Dynamics of linear operators 1st Edition Frédéric Bayart all chapter instant download
Dynamics of linear operators 1st Edition Frédéric Bayart all chapter instant download
Dynamics of linear operators 1st Edition Frédéric Bayart all chapter instant download
com
https://ebookname.com/product/dynamics-of-linear-
operators-1st-edition-frederic-bayart/
OR CLICK BUTTON
DOWNLOAD EBOOK
https://ebookname.com/product/the-swan-system-of-the-c2-molecule-john-
g-phillips/
ebookname.com
https://ebookname.com/product/warships-in-the-war-of-the-
pacific-1879-83-1st-edition-angus-konstam/
ebookname.com
https://ebookname.com/product/cultural-competence-for-public-managers-
managing-diversity-in-today-s-world-espiridion-borrego/
ebookname.com
A Vineyard in My Glass 1st Edition Gerald Asher
https://ebookname.com/product/a-vineyard-in-my-glass-1st-edition-
gerald-asher/
ebookname.com
https://ebookname.com/product/logical-creative-thinking-methods-1st-
edition-ding/
ebookname.com
https://ebookname.com/product/handling-of-radiation-accident-patients-
by-paramedical-and-hospital-personnel-second-edition-thomas-a-carder/
ebookname.com
https://ebookname.com/product/exploring-religion-in-ancient-egypt-
blackwell-ancient-religions-1st-edition-quirke/
ebookname.com
https://ebookname.com/product/the-italian-fashion-system-1st-edition-
elisabetta-merlo/
ebookname.com
Great Writers on Organizations 3rd Omnibus Edition Derek
S. Pugh And David J. Hickson
https://ebookname.com/product/great-writers-on-organizations-3rd-
omnibus-edition-derek-s-pugh-and-david-j-hickson/
ebookname.com
This page intentionally left blank
CAMBRIDGE TRACTS IN MATHEMATICS
General Editors
B . B O L L O B Á S , W . F U L T O N , A . K A T O K ,
F. K I R WA N , P. S A R N A K , B . S I M O N , B . T O TA R O
F R É D É R I C B A Y A R T
Université de Clermont-Ferrand, France
É T I E N N E M A T H E R O N
Université d’Artois, France
CAMBRIDGE UNIVERSITY PRESS
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore,
São Paulo, Delhi, Dubai, Tokyo
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/9780521514965
© F. Bayart and E. Matheron 2009
Introduction page ix
v
vi Contents
Appendices 310
A Complex analysis 310
B Function spaces 311
C Banach space theory 314
D Spectral theory 316
References 321
Notation 331
Author index 333
Subject index 335
Introduction
Linear dynamics is a young and rapidly evolving branch of functional analysis, which
was probably born in 1982 with the Toronto Ph.D. thesis of C. Kitai [158]. It has
become rather popular, thanks to the efforts of many mathematicians. In particular,
the seminal paper [123] by G. Godefroy and J. H. Shapiro, the authoritative survey
[133] by K.-G. Grosse-Erdmann and the beautiful notes [222] by J. H. Shapiro have
had a considerable influence on both its internal development and its diffusion within
the mathematical community. After more than two decades of active research, this
would seem to be the proper time to write a book about it.
As the name indicates, linear dynamics is mainly concerned with the behaviour
of iterates of linear transformations. On finite-dimensional spaces, things are rather
well understood since linear transformations are completely described by their Jor-
dan canonical form. However, a new phenomenon appears in an infinite-dimensional
setting: linear operators may have dense orbits. In fact, quite a lot of natural operators
have this property.
To settle some terminology, let us recall that if T is a continuous linear operator
acting on some topological vector space X, the T -orbit of a vector x ∈ X is the
set O(x, T ) := {x, T (x), T 2 (x), . . . }. The operator T is said to be hypercyclic
if there exists some vector x ∈ X whose T -orbit is dense in X. Such a vector
x is said to be hypercyclic for T . Hypercyclicity is the main topic of the present
book.
From the definition of hypercyclicity, it is immediately apparent that linear
dynamics lies at the intersection of at least three different domains of mathematics.
1. Topological dynamics The definition of hypercyclicity does not require any
linear structure. It makes sense for an arbitrary continuous map T : X → X acting
on a topological space X and, in fact, continuous maps with dense orbits are in some
sense the main objects of study in topological dynamics.
However, the usual setting of topological dynamics is that of compact topologi-
cal spaces, and compactness is essential at many points in the discussion. In linear
dynamics, the underlying space is never compact or even locally compact because
hypercyclicity turns out to be a purely infinite-dimensional property. Thus, it would
ix
x Introduction
seem to be hard to use sophisticated tools from topological dynamics. Nevertheless,
when the linear structure is added interesting phenomena appear.
