, x)1 = I>' - (Tx, x)1 = >. - (Tx, x) ~ >. - MT > O.
Once again>. E res (T). Finally, Sp(T) C [mT' MT).
Since (Tx, x) E R for x E H; therefore, in virtue of 6.5.2
where the operator T>., the part of T in Hi:, is hermitian and compact, with
Sp(T>.) = Sp(T) \ {.x}.
<I The subspace H>. is finite-dimensional in view of the compactness of T.
Furthermore, H>. is invariant under T. Consequently, the orthogonal complement
Hi: of H>. is an invariant subspace of T* (coincident with T), since h E Hi: '*
(Vx E H>.) x..l h =} (Vx E H>.) 0 = (h, Tx) = (T*h, x) =} T*h E Hi:.
The part of T in H>. is clearly oX. The part T>. of T in Hi: is undoubtedly
compact and hermitian. Obviously, for Jl t= oX, the operator
<I Using 6.5.6 and 6.6.6 as many times as need be, for every finite subset ()
of Sp(T) obtain the equality
L Ak(ek) = L Ake~
00 00
T = @ ek,
k=I k=I
where the series is summed in operator norm. <It>
T= Lllke~@!k.
k=I
<l Put S:= T*T. It is clear that S E B(HI ) and S is compact. Furthermore,
(Sx, x) = (T*Tx, x) = (Tx, Tx) = IITxI12. Consequently, in virtue of 6.4.6, Sis
hermitian and Ho := ker S = ker T. Observe also that SpeS) c 1R+ by Theorem
6.5.5.
Let (ekhEN be an orthonormal basis for Ht comprising all eigenvalues of S
and let (Ak )kEN be a corresponding decreasing sequence of strictly positive eigen-
values Ak > 0, kEN (cf. 6.6.8). Then every element x E HI expands into the
98 Chapter 6. Hilbert Spaces
Fourier series
00
Therefore, considering that T PHo = 0 and assigning /-lk:= y'):k and ik:= /-l;lTek,
find
00 00 00
Ten Tem)
( in, im ) = ( - 1
-, - - = --(Ten, Te m )
/-In /-lm /-In/-lm
= _l_(T*Te n , em) = _l_(Se n , em)
/-In/-lm /-In/-lm
= _-l--(..\ne n , em) = /-In (en, em).
/-In, /-lm /-lm
Successively using the Pythagoras Theorem and the Bessel inequality, argue as
follows:
00 00
k=n+l k=n+l
Exercises
6.1. Describe the extreme points of the unit ball of a Hilbert space.
6.2. Find out which classical Banach spaces are Hilbert spaces and which are not.
6.3. Is a quotient space of a Hilbert space also a Hilbert space?
6.4. Is it true that each Banach space may be embedded into a Hilbert space?
6.5. Is it possible that the (bounded) endomorphism space of a Hilbert space presents
a Hilbert space?
6.6. Describe the second (= repeated) orthogonal complement of a set in a Hilbert
space.
6.7. Prove that no Hilbert basis for an infinite-dimensional Hilbert space is a Hamel
basis.
6.S. Find the best approximation to a polynomial of degree n +1 by polynomials of
degree at most n in the L2 space on an interval.
6.9. Prove that z ..l. y if and only if Ilz+yW = IIzW + IlyW and liz +iY112 = IIzll2 + IIYI12.
6.10. Given a bounded operator T, prove that
6.11. Reveal the interplay between hermitian forms and hermitian operators (cf. 6.5.3).
6.12. Find the adjoint of a shift operator, a multiplier, and a finite-rank operator.
6.13. Prove that an operator between Hilbert spaces is compact if and only if so is its
adjoint.
6.14. Assume that an operator T is an isometry. Is T* an isometry too?
6.15. A partial isometry is an operator isometric on the orthogonal complement of its
kernel. What is the structure of the adjoint of a partial isometry?
6.16. What are the extreme points of the unit ball of the endomorphism space of
a Hilbert space?
6.17. Prove that the weak topology of a separable Hilbert space becomes metrizable
if restricted onto the unit ball.
6.lS. Show that an idempotent operator P in a Hilbert space is an orthoprojection
if and only if P commutes with P*.
6.19. Let (akl)k,IEN be an infinite matrix such that akl ~ 0 for all k and l. Assume
further that there are also Pk and (3, "f > 0 satisfying
00 00
Then there is some T in B(l2) such that (ek, ed akl and IITII = y'jh, where ek is the
characteristic function of k, a member of N.
Chapter 7
Principles of Banach Spaces
Now putting Xo := _1/ 2Ul + Xo and €:= 4 and applying the argument of the
preceding paragraph, find an element U2 of U satisfying 111/4 u2 + 1h U l - Xo :S II
1 he b = 1/ 8 b. Proceeding by induction, construct a sequence (Un)nEN ofthe points
of U which possesses the property that the series 2::::'=11 hnUn converges to Xo.
Since 2::::'=1 1/ 2n = 1 and the set U is ideally convex, deduce Xo E U. t>
7.1.6. For every ideally convex set U in a Banach space the following four
sets coincide: the core of U, the interior of U, the core of the closure of U and the
interior of the closure of U.
<I It is clear that int U c core U C core cl U. If U E core cl U then cl (U - u)
equal to cl U -u is an absorbing set. An ideally convex set translates into an ideally
convex set (cf. 7.1.4 (1)). Consequently, U -u is a neighborhood of zero by Banach's
Fundamental Principle. By virtue of 5.2.10, u belongs to int U. Thus, int U =
core U = core cl U. Using 7.1.1, conclude that int cl U = int U. t>
7.1.7. The core and the interior of a closed convex set in a Banach space
coincide.
<I A closed convex set is ideally convex. t>
7.1.8. REMARK. Inspection ofthe proof of 7.1.5 shows that the condition for
the ambient space to be a Banach space in 7.1.7 is not utilized to a full extent.
There are examples of incomplete normed spaces in which the core and interior of
each closed convex set coincide. A space with this property is called barreled. The
concept of barreledness is seen to make sense also in multinormed spaces. Barreled
multinormed spaces are plentiful. In particular, such are all Frechet spaces.
7.1.9. COUNTEREXAMPLE. Each infinite-dimensional Banach space contains
absolutely convex, absorbing and not ideally convex sets.
<I Using, for instance, a Hamel basis, take a discontinuous linear functional f.
Then the set {If I :S I} is what was sought. t>
7.2.6. Let X be a Banach space and let U be a subset of X'. Then tbe
following statements are equivalent:
(1) U is bounded in X';
(2) for every x in X tbe numeric set {(x I x'): x' E U} is bounded in IF.
<l This is a particular case of 7.2.5. t>
104 Chapter 7. Principles of Banach Spaces
Take 1£ > 0 and choose, first, x E E such that 2suPn IITnllllx - xII ::; "12, and,
second, n E N such that IITmx - TkXII ::; ., 12 for m, k ~ n. By virtue of what was
proven IITmx - TkXIl ::; 1£; i.e., (TnX)nEN is a Cauchy sequence in Y. Since Y is
a Banach space, conclude that x E E. Thus, (2) => (1) is proven.
To complete the proof it suffices to observe that
<I~: If (Xl, yd, (X2' Y2) E F and 01, 02 ~ 0, 01 +02 = 1, then 01Y+02Y2 E
F(OIXI + 02X2) since YI E F(XI) and Y2 E F(X2)'
=>: If either Xl or X2 fails to enter in dom F then there is nothing to prove. If
YI E F(xd and Y2 E F(X2) with Xl, X2 E dom F then OI(XI, yJ)+02(X2, Y2) E F
for OJ, 02 ~ 0, 01 + 02 = 1 (cf. 3.1.2 (8)). I>
7.3.2. REMARK. Let X and Y be Banach spaces. It is clear that there are
many ways for furnishing the space X X Y with a norm so that the norm topology
be coincident with the product of the topologies TX and Ty. For instance, it is
possible to put lI(x, y)II:= IIxlix + lIylly; i.e., to define the norm on X X Y as that
of the I-sum of X and Y. Observe immediately that the concept of ideally convex
106 Chapter 7. Principles of Banach Spaces
set has a linear topological character: the class of objects distinguished in a space
is independent of the way of introducing the space topology; in particular, the class
remains invariant under passage to an equivalent (multi)norm. In this connection
the next definition is sound.
7.3.3. DEFINITION. A correspondence F C X x Y, with X and Y Banach
spaces, is called ideally convex or, briefly, ideal if F is an ideally convex set.
7.3.4. Ideal Correspondence Lemma. The image of a bounded ideally
convex set under an ideal correspondence is an ideally convex set.
<l Let F c X x Y be an ideal correspondence and let U be a bounded ideally
convex set in X. If U n dom F = 0 then F(U) = 0 and nothing is left unproven.
Let now (Yn)nEN C F(U); i.e., Yn E F(xn), where Xn E U and n E N. Let, finally,
(an) be a sequence of positive numbers such that L::"=l an = 1 and, moreover,
there is a sum of the series y:= L::"=l anYn in Y. It is beyond a doubt that
Here use was made of the fact that II)..xll = 1; i.e., Ax E Bx. l>
7.4. Open Mapping and Closed Grapb Tbeorems 107
7.4.8. Corollary. Suppose that X and Yare Banach spaces and T is a linear
operator from X to Y. The following conditions are equivalent:
(1) T E B(X, Y)j
(2) for every sequence (Xn)nEN in X, together with some x in X and y
in Y satisfying Xn --+ x and TX n --+ y, it happens that y = Tx.
<I (2): This is a reformulation of the closure property of T. t>
~l~
Z
is commutative for some ~ in B(Y, Z) if and only if ker ACker B.
<l Only -<= needs examining. Moreover, in the case im A = Y the sole member
!to of £(Y, Z) such that !tOA = B is continuous. Indeed, ~-l(U) = A(B-l(U))
for every open set U in Z. The set B-l(U) is open in virtue of the boundedness
property of B, and A(B-l(U)) is open by the Banach Homomorphism Theorem.
In the general case, construct !to E B(im A, Z) and take as ~ the operator !tOP,
where P is some continuous projection of Y onto im A. (Such a projection is
available by the Complementation Principle.) I>
7.4.13. Phillips Theorem. Suppose that X, Y, and Z are Banach spaces.
Take A E B(Y, X) and B E B(Z, X). Suppose further that ker A is a comple-
mented subspace ofY. The diagram
A
X---Y
Z
is commutative for some ~ in B(Z, Y) if and only if im A :::> im B.
<l Once again only -<= needs examining. Using the definition of complemented
subspace, express Y as the direct sum of ker A and Yo, where Yo is a closed
subspace. By 5.5.9 (1), Yo is a Banach space. Consider the part AD of A in Yo.
Undoubtedly, im AD = im A :::> im B. Consequently, by 2.3.13 and 2.3.14 the
equation Ao!tO = B has a unique solution !to := A;;-l. It suffices to prove that
the operator !to, treated as a member of £(Z, Yo), is bounded.
The operator !to is closed. Indeed (cf. 7.4.8), if Zn -+ Z and A;;-l BZn -+ Y
then BZn -+ Bz, since B is bounded. In addition, by the continuity of AD, the
correspondence A;;-l C X X Yo is closed; and so 7.3.10 yields the equality y =
A;;-l Bz. I>
7.4. Open Mapping and Closed Graph Theorems 111
~ 1~
(X,II·lId
By the Phillips Theorem some continuous operator ~ makes the diagram com-
mutative. Such an operator is unique: it is Ix. [>
7.4.17. Graph Norm Principle. Let X and Y be Banach spaces and let T
in 2'(X, Y) be a closed operator. Given x E X, define the graph norm of x as
IIxllgr T:= IIxlix + IITxlly· Then II . IIgr T '" II . IIx.
<l Observe that the space (X, 1I·lIgr T) is complete. Further, 1I·lIgr T 2: 1I·lIx.
It remains to refer to the Two Norm Principle. [>
¢=:Let Z be a some Banach range, i.e. Z = T(Y) for some Banach space Y
and T E B(Y, Z). Passing, if need be, to the monoquotient of T, we may assume
that T is an isomorphism. Put IIzllo:= IIT-lzlly. It is clear that (Z, 11·110) is
a Banach space and IIzll = IITT- I zll :s; IITIlIIT- l zll = IITllllzllo; i.e., 11·110 r II· liz.
Applying this construction to Xl and X 2 , obtain Banach spaces (Xl, II· lid and
(X2' 11·112). Now 1I·lIk r 1I·lIx on Xk for k:= 1,2.
Given Xl E Xl and X2 E X 2, put IIXI + x2110:= IIxllll + IIx2112' Thereby we
introduce in X some norm II . II that is stronger than the initial norm II . IIx.
By construction (X, II ·110) is a Banach space. It remains to refer to 7.4.16. [>
112 Chapter 7. Principles of Banach Spaces
Ox(J 0 T) = OTx(J) 0 T.
<I The right side of the sought formula is included into its left side for obvious
reasons. If I in X, belongs to ox(J 0 T), then by the Hahn-Banach Theorem there
is an element II of the algebraic sub differential of f at Tx for which 1 = II 0 T.
It suffices to observe that, in virtue of 7.5.7, II is an element of Y' and so it is
a member of the topological sub differential OTx(J). t>
7.5.10. Balanced Hahn-Banach Theorem. Suppose that X and Yare
normed spaces. Given T E B(X, Y), let p : Y -+ JR. be a continuous seminorm.
Then
lol(p 0 T) = 181(p) 0 T.
<I If I E 181(p 0 T) then I = II 0 T for some it in the algebraic balanced
sub differential of p (cf. 3.7.11). From 7.5.2 it follows that it is continuous. Thus,
181(p 0 T) c 181(p) 0 T. The reverse inclusion raises no doubts. t>
7.5.11. Continuous Extension Principle. Let Xo be a subspace of X and
let 10 be a continuous linear functional on Xo. Then there is a continuous linear
functional 1 on X extending 10 and such that 11111 = 111011.
<I Take p:= 11101111 . II, and consider the identical embedding t : Xo -+ X.
On account of 7.5.10, 10 E 181(p 0 t) = 181(p) 0 t = 111011181(11' II) 0 to It suffices to
observe that 181(11'lIx) = Bx'. t>
114 Chapter 7. Principles of Banach Spaces
<I It is clear that U E X' & ker f J Xo) =} ker f J cl Xo. If Xo 1:. cl Xo
then there is an open convex neighborhood about Xo disjoint from cl Xo. In virtue
of 7.5.12 and 7.5.13 there is a functional fo, a member of (XIR)' , such that ker fo J
cl Xo and fo(xo) = 1. From the properties of the complexifier infer that the
functionallRe- 1 fo vanishes on Xo and differs from zero at the point Xo. It is also
beyond a doubt that the functional is continuous. e>
7.6.4. EXAMPLES.
(1) Let X and Y be Hilbert spaces. Take T E B(X, Y). Observe first
that, in a plain sense, T E B(X, Y) {:} T E B(X*, Y*). Denote the prime mapping
of X by (.)~ : X* -+ X', i.e., x 1--+ x' := (., x); and denote the prime mapping
of Y by (-)y : Y* -+ Y', i.e., y 1--+ y':= (., y).
7.6. Prime Principles 115
The adjoint of T, the member T* of B(Y, X), and the dual of T, the mem-
ber T' of B(Y', X'), are related by the commutative diagram:
T·
X*+--Y*
O~! ! Oy
X'~Y'
<I Indeed, it is necessary to show the equality T'y' = (T*y)' for y E Y. Given
x EX, by definition observe that
X~Y X'~Y'
R". /s R',,- /s'
Z Z'
are commutative or not simultaneously. Since R = ST ::::} R' = (ST)' = T' S';
therefore, the commutativity of the triangle on the left entails the commutativity
of the triangle on the right. If the latter commutes then by what was already
proven R" = S"T". Using 7.6.6, argue as follows: (Rx)" = R"x" = S"T"x" =
S"(T"x") = S"(Tx)" = (STx)" for all x E X. Consequently, R = ST. t>
7.6.8. DEFINITION. Let Xo be a subspace of X and let .;to be a subspace
of X'. Put
,
X k-l T~ X' T~+l ,
... +- f--- k f--- X k +1 +- '" .
<l =?: Since im Tk+l = ker Tk+2; therefore, Tk+l is normally solvable. Using
the Polar Lemma, conclude that
118 Chapter 7. Principles of Banacb Spaces
=}: By the Hausdorff Theorem Tk+l is normally solvable. Once again appeal-
ing to 7.6.11 (2), infer that
Here account was taken of the fact that ker Tk+l itself is a closed subspace. I>
o -+ ker T -+ X
T
----+ Y -+ coker T -+ 0
is exact. I>
Exercises
7.1. Find out which linear operators are ideal.
7.2. Establish that a separately continuous bilinear form on a Banach space is jointly
continuous.
7.3. Is a family of lower semicontinuous sublinear functionals on a Banach space uni-
formly bounded on the unit ball?
7.4. Let X and Y be Banach spaces and T : X ..... Y. Prove that IITxlly ~ tllxllx for
some strictly positive t and all x E X if and only if ker T =
0 and im T is a complete set.
7.5. Find conditions for normal solvability of the operator of multiplication by a function
in the space of continuous functions on a compact set.
7.6. Let T be a bounded epimorphism of a Banach space X onto h(C). Show that
ker T is complemented.
7.7. Establish that a uniformly closed subspace of C([a, bJ) composed of continuously
differentiable functions (i.e., elements of C(l)([a, b])) is finite-dimensional.
7 .S. Let X and Y be different Banach spaces, with X continuously embedded into Y.
Establish that X is a meager subset of Y.
Exercises 119
is strictly positive.
