Published by

World Scientific Publishing Co. Pte. Ltd.

5 Toh Tuck Link, Singapore 596224
USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.

GLOBAL ATTRACTORS OF NON-AUTONOMOUS DISSIPATIVE

DYNAMICAL SYSTEMS
Copyright © 2004 by World Scientific Publishing Co. Pte. Ltd.
All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means, electronic or
mechanical, including photocopying, recording or any information storage and retrieval system now known or to
be invented, without written permission from the Publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center,
Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from
the publisher.

ISBN 981-256-028-9

Printed in Singapore.
 Dedicated to my wife Ivanna
and my children Olga and Anatoli
vi
 Preface

In the qualitative theory of differential equations non-local problems play an im-

portant role, especially in regard to questions of boundedness, periodicity, almost
periodicity, Poisson stability, asymptotic behaviour, dissipativity, etc.
The present work takes a similar approach and is dedicated to the study of
abstract non-autonomous dissipative dynamical systems and their application to
differential equations.
In applications there often occur systems

u0 = f (t, u), (0.1)

which have every one of their solutions driven into fixed bounded domain and kept
there under further increase of time, becouse of natural dissipation. Such systems
are called dissipative ones [135]-[137],[270]-[272],[325], [326]. Solutions of dissipative
systems are called limit (finally) bounded [325, 326].
Dynamical systems occur in hydrodynamics studying turbulent phenomena,
meteorology, oceanography, theory of oscillations, biology, radio engineering and
other domains of sciences and engineering technics related to the study of asymp-
totic behaviour. Lately interest in dissipative systems increased even more
because of intensive elaboration of strange attractors (see, e.g.,[143, 239, 296,
306]).
The study of the dissipative systems there are dedicated plenty of works, begin-
ning from the classical works of N. Levinson. Among works on dissipative systems
of ordinary differential equations two directions can be made out.
To the first belong works which contain some conditions assuring the dissipativ-
ity of system (0.1), some class or concrete system, representing theoretical or applied
interest. Examples are works of P. V. Atrashenok [9], B. P. Demidovich [135]-[137],
V. I. Zubov [336]-[338], V. M. Matrosov [251], V. V. Nemytski [259]-[260], V. A.
Pliss [270], V. N. Schennikov [298, 299], C. Corduneanu [125], N. Levinson [237], N.
Pavel [264]-[266], R. Reissig [272], T. Talpalaru [309] and a lot of other authors.
To the second direction belong works in which are studied inner conditions

vii
viii Global Attractors of Non-autonomous Dissipative Dynamical Systems

of dissipative systems, that is, conditions relating to the character of behaviour

of solution of the system when assuming its dissipativity, for different classes of
differential equations. Among them are the works of V. M. Gershtein [159]-[163],
V. V. Zhikov [331]-[332], I. L. Zinchenko [334], M. A. Krasnoselsky [159], S. Yu.
Pilyugin [269], V. A. Pliss [270]-[271], M. L. Cartwright and J. E. Littlewood [44]-
[46], N. Levinson [237], G. Fusco and M. Oliva [157], J. Skowronski and S. Ziemba
[307] and other authors.
For a compact map on a Banach space, V. M. Gerstein [162], V. M. Ger-
stein and M. A. Krasnoselskii [159] investigated the existence of maximal com-
pact invariant set and studied some of its properties. In works of G. Billoti,
G. Cooperman, J. LaSalle, O. Lopes, P. Massat, M. Slemrod, J. Hale [175, 26,
124], [171]-[176],[235],[246]-[248] and a lot of other authors many important results
obtained for ordinary differential equations [270]-[271] are generalized to functional-
differential equations.
Of late in the theory of partial differential equations there appeared works of
A. V. Babin and M. I. Vishik [12]-[16], Ju. S. Ilyashenko [194]-[197], O. A. La-
dyzhenskaya [230], A. N. Sharkovsky [296], R. Temam [314] and other authors,
in which, are studied evolutionary equations with maximal attractors (dissipative
evolutionary equations).
We note that all works mentioned above (with rare exceptions) studied periodical
or autonomous systems. These results are presented in monographs [175], [270]-
[272],[296].
If the right hand side f of the equation (0.1) is non-periodic, e.g. quasi-periodic
(almost periodic by Bohr, recurrent in sense the of Birkhoff, almost periodic by
Levitan, stable by Poisson) or depending on time in more complicated way, then
the situation essentially complicates already in the class of almost periodic systems.
It is caused at least by two reasons.
First, the definition of dissipativity in the non-autonomous case needs to be made
more precise because Levinson’s definition in the class of non-periodical systems
divides on some non equivalent notions and we need to choose one which allows
us develop a general theory which would contain as particular case most essential
results obtained for periodical dissipative systems.
Second, in the study of periodic dissipative systems an important role is played
by discrete dynamical system (cascade) generated by degrees of Poincaré’s trans-
formation (mapping). For non-periodic systems there is no Poincaré’s transforma-
tion and, consequently, the approach created for research on periodical dissipative
systems is not useful in the more general case. That is why to study non-periodical
dissipative systems we need new ideas; that is, making a theory of non-autonomous
dissipative dynamical systems demands making corresponding methods of research.
Our approach to the study of dissipative systems of differential equations consists
of drawing to the study of non-autonomous dissipative systems ideas and methods
 Preface ix

developed in the theory of abstract dynamical systems. We select one class of

dynamical systems (called in this work, dissipative) modelling the properties of
dissipative differential equations. The selected class is systematically researched
and then the general results obtained are applied to the study of dissipative systems
of differential and some other classes of equations.
The idea of applying methods of the theory of dynamical systems to the study
of non-autonomous differential equations is not new. It has been successfully
applied to the resolution of different problems in the theory of linear and non-
linear non-autonomous differential equations for more than thirty years. First
this approach to non-autonomous differential equations was introduced in works
of V. M. Millionshchikov [253]-[255], B. A. Shcherbakov [300, 302], L. G. Dey-
seach and G. R. Sell [139], R. K. Miller [252], G. Seifert [289], G. R. Sell [290,
291], later in works of V. V. Zhikov [331], I. U. Bronshtein [32], R. A. Johnson
[203]-[204] and many other authors. This approach consists of naturally linking
with equation (0.1) a pair of dynamical systems and a homomorphism of the first
onto the second. In one dynamical system is put the information about right hand
side of equation (0.1) and in the other about the solutions of equation (0.1).
We note that there exists the another approach offered in the works of V. I.
Zubov [336] and then developed in works of C. M. Dafermos [128]-[131], J. K. Hale
[175], I. Hitoshi [189] and many other authors (for details, see survey [303]). It con-
sists of linking with every non-autonomous differential equation a two-parametric
family of mappings (by terms of some authors – process).
In the beginning a these two approaches developed independently but later there
was found a relation between them (see, e.g. [130],[187]).
This author adheres to the first approach because in his opinion it is better
adapted for resolving those problems which are studied in this work.
The proposed work consists of fourteen chapters.
In the first chapter for autonomous dynamical systems there are introduced and
studied different kinds of dissipativity: point, compact, local, bounded and weak
one. Criteria of point, compact and local dissipativity are given. It is shown that
for dynamical systems in locally compact spaces all three types of dissipativity
are equivalent. Examples are given showing that in the general case the notions of
point, compact and local dissipativity are different. The notion of Levinson’s center,
which is an important characteristic of compact dissipative system, is introduced.
The solution of one J. K. Hale’s problem for locally bounded dynamical systems is
given.
The second chapter is dedicated to non-autonomous dissipative dynamical
systems. It is noted that in the general case Levinson’s center of non-autonomous
dissipative dynamical system is not orbitally stable. The question of stability of
Levinson’s center of non-autonomous system is studied. A simple geometric descrip-
tion ensuring its stability is given, as is a description of Levinson’s center of non-
x Global Attractors of Non-autonomous Dissipative Dynamical Systems

autonomous systems satisfying the condition of uniform positive stability. There are
pointed conditions of keeping the property of dissipativity under homomorphisms.
It is selected a class of dynamical systems which allow full description of Levinson’s
center’s structure, called in this work systems with convergence. Some criteria of
convergence in terms of Lyapunov’s functions depending on two space variables are
given. It is shown that for non-autonomous dissipative dynamical systems with
finite-dimensional phase space all three types of dissipativity are equivalent. There
are given series of conditions that are equivalent to dissipativity in finite-dimensional
space. At last, it is proved that for linear systems dissipativity reduces to conver-
gence. Also there are given series of conditions equivalent to dissipativity of linear
systems.
The third chapter deals mostly with one special class of non-autonomous dissi-
pative dynamical systems called in the work C-analytic. It is proved that C-analytic
dissipative dynamical system has the property of uniform positive stability on com-
pact subsets. Full description of Levinson’s center of these systems is given. One
general construction allowing to connect with given non-autonomous dynamical
system an autonomous dynamical system in space of continuous sections is given.
With the help of these constructions are studied quasi-periodic solutions of analytic
systems with quasi-periodic coefficients. In conclusion conditions are given which
guarantee the dissipativity of weakly nonlinear systems of differential equations,
as is a condition which assure the existence of almost periodic solution of weakly
nonlinear system with almost periodic coefficients in Levinson’s center.
The fourth chapter is dedicated to a study of Levinson’s center’s structure
with condition of hyperbolicity on closure of recurrent motion’s set. There we
establish some topological properties of Levinson’s center of compact dissipative
dynamical system. In particular, it is shown that Levinson’s center is indecompos-
able if the phase space of dynamical systems is also decomposable. It is proved
that in connected and locally connected space Levinson’s center of compact dissi-
pative dynamical system both with continuous and discrete time is a connected set.
There we establish some properties of a set of chain recurrent motions of dissipative
system. A theorem is proved about the spectral decomposition of Levinson’s cen-
ter which is analogous to known Smale’s theorem. For one-dimensional dissipative
dynamical systems it was proved theorem precising theorem about spectral decom-
position of Levinson’s center and, particularly, it was shown that Levinson’s center
of such systems contains a local maximal hyperbolic Markov set. In the end of the
chapter an application of obtained results to periodic systems is given.
In the fifth chapter we develop the method of Lyapunov’s functions for research
of non-autonomous dissipative dynamical systems in finite-dimensional space. Cri-
teria of dissipativity in terms of Lyapunov’s functions, with the help of which we
can get sufficient tests for dissipativity suitable in applications, are discassed. With
the help of Lyapunov’s functions there were proved series of tests for dissipativity of
 Preface xi

multi-dimensional non-autonomous differential equations. On the basis developed

for research of non-autonomous dissipative systems methods a criterion of asymp-
totic stability of zero section of non-autonomous systems has been obtained. In
particular, it is proved the analog of known Barbashin–Krasovsky’s theorem for
non-autonomous dynamical systems. There are established some tests for conver-
gence of systems of differential equations with the help of Lyapunov’s functions
depending on two space variable. There were proved tests for dissipativity and
convergence of some systems of differential equations of the 2nd and 3rd order
appearing in applications.
The sixth chapter is dedicated to some applications of general results obtained
in previous chapters to difference equations, equations with impulses, functional-
differential equations and evolutionary equations x0 +Ax = f with uniformly mono-
tone operator A. In particular, there are given tests for dissipativity and con-
vergence of weakly nonlinear systems of difference equations and equations with
impulses. It is proved the criterion of asymptotic stability of linear functional-
differential equations. It is established test for convergence of evolutionary equation
x0 + Ax = f with uniformly monotone operator A.
In the seventh chapter we systematically study of the problem of upper semi-
continuity of compact global attractors and compact pullback attractors of abstract
non-autonomous dynamical systems for small perturbations. Several applications
of our results are given for different classes of evolutionary equations.
The eighth chapter is devoted to the study of the relationship between the
global attractor of the skew–product system and the pullback and forward attractors
of the cocycle system. We also note that forward attractors are stronger than
global attractors if we suppose a compact set of non-autonomous perturbations.
An example is presented in which the cartesian product of the component subsets
of a pullback attractor is not a global attractor of the skew–product flow. This
set is, however, a maximal compact invariant subset of the skew–product flow. By
a generalization of some stability results of V. I. Zubov [336] it is asymptotically
stable. Thus a pullback attractor always generates a local attractor of the skew–
product system, but this need not be a global attractor. If, however, the pullback
attractor generates a global attractor in the skew–product flow and if, in addition, its
component subsets depend lower continuously on the parameter, then the pullback
attractor is also a forward attractor. Several examples illustrating these results are
presented in the final section.
In the ninth chapter we systematically study the global pullback attractors of
C−analytic cocycles. For the large class of C−analytic cocycles we give the de-
scription of the structure of their pullback attractors. Particularly we prove that
it is trivial, i.e. the fibers of these attractors contain only one point. Several ap-
plications of these results are given (ODEs, Caratheodory’s equations with almost
periodic coefficients, almost periodic ODEs with impulse).
xii Global Attractors of Non-autonomous Dissipative Dynamical Systems

The tenth chapter is dedicated to the investigation of the effect of time discretiza-
tion on the pullback attractor of a non-autonomous ordinary differential equation
for which the vector fields depend on a parameter that varies in time rather than
depending directly on time itself. The parameter space is assumed to be compact
so the skew product flow formalism as well as cocycle formalism also applies and
the vector fields have a strong dissipative structure that implies the existence of
a compact set that absorbs all compact sets under the resulting non-autonomous
dynamics. The numerical scheme considered is a general 1–step scheme such as the
Euler scheme with variable time-steps. Our main result is to show that the numeri-
cal scheme interpreted as a discrete time non-autonomous dynamical system, hence
discrete time cocycle mapping and skew product flow on an extended parameter
space, also possesses a cocycle attractor and that its component subsets converge
upper semi–continuously to those of the cocycle attractor of the original system
governed by the differential equation. We will also see that the corresponding skew
product flow systems have global attractors with the cocycle attractor component
sets as their cross-sectional sets in the original state space. Finally, we investigate
the periodicity and almost periodicity of the discretized pullback attractor when
the parameter dynamics in the ordinary differential equation is periodic or almost
periodic and the pullback attractor consists of singleton valued component sets, i.e.
the pullback attractor is a single trajectory.
In the eleventh chapter we study the non-autonomous Navier-Stokes equations.
It is proved that such systems admit compact global attractors. This problem is
formulated and solved in the terms of general non-autonomous dynamical systems.
We give conditions of convergence of non-autonomous Navier-Stokes equations . A
test of existence of almost periodic (quasi periodic, recurrent, pseudo recurrent)
solutions of non-autonomous Navier-Stokes equations is given. We prove the global
averaging principle for non-autonomous Navier-Stokes equations.
The twelfth chapter is devoted to the investigation of the global attractors of
general V - monotone non-autonomous dynamical systems and their applications to
different class of differential equations (ODEs, ODEs with impulse, some class of
evolution partial differential equations).
In the thirteenth chapter we study the linear almost periodic dynamical systems.
The bounded solutions, relation between different types of stability and uniform
exponential stability for those systems are studied. We give several applications the
obtained results for ODEs, PDEs and functional-differential equations.
Chapter 14 is devoted to the study of quasi-linear triangular maps: chaos, almost
periodic and recurrent solutions, integral manifolds, chaotic sets etc. This problem
is formulated and solved in the framework of non-autonomous dynamical systems
with discrete time. We prove that such systems admit an invariant continuous
section (an invariant manifold). Then, we obtain the conditions of the existence of
a compact global attractor and characterize its structure. We give a criterion for the
 Preface xiii

existence of almost periodic and recurrent solutions of the quasi-linear triangular

maps. Finally, we prove that quasi-linear maps with chaotic base admits a chaotic
compact invariant set.
The results given in this work belong mostly to the author. The results of
chapters eight, ten and twelve are jointly obtained by the author, Peter Kloeden and
Bjoern Schmalfuss [96, 97, 100, 106]. The results of chapter eleven and paragraph
six of chapter six are obtained jointly with Jinqiao Duan [104, 107]. The results of
chapter fourteen are obtained jointly with Cristiana Mammana [108].
The reader needs no deep knowledge of special branches of mathematics, al-
though it should easier for readers who know the fundamentals of the qualitative
theory of differential equations.
The quality of English of this exposition is adversely affected by its not being
the native tongue of the author, for which fact the author asks for the reader’s kind
indulgence and understanding.
Acknowledgment: The research described in this publication, in part, was
possible due to Award No. MM1-3016 of the Moldovan Research and Development
Association (MRDA) and the U.S. Civilian Research & Development Foundation
for the Independent States of the Former Soviet Union (CRDF).

June 2003
D. N. Cheban
[email protected]
http://www.usm.md/davidcheban
xiv Global Attractors of Non-autonomous Dissipative Dynamical Systems
 Contents

Preface vii

Notations xxi

1. Autonomous dynamical systems 1

1.1 Some notions, notations and facts from theory of dynamical systems 1
1.2 Limit properties of dynamical systems . . . . . . . . . . . . . . . . . 8
1.3 Center of Levinson . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4 Dissipative systems on the local compact spaces . . . . . . . . . . . . 16
1.5 Criterions of compact dissipativity . . . . . . . . . . . . . . . . . . . 18
1.6 Local dissipative systems . . . . . . . . . . . . . . . . . . . . . . . . 25
1.7 Global attractors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
1.8 On a Problem of J. Hale . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.9 Connectedness of the Levinson’s center . . . . . . . . . . . . . . . . . 38
1.10 Weak attractors and center of Levinson . . . . . . . . . . . . . . . . 43
1.11 Asymptotic stability . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

2. Non-autonomous dissipative dynamical systems 53

2.1 On the stability of Levinson’s center . . . . . . . . . . . . . . . . . . 53
2.2 The positively stable systems . . . . . . . . . . . . . . . . . . . . . . 60
2.3 Behaviour of dissipative dynamical systems under homomorphisms . 64
2.4 Non-autonomous dynamical systems with convergence . . . . . . . . 68
2.5 Tests for convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
2.6 Global attractors of non-autonomous dynamical systems . . . . . . . 89
2.7 Global attractor of cocycles . . . . . . . . . . . . . . . . . . . . . . . 93
2.8 Global attractors of non-autonomous dynamical system with minimal
base . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
2.9 Homogeneous dynamical systems . . . . . . . . . . . . . . . . . . . . 101

xv
xvi Global Attractors of Non-autonomous Dissipative Dynamical Systems

2.10 Power-law asymptotic of homogeneous systems . . . . . . . . . . . . 107

2.11 Linear systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

3. Analytic dissipative systems 119

3.1 Skew-product dynamical systems and cocycles . . . . . . . . . . . . . 119
3.2 C-analytic systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
3.3 Converse of Lyapunov’s theorem for C-analytic systems . . . . . . . 128
3.4 On the structure of compact attracting sets of C-analytic systems . . 132
3.5 Dynamical systems in spaces of sections . . . . . . . . . . . . . . . . 136
3.6 Quasi-periodic solutions . . . . . . . . . . . . . . . . . . . . . . . . . 139
3.7 The analogy of Cameron-Johnson’s theorem . . . . . . . . . . . . . . 142
3.8 Almost periodic solutions of the weak nonlinear dissipative systems . 145

4. The structure of the Levinson center of system with the condition

of the hyperbolicity 155
4.1 The chain recurrent motions . . . . . . . . . . . . . . . . . . . . . . . 155
4.2 The spectral decomposition of the Levinson’s center . . . . . . . . . 157
4.3 One-dimensional systems with hyperbolic center . . . . . . . . . . . 159
4.4 The dissipative cascades . . . . . . . . . . . . . . . . . . . . . . . . . 164
4.5 The periodic dissipative systems . . . . . . . . . . . . . . . . . . . . 167

5. Method of Lyapunov functions 171

5.1 Criterions of dissipativity in term of Lyapunov functions . . . . . . . 171
5.2 Some criterions of dissipativity of differential equations . . . . . . . . 178
5.3 Theorem of Barbashin–Krasovskii for non-autonomous dynamical
systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
5.4 Equations with convergence . . . . . . . . . . . . . . . . . . . . . . . 192
5.5 Dissipativity and convergence of some equations of 2nd and
3rd order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
5.6 Construction of Lyapunov function for homogeneous systems . . . . 208
5.7 Differentiable homogeneous systems . . . . . . . . . . . . . . . . . . 213
5.8 Global attractors of quasi-homogeneous systems . . . . . . . . . . . . 220

6. Dissipativity of some classes of equations 225

6.1 Difference equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
6.2 Equations with impulse . . . . . . . . . . . . . . . . . . . . . . . . . 228
6.3 Convergent periodic equations with impulse . . . . . . . . . . . . . . 232
6.4 Asymptotic stability of linear functional differential equations . . . . 236
6.5 Convergence of monotone evolutionary equations . . . . . . . . . . . 238
6.6 Global attractors of non-autonomous Lorenz systems . . . . . . . . . 244
 Contents xvii

6.6.1 Non-autonomous Lorenz systems . . . . . . . . . . . . . . . . 245

6.6.2 Non-autonomous dissipative dynamical systems and their
attractors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
6.6.3 Almost periodic and recurrent solutions of non-autonomous
Lorenz systems . . . . . . . . . . . . . . . . . . . . . . . . . . 251
6.6.4 Uniform averaging principle . . . . . . . . . . . . . . . . . . . 253
6.6.5 Global averaging principle for the non-autonomous Lorenz
systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259

7. Upper semi-continuity of attractors 263

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
7.2 Maximal compact invariant sets . . . . . . . . . . . . . . . . . . . . . 263
7.3 Upper semi-continuity . . . . . . . . . . . . . . . . . . . . . . . . . . 266
7.4 Connectedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
7.5 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
7.5.1 Quasi-homogeneous systems . . . . . . . . . . . . . . . . . . . 274
7.5.2 Monotone systems . . . . . . . . . . . . . . . . . . . . . . . . 276
7.5.3 Quasi-linear systems . . . . . . . . . . . . . . . . . . . . . . . 278
7.5.4 Non-autonomously perturbed systems . . . . . . . . . . . . . 279
7.5.5 Non-autonomous 2D Navier-Stokes equations . . . . . . . . . 280
7.5.6 Quasi-linear functional-differential equations . . . . . . . . . . 284

8. The relationship between pullback, forward and global attractors 287

8.1 Pullback, forward and global attractors . . . . . . . . . . . . . . . . 288
8.2 Asymptotic stability in α-condensing semi–dynamical systems . . . . 294
8.3 Uniform pullback attractors and global attractors . . . . . . . . . . . 299
8.4 Examples of uniform pullback attractors . . . . . . . . . . . . . . . . 300
8.4.1 Periodic driving systems . . . . . . . . . . . . . . . . . . . . . 300
8.4.2 Pullback attractors with singleton component sets . . . . . . 302
8.4.3 Distal dynamical systems . . . . . . . . . . . . . . . . . . . . 304

9. Pullback attractors of C-analytic systems 307

9.1 C-analytic cocycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
9.2 Some general facts about non-autonomous dynamical systems . . . . 310
9.3 Positively uniformly stable cocycles . . . . . . . . . . . . . . . . . . . 314
9.4 The compact global pullback attractors of C-analytic cocycles with
compact base . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
9.5 The uniform dissipative cocycles with noncompact base . . . . . . . 319
9.6 The compact and local dissipative cocycles with noncompact base . 323
9.7 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
xviii Global Attractors of Non-autonomous Dissipative Dynamical Systems

9.7.1 ODEs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326

9.7.2 Caratheodory differential equations . . . . . . . . . . . . . . . 328
9.7.3 ODEs with impulses . . . . . . . . . . . . . . . . . . . . . . . 329

10. Pullback attractors under discretization 331

10.1 Non-autonomous dynamical systems and pullback attractors . . . . . 333
10.2 Non-autonomous quasi-linear differential equation . . . . . . . . . . 334
10.3 Cocycle property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
10.4 Main result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
10.4.1 Existence of an absorbing set . . . . . . . . . . . . . . . . . . 341
10.4.2 Upper semi–continuity of the pullback attractor component
sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
10.4.3 Upper semi–continuous convergence of the discretized
pullback attractors . . . . . . . . . . . . . . . . . . . . . . . . 345
10.4.4 Upper semi–continuous convergence of the discretized
global attractors . . . . . . . . . . . . . . . . . . . . . . . . . 346
10.5 Singleton set-valued pullback attractor case . . . . . . . . . . . . . . 348
10.6 Appendix: Proof of Lemma 10.4 . . . . . . . . . . . . . . . . . . . . 353

11. Global attractors of non-autonomous Navier-Stokes equations 357

11.1 Non-autonomous Navier-Stokes equations . . . . . . . . . . . . . . . 358
11.2 Attractors of non-autonomous dynamical systems . . . . . . . . . . . 367
11.3 Almost periodic and recurrent solutions of non-autonomous Navier-
Stokes equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
11.4 Uniform averaging for a finite interval . . . . . . . . . . . . . . . . . 374
11.5 The global averaging principle for Navier-Stokes equations . . . . . . 379

12. Global attractors of V -monotone dynamical systems 385

12.1 Global attractors of V -monotone NDS . . . . . . . . . . . . . . . . . 385
12.2 On the structure of Levinson center of V -monotone NDS . . . . . . . 389
12.3 Almost periodic solutions of V -monotone systems . . . . . . . . . . . 391
12.4 Pullback attractors of V -monotone NDS . . . . . . . . . . . . . . . . 394
12.5 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395
12.5.1 Finite-dimensional systems . . . . . . . . . . . . . . . . . . . . 395
12.5.2 Caratheodory’s differential equations . . . . . . . . . . . . . . 399
12.5.3 ODEs with impulse . . . . . . . . . . . . . . . . . . . . . . . . 400
12.5.4 Evolution equations with monotone operators . . . . . . . . . 402

13. Linear almost periodic dynamical systems 407

13.1 Bounded motions of linear systems . . . . . . . . . . . . . . . . . . . 407
 Contents xix

13.2 Bounded solutions of linear equations . . . . . . . . . . . . . . . . . 416

13.3 Finite-dimensional systems . . . . . . . . . . . . . . . . . . . . . . . 422
13.4 Relationship between different types of stability . . . . . . . . . . . . 426
13.5 Linear α-condensing systems . . . . . . . . . . . . . . . . . . . . . . 434
13.6 Exponential stable systems . . . . . . . . . . . . . . . . . . . . . . . 437
13.7 Linear system with a minimal base . . . . . . . . . . . . . . . . . . . 439
13.8 Some classes of uniformly exponentially stable equations . . . . . . . 441
13.9 Linear periodic systems . . . . . . . . . . . . . . . . . . . . . . . . . 450
13.9.1 Exponential stable linear periodic dynamical systems . . . . . 451
13.9.2 Some classes of linear uniformly exponentially stable
periodic differential equations . . . . . . . . . . . . . . . . . . 456

14. Triangular maps 461

14.1 Triangular maps and non-autonomous dynamical systems . . . . . . 462
14.2 Linear non-autonomous dynamical systems . . . . . . . . . . . . . . 463
14.3 Quasi-linear non-autonomous dynamical systems . . . . . . . . . . . 467
14.4 Global attractors of quasi-linear triangular systems . . . . . . . . . . 469
14.5 Almost periodic and recurrent solutions . . . . . . . . . . . . . . . . 475
14.6 Pseudo recurrent solutions . . . . . . . . . . . . . . . . . . . . . . . . 477
14.7 Chaos in triangular maps . . . . . . . . . . . . . . . . . . . . . . . . 478

Bibliography 481

Index 501
xx Global Attractors of Non-autonomous Dissipative Dynamical Systems
 Notations

∀ for all;
∃ there exists;
:= is equal (coincide) by the definition;
0 the number 0 and also the null element of all additive group
(semigroup);
N the set of all natural numbers;
Z the set of all integers;
Q the set of all rational numbers;
R the set of all real numbers;
C the set of all complex numbers;
S one of the sets R or Z;
S+ (S− ) the set of all nonnegative (non-positive) numbers from the
set S;
X ×Y the product space of two sets;
n
M the product space of n copies of the set M ;
n
E the real or complex n-dimensional Euclidean space;
{xn } a sequence;
x∈X x is an element of the set X;
∂X the boundary of the set X;
X⊆Y the set X is a part of or coincides with the set Y ;
S
X Y the union of the two sets X and Y ;
X \Y the complement of the set Y in the set X;
T
X Y the intersection of the sets X and Y ;
∅ the empty-set;
(X, ρ) the complete metric space with the distance ρ;

xxi
xxii Global Attractors of Non-autonomous Dissipative Dynamical Systems

2X the space of all closed bounded subsets of the metric

space X;
M the closure of the set M ;
−1
f the inverse mapping of the mapping f ;
f (M ) the image of the set M ⊆ X by the mapping f : X → Y ,
i.e., {y ∈ Y : y = f (x), x ∈ M };
f ◦g the composition of the mappings f and g, i.e., (f ◦ g)(x) =
f (g(x));
f |M the restriction of the mapping f on the set M ;
f (·, x) the partial mapping, defined by the function f if the second
variable takes a value x;
IdX the identity mapping from the X into itself;
Im(f ) range of values of the function f ;
D(f ) the domain of the definition of the function f ;
|x| or ||x|| the norm of the element x;
(x, y) the ordered pair;
C(X, Y ) the set of all continuous mappings from X into Y with
compact-open topology;
C k (U, M ) the set of all k-time continuously differentiable functions
f : U → M;
f :X→Y the mapping from X into Y ;
B(M, ε) the open ε-neighborhood of the set M in the metric
space X;
B[M, ε] the closed ε-neighborhood of the set M in the metric
space X;
{x, y, ..., z} the set consisting from the elements x, y, ..., z;
1, n the set consisting from the elements 1, 2, ..., n;
{x ∈ X|R(x)} the set of all elements of the set X, possessing the
property R;
f −1 (M ) the pro-image of the set M ⊆ Y by the mapping f : X → Y ,
i.e. {x ∈ X : f (x) ∈ M };
F (t, ·) := f t the partial mapping, defined by the function f if the first
variable takes the value t;
ρ(ξ, η) the distance in the metric space X;
lim xn the limit of the sequence;
n→+∞
 Notations xxiii

εk ↓ 0 the monotone decreasing to the 0 sequence;

lim f (x) the limit of the mapping f as x → a;
x→a
S
{Mλ : λ ∈ Λ} the union of the family of sets {Mλ }λ ∈ Λ;
T
{Mλ : λ ∈ Λ} the intersection of the family of sets {Mλ }λ ∈ Λ;

(H, h·, ·i) Hilbert space with the scalar product h·, ·i;
dϕ
ϕ̇ := dt the derivative of the function ϕ;

cardM the cardinality of the set M ;

C(X) the family of all compacts from X;
B(X) the family of all bounded subsets from X;
λ(A) the measure of non-compactness of the set A ∈ B(X);
(X, T, π) a dynamical system;
(X, P ) cascade, generated by the mapping P ;
ωx (αx ) ω(α)-limit set of the point x;
ϕx the whole trajectories of (X, T, π) with condition ϕx (0) = x.
Φx the set of all whole trajectories ϕx of (X, T, π).
α ϕx α-limit set of the whole trajectory ϕx ∈ Φx ;
W s (M ) the stable manifold of the set M ;
+
M st.L the set M is stable in sense of Lagrange
in the positive direction;
Dx+ (Jx+ ) positive extension (positive limit extension) of the point x;
t
xt = π x = π(t, x) the state of point x at the moment t;
pri : X1 × X2 → Xi the projection of X1 × X2 on the i-th (i = 1, 2) component
Xi ;
D+ (M ) the positive extension of the set M ;
+
J (M ) the positive limit extension of the set M ;
Σ+
x the positive semi-trajectory of the point x;
+
Σ (M ) the positive semi-trajectory of the set M ;
+
H (x) the closure of positive semi-trajectory of the point x;
Σx the trajectory of the point x;
H(x) the closure of trajectory of the point x;
Ω the closure of the union of all ω-limit points of (X, T, π);
β(A, B) the semi-deviation of the set A from the set B (A, B ∈ 2X );
 Chapter 1

Autonomous dynamical systems

1.1 Some notions, notations and facts from theory of dynamical

systems

1. Below we give some notions, denotations and facts from theory of dynamical
systems [32, 33, 258, 261, 300, 302, 304] which we will use in this book.
Let X be a topological space, R (Z) be a group of real (integer) numbers, R+
(Z+ ) be a semi-group of the nonnegative real (integer) numbers, S be one of the
two sets R or Z and T ⊆ S (S+ ⊆ T) be a sub-semigroup of additive group S.

Definition 1.1 Triplet (X, T, π), where π : T × X → X is a continuous mapping

satisfying the following conditions:

π(0, x) = x; (1.1)

π(s, π(t, x)) = π(s + t, x); (1.2)

is called a dynamical system. If T = R (R+ ) or Z (Z+ ), then the dynamical system

(X, T, π) is called a group (semi-group). In the case, when T = R+ or R the
dynamical system (X, T, π) is called a flow, but if T ⊆ Z, then (X, T, π) is called a
cascade (discrete flow).

Sometimes, briefly, we will write xt instead of π(t, x).

Below X will be a complete metric space with metric ρ.

Definition 1.2 The function π(·, x) : T → X is called a motion passing through

the point x at the moment t = 0 and the set Σx := π(T, x) is called a trajectory of
this motion.

Definition 1.3 A nonempty set M ⊆ X is called positively invariant (nega-

tively invariant, invariant) with respect to dynamical system (X, T, π) or, sim-
ple, positively invariant (negatively invariant, invariant), if π(t, M ) ⊆ M (M ⊇
π(t, M ), π(t, M ) = M ) for every t ∈ T.

1
2 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Definition 1.4 A closed positively invariant set, which does not contain own
closed positively invariant subset, is called minimal.

It easy to see that every positively invariant minimal set is invariant.

Definition 1.5 A closed positively invariant (invariant) set is called indecompos-

able, if it can not be represented in the form of union of two nonempty disjoint
positively invariant (invariant) subsets.

Definition 1.6 Let M ⊆ X. The set

\ [
ω(M ) := π(τ, M )
t≥0 τ ≥t

is called ω-limit for M . If T = S, then the set

\ [
α(M ) := π(τ, M )
t≤0 τ ≤t

is called α-limit for M .

Denote by Σ(M ) := π(S, M ) (Σ+ (M ) := π(S+ , M )) and H(M ) := π(S, M )

( H + (M ) := π(S+ , M ) ). If M = {x}, then we put ωx := ω({x}), αx :=
α({x}), Σ+ + + +
x := Σ ({x}), H (x) := H ({x}) and H(x) := H({x}).
Let (X, T, π) be a dynamical system.

Definition 1.7 The point x ∈ X is called a τ -periodic (τ > 0, τ ∈ T) point, if

xt = x (xτ = x respectively) for all t ∈ T, where xt := π(t, x).

Definition 1.8 The number τ ∈ T is called ε > 0 shift (almost period) if

ρ(xτ, x) < ε (respectively ρ(x(τ + t), xt) < ε for all t ∈ T).

Definition 1.9 The point x ∈ X is called almost recurrent (almost periodic) if

for any ε > 0 there exists a positive number l such that on any segment of length
l, will be found an ε shift (almost period) of point x ∈ X.

Definition 1.10 If a point x ∈ X is almost recurrent and the set H(x) =

{xt | t ∈ T} is compact, then x is called recurrent.

Definition 1.11 The point x ∈ X is called stable in the sense of Poisson in the
positive direction if x ∈ ωx . If the dynamical system (X, T, π) is a two-sided one
and x ∈ αx , then the point x is called stable in the sense of Poisson in the negative
direction. If the point x is stable in the sense of Poisson in both directions, it is
called Poisson stable.

Definition 1.12 The continuous mapping ϕ : S → X with the properties:

 Autonomous dynamical systems 3

(1) ϕ(0) = x;
(2) π t ϕ(s) = ϕ(t + s) for all t ∈ T and s ∈ S
is called a whole trajectory of one-sided dynamical system (X, T1 , π).

Denote by Φx the family of all whole trajectory of dynamical system (X, T1 , π)

passing through the point x ∈ X at the moment t = 0.

Definition 1.13 The motion π(·, x) : T1 → X of one-sided dynamical system

(X, T1 , π) is called extendable on S, if there exists a whole trajectory ϕ passing
through point x.

We denote by αϕx := {y : ∃tn → −∞, tn ∈ S and ϕ(tn ) → y}, where ϕ is an

extension on S the motion π(·, x).

Definition 1.14 The set W s (Λ) (W u (Λ)), defined by equality

W s (Λ) := {x ∈ X| lim ρ(xt, Λ) = 0}

t→+∞

(W u (Λ) := {x ∈ X| lim ρ(xt, Λ) = 0}),

t→−∞

is called a stable (unstable) manifold of set Λ ⊆ X.

Definition 1.15 The set M is called:

− orbital stable, if for every ε > 0 there exists δ = δ(ε) > 0 such that ρ(x, M ) < δ
implies ρ(xt, M ) < ε for all t ≥ 0;
− attracting, if there exists γ > 0 such that B(M, γ) ⊂ W s (M );
− asymptotic stable, if it is orbital stable and attracting;
− global asymptotic stable, if it is asymptotic stable and W s (M ) = X;
− uniform attracting, if there exists γ > 0 such that

lim sup ρ(xt, M ) = 0.

t→+∞ x∈B(M,γ)

Theorem 1.1 [305] Let M ⊆ X be a nonempty compact invariant set, then every
motion in M is extendable on S.

Theorem 1.2 [305] The compact set M ⊆ X is negative invariant, if and only if
for every point x ∈ M there exists a whole trajectory ϕ passing through point x ∈ M
such that ϕ(t) ∈ M for all t ≤ 0.

Theorem 1.3 [305] Union and closure of positively invariant (negatively invari-
ant, invariant) sets is positively invariant (negatively invariant, invariant).

Theorem 1.4 [305] The following statements are true:

4 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(1) ∀ x ∈ X the set ωx is positively invariant;

(2) ∀ x ∈ X and ϕ ∈ Φx the set αϕx is positively invariant.

2. Below we indicate one general method of construction of dynamical systems

on the space of continuous functions. In this way we will get many well known
dynamical systems on the functional spaces (see, for example, [32, 300]).
Let (X, T, π) be a dynamical system on X, Y be a complete pseudo metric
space and P be a family of pseudo metric on Y . Denote by C(X, Y ) the family
of all continuous functions f : X → Y equipped with the compact-open topology.
This topology is given by the following family of pseudo metric {dpK } (p ∈ P, K ∈
C(X)), where

dpK (f, g) := sup p(f (x), g(x))

x∈K

and C(X) a family of all compact subsets of X. We define for all τ ∈ T the mapping
στ : C(X, Y ) → C(X, Y ) by the following equality: (στ f )(x) := f (π(τ, x)) (x ∈ X).
We note that the family of mappings{στ : τ ∈ T} possesses the following properties:

a. σ0 = idC(X,Y ) ;
b. ∀τ1 , τ2 ∈ T στ1 ◦ στ2 = στ1 +τ2 ;
c. ∀τ ∈ T στ is continuous.

Lemma 1.1 The mapping σ : T × C(X, Y ) → C(X, Y ), defined by the equality

σ(τ, f ) := στ f (f ∈ C(X, Y ), τ ∈ T), is continuous.

Proof. Let f ∈ C(X, Y ), τ ∈ T and {fν }, {τν } are arbitrary directions which
converge to f and τ respectively. Then

dpK (σ(τν , fν ), σ(τ, f )) = sup p(σ(τν , fν )(x), σ(τ, f )(x)) =

x∈K

sup p((fν (π(τν , x)), f (π(τ, x))) : x ∈ K} ≤

x∈K

sup p(fν (π(τν , x)), f (π(τν , x))) + sup p(f (π(τν , x)), f (π(τ, x))) ≤
x∈K x∈K

sup p(fν (π(s, x)), f (π(s, x))) + sup p(f (π(τν , x)), f (π(τ, x)))
x∈K, s∈Q x∈K

≤ sup p(fν (m), f (m)) + sup p(f (π(τν , x)), f (π(τ, x))) =
m∈π(Q,K) x∈K

dpπ(Q,K) (fν , f ) + sup p(f (π(τν , x)), f (π(τ, x))), (1.3)

x∈K
 Autonomous dynamical systems 5

where Q = {τν }. Passing to limit in the inequality (1.3) we obtain the necessary
statement. 

Corollary 1.1 The triple (C(X, Y ), T, σ) is a dynamical system on C(X, Y ).

Remark 1.1 Lemma 1.1 is true, if X is an arbitrary Hausdorff topological space.

Consider now same examples of dynamical systems of form (C(X, Y ), T, σ) useful

in the applications.

Example 1.1 Let X = T and we denote by (X, T, π) a dynamical system

on T, where π(t, x) := x + t. Dynamical system (C(T, Y ), T, σ) is called Bebu-
tov’s dynamical system [300] (dynamical system of translations or shifts dynamical
system).

Definition 1.16 We will say that the function ϕ ∈ C(T, Y ) possesses the prop-
erty A, if with this property possesses the motion σ(·, ϕ) : T → C(T, Y ) in the
dynamical system of Bebutov (C(T, Y ), T, σ), generated by the function ϕ. In qual-
ity of property A may be periodicity, almost periodicity, recurrence, asymptotic
almost periodicity etc.

Example 1.2 Let X := T × W , where W some metric space and by (X, T, π)

we denote a dynamical system on X defined in the following way: π(t, (s, w)) :=
(s + t, w). Using the general method proposed above we can define on C(T × W, Y )
a dynamical system of translations (C(T × W, Y ), T, σ).
The function f ∈ C(T × W, Y ) is called almost periodic (recurrent, asymptotic
almost periodic etc) with respect to t ∈ T uniform on w on every compacts from
W , if the motion σ(·, f ) is almost periodic (recurrent, asymptotic almost periodic,
etc.) in the dynamical system (C(T × W, Y ), T, σ).

Example 1.3 Let W := Cn , Y := Cm and A(T × Cn , Cm ) be a family of all

functions f ∈ C(T × Cn , Cm ) which are holomorphic with respect to second vari-
able. It easy to verify that the set A(T × Cn , Cm ) is closed and invariant subset of
dynamical system (C(T × Cn , Cm ), T, σ) and, consequently, on A(T × Cn , Cm ) it is
induced a dynamical system of translations (A(T × Cn , Cm ), T, σ).

Example 1.4 Let W and X be complete pseudo metric spaces [213]. Denote by
P(T × W, X) a set of all continuous functions f : T × W → X which are bounded
on every bounded subset from T × W and continuous w.r.t. t ∈ T uniformly on
w on bounded subsets W equipped with the topology of uniform convergence on
bounded subsets from T×W . This topology can by generate by the following family
of pseudo metrics {dpB } (p ∈ P, B ∈ B(T × W )), where

dpB (f, g) = sup p(f (t, w), g(t, w)), (1.4)

(t,w)∈B
6 Global Attractors of Non-autonomous Dissipative Dynamical Systems

B(T × W ) is a family of all bounded subsets of T × W and P is a family of pseudo

metric on X. We note that the set P(T × W, X) endowed with the family of
pseudo metric {dpB |p ∈ P, B ∈ B(T × W )} become a complete pseudo metric
space, invariant with respect to translations on t ∈ T. For each τ ∈ T we denote
by fτ the translation of function f ∈ P(T × W, X) on τ w.r.t. variable t ∈ T,
i.e. fτ (t, w) = f (t + τ, w) ( (t, w) ∈ T × W ). Now we define a mapping σ :
T × P(T × W, X) → P(T × W, X) as following: σ(τ, f ) := fτ for all f ∈ P(T × W, X)
and τ ∈ T. It is easy to see that σ(0, f ) = f and σ(τ2 , σ(τ1 , f )) = σ(τ1 + τ2 , f ) for
all f ∈ P(T × W, X) and τ1 , τ2 ∈ T. Using the same reasoning as in the lemma 1.1
it is possible to verify the mapping σ is continuous and, consequently, the triplet
(P(T × W, X), T, σ) is a dynamical system on P(T × W, X).

Remark 1.2 If the function f ∈ C(T × W, X) is uniform continuous on every

bounded subset from T×W , then f ∈ P(T×W, X). Denote by U (T×W, X) the set of
all functions f ∈ C(T × W, X) which are bounded and uniform continuous on every
bounded subset from T×W. Then the set U (T×W, X) is invariant w.r.t. translations
on t ∈ T and it is closed in P(T × W, X) and, consequently, on U (T × W, X) it is
induced a dynamical system of translations (U (T × W, X), T, σ) (see, for example,
[32, 300] and [302]).

Definition 1.17 Let (X, T1 , π) and (Y, T2 , σ) (S+ ⊆ T1 ⊆ T2 ⊆ S) be

two dynamical systems. The mapping h : X → Y is called a homomorphism
(respectively isomorphism) of dynamical system (X, T1 , π) on (Y, T2 , σ), if the
mapping h is continuous (respectively homeomorphic) and h(π(x, t)) = σ(h(x), t)
( t ∈ T1 , x ∈ X). In this case a dynamical system (X, T1 , π) is an extension of
dynamical system (Y, T2 , σ) by homomorphism h, but a dynamical system (Y, T2 , σ)
is called a factor of dynamical system (X, T1 , π) by homomorphism h. Dynamical
system (Y, T2 , σ) is called also a base of extension (X, T1 , π).

Definition 1.18 The triplet h(X, T1 , π), (Y, T2 , σ), hi, where h is a homo-
morphism from (X, T1 , π) on (Y, T2 , σ) and (X, h, Y ) is a local-trivial fibering [190],
is called a non-autonomous dynamical system.

Remark 1.3 In the latter years in the works of I.U.Bronsteyn and his col-
laborators (see, for example, [32, 33, 34, 158]) an extension is called a triplet
h(X, T, π), (Y, T, h), hi, i.e. the object which we name here a non-autonomous
dynamical system.

Definition 1.19 The triplet hW, ϕ, (Y, T2 , σ)i (or shortly ϕ), where (Y, T2 , σ) is
a dynamical system on Y , W is a complete metric space and ϕ is a continuous
mapping from T1 × W × Y in W , possessing the following conditions:
a. ϕ(0, u, y) = u (u ∈ W, y ∈ Y );
 Autonomous dynamical systems 7

b. ϕ(t + τ, u, y) = ϕ(τ, ϕ(t, u, y), σ(t, y)) (t, τ ∈ T1 , u ∈ W, y ∈ Y ),

is called [6, 290, 291] a cocycle on (Y, T2 , σ) with fiber W .

Definition 1.20 Let X := W × Y and we define a mapping π : X × T1 → X

as following: π((u, y), t) := (ϕ(t, u, y), σ(t, y)) (i.e. π = (ϕ, σ)). Then it easy
to see that (X, T1 , π) is a dynamical system on X which is called a skew-product
dynamical system [2, 292] and h = pr2 : X → Y is a homomorphism from (X, T1 , π)
on (Y, T2 , σ) and, consequently, h(X, T1 , π), (Y, T2 , σ), hi is a non-autonomous
dynamical system.

Thus, if we have a cocycle hW, ϕ, (Y, T2 , σ)i on dynamical system (Y, T2 , σ) with
the fiber W , then it generates a non-autonomous dynamical system h(X, T1 , π),
(Y, T2 , σ), hi (X := W × Y ), which is called a non-autonomous dynamical system,
generated by cocycle hW, ϕ, (Y, T2 , σ)i on (Y, T2 , σ).
Non-autonomous dynamical systems (cocycles) play a very important role in the
study of non-autonomous evolutionary differential equations. Under appropriate
assumptions every non-autonomous differential equation generates some cocycle
(non-autonomous dynamical system). Below we give an example of this type.

Example 1.5 Let E n be an n-dimensional real or complex Euclidean space. Let

us consider a differential equation

u0 = f (t, u), (1.5)

where f ∈ C(R × E n , E n ). Along with equation (1.5) we consider its H-class

[32, 137, 238, 300, 302], i.e., the family of equations

v 0 = g(t, v), (1.6)

where g ∈ H(f ) = {fτ : τ ∈ R}, fτ (t, u) = f (t + τ, u) for all (t, u) ∈ R × E n and by

bar we denote the closure in C(R × E n , E n ). We will suppose also that the function
f is regular, i.e. for every equation (1.6) the conditions of existence, uniqueness and
extendability on R+ are fulfilled. Denote by ϕ(·, v, g) the solution of equation (1.6),
passing through the point v ∈ E n at the initial moment t = 0. Then it is correctly
defined a mapping ϕ : R+ × E n × H(f ) → E n , verifying the following conditions
(see, for example, [32, 290, 291]):

1) ϕ(0, v, g) = v for all v ∈ E n and g ∈ H(f );

2) ϕ(t, ϕ(τ, v, g), gτ ) = ϕ(t + τ, v, g) for every v ∈ E n , g ∈ H(f ) and t, τ ∈ R+ ;
3) the mapping ϕ : R+ × E n × H(f ) → E n is continuous.

Denote by Y := H(f ) and (Y, R+ , σ) a dynamical system of translations

(semigroup system) on Y , induced by dynamical system of translations (C(R ×
E n , E n ), R, σ). The triplet hE n , ϕ, (Y, R+ , σ)i is a cocycle on (Y, R+ , σ) with the
8 Global Attractors of Non-autonomous Dissipative Dynamical Systems

fiber E n . Thus the equation (1.5) generates a cocycle hE n , ϕ, (Y, R+ , σ)i and a
non-autonomous dynamical system h(X, R+ , π), (Y, R+ , σ), hi, where X := E n × Y ,
π := (ϕ, σ) and h := pr2 : X → Y .

1.2 Limit properties of dynamical systems

Let M ⊆ X and ω(M ) be ω–limit set of M . Directly from the definition of ω(M )
follows

Lemma 1.2 The following statements take place:

(1) the point y ∈ ω(M ) if and only if there exist sequences {xn } ⊆ M and {tn } ⊆ T
such that tn → +∞ and y = lim xn tn ;
n→+∞
(2) the set ω(M ) is closed and positively invariant;
(3) if A ⊆ B then ω(A) ⊆ ω(B);
(4) ω(A ∪ B) ⊆ ω(A) ∪ ω(B) for any pair of subsets A and B from X;
(5) if set A is positively invariant (negatively invariant, invariant) then ω(A) ⊆ A
( A ⊆ ω(A), ω(A) = A);
(6) ∪{ωx | x ∈ M } ⊆ ω(M );
(7) if Σ+ (M ) is relatively compact, then ω(M ) is invariant.

Lemma 1.3 Let B ⊆ X then the following conditions are equivalent:

(1) for every {xk } ⊆ B and tk → +∞ sequence {xk tk } is relatively compact;

(2)(a) ω(B) is not empty and is compact;
(b) ω(B) is invariant and the equality below takes place

lim sup ρ(xt, ω(B)) = 0; (1.7)

t→+∞ x∈B

(3) there exists a non empty compact K ⊆ X such that

lim sup ρ(xt, K) = 0.

t→+∞ x∈B

Proof. Let us show that from 1. follows 2. Let {xk } ⊆ B and tk → +∞,
then according to 1. the sequence {xk tk } can be considered convergent. Assume
x = lim xk tk , then x ∈ ω(B) and consequently ω(B) 6= ∅. Let us show ω(B) is
k→+∞
compact. Let εk ↓ 0 and {yk } ⊆ ω(B), then there exist xk ∈ B and tk ≥ k such
that

ρ(xk tk , yk ) < εk .

According to the condition (1), the sequence {xk tk } is relatively compact and since
εk ↓ 0, {yk } also is relatively compact. From the definition of ω(B) follows its
 Autonomous dynamical systems 9

positive invariance and consequently for ω(B) to be invariant it is sufficient to

show that it is negatively invariant. Let y ∈ ω(B) and t ∈ T, then there exist
{xk } ⊆ B and tk → +∞ such that y = lim xk tk = lim xk (tk − t + t) =
k→+∞ k→+∞
lim π t (xk (tk − t)). As tk − t → +∞ then according to condition (1), the sequence
k→+∞
{xk (tk − t)} can be considered convergent. Assume yt = lim xk (tk − t), then
k→+∞
y = π t yt and yt ∈ ω(B), i.e. y ∈ π t ω(B). Like so the invariance of ω(B) is proved.
Now let us show that the equality (1.7) takes place. If we suppose that (1.7) does
not take place, then there will exist ε0 > 0, tk → +∞ and xk ∈ B such that

ρ(xk tk , ω(B)) ≥ ε0 . (1.8)

According to the condition (2), the sequence {xk tk } can be considered convergent.
Let y = lim xk tk , then y ∈ ω(B). On the other hand, passing to the limit in (1.8)
k→+∞
when k → +∞ we will obtain y ∈ / ω(B). This contradiction completes the proof of
the implication condition (1) ⇒ condition (2).
It is evident that (2) ⇒ (3) and (3) ⇒ (1). The lemma is proved. 

Corollary 1.2 Let M ⊆ X be nonempty and Σ+ (M ) be relatively compact, then

ω(M ) 6= ∅, is compact, invariant and

lim sup ρ(xt, ω(M )) = 0. (1.9)

t→+∞ x∈M

The statement inverse to the corollary 1.2 takes place, but before formulating it
we will review the notion of the measure of non-compactness.

Definition 1.21 The mapping λ : B(X) → R+ , satisfying the following condi-

tions:
(1) λ(A) = 0 if and only if A ∈ B(X) is relatively compact;
(2) λ(A ∪ B) = max(λ(A), λ(B)) for every A, B ∈ B(X).
is called [175, 179, 278] a measure of non-compactness on X.

Definition 1.22 The measure of non-compactness of Kuratowsky λ : B(X) →

R+ is defined by the equality λ(A) := inf{ε > 0 | A admits finite ε-covering }.

Lemma 1.4 Let M ⊆ X be nonempty and relatively compact. If set ω(M ) is

nonempty, compact and the equality (1.9) tales place, then Σ+ (M ) is relatively
compact.

Proof. Let ε > 0, then from the equality (1.9) follows the existence of a positive
number L(ε) such that
[
Mε := π t M ⊆ B(ω(M ), ε). (1.10)
t≥L(ε)
10 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Denote by λ(K) the measure of non-compactness of Kuratowsky of the set K. From

the inclusion (1.10) follows that

λ(Σ+ (M )) = λ(π(M, [0, L(ε)]) ∪ Mε ) =

max(λ(π(M, [0, Lε ]), λ(Mε )) = λ(Mε ) ≤ 2ε.

From this we obtain the equality λ(Σ+ (M )) = 0. The lemma is proved. 

Theorem 1.5 Let M ⊆ X be nonempty and relatively compact. For the set
Σ+ (M ) to be compact it is necessary and sufficient the fulfillment of the following
3 conditions:

(1) ω(M ) 6= ∅;
(2) ω(M ) is compact;
(3) the equality (1.9) takes place.

The formulated statement follows from the Corollary 1.2 and Lemma 1.4.

1.3 Center of Levinson

Let M be some family of subsets from X.

Definition 1.23 Dynamical system (X, T, π) will be called M-dissipative if for

every ε > 0 and M ∈ M there exists L(ε, M ) > 0 such that π t M ⊆ B(K, ε) for any
t ≥ L(ε, M ), where K is a certain fixed subset from X depending only on M. In
this case K we will call the attractor for M.

For the applications the most important ones are the cases when K is bounded or
compact and M = {{x} | x ∈ X} or M = C(X), or M = {B(x, δx ) | x ∈ X, δx > 0},
or M = B(X).

Definition 1.24 The system (X, T, π) is called:

− point dissipative if there exist K ⊆ X such that for every x ∈ X

lim ρ(xt, K) = 0; (1.11)

t→+∞

− compact dissipative if the equality (1.11) takes place uniformly w.r.t. x on the
compacts from X;
− locally dissipative if for any point p ∈ X there exist δp > 0 such that the
equality (1.11) takes place uniformly w.r.t. x ∈ B(p, δp );
− bounded dissipative if the equality (1.11) takes place uniformly w.r.t. x on
every bounded subset from X.
 Autonomous dynamical systems 11

As we have already mentioned above, for applications our main interest rests on
cases in which K is compact and bounded. According to this the system (X, T, π)
is called point k (b)-dissipative if (X, T, π) is point dissipative and the set K from
(1.11) is compact (bounded). Also like that is defined the notion of the compact k
(b)-dissipative system and of other types of k (b)-dissipativity.
From the definitions above follows that bounded k (b)-dissipativity implies local
k (b)-dissipativity (see Lemma 1.5), local k (b)-dissipativity implies compact k (b)-
dissipativity, and compact k (b)-dissipativity implies point k(b)-dissipativity.
As it will be shown below in general case introduced by us classes are different.
We will take interest in the following tasks:
(1) What connection does exist between different types of dissipativity?
(2) Conditions under which a dissipative system admits maximal compact invariant
set.
(3) Structure of the center of Levinson (maximal compact invariant attractor) for
different classes of dynamical systems.
(4) Conditions of asymptotic and uniform asymptotic stability of the center of
Levinson.

Lemma 1.5 Let (X, T, π) be local k (b)-dissipative, then it is compact k (b)-

dissipative.

Proof. Let (X, T, π) be local k (b)-dissipative. Then there exist a nonempty

compact (bounded) set K ⊆ X, such that for any ε > 0 and x ∈ X there exist
δ(x) > 0 and l = l(ε, x) > 0 for which

ρ(yt, K) < ε (1.12)

for every t ≥ l and y ∈ B(x, δ(x)).

Let M is a nonempty compact from X. Then for every x ∈ M there exist δ =
δ(x) > 0 and l = l(ε, x) > 0 such that the inequality (1.12) takes place. Consider
the open covering {B(x, δ(x)) | x ∈ M } of the set M . By virtue of the compactness
of M and completeness of the space X, from the constructed covering of the set M
we can extract {B(xi , δ(xi )) | i ∈ 1, m}. Assume L(ε, M ) = max{l(ε, xi ) | i ∈ 1, m},
then from (1.12) follows that ρ(xt, K) < ε for all x ∈ M and t ≥ L(ε, M ). The
lemma is proved. 

Lemma 1.6 Let M ⊆ X and the set Σ+ (M ) be relatively compact. If ω(M ) ⊆ M,

then

ω(M ) = ∩{π t M | t ∈ T}. (1.13)

Proof. Denote by I(M ) := ∩{π t M |t ∈ T}. It is clear that I(M ) ⊆ ω(M ). Let us
show that the inverse inclusion also takes place if ω(M ) ⊆ M . In fact, according to
12 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Lemma 1.3 and Corollary 1.2 the set ω(M ) is invariant and, consequently, ω(M ) =
π t ω(M ) ⊆ π t M for every t ∈ T. From what follows the inclusion ω(M ) ⊆ I(M ).
The lemma is proved. 

Remark 1.4 Below we will mainly consider k-dissipative dynamical systems.

That’s why the sign k- we will drop out everywhere where it won’t lead to mis-
understanding.

Let (X, T, π) be compact dissipative and K is a nonempty compact set that is

an attractor for compact subsets X. Then for every compact M ⊆ X the equality

lim sup ρ(xt, K) = 0

t→+∞ x∈M

holds. Hence ω(K) ⊆ K and, consequently,

J := ω(K) = ∩{π t K | t ∈ T}. (1.14)

Let us show that the set J does not depend on the choice of the set K attracting
all compact subsets of the space X. In fact, if we denote by J(K) := ω(K) and
K1 every other compact set attracting all compacts from X, then there will be
L = L(K, K1 , ε) > 0 such that π t J(K) ⊆ K1 and π t J(K1 ) ⊆ K for all t ≥ L. As
J(K) = ω(K) ⊆ K and J(K1 ) = ω(K1 ) ⊆ K1 then from the invariance of J(K1 )
and J(K) follows that J(K) ⊆ K1 , J(K1 ) ⊆ K, J(K) ⊆ π t K1 and J(K1 ) ⊆ π t K
for all t ∈ T and consequently J(K) = J(K1 ).
So, the set J defined by the equality (1.14) does not depend on the choice of
the attractor K but is defined only by inner properties of the dynamical system
(X, T, π).

Definition 1.25 The set J defined by the equality (1.14), according to [332]
we will call the center of Levinson of the compact dissipative dynamical system
(X, T, π).

Theorem 1.6 [102, 175, 232] Let (X, T, π) be compact dissipative system and let
J be its center of Levinson. Then:

(1) J is compact invariant set;

(2) J is orbitally stable;
(3) J is the attractor of the family of all compacts of X;
(4) J is the maximal compact invariant set in (X, T, π).

Proof. The first statement of the Theorem directly follows from the definition of J
and from Lemma 1.3. Let us show that the set J is orbitally stable. If we suppose
the contrary we will obtain the existence of ε0 > 0, δn → 0 (δn > 0), xn ∈ B(J, δn )
 Autonomous dynamical systems 13

and tn → +∞ such that

ρ(xn tn , J) ≥ ε0 . (1.15)

As xn ∈ B(J, δn ), δn → 0 and J is compact, without loss of generality the sequence

{xn } can be considered convergent. By virtue of the compact dissipativity of the
system (X, T, π) the set Σ+ ({xn }) is relatively compact. Note that together with the
set K the set K 0 = K ∪H + ({xn }) is the attractor for the family of all compacts from
X and consequently ω(K 0 ) = ω(K) = J. In particular, ω(H + ({xn })) ⊆ ω(K) = J.
By compactness of H + ({xn }) the sequence {xn tn } can be considered convergent.
Let p = lim xn tn , then p ∈ ω(H + ({xn })). On the other hand from (1.15) follows
n→+∞
that p ∈
/ J. The obtained contradiction proves the second statement of the theorem.
Now let M ∈ C(X). Then by virtue of the compact dissipativity of (X, T, π) the
set Σ+ (M ) is relatively compact and according to Theorem 1.5 the conditions (1)-
(2) are fulfilled. In particular, for every ε > 0 there exists L(ε) > 0 such that π t M ⊆
B(ω(M ), ε) for all t ≥ L(ε). Together with the set K the set K 0 = K ∪ ω(M ) also is
the attractor of the compact subsets from X and consequently ω(K 0 ) = ω(K) = J.
So, ω(M ) ⊆ ω(K 0 ) = J, and hence

β(π t M, J) ≤ β(π t M, ω(M )) < ε,

i.e. lim β(π t M, J) = 0 for all M ⊆ C(X) where β(A, B) is semi-deviation of the
t→+∞
set A from the set B.
At last let us prove the 4th statement of the theorem. Let J1 be compact
invariant subset from X. Then according to the 3rd statement of the theorem we
have

lim β(π t J1 , J) = 0. (1.16)

t→+∞

By virtue of the invariance of J1 the equality π t J1 = J1 holds for all t ∈ T. From

this and from the equality (1.16) we obtain J1 ⊆ J. The theorem is completely
proved. 

Denote by {Kλ |λ ∈ Λ} the family of all nonempty compact positive invariant

sets that attract all compacts from X.
The following theorem holds.

Theorem 1.7 Let (X, T, π) be compact dissipative dynamical system and let J be
its Levinson center.

J = J 0 := ∩{Kλ |λ ∈ Λ},

i.e. J is the least compact positive invariant set attracting all compacts from X.
14 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Proof. Assume

K := ∩{Kλ |λ ∈ Λ}.

First of all note that J ⊆ K and consequently K 6= ∅. In fact, for every λ ∈ Λ we

have J = ω(Kλ ) ⊆ Kλ , i.e. J ⊆ K.
Let us show now that the reverse inclusion also holds. As J attracts all com-
pacts from X and is nonempty and positive invariant, then J ∈ {Kλ |λ ∈ Λ} and
consequently K ⊆ J. The theorem is proved. 

Lemma 1.7 Let (X, T, π) be compact dissipative dynamical system, J be its Levin-
son’s center and K be a nonempty compact attracting all compacts from X, then
\
J= π t K.
t∈T

Proof. As K is the attractor of all compacts from X, then ω(K) ⊆ K and according
to Lemma 1.6
\
J := ω(K) = π t K.

The lemma is proved. 

Lemma 1.8 Let M be compact positive invariant asymptotically stable set, then
the following statements hold:

(1) the domain of attraction W s (M ) of the set M is open;

(2) the equality

lim β(π t K, M ) = 0 (1.17)

t→+∞

takes place for every compact K from W s (M ).