2. Operator theory The word “hypercyclic” comes from the much older notion of
a cyclic operator. An operator T ∈ L(X) is said to be cyclic if there exists a vector
x ∈ X such that the linear span of O(x, T ) is dense in X. This notion is of course
related to the famous invariant subspace problem: given an operator T ∈ L(X), is
it possible to find a non-trivial closed subspace F ⊂ X which is T -invariant (i.e. for
which T (F ) ⊂ F )? Here, non-trivial means that F = {0} and F = X. Clearly, the
closed linear span of any T -orbit is an invariant subspace for T ; hence, T lacks non-
trivial invariant closed subspaces if and only if (iff) every non-zero vector x ∈ X is
cyclic for T .
Similarly, the invariant subset problem asks whether any operator T ∈ L(X) has
a non-trivial closed invariant subset. Since the closure of any T -orbit is a T -invariant
closed set, an operator T lacks non-trivial invariant closed sets iff all non-zero vectors
x ∈ X are hypercyclic for T . In the language of topological dynamics, this means
that (X \ {0}, T ) is a minimal dynamical system.
Despite considerable efforts, the invariant subspace problem remains largely open,
most notably for Hilbert space operators. Since P. Enflo’s negative solution on
a rather peculiar Banach space [104], the most impressive achievement has been
C. J. Read’s construction of an operator T on 1 (N) for which every non-zero vector
x ∈ 1 (N) is hypercyclic ([202]). This means that the invariant subset problem has a
negative solution on the space 1 .
3. Universality Let (Ti )i∈I be a family of continuous maps Ti : X → Y between
two fixed topological spaces X and Y . The family (Ti ) is said to be universal if there
exists x ∈ X such that the set {Ti (x); i ∈ I} is dense in Y . The first example of
universality seems to go back to M. Fekete in 1914 (quoted in [191]) who discovered
the existence of a universal Taylor series n≥1 an tn : for any continuous function g
on [−1 , 1] with g(0) = 0, there exists an increasing sequence of integers (nk ) such
nk
that n=1 an tn → g(t) uniformly as k → ∞. (Here X = CN , Y is the space of all
i
continuous functions on [−1, 1] vanishing at 0, and Ti ((an )) = n=1 an tn , i ≥ 1).
Since then, universal families have been exhibited in a huge number of situations;
see [133].
Hypercyclicity is of course a particular instance of universality, in which X = Y
is a topological vector space and (Ti )i∈N is the sequence of iterates of a single linear
operator T ∈ L(X). Nevertheless, it is worth keeping in mind that a number of
results pertaining to hypercyclic operators can be formulated (and proved!) in the
more general setting of universal families. When working with the iterates of an
operator, however, more specific tools can be used. In particular, spectral theory is
often helpful.
One particularly seductive feature of linear dynamics is the diversity of ideas and
techniques that are involved in its study, owing to its strong connections with a num-
ber of distinct branches of mathematics. For some of them, e.g. topology, operator
Introduction xi
theory, and approximation theory, this is rather obvious. More unexpectedly, Banach
space geometry and probability theory also play quite an important role. Even num-
ber theory may be useful at times! For that reason, we believe that linear dynamics is
an extremely attractive area, where many beautiful results are still to be discovered.
We hope that the present book will give some substance to this affirmation.
It is now time to describe the contents of the book in more detail.
Chapter 1 contains the basics of linear dynamics. We introduce hypercyclicity and
the weaker (typically linear) notion of supercyclicity. An operator T ∈ L(X) is said
to be supercyclic if there is some vector x ∈ X such that the cone generated by
O(x, T ) is dense in X. Our approach is based on the Baire category theorem. We
start with the well-known equivalence between hypercyclicity and topological tran-
sitivity: an operator T acting on some separable completely metrizable space X is
hypercyclic iff for each pair of non-empty open sets (U, V ) in X, one can find n ∈ N
such that T n (U ) ∩ V = ∅; in this case, there is in fact a residual set of hypercyclic
vectors. From this, one gets immediately the so-called Hypercyclicity Criterion, a
sufficient set of conditions for hypercyclicity with a remarkably wide range of appli-
cation. The analogous Supercyclicity Criterion is proved along the same lines. Next,
we show that hypercyclicity and supercyclicity induce noteworthy spectral proper-
ties. Then we discuss the algebraic and topological properties of HC(T ), the set of
all hypercyclic vectors for a given hypercyclic operator T ∈ L(X). We show that
HC(T ) always contains a dense linear subspace of X (except 0) and that HC(T ) is
homeomorphic to X when X is a Fréchet space and hence to the separable Hilbert
space. Finally, several fundamental examples are treated in detail: weighted shifts
on p spaces, operators commuting with translations on the space of entire functions
H(C), and composition operators acting on the Hardy space H 2 (D). We will come
back to these examples several times in the book.