7.10. Let (amn) be a double countable sequence such that there is a sequence (z(m») of
elements of 11 for which all the series 2::=1 amnZ~m) fail to converge in norm. Prove that there
is a sequence z in It such that the series 2::=1 amnz n fail to converge in norm for all mEN.
7.11. Let T be an endomorphism of a Hilbert space H which satisfies (Tz I y) = (z I Ty)
for all z, y E H. Establish that T is bounded.
7.12. Let a closed cone X+ in a Banach space X be reprodu.cing: X = X+ - X+. Prove
that there is a constant t > 0 such that for all z E X and each presentation z = Zl - Z2 with
Zl, Z2 E X+, the estimates hold: IIz111 $ tllzll and IIz211 $ tllzll.
7.13. Let lower semicontinuous sublinear functionals p and q on a Banach space X be
such that the cones dom p and dom q are closed and the subspace dom p - dom q = dom q - dom p
is complemented in X. Prove that
i.e., 8(y~x) E B(X, Y)'. Furthermore, given T E B(X, Y), infer that
Therefore, AB is a norming set for B(X, Y). If Ay is a fully norming set then AB
is also a fully norming set. Indeed, if U is such that the numeric set {ly'(Tx)1 :
8.1. Holomorphic Functions and Contour Integrals 121
. loJ(z)-loJ(zo)
11m ;
Z--+Zo Z - Zo
i.e., the function 10 J : ~ -+ C is holomorphic for 1 E A.
<l (1) =} (2): This is obvious.
(2) =} (1): For the sake of simplicity assume that Zo = 0 and J( zo) = O.
Consider the disk of radius 2c: centered at zero and included in ~; i.e., 2c:1D> C ~,
where ID>:= Be:= {z E C: Izi ::; I} is the unit disk. As is customary in complex
analysis, treat the disk c:1D> as an (oriented) compact manifold with boundary c:'Ir,
where 'Iris the (properly oriented) unit circle 'Ir:= {z E C: Izi = I} in the complex
plane CIR. Take Zl, Z2 E c:1D> \ 0 and the holomorphic function 10 J (the functionall
lies in A). Specifying the Cauchy Integral Formula, observe that
10 J(Zk)
Zk
=~
21l"Z
j 10 J(z) dz
Z(Z-Zk)
(k:= 1, 2).
2<:11"
If now Zl i- Z2 then, using the condition Iz - zkl ;::: c: (k:= 1, 2) for z E 2c:'Ir and
the continuity property of the function 10 J on ~, find
_1_. _1
Zl - Z2 21l"i
j 10 J(z) ( 1
z(z - zt)
_ 1
z(z - Z2)
) dz
2<:11"
This final inequality ensures that the sought limit exists. I>
8.1.4. DEFINITION. A mapping J: ~ -+ X satisfying 8.1.3 (1) (or, which is
the same, 8.1.3 (2) for some fully norming set A) is called holomorphic.
8.1.5. REMARK. A meticulous terminology is used sometimes. Namely, if J
satisfies 8.1.3 (1) then J is called a strongly holomorphic Junction. If J satisfies
8.1.3 (2) with A:= Bx' then J is called weakly holomorphic. Under the hypotheses
of 8.1.3 (2) and 8.1.2 (4), i.e. for J : ~ -+ B(X, Y), Ay:= BY' and the corre-
sponding A:= AB, the expression, "J is weakly operator holomorphic," is employed.
With regard to this terminology, the Dunford-Hille Theorem is often referred to
as the Holomorphy Theorem and verbalized as follows: "A weakly holomorphic
function is strongly holomorphic."
8.1.6. REMARK. It is convenient in the sequel to use the integrals of the
simplest smooth X-valued forms J(z)dz over the simplest oriented manifolds, the
boundaries of elementary planar compacta (cf. 4.8.5) which are composed of finitely
many disjoint simple loops. An obvious meaning is ascribed to the integrals:
Namely, given a loop 'Y, choose an appropriate (smooth) parametrization 'l1 : '[' -+ 'Y
(with orientation accounted for) and put
J
"Y
J(z)dz:= J
'll'
J 0 'l1d'l1,
with the integral treated for instance as a suitable Bochner integral (cf. 5.5.9
(6)). The soundness of the definition is beyond a doubt, since the needed Bochner
integral exists independently of the choice of the parametrization 'l1.
8.1.7. Cauchy-Wiener Integral Theorem. Let ~ be an nonempty open
subset of tbe complex plane and let J : ~ -+ X be a bolomorpbic X-valued
function, witb X a Banacb space. Assume furtber tbat F is a rougb draft for tbe
J
pair (0, ~). Tben
J(z)dz = O.
8F
J
Moreover,
J(zo) = ~ J(z) dz
271'z Z - Zo
8F
for Zo E int F.
8.1. Holomorphic Functions and Contour Integrals 123
<l By virtue of 8.1.3 the Bochner integrals exist. The sought equalities follow
readily from the validity of their scalar versions basing on the Cauchy Integral
Formula, the message of 8.1.2 (1) and the fact that the Bochner integral commutes
with every bounded functional as mentioned in 5.5.9 (6). t>
8.1.8. REMARK. The Cauchy-Wiener Integral Theorem enables us to infer
analogs of the theorems of classical complex analysis for X -valued holomorphic
functions on following the familiar patterns.
8.1.9. Taylor Series Expansion Theorem. Let J : ~ ~ X be a holo-
morphic X -valued function, with X a Banach space, and take Zo E ~. In every
open disk U:= {z E C: Iz - Zo 1< E} such that cl U lies in ~, the Taylor series
expansion holds (in a compactly convergent power series, cf. 7.2.10):
00
J(z) = ~ J J(u) du (z E cl U)
27rZ U - z
au'
1 1 ~ (z - zo)n
u-z (u-z o )(l- ~=~~) - n=o(u-zo)n+l·
The last series converges uniformly in u E aU'. (Here U' = U + qD for some
q > 0 such that cl U' c ~.) Taking it into account that sup IIJ(aU')1I < +00
and integrating, arrive at the sought presentation of J(z) for z E cl U. Applying
this to U' and using 8.1. 7, observe that the power series under study converges
in norm at every point of U'. This yields uniform convergence on every compact
subset of U', and so on U. t>
8.1.10. Liouville Theorem. If an X-valued function J :C ~ X, with X
a Banach space, is holomorphic and sup IIJ(C)II < +00 then J is a constant map-
ping.
124 Chapter 8. Operators in Banach Spaces
<I Considering the disk e;ll) with e; > 0 and taking note of 8.1.9, infer that
l/cnll :::; sup II/(z)II' e;-n :::; sup II/(C)II' e;-n
zEe1I'
for all T E B(X), with X a complex Banach space; i.e., the spectral radius of
an operator T coincides with the radius of the spectrum ofT.
<I It is an easy matter that the spectral radius r(T) is greater than the radius
of the spectrum of T. So, there is nothing to prove if r(T) = O. Assume now that
r(T) > O. Take A E C so that IAI > sup{IILI: IL E Sp(T)}. Then the disk of radius
IAI- 1 lies entirely in the domain of the holomorphic function (cf. 5.6.15)
R(T, z-1), for z i= 0 and z-1 E res (T)
I(z):= {
0, for z = o.
Using 8.1.9 and 5.6.17, conclude that IAI- 1 < r(T)-1. Consequently, IAI > r(T). I>
8.1.13. Let K be a nonempty compact subset ofC. Denote by H(K) the set
of all functions holomorphic in a neighborhood of K (i.e., I E H(K) {::} f :
dom f ---t Cis a holomorphic function with dom I:J K). Given ft, h E H(K), let
the notation ft '" h mean that it is possible to find an open subset ~ of dom 11 n
dom h satisfying K C ~ and ft I~ = h I~· Then'" is an equivalence in H(K). <II>
8.1.14. DEFINITION. Under the hypotheses of 8.1.13, put £,(K):= H(K)/",.
The element 7 in £,(K) containing a function I in H(K) is the germ of Ion K.
8.1.15. Let 7, 9 E £,(K). Take ft, h E 7 and g1, g2 E g. Put
<I Choose open sets ~1 and ~2 such that K C ~1 C dom h n dom 12 and
K C ~2 C dom gl ndom g2, with hlgo)l = 12lgo)l and gIigo)2 = g21go)2· Observe now
that 'PI and 'P2 agree on ~1 n ~2. t>
8.1.16. DEFINITION. The coset, introduced in 8.1.15, is the sum of the germs
11 and 12. It is denoted by 11 + 12. The product of germs and multiplication of
a germ by a complex number are introduced by analogy.
8.1.17. The set £(l<) with operations defined in 8.1.16 is an algebra. <II>
J h(z)R(z)dz = J 12(z)R(z)dz.
8F1 8F2
<I Let K C ~ C int Fl n int F2, with '?J open and h Igo) = 12 19o). Choose
a rough draft F for the pair (K, D). Since hR is holomorphic on dom h \ l< and
12R is holomorphic on dom 12 \ K, infer the equalities
J h(z)R(z)dz = J h(z)R(z)dz,
J J
8F 8Fl
12(z)R(z)dz = 12(z)R(z)dz
8F 8F2
(from the nontrivial fact of the validity of their scalar analogs). Since hand 12
agree on '?J, the proof is complete. t>
8.1.20. DEFINITION. Under the hypotheses of 8.1.19, given an element h
in £(K), define the contour integral of h with kernel R as the element
f h(z)R(z)dz:= J
8F
JCz)R(z)dz,
where h = 1 and F is a rough draft for the pair (K, dom f).
126 Chapter 8. Operators in Banach Spaces
8.1.21. REMARK. The notation h(z) in 8.1.20 is far from being ad hoc. It is
well justified by the fact that w: = It (z) = h (z) for every point z in K and every
two members It and h of a germ h. In this connection the element w is said to
be the value of h at z, which is expressed in writing as h(z) = w. It is also worth
noting that the function R in 8.1.2 may be assumed to be given only in U \ K,
where int U ::) K.
f%Th:= ~
27l'z
f h(z)R(T, z)dz
and called the Riesz-Dunford integral (of the germ h). If f is a function holomor-
phic in a neighborhood about Sp(T) then put f(T):= f%T f:= f%T f. We also use
more suggestive designations like
f(T) = ~
27l'Z
f fez) dz.
z- T
JJ
8F2 8F1
JJ/( )/ (
8F1 8F2
2 .~
-- _1 2 . 1 Zl 2 Z2 )R(T, zt}-R(T, z2)dZ2 dZl
11"Z 11"Z Z2 - Zl
8F1 8F2
1J (1 J h
--2'
11"Z
h(Z2) -2'
11"Z
(Zl)
Z2 - Zl
dZl ) R(T, z2)dz2
= 2~i J
8F2 8Fl
Choose a circle 'Y : = c'll" that lies in res (T) as well as in the (open) disk of
128 Chapter 8. Operators in Banach Spaces
convergence of the series J(z) = L:~=o cnz n . From 5.6.16 and 5.5.9 (6) derive
=
1 ~ JJ( ) -n-ITnd
27l"i ~""Y z Z Z =~
~ ( 27l"1 i J
""y
J(z) d ) Tn
zn+t Z
~ Tn
= ~ Cn
J
for Zl E 8F1 . Therefore, by 8.2.3
1 R(T, Z2)
R(g(T), zt} = -2.
7rt
( ) dZ 2
ZI - g Z2
(Zl E 8Ft).
8F2
1
~g(T) J = -2 .
7rt
J J(zt}(T) dZ1 = -2
Zl - g
1 1
. -2'
7rt 7rt
J (J J( zt} R(T, Z2)
ZI -
)
g ( Z2 ) dZ2 dZ1
8F1 8F1 8F2
Since g(Z2) E int Fl for Z2 E 8F2 by construction; therefore, the Cauchy Integral
Formula yields the equality
Consequently,
0)
matrix form
T (TO'
rv
o TO', ,
with the part TO' of T in Xu and the part TO', of T in Xu' satisfying
<l Since Xu + xu' = XSp(T) = Ie, in view of 8.2.3 and 8.2.10 it suffices
to establish the claim about the spectrum of TO'.
From 8.2.5 and 8.2.3 obtain
PtTT rv (~ ~).
Let A be a nonzero complex number. Then
A_ P T rv (A - TtT 0)
0' 0 A j
8.2. Tbe Holomorpbic Functional Calculus 131
i.e., the operator ,\ - PuT is not invertible if and only if the same is true of the
operator ,\ - Tu. Thus,
Sp(Tu) \ 0 c Sp(P.,.T) \ 0 = (a U 0) \ 0 c a.
Suppose that 0 E Sp(Tu) and 0 ~ a. Choose disjoint open sets (gu and (gul
so that a C (gu, 0 ~ (gu and a' C (gul, and put
1
z E (gu => h(z):= -;
z
z E (gul => h(z):= O.
Here xua):= xu! and xula):= X.,.I! for! E H(Sp(T» are tbe representations
induced by restricting! onto a and a'. <ll>
132 Chapter 8. Operators in Banach Spaces
8.3.7. The linear span of the set of (bounded) rank-one operators comprises
all finite-rank operators:
T E F(X, Y)
n
<=> (3 x~, ... ,x~ E X') (3 Yl, ... ,Yn E Y) T = I>~ ® Yk. <11>
k=l
8.3.12. REMARK. For a long time there was an unwavering (and yet un-
provable) belief that every Banach space possesses the approximation property.
Therefore, P. Enflo's rather sophisticated example of a Banach space lacking the
approximation property was acclaimed as sensational in the late seventies of the
current century. Now similar counterexamples are in plenty:
8.3.13. SZANKOWSKI COUNTEREXAMPLE. The space B(lz) lacks the approx-
imation property.
8.3.14. DAVIS-FIGIEL-SZANKOWSKI COUNTEREXAMPLES. The spaces Co
and lp with p =I- 2 have closed subspaces lacking the approximation property.
II x o - x,,11 = IIxo - =
II: ::1111
= Ilx' ~ xll ll(lI x - x'llxo + x') - xii ~ ~~,~oli ~ 1 - e. C>
8.4.2. Riesz Criterion. Let X be a Banach space. The identity operator
in X is compact if and only if X is finite-dimensional.
<l Only the implication =} needs proving. If X fails to be finite-dimensional,
then select a sequence of finite-dimensional subspaces Xl C X 2 C ... in X such
that X n+ I =I- Xn for all n E N. In virtue of 8.4.1 there is a sequence (x n ) satisfying
Xn+I E Xn+I, IIxn+III = 1 and d(Xn+I, Xn) ~ 112, namely, some sequence of 112-
perpendiculars to Xn in X n+I . It is clear that d(xm, Xk) ~ 112 for m =I- k. In other
words, the sequence (x n ) lacks Cauchy subsequences. Consequently, by 8.3.1 the
operator Ix is not compact. c>
8.4.3. Let T E X(X, Y), with X and Y Banach spaces. The operator Tis
normally solvable if and only if T has finite rank.
<l Only the implication =} needs examining.
Let Yo := im T be a closed subspace in Y. By the Banach Homomorphism
Theorem, the image T(Bx) of the unit ball of X is a neighborhood of zero in Yo.
Furthermore, in virtue of the compactness property of T, the set T(Bx) is rela-
tively compact in Yo. It remains to apply 8.4.2 to Yo. c>
8.4. The Riesz-Schauder Theory 135
8.4.6. Schauder Theorem. Let X and Y be Banach spaces (over the same
ground field IF). Then
<l =}: Observe first of all that the restriction mapping x' f--t X'IBx implements
an isometry of X' into loo(Bx). Therefore, to check that ]{'(Byt) is relatively
compact we are to show the same for the set V:= {]{'y'IBx : y' E By,}. Since
136 Chapter 8. Operators in Banach Spaces
K'y'IBX(X) = y' 0 KIBx(X) = y'(Kx) for x E Bx and y' E By', consider the
o
compact set Q:= cl K(Bx) and the mapping K : C(Q, IF) -+ loo(Bx) defined
o 0
a(T):= nul T:= dim ker T; (3(T):= def T:= dim coker T.
The integer ind T := aCT) - (3(T), a member of Z, is the index or, fully, the
Fredholm index of T.
8.5.2. REMARK. In the Russian literature, a Fredholm operator is usually
called a Noether operator, whereas the term "Fredholm operator" is applied only
to an index-zero Fredholm operator.
8.5.3. Every Fredholm operator is normally solvable.
<J Immediate from the Kato Criterion. t>
<I By virtue of 2.3.5 (6), 8.5.3, 5.5.4 and the Sequence Prime Principle, the
next pairs of sequences are exact simultaneously:
+L
n
, L...t
y , =Yo+ ""'
k=l
n
I-'kYk'
<I This is a reformulation of 8.5.5 with account taken of the Polar Lemma. t>
8.5. Fredholm Operators 139
8.5.7. EXAMPLES.
(1) If T is invertible then T is an index-zero Fredholm operator.
(2) Let T E ~(IF n, IF m). Let rank T:= dim im T be the rank of T.
Then aCT) = n-rank T and J3(T) = m-rank T. Consequently, T E § r(IF n, IF m)
and ind T = n - m.
(3) Let T E B(X) and X = Xl EB X 2 • Assume that this direct sum
decomposition of X reduces T to matrix form
T~ (~l ~2).
Undoubtedly, T is a Fredholm operator if and only if its parts are Fredholm
operators. Moreover, aCT) = a(Tt) + a(Tz) and J3(T) = J3(Tt) + J3(Tz); i.e.,
ind T = ind TI + ind T z . <ll>
8.5.8. Fredholm Theorem. Let K E X(X). Then 1 - K is an index-zero
Fredholm operator.
<l First, settle the case IF := C. If 1 rt. Sp(K) then 1 - K is invertible and
ind(l - K) = o. If 1 E Sp(K) then in virtue of the Riesz-Schauder Theorem
and the Spectral Decomposition Theorem there is a decomposition X = Xl EB Xz
such that Xl is finite-dimensional and 1 rt. Sp(Kz), with Kz the part of Kin X z .