Proof. Show that W s (M ) is open. Really, as M is attracting set, then there

exists δ > 0 such that B(M, δ) ⊂ W s (M ). Let us show that for every point p ∈
W s (M )\B(M, δ) there exists η > 0 such that B(p, η) ⊂ W s (M ). As p ∈ W s (M ),
there will be tp > 0 such that ptp ∈ B(M, δ). By openness of B(M, δ) there is γ > 0
for which B(ptp , γ) ⊂ B(M, δ). According to continuity of the mapping π(tp , ·) :
X −→ X there exists η > 0 such that the inclusion π tp B(p, η) ⊂ B(M, δ) ⊂ W s (M )
holds and consequently the set W s (M ) is open.
Let us prove the second statement of the lemma. Let ε > 0 and K be a compact
from W s (M ). For the number ε > 0 we choose δ(ε) > 0 taking into account the
condition of the stability of M . As M attracts points from W s (M ), for every point
x ∈ K there are γ(x, ε) > 0 and l(x, ε) > 0 such that

π t B(x, γ(x, ε)) ⊆ B(M, ε) (1.18)

 Autonomous dynamical systems 15

for all t ≥ l(x, ε). By the compactness of K from the open covering {B(x, γ(x, ε)) |
x ∈ K} ⊇ K we can extract sub-covering ∪{B(xi , γ(x, ε)) | i = 1, n} ⊇ K. Assume
L(M, ε) := max{l(xi , ε) | i = 1, n}, then π t M ⊆ B(K, ε) for all t ≥ L(M, ε), i.e. M
attracts K. The lemma is proved. 

Denote by {Mλ | λ ∈ Λ} the family of all nonempty compact positively invariant

T
and globally asymptotically stable sets from X and J 00 := {Mλ |λ ∈ Λ}.

Theorem 1.8 Let (X, T, π) be compact dissipative and J be its center of Levin-
son, then J = J 00 , i.e. the center of Levinson J of the compact dissipative system
(X, T, π) is the least compact positively invariant globally asymptotically stable set
in X.

Proof. According to Lemma 1.8 the inclusion J 0 ⊆ J 00 takes place and according to
Theorem 1.7 J = J 0 ⊆ J 00 . The reverse inclusion also holds. For that it is sufficient
to note that for certain λ0 ∈ Λ the equality Mλ0 = J holds and consequently
J 00 = ∩{Mλ |λ ∈ Λ} ⊆ J. The theorem is proved. 

Theorem 1.9 Let (X, T, π) be compact dissipative dynamical system and K be a

nonempty compact invariant set from X, then the following statements are equiva-
lent:

1. K is the center of Levinson of (X, T, π);

2. K is globally asymptotically stable;
3. K is maximal compact invariant set in X.

Proof. According to Theorem 1.6 from 1. follows 2. Now we will show that the
inverse statement holds. Denote by J the center of Levinson of (X, T, π), then
according to Theorem 1.8 J ⊆ K. On the other hand, the center of Levinson
according to Theorem 1.6 is the attractor of all compact subsets from X and by
invariance of K we have K ⊆ J. So, J = K.
From Theorem 1.6 follows that 1. implies 3. To complete the proof of the
theorem it is sufficient to show that the inverse implication also holds. Let J be the
center of Levinson of (X, T, π). Then from one hand in virtue of Theorem 1.6 the
set J is compact and invariant and by the maximality of K we have J ⊆ K. On
the other hand, according to Theorem 1.6 the center of Levinson is the attractor
of the family C(X) and by the invariance of K we have K ⊆ J and consequently
K = J. The theorem is proved. 
16 Global Attractors of Non-autonomous Dissipative Dynamical Systems

1.4 Dissipative systems on the local compact spaces

As it was mentioned above from the local dissipativity of the system (X, T, π) follows
its compact dissipativity. On the other hand if the space X is locally compact then
is is easy to see that the converse also is true. So, in locally compact spaces compact
and local dissipativity are equivalent. In fact, as it will be shown bellow a stronger
statement holds, namely: in locally compact spaces point dissipativity implies local
dissipativity.

Definition 1.26 The dynamical system (X, T, π) we will call:

− locally completely continuous if for every point p ∈ X there exist δ = δ(p) > 0
and l = l(p) > 0 such that π l B(p, δ) is relatively compact;
− weakly dissipative if there exist a nonempty compact K ⊆ X such that for every
ε > 0 and x ∈ X there is τ = τ (ε, x) > 0 for which xτ ∈ B(K, ε). In this case
we will call K weak attractor.

Note that every dynamical system (X, T, π) defined on the locally compact met-
ric space X is locally completely continuous.

Lemma 1.9 Let K ⊆ X be a nonempty compact, pi ∈ X and δi > 0 (i = 1, m).

If K ⊆ ∪{B(pi , δi ) | i = 1, m}, then there exists γ > 0 such that

B(K, γ) ⊆ ∪{B(pi , δi ) | i = 1, m}. (1.19)

Proof. Assume that the inclusion (1.19) does not hold for any γ > 0. Then there
are γn ↓ 0 and rn ∈ B(K, γn ) such that rn 6∈ ∪{B(pi , δi ) | i = 1, m}. As γn ↓ 0 then
for every point rn there exists the point qn ∈ K such that

ρ(rn , qn ) < γn . (1.20)

By the compactness of K the sequence {qn } can be considered converging to some

point q ∈ K. Then according to (1.20) rn → q, which contradicts the choice of the
sequence {rn }. 

Lemma 1.10 Let (X, T, π) be weakly dissipative and locally completely continuous
and K ⊂ X be the weak attractor of (X, T, π), then:

1) there exist a0 > 0 and l0 > 0 such that π t B(K, a0 ) is relatively compact for
every t ≥ l0 ;
2) there exist L0 ≥ l0 such that for all t ≥ L0

π t B(K, a0 ) ⊆ π l0 B(K, a0 ).
 Autonomous dynamical systems 17

Proof. Let x ∈ K. By virtue of the local complete continuity of (X, T, π) for the
point x ∈ K there exist l(x) > 0 and δx > 0 such that π t B(x, δx ) is relatively
compact for all t ≥ l(x). It is clear that {B(x, δx ) | x ∈ K} is an open covering of
K and by its compactness we can extract finite sub-covering {B(xi , δxi ) | i ∈ 1, n}
from the constructed covering. Let l0 = max{l(xi ) | i ∈ 1, n}. According to Lemma
1.9 there exists a0 > 0 such that

K ⊆ B(K, a0 ) ⊆ ∪{B(xi , δxi ) | i ∈ 1, n}.

Consequently

π t B(K, a0 ) ⊆ ∪{π t B(xi , δxi ) | i ∈ 1, n},

that is why the set π t B(K, a0 ) is relatively compact for all t ≥ l0 .

Let us now prove the 2nd statement of the lemma. Let a0 and l0 be positive
numbers from the previous point. Supposing that the 2nd statement of the lemma
is not true, we will find that there exist {xk } ⊂ B(K, a0 ) and tk → +∞ such that

(1) xk ∈ B(K, a0 ).
(2) xk t ∈ X \ B(K, a0 ) for all 0 < t < tk .
(3) xk tk ∈ B(K, a0 ).

From 2. follows that

4. x0k t ∈ X \ B(K, a0 ) for all 0 < t < tk − l0 , where x0k = xk l0 .

By virtue of the relative compactness of π l0 B(K, a0 ) the sequence x0k = xk l0

can be considered convergent. Assume x0 = lim x0k , then from 4. follows that
k→+∞
x0 t ∈ X \ B(K, a0 ) for all 0 < t < +∞ and consequently ∅ 6= ωx0 ⊂ X \ B(K, a0 ).
So ωx0 ∩ K = ∅ that contradicts the weak dissipativity of (X, T, π) and to the fact
that K is a weak attractor. The contradiction obtained completes the proof of the
lemma. 

Theorem 1.10 For the locally completely continuous dynamical systems the weak,
point, compact and local dissipativity are equivalent.

Proof. It is clear that for the proof of the formulated theorem it is sufficient to show
that from the weak dissipativity of (X, T, π) follows its local dissipativity if (X, T, π)
is locally completely continuous. Let K 6= ∅ be compact and be the weak attractor
of (X, T, π). Denote by a0 the number from Lemma 1.10. If x ∈ X then there exists
τ > 0 such that xτ ∈ B(K, a0 ). Let γ > 0 be such that B(xτ, γ) ⊂ B(K, a0 ). By
the continuity of the mapping π τ : X → X in the point x there exists α > 0 such
18 Global Attractors of Non-autonomous Dissipative Dynamical Systems

that

π τ B(x, α) ⊂ B(xτ, γ) ⊂ B(K, a0 ).

Assume M = B(K, a0 ). According to Lemmas 1.3 and 1.10 the set M 6= ∅ is

compact and

lim β(π t B(x, α), M ) = 0.

k→+∞

So, we constructed the nonempty compact M ⊂ X attracting every point with some
α-neighborhood that is (X, T, π) is locally dissipative. The theorem is proved. 

1.5 Criterions of compact dissipativity

Assume Ω := ∪{ωx |x ∈ X}. Let (X, T, π) be compact dissipative dynamical system

and J be its Levinson center. It is clear that Ω ⊆ J. At once plain examples
show that in general case Ω 6= J. At the same time the set Ω is an important
characteristic of dissipative dynamical system. For example, from Theorem 1.10
follows that in the locally compact space X the dynamical system (X, T, π) is point
dissipative if and only if Ω is not empty, compact and ωx 6= ∅ for all x ∈ X. In view
of that has been said above, we take on interest in such questions:

a. Which are the conditions necessary for the coincidence of the sets Ω and J?
b. What is the connection between Ω and J in general case?

From the results given bellow follow the answers to the formulated questions.
Denote by
\[
D+ (M ) := {π t B(M, ε)|t ≥ 0},
ε>0

\ \[
J + (M ) := {π τ B(M, ε)|τ ≥ t},
ε>0 t≥0

Dx+ := D+ ({x}) and Jx+ := J + ({x}).

Directly from the definition D + (M ) and J + (M ) follows the following lemma.

Lemma 1.11 The following statements hold:

(1) p ∈ D+ (M ) (p ∈ J + (M )) if and only if there exist {xn } and {tn }(tn → +∞)
such that ρ(xn , M ) → 0 and xn tn → p;
(2) the set D+ (M ) (J + (M )) is closed and positive invariant.

Lemma 1.12 Let y ∈ ωx , then Jx+ ⊆ Jy+ .

 Autonomous dynamical systems 19

Proof. Let y ∈ ωx and p ∈ Jx+ , then there exist sequences {τn0 } → +∞, xτn0 → y,
{t0n } and {xn } such that xn → x, t0n → +∞ and xn t0n → p. According Lemma 1.11
we can consider that t0n −τn0 > n for all n ∈ N. For every k ∈ N consider the sequence
{xn τk0 }. By the axiom of continuity xn τk0 → xτk0 for n → +∞ (for every k ∈ N) and
consequently for every k ∈ N there is nk ≥ k such that ρ(xn τk0 , xτk0 ) ≤ k −1 for all n ≥
nk . As xτn0 → y, we have ρ(y, xnk τk0 ) ≤ ρ(y, xτk0 ) + ρ(xτk0 , xnk τk0 ) ≤ ρ(y, xτk0 ) + k −1 .
Note that xnk t0nk = xnk τn0 k (t0nk − τn0 k ), xnk τn0 k → y and t0nk − τn0 k > nk ≥ k. From
this follows that p ∈ Jy+ , i.e. Jx+ ⊆ Jy+ . The lemma is proved. 

Corollary 1.3 If y ∈ ωx then Jy+ = Dy+ .

Proof. In fact, as Jy+ ⊆ Dy+ , it is sufficient to show that Dy+ ⊆ Jy+ . Note that
Dy+ = Σ+ + + + + + +
y ∪ Jy . Since y ∈ ωx , Σy ⊆ ωx ⊆ Jx ⊆ Jy . Hence, Dy = Σy ∪ Jy ⊆
+

Jy+ ∪ Jy+ = Jy+ . 

Lemma 1.13 If xn → x, yn → y for n → +∞ and xn ∈ Dy+n (xn ∈ Jy+n ), then

x ∈ Dy+ (x ∈ Jy+ ).

Proof. Let ε > 0 and δ > 0. From the inclusion xn ∈ Dy+n (xn ∈ Jy+n ) for every n
follows the existence of the point zn ∈ B(yn , 2δ ) and of the number tn ≥ 0 (tn ≥ n)
such that
ε
ρ(xn , zn tn ) < . (1.21)
2
From xn → x and yn → y follows the existence of such integer n0 that for all n > n0
simultaneously the inequalities
δ ε
ρ(y, yn ) < and ρ(xn , x) < (1.22)
2 2
hold. From zn ∈ B(yn , δ2 ) and (1.22) we obtain ρ(y, zn ) < δ, and from (1.21) and
(1.22) follows the inequality ρ(x, zn tn ) < ε, i.e. x ∈ Dy+ (x ∈ Jy+ ). The lemma is
proved. 

Lemma 1.14 If the set M ⊆ X is compact, then D + (M ) = ∪{Dx+ |x ∈ M } and

J + (M ) = ∪{Jx+ |x ∈ M }.

Proof. Since ∪{Dx+ |x ∈ M } ⊆ D+ (M ), then to prove the lemma it is sufficient

to show that D+ (M ) ⊆ ∪{Dx+ |x ∈ M }. Let y ∈ D+ (M ), then there exist {xn }
and tn ≥ 0 such that ρ(xn , M ) → 0 and y = lim xn tn . As M is compact
n→+∞
the sequence {xn } can be considered convergent. Assume x := lim xn , then
n→+∞
x ∈ M . So, xn → x, y ∈ Dx+
and yn → y (yn := xn tn ). According to Lemma 1.13
y ∈ Dx+ ⊆ ∪{Dx+ |x ∈ M }. The second statement is established in the same way.
The lemma is proved. 
20 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Lemma 1.15 If M is nonempty compact negative invariant set, then M ⊆ J + (Ω).

Proof. Let M be nonempty compact negatively invariant set and x ∈ M , then

according to Theorem 1.2 there exists an entire trajectory ϕ : S −→ X such that
ϕ(0) = x and ϕ(t) ∈ M for all t ≥ 0. As αϕx 6= ∅, is closed and αϕx ⊆ M , then it is
compact. Let y ∈ αϕx , then ωy ⊆ αϕx . If p ∈ ωy ⊆ Ω, then there exists tn → +∞
such that xk = ϕ(−tk ) → p, x = π tk ϕ(−tk ) and consequently x ∈ Jp+ ⊆ J + (Ω).
The lemma is proved. 

Corollary 1.4 If (X, T, π) is point dissipative, Ω 6= ∅ and it is compact, then

Ω ⊆ J + (Ω).

Proof. In fact, since ωx is invariant for all x ∈ X and Ω = ∪{ωx | x ∈ X}, then Ω
is also invariant and according to Lemma 1.15 Ω ⊆ J + (Ω). 

Lemma 1.16 If the dynamical system (X, T, π) is point dissipative and D + (Ω)
(J + (Ω)) is compact, then D + (Ω) = D+ (D+ (Ω)) (J + (Ω) = J + (J + (Ω))).

Proof. As Ω ⊆ D + (Ω), then D+ (D+ (Ω)) ⊆ D+ (Ω) and, consequently, to prove

Lemma 1.16 it is sufficient to establish the inverse inclusion D + (D+ (Ω)) ⊆ D+ (Ω).
Let x ∈ D+ (Ω), then ωx ⊆ Ω, and if y ∈ ωx ⊆ Ω then according to Lemma 1.12 and
Corollary 1.3 Jx+ ⊆ Jy+ = Dy+ ⊆ D+ (Ω). Since Dx+ = Σ+ + + +
x ∪ Jx ⊂ D (Ω) ∪ D (Ω) =
+ + +
D (Ω) for all x ∈ D (Ω) and D (Ω) is compact, then according to Lemma 1.14
D+ (D+ (Ω)) = ∪{Dx+ |x ∈ D+ (Ω)} ⊆ D+ (Ω). By analogy with this the proof of
the second equation is executed but we should keep in mind that Ω ⊆ J + (Ω). The
lemma is proved. 

Lemma 1.17 If (X, T, π) is compact dissipative, then D + (D+ (Ω)) = D+ (Ω)

(J + (J + (Ω)) = J + (Ω)).

Proof. Let (X, T, π) be compact dissipative and let J be its Levinson’s center.
Then Ω ⊆ J. By virtue of orbital stability of J we have D + (J) = J (J + (J) ⊆ J)
and, consequently, D + (Ω) ⊆ J (J + (Ω) ⊆ J). As D+ (Ω) (J + (Ω)) is closed and J is
compact, D+ (Ω) (J + (Ω)) is also compact and to finish the proof of the lemma it is
sufficient to refer to Lemma 1.16. 

It is known [25] that for compact positive invariant set M the equality D + (M ) =
M holds if and only if M is orbital stable (X is locally compact and T = R).
Below we will show that this statement is valid also for compact dissipative systems
on arbitrary metric space both with continuous and discrete time. Namely, the
following lemma holds.
 Autonomous dynamical systems 21

Lemma 1.18 Let (X, T, π) be compact dissipative. For the compact positive in-
variant set M ⊆ X to be orbital stable, it is necessary and sufficient that the equality
D+ (M ) = M holds.

Proof. If M is orbitally stable, then by standard reasoning (see, for example, [25])
is shown that D+ (M ) = M for every systems (including dissipative ones).
Let now D+ (M ) = M and show that M is orbital stable. If we suppose the
contrary then there exist ε0 > 0, xn → x ∈ M , and tn ≥ 0 such that

ρ(xn tn , M ) ≥ ε0 . (1.23)

Since the system (X, T, π) is compact dissipative, the set Σ+ (K) is relatively com-
pact where K := {xn } and, consequently, the sequence {xn tn } can be considered
convergent. Assume y := lim xn tn . Then on the one hand y ∈ D + (M ) = M .
n→+∞
On the other hand, from (1.23) follows that ρ(y, M ) ≥ ε0 > 0. The obtained
contradiction proves the lemma. 

Theorem 1.11 If the dynamical system (X, T, π) is compact dissipative, then

J = J + (Ω).

Proof. As Ω ⊆ J and J is asymptotically stable, then J + (Ω) ⊆ J + (Ω).

From Lemmas 1.17 and 1.18 follows that the set J + (Ω) is orbital stable because
D+ (J + (Ω)) = J + (Ω). Let x ∈ J \ J + (Ω) and dx := ρ(x, J + (Ω)) ≥ 0. If dx = 0 for
all x ∈ J \J + (Ω), then the theorem is proved. Suppose that for some x0 ∈ J \J + (Ω)
we have dx0 > 0. For the number 0 < ε < 2−1 dx0 , choose δ(ε) > 0 out of the condi-
tion of orbital stability of J + (Ω). Since x0 ∈ J, according to Theorem 1.2 there ex-
ists a continuous mapping ϕ : S −→ J such that π t ϕ(s) = ϕ(t+s) for all t ∈ T, s ∈ S
and ϕ(0) = x0 . Since J is compact, the set α of limit points αϕx0 of the motion ϕ is
not empty, compact and αϕx0 ∩Ω 6= ∅ and, consequently, there exists tn → −∞ such
that ρ(x0 tn , Ω) → 0. Choose n0 such that ρ(x0 tn , Ω) < δ (n ≥ n0 ). Then we have
ρ(x0 tn , A, AA?, J + (Ω)) < δ and, consequently, ρ(x0 (t+tn ), J + (Ω)) < ε for all t0 ≥ 0
and n ≥ n0 . In particular, when t = −tn , we have dx0 = ρ(x0 , J + (Ω)) < ε < 2−1 dx0 .
The obtained contradiction completes the proof of the theorem. 

Corollary 1.5 Under the conditions of Theorem 1.11 the equality J = D + (Ω)
holds.

Proof. The formulated statement follows from Theorem 1.11 taking into account
that J ⊆ J + (Ω) ⊆ D+ (Ω) ⊆ J. 

Corollary 1.6 If (X, T, π) is compact dissipative, then J = Ω if and only if Ω is

orbital stable.
22 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Proof. This statement follows from Theorem 1.11 and Lemma 1.18. 

Theorem 1.12 For the dynamical system (X, T, π) to be compact dissipative, it is

necessary and sufficient that there exists a nonempty compact set K ⊆ X satisfying
the condition: for every ε > 0 and x ∈ X, there exist δ(ε, x) > 0 and l(ε, x) > 0
such that

π t B(x, δ(ε, x)) ⊆ B(K, ε) (1.24)

for all t ≥ l(ε, x).

Proof. Let (X, T, π) be compact dissipative, J be its center of Levinson, ε > 0 and
x ∈ X. According to Theorem 1.6 J is orbital stable. Denote by γ(ε) > 0 a number
defined by ε out of the condition of orbital stability of J. The set J is globally
asymptotically stable, hence for x ∈ X and γ(ε) > 0 there exists l(ε, x) such that
xt ∈ B(J, γ) for all t ≥ l(ε, x). Since B(J, γ) is open, there exists α = α(ε, x) > 0
for which B(π(x, l(ε, x)), α) ⊆ B(J, γ). By virtue of continuity of the mapping
π(l(ε, x), ·) : X −→ X for all x ∈ X and α > 0, there exists δ = δ(ε, x) > 0 such
that

π(l(ε, x), B(x, δ)) ⊆ B(π(l(ε, x), x), α). (1.25)

From the inclusion (1.25) and by the choice of γ we have π t B(x, δ) ⊆ B(J, ε) for all
t ≥ l(ε, x).
Let now K ⊆ X be a nonempty compact satisfying the condition (1.24). If ε > 0
and M is a nonempty compact from X, then for every x ∈ M there exist δ(ε, x) > 0
and l(ε, x) > 0 such that (1.24) holds. Consider open covering {B(x, δ(ε, x)) | x ∈
M } of the set M . By virtue of compactness of M and of the completeness of the
space X we can extract from this covering a finite sub-covering {B(xi , δ(ε, xi )) | i ∈
1, n}. Assume L(ε, M ) := max{l(ε, xi ) | i = 1, n}. From (1.24) follows that π t M ⊆
B(K, ε) for all t ≥ L(ε, M ). The theorem is proved. 

Theorem 1.13 Let (X, T, π) be point dissipative. For (X, T, π) to be compact

dissippative it is necessary and sufficient that there exists a nonempty compact set
M possessing the following properties:

(1) Ω ⊆ M ;
(2) M is orbital stable.

In this case J ⊆ M where J is the center of Levinson of (X, T, π).

Proof. Necessity of the conditions of the theorem is obvious. To prove the formu-
lated statement it is enough to show the sufficiency of the conditions (1) and (2).
Reasoning in the same way as in the first part of Theorem 1.12 we establish that
 Autonomous dynamical systems 23

for every ε > 0 and x ∈ X there exist δ(ε, x) > 0 and l(ε, x) > 0 satisfying the
condition (1.24). In this case as K we should take the set M from Theorem 1.13.
Applying Theorem 1.12 we obtain compact dissipativity of (X, T, π).
Now let us show that J ⊆ M . Since Ω ⊆ M and M is orbital stable, D + (Ω) ⊆
D+ (M ) = M . To finish the proof of the theorem it is sufficient to refer to Corollary
1.5. 

Theorem 1.14 Let (X, T, π) be point dissipative. For (X, T, π) to be compact

dissipative it is necessary and sufficient that the set D + (Ω) (J + Ω)) be compact and
orbital stable. In this case J = D + (Ω) (J = J + (Ω)) where J is the center of
Levinson of the dynamical system (X, T, π).

Proof. The formulated statement follows from Theorems 1.11, 1.12, 1.13 and
Corollary 1.5. 

Theorem 1.15 Let (X, T, π) be point dissipative. For (X, T, π) to be compact

dissipative it is necessary and sufficient that Σ+ (K) be relatively compact for any
compact K ⊆ X.

Proof. The necessity of the conditions of the theorem follows from Theorem 1.5.
Let us prove the sufficiency. First of all, note that under the conditions of the
theorem the set J + (Ω) 6= ∅ and is compact. In fact, according to Corollary 1.4
Ω ⊆ J + (Ω) and, consequently, J + (Ω) 6= ∅. Show now that J + (Ω) is compact. Let
{yk } ⊆ J + (Ω) and εk ↓ 0, then there exist pk ∈ Ω, yk ∈ Jp+k , pk ∈ B(pk , εk ) and
tk > 0 such that

ρ(yk , pk tk ) < εk . (1.26)

Since {pk } ⊆ Ω, Ω is compact and εk ↓ 0, the sequence {pk } is relatively compact

and, consequently, under the conditions of the theorem the sequence {pk tk } is also
compact. Then from the inequality (1.26) follows that the sequence {yk } is relatively
compact, i.e. J + (Ω) is compact.
Let us show that J + (Ω) attracts all compacts from X. Let K be an empty com-
pact from X. Then under the conditions of the theorem the set Σ+ (K) is relatively
compact and according to Theorem 1.5 the set Ω(K) is not empty and the equality

lim β(π t K, Ω(K)) = 0 (1.27)

t→+∞

holds. According to Lemma 1.3 the set Ω(K) is invariant and by Lemma 1.15
Ω(K) ⊆ J + (Ω) and, consequently, from (1.27) follows that J + (Ω) attracts K. The
theorem is proved. 
At the end of this chapter, we give an example of point dissipative dynamical
system that is not compact dissipative.
24 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Example 1.6 Let ϕ : R −→ R be a continuous function equal to zero on R− and

defined on R+ by the equality
( n −1 o
exp (t − 1)2 − 1 +1 0≤t<2
ϕ(t) = (1.28)
0, 2 ≤ t < +∞.

Assume X = {ϕ(at + b)| a, b; t ∈ R}. Note that X ⊂ C(R, R) is closed and invariant
w.r.t. the shifts subset of C(R, R) and, consequently, the dynamical system of
Bebutov induced on X a group dynamical system of shifts (X, R, σ). We remark on
some properties of the constructed dynamical system:

(1) For any function ψ ∈ X, the set {σ(t, ψ) : t ∈ R} is relatively compact and
there exists c = c(ψ) ∈ [0, 1] such that ωψ = αψ = {ϕc } where ϕc (t) = c for all
t ∈ R;
(2) (X, R, σ) is point dissipative and Ω = {ϕc | 0 ≤ c ≤ 1};
(3) D+ (Ω) = X.

Note that X is not compact. In fact, if that were not so, then according to
Theorem of Arzela-Ascoli the functions from X would be equicontinuous on the
segment [0, 2]. Let us construct a subset from X that does not possess this property.
Let ϕn (t) = ϕ(nt). Assume t1n = n−1 and t2n = 2n−1 . It is clear that t1n , t2n ∈ [0, 2],
t1n − t2n → 0 for n → +∞ and |ϕn (t1n ) − ϕn (t2n )| = |ϕ(1) − ϕ(2)| = 1, i.e. the
sequence {ϕn } is not equicontinuous on the segment [0, 2]. So, D + (Ω) = X is
not compact and according to Theorem 1.14 the dynamical system (X, R, σ) is not
compact dissipative.

In relation to Example 1.6, note that according to Theorem 1.14 examples of

point dissipative but not compact dissipative dynamical systems can be either of
two types: D+ (Ω) is not compact (as in Example 1.6), or D + (Ω) is compact but is
not orbital stable. Let us show the latter case.

Example 1.7 Let ϕ be the function from the previous example. Assume X =
{ϕ(at + b)| a, b ∈ R and t ∈ R}. By (X, R, σ) denote a semigroup dynamical
system of shifts on X. We note the following properties of the constructed dynamical
system:

(1) For any function ψ ∈ X, the set {σ(t, ψ) | t ∈ R+ } is relatively compact and
there exists c ∈ [0, 1] such that ωψ = {ϕc };
(2) D+ (Ω) = Ω and, consequently, D + (Ω) is compact, as Ω = {ϕc | 0 ≤ c ≤ 1}.

Show that D+ (Ω) is not orbital stable. It is clear that to prove this statement
it is sufficient to construct the sequence {ψn } ⊆ X and tn ≥ 0 such that ψn →
ψ0 ∈ Ω and inf{ρ(σ(tn , ψn ), Ω)| n ≥ 0} > 0. Assume ψn (t) = ϕ(nt + 1 − n2 )
(t ∈ R) and tn = n. Then σ(tn , ψn )(s) = ϕ(ns + 1) is not convergent in X, hence
 Autonomous dynamical systems 25

inf{ρ(σ(tn , ψn ), Ω) | n ≥ 0} > 0 (in fact, the sequence, ϕ(ns + 1) does not contain
sub-sequences convergent in X). So, for semigroup point dissipative dynamical
system (X, R+ , σ) the set D+ (Ω) is compact but not orbital stable. According to
Theorem 1.14 (X, R+ , σ) is not compact dissipative.

Remark 1.5 As it was proved in Lemma 1.18, for the compact positively invariant
set M ⊆ X the equality D + (M ) = M assures the orbital stability of M if the
dynamical system (X, T, π) is compact dissipative. Example 1.7 shows that for point
dissipative systems the corresponding fact does not hold, i.e., in this case the well
known Theorem of Ura [25] does not hold.

1.6 Local dissipative systems

Theorem 1.16 For the compact dissipative dynamical system (X, T, π) to be lo-
cally dissipative is necessary and sufficient that for every point p ∈ X there exists
δp > 0 such that

lim β(π t B(p, δp ), J) = 0, (1.29)

t→+∞

where J is the center of Levinson of (X, T, π).

Proof. The sufficiency of the theorem is obvious. Let (X, T, π) be locally dissipative
dynamical system. Then there exists a nonempty compact K ⊆ X such that for
every point p ∈ X there exists δp > 0 for which the following equality holds:

lim β(π t B(p, δp ), K) = 0. (1.30)

t→+∞

From Lemma 1.3 follows that the set Kp := ω(B(p, δp )) is nonempty, compact,
invariant and

lim β(π t B(p, δp ), Kp ) = 0. (1.31)

t→+∞

From the equality (1.30) follows the inclusion Kp ⊆ K and consequently ω(Kp ) ⊆
ω(K) ⊆ J. By virtue of the invariance of the set Kp we have Kp = ω(Kp ); hence
Kp ⊆ J. From the last inclusion and the equality (1.31) follows the equality (1.29).
The theorem is proved. 

Theorem 1.17 For the compact dissipative dynamical system (X, T, π) to be lo-
cal dissipative, it is necessary and sufficient that its Levinson center J would be
uniformly attracting set.

Proof. Let (X, T, π) be local dissipative, J be its center of Levinson and p ∈ J.

Then according to Theorem 1.16 there exists δp > 0 such that the equality (1.29)
26 Global Attractors of Non-autonomous Dissipative Dynamical Systems

holds. By compactness of J from its open covering {B(p, δp )| p ∈ J}, it is possible

to extract finite sub-covering {B(pi , δpi )| i ∈ 1, m}. According to Lemma 1.9 there
exists γ > 0 such that B(J, γ) ⊂ U {B(pi , δpi )| i ∈ 1, m}. It is clear that for the
γ > 0 the equality

lim β(π t B(J, γ), J) = 0 (1.32)

t→+∞

holds, i.e., J is uniformly attracting set.

Let now (X, T, π) be compact dissipative and its Levinson center be uniformly
attracting set, i.e., there exists γ > 0 such that the equality (1.32) holds. For x ∈ X
there exists l = l(x) > 0 for which

ρ(xt, J) < γ (1.33)

for all t ≥ l. According to (1.32) for every positive number ε we can choose L(ε) > 0
such that

ρ(yt, J) < ε (1.34)

for all t ≥ L(ε) and y ∈ B(J, γ). Since B(J, γ) is open by virtue of (1.33) we
can select η = η(x) > 0 such that the inclusion B(xl, η) ⊂ B(J, γ) holds. By
the continuity of the mapping π(l, ·) : X → X for η we can find δ = δx > 0
such that yl ∈ B(xl, η) and yl ∈ B(J, γ) for all y ∈ B(x, δx ). By virtue of (1.34)
y(t + l) ∈ B(J, ε) for all t ≥ L(ε) and y ∈ B(x, δx ). Let L(ε, x) := l(x) + L(ε), then
yt ∈ B(J, ε) for all t ≥ L(ε, x) and y ∈ B(x, δx ), i.e. (X, T, π) is local dissipative.
The theorem is completely proved. 

Lemma 1.19 Let M ⊆ X be a nonempty compact positive invariant set in

(X, T, π). If M is uniformly attracting it is orbital stable.

Proof. Suppose that the conditions of the lemma are fulfilled but M is not orbital
stable. then there exist ε0 > 0, δn ↓ 0, xn ∈ B(M, δn ) and tn → +∞ such that

ρ(xn tn , M ) ≥ ε0 . (1.35)

Since M is uniformly attracting, for the number ε0 there exists positive number
L(ε0 ) such that
ε0
ρ(xt, M ) < (1.36)
2
for all x ∈ B(M, γ) and t ≥ L(ε0 ) where γ > 0 is such that

lim β(π t B(M, γ), M ) = 0.

t→+∞
 Autonomous dynamical systems 27

As xn ∈ B(M, δn ) and δn ↓ 0, the sequence {xn } can be considered convergent.

Assume x0 = lim xn . Then x0 ∈ M and tn ≥ L(ε0 ) for sufficiently large n.
n→+∞
According to (1.36) holds the equality
ε0
ρ(xn tn , M ) < . (1.37)
2
The inequalities (1.35) and (1.37) are contradictory. The obtained contradiction
completes the proof of the lemma. 

Theorem 1.18 For a point dissipative dynamical system (X, T, π) to be local

dissipative, it is necessary and sufficient that the following two conditions hold:

1. D+ (Ω) (J + (Ω)) is compact;

2. D+ (Ω) (J + (Ω)) is uniformly attracting set.

Proof. Let (X, T, π) be point dissipative and the conditions 1. and 2. of the
theorem are fulfilled. Since according to Lemma 1.19 the set D + (Ω) (J + (Ω)) is
orbital stable, then according to Theorem 1.14 the dynamical system (X, T, π) is
compact dissipative and D + (Ω) (J + (Ω)) coincides with its center of Levinson J.
Now to finish the proof of the theorem it is sufficient to refer to Theorem 1.17.
Let (X, T, π) be local dissipative. As according to Lemma 1.5 the dynamical
system (X, T, π) is compact dissipative. Then according to Theorem 1.14 D + (Ω)
(J + (Ω)) is compact and orbital stable. By Theorem 1.11 and Corollary 1.5 the
equality J = D+ (Ω) (J = J + (Ω)) holds, where J is the center of Levinson of
(X, T, π). From Theorem 1.17 follows that D + (Ω) (J + (ω)) is a uniformly attracting
set. The theorem is completely proved. 

Definition 1.27 We will call the dynamical system (X, T, π) local asymptotically
condensing if for every point p ∈ X there exist δp > 0 and a nonempty compact
Kp ⊆ X such that

lim β(π t B(p, δp ), Kp ) = 0. (1.38)

t→+∞

The following holds:

Theorem 1.19 Let (X, T, π) be point dissipative. For (X, T, π) to be local dissipa-
tive it is necessary and sufficient that (X, T, π) be locally asymptotically condensing.

Proof. If (X, T, π) is locally dissipative, then it is clear that (X, T, π) is asymp-

totically condensing too. Suppose that (X, T, π) is point dissipative and locally
asymptotically condensing and let us show that (X, T, π) is local dissipative. First
of all, let us show that for every compact K ∈ C(X) the set Σ+ (K) is relatively
compact. Let p ∈ K, δp > 0 and Kp ∈ C(X) be such that the equality (1.38) is
28 Global Attractors of Non-autonomous Dissipative Dynamical Systems

fulfilled. By compactness of K from its open covering {B(p, δp )| p ∈ K} we can ex-

tract finite subcovering {B(pi , δpi )| 1 ≤ i ≤ n}. Assume W = Kp1 ∪ Kp2 ∪ . . . ∪ Kpn .
Then W is compact and

lim β(π t K, W ) = 0. (1.39)

t→+∞

From the equality (1.39) follows relative compactness of Σ+ (K). According to

Theorem 1.15 the dynamical system (X, T, π) is compact dissipative. Let now J be
the center of Levinson of (X, T, π), and that p ∈ J, δp > 0, and Kp ∈ C(X) are such
that the equality (1.38) holds. By Lemma 1.3 the set ω(B(p, δp )) 6= ∅, is compact,
invariant and the equality

lim β(π t B(p, δp ), ω(B(p, δp ))) = 0 (1.40)

t→+∞

holds. Since J is maximal compact invariant set of (X, T, π), then ω(B(p, δp )) ⊆ J
and from the equality (1.40) follows the equality (1.29). According to Theorem 1.16
the dynamical system (X, T, π) is local dissipative. The theorem is proved. 

At the end of this paragraph, we give an example of compact dissipative

dynamical system that is not local dissipative.

Example 1.8 Consider a linear differential equation

ẋ = Ax (1.41)

in the Hilbert space H := L2 [0, 1] with the continuous operator A : L2 [0, 1] →

L2 [0, 1] defined by the equality (Aϕ)(τ ) = −τ ϕ(τ ) for all τ ∈ [0, 1] and ϕ ∈ L2 [0, 1].
Note that the spectrum of the operator A coincides with the segment [−1, 0]. Denote
by U (t) the Cauchy operator of the equation (1.41). It is clear that (U (t)ϕ)(τ ) =
e−τ t ϕ(τ ) for all t ∈ R, τ ∈ [0, 1] and ϕ ∈ L2 [0, 1] (see [250, p.404]). Denote by
(H, R, π) the dynamical system generated by the equation (1.41) i.e. π(ϕ, t) :=
U (t)ϕ for every t ∈ R and ϕ ∈ L2 [0, 1]. According to Theorem of Lebesgue on the
limit passage under the sign of integral we have
Z 1
2
||π(t, ϕ)|| = e−2τ t |ϕ(τ )|2 dτ → 0
0

as t → +∞. Hence, the dynamical system (H, R, π) is point dissipative and ωϕ =

{0} for every ϕ ∈ H and consequently Ω = ∪{ωϕ |ϕ ∈ H} = {0}. Further note that
Z 1 Z 1
2 −2τ t 2
||π(t, ϕ)|| = e |ϕ(τ )| dτ ≤ |ϕ(τ )|2 dτ = ||ϕ||2
0 0

for all t ≥ 0. According to Theorem 1.14 the dynamical system (H, R, π) is compact
dissipative and its Levinson center J = {0}.
 Autonomous dynamical systems 29

Let us show that the constructed dynamical system (H, R, π) is not local dissi-
pative. In fact, if (H, R, π) were local dissipative then according to Theorem 1.17
there would be γ > 0 such that

lim sup ||π(t, ϕ)|| = 0. (1.42)

t→+∞ ||ϕ||≤γ

By virtue of the linearity of the system (H, R, π) the equality (1.42) is equivalent
to the condition

lim sup ||π(t, ϕ)|| = 0. (1.43)

t→+∞ ||ϕ||≤1

√
Define the function ϕn ∈ H (n = 1, 2, . . .) by the following rule: ϕn (τ ) := nχn (τ )
for all τ ∈ [0, 1] where χn is the characteristic function of the set [0, n−1 ] ⊆ [0, 1].
Note that ||ϕn || = 1 and
Z 1
n
2
||π(tn , ϕn )|| = ne−2τ t dτ
0
n h − 2tnn
i
= 1−e = 1 − e−1 6→ 0, (1.44)
2tn

where tn = n2 . However, (1.43) and (1.44) cannot take place simultaneously. The
obtained contradiction shows that our assumption about the local dissipativity of
(H, R, π) is not true. The necessary example is constructed.

1.7 Global attractors

Definition 1.28 A nonempty compact set I ⊂ X we will call the global attractor
of the dynamical system (X, T, π) if the following conditions are fulfilled:

a. I is invariant w.r.t. (X, T, π);

b. I attracts all the bounded subsets from X.

Theorem 1.20 The following conditions are equivalent:

1. the dynamical system (X, T, π) admits a compact global attractor;

2. the dynamical system (X, T, π) is bounded k-dissipative;
3. the dynamical system (X, T, π) is compact k-dissipative and its center of Levin-
son attracts all bounded subsets of X.

Proof. The implication 1. ⇒ 2. is obvious. Let us show that from 2. follows 3. Let
(X, T, π) be bounded k-dissipative, then (X, T, π) is compact k-dissipative. Denote
by J the center of Levinson of (X, T, π) and show that J attracts all bounded subsets
30 Global Attractors of Non-autonomous Dissipative Dynamical Systems

from X. Let B ∈ B(X) and let K be a nonempty compact from X attracting all
bounded subsets from X, then

lim β(π t B, K) = 0. (1.45)

t→+∞

From the equality (1.45) and Lemma 1.3 follows that ω(B) 6= ∅, is compact, invari-
ant and

lim β(π t B, ω(B)) = 0. (1.46)

t→+∞

According to Theorem 1.6 J is maximal compact invariant set of (X, T, π) and

consequently ω(B) ⊆ J. From the last inclusion and the equality (1.46) follows
that

lim β(π t B, J) = 0. (1.47)

t→+∞

At last, it is clear that from 3. follows 1. The theorem is proved. 

Definition 1.29 Following to [206] we will say that the dynamical system
(X, T, π) satisfies the condition of Ladyzhenskaya if for every bounded set B ⊆ X
there exists a nonempty compact K ⊆ X such that the equality (1.45) holds.

Theorem 1.21 The following conditions are equivalent:

1. (X, T, π) is bounded k-dissipative;
2. (X, T, π) is point b-dissipative and satisfies the condition of Ladyzhenskaya.

Proof. According to Theorem 1.20 from 1. follows 2. Let us now prove that from 2.
follows 1. First of all note that from the condition 2. follows compact k-dissipativity
of (X, T, π). Let K ∈ C(X), then from the condition of Ladyzhenskaya, Lemma 1.3
and Theorem 1.5 follows relative compactness of Σ+ (K). According to Theorem
1.15 the dynamical system (X, T, π) is compact k-dissipative. Denote by J the
center of Levinson of (X, T, π) and let B ∈ B(X). Then by virtue of the condition
of Ladyzhenskaya and Lemma 1.3 ω(B) 6= ∅, is compact, invariant and the equality
(1.46) holds. As ω(B) ⊆ J, from the equality (1.46) follows that J attracts B. The
theorem is proved. 

Definition 1.30 We will call the dynamical system (X, T, π):

− completely continuous if for every B ∈ B(X) there exists l = l(B) > 0 such
that π l B is relatively compact;
− weakly b-dissipative if there exists a nonempty bounded set B0 ⊆ X such that
Σ+x ∩B0 6= ∅ for every point x from X (i.e. for every x ∈ X there is τ = τ (x) ≥ 0
such that xτ ∈ B0 ). In this case the set B0 we will call a weak b-attractor of
the dynamical system (X, T, π).
 Autonomous dynamical systems 31

Theorem 1.22 Let (X, T, π) be weakly b-dissipative and completely continuous.

Then (X, T, π) admits a compact global attractor.

Proof. Let B0 ∈ B(X) be a weak attractor of the system (X, T, π), l0 = l(B0 ) > 0
be such that π l0 B0 is relatively compact and let K := π l0 B0 . Under the conditions
of Theorem 1.22 for every point x ∈ X there exists τ = τ (x) ≥ 0 such that xτ ∈ B 0
and consequently x(τ + l0 ) ∈ K. So, the dynamical system (X, T, π) is weakly
k-dissipative and the compact K is its weak k-attractor. According to Theorem
1.10 the dynamical system (X, T, π) is compact k-dissipative. Let J be the center
of Levinson of (X, T, π), B ∈ B(X) and l = l(B) > 0 be such that π l B is relatively
compact. According to Theorem 1.6 the center of Levinson J attracts the compact
π l B and consequently holds the equality (1.47). The theorem is proved. 

Definition 1.31 The dynamical system (X, T, π) is called (see [179, 234]) asymp-
totically compact if for every bounded positive invariant set B ⊆ X (i.e. π t B ⊆ B
for all t ≥ 0) there exists a nonempty compact K ⊆ X such that the equality (1.45)
holds.

Theorem 1.23 Let (X, T, π) be compact k-dissipative and asymptotically com-

pact. Then (X, T, π) is local k-dissipative.

Proof. Let J be the center of Levinson of (X, T, π). According to Theorem 1.17
for (X, T, π) to be local k-dissipative it is necessary and sufficient to show that
the set J is uniformly attracting. Let ε0 = 1 and δ0 = δ(1) > 0 be such numbers
that from x ∈ B(J, δ0 ) follows xt ∈ B(J, 1) for all t ≥ 0 (according to Theorem 1.6
such δ0 exists). It is clear that the set B := ∪{π t B(J, δ0 )| t ≥ 0} is bounded and
positively invariant. By virtue of the asymptotic compactness of (X, T, π) there
exists a nonempty compact K ⊆ X such that the equality (1.45) holds . According
to Lemma 1.3 the set ω(B) 6= ∅, is compact, invariant and the equality (1.46) holds
. Since J is the maximal compact invariant set of (X, T, π), then ω(B) ⊆ J and,
in particular, from (1.46) follows (1.47). From the equality (1.47) and from the
inclusion B(J, δ0 ) ⊆ B follows the equality below

lim β(π t B(J, δ0 ), J) = 0.

t→+∞

The theorem is proved. 

Lemma 1.20 Let (X, T, π) be b-dissipative w.r.t. the family M and asymptotically
compact. Then (X, T, π) is also k-dissipative w.r.t. the family M.

Proof. There exists a bounded closed set B0 ⊆ X such that for every M ∈ M
there is L > 0 for which π t M ⊆ B0 for all t ≥ L. Assume B := {x ∈ B0 | xt ∈ B0
for all t ≥ 0}. It is clear that B 6= ∅, is bounded and positively invariant. By the
32 Global Attractors of Non-autonomous Dissipative Dynamical Systems

asymptotic compactness of (X, T, π) there exists a compact set K ⊆ B0 such that

the equality (1.45) holds. According to Lemma 1.3 the set ω(B) 6= ∅, is compact
and the equality (1.46) holds. Note that ML := ∪{π t M | t ≥ L} ⊆ B and ML
is positively invariant. By virtue of the asymptotic compactness of (X, T, π) and
according to Lemma 1.3 the set ω(ML ) 6= ∅, is compact, ω(ML ) ⊆ ω(B) and ω(ML )
attracts ML . Now to finish the proof of the lemma it is sufficient to note that the
set M is attracted by the compact ω(ML ), hence by the compact ω(B) too. In such
way we constructed the nonempty compact ω(B) attracting all set from the family
M. The lemma is proved. 

Corollary 1.7 Let (X, T, π) be point (compact, local, bounded) b-dissipative and
asymptotically compact. Then (X, T, π) is point (compact, local, bounded) k-
dissipative.

Corollary 1.8 Let (X, T, π) be asymptotically compact. Then local b-dissipativity

and compact b-dissipativity are equivalent.

Proof. The formulated statement follows from Corollary 1.7 and Theorem 1.23. 

Theorem 1.24 Let (X, T, π) satisfy the condition of Ladyzhenskaya. Then the
following conditions are equivalent:

1. the dynamical system (X, T, π) is weakly b-dissipative;

2. there exists a bounded set B1 ⊆ X absorbing all points from X, i.e. for every
x ∈ X there exists τ = τ (x) ≥ 0 such that xt ∈ B1 for all t ≥ τ ;
3. the dynamical system (X, T, π) is point k-dissipative;
4. the dynamical system (X, T, π) is weakly k-dissipative;
5. the dynamical system (X, T, π) is bounded k-dissipative;
6. there exists a bounded set B2 ⊆ X absorbing all bounded subsets of X, i.e.
for every B ∈ B(X) there exists L = L(B) > 0 such that π t B ⊆ B2 for all
t ≥ L(B).

Proof. It is clear that from 2. follows 1.. Let us show from 1. follows 3.. Let B0
be a weak attractor of the system (X, T, π). As (X, T, π) satisfies the condition of
Ladyzhenskaya, the set K1 := ω(B0 ) 6= ∅, is compact and attracts the set B0 . Since
for every x ∈ X there exists τ = τ (x) > 0 such that xτ ∈ B0 and consequently
lim ρ(xt, K1 ) = 0.
t→+∞
Obviously, from 3. follows 4. Let the condition 4. be fulfilled and K2 ⊆ X
be a nonempty compact such that ωx ∩ K2 6= ∅ for every x ∈ X, γ > 0 and
K3 := ω(B(K2 , γ)). Under the conditions of the theorem the set K3 6= ∅, is compact
and attracts B(K2 , γ). According to the condition 4. for every x ∈ X there exists
τ (x) > 0 such that xτ ∈ B(K2 , γ) and consequently x is attracted by the set K3 .
 Autonomous dynamical systems 33

So, (X, T, π) is point k-dissipative and according to Theorem 1.21 it is bounded

k-dissipative.
To finish the proof of the theorem it is sufficient to note that from 5. follows 6.
and from 6. follows 2. The theorem is proved. 

1.8 On a Problem of J. Hale

Let P : X → X be a continuous mapping and (X, P ) be a discrete dynamical

system, generated by the positive powers of mapping P .

Definition 1.32 A continuous mapping P : X → X is called a λ-contraction of

order k ∈]0, 1[, if λ(P (A)) ≤ kλ(A) for all A ∈ B(X), where λ is some measure of
non-compactness on X.

In the work [174] it was formulated the following problem

Problem (Problem of J. Hale). Let a discrete dynamical system (X, P ) be
pointwise b-dissipative and the mapping P be a λ-contraction. In the dynamical
system (X, P ), does a maximal compact invariant set exist?
In the relation to the problem of J. Hale it is interesting to note the following
problem.
Problem. Let a discrete dynamical system (X, P ) be pointwise b-dissipative.
To find the conditions which guarantee existence in the dynamical system (X, P ) a
maximal compact invariant set.
Below we give some results in relation to this problem.
The example 1.6 shows, that there exist pointwise k-dissipative dynamical
systems which do not admit a maximal compact invariant set. On the other hand,
by Theorem 1.6 for the compact k-dissipative dynamical system, its Levinson center
is a maximal compact invariant set.
The following assertion holds.

Lemma 1.21 Let (X, T, π) be a pointwise k-dissipative dynamical system and

J + (Ω) be a compact set, then in (X, T, π) there is a maximal negatively invariant
compact set I ⊆ J + (Ω).

Proof. Denote by I ∗ the family of all nonempty negatively invariant compact

subsets A ⊆ J + (Ω). We note that I ∗ 6= ∅, because according to the corollary 1.4
Ω ∈ I ∗.
Denote by I a closure of union of all negatively invariant compact subsets of
+
J (Ω). By Theorem 1.3 the set I is negative invariant and, consequently, it is
unknown. The lemma is proved. 

Corollary 1.9 Under the conditions of Lemma 1.21 the set I is invariant.
34 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Proof. In fact, since the set I is negatively invariant, then for every t ∈ T the
set π t I also is negatively invariant and compact and by Lemma 1.15 π t I ⊆ J + (Ω).
According to Lemma 1.21 I is a maximal compact negatively invariant set and,
consequently, π t I ⊆ I i.e. π t I = I for all t ∈ T. 
Thus, if (X, T, π) is pointwise k-dissipative and J + (Ω) is compact, then in
(X, T, π) there exists a maximal compact invariant set I ⊆ J + (Ω). On the other
hand according to the example 1.7, there are pointwise k-dissipative dynamical
systems, but not compact k-dissipative, for which the set J + (Ω) is compact.

Lemma 1.22 Let T = R+ . Dynamical system (X, T, π) is asymptotic compact,

if and only if for some t0 > 0 the discrete dynamical system (X, P ) is asymptotic
compact, where P := π t0 : X −→ X.

Proof. It easy to see that if the dynamical system (X, T, π) is asymptotic compact,
then the discrete dynamical system (X, P ) is asymptotic compact too, where P :=
π t0 : X −→ X and t0 is a arbitrary positive number. Let M ⊆ X be a bounded
positively invariant set of (X, T, π). Then P (M ) ⊆ M and according to asymptotic
compactness of discrete dynamical system (X, P ) there exists a compact K0 ∈ C(X)
such that

lim β(P n (M ), K0 ) = 0. (1.48)

n→+∞

Let K := π([0, t0 ], K0 ). Then by the continuity of mapping π : T × X −→ X and

by compactness of the set [0, t0 ] × K0 the set K is also compact. We will show that
the following equality lim β(π t M, K) = 0 holds. In fact, if we suppose that it is
t→+∞
false, then there exist ε0 > 0, tn → +∞ and {xn } ⊆ M such that

ρ(xn tn , K) ≥ ε0 . (1.49)

Let tn = kn t0 +τn , where kn is a integer part of the number tn /t0 and τn ∈ [0, t0 ],
then π(tn , xn ) = π(τn , π(kn t0 , xn )). Taking into account the equality (1.48) we
may suppose that the sequence {π(kn t0 , xn )} is convergent. Let τ0 := lim τn
n→+∞
and y0 := lim π(kn t0 , xn ), then lim π(tn , xn ) = π(τ0 , y0 ). Since y0 ∈ K0 ,
n→+∞ n→+∞
then π(τ0 , y0 ) ∈ K. On the other hand, passing to limit in the inequality (1.49) as
n → +∞, we obtain ρ(π(τ0 , y0 ), K) ≥ ε0 , i.e. y0 τ0 ∈
/ K. The obtained contradiction
completes he proof of Lemma. 

Definition 1.33 Recall (see [175, 179] and [278]) that the mapping P : X −→ X
is called condensing with respect to measure of the non-compactness λ, if λ(P (A)) <
λ(A) for all A ∈ B(X) with condition λ(A) > 0.

Lemma 1.23 [175] Let P : X → X be a continuous λ-condensing mapping. Then

the discrete dynamical system (X, P ) is asymptotic compact.
 Autonomous dynamical systems 35

Proof. At first we will show, that for every bounded set B ⊂ X with condition
P (B) ⊆ B and every sequence {kj } with integer kj → +∞ the set {P kj xj } is
relatively compact. Let C = {{P kj xj } : {xj } ⊆ B, {kj } are integers, 0 ≤ kj →
+∞}. We put η := sup{λ(h) : h ∈ C}. Since h ⊆ B, if h ∈ C, then η <
λ(B).We will prove the existence h∗ ∈ C such that λ(h∗ ) = η. Let {hl } be a
sequence of elements of the set C for which λ(hl ) → η. We will define e hl :=
 k
P j xj : P kj xj ∈ hl , kj > l . Then λ(e hl ) = λ(hl ). If h∗ = ∪{ehl | l = 1, 2, . . .}
is ordered in arbitrary way, then h∗ ∈ C and λ(h∗ ) ≥ λ(hl ) for every l. Thus

λ(h∗ ) = η. We put e h∗ := P kj −1 xj : P kj xj ∈ h∗ . Then eh∗ ∈ C and λ(e h∗ ) ≤ η.
ButP (eh ) = h , so that λ(P (e
∗ ∗
h )) ≥ λ(e
∗
h ) and, consequently, λ(e
∗ ∗
h ) = 0 because
e ∗ ∗
the mapping P is λ-condensing. Therefore h and h are relatively compact and
η = 0. Thus, each sequence in C is relatively compact.
By Lemma 1.3 the set ω(B) 6= ∅, compact, invariant and the equality

lim β(P n (B), ω(B)) = 0 (1.50)

n→+∞

holds. Lemma is proved. 

Definition 1.34 A dynamical system (X, T, π) is called locally bounded, if for

every x ∈ X there are δx > 0 and lx > 0 such that the set ∪{π t B(x, δx ) : t ≥ lx } is
bounded.

The example 1.6 shows, that there exist pointwise dissipative and locally
bounded dynamical systems that are not compact dissipative.

Definition 1.35 A dynamical system (X, T, π) is called λ-condensing, if

λ(π t B) < λ(B) for all B ∈ B(X), t ∈ T, λ(B) > 0 and π t B ∈ B(X).

Theorem 1.25 Let (X, T, π) be a pointwise b-dissipative, locally bounded and

λ-condensing dynamical system, then (X, T, π) is locally k-dissipative.

Proof. Let (X, T, π) be a λ-condensing dynamical system. Then by Lemmas 1.22

and 1.23 the dynamical system (X, T, π) is asymptotic compact. From the corollary
1.7 it follows that the dynamical system (X, T, π) is pointwise k-dissipative. Let
p ∈ X, δp > 0 and lp > 0 be numbers from the definition of local boundedness
of dynamical system (X, T, π). Consider a set Mp := ∪{π t B(p, δp ) : t ≥ lp }. By
asymptotic compactness of dynamical system (X, T, π) there exists a compact set
Kp ∈ C(X) such that

lim β(π t B(p, δp ), Kp ) = 0. (1.51)

t→+∞

According to Theorem 1.19 the dynamical system (X, T, π) is local dissipative. The
theorem is proved. 
36 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Theorem 1.26 Let (X, T, π) be a pointwise b-dissipative and λ-condensing

dynamical system. If for any p ∈ Ω there exist δp > 0 and lp > 0 such that
∪{π t B(p, δp )| t ≥ lp } ∈ B(X), then (X, T, π) is local k-dissipative.

Proof. Let (X, T, π) be pointwise dissipative and λ-condensing. Then by Lem-

mas 1.22 and 1.23 the dynamical system (X, T, π) is asymptotic compact. Ac-
cording to Corollary 1.7 the dynamical system (X, T, π) is pointwise k-dissipative
and, consequently, the set Ω is nonempty and compact. From the open cov-
ering {B(p, δp )| p ∈ Ω} of the compact set Ω we choice a finite sub-covering
{B(pi , δpi ) | i ∈ 1, m}. We put l0 := max{lpi | i ∈ 1, m}. By Lemma 1.9, there
exists γ > 0 such that B(Ω, γ) ⊂ ∪{B(pi , δpi )| i = 1, m}. If x ∈ X, then there
exists l(x) > 0 such that xl ∈ B(Ω, γ). Since the set B(Ω, γ) is open, then there
exists a number ε > 0 such that B(xl, ε) ⊂ B(Ω, γ). According to the continuity of
mapping π l : X −→ X, there exists a number δx > 0 such that yl ∈ B(xl, ε) for all
y ∈ B(x, δ) and, consequently, the set ∪{π t B(x, δ)| t ≥ l0 + l(x)} is bounded. To
finish the proof it is sufficient to use Theorem 1.25. 
Theorems 1.25 and 1.26 give a solution of the problem of J. Hale for the local
bounded dynamical systems.

Definition 1.36 A dynamical system (X, T, π) is called compact bounded, if for

every compact K ∈ C(X) the set Σ+ (K) is bounded, i.e. Σ+ (K) ∈ B(X).

The following theorem holds:

Theorem 1.27 Let (X, T, π) be a pointwise b-dissipative, compact bounded and

λ-condensing dynamical system. Then (X, T, π) is compact b-dissipative.

Proof. By Lemmas 1.22 and 1.23 the dynamical system (X, T, π) is asymptotic
compact and, consequently, (X, T, π) is pointwise k-dissipative. Since the dynamical
system (X, T, π) is compact bounded, asymptotic compact and the set Σ+ (K) is pos-
itively invariant, then the set Σ+ (K) is relatively compact for every K ∈ C(X) and
according to Theorem 1.15 the dynamical system (X, T, π) is compact k-dissipative.


Definition 1.37 A dynamical system (X, T, π) is called bounded, if for every

B ∈ B(X) the set Σ+ (B) is bounded, i.e. Σ+ (B) ∈ B(X).

Theorem 1.28 Let (X, T, π) be a bounded and λ-condensing dynamical system.

Then (X, T, π) is bounded k-dissipative.

Proof. This assertion can be proved using the same arguments as in the proof of
Theorem 1.25. 
 Autonomous dynamical systems 37

At the conclusion of this paragraph we give an example of pointwise dissipative

dynamical system that is local bounded and does not admit a maximal compact
invariant set.

Example 1.9 Let ϕ ∈ C(R, R) be a function possessing the following properties:

1. ϕ(0) = 0;
2. supp(ϕ) = [0, 2];
3. ϕ ∈ C ∞ (R, R);
4. ϕ(1) = 1;
5. the function ϕ is monotone creasing from 0 to 1 and it is decreasing from 1 to
2;
6. xϕ(x−1 ) → 0 as x → +∞.

A function ϕ with the properties 1. − 6. can be constructed as following. Let

(

exp [t2 − 1]−1 + 1 , |t| < 1
ϕ0 (t) :=
0 , |t| ≥ 1.

Then the function ϕ(t) := ϕ0 (t + 1) is unknown. We put X := {aϕ at + h | h ∈
R, a > 0} ∪ {θ}, where θ is function from C(R, R) identical equal to 0. It is possible
to show that the set X is closed in C(R, R) and it is invariant with respect to
shifts. Thus on the set X is induced a dynamical system (on C(R, R) is defined a
dynamical system of translations or Bebutov’s dynamical system), which we denote
by (X, R, σ). We will indicate some properties of this system:

(1) for every function ψ ∈ X the set {σ(t, ψ) : t ∈ R} is relatively compact and
ωψ = αψ = {θ};
(2) the dynamical system (X, R, σ) is pointwise dissipative and Ω :=
∪{ωx | x ∈ X} = {θ};
(3) D+ (Ω) = X and, consequently, the dynamical system (X, R, σ) is not compact
dissipative because the set X, evidently, is not compact;
(4) the dynamical system (X, R, σ) is not local bounded;

In fact, for every δ > 0 we have Σ+ (B(θ, δ)) := ∪{π t B(θ, δ) : t ≥ 0} = X. From
this equality it follows the set Σ+ (B(θ, δ)) is not bounded for all δ > 0.

5. the dynamical system (X, R, σ) does not admit a maximal compact invariant
set.

The necessary example is constructed.

38 Global Attractors of Non-autonomous Dissipative Dynamical Systems

1.9 Connectedness of the Levinson’s center

Below we give the conditions of of the Levinson center for a compact dissipative
dynamical system.

Lemma 1.24 Let M1 , M2 ∈ C(X), W s (M1 ) 6= ∅ and W s (M2 ) 6= ∅, then:

(1) if W s (M1 ) ∩ W s (M2 ) 6= ∅, then M1 ∩ M2 6= ∅.

(2) if M1 ∩ M2 = ∅, then W s (M1 ) ∩ W s (M2 ) = ∅.
(3) if the sets M1 and M2 are attracting and W s (M1 ) ∩ W s (M2 ) = ∅, then M1 ∩
M2 = ∅.
(4) if the sets M1 and M2 are attracting and M1 ∩ M2 6= ∅, then W s (M1 ) ∩
W s (M2 ) 6= ∅.

Proof. The formulated affirmation follows from the corresponding definitions. 

Definition 1.38 A closed (open) positively invariant set M ⊆ X of a dynamical

system (X, T, π) is called indecomposable if it cannot be represented in the form of
a union of two own closed (open) positively invariant subsets.

Lemma 1.25 Let M be a nonempty compact positively invariant and indecom-

posable subset of X and T = R+ . Then M is connected.

Proof. Suppose that the set M is not connected. Then there are nonempty closed
subsets M1 and M2 ⊆ X such that M = M1 t M2 . Let x ∈ M1 and Ti =
{t ∈ T | xt ∈ Mi } (i = 1, 2). Then according to the continuity of the mapping
π(·, x) : T → X the sets T1 and T2 are closed, T1 ∩ T2 = ∅ and T = T1 ∪ T2 . From
the connectedness of the set R+ , it follows that T2 = ∅ and, consequently, the set
M1 is positively invariant. Using the same arguments we can prove that the set M 2
is also connected. But this fact contradicts to the indecomposability of the set M .
The lemma is proved. 

Lemma 1.26 Let M ∈ C(X) be orbitally stable, M = M1 t M2 and each of the

sets M1 and M2 be nonempty and positively invariant. Then

1) the sets M1 and M2 are orbitally stable;

2) W s (M ) = W s (M1 ) ∪ W s (M2 ).
3) if the set M is asymptotically stable, then the sets M1 and M2 are also asymp-
totically stable.