Chapter 2 contains some rather impressive results showing that hypercyclicity is
not a mere curiosity. We first prove that hypercyclic operators can be found in any
infinite-dimensional separable Fréchet space. The key point here is that operators of
the form “identity plus a backward shift” are always hypercyclic, and even topolog-
ically mixing; that is, for each pair of non-empty open sets (U, V ), all but finitely
many n ∈ N satisfy T n (U ) ∩ V = ∅. Then we show that any countable, dense,
linearly independent set in a separable infinite-dimensional Banach space is an orbit
of some hypercyclic operator. Next, we discuss the size of the set of all hypercyclic
operators on some given infinite-dimensional separable Banach space X. This set
is always dense in L(X) with respect to the strong operator topology, but nowhere
dense with respect to the norm topology (at least when X is a Hilbert space). Then,
we show that linear dynamics provides a universal model for topological (non-linear)
dynamics: there exists a single hypercyclic operator T acting on the separable Hilbert
space H such that any continuous self-map of a compact metric space is topologi-
cally conjugate to the restriction of T to some invariant compact set K ⊂ H. We
conclude the chapter by showing that any Hilbert space operator is the sum of two
hypercyclic operators.
xii Introduction
In Chapter 3, we present several elegant and useful results that point out
a kind of “rigidity” in linear dynamics: the powers and rotations of hyper-
cyclic operators remain hypercyclic; every single operator in a hypercyclic C0 -
semigroup is already hypercyclic; and any orbit of an arbitrary operator is either
nowhere dense or everywhere dense in the underlying topological vector space.
Besides obvious formal similarities, these results have another interesting com-
mon feature: the proof of each ultimately relies on some suitable connectedness
argument.
Chapter 4 is devoted to the Hypercyclicity Criterion. It turns out that a linear op-
erator T ∈ L(X) satisfies the Hypercyclicity Criterion iff it is topologically weakly
mixing, which means that the product operator T × T is hypercyclic on X × X. The
Hypercyclicity Criterion Problem asks whether every hypercyclic operator has to
be weakly mixing. In a non-linear context, very simple examples show that the an-
swer is negative. However, the linear problem proves to be much more difficult. It
was solved only recently by C. J. Read and M. De La Rosa, who showed that a
counterexample exists in some suitably manufactured Banach space. We present in
Chapter 4 a variant of their construction which allows us to exhibit counterexam-
ples in a large class of separable Banach spaces, including separable Hilbert spaces.
The chapter also contains various characterizations of the weak mixing property in-
volving the sets of natural numbers N(U, V ) := {n ∈ N; T n (U ) ∩ V = ∅}, where
U, V are non-empty open sets in X. These characterizations are undoubtly quite well
known to people working in topological dynamics, but perhaps less so to the operator
theory community.
In Chapter 5, we give a rather detailed account of the connections between linear
dynamics and measurable dynamics, i.e. ergodic theory. The basic idea is the fol-
lowing: if an operator T turns out to be ergodic with respect to some measure with
full support then T is hypercyclic by Birkhoff’s ergodic theorem. Accordingly, it is
desirable to find conditions ensuring the existence of such an ergodic measure. We
concentrate on Gaussian measures only, since they are by far the best understood
infinite-dimensional measures. We start with a general and essentially self-contained
discussion of Gaussian measures and covariance operators on Banach spaces. Then
we show how one can construct an ergodic Gaussian measure for an operator T
provided that T has “sufficiently many” eigenvectors associated with unimodular
eigenvalues. The geometry of the underlying Banach space turns out to be quite im-
portant here, which should not be too surprising to anyone who has heard about
probability in Banach spaces.
In Chapter 6, we discuss some variants or strengthenings of hypercyclicity. We
first show that an operator is hypercyclic whenever it has an orbit passing “not too
far” from any point of the underlying space. Then, we consider chaotic and fre-
quently hypercyclic operators (the latter being implicitly present in Chapter 5).
Chaoticity and frequent hypercyclicity are qualitative strengthenings of hypercyclic-
ity, both strictly stronger because operators with one or the other property are shown
to be weakly mixing. There are interesting similarities and differences between
Introduction xiii
hypercyclicity and these two variants. For example, on the one hand any rotation
and any power of a chaotic or frequently hypercyclic operator has the same prop-
erty; on the other hand, some separable Banach spaces do not support any chaotic
or frequently hypercyclic operator. Moreover, we show the existence of frequently
hypercyclic operators which are not chaotic and Hilbert space operators which are
both chaotic and frequently hypercyclic but not topologically mixing.
In Chapter 7, we discuss in some detail the problem of the existence of com-
mon hypercyclic vectors for uncountable families of operators. By the Baire
category theorem, any countable family of hypercyclic operators has a residual
set of common hypercyclic vectors, but there is no obvious result of that kind
for uncountable families. We present several positive criteria which prove to be
efficient in various situations. These criteria may be viewed as kinds of “uncount-
able Baire category theorems”, applying, of course, to very special families of
open sets. Then we consider the particular case of weighted shifts and show that
continuous paths of weighted shifts may or may not admit common hypercyclic
vectors.