Furthermore,
1- K ~ (1 -0KI 1 _0 K z ) .
k=l
<l For n = 1 the exactness of the sequence 0 -+ Xl -+ 0 means that Xl = 0,
and for n = 2 the exactness of 0 -+ Xl -+ X 2 -+ 0 amounts to isomorphy between
Xl and X 2 (cf. 2.3.5 (4)). Therefore, the Euler identity is beyond a doubt for
n:= 1,2.
Suppose now that for m ::; n - 1, where n > 2, the desired identity is already
established. The exact sequence
o -+ Xl X 2 -+ ...
-+ -+ X n- 2
Tn
---+
-2 X n-l Tn _ 1
----+
X n-+ 0
reduces to the exact sequence
k
o -+ Xl -+ X2 -+ ... -+ Xn- 2 ----t
Tn _ 2 'T'
er .Ln-l -+ O.
By hypothesis,
n-2
L:( _l)k dimXk + (-It- 1 dimker Tn- l = O.
k=l
Furthermore, since Tn - l is an epimorphism,
dimXn- 1 = dimker Tn- l + dimXn .
Finally,
= [>
k=l
8.5.18. Atkinson Theorem. The index of the product of Fredholm opera-
tors equals the sum of the indices of the factors.
<l Let T E $r(X, Y) and S E $r(Y, Z). By virtue of 8.5.16, ST E
$r(X, Z). Using the Snowflake Lemma, obtain the exact sequence of finite-
dimensional spaces
o -+ ker T -+ ker ST -+ ker S -+ coker T -+ coker ST -+ coker S -+ o.
Applying 8.5.17, infer that
aCT) - neST) + o(S) - j3(T) + j3(ST) - j3(S) = 0;
whence ind (ST) = ind S + ind T. [>
142 Chapter 8. Operators in Banach Spaces
ST = 1 + Kx; TS = 1+Ky
with some Kx E X(X) and Ky E X(Y) (that S exists is ensured by 8.5.14).
It is clear that
SeT + K) = ST + SK = 1 + Kx + SK E 1 + X(X);
(T + K)S = TS + KS = 1 + Ky + KS E 1 + X(Y);
i.e., S is an approximate inverse of T+K. By virtue of 8.5.14, T+K E $r(X, Y).
Finally, from 8.5.19 infer the equalities ind (T + K) = - ind S and ind T =
-ind S. I>
8.5.21. Bounded Index Stability Theorem. The property of being
a Fredholm operator and the index of a Fredholm operator are preserved under
sufficiently small bounded perturbations: the set $r(X, Y) is open in the space
of bounded operators, and the index of a Fredholm operator ind : $ r( X, Y) - t Z
is a continuous function.
<l Let T E $r(X, Y). By 8.5.14 there are operators S E B(Y, X), Kx E
X(X) and Ky E X(Y) such that
ST=l+Kx; TS=l+Ky.
= -ind«1 + SV)-lS)
ind(T + V)
= -ind(1 + SV)-l - ind S = -ind S = ind T
(because (1 + SV)-l is a Fredholm operator by 8.5.7 (1». This means that the
Fredholm index is continuous. t>
8.5.22. Nikol'skiY Criterion. An operator is an index-zero Fredholm oper-
ator if and only if it is the sum of an invertible operator and a compact operator.
<I ~: Let T E $r(X, Y) and ind T = O. Consider the direct sum decom-
positions X = Xl $ ker T and Y = im T $ Yi. It is beyond a doubt that the
operator Tl , the restriction of T to Xl, implements an isomorphism between Xl
and im T. Furthermore, in virtue of 8.5.5, dim Yl = f3(T) = aCT); i.e., there is
a natural isomorphism Id : ker T --4 Yl . Therefore, T admits the matrix presen-
tation
T ~ (~l ~) = (~ I~) + (~ _~d ) .
¢:::: If T:= S + K with K E £(X, Y) and S-l E B(Y, X) then, by 8.5.20
and 8.5.7 (1), ind T = ind(S + K) = ind S = O. t>
8.5.23. REMARK. Let Inv (X, Y) stand as before for the set of all invertible
operators from X to Y (this set is open by Theorem 5.6.12). Denote by §(X, Z)
the set of all Fredholm operators acting from X to Y and having index zero. The
Nikol'skil Criterion may now be written down as
Exercises
8.1. Study the Riesz-Dunford integral in finite-dimensions.
8.2. Describe the kernel of the Riesz-Dunford integral.
8.3. Given n E N, let In be a function holomorphic in a neighborhood U about the
spectrum of an operator T. Prove that the uniform convergence of (In) to zero on U follows
from the convergence of (In (T)) to zero in operator norm.
8.4. Let rT be an isolated part of the spectrum of an operator T. Assume that the
part rT' := Sp(T) \ rT is separated from rT by some open disk with center a and radius r so that
rT C {z E IC: Iz - al < r}. Considering the Riesz projection P", prove that
introduce the concept of ascent and the denotation a(T). Demonstrate that, for an operator T
with finite descent and finite ascent, the two quantities, a(T) and d(T), coincide.
8.11. An operator T is a Riesz-Schauder operator, if T is a Fredholm operator and has
finite descent and finite ascent. Prove that an operator T is a Riesz-Schauder operator if and
only if T is of the form T = U + V, where U is invertible and V is of finite rank (or compact)
and commutes with U.
Exercises 145
8.12. Let T be a bounded endomorphism of a Banach space X which has finite descent
and finite ascent, r := aCT) = d(T). Prove that the subspaces im (Tr) and ker (Tr) are closed,
the decomposition X = ker (Tr) EB im (Tr) reduces T, and the restriction of T to im (Tr) IS
invertible.
8.13. Let T be a normally solvable operator. If either of the next quantities is finite
Prove that
8.14. Let T be a bounded endomorphism. Prove that T belongs to <1>+ (X) if and only
if for every bounded but not totally bounded set U, the image T(U) is not a totally bounded set
in X.
8.15. A bounded endomorphism T in a Banach space is a Riesz operator, if for every
nonzero complex>. the operator (>. - T) is a Fredholm operator. Prove that T is a Riesz operator
if and only if for all >. E C, >. I' 0, the following conditions are fulfilled:
(1) the operator (>. - T) has finite descent and finite ascent;
(2) the kernel of (>. - T)k is finite-dimensional for every kEN;
(3) the range of (>. - T)k has finite deficiency for kEN,
and, moreover, all nonzero points of the spectrum of T are eigenvalues, with zero serving as the
only admissible limit point (that is, for whatever strictly positive e, there are only finitely many
points of Sp(T) beyond the disk centered at zero with radius e).
8.16. Establish the isometric isomorphisms: (X/Y)' ~ yl. and X'/Yl. ~ Y' for Ba-
nach spaces X and Y such that Y is embedded into X.
8.17. Prove that for a normal operator T in a Hilbert space and a holomorphic func-
tion I, a member of H(Sp(T)), the operator I(T) is normal. (An operator is normal if it
commutes with its adjoint, cf. 11.7.1.)
8.18. Show that a continuous endomorphism T of a Hilbert space is a Riesz operator
if and only if T is the sum of a compact operator and a quasinilpotent operator. (Quasinilpotency
of an operator means triviality of its spectral radius.)
8.19. Given two Fredholm operators Sand T, members of $"r(X, Y), with ind S =
ind T, demonstrate that there is a Jordan arc joining Sand T within $"r(X, Y).
Chapter 9
An Excursion into General Topology
U E t( T )( x) {:} (3 V E Op (T » x E V & U :J V.
for x E X.
<l Straightforward from 9.1.9 (5). I>
9.1.11. Let 1'1, 1'2 E T(X). The following statements are equivalent:
(1) 1'1 21'2;
(2) Op(Tl):J Op(T2);
(3) Cl (1'1) :J Cl ( 1'2). <ll>
148 Chapter 9. An Excursion into General Topology
9.1.12. REMARK. As follows from 9.1.8 (5) and 9.1.11, the topology of a space
is uniquely determined from the collection of its open sets. Therefore, the set
Op (X) itself is legitimately called the topology of the space X. In particular,
the collection of open sets of a pretopological space (X, T) makes X into the
topological space (X, t( T» with the same open sets in stock. Therefore, given
a pretopology T, the topology t( T) is usually called the topology associated with T.
9.1.13. Theorem. The set T (X) of all topologies on X with the relation
"to be stronger" presents a complete lattice. Moreover, for every subset C of T (X)
the equality holds:
sUPT (X) C = sup 3'(X) C.
<I Evidently, t(SUP3'(X) C) ~ sUP3'(X) t(C) ~ sUP3'(X) C ~ t(SUP3'(X) C).
Thus, T := sUP3'(X) C belongs to T (X). It is clear that T ~ C. Furthermore,
if TO ~ C and TO E T (X) then TO ~ T and so T = SUPT (X) C. It remains to refer
to 1.2.14. I>
9.1.14. REMARK. The explicit formula for the greatest lower bound of C is
more involved:
infT (X) C = t(inf 3'(X) C).
However, the matter becomes simpler when the topologies are given by means of
their collections of open sets in accordance with 9.1.12. Namely,
n
In other words,
Op (infT (X) C) = 0p(T).
rEtC
9.2. Continuity
9.2.1. REMARK. The presence of a topology on a set obviously makes it pos-
sible to deal with such things as the interior and closure of a subset, convergence
of filters and nets, etc. We have already made use of this circumstance while in-
troducing multinormed spaces. Observe for the sake of completeness that in every
topological space the following analogs of 4.1.19 and 4.2.1 are valid:
9.2.2. BirkhoffTheorem. For a nonempty subset U and a point x ofa topo-
logical space the following statements are equivalent:
(1) x is an adherent point ofU;
(2) there is some filter containing U and converging to x;
(3) there is a net of elements of U which converges to x. <II>
9.2. Continuity 149
(X,T) ~ (Y,w)
Ix i !!y
(X,TI) ~ (Y,Wl)
9.2.16. REMARK. The messages of 9.2.14 and 9.2.15 are often referred to as
the theorems on topologizing by a family of mappings.
9.2.17. EXAMPLES.
(1) Let (X, T) be a topological space and let Xo be a subset of X.
Denote the identical embedding of Xo into X by l : Xo -+ X. Put To:= [-l(T).
The topology TO is the induced topology (by T in Xo), or the relative or subspace
topology; and the space (X 0, TO) is a subspace of (X, T).
(2) Let (X{, T{){E3 be a family of topological spaces and let X :=
I1{E3 X{ be the product of (X{){E3. Put T:= sUP{E3 Pre1( T{), where Pr{ : X -+
X{ is the coordinate projection (onto X{); i.e., Pr{ x = x{ (~ E 3). The topology T
is the product topology or the product of the topologies (T{ ){E3, or the Tychonoff
topology of X. The space (X, T) is the Tychonoff product of the topological spaces
under study. In particular, if X{:= [0, 1] for all ~ E 3 then X:= [0, 1]3 (with the
Tychonoff topology) is a Tychonoff cube. When 3:= N, the term "Hilbert cube"
is applied.
9.3.2. DEFINITION. A topological space satisfying one (and hence all) of the
equivalent conditions 9.3.1 (1)-9.3.1 (3), is called a separated space or a T 1-space.
The topology of a Trspace is called a separated topology or (rarely) a T1-topology.
9.3.3. REMARK. By way of expressiveness, one often says: "A T 1-space is
a space with closed points."
152 Chapter 9. An Excursion into General Topology
{f < t} C Ut C {f ::; t} (t E T)
if and only if
(t, sET & t < s) =} cl Ut C int Us.
<l =}: Take t < s. Since {f ::; t} is closed and {f < s} is open, the inclusions
hold:
cl Ut C {f ::; t} C {f < s} C int Us.
W t = int Ut C Ut C Us ::} I ~ hj
lit = cl Ut = Ws ::} h ~ gj
C int Us
Ut C Us C cl Us = Vs ::} 9 ~ f.
So, 1= 9 = h. Taking account of 3.8.4 and 9.1.5 and given t E iR, find
°
that F is a closed set in X and G is a neighborhood ofF. Then there is a continuous
function I : X ---t [0, 1] such that I(x) = for x E F and I(x) = 1 for x rf. G.
<I Put Ut := 0 for t < °and Ut := X for t > 1. Consider the set T of the
dyadic-rational points ofthe interval [0, l]j i.e., T:= UnENTn with Tn:= {k2- n+1 :
k:= 0, 1, ... ,2 n - 1 }. It suffices to define Ut for all tin T so that the family t 1--+ Ut
(t E T:= T U (iR \ [0, 1])) satisfy the criterion of 9.3.12. This is done by way
of induction.
If t E T17 i.e., t E {O, l}j then put Uo := F and U1 := G. Assume now
that, for t E Tn and n ~ 1, some set Ut has already been constructed, satisfying
cl U, C int Us whenever t, s E Tn and t < s. Take t E T n+1 and find the two
points t, and tr in Tn nearest to t:
It remains to show that the resulting family satisfies the criterion of 9.3.12.
To this end, take t, s E Tn+! with t < s. If tr = Sl, then for s > Sl
by construction
9.4. Compactness 155
9.4. Compactness
9.4.1. Let ~ be a Elterbase on a topological space and let
cl~:=n{clB: BE~}
be tbe set of adberent points of ~ (also called tbe adberence of ~). Tben
(1) cl ~ = cl fil ~;
(2) ~ ~ x =} x E cl ~;
(3) (~is an ultraElter and x E cl ~) =} ~ - t x.
<l Only (3) needs demonstrating, since (1) and (2) are evident. Given U E r( x)
and B E ~, observe that Un B =I- 0. In other words, the join § := rex) V
~ is available. It is clear that § - t x. Furthermore, § = ~, because ~ is
an ultrafilter. I>
9.4.2. DEFINITION. A subset C of a topological space X is a compact set
in X if each open cover of C has a finite sub cover (d. 4.4.1).
156 Chapter 9. An Excursion into General Topology
fJO:= { n
GE'£o
X \ G: tffo is a finite subset of tff} .
L:~=I hk(X) = 1 for every point x in some neighborhood about F. (As usual,
G~:= X \ Gd
9.4.20. DEFINITION. A topology is called locally compact if each point pos-
sesses a compact neighborhood. A locally compact space is a set furnished with
a Hausdorff locally compact topology.
9.4.21. A topological space is locally compact if and only if it is bomeomor-
pbic witb a punctured compactum (= a compactum witb a deleted point), i.e. tbe
complement of a singleton to a compactum.
<I {=: In virtue of the Weierstrass Theorem it suffices to observe that each
point of a punctured compactum possesses a closed neighborhood (since every
compactum is regular). It remains to make use of 9.4.9 and 9.4.6.
=>: Put the initial space X in X":= Xu {oo}, adjoining to X a point 00 taken
elsewhere. Take the complements to X· of compact subsets of X as a base for the
neighborhood filter about 00. A neighborhood of a point x of X in X· is declared
to be a superset of a neighborhood of x in X. If 21 is an ultrafilter in X· and K
is a compactum in X then 21 converges to a point in K provided that K E 21.
If 21 contains the complement of each compactum K in X to X, then 21 converges
to 00. t>
9.4.22. REMARK. If a locally compact space X is not compact in its own
right then X· of 9.4.20 is the one-point or Alexandroff compactijication of X.
The uniformity J- l ('f/y ) is the inverse image of 'f/y under J. Therefore, the
inverse image of a uniform topology is uniformizable.
(4) Let (Xe, o/I'e)eEs be a family of uniform spaces. Assume further that
X:= IleES Xe is the product of the family. Put 'f/x := sUPeES Pre-I (o/I'e). The
uniformity o/I'x is the Tychonoff uniformity. It is beyond a doubt that the uniform
topology T('f/X) is the Tychonoff topology of the product of (Xe, T(o/I'e))eES. <II>
(5) Each Hausdorff compact space is uniformizable in a unique fashion.
<I By virtue of 9.4.15 such a space X may be treated as a subspace of a Ty-
chonoff cube. From 9.5.5 (3) and 9.5.5 (4) it follows that X is uniformizable.
Since each entourage of a uniform space includes a closed entourage; therefore, the
compactness property of the diagonal Ix of X 2 implies that every neighborhood
of Ix belongs to 'f/x. On the other hand, each entourage is always a neighborhood
of the diagonal. I>
(6) Assume that X and Yare nonempty sets, 'f/y is a uniformity on Y
and fJB is an upward-filtered subset of 9i'(X). Given B E fJB and () E o/I'y, put
UB,(J:= {(f, g) E YX x yX : go IB 0 J- l C ()}.
Then 'f/ := fil {UB,(J: B E fJB, () E 'f/y} is a uniformity on Y x. This uniformity
has a cumbersome (but exact) title, the "uniformity of uniform convergence on the
9.5. Uniform and Multimetric Spaces 161
members of !II." Such is, for instance, the uniformity of the Arens multinorm
(cf. 8.3.8). If !II is the collection of all finite subsets of X, then all coincides
with the Tychonoff uniformity on Y x. This uniformity is called weak, and the
corresponding uniform topology is called the topology of pointwise convergence
or, rarely, that of simple convergence. If ~ is a singleton {X}, then the uniformity
all is called strong and the corresponding topology 7(0//) in yX is the topology
of uniform convergence on X.
9.5.6. REMARK. It is clear that, in a uniform (or uniformizable) space, the
concepts make sense such as uniform continuity, total boundedness, completeness,
etc. It is beyond a doubt that in such a space the analogs of 4.2.4-4.2.9,4.5.8,4.5.9,
and 4.6.1-4.6.7 are preserved. It is a rewarding practice to ponder over a possibility
of completing a uniform space, to validate a uniform version of the Hausdorff
Criterion, to inspect the proof of the Ascoli-Arcela Theorem in an abstract uniform
setting, etc.