Proof. Let d = ρ(M1 , M2 ) > 0. For the number d/2 we will choice a number δ2
such that the inequality ρ(x, Mi ) < δ2 (i = 1, 2) implies ρ(xt, Mi ) < d/2 for all
t ∈ [0, 1]. We take positive numbers ε > 0 (ε < d/2) and δ1 = δ1 (ε, M ) out of the
 Autonomous dynamical systems 39

condition of orbital stability of the set M . Let δ(ε) := min(δ1 , δ2 ) and show that
from the inequality ρ(x, M1 ) < δ(ε) follows ρ(xt, M1 ) < ε for all t ≥ 0. In fact, if
ρ(x, M1 ) < δ(ε), then ρ(x, M ) < δ(ε) and, consequently, ρ(xt, M ) < ε for all t ≥ 0.
Denote by

Ti = {t ∈ T | ρ(xt, Mi ) < ε} (i = 1, 2).

Note, that if the set T is connected, then T2 = ∅ and, consequently, T1 = T.

Let us consider the case when the set T is discrete. We will show that then the
set T2 is also empty. Suppose that T2 6= ∅ and denote by t0 = inf T2 . It is clear
that t0 > 1, t0 ∈ T2 and t0 − 1 ∈ T1 . Note, that for the point y := x(t0 − 1) we
have ρ(y, M1 ) < ε and, consequently,

ρ(y, M1 ) < d/2. (1.52)

On the other hand,

ρ(y, M1 ) ≥ ρ(M1 , M2 ) − ρ(y, M2 ) > d/2. (1.53)

The inequalities (1.52) and (1.53) are contradictory. Thus, our assumption that
T2 6= ∅ is not true and, consequently, T = T1 , i.e. the set M1 is orbitally stable.
On the analogy with this we can prove the orbital stability of the set M2 . Hence,
the first statement of the theorem is proved.
Now let us prove that W s (M ) = W s (M1 ) ∪ W s (M2 ). Obviously, it is suf-
ficient to show that W s (M ) ⊂ W s (M1 ) ∪ W s (M2 ). Let 0 < ε < d/2 and
δ(ε) = min(δ(ε), δ1 (ε), δ2 (ε)), where the number δ(ε) (δ1 (ε), δ2 (ε)) is chosen out
of the orbital stability of the sets M (M1 , M2 ). Then for every point x ∈ W s (M )
there exists t0 ≥ 0 such that ρ(xt0 , M ) < δ(ε). One of the following two conditions
is true:

ρ(xt0 , M1 ) < δ(ε) or ρ(xt0 , M2 ) < δ(ε).

If the first condition is true, then ρ(xt, M1 ) < ε for all t ≥ t0 , ωx ⊆ B(M1 , ε) and
ωx ∩ M2 = ∅, i.e. ωx ⊆ M1 and x ∈ W s (M1 ). If the second condition holds, then
ρ(xt, M2 ) < ε for all t ≥ t0 , ωx ⊆ B(M2 , ε) and ωx ∩ M1 = ∅, i.e. ωx ⊆ M2 and
x ∈ W s (M2 ).
Now it remains to prove the third statement. For this it is sufficient to prove
that each of the sets M1 and M2 is attracting. Since the set M is asymptotic
stable, then there exists a number γ > 0 such that B(M, γ) ⊂ W s (M ), B(M, γ) =
B(M1 , γ) t B(M2 , γ) and γ < d/2, where d = ρ(M1 , M2 ) > 0. For the number
γ we will choose δ = δ(M1 , γ) > 0 (δ = δ(M2 , γ) > 0) out of the condition of
orbital stability of the set M1 (M2 ). Then if x ∈ B(M1 , δ) (x ∈ B(M2 , δ)), we have
xt ∈ B(M1 , γ) (xt ∈ B(M2 , γ)) for all t ≥ 0 and, consequently, ωx ⊆ B(M1 , γ)
40 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(ωx ⊆ B(M2 , γ), i.e. x ∈ W s (M1 ) (x ∈ W s (M2 )). Thus B(M1 , δ) ⊆ W s (M1 )
(B(M2 , δ) ⊆ W s (M2 )). The lemma is completely proved. 

Theorem 1.29 Let (X, T, π) be a compact dissipative dynamical system and J

be its Levinson center. If the space X is indecomposable, then the set J is also
indecomposable.

Proof. We will show that if the space X is indecomposable, then the set J possesses
the same property. Suppose that it is not true. Then there exist nonempty compact
positively invariant subsets J1 , J2 ⊆ J such that J = J1 t J2 . Since the set J is
asymptotically stable, then by Lemma 1.26 each of the sets J1 and J2 possesses the
same property, X = W s (J) = W s (J1 ) ∪ W s (J2 ) and, in addition, by Lemmas 1.8
and 1.26 each of the sets W s (J1 ) and W s (J2 ) is open. From Lemma 1.24 it follows
that W s (J1 ) ∩ W s (J2 ) = ∅. The last equality contradicts the indecomposability of
the set J. 

Lemma 1.27 Let M be a nonempty compact positively invariant attracting subset

of X. If the set M is indecomposable and the set W s (M ) is closed, then the set
W s (M ) is indecomposable too.

Proof. Suppose that the conditions of the lemma are fulfilled, but the set W s (M )
is decomposable. Then the set W s (M ) can be presented in the form of a disjoint
union of two own closed positively invariant subsets A1 and A2 . It is clear that
M ⊆ A1 ∪ A2 . Note, that for every point x ∈ A1 ∅ 6= ωx ⊆ M ∩ A1 and for every
point y ∈ A2 ∅ 6= ωx ⊂ M ∩ A2 and, consequently, Mi = M ∩ Ai 6= ∅ (i = 1, 2) and
M = M1 ∪ M2 . The last equality contradicts the indecomposability of the set M  .

Corollary 1.10 If the set M ∈ C(X) is attracting and connected, then the set
W s (M ) is connected too.

Proof. The formulated statement follows from Lemmas 1.25 and 1.27. 

Corollary 1.11 Let (X, R+ , π) be a compact dissipative dynamical system and

J be its Levinson center. The set J is connected if and only if the space X is
connected.

Proof. This statement follows dierectly from Theorem 1.29, Lemma 1.25 and
Corollary 1.10. 

Remark 1.6 Different versions of this statement are contained in the works [20,
175, 179, 232, 294] (see also the references within).
 Autonomous dynamical systems 41

Corollary 1.12 For a compact dissipative dynamical system (X, T, π) from the
indecomposability of its Levinson center follows the indecomposability of the space
X.

Theorem 1.30 Let (X, T, π) be a compact dissipative dynamical system and J

be its Levinson center. The set J is indecomposable if and only if the space X is
indecomposable.

Proof. This affirmation follows from Theorem 1.29 and Corollary 1.10. 

Lemma 1.28 Let M ∈ C(X) be an asymptotically stable set. If the set W s (M )

is indecomposable, then the set M is indecomposable too.

Proof. Suppose that the statement of the lemma is not true, i.e. the compact
set M is asymptotically stable and the set W s (M ) is indecomposable, but the set
M is decomposable. Then there exist nonempty subsets M1 and M2 of the set M
such that M = M1 t M2 . By Lemma 1.24 W s (M1 ) ∩ W s (M2 ) = ∅. Since the
set M is asymptotically stable, then according to Lemma 1.26 the sets M1 , M2
and W s (M ) = W s (M1 ) ∪ W s (M2 ) possess the same property. The sets W s (M1 )
and W s (M2 ) are positively invariant and by Lemmas 1.8 and 1.26 each of the sets
W s (M1 ) and W s (M2 ) is open. That contradicts the indecomposability of the set
W s (M ). 

Corollary 1.13 Let M ∈ C(X) be an asymptotically stable set. The set M is

connected if and only if the set W s (M ) is connected.

Theorem 1.31 Let X be locally connected. If M ∈ C(X) is a nonempty positively

invariant uniform attracting set, then the set M has a finite number of components
of connectivity.

Proof. By Lemma 1.8 the set W s (M ) is open. Since the space X is locally
connected, then the set W s (M ) is also locally connected. We will cover the set M
by a finite number of open connected sets H1 , H2 , . . . , Hm satisfying the condition
H = H1 ∪ H2 ∪ · · · ∪ Hm ⊂ W s (M ). Let V1 , V2 , . . . , Vk be the components of
connectivity of the open set H. Then the sets V1 , V2 , . . . , Vk are mutually disjoint
and they are contained in W s (M ). Denote by Mi := M ∩ Vi (i = 1, k) and note
that M = M1 ∪M2 ∪· · · ∪Mk , the sets M1 , M2 , . . . , Mk are mutually disjoint and in
virtue of Lemma 1.24 W s (M1 ), W s (M2 ), . . . , W s (Mk ) also are mutually disjoint.
If T = R+ , then each of the sets Mi is closed, positively invariant and in view
of Lemma 1.26 each of the sets Mi is asymptotically stable, W s (M ) = W s (M1 ) ∪
W s (M2 ) ∪ · · · ∪ W s (Mk ) and Vi ⊂ W s (Mi ). Every set Mi (i = 1.k) is positively
invariant and by Lemma 1.8 they are open. We will prove the connectedness of
42 Global Attractors of Non-autonomous Dissipative Dynamical Systems

the sets Mi (i = 1.k). Assuming the contrary, we obtain that the set Mi can be
presented in the form of a union of two closed nonempty disjoint positively invariant
sets Mi0 and Mi00 in Vi . By virtue of Lemma 1.24 W s (Mi0 ) ∩ W s (Mi00 ) = ∅. Note,
that the sets Mi0 and Mi00 are asymptotically stable and W s (Mi ) = W s (Mi0 ) ∪
W s (Mi00 ). The sets Mi0 and Mi00 are open and positively invariant. That contradicts
the connectivity of the set Vi .
Let now (X, f ) be a discrete dynamical system. Since the set M is uniformly
attracting, then for a sufficiently large number m we have f m (H) ⊂ H and, con-
sequently, f m (Vi ) ⊂ V1 ∪ V2 ∪ · · · ∪ Vk . According to the connectivity of the sets
V1 , V2 , . . . , Vk and the continuity of the mapping f , the sets f m (Vi ) i = 1, k
are connected and, consequently, for every number i there exists a unique num-
ber j such that f m (Vi ) ⊂ Vj . Let as above Mi := M ∩ Vi (i = 1, k), then
f m (Mi ) ⊆ f m (M ) ∩ f m (Vi ) ⊆ M ∩ Vj = Mj . Since the set M is positively in-
variant, then f m (M ) ⊆ M and, consequently, f m (Mi ) ⊆ Mj . Thus, the mapping
f m permutes the sets M1 , M2 , . . . , Mk and, consequently, there exists a number n
such that f n (Mi ) ⊂ Mi (i = 1, k). Thus, every set M1 , M2 , . . . , Mk is positively
invariant with respect to mapping f n . Then in view of Lemma 1.26 each of the sets
M1 , M2 , . . . , Mk is asymptotically stable with respect to the discrete dynamical
system (X, f n ) and W s (M ) = W s (M1 ) ∪ W s (M2 ) ∪ · · · ∪ W s (Mk ). Note, that
f n (Vi ) ⊆ Vi and Vi ⊂ W s (Mi ) (i = 1, k). Using the similar arguments as ones
above, for the case T = R+ we obtain the connectivity of the set Mi . Theorem is
completely proved. 

Theorem 1.32 Let (X, Z+ , f ) be a locally dissipative dynamical system and J be

its Levinson center. If the space X is connected and locally connected, then the set
J is also connected.

Proof. If the space X is connected and locally connected, then by Theorems

1.17 and 1.31 the center of Levinson of the discrete dynamical system (X, Z+ , f )
consists of finite numbers of components of the connectivity J1 , J2 , . . . , Jk . In
the proof of Theorem 1.31 there has been established the existence of a number n
such that f n (Ji ) ⊆ Ji (i = 1, k). Put P := f n and consider the discrete dynamical
system (X, P ) generated by the the positive powers of the mapping P . Note, that
the system (X, P ) is compactly dissipative and satisfies the following condition:
J = J1 t J2 t · · · t Jk . In addition, we have P (Ji ) ⊆ Ji (i ∈ 1, k). Thus, the center
of Levinson J˜ of the discrete dynamical system (X, P ) is decomposable, but this
fact contradicts Theorem 1.29 because the space X is connected and, consequently,
indecomposable. The theorem is proved. 

Remark 1.7 We note that only the connectedness of the space X, without local
connectivity, does not guarantee the connectivity of the center of Levinson J of the
discrete dynamical system (X, f ) (see [166]).
 Autonomous dynamical systems 43

Definition 1.39 A metric space X is said to be possessing the property (S) if for
every compact K ⊆ X there exists a connected compact I ⊆ X such that K ⊆ I.

Remark 1.8 a) If the space X possesses the property (S), then it is connected.
b) It is possible to construct an example of a connected space that does not possess
the property (S).
c) Every linear metric space possesses the property (S).
In fact, if K ⊆ X is a compact set, then the set L(K) = {λx + (1 − λ)y | x, y ∈
K, 0 ≤ λ ≤ 1} is a connected compact and K ⊆ L(K).

Theorem 1.33 Let (X, Z+ , π) be a compact dissipative dynamical system, J be

its Levinson center and the space X possesses the property (S). Then the set J is
connected.

Proof. Let X be a space with the property (S). Suppose that the set J is not
connected. Then there are such subsets U1 and U2 that J ⊂ U1 ∪ U2 , J ∩ Ui 6= ∅
(i ∈ 1, 2) and U1 ∩ U2 = ∅. Let I be some connected compact containing the
set J. Then the set π(t, I) is a connected compact for every t ∈ Z+ . As J ⊂ I
and π(t, J) = J (t ∈ Z+ ), then J ⊂ π(t, I) and, consequently, π(t, I) ∩ Ui 6= ∅
(i = 1, 2, t ∈ Z+ . Let tn → +∞. By the connectivity of the set π(tn , I) there exists
xn ∈ π(tn , I) such that xn 6∈ U1 ∪ U2 . Since the dynamical system (X, Z+ , π) is
compactly dissipative, then we can suppose that the sequence {xn } is convergent.
Let xn → x as tn → +∞. Thus, we have x 6∈ U1 ∪ U2 and, on the other hand,
x ∈ J ⊆ U1 ∪ U2 . The obtained contradiction proves the theorem. 

Corollary 1.14 Let X be a complete linear metric space. The Levinson center
of the compact dissipative dynamical system (X, Z+ , f ) is a connected set.

Remark 1.9 a) In the case when X is a Banach space, the statement containing
in Corollary 1.14 was proved in the work [248].
b) There exist examples of compact dissipative dynamical systems (X, Z + , π)
with a disconnected phase space X and with a connected Levinson center.

1.10 Weak attractors and center of Levinson

Let (X, T, π) be a compact dissipative dynamical system. Recall that a compact set
M ⊆ X is called a weak attractor of the dynamical system (X, T, π) if ωx ∩M 6= ∅ for
all x ∈ X. In this paragraph we establish the relationship between weak attractors
of the dynamical system (X, T, π) and its Levinson center.
44 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Definition 1.40 The point p ∈ X is uniformly stable in the sense of Lagrange

in positive direction if the sequence {xk tk } is relatively compact for each {xk } → p
and {tk } → +∞.

Theorem 1.34 The point p ∈ X is uniformly stable in the sense of Lagrange in

positive direction if and only if the following conditions are fulfilled:
T T S τ
1. the set Jp+ := π B(p, ε) 6= ∅ and is compact;
ε>0 t≥0 τ ≥t
2. the set Jp+ is invariant;
3.

lim ρ(xt, Jp+ ) = 0. (1.54)

x→p,t→+∞

Proof. Let p ∈ X, xn → p and tn → +∞. Then under the conditions of the

theorem the sequence {xn tn } is relatively compact. If y is a limit point of the
sequence {xk tk }, then y ∈ Jp+ 6= ∅. We will show that the set Jp+ is compact. Let
{ym } ⊆ Jp+ , εm ↓ 0, km ∈ N and xm m
k → p and tk → +∞ as k → +∞ (for every
m ∈ N) such that:

a. xm m
k tk → ym as k → +∞ (m ∈ N);
b. ρ(xm
km , p) < εm ;
c. tm m m
km > m and ρ(xkm tkm , ym ) < εm .

Let xm := xm m
km and tm := tkm > m, then xm → p and tm → +∞ and, consequently,
under the conditions of the theorem we may suppose that the sequence {xm tm } is
convergent and y := lim xm tm . Then
m→+∞

ρ(ym , y) ≤ ρ(ym , xm tm ) + ρ(xm tm , y) < εm + ρ(xm tm , y). (1.55)

Passing to limit in the inequality (1.55) when m → +∞, we obtain y = lim ym ,

m→+∞
Jp+
i.e., the set is compact.
We will show now that the set Jp+ is invariant, i.e. π t Jp+ = Jp+ for all t ∈ T .
First, we note that from the definition of Jp+ follows its positive invariance, i.e.,
π t Jp+ ⊆ Jp+ for all t ∈ T . Thus, to prove that the set Jp+ is invariant is sufficient
to prove that Jp+ ⊆ π t Jp+ for all t ∈ T . Let q ∈ Jp+ and t ∈ T , then there exist
xk → p and tk → +∞ such that q = lim xk tk = lim π t (π tk −t xk ). Since
k→+∞ k→+∞
tk − t → +∞ and xk → p, the sequence {xk (tk − t)} is relatively compact. Without
loss of generality we can suppose that the sequence {xk (tk − t)} is convergent. We
put z = lim xk (tk − t), then q = π t z, i.e. Jq+ ⊆ π t Jq+ .
k→+∞
Finally, let us show that the equality (1.54) is true. If we suppose that it is not
so, then there exist ε0 > 0, xk → p and tk → +∞ such that

ρ(xk tk , Jp+ ) ≥ ε0 . (1.56)

 Autonomous dynamical systems 45

According to the stability in the sense of Lagrange in positive direction of the point
p, we may suppose that the sequence {xk tk } is convergent. Let q := lim xk tk ,
k→+∞
then q ∈ Jp+ . Passing to limit in the inequality (1.56) as k → +∞, we obtain
ρ(q, Jq+ ) ≥ ε0 , i.e. q ∈
/ Jp+ . The obtained contradiction proves our statement.
To finish the proof of the theorem it is sufficient to note that from the conditions
3. and 1. follows the uniform stability in the sense of Lagrange in positive direction
of the point p. The theorem is proved. 

Lemma 1.29 Let M 6= ∅ and be compact. Then the following equality

J + (M ) = ∪{Jm
+
|m ∈ M } (1.57)

holds.

+
Proof. It is easy to see that Jm ⊆ J + (M ) for all m ∈ M and, consequently,
S +
{Jm | m ∈ M } ⊆ J + (M ). We will show that the inverse inclusion also is true.
Let x ∈ J + (M ), then there exist {xn } and tn → ∞ such that x = lim xn tn and
n→+∞
ρ(xn , M ) → 0 as n → +∞. By virtue of compactness of the set M we can suppose
+
that the sequence {xn } is convergent. Let lim xn = m ∈ M , then x ∈ Jm and,
n→+∞
consequently, J + (M ) ⊆ ∪{Jp+ |p ∈ M }. The lemma is proved. 

Theorem 1.35 Let (X, T, π) be compactly dissipative, J be its Levinson center

and M be a compact weak attractor of the dynamical system (X, T, π). Then J =
J + (M ).

Proof. Since the set M is compact, then according to Lemma 1.29 the equality
(1.57) holds. Thus, the dynamical system (X, T, π) is compactly dissipative, then
every point p ∈ X is uniformly stable in the sense of Lagrange in positive direction.
+
By virtue of Theorem 1.34 for each m ∈ M the set Jm 6= ∅, is compact and invariant
+ +
and, consequently, Jm ⊆ J. Hence, J (M ) ⊆ J. Show that the reverse inclusion
is also true. Let x ∈ J, then by the invariance of the set J and Theorem 1.1 there
exists a whole trajectory ϕ passing through point x at the initial moment t = 0 and
αϕx 6= ∅, is compact and positively invariant. If p ∈ αϕx , then ωp ⊆ αϕx . Since
ωp ∩ M 6= ∅, then there are q ∈ ωp ∩ M and tn → +∞ such that ϕ(−tn ) → q and,
consequently, x = π tn ϕ(−tn ) ∈ Jq+ ⊆ J + (M ), i.e. J ⊆ J + (M ). The theorem is
proved. 

Lemma 1.30 Let M be a weak attractor of the compact dissipative dynamical

system (X, T, π). Then ω(M ) is also a weak attractor of this system.

Proof. As (X, T, π) is compactly dissipative, then according to Lemma 1.3 ω(M ) 6=

∅, is compact and invariant. We will show that the set ω(M ) is a weak attractor
46 Global Attractors of Non-autonomous Dissipative Dynamical Systems

of the system (X, T, π). If we suppose that it is not true, then there exists x0 ∈ X
such that

ωx0 ∩ ω(M ) = ∅. (1.58)

Since ωx0 ∩ M 6= ∅, then there is m ∈ ωx0 ∩ M . According to the compactness

and invariance of the set ωx0 , we may suppose that the point m is recurrent and,
consequently, m ∈ ωm ⊆ ω(M ). The last inclusion contradicts to the condition
(1.58). The obtained contradiction finishes the proof of the lemma. 

Lemma 1.31 Let M be a weak attractor of the system (X, T, π) and let a point
p ∈ X be recurrent, then p ∈ ω(M ).

Proof. If the point p ∈ X is recurrent, then ωp ∩ M 6= ∅. Let q ∈ ωp ∩ M . Since

the point p is recurrent, there exists a sequence tn → +∞ such that qtn → p and,
consequently, p ∈ ω(M ). The lemma is proved. 

Lemma 1.32 Let M be a weak attractor of a dynamical system (X, T, π). Then
the closure of the set M consisting of all recurrent points of the compact dissipative
dynamical system (X, T, π) is also a weak attractor and M ⊂ M .

Proof. First, we note that M is a a weak attractor of the system(X, T, π). Really, let
x ∈ X. Then ωx contains at least one recurrent point and, consequently, ωx ∩M 6= ∅.
Let now M be a closed, invariant and weakly attracting set of the system (X, T, π)).
Then ω(M ) = M and, consequently, according to Lemma 1.31 M ⊆ ω(M ) = M .
The lemma is proved. 

Remark 1.10 For a compact dissipative dynamical system (X, T, π) the maximal
compact invariant weak attractor is its Levinson center J.

Lemma 1.33 Let (X, T, π) be a compact dissipative dynamical system, J be its

S
center of Levinson and Ω(J) = {ωx |x ∈ J}. Then Ω(J) is a weak attractor of the
system (X, T, π).

Proof. Let p ∈ X be a recurrent point, then ωp is a nonempty compact minimal

invariant set. Since J is a maximal compact invariant set of the dynamical system
(X, T, π), then ωp ⊆ J. Thus, p ∈ ωp ⊆ J and, consequently, p ∈ Ω(J). This
means that M ⊆ Ω(J) (M is a closure of all recurrent points) and, consequently,
M ⊆ Ω(J) also is a weak attractor. The lemma is proved. 

Corollary 1.15 J = J + (Ω).

Proof. The formulated statement follows from Lemma 1.33 and Theorem1.35. 
 Autonomous dynamical systems 47

1.11 Asymptotic stability

In this paragraph we study the problem of asymptotic stability of compact posi-

tively invariant sets of the dynamical systems with infinite-dimensional (non locally
compact) phase space. The different conditions that are equivalent to asymptotic
stability are given. The obtained results are the local version of the results proved
before for dissipative dynamical systems.

Lemma 1.34 Let M ⊆ X be a compact and asymptotically stable set of a

dynamical system (X, T, π). Then there exists δ0 > 0 such that

lim β(π t K, M ) = 0 (1.59)

t→+∞

for every compact K ∈ C(B(M, δ0 )).

Proof. Since M is asymptotically stable, then there exists δ0 such that B(M, δ0 ) ⊂
W s (M ). We will show that for every compact K ∈ C(B(M, δ0 )) the equality (1.59)
holds. For this we note, that from the asymptotic stability of the set M follows the
existence for all ε > 0 and x ∈ B(M, δ0 ) of such positive numbers δ(ε, x) and l(ε, x)
that

ρ(yt, M ) < ε (1.60)

for all t ≥ l(ε, x) and y ∈ B(x, δ). We will show that from (1.60) follows (1.59). If
we suppose that it is not true, then there are K0 ∈ C(B(M, δ0 )), tn → +∞ and
ε0 > 0 such that

β(πnt K0 , M ) ≥ ε0 . (1.61)

From the inequality (1.61) follows that there are xn ∈ K0 such that

ρ(xn tn , M ) ≥ ε0 . (1.62)

Since the set K0 is compact, then we can suppose that the sequence {xn } is con-
vergent. We put x0 := lim xn , then from the inequality (1.60) follows that for n
n→+∞
large enough the inequality

ρ(xn tn , M ) < ε0 , (1.63)

holds. But the inequality (1.63) contradicts (1.62). The obtained contradiction
proves our statement. 

Theorem 1.36 Let (X, T, π) be a dynamical system, M ⊆ X be a compact and

positively invariant subset. Then the following conditions are equivalent:

1. the set M is asymptotically stable;

48 Global Attractors of Non-autonomous Dissipative Dynamical Systems

2. the set W s (M ) is open and for each ε > 0 and x ∈ W s (M ) there are δ =
δ(ε, x) > 0 and τ = τ (ε, x) > 0 such that

β(π t B(x, δ), M ) < ε (1.64)

for all t ≥ τ ;
3. the set W s (M ) is open and M attracts all compact subsets from W s (M ), i.e.
for every compact K ∈ C(W s (M )) the equality (1.59) holds.
4. the set W s (M ) is open, every compact K ⊆ W s (M ) is stable in the sense of
Lagrange in positive direction and ∅ 6= Jx+ ⊆ M for all x ∈ W s (M ).

Proof. We will show that 1. implies 2. First of all, let us prove that the set
W s (M ) is open. In fact, since M is a attractive set, then there exists δ > 0 such
that B(M, δ) ⊂ W s (M ). Let now q ∈ W s (M ) \ B(M, δ), then there is τq > 0 such
that qτq ∈ B(M, δ). Since the set B(M, δ) is open, then there exists γ > 0 such
that

B(qτq , γ) ⊂ B(M, δ). (1.65)

According to the continuity of the mapping π τq : X → X there exists η > 0 such

that the inclusion

π τq B(q, η) ⊂ B(qτq , γ)

holds. By Lemma 1.34 B(q, η) ⊂ W s (M ) and, consequently, the set W s (M ) is

open.
Denote by ξ(ε) > 0 a number, chosen for ε > 0 out of the condition of orbital
stability of the set M . Since W s (M ) = {x ∈ X : lim ρ(xt, M ) = 0 for t → +∞},
t→+∞
then for all ξ and x ∈ W s (M ) there exists τ = τ (ε, x) > 0 such that ρ(xt, M ) < ε
for all t ≥ τ . From the continuity of the mapping π τ (ε,x) : X → X follows that for
x ∈ W s (M ) and ξ > 0 there is δ = δ(ε, x) > 0 such that

π τ B(x, δ) ⊆ B(xτ, ξ). (1.66)

By the choice of the number ξ and from (1.66) we have β(π t B(x, δ), M ) < ε for all
t ≥ τ.
Now we will show that from 2. follows 3. Let ε > 0 and ∅ 6= K ⊂ W s (M ) be
compact. Then for every point x ∈ K there exist δ = δ(ε, x) > 0 and τ = τ (ε, x) > 0
such that the inequality (1.64) holds for all t ≥ τ (ε, x). Consider an open covering
{B(x, δ(ε, x))}x∈K of the set K. Since the set K is compact and the space X is
complete, we can extract from this covering a finite sub-covering {B(xi , δ(ε, xi )}ni=1 .
Let l(ε, K) := max{τ (ε, xi ) | i ∈ 1, n}, then from the inequality (1.64) follows that
β(π t K, M ) < ε for all t ≥ l(ε, K).
 Autonomous dynamical systems 49

Let us prove now that 3. implies 4. Really, since the set M attracts every
compact K ⊂ W s (M ), then by Theorem 1.5 the set Σ+ (K) = ∪{π t K | t ≥ 0} is
relatively compact and, consequently, Jx+ ⊇ ωx 6= ∅ for every point x ∈ W s (M ).
We note that Jx+ ⊆ M for all x ∈ W s (M ). In fact, if y ∈ Jx+ , then there exist
xn → x and tn → +∞ such that xn tn → y. Since the set K1 := {xn } is compact
and the set M attracts K1 , then y ∈ M , i.e. Jx+ ⊂ M .
Finally, we will prove that from 4. follows 1. For this aim we note that from
the openness of W s (M ) and compactness of M follows the existence of δ > 0 such
that B(M, δ) ⊂ W s (M ), i.e. the set M is attracting. Now show that the set M is
orbitally stable. Since the set M is positively invariant, then

M ⊆ D+ (M ) = Σ+ (M ) ∪ J + (M ) ⊆ M ∪ M = M

and, consequently, D + (M ) = M . To finish the proof of the theorem is sufficient

to note that under the conditions of the theorem from the equality D + (M ) = M
follows the orbital stability of the set M . If we suppose that it is not true, then
there are ε0 > 0, ε0 > δn → 0, xn ∈ B(M, δn ) and tn → +∞ such that

ρ(xn tn , M ) ≥ ε0 . (1.67)

As xn ∈ B(M, δn ), δn → 0 and the set M is compact, then the set K := {xn } ⊆

B(M, ε0 ) is compact and, consequently, the set Σ+ (K) is relatively compact. Thus,
we may suppose that the sequence {xn tn } is convergent. Let y = lim xn tn , then
n→+∞
y ∈ D+ (M ) = M . From the equality (1.67), it follows that y ∈
/ M . The obtained
contradiction finishes the proof of the theorem. 

Theorem 1.37 Let M ⊆ X be a nonempty compact positively invariant and

asymptotically stable subset of a dynamical system (X, T, π), then the following
affirmations hold:
(1) ω(M ) ⊆ M ;
(2) the set ω(M ) is invariant;
T t
(3) ω(M ) = π M;
t≥0
(4) ω(M ) is a maximal compact invariant set in W s (M ).

Proof. The first statement of the theorem is true because the set M is positively
invariant and closed.
Let us prove the second statement. Since the set M is stable in the sense of
Lagrange in positive direction and ω(M ) ⊆ M , then by Lemma 1.3 the set ω(M )
is invariant.
50 Global Attractors of Non-autonomous Dissipative Dynamical Systems

T
To prove the third statement of the theorem we note that π t M ⊆ ω(M ).
t≥0
Since from the inclusion ω(M ) ⊆ M and the invariance of the set ω(M ) follows
T t T t
that ω(M ) ⊆ π M, then ω(M ) = π M.
t≥0 t≥0
Finally we will show that the fourth statement is true. Let K ⊂ W s (M ) be an
arbitrary compact invariant set. By Theorem 1.36 the set M attracts the compact
K and, consequently, K ⊆ ω(K) ⊆ ω(M ). The theorem is proved. 

Theorem 1.38 Let M ⊆ X be a nonempty compact asymptotically stable set.

Then the following statements hold:

(1) the set ω(M ) is orbitally stable;

(2) the set ω(M ) attracts every compact from W s (M );
(3) J + (A) = ω(M ), where A = ∪{αϕx | ϕ ∈ Φx , x ∈ ω(M )}.

Proof. Show that the set ω(M ) is orbitally stable. If we suppose that it is not
true, then there exist ε0 > 0 and δn → 0 such that for certain xn ∈ B(ω(M ), δn )
and tn → +∞

ρ(xn tn , ω(M )) ≥ ε0 . (1.68)

Since xn ∈ B(ω(M ), δn ) and the set ω(M ) is compact, then we can suppose that the
sequence {xn } is convergent. By Theorem 1.36 the set H + (K) := ∪{π t K | t ≥ 0}
is compact. We note that along with the set M the set M 0 = M ∪ H + (K) also
is attracting for the family of compact subsets from W s (M ) and, consequently,
ω(M 0 ) = ω(M ). In particular, ω(H + (K)) ⊆ ω(M 0 ) = ω(M ). By the compactness
of the set H + (K) we my suppose that the sequence {xn tn } is convergent. Let
y := lim xn tn , then y ∈ ω(H + (K)) ⊆ ω(M ). But from the inequality (1.68)
n→+∞
follows that y ∈/ ω(M ). The obtained contradiction proves the first statement of
the theorem.
Let us prove now the second assertion. Let K1 be an arbitrary compact subset
from W s (M ). By Theorem 1.5 the set K1 is stable in the sense of Lagrange in
positive direction and the set ω(K1 ) 6= ∅, is compact and β(π t K1 , ω(K1 )) → 0 as
t → +∞. According to Lemma 1.3 the set ω(K1 ) is invariant, and by Theorem
1.37 we have ω(K1 ) ⊆ ω(M ) because ω(M ) is a maximal compact invariant set in
W s (M ).
We will show that J + (A) = ω(M ). Since A ⊆ ω(M ), then J + (A) ⊆ J + (ω(M )).
By the asymptotic stability of the set ω(M ) we have J + (ω(M )) ⊆ ω(M ) and,
consequently, J + (A) ⊆ ω(M ). Now we will prove the reverse inclusion. Let x ∈
ω(M ) and ϕ ∈ Φx , then ∅ 6= αϕx ⊂ A and, consequently, there are y ∈ αϕx and
tn → +∞ such that ϕ(−tn ) → y ∈ A and x = π tn ϕ(−tn ). Thus, x ∈ Jy+ ⊆ J + (A),
i.e. ω(M ) ⊆ J + (A). The theorem is completely proved. 
 Autonomous dynamical systems 51

Denote by {Mλ | λ ∈ Λ} a family of nonempty compact positively invariant sets

attracting all the compacts from W s (M ) and I := ∩{Mλ | λ ∈ Λ}.

Theorem 1.39 Let M be a nonempty compact positively invariant and asymptot-

ically stable set. Then I = ω(M ) (i.e. the set ω(M ) is the least compact positively
invariant set attracting all the compact sets from W s (M )).

Proof. First of all we will prove that ω(M ) ⊆ I and, consequently, I 6= ∅. In fact,
for all λ ∈ Λ we have ω(M ) = ω(Mλ ) ⊆ Mλ , i.e. ω(M ) ⊆ I. To finish the proof of
the theorem it is sufficient to show that I ⊆ ω(M ). Since the set ω(M ) is nonempty,
compact and attracts all the compacts from W s (M ), then ω(M ) ∈ {Mλ | λ ∈ Λ}
and, consequently, I ⊆ ω(M ). The theorem is proved. 
52 Global Attractors of Non-autonomous Dissipative Dynamical Systems
 Chapter 2

Non-autonomous dissipative dynamical

systems

2.1 On the stability of Levinson’s center

Let us consider a non-autonomous dynamical system

h(X, T1 , π), (Y, T2 , σ), hi. (2.1)

Denote by 2X a family of all bounded closed subsets from X, equipped with the
Hausdorff distance. Let K ⊆ X be a compact subset of X.

Definition 2.1 We will say that the set K satisfies the condition (C) if the
mapping F : Y → 2X defined by the equality

F (y) := Ky := {x ∈ K | h(x) = y}, (2.2)

is continuous.

Note, that the condition (C) means that the mapping h : X → Y is open on K.
This condition plays an important role in our reasoning; therefore we give a simple
test, which guarantees that on a compact invariant set we have this property.

Lemma 2.1 A nonempty compact invariant set K ⊆ X possesses the property

(C), if at least one of the following two conditions is fulfilled:

(1) there exist y0 ∈ Y and τ > 0 (τ ∈ T2 ) such that y0 τ = y0 and Y =

{σ(t, y0 ) | 0 ≤ t < τ }, i.e. τ is the least positive period for the point y0 ;
(2) the set Y is compact minimal one and the non-autonomous dynamical system
(2.1) is distal on the set K, i.e., for all different points x1 , x2 ∈ K such that
h(x1 ) = h(x2 ), we have

inf ρ(x1 t, x2 t) > 0. (2.3)

t∈T1

Proof. The first statement of the lemma is evident. The second statement follows
from the proposal 4 from [238, p.107]. 

53
54 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Definition 2.2 The non-autonomous dynamical system (2.1) is said to be point-

wise (compactly, locally, boundedly) dissipative if the autonomous dynamical
system (X, T1 , π) possesses this property.

Definition 2.3 Let the non-autonomous dynamical system (2.1) be compactly di-
issipative and the set J be the center of Levinson of the dynamical system (X, T1 , π).
The set J is said to be the Levinson’s center of the non-autonomous dynamical
system (2.1).

Above we have established that the Levinson’s center J is orbitally stable with
respect to the autonomous dynamical system (X, T1 , π). Considering the set J as
the Levinson’s center of the non-autonomous dynamical system (2.1), the orbital
stability with respect to the non-autonomous system (2.1) is more naturally defined
as follows:

Definition 2.4 The set K ⊆ X is said to be orbitally stable with respect to

the non-autonomous system (2.1) if for all ε > 0 there exists δ(ε) > 0 such that
ρ(x, Ky ) < δ (x ∈ X, y = h(x)) implies ρ(xt, Kyt ) < ε for all t ≥ 0. If in addition

a. there is γ > 0 such that ρ(xt, Kyt ) → 0 as t → +∞ for all x ∈ Ky with

the condition that ρ(x, Ky ) ≤ γ, then the set K is said to be asymptotically
orbitally stable;
b. if ρ(xt, Kyt ) → 0 as t → +∞ for all x ∈ Xy , then the set K ⊆ X is said to
be globally asymptotically stable with respect to the non-autonomous system
(2.1).

Definition 2.5 The set K ⊆ X is called [332] globally asymptotically stable in

the sense of Lyapunov-Barbashin if

lim ρ(xt, Kh(x)t ) = 0 (2.4)

t→+∞

for all x ∈ X, moreover the equality (2.4) holds uniformly with respect to the
variable x on compact subsets of X.

In the work [332] there it was shown that the Levinson’s center of a non-
autonomous dynamical system, generally speaking, is not globally asymptotically
stable in the sense of Lyapunov-Barbashin. We will establish below that the Levin-
son’s center J of the non-autonomous system (2.1) is globally asymptotically stable
if the set J possesses the property (C).
The following assertion may be made.

Lemma 2.2 Let K ⊆ X. If the set Σ+ (K) is relatively compact and the set
M (ω(K) ⊆ M, h(M ) = Y ) possesses the property (C). Then for each x ∈ K the
 Non-autonomous Dissipative Dynamical Systems 55

equality

lim sup ρ(xt, Myt ) = 0 (2.5)

t→+∞ x∈Ky

holds uniformly with respect to y ∈ h(H + (K)), where My := M ∩ h−1 (y).

Proof. If we suppose the contrary, then there are ε0 > 0, {yn } ⊆ h(H + (K)), xn ∈
Myn and tn → +∞ such that

ρ(xn tn , Myn tn ) ≥ ε0 . (2.6)

Since the set Σ+ (K) is relatively compact, then we may suppose that the sequences
{xn tn } and {yn tn } are convergent. Let x̄ := lim xn tn and ȳ := lim yn tn . We
n→+∞ n→+∞
note that

h(x̄) = lim h(xn tn ) = lim h(xn )tn = ȳ (2.7)

n→+∞ n→+∞

and, consequently, x̄ ∈ Xȳ . On the other hand, according to Lemma 1.2 x̄ ∈ ω(K).
Thus, x̄ ∈ ωȳ (K) ⊆ Mȳ .
Since the set M satisfies the condition (C), then passing to limit in the inequality
(2.6) as n → +∞, we obtain

ρ(x̄, Mȳ ) ≥ ε0 . (2.8)

The inequality (2.8) contradicts the inclusion x̄ ∈ Mȳ . The lemma is proved. 

Theorem 2.1 Let h(X, T1 , π), (Y, T2 , σ), hi be a compact dissipative non-
autonomous dynamical system and J be its Levinson center satisfying the condition
(C). If Y = h(J), then:

(1) the set J is orbitally stable in the positive direction with respect to the non-
autonomous dynamical system (2.1).
(2) the set J is a compact attractor of this system, i.e., the following equality

lim ρ(xt, Jh(x)t ) = 0, (2.9)

t→+∞

holds uniformly with respect to x on every compact subset from X.

Proof. We will show that the Levinson center J is orbitally stable in positive
direction with respect to the non-autonomous dynamical system (2.1). If it is not
true, then there exist ε0 > 0, δn ↓ 0, xn ∈ B(Jh(xn ) , δn ) and tn ≥ 0 such that

ρ(xn tn , Jh(xn )tn ) ≥ ε0 . (2.10)

Under the conditions of the theorem we may suppose that the sequences {xn } and
{h(xn )} are convergent. Let x0 := lim xn . Since the non-autonomous dynamical
n→+∞
56 Global Attractors of Non-autonomous Dissipative Dynamical Systems

system (2.1) is compactly dissipative, then the set H + ({xn }) is compact and, conse-
quently, we may suppose that the sequences {xn tn } and {h(xn )tn } are convergent.
Put x̄ := lim xn tn and ȳ := lim h(xn )tn .
n→+∞ n→+∞
We will show now that the sequence {tn } figuring in the equality (2.10) tends to
+∞. If we suppose that it is not true, then we may suppose that it is convergent.
Denote by t0 := lim tn and, passing to limit in the inequality (2.10) and taking
n→+∞
into consideration that the set J satisfies the condition (C), we obtain

ρ(x0 t0 , Jh(x0 )t0 ) ≥ ε0 . (2.11)

On the other hand, ρ(xn , Jh(xn ) ) < δn and, consequently, x0 ∈ Jh(x0 ) . Since the
set J is invariant, then from the last inclusion follows that x0 t0 ∈ Jh(x0 )t0 . But it
contradicts the inequality (2.11). Thus, tn → +∞ and, consequently, x̄ ∈ ω(K0 ) ⊆
J. From the equality (2.5) follows that x̄ ∈ Xȳ , i.e. x̄ ∈ ωȳ (K) ⊆ Jȳ . From the
inequality (2.6) it follows that x̄ ∈ Xȳ , i.e., x̄ ∈ ωȳ (K) ⊆ Jȳ .
Passing to limit in the inequality (2.10) as n → +∞ and taking into account
that x̄ ∈ Jȳ and the fact that the set J possesses the property (C), we obtain ε0 ≤ 0.
But this contradicts the choice of the number ε0 . Thus, the orbital stability of the
set J in the positive direction is proved.
Now we will prove the second statement of the theorem. Let K ⊆ X be an
arbitrary compact, then under the conditions of the theorem the set H + (K) is
compact and ω(K) ⊆ J. To finish the proof of the theorem it is sufficient to refer
to Lemma 2.2. 

Remark 2.1 We note, that the problem of the stability in the sense of Lyapunov-
Barbashin of the Levinson center of a non-autonomous dynamical system has been
studied (in one special case) in the work [332]. We also note that in the work [332]
stability means the equality (2.9), but not orbital stability. Using our terminology we
can formulate the results of the work [332] in the following way. Let X be a locally
compact space, the non-autonomous dynamical system (2.1) be pointwise dissipative
and Y be a compact minimal set. If the Levinson center J of the dynamical system
(X, T, π) satisfies the condition (C), then the non-autonomous dynamical system
h(X, T1 , π), (Y, T2 , σ), hi is compactly dissipative and its Levinson center attracts
every compact subset from X with respect to the non-autonomous dynamical system
(2.1).
This statement follows from Theorems 1.10 and 2.1. However, from the same
theorems follows that the Levinson center J is orbitally stable in the positive direc-
tion.

Theorem 2.2 Let h(X, T1 , π), (Y, T2 , σ), hi be a compact dissipative non-
autonomous dynamical system, its Levinson center J satisfy the condition (C) and
h(J) = Y , then the following conditions are equivalent:
 Non-autonomous Dissipative Dynamical Systems 57

1. the set J is globally asymptotically stable, i.e., the set J is orbitally stable and
the equality

lim ρ(xt, Jh(x)t ) = 0 (2.12)

t→+∞

holds for all x ∈ X;

2. the set J is globally asymptotically stable in the sense of Lyapunov-Barbashin.

Proof. We will show that under the conditions of the theorem 1. implies 2. In fact,
if it is not true, then there are ε0 > 0, K0 ∈ C(X), xn ∈ K0 and tn → +∞ such
that

ρ(xn tn , Jh(xn )tn ) ≥ ε0 .

Since the dynamical system h(X, T1 , π), (Y, T2 , σ), hi is compactly dissipative we
may suppose that the sequences {xn tn } and {h(xn )tn } are convergent. Let x̄ =
lim xn tn and ȳ = lim h(xn )tn , then x̄ ∈ Jȳ = J ∩ Xȳ . On the other hand,
n→+∞ n→+∞
passing to limit in the inequality (2.12) as n → +∞ and taking into account that
Jh(xn )tn → Jȳ in the Hausdorff metric, we obtain ρ(x̄, Jȳ ) ≥ ε0 , i.e. x̄ ∈
/ Jȳ . The
obtained contradiction completes the proof of our statement.
Now we will show that 2. implies 1. For this aim it is sufficient to prove that
from 2. follows the orbital stability of the set J. If we suppose that it is not so,
then there exist ε0 > 0, δn ↓ 0, xn ∈ B(J, δn ) and tn → +∞ such that

ρ(xn tn , Jh(xn )tn ) ≥ ε0 . (2.13)

Since the set J is compact, then the sequence {xn } is relatively compact and by
the condition 2. for the number ε0 > 0 and the compact K0 = {xn } there exists a
number L = L(ε0 , K0 ) > 0 such that
ε0
ρ(xn t, Jh(xn )t ) < (2.14)
2
for all t ≥ L. The inequalities (2.13) and (2.14) are contradictory. The obtained
contradiction completes the proof of the theorem. 

Remark 2.2 Note, that in the proof of the second part of the theorem 2.2 we did
not use the fact that the set J satisfies the condition (C). For us it is not clear how
important is the condition (C) in the first part of the theorem.

Below we give an example of the non-autonomous dynamical system with the

unstable (orbital) Levinson center.

Example 2.1 Consider Opial’s example [263] of the scale almost periodic differ-
ential equation

ṗ = f (t, p), (2.15)

58 Global Attractors of Non-autonomous Dissipative Dynamical Systems

which has all the solutions bounded, but does not admit an almost periodic solution.
We fix some minimal set M ⊂ X = R × H(f ) (H(f ) = {fτ | τ ∈ R}), where by bar
is denoted a closure in the space C(R × R, R) and let ϕ∗ and ϕ∗ : H(f ) → M be
the mappings defined by the equalities: ϕ∗ (g) := sup{p | (p, g) ∈ M } and ϕ∗ (g) :=
inf{p | (p, g) ∈ M } (g ∈ H(f )).
Recall that the equation (2.15) may be obtained in the following way. Let
ẏ = g(t, y) be the equation from the example of Danjua [263],[332] and the function
g be periodic with respect to each variable and realizing on the torus a non-ergodic
case (the function g can be chosen from the class C 1 on the torus), P be a perfect
nowhere dense limit set on R: t = 0 and α be the number of rotations. Denote by
p(t) = y(t) − αt, then we obtain

ṗ(t) = g(t, p(t) + αt) − α = f (t, p(t)).

Using the theory of the equations on the torus (see, for example, [270, Ch.2]), it
is possible to show that the solution p(t) of the equation (2.15) is recurrent if and
only if when p(0) ∈ P .
We need to change now the function f (x) = f (p, h) outside of the set M so that
the new system would be dissipative and the function p∗ (g) (p∗ (g) := sup{p | (p, g) ∈
J}) would be discontinuous.
We put F (x) = f (x) − ρ(x, M ); then (conserving the continuity) we will change
this function for the negative numbers p large enough by the module such that
F (p, h) ≥ 1 (h ∈ H). It is clear that the non-autonomous dynamical system
generated by the equation ṗ = F (p, σt h) is dissipative (σt h is an irrational hull of
the torus).
Suppose that the function p∗ (h) is continuous. Since p∗ (h) ≥ ϕ∗ (h) and the
solutions p∗ (σt h) are almost periodic, then inf{p∗ (h) − ϕ∗ (h) | h ∈ H(f )} > 0.
From the definition of the function F (x) follows the inequality

F (p∗ (h), h) ≤ f (p∗ (h), h) − c1 (c1 > 0, h ∈ H). (2.16)

We put p(t) = p∗ (σt h) and let q(t) be the solution of the equation (2.15) with the
initial condition q(0) = p(0). From the relation (2.16) we obtain the inequalities

inf (q(t) − p(t)) > 0, inf (p(t) − q(t)) > 0. (2.17)

t≥1 t≤−1

From the inequalities (2.17), it follows that the function q(t) is not recurrent, i.e.
q(0) ∈/ P . Let q1 , q2 (q1 > q2 ) be the ends of the corresponding adjoining interval
and q1 (t), q2 (t) be the solution of the equation (2.15) with initial conditions q1 (0) =
q1 , q2 (0) = q2 . Since the functions q1 (t), p(t) are simultaneously recurrent, then
there exists a sequence tm → +∞ such that q1 (t + tm ) → q1 (t), p(t + tm ) → p(t)
for all t ∈ R. Then from the inequality (2.17) we obtain the inequality inf{q1 (t) −
 Non-autonomous Dissipative Dynamical Systems 59

p(t) | t ∈ R} > 0. In the same way we can establish the inequality

inf{p(t) − q2 (t) | t ∈ R} > 0.

But this inequality contradicts the relation lim (q1 (t) − q2 (t)) = 0 (see [263]). The
t→+∞
example needed is constructed.

Note that the example given above belongs to V. V. Zhikov [332]. An analogous
example was published also in the work [152].

Lemma 2.3 Let M ⊆ X be a nonempty compact set and the mapping F : Y →

2M (F (y) = My ) be continuous in the metric of Hausdorff. Then for all δ > 0 there
is γ = γ(δ) > 0 such that the following inclusion
[
e
B(M, γ) ⊂ B(M, δ) = By (M, δ) (2.18)
y∈Y

holds, where By (M, δ) = {x | x ∈ Xy , ρ(x, M ) < δ}.

Proof. Suppose that the lemma is not true, then there are δ0 > 0, γn ↓ 0 and
xn ∈ B(M, γn ) such that xn ∈ e
/ B(M, δ), i.e.

ρ(xn , Myn ) ≥ δ0 (2.19)

for all n, where yn = h(xn ). Since the set M is compact and the mapping F :
Y → 2M is continuous, then we may suppose that xn → x0 and Myn → My0 where
y0 = h(x0 ). Passing to limit in the inequality (2.19) as n → +∞, we obtain

x0 ∈
/ B(My0 , δ0 ). (2.20)

On the other hand, xn ∈ B(M, γn ) and since γn → 0, then x0 ∈ M and, conse-

quently, x0 ∈ My0 . This inclusion contradicts the relation (2.20). The obtained
contradiction completes the proof of the lemma. 

Lemma 2.4 Let M ⊆ X be a nonempty compact orbitally asymptotically stable

with respect to the non-autonomous dynamical system (2.1) invariant set, then it is
orbitally asymptotically stable also with respect to the autonomous dynamical system
(X, T1 , π).

Proof. This statement follows directly from Lemma 2.3 and the corresponding
definitions. 

Theorem 2.3 Under the conditions of Lemma 2.3, if the set M is or-
bitally asymptotically stable with respect to the dynamical system (X, T 1 , π), then
the set M is orbitally asymptotically stable with respect to dynamical system
h(X, T1 , π), (Y, T2 , σ), hi too.
60 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Proof. We will show that the set M is orbitally stable with respect to
h(X, T1 , π), (Y, T2 , σ), hi. If we suppose that it is false, then there exist ε0 > 0, δn ↓
0, xn ∈ B(Myn , δn ) (yn = h(xn )) and tn → +∞ such that

ρ(xn tn , Myntn ) ≥ ε0 . (2.21)

According to Theorem 1.36 we may suppose that the sequence {xn tn } is convergent.
Let x̄ := lim xn tn , x0 := lim xn , y0 := h(x0 ) and ȳ := lim yn tn . Note, that
n→+∞ n→+∞ n→+∞
x̄ ∈ Jȳ+ . According to Theorem 1.36 x̄ ∈ Jȳ+ ⊆ M and, consequently, x̄ ∈ Mȳ . On
the other hand, passing to limit in the inequality (2.21) as n → +∞, we obtain
x̄ ∈
/ Mȳ . The obtained contradiction shows that the set M is orbitally stable with
respect to the non-autonomous dynamical system h(X, T1 , π), (Y, T2 , σ), hi.
Let now x ∈ W s (M ), then lim ρ(xt, M ) = 0. By Lemma 2.2 the following
t→+∞
equality

lim ρ(xt, Mh(x)t ) = 0 (2.22)

t→+∞
S
holds and, consequently, W s (M ) = Wys (M ). The theorem is proved.
y∈Y 

2.2 The positively stable systems

Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical system. Denote by

C(X, Y ; h) the set of all functions ξ : X → X, satisfying the following conditions:


(1) ξ Xh(x) ⊆ Xh(ξ(x)) for all x ∈ X,
(2) the mapping ξ : Xh(x) −→ Xh(g(x)) is continuous on Xh(x) for each x ∈ X.

We define the topology on C(X, Y ; h) by a family of pseudo-metrics

ρy,K (f, g) := max ρ(f (x), g(x)) (2.23)

x∈Ky

where K is an arbitrary nonempty compact from X and y ∈ Y . Note that the

family of pseudo-metrics (2.23) defines on C(X, Y ; h) a topology of the convergence
uniform (on every fiber) on the compact subsets from X.

Lemma 2.5 The space C(X, Y ; h) is a topological semigroup with respect to com-
position (on the fibers).

Proof. To prove this statement it is sufficient to verify the continuity of the mapping

F : C(X, Y ; h) × C(X, Y ; h) → C(X, Y ; h),

defined by the equality F (f, g) := f ◦ g. If we suppose that this statement is not

true, then there is a point (f0 , g0 ) such that the mapping F in this point is not
 Non-autonomous Dissipative Dynamical Systems 61

continuous, i.e., there is a direction (fλ , gλ ) → (f0 , g0 ) such that fλ ◦ gλ 9 f0 ◦ g0 .

Then there are ε0 > 0, a compact K 0 ⊆ X and y0 ∈ Y such that

max ρ(fλ (gλ (x)), f0 (g0 (x))) ≥ ε0 . (2.24)

x∈Ky00

From the inequality (2.24) follows the existence {xλ } ⊆ Ky00 such that

ρ(fλ (gλ (xλ )), f0 (g0 (xλ ))) ≥ ε0 . (2.25)

Since the set K 0 is compact, then we may suppose that the direction {xλ } is con-
vergent. Let xλ → x0 , then g0 (xλ ) → g0 (x0 ). We will show that gλ (xλ ) → g0 (x0 ).
In fact,

ρ(gλ (xλ ), g0 (x0 )) ≤ max0 ρ(gλ (x), g0 (x)) + ρ(g0 (xλ ), g0 (x0 )) (2.26)
x∈Ky0

and passing to the limit in (2.26), we obtain the unknown affirmation. Analogously
can be proved that fλ (gλ (xλ )) → f0 (g0 (x0 )). Passing to the limit in (2.25) we
obtain ε0 ≤ 0, that contradicts the choice of the number ε0 . The lemma is proved.
Let α > 0 and Tα := {t ∈ T : t ≥ α}. We put P α := P α (X, Y ; h) :=
{π t : t ∈ Tα }, where by the bar we note the closure in C(X, Y ; h).

Lemma 2.6 The set P α is a closed subsemigroup of the semigroup C(X, Y ; h).

Proof. The formulated statement immediately follows from the lemma 2.5 and
from the definition of the set P α . 

Definition 2.6 The non-autonomous dynamical system (2.1) is said to be uniform

stable in the positive direction on the compact subsets from X, if for all ε > 0 and
compact K ⊆ X there is δ = δ(ε, K) > 0 such that for all x1 , x2 ∈ K with condition
h(x1 ) = h(x2 ), from the inequality ρ(x1 , x2 ) < δ follows ρ(x1 t, x2 t) < ε for all t ≥ 0.

Let Py := {ξ ∈ P α : ξXy ⊆ Xy }. It easy to check that Py is a closed subsemi-

group of the semigroup P α .
The following statement holds.

Lemma 2.7 If the dynamical system (2.1) is compact dissipative and uniform
stable in the positive direction on the compact subsets from X, then P α is a compact
semigroup and Py is its nonempty subsemigroup for all Poisson stable, in the positive
direction point y ∈ Y .

Proof. At first we will prove that under the conditions of the lemma the semigroup
P α is compact. Let K ⊆ X be an arbitrary compact. By the compact dissipativity
of the non-autonomous dynamical system (2.1) the set Σ+ (K) is compact. In view
62 Global Attractors of Non-autonomous Dissipative Dynamical Systems

of the uniform stability in the positive direction of the system (2.1) the family of
the mappings {π t : t ∈ Tα } ⊆ C(X, Y ; h) are equicontinuous on Ky (y ∈ Y ). Then
according to the theorem of Arzela-Ascoli, taking into account the topology on the
set C(X, Y ; h), we conclude that {π t : t ∈ Tα } is relatively compact in C(X, Y ; h)
and, consequently, its closure is a compact.
We will show that that the set Py is nonempty. Let tn → +∞ such that ytn → y.
Consider the sequence {π tn } ⊆ P α . By virtue of the compactness of the set P α we
can suppose that the sequence {π tn } is convergent in the space C(X, Y ; h). We put
ξ := lim π tn and we will show that ξ ∈ Py . To this end, it is sufficient to prove
n→+∞
that ξXy ⊆ Xy . Let x ∈ Xy , then ξ(x) = lim xtn and, consequently,
n→+∞

h(ξ(x)) = h( lim xtn ) = lim h(xtn ) = lim h(x)tn = lim ytn = y.

n→+∞ n→+∞ n→+∞ n→+∞

The lemma is proved. 

Lemma 2.8 Let Y be a compact minimal set and the non-autonomous dynamical
system (2.1) satisfies the conditions of the lemma 2.7. Then the Levinson’s center
J of the system (2.1) is bilateral distal, i.e. for all x1 and x2 from J (x1 6= x2 ,
h(x1 ) = h(x2 )) the following inequality inf{ρ(ϕ1 (s), ϕ2 (s)) : s ∈ S} > 0 holds, where
ϕi is an extension on the S of motion π(·, xi ) (i = 1, 2).

Proof. According to the theorem 1.6 all the motions on the set J are extendable in
the negative direction. From the uniform stability in the positive direction of the
set J it follows its distality in the negative direction. Since the set Y is minimal,
then by the lemma 22.4 [32] (see also [31] and the lemma 1 from [238, p.104]) the
set J is bilateral distal. 

Theorem 2.4 Suppose that a non-autonomous dynamical system h(X, T1 , π),

(Y, T2 , σ), hi is compact dissipative, uniform stable in the positive direction on the
compact subsets from X and the set Y is minimal, then:

(1) all the motions on the Levinson’s center J of the system (2.1) can be extended
in the negative direction and the set J is bilateral distal;
(2) the set J consists of recurrent trajectories and every pair of points x 1 , x2 ∈
Jy := Xy ∩ J (y ∈ Y ) is mutually recurrent;
(3) if the fibers Xy are connected, then for each y ∈ Y the set Jy is connected and
for the distinct points y1 and y2 the sets Jy1 and Jy2 are homeomorphic;
(4) the set J is orbital stable with respect to (2.1), i.e. for all ε > 0 there is δ(ε) > 0
such that the inequality ρ(x, Jh(x) ) < δ implies ρ(xt, Jh(x) t) < ε for all t ≥ 0;
(5) the equality lim ρ(xt, Jh(x)t ) = 0 holds for each x ∈ X.
t→+∞

Proof. The first statement of theorem coincides with the lemma 2.8.
 Non-autonomous Dissipative Dynamical Systems 63

The second statement of theorem follows from the first one and from the lemma
1 from [238, p.104].
We will prove the third statement of the theorem. Let y ∈ Y , then under the
conditions of the theorem the point y is stable in the sense of Poisson and by the
lemma 2.7 Py is a nonempty compact topological subsemigroup of the semigroup
P α . According to the lemma 4.11 [32] in the semigroup Py there exists at least
one idempotent, i.e., an element u ∈ Py , such that u ◦ u = u. We wills show that
for idempotent u there is a sequence tn → +∞ such that u = lim π tn . In fact
n→+∞
since u ∈ Py ⊆ P α , then there is a sequence tn ≥ α such that u = lim π tn . If
n→+∞
the sequence {tn } is unbounded, then from this sequence it is possible to extract
a subsequence which converges to +∞ and our statement is proved. Suppose now
that the sequence {tn } is bounded, then we can suppose that it is convergent. Let
0
t0 := lim tn (t0 ≥ α) and t0n := nt0 . It easy to see that as t0n → +∞, π tn → u.
n→+∞
Thus in the semigroup Py there is an idempotent u and tn → +∞ such that
u = lim π tn . From the first two assertions of the theorem 2.4 it follows that
n→+∞
u(Xy ) ⊆ Jy . We will show that in fact we have the equality Jy = u(Xy ). To this
end it is sufficient to show that Jy = u(Jy ). But this equality follows from the
bilateral distality on the set J and, consequently, every idempotent from Py on the
set Jy acts as a identity mapping. The connectedness of the set Jy follows from the
continuity of the mapping u, the connectedness of the fiber Xy and the equality
Jy = u(Xy ).
We will show that the second part of the third statement of the theorem is true.
Let y1 and y2 ∈ Y . Since the set Y is minimal, then there is {tn } ⊆ T such that
y2 = lim y1 tn . Consider the sequence {π tn } on Jy1 . Since π tn : Jy1 → J and the
n→+∞
set J is bilateral uniform stable [238, p.107], then the family of mappings {π tn } is
equicontinuous and, consequently, we may suppose that it is uniform convergent.
Let ξ := lim π tn . From the third statement of the theorem and from the proposal
n→+∞
4 from [238, p.107] it follows that ξ(Jy1 ) = Jy2 . Thus, we will show that the
continuous mapping ξ acts from Jy1 on Jy2 . From the bilateral distality of the set
J it follows that ξx1 6= ξx2 , if x1 6= x2 (x1 , x2 ∈ Jy1 ). This means that the mapping
ξ is a homeomorphism from Jy1 onto Jy2 .
The fourth assertion of the theorem it follows from the uniform stability in the
positive direction of the system (2.1) on the compact subsets from X. If we suppose
that it is false, then there are ε0 > 0, δn ↓ 0, xn ∈ X and tn → +∞ such that

ρ(xn , Jh(xn ) ) < δn and ρ(xn tn , Jh(xn )tn ) ≥ ε0 . (2.27)

From the relation (2.27) follows that the set K0 = {xn } is compact. For the number
ε0 ε0


2 and the compact K = K0 ∪J we will choose δ = δ 2 , K > 0 from the condition
of the uniform stability in the positive direction of the system (2.1) on the compact
64 Global Attractors of Non-autonomous Dissipative Dynamical Systems

subsets from X. Let xn ∈ Jh(xn ) be a sequence such that ρ(xn , Jh(xn ) ) = ρ(xn , xn ).
Then for sufficiently large n we have ρ(xn , xn ) < δ and, consequently,
ε0
ρ(xn tn , Jh(xn )tn ) ≤ ρ(xn tn , xn tn ) < . (2.28)
2
But the inequalities (2.28) and (2.27) are contradictory. The obtained contradiction
proves our assertion.
Now we will prove the fifth statement of the theorem, i.e., if x ∈ X, then
lim ρ(xt, Jh(x)t ) = 0. Suppose that it is not true, then there are x0 ∈ X, ε0 > 0
t→+∞
and tn → +∞ such that

ρ(x0 tn , Jh(x0 )tn ) ≥ ε0 . (2.29)

Under the conditions of the theorem we may suppose that the sequence {π tn } is
convergent in the space C(X, Y ; h). Let ξ = lim π tn , x̄ = ξ(x0 ) and ȳ = ξ(y0 ).
n→+∞
Passing to the limit in the inequality (2.29) and taking in the consideration that
ξ(Jh(x0 ) ) = Jh(ξ(x0 )) , we obtain

ρ(x, Jy ) ≥ ε0 . (2.30)

On the other hand x ∈ ωx and x ∈ Xy and, consequently, x ∈ Jy . This inclusion

contradicts the inequality (2.30). The obtained contradiction completes the proof
of the theorem. 