Chapter 8 is centred around the following question: when does a given hyper-
cyclic operator admit a hypercyclic subspace, i.e. when is it possible to find an
infinite-dimensional closed subspace of the underlying space consisting entirely of
hypercyclic vectors (except 0)? If the operator T acts on a complex Banach space and
satisfies the Hypercyclicity Criterion then there is a complete and very simple char-
acterization: such a subspace can be found iff the essential spectrum of T intersects
the closed unit disk. We prove this result in two different ways and then give several
natural examples. We also prove some results related to the existence of non-trivial
algebras of hypercyclic vectors.
Chapter 9 is entirely devoted to supercyclicity. We prove the so-called Angle Cri-
terion, a geometrical result which is often useful for showing that a given operator is
not supercyclic. Then, we illustrate this criterion with two nice examples: the com-
position operators on H 2 (D) associated with non-automorphic parabolic maps of the
disk and the classical Volterra operator acting on L2 ([0, 1]).
In Chapter 10, we consider hypercyclicity or supercyclicity with respect to the
weak topology of a given Banach space X. We start with a detailed discussion of
weakly dense sequences which are not dense with respect to the norm topology.
Then we concentrate on weak hypercyclicity or supercyclicity for bilateral weighted
shifts acting on p (Z). In particular we show that there exist bilateral weighted
shifts which are weakly hypercyclic but not hypercyclic, that weak hypercyclicity or
supercyclicity of a weighted shift really depends on the exponent p (unlike norm hy-
percyclicity and supercyclicity), and that the unweighted shift is weakly supercyclic
on p (Z) iff p > 2. Then we consider unitary operators. We show that, surprisingly
enough, there exist Borel probability measures μ on T for which the “multiplica-
tion by the variable” operator Mz is weakly supercyclic on L2 (μ). This holds if the
support of μ is “very small” but it is also possible to require that the Fourier coef-
ficients of μ vanish at infinity, in which case the support of μ is rather “large”. We
xiv Introduction
conclude the chapter by discussing the notions of weak sequential hypercyclicity and
supercyclicity, which are still not well understood.
Chapter 11 is devoted to the universality properties of the Riemann zeta function.
We show that any holomorphic function in the strip {1/2 < Re(s) < 1} and without
zeros can be uniformly approximated on compact sets by imaginary translates of the
zeta function. This remarkable result is due to S. M. Voronin. The proof is very much
in the spirit of the whole book, being a mixture of analytic number theory, function
theory, Hilbert space geometry, and ergodic theory.
In Chapter 12, we try to give a reader-friendly description of one of the many op-
erators constructed by C. J. Read in connection with the invariant subspace problem.
We concentrate on the simplest example: an operator without non-trivial invariant
subspaces on the space X = 1 (N). Even so, the construction is quite involved but is
presented at a relatively slow pace in a reasonable number of pages. When working
on that chapter, our hope was, of course, to be able to solve the problem on a sepa-
rable Hilbert space! The final result is much less impressive; nevertheless, we hope
that the chapter will be useful to some people. Typically, a non-expert interested in
the invariant subspace problem (just like us) may find our exposition convenient.
From this outline, it should be clear that we have not written an encyclopaedic
treatise on linear dynamics. This book is rather a selection of results and ideas made
mostly according to our personal tastes but also because they fit together to give a
reasonably accurate global picture of the subject. As a result, the chapters have few
overlaps and can be read more or less independently.
At the end of the book, we have added four appendices on complex analysis, func-
tion spaces, Banach space theory, and spectral theory. The reader will find there only
definitions and results that are explicitly needed in the main body of the book. Sev-
eral proofs are given. One reason is that the reader may find it more convenient to
have grouped together the proofs of important results that are used several times,
rather than to look for them in various sources. Another reason is that some results
definitely need to be proved, but it seemed better to postpone the proofs to the ap-
pendices in order to keep the reading of the book reasonably fluent. Concerning the
appendix on spectral theory, we must confess that the first reason for including these
proofs was that this was useful for us, since we are very far from being experts in
that area.
As a rule, we have tried to give the simplest and most natural proofs that we were
able to produce. However, this does not mean that we have refrained from stating
a result in great generality whenever this seemed to be both possible and desirable.
We hope that various kinds of readers will find our book useful, e.g. Ph.D. students,
specialists in the area and non-specialists wanting to get a flavour of the subject.
We felt that it should be accessible to a rather large audience, including graduate
students with an interest in functional analysis. Perhaps ambitiously, we hope that
some of these people actually enjoy reading the book!