9.5.7. DEFINITION. Let X be a set and put JR+ := {x E JR.: x ~ a}.
A mapping d : X2 ~ JR+ is called a semimetric or a pseudometric on X, provided
that
(1) d(x, x) = a (x E X)j
(2) d(x, y) = dey, x) (x, y E X)j
(3) d(x, y) S; d(x, z) + d(z, y) (x, y, z EX).
A pair (X, d) is a semimetric space.
9.5.B. Given a semimetric space (X, d), let o//d:= fil {{d S; c}: c > a}. Then
o//d is a uniformity. <It>
9.5.9. DEFINITION. Let 9Jt be a (nonempty) set of semimetrics on X. Then
the pair (X, 9Jt) is a multimetric space with multimetric 9Jt. The multimetric
uniformity on X is defined as o//!Ul := sup{ 'Wd: dE 9Jt}.
9.5.10. DEFINITION. A uniform space is called multimeirizable, if its unifor-
mity coincides with some multimetric uniformity. A multimetrizable topological
space is defined by analogy.
9.5.11. Assume that X, Y, and Z are sets, T is a dense subset of i", and
(UdtET and (Vt)tET are increasing families of subsets of X x Z and Z x Y, re-
spectively. Then there are unique functions f : X x Z ~ i", g : Z x Y ~ i" and
h : X x Y ~ i" such that
<I The sought functions exist by 3.8.2. The claim of uniqueness is straightfor-
ward from 3.8.4. The presentation of h via f and 9 raises no doubts. t>
9.5.12. DEFINITION. Let f : X x Z -+ iR and 9 : Z x Y -+ R. The function h,
given by 9.5.11, is called the V-convolution (read: vel-convolution) of f and 9 and
is denoted by
Since
d(x, y) ~ hn(x, y) = fnD+ln(x, y) ~ fn(x, z) + fn(z, y),
for n E N, it follows that d(x, y) ~ d(x, z) + d(z, y). The validity of 9.5.7 (1)
and 9.5.7 (2) is immediate.
We are left with proving that 1 hf ~ d. To this end, show that fn ~ 1 hI for
n E N. Proceed by way of induction.
The desired inequalities are obvious when n:= 1,2. Assume now that f ~
it ~ ... ~ fn ~ Ihf and at the same time fn+l(X, y) < 1 hf(x, y) for some
(x, y) in X2 and n ~ 2.
9.5. Uniform and Multimetric Spaces 163
0, t<O
Ix, t=O
Ut := Vlnf{nEN: t~2-n}, O<t<l
VI, t=l
X2 , t> l.
{f < t} C Ut C {f ~ t} (t E R).
If Wt := Uu for t E R then
for s < t. Consequently, in virtue of 3.8.3 and 9.2.1 the mapping f is a 2-ultra-
metric.
Using 9.5.15, find a semimetric d v such that 1 Iz/ ~ d v ~ f. Clearly,
%'dv = fil {Vn: n EN}. Also, it is also beyond a doubt that %'m = %'X for
the multimetric 001:= {d v : V E %'x}. t>
9.5.17. Corollary. A topological space is uniformizable if and only if it is
a T 31 h -space. <It>
9.5.1B. Corollary. A Tychonoff space is the same as a separated multimetric
space. <It>
composes a point finite cover of X, and L.fEFf(x) = 1 for all x E U. The empty
family of functions in this context is treated as summable to unity at each point.
The term "continuous partition of unity" and the like are understood naturally.
9.6.7. DEFINITION. Let C be a cover of a subset U of a topological space
and let F be a continuous partition of unity on U. If the family of supports
{supp (I): f E F} refines C then F is a partition of unity 8ubordinate to C.
A possibility of finding such an F for C is also verbalized as follows: "C admits
a partition of unity."
9.6.8. Each locally finite open cover of a normal space admits a partition
of unity.
e
By the Lefschetz Lemma, such a cover {Ue: E 3} has an open refinement
e e
<I
{~: E 3} satisfying the condition cl Ve C Ue for all E 3. By the Urysohn
Theorem, there is a continuous function ge : X - t [0, 1) such that ge( x) = 1 for
x E Ve and ge(x) = 0 for x E X \ Ue· Consequently, supp(ge) CUe. In virtue
of 9.6.5 the family (ge)eEB is summable pointwise to a continuous function g.
Moreover, g(x) > 0 for all x E X by construction. Put Ie:= gdg (e E 3). The
family (Ie )eEB is what we need. [>
9.6.9. DEFINITION. A topological space X is called paracompact, if each cover
of X has a locally finite open refinement.
9.6.10. REMARK. The theory of paracompactness contains deep and surpris-
ing facts.
9.6.11. Theorem. Every metric space is paracompact.
9.6.12. Theorem. A Hausdorff topological space is paracompact if and only
if its every open cover admits a partition of unity.
9.6.13. REMARK. The metric space RN possesses a number of additional
structures providing a stock of well-behaved, 8mooth (= infinitely differentiable)
functions (cf. 4.8.1).
9.6.14. DEFINITION. A mollifier or a mollifying kernel on RN is a real-valued
smooth function a having unit (Lebesque) integral and such that a(x) > 0 for
Ixl < 1 and a(x) = 0 for Ixl ~ 1. In this event, supp(a) = {x E RN: Ixl:S; I} is
the (unit Euclidean) ball B:= ERN.
9.6.15. DEFINITION. A delta-like 8equence is a family of real-valued (smooth)
functions (b.:)",>o such that, first, lim",-+o(sup I supp (b",) I) = 0 and, second, the
equality holds JRN b",(x)dx = 1 for all c > O. The terms "6-sequence" and "6-like
sequence" are also in current usage. Such a sequence is often assumed countable
without further specification.
9.6.16. EXAMPLE. The function a(x) := texp{-{lxI 2 - 1)-1) is taken as
a most popular mollifier when extended by zero beyond the open ball int B, with
9.6. Covers, and Partitions of Unity 167
J
is defined as
f*g(x):= f(x-y)g(y)dy (XEJRN).
IRN
9.6.1B. REMARK. The role of a mollifying kernel and the corresponding delta-
like sequence (a., ).,>0 becomes clear from inspecting the aftermath of applying
the smoothing process f ~ (f * a., ).,>0 to a function f belonging to ~ ,loc (JR N)
(cf. 10.10.7 (5)).
9.6.19. The following statements are valid:
(1) to every compact set K in the space JR N and every neighborhood U
of K there corresponds a truncator (= a bump function) 'I/J:= 'l/JK,U,
i.e. a smooth mapping'I/J : JRN -T [0, 1] such that K C int {'IjJ = I}
and supp('I/J) C U;
(2) assume that U1 , ... ,Un E Op(JR N ) and U1 U ... U Un is a neighbor-
hood of a compact set K; there are smooth functions 'l/Jl, ... ,'l/Jn :
JRN -T [0, 1] such that SUpp('l/Jk) C Uk and L:~=l 'l/Jk(X) = 1 for x
in some neighborhood of K.
<l (1) Put c::= d(K, JRN \ U):= inf{lx - yl: x E K, y f/. U}. It is clear that
c: > O. Given f3 > 0, denote the characteristic function of K + c:lm by Xp. Take
a delta-like sequence (b'Y )'Y>o of positive functions and put 'I/J := XP * b'Y. When
"( :::; f3 and f3 + "( :::; c: with "(:= sup 1supp (b'Y )1, observe that 'I/J is a sought function.
(2) By the Diedonne Lemma there are closed sets Fk, with Fk C Uk, compos-
ing a cover of K. Put Kk:= FknK and choose some truncators 'l/Jk:= 'l/JKk,Uk. The
functions 'l/Jk/ L:~=l 'l/Jk (k:= 1, ... ,n), defined on {L:~=l 'l/Jk > O}, meet the claim
after extension by zero onto {L:~=l 'l/Jk = O} and multiplication by a truncator
corresponding to an appropriate neighborhood of K. I>
9.6.20. Countable Partition Theorem. Let tff be a family of open sets
in JR N and n := utff. There is a countable partition of unity which is composed
of smooth compactly-supported functions on JRN and subordinate to the cover tff
ofn.
<l Refine from tff a countable locally finite cover A of n
with compact sets so
that the family (a:= int a)"'EA be also an open cover of n. Choose an open cover
n
(V"')",EA of from the condition cl V", c a for a E A. In virtue of 9.6.19 (1) there
are truncators ~'" := 'l/Jcl Va,a. Putting 'I/J",(x) := ~",(x)/ L:"'EA ~",(x) for x E n
and 'I/J",( x):= 0 for x E JR N \ n, arrive to a sought partition. I>
168 Chapter 9. An Excursion into General Topology
9.S. Show that a Tychonoff product is a Hausdorff space if and only if so is every factor.
9.9. Establish compactness criteria for subsets of classical Banach spaces.
9.10. A Hausdorff space X is called H-closed, if X is closed in every ambient Hausdorff
space. Prove that a regular H-closed space is compact.
9.11. Study possibilities of compactifying a topological space.
9.12. Prove that the Tychonoff product of uncountably many real axes fails to be a nor-
mal space.
9.13. Show that each continuous function on the product of compact spaces depends
on at most count ably many coordinates in an evident sense (specify it!).
9.14. Let A be a compact subset and let B be a closed subset of a uniform space, with
An B = 0. Prove that V(A) n V(B) = 0 for some entourage V.
9.15. Prove that a completion (in an appropriate sense) of the product of uniform spaces
is uniformly homeomorphic (specify!) to the product of completions of the factors.
9.16. A subset A of a separated uniform space is called precompact if a completion of A
is compact. Prove that a set is precompact if and only if it is totally bounded.
9.17. Which topological spaces are metrizable?
9.18. Given a uniformizable space, describe the strongest uniformity among those in-
ducing the initial topology.
9.19. Verify that the product of a paracompact space and a compact space is paracom-
pact. Is paracompactness preserved under general products?
Chapter 10
Duality and Its Applications
+ : (X x X, r x T) -+ (X, T),
. : (IF x X, 'TJF x r) -+ (X, T).
In this regard a vector topology is often called a linear topology and a topological
vector space, a linear topological space. This terminology should be used on the
understanding that a topology may possess the "linearity" property while failing
to be linear. For instance, such is the discrete topology of a nonzero vector space.
170 Chapter 10. Duality and Its Applications
we are left with examining the continuity property of scalar multiplication at zero
in the second argument with the first argument fixed. In other words, it is neces-
sary to show that AJV :J JV for A E IF. With this in mind, find n E N such that
:s
IAI n. Let V in JV and Win JV be such that W is balanced and WI + ... + Wn C
V, where Wk:= W. Then AW = n (A/nW) C nW C WI + ... + Wn C V. l>
10.1.5. Theorem. The set VT (X) of all vector topologies on X presents
a complete lattice. Moreover,
It is obvious that PB is continuous (cf. 7.5.1); i.e., PB E Wlr, and so {PB < I} E
w(O). Consequently, V E w(O). Using 5.2.10, infer that w(x) = x + w(O) ::::>
x + T(O) = T(X); i.e., w 2: T. Furthermore, T 2: w by definition. I>
10.2.9. DEFINITION. A vector space, endowed with a separated locally con-
vex topology, is a locally convex space.
10.2.10. REMARK. Theorem 10.2.8 in slightly trimmed form is often verbal-
ized as follows: "The concept of locally convex space and the concept of sepa-
rated multinormed space have the same scope." For that reason, the terminology
10.3. Duality Between Vector Spaces 173
10.2.13. Prime Theorem. The prime mapping 7 t-t (X, 7)' from LCT (X)
to Lat (X#) preserves suprema; i.e.,
If f E (X, sup g)" then in view of 10.2.12 and 9.1.13 there are topologies
71, ... ,7n E g such that f E (X, 71 V ... V 7 n )'. Using 10.2.12 and 5.3.7, find
PI E 9JtT1 , · · · ,Pn E 9JtTn satisfying f E 181(pl V ... VPn). Recalling 3.5.7 and 3.7.9,
observe that 181(pl + ... + Pn) = 181(pd + ... + 181(Pn). Finally,
f E (X, 7d' + ... + (X, 7 n )' = (X, 7t)' V ... V (X, 7 n )'. [>
The mappings (. 1and I,) are the bra-mapping and the ket-mapping of the initial
bilinear form. By analogy, a member of (X 1 is a bra-functional on X and a member
of 1Y) is a ket-functional on Y.
10.3.2. The bra-mapping and the ket-mapping are linear operators. <][>
174 Chapter 10. Duality and Its Applications
{I(·I y)l: y E Y}. Likewise, a(Y, X) is the topology of the multinorm {I(x 1·)1 :
x EX}. <ll>
10.3.8. Tbe spaces (X, a(X, Y)) and (Y, a(Y, X)) are locally convex.
<l Immediate from 10.2.4 and 10.2.5. I>
10.4.3. Let reX, Y) stand for the least upper bound of the set of all locally
convex topologies on X compatible with duality (between X and Y). Then the
topology reX, Y) is also compatible.
<l Denote the set of all compatible topologies on X by tC. Theorem 10.2.13
readily yields the equalities
(X, reX, Y))' = sup{(X, r)': r E tC} = sup{1 Y): r E tC} = IY),
because tC is nonempty by 10.4.2. I>
10.4.4. DEFINITION. The topology reX, Y), constructed in 10.4.3 (i.e., the
finest locally convex topology on X compatible with duality by pairing X f-t Y),
is the Mackey topology (on X induced by X f-t Y).
10.4.5. Mackey-Arens Theorem. A locally convex topology r on X is
compatible with duality between X and Y if and only if
IY) = (X, a(X, Y)) C (X, r)' C (X, reX, Y))' = IY).
The remaining claim is obvious. I>
10.5. Po lars
10.5.1. DEFINITION. Let X and Y be sets and let F C X x Y be a corre-
spondence. Given a subset U of X and a subset V of Y, put
The set 7l"(U) is the (direct) polar of U (under F), and the set 7l"-l(V) is the
(reverse) polar of V (under F).
10.5.2. The following statements are valid:
(1) 7l"(u):= 7l"({u}) = F(u) and 7l"(U) = n"EU7l"(U);
(2) 7l"(UeESUe) = neES7l"(Ue);
(3) 7l"Fl(V) = 7l"F-l(V);
(4) U1 c U2 ::::} 7l"(U1 ) J 7l"(U2 );
(5) U x V c F::::} (V C 7l"(U) & U C 7l"-l(V);
(6) U C 7l"-l(7l"(U». <It>
178 Chapter 10. Duality and Its Applications
To refer to direct or inverse polars under pol, we use the unified term "polar" (with
respect to X +-t Y) and the unified designations 7r(U) and 7r(V). In the case of the
correspondence abs pol, we speak of absolute po lars (with respect to X +-t Y) and
write UO and VO (for U C X and V C Y).
10.5.8. Bipolar Theorem. The bipolar 7r 2 (U):= 7r( 7r(U)) is the (inclusion)
least weakly closed conical segment greater than U.
<l Straightforward from 1004.8 and the Akilov Criterion. I>
If 5 is absolutely convex, then the absolute polar 5° is the inverse image of the
(algebraic) balanced subdifferentiaI181(ps) under the ket-mapping; i.e.,
<l If Y E YlR and Y E I 8(ps) )Jil then I Y)lR belongs to 8(ps). Hence, Re (x I y) =
(xIY)lR = IY)lR(x) ~ ps(x) ~ 1 for x E 5, because 5 C {ps ~ 1} by the Gauge
Theorem. Consequently, Y E 7r(5).
If, in turn, Y E 7r(5) then I Y)lR belongs to 8(ps). Indeed, 1 > ps(a-1x) for
all x in XlR and a> ps(x); i.e., a-Ix E {ps < 1} C 5. Whence (a-1xIY)lR =
Re (a- l x I y) = a-IRe (x I y) ~ 1. Finally, observe that I Y)lR(x) ~ a. Since a
is arbitrary, this inequality means that I Y)lR(x) ~ ps(x). In other words, Y E
I 8(Ps))JiI, which implies that 7r(5) = 18(ps))Ji I . The remaining claim follows
from the properties of the complexifier (cf. 3.7.3 and 3.7.9). [>
10.6.7. Alaoglu-Bourbaki Theorem. The polar of a neighborhood of zero
of each compatible topology is a weakly compact convex set.
<l Let U be a neighborhood of zero in a space X and let 7r(U) be the polar of U
(with respect to X ~ X'). Since U J {p ~ 1} for some continuous seminorm p,
by 10.5.2 (4), 7r(U) C 7r( {p ~ 1}) = 7r(Bp) = B;. Using 10.6.6 and recalling that
p is the Minkowski functional of B p, obtain the inclusion 7r(U) C 181(p). By virtue
of 10.6.2 the topological balanced sub differential 181(p) is a(X', X)-compact.
By definition 7r(U) is weakly closed. To infer the a(X', X)-compactness of 7r(U),
it remains to appeal to 9.4.9. The convexity property of 7r(U) is beyond a doubt. [>
by the topology u(X", X'), conclude that Ex" = Ex (because of the obvious
convexity and closure properties of Ex, the latter following from compactness
since X is separated). Thus, X is reflexive. I>
10.7.2. Corollary. A space X is reflexive if and only if every bounded closed
convex set in X is weakly compact. <11>
10.7.3. Corollary. Every closed subspace of a reflexive space is reflexive.
<1 By the Mazur Theorem, such a subspace and, hence, the unit ball of it are
weakly closed. It thus suffices to apply the Kakutani Criterion twice. I>
10.7.4. Pettis Theorem. A Banach space and its dual are (or are not) re-
flexive simultaneously.