2.3 Behaviour of dissipative dynamical systems under homomorphisms

In this paragraph we determine the conditions under which the properties of the
point, compact and local dissipativity are preserved under homomorphisms.
Let X and Y be complete metric spaces, (X, T1 , π) and (Y, T2 , σ) be dynamical
systems on X and Y correspondingly, ΩX (ΩY ) is closure of the set of all ω-limit
points of dynamical system h(X, T1 , π), (Y, T2 , σ), hi and JX (JY ) is its Levinson’s
center.

Lemma 2.9 Let h : X → Y be a homomorphism of dynamical system (X, T1 , π)

onto (Y, T2 , σ), then:

(1) h(ΩX ) ⊆ ΩY ;
(2) if the set M is compact, then h(D + (M )) ⊆ D+ (h(M )) and h(J + (M )) ⊆
J + (h(M ));
(3) if (X, T1 , π) is point dissipative, then h(D + (ΩX )) ⊆ D+ (ΩY ) and h(J + (ΩX )) ⊆
J + (ΩY ).
 Non-autonomous Dissipative Dynamical Systems 65

Proof. Let x ∈ ΩX , then there exists xn ∈ ωx̃n (x̃n ∈ X) such that x = lim xn .
n→+∞
Since h(ωx̃n ) ⊆ ωh(x̃n) , h(xn ) ∈ ωh(x̃n ) and, consequently, h(x) = lim h(xn ) ∈
n→+∞
Ωh(X) ⊆ ΩY .
Let us prove the second statement of Lemma. Let x ∈ ΩX , then according to
Lemma 1.14 there exists x̃ ∈ M such that x ∈ Dx̃+ and, consequently, there exist
xn → x̃ and tn ≥ 0 such that x = lim xn tn . Note that h(xn ) → h(x̃) ∈ h(M )
n→+∞
and h(x) = lim h(xn )tn ∈ D+ (h(M )). And analogously we establish the second
n→+∞
inclusion h(J + (M )) ⊆ J + (h(M )).
Finally, note that the first and the second statements of Lemma imply the third.


Theorem 2.5 The following statements hold:

(1) if h is a homomorphism (X, T1 , π) onto (Y, T2 , σ) and dynamical system

(X, T1 , π) is point dissipative, then (Y, T2 , σ) also is point dissipative and
h(ΩX ) = ΩY ;
(2) if h is a homomorphism (X, T1 , π) onto (Y, T2 , σ) and dynamical system
(X, T1 , π) is compact dissipative, then (Y, T2 , σ) also is compact dissipative and
h(JX ) = JY ;
(3) if dynamical system (X, T1 , π) is locally dissipative and h is an open homo-
morphism (X, T1 , π) onto (Y, T2 , σ), then (Y, T2 , σ) also is locally dissipative.

Proof. As Y = h(X) and all positive semi-trajectories of system (X, T2 , σ) are

relatively compact, positive semi-trajectories of system (Y, T2 , π) also are relatively
compact. That is why for the point dissipativity it is sufficient to show that ΩY =
h(ΩX ). Let y ∈ ΩY , then there exist {yn } and {ỹn } such that yn ∈ ωỹn and
lim yn = y. Since Y = h(X), there exists x̃n ∈ X such that ỹn = h(x̃n ) and,
n→+∞
consequently, h(ωx̃n ) = ωỹn . So, there is xn ∈ ωx̃n ⊆ ΩX for which h(xn ) = yn . As
(X, T1 , π) is point dissipative, the set ΩX is compact and, consequently, the sequence
{xn } can be considered convergent. Assume that x := lim xn , then x ∈ ΩX and
n→+∞
h(x) = y. Thus, ΩY ⊆ h(ΩX ). To finish the proof of the first statement of the
theorem it is sufficient to refer to Lemma 2.9.
Now let us prove the second statement. If h is a homomorphism of compact
dissipative dynamical system (X, T1 , π) onto (Y, T2 , σ), then according to the first
statement of the theorem the system (Y, T2 , σ) is point dissipative. According to
Theorem 1.15 to prove the compact dissipativity of (Y, T2 , σ) it is sufficient to show,
that for every non-empty compact N ⊂ Y the set Σ+ N = {σ(t, y)|t ≥ 0, y ∈ N } is
+
relatively compact. Let {ỹn } ⊂ ΣN be an arbitrary sequence. Then there exist
{yn } ⊆ N and {tn } ⊂ T2 (tn ≥ 0) such that ỹn = σ(tn , yn ). If {tn } is bounded,
then the sequence {ỹn } is relatively compact; so without loss of generality we can
consider that tn → +∞ and yn → y ∈ Y . As (X, h, Y ) is locally trivial, then there
66 Global Attractors of Non-autonomous Dissipative Dynamical Systems

exists a sequence {xn } ⊆ X such that xn → x and h(xn ) = yn . By virtue of compact

dissipativity of (X, T1 , π) the sequence {xn tn } can be considered convergent, and,
consequently, {h(xn tn )} = {h(xn )tn } = {yn tn } also is convergent. Thus, (Y, T2 , σ)
is compact dissipative. Let us show that h(JX ) = JY . According to Lemma 2.9
h(J + (ΩX )) ⊆ J + (ΩY ). We will show that the reverse inclusion holds J + (ΩY ) ⊆
h(J + (ΩX )). Let q ∈ J + (ΩY ), then it will exist y ∈ ΩY such that q ∈ Jy+ . According
to the first statement of Theorem 2.5 we have h(ΩX ) = ΩY and, consequently,
there exists x ∈ ΩX such that h(x) = y. Let yn → y and tn → +∞ are such
that yn tn → q. Since (X, T1 , π) is locally trivial, then there exists {xn } → x such
that h(xn ) = yn . As (X, T1 , π) is compact dissipative, the sequence {xn tn } can be
considered convergent. Assume that p = lim xn tn , then p ∈ Jx+ ⊆ J + (ΩX ) and,
n→+∞
consequently, h(p) = lim h(xn tn ) = lim yn tn = q, i.e J + (ΩY ) ⊆ h(J + (ΩX )).
n→+∞ n→+∞
Thus, h(J + (ΩX )) = J + (ΩY ), and to finish the proof of the second statement of the
theorem it is sufficient to note that according to Theorem 1.11 JX = J + (ΩX ) and
JY = J + (ΩY ).
To prove the third statement of the theorem let us note that according to
the second statement the dynamical system (Y, T2 , σ) is compact dissipative and
h(JX ) = JY . By Theorem 1.18 to demonstrate the local dissipativity of (Y, T2 , σ)
it is sufficient to show that JY is a uniformly attracting set. First of all note that
by virtue of the compactness of JX and the equality h(JX ) = JY for any ε > 0
there is δ(ε) > 0 such that

ρ(h(x), JY ) < ε (2.31)

for all x ∈ B(JX , δ). In fact, if we suppose that is not true, then there exist
ε0 > 0, δn ↓ 0 and xn ∈ B(JX , δn ) such that

ρ(h(xn ), JY ) < ε. (2.32)

The sequence {xn } can be considered convergent. Assuming x0 = lim xn and

n→+∞
going over to the limit in the inequality (2.32), as n → +∞, we have

ρ(h(x0 ), JY ) ≥ εx ,

¯ JY and x0 ∈ JX . It contradicts the equality h(JX ) = JY . So, let ε > 0

i.e., h(x0 )∈
and δ(ε) > 0 are such that the equality (2.31) is fulfilled. By virtue of locally
dissipativity of (X, T1 , π) it follows from theorem 1.18 that there exists γ > 0 such
that the equality

lim β(π t B(JX , γ), JX ) = 0 (2.33)

is fulfilled. As the homomorphism h is open, the set V = h(B(JX , γ)) ⊃ JY is open.

Let α > 0 be such that B(JY , α) ⊂ V . From the equality (2.33) it follows that for
 Non-autonomous Dissipative Dynamical Systems 67

δ(ε) > 0 there exists L(ε) = L(δ(ε)) > 0 such that

β(π t B(JX , γ), JX ) < δ(ε) (2.34)

for all t ≥ L(ε). The inequalities (2.31) and (2.34) imply

β(σ t B(JY , α), JY ) < ε

for all t ≥ L(ε). The theorem is completely proved. 

Corollary 2.1 Let h(X, T1 , π), (Y, T2 , σ), hi be a compact dissipative non-
autonomous dynamical system and Y be a compact invariant set (i.e., σ t Y = Y
for all t ∈ T2 ), then:
T
(1) h(JX ) = Y and, consequently, Jy = JX Xy 6= ∅ for all y ∈ Y , where Xy =
h−1 (y);
(2) Jy = {x ∈ Xy | there exists whole relatively compact trajectory ϕx of the
dynamical system (X, T1 , π) such that ϕ(0) = x}.

Remark 2.3 We don’t know whether or not the condition of the openness of h is
so essential in Theorem 2.5, but apparently we can’t completely reject it.

Definition 2.7 Recall that ϕ : Y → X is called a continuous section of the

homomorphism h : X → Y , if ϕ is continuous and h ◦ ϕ = idY .

Definition 2.8 A continuous section is called invariant if ϕ ◦ σ t = π t ◦ ϕ for all

t ∈ T1 .

Theorem 2.6 Let the homomorphism h : X → Y admits a continuous invariant

section, then:

(1) if (X, T1 , π) is point dissipative, then (Y, T2 , σ) also is point dissipative,

h(ΩX ) = ΩY and h(D+ (ΩX )) = D+ (ΩY );
(2) if (X, T1 , π) is compact dissipative, then (Y, T2 , σ) also is compact dissipative
and h(JX ) = Jy ;
(3) if (X, T1 , π) is locally dissipative, then (Y, T2 , σ) also is locally dissipative.

Proof. Let ϕ be a continuous invariant section of h. Since h ◦ ϕ = idY , then

h(X) = Y and the first and the second statements follow from Theorem 2.5, except
the equality h(D + (ΩX )) = D+ (ΩY ). To prove it note that by virtue of compactness
ΩX from Lemma 2.9 the inclusion h(D + (ΩX )) ⊆ D+ (ωY ) holds.
Let us show that within the conditions of Theorem 2.6 the reverse inclusion also
holds. In fact, according to Theorem 2.5 the dynamical system (Y, T2 , σ) is point
dissipative and ΩY = h(ΩX ) is compact. Since ϕ : Y → X is a homomorphism
68 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(Y, T2 , σ) onto (X, T1 , π), then according to Lemma 2.9 ϕ(D + (ΩY )) ⊆ D+ (ΩX )
and, consequently, D + (ΩY ) = h ◦ ϕ(D+ (ΩX )) ⊆ h(D+ (ωX )).
Let (X, T1 , π) be locally dissipative, then from the foregoing proof the dynamical
system (Y, T2 , σ) is compact dissipative. According to Theorem 1.18 for the locally
dissipativity of (Y, T2 , σ) it is sufficient to show that its Levinson’s center JY is a
uniformly attracting set. As well as in Theorem 2.5, for ε > 0 (η > 0) we will select
δ(ε) > 0 (ξ(η) > 0) such that

ρ(h(x), JY ) < ε (ρ(ϕ(y), JX ) < η) (2.35)

for all x ∈ B(JX , δ) (y ∈ B(JY , ξ)). Since X, T1 , π) is locally dissipative then

there exists γ > 0 such that

lim β(π t B(JX , γ), JX ) = 0. (2.36)

t→+∞

Assume that ν = ξ(γ) > 0 and let us show that

lim β(σ t B(JY , ν), JY ) = 0.

t→+∞

Let ε > 0, then according to (2.36) there exists L(ε) > 0 such that

π t B(JX , γ) ⊆ B(JX , δ) (2.37)

for all t ≥ L(ε). By virtue of the choice of ν = ξ(γ) > 0 we have

ϕ(B(JX , γ)) ⊆ B(JX , γ). (2.38)

From the inclusions (2.37) and (2.38) it follows that

π t ϕ(B(JX , γ)) ⊆ B(JX , δ) (2.39)

for all t ≥ L(ε). From (2.35) and (2.39) we obtain

h(π t ϕ(B(JY , ν))) ⊂ B(JY , ε) (2.40)

for all t ≥ L(ε). Since h ◦ π t ◦ ϕ = σ t ◦ h ◦ ϕ = σ t (h ◦ ϕ = idY ) for all t ∈ T1 , then

from (2.40) we have

σ t B(JY , ν) ⊂ B(JY , ε)

for all t ≥ L(ε). The theorem is completely proved. 

2.4 Non-autonomous dynamical systems with convergence

Let (X, ρ) and (Y, d) be two complete metric spaces, R(Z) be a group of real (integer)
numbers, S = R or Z, S+ = {t ∈ S | t ≥ 0} and T(S+ ⊆ T) be a subgroup of group
S.
 Non-autonomous Dissipative Dynamical Systems 69

By (X, T, π) we denote a dynamical system on X and xt = π(t, x) = π t x.

Recall, that a dynamical system (X, T, π) is called compactly dissipative if there
exists a nonempty compact K ⊆ X such that

lim ρ(xt, K) = 0 (2.41)

t→+∞

for all x ∈ X; moreover equality (2.41) holds uniformly with respect to x ∈ X on

each compact subset from X. In this case the set K is called an attractor of the
family of all compact subsets C(X) from the space X.
We denote by
\ [
J = Ω(K) = π τ K,
t≥0 τ ≥t

then the set J does not depend of the choice of the attractor K and is characterized
by the properties of the dynamical system (X, T, π) . The set J is called a Levinson’s
center of the dynamical system (X, T, π).
Let us mention some facts that we will use below.
Let Y be a compact metric space, (X, T1 , π) ((Y, T2 , σ)) be a dynamical system
on X (Y ), (T1 ⊆ T2 ) and h : X → Y be a homomorphism of (X, T1 , π) onto
(Y, T2 , σ). Then the triple h(X, T1 , π), (Y, T2 , σ), hi is called a non-autonomous
dynamical system.
Let W and Y be complete metric spaces, (Y, S, σ) be a group dynamical system
on Y and hW, ϕ, (Y, S, σ)i be a cocycle over (Y, S, σ) with the fiber W (or, by
short, ϕ), i.e. ϕ is a continuous mapping of W × Y × T into W satisfying the
following conditions: ϕ(0, w, y) = w and ϕ(t + τ, w, y) = ϕ(t, ϕ(τ, w, y), σ(τ, y)) for
all t, τ ∈ T, w ∈ W and y ∈ Y .
We denote X = W × Y and define on X a skew product dynamical system
(X, T, π) by the equality π = (ϕ, σ), i.e. π(t, (w, y)) = (ϕ(t, w, y), σ(t, y)) for all
t ∈ T and (w, y) ∈ W × Y . Then the triple h(X, T, π), ((Y, S, σ), hi, where h = pr2 ,
is a non-autonomous dynamical system.
For any two bounded subsets A and B from X we denote by β(A, B) a semi-
deviation of A to B, i.e. β(A, B) = sup{ρ(a, B)|a ∈ A} and ρ(a, B) = inf{ρ(a, b)|b ∈
B}.

Definition 2.9 A cocycle ϕ over (Y, S, σ) with the fiber W is called compactly
dissipative if there exists a nonempty compact K ⊆ W such that

lim sup{β(U (t, y)M, K) : y ∈ Y } = 0

t→+∞

for all M ∈ C(W ) where U (t, y) = ϕ(t, ·, y).

Definition 2.10 By a whole trajectory of the semigroup dynamical system

(X, T, π) (of the cocycle hW, ϕ, (Y, S, σ)i over (Y, T, σ) with the fiber W ) pass-
70 Global Attractors of Non-autonomous Dissipative Dynamical Systems

ing through the point x ∈ X ( (u, y) ∈ W × Y ) we mean a continuous mapping

γ : S → X(ν : S → W ) satisfying the conditions : γ(0) = x(ν(0) = w) and
γ(t + τ ) = π t γ(τ ) (ν(t + τ ) = ϕ(t, ν(τ ), yτ )) for all t ∈ T and τ ∈ S.

Definition 2.11 h(X,T1 ,π),(Y,T2 ,σ),hi is said to be convergent if the following

conditions are valid:

(1) the dynamical systems (X, T1 , π) and (Y, T2 , σ) are compactly dissipative;
T
(2) the set JX Xy contains no more than one point for all y ∈ JY where Xy :=
h−1 (y) := {x|x ∈ X, h(x) = y} and JX (JY ) is the Levinson’s center of the
dynamical system (X, T1 , π)((Y, T2 , σ)).

˙
Let M ⊆ X and M ×M := {(x1 , x2 ) | x1 , x2 ∈ M, h(x1 ) = h(x2 )}.

Lemma 2.10 Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical

system, K ⊆ X be a compact invariant set and M := h(K). If the equality

lim sup ρ(x1 t, x2 t) = 0 (2.42)

t→+∞ ˙
(x1 ,x2 )∈K ×K
T
holds, then the set Ky := K Xy contains a single point for all y ∈ M .

Proof. Suppose that there exists y0 ∈ M such that Ky0 contains at least two points
x̄1 and x̄2 (x̄1 6= x̄2 ). Since the set K is invariant, then there exists a trajectory ϕi
passing trough the point x̄i (i = 1, 2) such that ϕi (S) ⊆ K. Let 0 < ε < ρ(x̄12,x̄2 ) and
L(ε) > 0, so that

ρ(x1 t, x2 t) < ε

˙
for all t ≥ L(ε) and (x1 , x2 ) ∈ K ×K. Thus, we have

ρ(x̄1 , x̄2 ) = ρ(π t ϕ1 (−t), π t ϕ2 (−t)) < ε

for all t ≥ L(ε). The obtained contradiction shows that Ky contains a single point
for all y ∈ M. The lemma is proved. 

Definition 2.12 A dynamical system (X, T, π) is said to be satisfying the con-

S
dition (A) if the set {π t K | t ≥ 0} is relatively compact for every K ∈ C(X) =
{K | K ⊆ X and K is compact }.

We denote by LY := {x | x ∈ X is so that at least one entire trajectory of the

dynamical system (X, T, π) passes through x}.

Remark 2.4 For a compactly dissipative system (X, T, π) we have LX = JX

where JX is the Levinson’s center of (X, T, π) .
 Non-autonomous Dissipative Dynamical Systems 71

Theorem 2.7 Let (X, T1 , π) be a dynamical system satisfying the condition (A)
and let (Y, T2 , σ) be compactly dissipative, then the following conditions are equiva-
lent:
T
1. the set LX Xy contains no more than one point for all y ∈ JY ;
2. every semi-trajectory Σ+x = {xt | t ≥ 0} is asymptotically stable, i.e.

2.a. for all ε > 0 and p ∈ X there exists δ(ε, p) > 0 such that ρ(x, p) <
δ (h(x) = h(p)) implies ρ(xt, pt) < ε for any t ≥ 0;
2.b. there exists γ(p) > 0 such that ρ(x, p) < γ(p) (h(x) = h(p)) implies
lim ρ(xt, pt) = 0.
t→+∞

3. 3.a for all ε > 0 and K ∈ C(X) there exists δ(ε, K) > 0 such that ρ(x 1 , x2 ) <
δ (h(x1 ) = h(x2 ); x1 , x2 ∈ K) implies ρ(x1 t, x2 t) < ε for all t ≥ 0;
˙
3.b lim ρ(x1 t, x2 t) = 0 for all (x1 , x2 ) ∈ X ×X.
t→+∞
4. the equality (2.42) holds for all K ∈ C(X).

Proof. We will prove that condition 1. implies condition 2. If we suppose that it is

not so, then there are p0 ∈ X, ε0 > 0, pn → p0 (h(pn ) = h(p0 )) and tn → +∞ such
that

ρ(pn tn , p0 tn ) ≥ ε0 . (2.43)

Since (X, T1 , π) satisfies the condition (A), then we may suppose that the sequences
{pn tn } and {p0 tn } are convergent. Letting p̄ = lim pn tn , p̄0 = lim p0 tn
n→+∞ n→+∞
and taking into consideration (2.43), we will have p̄ 6= p̄0 . On the other hand,
h(p̄) = lim h(pn )tn = lim h(p0 )tn = h(p̄0 ) = ȳ ∈ JY and according to Lemma
n→+∞ T n→+∞
1.3 p̄, p̄0 ∈ LX Xȳ , but by virtue of condition 1. we have p̄ = p̄0 . The obtained
contradiction proves the necessary assertion.
Now we will note that condition 1. implies condition 2.b. To prove this impli-
cation is sufficient to show that

lim ρ(x1 t, x2 t) = 0
t→+∞

˙
for all (x1 , x2 ) ∈ X ×X. Assuming the contrary we obtain

ρ(x01 tn , x02 tn ) ≥ ε0 . (2.44)

The dynamical system (X, T1 , π) satisfies the condition (A) and, consequently, we
may assume that the sequences {x0i tn }(i = 1, 2) and {y0 tn } (y0 = h(x01 ) = h(x02 ))
are convergent. We denote by x̄0i := lim x0i tn and ȳ0 := lim y0 tn , then
T n→+∞ n→+∞
x̄01 , x̄02 ∈ LX Xȳ0 and according to the condition 1. x̄01 = x̄02 . The last equality
and the inequality (2.44) are contradictory. This contradiction proves the necessary
assertion.
72 Global Attractors of Non-autonomous Dissipative Dynamical Systems

We will show that condition 2. implies condition 3. Note that

lim ρ(xt, pt) = 0 (2.45)

t→+∞

for all p ∈ X and x ∈ Xq (q = h(p)). In fact, we denote by Gq := {x | x ∈ X

such that the equality (2.45) holds } and suppose that Gq 6= Xq . By virtue of the
Condition 2, Gq is open in Xq . Let Γq := ∂Gq (∂Gq is the boundary of Gq ) and
T
p̄ ∈ Γq , then B(p̄, γ(b̄)) (Xq \ Gq ) 6= ∅ (B(p̄, γ(b̄)) := {x | h(x) = h(p̄), ρ(x, p̄) <
γ(p̄)}). It is easy to see that the last relations are not satisfied simultaneously and,
consequently, Γq = ∅ for all q ∈ Y , i.e. Xq = Gq . Let K ∈ C(X) and ε > 0, then
there exists δ(ε, K) > 0 such that ρ(x1 , x2 ) < δ(h(x1 ) = h(x2 ); x1 , x2 ∈ K) implies
ρ(x1 t, x2 t) < ε for any t ≥ 0. Assuming the contrary, we obtain K0 ∈ C(X), ε0 >
0, δn → 0 (δn > 0), {xin } ⊆ K0 (i = 1, 2) and tn → +∞ such that ρ(x1n , x2n ) < δn
and

ρ(x1n tn , x2n tn ) ≥ ε0 . (2.46)

Since K0 is a compact subset of X we may suppose that the sequences {xin } (i = 1, 2)

are convergent and we denote by x̄ := lim x1n = lim x2n (x̄ ∈ K0 ). According to
n→+∞ n→+∞
the condition 2., for ε0 > 0 and x̄ ∈ K0 there exists δ( ε30 , x̄) > 0 such that ρ(x, x̄) <
δ( ε30 , x̄) (h(x) = h(x̄)) implies ρ(xt, x̄t) < ε30 for all t ≥ 0. Since xin → x̄ (i = 1, 2),
then there exists n̄ such that ρ(xin , x̄) < δ( ε30 , x̄) (n ≥ n̄) and, consequently,

2ε0
ρ(x1n t, x2n t) ≤ (2.47)
3
for all t ≥ 0 and n ≥ n̄. But the inequalities (2.47) and (2.46) are contradictory.
Thus, we showed that condition 2. implies condition 3.
We will prove that condition 3. implies condition 4. If we suppose the contrary,
then there exist ε0 > 0, K0 ∈ C(X), tn → +∞ and {xin } ⊆ K0 (i = 1, 2; h(x1n ) =
h(x2n )) such that the inequality (2.46) holds. We may assume without loss of gen-
erality that the sequences {xin } (i = 1, 2) are convergent, because K0 is compact.
Let xi := lim xin , 0 < ε < ε0 and δ( ε3 , K0 ) > 0 be chosen according to condition
n→+∞
3.a. Since h(x1 ) = h(x2 ) and x1 , x2 ∈ K0 , then for 3ε there exists L( 3ε , x1 , x2 ) > 0
such that ρ(x1 t, x2 t) < ε3 for all t ≥ L( 3ε , x1 , x2 ) and, consequently,

ρ(x1n tn , x2n tn ) ≤ ρ(x1n tn , x1 tn ) + ρ(x1 tn , x2 tn ) + ρ(x2 tn , x2n tn ) < ε (2.48)

for sufficiently large n. The inequalities (2.47) and (2.48) are contradictory. Hence,
the necessary assertion is proved.
Finally, we note that 4. implies 1. In fact, if we suppose that there exists
T
y0 ∈ JY such that LX Xy0 contains at least two points x1 and x2 (x1 6= x2 )
and denoting by K a compact invariant set such that x1 , x2 ∈ K, we will have
 Non-autonomous Dissipative Dynamical Systems 73

T
x1 , x2 ∈ Ky0 = K Xy0 . On the other hand, according to Lemma 2.10 Ky0 contains
no more than one point. The obtained contradiction proves Theorem 2.7. 

Corollary 2.2 Let (X, T1 , π) and (Y, T2 , σ) be two compactly dissipative

dynamical systems, then the following conditions are equivalent:

(1) the non-autonomous dynamical system h(X, T1 , π), (Y, T2 , σ), hi is convergent;
P+
(2) every semi-trajectory x (x ∈ X) is asymptotically stable;
(3) 3.a and 3.b from Theorem 2.7 are fulfilled;
(4) the equality (2.42) holds for all K ∈ C(X).

Definition 2.13 Recall that the dynamical system (X, T, π) is called locally com-
pact if for every x ∈ X there exist δ = δx > 0 and l = lx > 0 such that π l B(x, δ) is
relatively compact.

Theorem 2.8 Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical

system, (Y, T2 , σ) be compactly dissipative and (X, T1 , π) be locally compact. For
h(X, T1 , π), (Y, T2 , σ), hi to be convergent, it is necessary and sufficient that every
semi-trajectory Σ+ x of the system (X, T1 , π) would be relatively compact and that the
dynamical system h(h−1 (JY ), T1 , π), (JY , T2 , σ), hi, where JY is a Levinson center
of (Y, T2 , σ), would be convergent.

Proof. The necessity of the theorem is evident.

Sufficiency. We will prove that the dynamical system (X, T1 , π) is point dissi-
pative. For this aim under the conditions of the theorem it is sufficient to show
S
that ΩX = {ωx | x ∈ X} is compact. Note that h(ωx ) ⊆ ωh(x) ⊆ JY and,
consequently, ωx ⊆ h−1 (JY ). Since ωx is compact and invariant, then ωx ⊆ J˜
where J˜ is the center of Levinson of the system (h−1 (JY ), T1 , π). Thus, ΩX ⊆ J,˜
and, consequently, ΩX is compact. According to Theorem 1.10 for a locally com-
pact dynamical system point dissipativity and compact dissipativity are equivalent.
Hence, the dynamical system (X, T1 , π) is compactly dissipative. Let JX be the
center Levinson of (X, T1 , π), then h(JX ) ⊆ JY and, consequently, JX ⊆ h−1 (JY ).
Since J˜ is a maximal compact invariant set in (h−1 (JY ), T1 , π), then JX ⊆ J˜ and
T T T
JX Xy ⊆ J˜ Xy for all y ∈ JY . From this inclusion follows that the set JX Xy
contains at most one point for any y ∈ JY . The theorem is proved. 

Theorem 2.9 Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical

system, (Y, T2 , σ) be a compactly dissipative dynamical system and let there ex-
ists y0 ∈ Y such that Y = H + (y0 ). For the non-autonomous dynamical system
h(X, T1 , π), (Y, T2 , σ), hi to be convergent, it is necessary and sufficient that the
following conditions would be fulfilled:

(1) the dynamical system (X, T1 , π) satisfies the condition (A);

74 Global Attractors of Non-autonomous Dissipative Dynamical Systems

T
(2) LX Xy contains at most one point for every y ∈ JY = ωy0 .

Proof. The necessity of the conditions 1 and 2 is evident.

Sufficiency. Let x0 ∈ Xy0 , then h(H + (x0 )) = H + (y0 ) and h(ωx0 ) = ωy0 . Note
that h(ΩX ) ⊆ ΩY ⊆ JY = ωy0 and since ωx0 ⊆ ΩX , then h(ΩX ) = ωy0 . On the
T
other hand, ΩX ⊆ LX , LX Xy contains at most one point for all y ∈ JY = ωy0 .
Thus, ΩX = ωx0 is compact and, consequently, (X, T1 , π) is pointwise dissipative.
Since (X, T1 , π) is point dissipative and satisfies the condition (A), then by Theorem
1.15 (X, T1 , π) is compactly dissipative. Let JX be the center of Levinson of the
T
dynamical system (X, T1 , π), then JX ⊆ LX and, consequently, JX Xy contains
at most one point for every y ∈ JY . The theorem is proved. 

Definition 2.14 A point y0 ∈ Y is called [103],[301] asymptotically stationary

(asymptotically τ -periodic, asymptotically almost periodic, asymptotically recur-
rent) if there exists a stationary (τ − periodic, almost periodic, recurrent) point
q ∈ Y such that

lim d(y0 t, qt) = 0.

t→+∞

Remark 2.5 a. Let Y = H + (y0 ) = {y0 t | t ≥ 0} be compact, then the dynamical

system (Y, T, σ) is compactly dissipative and JY = ωy0 .
b. Let y0 be asymptotically stationary (asymptotically τ -periodic, asymptotically
almost periodic, asymptotically recurrent) and Y = H + (y0 ), then (Y, T, σ) is com-
pactly dissipative and the set JY = ωy0 is minimal.

Definition 2.15 A point x ∈ X is called [49],[103],[301] comparable in limit with

regard to the recurrence property with a point y ∈ Y if the inclusion Ly ⊆ Lx holds,
where Ly := {{tn }|tn → +∞ and {ytn } is convergent}.

It is known [49, 103, 301] that if Ly ⊆ Lx , then the point x possesses the same
character of the recurrence property in limit as the point y ∈ Y. In particular, if
the point y ∈ Y is asymptotically stationary (asymptotically τ -periodic, asymptot-
ically almost periodic, asymptotically recurrent) and Ly ⊆ Lx , then the point x
will be asymptotically stationary (asymptotically τ -periodic, asymptotically almost
periodic, asymptotically recurrent) .

Theorem 2.10 Let y0 ∈ Y be asymptotically stationary (asymptotically τ -

periodic, asymptotically almost periodic, asymptotically recurrent) and Y = H + (y0 ).
Then the non-autonomous dynamical system h(X, T1 , π), (Y, T2 , σ), hi will be con-
vergent if and only if the following conditions are fulfilled:

a. the dynamical system (X, T1 , π) satisfies the condition (A);

 Non-autonomous Dissipative Dynamical Systems 75

b. every point x ∈ X is comparable in limit with regard to the recurrence prop-

erty with the point y = h(x) and, in particular, x is asymptotically stationary
(asymptotically ω- periodic, asymptotically almost periodic, asymptotically re-
current);
c. for any ε > 0 and K ∈ C(X) there exists δ = δ(ε, K) > 0 such that ρ(x1 , x2 ) <
δ (h(x1 ) = h(x2 ); x1 , x2 ∈ K) implies ρ(x1 t, x2 t) < ε for all t ≥ 0;
˙
d. the equality lim ρ(x1 t, x2 t) = 0 holds for all (x1 , x2 ) ∈ X ×X.
t→+∞

Proof. The necessity of the conditions a,c and d is assured by Remark 2.5. Now, let
us show that under the conditions of the theorem the condition b. holds. Let x ∈ X
and y = h(x), then, according to the convergence of the non-autonomous dynamical
system h(X, T1 , π), (Y, T2 , σ), hi, the set H + (x) := {xt | t ≥ 0} is compact. We note
T T
that ωx Xq ⊆ JX Xq for all q ∈ ωy and since h(X, T1 , π), (Y, T2 , σ), hi is conver-
T
gent, then ωx Xq contains a single point. According to Theorem 1 [49] the point
x is comparable in limit with regard to the recurrence property with the point y.
If y ∈ H + (y0 ), then it is evident that the point y will be asymptotically stationary
(asymptotically ω- periodic, asymptotically almost periodic, asymptotically recur-
rent) and, consequently, the point x possesses the same character of recurrence
property in limit as the point y does.
We will show that the Conditions a, b, c and d imply the convergence of the
non-autonomous dynamical system h(X, T1 , π), (Y, T2 , σ), hi. First of all, according
to the Condition b we have that ωx 6= ∅ is compact, minimal and h(ωx ) = ωy0 for
T
all x ∈ X. We note that ωx Xq contains a single point for every q ∈ ωy0 . In the
T
opposite case there exist q0 ∈ ωy0 , p1 , p2 ∈ ωx Xq0 (p1 6= p2 ) and tin → +∞ (i =
1, 2) such that xtin → pi (i = 1, 2) as n → +∞. We note that ytin → q0 (i = 1, 2)
as n → +∞, where y = h(x). Let t̄2n−1 = t1n and t̄2n = t2n for every n ∈ N, then
{t̄n } ∈ Ly and, consequently, {t̄n } ∈ Lx , i.e. {xtn } is convergent; therefore p1 = p2 .
The last equality contradicts the choice of the points p1 and p2 . The obtained
T
contradiction proves the necessary assertion. Now we will prove that ω x1 Xq =
T T
ωx2 Xq for all x1 , x2 ∈ X and q ∈ ωy0 . Let q ∈ ωy0 , {pi } = ωxi Xq (i = 1, 2) and
{tn } ∈ Lq be such that qtn → q. By virtue of the Condition c and the minimality
of ωxi (i = 1, 2) we have ρ(p1 tn , p2 tn ) → 0 as n → +∞ and, consequently, p1 = p2 .
Thus, ωx1 = ωx2 for all x1 , x2 ∈ X and, consequently, (X, T1 , π) is point dissipative
and since (X, T1 , π) satisfies the condition (A), then according to Theorem 2.9
(X, T1 , π) is compactly dissipative. To finish the proof of the theorem is sufficient
to apply Theorem 2.7 and Remark 2.4. 

Corollary 2.3 Under the conditions of Theorem 2.11 if the space X is locally
compact, then the Condition a results from the Conditions b, c and d.
76 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Theorem 2.11 Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical

system, (Y, T2 , σ) be compactly dissipative and let its Levinson’s center JY be min-
P+
imal (i.e., every semi-trajectory y (y ∈ JY ) is dense in JY ). Then the following
conditions are equivalent:

(1) the non-autonomous dynamical system h(X, T1 , π), (Y, T2 , σ), hi is convergent;
(2) the dynamical system (X, T1 , π) satisfies the condition (A) and for every K ∈
C(X) the equality (2.42) holds.

Proof. By virtue of Corollary 2.2 Condition 1 implies Condition 2. Let us show

P+ S P+
that the converse assertion holds. Let K ∈ C(X), then K := { x |x ∈ K} is
relatively compact and according to Lemma 1.3 the set
\ [
Ω(K) = πτ K
t≥0 τ ≥t

is non-empty, compact, invariant and, consequently, h(Ω(K)) ⊆ Ω(h(K)) ⊆ JY ,

because JY is the maximal compact invariant set in Y . Thus, JY is minimal, then
the equality

h(Ω(K)) = JY

holds. We note that Ω(K1 ) = Ω(K2 ) for all K1 and K2 from C(X). In fact,
S
since M := Ω(K1 ) Ω(K2 ) is compact and invariant and JY is minimal, we have
T
h(M ) = JY . On the other hand, according to Lemma 2.10 the set My = M Xy
T T
contains a single point for every y ∈ JY . We have Ω(Ki ) Xy ⊆ M Xy (i = 1, 2)
T T T
and Ω(K1 ) Xy = Ω(K2 ) Xy = M Xy for any y ∈ JY and, consequently,
Ω(K1 ) = Ω(K2 ) for all K1 and K2 from C(X). From this follows that (X, T1 , π)
is compactly dissipative and according to Theorem 2.7 h(X, T1 , π), (Y, T2 , σ), hi is
convergent. 

Corollary 2.4 Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical

system, (Y, T2 , σ) be compactly dissipative, JY be minimal and (X, T1 , π) satisfy
the condition (A). Then the Conditions 1-4 from Corollary 2.2 are equivalent.

Theorem 2.12 Let (Y, T2 , σ) be compactly dissipative and h(LX ) = JY . For the
non-autonomous dynamical system h(X, T1 , π), (Y, T2 , σ), hi to be convergent, it is
necessary, and if JY = Y it is sufficient, that there would be fulfilled the following
conditions:

1. Σ+ is relatively compact for all x ∈ X and LX is relatively compact;

x T
2. LX Xy contains only one point for any y ∈ Y ;
T
3. for every ε > 0 there exists δ(ε) > 0 such that ρ(x, xy ) < δ ({xy } = LX Xy
and h(x) = y ∈ JY ) implies ρ(xt, xyt ) < ε for all t ≥ 0 and x ∈ X.
 Non-autonomous Dissipative Dynamical Systems 77

Proof. Necessity. Let h(X, T1 , π), (Y, T2 , σ), hi be convergent, then (X, T1 , π) is
compactly dissipative and JX = LX . It is easy to see that the Conditions 1 and 2
are fulfilled. We will show that Condition 3 is fulfilled too. Suppose that it is not
so, then there exist ε0 > 0, δn ↓ 0, {xn } from X, {yn } from JY and tn → +∞ such
that ρ(xn , xyn ) < δn (yn = h(xn )) and

ρ(xn tn , xyn tn ) ≥ ε0 . (2.49)

Since the sets JY and JX are compact, then we can suppose that the sequences {yn }
and {xyn } are convergent. Let y0 = lim yn , then xy0 = lim xyn = lim xn .
n→+∞ n→+∞ n→+∞
By the compact dissipativity of the dynamical system (X, T1 , π) we can suppose
that the sequence {xn tn } is convergent and let x̄ = lim xn tn . Since yn tn ∈ JY ,
n→+∞
then the sequence {yn tn } can be considered convergent too. Let ȳ = lim yn tn
n→+∞
and we note that h(x̄) = lim h(xn )tn = lim yn tn = ȳ, x̄ ∈ JX and hence
T n→+∞ n→+∞
x̄ ∈ JX Xȳ = {xȳ }, i.e., x̄ = xȳ . On the other hand, passing to limit in (2.49) as
n → +∞ we obtain ρ(x̄, xȳ ) ≥ ε0 . The obtained contradiction proves the statement
required.
Sufficiency. Suppose that Conditions 1, 2 and 3 of the theorem are fulfilled. For
the convergence of the system h(X, T1 , π), (Y, T2 , σ), hi it is sufficient to show that
under the conditions of the theorem the dynamical system (X, T1 , π) is compactly
dissipative. By the condition of the theorem (X, T1 , π) is point dissipative and
S
ΩX = {ωx | x ∈ X} ⊆ LX . Note, that the set LX is closed. We will show that
the set LX is orbitally stable. If we suppose that it is not true, then there exist
ε0 > 0, xn → x0 ∈ LX and tn → +∞ such that

ρ(xn tn , LX ) ≥ ε0 .

Since yn → y0 = h(x0 ), where yn = h(xn ), then under the conditions of the theorem
xyn → xy0 = x0 and, consequently, ρ(xn , xyn ) ≤ ρ(xn , x0 ) + ρ(x0 , xyn ) → 0. From
this relation and Condition 3 of the theorem follows that ρ(xn tn , xyn tn ) → 0. This
relation and the inequality (2.49) are contradictory. Thus, LX is compact, invariant
and orbitally stable. Since ΩX ⊆ LX , then J + (ΩX ) ⊆ LX . Let us show that
J + (ΩX ) = JX . In fact, let x̄ ∈ LX and ϕ : S → LX be a whole trajectory
T S
of (X, T1 , π) passing through point x̄ for t = 0. Denote by αϕx̄ := ϕ(τ ),
T T t≤0 τ ≤t
then ΩX αϕx̄ 6= ∅. Let p ∈ αϕx̄ ΩX , then there exists tn → −∞ such that
ϕ(tn ) → p and, consequently, π −tn ϕ(tn ) = ϕ(0) = x̄, i.e., x̄ ∈ Jp+ ⊆ J + (ΩX ).
Thus, LX ⊆ J + (ΩX ), i.e. LX = J + (ΩX ) is compact and orbitally stable and
by Theorem 1.13 the dynamical system (X, T1 , π) is compactly dissipative. The
theorem is proved. 

Theorem 2.13 Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical

system and Y be a compact minimal set, then the following conditions are equivalent:
78 Global Attractors of Non-autonomous Dissipative Dynamical Systems

1. h(X, T1 , π), (Y, T2 , σ), hi is convergent;

P+
2. every semi-trajectory x (x ∈ X) is relatively compact and asymptotically sta-
ble;
P+
3. (a) every semi-trajectory x (x ∈ X) is relatively compact;
˙
(b) lim ρ(x1 t, x2 t) = 0 for all (x1 , x2 ) ∈ X ×X;
t→+∞
(c) for any ε > 0 and K ∈ C(X) there exists δ(ε, K) > 0 such that ρ(x1 , x2 ) <
δ(h(x1 ) = h(x2 ); x1 , x2 ∈ K) implies ρ(x1 t, x2 t) < ε for all t ≥ 0.
P+
4. every semi-trajectory x (x ∈ X) is relatively compact and the equality (2.42)
holds for all K ∈ C(X).

Proof. The implications 1. → 2. → 3. → 4. are proved by a slight modification of

the proof of Theorem 2.7. To finish the proof of our theorem is sufficient to establish
that Condition 4 implies Condition 1. Let x0 ∈ X. Since Σ+ x0 is relatively compact,
then ωx0 6= ∅, is compact and all the motions in ωx0 are extendable to the left. We
set M := ωx0 . Since h(ΩX ) ⊆ ΩY ⊆ Y and Y is minimal, then h(M ) = Y and,
T
consequently, My := M Xy 6= ∅ for all y ∈ Y . By Lemma 2.1 the set My consists
of the one point for any y ∈ Y . We will show now that for every x ∈ X the equality
S T
ωx = M is true. In fact, let K := ωx M , then by Lemma 2.10 Ky := K Xy
contains only one point for every y ∈ Y . Since h(ωx ) = h(ωx0 ) = h(K) = Y , then
T T T
ωx Xy = ωx0 Xy = K Xy for all y ∈ Y and, consequently, ωx = ωx0 = M for
any x ∈ X. Thus, (X, T1 , π) is point dissipative. Let now K ∈ C(X), then Σ+ (K)
is relatively compact. In fact, if {xn } ⊆ K and tn → +∞, then by the condition 4.
we have

lim ρ(xn tn , Mh(xn)tn ) = 0,

n→+∞

and, consequently, the sequence {xn tn } is relatively compact. According to

Theorem 1.15 the dynamical system (X, T1 , π) is compact dissipative. Let JX be
the center of Levinson of the dynamical system (X, T1 , π). By Lemma 2.10 the set
T
JX Xy contains only one point for each y ∈ Y . The theorem is completely proved.


Theorem 2.14 Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical

system, M 6= ∅ be a compact positively invariant set. Suppose that the following
conditions are fulfilled:

(1) h(M ) = Y ;
T
(2) M Xy contains a single point for all y ∈ Y ;
(3) M is globally asymptotically stable, i.e. for any ε > 0 there exists δ(ε) > 0 such
T
that ρ(x, p) < δ (x ∈ Xy , p ∈ My := M Xy ) implies ρ(xt, pt) < ε for all t ≥ 0
and lim ρ(xt, Mh(x)t ) = 0 for all x ∈ X.
t→+∞

Then the non-autonomous dynamical system h(X, T1 , π), (Y, T2 , σ), hi is convergent.
 Non-autonomous Dissipative Dynamical Systems 79

Proof. We note that under the conditions of the theorem the dynamical system
(X, T1 , π) is point dissipative and ΩX ⊆ M. We will show that set M is orbitally
stable in (X, T1 , π) . Suppose that it is not true, then there exist ε0 > 0, δn →
0, xn ∈ B(M, δn ) and tn → +∞ such that

ρ(xn tn , M ) ≥ ε0 . (2.50)

Since M is compact, then we may suppose that the sequence {xn } is convergent.
Let x0 := lim xn , xyn ∈ Myn , ρ(xn , M ) = ρ(xn , xyn ) and y0 = h(x0 ), then x0 =
n→+∞
lim xyn and x0 ∈ My0 . Let qn = h(xn ) and note that
n→+∞

ρ(xn , xqn ) ≤ ρ(xn , xyn ) + ρ(xyn , xqn ) → 0 (2.51)

as n → +∞, because qn → y0 and xqn → x0 . Taking into account (2.51) and the
asymptotic stability of the set M , we have

ρ(xn tn , xqn tn ) = ρ(xn tn , xqn tn ) → 0. (2.52)

But the equalities (2.50) and (2.52) are contradictory. Hence, the set M is orbitally
stable in (X, T1 , π) and by virtue of Theorem 1.13 the dynamical system (X, T1 , π)
is compactly dissipative and JX ⊆ M . To finish the proof of Theorem it is sufficient
T T
to note that h(JX ) = JY and for all y ∈ JY we have JX Xy ⊆ M Xy and,
T
consequently, JX Xy contains a single point for any y ∈ JY . The theorem is
proved. 

Remark 2.6 If there exists y0 ∈ Y such that Y = H + (y0 ), then it is evident that
Theorem 2.14 is invertible. For this aim we may take set H + (x0 ), where x0 ∈ Xy0 ,
in the quality of set M appearing in the theorem.

2.5 Tests for convergence

Note that in Theorems 2.5 and 2.6 there are contained some conditions under which
the property of compact dissipativity is kept under the homomorphisms.
Below we make some assertions which ensure dissipativity of the system
(X, T1 , π) if the dynamical system (Y, T2 , σ) has this property and there exists
a homomorphism h of (X, T1 , π) onto (Y, T2 , σ).

Definition 2.16 Let h be a homomorphism of (X, T1 , π) onto (Y, T2 , σ). A set

M ⊆ X is called uniformly stable in the positive direction with respect to the homo-
morphism h if for all ε > 0 there exists δ = δ(ε, M ) > 0 such that for any x1 , x2 ⊆ M
for which h(x1 ) = h(x2 ) the inequality ρ(x1 , x2 ) < δ implies ρ(x1 t, x2 t) < ε for every
t ≥ 0.
80 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Definition 2.17 A dynamical system (X, T1 , π) is called uniformly stable (in the
positive direction) with respect to the homomorphism h on compact subsets from
X, if every compact M ∈ C(X) is uniformly stable in the positive direction with
respect to the homomorphism h.

Lemma 2.11 Suppose that a homomorphism h satisfies the following conditions:

(1) there is a continuous invariant section ϕ : Y → X of the homomorphism h;
(2) lim ρ(x1 t, x2 t) = 0 for all x1 , x2 ∈ X (h(x1 ) = h(x2 ));
t→+∞
(3) the dynamical system (X, T1 , π) is uniformly stable in the positive direction on
compact subsets from X with respect to the homomorphism h.
Then:
(1) if (Y, T2 , σ) is point dissipative then (X, T1 , π) is point dissipative too. More-
over, ΩX and ΩY are homeomorphic;
(2) if (Y, T2 , σ) is compactly dissipative then (X, T1 , π) is also compactly dissipative
and, moreover, ΩX and ΩY are homeomorphic.

Proof. Let (Y, T2 , σ) be point dissipative. Then ΩY is a nonempty, compact and

invariant set and, consequently, M := ϕ(ΩY ) ⊆ ΩX is also non-empty, compact
and invariant. For x ∈ X and y := h(x) we have lim ρ(xt, ϕ(y)t) = 0. Hence,
t→+∞
Σ+x is relatively compact compact and ωx ⊆ ωϕ(y) ⊆ ϕ(ΩY ) = M . Thus, ΩX ⊆ M
and (X, T1 , π) is point dissipative and ϕ(ΩY ) = ΩX . Since ϕ : ΩY → ΩX separates
points and the set ΩY is compact, then ΩY and ΩX are homeomorphic.
Now we will prove the second statement of the lemma. Let (Y, T2 , σ) be com-
pactly dissipative. By the statement above, the dynamical system (X, T1 , π) is
point dissipative. Let M := ϕ(JY ) = ϕ(D+ (ΩY )) ⊆ D+ (ΩX ). We will show that
the set M is orbitally stable. If we suppose that it is not true, then there are
ε0 , xn → x0 ∈ M and tn → +∞ such that

ρ(xn tn , M ) ≥ ε0 . (2.53)

Note, that h(xn ) = yn → y0 = h(x0 ) ∈ h(M ) = h(ϕ(JY )) = JY and by compact

dissipativity of (Y, T2 , σ) we can suppose that the sequence {yn tn } is convergent.
Let y := lim yn tn , then y ∈ JY and ϕ(y) = lim ϕ(h(xn ))tn . Since ϕ : JY →
n→+∞ n→+∞
ϕ(JY ) = M separates points and h ◦ ϕ = idY , then ϕ : JY → M = ϕ(JY ) is a
homomorphism and ϕ ◦ h(x) = x for all x ∈ M and, consequently, ϕ(h(xn )) →
ϕ(h(x0 )) = x0 ∈ M . From this relation follows that

lim ρ(xn , ϕ ◦ h(xn )) = 0. (2.54)

n→+∞
S S
Let K := M {xn } {ϕ ◦ h(xn )}, ε > 0 and δ(ε, K) > 0 be the numbers from
the condition of uniform stability in the positive direction on compact subsets from
 Non-autonomous Dissipative Dynamical Systems 81

X of the dynamical system (X, T1 , π) with respect to the homomorphism h. The

equality (2.54) implies that for sufficiently large n the inequality ρ(xn , ϕ◦h(xn )) < δ
holds and, consequently, ρ(xn t, ϕ ◦ h(xn )t) < ε for all t ≥ 0. In particular,

ρ(xn tn , ϕ ◦ h(xn )tn ) < ε (2.55)

for sufficiently large n. Taking into account that ε is arbitrary, we get that from
(2.55) follows lim xn tn = lim ϕ ◦ h(xn )tn = ϕ(y) ∈ M . This equality and
n→+∞ n→+∞
(2.53) are contradictory. The obtained contradiction shows that M is orbitally
stable. Thus, (X, T1 , π) is point dissipative, ΩX ⊆ M , the set M is non-empty,
compact, invariant and orbitally stable. By Theorem 1.13, the dynamical system
(X, T1 , π) is compactly dissipative and JX ⊆ M = ϕ(JY ) ⊆ D+ (ΩX ). According
to Theorem 1.11 and Corollary 1.5, we have JX = ϕ(JY ) = D+ (ΩX ). Thus, ϕ is a
homeomorphism of JY onto JX . The lemma is proved. 

Corollary 2.5 Under the conditions of Lemma 2.11, if (Y, T2 , σ) is compactly

dissipative, then h(X, T1 , π), (Y, T2 , σ), hi is convergent.

Proof. This assertion follows from Lemma 2.11 and Corollary 2.2. 

Let (Y, S, σ) be a two-sided dynamical system, (X, S+ , π) be a semi-group

dynamical system and h : X → Y be a homomorphism of (X, S+ , π) onto (Y, S, σ).
Consider a non-autonomous dynamical system h(X, S+ , π), (Y, S, σ), hi and denote
by Γb (Y, X) the family of all continuous and bounded sections of the homomorphism
h. By equality

d(ϕ1 , ϕ2 ) := sup ρ(ϕ1 (y), ϕ2 (y)) (2.56)

y∈Y

there is defined a metric on Γb (Y, X).

Let X ×X˙ := {(x1 , x2 ) : x1 , x2 ∈ X, h(x1 ) = h(x2 )} and let V : X ×X
˙ → R+ be
a mapping satisfying the following conditions:

˙
1. a(ρ(x1 , x2 )) ≤ V (x1 , x2 ) ≤ b(ρ(x1 , x2 )) for all (x1 , x2 ) ∈ X ×X, where a, b
are two functions from A (A := {a | a : R+ → R+ , a is continuous, strongly
increasing and a(0) = 0 }) and Im(a) = Im(b) ;
˙
2. V (x1 , x2 ) = V (x2 , x1 ) for all (x1 , x2 ) ∈ X ×X;
3. V (x1 , x2 ) ≤ V (x1 , x3 ) + V (x3 , x1 ) for all x1 , x2 , x3 ∈ X such that h(x1 ) =
h(x2 ) = h(x3 ).

From the conditions 1–3 follows that the function V on each fiber Xy = h−1 (y)
defines some metric which is topologically equivalent to ρ.
82 Global Attractors of Non-autonomous Dissipative Dynamical Systems

˙ → R+ satisfies the conditions

Lemma 2.12 Suppose that the function V : X ×X
1–3. Then by equality

p(ϕ1 , ϕ2 ) := sup{V (ϕ1 (y), ϕ2 (y))| y ∈ Y } (2.57)

on Γb (Y, X) there is defined a complete metric that is topologically equivalent to

(2.56).

Proof. From Condition 1 it follows that the function V vanishes only on the
diagonal of X ×X,˙ i.e. V (x1 , x2 ) = 0 if and only if x1 = x2 . Taking into account
(2.57), we have d(ϕ, ψ) = 0 if and only if ϕ = ψ.
From Property 2 of the function V it follows that the function d is symmetric
(i.e., d(ϕ, ψ) = d(ψ, ϕ)). Finally, from Condition 3. it follows that the function d
satisfies the triangle inequality.
From the condition 1. follows that the metrics (2.56) and (2.57) are topologically
equivalent.
Thus, to complete the proof of the lemma it is sufficient to establish that the
space (Γb (Y, X), d) is complete. Let {ϕn } be a Cauchy’s sequence in (Γb (Y, X), d),
i.e., d(ϕn , ϕm ) → 0 as n, m → +∞. For every ε > 0 there exists N1 (ε) > 0 such
that

ρ(ϕn (y), ϕm (y)) < ε (2.58)

for any y ∈ Y and n, m > N1 (ε). From this relation and from the completeness
of (X, ρ) it follows that for each y ∈ Y there exists lim ϕn (y) in (X, ρ). We set
n→+∞
ϕ(y) := lim ϕn (y). We will show that this equality holds uniformly with respect
n→+∞
to y ∈ Y and, consequently, ϕ ∈ Γb (Y, X). In fact, passing to limit in the equality
(2.58) as m → +∞ we have ρ(ϕn (y), ϕ(y)) ≤ ε for all n > N1 (ε) and y ∈ Y. Thus
ϕ ∈ (Γb (Y, X), d) and d(ϕn , ϕ) → 0 as n → +∞. The lemma is proved. 

Denote by S t : Γb (Y, X) → Γ(Y, X) the mapping, defined by the equality

(S ϕ)(y) = π t ϕ(σ −1 y) for all t ∈ S+ , ϕ ∈ Γb (Y, X) and y ∈ Y . It is easy to
t

check that the family of mappings {S t }t≥0 is a commutative semigroup.

Lemma 2.13 ˙
If there is a function V : X ×X → R+ satisfying Conditions 1–3
and

4. V (x1 t, x2 t) ≤ N e−νt V (x1 , x2 ) (∀(x1 , x2 ) ∈ X ×X,

˙ t ≥ 0), where N , ν > 0,

then the semigroup {S t }t≥0 has a unique fixed point ϕ ∈ Γb (Y, X) which is an in-
variant section of h.
 Non-autonomous Dissipative Dynamical Systems 83

Proof. Note that

p(S t ϕ1 , S t ϕ2 ) = sup{V (π t ϕ1 (σ −t y), π t ϕ2 (σ −t y))| y ∈ Y }

≤ N e−νt sup{V (ϕ1 (σ −t y), ϕ2 (σ −t y)) | y ∈ Y } ≤ N e−νt p(ϕ1 , ϕ2 )

and, consequently, S t is a contraction for sufficiently large t. From this fact and
taking into consideration that {S t } is commutative we obtain the existence of a
unique fixed point ϕ of the semigroup {S t }t≥0 , which is an invariant section of h.
The lemma is proved. 

Theorem 2.15 Let h(X, S+ , π), (Y, S, σ), hi be a non-autonomous dynamical

system and let the following conditions be fulfilled:

1. (Y, S, σ) is compactly dissipative;

2. Γb (Y, X) 6= ∅;
˙ → R+ satisfying Conditions 1–4 of Lemma 2.13.
3. there exists V : X ×X

Then (X, S+ , π) is compactly dissipative and the sets JX and JY are homeomorphic.

Proof. Under the conditions of the theorem, by Lemma 2.13 the semi-group {S t }t≥0
has a unique fixed point ϕ that is an invariant section of h.
In addition,

a(ρ(x1 t, x2 t)) ≤ V (x1 t, x2 t) ≤ N e−νt V (x1 , x2 ) ≤ N e−νt b(ρ(x1 , x2 )).

Therefore, ˙
lim a(ρ(x1 t, x2 t)) = 0 for all (x1 , x2 ) ∈ X ×X and, consequently,
t→+∞
lim ρ(x1 t, x2 t) = 0. Note, that the system (X, S+ , π) is uniformly stable with
t→+∞
respect to h. In fact, let ε > 0 and δ(ε) = b−1 (N a(ε)). Then ρ(x1 , x2 ) < δ(ε)
(h(x1 ) = h(x2 )) implies ρ(x1 t, x2 t) < ε for all t ≥ 0. To finish the proof of the
theorem it is sufficient to refer to Lemma 2.11. 

Corollary 2.6 Under the conditions of Theorem 2.15 the non-autonomous

dynamical system h(X, S+ , π), (Y, S, σ), hi is convergent.

Proof. This assertion follows from Theorem 2.15 and Corollary 2.5. 
The function V : X ×X ˙ → R+ is said to be continuous, if xin → xi (i = 1, 2 and
h(xn ) = h(xn )) implies V (x1n , x2n ) → V (x1 , x2 ).
1 2

Theorem 2.16 Let (X, T1 , π) and (Y, T2 , σ) be compactly dissipative. The

dynamical system h(X, T1 , π), (Y, T2 , σ), hi is convergent if and only if there is a
˙ → R+ satisfying the following conditions:
continuous function V : X ×X

(1) V is positively defined, i.e. V (x1 , x2 ) = 0 if and only if x1 = x2 ;

84 Global Attractors of Non-autonomous Dissipative Dynamical Systems

˙
(2) V (x1 t, x2 t) ≤ V (x1 , x2 ) for all t ≥ 0 and (x1 , x2 ) ∈ X ×X;
(3) V (x1 t, x2 t) = V (x1 , x2 ) for all t ≥ 0 if and only if x1 = x2 .

Proof. Suppose that the conditions of the theorem are fulfilled. We will show that
for all ε > 0 and compact K ⊆ X there is δ(ε, K) > 0 such that ρ(x1 , x2 ) < δ
(x1 , x2 ∈ K and h(x1 ) = h(x2 )) implies ρ(x1 t, x2 t) < ε for all t ≥ 0. Indeed, if we
suppose that it is not true, then there exist ε0 > 0, compact K0 ⊆ X, sequences
δn ↓ 0, {xin } ⊆ K0 (i = 1, 2 and h(x1n ) = h(x2n )) and tn → +∞ such that

ρ(x1n , x2n ) < δn and ρ(x1n tn , x2n tn ) ≥ ε0 . (2.59)

By the compact dissipativity of (X, T1 , π) we can assume that the sequences {xin tn }
(i = 1, 2) are convergent. Let us put x̄i := lim xin tn (i = 1, 2). It is clear that
n→+∞
h(x̄1 ) = h(x̄2 ). Since {xin } ⊆ K, then we may suppose that they are convergent too.
By virtue of (2.59), lim x1n = lim x2n = x̄ and
n→+∞ n→+∞

0 ≤ V (x̄1 , x̄2 ) = lim V (x1n tn , x2n tn ) ≤ lim V (x1n , x2n ) = V (x̄, x̄) = 0,
n→+∞ n→+∞

from which it follows that x̄1 = x̄2 . The last equality contradicts (2.59).
˙ (h(x1 ) = h(x2 )) the equality
We will show now that for all (x1 , x2 ) ∈ X ×X

lim ρ(x1 t, x2 t) = 0
t→+∞

holds. If we suppose that it is not true, then there exist y0 ∈ Y, x̄1 , x̄2 ∈ Xy0 , ε0 > 0
and tn → +∞ such that

ρ(x̄1 tn , x̄2 tn ) ≥ ε0 . (2.60)

Let (X ×X, T1 , π×π) := (X, T1 , π)×(X, T1 , π) be a direct product of (X, T1 , π) and

(X, T1 , π). In view of the compact dissipativity of (X, T1 , π), the point (x̄1 , x̄2 ) ∈
X × X is L+ stable in (X × X, T1 , π × π), and from Condition 2 of the theorem
follows the existence of a finite limit

V0 = lim V (x̄1 t, x̄2 t). (2.61)

t→+∞

Let (p, q) ∈ ω(x̄1 ,x̄2 ) , then from (2.61) it follows that V (p, q) = V0 . By the invariance
of ω(x̄1 ,x̄2 ) we have V (pt, qt) = V (p, q) for all t ∈ T, and, according to Condition 3
of the theorem, p = q, i.e.,

ω(x̄1 ,x̄2 ) ⊆ ∆X := {(x, x) | x ∈ X}. (2.62)

We may suppose that the sequences {x̄i tn } (i = 1, 2) are convergent. Let us put
p̄ = lim x̄1 tn and q̄ = lim x̄2 tn , then (p̄, q̄) ∈ ω(x̄1 ,x̄2 ) and from (2.60) it
n→+∞ n→+∞
follows that p̄ 6= q̄. The last equality contradicts the inclusion (2.62). The obtained
 Non-autonomous Dissipative Dynamical Systems 85

contradiction proves the required statement. By Theorem 2.7, the non-autonomous

dynamical system (2.1) is convergent.
Conversely, let h(X, T1 , π), (Y, T2 , σ), hi be convergent. We will define the func-
˙ → R+ by the equality
tion V : X ×X

V (x1 , x2 ) := sup{ρ(x1 t, x2 t) | t ≥ 0}. (2.63)

It is clear that this function satisfies Conditions 1 and 2 of the theorem. Let us
show that V satisfies Condition 3 of the theorem. Let V (x1 t, x2 t) = V (x1 , x2 ) for
all t ≥ 0. If we suppose that x1 6= x2 , then δ := V (x1 , x2 ) > 0. In view of Theorem
2.7, ρ(x1 t, x2 t) → 0 as t → +∞. Therefore, there exists t0 = t0 (δ) > 0 such that
ρ(x1 t, x2 t) < 2δ for all t ≥ t0 . Hence,
δ
V (x1 t0 , x2 t0 ) = sup{ρ(x1 t, x2 t) | t ≥ t0 } ≤ < V (x1 , x2 ).
2
The last inequality contradicts our assumption. We will prove that for every com-
pact K ⊆ X

lim sup{ρ(x1 t, x2 t) | x1 , x2 ∈ K, h(x1 ) = h(x2 )} = 0. (2.64)

t→+∞

In fact, if suppose that it is not true, then there exit ε0 > 0, compact K ⊆ X,
δn ↓ 0 and sequences {xin } ⊆ K (i = 1, 2 and h(x1n ) = h(x2n )) tn → +∞ such
that Condition (2.46) holds. Without loss of generality we may suppose that the
sequences {xin } and {xin tn } (i = 1, 2) are convergent. Let lim xin = xi (i = 1, 2).
n→+∞
We put x̄i := lim xin tn (i = 1, 2). Note that h(x̄1 ) = h(x̄2 ) = y, x̄1 , x̄2 ∈ ω(K) ⊆
n→+∞
JX and, hence, x̄1 = x̄2 . The last equality contradicts (2.46). Thus, the equality
(2.64) is proved. From the equalities (2.63) and (2.64), it follows that

V (x1 , x2 ) = ρ(x1 τ, x2 τ ) (2.65)

for some τ (x1 , x2 ) ∈ [0, l(K)].