Each chapter ends with some comments and a set of exercises. Some exercises are
quite easy, but these are not necessarily the less interesting ones. Some others outline
Introduction xv
proofs of useful published results that could have been included in the main body of
the book but were relegated to exercises owing to the lack of space. A few exercises
have the aim of proving new results which will probably not have been published
elsewhere. Finally, some exercises are devoted to results which are used in the text
but for which we did not give full proofs in order to make the presentation more
digestible. We have worked out each exercise rather carefully and included a number
of explicit hints. In that way, we believe that any motivated reader will succeed in
finding solutions without excessive effort.
Just like any other, this book cannot pretend to be perfect and is bound to suffer
from flaws. The authors take responsibility for these, including any mathematical
errors. In particular, if some result has not been included this by no means indicates
that it did not deserve to be mentioned. Most likely the omission was due to a lack
of space; and if the result does not even appear in the comments this simply means
that the authors were not aware of it. We would be very grateful to anyone pointing
out unfortunate omissions, mathematical inaccuracies or troublesome typos to us.
Acknowledgements We are very honoured that Cambridge University Press ac-
cepted our book for publication. For this many thanks are due to Roger Astley, whose
quite positive reception of a first, very preliminary, draft of the manuscript and al-
ways optimistic emails were extremely encouraging for us. We also thank Susan
Parkinson and Anna-Marie Lovett for their help during the production process.
Many colleagues and friends, one student and one relative have read parts of the
manuscript and/or made a number of useful suggestions for improving the text. In
alphabetical order, our warmest thanks to Richard Aron, Gilles Bailly-Maitre, Juan
Bès, Josée Besenger, Monsieur Bony, Isabelle Chalendar, Kit Chan, Éric Charpentier,
Bernard Chevreau, George Costakis, Sylvain Delpech, Robert Deville, Eva Gallardo,
Frédéric Gaunard, Gilles Godefroy, Sophie Grivaux, Andreas Hartmann, General
Mickael F. Jourdan, Alexandre Matheron, Jonathan Partington, Hervé Queffélec,
Elizabeth Strouse and Frédérique Watbled.
Finally, a more personal reading was done by, in decreasing size order, Véronique,
Armelle, Mathilde, Juliette, Solène and Émile. For that, and so much more, we
dedicate the book to them.
1
Hypercyclic and supercyclic operators
Introduction
The aim of this first chapter is twofold: to give a reasonably short, yet significant
and hopefully appetizing, sample of the type of questions with which we will be
concerned and also to introduce some definitions and prove some basic facts that
will be used throughout the whole book.
Let X be a topological vector space over K = R or C. We denote by L(X) the set
of all continuous linear operators on X. If T ∈ L(X), the T -orbit of a vector x ∈ X
is the set
O(x, T ) := {T n (x); n ∈ N}.
The operator T is said to be hypercyclic if there is some vector x ∈ X such
that O(x, T ) is dense in X. Such a vector x is said to be hypercyclic for T (or
T -hypercyclic), and the set of all hypercyclic vectors for T is denoted by HC(T ).
Similarly, T is said to be supercyclic if there exists a vector x ∈ X whose projective
orbit
K · O(x, T ) := {λT n (x); n ∈ N, λ ∈ K}
is dense in X; the set of all supercyclic vectors for T is denoted by SC(T ). Finally,
we recall that T is said to be cyclic if there exists x ∈ X such that
K[T ]x := span O(x, T ) = {P (T )x; P polynomial}
is dense in X.
Of course, these notions make sense only if the space X is separable. Moreover,
hypercyclicity turns out to be a purely infinite-dimensional phenomenon ([206]):
P ROPOSITION 1.1 There are no hypercyclic operators on a finite-dimensional
space X = {0}.
P ROOF Suppose on the contrary that T is a hypercyclic operator on KN , N ≥ 1.
Pick x ∈ HC(T ) and observe that (x, T (x), . . . , T N −1 (x)) is a linearly independent
family and hence is a basis of KN . Indeed, otherwise the linear span of O(x, T )
would have dimension less than N and hence could not be dense in KN . For any
α ∈ R+ , one can find a sequence of integers (nk ) such that T nk (x) → αx. Then
T nk (T i x) = T i (T nk x) → αT i x for each i < N , and hence T nk (z) → αz for
any z ∈ KN . It follows that det(T nk ) → αN , i.e. det(T )nk → αN . Thus, putting
a := |det(T )|, we see that the set {an ; n ∈ N} is dense in R+ . This is clearly
impossible.
The most general setting for linear dynamics is that of an arbitrary (separable)
topological vector space X. However, we will usually assume that X is an F -space,
1
2 Hypercyclic and supercyclic operators
i.e. a complete and metrizable topological vector space. Then X has a translation-
invariant compatible metric (see [210]) and (X, d) is complete for any such metric
d. In fact, in most cases X will be a Fréchet space, i.e. a locally convex F -space.