<1 If X is reflexive then u(X', X) coincides with u(X', X"). Therefore,
by the Alaoglu-Bourbaki Theorem, Ex' is u(X', X")-compact. Consequently,
X, is reflexive. If, in turn, X' is reflexive, then so is X" by what was proven.
However, X, as a Banach space, is a closed subspace of X". Thus, X is reflexive
by 10.7.3. I>
10.7.5. James Theorem. A Banach space is reflexive if and only if each
continuous (real) linear functional attains its supremum on the unit ball of the
space.
10.8.4. EXAMPLES.
(1) 0, C(Q, 1R), and the closure of a sublattice.
(2) The intersection of each family of sublattices is also a sublattice.
(3) Let L be a sublattice and let Qo be a subset of Q. Put
Therefore,
LQo = {J E C(Q, 1R): flQo E LIQJ·
Moreover, if L contains a constant function other than zero (i.e. a nonzero multiple
of the constantly-one function 1) then as 0:1 and 0:2 in the above presentation
of L{ql,Q2} the unity, 1, may be taken. <It>
10.8.5. Let L be a sublattice ofC(Q, 1R). A function f in C(Q, 1R) belongs
to the closure of L if and only if for all e > 0 and (x, y) E Q2 there is a function
l:= fZ,y,E in L satisfying the conditions
clL=
The inclusion of cl L into cl (L {q, ,q2}) for all (ql, q2) E Q2 raises no doubts.
<l
If IE cl (L{Q"Q2}) for all such ql and q2, then by 10.8.5, I E cl L. I>
10.8.8. Corollary. Every vector sublattice L of C(Q, JR.) is expressible as
clL=
<l Observe that every set of the form L {Q, ,Q2} is closed. I>
for every pair (ql, q2) in Q2 (cf. 10.8.4 (6)). It remains to appeal to 10.8.8. I>
.Af(/1-) = n
qEQ
.Af(/1-){q}'
.Af(/1-) = n
qEsupp (,,)
.Af(/1-){q} = {J E C(Q, JR): fl supp (,,) = o}.
The claim of uniqueness follows from the normality of Q (cf. 9.4.14) and the
Urysohn Theorem. t>
10.8.12. DEFINITION. The set supp (/1-) under discussion in 10.8.11 is the
support of /1- (cf. 10.9.4 (5)).
10.8.13. REMARK. If /1- is positive then
.Af(/1-) = {f E C(Q, JR): /1-(lfl) = o}.
Consequently, when /1-(fg) = 0 for all g E C(Q, JR), observe that
supp "
= o. fl ()
By analogy supp(/1-) = 0 {:} .Af(/1-) = C(Q, JR) {:} /1- = O. Therefore, it is quite
convenient to work with the support of a positive functional.
Let F be a closed subset of Q. It is said that F supports or carries /1- or
that X \ F lacks /1- or is void of /1- if /1-(lfl) = 0 for every continuous function f
with supp(f) c Q \ F. The support supp(/1-) of /1- carries /1-; moreover, supp(/1-)
is included in every closed subset of Q supporting /1-. In other words, the support
of /1- is the complement of the greatest open set void of /1- (cf. 10.10.5 (6)).
It stands to reason to observe that in virtue of 3.2.14 and 3.2.15 to every
bounded functional/1- there correspond some positive (and hence bounded) func-
tionals /1-+, /1--, and 1/1-1 defined as
/1-+(f) = sup /1-[0, fl; /1--(f) =- inf /1-[0, fl; 1/1-1 = /1-+ + /1--,
given f E C(Q, JR)+.
Moreover, C(Q, JR)' is a Kantorovich space (cf. 3.2.16). <It>
10.8. The Space C(Q, JR) 185
10.8.15. Considering a E C(Q, JR) with 0:::; a:::; 1, define af-l: I t-+ f-l(af) for
IE C(Q, JR) and f-l E C(Q, JR)'. Then laf-ll = alf-ll·
<l Given I E C(Q, JR)+, infer that
Furthermore,
Put f-ll:= att and f-l2:= (1 - a)tt. It is clear that f-ll + f-l2 = f-l and the functionals
f-ll and f-l2 are both nonzero. Moreover,
and ttl, tt2 E AJ.., conclude that f-ll = II f-lI/! tt· By 10.8.15, alf-ll = laf-ll = Ittll =
lIf-lllllttl· Consequently, If-ll((a -1if-llIl1)g) = 0 for all g E C(Q, JR). Using 10.8.13
and 10.8.14, infer that the function a is constant on the support of f-l. [>
186 Chapter 10. Duality and Its Applications
for all t E [-1, 1]. Then IPn (0) I ::; 1 hn. Therefore, the polynomial
maintains the inequality IPnCt) -Itll ::; 1/ n when -1 ::; t ::; 1. By construction, Pn
lacks the constant term. Now, if a function a lies in a subalgebra A of C(Q, R)
and lIall ::; 1, then
IPn (a( q» - Ia( q) II ::; 1 / n (q E Q).
Moreover, the function q t-+ Pn(a(q» is clearly a member of A. t>
It is beyond a doubt that im ~ is a sub algebra of C( Qo, 1R) separating the points
of Qo and containing constant functions. In virtue of 10.8.17 it is thus sufficient
(and, clearly, necessary) to examine the closure of im ~.
Consider the monoquotient "£ of the operator ~ and the coset mapping c.p :
C(Q, 1R) -+ coim ~. Given f E C(Q, 1R), put
By definition flQo = glQo; i.e., 1:= c.p(f) = c.p(g). Consequently, Ilgll > 11111·
Furthermore,
Therefore,
i.e., t IS an isometry. Successively applying 5.5.4 and 4.5.15, infer first that
coim ~ is a Banach space and second that im"£ is closed in C(Qo, 1R). It suf-
fices to observe that im ~ = im "£. [>
Ul . U2( ql, q2):= Ul Q9 U2( ql, q2):= Ul (ql )U2( q2) with Us E K( Qs) is
dense in C(Q, IF);
(3) if n is compact then K(n) = C(n, IF); if n fails to be compact
and is embedded naturally into C(n', IF), with n' := n U {oo} the
Alexandroff compactification of n, then the space K(n) is dense
in the hyperplane {J E C(n', IF): f( 00) = OJ;
(4) the mapping E E Open) ~ K(E) E Lat(K(n)) preserves suprema;
(5) for E', E" E Op (n) the following sequence is exact:
J
fl
f dJ.L:= J
f dJ.L:= J
f(x)dJ.L(x):= J.L(f) (f E K(n)).
The scalar J.L(f) is the integral of f with respect to J.L. In this connection J.L is also
called an integral.
10.9.4. EXAMPLES.
(1) For q E n the Dirac measure f ~ f(q) (f E K(n)) presents a Radon
measure. It is usually denoted by the symbol bq and called the delta-function at q.
Suppose also that n is furnished with group structure so that the taking
of an inverse q E n ~ q-l E n and the group operation (s, t) E n x n ~ st E n
are continuous; i.e., n is a locally compact group. By b we denote be where e is the
unity of n (recall the concurrent terms: "identity," "unit," and "neutral element,"
10.9. Radon Measures 189
all meaning the same: es = se = s for all sED). The nomenclature pertinent
to addition is routinely involved in abelian (commutative) groups.
Given a E D, acknowledge the operation of (left or right) translation or shift
by a in K(D) (in fact, every function is shifted in D x IF):
J
!1
f(a- 1 x)dx = J
!1
f(x)dx (1 E K(D), a ED).
Observe that, in current usage, the word "measure" often means a positive mea-
sure, whereas a "general" measure is referred to as a signed measure or a charge.
If fL and v are measures and IfLl /\ Ivl = 0 then fL and v are called disjoint or
independent (of one another). A measure v is absolutely continuous with respect
to fL on condition that v is independent of every measure independent of fl. Such
a measure v may be given as v = ffL, where f E L 1 .1oc (fL) and the measure ffL
having density f with respect to fL acts by the rule (ffL)(9):= fL(f9) (g E K(n))
(this is the Radon-Nikodym Theorem).
(4) Given n' E Op (n) and fL E vIl(n), consider the restriction fLrI' :=
fLIK(rI') of fL to K(n'). The restriction operator fL I-t fLrI' from vIl(n) to vIl(n'),
viewed as depending on a subset of n, meets the agreement condition: if nil c
n' c nand fL E vIl(n) then fLrI" = (fLrI' )rl lI . This situation is verbalized as
follows: "The mapping vIl : E E Op (n) I-t vIl(E) and the restriction operator
(referred to jointly as the functor vIl) defines a presheaf (of vector spaces over n)."
It stands to reason to convince oneself that the (values of the) restriction operator
need not be an epimorphism.
(5) Let E E Open) and fL E vIl(n). It is said that E lacks fL or is
void of fL or that n \ E supports or carries fL if fLE = O. By 10.9.2 (4) there
is a least closed set supp (fL) supporting fL, the support of fl. It may be shown
that SUPP(fL) = supp(lfLl). The above definition agrees with 10.8.12. The Dirac
measure bq is a unique Radon measure supported at {q} to within scaling.
(6) Let n k be a locally compact space and fLk E vIl(n k ) (k := 1,2).
There is a unique measure fL on the product n l x n 2 such that
J
rllXrl2
Ul(X)U2(y)dfL(X,y) = J
rI,
ul(x)dfLl(X) J
rl2
u2(y)dfL2(Y)
with Uk E K(nk). The next designations are popular: fLl x fL2 := fLl 18) fL2 := fl·
Using 10.9.2 (4), infer that for f E K(nl x n 2 ) the value fLl x fL2(f) may be
calculated by repeated integration (this is the Fubini Theorem).
(7) Let G be a locally compact group, and fL, v E M( G). For f E K( G)
the function j(s, t):= fest) is continuous and l(fL x v)(i)1 :S IlfLllllvllllflloo. This
defines the Radon measure fL * v(f):= (fL X v)(j) (f E K( G)), the convolution of fL
and v. Using vector integrals, obtain the presentations:
fL *v = J
GxG
bs * bt dfL(S)dv(t) = J
G
bs * vdfL(S) = J*
G
fL bt dv(t).
I*g = /(8"
G
*g)I(x)dm(x) =/
G
T,,(g)I(x)dm(x).
10.9.6. Tbe topology TK(fI) is strongest among all locally convex topologies
making tbe identical embedding of K(Q) to K(n) continuous for every Q witb
Q ~ n (i.e., TK(fI) is tbe so-called inductive limit topology (cf. 9.2.15)).
<l If T is the inductive limit topology and J-l E (K(n), T)' then by definition
J-l E Al(n), because J-l 0 LK(q) is continuous for Q ~ n. In turn, for J-l E Al(n)
the set Vq:= {f E K(Q): IJ-l(f) I ::; 1} is a neighborhood of zero in K(Q). From
the definition of T, infer that U {Vq: Q ~ n} = {f E K(n): IJiU)I::; 1} is
a neighborhood of zero in T. Thus, J-l E (K(n), T)' and T is compatible with
duality. Therefore, T ::; TK(fI).
On the other hand, if p is a seminorm in the mirror of the Mackey topology
then p is the supporting function of a subdifferentiallying in Al(n). Consequently,
192 Chapter 10. Duality and Its Applications
where tn is the "restriction" of t to K n , the symbol (J'n stands for addition and
IT
k<1
k,IE{l, ... ,n}
- -
By hypothesis, im tn = ker (J'n· If (J'n+! (J, fn+l) = 0 where f:= (II,· .. ,fn), then
(J'n1 = - fn+l and fn+l E K((EI U ... U En) n En+d. Since (J'n is an epimorphism
by 10.9.2 (5), there are some Ok E K(Ek n En+I) such that (J'nO = - fn+l for
0:= (0 1 , ... ,On)' Whence il- 0) E ker (J'n, and by hypothesis there is a member
x of Kn such that tnX = f - O. Clearly,
IT K(Ek n En+d
n
K n+l = Kn x
k=l
(to within isomorphism), x:= (x, 01 , ... ,On) E Kn+I, and tn+lX = (1, fn+d.
Passing to the diagram prime (cf. 7.6.13), arrive at the exact sequence
o --+ ~(E' nEil) t(E~EII») ~(E') x ~(E") O'(E~EII») ~(E' U E") --+ O.
(4): Take q' E ]RN and q" E ]RM. Let a(q', q"):= a'(q')a"(q"), where a'
and a" are mollifiers on ]RN and ]RM, respectively. Given f E 9&(Q), n E N, and
c > 0, choose X from the condition Ilf - f *axlin,Q :=:; eh. Using the equicontinuity
property of the family $:= {GO< f(q)Tq(a x ): lal:=:; n, q E Ql x Q2}, find finite
sets 6,.' and 6,.", with 6,.' c Ql and 6,." C Q2, so that the integral of each function
in $ be approximated to within 1 h(N + 1)-n c by a Riemann sum of it using
the points of 6,.' x 6,.". This yields a function 1, a member of the linear span of
9&(Ql) x 9&(Q2), which was sought; i.e., Ilf -11In,Q :=:; c.
(5): Check this is as in 10.9.2 (4), on replacing 9.4.18 with 9.6.19 (2). I>
10.10.3. REMARK. The Generalized Weierstrass Theorem may be applied
to the demonstration of 10.10.2 (4), when combined with due truncation providing
that the constructed approximation has compact support.
10.10.4. DEFINITION. A functional u, a member of 9&(n, If)#, is a distribu-
tion or a generalized function whenever u I~(Q) E 9&'(Q):= 9&(Q)' for all Q <s n.
This is expressed in writing as u E 9&'(n):= 9&'(n, If). Sometimes a reference is
appended to the nature of the ground field If.
The usual designations are as follows: (u, J):= (J Iu):= u(J). Often we use
the most suggestive and ubiquitous symbol
10.10.5. EXAMPLES.
(1) Let 9 E Ll,loc(]RN) be a locally integrable function. The mapping
The notions of the order of a distribution and of a distribution of finite order are
understood in a matter-of-fact fashion. Evidently, it is false that every distribution
has finite order.
196 Chapter 10. Duality and Its Applications
+
(9) Let u, v E ~'(R,N). Given f E ~(R,N), put f := f 0 +. It is
+
clear that f E Coo(R, N X R, N). Say that the distributions U and v convolute or
admit convolution or are convolutive provide that the product U x v applies to
+
the function f C Coo(R,N X R,N) for each f in ~(R,N). It is easy (cf. 10.10.10)
+
that the resulting linear functional f I--t (u X v)( f) (f E ~(R, N)) is a distribution
called the convolution of U and v and denoted by U * v. It is beyond a doubt that
the convolutions of functions (cf. 9.6.17) and measures on R,N (d. 10.9.4 (7)) are
particular cases of the convolution of distributions. In some classes, every two
distributions convolute. For instance, the space G"'(R,N) of compactly-supported
distributions with convolution as multiplication presents an (associative and com-
mutative) algebra with unity the delta-function 8. Further, [)OI U = [) 0I 8 * U and
[)OI(U * v) = [)OIU * V = U * [)OI V • Moreover, the following remarkable equality, the
Lions Theorem of Supports, holds:
10.10.6. DEFINITION. The spaces ~(n) and ~/(n) are set in duality (in-
duced by the duality bracket ~(n) H ~(n)#). Moreover, ~'(n) is furnished
with the topology of the distribution space a( ~'(n), ~(n)), and ~(n) is furnished
with the topology of the test function space, the Mackey topology T~ := T~(n) :=
T(~(n), ~/(n)).
L Uk(f).
00
U(f):=
k=l
U= t L Va 0 8a = t L (-1)laI8ava
lal:<:;m lal:<:;m
'l/Jk U = L 8 a flk,a,
lal:<:;m
10.11. The Fourier Transform of a Distribution 201
L fLk,Ot(J)
00
fLOt(J) :=
k=1
exists for all f E K(n). Moreover, the resulting distribution fLOt is a Radon mea-
sure. Once again appealing to 10.10.10, infer that
Put 1jJ( x) := X(f) -1 X( i5 x * f). This soundly defines some continuous mapping
1jJ : R N -+ c.. Moreover,
1jJ(x + y)
= xU * 9)-1 X( i5 x +y * U * g)) = X(f)-1 X(g )-1 X( i5x * f * i5y * g)
= X(f)-I X(i5 x * f)X(g)-I X(i5 y * g) = 1jJ(x)1jJ(y)
for x, y ERN; i.e., 1jJ is a (unitary) group character: 1jJ E X(RN). Calculus shows
that 1jJ = et for some (obviously, unique) t E JR N. Further, by the properties of the
Bochner integral
= J
]RN
X(i5 x *g)f(x)dx = J
]RN
f(x)X(g)1jJ(x)dx = X(g) J
]RN
f(x)1jJ(x)dx.
J
Thus,
xU) = f(x)1jJ(x)dx U E L 1(JRN)).
]RN
j;g(t) = uf*g(et)
= JJ
]RN ]RN
f(x)g(y)et(x+y)dxdy= J f(x)et(x)dx
]RN]RN
J g(y)et(y)dy
10.11.2. REMARK. The essential steps of the above argument remain valid
for every locally compact abelian group G, and so there is a one-to-one correspond-
ence between the character space X( Ll (G)) of the group algebra and the set X( G)
comprising all (unitary) group characters of G. Recall that such a character is
a continuous mapping 1jJ : G -+ C satisfying
J
]RN
J
]RN
Since the character groups of isomoryhic groups are also isomorphic, there are
grounds to using the same notation f for generally distinct functions § f, § _ f,
and § ±21r f. The choice of the symbol ~ for §21r (§-21r) dictates the denota-
tion v for §-21r (§21r) (cf. 10.11.12).
10.11.5. EXAMPLES.
(1) Let f : Jlt - t Jlt be the characteristic function of the interval [-1, 1].