˙
Let (x1n , x2n ) → (x1 , x2 ) ∈ X ×X. The sequence {V (x1n , x2n )} = {ρ(x1n τn , x2n τn )}
is bounded, where τn = τ (x1n , x2n ). We will show that it has a unique limit point.
Denote by Ṽ one of the limit points of the sequence {V (x1n , x2n )}. There exists a
subsequence {V (x1kn , x2kn )} such that V (x1kn , x2kn ) → Ṽ as n → +∞. Since {τkn } is
bounded, then we may suppose that it is convergent. Then Ṽ = ρ(x1 τ 0 , x2 τ 0 ), where
τ 0 = lim τkn . Let us show that ρ(x1 τ, x2 τ ) = ρ(x1 τ 0 , x2 τ 0 ) (τ = τ (x1 , x2 ) ≥ 0 is
n→+∞
chosen out of Condition (2.65)). If we suppose that ρ(x1 τ, x2 τ ) 6= ρ(x1 τ 0 , x2 τ 0 ), then
ρ(x1 τ, x2 τ ) > ρ(x1 τ 0 , x2 τ 0 ). Let ε > 0 be such that ρ(x1 τ, x2 τ ) + 2ε < ρ(x1 τ, x2 τ ).
Then for a sufficiently large k ∈ N we have |ρ(x1kn τ, x2kn τ ) − ρ(x1 τ, x2 τ )| < ε and
|ρ(x1kn τkn , x2kn τkn ) − ρ(x1 τ 0 , x2 τ 0 )| < ε, and, consequently,

ρ(x1kn τ, x2kn τ ) > ρ(x1kn τkn , x2kn τkn ) = V (x1kn , x2kn ). (2.66)
86 Global Attractors of Non-autonomous Dissipative Dynamical Systems

The inequality (2.66) contradicts the choice of the number τkn . Thus, V (x1 , x2 ) =
ρ(x1 τ 0 , x2 τ 0 ) and, consequently, lim V (x1n , x2n ) = V (x1 , x2 ), i.e., the function V
n→+∞
is continuous. The theorem is proved. 

Theorem 2.16 is a generalization for the abstract non-autonomous dynamical

systems of the theorems 7.2 and 7.3 from [270].

Theorem 2.17 Let dynamical systems (X, T1 , π) and (Y, T2 , σ) be compactly dis-
sipative. A non-autonomous dynamical system h(X, T1 , π),(Y, T2 , σ), hi is conver-
˙
gent if and only if there exists a continuous function V : X ×X → R+ , satisfying
the following conditions:

(1) V is positively defined;

˙ \ ∆X , where ∆X :=
(2) V (x1 t, x2 t) < V (x1 , x2 ) for all t > 0 and (x1 , x2 ) ∈ X ×X
{(x, x) | x ∈ X}.

Proof. The sufficiency of Conditions of the theorem it follows from the previous
theorem. It is sufficient to remark that from the second condition of Theorem 2.17
it follows the second and third conditions of Theorem 2.60.
Necessity. Let h(X, T1 , π), (Y, T2 , σ), hi be convergent and the function V :
X ×X˙ → R+ be defined by the equality (2.63). Then it is continuous and sat-
isfies Conditions 1–3 of Theorem 2.16. Let us put
Z +∞
V (x1 , x2 ) := V(x1 t, x2 t)e−t dt. (2.67)
0

From the definition of the function V follow its continuity and positive definiteness.
The function V satisfies Condition 2 of Theorem 2.17. In fact, if we suppose the
contrary, then there exit (x̄1 , x̄2 ) ∈ X ×X ˙ and t0 > 0 such that V (x̄1 t0 , x̄2 t0 ) =
V (x̄1 , x̄2 ) and x̄1 6= x¯2 . Then from (2.67) and from the last equality, it follows that

V(x̄1 (t0 + t), x̄2 (t0 + t)) = V(x̄1 t, x̄2 t) (2.68)

for all t ≥ 0. By virtue of (2.68), the function ϕ(t) = V(x̄1 t, x̄2 t) is t0 periodic.
It is obvious that ϕ is continuous and non-increasing. Therefore, ϕ is stationary
and, hence, V(x̄1 t, x̄2 t) = V(x̄1 , x̄2 ) for all t ≥ 0. Since the function V satisfies
Conditions 1–3 of Theorem 2.16, then the equality V(x̄1 t, x̄2 t) = V(x̄1 , x̄2 ) for all
t ≥ 0 implies x̄1 = x̄2 . The last equality contradicts the choice of (x̄1 , x̄2 ). The
obtained contradiction completes the proof of the theorem. 

Theorem 2.18 Let (X, T1 , π) and (Y, T2 , σ) be compactly dissipative and there
˙ → R+ , satisfying the following conditions:
exist a continuous function V : X ×X

(1) V is positively defined;

 Non-autonomous Dissipative Dynamical Systems 87

˙
(2) V (x1 t, x2 t) ≤ ω(V (x1 , x2 ), t) for all (x1 , x2 ) ∈ X ×X and t ≥ 0, where ω :
R+ × R+ → R+ is a non-increasing in first variable function and ω(x, t) → 0,
as t → +∞, for every x ∈ R+ .

Then h(X, T1 , π), (Y, T2 , σ), hi is convergent.

Proof. According to Corollary 2.2 for the convergence of h(X, T1 , π), (Y, T2 , σ), hi
it is sufficient to show that the equality (2.42) holds for every compact K ∈ C(X).
First of all, we will show that the equality

lim sup V (x1 t, x2 t) = 0 (2.69)

t→+∞ ˙
(x1 ,x2 )∈K ×K

holds for all compact K ∈ C(X). Indeed, since K is compact, then there exists
˙
α > 0 such that V (x1 , x2 ) ≤ α for all (x1 , x2 ) ∈ K ×K and V (x̄1 , x̄2 ) = α for some
˙
(x̄1 , x̄2 ) ∈ K ×K, and, consequently,

V (x1 t, x2 t) ≤ ω(V (x1 , x2 ), t) ≤ ω(α, t). (2.70)

Since ω(α, t) → 0 as t → +∞, then from (2.70) it follows (2.69). Let us show that
(2.69) implies (2.42). If we suppose the contrary, then there exit K ∈ C(X), ε0 >
0, {xin } ⊆ K (i = 1, 2) and tn → +∞ such that

ρ(x1n tn , x2n tn ) ≥ ε0 . (2.71)

In view of the compact dissipativity of (X, T1 , π) we may suppose that the sequences
{xin tn } (i = 1, 2) are convergent. Let us set x̄i = lim xin tn , then from the
n→+∞
inequality (2.71), it follows that x̄1 6= x̄2 . On the other hand, by the inequality
(2.70)

0 ≤ V (x1n tn , x2n tn ) ≤ V (α, tn ) → 0

as n → +∞ and, hence,

V (x̄1 , x̄2 ) = lim V (x1n tn , x2n tn ) = 0.

n→+∞

From the last equality, it follows that x̄1 = x̄2 . The obtained contradiction com-
pletes the proof of the theorem. 

˙ →
Remark 2.7 a. Condition 2 of Theorem 2.18 holds, if the function V : X ×X
R+ satisfies one of the following conditions:

(1) V (x1 t, x2 t) ≤ N e−νt V (x1 , x2 ) for all t ≥ 0 and (x1 , x2 ) ∈ X ×X˙ (ω(x, t) =
−νt
N xe );
2V (x1 ,x2 ) ˙ 2x
(2) V (x1 t, x2 t) ≤ 2+V (x1 ,x2 )t for all t ≥ 0 and (x1 , x2 ) ∈ X ×X (ω(x, t) = 2+xt );
88 Global Attractors of Non-autonomous Dissipative Dynamical Systems

b. All results of this section are true also in the case when the spaces X and Y
are not metric but pseudo-metric.

In conclusion, we note that convergent systems are in some sense the simplest
dissipative dynamical systems. If h(X, T1 , π), (Y, T2 , σ), hi is a convergent non-
autonomous dynamical system and JX (JY ) is a Levinson center of the dynamical
system (X, T1 , π) ((Y, T2 , σ)), then JX and JY are homeomorphic. Although the
center of Levinson of a convergent system can by completely described, it may
be sufficiently complicated. We will give an example which illustrates the above
cooment.

Example 2.2 Let Y := R and (Y, Z+ , σ) be a cascade generated by positive

powers of the odd function g, defined on R+ in the following way:
 1

 −2y , 0≤y≤ 2


g(y) = 2(y − 1) , 12 < y ≤ 1 .



1
2 (y − 1) , 1 < y < +∞.

It is easy to check that (Y, Z+ , σ) is dissipative and JY ⊆ [−1, 1]. Let us put
X := R2 and denote by (X, Z+ , π) a cascade generated by the positive powers of
the mapping P : R2 → R2
! !
u f (u, y)
P = , (2.72)
y g(y)

1
where f (u, y) := 10 u + 21 y. Finally, let h = pr2 : X → Y . From (2.72), it
follows that h is a homomorphism of (X, Z+ , π) onto (Y, Z+ , σ) and, consequently,
h(X, Z+ , π), (Y, Z+ , σ), hi is a non-autonomous dynamical system. Note that

|(u1 , y) − (u2 , y)| = |u1 − u2 | = 10|P (u1 , y) − P (u2 , y)|. (2.73)

From (2.73), it follows that

|P n (u1 , y) − P n (u2 , y)| ≤ N e−νn |hu1 , yi − hu2 , yi| (2.74)

for all n ∈ Z+ , where N = 1 and ν = ln 10. We will show that the dynamical
system h(X, Z+ , π), (Y, Z+ , σ), hi is convergent. It is possible to check directly
that the dynamical system (X, Z+ , π) is dissipative; therefore, for convergence of
T
h(X, Z+ , π), (Y, Z+ , σ), hi it is sufficient to show that JX Xy contains only one
point for all y ∈ JY . If we suppose that it is not true, then there exit y0 ∈ JY
T
and x1 , x2 ∈ JX Xy0 with x1 6= x2 . Let ϕi be a motion of the dynamical system
(X, Z+ , π) defined on Z and passing through the point xi (i = 1, 2). From (2.74),
 Non-autonomous Dissipative Dynamical Systems 89

it follows that

|ϕ1 (−n) − ϕ2 (−n)| ≥ N −1 eνn |u1 − u2 | (2.75)

for all n ∈ Z+ , where (ui , y0 ) = xi (i = 1, 2). On the other hand,

sup |ϕ1 (n) − ϕ2 (n)| < +∞. (2.76)

n∈Z

The inequalities (2.75) and (2.76) are contradictory. The obtained contradic-
tion proves the required statement. Thus, the dynamical system h(X, Z+ , π),
(Y, Z+ , σ), hi is convergent and, consequently, JX and JY are homeomorphic.

Note that this example is an just a slight modification of an example from

[297, p.39–42]. In addition, they have the same attractors and the corresponding
systems on the attractors act equally. Hence, the Levinson centers JX and JY are
intermixing (strange) attractors.
Thus, the Levinson center JX of the convergent dynamical system h(X, T1 , π),
(Y, T2 , σ), hi is completely defined by the structure of JY , but the last one can be
organized in a very complicated way.

2.6 Global attractors of non-autonomous dynamical systems

Let Y be a compact metric space, (X, h, Y ) be a locally trivial Banach fibering

[190] and | · | be a norm on (X, h, Y ) coordinated with the metric ρ on X (that is
ρ(x1 , x2 ) = |x1 − x2 | for any x1 , x2 ∈ X such that h(x1 ) = h(x2 ) ).

Theorem 2.19 Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical

system and for any bounded set M ∈ B(X) there exists l = l(M ) > 0 such that
π l (M ) would be relatively compact (that is the dynamical system (X, T 1 , π) is com-
pletely continuous). Then the following conditions are equivalent:

1. there exits a positive number r such that for any x ∈ X there will be τ = τ (x) ≥
0 for which |xτ | < r;
2. the dynamical system h(X, T1 , π), (Y, T2 , σ), hi is compactly dissipative and

lim sup ρ(xt, J) = 0 (2.77)

t→+∞ |x|≤R

for any R > 0, where J is the Levinson center of (X, T1 , π), that is the non-
autonomous system h(X, T1 , π), (Y, T2 , σ), hi admits a compact global attractor.

Proof. That Condition 2 implies Condition 1 is evident. Let us show that under
conditions of Theorem 2.19, the converse also holds. Suppose that A(r) := {x ∈
E | |x| ≤ r} where r > 0 is the number appearing in Condition 1. As Y is
90 Global Attractors of Non-autonomous Dissipative Dynamical Systems

compact and the Banach fibering (X, h, Y ) is locally trivial, then its null section
Θ = {θy | y ∈ Y , where θy is the null element of the fiber Ey := h−1 (y)} is compact
and, hence, the set A(r) is bounded, as A(r) ⊆ S(Θ, r) = {x ∈ E| |ρ(x, θ) ≤ r}.
According to the condition of the theorem for a bounded set M there exits a positive
number l such that π l M is relatively compact. Let x ∈ M and τ = τ (x) ≥ 0 be
such that xτ ∈ M , then x(τ + l) ∈ K := π l M . Thus, the non-empty compact
K is a weak attractor of the system (X, T1 , π) and, according to Theorem 1.10,
the dynamical system (X, T1 , π) is compactly dissipative. Let J be the Levinson
center of (X, T1 , π) and R > 0, then the set A(R) := {x ∈ E | |x| ≤ R}, as it
was noticed above, is bounded, and for it there exists a number l > 0 such that
π l A(R) is relatively compact and since (X, T1 , π) is compactly dissipative, then its
Levinson center J, according to Theorem 1.6, attracts the set π l A(R) and, hence,
the equality (2.77) holds. The theorem is proved. 

Corollary 2.7 Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical

system and suppose that the vector fibering of (X, T1 , π) is finite-dimensional, then
Conditions 1. and 2. of Theorem 2.19 are equivalent .

Proof. This assertion it follows from Theorem 2.19 as for any r > 0 the set
{x ∈ E| |x| ≤ r} is compact, if the vector fibering (X, h, Y ) is finite-dimensional
and Y is compact. 

Recall that the dynamical system (X, T1 , π) is called asymptotically compact, if

for any bounded closed positively invariant set M ∈ B(X) there exits a non-empty
compact K such that the equality

lim β(π t M, K) = 0
t→+∞

holds.

Theorem 2.20 Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical

system and for every bounded set A ⊆ X there exists a nonempty compact K A ⊆ X
such that

lim β(π t A, KA ) = 0. (2.78)

t→+∞

Then the conditions 1 and 2 of Theorem 2.19 are equivalent.

Proof. Since Y is compact and (X, h, Y ) is locally trivial, then for any R > 0 the set
{x ∈ E | |x| ≤ R} is bounded. According to Condition 1 of Theorem 2.19, for any
x ∈ E there exits τ = τ (x) ≥ 0 such that xτ ∈ A(r) := {x ∈ E | |x| ≤ r}. According
to Theorem 1.24 the dynamical system (X, T1 , π) is compactly dissipative. Let J
be the Levinson center of (X, T1 , π) and R > 0. As the set M := A(R) := {x ∈
 Non-autonomous Dissipative Dynamical Systems 91

E | |x| ≤ R} is bounded, then according to Condition of the theorem and of Lemma

1.3 the set Ω(M ) 6= ∅, is compact, invariant and the equality

lim β(π t A, ω(A)) = 0

t→+∞

holds. As J is the maximal compact invariant set in (X, T1 , π), then Ω(M ) ⊆ J
and, hence, the equality

lim β(π t A, J) = 0
t→+∞

holds. The theorem is proved. 

Corollary 2.8 Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical

system and suppose that at least one of the following conditions is fulfilled:

(1) for every bounded set A ⊆ X there exits a positive number l such that π l A is
relatively compact;
(2) the dynamical system h(X, T1 , π) is asymptotically compact.

Then Conditions 1 and 2 of Theorem 2.19 are equivalent.

Proof. This assertion it follows directly from Theorems 2.19 and 2.20 . 

Theorem 2.21 Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical

system and (X, T1 , π) be asymptotically compact, then the following conditions are
equivalent:

1. there exits a positive number R0 and for any R > 0 there exists l(R) > 0 such
that

|π t x| ≤ R0 (2.79)

for all t ≥ l(R) and |x| ≤ R;

2. the dynamical system h(X, T1 , π), (Y, T2 , σ), hi admits a compact global attrac-
tor, i.e it is compactly dissipative and for its Levinson center J the equality
(2.77) holds for any R > 0.

Proof. Evidently Condition 2 implies Condition 1, that is why for proving the
theorem it is sufficient to show that from Condition 1 follows Condition 2. Let
M0 ∈ B(X), then there exits R > 0 such that M0 ⊆ A(R) := {x ∈ E| |x| ≤ R}.
According to Condition 1, for a given R there exists l = l(R) > 0 such that (2.79)
S
holds and, in particular, the set M := {π t M0 |t ≥ l(R)} is bounded and positively
invariant. As (X, T1 , π) is asymptotically compact, for the set M there exists a
nonempty compact K for which the equality (2.78) holds. To finish the proof of the
theorem it is sufficient to cite Theorem 2.20. The theorem is proved. 
92 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Theorem 2.22 Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical

system and let mappings π t := π(t, ·) : X → X(t ∈ T1 ) be represented like a
sum π(t, x) = ϕ(t, x) + ψ(t, x) for all t ∈ T1 and x ∈ X Let also the following
conditions be fulfilled:

(1) |ϕ(t, x)| ≤ m(t, r) for all t ∈ T1 , r > 0 and |x| ≤ r, where m : T1 × R+ → R+
and m(t, r) → 0 as t → +∞;
(2) the mappings ψ(t, ·) : X → X(t > 0) are conditionally completely continuous,
that is ψ(t, A) is relatively compact for any t > 0 and a bounded positively
invariant set A ⊆ X.

Then the dynamical system (X, T1 , π) is asymptotically compact.

S
Proof. Let A ⊆ X be a bounded set such that Σ+ (A) := {π t A | t ≥ 0} is also
bounded, r > 0 and A ⊆ {x ∈ X| |x| ≤ r}. Let us show that for any {xk } ⊆ A and
tk → +∞, the sequence {xk tk } is relatively compact. We will convince ourselves
that the set M := {xk tk } may be covered by a compact ε net for any ε > 0.
Let ε > 0 and l > 0 be such that m(l, r) < ε/2 and let us represent M in the
form of the union M1 ∪ M2 where M1 := {xk tk }kk=1 1
, M2 := {xk tk }+∞
k=k1 +1 and
k1 := max{k | tk < l}. The set M2 is a subset of the set π l (Σ+ (A)), the elements
of which we can represent in the form of ϕ(l, x) + ψ(l, x)(x ∈ Σ+ (A)). As the set
ψ(Σ+ (A), l) is relatively compact, then it may be covered by a finite ε/2 net. Let
us notice that for any y ∈ ϕ(l, Σ+ (A)) there exits x ∈ Σ+ (A) such that y = ϕ(l, x)
and |y| = |ϕ(l, x)| ≤ m(l, r) < ε/2. That is why the null section Θ of the fibering
of (X, h, Y ) is an ε/2 net of the set ϕ(l, Σ+ (A)). Thus, M2 , and, subsequently,
M is covered by a compact ε net and as the space X is complete, then the set
M = {xk tk } is relatively compact. Now to finish the proof of the theorem it is
sufficient to cite Lemma 1.3. The theorem is proved. 

At the end of this section we will study dissipative dynamical systems in finite-
dimensional space. We will give some conditions which are equivalent to dissipativ-
ity.
Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical system, Y be com-
pact, (X, h, Y ) be a finite-dimensional vectorial fibering [190] and | · | be a Rieman-
nian metric on (X, h, Y ).

Lemma 2.14 Under the conditions mentioned above all the types of dissipativity
of the non-autonomous dynamical system h(X, T1 , π), (Y, T2 , σ), hi are equivalent.

Proof. Since the space Y is compact and (X, h, Y ) is finite-dimensional and locally
trivial [190], then the space X is locally compact and possesses the Heine-Borel
property. To finish the proof of the lemma it is sufficient to cite Theorem 1.10. 
 Non-autonomous Dissipative Dynamical Systems 93

Theorem 2.23 Let h(X, T1 , π), (Y, T2 , σ), hi be a finite-dimensional non-

autonomous dynamical system, then the following conditions are equivalent:

1. there exits a positive number R such that for all x ∈ X the following inequality

lim sup |xt| < R (2.80)

t→+∞

holds;
2. there exits a positive number r such that for all x ∈ X there exists τ ≥ 0 such
that |xτ | < r;
T
3. there exists a nonempty compact K1 ⊂ X such that ωx K1 6= ∅ for any x ∈ X;
4. there exists a nonempty compact K2 ⊂ X such that ∅ 6= ωx ⊂ K2 for every
x ∈ X;
5. there exits a positive number R0 such that for all R1 > 0 there exits l(R1 ) > 0
such that

|xt| < R0 (2.81)

for all t ≥ l(R1 ) and |x| ≤ R1 .

Proof. It is clear that 5. ⇒ 1. ⇒ 4. ⇒ 3. ⇒ 2. According to Theorem 2.19 from 2.

it follows 5. The theorem is completely proved. 

2.7 Global attractor of cocycles

Definition 2.18 The family {Iy | y ∈ Y } (Iy ⊂ W ) of nonempty compact subsets

W is called (see, for example, [6] and [153]) a compact pullback attractor (uniform
pullback attractor ) of a cocycle ϕ, if the following conditions hold:
S
(1) the set I := {Iy | y ∈ Y } is relatively compact;
(2) the family {Iy | y ∈ Y } is invariant with respect to the cocycle ϕ, i.e.
ϕ(t, Iy , y) = Iσ(t,y) for all t ∈ T+ and y ∈ Y ;
(3) for all y ∈ Y (uniformly in y ∈ Y ) and K ∈ C(W )

lim β(ϕ(t, K, y−t ), Iy ) = 0,

t→+∞

where β(A, B) := sup{ρ(a, B) : a ∈ A} is a semi-distance of Hausdorff.

Below in this section we suppose that T2 = S.

Definition 2.19 The family {Iy | y ∈ Y }(Iy ⊂ W ) of nonempty compact subsets

is called a compact global attractor of the cocycle ϕ, if the following conditions are
fulfilled:
S
(1) the set I := {Iy | y ∈ Y } is relatively compact;
94 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(2) the family {Iy | y ∈ Y } is invariant with respect to the cocycle ϕ;

(3) the equality

lim sup β(ϕ(t, K, y), I) = 0

t→+∞ y∈Y

holds for every K ∈ C(W ).

Let M ⊆ W and
\ [
ωy (M ) := ϕ(τ, M, σ −τ y) (2.82)
t≥0 τ ≥t

for all y ∈ Y, where y −t := σ(−t, y).

Lemma 2.15 The following assertions hold:

(1) The point p ∈ ωy (M ) if and only if there exit tn → +∞ and {xn } ⊆ M such
that p = lim ϕ(tn , xn , y −tn );
n→+∞
(2) U (t, y)ωy (M ) ⊆ ωyt (M ) for all y ∈ Y and t ∈ T+ , where U (t, y) := ϕ(t, ·, y);
(3) for any point w ∈ ωy (M ) the motion ϕ(t, w, y) is defined on S;
(4) if there exits a nonempty compact K ⊂ W such that

lim β(ϕ(t, M, y −t ), K) = 0,
t→+∞

then ωy (M ) 6= ∅, is compact,

lim β(ϕ(t, M, y −t ), ωy (M )) = 0 (2.83)

t→+∞

and

U (t, y)ωy (M ) = ωyt (M ) (2.84)

for all y ∈ Y and t ∈ S+ .

Proof. The first assertion of the lemma directly it follows from the equality (2.82).
Let w ∈ ωy (M ), then there exit tn → +∞ and xn ⊆ M such that w =
lim ϕ(tn , xn , y −tn ) and, hence,
n→+∞

ϕ(t, w, y) = lim ϕ(t, ϕ(tn , xn , y −tn ), y) = lim ϕ(t + tn , xn , y −tn ). (2.85)

n→+∞ n→+∞

Thus, ϕ(t, w, y) ∈ ωyt (M ), that is U (t, y)ωy (M ) ⊆ ωyt (M ) for all y ∈ Y and
t ∈ T+ .
From the equality (2.85), it follows that the motion ϕ(t, w, y) is defined on S
like ϕ(t + tn , xn , y −tn ) is defined on [−tn , +∞) and tn → +∞.
The fourth assertion of the lemma is proved as in Theorem 2.4 and Lemma 1.1.3
from [217]. 
 Non-autonomous Dissipative Dynamical Systems 95

Definition 2.20 A cocycle ϕ over (Y, S, σ) with the fiber W is said to be com-
pactly dissipative, if there exits a nonempty compact K ⊆ W such that

lim sup{β(U (t, y)M, K) | y ∈ Y } = 0 (2.86)

t→+∞

for any M ∈ C(W ).

Lemma 2.16 Let Y be compact and hW, ϕ, (Y, S, σ)i be a cocycle over (Y, S, σ)
with the fiber W . For hW, ϕ, (Y, S, σ)i to be compactly dissipative, it is necessary
and sufficient that the semigroup autonomous system (X, S+ , π) should be compactly
dissipative.

Proof. This assertion follows from the corresponding definitions. 

Theorem 2.24 Let Y be compact, hW, ϕ, (Y, S, σ)i be compactly dissipative and
K be the nonempty compact subset of W appearing in the equality (2.86), then:
1. Iy = ωy (K) 6= ∅, is compact, Iy ⊆ K and

lim β(U (t, y −t )K, Iy ) = 0

t→+∞

for every y ∈ Y ;
2. U (t, y)Iy = Iyt for all y ∈ Y and t ∈ S+ ;
3.

lim β(U (t, y −t )M, Iy ) = 0 (2.87)

t→+∞

for all M ∈ C(W ) and y ∈ Y ;

4.

lim sup{β(U (t, y −t )M, I) | y ∈ Y } = 0 (2.88)

t→+∞

for any M ∈ C(W ), where I := ∪{Iy | y ∈ Y };

5. Iy = pr1 Jy for all y ∈ Y , where J is the Levinson center of (X, T+ , π), and
hence I = pr1 J;
6. the set I is compact;
7. the set I is connected if one of the next two conditions is fulfilled:
(a) S+ = R+ and the spaces W and Y are connected;
(b) S+ = Z+ and the space W × Y possesses the (S)-property or it is connected
and locally connected.

Proof. The first two assertions of the theorem follow from Lemma 2.15. If we
suppose that the equality (2.87) does not hold, then there exist ε0 > 0, y0 ∈
Y, M0 ∈ C(W ), {xn } ⊆ M0 and tn → +∞ such that

ρ(U (tn , y0−tn )xn , Iy0 ) ≥ ε0 . (2.89)

96 Global Attractors of Non-autonomous Dissipative Dynamical Systems

According to the equality (2.87), for ε0 and y0 ∈ Y there exists t0 = t0 (ε0 , y0 ) > 0
such that
ε0
β(U (t, y0−t )K, Iy0 ) < (2.90)
2
for all t ≥ t0 . Let us notice that

U (tn , y0−tn )xn = U (t0 , y0−t0 )U (tn − t0 , y0−tn )xn .

As hW, ϕ, (Y, S, σ)i is compactly dissipative, then the sequence {U (tn −

t0 , y0−tn )xn } may be considered convergent. Suppose x = lim ϕ(tn −t0 , xn , y0−tn ),
n→+∞
then, according to Lemma 2.15, x ∈ ωy−t0 (M0 ) and U (t0 , y0−t0 )x ∈ ωy0 (M0 ). From
0
the equality (2.86), it follows that x ∈ K. Passing to limit in (2.89) as n → +∞,
we get

U (t0 , y0−t0 )x ∈
/ B(Iy0 , ε0 ). (2.91)

On the other hand, as x ∈ K, from (2.90) we have

ε
U (t0 , y0−t0 )x ∈ B(Iy0 , ).
2
This contradicts (2.91). The obtained contradiction proves the assertion we need.
Let us prove now the equality (2.88). If we suppose that it does not hold, then
there exist ε0 > 0, M0 ∈ C(W ), yn ∈ Y, {xn } ⊆ M0 and tn → +∞ such that

ρ(U (tn , yn−tn )xn , I) ≥ ε0 . (2.92)

As Y is compact, then we may consider the sequences {yn } and {yn tn } convergent.
Suppose y0 := lim yn and y := lim yn tn . According to (2.87), for given ε0 > 0
n→+∞ n→+∞
and y0 ∈ Y , there exists t0 = t0 (ε0 , y0 ) such that
ε0
β(U (t0 , y0 −t )M0 , Iy0 ) < (2.93)
2
for all t ≥ t0 (ε0 , y0 ). Let us note that

U (tn , yn−tn )xn = U (t0 , yn−t0 )U (tn − t0 , yn−tn )xn . (2.94)

As hW, ϕ, (Y, T, σ)i is compactly dissipative, then the sequence {U (tn − t0 , yn−tn )xn }
may be considered a convergent one. Suppose x0 = lim ϕ(tn − t0 , xn , yntn ) and let
n→+∞
us notice that according to (2.93) x0 ∈ K. From the equality (2.94), it follows that
U (tn , yn−tn )xn → U (t0 , y −t0 )x0 and, hence, from (2.92) we have

U (t0 , y0 −t0 )x0 ∈

/ B(Iy0 , ε0 ).

On the other hand, from (2.93) and from x0 ∈ K, it follows that

ε0
U (t0 , y0 −t0 )x0 ∈ B(Iy0 , ).
2
 Non-autonomous Dissipative Dynamical Systems 97

This last inclusion contradicts (2.93), and this finishes the proof of the fourth
assertion of the theorem.
Let us prove the fifth assertion of the theorem. In order to do it let us notice
that w ∈ Iy , if ϕ(t, w, y) is defined on S and ϕ(S, w, y) is relatively compact. In
fact, as w = ϕ(t, ϕ(−t, w, y), y −t ) for all t ∈ S, then from the equality (2.87) it
follows the inclusion we need. Thus, we get the following description of the set
Iy : Iy = {w ∈ W | there exists at least one whole trajectory of hW, ϕ, (Y, S, σ)i,
passing through the point (x, y)}. Now it remains to notice that the Levinson’s
center J is compact and consists of the whole trajectories of (X, S+ , π), and, hence,
pr1 Jy ⊆ Iy for all y ∈ Y .
The compactness of the set I it follows from the equality I = pr1 J, from com-
pactness of J and from continuity of pr1 : X → W .
The last assertion it follows from the next observation: under the conditions of
the theorem Levinson’s center J of the dynamical system (X, S+ , π), according to
Corollary 1.11 and Theorem 1.33, is connected. Hence, I, as a continuous image of
a connected set, also is connected. The theorem is completely proved. 

Remark 2.8 Theorem 2.24 refines and generalises the main results of [110] and
[217].

Theorem 2.25 Under the conditions of Theorem 2.24 the following statements
hold true:

(1) w ∈ Iy (y ∈ Y ) if and only if there exits a whole trajectory ν : S → W of the

cocycle ϕ, satisfying the following conditions: ν(0) = w and ν(S) is relatively
compact;
(2) Iy (y ∈ Y ) is connected, if the space W possesses the property (S).

Proof. To prove the first assertion we note that the continuous function ν : S → W
is a whole trajectory of the cocycle hW, ϕ, (Y, S, σ)i if and only if γ = (ν, IdY ) is a
whole trajectory of the semi-group dynamical system (X, S, π) ( X = W × Y, π =
(ϕ, σ)). By Lemma 2.16, the dynamical system (X, S, π) is compactly dissipative
and according to Theorem 1.6 the set J is compact and invariant and, consequently,
the point (w, y) = x ∈ J if and only if through the point (w, y) = x passes the whole
trajectory γ = (ν, IdY ) of the dynamical system (X, S, π), which belongs to J, i.e.
γ(0) = (ν(0), y) = (w, y) and γ(s) ∈ J for all s ∈ S. To finish the proof of the first
statement of the theorem it is sufficient to cite Theorem 2.24 (item 5.).
To prove the second statement of the theorem we note that under the conditions
of Theorem 2.24 the set Iy 6= ∅ and is compact. Since the space W possesses the
property (S), then there exists a connected compact V ⊇ I. By (2.86), the following
98 Global Attractors of Non-autonomous Dissipative Dynamical Systems

equality

lim β(U (t, σ −t y)(V ), Iy ) = 0

t→+∞

holds. Note, that Iy ⊆ U (t, σ −t y)(V ) for all y ∈ Y and t ∈ S+ ; the mapping
U (t, σ −t y) : W → W is continuous and, consequently, the set U (t, σ −t y)(V ) is
compact and connected. To finish the proof of the theorem it is sufficient to cite
Lemma 3.12 from [126]. 

2.8 Global attractors of non-autonomous dynamical system with

minimal base

Everywhere in this paragraph we suppose that h(X, T1 , π), (Y, T2 , σ), hi is a non-
autonomous dynamical system, Y is a compact minimal set and (X, h, Y ) is a locally
trivial Banach fibering.

Theorem 2.26 Let the following conditions be satisfied:

(1) (X, T1 , π) is completely continuous, that is, for any bounded set A ⊆ X there
exits l = l(A) > 0 such that π l (A) is relatively compact;
(2) all motions (X, T1 , π) are bounded on T+ , that is, sup{|xt| |t ∈ T+ } < +∞
for any x ∈ E ;
(3) there exit y0 and R0 > 0 such that for any x ∈ Xy0 there exist τ = τ (x) ≥ 0
for which

|xτ | < R0 . (2.95)

Then the non-autonomous dynamical system h(X, T1 , π), (Y, T2 , σ), hi admits a
compact global attractor.

Proof. Let R > R0 , then for any x ∈ X there exits τ = τ (x) ≥ 0 such that
|xτ | < R. If it were not so, then there would be R0 > R0 any x10 ∈ E such that

|x10 τ | > R0 (2.96)

for all τ ≥ 0. As the dynamical system (X, T1 , π) is completely continuous and the
motion π(t, x10 ) is bounded on T+ , the point x10 is stable L+ and, as Y is minimal,
the set ωx10 ∩ Xy0 is nonempty. Thus, according to Condition (2.96) we have

|xt| ≥ R0 (2.97)

for all x ∈ ωx1 ∩ Xy0 and t ≥ 0. The inequality (2.97) contradicts (2.95). This
contradiction proves the assertion we need. Now to finish the proof of the theorem
it is sufficient to cite Theorem 2.19. 
 Non-autonomous Dissipative Dynamical Systems 99

Remark 2.9 1. For finite-dimensional systems (that is the vector fibering

(X, h, Y ) is finite-dimensional) Theorem 2.26 refines Theorem 2.6.1 from [209]. To
be exact, the condition of uniform boundedness is changed for ordinary boundedness
of trajectories of (X, T1 , π).
2. If the condition of minimality of Y in Theorem 2.26 is taken away, then
Theorem 2.26 is not true even in the class of linear non-autonomous systems. This
is proved by the example below.

Example 2.3 Let us consider a linear differential equation

x0 = a(t)x, (2.98)

where a ∈ C(R, R) is defined by the equality a(t) := −1 + sin t1/3 . Let us remark
the next properties of the function a and of the equation (2.98):
1. a0 (t) → 0 for t → +∞ ;
2. a(t) ∈ [−2, 0] for all t ∈ R ;
3. {aτ |τ ≥ 0} is relatively compact in C(R, R), where aτ (t) = a(t + τ )(t ∈ R);
4. ωa 6= ∅ and is compact ;
5. all functions from ωa are constant and b(t) = c ∈ [−2, 0](t ∈ R) for any b ∈ ωa ;
6. a(tn ) = 0 if and only if tn = −1 + ( π2 + 2πn)2 (n ∈ Z);
7. there exits {tnk } ⊆ {tn } such that a(t + tnk ) → b(t) and b(t) = 0 for all t ∈ R ;
8. for any b ∈ H + (a) = {aτ |τ ∈ R+ } the inequality

|ϕ(t, x, b)| ≤ |x| (2.99)

holds for all x ∈ R and t ∈ R+ , where ϕ(t, x, b) is the solution of the equation

y 0 (t) = b(t)y; (2.100)

passing through the point x ∈ R for t = 0.

9. if b ∈ ωa \ {0}, then b(t) = c < 0(t ∈ R) and, hence,

lim |ϕ(t, x, b)| = 0 (2.101)

t→+∞

for all x ∈ R ;
10. if b = 0(b ∈ ωa ), then ϕ(t, x, b) = x for all t ∈ R.
Suppose Y := H + (a) and denote by (Y, R+ , σ) the dynamical system of transla-
tions on Y . Let X := R×Y and (X, R+ , π) be a semigroup dynamical system on X,
where π = (ϕ, σ) (that is π(t, (x, b)) = (ϕ(t, x, b), bt ) for all (x, b) ∈ X and t ∈ R+ ).
Then h(X, R+ , π), (Y, R+ , σ)i is a non-autonomous dynamical system generated by
the equation (2.98), where h = pr2 : X → Y . From Properties 1-10 it follows that
for the non-autonomous dynamical system h(X, R+ , π), (Y, R+ , σ), hi generated by
the equation (2.98) all the conditions of Theorem 2.26 are carried out excepting the
minimality of Y and that it has no compact global attractor.
100 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Corollary 2.9 Let (X, T1 , π) be completely continuous and for any y ∈ Y let
there exist R(y) ≥ 0 such that

lim sup |xt| ≤ R(y) (2.102)

t→+∞

for any x ∈ Xy . Then the non-autonomous dynamical system h(X, T1 , π),

(Y, T2 , σ), hi admits a compact global attractor.

Proof. This assertion it follows from Theorem 2.26, if we will notice that from
condition (2.102) it follows the boundedness of every motion from (X, T1 , π) on T+.

Theorem 2.27 Let the next conditions are carrying out:

(1) (X, T1 , π) is asymptotic compact, that is for any bounded positively invariant
set A ⊂ X there exits a nonempty compact KA such that

lim β(π t A, KA ) = 0; (2.103)

t→+∞

(2) (X, T1 , π) is asymptotic bounded, that is for any bounded set A ⊂ E there exits
l = l(A) ≥ 0 such that ∪{π t A|t ≥ l} is bounded;
(3) there exit y0 ∈ Y and R0 > 0 such that (2.95) is fulfilled.
Then the non-autonomous dynamical system h(X, T1 , π), (Y, T2 , σ), hi admits the
maximal compact attractor.

Proof. First, let us notice, that in conditions of Theorem the dynamical system
(X, T1 , π) satisfies the condition of Ladyzhenskaya. Let R > R0 , then for any x ∈ E
there will be τ = τ (x) ≥ 0 such that |xτ | < R. If we suppose that it is not so, then
there exist x1 ∈ E and R0 > R0 such that

|x1 τ | ≥ R0 > R0 (2.104)

for all τ ≥ 0 and, hence, ωx1 ∩ Xy0 6= ∅. That is why for any x ∈ ωx1 ∩ Xy0 the
inequality (2.97) holds, but this contradicts (2.95). Thus the assertion we need is
proved. Now for ending the proof of the Theorem it is sufficient to cite theorem
1.24. 

Remark 2.10 Let us notice, that Theorem 2.27 without demanding the minimal-
ity of Y does not take place even in class of linear systems. The last assertion is
proved by the example 2.3.

Theorem 2.28 Let (X, h, Y ) be a finite-dimensional vectorial fibering, all the

motions (X, T1 , π) are bounded on T+ , Y be a compact minimal set and y0 ∈ Y ,
then the following conditions are equivalent:
1. the non-autonomous dynamical system h(X, T1 , π), (Y, T2 , σ), hi is dissipative;
 Non-autonomous Dissipative Dynamical Systems 101

2. there exits R > 0 such that

lim sup |xt| < R (2.105)

t→+∞

for all x ∈ Xy0 and all motions (X, T1 , π) are bounded on T+ ;

3. there exits a positive number r such that for any x ∈ Xy0 and l > 0 there will
be τ = τ (x) ≥ l for which

|xτ | < r (2.106)

and all the motions (X, T1 , π) are bounded on T+ ;

4. there exits a nonempty compact K1 ⊂ X such that ωx ∩ K1 6= ∅ for all x ∈ Xy0
and all the motions (X, T1 , π) are bounded on T+ ;
5. there exits a nonempty compact K2 ⊆ X such that ωx 6= ∅ and ωx ⊆ K2 for all
x ∈ X y0 ;
6. there exits a positive number R0 such that for any R1 > 0 there will be l(R1 ) > 0,
that

|xt| < R0 (2.107)

for all t ≥ L(R1 ), |x| ≤ R1 (x ∈ Xy0 ).

Proof. Implications 1.=⇒ 6. =⇒ 2. =⇒ 5. =⇒ 4. =⇒ 3. are evident. According to

Theorem 2.26 from 3. it follows 1. The theorem is proved. 

2.9 Homogeneous dynamical systems

The goal of the present section is to study the connection between uniform asymp-
totic stability and exponential asymptotic of the solutions of infinite-dimensional
systems. This problem is studied and solved within the framework of general non-
autonomous dynamical system with infinite-dimensional phase space.
Let (X, h, Y ) be a locally trivial fiber bundle [190] with the fiber E, (X, ρ) be a
complete metric space, T = S+ := {s ∈ S | = s ≥ 0}, where S := R or Z.

Definition 2.21 h(X, T1 , π),(Y, T2 , σ), hi (T1 ⊆ T2 ⊆ S) is said to be homoge-

neous of order k = 1, if π(t, λx) = λπ(t, x) for all λ > 0, x ∈ X and t ∈ T1 .

Definition 2.22 An autonomous dynamical system (X, R+ , π) is said to be ho-

mogeneous of order k (k ≥ 1), if π(t, λx) = λk π(λk−1 t, x) for all λ ≥ 0, x ∈ X and
t ∈ T1 .

If x ∈ X, then we put |x| := ρ(x, θh(x) ), where θy (y ∈ Y ) is null (trivial) element

of the linear space Xy and Θ := {θy | y ∈ Y } is a null (trivial) section of the vectorial
102 Global Attractors of Non-autonomous Dissipative Dynamical Systems

bundle (X, h, Y ). Denote by X s a stable manifold of h(X, T1 , π), (Y, T2 , σ)i, i.e.
X s = {x | x ∈ X, lim |π(t, x)| = 0}.
t→+∞

Lemma 2.17 Let h(X, T1 , π), (X, T2 , σ), hi be a homogeneous dynamical system
of order k = 1. Then

(1) if ΩX is compact, then ΩX ⊆ Θ;

(2) if J + (ΩX ) is compact, then J + (ΩX ) ⊆ Θ.

Proof. To prove the inclusion ΩX ⊆ Θ it is sufficient to show that ωx ⊆ Θ for all

x ∈ X. Let x ∈ X and p ∈ ωx , then by the homogeneity of h(X, T1 , π), (Y, T2 , σ), hi
of order k = 1 we have λp ∈ ωλx for all λ ≥ 0. If we suppose that for some x0 the
inclusion ωx0 ⊆ Θ it is not true, then there exits p ∈ ωx0 \ Θ and, hence, λp ∈ Ω for
all λ ≥ 0. The last fact contradicts the compactness of Ω.
Let now J + (ΩX ) be compact. We will show that J + (ΩX ) ⊆ Θ. If p ∈ J + (ΩX ),
then there exit q ∈ ΩX , qn → q and tn → ∞ such that p = lim π(tn , qn ) and
n→+∞
λp = lim λπ(tn , qn ) = lim π(tn , λqn ). Since λq ∈ ΩX along with the point q,
n→+∞ n→+∞
+
then λp ∈ Jλq ⊆ J + (ΩX ) for all λ ≥ 0 and p ∈ J + (ΩX ). If J + (ΩX ) * Θ, then
should be the point p0 ∈ J + (ΩX ) \ Θ and, consequently, λp0 ∈ J + (ΩX ) ∀λ ≥ 0.
The last inclusion contradicts the compactness of J + (ΩX ). The lemma is proved.

Corollary 2.10 Let h(X, T1 , π), (Y, T2 , σ), hi be a homogeneous non-autonomous

dynamical system of order k = 1. Then the following conditions are equivalent:

(1) h(X, T1 , π), (Y, T2 , σ), hi is compactly dissipative;

(2) h(X, T1 , π), (Y, T2 , σ), hi is convergent.

Theorem 2.29 Let h(X, T1 , π), (Y, T2 , σ), hi be homogeneous of order k = 1 and
(Y, T2 , σ) be pointwise k-dissipative. Then the following conditions are equivalent:

(1) h(X, T1 , π), (Y, T2 , σ), hi is pointwise dissipative (i.e. (X, T, π) is pointwise dis-
sipative);
(2) X s = X.

Proof. Let h(X, T1 , π), (Y, T2 , σ), hi be pointwise dissipative. To prove the equality
X s = X it is sufficient to show that ΩX ⊆ Θ. Since h(X, T1 , π), (Y, T2 , σ), hi is
pointwsice dissipative, then ΩX is compact and by Lemma 2.17 we have ΩX ⊆ Θ.
Let now X s = X; we will show that (X, T1 , π) is pointwise dissipative. First of
all, we note that {π(t, x) | t ≥ 0} is relatively compact for all x ∈ X. To convince
that it is so it is sufficient to show that from arbitrary sequence {π(tn , x)} (tn →
+∞) there can be chosen a convergent subsequence. Let y = h(x), then according to
the pointwise dissipativity of (Y, T2 , h) we may suppose that the sequence {σ(tn , y)}
 Non-autonomous Dissipative Dynamical Systems 103

is convergent. Let q = lim σ(tn , y), then q ∈ ωy . According to the local triviality
n→+∞
of the fiber bundle (X, h, Y ), the sequence {θσ(y,tn ) } → θq and, hence,



ρ(xtn , θq ) ≤ ρ xtn , θσ(y,tn ) + ρ θσ(y,tn ) , θq −→ 0

for n → +∞. In addition, from the reasoning mentioned above it follows that
ωx ⊆ Θ for all x ∈ X and, consequently, ΩX ⊆ Θ. Note, that by Lemma 2.9
h(ΩX ) ⊆ ΩY and, hence, ΩX ⊆ Θ ∩ h−1 (ΩY ). From the compactness of ΩY and
the local triviality of (X, h, Y ) it follows the compactness of Θ ∩ h−1 (ΩY ) and,
consequently, the compactness of ΩX too. The theorem is proved. 

Theorem 2.30 Let h(X, T1 , π), (Y, T2 , σ), hi be homogeneous of order k = 1 and
Y be compact. Then the following conditions are equivalent:
1. h(X, T1 , π), (Y, T2 , σ), hi is compactly dissipative, i.e. (X, T1 , π) is compactly
dissipative;
2. X s = X and the set Θ ∩ h−1 (JY ) is uniformly stable, i.e. for an arbitrary ε > 0
there exists δ(ε) > 0 such that |x| < δ implies |xt| < ε for all t ≥ 0;
3. if the fiber bundle (X, h, Y ) is normed (i.e. the metric ρ on the fibers (X, h, Y )
is compatible with the norm), then there exists a positive number N such that

|xt| ≤ N |x| (2.108)

for all x ∈ X, t ≥ 0 and X s = X.

Proof. Let h(X, T1 , π), (Y, T2 , σ), hi be compactly dissipative and J be the Levinson
center of the dynamical system (X, T1 , π). By Theorem 2.29 X s = X. The inclusion
can be proved JX ⊆ Θ in the same way that the inclusion ΩX ⊆ Θ (see the proof of
Theorem 2.29). We will show that the trivial section Θ of the fiber bundle (X, h, Y )
is uniformly stable. If we suppose that it is not true, then there exit ε0 > 0, δn ↓ 0,
|xn | < δn and tn → +∞ such that

|xn tn | ≥ ε0 . (2.109)

By virtue of the compactness of Y and the local triviality of (X, h, Y ) the set Θ
is compact and, consequently, we may suppose that the sequence {xn } converges.
Let x = lim xn . Since (X, T1 , π) is compactly dissipative, then we may suppose
n→+∞
that the sequence {xn tn } is convergent too and we put x = lim xn tn . Then
n→+∞
x ∈ JX ⊆ Θ and, consequently, |x| = 0. The last equality contradicts the inequality
(2.109). The obtained contradiction proves the required statement.
To prove the implication 2. ⇒ 1. it is sufficient to refer to Theorem 1.13.
Let now the fiber bundle (X, h, Y ) be normed. We will show that in this case
every one of Conditions 1. and 2. is equivalent to Condition 3. We will prove, for
example, that 2. implies 3. Since the trivial section Θ is stable, then for ε = 1 there
104 Global Attractors of Non-autonomous Dissipative Dynamical Systems

exits γ = δ(1) > 0 such that |π(t, x)| ≤ 1 for all t ≥ 0 and |x| ≤ q. We put N = γ −1 ,
then

|π(t, x)| = ||x|γ −1 π(t, x|x|−1 γ)| ≤ |x|γ −1 = N |x|.

for all t ≥ 0 and x ∈ X \ Θ.

Inversely. Let X s = X and let exist N > 0 such that the inequality (2.108)
holds for all x ∈ X and t ≥ 0. Let ε > 0 and δ(ε) < Nε such that |x| < δ implies
|xt| < ε for all t ≥ 0. The theorem is proved. 

Theorem 2.31 Let h(X, T1 , π), (Y, T2 , σ), hi be a homogeneous non-autonomous

dynamical system of order k = 1 and Y be compact. Then the following conditions
are equivalent:

1. h(X, T1 , π), (Y, T2 , σ), hi is locally dissipative, i.e. (X, T1 , π) is locally dissipa-
tive;
2. X s = X and the trivial section Θ is uniformly attracting, i.e. there exits γ > 0
such that

lim sup |π(t, x)| = 0; (2.110)

t→+∞ |x|≤γ

3. if the fiber bundle (X, h, Y ) is normed, then there exit positive numbers N and
ν such that

|π(t, x)| ≤ N e−νt |x| (2.111)

for all t ≥ 0 and x ∈ X.

Proof. We will prove that 1. ⇒ 2. ⇒ 3. If the non-autonomous dynamical system

h(X, T1 , π), (Y, T2 , σ), hi is locally dissipative and J is its Levinson’s center, then by
Theorem 2.30 X s = X and JX ⊆ Θ. In addition, according to Theorem 1.17 there
exists a positive number γ such that

lim β(π t B(J, γ), J) = 0. (2.112)

t→+∞

We will show that for the obtained γ > 0 the equality (2.110) holds. If we suppose
that it is not true, then there exit ε0 > 0, {xn } (|xn | ≤ γ) and tn → +∞ such that
the inequality (2.109) holds. From the equality (2.112) it follows that the sequence
{xn tn } can be considered convergent. Let x = lim xn tn . By (2.112) x ∈ J, and
n→+∞
since JX ⊆ Θ, then |x| = 0. On the other hand, passing to limit in (2.109) as
n → +∞, we obtain |x| ≥ ε0 > 0. The obtained contradiction proves the equality
(2.110).
 Non-autonomous Dissipative Dynamical Systems 105

From the homogeneity of the system h(X, T1 , π), (Y, T2 , σ), hi and the equality
(2.110) it follows that

lim sup |π(t, x)| = 0. (2.113)

t→+∞ |x|≤1

We put ψ(t) = sup{|π(t, x)| : |x| ≤ 1}. Note, that ψ(t + τ ) ≤ ψ(t)ψ(τ ) for all
t, τ ∈ T1 and ψ(t) → 0 as t → +∞. By the equality (2.110), the trivial section Θ
of the fiber bundle (X, h, Y ) is uniformly stable. Indeed, if we suppose that it is
not true, then there exit ε0 , δn ↓ 0, |xn | < δn and tn → +∞ such that the relation
(2.109) holds. We can suppose that δn ≤ γ, and according (2.110) for ε20 there exits
L0 > 0 such that
ε0
|xt| < (2.114)
2
for all t ≥ L0 . Since tn → +∞, then for the sufficiently large n the inequalities
(2.114) and (2.109) are contradictory. The obtained contradiction proves the re-
quired statement. By Theorem 2.30 the function ψ is bounded for t ≥ 0 and accord-
ing to Lemma 5.4 [21] there exist positive numbers N and ν such that ψ(t) ≤ N e−νt
for all t ≥ 0 and, consequently, the inequality (2.111) holds too.
Now we will show, that the inverse implications 3. ⇒ 2. ⇒ 1. hold too. It is
evident that 3. implies 2., therefore to finish the proof of the theorem it is sufficient
to show that from 2. it follows 1. First of all we note that from the equality (2.110),
according to the reasoning above, it follows the uniform stability of the trivial section
Θ. By Theorem 2.30 the dynamical system h(X, T1 , π), (Y, T2 , σ), hi is compactly
dissipative and JX ⊆ Θ. We will show that

lim β(π t B(J, γ), J) = 0.

t→+∞

If we suppose that it is not true, then there exit ε0 > 0, xn ∈ B(J, γ) and tn → +∞
such that

ρ(xn tn , Θy ) ≥ ε0 (2.115)

for all y ∈ Y . From (2.110) it follows, that the sequence {xn tn } is convergent and
lim xn tn = x ∈ Θ. On the other hand, passing to limit in (2.115) as n → +∞, we
n→+∞
obtain ρ(x, Θy ) ≥ ε0 for all y ∈ Y , i.e. x 6∈ Θ. The obtained contradiction proves
the equality (2.111) and, hence, the local dissipativity of (X, T, π). The theorem is
proved. 

Corollary 2.11 Let h(X, T1 , π), (Y, T2 , σ), hi be a homogeneous non-autonomous

dynamical system and (X, T, π) be asymptotically compact, then the following con-
ditions are equivalent:

(1) h(X, T1 , π), (Y, T2 , σ), hi is compactly dissipative;

106 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(2) h(X, T1 , π), (Y, T2 , σ), hi is locally dissipative;

(3) X s = X and the trivial section Θ of the fiber bundle (X, h, Y ) is uniformly
stable;
(4) X s = X and the trivial section Θ of the fiber bundle (X, h, Y ) is uniformly
attracting;
(5) if the fiber bundle (X, h, Y ) is normed, then there exits a positive number N > 0
such that

|xt| ≤ N |x|

for all x ∈ X, t ≥ 0, and X s = X.

(6) if the fiber bundle (X, h, Y ) is normed, then there exit positive numbers N and
γ such that

|xt| ≤ N e−γt |x|

for all t ≥ 0 and x ∈ X.

Proof. This statement it follows from Theorems 1.20, 2.30 and 2.31. 

Corollary 2.12 Let (X, T1 , π) be completely continuous. Then the following con-
ditions are equivalent:
(1) h(X, T1 , π), (Y, T2 , σ), hi is pointwise dissipative;
(2) h(X, T1 , π), (Y, T2 , σ), hi is compactly dissipative;
(3) h(X, T1 , π), (Y, T2 , σ), hi is locally dissipative;
(4) X s = X;
(5) X s = X and Θ is uniformly stable;
(6) X s = X and Θ is uniformly attracting;
(7) if (X, h, Y ) is normed, then there exits N > 0 such that

|xt| ≤ N |x|

for all x ∈ X, t ≥ 0 and X s = X;

(8) if (X, h, Y ) is normed, then there exit N > 0 and γ > 0 such that

|xt| ≤ N e−γt |x|

for all t ≥ 0 and x ∈ X.

Proof. Corollary 2.12 it follows from Theorems 1.10, 2.29 – 2.31. 

Remark 2.11 a) All the results of this section hold also for autonomous homo-
geneous (of order k ≥ 1) system with the continuous time (T = R+ or R).
b) Note, that the results of this section take place, when the bundle (X, h, Y ) is
not necessary linear (vectorial). It is sufficient that there would exist a fiber bundle
 Non-autonomous Dissipative Dynamical Systems 107

(W, h, Y ) such that X ⊆ W and λx ∈ X for all λ ≥ 0 and x ∈ X (in this case we
will call the bundle (X, h, Y ) homogeneous).

2.10 Power-law asymptotic of homogeneous systems

Let (X, h, Y ) be a local, trivial normed fiber bundle with the fiber E. We will say
that a non-autonomous dynamical system h(X, T1 , π), (Y, T2 , σ), hi admits a power-
law asymptotic of order m > 1, if there exist positive numbers ai and bi (i = 1, 2)
such that
1 1
(a1 t + b1 |x|1−m )− m−1 ≤ |π t x| ≤ (a2 t + b2 |x|1−m )− m−1 (2.116)

for all t ≥ 0 and x ∈ X.

Theorem 2.32 In order that a non-autonomous dynamical system h(X, T1 , π),

(Y, T2 , σ), hi would admit a power-law asymptotic of order m > 1, it is necessary
and sufficient that there would exist a continuous function V : X → R+ satisfying
the following conditions:

a) there exit positive numbers α and β such that

α|x| ≤ V (x) ≤ β|x| (2.117)

for all x ∈ X;
b) V is differentiable along the trajectories of the system (X, T1 , π) and

dV (π t x) 
= −|x|m (2.118)
dt t=0

for all x ∈ X.

Proof. Necessity. Suppose that the equality (2.116) holds and

Z
+∞

V (x) := |π τ x|m dτ. (2.119)

0

It is easy to see that under Condition (2.116) the equality (2.119) defines correctly
the mapping V : X → R+ . In addition, under Condition (2.116) the integral (2.119)
is uniformly (in |x| ≤ r, for each r > 0) convergent and, hence, the function V is
continuous. Further, from (2.116) and (2.119) it follows that α|x| ≤ V (x) ≤ β|x|
1 1
for all x ∈ X, where α = m−1 1−m
a1 b 1 and β = m−1 1−m
a2 b 2 .
Sufficiency. Let V : X → R+ be a continuous function satisfying Conditions a)
and b) of the theorem. We will show that the non-autonomous dynamical system
108 Global Attractors of Non-autonomous Dissipative Dynamical Systems

h(X, T1 , π), (Y, T2 , σ), hi admits a power-law asymptotic of order m. Indeed, let us
put ϕ(t) := V (π t x) and note that by (2.118)

dϕ(t)
= −|π t x|m
dt
1
for all t ≥ 0 and x ∈ X. From Condition (2.117) it follows that |π t x| ≥ βV (π t x)
for all t ≥ 0 and x ∈ X and, consequently,

dϕ(t) 1
≤ − ϕm (t).
dt β

Taking into account that ϕ(t) = V (π t x) > 0 for all t ≥ 0, if x 6= 0, we have

 − m−1
1
m−1
V (π t x) ≤ t + α−m+1 |x|−m+1 .
βm

On the other hand, |π t x| ≤ 12 V (π t x) and, hence,

− m−1
1
|π t x| ≤ a2 t + b2 |x|1−m

for all t ≥ 0 and x ∈ X, where a2 = β mm−1

α1−m and b2 = 1.
By analogy, there can be established a lower power-law asymptotic for |π t x|.
The theorem is completely proved. 

Theorem 2.33 Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical

system. Then the following conditions are equivalent:

(1) there exit positive numbers Ni and νi (i = 1, 2) such that

N1 e−ν1 t |x| ≤ |π t x| ≤ N2 e−ν2 t |x|

for all t ≥ 0 and x ∈ X;

(2) (a) there exit positive numbers α and β such that the inequality (2.117) holds
for all x ∈ X;
(b) the function V is differentiable along the trajectories of the system
(X, T1 , π) and

dV (π t x) 
= −|x|
dt t=0

for all x ∈ X.

Proof. This assertion can be proved using the same reasoning as in the proof of
Theorem 2.32. 
 Non-autonomous Dissipative Dynamical Systems 109

Lemma 2.18 Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical

system and let exist positive numbers a, b and m > 1 such that
1
|π t x| ≤ (at + b|x|1−m )− m−1 (2.120)

for all x ∈ X and t ≥ 0. Then

lim |π t x| = 0 (2.121)
t→+∞

uniformly in |x| ≤ r for each r > 0 and

|π t x| ≤ N |x| (2.122)
1
for all t ≥ 0 and x ∈ X, where N = b− 1−m .

1
Proof. Consider the function ω(r, t) = (at + br 1−m )− m−1 defined for all t ≥ 0 and
r ≥ 0. Note, that
∂ω(r, t) m
= br1−m (at + br1−m ) 1−m ≥ 0
∂t
for all t ≥ 0 and, consequently, for |x| ≤ r we have
1 1
|π t x| ≤ (at + b|x|1−m )− m ≤ (at + br1−m )− m ,

from which it follows (2.121).

The inequality (2.122) it follows from (2.120) and the equality
1 1
(at + b|x|1−m )− m−1 = |x|(at|x|m−1 + b)− m−1 .

The lemma is proved. 

Lemma 2.19 Let D be a family of functions m : R+ → R+ satisfying the condi-

tions:

a. there exists M > 0 such that 0 < m(t) ≤ M for all t ≥ 0 and m ∈ D;
b. m(t) → 0 as t → +∞ uniformly in m ∈ D, i.e. for any ε > 0 there exists
L(ε) > 0 such that m(t) < ε for all t ≥ L(ε) and m ∈ D .

Then we have the following assertions:

(1) if m(t + τ ) ≤ m(t)m(τ ) for all t, τ ≥ 0 and m ∈ D, then there exit positive
numbers N and ν such that

m(t) ≤ N e−νt (2.123)

for all t ≥ 0 and m ∈ D;

110 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(2) if m(t + τ ) ≤ m(t)m(τ mα−1 (t)) (α > 1) for all t, τ ≥ 0 and m ∈ D, then there
exist positive numbers a and b such that
1
m(t) ≤ M (a + bt)− α−1

for all t ≥ 0 and m ∈ D.

Proof. We will prove the first statement of the lemma. Let γ ∈ (0, 1) and τ > 0 be
such that m(τ ) ≤ γ < 1 for all m ∈ D. Let us put ν := −τ −1 ln γ and N := M eντ ,
then

m(nτ ) ≤ γ n

for all n = 1, 2, . . .. Let t ≥ 0 and n be a natural number, such that t ∈ [nτ, nτ + τ ].

Since

m(t) = m(t − nτ + nτ ) ≤ m(t − nτ )m(nτ ) ≤ M γ n ,

then the following inequality

m(t) ≤ eν(t−nτ ) m(t − nτ )γ n e−νt

holds too. Taking into account that eντ = γ −1 , by the choice of the number N we
have the required inequality (2.123).
To prove the second statement of the lemma we will choose τ > 0 so that the
inequality
1
m(t) ≤ (2.124)
2
would hold for all t ≥ τ and m ∈ D. Since 0 < m(t) ≤ M and m(t + τ ) ≤
m(t)m(τ mα−1 (t)) for all t, τ ≥ 0 and m ∈ D, then

m(t) ≤ M (0 ≤ t ≤ qτ, q = 2α−1 ) (2.125)

and
1
m(t) ≤ (τ ≤ t < +∞)
2
for all m ∈ D. Set t0 = 0, ti+1 = ti + τ qi (qi = q i ) and note that
1
m(ti ) ≤ (i ≥ 1) (2.126)
2
for all m ∈ D. Indeed, according to (2.124), we have m(t1 ) ≤ 2−1 . Moreover,

m(ti+1 ) = m(ti + τ qi ) ≤ m(ti )m(τ qi mα−1 (ti )) (2.127)

 Non-autonomous Dissipative Dynamical Systems 111

for all m ∈ D. Suppose that (2.126) is valid for all i ≤ n; then from (2.127) it
follows that
1
m(tn+1 ) ≤ n+1 ,
2
since τ qn mα−1 (tn ) ≥ τ (for all m ∈ D) in view of the choice of qn and the inductive
assumption (2.126). Thus, from (2.125) and (2.126) it follows that
1
m(t) ≤
2n
for all t ≥ tn and n ≥ 1. Note. that tn+1 = τ (q n+1 − 1)(q − 1)−1 and, hence,
 − α−1
1

−n 2α−1 − 1
2 =2 tn+1 + 1 .
τ
Now let t ∈ [tn , tn+1 ), then we have
 − α−1
1

−n 2α−1 − 1
2 <2 1+ t . (2.128)
τ
From (2.125) and (2.128) it follows that
 − α−1
1

1−α 1 − 21−α
m(t) ≤ M 2 + t
τ
for all t ≥ 0 and m ∈ D. By setting a = 21−α and b = τ −1 (1 − 21−α ), we obtain
the required assertion. The lemma is proved. 