Equivalently, a Fréchet space is a complete topological vector space whose topology
is generated by a countable family of seminorms.
An attractive feature of F -spaces is that one can make use of the Baire category
theorem. This will be very important for us. Incidentally, we note that the Banach–
Steinhaus theorem and Banach’s isomorphism theorem are valid in F -spaces, and
if local convexity is added then one can also use the Hahn–Banach theorem and its
consequences. If the reader feels uncomfortable with F -spaces and Fréchet spaces,
he or she may safely assume that the underlying space X is a Banach space, keeping
in mind that several natural examples live outside this context.
The chapter is organized as follows. We start by explaining how one can show
that a given operator is hypercyclic or supercyclic. In particular, we prove the so-
called Hypercyclicity Criterion, and the analogous Supercyclicity Criterion. Then we
show that hypercyclicity and supercyclicity both entail certain spectral restrictions
on the operator and its adjoint. Next, we discuss the “largeness” and the topological
properties of the set of all hypercyclic vectors for a given operator T . Finally, we treat
in some detail several specific examples: weighted shifts on p spaces, composition
operators on the Hardy space H 2 (D), and operators commuting with translations on
the space of entire functions H(C).
(i) T is hypercyclic;
(ii) T is topologically transitive; that is, for each pair of non-empty open sets
(U, V ) ⊂ X there exists n ∈ N such that T n (U ) ∩ V = ∅.
Since (Vj ) is a basis for the topology of X, this is equivalent to the topological
transitivity of T .
C OROLLARY 1.3 Let X be a separable F -space, and let T ∈ L(X). Assume that
T is invertible. Then T is hypercyclic if and only if T −1 is hypercyclic.
It is worth noting that T and T −1 do not necessarily share the same hypercyclic
vectors; see Exercise 1.11.
We illustrate Theorem 1.2 with the following historic example, also due to
Birkhoff [54].
E XAMPLE 1.4 (G. D. B IRKHOFF , 1929) Let H(C) be the space of all entire
functions on C endowed with the topology of uniform convergence on compact
sets. For any non-zero complex number a, let Ta : H(C) → H(C) be the
translation operator defined by Ta (f )(z) = f (z + a). Then Ta is hypercyclic
on H(C).
4 Hypercyclic and supercyclic operators
P ROOF The space H(C) is a separable Fréchet space, so it is enough to show that
Ta is topologically transitive. If u ∈ H(C) and E ⊂ C is compact, we set
u E := sup{|u(z)|; z ∈ E}.
Let U, V be two non-empty open subsets of H(C). There exist ε > 0, two closed
disks K, L ⊂ C and two functions f, g ∈ H(C) such that
U ⊃ h ∈ H(C); h − f K < ε ,
V ⊃ h ∈ H(C); h − g L <ε .
Let n be any positive integer such that K ∩(L+an) = ∅. Since C\(K ∪(L+an))
is connected, one can find h ∈ H(C) such that
this follows from Runge’s approximation theorem (see e.g. [209] or Appendix A).
Thus h ∈ U and Tan (h) ∈ V , which shows that Ta is topologically transitive.
Topologically transitive maps are far from being exotic objects. For example, the
map x → 4x(1 − x) is transitive on the interval [0, 1] and the map λ → λ2 is
transitive on the circle T (see e.g. R. L. Devaney’s classical book [94]). However, in
a topological setting one often needs a specific argument to show that a given map is
transitive.
Nevertheless, in a linear setting an extremely useful general criterion for hyper-
cyclicity does exist. This criterion was isolated by C. Kitai in a restricted form [158]
and then by R. Gethner and J. H. Shapiro in a form close to that given below, [119].
The version we use appears in the Ph.D. thesis of J. Bès [45].
D EFINITION 1.5 Let X be a topological vector space, and let T ∈ L(X). We say
that T satisfies the Hypercyclicity Criterion if there exist an increasing sequence of
integers (nk ), two dense sets D1 , D2 ⊂ X and a sequence of maps Snk : D2 → X
such that:
We will sometimes say that T satisfies the Hypercyclicity Criterion with respect
to the sequence (nk ). When it is possible to take nk = k and D1 = D2 , it is usually
said that T satisfies Kitai’s Criterion. We point out that in the above definition, the
maps Snk are not assumed to be linear or continuous.
where we have used (iii), (ii), and (iv). Thus T ml (x) − yl → 0 as l → ∞, which
concludes the proof.
R EMARK 1.7 We have in fact proved the following more precise result: if T ∈
L(X) satisfies the Hypercyclicity Criterion with respect to some sequence (nk )k≥0
then the family (T nk )k≥0 is universal, i.e. there exists some vector x ∈ X such that
the set {T nk (x); k ≥ 0} is dense in X. In fact, for any subsequence (nk ) of (nk ),
the family (T nk )k≥0 is universal: this is apparent from the above proofs.