204 Chapter 10. Duality and Its Applications
J liet)1 dt 2: J liet)1 dt = ?;
00
J liet)1 dt
J
[to,+oo) [k-n-,+oo) - [n1r,(n+l)1rj
2:
f 21 sin tid
(n+1)7r t=4
f 1
(n+1)7r=+00.
n=k [n1r,(n+l)1rj n=k
Thus, f 1. Ll(~).
(2) For fELl (~N) the function f is continuous, with the inequality
Ilflloo ~ Ilflh holding.
<l The continuity of f
follows from the Lebesque Dominated Convergence
Theorem; and the boundedness of f, from the obvious estimate
liet)1 ~ J
]RN
If(x)1 dx = Ilflll (t E ~N). I>
(5) §(f*) = (§~f)*, -;;] = exf, and = Txf for all f E Ll(~N) 0
and x E JRN.
<I We will only check the first equality. Since a*b = (ab*)* for a, b E C;
on using the properties of conjugation and integration, given t E JR N, infer that
(1
]RN ]RN
<l The first equality is straightforward from 10.11.1. The second, the multipli-
cation formula, follows on applying the Fubini Theorem:
J JJ]RN
[g =
]RN ]RN
f(x)et(x) dxg(t) dt
= J( J
]RN ]RN
g(t)et(X)dt) f(x)dx = J
]RN
fg· t>
JJ
]RN ]RN]RN
= f(y)g(t)et(x + y)dtdy
J J
]RN ]RN
]RN ]RN
~1 ($J)(t) = 8~1
IRN
J f(x)ei(x,t) dx = J
IRN
f(X) ix 1 ei (X,t) dx = $(ixd)(t). [>
II Jeitkxke-1hlxl~ dXk
N
iN(t) = (t E JRN).
k=l IR
Consequently, the matter reduces to the case N = 1. Now, given y E JR, observe
that
~
JI(y) = J e- 1/ 2X2 e'xy
. dx = J e- 1/ 2 ( X-'Y.)2 - 1/ 2Y 2 dx
J
IR IR
To calculate the integral A that we are interested in, consider (concurrently ori-
ented) straight lines '\1 and '\2 parallel to the real axis JR in the complex plane
CIR c::' JR 2 • Applying the Cauchy Integral Theorem to the holomorphic function
f(z):= exp ( - Z2 h) (z E C) and a rectangular with vertices on '\1 and '\2 and
properly passing to the limit, conclude that J)..J(z)dz = J)..2 f (Z)dz. Whence
it follows that
IR IR
10.11.6. DEFINITION. The Schwartz space is the set of tempered (or rapidly
decreasing) functions, cf. 10.11.17 (2),
(as usual, Ixl stands for the Euclidean norm of a vector x in jRN);
(4) ~(jRN) is dense in .9'(jRN); furthermore, the identical embedding
of ~(jRN) into .9'(jRN) is continuous and .9'(jRN)' C ~'(jRN);
(5) .9'(jRN) C L1(jRN).
<l We will check (4), because the other claims are easier.
Take f E .9'(jRN) and let '!/J be a truncator in ~(jRN) such that BeN = 1}.
Given x E jRN and e> 0, put '!/Je(x) := '!/J(ex), and fe = '!/Jef. Evidently,
fe E ~(jRN). Let E > 0 and a, {3 E (Z+)N. It is an easy matter to show
e
that sup{lIa1'('!/Je -1)1100: 'Y ~ (3, 'Y E (Z+)N} < +00 for 0 < ~ 1. Considering
that x aa Pf( x) ---+ 0 as Ixl ---+ +00, find r > 1 satisfying Ixaa p« '!/Je( x) -1 )f( x))1 < E
whenever Ixl > r. Moreover, fe(x) - f(x) = ('!/J(~x) -l)f(x) = 0 for Ixl ~ ~-l.
Therefore,
Thus,
1
Whence it is easily seen that E .9'(jRN) and the restriction of § to ~(Q) with
Q ~ jRN is continuous. It remains to refer to 10.10.7 (4) and 10.11.7 (4). t>
208 Chapter 10. Duality and Its Applications
10.11.7 derive that i, f, 9 E Ll(JR N ) and so, by 10.11.5 (7), (igf = f-* g. Put
ge(x):= g(c:x) for x E JRN and c: > o. Then for the same x, in view of 10.11.5 (4)
find
J
~N
g(ct)ict)et(x)dt= c:~ J
~N
f(y-x)g(D dy= J
~N
f(c:y-x)g(y)dy.
J
~N ~N
~N
J f g* = J
~N
ig* (f, 9 E .9"(JR N)).
10.11. The Fourier Transform of a Distribution 209
<I In view of 10.11.10 and 10.11.5 (5), only the sought identity needs examining.
Moreover, given f and g, from 10.11.5 (7) and 10.11.7 (4), obtain (1g)(0)
(J* g)(O). Using 10.11.5 (5), conclude that
J = (~(~-1
fg* = ((~-1 ~g*)(O)
J)g*((O) J)-*
= = dx = t>
~N ~N ~N
Jf(t) = 1
(27rth
J f(x)ei(x,t) dx'
'
J
~N
J- 1 f(x) = 1 f(t)e-i(x,t) dt
(27rt h
~N
10.11.17. EXAMPLES.
(1) Lp(JR. N) C .9"(JR.N) for 1 :S P :S +00.
<I Let f E Lp(JR. N ), 'IjJ E .9'(JR. N ), P < +00 and l/pl + l/p = 1. Using Holder
inequality, for suitable positive K, K', and K" successively infer that
J
1
IUf('IjJ)1 = 1('ljJlf)1 = J
]RN
f'IjJ :S IIfllpll'IjJllp':S Kpl('IjJ).
Let VOI:= (_1)1 01 1(1 +j • j2)n/J0I. Then Vol is tempered and u = L:IOII~n aOlvOI . [>
10.13. Let X and Y be Banach spaces and let T be a element of B(X, Y) satisfying
im T = Y. Demonstrate that Y is reflexive provided so is X.
10.14. Show that the spaces (X')" and (X")' coincide.
10.15. Prove that the space Co has no infinite-dimensional reflexive subspaces.
10.16. Let p be a continuous sublinear functional on Y, and let T E ~ (X, Y) be
a continuous linear operator. Establish the following inclusion of the sets of extreme points:
ext T'(8p) C T'(ext 8p).
10.17. Let p be a continuous semi norm on X and let .'!£ be a subspace of X. Prove
that f E ext (.'!£ 0 n 8p) if and only if the next equality holds:
X = cl .'!£ + {p - f ~ I} - {p - f ~ I}.
10.1S. Prove that the absolutely convex hull of a totally bounded subset of a locally
convex space is also totally bounded.
10.19. Establish that barreledness is preserved under passage to the inductive limit.
What happens with other linear topological properties?
Chapter 11
Banach Algebras
LA:=L:aI--+L a (aEA)
214 Chapter 11. Banach Algebras
isomorphism between the algebras A and L(A). The same natural terminology is
used for studying representations of arbitrary Banach algebras. Observe immedi-
ately that the canonical operator representation of A, in particular, justifies the
use of the symbol A in place of Ae for A E C, where e is the unity of A (cf. 5.6.5).
In other words, henceforth the isometric representation A 1--+ Ae is considered as
the identification of C with the subalgebra Ce of A.
for all b E B.
<I If b:= A - b E Inv (B) then surely b E Inv (A). Whence resB(b) C resA(b);
i.e.,
SpB(b) = C \ resB(b) J C \ resA(b) = SpA(b).
If A E aSpB(b) then b E OInv(B). Therefore, there is a sequence (b n ),
bn E Inv (B), convergent to b. Putting t:= sUPnENllb;lll, deduce that
216 Chapter 11. Banacb Algebras
In other words, if t < +00 then there is a limit a:= lim b;;1 in B. Multiplication
is jointly continuous, and so ab = ba = 1; i.e., bE Inv (B). Since Inv (B) is open
by the Banach Inversion Stability Theorem and 11.1.6, arrive at a contradiction
to the containment bE oInv (B).
r
Therefore, it may be assumed (on dropping to a subsequence, if need be) that
Ilb;;111 ---t +00. Put a n := Ilb;;1 1b;;1. Then
Whence it follows that b is not invertible. Indeed, in the opposite case for a:= -,;-1
it would hold that
Finally, conclude that).. - b does not belong to Inv (A); i.e., ).. E SpA(b). Since
).. is a boundary point of a larger set SpB(b); undoubtedly, ).. E oSpA(b). l>
11.2.5. Corollary. If SpB(b) lac1s,~)nterior points tben SpB(b) = SpA(b).
<l SpB(b) = oSPB(b) C oSPB(b) C oSPA(b) C SPA(b) c SpB(b) l>
=~
27rz
f 1 . f h(z)R(a, z) bdz
h(z)R(a, z)dzb = -2
7rZ
= ~fh(z)R(La,
27rz
1 . fh(Z)R(La, z)dzb
z)bdz = -2
7rZ
= gLah(b)
:j(SP~ §t•
. ~
B(A). L A
:.r~
L(A) L- 1 • A
Once again using the Urysohn Theorem, find h n E C(Q, lR) satisfying 0:::; h n :::; 1
withhnl Un =0 and hnl vn =1. Consider fn:= fh n . Since
::r:;p(a))
11.6.12. Theorem. For an element a of a commutative Banach algebra A
the following diagram of representations commutes:
A ~A. C(X(A), C)
In other words,
--
R(a, z)(X) = -(X) =
1 1
~() = -~(X) = R(a, z)(X).
1
z-a z-ax z-a
Therefore, appealing to the properties of the Bochner integral (cf. 5.5.9 (6)) and
given f E H(Sp(a)), infer that
If
= 27ri f(Z)C§A(R(a, z))dz = If --
27ri f(z)R(a, z))dz
Furthermore, given X E X(A) and using the Cauchy Integral Formula, derive the
following chain of equalities
= ~x
27rl
(f fez) dZ)
Z - a
= f(;)(x) = f(a)(x)· [>
Moreover, a* = x - iy.
<l Only the claim of uniqueness needs examining. If a = Xl + iYI then, using
the properties of involution (cf. 6.4.13), proceed as follows: a* = xi + (iyt)* =
xi - iyr = Xl - iYI. Thus, Xl = X and YI = y. [>
11.7. The Spectrum of an Element of a C*-Algebra 225
11.7.3. The unity of A is a hermitian element of A.
41* = 1*1 = 1*1** = (1*1)* = 1** = 1 t>
11.7.4. a E Inv(A) ¢} a* E Inv(A). Moreover, involution and inversion
commute.
4 For a E Inv(A) by definition aa- l = a-1a = 1. Consequently, a-ha* =
a*a- h = 1*. Using 11.7.3, infer that a* E Inv(A) and a*-l = a-h. Repeating
this argument for a:= a*, complete the proof. t>
11.7.5. Sp(a*) = Sp(a)*. 4t>
11. 7 .6. The spectrum of a unitary element of a C* -algebra is a subset of the
unit circle.
4 By Definition 6.4.13, lIa 2 11 = Ila*all ::; lIa*lIlIall for an arbitrary a. In other
words, lIall ::; lIa*lI· Therefore, from a = a** infer that lIall = lIa*lI. If now a is
a unitary element, a* = a-\ then lIal1 2 = lIa*all = lIa-1all = 1. Consequently,
lIall = lIa*1I = lIa-11l = 1. Whence it follows that Sp(a) and Sp(a- l ) both lie
within the unit disk. Furthermore, Sp(a- l ) = Sp(a)-l. t>
11.7.7. The spectrum of a hermitian element of a C* -algebra is real.
4 Take a in A arbitrarily. From the Gelfand-Dunford Theorem in an algebraic
setting derive
00 n * 00 (n)* 00 *)n
exp(a)* = ( "" ~ ) = ""
L...J n!
_a_ = "" ~ = exp(a*).
L...J n! L...J n!
n=O n=O n=O
If now h = h* is a hermitian element of A then, applying the holomorphic
functional calculus to a:= exp( ih), deduce that
A~B
* 1 1*
A~B
If 9t is also an isomorphism then 9t is a *-isomorphism of A and B. In the
presence of norms in the algebras, the naturally understood terms "isometric
*-representation" and "isometric *-isomorphism" are in common parlance.
11.8. Tbe Commutative Gelfand-Na1mark Tbeorem 227
The above connection between the continuous and holomorphic functional calcu-
luses for a normal element a of the C* -algebra A may be expressed as follows: The
next diagram commutes
sea) = sIB(a) 2: 0 for a 2: O. Thus, Sp(a) C 1R+ =* N(a) C 1R+, which ends the
proof. t>
11.9.4. DEFINITION. An element a of a C* -algebra A is called positive if a
is hermitian and Sp(a) C 1R+. The set of all positive elements of A is denoted
by A+.
11.9.5. In each C*-algebra A the set A+ is an ordering cone.
<l It is clear that N(a+b) C N(a)+N(b) and N(aa) = aN(a) for a, bE A and
a E 1R+. Hence, 11.9.3 ensures the inclusion alA+ + a2A+ C A+ for aI, a2 E 1R+.
Thus, A+ is a cone. If a E A+ n (-A+) then Sp(a) = O. Since a is a hermitian
element, from Theorem 11.8.6 deduce that Iiall = O. I>
11.9.6. To every hermitian element a of a C*-algebra A there correspond
some elements a+ and a_ of A+ such that
<3 By Lemmas 11.9.7 and 11.9.3, s(a*a) 2: 0 for all a E A. Putting a:= a + 1
and a:= a + i, successively infer that
Hence, (., .)" is a hermitian form. Since a*a 2: 0 for a E A in virtue of 11.9.7,
(a, a)" = s(a*a) 2: 0 by 11.9.3. Consequently, (., .)" is a positive-definite hermi-
tian form. r>
11.9.10. GNS-Construction Tbeorem. To every state s of an arbitrary
C*-algebra A there correspond a Hilbert space (H", (., . ),,), an element x" in Ha
and a *-representation vt" : A -+ B(H,,) such that sea) = (vt,,(a)xa, x"),, for all
a E A and the set {vt,,(a)xs : a E A} is dense in Hs.
<3 In virtue of 11.9.9, putting (a, b)s:= s b*a) for a, b E A, obtain a pre-
Hilbert space (A, (., . )B). Let pB(a):= (a, a)B stand for the seminorm of the
space. Assume that 'P" : A -+ AI ker PB is the coset mapping of A onto the
Hausdorff pre-Hilbert space AI ker Ps associated with A. Assume further that
lB : AI ker PB -+ HB is an embedding (for instance, by the double prime map-
ping) of AI ker PB onto a dense subspace of the Hilbert space HB associated with
(A, (., . ),,) (cf. 6.1.10 (4)). The inner product in H" retains the previous notation
( ., . )B. Therefore, in particular,
i.e., Lab E ker Ps. Uniqueness for La is provided by 2.3.9, since 'Ps is an epimor-
phism. Observe also that 'Ps is an open mapping (d. 5.1.3). Whence the continuity
of La is immediate. Therefore, in virtue of 5.3.8 the correspondence ts 0 La 0 (ts)-l
may be considered as a bounded linear operator from ts(A/ ker Ps) to the Banach
space Hs. By 4.5.10 such an operator extends uniquely to an operator ~s(a)
in B(Hs).
Demonstrate now that ~s : a I-t ~s (a) is a sought representation. By 11.1.6,
Lab = LaLb for a, b E A. Consequently,
Since Lab is a unique solution to the equation :t'Ps = 'PsLab, infer the equality
Lab = LaLb, which guarantees multiplicativity for ~s. The linearity of ~s may
be verified likewise. Furthermore,
i.e., = 1.
~s(l)
For simplicity, put 'l/Js := ts'Ps. Then, on account taken of the definition
of the inner product on Hs (cf. 6.1.10 (4)) and the involution in B(Hs) (d. 6.4.14
and 6.4.5), given elements a, b, and y in A, infer that
Now, since im tPs is dense in Hs it follows that 9l s(a*) = 9l s(a)* for all a in Aj
i.e., 9l s is a *-representation.
Let Xs:= tPs1. Then
H:= EB
sES(A)
Hs={h:=(hS)SES(A)E II
8ES(A)
Hs: L
sES(A)
IIhsllk.<+OO}.
(h, g) = L (h s, gs)s.
sES(A)
Whence it follows that the expression 9l(a)h : S 1-+ 9l s(a)h s defines an el-
ement 9l(a)h of H. The resulting operator 9l(a) : h 1-+ 9l(a)h is a member
of B(H). Moreover, the mapping 9l : a 1-+ 9l(a) (a E A) is a sought isometric
*-representation of A.
234 Chapter 11. Banach Algebras
Since hand g in H are arbitrary, conclude that 9l( a*) = 9l( a)* .
It remains to establish only that the *-representation 9l is an isometry, i.e. the
equalities 119l(a)1I = lIall for all a E A. First, assume a positive. From the Spectral
Theorem and the Weierstrass Theorem it follows that lIall E Sp(a). In virtue
of 11.9.3 (1) there is a state s E S(A) such that s(a) = Iiali. Using the properties
ofthe vector Xs corresponding to the *-representation 9l s (d. 11.9.10) and applying
the Cauchy-Bunyakovski'l-Schwarz inequality, infer that
Exercises
11.1. Give examples of Banach algebras and non-Banach algebras.
11.2. Let A be a Banach algebra. Take X E A# such that xU) = 1 and x(Inv (A)) C
Inv (C). Prove that X is multiplicative and continuous.
11.3. Let the spectrum Sp(a) of an element a of a Banach algebra A lie in an open
set U. Prove that there is a number e > 0 such that Sp(a + b) C U for all b E A satisfying
Ilbll ~ e.
Exercises 235
11.4. Describe the maximal ideal spaces of the algebras C(Q, iC) and C(1)([O, 1], iC)
with pointwise multiplication, and of the algebra of two-way infinite sum mabie sequences /1 (Z)
with multiplication
00
(a*b)(n):= L an_kbk·
k=-oo
11.5. Show that a member T of the endomorphism algebra B(X) of a Banach space X
has a left inverse if and only if T is a monomorphism and the range of T is complemented in X.