Theorem 2.34 Let a non-autonomous system h(X, T1 , π), (Y, T2 , σ), hi be homo-
geneous, Y be compact, and the fiber bundle (X, h, Y ) be normed. Then the following
conditions are equivalent:

a. the trivial section of the fiber bundle (X, h, Y ) is uniformly asymptotically

stable;
b. there exit positive numbers N and ν such that

|π t x| ≤ N e−νt |x| (2.129)

for all x ∈ X and t ≥ 0.

Proof. To prove the theorem it is sufficient to establish the implication a. ⇒ b.,

since the converse assertion is obvious.
Note, that from the uniform asymptotic stability of the trivial section Θ of
the fiber bundle (X, h, Y ) and the homogeneity of the non-autonomous system
h(X, T1 , π), (Y, T2 , σ), hi it follows the equality X s = X. Now, to complete the
proof of the theorem it is sufficient to refer to Theorem 2.31. 
112 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Theorem 2.35 Let X be a Banach space. For an autonomous homogeneous (of

order k > 1) dynamical system (X, R+ , π) the following conditions are equivalent:

1. the trivial motion of (X, R+ , π) is uniformly asymptotically stable;

2. there exit positive numbers α and β such that
1
|π t x| ≤ (α|x|1−k + βt)− k−1

for all t ≥ 0 and x ∈ X.

Proof. Let us show that under the conditions of the theorem assumption 1. implies
2. Let x ∈ X (x 6= 0). Then, in view of the homogeneity of order k > 1 of the
system (X, R+ , π), we have
x
|π(t, x)| = |x||π(t|x|k−1 , )| ≤ |x| sup |π(t|x|k−1 , y)| = |x|m(t|x|k−1 )
|x| |y|≤1

for all t ≥ 0 and x ∈ X, where

m(t) = sup |π(t, y)|.

|y|≤1

From the uniform asymptotic stability of the trivial motion of (X, R+ , π) and by
virtue of the homogeneity of order k > 1 of the system (X, R+ , π) we have X s =
X. From Theorem 2.29 it follows the boundedness of the function m, but from
the uniform asymptotic stability of the trivial motion of (X, R+ , π) it follows that
m(t) → 0 as t → +∞. In addition, note that

m(t + τ ) = sup |π(t + τ, y)| = sup |π(τ, π(t, y))|

|y|≤1 |y|≤1
  
 π(t, y) 

= m(t) sup π τ m k−1
(t), ≤ m(t)m(τ mk−1 (t))
|y|≤1 m(t) 

for all t, τ ≥ 0. According to Lemma 2.19, there exit positive numbers M , a and b
such that
1
m(t) ≤ M (a + bt)− k−1

for all t ≥ 0. Let us put α = M 1−k a and β = M 1−k b, then

1 1
|π(t, x)| ≤ |x|m(t|x|k−1 ) ≤ |x|M (a + bt|x|k−1 )− k−1 = (α|x|1−k + βt)− k−1

for all x ∈ X and t ≥ 0. To complete the proof of the theorem it is sufficient to

refer to Lemma 2.18. 

Remark 2.12 All results of this section hold for dynamical systems with homo-
geneous phase spaces (see Remark 2.11 b.).
 Non-autonomous Dissipative Dynamical Systems 113

2.11 Linear systems

Let (X, h, Y ) be a local trivial fiber bundle with the fiber E and h(X, T1 , π),
(Y, T2 , σ), hi be a linear non-autonomous dynamical system, i.e. the mapping
π t : Xy → Xσt y is linear for all y ∈ Y and t ∈ T1 . From the results of Chap-
ter 2 (section 10)follow the following statements.

Theorem 2.36 The following statements are equivalent:

(1) the dynamical system h(X, T1 , π), (Y, T2 , σ), hi is pointwise dissipative;
(2) X s = X.

Theorem 2.37 Let Y be compact. Then the following statements are equivalent:

(1) the dynamical system h(X, T1 , π), (Y, T2 , σ), hi is compactly dissipative;
(2) X s = X and the trivial section θ of the fiber bundle (X, h, Y ) is uniformly
stable, that is for all ε > 0 there exits δ(ε) > 0 such that |x| < δ implies
|xt| < ε for all t ≥ 0;
(3) if the fiber bundle (X, h, Y ) is normed, then there exits a positive number N
such that

|xt| ≤ N |x|

for all x ∈ X, t ≥ 0 and X s = X .

Theorem 2.38 Let Y be compact. Then the following statements are equivalent:

(1) the dynamical system h(X, T1 , π), (Y, T2 , σ), hi is locally dissipative;
(2) X s = X and the trivial section Θ of the fiber bundle (X, h, Y ) is uniformly
attracting, i.e. there exits γ > 0 such that

lim sup |π(t, x)| = 0;

t→+∞ |x|≤γ

(3) if the fiber bundle (X, h, Y ) is normed, then there exit positive numbers N and
ν such that

|π(t, x)| ≤ N e−νt |x| (2.130)

for all x ∈ X, t ≥ 0.

Remark 2.13 a. Using Banach-Steinhaus theorem, it is easy to check that for

linear autonomous or periodic system the notions of pointwise and compact dissi-
pativity are equivalent.
b. In the example 1.8 the linear autonomous dynamical system defined on the
Hilbert space X = L2 [0, 1] by the differential equation x0 = Ax with bounded operator
A is compactly dissipative but not locally dissipative.
114 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Theorem 2.39 Let h(X, T, π), (Y, T, σ), hi be a group (i.e. T = R or Z) linear
non-autonomous dynamical system, Y be compact, (X, h, Y ) be a normed fiber bun-
dle and let exist positive numbers M and a such that |π t x| ≤ M ea|t| |x| (x ∈ X,
t ∈ T). Then the following statements are equivalent:

1. h(X, T, π), (Y, T, σ), hi is locally dissipative, that is there exit N , ν> 0 such that
|xt| ≤ N e−νt |x| for all t ≥ 0 and x ∈ X;
2. there exits a function V : X → R+ possessing the following properties:
2.a. V is some norm on (X, h, Y );
2.b. there exit positive numbers α and β such that α|x| ≤ V (x) ≤ β|x| for all
x ∈ X;
2.c. V̇π (x) = −|x| (x ∈ X), where V̇π (x) := lim t−1 [V (xt) − V (x)] for T = R
t↓0
and V̇π (x) := V (π(1, x)) − V (x) for T = Z.

Proof. Let h(X, T, π), (Y, T, σ), hi be locally dissipative. Define a function V :
X → R+ by the formula
Z +∞
V (x) = |xt|dt (x ∈ X), (2.131)
0

if T = R. Directly from the equality (2.131) it follows that V is some norm on

(X, h, Y ). We will establish some properties of the function V .

1 N
a. |x| ≤ V (x) ≤ |x| (x ∈ X).
Ma ν
In fact,
Z +∞ Z +∞
N
V (x) = |xt|dt ≤ N e−νt |x|dt = |x|.
0 0 ν

On the other hand, |x| = |(xt)(−t)| ≤ M eat |xt| (x ∈ X, t ≥ 0) and, hence,

Z +∞ Z +∞
1 1
V (x) = |xt|dt ≥ e−at |x|dt = |x|.
0 M 0 Ma

b. V̇π (x) = −|x| (x ∈ X).

Indeed,
Z +∞ Z +∞
V (xt) − V (x) = |(xt)s|ds − |xs|ds
0 0
Z +∞ Z +∞ Z t
= |xs|ds |xs|ds = − |xs|ds
t 0 0
 Non-autonomous Dissipative Dynamical Systems 115

and, consequently,
Z t
lim t−1 [V (xt) − V (x)] = lim −t−1 |xs|ds = −|x| (x ∈ X).
t↓0 t↓0 0
P+∞
If T = Z, then we put V (x) = n=0 |xn| (x ∈ X). It is obvious that V defines
some norm on (X, h, Y ). It is easy to see that the function defined above possesses
the required properties. Thus, we established that 1. implies 2..
Now we will establish that 2. implies 1. Let x ∈ X and |x| 6= 0. Consider the
function ϕ(t) := V (xt). Since α|x| ≤ V (x) ≤ β|x| and V̇π (x) = −|x| (x ∈ X), then
we have
1 1
ϕ̇(t) = V̇π (xt) = −|xt| ≤ − V (xt) = − ϕ(t)
β β
and, hence,
1
ϕ(t) ≤ ϕ(0)e− β t (t ≥ 0). (2.132)

On the other hand,

1 1
V (xt) ≤ |xt| ≤ (x ∈ X, t ≥ 0). (2.133)
β αV (xt)

From the inequalities (2.132) and (2.133) it follows that

β − β1 t
|xt| ≤ e |x| (x ∈ X, t ≥ 0).
α
The theorem is proved. 

Definition 2.23 Let h(X, T, π), (Y, T, σ), hi be a linear non-autonomous

dynamical system. A non-autonomous dynamical system h(W,T,µ), (Z,T,λ),%i is
said to be linear non-homogeneous, generated by linear (homogeneous) dynamical
system h(X, T, π), (Y, T, σ), hi, if the following conditions hold:

1. there exits a homomorphism q of the dynamical system (Z, T, λ) onto (Y, T, σ);
2. the space Wy := (q ◦ρ)−1 (y) is affine for all y ∈ (q ◦%)(W ) ⊆ Y and the vectorial
space Xy = h−1 (y) is an associated space to Wy ([287, p.175]). The mapping
µt : Wy → Wσt y is affine and π t : Xy → Xσt y is its linear associated function
([287, p.179]), i.e. Xy = {w1 −w2 | w1 , w2 ∈ Wy } and µt w1 −µt w2 = π t (w1 −w2 )
for all w1 , w2 ∈ Wy and t ∈ T .

Remark 2.14 The definition of linear non-homogeneous system, associated by

the given linear system, is given in the work [33], but our definition is more general
and sometimes more flexible.
116 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Lemma 2.20 Let h(W, T, µ), (Z, T, λ), %i be a linear non-homogeneous non-
autonomous dynamical system and (Z, T, λ) be compactly dissipative. Then the
following statements are equivalent:

1. h(W, T, µ), (Z, T, λ), %i is compactly dissipative;

2. h(W, T, µ), (Z, T, λ), %i is convergent;

Proof. Let h(W, T, µ), (Z, T, λ), %i be compactly dissipative and JW be the Levinson
center of (W, T, µ). We will show that JW ∩ Wz contains at most one point for all
z ∈ JZ . If we suppose that it is not true, then there exit z0 ∈ JZ and w1 , w2 ∈
JW ∩ Wz0 such that w1 6= w2 . It is easy to see that w = w2 + λ(w1 − w2 ) =
(1 − λ)w2 + λw1 ∈ JW ∩ Wz0 for all λ ∈ R, but this contradicts the compactness of
JW . Thus, 1. implies 2. The reverse implication is obvious. The lemma is proved. 

Theorem 2.40 Let Γ(Z, W ) 6= ∅ and h(X, S+ , π), (Y, S, σ), hi be a linear ho-
mogeneous dynamical system, h(W, S+ , µ), (Z, S, λ), %i be a linear nonhomogeneous
dynamical system generated by h(X, S+ , π), (Y, S, σ), hi and q be a homomorphism
from (Z, S, λ) onto (Y, S, σ). If the spaces Y and Z are compact and (X, h, Y ) is a
normed fiber bundle, then the dynamical system h(W, S+ , µ), (Z, S, λ), ρi is locally
dissipative if and only if the dynamical system h(X, S+ , π), (Y, S, σ), hi is locally dis-
sipative.

Proof. Let h(W, S+ , µ), (Z, S, λ), ρi be locally dissipative and JW be Levinson center
of (W, S+ , µ). According to Lemma 2.20, the set JW ∩ Wz contains a single point
for all z ∈ JZ . Denote this point by wz . Since (W, S+ , µ) is locally dissipative, then
there exits γ > 0 such that

lim ρ(µt B(JW , γ), JW ) = 0. (2.134)

t→+∞

We will show that the equality (2.134) implies

lim ρ(µt w, µt wz ) = 0 (2.135)

t→+∞

for all z ∈ JZ and w ∈ B(wz , γ). Moreover, the equality (2.135) holds uniformly
in z and w. If we suppose that it is not true, then there exit ε0 > 0, tn → +∞,
zn ∈ JZ and xn ∈ B(wzn , γ) (h(xn ) = zn ) such that

ρ(µtn wn , µtn wzn ) ≥ ε0 . (2.136)

From the equality (2.135) and the compactness of Z it follows that we can sup-
pose that the sequences {µtn wn }, {zn tn } and {µtn wzn } convergent. We put
z = lim zn tn and w = lim µtn wn , then lim µtn wzn = wz . Passing to limit
t→+∞ t→+∞ t→+∞
in (2.136) as n → +∞, we obtain ρ(w, wz ) ≥ ε0 . This contradicts the relation
w ∈ Wz ∩ JW . The obtained contradiction proves the equality (2.136) .
 Non-autonomous Dissipative Dynamical Systems 117

Note, that the equality (2.135) implies

lim sup |xt| = 0,

t→+∞ |x|≤γ

if we remark that xt = µt w1 − µt w2 + µt wz − µt wz , where z = h(x), |x| ≤ γ and

x = w1 − w2 . From the equality (2.136) and Theorem 2.38 it follows the local
dissipativity of the dynamical system h(X, S+ , π), (Y, S, σ), hi.
Conversely. Let h(X, S+ , π), (Y, S, σ), hi be locally dissipative, then according to
Theorem 2.38 there exit positive numbers N and ν such that the equality (2.130)
holds. Denote by Γ(Z, W ) the space of all continuous sections of the homomorphism
% : Z → W endowed with the distance

d(ϕ, ψ) = max ρ(ϕ(z), ψ(z)) = max |ϕ(z) − ψ(z)|,

z∈Z z∈Z

which converts it in the complete metric space. Define the mapping S t : Γ(Z, W ) →
Γ(Z, W ) by the equality (S t ϕ)(z) := µt ϕ(λ−t z) (z ∈ Z, ϕ ∈ Γ(Z, W )) for all t ∈ S.
It is easy to check that the family of mappings {S t : t ∈ S+ } forms a commutative
group with respect to composition. Note, that

d(S t ϕ1 , S t ϕ2 ) = max |µt ϕ1 (λ−t z) − µt ϕ2 (λ−t z)|

z∈Z
= max |µ ϕ1 (z) − µ ϕ2 (z)| = max |π t (ϕ1 (z) − ϕ2 (z))| ≤
t t
z∈Z z∈Z
N e−νt max |ϕ1 (z) − ϕ2 (z)|. (2.137)
z∈Z

From the equality (2.137) it follows that

d(S t ϕ1 , S t ϕ2 ) ≤ N e−νt max |ϕ1 (z) − ϕ2 (z)|

z∈Z

for all ϕ1 , ϕ2 ∈ Γ(Z, W ) and t ≥ 0 and, consequently, the mapping S t is contraction

for the sufficient large t. From this it follows that the commutative semi-group {S t :
t ∈ S+ } of mappings has a unique fixed point ϕ ∈ Γ(Z, W ), i.e. (S t ϕ)(z) = ϕ(z)
for all z ∈ Z and t ∈ S+ and, consequently, µt ϕ(z) = ϕ(λt z) (z ∈ Z, t ∈ S). Let
w ∈ W and z = ρ(w), then

|µt w − µt ϕ(z)| = |π t (w − ϕ(z))| ≤ N e−νt |w − ϕ(z)| . (2.138)

From (2.138) it follows that the compactly invariant set J = ϕ(Z) ⊂ W is globally
uniformly asymptotically stable, and in addition, JZ = J ∩ WZ = {ϕ(z)} for all
z ∈ Z. According to Theorem 2.14, the dynamical system h(W, S, µ), (Z, S, λ), %i
is convergent and, consequently, it is compactly dissipative and J = ϕ(Z) is the
Levinson center of (W, S, µ). To prove that (W, S, µ) is locally dissipative it is
sufficient to show that the set J = ϕ(Z) is uniformly contracting. We will show
118 Global Attractors of Non-autonomous Dissipative Dynamical Systems

that the equality

lim β(µt B(J, γ), J) = 0.

t→+∞

holds for all γ > 0. If we suppose that it is not so, then there exit ε0 , γ0 , w0 ∈
B(J, γ0 ) and tn → +∞ such that

ρ(µtn wn , J) ≥ ε0 . (2.139)

From the inequality (2.138) and the compactness of Z it follows that the sequence
{µtn wn } can be considered convergent. Put w = lim µtn wn , then w ∈ J. Passing
t→+∞
to limit in (2.139) as n → +∞, we have w 6∈ J. The obtained contradiction proves
the theorem. 
 Chapter 3

Analytic dissipative systems

3.1 Skew-product dynamical systems and cocycles

Let Y be a compact metric space and W = E n . Consider the cocycle

hE n , ϕ, (Y, T2 , σ)i over the dynamical system (Y, T2 , σ) with the fiber E n and the
non-autonomous dynamical system h(X, T1 , π), (Y, T2 , σ), hi associated by it. Then
(X, h, Y ), where X := E n × Y , is a finite-dimensional vectorial fibering (a trivial
fibering with the fiber E n ). The function V : X × X → R+ defined by the equality
V ((u1 , y), (u2 , y)) := |u1 −u2 |, where |·| is a norm on E n defines on X a Riemannian
metric.
From Theorem 2.23 follows the following statement.

Theorem 3.1 Let Y be a compact and hE n , ϕ, (Y, T2 , σ)i be a cocycle over the
dynamical system (Y, T2 , σ) with the fiber E n . Then the following statements are
equivalent:

1. There exists a positive number R such that

lim sup |ϕ(t, u, y)| < R

t→+∞

n
for all u ∈ E ” y ∈ Y .
2. There is a positive number r1 such that for all u ∈ E n and y ∈ Y there exists
τ = τ (u, y) > 0 for which |ϕ(τ, u, y)| < r1 .
3. There is a positive number r2 such that

lim inf |ϕ(t, u, y)| < r2

t→+∞

for all u ∈ E n and y ∈ Y .

4. There exists a positive number R0 and for all R > 0 there is l(R) > 0 such that
|ϕ(t, u, y)| ≤ R0 for all t ≥ l(R), u ∈ E n , |u| ≤ R and y ∈ Y .

Taking into consideration Lemma 2.14 we obtain that every condition 1.-4. that
figures in Theorem 3.1 is equivalent to the dissipativity of the non-autonomous

119
120 Global Attractors of Non-autonomous Dissipative Dynamical Systems

dynamical system h(X, T1 , π), (Y, T2 , σ), hi associated by the cocycle hE n , ϕ,

(Y, T2 , σ)i over (Y, T2 , σ) with the fiber E n .
We will give below an example of a skew-product dynamical system which plays
an important role in the study of non-autonomous differential equations.

Example 3.1 Let E n be an n-dimensional real or complex Euclidean space. Con-

sider the differential equation

u0 = f (t, u), (3.1)

where f ∈ C(R × E n , E n ). Along with the equation (3.1) we consider its H-class
[32],[137], [238], [300],[302], i.e. the family of the equations

v 0 = g(t, v), (3.2)

where g ∈ H(f ) = {fτ : τ ∈ R} and fτ (t, u) = f (t + τ, u), where the bar indicating
closure in the compact-open topology. We will suppose that the function f is
regular, that is for every equation (3.2) the conditions of existence, uniqueness and
extendability on R+ are fulfilled. Denote by ϕ(·, v, g) the solution of (3.2) passing
through the point v ∈ E n for t = 0. Then the mapping ϕ : R+ × E n × H(f ) → E n
satisfies the following conditions (see, for example, [32],[290],[291]):

1) ϕ(0, v, g) = v for all v ∈ E n and g ∈ H(f );

2) ϕ(t, ϕ(τ, v, g), gτ ) = ϕ(t + τ, v, g) for each v ∈ E n , g ∈ H(f ) and t, τ ∈ R+ ;
3) ϕ : R+ × E n × H(f ) → E n is continuous.

Denote by Y := H(f ) and (Y, R+ , σ) a dynamical system of translations on

Y , induced by the dynamical system of translations (C(R × E n , E n ), R, σ). The
triple hE n , ϕ, (Y, R+ , σ)i is a cocycle over (Y, R+ , σ) with the fiber E n . Hence,
the equation (3.1) generates a cocycle hE n , ϕ, (Y, R+ , σ)i and the non-autonomous
dynamical system h(X, R+ , π), (Y, R+ , σ), hi, where X := E n × Y , π := (ϕ, σ) and
h := pr2 : X → Y .

Definition 3.1 Recall that the equation (3.1) is called dissipative [137],
[271],[325],[326], if for all t0 ∈ R and x0 ∈ E n there exists a unique solution
x(t; x0 , t0 ) of the equation (3.1) passing through the point (x0 , t0 ) and if there exists
a number R > 0 such that lim sup |x(t; x0 , t0 )| < R for all x0 ∈ E n and t0 ∈ R.
t→+∞
In other words, for every solution x(t; x0 , t0 ) there is an instant t1 = t0 + l(t0 , x0 ),
such that |x(t; x0 , t0 )| < R for all t ≥ t1 . If the number l(t0 , x0 ) can be chosen
independent on t0 , then the equation (3.1) is called uniformly dissipative [137].

Below we will establish the relation between the dissipativity of the equation
(3.1) and the dissipativity of the non-autonomous dynamical system generated by
the equation (3.1).
 Analytic dissipative systems 121

Lemma 3.1 Let f ∈ C(R × E n , E n ) be regular and H(f ) be compact. Then the
following affirmations are equivalent:

(1) Equation (3.1) is uniformly dissipative; that is, there are numbers R > 0 and
l(x0 ) > 0 such that

|x(t; x0 , t0 )| < R (t ≥ t0 + l(x0 ), x0 ∈ E n ). (3.3)

(2) There is a positive number r such that

lim sup |ϕ(t, x0 , g)| < r (x0 ∈ E n , g ∈ H(f )). (3.4)

t→+∞

Proof. Let x0 ∈ E n and g ∈ H(f ). If g = fτ for some τ ∈ R. Then

ϕ(t, x0 , fτ ) = x(t + τ, x0 , τ ) (3.5)

and the inequality (3.4) follows from (3.3). For this aim it is sufficient to take r > R.
Consider the case when g = lim fτk and τk → +∞ (or −∞). By Theorem 3.2
k→+∞
from [186] follows that the sequence {ϕ(t, x0 , fτk )} converges to a certain solution
of (3.2) and this solution passes through the point (x0 , 0) ∈ E n × R. Note, that the
solution passing through the point (x0 , 0) is unique because the function f is regular.
Thus, ϕ(t, x0 , g) = lim ϕ(t, x0 , fτk ) uniformly with respect to t on every compact
k→+∞
from R+ . From (3.3) and (3.5) follows that |ϕ(t, x0 , g)| = lim |ϕ(t, x0 , fτk )| ≤ R
k→+∞
(t ≥ l(x0 )) and, consequently, the inequality (3.4) holds.
Conversely. Suppose that the inequality (3.4) holds. Then the non-autonomous
dynamical system generated by the equation (3.1) is dissipative and by Theorem
3.1 (the condition 4.) there is R0 > 0 such that for all x0 ∈ E n there exists
l(x0 ) = l(|x0 |) > 0 for which

|ϕ(t, x0 , g) ≤ R0 (3.6)

for all t ≥ l(x0 ) and g ∈ H(f ). Note that

x(t; x0 , t0 ) = ϕ(t − t0 , x0 , fto ), (3.7)

and taking into account (3.6) we obtain (3.3). For this aim it is sufficient to take
R > R0 . The lemma is proved. 
Thus, for the equation (3.1) (f is regular and H(f ) is compact) we established
that it is uniformly dissipative if and only if the non-autonomous dynamical system
generated by this equation is dissipative.
We note that the notions of the dissipativity and uniform dissipativity for the
equation (3.1) with periodic right hand side are equivalent, but for the equations
with almost periodic right hand side they are different. This assertion can be
illustrated by the following example.
122 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Example 3.2 Consider the linear scalar equation

x0 = a(t)x, (3.8)

where the function a ∈ C(R, R) is almost periodic such that all the solutions of
the equation (3.8) tend to 0 as t → +∞, but the trivial solution of (3.8) is not
uniformly stable. The equations (3.8) with the indicated above properties exist
(see, for example [123] and also the example 13.2). It is clear that along with the
equation (3.8) all translations of the equation (3.8), that is

x0 = aτ (t)x,

where aτ (t) = a(t + τ ) (t, τ ∈ R), possess the same property. Therefore, from
the relation (3.7) follows that lim |x(t; x0 , t0 )| = 0 (x(t; x0 , t0 ) is the solution of
t→+∞
the equation (3.8) passing through the point (x0 , t0 ) ), i.e. the equation (3.8) is
dissipative.
We will prove that the equation (3.8) is not uniformly dissipative. In fact, if we
suppose that is not true, then by Lemma 3.1 the non-autonomous dynamical system
generated by the equation (3.8) will be dissipative. In virtue of Theorem 2.38 the
trivial solution of (3.8) is uniformly asymptotically stable. But this contradicts to
the choice of the function a. The desired example is constructed.

Taking into account the fact established above, we will call the equation (3.1)
dissipative if the non-autonomous dynamical system generated by this equation is
dissipative.
An important class of dissipative equations are equations with convergence [137],
[270].
The equation (3.1) possesses the property of convergence if it has a unique
bounded on R solution which is globally uniformly asymptotically stable. This
unique bounded solution is called a limit regime for the equation (3.1).
From the results of [8],[290],[291] follows that, if the equation (3.1) is convergent,
then every equation from the family (3.2) possesses the same property. Thus, the
non-autonomous dynamical system generated by the equation (3.1) is convergent.
On the other hand, it is easy to construct examples of the equation which is not
convergent, but the non-autonomous dynamical system associated by this equation
is convergent. Note, that it is possible that the equation (3.1) has the limit regime
which is not a solution of this equation, but if the right hand side of the equation
(3.1) is stable in the sense of Poisson in positive direction, then the phenomenon
indicated above is not true. That is in this case the equation (3.1) is convergent if
and only if the corresponding non-autonomous dynamical system generated by this
equation is convergent.
Below we will call the equation (3.1) convergent if the non-autonomous
dynamical system generated by this equation is convergent.
 Analytic dissipative systems 123

Example 3.3 Consider the system of differential equations

(
u0 = F (y, u)
(3.9)
y 0 = G(y),

where Y ⊆ E m , G ∈ C(Y, E n ) and F ∈ C(Y × E n , E n ). Suppose that for the

system (3.9) the conditions of the existence, uniqueness and extendability on R+
are fulfilled. Denote by (Y, R+ , σ) a dynamical system on Y generated by the second
equation of the system (3.9) and by ϕ(t, u, y) – the solution of the equation

u0 = F (σ(t, y), u) (3.10)

passing through the point u ∈ E n for t = 0. Then the mapping ϕ : R+ × E n × Y →

E n satisfies the conditions 1) and 2) from Example 3.1 and, consequently, the system
(3.9) generates a non-autonomous dynamical system h(X, R+ , π), (Y, R+ , σ), hi
(where X := E n × Y , π := (ϕ, σ) and h := pr2 : X → Y ).
We will give some generalization of the system (3.9). Namely, let (Y, R+ , σ) be
a dynamical system on the metric space Y . Consider the system
(
u0 = F (σ(y, t), u)
(3.11)
y ∈ Y,

where F ∈ C(Y × E n , E n ). Suppose that for the equation (3.10) the conditions
of the existence, uniqueness and extendability on R+ are fulfilled. The system
h(X, R+ , π), (Y, R+ , σ), hi, where X := E n × Y , π := (ϕ, σ), ϕ(·, x, y) is the solution
of (3.10) and h := pr2 : X → Y is a non-autonomous dynamical system generated
by the equation (3.11).

3.2 C-analytic systems

Definition 3.2 We will call that the non-autonomous dynamical system

h(X, T1 , π), (Y, T2 , σ), hi is C-analytic, if the following conditions are fulfilled:

a. X := Cn × Y and π((z, y), t) := (ϕ(t, z, y), σ(t, y)) for all t ∈ T and (z, y) ∈
Cn × Y i.e. the non-autonomous dynamical system h(X, T1 , π), (Y, T2 , σ), hi is
generated by the cocycle hCn , ϕ, (Y, T2 , σ)i over (Y, T2 , σ) with the fiber Cn ;
b. for each y ∈ Y and t ∈ T1 the mapping U (t, y) : Cn → Cn (U (t, y) := ϕ(t, ·, y))
is holomorphic.

It follows from the results above that, when Y is compact, all three types of dis-
sipation properties for C-analytic systems are equivalent. In addition a C-analytic
system is dissipative if and only if there is a positive number r (independent of both
z and y) such that lim sup |ϕ(t, z, y)| ≤ r, where | · | denotes the norm in Cn .
t→+∞
124 Global Attractors of Non-autonomous Dissipative Dynamical Systems

An important property of C-analytic dissipative systems is described in

Theorem 3.2 If Y is compact, then the C-analytic dissipative dynamical system

h(X, T1 , π), (Y, T2 , σ), hi is uniformly stable in the positive direction on compacts of
Cn , i.e., for arbitrary ε > 0 and r > 0, there is a δ = δ(ε, r) > 0 such that the
inequality |z1 − z2 | < δ (|z1 |, |z2 | ≤ r) implies that |U (t, y)(z1 ) − U (t, y)(z2 )| ≤ ε for
all y ∈ Y and t ∈ T+ .

Proof. We first prove that, for arbitrary a > 0, there is a compact Ka ⊂ Cn

such that U (t, y)B[0, a] ⊆ Ka for all t ≥ 0 and y ∈ Y . In fact, Theorem 3.1
implies that there is b > 0 such that, for arbitrary a > 0, there is la ≥ 0 for which
U (t, y)B[0, a] ⊆ B[0, b] for t ≥ la and y ∈ Y . Let Ka := ϕ([0, la ], B[0, a], Y )∪B[0, b];
then Ka is the required set.
Now let r > 0; then the foregoing implies that the family of holomorphic map-
pings {U (t, y)} of the space C n into itself is uniformly bounded (with respect to
t ≥ 0 and y ∈ Y ) on the ball B[0, r], and the generalized Vitali theorem [167]
implies that this family is equicontinuous in B[0, r], i.e. for arbitrary ε > 0 there is
δ(ε, r) > 0 such that |U (t, y)z1 − U (t, y)z2 | < ε for all (t, y) ∈ T+ × Y if |z1 − z2 | < δ
(|z1 |, |z2 | < r). This proves the theorem. 

Theorem 3.3 Let h(X, T1 , π), (Y, T2 , σ), hi be a C-analytic non-autonomous dis-
sipative dynamical system and let Y be a compact minimal set; then

(1) for each y ∈ Y the set Jy := J ∩ Xy (Xy := h−1 (y)) contains only the single
point xy , where J is the centre of Levinson of the system (X, T1 , π);
(2) for all ε > 0 there is δ(ε) such that the inequality ρ(x, xy ) < δ (x ∈ Xy ) implies
that ρ(xt, xyt ) < ε for t ≥ 0.
(3) lim ρ(xt, xh(x)t ) = 0 for x ∈ X, uniformly with respect to x on compacts from
t→+∞
X.

Proof. Since Y is compact, it follows from the preceding remark that the non-
autonomous system h(X, T1 , π), (Y, T2 , σ), hi is compact dissipative, and Theorem
3.2 implies that it is uniformly positively stable on compacts from X. Hence, if
Y is a compact minimal set, then all conditions of Theorem 2.4 are satisfied, and
applying this theorem in the case under consideration we obtain assertions 2. and
3. To complete the proof of the theorem it remains to establish 1.
Thus let y ∈ Y . According to Theorem 2.4, the set Jy is connected because
Xy := Cn × {y} (y ∈ Y ) is connected. In proving Theorem 2.4 we established that
the semigroup Py contains an idempotent element u such that Jy = u(Xy ), and
that there are tk → +∞ for which u = lim π tk . Let x ∈ Xy , that is, x := (z, y)
k→+∞
(z ∈ Cn ); then u(x) = u(z, y) = (ξy (z), y), where ξy (z) := lim U (tk , y)z and this
k→+∞
 Analytic dissipative systems 125

relation holds uniformly on compacts in Cn . Since the mappings U (tk , y) : Cn → Cn

(k = 1, 2, . . .) are holomorphisms, the same is true of their limit ξy . In proving
Theorem 2.4 we showed that the idempotent element u acts on Jy in the same
fashion as the identity mapping, i.e. u(x) = x (x ∈ Jy ); hence ξy (z) = z for all z
such that (z, y) ∈ Jy . Thus, if (z, y) ∈ Jy , then z ∈ Cn is a fixed point of ξy . Since
ξy is a holomorphism, it has at most a finite number of fixed points in the interior
of each bounded set. It follows that Jy contains a finite number of points, and its
connectivity implies that it contains only one point xy . This proves the theorem.

Corollary 3.1 If Y is compact and minimal, then C-analytic non-autonomous

dynamical system h(X, T1 , π), (Y, T2 , σ), hi is convergent.

Corollary 3.2 Under the conditions of Theorem 3.3 the dynamical system
(X, T1 , π) induces on the centre of Levinson J of the system (X, T1 , π) a dynamical
system (J, T2 , π
b ) which is isomorphic to (Y, T2 , σ). In particular, the point xy is
recurrent (almost periodic, quasi-periodic, periodic) if y ∈ Y is recurrent (almost
periodic, quasi-periodic, periodic).

Proof. This follows from Theorem 3.3 (see also Corollary 3.1). It is sufficient to
note that the mapping p : Y → J, defined by the relation p(y) := xy (Jy = {xy }),
is a homomorphism of (Y, T2 , σ) onto (J, T2 , π
b). 
The general results obtained in this section can be used in the study of the
dissipative differential equations with the recurrent (with respect to time t) right
side if it is holomorph with respect to spatial variable z. Before formulate the
respectively results we will give the following example

Example 3.4 Denote by A(R × Cn , Cn ) the set of all continuous with respect to
t ∈ R and holomorphic with respect to z ∈ Cn functions f : R × Cn → Cn , equipped
with the topology of convergence uniform on the compacts from R × Cn . Consider
the equation (3.1) with right side f from A(R × Cn , Cn ). Along with equation (3.1)
we consider its H - class, i.e. the family of equations

ż = g(t, z) (g ∈ H(f )), (3.12)

where H(f ) = {fτ : τ ∈ R}, fτ (t, z) = f (t + τ, z) and the bar indicating closure in
the space A(R × Cn , Cn ). Let ϕ(·, z, g) be the solution of (3.12) passing through
the point z for t = 0 and defined on R+ .
Note the following properties of the mapping ϕ : R+ × Cn × H(f ) → Cn
a. ϕ(0, z, g) = z (z ∈ Cn , g ∈ H(f ));
b. ϕ(τ, ϕ(t, z, g), gt ) = ϕ(t + τ, z, g) (τ ∈ R+ , t ∈ R, z ∈ Cn and g ∈ H(f ));
c. the mapping ϕ : R+ × Cn × H(f ) → Cn is continuous and, for fixed t ∈ R+ and
g ∈ H(f ) the mapping U (t, g) := ϕ(t, ·, g) : Cn → Cn is holomorphic [122].
126 Global Attractors of Non-autonomous Dissipative Dynamical Systems

We write (Y, R, σ) (Y := H(f )) for the dynamical system of translations on H(f )

and define on X := Cn × Y the semi-group dynamical system (X, R+ , π) as follows:
π(τ, (z, g)) := (ϕ(τ, z, g), gτ ). Put h := pr2 : X → Y (h is a homomorphism of
(X, R+ , π) onto (Y, R, σ)). Equation (3.12) therefore generates the non-autonomous
dynamical system h(X, R+ , π), (Y, R, σ), hi. Property c. of ϕ implies that the non-
autonomous dynamical system constructed is C-analytic and application to it of
Theorem 3.3 yields

Theorem 3.4 If f : R × Cn → Cn is analytic in z ∈ Cn and continuous in t ∈ R

and f recurrent with respect to t ∈ R uniformly with respect to z on compacts from
Cn (in particular, it is almost periodic or periodic), then the dissipativity of equation
(3.12) implies that it is convergent, i.e. (3.12) has a unique solution p : R → C n
, bounded on R and uniformly asymptotically stable in the large; moreover, p is
recurrent (almost periodic or periodic) if f has this property.

Proof. This is a direct consequence of Theorem 3.3 Corollaries 3.1 and 3.2. It should
be kept in mind that the recurrence of f means that the set H(f ) is compact and
minimal with respect to the dynamical system of translations on H(f ). 

Remark 3.1 The ”local” variant of Theorem 3.3 (and so of Theorem 3.4) holds
too, i.e., Theorem 3.4 remains in force when Cn is replaced everywhere by a domain
W ⊆ Cn (W can be both bounded and unbounded and can coincide with Cn ). Hence
Theorem 3.4 (see also Corollary 3.2) implies that the autonomous equation ż = f (z)
cannot have limit cycles in W ⊆ C if f is holomorphic on W .

Definition 3.3 h(X, T1 , π), (Y, T2 , σ), hi is said to be C-analytic on the set M ⊆
X, if M is invariant and the system h(M, T1 , π), (Ye , T2 , σ), hi is C-analytic, where
Ye := h(M ), but (M, T1 , π) is a restriction (X, T1 , π) on M (analogous is defined
(Ye , T2 , π)).

Theorem 3.5 Suppose that the following conditions are fulfilled:

1. (Y, T2 , σ) is compact dissipative and its Levinson’s centre JY is a minimal set;

2. (X, T1 , π) is compact dissipative and JX is its Levinson’s centre;
3. h(X, T1 , π), (Y, T2 , σ), hi is C-analytic on JX .

Then the set JX ∩Xy contains only one point for every y ∈ JY and, consequently,
the system h(X, T1 , π), (Y, T2 , σ), hi is convergent.

Proof. Since JX is nonempty, compact and JY is minimal, then h(JX ) = JY . From

the C-analyticity of h(X, T1 , π), (Y, T2 , σ), hi on JY follows, according to Theorem
3.3, that JX ∩ Xy contains only one point for any y ∈ JY . Thus a non-autonomous
dynamical system h(X, T1 , π), (Y, T2 , σ), hi is convergent. 
 Analytic dissipative systems 127

Corollary 3.3 Suppose that for certain y0 ∈ Y H + (y0 ) = Y is compact and ωy0
is minimal. If the conditions 2. and 3. of Theorem 3.5 are fulfilled, then the system
h(X, T1 , π), (Y, T2 , σ), hi is convergent.

Proof. This assertion follows from Theorem 3.4. In fact, under the conditions of
the Corollary 3.3 the system (Y, T2 , σ) is compact dissipative and JY = ωy0 . 

Corollary 3.4 Let y0 be asymptotic recurrent (i.e. there is a recurrent point

q0 ∈ Y such that lim ρ(y0 t, q0 t) = 0). If Y := H + (y0 ) and the conditions of
t→+∞
Theorem 3.5 are fulfilled, then the system h(X, T1 , π), (Y, T2 , σ), hi is convergent
and each point x ∈ X is asymptotic recurrent.

Proof. The first statement of Corollary follows from Theorem 3.5 and Corollary
3.3. Let now x ∈ Xy and y ∈ Y = H + (y0 ). Since the point y0 is asymptotic
recurrent, then the point y is so asymptotic recurrent, i.e. there is a recurrent
point q ∈ ωy0 such that ρ(yt, qt) → 0 as t → +∞. Taking into account that
h(X, T1 , π), (Y, T2 , σ), hi is convergent, then JX ∩ Xq consists from the single point
p. Since in our conditions ρ(xt, pt) → 0 as t → +∞, then the point x is asymptotic
recurrent. 
Consider the equation

ż = f (t, z) + r(t, z) (z ∈ C). (3.13)

It follows directly from Theorem 3.5 and Corollary 3.4 the following theorem.

Theorem 3.6 Let f, r ∈ C(R × Cn , Cn ) and the following conditions are fulfilled:
(1) equation (3.13) is dissipative;
(2) lim |r(t, z)| = 0 uniformly in z on compacts from Cn ;
t→+∞
(3) f is holomorphic in z ∈ Cn and recurrent (almost periodic, periodic) in t ∈ R
uniformly in z on the compacts from Cn .
Then the equation

ż = f (t, z)

has a unique bounded on R solution p ∈ C(R, Cn ), which is recurrent (almost peri-

odic, periodic) in t ∈ R and lim |ϕ(t) − p(t)| = 0 for all solution ϕ of (3.13).
t→+∞

In the conclusion of this section we will give the following

Example 3.5 Consider the system of differential equations

(
ż1 = −2z1 + z22
(z1 , z2 ∈ C). (3.14)
ż2 = −z2
128 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Dynamical system defined by this system is C-analytic and dissipative. In ad-

dition, from [7, p.167-174] follows that the system (3.14) can not linearized by
biholomorphic transformation.
Thus, there are the nonlinear C-analytic dissipative systems, which can not
linearized by biholomorphic transformation.

Remark 3.2 For real analytic system the assertion analogous to example 3.5 does
not hold, as the example below shows.

Example 3.6 Consider the system of the differential equations

(
ẋ = (1 − x2 − y 2 )x
(3.15)
ý = (1 − x2 − y 2 )y

It is possible to check by directly integration that the system (3.15) is dissipative

and its Levinson’s centre J = {(x, y) : x2 + y 2 ≤ 1} contains a continuum points.
Thus , the analogous of Theorem 3.3 for real analytic system is not true.

3.3 Converse of Lyapunov’s theorem for C-analytic systems

Let R (C) be the set of all real (complex) numbers and Cn be n-dimensional complex
space with norm kzk := max{|zi | : i = 1, ..., n}. Denote by A(R × W, Cn ) the
space of all continuous functions f : R × W → Cn which are holomorphic in the
second variable, where W is a domain in Cn containing 0 ∈ Cn .
We consider the differential equation
dz
= f (t, z), (3.16)
dt
where f ∈ A(R × Cn , Cn ) and f (t, 0) = 0. along with (3.16) we consider the
variational equation for the zero solution
du
= A(t)u, (3.17)
dt
∂f
in which A(t) := (t, 0). We denote by z(t; t0 , v) the solution of (3.16) passing
∂t
through v ∈ Cn for t = t0 .

Definition 3.4 We recall (see, for example, [273]) that the solution z = 0 of
(3.16) is said to be stable if for any ε > 0 and t0 ∈ R there exists a δ = δ(ε, t0 ) > 0
such that for all z ∈ B(0, δ) = {w| kwk < δ} and t ≥ t0 one has kz(t; t0 , v)k ≤ ε.
If one can choose δ(ε, t0 ) independent of t0 ∈ R, then the solution z = 0 is called
uniformly stable.
 Analytic dissipative systems 129

Definition 3.5 The solution z = 0 of (3.16) is said to be attracting if for any

t0 ∈ R one can find an η = η(t0 ) > 0 and for any ε > 0 and kvk < η one can find a
σ = σ(t0 , ε, v) > 0 such that kz(t; t0 , v)k < ε for all t ≥ t0 + σ. If one can choose η
independent of t0 and σ, but dependent only on ε, then the solution z = 0 is called
uniformly attracting.

Definition 3.6 The solution z = 0 is said to be asymptotically stable (uniformly

asymptotically stable) if it is stable (uniformly stable) and is attracting (uniformly
attracting).

Theorem 3.7 The following assertions hold:

(1) if the zero solution of (3.16) is stable (uniformly stable), then the zero solution
of (3.17) is also stable (uniformly stable);
(2) if the zero solution of (3.16) is attracting (uniformly attracting), then so is the
zero solution of (3.17).

Proof. Let the solution z = 0 of (3.16) be stable (uniformly stable), ε > 0, t0 ∈ R

and δ = δ(ε, t0 ) > 0 (δ = δ(ε) > 0) be a positive number from the stability (uniform
stability) condition. By the theorem on analytic dependency on the initial data [122]
the function z(t; t0 , v) is holomorphic in v for fixed t and t0 in the domain of the
definition of z(t; t0 , v). We compute the linear part of the Taylor expansion at the
point v = 0 of the function z(t; t0 , v). According to the variation of parameters
formula we have
Zt
z(t; t0 , v) = U (t, t0 )v + U (t, τ )F (τ, z(τ ; t0 , v))dτ,
t0

−1
where U (t, τ ) := U (t, A)U (τ, A), U (t, A) being the Cauchy operator of (3.17) and
F (t, z) := f (t, z) − A(t)z.
Rt
We note that z(t; t0 , v) = v + f (τ, z(τ ; t0 , v))dτ and hence
t0

 Zt  
∂z  ∂f  ∂z 

(t; t0 , v) =E+ 
(τ, z(τ ; t0 , v)) (τ ; t0 , v) dτ,
∂v v=0 ∂z v=0 ∂v v=0
t0

where E is the identity matrix. Thus, V (t, t0 ) = U (t, t0 ) satisfies the matrix
differential equation ∂V ∂t = A(t)V and the initial condition V (t0 , t0 ) = E so
V (t, t0 ) = U (t, t0 ). Since
 t 
Z Zt
∂  ∂F
U (t, τ )F (τ, z(τ ; t0 , v))dτ  = U (t, τ ) (τ, 0) · U (τ, t0 )dτ = 0,
∂v ∂z
t0 v=0 t0
130 Global Attractors of Non-autonomous Dissipative Dynamical Systems

the linear part of the Taylor expansion in v of z(t; t0 , v) for any t0 ∈ R and t ≥ t0
is U (t; t0 )v. On the other hand, this linear part can be computed from the Cauchy
integral formula [207]. Since in Cn the l∞ norm is chosen, the domain B(0, δ) is an
open polydisk and one can write
Z Z n
X
1 z(t; t0 , v) wk
U (t, t0 )w = · . . . · × dv1 · . . . · dvn . (3.18)
(2πi)n v1 · v2 · . . . · vn vk
k=1
|v1 |= 2δ |vn |= 2δ

The stability (uniform stability) of the zero solution of (3.17) follows directly from
(3.18) if the zero solution of (3.16) is stable (uniformly stable).
If the zero solution of (3.16) is attracting (uniformly attracting), then passing to
the limit in (3.18) and considering Lebesgue’s theorem on passage to the limit under
the integral sign as t → +∞ we get the analogous property for the zero solution of
(3.17). The theorem is proved. 

Corollary 3.5 If the zero solution of (3.16) is asymptotically stable (uniform

asymptotically stable), then so is the zero solution of (3.17).

Remark 3.3 1. If (3.16) is autonomous, Theorem 3.7 is a result of Lyubich

[241].
2. For a real analytic system the assertion analogous to Theorem 3.7, does not
hold, as the example ẏ = −y 3 shows.
3. Theorem 3.7 is true also for the difference equations.

Along with equation (3.16) we consider its H-class

dz
= g(t, z) (g ∈ H(f )). (3.19)
dt
Let Θ := {(0, g)|g ∈ H(f )} ⊆ Cn × H(f ), W s (Θ) := {(v, g)|(v, g) ∈ Cn ×
H(f ), lim kϕ(t, v, g)k = 0} and h(X, R+ , π), (Y, R, σ), hi be a non-autonomous
t→+∞
dynamical system generated by the equation (3.16) (see the example 3.1). If the
zero solution of (3.16) is uniform asymptotically stable and the set H(f ) is compact,
then the compact invariant set Θ ⊆ X = Cn × H(f ) is asymptotically stable and
its domain of attraction W s (Θ) is open. The following theorem holds.

Theorem 3.8 Let f ∈ A(R × Cn , Cn ), H(f ) be compact and the zero solution of
(3.16) be uniform asymptotically stable, then the set W s (Θ) is unbounded.

Proof. Let as assume that W s (Θ) is bounded, then the set M = W s (Θ) ⊆ X is
compact invariant subset (X, R+ , π). In virtue of Theorem 3.7 the zero solution of
(3.17) is also uniform asymptotically stable and hence one can find positive numbers
N and ν such that

kU (t, t0 )k ≤ N e−ν(t−t0 ) (3.20)

 Analytic dissipative systems 131

for all t ≥ t0 (t, t0 ∈ R). It follows from (3.20) that

kU (t, A)k ≥ N −1 e−νt (3.21)

for all t ≤ 0.
On the other hand, it follows from (3.18) and the boundedness of W s (Θ) that

sup kU (t, A)k < +∞,

t∈R

which contradicts (3.21). The theorem is proved. 

Corollary 3.6 Let f ∈ A(R × Cn , Cn ), H(f ) be compact and the zero solu-
tion of (3.16) be uniform asymptotically stable. Then the set Wfs (0) = {v | v ∈
Cn , lim kϕ(t, v, f )k = 0} is unbounded.
t→+∞

Let Ω be a metric space and (Ω, R, σ) be a dynamical system on Ω. Consider

the differential equation

ẋ = f (ωt, x) (3.22)

where ωt := σ(t, ω), f ∈ A(Ω × Cn , Cn ), x ∈ Cn and f (ω, 0) = 0 for all ω ∈ Ω. The

results presented above are true also for the equation (3.22). We will formulated
some of them.

Theorem 3.9 If the zero solution of (3.22) is uniform asymptotically stable, then
the zero solution of

ẏ = A(ωt)y, (3.23)

∂f (ω, x) 
where A(ω) := , is also uniform asymptotically stable.
∂x x=0

Theorem 3.10 Let the zero solution of (3.22) be uniform asymptotically stable
and Ω be compact. Then the set

W s (Θ) = {(x, ω) ∈ Cn × Ω | lim |ϕ(t, x, ω)| = 0}

t→+∞

is unbounded, where ϕ(t, x, ω) is the solution of (3.22) passing through the point
x ∈ Cn for t = 0 and defined on R+ .

Proof. These statements can be proved as Theorem 3.7 and 3.8. 

132 Global Attractors of Non-autonomous Dissipative Dynamical Systems

3.4 On the structure of compact attracting sets of C-analytic systems

The present section is devoted to the structure of compact invariant sets of the
system
dz
= f (t, z) (3.24)
dt
with right-hand side f holomorphic with respect to z and almost periodic with re-
spect to t. We show that an asymptotically stable compact invariant set of system
(3.24) consists of almost periodic solutions to this system. This problem is studied
and solved in the frame of general non-autonomous dynamical systems. We give
also the description of the structure of uniform asymptotically stable compact in-
variant set of arbitrary C–analytic non-autonomous dynamical system with compact
minimal base.
We recall that the set K ⊆ X is called orbitally stable with respect to non-
autonomous system (2.1), if for any ε > 0 there exist δ(ε) > 0 such that the
inequality ρ(x, Ky ) < δ (x ∈ X, y = h(x)) implies ρ(xt, Kyt ) < ε for all t ≥ 0.
In addition, if there exist γ > 0 such that ρ(xt, Kyt ) → 0 for t → +∞ and every
x ∈ Ky such that ρ(x, Ky ) ≤ γ then it is said that K orbitally asymptotically stable.

Theorem 3.11 Assume that h(X, S+ , π), (Y, S, σ), hi is a non-autonomous C-

analytic dynamical system and the following conditions are satisfied:

(1) Y is compact and locally connected and (Y, S, σ) is minimal, i.e., Y = H(y) =
{yt | t ∈ S} for all y ∈ Y ;
(2) M ⊂ X is nonempty, compact, orbitally asymptotically stable, and invariant
with respect to h(X, S+ , π), (Y, S, σ), hi;
(3) M orbitally asymptotically stable.

Then h(M, S, π), (Y, S, σ), hi is a distal covering of finite multiplicity, i.e.,

(a) for any y ∈ Y the set My = {m | m ∈ M, h(m) = y} consists of finitely many

points;
(b) inf ρ(x1 t, x2 t) > 0 for all x1 , x2 ∈ M (x1 6= x2 , h(x1 ) = h(x2 )).
t∈S

Proof. Since the space Y is locally connected and the bundle (X, h, Y ) is locally
trivial, the space X is locally connected as well. By Proposition 1.18 [33], the set
M has finitely many connected components M1 , M2 , . . . , Mp . We write

Wys (M ) := {x | x ∈ Xy = h−1 (y), lim ρ(xt, Myt ) = 0}

t→+∞

and note that Wys (M ) is open in Xy . By theorem 1.36, for any compact set K ⊂
W s (M ) := ∪{Wys (M ) | y ∈ Y }, the set Σ+ (K) := ∪{π t K | t ≥ 0} is relatively
compact.
 Analytic dissipative systems 133

We set Ki = W s (Mi ) ∩ K and note that Ki (i ∈ 1, p) are closed, Ki ∩ Kj = ∅,

if i 6= j, and K = K1 ∪ K2 ∪ . . . Kp . Let γ > 0 (γ < α = min ρ(Ki , Kj )) and
i6=j
l > 0 such that we have B(M, γ), B(K, γ) ⊂ W s (M ) and |xt| ≤ l for all t ≥ 0
and x ∈ B(K, γ). By the Cauchy integral formula [167] there exists a number
L = L(l, γ) > 0 such that

ρ(π t x1 , π t x2 ) ≤ Lρ(x1 , x2 ) (3.25)

for all t ≥ 0 and x1 , x2 ∈ Ki (i ∈ {1, 2, . . . , p}, h(x1 ) = h(x2 )).

Let us now show that for any ε > 0 and any compact set K ⊂ W s (M ), there
exists δ(ε, K) > 0 such that ρ(x1 , x2 ) < δ (x1 , x2 ∈ K, h(x1 ) = h(x2 )) the relation
implies

ρ(x1 t, x2 t) < ε (3.26)

for all t ≥ 0. If we suppose the contrary, then there exist ε0 , δn ↓ 0, K0 ⊂

W s (M ), {xin } ⊂ K0 (i = 1, 2; h(x1n ) = h(x2n )) and tn → +∞ such that

ρ(x1n , x2n ) < δn and ρ(x1n tn , x2n tn ) ≥ ε0 . (3.27)

Since K0 is compact we can assume that the sequences {xin } (i = 1, 2) are conver-
gent. Let x0 := lim x1n = lim x2n and x0 ∈ Ky0 := K ∩ Xy0 . Hence Wys0 (M ) =
n→+∞ n→+∞
p
S
Wys0 (Mi ), there exists i0 ∈ {1, 2, . . . , p} such that x0 ∈ Wys0 (Mi0 ). Since
i=1
Wys0 (Mi0 ) is open set in Xy , there exists δ0 > 0 such that By0 (x0 , δ0 ) := {x | x ∈
ε0
Xy0 , ρ(x, x0 ) < δ0 } ⊂ Wys0 (Mi0 ). For sufficiently large n we have ρ(x1n , x2n ) < 2L .
i
Without loss of generality we can assume that {xn } ⊂ Ki0 (i = 1, 2); then by (3.25)
we have the estimate
ε0
ρ(x1n tn , x2n tn ) ≤ Lρ(x1n , x2n ) < .
2
which contradicts (3.27). The obtained contradiction completes the proof of our
statement.
Now we will prove the first statement of the theorem. Note that it follows from
inequality (3.25) implies that an invariant set M is distal in the negative direction
with respect to h. Now, since Y is minimal, it follows from Lemma 2.8 [32] (see
also Lemma 1 [238, p.104] ) that M is two-sided distal.
Let α > 0, y ∈ Y . We set Eyα := {π t |Wys (M ) | t ≥ α}, where the bar denotes
the closure in the compact-open topology, and Py := {ξ | ξ ∈ Eyα , ξ(Wys (M )) ⊆
Wys (M )}. As in Lemma 2.7, we can establish that Py is a nonempty compact
topological semigroup. By Lemma 4.11 [32], there exists an idempotent u ∈ Py ,
i.e., u2 = u. Since u ∈ Py ⊆ Eyα and Y is minimal, there exists tn → +∞
such that u = lim π tn |Wys (M ) and hence u(Wys (M )) ⊆ My . Indeed, we have
n→+∞
My = u(Wys (M )), because M is two-sided distal and the idempotent u actcs on My
134 Global Attractors of Non-autonomous Dissipative Dynamical Systems

as the identity mapping. Now let x ∈ Wys (M ) and p = u(x) ∈ My . Let us show
that

lim ρ(xt, pt) = 0. (3.28)

t→+∞

Indeed, since u(p) = u(u(x)) = u(x), such that ρ(xtn , ptn ) → 0 as n → +∞. Let
ε > 0, K = H + (x) = {xt | t ≥ 0} and δ = δ(ε, K) > 0 be such that (3.26) holds for
ρ(x1 , x2 ) < δ (x1 , , x2 ∈ K and h(x1 ) = h(x2 )). Choose n0 such that ρ(xtn , ptn ) < δ
for all n ≥ n0 ; then we have

ρ(x(tn + t), p(tn + t)) < ε

for all t ≥ 0 and thus relation (3.28) holds.

Consider the open covering {B(p, γ) | p ∈ My } of the set My . Since My is com-
pact, this covering contains a finite sub-covering {B(pi , γ)| i = 1, m}. Note that
u(B(pi , γ)) ⊆ My and pi ∈ u(B(pi , γ)). Since My ⊆ ∪{B(pi , γ) | i ∈ 1, m}, we have
my = u(My ) ⊆ ∪{u(B(pi , γ))|i ∈ 1, m} ⊆ My , and therefore ∪{u(B(pi , γ))|i =
1, m} = My . Since u is holomorphic and u(x) = x for all x ∈ My , the set
My is analytic. Since My is compact, it consists of finitely many points [167].
Let My = {x1 , x2 , . . . , xm(y) }. Consider the non-autonomous dynamical system
h(M, S, π), (Y, S, σ), hi. As proved above, this system is two-sided distal. By Propo-
sition 4 [238], the mapping H : Y → 2M (H(y) := My ) is continuous in the Haus-
dorff metric, and therefore cardMy = n(y) does not depend on y ∈ Y . Thus, the
theorem is completely proved. 

Corollary 3.7 Assume that the conditions of Theorem 3.11 are satisfied and that
Y consists of recurrent (Bohr almost periodic, quasi-periodic, periodic, stationary)
motions. Then the set M consists of recurrent (Bohr almost periodic, quasi-periodic,
periodic, stationary respectively) motions.

Proof. This assertion follows from Theorem 3.11, Theorem 3 [238] and Lemma
1[238]. 

Theorem 3.12 Under the conditions of Theorem 3.11 the following assertions
hold:

(1) the set M is uniform stable in the sense of Lyapunov, i.e. ∀ε > 0 ∃ δ(ε) > 0
such that ρ(x, p0 ) < δ (x ∈ Xy , p ∈ My ) implies ρ(xt, pt) < ε for all t ≥ 0;
(2) for each x ∈ W s (M ) there is p ∈ My (y := h(x)) such that

lim ρ(xt, pt) = 0.

t→+∞

In particular, if Y consists from the recurrent (almost periodic, quasi-periodic, pe-

riodic, stationary) points, then the set W s (M ) consists of asymptotically recurrent
 Analytic dissipative systems 135

(asymptotically almost periodic, asymptotically quasi-periodic, asymptotically peri-

odic, asymptotically stationary) points.

Proof. The first statement of Theorem follows directly from (3.26). The second
assertion follows from Theorem 3.11 and Lemma 4 [48]. 

Corollary 3.8 Let (Cn , S, π) be an autonomous C-analytic dynamical system,

M ⊂ C be a nonempty compact invariant orbitally asymptotically stable set, then
M consists of finitely many stationary points {x1 , x2 , . . . , xp }.

Let f ∈ A(R × Cn , Cn ). Consider the differential equation

dz
= f (t, z). (3.29)
dt
Along with equation (3.29), we consider its H-class
dz
= g(t, z) (g ∈ H(f )). (3.30)
dt
Definition 3.7 The set M ⊆ Cn is said to be invariant for equation (3.29), if for
any point v ∈ M there exists g ∈ H(f ), such that the solution ϕ(t, v, g) to equation
(3.30) is defined on the entire R and ϕ(t, v, g) ∈ M for all t ∈ R.

Definition 3.8 A compact set M ⊂ Cn invariant with respect to (3.29) is said

to be orbitally stable if for any ε > 0, there exists δ(ε) > 0 such that the relation
ρ(v, Mg ) < δ implies ρ(ϕ(t, v, g), Mgt ) < ε for all t ≥ 0, where Mg = {x | x ∈
M, ϕ(t, x, g) ∈ M for all t ∈ R} and gt - is a translation of g in first variable on t.

We write Wgs (M ) = {v | v ∈ Cn , lim ρ(ϕ(t, v, g), Mgt ) = 0} and W s (M ) =

t→+∞
∪{Wgs (M ) | g ∈ H(f )}.

Definition 3.9 A compact set M , invariant with respect to equation (3.29), is

said to be asymptotically orbitally stable if it is orbitally stable and there exists
γ > 0 such that B(M, γ) = {x | ρ(x, M ) < γ} ⊆ W s (M ).

Applying Theorem 3.11 and Corollary 3.7, 3.8 to the constructed non-
autonomous C-analytic system generated by equation (3.29), we obtain the following
statement.

Theorem 3.13 Let f ∈ A(R × Cn , Cn ) be regular and recurrent (Bohr almost

periodic, quasi-periodic, periodic, stationary) with respect to t ∈ R uniformly with
respect to z on compact sets from Cn , M ⊆ Cn and let the following conditions be
satisfied:

(1) M is a compact set invariant with respect to equation (3.29);

136 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(2) M is orbitally asymptotically stable,

Then for any g ∈ H(f ), the set Mg consists of finitely many recurrent (Bohr al-
most periodic, quasi-periodic, periodic, stationary respectively) solutions of equation
(3.30).

Remark 3.4 A statement similar to Theorem 3.13, is valid for difference equa-
tions as well.

3.5 Dynamical systems in spaces of sections

Let h(X, S+ , π), (Y, S, σ), h)i be a non-autonomous dynamical system. We recall
that a mapping γ : Y → X is called a section (selector) of the homomorphism h, if
h(γ(y)) = y for all y ∈ Y . The section γ of the homomorphism h is called invariant
if γ(σ(t, y)) = π(t, γ(y)) for all y ∈ Y and t ∈ S.
Denote by Γ = Γ(Y, X) family of all continuous sections of h, i.e. Γ(Y, X) =
{γ ∈ C(Y, X) : h ◦ γ = IdY }. We will suppose that Γ(Y, X) 6= ∅. This condition is
fulfilled in the many important case for the applications.
We endow the space Γ = Γ(Y, X) with the uniform-convergence on compacts in
Y.
Let K ⊆ Y be a nonempty compact from Y , then the relation

pK (γ1 , γ2 ) := max ρ(γ1 (y), γ2 (y))

y∈K

defines a pseudo-metric pK on Γ, and the family of pseudo-metrics {pK : K ∈

C(Y )}, where C(Y ) is the family of nonempty compact subsets of Y , defines a
uniform structure on Γ (a uniform-convergence topology on compacts from Y ).
S
∞
We consider the special case in which Y = Qk , where {Qk } is an expanding
k=1
sequence of nested compacts. In this situation it is sufficient to use a countable
number of pseudo-metrics P 0 = {pk : k = 1, 2 . . .}, where

pk (γ1 , γ2 ) := max ρ(γ1 (y), γ2 (y)).

y∈Qk

It can be easily verified that this family of pseudo-metrics separates the points of
Γ, and p1 (γ1 , γ2 ) ≤ p2 (γ1 , γ2 ) ≤ . . . pk (γ1 , γ2 ) ≤ . . .. Hence the topology defined by
this family of pseudo-metrics is metrizable. For example the topolgy defined by the
metric
∞
X 1 pk (γ1 , γ2 )
d(γ1 , γ2 ) := ,
1
2k 1 + pk (γ1 , γ2 )

is consistent with the topology of Γ defined by the family of pseudo-metrics P 0 .