6 Hypercyclic and supercyclic operators
Theorem 1.6 will ensure the hypercyclicity of almost (!) all the hypercyclic oper-
ators in this book. We give two historical examples, due to G. R. MacLane [176] and
to S. Rolewicz [206]. The latter was the first example of a hypercyclic operator that
acts on a Banach space.
E XAMPLE 1.9 (S. ROLEWICZ , 1969) Let B : 2 (N) → 2 (N) be the backward
shift operator, defined by B(x0 , x1 , . . . ) = (x1 , x2 , . . . ). Then λB is hypercyclic
for any scalar λ such that |λ| > 1.
Abend war es. Seidiger Dunst stand über dem Garten, an den
Bäumen hingen ein paar Lichter, die von der schmalen Mondsichel
kamen.
Um das Haus ging ein Klappern. Drüben bei Agnes Elisabeths
Wohnzimmer fing es an, kam hinten herum ... Plötzlich erschien
Agnes Elisabeth vor Julies Fenster und schlug den Laden zu.
»Was soll das?« rief Julie hinaus. »Warum tust du das?«
»Das muß so sein! Man ist dann sicherer!«
Julie kam ans Fenster und lehnte den Laden wieder zurück.
»Aber, Agnes Elisabeth, das haben wir früher nicht getan!«
»Es ist auch anders geworden!«
»Bei mir müssen die Fenster aufbleiben! Ich fürchte mich nicht!«
In Agnes Elisabeths Stimme kam ein weinerlicher Ton:
»So schließe wenigstens die Tür ab, Julie! Es ist nicht mehr wie
früher!« Sie stand auf dem Rasen, einen Rechen in der Hand, mit
dem sie die Wege noch einmal nachgesehen haben mochte. Julie
beugte sich zu ihr hinaus.
»Wenn du dich fürchtest, soll ich dann nicht lieber bei dir
schlafen?!«
Agnes Elisabeth tat, als höre sie nichts, und begann den Weg
unter dem Fenster zu harken.
Julie ging ins Zimmer zurück. Die Angst um die Schwester war
wieder da. Eine Angst, die an hundert Dingen hing und sich doch
eigentlich auf nichts gründete, ein Zittern vor etwas Grausigem, das
irgendwo wartete, sich nicht greifen ließe und doch da wäre und
drohte.
Sie mußte sich beschäftigen. Sie schloß ihren Schreibtisch auf,
stellte alle Sachen wieder an ihren alten Platz, ging an den Koffer,
packte die letzten Stücke aus. Dabei wartete sie immer darauf, daß
Agnes Elisabeth käme, ihr gute Nacht zu sagen. Sie wollte mit ihr
über all dies Seltsame sprechen, wollte sie fragen, ob sie ihr nicht
helfen könne. Wie ein kleines Mädchen kam sie sich vor, als sie
daran dachte, daß sie helfen solle. Sie wußte nicht, wo sie mit helfen
beginnen sollte.
Agnes Elisabeth kam nicht. Vielleicht besprach sie mit Gesche
noch das Essen für morgen oder sah nach dem neu eingekochten
Saft. Schließlich dauerte es Julie zu lange, und sie ging hinaus. Auf
der Diele brannte ein Nachtlicht. Die Dunkelheit erschien dunkler
dadurch. In der Küche war niemand zu sehen; nur die Glut
verglimmender Torfstücke kroch noch über den Herd.
Julie ging nach Agnes Elisabeths Zimmer hinüber, faßte die Klinke
und wollte eintreten.
Die Tür war verschlossen. Drinnen sprang jemand auf.
»Wer ist da?«
»Ich bin es ... Julie! Ich wollte dir gute Nacht sagen!«
»Ich bin schon zu Bett,« sagte es drinnen nach einer Weile. »Ich
bin müde.«
»Du bist doch nicht krank?«
»Warum sollte ich krank sein? Geh nur zu Bett, Julie!«
»Wenn du mich brauchst, Agnes Elisabeth, weckst du mich, hörst
du?!«
Von drinnen kam keine Antwort.
Julie ging in ihr Zimmer zurück.
Sie trat ans Fenster und sah in den Garten hinaus. Da waren mit
einem Male viele weiße Laken. Große grelle Flecke! Sie erschrak.
Dann fiel ihr ein: die hatte wohl Agnes Elisabeth noch aufgehängt.
Sie zog ihren Stuhl heran und lehnte die Arme auf die Fensterbank.
Lange Zeit saß sie so.
Plötzlich fuhr sie zusammen. Irgendwo dort hinten war ein
Schatten über die Laken gegangen. Sie beugte sich hinaus. Es war
nichts zu sehen; der Garten lag totenstill. Sie ging ins Zimmer zurück
an ihren Schreibtisch. Hier lag alles, womit sie sich die letzten Jahre
beschäftigt hatte. Jetzt würde sie zu solcher Arbeit fürs erste nicht
mehr kommen.