11.6. Show that a member T of the endomorphism algebra B(X) of a Banach space X
has a right inverse if and only if T is an epimorphism and the kernel of T is complemented in X.
11.7. Assume that a Banach algebra A has an element with disconnected spectrum
(having a proper clop en part). Prove that A has a nontrivial idempotent.
11.8. Let A be a unital commutative Banach algebra and let E be some set of maximal
ideals of A. Such a set E is a boundary of A if 11;:J1100 = sup I;:J(E) 1for all a E A. Prove that the
intersection of all closed boundaries of A is also a boundary of A. This is the Shilov boundary
of A.
11.9. Let A and B be unital commutative Banach algebras, with B C A and 1B = 1A.
Prove that each maximal ideal of the Shilov boundary of B lies in some maximal ideal of A.
11.10. Let A and B be unital CO-algebras and let T be a morphism from A to B.
Assume further that a is a normal element of A and f is a continuous function on SPA (a).
Demonstrate that SPB(Ta) C SPA(a) and Tf(a) = f(Ta).
11.11. Let f E A', with A a commutative CO-algebra. Show that f is a positive form
(i.e., f(a'a) ~ 0 for a E A) if and only if IIfll = f(l).
11.12. Describe extreme rays of the set of positive forms on a commutative CO-algebra.
11.13. Prove that the algebras C( Ql, iC) and C( Q2, iC), with Ql and Q2 compact, are
isomorphic if and only if Ql and Q2 are homeomorphic.
11.14. Let a normal element a of a CO-algebra has real spectrum. Prove that a is
hermitian.
11.15. Using the continuous functional calculus, develop a spectral theory for normal
operators in a Hilbert space. Describe compact normal operators.
11.16. Let T be an algebraic morphism between CO-algebras, and IITII :c::; 1. Then
T(a') = (Ta)' for all a.
11.17. Let T be a normal operator in a Hilbert space H. Show that there are a hermitian
operator S in H and a continuous function f : SpeS) -> rc such that T = f(S). Is an analogous
assertion valid in CO-algebras?
11.18. Let A and B be CO-algebras and let p be a *-monomorphism from A to B. Prove
that p is an isometric embedding of A into B.
11.19. Let a and b be hermitian elements of a CO-algebra A. Assume that ab = ba and,
moreover, a :c::; b. Prove that f(a) :c::; feb) for (suitable restrictions of) every increasing continuous
scalar function f over R
References
17. Approximation of Hilbert Space Operators, Pitman, Boston etc. Vol. 1: Her-
rero D. A. (1982); Vol. 2: Apostol C. et al. (1984).
18. Arkhangel'skil A. V., Topological Function Spaces [in Russian], Moscow Uni-
versity Publ. House, Moscow (1989).
19. Arkhangel'skil A. V. and Ponomarev V. 1., Fundamentals of General Topology
in Problems and Exercises [in Russian], Nauka, Moscow (1974).
20. Arveson W., An Invitation to C*-Algebra, Springer-Verlag, Berlin etc. (1976).
21. Aubin J.-P., Applied Abstract Analysis, Wiley-Interscience, New York (1977).
22. Aubin J.-P., Applied Functional Analysis, Wiley-Interscience, New York
(1979).
23. Aubin J.-P., Mathematical Methods of Game and Economic Theory, North-
Holland, Amsterdam (1979).
24. Aubin J.-P., Nonlinear Analysis and Motivations from Economics [in French],
Masson, Paris (1984).
25. Aubin J.-P., Optima and Equilibria. An Introduction to Nonlinear Analysis,
Springer-Verlag, Berlin etc. (1993).
26. Aubin J.-P. and Ekeland 1., Applied Nonlinear Analysis, Wiley-Interscience,
New York etc. (1984).
27. Aubin J.-P. and Frankowska H., Set-Valued Analysis. Systems and Control,
Birkhauser, Boston (1990).
28. Baggett L. W., Functional Analysis. A Primer, Dekker, New York etc. (1991).
29. Baggett L. W., Functional Analysis, Dekker, New York etc. (1992).
30. Baiocchi C. and Capelo A., Variational and Quasivariational Inequalities. Ap-
plication to Free Boundary Problems [in Italian], Pit agora Editrice, Bologna
(1978).
31. Balakrishnan A. V., Applied Functional Analysis, Springer-Verlag, New York
etc. (1981).
32. Banach S., Theorie des Operationes Lineares, Monografje Mat., Warsaw
(1932).
33. Banach S., Theory of Linear Operations, North-Holland, Amsterdam (1987).
34. Bauer H., Probability Theory and Elements of Measure Theory, Academic
Press, New York etc. (1981).
35. Beals R., Advanced Mathematical Analysis, Springer-Verlag, New York etc.
(1973).
36. Beauzamy B., Introduction to Banach Spaces and Their Geometry, North-
Holland, Amsterdam etc. (1985).
37. Berberian St., Lectures in Functional Analysis and Operator Theory, Springer-
Verlag, Berlin etc. (1974).
38. Berezanskil Yu. M. and Kondrat'ev Yu. G., Spectral Methods in Infinite-
Dimensional Analysis [in Russian], Naukova Dumka, Kiev (1988).
References 239
39. Berezanskil Yu. M., Us G. F., and Sheftel' Z. G., Functional Analysis. A Lec-
ture Course [in Russian], Vishcha Shkola, Kiev (1990).
40. Berg J. and Lofstrom J., Interpolation Spaces. An Introduction, Springer-
Verlag, Berlin etc. (1976).
41. Berger M., Nonlinearity and Functional Analysis, Academic Press, New York
(1977).
42. Besov O. V., Il'in V. P., and Nikol'skil S. M., Integral Representations and
Embedding Theorems [in Russian], Nauka, Moscow (1975).
43. Bessaga Cz. and Pelczynski A., Selected Topics in Infinite-Dimensional To-
pology, Polish Scientific Publishers, Warsaw (1975).
44. Birkhoff G., Lattice Theory, Amer. Math. Soc., Providence (1967).
45. Birkhoff G. and Kreyszig E., "The establishment of functional analysis," His-
toria Math., 11, No.3, 258-321 (1984).
46. Birman M. Sh. et al., Functional Analysis [in Russian], Nauka, Moscow
(1972).
47. Birman M. Sh. and Solomyak M. Z., Spectral Theory of Selfadjoint Operators
in Hilbert Space [in Russian), Leningrad University Publ. House, Leningrad
(1980).
48. Boccara N., Functional Analysis. An Introduction for Physicists, Academic
Press, New York etc. (1990).
49. Bogolyubov N. N., Logunov A. A., Oksak A. I., and Todorov I. T., General
Principles of Quantum Field Theory [in Russian], Nauka, Moscow (1987).
50. Bollobas B., Linear Analysis. An Introductory Course, Cambridge University
Press, Cambridge (1990).
51. Bonsall F. F. and Duncan J., Complete Normed Algebras, Springer-Verlag,
Berlin etc. (1973).
52. Boos B. and Bleecker D., Topology and Analysis. The Atiyah-Singer Index
Formula and Gauge-Theoretic Physics, Springer-Verlag, Berlin etc. (1985).
53. Bourbaki N., Set Theory [in French], Hermann, Paris (1958).
54. Bourbaki N., General Topology. Parts 1 and 2, Addison-Wesley, Reading
(1966).
55. Bourbaki N., Integration. Ch. 1-8 [in French), Hermann, Paris (1967).
56. Bourbaki N., Spectral Theory [in French), Hermann, Paris (1967).
57. Bourbaki N., General Topology. Ch. 5-10 [in French), Hermann, Paris (1974).
58. Bourbaki N., Topological Vector Spaces, Springer-Verlag, Berlin etc. (1981).
59. Bourbaki N., Commutative Algebra, Springer-Verlag, Berlin etc. (1989).
60. Bourgain J., New Classes of 2'P-Spaces, Springer-Verlag, Berlin etc. (1981).
61. Bourgin R. D., Geometric Aspects of Convex Sets with the Radon-Nikodym
Property, Springer-Verlag, Berlin etc. (1983).
62. Bratteli O. and Robertson D., Operator Algebras and Quantum Statistical
Mechanics. Vol. 1, Springer-Verlag, New York etc. (1982).
240 References
84. Dautray R and Lions J.-L., Mathematical Analysis and Numerical Methods
for Science and Technology. Vol. 2 and 3, Springer-Verlag, Berlin etc. (1988,
1990).
85. Day M., Normed Linear Spaces, Springer-Verlag, Berlin etc. (1973).
86. De Branges L., "The Stone-Weierstrass theorem," Proc. Amer. Math. Soc.,
10, No.5, 822-824 (1959).
87. De Branges L. and Rovnyak J., Square Summable Power Series, Holt, Rine-
hart and Winston, New York (1966).
88. Deimling K., Nonlinear Functional Analysis, Springer-Verlag, Berlin etc.
(1985).
89. DeVito C. L., Functional Analysis, Academic Press, New York etc. (1978).
90. De Wilde M., Closed Graph Theorems and Webbed Spaces, Pitman, London
(1978).
91. Diestel J., Geometry of Banach Spaces - Selected Topics, Springer-Verlag,
Berlin etc. (1975).
92. Diestel J., Sequences and Series in Banach Spaces, Springer-Verlag, Berlin
etc. (1984).
93. Diestel J. and Uhl J. J., Vector Measures, Amer. Math. Soc., Providence
(1977).
94. Dieudonne J., Foundations of Modern Analysis, Academic Press, New York
(1969).
95. Dieudonne J., A Panorama of Pure Mathematics. As Seen by N. Bourbaki,
Academic Press, New York etc. (1982).
96. Dieudonne J., History of Functional Analysis, North-Holland, Amsterdam
etc. (1983).
97. Dinculeanu N., Vector Measures, Verlag der Wissenschaften, Berlin (1966).
98. Dixmier J., C*-Algebras and Their Representations [in French], Gauthier-
Villars, Paris (1964).
99. Dixmier J., Algebras of Operators in Hilbert Space (Algebras of Von Neu-
mann) [in French], Gauthier-Villars, Paris (1969).
100. Donoghue W. F. Jr., Distributions and Fourier Transforms, Academic Press,
New York etc. (1969).
101. Doob J., Measure Theory, Springer-Verlag, Berlin etc. (1993).
102. Dorau Rand Belfi V., Characterizations of C*-Algebras. The Gelfand-
Nalmark Theorem, Dekker, New York and Basel (1986).
103. Dowson H. R, Spectral Theory of Linear Operators, Academic Press, London
etc. (1978).
104. Dugundji J., Topology, Allyn and Bacon, Boston (1966).
105. Dunford N. and Schwartz G. (with the assistance ofW. G. Bade and R G. Bar-
tle), Linear Operators. Vol. 1: General Theory, Interscience, New York
(1958).
242 References
106. Edmunds D. E. and Evans W. D., Spectral Theory and Differential Operators,
Clarendon Press, Oxford (1987).
107. Edwards R, Functional Analysis. Theory and Applications, Holt, Rinehart
and Winston, New York etc. (1965).
108. Edwards R, Fourier Series. A Modern Introduction. Vol. 1 and 2, Springer-
Verlag, New York etc. (1979).
109. Efimov A. V., Zolotarev Yu. G., and Terpigorev V. M., Mathematical Analysis
(Special Sections). Vol. 2: Application of Some Methods of Mathematical and
Functional Analysis [in Russian], Vysshaya Shkola, Moscow (1980).
110. Ekeland 1. and Temam R, Convex Analysis and Variational Problems, North-
Holland, Amsterdam (1976).
111. Emch G., Algebraic Methods in Statistic Mechanics and Quantum Field The-
ory, Wiley-Interscience, New York etc. (1972).
112. Enflo P., "A counterexample to the approximation property in Banach
spaces," Acta Math., 130, No. 3-4, 309-317 (1979).
113. Engelkind R, General Topology, Springer-Verlag, Berlin etc. (1985).
114. Erdelyi 1. and Shengwang W., A Local Spectral Theory for Closed Operators,
Cambridge University Press, Cambridge (1985).
115. Faris W. G., Selfadjoint Operators, Springer-Verlag, Berlin etc. (1975).
116. Fenchel W., Convex Cones, Sets and Functions, Princeton University Press,
Princeton (1953).
117. Fenchel W., "Convexity through ages," in: Convexity and Its Applications,
Birkhauser, Basel etc., 1983, pp. 120-130.
118. Floret K., Weakly Compact Sets, Springer-Verlag, Berlin etc. (1980).
119. Folland G. B., Fourier Analysis and Its Applications, Wadsworth and Brooks,
Pacific Grove (1992).
120. Friedrichs K. 0., Spectral Theory of Operators in Hilbert Space, Springer-
Verlag, New York etc. (1980).
121. Gamelin T., Uniform Algebras, Prentice-Hall, Englewood Cliffs (1969).
122. Gelbaum B. and Olmsted G., Counterexamples in Analysis, Holden-Day, San
Francisco (1964).
123. Gelfand 1. M., Lectures on Linear Algebra [in Russian], Nauka, Moscow
(1966).
124. Gelfand 1. M. and Shilov G. E., Generalized Functions and Operations over
Them [in Russian], Fizmatgiz, Moscow (1958).
125. Gelfand 1. M. and Shilov G. E., Spaces of Test and Generalized Functions
[in Russian], Fizmatgiz, Moscow (1958).
126. Gelfand 1. M., Ralkov D. A., and Shilov G. E., Commutative Normed Rings,
Chelsea Publishing Company, New York (1964).
127. Gelfand 1. M. and Vilenkin N. Ya., Certain Applications of Harmonic Anal-
ysis. Rigged Hilbert Spaces, Academic Press, New York (1964).
References 243
150. Havin V. P., "Methods and structure of commutative harmonic analysis," in:
Current Problems of Mathematics. Fundamental Trends. Vol. 15 [in Russian],
VINITI, Moscow, 1987, pp. 6-133.
151. Havin V. P. and Nikol'ski'l N. K. (eds.), Linear and Complex Analysis Problem
Book 3. Parts 1 and 2, Springer-Verlag, Berlin etc. (1994).
152. Helmberg G., Introduction to Spectral Theory in Hilbert Space, North-Hol-
land, Amsterdam etc. (1969).
153. Herve M., The Fourier Transform and Distributions [in French], Presses Uni-
versitaires de France, Paris (1986).
154. Heuser H., Functional Analysis, Wiley, New York (1982).
155. Heuser H., Functional Analysis [in German]' Teubner, Stuttgart (1986).
156. Hewitt E. and Ross K. A., Abstract Harmonic Analysis. Vol. 1 and 2,
Springer-Verlag, New York (1994).
157. Hewitt E. and Stromberg K., Real and Abstract Analysis, Springer-Verlag,
Berlin etc. (1975).
158. Heyer H., Probability Measures on Locally Compact Groups, Springer-Verlag,
Berlin etc. (1977).
159. Hille E. and Phillips R., Functional Analysis and Semi groups , Amer. Math.
Soc., Providence (1957).
160. Hochstadt H., "Edward Helly, father of the Hahn-Banach theorem," Math.
Intelligencer, 2, No.3, 123-125 (1980).
161. Hoffman K., Banach Spaces of Analytic Functions, Prentice-Hall, Englewood
Cliffs (1962).
162. Hoffman K., Fundamentals of Banach Algebras, University do Parana, Curi-
taba (1962).
163. Hog-Nlend H., Bornologies and Functional Analysis, North-Holland, Amster-
dam etc. (1977).
164. Holmes R. B., Geometric Functional Analysis and Its Applications, Springer-
Verlag, Berlin etc. (1975).
165. Hormander L., Introduction to Complex Analysis in Several Variables, D. Van
Nostrand Company Inc., Princeton (1966).
166. Hormander L., The Analysis of Linear Differential Equations. Vol. 1, Springer-
Verlag, New York etc. (1983).
167. Horn R. and Johnson Ch., Matrix Analysis, Cambridge University Press,
Cambridge etc. (1986).
168. Horvath J., Topological Vector Spaces and Distributions. Vol. 1, Addison-
Wesley, Reading (1966).
169. Husain T., The Open Mapping and Closed Graphs Theorems in Topological
Vector Spaces, Clarendon Press, Oxford (1965).
170. Husain T. and Khaleelulla S. M., Barrelledness in Topological and Ordered
Vector Spaces, Springer-Verlag, Berlin etc. (1978).
References 245
171. Hutson W. and Pym G., Applications of Functional Analysis and Operator
Theory, Academic Press, London etc. (1980).
172. Ioffe A. D. and Tikhomirov V. M., Theory of Extremal Problems, North-
Holland, Amsterdam (1979).
173. Istratescu V. I., Inner Product Structures, Reidel, Dordrecht and Boston
(1987).
174. James R. C., "Some interesting Banach spaces," Rocky Mountain J. Math.,
23, No.2. 911-937 (1993).
175. Jameson G. J. 0., Ordered Linear Spaces, Springer-Verlag, Berlin etc. (1970).
176. Jarchow H., Locally Convex Spaces, Teubner, Stuttgart (1981).
177. Jones D. S., Generalized Functions, McGraw-Hill Book Co., New York etc.
(1966).
178. Jonge De and Van Rooij A. C. M., Introduction to Riesz Spaces, Mathema-
tisch Centrum, Amsterdam (1977).
179. Jorgens K., Linear Integral Operators, Pitman, Boston etc. (1982).
180. Kadison R. V. and Ringrose J. R., Fundamentals of the Theory of Operator
Algebras. Vol. 1 and 2, Academic Press, New York (1983, 1986).