 Analytic dissipative systems 137

Remark 3.5 a. It is easily shown that Γ is a complete space if and only if X is

a complete space.
b. If (X, h, Y ) is a vector fibering, then the section space Γ = Γ(Y, X) is naturally
endowed with a vector-space structure, and all operations are continuous with respect
to the topology defined on Γ, i.e., in this case Γ is covered into a topological vector
space.
c. If Y is compact and (X, h, Y ) is a Banach fibering, then the relation

||γ|| := max |γ(y)|

y∈Y

defines a norm on Γ (here |·| : X → R+ is a norm on (X, h, Y ), and consistent with

the metric) and the topology generated by this norm is consistent with the original
topology on Γ.

Let µ : S+ × Γ → Γ denote the mapping defined by the relation: µ(t, γ)(y) :=

π t γ(σ −t y) for all y ∈ Y . It is easily verified that µ(t, γ) ∈ Γ.
The following assertion holds.

Theorem 3.14 The triple (Γ, S+ , µ) is a (semigroup) dynamical system on Γ,

i.e.,

(1) µ(0, γ) = γ for all γ ∈ Γ.

(2) µ(τ, µ(t, γ)) = µ(t + τ, γ) for all γ ∈ Γ and t, τ ∈ S+ .
(3) µ : S+ × Γ → Γ is continuous.

Proof. The first two assertions can be established directly. The nontrivial part of
Theorem 3.14 is the continuity of µ. Let tν → t0 and γν → γ0 ({tν } (respectively
{γν }) is a directedness in S+ (respectively Γ)). We shall prove that µ(tν , γν ) →
µ(t0 , γ0 ) in the space Γ, i.e.,

lim max ρ(π tν γν (σ −tν y), π t0 γ0 (σ −t0 y)) = 0.

ν y∈K

for each compact K ⊆ Y . Assume the contrary: There is ε0 > 0 and a compact
K0 ⊆ Y such that

max ρ(π tν γν (σ −tν y), π t0 γ0 (σ −t0 y)) ≥ ε0 .

y∈K0

Since K0 is compact, there is {yν } ⊆ K0 for which

ρ(π tν γν (σ −tν yν ), π t0 γ0 (σ −t0 yν )) ≥ ε0 .

Again since K0 is compact, we assume that the sequence {yν } is convergent. Let
y0 := lim yν , then σ −tν yν → σ −t0 y0 and hence
ν

lim γ0 (σ −tν yν ) = γ0 (σ −t0 y0 ). (3.31)

ν
138 Global Attractors of Non-autonomous Dissipative Dynamical Systems

From (3.31) it follows that

lim π tν γ0 (σ −tν yν ) = π0t γ0 (σ0−t y0 ). (3.32)

ν

However

ε0 ≤ ρ(π tν γν (σ −tν yν ), π t0 γ0 (σ −t0 yν )) ≤ (3.33)

tν −tν t0 −t0
ρ(π γν (σ yν ), π γ0 (σ y0 ))
t0 −t0 t0 −t0
+ρ(π γ0 (σ y0 ), π γ0 (σ yν )).

The second term on the right here tend to zero when yν → y0 . We shall prove that
the first term also tends to zero yν → y0 and tν → t0 . At first we shall establish
that lim γν (σ −tν yν ) = γ0 (σ −t0 y0 ). In fact
ν

ρ(γν (σ −tν yν ), γ0 (σ −t0 y0 )) ≤ ρ(γν (σ −tν yν ), γ0 (σ −tν yν )) +

ρ(γ0 (σ −tν yν ), γ0 (σ −t0 y0 )) ≤ max0 ρ(γν (y), γ0 (y)) + (3.34)
y∈K

ρ(γ0 (σ −tν yν ), γ0 (σ −t0 y0 )),

where K 0 = σ(K0 , T0 ) and T0 = {−tν }. Letting n → +∞ in (3.34) and using (3.32)

and the fact that γν → γ0 in topology Γ, we obtain the required result.
Hence γν (σ −tν yν ) → γ0 (σ −t0 y0 ) and tν → t0 , and thus

ρ(π tν γν (σ −tν yν , π t0 γ0 (σ −t0 y0 )) → 0. (3.35)

Letting tν → t0 and yν → y0 in (3.33) and employing (3.35), we find that ε0 ≤ 0,

in contradiction with the choice of ε0 . The mapping µ : Γ × S+ → Γ is therefore
continuous, and this completes the proof of the theorem. 

Remark 3.6 a. If h(X, S, π), (Y, S, σ), hi is a group non-autonomous dynamical

system, then the group dynamical system (Γ, S, µ) is defined on Γ.
b. A continuous section γ ∈ Γ is invariant if and only if γ ∈ Γ is a stationary
point of the system (Γ, S+ , µ).

We consider a special case of the foregoing construction. Let hW, ϕ, (Y, S, σ)i
be a cocycle over (Y, S, σ) with fiber W and h(X, S+ , π), (Y, S, σ), hi be a non-
autonomous dynamical system generated by this cocycle. Then h ◦ γ = IdY and
because h = pr2 it follows: γ = (ψ, IdY ), where γ ∈ Γ(Y, X) and ψ : Y → W . Hence
to each section γ there corresponds a mapping ψ : Y → W and conversely. There
being a one-to-one relation between Γ(Y, W × Y ) and C(Y, W ), where C(Y, W ) is
the space of continuous functions ψ : Y → W, with the uniform-convergence topol-
ogy on compacts of Y , we identify these two objects from now on. The dynamical
system (Γ, S+ , µ) induces, in a natural fashion a dynamical system (C(Y, W ), S+ , Q)
 Analytic dissipative systems 139

on C(Y, W ). Namely

(µt γ)(y) = π t γ(σ −t y) = π t (ψ, IdY )(σ −t y) =

π t (ψ(σ −t y), σ −t y) = (U (t, σ −t y)ψ(σ −t y), y) = ((Qt ψ)(y), y).

Hence µt (ψ, IdY ) = (Qt ψ, IdY ), and so the mapping Q : C(Y, W ) × S+ → C(Y, W ),
defined by the relation Q(ψ, t) = Qt ψ, with (Qt ψ)(y) = U (t, σ −t y)ψ(σ −t y) (y ∈ Y ),
has the following properties:

a. Q0 = IdC(Y,W ) ;
b. Qt Qτ = Qt+τ ( t, τ ∈ S+ )
c. Q is continuous,

i.e., (C(Y, W ), S+ , Q) is a semigroup dynamical system on C(Y, W ). By virtue of

the foregoing, we identify the systems (Γ, S+ , µ) and (C(Y, W ), S+ , Q) in the sequel.

3.6 Quasi-periodic solutions

In this section we investigate dissipative differential equation (3.1) with quasi-

periodic coefficients, and prove that, if its right side is analytic, then there is always
at least one quasi-periodic solution with the same frequency basis as the right side.
Consider the system
(
ẋ = F (x, ϕ),
(3.36)
ϕ̇ = iωϕ,

where ω = (ω1 , ω2 , . . . , ωm ) ∈ Rm (ω1 , ω2 , . . . , ωm are linearly independent

with respect to the integers), i2 = −1, ϕ = (ϕ1 , ϕ2 , . . . , ϕm ) ∈ Cm , ωϕ =
col(w1 ϕ1 , . . . , ωm ϕm ), x ∈ Cm and F ∈ C(Cn × Cm , Cm ). We denote by T m
m-dimensional torus, i.e. T m := {ϕ ∈ Cm : |ϕk | = 1, k = 1, m}. We are interested
in the system (3.36), when the parameter ϕ runs over the points of the torus T m
and a small neighborhood

Dδ := {ϕ ∈ C m : |ϕk | < 1 + δ, k ∈ 1, m},

where δ > 0 is sufficiently small.

The system (3.36) is clearly equivalent to the equation

ẋ = F (x, ϕ0 eiωϕ ), (3.37)

where ϕ0 ∈ Cm , or to a subset (for example Dδ ).

Definition 3.10 We will say that the system (3.36) (or equation (3.37)) is an-
alytic on T m if there is a positive number δ, for which F : Cn × Dδ → Cn is
analytic.
140 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Let (Cm , R, σ) denote the dynamical system on Cm determined by the second

equation of the system (3.36). Then


σ(ϕ, t) := col ϕ1 eiω1 t , ϕ2 eiω2 t , . . . , ϕm eiωm t .

It is easily seen that, for arbitrary δ > 0 the sets Dδ and Dδ , where the bar indicates
the closure of Dδ in Cm , are invariant with respect to (Cm , R, σ).

Definition 3.11 The system (3.36) (and the equation (3.37)) is said to be dis-
sipative on T m , if there exists δ > 0 (sufficiently small) such that the dynamical
system (Cm , R, σ), generated by (3.36) ( ϕ ∈ D δ ) is dissipative.

We sharpen this definition. Suppose that, for the system (3.36), conditions
for existence, uniqueness, and continuability to the right are satisfied. Then it is
known that (3.36) determines a semigroup system (Cn × Dδ , R+ , π) on Cn × D δ ;
then π((x, ϕ0 ), t) := (ψ(t, x, ϕ0 ), ϕ0 eiωt ), where ψ(·, x, ϕ0 ) is the solution of equation
(3.37) passing through the point x ∈ Cn for t = 0. It is easily to verify that the triple
h(Cn × Dδ , R+ , π), (Dδ , R, σ), hi (h := pr2 : Cn × Dδ → Dδ ) is a non-autonomous
dynamical system. Hence, if the system of differential equations (3.36) is dissipative
on T m , then the non-autonomous dynamical system h(Cn ×Dδ , R+ , π), (D δ , R, σ), hi
generated by it ( for sufficiently small δ) is dissipative, i.e. there is a positive number
r for which

lim sup |ψ(t, x, ϕ)| < r

t→+∞

for x ∈ C and ϕ ∈ Dδ , and r is independent of both x ∈ Cn and ϕ ∈ Dδ .

n

Theorem 3.15 If the equation (3.37) is dissipative on T m and the right side of
(3.37) is analytic on T m , then the equation (3.37) has at least one quasi-periodic
solution with the same frequency basis ω1 , ω2 , . . . , ωm as the right side F .

Proof. If (3.37) is dissipative and F is analytic on T m , then there is δ > 0 (suf-

ficiently small) such that F : Cn × Dδ → Cn is analytic and the non-autonomous
dynamical system h(Cn × Dδ , R+ , π), (Dδ , R, σ), hi is dissipative. Let A(Dδ , Cn )
denote the set of all analytic functions γ : Dδ → Cn endowed with the uniform-
convergence topology on compacts from Dδ . Clearly A(Dδ , Cn ) is a closed subset
of the space C(Dδ , Cn ) of all continuous functions γ : Dδ → Cn with the uniform-
convergence topology on compacts from Dδ . Since Dδ is an invariant subset of
(Dδ , R, σ), the non-autonomous system h(Cm ×Dδ , R+ , π), (Dδ , R, σ), hi is correctly
defined and, by virtue of Remark 3.6, C(Dδ , Cn ) is the space of all continuous sec-
tions for this system. Theorem 3.14 implies that the triple (C(Dδ , Cn ), R+ , Q),
where (Qt γ)(ϕ) = ψ(t, γ(σ −t ϕ), σ −t ϕ), is a dynamical system on C(Dδ , Cn ).
We shall prove that A(Dδ , Cn ) is an invariant subset of the dynamical system
 Analytic dissipative systems 141

(C(Dδ , Cn ), R+ , Q). In fact, if γ ∈ A(Dδ , Cn ), then general properties [122] of

solutions of equations with analytic right sides imply that Qt γ ∈ A(Dδ , Cn ) for
all t ≥ 0. Hence a semigroup dynamical system (A(Dδ , Cn ), R+ , Q) is defined on
A(Dδ , Cn ). We note one property of this system following from the dissipativity
of equation (3.37) on T m . Let Ar := {γ ∈ A(Dδ , Cn ) : γ(Dδ ) ⊆ B(0, r)}, where
B(0, r) is the open ball in Cn of radius r centered at the origin. It follows from Vi-
tali’s theorem [167] that, for each r > 0, the sets Ar are precompact in A(Dδ , Cn ).
It is clear that the sets Ar are convex. Theorem 3.1 implies that there is b > 0 such
that, for each r > 0, there is lr ≥ 0 for which

Qt Ar ⊆ A b

for all t ≥ lr . Let r > b (for example r = b + 1), and U = IntAr (the interior of
Ar ), then Qt U ⊆ Ab ⊂ U . According to Theorem 1 [205] there exists a stationary
point γ ∈ A(Dδ , Cn ) of the dynamical system (A(Dδ , Cn ), R+ , Q), i.e.

ψ(t, γ(σ −t ϕ), σ −t ϕ) = γ(ϕ) (ϕ ∈ Dδ , t ≥ 0). (3.38)

Hence ψ(t, γ(ϕ), ϕ) = γ(σ t ϕ) (ϕ ∈ Dδ , t ≥ 0), that is, ψ(t, γ(ϕ0 ), ϕ0 ) =

γ(σ t ϕ0 ) = γ(ϕ0 eiωt ) is a quasi-periodic solution of (3.37) with the frequency basis
ω1 , ω2 , . . . , ωm . This proves the theorem. 

Remark 3.7 a. We note that the proof of Theorem 3.15 is of a general nature,
and it can be directly applied to equations of more general form. For example,
instead of (3.36) we could consider the system
(
ẋ = F (x, ϕ)
(x ∈ Cn , ϕ ∈ Cm ). (3.39)
ϕ̇ = G(ϕ)

We must assume that T m is a minimal quasi-periodic set for the second equation
of (3.39), and that there is a neighborhood Dδ of T m that is an invariant set of the
dynamical system generated by second equation of (3.39). The other conditions, in
Theorem 3.15, can clearly be modified for this case.
b. Theorem 3.15 (see also Theorem 3.4) established a conjecture of Bronstein
(see [32, p.278]). It should be added that the hypothesis was concerning equation
with almost periodic right sides, satisfying the uniform positive stability condition.
Hence the latter condition is superfluous for analytic systems.
c. In the author’s opinion, the importance of the condition that the right side F
be analytic is doubtful in Theorem 3.15. We are inclined to believe that Theorem
3.15 still applies when F is sufficiently smooth. In this connection we note two
examples in [152, 332]. One is an example of a dissipative equation of the form
(3.1) with quasi-periodic coefficients but with no almost-periodic solutions. Both in
142 Global Attractors of Non-autonomous Dissipative Dynamical Systems

[152, 332] the construction is based on Denjoy’s results concerning flows on a two-
dimensional torus, and analysis of the examples shows that, if the right hand side
of equation (3.1) is sufficiently smooth, then this situation does not prevail.

Theorem 3.16 Suppose that the conditions of Theorem 3.15 are fulfilled and
F (x, ϕ) = F (x, ϕ) (where by bar we note the complex conjugation). Then there
exists γ ∈ A(Dδ , Cn ) such that

(1) γ(x) = γ(x) for all x ∈ Dδ ;

(2) γ(ϕ0 eiωt ) = ψ(t, γ(ϕ0 ), ϕ0 ) for all ϕ0 ∈ Dδ and t ∈ R.

Proof. The proof of this statement uses the same argument as Theorem 3.15. 

3.7 The analogy of Cameron-Johnson’s theorem

The well-known Cameron-Johnson’s theorem [39], [204] asserts that the equation

x0 = A(t)x (3.40)

with a recurrent (almost periodic by Bhor) matrix may be reduced by the Lyapunov-
Perron transformation to the equation y 0 = B(t)y with a skew-symmetric matrix
B(t), if all solutions of the equation (3.40) and all its limit equations are bounded
on the whole real axis. The generalization of this result on the linear C-analytic
equations in the Hilbert space is our main scope in this section.
Let R = (−∞, +∞), Cm is an m-dimensional complex Euclidean space, G be
a domain from Cm , H be a real or complex Hilbert space with the scalar product
h·, ·i and with the norm | · |2 = h·, ·i. Denote by Hw a space H equipped with
weak topology, and by [H] ( respectively [Hw ] ) the family of all linear continuous
operators acting into H (respectively Hw ) and equipped with operational norm
(with weak topology). Assume that H(G, [H])) (resp. H(G, [Hw ]), H(G, Cm )) is the
family of all holomorphic functions h : G → [H] (resp. h : G → [Hw ], h : G → Cm )
equipped with compact-open topology.
Consider the system
( 0
x = A(z)x
(z ∈ G), (3.41)
z 0 = Φ(z)

where Φ ∈ H(G, Cm ) and A ∈ H(G, [H]). We suppose that the second equation of
the system (3.41) generates the dynamical system (G, R, σ) on G. Denote by U (t, z)
the Cauchy’s operator of the equation

x0 = A(zt)x (z ∈ G), (3.42)

 Analytic dissipative systems 143

where zt = σ(t, z). From the general property of solutions of differential equations
(see, for example, [122] and [132],[137]), it follows that the family of operators
{U (t, z)|t ∈ R, z ∈ G} satisfies the following conditions:

1. U (0, z) = I (∀z ∈ G), where I is a unit operator on H ;

2. U (t + τ, z) = U (t, zτ )U (τ, z) (∀t, τ ∈ R and z ∈ G );
3. the mapping U : R × G → [H] (U : (t, z) 7→ U (t, z)) is continuous and for
every t ∈ R the mapping U (t, ·) : G → [H] is holomorphic.

The following assertion holds.

Theorem 3.17 Suppose that there exists a positive constant C such that

kU (t, z)k ≤ C (3.43)

for all t ∈ R and z ∈ G. Then there exists P ∈ H(G, [H]) such that

a. P - is bi-holomorphic, i.e. the operator P (z) is invertible for all z ∈ G and the
mapping P −1 : G → [H] (P −1 : z 7→ P −1 (z)) is holomorphic;
b. P∗ (z) = P (z) for all z ∈ G (P∗ (z) is an adjoint for P (z) operator);
c. the operator P (z) is positively defined for all z ∈ G;
d. C −1 |x| ≤ |P (z)x| ≤ C|x| for all z ∈ G and x ∈ H;
e. the change of variables x = P (zt)y alters the equation (3.42) into

y 0 = B(zt)y (3.44)

with a skew-Hermitian operator B ∈ H(G, [H]), i.e. B∗ (z) = −B(z) for all z ∈ G.

Proof. Let S t : H(G, [H]) → H(G, [H]) ( resp. H(G, [Hw ]) → H(G, [Hw ])) be a
mapping defined by the equality (S t f )(z) = U∗ (t, z)f (zt)U (t, z) for all z ∈ G and
f ∈ H(G, [H]) (resp. f ∈ H(G, [Hw ])). It is easy to verify that the family of
mappings {S t |t ∈ R} satisfies the following conditions:

4. the operator S t maps H(G, [H]) (or f ∈ H(G, [Hw ])) into itself for each t ∈ R;
5. S 0 = I, where I is a unit mapping on H(G, [H]) ( resp. H(G, [Hw ]));
6. S t S τ = S t+τ for all t, τ ∈ R;
7. the mapping S t ( t ∈ R) is linear and continuous on H(G, [H]) ( resp.
H(G, [Hw ])).

From 4.-7. follows that the family of mappings {S t |t ∈ R} is a commutative group

of linear continuous operators on H(G, [H]) (or H(G, [Hw ])).
Denote by K := {A ∈ [H]| kAk ≤ C 2 } and note that the set K is weakly
closed. According to Tihonoff theorem, the set K is weakly compact because every
bounded and closed set in H is weakly compact. From Cauchy’s integral formula
(see, for example, [288, p.339]) and the Arzela-Ascoli theorem [288] follows that the
144 Global Attractors of Non-autonomous Dissipative Dynamical Systems

set Γ = {γ ∈ H(G, [H]| kγ(z)k ≤ C 2 } is compact in the space H(G, [Hw ]). We
suppose that V := conv{S t Q | t ∈ R}, where Q(z) = I(∀z ∈ G), I is a unit operator
in H and conv is a weak closure of the convex hull of {S t Q | t ∈ R}. Note, that the
set V is invariant, i.e. S t V ⊆ V for all t ∈ R because the family of operators {S t }
is linear and continuous in H(G, [Hw ]). It is easy to see that

hS t Q(z)x, xi = hU∗ (t, z)U (t, z)x, xi = |U (t, z)x|2 (3.45)

for all z ∈ G, t ∈ R and x ∈ H. From (3.43) and (3.45) follows that

C −2 |x|2 ≤ hS t Q(z)x, xi ≤ C 2 |x|2

and, consequently,

C −2 |x|2 ≤ hR(z)x, xi ≤ C 2 |x|2 (3.46)

for all R ∈ V, z ∈ G and x ∈ H. In addition, the equality R∗ (z) = R(z) holds

for all z ∈ G and R ∈ V. According to the equality (3.46) we have kR(z)k ≤ C 2
for all z ∈ G and, consequently, V ⊆ Γ. According to the theorem of Markov-
Kakutany [28], the group {S t | t ∈ R} admits in V at least one fix point, i.e. there
exists R ∈ V such that U∗ (t, z)R(zt)U (t, z) = R(z) for all z ∈ G and t ∈ R. We
note that the operator R(z) is self-adjoint and, according to the equality (3.46),
is positively defined. From Theorem 12.33 [274] (see also [182, p.65]) follows that
there exists a unique reversible self-adjoint and positively defined operator M (z)
such that M 2 (z) = R(z) for all z ∈ G and, according to the inequality (3.46), we
have

C −1 |x| ≤ |M (z)x| ≤ C|x|

for all z ∈ G and x ∈ H. Finally, according to the equality

I
1
Rα (z) = − λα (R(z) − λI)−1 dλ
2πi
L

(α = ± 12 ), where L is a simple contour, enclosing the spectrum of the operator

R(z), we have that the mapping M : G → [H] is biholomorhic.
Note, that M (zt)U (t, z)M −1 (z) = M −1 (zt)U∗ −1 (t, z)M (z) since

U∗ (t, z)M 2 (zt)U (t, z) = M 2 (z)

for all t ∈ R and z ∈ G. We set V(t, z) = M (zt)U (t, z)M −1 (z) and note that the
family of operators {V(t, z) | t ∈ R, z ∈ G} satisfies the conditions 1.-3. and in
addition:

8. V∗ (t, z) = V −1 (t, z) for all t ∈ R and z ∈ G;

9. U (t, z)P (z) = P (zt)V(t, z) for all t ∈ R and z ∈ G, where P (z) = M −1 (z).
 Analytic dissipative systems 145

Let
d
B(z) := V(t, z)|t=0 = [Ṁ (z) + M (z)A(z)]P (z),
dt
d dM (z)
where Ṁ (z) := dt M (zt)|t=0 = dz Φ(z). Then B ∈ H(G, [H]) and

dV(t, z)
= B(zt)V(t, z) (3.47)
dt
for all t ∈ R and z ∈ G. Thus, V(t, z) is a Cauchy operator of (3.44). On the other
hand, by the condition 8., we have
( −1
V 0 ∗ (t, z) = −B∗ (zt)V∗ −1 (t, z)
(3.48)
V∗ −1 (0, z) = I

for all t ∈ R and z ∈ G. From the equalities (3.47), (3.48) and condition 8. follows
that B∗ (z) = −B(z) for all z ∈ G. Finally, to finish the proof of the theorem it is
sufficient to remark that the change of variables x = P (zt)y reduces equation (3.42)
to the equation (3.44). The theorem is proved. 

Remark 3.8 a. In the case, when the space H is finite-dimensional, the statement
close to Theorem 3.17 was obtained by V. V. Glavan [165].
b. If the point z ∈ G is stationary ( ω-periodic, quasi-periodic, almost periodic
etc.), then according to [300], the operator-function P (zt) will also be stationary
(ω-periodic, quasi-periodic, almost periodic etc.).
c. Theorem 3.17 holds for the system of difference equations
(
x(k + 1) = A(zk)x(k)
(k ∈ Z),
z(k + 1) = Φ(zk)

where A ∈ H(G, [H]) and Φ ∈ H(G, Cm ) and also for the system of differential and
difference equations with multi-dimensional time.

3.8 Almost periodic solutions of the weak nonlinear dissipative

systems

1. In this item we will study bounded and uniformly compatible solutions of the
linear and nonlinear system of differential equations.
Denote by Cb (R, E n ) the set of all bounded on R functions ϕ : R → E n with
the sup-norm. Consider a linear homogeneous equation

ẋ = A(t)x, (3.49)
146 Global Attractors of Non-autonomous Dissipative Dynamical Systems

where A ∈ Cb (R, [E n ]). Along with the equation (3.49) we consider the non-
homogeneous equation

ẋ = A(t)x + f (t) . (3.50)

Definition 3.12 Recall, that the equation (3.49) is weakly regular if for every
function f ∈ Cb (R, E n ) the equation (3.50) has at least one solution ϕ ∈ Cb (R, E n ).

Let ϕ ∈ C(R, E) (respectively f ∈ C(×U, E)), denote by Mϕ (respectively Mf )

the family of all sequences {tn } such that the functional sequence {ϕtn } (respectively
{ftn }) converges in the space C(R, E) (respectively C(×U, E)), where ϕτ (t) :=
ϕ(t + τ ) for all t ∈ R (respectively fτ (t, x) := f (t + τ, x) for all (t, x) ∈ R × U ).

Definition 3.13 Recall [300, 302] that the solution ϕ ∈ C(R, E n ) of equation

x0 = f (t, x) (f ∈ C(R × E n , E n )) (3.51)

is called uniformly compatible if MfQ ∈ Mϕ , fQ is a restriction on R × Q of the

function f and Q := ϕ(R).

It is known [300, 302] that, if the function f is τ -periodic (quasi-periodic, almost

periodic, recurrent) w.r.t. variable t ∈ R uniformly w.r.t. x on the compacts from
E n and ϕ is a uniformly compatible solution of equation (3.51), then the function
ϕ is is τ -periodic (quasi-periodic, almost periodic, recurrent) too.

Theorem 3.18 Let A ∈ Cb (R, [E n ]). If the set H(A) is compact, then the follow-
ing conditions are equivalent:

1. the equation (3.49) is weakly regular;

2. the equation (3.50) has at least one bounded on R uniformly compatible solution
for every function f ∈ Cb (R, E n ).

Proof. It is easy to see that from 2. follows 1.. Therefore, to prove Theorem 3.18
it is sufficient to show, that the condition 1. implies 2. We will consider two cases:
a. Let A ∈ C(R, [E n ]) be a Poisson stable at least in one direction matrix, i.e.
there a sequence |tk | → +∞ such that Atk → A in C(R, [E n ]). In this case from
Theorem 3.18 and Lemma 4.2.1 [51] follows that the equation (3.49) satisfies the
condition of exponential dichotomy on R and, according to Theorem 22.23 [32], the
equation (3.50) has a unique bounded on R and uniformly compatible solution.
b. The matrix-function A ∈ C(R, [E n ]) is not stable in the sense of Poisson
neither in positive nor in negative direction. Let ϕ be a bounded on R solution
of the equation (3.50) and {tk } ∈ M(A,f ) . There is (B, g) ∈ H(A, f ) such that
Atk → B and ftk → g. Note, that in this case for the sequence {tk } it is possible
only the following two cases:
 Analytic dissipative systems 147

(1) the sequence {tk } converges to t0 ∈ R and, consequently, {tk } ∈ Mϕ ;

(2) |tk | → +∞.

Then the matrix-function B := lim Atk is ω-limit or α-limit for A and in this case
k→+∞
we have {tk } ∈ Mϕ . In fact, the sequence {ϕtk } is compact in C(R, E n ) because
the function ϕ is bounded. In addition, every limit function for the sequence {ϕtk }
is a bounded on the R solution of the equation

ẋ = B(t)x + g(t) . (3.52)

By Lemma 4.2.1 from [51] the equation

ẏ = B(t)y

satisfies the condition of the exponential dichotomy on R and, consequently, the

equation (3.52) cannot have more than one bounded on R solution. From this fact
follows that the sequence {ϕtk } converges in the space C(R, E n ), i.e. {tk } ∈ Mϕ .
Finally, we will shows that, in case the sequence {tk } is unbounded, it does not
contain bounded sub-sequences. If we suppose that it is not true, then there is a
0 0 0
subsequence {tk } ⊆ {tk } such that tk → t ∈ R and in this case we have
0
B := lim Atk = lim At0 = At0 (t ∈ R).
k→+∞ k→+∞ k

From this equality follows that the matrix-function A is stable in the sense of Poisson
at least in one direction, but this contradicts to our condition. The theorem is
proved. 
Denote by E0 := {u ∈ E n : sup{kϕ(t, u, A)k : t ∈ R} < +∞} and by P0 a
projection mapping the space E n onto E0 . The following statement holds.

Lemma 3.2 Let A ∈ C(R, [E n ]). If the matrix-function A is bounded on R and

the equation (3.49) is weakly regular, then for every f ∈ Cb (R, E n ) there exists
a unique solution ϕ ∈ Cb (R, E n ) of the equation (3.50) satisfying the following
conditions:

(1) P0 ϕ(0) = 0;
(2) there exists a positive constant K (a constant of the weak regularity of (3.49)),
which does not depend on the function f , such that kϕk 6 Kkf k.

Proof. We prove the formulated statement using the same arguments as in Lemma
6.3 [186, p.515]. 

Corollary 3.9 Let A ∈ C(R, [E n ]). If the set H(A) is compact, then the following
conditions are equivalent:

(1) the equation (3.49) is weakly regular;

148 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(2) for every function f ∈ Cb (R, E n ) there is a unique bounded on R uniformly

compatible solution ϕ of the equation (3.50) satisfying the following conditions:

a. P0 ϕ(0) = 0 and b. kϕk ≤ Kkf k,

where K is a constant of the weak regularity of the equation (3.49).

Theorem 3.19 Let A ∈ C(R, [E n ]) be a bounded on R matrix-function, the equa-

tion (3.49) be weakly regular and F ∈ C(R×E n , E n ) be a function with the following
property: |F (t, u)| ≤ c(|u|) (∀t ∈ R and u ∈ E n ), where c : R+ → R+ is a non-
decreasing function. If {r > 0 : Kc(r) ≤ r} 6= ∅, where K is a constant of the
weak regularity of the equation (3.49), then the equation

u̇ = A(t)u + F (t, u) (3.53)

has at least one bounded on R solution.

Proof. For every function f ∈ Cb (R, E n ) the equation (3.50) has a unique solution
ψ ∈ Cb (R, E n ) such that

P0 ψ(0) = 0 and kψk ≤ Kkf k , (3.54)

where K is a constant of the weak regularity of the equation (3.49). Let ϕ ∈

Cb (R, E n ). Consider a differential equation

u̇ = A(t)u + F (t, ϕ(t)) . (3.55)

Since the function f (t) := F (t, ϕ(t)) is bounded on the R (|f (t)| = |F (t, ϕ(t))| ≤
c(|ϕ(t)|) ≤ c(kϕk)), then the equation (3.55) has a unique bounded on R solution
ψϕ satisfying the condition (3.54) and, consequently,

kψϕ k ≤ Kkf k = K sup |F (t, ϕ(t))| ≤ Kc(kϕk) . (3.56)

t∈R

Let us define an operator Φ : Cb (R, E n ) → Cb (R, E n ) in the following way:

(Φϕ)(t) = ψϕ (t) (ϕ ∈ Cb (R, E n ), t ∈ R ). We will show that if the number r0 > 0
satisfies the inequality Kc(r0 ) ≤ r0 , then the ball B[0, r0 ] := {ϕ ∈ Cb (R, E n ) :
kϕk ≤ r0 } is invariant with respect to the mapping Φ. In fact, if ϕ ∈ B[0, r0 ], then
kΦϕk ≤ Kc(kϕk) ≤ Kc(r0 ) ≤ r0 . Consider now the space Cb (R, E n ) in the quality
of the subset embedded in the space C(R, E n ). First of all, we note that every ball
B[0, r] ⊂ Cb (R, E n ) is a convex, bounded and closed subset of the space C(R, E n ).
Note, that the mapping Φ : B[0, r0 ] → B[0, r] is continuous in the topology of
the space C(R, E n ). In fact, let {ϕk } ⊆ B[0, r] and ϕk → ϕ in C(R, E n ). Consider
the sequence (Φϕk )(t) := ψϕk (t ∈ R ). Note, that fk (t) := F (t, ϕk (t)) → f (t) :=
 Analytic dissipative systems 149

F (t, ϕ(t)) in the topology of the space C(R, E n ). If we suppose the contrary, then
there are ε0 > 0 and L0 > 0 such that

max |F (t, ϕk (t)) − F (t, ϕ(t))| ≥ ε0 .

|t|≤L0

Hence, there exists a sequence {tk } ⊆ [−L0 , L0 ] such that

|F (tk , ϕk (tk )) − F (tk , ϕ(tk ))| ≥ ε0 . (3.57)

Since the sequence {tk } is bounded, then we may suppose that it is convergent.
Note, that ϕk (tk ) → ϕ(t0 ), where t0 := lim tk . In fact,
k→+∞

|ϕk (tk ) − ϕ(t0 )| ≤ |ϕk (tk ) − ϕ(tk )| + |ϕ(tk ) − ϕ(t0 )|

≤ max |ϕk (t) − ϕ(t)| + |ϕ(tk ) − ϕ(t0 )| .
|t|≤L0

Passing to limit in this inequality, as k → +∞, we obtain the necessary state-

ment. From the inequality (3.57) (taking into account that the sequence {ϕk (tk )}
converges to ϕ(t0 ) as k → +∞) we obtain ε0 ≤ 0 and this contradicts to the
choice of the number ε0 . Thus, fk → f in the space C(R, E n ) and, in addition,
kfk k = sup |F (t, ϕk (t))| ≤ c(kϕk k) ≤ c(r0 ), i.e. |fk (t)| ≤ c(r0 ) (t ∈ R ) and, conse-
t∈R
quently, |f (t)| ≤ c(r0 ) (t ∈ R ). On the other hand, the function Φϕk is a bounded
on R solution of the equation

u̇ = A(t)u + fk (t) .

In addition, the functions {Φϕk } and their derivatives are uniformly bounded on R
and, consequently, the sequence {Φϕk } is relatively compact in the space C(R, E n ).
Since fk → f in C(R, E n ), then every limit function of the sequence {Φϕk } is
a bounded on R solution of the equation (3.50), satisfying the conditions (3.54).
But by Corollary 3.9 the equation (3.50) has at most one bounded on R solution
satisfying the conditions (3.54). From this fact follows that the sequence {Φϕ k } is
convergent in the space C(R, E n ) and the continuity of the mapping Φ : B[0, r0 ] →
B[0, r0 ] is established.
Now we will show that the mapping Φ : B[0, r0 ] → B[0, r0 ] is completely con-
tinuous in the topology of the space C(R, E n ). For this aim we note that

(Φϕ)0 (t) = A(t)(Φϕ)(t) + F (t, ϕ(t))

and, consequently,

|(Φϕ)0 (t)| ≤ a|(Φϕ)(t)| + |F (t, ϕ(t))| ≤ ar0 + c(r0 ) (t ∈ R) ,

where a = sup{kA(t)k : t ∈ R}. From this follows that the set Φ(B[0, r0 ]) is
relatively compact in the topology of the space C(R, E n ). According to the theorem
of Tihonoff-Shauder, the mapping Φ has at least one fixed point ϕ ∈ B[0, r0 ]. It is
150 Global Attractors of Non-autonomous Dissipative Dynamical Systems

easy to see that the function ϕ ∈ B[0, r0 ] is a bounded on R solution of the equation
(3.53). The theorem is proved. 

Remark 3.9 The statement close to Theorem 3.19 was proved in the work [257],
but under more strongly assumptions on the nonlinear perturbation F .

Theorem 3.20 Let A ∈ C(R, [E n ]), F ∈ C(R × E n , E n ) and the following con-
ditions be hold:

a. a = sup{kA(t)k : t ∈ R} < +∞;

b. the equation (3.49) is weakly regular;
c. |F (t, u)| ≤ c(|u|) ( t ∈ R, u ∈ E n ) and {r > 0 : Kc(r) ≤ r} 6= ∅, where
c : R+ → R+ is a non-decreasing function and K is a constant of the weak
regularity of the equation (3.49);
d. the restriction F0 of the function F on R × B[0, r0 ] satisfies the condition of
Lipschitz and Lip(F0 ) < K −1 , where r0 is some positive number satisfying the
inequality Kc(r0 ) ≤ r0 .

Then the equation (3.53) has at least one bounded on R and uniformly compatible
solution.

Proof. Denote by Br0 (M) := {ϕ ∈ Cb (R, E n ) : |ϕ(t)| ≤ r0 (t ∈ R )} and

M ⊆ Mϕ , where M := M(A,F ) . Br0 (M) is a closed subset of Cb (R, E n ) [300,
p.60]. The operator Φ : Cb (R, E n ) → Cb (R, E n ), defined as well as in the proof
of Theorem 3.19, maps Br0 (M) into itself. In fact, if ϕ ∈ Br0 (M), then the func-
tion f (t) := F (t, ϕ(t)) (t ∈ R) belongs also to Br0 (M). According to Corollary
3.9, Φϕ is a unique bounded on R uniformly compatible solution of the equation
(3.50), satisfying the conditions (3.54). In addition, as well as in Theorem 3.14,
Φ(B[0, r0 ]) ⊆ B[0, r0 ] and, consequently, Φ(Br0 (M)) ⊆ Br0 (M). Let ϕ1 and ϕ2 be
the function from Br0 (M). Note, that

(Φϕ1 − Φϕ2 )0 (t) = A(t)[(Φϕ1 − Φϕ2 )(t)] + F (t, ϕ1 (t)) − F (t, ϕ2 (t))

and, therefore,

kΦϕ1 − Φϕ2 k ≤ K sup |F (t, ϕ1 (t)) − F (t, ϕ2 (t))| ≤ KLip(F0 )kϕ1 − ϕ2 k .

t∈R

From the latter inequality follows that the mapping Φ : Br0 (M) → Br0 (M) is a
contraction, therefore it has a unique fixed point, which is a uniformly compatible
solution of the equation (3.53). The theorem is proved. 

Corollary 3.10 Under the conditions of Theorem 3.20 if the functions A ∈

C(R, [E n ]) and F0 = F |R×B[0,r0 ] are mutually τ -periodic (almost periodic, recur-
rent, stable in the sense of Poisson, asymptotically τ -periodic, asymptotically almost
 Analytic dissipative systems 151

periodic, asymptotically recurrent). Then the equation (3.53) admits at least one
τ -periodic (almost periodic, recurrent, stable in the sense of Poisson, asymptotically
τ -periodic, asymptotically almost periodic, asymptotically recurrent) solution.

Remark 3.10 The problem of existence of uniformly compatible solutions of weak

nonlinear equations was studied before in the works [32],[300]. The principal dif-
ference of Theorem 3.20 from the corresponding results [32],[300] consists of the
following: under the conditions of Theorem 3.20 and Corollary 3.10 the equation
(3.49) does not obligatory satisfy the condition of the exponential dichotomy on R.
In addition, the nonlinear term is not obligatory small with respect to the variable
u ∈ En.

2. In this item we will study uniformly compatible solutions of weak nonlinear

dissipative systems.

Theorem 3.21 Let A ∈ C(R, [E n ]), F ∈ C(R × E n , E n ) and the following condi-
tions be fulfilled:

(1) a = sup{kA(t)k : t ∈ R} < +∞;

(2) there are positive numbers N and ν such that

kU (t, A)U −1 (τ, A)k ≤ N e−ν(t−τ ) (t ≥ τ, t, τ ∈ R);

(3) |F (t, u)| ≤ M + ε|u| ( u ∈ E n , t ∈ R ) and 0 ≤ ε ≤ ε0 < ν 2 (N a)−1 .

Then the equation (3.53) is dissipative, i.e. there is a number R0 > 0 such that

lim sup |ϕ(t, ν, B, G)| < R0 ( ν ∈ E n , (B, G) ∈ H(A, F ) ) ,

t→+∞

where ϕ(·, ν, B, G) is a solution of the equation

ν̇ = B(t)ν + G(t, ν) , (3.58)

satisfying the initial condition ϕ(0, ν, B, G) = ν.

Proof. For all B ∈ H(A) we will define on E n a norm | · |B by the equality

Z
+∞

|u|B := |U (t, B)u|dt .

0

As well as in the proof of Theorem 2.39, it is possible to check that

1 N
|u| ≤ |u|B ≤ |u| ( u ∈ E n ).
a ν
152 Global Attractors of Non-autonomous Dissipative Dynamical Systems

We put
Z
+∞

u(t) := |ϕ(t, u, B, G)|Bt = |U (s, Bt )ϕ(t, u, B, G)|ds .

0

Since
Zt


ϕ(t, u, B, G) = U (t, B) u + U −1 (τ, B)G(τ, ϕ(τ, u, B, G))dτ ,
0

then
Z +∞ Z t
 
u(t) = |U (s, Bt )U (t, B) u + U −1 (τ, B)G(τ, ϕ(τ, u, B, G))dτ |ds
0 0
Z +∞ Zt
= |U (s + t, B)(u + U −1 (τ, B)G(τ, ϕ(τ, u, B, G))dτ )|ds
0
0
Z +∞ Z +∞ Z t
≤ |U (s + t, B)u|ds + | U (s + t, B)U −1 (τ, B) ×
0 0 0
N
G(τ, ϕ(τ, u, B, G))dτ )|ds ≤ e−νt |u| +
ν
Z +∞Z t
N e−ν(s+t−τ ) (M + ε|ϕ(τ, u, B, G)|)dτ ds
0 0
Zt
N N M τt

= e−νt |u| + e−νt (e − 1) + ε |ϕ(τ, u, B, G)|eντ dτ
ν ν ν
0
Z t
N NM N ε −νt
≤ e−νt |u| + 2 (1 − e−νt ) + e a|ϕ(τ, u, B, G)|Bτ eντ dτ
ν ν ν 0

Z t
N NM aN ε −νt
= e−νt |u| + 2 (1 − e−νt ) + e u(τ )eντ dτ. (3.59)
ν ν ν 0

Let ϕ(t) := u(t)eνt . From the inequality (3.59) follows that

Zt
N NM aεN
ϕ(t) ≤ |u| + 2 (eνt − 1) + ϕ(τ )dτ
ν ν ν
0

and by Theorem 9.3 from [132] ϕ(t) ≤ ψ(t) (t ∈ R), where ψ is a solution of the
integral equation
Zt
N NM aεN
y(t) = |u| + 2 (eνt − 1) + y(τ )dτ .
ν ν ν
0
 Analytic dissipative systems 153

Solving the latter equation, we find that

N N M  aεN t NM
ψ(t) = |u| + 2 e ν + 2 eνt
ν ν − aεN ν − aεN
and, consequently,
N N M  aεN t NM
u(t)eνt ≤ |u| + e ν + 2 eνt . (3.60)
ν ν 2 − aεN ν − aεN
From the inequalities (3.59) and (3.60) we obtain

|ϕ(t, u, B, G)| ≤ a|ϕ(t, u, B, G)|Bt

N N M  − ν 2 −aεN t aN M
≤a |u| + 2 e ν + 2 .
ν ν − aεN ν − aεN
Therefore,
aM N
lim sup |ϕ(t, u, B, G)| ≤
t→+∞ ν 2 − aεN
aεN
(u ∈ E n , (B, G) ∈ H(A, F ) and ν − ν > 0). The theorem is proved. 

Remark 3.11 Simple examples show that under the conditions of Theorem 3.22
dissipativity does not reduce to convergence. This statement can be illustrated by
the following example ẋ = −x + 2x(1 + x2 )−1 .

Theorem 3.22 Let A ∈ C(R, [E n ]), F ∈ C(R × E n , E n ) and the following con-
ditions be fulfilled:

(1) a := sup{ kA(t)k : t ∈ R} < +∞

(2) there exist positive numbers N and ν such that

kU (t, A)U −1 (τ, A)k ≤ N e−ν(t−ν) (t ≥ τ, t, τ ∈ R) .

2
ν ν
(3) |F (t, u)| ≤ M + ε|u| (u ∈ E n ) and 0 ≤ ε < ε0 := min( N , N a ), where M is
some positive number.
(4) the restriction F0 of the function F on R×B[0, r0 ], where r0 is a positive number
NM
appearing in Theorem 3.20 (for example, r0 = ν−ε 0N
), satisfies the condition
of Lipschitz with the Lipschitz constant Lip(F0 ) < N ν −1 .

Then the equation (3.52) is dissipative and in the ball B[0, r0 ] there exists a
unique bounded on R uniformly compatible solution of this equation.

Proof. The formulated assertion immediately follows from Theorems 3.20 and 3.21


Remark 3.12 The theorems 3.20 – 3.22 are true also for the differential equations
(3.53) in arbitrary Banach spaces.
154 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Corollary 3.11 If under the conditions of Theorem 3.22 the functions A and F0
are almost periodic, then the equation (3.53) is dissipative and its Levinson center
contains at least one almost periodic solution.

Remark 3.13 Under the conditions of Theorem 3.22 and Corollary 3.11, dissi-
pativity does not reduce to convergence.

We will give an example which confirms this statement.

Example 3.7 Consider the scalar equation

ẋ = −kx + αF (x) , (3.61)

where
(
x2
2 for |x| ≤ 10
F (x) :=
50 + 10[1 − exp(10 − |x|)] |x| > 10

for 0 < k ≤ 5α. In this connection we can take r0 = αk . It is easy to check that for
the equation (3.61) all the conditions of Theorem 3.22 are fulfilled for the chosen
α, k, and F . In addition, its Levinson center contains at least two fixed points
(in reality there are 3) and, consequently, the equation (3.61) is not convergent.
 Chapter 4

The structure of the Levinson center of

system with the condition of the
hyperbolicity

4.1 The chain recurrent motions

Let Σ ⊆ X be a compact invariant set, Σ, ε > 0 and t > 0.

Definition 4.1 The collection {x = x0 , x1 , x2 , . . . , xk = y; t0 , t1 , . . . , tk } of the

points xi ∈ Σ and the numbers ti ∈ T such that ti ≥ t and ρ(xi ti , xi+1 ) < ε (i =
0, 1, . . . , k − 1) is called a (ε, t, π)-chain joining the points x and y. We denote by
P (Σ) the set {(x, y) : x, y ∈ Σ, ∀ ε > 0 ∀ t > 0 ∃ (ε, t, π)-chain joining x and y}.

The relation P (Σ) is closed, invariant and transitive [33]. The point x ∈ Σ is
called chain recurrent, if (x, x) ∈ P (Σ). Let R(Σ) = {x ∈ Σ : (x, x) ∈ P (Σ)}. On
R(Σ) we will introduce a relation ∼ as follows: x ∼ y if and only if (x, y) ∈ P (Σ)
and (y, x) ∈ P (Σ). It is easy to check that the introduced relation ∼ on R(Σ) is a
relation of equivalence and, consequently, it is easy to decompose it on the classes
of equivalence {Rλ : λ ∈ L} i.e. R(Σ) = t{Rλ ∈ L}. By Proposal 2.6 from [33]
the defined above components of the decomposition of the set R(Σ) are closed and
invariant.

Lemma 4.1 If the compact invariant set Σ from X contains only a finite numbers
of minimal set, then the relation ∼ decomposes the set R(Σ) on the finite numbers
of different classes of equivalence.

Proof. Let R(Σ) = t{Rλ : λ ∈ L}. Since Rλ ⊆ Σ are closed and invariant and Σ is
compact, then by the theorem of Birkhoff the set Rλ contains at least one minimal
set. Consequently, the number of different classes of equivalence is not more that
the number of minimal sets. The lemma is proved. 

We will indicate below the condition when the number of classes of equivalence
{Rλ : λ ∈ L} is finite in case, if the set Σ contains infinite number of different
minimal sets.

155
156 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Definition 4.2 By analogy with the work [2] in the collection {R1 , R2 , . . . , Rk }
we will introduce a relation of partial order as follows: Ri < Rj , if there exist
T
i1 , i2 , . . . , ir such that i1 = i, ir = j and W s (Rip ) W u (Rip+1 ) 6= ∅ for all p =
1, 2, . . . , r − 1.

Definition 4.3 The ordered collection of r (r ≥ 2) different indexes {i1 , i2 , . . . , ir }

satisfying the condition Ri1 < Ri2 < · · · < Rir < Ri1 is called an r-
cycle in the collection {R1 , R2 . . . , Rk }. The 1-cycle is called such index i that
T
W s (Ri ) W u (Ri ) \ Ri 6= ∅.

Note, that the introduced above notion of partial order is a slight modification
of the corresponding notion from [33, p.61].

Definition 4.4 Following the works [2],[261],[271], the collection of points

{x1 , x2 , . . . , xn } ⊆ X (or K := H(x1 ) ∪ H(x2 ) ∪ · · · ∪ H(xn )) is said to be a
generalized homoclinic contour, if ωxi ∩ αxi+1 6= ∅ for all i = 1, 2, . . . , n, where
xn+1 = x1 .

Lemma 4.2 Let Σ ⊆ X be a compact invariant set and {x1 , x2 , . . . , xn } be a

generalized homoclinic contour. Then (xi , xj ) ∈ P (Σ) for all i, j = 1, 2, . . . , n.

Proof. Let i, j ∈ {1, 2, . . . , n}. Suppose, for example, that i ≤ j and 0 ≤ p =

j − i ≤ n. We will show that (xi , xj ) ∈ P (Σ). Let ε > 0 and t > 0. Since
ωxk ∩ αxk+1 6= ∅ for all k = i + 1, . . . , j − 1, then for the numbers ε > 0, t > 0
and the point xk (k = i, i + 1, . . . , j − 2) there exist t0k > t, t00k < −t and pk ∈
ωxk ∩ αxk such that ρ(xk t0k , pk ) < 3ε and ρ(xk t00k , pk−1 ) < 3ε . Let x0 := xi , x1 :=
xi+1 t0i+1 , . . . , xp−1 := xj t00j , xp := xj ; t0 := t0i , t1 := t0i+1 − t00i+1 , . . . , tp−1 := −t00j . It
is clear that {x0 , x1 , . . . , xp ; t0 , t1 , . . . , tp−1 } is a (ε, t, π)-chain joining the points xi
and xj . The lemma is proved. 
In Lemmas 4.3-4.5 we will suppose that R(Σ) consists of finite number of dif-
ferent classes of equivalence R1 , R2 . . . , Rk , i.e. R(Σ) = R1 t R2 t · · · t Rk . Let
us establish some properties of the sets Ri (i = 1, 2, . . . , k).

Lemma 4.3 In the collection {R1 , R2 . . . , Rk } there is no r-cycles (r ≥ 1).

Proof. Assuming the contrary, we obtain that there exist r ≥ 1 and i1 , i2 , . . . , ir

such that Ri1 < Ri2 < · · · < Rir < Ri1 . As well as in Lemma 4.2, it is possible
to prove that the set M := Ri1 ∪ · · · ∪ Rir belongs to one class of equivalence and,
consequently, r = 1. We will show that in the collection {R1 , R2 . . . , Rk } 1-cycles
are absent too. In fact, if we suppose that there exist 1 ≤ i ≤ r and a point x from
(W s (Ri ) ∩ W u (Ri )) \ Ri , then H(x0 ) ∪ Ri ⊆ (Σ). Moreover, H(x0 ) ∪ Ri belongs to
one class of equivalence and, consequently, H(x0 ) ∪ Ri ⊆ Ri . The latter contradicts
to the choice of the point x0 . The lemma is proved. 
 The structure of the Levinson center with the condition of the hyperbolicity 157

Lemma 4.4 The sets Ri (i = 1, 2, . . . , k) are locally maximal in Σ, i.e. for Ri

there exists a neighborhood Ui of the set Ri in Σ such that Ri is a maximal closed
invariant set in Ui .

Proof. Since Ri ∩ Rj 6= ∅ for all i 6= j, then there exist neighborhoods Ui of the

sets Ri such that U i ∩ U j = ∅ for i 6= j. Note, that in Ui the set Ri is maximal.
In fact, if we suppose that there exists a compact invariant set Λi ⊂ Ui such that
Λi * Ri , then Λi \ Ri 6= ∅. Let x ∈ Λi \ Ri . Since the sets αx and ωx are chain
recurrent [33, p.33], then αx , ωx ⊆ Ri , i.e. x ∈ (W s (Ri ) ∩ W u (Ri )) \ Ri and,
consequently, x ∈/ R(Σ). On the other hand, reasoning in the same way that in the
proof of Lemma 4.2, it is possible to show that x is chain recurrent, i.e. x ∈ R(Σ).
The obtained contradiction finishes the proof of the lemma. 

Lemma 4.5 The sets Ri (i = 1, 2, . . . , k) are indecomposable.

Proof. Suppose that for some i the set Ri can be represented in the form of a
union of two own closed invariant subsets A1 and A2 , i.e. Ri = A1 t A2 . Since
A1 ∩ A2 = ∅ and Ri ∩ Rj = ∅ for all j 6= i, then there are neighborhoods U1 and U2
of the sets A1 and A2 respectively, such that U 1 ∩ U 2 = ∅ and (U 1 ∪ U 2 ) ∩ Rj = ∅
for all j 6= i. Let ε0 > 0 be such that B(A1 , ε0 ) ⊂ U 1 and B(A2 , ε0 ) ⊂ U 2 .
We will take arbitrary points a1 ∈ A1 , a2 ∈ A2 and numbers 0 < ε < ε0 and
t > 0. Since a1 , a2 ∈ Ri , then for the numbers ε and t there exists (ε, t, π)-chain
{a1 = x0 , x1 , . . . , xn = a2 ; t1 , t2 , . . . , tn } joining the points a1 and a2 . By Theorem
2.24 from [33] we can suppose that xk ∈ Ri (k = 0, 1, . . . , n). According to the
choice of the number ε and the invariance of the set A1 , we have xk ∈ A1 for all
k = 0, 1, . . . , n, i.e. xn = a2 ∈ A1 . The latter contradicts to our assumption. The
lemma is proved. 

4.2 The spectral decomposition of the Levinson’s center

Let h(X, S, π), (Y, S, σ), hi be a two-sided non-autonomous dynamical system with
the fibers Xy := {x ∈ X | h(x) = y} (y ∈ Y ). And let every fiber Xy be homeo-
morphic to V, where V is some n-dimensional manifold (for example, V = E n ) and
p ∈ Xy . Following the work [33, p.213], we denote by Wδs (p) := {x ∈ Xy | ρ(xt, pt) ≤
δ, t ≥ 0} and Wδu (p) := {x ∈ Xy | ρ(xt, pt) ≤ δ, t ≤ 0} (δ > 0).

Definition 4.5 The compact invariant set Λ ⊆ X is said to be hyperbolic (or non-
autonomous dynamical system h(X, S, π), (Y, S, σ), hi has a hyperbolic structure on
Λ), if there are positive numbers N , ν, δ and γ such that
H1. Wδs (p) ∩ Wδu (p) = {p} for all p ∈ Λ, Wδs (p) and the sets Wδu (p) are sub-
manifolds from Xy (y = h(p)), which are homeomorphic to the closed desk
158 Global Attractors of Non-autonomous Dissipative Dynamical Systems

in Rk and Rn−k respectively. In addition, if ρ(p, q) ≤ γ (p, q ∈ Xy ), then

Wδs (p) ∩ Wδu (p) 6= ∅;
H2. π t Wδs (p) ⊆ Wδs (π t p) for all t ≥ 0 and π t Wδu (p) ⊇ Wδu (π t p) for all t ≤ 0 and
and each p ∈ Λ;
H3. the manifolds Wδs (p) and Wδu (p) depend continuously on the point p ∈ Λ in
the distance of Hausdorff;
H4. (a) ρ(p1 t, p2 t) ≤ N exp(−νt)ρ(p1 , p2 ) for all p1 , p2 ∈ Wδs (p) and t ≥ 0;
(b) ρ(p1 t, p2 t) ≤ N exp(νt)ρ(p1 , p2 ) for all p1 , p2 ∈ Wδu (p) and t ≤ 0.

Remark 4.1 a. If the set Λ ⊆ X is hyperbolic in usual sense [33, 261], then
under some additional conditions of smoothness, the set Λ will also be hyperbolic in
the sense of the definition above. The inverse statement, generally speaking, is not
true.
b. For a discrete autonomous dynamical system there was introduced a close
notion (axiom A# ) in the work [2].

Denote by M(Σ) a closure of all the recurrent points of the set Σ. The following
statement takes place.

Theorem 4.1 Let Y be a compact minimal set, Σ be a compact invariant set

from X and let one of the following two conditions be fulfilled:

a. the number of minimal sets in Σ is finite;

b. if the set Σ contains an infinite number of minimal sets, then on the set M(Σ)
(except, maybe, a finite number of isolated minimal sets) the non-autonomous
dynamical system h(X, S, π), (Y, S, σ), hi has a hyporbolic structure.

Then the relation ∼ decomposes the set R(Σ) on finite number of different classes
of equivalence.

Proof. If the set Σ contains only a finite number of minimal sets, then the statement
of the theorem coincide with the lemma 4.1
Let now the condition b. of the theorem be fulfilled. And we suppose that the
theorem is not true. Then there is a denumerable family of classes of equivalence
{Rk : k = 1, 2, . . . }. Since the sets Rk are closed, invariant and the set Σ is compact,
then every set Rk contains at least one compact minimal subset Mk ⊆ Rk . Since the
set Y is minimal, then h(Mk ) = Y for all k = 1, 2, . . . . Let y ∈ Y and pk ∈ Mk ∩Xy .
According to the compactness of the set Σ, we can suppose that the sequence {p k }
is convergent. Put p := lim pk and we note that p ∈ Λ := ∪{Mk : k = 1, 2, . . . }.
k→+∞
Without loss of generality we can suppose that the set Λ is hyperbolic. Let γ > 0 be
the number figuring in the condition H1. for the set Λ. The sequence {pk } ⊆ Λ is
convergent; therefore, starting from some k0 , the manifolds Wδs (pk1 ) and Wδu (pk2 )
 The structure of the Levinson center with the condition of the hyperbolicity 159

are a nonempty intersection for all k1 , k2 ≥ k0 . Let now k1 6= k2 and k1 , k2 ≥ k0 .

Chose points x1 ∈ Wδs (pk1 ) ∩ Wδu (pk2 ), x2 ∈ Wδu (pk1 ) ∩ Wδs (pk2 ) and consider the
homoclinic contour K := H(x1 ) ∪ H(x2 ). By Corollary 2.19 [33] and taking in
consideration that on an ω-limit (α-limit) set every two points are equivalent in the
sense of the relation ∼, we obtain the equivalence of every two points of homoclinic
contour K, i.e. the set K belongs to one class of equivalence. Thus, beginning with
k0 , the points pk1 and pk2 are equivalent for all k1 , k2 ≥ k0 . On the other hand,
the point {pk } was chosen from the different classes of equivalence. The obtained
contradiction proves our theorem. 

Theorem 4.2 Under the conditions of Theorem 4.1 on the center of Levinson J
of the dynamical system h(X, S, π), (Y, S, σ), hi the following assertions hold:
(1) the relation ∼ decomposes the set R(J) on finite number of different classes of
equivalence, i.e. R(J) = R1 t R2 t · · · t Rk ;
(2) the sets Ri (i = 1, 2, . . . , k) are closed, invariant, indecomposable, locally max-
imal and in the collection {R1 , R2 , . . . , Rk } there is no r (r ≥ 1) cycles;
(3) J = ∪{W u (Ri ) : i = 1, k}, where W u (Ri ) := {x ∈ X : lim ρ(xt, Ri ) = 0}
t→−∞
(i = 1, k).

Proof. The first and the second statements of Theorem 4.2 directly follow from
Lemmas 4.3-4.5 and Theorem 4.1. Now we will prove the third statement. Let
x ∈ W u (Ri ) (i = 1, k), then by the compact dissipativity of the dynamical system
(X, S, π) the set H(x) is compact and, consequently, x ∈ J.
Inverse. Let x ∈ J. In view of the compactness and invariance of the set J, the
sets αx and ωx are nonempty, compact, indecomposable and chain recurrent [33].
Since under the conditions of Theorem 4.1 R(J) = R1 t R2 t · · · t Rk and the sets
Ri (i = 1, k) are disjoint, closed and invariant, then there are i, j ∈ {1, 2 . . . , k}
such that αx ⊆ Ri and ωx ⊆ Rj . Thus x ∈ W u (Rj ). The theorem is proved. 

Remark 4.2 Theorems 4.1 and 4.2 are true also in the case, if we replace the
condition of the minimality of the set Y by the following condition: the set Y
contains only finite number of minimal sets.

4.3 One-dimensional systems with hyperbolic center

In this section we study the structure of the Levinson center for the scalar dissipative
equation u0 = f (t, u) with recurrent (almost periodic, periodic) right-hand side.
Although the problem under study concerns the equation u0 = f (t, u), its solution
is given within the framework of general non-autonomous dynamical system.
160 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Let Λ ⊆ X be a hyperbolic set of non-autonomous dynamical system

h(X, S, π), (Y, S, σ), hi, (X, h, Y ) be a finite-dimensional vector bundle with the fiber
Rn , and |·| be a Riemannian metric on (X, h, Y ), that is compatible with the metric
ρ on X.

Remark 4.3 a. If p ∈ Λ is such that k = n (k = 0), then the set H(p) is

exponentially stable on S+ (S− ).
b. For n = 1 the hyperbolic set H(p) is exponentially stable either on S + or on
S− .

Everywhere in this section, we will suppose that the set Y is compact minimal
and n = 1.

Lemma 4.6 If the compact minimal set M of X is hyperbolic, then h|M : M → Y

is a homeomorphism (or, what is the same thing, My := M ∩ Xy consists of exactly
one point for each y ∈ Y ).

Proof. Let M ⊆ X be a compact minimal subset of X. If M is hyperbolic,

then, by Remark 4.3 b., we can assume that M is exponentially stable on S+ ,
and,consequently, M is uniformly stable in the positive direction, i.e. for each ε > 0
there exists a γ(ε) > 0 such that ρ(m1 , m2 ) < γ (m1 , m2 ∈ M and h(m1 ) = h(m2 ))
implies the inequality ρ(m1 t, m2 t) < ε for all t ≥ 0. To complete the proof of the
Lemma it is sufficient to refer to Theorem 3 from [331, p.110]. 

Let Σ be a nonempty compact invariant set in X, {Mα | α ∈ A} by the family

of all minimal subsets of Σ, and M(Σ) := ∪{Mα | α ∈ A}.

Lemma 4.7 If M(Σ) is hyperbolic, then Σ contains a finite number of minimal

sets M1 , M2 , . . . , Mk and, consequently, M(Σ) = M1 t M2 t · · · t Mk .

Proof. Let us assume the contrary, i.e. let there exist a countable family {Mk : k =
1, 2, . . . } of distinct minimal sets in Σ. Let y ∈ Y and pk ∈ Mk ∩ Xy (k = 1, 2, . . . ).
Since pk ∈ M(Σ), we can assume that {pk } is convergent. Let pk → p, then
p ∈ M(Σ). Since M(Σ) is hyperbolic, we can assume that (see Remark 4.3 b.)
H(p) is exponentially stable on S+ , and, consequently, there exists a k0 ∈ N such
that ρ(pk1 t, pk2 t) ≤ N exp(−νt)ρ(pk1 , pk2 ) for all k1 , k2 ≥ k0 and t ≥ 0. It follows
from the last inequality that ρ(pk1 t, pk2 t) → 0 as t → +∞ for all k1 , k2 ≥ k0 . But
this is not possible because by Lemma 4.6 pk1 and pk2 are jointly recurrent. This
contradiction proves the Lemma 4.7. 