Da waren auch Peters Aquarelle. Sie wollte sie morgen
aushängen; jetzt war sie zu müde. Sie begann sich langsam
auszukleiden.
Plötzlich ging sie ans Fenster und ließ das Rouleau herunter. —
Es war doch vielleicht besser. Dann legte sie sich hin und löschte
das Licht.
XXXII
Agnes Elisabeth kniete auf dem Wege und wühlte mit beiden
Händen zwischen den Johannisbeersträuchern.
Ob sie dieses Mal wohl dabei sein durfte?! Bei Marianne hatte
man sie nicht hineingelassen. Aber hier im eignen Hause! Sie wollte
das doch gern einmal sehen, wie so das Leben anfängt! Diese
dumme Wolfsmilch! Fünf Blätter, drei Krauseminzen, vier schwarze
Johannisbeerblätter, bei zunehmendem Mond, er wurde jetzt voller;
dann könnte sie endlich wieder schlafen! Auf den Boden mußte sie
gehen, wegen der Wiege und der Kindersachen und der
Wachspuppe ...! Und der Doktor ...! Wann würde das Kind kommen?
Kam es nicht oft vor, daß die Mutter dabei starb? Dann wollte s i e
das Kind haben! Dann wollte sie aber lachen! Daß s i e ein Kind
hatte!
Als sie endlich aufstand, war alles Unkraut wieder eingepflanzt.
Julie ging inzwischen durch den Garten. Sie lehnte sich ans Gitter
und sah die Chaussee hinunter. Ein paar Torfwagen rumpelten
vorüber, ein kleines Bauernmädchen knickste. Alles war wie früher.
Eben wollte sie ins Haus, da kam um die Ecke dort drüben etwas
Gelbes; ein volles Rosa wölbte sich darüber und zuoberst nickte ein
Geranke von Federn, Blumen und Stroh.
Julie in ihrer Kurzsichtigkeit meinte zuerst, das Ganze sei
Marianne. Sie atmete auf, als sich das eine als Kinderwagen, das
andere als Mariannes Hut herausstellte.
Nun löste sich etwas Schwarzes und steuerte nach der Tür
herüber. Das Gelbe mit dem rosa Verdeck aber zog vorüber.
Dahinter kam Marianne, eine Weile nichts, als Marianne, breit und
voll und himmelblau ... Sie blickte zur Seite und wollte Julie nicht
sehen.
Das war eine Demonstration!
Julie nahm sie belustigt entgegen.
»Guten Tag, Julie!« sagte Lukas. »Ich mußte doch kommen und
dich begrüßen.«
Julie gab ihm die Hand.
»Das ist nett von dir! Wie geht es dir, Lukas?«
»Gut!« Sein Gesicht strahlte. Er war froh, daß er standhaft
geblieben war; wenn es auch nicht gerade gegen Mariannes Willen,
so war es doch ohne ihre Zustimmung geschehen, daß er sie
verlassen hatte, um Julie zu begrüßen.
»Wie geht es Marianne und eurer Kleinen?« fragte Julie.
»Der Kleinen? Oh, das ist ein Prachtmädel! Gesund und fidel,
wiegt schon vierundzwanzig Pfund! Auch Marianne geht es, Gott sei
Dank, recht gut. Sie hat viel im Haushalt zu tun. Es ist gemütlich bei
uns. Du solltest ...« Er brach plötzlich ab und wurde rot.
Julie mußte lächeln; sie war eingehüllt in eine Wolke von
Fliederparfüm. Das war Mariannes Seife, und auch Lukas mußte sie
natürlich benutzen.
»Nicht wahr, Julie,« begann er stockend, »du trägst es Marianne
nicht nach, daß sie ...? Sie wird schon zur Einsicht kommen. Im
Grunde ist sie doch so gut ...!«
Julie nickte freundlich.
»Ich kenne sie. Aber willst du nicht hereinkommen, Lukas?«
»Ich habe keine Zeit. Pastor Gerlach will mich heute noch sehen.
Wir richten eine Schulbibliothek ein; da gibt es viel zu besprechen.
Du hast wohl keine alten Bücher, die du entbehren kannst?«
»Ich will einmal nachsehen!«
»Wir nehmen alles mit Dank an!« Lukas schüttelte ihr die Hand
und ging eiligen Schrittes der Kirche zu.
Der Fliedergeruch dehnte sich in der Sommerluft, wurde flacher
und dünner und war verschwunden. Julie verstand jetzt besser als
früher, daß man Lukas Allm lieb haben konnte.
Aber warum gerade Marianne ihn lieb hatte?!
XXXIII