181. Kahane J .-P., Absolutely Convergent Fourier Series [in French], Springer-
Verlag, Berlin etc. (1970).
182. Kamthan P. K. and Gupta M., Sequence Spaces and Series, Dekker, New
York and Basel (1981).
183. Kantorovich L. V. and Akilov G. P., Functional Analysis in Normed Spaces,
Pergamon Press, Oxford etc. (1964).
184. Kantorovich L. V. and Akilov G. P., Functional Analysis, Pergamon Press,
Oxford and New York (1982).
185. Kantorovich L. V., Vulikh B. Z., and Pinsker A. G., Functional Analysis
in Semiordered Spaces [in Russian], Gostekhizdat, Moscow and Leningrad
(1950).
186. Kashin B. S. and Saakyan A. A., Orthogonal Series [in Russian], Nauka,
Moscow (1984).
187. Kato T., Perturbation Theory for Linear Operators, Springer-Verlag, Berlin
etc. (1995).
188. Kelly J. L., General Topology, Springer-Verlag, New York etc. (1975).
189. Kelly J. L. and Namioka I., Linear Topological Spaces, Springer-Verlag, Berlin
etc. (1976).
190. Kelly J. L. and Srinivasan T. P., Measure and Integral. Vol. 1, Springer-
Verlag, New York etc. (1988).
191. Kesavan S., Topics in Functional Analysis and Applications, Wiley, New York
etc. (1989).
192. Khaleelulla S. M., Counterexamples in Topological Vector Spaces, Springer-
Verlag, Berlin etc. (1982).
246 References
193. Khelemskil A. Ya., Banach and Polynormed Algebras: General Theory, Rep-
resentation and Homotopy [in Russian], Nauka, Moscow (1989).
194. Kirillov A. A., Elements of Representation Theory [in Russian], Nauka, Mos-
cow (1978).
195. Kirillov A. A. and Gvishiani A. D., Theorems and Problems of Functional
Analysis [in Russian], Nauka, Moscow (1988).
196. Kislyakov S. V., "Regular uniform algebras are not complemented," Dokl.
Akad. Nauk SSSR, 304, No.1, 795-798 (1989).
197. Knyazev P. N., Functional Analysis [in Russian]' VysheIshaya Shkola, Minsk
(1985).
198. Kollatz L., Functional Analysis and Numeric Mathematics [in German]'
Springer-Verlag, Berlin etc. (1964).
199. Kolmogorov A. N., Selected Works. Mathematics and Mechanics [in Russian],
Nauka, Moscow (1985).
200. Kolmogorov A. N. and Fomin S. V., Elements of Function Theory and Func-
tional Analysis [in Russian], Nauka, Moscow (1989).
201. Korner T. W., Fourier Analysis, Cambridge University Press, Cambridge
(1988).
202. Korotkov V. B., Integral Operators [in Russian], Nauka, Novosibirsk (1983).
203. Kothe G., Topological Vector Spaces. Vol. 1 and 2, Springer-Verlag, Berlin
etc. (1969, 1980).
204. Kostrikin A. I. and Manin Yu. I., Linear Algebra and Geometry [in Russian],
Moscow University Publ. House, Moscow (1986).
205. Krasnosel'skil M. A., Positive Solutions to Operator Equations. Chapters
of Nonlinear Analysis [in Russian], Fizmatgiz, Moscow (1962).
206. Krasnosel'skil M. A., Lifshits E. A., and Sobolev A. V., Positive Linear Sys-
tems. The Method of Positive Operators [in Russian], Nauka, Moscow (1985).
207. Krasnosel'skil M. A. and Rutitskil Ya. B., Convex Functions and Orlicz
Spaces, Noordhoff, Groningen (1961).
208. Krasnosel'skil M. A. and Zabrelko P. P., Geometric Methods of Nonlinear
Analysis [in Russian]' Nauka, Moscow (1975).
209. Krasnosel'skil M. A. et al., Integral Operators in the Spaces of Integrable
Functions [in Russian], Nauka, Moscow (1966).
210. KreIn S. G., Linear Differential Equations in Banach Space [in Russian],
Nauka, Moscow (1967).
211. KreIn S. G., Linear Equations in Banach Space [in Russian]' Nauka, Moscow
(1971).
212. KreIn S. G., Petunin Yu. I., and Semenov E. M., Interpolation of Linear
Operators [in Russian], Nauka, Moscow (1978).
213. Kreyszig E., Introductory Functional Analysis with Applications, Wiley, New
York (1989).
References 247
257. Morris S., Pontryagin Duality and the Structure of Locally Compact Abelian
Groups, Cambridge University Press, Cambridge (1977).
258. Motzkin T. S., "Endovectors in convexity," in: Proc. Sympos. Pure Math.,
7, Amer. Math. Soc., Providence, 1963, pp. 361-387.
259. Na'lmark M. A., Normed Rings, Noordhoff, Groningen (1959).
260. Napalkov V. V., Convolution Equations in Multidimensional Spaces [in Rus-
sian], Nauka, Moscow (1982).
261. Narici L. and Beckenstein E., Topological Vector Spaces, Dekker, New York
(1985).
262. Naylor A. and Sell G., Linear Operator Theory in Engineering and Science,
Springer-Verlag, Berlin etc. (1982).
263. Neumann J. von, Mathematical Foundations of Quantum Mechanics [in Ger-
man], Springer-Verlag, Berlin (1932).
264. Neumann J. von, Collected Works, Pergamon Press, Oxford (1961).
265. Neveu J., Mathematical Foundations of Probability Theory [in French], Mas-
son et Cie, Paris (1964).
266. Nikol'ski'l N. K., Lectures on the Shift Operator [in Russian], Nauka, Moscow
(1980).
267. Nikol'ski'l S. M., Approximation to Functions in Several Variables and Em-
bedding Theorems [in Russian], Nauka, Moscow (1977).
268. Nirenberg L., Topics in Nonlinear Functional Analysis. 1973-1974 Notes
by R. A Artino, Courant Institute, New York (1974).
269. Oden J. T., Applied Functional Analysis. A First Course for Students of Me-
chanics and Engineering Science, Prentice-Hall, Englewood Cliffs (1979).
270. Palais R. (with contributions by M. F. Atiyah et al.), Seminar on the Atiyah-
Singer Index Theorem, Princeton University Press, Princeton (1965).
271. Palamodov V. P., Linear Differential Operators with Constant Coefficients,
Springer-Verlag, Berlin etc. (1970).
272. Pedersen G. K., Analysis Now, Springer-Verlag, New York etc. (1989).
273. Pelczynski A., Banach Spaces of Analytic Functions and Absolutely Summing
Operators, Amer. Math. Soc., Providence (1977).
274. Petunin Yu. I. and Plichko A. N., Theory of Characteristics of Subspaces and
Its Applications [in Russian], Vishcha Shkola, Kiev (1980).
275. Phelps R., Convex Functions, Monotone Operators and Differentiability,
Springer-Verlag, Berlin etc. (1989).
276. Pietsch A., Nuclear Locally Convex Spaces [in German]' Akademie-Verlag,
Berlin (1967).
277. Pietsch A., Operator Ideals, VEB Deutschen Verlag der Wissenschaften,
Berlin (1978).
278. Pietsch A., Eigenvalues and S-Numbers, Akademie-Verlag, Leipzig (1987).
250 References
279. Pisier G., Factorization of Linear Operators and Geometry of Banach Spaces,
Amer. Math. Soc., Providence (1986).
280. Plesner A. I., Spectral Theory of Linear Operators [in Russian), Nauka, Mos-
cow (1965).
281. Prossdorf S., Some Classes of Singular Equations [in German)' Akademie-
Verlag, Berlin (1974).
282. Radjavi H. and Rosenthal P., Invariant Subspaces, Springer-Verlag, Berlin
etc. (1973).
283. Radyno Ya. V., Linear Equations and Bornology [in Russian), The Belorus-
sian State University Publ. House, Minsk (1982).
284. Ralkov D. A., Vector Spaces [in Russian), Fizmatgiz, Moscow (1962).
285. Reed M. and Simon B., Methods of Modern Mathematical Physics, Academic
Press, New York and London (1972).
286. Reshetnyak Yu. G., Vector Measures and Some Questions of the Theory
of Functions of a Real Variable [in Russian], Novosibirsk University Publ.
House, Novosibirsk (1982).
287. Richards J. Ian and Joun H. K., Theory of Distributions: a Nontechnical
Introduction, Cambridge University Press, Cambridge (1990).
288. Richtmyer R., Principles of Advanced Mathematical Physics. Vol. 1, Springer-
Verlag, New York etc. (1978).
289. Rickart Ch., General Theory of Banach Algebras, D. Van Nostrand Company
Inc., Princeton (1960).
290. Riesz F. and Szokefalvi-Nagy B., Lectures on Functional Analysis [in French],
Akademi Kiad6, Budapest (1972).
291. Robertson A. and Robertson V., Topological Vector Spaces, Cambridge Uni-
versity Press, Cambridge (1964).
292. Rockafellar R., Convex Analysis, Princeton University Press, Princeton
(1970).
293. Rolewicz S., Functional Analysis and Control Theory [in Polish), Panstwowe
Wydawnictwo Naukowe, Warsaw (1977).
294. Rolewicz S., Metric Linear Spaces, Reidel, Dordrecht etc. (1984).
295. Roman St., Advanced Linear Algebra, Springer-Verlag, Berlin etc. (1992).
296. Rudin W., Fourier Analysis on Groups, Interscience, New York (1962).
297. Rudin W., Functional Analysis, McGraw-Hill Book Co., New York (1973).
298. Sadovnichil V. A., Operator Theory [in Russian], Moscow University Publ.
House, Moscow (1986).
299. Sakai S., C*-Algebras and W*-Algebras, Springer-Verlag, Berlin etc. (1971).
300. Samuelides M. and Touzillier L., Functional Analysis [in French], Toulouse,
Cepadues Editions (1983).
301. Sard A., Linear Approximation, Amer. Math. Soc., Providence (1963).
References 251
302. Schaefer H. H., Topological Vector Spaces, Springer-Verlag, New York etc.
(1971).
303. Schaefer H. H., Banach Lattices and Positive Operators, Springer-Verlag,
Berlin etc. (1974).
304. Schapira P., Theory of Hyperfunctions [in French], Springer-Verlag, Berlin
etc. (1970).
305. Schechter M., Principles of Functional Analysis, Academic Press, New York
etc. (1971).
306. Schwartz L. (with participation of Denise Huet), Mathematical Methods for
Physical Sciences [in French], Hermann, Paris (1961).
307. Schwartz L., Theory of Distributions [in French], Hermann, Paris (1966).
308. Schwartz L., Analysis. Vol. 1 [in French], Hermann, Paris (1967).
309. Schwartz L., Analysis: General Topology and Functional Analysis [in French],
Hermann, Paris (1970).
310. Schwartz L., Hilbertian Analysis [in French], Hermann, Paris (1979).
311. Schwartz L., "Geometry and probability in Banach spaces," Bull. Amer.
Math. Soc., 4, No.2. 135-141 (1981).
312. Schwarz H.-U., Banach Lattices and Operators, Teubner, Leipzig (1984).
313. Segall. and Kunze R., Integrals and Operators, Springer-Verlag, Berlin etc.
(1978).
314. Semadeni Zb., Banach Spaces of Continuous Functions, Warsaw, Polish Sci-
entific Publishers (1971).
315. Shafarevich I. R., Basic Concepts of Algebra [in Russian], VINITI, Moscow
(1986).
316. Shilov G. E., Mathematical Analysis. Second Optional Course [in Russian],
Nauka, Moscow (1965).
317. Shilov G. E. and Gurevich B. L., Integral, Measure, and Derivative [in Rus-
sian], Nauka, Moscow (1967).
318. Sinclair A., Automatic Continuity of Linear Operators, Cambridge University
Press, Cambridge (1976).
319. Singer I., Bases in Banach Spaces. Vol. 1 and 2, Springer-Verlag, Berlin etc.
(1970,1981).
320. Smirnov V. I., A Course of Higher Mathematics. Vol. 5, Pergamon Press,
New York (1964).
321. Sobolev S. L., Selected Topics of the Theory of Function Spaces and Gener-
alized Functions [in Russian], Nauka, Moscow (1989).
322. Sobolev S. L., Applications of Functional Analysis in Mathematical Physics,
Amer. Math. Soc., Providence (1991).
323. Sobolev S. L., Cubature Formulas and Modern Analysis, Gordon and Breach,
Montreux (1992).
252 References
324. Steen L. A., "Highlights in the history of spectral theory," Amer. Math.
Monthly, 80, No.4., 359-381 (1973).
325. Steen L. A. and Seebach J. A., Counterexamples in Topology, Springer-Verlag,
Berlin etc. (1978).
326. Stein E. M., Singular Integrals and Differentiability Properties of Functions,
Princeton University Press, Princeton (1971).
327. Stein E. M., Harmonic Analysis, Real-Variable Methods, Orthogonality, and
Oscillatory Integrals, Princeton University Press, Princeton (1993).
328. Stein E. M. and Weiss G., Introduction to Harmonic Analysis on Euclidean
Spaces, Princeton University Press, Princeton (1970).
329. Stone M., Linear Transformations in Hilbert Space and Their Application
to Analysis, Amer. Math. Soc., New York (1932).
330. Sundaresan K. and Swaminathan Sz., Geometry and Nonlinear Analysis in
Banach Spaces, Berlin etc. (1985).
331. Sunder V. S., An Invitation to Von Neumann Algebras, Springer-Verlag, New
York etc. (1987).
332. Swartz Ch., An Introduction to Functional Analysis, Dekker, New York etc.
(1992).
333. Szankowski A., "B(H) does not have the approximation property," Acta
Math., 147, No. 1-2,89-108 (197«;l)"
334. Takeuti G. and Zaring W., Introduction to Axiomatic Set Theory, Springer-
Verlag, New York etc. (1982).
335. Taldykin A. T., Elements of Applied Functional Analysis [in Russian], Vys-
shaya Shkola, Moscow (1982).
336. Taylor A. E. and Lay D. C., Introduction to Functional Analysis, Wiley, New
York (1980).
337. Taylor J. L., Measure Algebras, Amer. Math. Soc., Providence (1973).
338. Tiel J. van, Convex Analysis. An Introductory Theory, Wiley, Chichester
(1984).
339. Tikhomirov V. M., "Convex analysis," in: Current Problems of Mathematics.
Fundamental Trends. Vol. 14 [in Russian], VINITI, Moscow, 1987, pp. 5-101.
340. Trenogin V. A., Functional Analysis [in Russian], Nauka, Moscow (1980).
341. Trenogin V. A., Pisarevskil B. M., and Soboleva T. S., Problems and Exercises
in Functional Analysis [in Russian], Nauka, Moscow (1984).
342. Treves F., Locally Convex Spaces and Linear Partial Differential Equations,
Springer-Verlag, Berlin etc. (1967).
343. Triebel H., Theory of Function Spaces, Birkhauser, Basel (1983).
344. Uspenskil S. V., Demidenko G. V., and PerepeIkin V. G., Embedding Theo-
rems and Applications to Differential Equations [in Russian], Nauka, Novosi-
birsk (1984).
References 253
367. Young L., Lectures on the Calculus of Variations and Optimal Control Theory,
W. B. Saunders Company, Philadelphia etc. (1969).
368. Zaanen A. C., Riesz Spaces. Vol. 2, North-Holland, Amsterdam etc. (1983).
369. Zemanian A. H., Distribution Theory and Transform Analysis, Dover, New
York (1987).
370. Ziemer W. P., Weakly Differentiable Functions. Sobolev Spaces and Functions
of Bounded Variation, Springer-Verlag, Berlin etc. (1989).
371. Zimmer R. J., Essential Results of Functional Analysis, Chicago University
Press, Chicago and London (1990).
372. Zorich V. A., Mathematical Analysis. Part 2 [in Russian], Nauka, Moscow
(1984).
373. Zuily C., Problems in Distributions and Partial Differential Equations, North-
Holland, Amsterdam etc. (1988).
374. Zygmund A., Trigonometric Series. Vol. 1 and 2, Cambridge University Press,
New York (1959).
Notation Index
1. A.A. Hanns and D.R. Wyman: Mathematics and Physics of Neutron Radiography.
1986 ISBN 90-277-2191-2
2. H.A. Mavromatis: Exercises in Quantum Mechanics. A Collection of Illustrative
Problems and Their Solutions. 1987 ISBN 90-277-2288-9
3. V.I. Kukulin, V.M. Krasnopol'sky and J. Horacek: Theory of Resonances. Principles
and Applications. 1989 ISBN 90-277-2364-8
4. M. Anderson and Todd Feil: Lattice-Ordered Groups. An Introduction. 1988
ISBN 90-277-2643-4
5. J. Avery: Hyperspherical Harmonics. Applications in Quantum Theory. 1989
ISBN 0-7923-0165-X
6. H.A. Mavromatis: Exercises in Quantum Mechanics. A Collection of Illustrative
Problems and Their Solutions. Second Revised Edition. 1992 ISBN 0-7923-1557-X
7. G. Micula and P. Pavel: Differential and Integral Equations through Practical
Problems and Exercises. 1992 ISBN 0-7923-1890-0
8. W.S. Anglin: The Queen of Mathematics. An Introduction to Number Theory. 1995
ISBN 0-7923-3287-3
9. Y.G. Borisovich, N.M. Bliznyakov, T.N. Fomenko and Y.A. Izrailevich: Introduction
to Differential and Algebraic Topology. 1995 ISBN 0-7923-3499-X
10. J. Schmeelk, D. Taka~i and A. Takoci: Elementary Analysis through Examples and
Exercises. 1995 ISBN 0-7923-3597-X
11. 1.S. Golan: Foundations of Linear Algebra. 1995 ISBN 0-7923-3614-3