Lemma 4.8 Under the conditions of Lemma 4.7, for each point x ∈ Σ there exist
recurrent points p1 and p2 (h(p1 ) = h(p2 ) = h(x)) such that the following conditions
 The structure of the Levinson center with the condition of the hyperbolicity 161

are fulfilled:

a. lim ρ(xt, p1 t) = 0 and b. lim ρ(xt, p2 t) = 0. (4.1)

t→+∞ t→−∞

Proof. Let x ∈ Σ, ω and α-limit sets ωx and αx are nonempty, compact and
contain the minimal sets M1 and M2 , respectively. Since Y is minimal, it follows
that h(M1 ) = h(M2 ) = Y , and, since M(Σ) is hyperbolic, the sets M1y and M2y
consist of exactly one point each for an arbitrary y ∈ Y .
Let x be non-recurrent (in the contrary case Lemma is obvious) and Miy = {pi }
(i = 1, 2). Then there exist t1n → +∞ and t2n → −∞ such that xt1n → p1 and
xt2n → p2 . By Lemma 4.6 pi tin → pi (i = 1, 2) as n → +∞. Hence

ρ(xtin , pi tin ) → 0 (4.2)

a. Let M1 be exponentially stable on S+ . Then (4.2) implies (4.1a). Now we

show that (4.1b) holds. Indeed, if M1 is exponentially stable on S+ , then M2 is
exponentially stable on S− . If we assume the contrary, then, by (4.2) ρ(xt, p2 t) → 0
as t → +∞. It follows from Lemma 4.6 that p1 = p2 = p. Let ε = ρ(x, p) > 0 and
γ(ε) > 0 be chosen for ε from the condition of uniform stability of H(p) on S+ . Then
it follows from (4.2) that for sufficiently large n we have ρ(xt2n , p2 t2n ) < γ(ε) and,
consequently, ρ(x(t + t2n ), p2 (t + t2n )) < ε for all t ≥ 0. In particular, ρ(x, p) < ε,
which contradicts the choice of ε. This contradiction proves the exponential stability
of the set M2 on S− . From (4.2) it follows (4.1b).
b. We will show that no M1 ⊆ ωx cannot be exponentially stable on S− . Indeed,
if we assume the contrary, then it follows from (4.2) that ρ(xt, p1 t) → 0 as t → −∞
and, consequently, αx = M1 ⊆ ωx . Thus M2 = M1 = αx ⊆ ωx . Let p := p1 = p2 ∈
M1 ∩ Xy = M2 ∩ Xy and ε = ρ(x, p) > 0. Further, reasoning in the same manner
as at the end of the preceding item, we obtain a contradiction, which proves the
exponential stability on S+ of the minimal set M1 ⊆ ωx . The Lemma is completely
proved. 
Under the conditions of Lemma 4.7, the family Σ consists of only a finite number
of distinct minimal sets M1 , M2 , . . . , Mk and M(Σ) = M1 ∪M2 ∪· · ·∪Mk . Let y ∈ Y
and {γi (y)} = Mi ∩ Xy . We will suppose that M1 , M2 , . . . , Mk are ordered such
that γ1 (y) < γ2 (y) < · · · < γk (y). Let us set µy := inf Σy an νy := sup Σy . It is
obvious that µy , νy ∈ Σy .

Lemma 4.9 Under the conditions of Lemma 4.7 the equalities µy = γ1 (y) and
νy = γk (y) hold for all y ∈ Y .

Proof. We prove, for example, the first equality (the second equality is proved in the
same manner). If we suppose that µy 6= γ1 (y) for certain y ∈ Y , then µy < γ1 (y).
By Lemma 4.8, there exist recurrent points p1 and p2 from M(Σ), such that (4.1a)
162 Global Attractors of Non-autonomous Dissipative Dynamical Systems

and (4.1b) hold. We show that p1 , p2 ≤ γ1 (y). In fact, from (4.1) it follows the
existence of t1n → +∞ and t2n → −∞ such that µy t1n → p1 and µy t2n → p2 .
Moreover, by Lemma 4.6, γi (y)tin → γi (y) (i = 1, 2). Since µy < γ1 (y), it follows
that µy t1n < γ1 (y)t1n and, consequently, p1 ≤ γ1 (y). Analogously, p2 ≤ γ1 (y).
Thus, p1 = p2 = γ1 (y). Since H(γ1 (p)) = M1 is hyperbolic, it follows that M1 is
exponentially stable either on S+ or on S− . Further, reasoning in the same manner
as in Lemma 4.8, we obtain a contradiction. The lemma is proved. 

Theorem 4.3 Let Σ ⊆ X be a nonempty compact invariant subset of X and

M(Σ) ⊆ Σ be hyperbolic. Then the following statements are valid:

(1) Σ consists of only a finite number of distinct minimal sets M1 , M2 , . . . , Mk and

M(Σ) = M1 ∪ M2 ∪ · · · ∪ Mk ;
(2) each minimal set Mi (i = 1, 2, . . . , k) is homeomorphic to Y and, in particular,
Mi is almost periodic (periodic) if Y has this property;
(3) for each point x ∈ Σ there exist points p1 and p2 from M(Σ) such that relations
(4.1) are fulfilled;
(4) if M1 , M2 , . . . , Mk are ordered such that γi (y) < γj (y) for all y ∈ Y (where
{γi (y)} := Mi ∩ Xy ) if and only if i < j, then ∂Σ = M1 ∪ Mk , where ∂Σ is the
boundary of Σ.

Proof. This assertion follows from Lemmas 4.6-4.9. 

Theorem 4.4 Let h(X, S, π), (Y, S, σ), hi be compact dissipative and J be its
Levinson center. If M(J) is hyperbolic, then the following statements are valid:

(1) J contains only a finite number of the minimal sets M1 , M2 , . . . , Mk , each of

which is homeomorphic to Y , and M(J) = M1 ∪ M2 ∪ · · · ∪ Mk ;
(2) for each point x ∈ J there exist points p1 and p2 in M(J) such that relations
(4.1) are fulfilled;
(3) if x ∈
/ J, then
(a) ρ(xt, γ1 (y)t) → 0 as t → +∞ if x < γ1 (y), where y = h(x), and
(b) ρ(xt, γk (y)t) → 0 as t → +∞ if x > γk (y).
(4) ∂J = M1 ∪Mk and ∂J is uniformly asymptotically stable in the positive direction
and, in particular, k is odd.

Proof. This statement follows in essence from Theorem 4.4 and properties of the
Levinson center. 

Remark 4.4 Although Theorems 4.3 and 4.4 are proved for one-dimensional
systems, yet in fact we have always used only a consequence of one-dimensionality
(see Remark 4.3 b.). Therefore, Theorems 4.3 and 4.4 are valid with insignificant
 The structure of the Levinson center with the condition of the hyperbolicity 163

changes for n ≥ 2 also if besides the hyperbolicity of M we require that for each
m ∈ M the set H(m) is exponentially stable either on S+ or on S− . In addition,
the minimal sets M1 , M2 , . . . , Mk in Theorems 4.3 and 4.4, generally speaking, not
homeomorphic to Y but are only distal covering of Y of finite multiplicity, which
also ensures their almost periodicity (periodicity) if Y is almost periodic (periodic).

Using the known connection between non-autonomous dynamical system and

differential equations, we can obtain statements, analogous to Theorems 4.3 and
4.4, for differential equations.
Let us consider the differential equation

u0 = f (t, u), (4.3)

where f ∈ C 0,2 (R × Rn , Rn ) and C 0,2 (R × Rn , Rn ) is the space of all continuous

functions f : R × Rn → Rn , continuously differentiable twice in u ∈ Rn , equipped
with the topology of uniform convergence of functions and their derivatives up to
second order inclusive on compacts from R × Rn . Along this equation (4.3) we
consider the family of equations

u0 = g(t, u), (4.4)

where g ∈ H(f ) := {fτ | τ ∈ R}, and fτ (t, u) := f (t+τ, u), the bar denoting closure
in f ∈ C 0,2 (R × Rn , Rn ). Let us suppose that the condition of non-local extend-
ability is fulfilled for each equation (4.4). Let ϕ(t, v, g) denote the solution of (4.4)
that passes through v ∈ Rn for t = 0. Let Y := H(f ) and (Y, R, σ) be a dynamical
system of translations on Y. Further, let X := Rn × Y. Then we can define the
dynamical system (X, R, π) by the rule π(τ, (v, g)) := (ϕ(τ, v, g), gτ ). Finally, we
set h := pr2 : X → Y. Then we obtain the non-autonomous dynamical system
h(X, R, π), (Y, R, σ), hi. Applying Theorems 4.3 and 4.4 to the so-constructed
dynamical system, we get the corresponding statements for equation (4.3). For
example, we formulate the statement that follows from Theorem 4.4. Fist of all, let
us recall certain notion related to equation (4.3).
We recall that the equation (4.3) is called dissipative if the non-autonomous
dynamical system, generated by it, is dyssipative. Let J ⊂ Rn × H(f ) be the
Levinson center of (4.3) and Σ ⊆ J.For each point (v, g) ∈ Σ let us consider the
equation in variations for (4.4) along the solution ϕ(t, v, g), i.e., the linear equation

w0 = gv0 (t, ϕ(t, v, g))w. (4.5)

The family of equation (4.5), where (v, g) ∈ Σ, is said to satisfy the exponential
dichotomy on R [132] and the corresponding constants can be chosen independent
of (v, g) ∈ Σ.
164 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Theorem 4.5 Let f ∈ C 0,2 (R × Rn , Rn ) and f, fu0 and fu002 be jointly recurrent
(almost periodic, τ -periodic) in t ∈ R uniformly in x on compacts from R. Suppose
that the equation (4.3) is dissipative, J ⊂ R × H(f ), be its Levinson center, and
the family of equation (4.5), where (v, g) ∈ M(J), satisfy the exponential dichotomy
condition on R. Then for each g ∈ H(f ), (4.4) has only a finite number of uniformly
compatible solutions [300]: These are q1 , q2 , · · · , qk (k is an odd number and depends
only of f ); all the remaining solutions, bounded on R, converge to the indicated
solutions as t → +∞ or −∞. But if a certain solution of (4.4) is bounded on R +
only (since (4.3) is dissipative, all solutions of (4.4) is bounded on R + ), then it
converges to one of the solutions q1 and qk as t → +∞.

Remark 4.5 A statement, analogous to Theorem 4.5, holds also for the difference
equations.

4.4 The dissipative cascades

Let M be a differentiable manifold, U be an open subset of M , f : U → f (U ) ⊆ M

be a diffeomorphism of the class C r (r ≥ 1), Λ ⊂ U be a maximal compact invariant
set of the diffeomorphism f . Suppose that U is a domain of attraction for Λ, i.e.
W s (Λ) = U . In this case, the dynamical system generated by the positive powers
of the diffeomorphism f is dissipative and its Levinson center J = Λ.

Lemma 4.10 Let x0 ∈ U . If ωx0 is a compact and hyperbolic set of the diffeo-
morphism f : U → M , then for every x ∈ ωx0 and δ > 0 there is p ∈ P er(f ) such
that ρ(x, p) < δ and Σp ⊆ B(ωx0 , δ), where P er(f ) is a set of all the periodic points
of the diffeomorphism f , Σp is a trajectory of the point p.

Proof. Let δ > 0. We may take the number δ to be so small such that B(ωx0 , δ) ⊂
e (ωx0 ), where U(ω
U e x0 ) is a neighborhood of the hyperbolic set ωx0 figuring in the
theorem of Anosov about the family of ε-trajectories [261, p.220]. In the cited
above theorem we will chose a number ε (ε < 2δ ) that corresponds to the number
δ ε n ε
2 . Then for 2 there is l > 0 such that f x0 ∈ B(ωx0 , 2 ) for all n ≥ l. If x ∈ ωx0 ,
then there is a number n0 ≥ l such that f n x0 ∈ B(x, 2ε ). Put x1 = f n0 x0 . There
exists a number N > 0 such that f N x1 ∈ B(x, 2ε ). Consider the space XN =
{0, 1, . . . , N − 1} with discrete topology, the homeomorphism of the shift τ : XN →
XN , τ (k) = k + 1 mod(N ), and the mapping ϕ : XN → M defined by the equality
ϕ(k) = f k x1 (k = 0, 1, . . . , N − 1). Note, that ϕ : XN → B(ωx0 , 2ε ) ⊂ U(ω e x0 ).
N
Further: P (ϕ ◦ τ, f ◦ ϕ) := sup ρ(ϕ(τ (k)), f (ϕ(k))) ≤ ρ(x1 , f x1 ). Since
0≤k≤N −1
ρ(x1 , f N x1 ) ≤ ρ(x1 , x) + ρ(x, f N x1 ) < ε, then P (ϕ ◦ τ, f ◦ ϕ) < ε. According to
Anosov’s theorem of the family of ε-trajectories, there exists a continuous mapping
 The structure of the Levinson center with the condition of the hyperbolicity 165

ψ : XN → U e (ωx0 ) such that ψ ◦ τ = f ◦ ψ and P (ϕ, ψ) < δ . To finish the

2
proof of the lemma it is sufficient to note that ψ(0) ∈ P er(f ), ρ(ψ(0), x) < 2δ and
ψ(k) ∈ B(ωx0 , δ) for all k = 0, 1, . . . , N − 1. 

Corollary 4.1 If the set Σ is locally maximal and ωx0 ⊆ Σ, then under the
conditions of Lemma 4.10, in the set Σ there exists at least one periodic point f .

Corollary 4.2 Let Λ be a locally maximal compact invariant set. If Σ ⊆ ωx0 ⊆ Λ

is a compact minimal set, which does not contain periodic trajectories, then under
the conditions of Lemma 4.10 in the set Λ there exists an infinite number of different
hyperbolic points pk → p ∈ Σ ⊆ ωx0 , moreover, Σpk ⊆ B(ωx0 , εk ) and εk ↓ 0.

Definition 4.6 The set M ⊆ X of the dynamical system (X, S, π) is called quasi-
minimal, if there is a stable in the sense of Poisson point p ∈ M such that M = H(p).

Theorem 4.6 Let f : U → M be a diffeomorphism of the class C k (k ≥ 1) and Λ

be a compact invariant set of the diffeomorphism f . If the set M0 (Λ) is hyperbolic,
then:

(1) the relation ∼ decomposes the set R(Λ) on finite number of classes of equiva-
lence, i.e. R(Λ) = R1 t R2 t · · · t Rk ;
(2) the sets Ri (i = 1, k) are closed, invariant, indecomposable, locally maximal
and in the collection {R1 , R2 , . . . , Rk } there is no r-cycles, where r ≥ 1.

Proof. This statement follows from Theorem 4.1. 

Corollary 4.3 Let M be a differentiable manifold and f : M → M be a diffeo-

morphism of the class C k (k ≥ 1). If the set of the chain recurrent points R(f ) of
the diffeomorphism f is compact and hyperbolic, then:

(1) the relation ∼ decomposes the set R(f ) on finite number of classes of equiva-
lence, i.e. R(f ) = R1 t R2 t · · · t Rk ;
(2) the sets Ri (i = 1, k) are closed, invariant, quasi-minimal, locally maximal and
in the collection {R1 , R2 , . . . , Rk } there is no r (r ≥ 1)-cycles;
(3) if, in addition, in the set M there exists an infinte number of different minimal
sets, then in the set M there is a nontrivial (non periodic) minimal set.

Proof. The formulated assertion follows from Theorem 4.6, Theorems 7.5 (and its
discrete analog) 7.8 from [2] and Lemma 6.4 from [261]. 

Theorem 4.7 Let f : U → M be a dissipative diffeomorphism and J be its

Levinson center. If one of the following two condition is fulfilled:

a. in the set Λ there is only a finite number of minimal sets;

166 Global Attractors of Non-autonomous Dissipative Dynamical Systems

b. in the set Λ there is an infinite number of different minimal sets and the set
M0 (Λ) is hyperbolic [261],

then:

(1) the relation ∼ decomposes the set of the chain recurrent points R(f ) of the
diffeomorphism f on finite number of classes of equivalence, i.e. R(f ) = R1 t
R2 t · · · t R k ;
(2) the sets Ri (i = 1, 2, . . . , k) are closed, invariant, indecomposable, locally max-
imal and in the collection {R1 , R2 , . . . , Rk } there is no r (r ≥ 1)-cycles;
(3) J = ∪{W u (Ri ) : i ∈ 1, k}.

Proof. This statement follows from Theorem 4.2. In fact, if the set M0 (Λ) 6= ∅
and it is hyperbolic, then from the results [261] (see Theorems 0.1, 5.4 and the
theorem from the first appendix) follows, that on the set M(Λ), except, maybe,
a finite number of isolated minimal sets, a dynamical system generated by the
positive powers of the diffeomorphism f has a hyperbolic structure in the sense of
our definition. 

Remark 4.6 a. Theorem 4.7 is an analog of the well-known theorem of Smale

about spectral decomposition [261] for A-diffeomorphisms.
b. The analog of the theorem 4.7 is true also for the continuous dynamical system
(flows).

Theorem 4.8 Let f : U → M be a dissipative diffeomorphism and let its center of

Levinson contain an infinite number of minimal sets. If the set M0 (Λ) is hyperbolic,
then the statements 1.-3. of Theorem 4.6 hold. In addition, in the set Λ there is an
infinite number of hyperbolic periodic trajectories.

Proof. According to Theorem 4.7 it is sufficient to show, that in the set Λ there is
an infinite number of different hyperbolic periodic trajectories. Denote by U e ⊂U a
neighborhood of the set M0 (Λ) such that every compact, invariant with respect to
f set Λ0 ⊂ U e is hyperbolic. By Theorem 2 [261, p.224] there is such a neighborhood.
According to [226, p.52] a metric space 2Λ is compact. Since in the set Λ there is
an infinite number of the minimal sets {Mi } and Mi ⊆ R = R1 t R2 t · · · t Rk ,
then we can suppose that Mi ⊆ R1 (i = 1, 2, . . . ). Taking into consideration the
compactness of the space 2Λ , we may suppose that the sequence {Mi } is convergent
in the metric of Hausdorff. In particular, beginning from a number i0 , we have
Mi ⊆ U e for i ≥ i0 . In view of the choice of the set U,
e all the minimal sets Mi
(i ≥ i0 ) are hyperbolic. The following two cases are possible:

a. every minimal set Mi (i ≥ i0 ) is periodic and in this case the theorem is proved.
b. between the sets Mi (i ≥ i0 ) there is a nontrivial minimal set M .
 The structure of the Levinson center with the condition of the hyperbolicity 167

Then, according to Corollary 4.2, there exists an infinite number of different periodic
points pk → p ∈ M , moreover, Σpk ⊆ B(M, εk ) and εk ↓ 0. Since Mi ⊆ R1 (i ≥ i0 )
R1 is closed and locally maximal, then starting from a number m0 we have pm ∈ R1 .
The theorem is proved. 

Definition 4.7 Recall [2]-[3], that an invariant set Λ of the diffeomorphism f :

U → M is called factor-markovian, if there is a TMC (topological markovian chain
[2]) T : ΩΠ → ΩΠ and a continuous mapping ϕ : ΩΠ → Λ such that ϕ ◦ T = f ◦ ϕ.
If the mapping ϕ is a homeomorphism, then the set Λ is called markovian.

Corollary 4.4 Under the conditions of Theorem 4.8 in the set R there is a locally
maximal hyperbolic markovian set Λ ⊆ R and, in particular, in the set R there is a
transversal homoclynic trajectory.

Proof. The formulated statement follows from Theorem 4.8 and also from Theorems
7.7 [2] and 2.4 [271]. 

4.5 The periodic dissipative systems

1. Let

h(X, T1 , π), (Y, T2 , σ), hi (4.6)

be a non-autonomous dynamical system. In this paragraph we will suppose that

the dynamical system (Y, T2 , σ) is periodic, i.e. there exist τ ∈ T2 (τ > 0) and
q ∈ Y such that σ(q, τ ) = q and Y = {σ(q, t) : 0 ≤ t < τ }. In this connection
the non-autonomous dynamical system (4.6) is said to be τ -periodic. Denote by
P : Xq → Xq a mapping defined by the equality P (x) = π(τ, x) and by (Xq , P )
denote a cascade generated by the positive powers of the mapping P .

Lemma 4.11 A non-autonomous dynamical system (4.6) is pointwise (com-

pact,local) dissipative if and only if the cascade (Xq , P ) is pointwise (compactly,
locally) dissipative. In this case the following relations hold:

J = {π t Jq : 0 ≤ t < τ } and Jq := J ∩ Xq ,

where J is the Levinson center of the dynamical system (4.6), but Jq is the Levinson
center of the dynamical system (Xq , P ).

Proof. This assertion follows immediately from the corresponding definitions. 

Corollary 4.5 If (X, h, Y ) is a finite-dimensional vectorial fibering, then the

center of Levinson J of the non-autonomous dynamical system (4.6) and Jq are
connected.
168 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Proof. This statement follows from Theorems 1.32, 1.33 and Corollary 1.14. 

Corollary 4.6 Under the conditions of Corollary 4.5, there exists r > 0 and for
all a > 0 there is k(a) ∈ Z+ such that P k B[0, a] ⊆ B[0, r] for all k ≥ k(a), where
B[0, a] := {x ∈ Xq : |x| ≤ a}.

Proof. Corollary 4.6 follows from Theorem 2.23 and Lemma 4.11. 

Corollary 4.7 Under the conditions of Corollary 4.5 there exists a fixed point of
the mapping P : Xq → Xq .

Proof. Corollary 4.7 follows from the theorem on the fixed point of Brauder [35]
and Corollary 4.6. 

Theorem 4.9 Let (X, h, Y ) be a finite-dimensional fibering. There exists a

nonempty compact invariant set J (Levinson center) of the dynamical system (4.6)
possessing the following properties:

(1) for every y ∈ Y the set Jy = J ∩ Xy is connected;

(2) for every  > 0 there exists δ() > 0 such that the inequality ρ(x, J y ) < δ
(x ∈ Xy ) implies ρ(xt, Jyt ) <  for all t ≥ 0;
(3) for all y ∈ Y and x ∈ Xy lim ρ(xt, Jyt ) = 0;
t→+∞
(4) in the center of Levinson J of the system (4.6) there is at least one τ -periodic
point.

Proof. The formulated statement follows from Lemma 2.1, Theorem 2.1 and Corol-
lary 4.7. 

Theorem 4.10 If Xq is a complex n-dimensional space (i.e. Xq = Cn ) and the

mapping P : Xq → Xq is holomorphic, then Jq = {p} and lim ρ(xt, pt) = 0 for
t→+∞
all x ∈ Xq , i.e. the dynamical system (4.6) is convergent.

Proof. The proof of this assertion can be obtained with the help of a slight modi-
fication of the proof of Theorem 3.3. 

2. Consider a differential equation

u̇ = f (t, u), (t ∈ R, u ∈ E n ) (4.7)

where f is a τ -periodic function with respect to the variable t ∈ R and f is contin-

uously differentiable w.r.t. u ∈ E n . If the function f ∈ C(R × E n , E n ) is regular,
then (see Example 3.1) the equation (4.7) generates a non-autonomous dynamical
system (and this system is τ -periodic). Applying to this non-autonomous dynamical
 The structure of the Levinson center with the condition of the hyperbolicity 169

system Theorems 4.6, 4.7 and 4.8, we will obtain the corresponding statements for
the equation (4.7).
Let P : E n → E n be a Poincare transformation for the equation (4.7), i.e.
P (u) = ϕ(τ, u, f ).

Definition 4.8 The set of chain recurrent points for the equation (4.7) is said to
be a set of chain recurrent points of the cascade (E n , P ).

It is well known (see, for example, [33, p.56]), that for the equation (4.7) with
the differentiable (w.r.t. variable u) function f the set of chain recurrent points R
coincides with the set Π that generates periodic (with the period kτ , where k ∈ N)
solutions of the equation (4.7) (for the definition of the set Π see [271, p.278]).

Theorem 4.11 Suppose, that for the equation (4.7) the set of chain recurrent
points R is compact and the set M0 (R) is hyperbolic, then

(1) the relation ∼ decomposes the set R on finite number of classes of equivalence,
i.e. R = R1 t R2 t · · · t Rk ;
(2) the sets Ri (i = 1, k) are closed, invariant, indecomposable, locally maximal
and in the collection {R1 , R2 , . . . , Rk } there is no r (r ≥ 1)-cycles;

Corollary 4.8 Suppose, that for the equation (4.6) the set Π, which generates
periodic solutions (with the period kτ , where k ∈ N), is compact and hyperbolic,
then:

(1) the relation ∼ decomposes the set Π on finite number of classes of equivalence,
i.e. Π = Π1 t Π2 t · · · t Πk ;
(2) the sets Πi (i = 1, k) are closed,invariant, quasi-minimal, locally maximal and
the collection {Π1 , Π2 , . . . , Πk } does not contain r-cycles;
(3) in addition, if in the set Π there is an infinite number of minimal sets, then in
Π there exists a nontrivial minimal set.

This assertion generalizes Theorem 3.3 from [271, p.280].

Theorem 4.12 Let the equation (4.7) be dissipative and the set J be its Levinson’s
center. Suppose that one of the following two conditions is fulfilled:

a. in the set J there is a finite number of minimal sets;

b. in the set J there is an infinite number of minimal sets and the set M0 (J) is
hyperbolic.

Then:

(1) the relation ∼ decomposes the set of chain recurrent points of the equation (4.7)
R on finite number of classes of equivalence, i.e. R = R1 t R2 t · · · t Rk ;
170 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(2) the sets Ri (i = 1, k) are closed, invariant, indecomposable, locally maximal

and in the collection {R1 , R2 , . . . , Rk } there is no r-cycles (r ≥ 1);
(3) J = ∪{W u (Ri ) : i ∈ 1, k}.

Theorem 4.13 Suppose that the equation (4.7) is dissipative, J is its Levinson
center. If the set J contains an infinite number of minimal sets and the set M0 (J)
is hyperbolic, then the statements 1.-3. of Theorem 4.12 hold. In addition, in the
set J there is an infinite number of hyperbolic periodic solutions.

Corollary 4.9 Under the conditions of Theorem 4.13 in the set J there is a locally
maximal markovian set Λ ⊆ J and, in particular, in the set J there is a transversal
homoclinic periodic trajectory.

This statement follows from Corollary 4.4.

 Chapter 5

Method of Lyapunov functions

5.1 Criterions of dissipativity in term of Lyapunov functions

In this section we suppose that (X, h, Y ) is a Banach fiber bundle.

Theorem 5.1 Let Y be compact and (X, T1 , π) be completely continuous.

The non-autonomous dynamical system h(X, T1 , π), (Y, T2 , σ), hi is boundedly k-
dissipative if and only if there is a number r > 0 and a continuous function
V : Er → R (Er := {x ∈ E | |x| ≥ r}) with the following properties:

1. the set {x | V (x) ≤ c} is bounded for any c ∈ R;

2. if xτ ∈ Er for all τ ∈ [0, t], then V (xt) ≤ V (x);
3. level lines of the function V do not contain ω-limit points of the dynamical
system (X, T1 , π).

Proof. Let K1 = {x ∈ E | |x| ≤ r}, where r is the number from the statement
of the theorem. As Y is compact and the fiber bundle (X, h, Y ) is locally trivial,
then the set K1 is bounded. As (X, T1 , π) is completely continuous, then for the
bounded set K1 there will be a number l such that π l K1 is relatively compact.
Suppose K = π l K1 and let us show that ωx ∩ K 6= ∅ for any x ∈ E. Suppose it
is not so, then there will be x0 ∈ E for which ωx0 ∩ K 6= ∅ and, consequently, for
some t0 ≥ 0 we will have x0 t ∈
/ K1 for all t ≥ t0 . If this were not so, then we should
have a sequence tn → +∞ such that |x0 tn | ≤ r and, hence, {x0 (tn + l)} ⊆ K. We
consider the sequence {x0 (tn + l)} to be convergent. Let x = lim x0 (tn + l), then
n→+∞
x ∈ ωx0 ∩ K, but this contradicts the choice of x0 . Thus x0 t ∈ / K1 for all t ≥ t0 and,
hence, V (x0 t) ≤ V (x0 t0 ) for all t ≥ t0 . According to condition 1. of the theorem,
the set Σ+ x0 = {x0 t | t ≥ 0} is bounded and by virtue of complete continuity of
(X, T1 , π) it is relatively compact. From here it results that ωx0 6= ∅ is compact,
invariant and ωx0 ⊆ Er . Let p ∈ ωx0 and ϕ : S → ωx0 is the whole trajectory of
the system (X, T1 , π) passing through the point p ∈ ωx0 . Consider the function
µ : S+ → R defined by the equality µ(t) = V (ϕ(−t)). It is continuous, bounded

171
172 Global Attractors of Non-autonomous Dissipative Dynamical Systems

and non-decreasing and, consequently, there is c0 = lim µ(t). From here it results
t→+∞
that on the set αϕp := {q | ∃tn → +∞, ϕ(−tn ) → q} the function V is continuous,
−1
that is αϕ
p ⊆ V (c0 ) ∩ ωx0 . This contradicts the condition 3. of the theorem.
So ωx ∩ K 6= ∅ for all x ∈ E. According to Theorems 1.22 the non-autonomous
dynamical system h(X, T1 , π), (Y, T2 , σ), hi be boundedly k-dissipative.
Let now h(X, T1 , π), (Y, T2 , σ), hi be boundedly k-dissipative, J be its Levinson
center and r > r0 := sup{|x| : x ∈ J}. Let us define a function V : Er → R by the
equality

V (x) := sup{|xt| : t ≥ 0}. (5.1)

Directly from the definition of V it results that:

(1) V (x) ≥ |x| for all x ∈ Er and, therefore, it satisfies the condition 1. of the
theorem;
(2) if xτ ∈ Er for all τ ∈ [0, t] , then V (xt) ≤ V (x).

Let us show that the function V , defined by the equality (5.1), is continuous. Let
xn → x(x, xn ∈ Er ) and R := sup{|xn | | n ∈ N}, then from Theorem 2.19 for the
number R > 0 there exists l(R) > 0 such that

|xt| ≤ r (t ≥ l(R) and |x| ≤ R). (5.2)

From (5.1) and (5.2) it results that

V (x) = |xτ |

for some τ (x) ∈ [0, l(R)] (|x| ≤ R). The sequence {V (xn )} = {|xn τn |} is relatively
compact, where τn = τ (xn ). Let us show that it has a unique limit point. Let V 0
be one of the limit points of the sequence {V (xn )}, then there is a subsequence
{V (xkn )} such that V (xkn ) → V 0 for k → +∞. As {τkn } is bounded, then we
can consider it to be convergent. Then V 0 = |xτ 0 |, where τ 0 = lim τkn . Let us
n→+∞
show that |xτ 0 | = |xτ | = V (x). If we suppose that |xτ 0 | 6= |xτ |, then |xτ | > |xτ 0 |.
Let  > 0 be such that |xτ 0 | + 2 < |xτ |. Then for k sufficiently large we have
||xkn τ | − |xτ || <  and ||xkn τkn | − |xτ 0 || < , hence,

|xkn τ | > |τ | −  > |xτ 0 | +  > |xkn τkn | = V (xkn ). (5.3)

Inequality (5.3) contradicts the choice of the number τkn . Thus V (x) = |xτ 0 | and,
hence V (x) = lim V (xn ), so that V is continuous.
n→+∞
Finally, let us show that level lines of the function V do not contain ω-limit points
(X, T1 , π). Really, for any x ∈ E, we have ωx ⊆ J. According to the definition of
the set Er it has not common points with J and therefore, ωx ∩ V −1 (c) = ∅, for
any x ∈ Er and c ∈ R. This proves the theorem. 
 Method of Lyapunov functions 173

Lemma 5.1 Let (X, T1 , π) be completely continuous, h(X, T1 , π), (Y,T2 ,σ),hi be a
non-autonomous dynamical system and there exists r > 0 and a continuous function
V : Er → R which satisfies conditions:

1. the set {x ∈ E : V (x) ≤ c} is bounded for any c ∈ R;

2. if xτ ∈ Er for all τ ∈ [0, t], then V (xt) ≤ V (x).

Then the following conditions are equivalent:

3. level lines of the function V do not contain ω-limit points of the dynamical
system (X, T1 , π);
4. level lines of the function V do not contain positive semi-trajectories of the
dynamical system (X, T1 , π).

Proof. Let us show that Condition 3. implies Condition 4. If we suppose that it

is not so, then there will be c0 and x0 ∈ V −1 (c0 ) such that x0 t ∈∈ V −1 (c0 ) for all
t ≥ 0. In conditions of the lemma the level lines of the function V are bounded sets,
therefore Σ+ x0 is bounded. As (X, T1 , π) is completely continuous then the set Σx0
+

is relatively compact and, hence, ωx0 6= ∅, is compact and ωx0 ⊆ V −1 (c0 ). The last
is a contradiction to the condition 3. This contradiction proves the implication we
need.
Let us show that from the condition 4. follows the condition 3. Suppose the
opposite, that there is c0 ∈ R and x0 ∈ Er such that ωx0 t ∩ V −1 (c0 ) 6= ∅. Reasoning
−1
like in Theorem 5.1 we can show that ωx0 is compact and αϕ p ⊆ ω x0 ∩ V (c0 ) 6= ∅
for some complete motion ϕ (ϕ(0) = p). As αp is positive invariant, then V −1 (c0 )
ϕ

contains positive semi-trajectories of the system (X, T1 , π), but this contradicts the
condition 4. Lemma is proved. 

Corollary 5.1 Let Y be compact and the dynamical system (X, T1 , π) be

completely continuous. The non-autonomous dynamical system h(X, T1 , π),
(Y, T2 , σ), hi is boundedly k-dissipative if and only if there are number r > 0 and
continuous function V : Xr → R, satisfying the conditions:

(1) V (x) ≥ a(|x|) for all x ∈ Xr , where a ∈ A :={a | a : R+ →R+ , a is a continu-

ous and strongly monotonically increasing function } and Im V ⊆ Im (a);
(2) if xτ ∈ Xr for all τ ∈ [0, t], then V (xt) ≤ V (x);
(3) level lines of the function V do not contain positive semi-trajectories of
dynamical system (X, T1 , π).

Theorem 5.2 Let (X, T1 , π) be completely continuous. A non-autonomous

dynamical system h(X, T1 , π), (Y, T2 , σ), hi is boundedly k-dissipative if there is a
number r > 0 and a continuous function V : Er → R with the following properties:

(1) the set {x ∈ E | V (x) ≤ c} is bounded for any c ∈ R;

174 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(2) if xτ ∈ Er for all τ ∈ [0, t], then V (xt) < V (x).

Proof. Let K1 = {x ∈ E : |x| ≤ r}, l > 0, be such that π l K1 is relatively compact

and K = π l K1 . Let us show that ωx ∩ K 6= ∅ for all x ∈ E. Like in Theorem
5.1 it can be shown that in the opposite case we should have a point x0 ∈ E such
that Σ+ x0 is relatively compact, ωx0 6= ∅ is compact and ωx0 ∩ K = ∅. Furthermore
there will be c0 ∈ R and a complete motion ϕ : S → ωx0 (ϕ(0) = p) such that
−1
αϕp ⊆ ω x0 ∩ V (c0 ) 6= ∅. As αϕp is positively invariant, then together with the
ϕ
point x the set αp contains xt for all t > 0, and, hence, V (xt) < V (x) = c0 . The
contradiction we have got proves the assertion we need. Thus ωx ∩ K 6= ∅ for
all x ∈ E and according to Theorem 2.19 the non-autonomous dynamical system
h(X, T1 , π), (Y, T2 , σ), hi is boundedly k-dissipative. The Theorem is proved. 

Theorem 5.3 Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical

system and T1 = R+ . Assume that there is a number r > 0 and function
V : Er → R+ with the following properties:
1. the function V is bounded on bounded sets and for any c ∈ R the set {x ∈
Er | V (x) ≤ c} is bounded;
2. Vπ0 (x) ≤ −c(|x|) for all x ∈ Er , where c : R+ → R+ is strictly positive on
[r, +∞), and Vπ0 (x) = lim sup t−1 [V (xt) − V (x)].
t→+0

Then there exists R0 > 0 such that for any R > 0 there will be l(R) > 0 for which
|xt| ≤ R0 for all t ≥ l(R) and |x| ≤ R.

Proof. First let us show that for any x ∈ Er there will be τ = τ (x) > 0 such
that |xτ | < r. If that assertion is not true then there exists x0 ∈ Er such that
|x0 τ | ≥ r for all τ ≥ 0. Consequently, V (x0 τ ) ≤ V (x0 ) for all τ ≥ 0 and the
set Σ+ x0 := ∪{x0 τ | τ ≥ 0} is bounded. Let b0 := sup{|x0 τ | |τ ≥ 0} and ν =
inf{c(α) | r ≤ α ≤ b0 } and we note that Vπ0 (x0 t) ≤ −ν, therefore

V (x0 t) ≤ V (x0 ) − νt (5.4)

for all t > 0. The right hand side of (5.4) is negative for t sufficiently large, but this
contradicts positive definiteness of V on Er . This contradiction proves the assertion
we need.
Let b(R) := sup{V (x) : x ∈ Er , |x| ≤ R}, M (R) := {x ∈ Er | V (x) ≤ b(R)}
and R0 := sup{|x| | x ∈ M (R)}. According to condition 1. of the theorem the
set M (R) is bounded and, hence, R0 < +∞. Let us show that |xt| ≤ R0 for all
|x| ≤ r and t ≥ 0. In fact, if |x| ≤ r, |xβx | = |xγx | = r and |xt| > r for all
t ∈]βx , γx [, then V (xt) ≤ V (xβx ) ≤ b(R) and, hence, xt ∈ M (R). From this it
follows that |xt| ≤ R0 for all t ∈]βx , γx [ and |x| ≤ r, and therefore |xt| ≤ R0
for all t ≥ 0 and |x| ≤ r. According to the proved above for any x ∈ Er there
 Method of Lyapunov functions 175

will be τ (x) > 0 such that |xτ | < r. Suppose νx := sup{t | xτ ∈ Er for all
τ ∈ [0, t]}, then |xνx | = r and |xt| > r for all t ∈]0, νx [. Let us notice that for
any R > 0 the set B(R) := {xt | x ∈ Er , |x| ≤ R and t ∈]0, νx [} is bounded.
Indeed, V (xt) ≤ V (x) ≤ b(R) for all |x| ≤ R and t ∈]0, νx [, therefore xt ∈ M (R).
According to condition 1. of the theorem the set M (R) is bounded. Let us show
now that L(R) := sup{νx | |x| ≤ R} is finite for any R > 0. If the assertion is
not true, then there will be R > 0 and {xn } such that |xn | ≤ R and νxn → +∞.
Suppose d(R) := sup{|x| | x ∈ B(R)}. As the set B(R) is bounded, then we have
d(R) < +∞. Suppose γ(R) := inf{c(ν) | r ≤ ν ≤ d(R)} and let us notice that
Vπ0 (xt) ≤ −γ for all |x| ≤ R and t ∈]0, νx [ and, hence, V (xt) ≤ V (x) − γ(R)t. In
particular,

V (xn νxn ) ≤ b(R) − γ(R)νxn (5.5)

for all n. The right hand side of (5.5) is negative for n sufficiently large and this
contradicts the positive definiteness of V on Er . From this contradiction it follows
that for every R > 0 the number L(R) > 0 and is finite. Now for finishing the proof
of the theorem it is sufficient to notice that |xt| ≤ R0 for all |x| ≤ R and t ≥ L(R).


Corollary 5.2 Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical

system and T1 = R+ . Assume that there is a function V : E → R which satis-
fies the following conditions:

(1) V is bounded on the bounded sets from E;

(2) V (x) ≥ γ1 |x|m − D1 (γ1 , D1 , m > 0) for all x ∈ E;
(3) Vπ0 (x) ≤ −γ2 V (x) + D2 (γ2 , D2 > 0) for all x ∈ E.

Then there exists R0 > 0 such that for any R > 0 there will be l(R) > 0 for which
|xt| ≤ R0 for all |x| ≤ R and t ≥ l(R).

Proof. This assertion follows from Theorem 5.3. To prove that it is sufficient to
notice that for |x| sufficiently large conditions 1.–3. of Corollary 5.1 guarantee the
fulfillment of the conditions of Theorem 5.3. 

Corollary 5.3 Under the conditions of Theorem 5.3, if Y is compact and

(X, T1 , π) is asymptotic compact, then the non-autonomous dynamical system
h(X, T1 , π), (Y, T2 , σ), hi is boundedly k-dissipativ.

Proof. This statement follows from Theorems 5.3 and 2.21. 

Remark 5.1 The statement, analogously to Corollary 5.3, holds also for the
dynamical system with discrete time T1 = Z+ .
176 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Remark 5.2 Unlike Theorems 5.1 and 5.2 in Theorem 5.3 we do not demand the
continuity of the function V , but instead of that we impose certain condition on the
velocity of decreasing of the function V along trajectories of the system (X, T 1 , π).
In additional, when T1 = R+ the completeness of the space is not necessary (in the
applications there are examples, when Y is a fortiori incomplete).

Theorem 5.4 Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical

system, T1 = R+ , (X, T1 , π) be asymptotically compact and there is a continuous
function V : E → R satisfying the following conditions:
1. for all c ∈ R the set {x ∈ E| V (x) ≤ c} is bounded;
2. along trajectories of the system (X, T1 , π) the function V is non-increasing, i.e.
V (xt) ≤ V (x) for all x ∈ E and t ≥ 0;
3. there exists r > 0 such that V (xt) < V (x), if xτ ∈ Er for all τ ∈ [0, t] and
t > 0.

Then the non-autonomous dynamical system h(X, T1 , π), (Y, T2 , σ), hi is locally k-
dissipative.

Proof. First let us notice that as V is continuous, then in a point p ∈ X there is

Cp ∈ R and δp > 0 such that V (x) ≤ Cp for any x ∈ B(p, δp ). According to the
condition 2. V (xt) ≤ V (x) ≤ Cp for all x ∈ B(p, δp ) and t ≥ 0. According to the
condition 1. the set Mp := Σ+ (B(p, δp )) = ∪{π t B(p, δp )|t ≥ 0} is bounded and as
(X, T1 , π) is asymptotically compact, then there is a nonempty compact Kp such
that the equality

lim β(π t Mp , Kp ) = 0 (5.6)

t→+∞

takes place. Define K := {x ∈ X | |x| ≤ r}, where r > 0 is the number from the
condition 3. of the theorem. As Y is compact and as the fiber bundle (X, h, Y ) is
locally trivial, then the set K is bounded. Let us notice that ωp ∩ K 6= ∅ for any
point p ∈ X. In fact, if this were not so, we should have a point p0 ∈ X such that
ωp0 ∩ K = ∅. According to the equality (5.6) Σ+ t
p0 := ∪{π p0 | t ≥ 0} is relatively
compact and, hence, ωp0 6= ∅ is compact and invariant. The function µ : R+ → R
defined by the equality µ(t) := V (p0 t) is continuous, bounded and monotone non-
increasing and, hence, there is lim V (p0 t) = c0 . As V is continuous, we have
t→+∞
ωp0 ⊆ V −1 (c0 ). Let x ∈ ωp0 , then V (xt) = V (x) = c0 for all t > 0. The last
contradicts condition 3. of the theorem. This contradiction shows that ωp ∩ K 6= ∅
for all p ∈ X. Let us show that as a matter of fact ωp ⊆ K for all p ∈ X. Indeed, if
it is not so, then there will be t0 > 0 such that xτ ∈ ωp0 \ K for all τ ∈ [0, t0 ] and,
hence, V (xt0 ) < V (x). On the other hand, reasoning as we do above, we will have
V (x) = c0 for some c0 ∈ R and all x ∈ ωp0 . This contradicts the last inequality.
Thus Ω ⊆ K, where Ω := ∪{ωp |p ∈ X}, and, as Ω is invariant and as (X, T1 , π)
 Method of Lyapunov functions 177

is asymptotically compact, we conclude that Ω is compact and, hence, the system

(X, T1 , π) is pointwise k-dissipative. Taking into consideration (5.6) and Theorem
1.19 we conclude that the system (X, T1 , π) is locally k-dissipative. Theorem is
proved. 

Remark 5.3 a) From condition 3. it follows that level lines of the function V do
not contain ω-limit points of the dynamical system (X, T1 , π).
b) Theorem 5.4 takes place for discrete dynamical systems, too, if condition 3.
is changed by the following: level lines of the function V do not contain ω-limit
points of the dynamical system (X, T1 , π).

Proof. For proving the last assertion let us notice that in conditions of Theorem
11.2 the set K is a weak attractor and that is why Ω ⊆ K. Indeed, if this were not
so, then we should have x0 ∈ Er such that ωx0 \ K 6= ∅. Let p ∈ ωx0 \ K, then there
exits c0 such that ωx0 ⊆ V −1 (c0 ). Let c00 = V (p), then c00 = V (p) = c0 and, hence,
p ∈ V −1 (c0 ) ∩ Er ∩ ωxo . The last contradicts the condition of the theorem. 

Definition 5.1 h(X, T1 , π), (Y, T2 , σ), hi is said to be bounded, if for arbitrary
R > 0 there is C(R) > 0 such that |xt| ≤ C(R) for all |x| ≤ R and t ≥ 0.

Theorem 5.5 Let h(X, T1 , π), (Y, T2 , σ), hi be a non-autonomous dynamical

system, T1 = R+ and (X, T1 , π) is asymptotically compact. If there exist r > 0
and a continuous bounded on the bounded sets from X function V : Xr → R which
satisfies the following conditions:

1. for any c ∈ R the set {x ∈ Er | V (x) ≤ c} is bounded;

2. if xτ ∈ Er for all τ ∈ [0, t], then V (xt) ≤ V (x);
3. level lines of V do not contain ω-limit points of the dynamical system (X, T 1 , π).

Then the non-autonomous dynamical system h(X, T1 , π), (Y, T2 , σ), hi is boundedly
k-dissipative and bounded, i.e. for any R > 0 there exists C(R) ≥ 0 such that
|x| ≤ C(R) for all |x| ≤ R and t ≥ 0.

Proof. Like in Theorem 5.3 the existence of C(R) ≥ 0 such that |xt| ≤ C(R) for
all t ≥ 0 and |x| ≤ R (R ≥ r) is proved.
Let us show that under the conditions of Theorem 5.5 the non-autonomous
system (X, T1 , π), (Y, T2 , σ), hi is boundedly k-dissipative. As h(X, T1 , π),
(Y, T2 , σ), hi is bounded and asymptotically compact, then according to Lemma 1.4
Ω(K1 ) 6= ∅, is compact and attracts K1 := {x ∈ E | |x| ≤ r}. Suppose K := Ω(K1 )
and let us show that ωx ∩ K 6= ∅ for any x ∈ X. If this were not so, we should
have x0 ∈ X such that ωx0 ∩ K = ∅. Under the conditions of Theorem 5.5 the set
Σ+x0 := {x0 t | t ≥ 0} is relatively compact and, hence, ωx0 6= ∅ and is compact.
178 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Suppose d = ρ(ωx0 , K) > 0, then

d d
B(ωx0 , ) ∩ B(K, ) = ∅ (5.7)
3 3
and there will be l > 0 such that
d
π t K1 ⊆ B(K, ) (5.8)
3
for all t ≥ l. Let us show that x0 t ∈
/ K1 for all t ≥ t0 , where t0 is some nonnegative
number. If it is not so then there will be tn → +∞, such that x0 tn ∈ K1 and
according to (5.8) x0 (tn + l) ∈ B(K, d3 ). We consider the sequence {x0 (tn + l)} to
be convergent. Suppose p = lim x0 (tn + l), then p ∈ ωx0 ∩ B(K, d3 ). The last
n→+∞
contradicts (5.7), and it proves the assertion we need. Further, reasoning like in
Theorem 5.1 we will have that for some c0 ∈ R the set ωx0 ∩ V −1 (c0 ) 6= ∅. The last
contradicts condition 3. of the theorem. The contradiction shows that ωx ∩ K 6= ∅
for all x ∈ X. As the dynamical system h(X, T1 , π), (Y, T2 , σ), hi is bounded and
asymptotically compact, then (X, T1 , π) satisfies the condition of Ladyzhenskaya.
Now for finishing the proof of Theorem 5.5 it is sufficient to refer to Theorem 2.20.
Theorem is proved. 

Remark 5.4 a) The condition 1. of Theorem 5.5 is equivalent to positive defi-

niteness V on Xr and to its boundedness on bounded subsets from X;
b) The assertion which is the opposite to Theorem 5.5 takes place and can be
proved like Theorem 5.1.
c) Lemma 5.1 remains correct if the condition of complete continuity of
(X, T1 , π) is changed by the condition of Ladyzhenskaya. That is why Theorem
5.5 is correct in the case when condition 3. is changed by the following: level lines
of the function V do not contain ω- limit points of the dynamical system (X, T1 , π).
d) Theorem 5.5 remains valid if conditions 2. and 3. are changed by the follow-
ing: if xτ ∈ Er for all τ ∈ [0, t] (t > 0), then V (xt) < V (x), and it can be proved
like Theorem 5.2.
e) Theorem 5.5 takes place for discrete dynamical systems if the function V is
defined on the whole space and condition 2. takes place for all t ≥ 0 and x ∈ E.
For autonomous systems with discrete time this assertion is contained in [174].

5.2 Some criterions of dissipativity of differential equations

1. Let E n be n-dimensional Euclidean space, | · | is the norm on E n generated by

the scalar production of h·, ·i. Consider the differential equation

u̇ = f (t, u), (5.9)

 Method of Lyapunov functions 179

where f ∈ C(R × E n , E n ) is a regular function. Along with equation (5.9) let us

consider its H-class

v̇ = f (t, v) (g ∈ H(f )). (5.10)

As it was mentioned above, the equation (5.9) naturally generates a non-

autonomous dynamical system (see example 3.1). Applying to it the results con-
cerning general non-autonomous dynamical system we obtain a series of statements
about the equation (5.9). From theorems 5.1, 5.2, Lemma 5.1 and Corollary 5.1
follow Theorems below.
Let r > 0 and Ern := {u ∈ E n | |u| ≥ r}.

Theorem 5.6 Let H(f ) be compact. For the equation (5.9) to be dissipative
it is necessary and sufficient that would exist r > 0 and continious function V :
H(f ) × Ern → R+ satisfying the conditions:
(1) V (g, v) ≥ a(|v|) for all v ∈ Ern and g ∈ H(f ), where a ∈ A and Im V ⊆ Im a;
(2) if |ϕ(τ, v, g)| ≥ r for all τ ∈ [0, t] then V (gt , ϕ(t, v, g)) ≤ V (g, v) ( g ∈ H(f ));
(3) the lines of the level of the function V do not contain positive semitrajectories
of equation (5.10).

Theorem 5.7 Let H(f ) be compact. For the equality (5.9) to be dissipative it
is necessary and sufficient that there would exist r > 0 and a continious function
V : H(f ) × Ern → R+ , satisfying the following conditions:
(1) V (g, v) ≥ a(|v|) for all v ∈ Ern g ∈ H(f ), where a ∈ A and Im V ⊆ Im a;
(2) if |ϕ(τ, v, g)| ≥ r for all τ ∈ [0, t] then V (gt , ϕ(t, v, g)) < V (g, v) (g ∈ H(f )).

Theorem 5.8 Let H(f ) be compact. Then the following conditions are equivalent:
1. the equation (5.9) is dissipative;
2. there exists a number r > 0 possessing the following property: for every v ∈ E n
and g ∈ H(f ) there exists τ = τ (v, g) > 0 such that, |ϕ(τ, v, g)| < r;
3. there exists such number R1 > 0 that lim inf |ϕ(t, v, g)| < R1 for all v ∈ E n
t→+∞
and g ∈ H(f );
4. there exists a number R0 > 0 and for every R > 0 there is l(R) > 0, such that
|ϕ(t, v, g)| ≤ R0 for all t ≥ l(R), |v| ≤ R and g ∈ H(f ).

Proof. The statement formulated above follows from Theorems 2.19 and 3.1. In the
case when f is periodic the equivalence of the conditions 1., 2. and 4. is established
in [270]. The equivalence of the conditions 1. and 2. for nonperiodic equationsis
established in [163]. 

Definition 5.2 We will say that the function F ∈ C(R × E n , E n ) satisfies the
local condition of Lipschitz u ∈ E n uniformly on t ∈ R or just satisfies the condition
180 Global Attractors of Non-autonomous Dissipative Dynamical Systems

of Lipschitz if for every r > 0 there exists L(r) > 0 such that |F (t, u1 ) − F (t, u2 )| ≤
L|u1 − u2 | for all t ∈ R and u1 , u2 ∈ B[0, r].

Theorem 5.9 Let f ∈ C(R × E n , E n ) satisfy the condition of Lipschitz, H(f )

be compact and let exist R > 0 and continuous differentiable function V ∈ C(R ×
Ern , R+ ) satisfying the following conditions:
a. V satisfies the condition of Lipschitz;
b. a(|u|) ≤ V (t, u) ≤ b(|u|) (a, b ∈ A, Im a = Im b, t ∈ R and u ∈ Ern );
c. Mf ⊆ MV ∩ M ∂V ∩ Mgradu V (where Mf := {{tn } : {ftn } is convergent } and
∂t
by analogy for MV , M ∂V and Mgradu V );
∂t
∂V
d. V̇f (t, u) := (t, u) + hgradu V, f (t, u)i ≤ 0 (t ∈ R and u ∈ Ern );
∂t
e. whatever would be (Ve , g) ∈ H(V, f ) := {(Vτ , fτ ) : τ ∈ R} the lines of level of
the function Ve do not contain positive semi-trajectories of the equation (5.10)
lying out of the ball B[0, r].
Then the equation (5.9) is dissipative.

Proof. We will execute the proof in few steps.

1. Let g ∈ H(f ). There exists {tk } ⊂ R such that ftk → g. Since Mf ⊆ MV ,
the sequence {Vtk } is also convergent. Assume Ve := lim Vtk . From the conditions
k→+∞

a.– e. follows that the function Ve satisfies the condition of Lipschitz, Vė g (t, v) :=
∂ Ve
(t, v) + hgradv Ve , gi ≤ 0 and besides a(|v|) ≤ Ve (t, v) ≤ b(|v|) for all v ∈ Ern and
∂t
t ∈ R.
2. Let v ∈ Ern and g ∈ H(f ). Denote by ϕ(·, v, g) the solution of the equation
(5.10) passing through the point v for t = 0. Let us show that this solution is
extendable to the right onto the whole semi-axis R+ . Obviously, for this it is
sufficient to show that it is bounded on the whole domain of its existence [0, tv ].
Let R = R(r) > 0 be such that a(R) > b(r), T1 (v) := {t ∈ [0, tv [: |ϕ(t, v, g)| ≤
r} and T2 (v) := [0, tv [ \ T1 (v). It is clear that T2 (v) is open and, consequently,
S
T2 (v) = ]tα , tβ [ (β = β(α)). For every t ∈ T2 (v) there exists α such that
α
t ∈]tα , tβ [, |ϕ(tα , v, g)| = |ϕ(tβ , v, g)| = r, |ϕ(t, v, g)| > r and, consequently,

a(|ϕ(t, v, g)|) ≤ Ve (t, ϕ(t, v, g)) ≤ Ve (tα , ϕ(tα , v, g))

≤ b(|ϕ(tα , v, g)|) = b(r) ≤ a(R) (5.11)

From the inequality (5.11) follows that |ϕ(t, v, g)| ≤ R and we have

sup{|ϕ(t, v, g)| : t ∈ [0, tv [} ≤ r0 := max(r, R(r)).

Since f ∈ C(R × E n , E n ) satisfies the condition of Lipshitz, it is regular.

 Method of Lyapunov functions 181

3. Let h(X, R+ , π), (Y, R+ , σ), hi be a non-autonomous dynamical system gener-

ated by the equation (5.9). Define on Xr := Ern × H(f ) a function V : Xr → R+ by
the equality V(v, g) = Ve (0, v) where Ve is the function defined in the point 1. Note
that the function V(·, g) is defined uniquely by g ∈ H(f ) in virtue of the inclusion
Mf ⊆ M V .
4. Show that the introduced in the previous point function V satisfies all the
conditions of Corollary 5.1. Let us establish its continuity on Xr = Ern × H(f ). Let
gk → g and vk → v ((vk , gk ) ∈ Ern × H(f )). Then

|V(vk , gk ) − V(v, g)| ≤ |V(vk , gk ) − V(v, gk )| (5.12)

+|V(v, gk ) − V(v, g)| ≤ Lgk (R)|vk − v| + |V(v, gk ) − V(v, g)|,

where Lgk (R) is the constant of Lipschitz for V (·, gk ) on B(0, R) and R := sup |vk |.
k∈N
Since Lgk (R) ≤ Lf (R) for all g ∈ H(f ) (see, for example, [95]) and Mf ⊆ MV ,
V(v, gk ) → V(v, g) and passing to limit in (5.12) with k → +∞ we establish the
continuity of V in the point (v, g) ∈ Xr .
Other conditions of Corollary 5.1 under the conditions of the theorem are obvi-
ously fulfilled. To finish the proof it is sufficient to refer to Corollary 5.1. 

Note, that in the case of periodicity of f the statement of Theorem 5.9 reinforces
Theorem 2.5 from [270].
2. Consider the linear homogeneous equation

u̇ = A(t)u, (5.13)

where A ∈ C(R, [E n ]) and [E n ] is the set of all linear operators A : E n → E n . By

U (t, A) we denote the operator of Cauchy of the equation (5.13). Along with the
equation (5.13) we will consider the perturbed equation

u̇ = A(t)u + F (t, u) (F ∈ C(R × E n , E n )). (5.14)

The following theorem takes place.

Theorem 5.10 For the equation (5.14) to be dissipative the fulfillment of the
following condition is sufficient:

(1) the function F satisfies the condition of Lipschitz;

(2) H(A, F ) := {(Aτ , Fτ ) : τ ∈ R} is compact in C(R, [E n ]) × C(R × E n , E n );
(3) there exist positive numbers N and ν such that

kU (t, A)U −1 (τ, A)k ≤ N e−ν(t−τ ) (t ≥ τ ); (5.15)

(4) |F (t, u)| ≤ A + ε|u| (∀u ∈ E n , t ∈ R) where A > 0 and 0 ≤ ε ≤ ε0 < νN −2 .

182 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Proof. Let (B, G) ∈ H(A, F ). Define WB ∈ [E n ] by the equality

Z +∞
WB := U ∗ (τ, B)U (τ, B)dτ. (5.16)
0

From (5.15) and (5.16) follows that ||WB || ≤ N 2 (2ν)−1 . Denote by a :=

sup{||A(t)|| : t ∈ R} and note that

|u| = |U (−τ, Bτ )U (τ, B)u| ≤ eaτ |U (τ, B)u|.

From this
Z
+∞ Z
+∞
1
hWB u, ui = 2
|U (τ, B)u| dτ ≥ e−2aτ |u|2 dτ = |u|2 .
2a
0 0

So, we established the following inequality

1 N2 2
|u|2 ≤ hWB u, ui ≤ |u| (u ∈ E n , B ∈ H(A)).
2a 2ν
Let Y := H(A, F ), (Y, R, σ) be the dynamical system of shifts on Y, X := E n ×
Y and π t (v; B, g) := (ϕ(t, v, B, g); Bt , gt ) (v ∈ E n , (B, G) ∈ H(A, F )), where
ϕ(·, v, B, G) is the solution of the equation

v̇ = B(t)v + G(t, v), (5.18)

passing through the point v as t = 0. Let us consider a non-autonomous dynamical

system h(X, R, π), (Y, R, σ), hi where h := pr2 : X → Y . Define on X a function
V : X → R+ by the following rule:


V (v; B, G) := WB v, v (5.19)

for all v ∈ E n and (B, G) ∈ H(A, F ). Note that this function possesses the following
properties:
1 2 N2 2
a. |v| ≤ V (v; B, G) ≤ |v| (v ∈ E n , (B, g) ∈ H(A, F )).
2a 2a
b. V is continuous.

In fact, if vk → v and (Bk , Gk ) → (B, G) then



|V (vk ; Bk , Gk ) − V (v; B, G)| = | WBk vk , vk − WB v, v | (5.20)




= | WBk (vk − v), vk + WBk v, vk − v + (WBk − WB )v, v |
N2 N2
≤ |vk ||vk − v| + |v||v − vk | + ||WBk − WB |||v|2 .
2ν 2ν
From the inequality (5.20) we can easily see that for V to be continuous it is sufficient
to show that ||WBk − WB || → 0 if Bk → B. Let us establish the latter. From (5.16)
 Method of Lyapunov functions 183

follows that WB is a self-conjugate operator. That is why


||WBk − WB || = sup | (WBk − WB )v, v | ≤ (5.21)
|v|=1

Z
+∞ Zl
≤ sup ||[U (τ, Bk ) − U (τ, B)]v||2 dτ = sup |[U (τ, Bk ) − U (τ, B)]v|2 dτ
|v|≤1 |v|≤1
0 0

Z
+∞

+ sup |[U (τ, Bk ) − U (τ, B)]v|2 dτ ≤ sup ||U (τ, Bk ) − U (τ, B)||2 l
|v|≤1 0≤τ ≤l
l

Z
+∞

+2 (||U (τ, Bk )||2 + ||U (τ, B)||2 )dτ.

l

From (5.15) follows the inequality ||U (τ, B)|| ≤ N e−ντ for all τ ∈ R+ and
B ∈ H(A) (see [51, p.70]). Hence, from (5.21) we obtain
2N 2 −2vl
||WBk − WB || ≤ sup ||U (τ, Bk ) − U (τ, B)||2 l + e . (5.22)
0≤τ ≤l v
Passing to limit as k → +∞ and taking in consideration that the first term in the
right hand side of (5.22) vanishes with any fixed l > 0 we establish that
2N 2 −2vl
lim sup ||WBk − WB || ≤ e . (5.23)
k→+∞ v
In this case (5.23) takes place for all l > 0. Making l → +∞ and taking in
consideration that the left hand side of the equation (5.23) does not depend on l
we will obtain the result needed. So, the continuity of V is established.
c. Let now |ϕ(τ ; v, B, G)| ≥ r (r > N 2 A(v − ε0 N 2 )−1 ) for all τ ∈ [0, t]. Then
V (ϕ(t; v, B, G), Bt , Gt ) < V (v; B, G). To establish this inequality we calculate the
derivative of the function


V(τ ) := V (ϕ(τ ; v, B, G), Bτ , Gτ ) = WB ϕ(τ ; v, B, G), ϕ(τ ; v, B, G)

in τ ∈ R. By standard reasoning, taking into account the identity

d
WBτ + B ∗ (τ )WBτ + WBτ B(τ ) = −E (5.24)
dτ
(E is the unit operator in E n ), we have
d
V (ϕ(τ ; v, B, G), Bτ , Gτ ) = −|ϕ(τ ; v, B, G)|2 +
dτ
 (5.25)
2Re G(τ, ϕ(τ ; v, B, G)), WBτ ϕ(τ, v, B, G) .
184 Global Attractors of Non-autonomous Dissipative Dynamical Systems

From (5.25) we obtain

d
V (ϕ(τ ; v, B, G), Bτ , Gτ ) ≤ −|ϕ(τ ; v, B, G)|2
dτ
N2
+ 2 |ϕ(τ ; v, B, G)|(A + ε0 |ϕ(τ ; v; B, G)|)
v
N2 N 2A 1
≤ |ϕ(τ ; v, B, G)|2 (−1 + ε0 + ).
v v r
From this follows that V (ϕ(t; v, B, G), Bt , Gt ) < V (v, B, G). In the same way
as in Theorem 5.9 we prove the possibility of non-local extension to the right of
all the solutions of the equation (5.18); their uniqueness follows from the condition
of Lipschitz for F . To finish the proof of the theorem it is sufficient to refer to
Theorem 5.9 (see also Theorem 5.2). 

Remark 5.5 a. Theorem 5.10 coincides with Theorem 3.21, which is obtained
with the help of a set of a priori estimations. However, Theorem 3.21 takes place
also for the equations defined in some Banach space.
b. Theorem 5.11 remains valid also in the case when we replace the condition
4. by: |F (t, u)| ≤ A + B|u|α (∀u ∈ E n , t ∈ R) where A and B are some positive
numbers and 0 ≤ α ≤ 1. We execute the proof of this statement by the same scheme
that Theorem 5.10 choosing r such that the condition r > r0 would take place, where
r0 is the solution of the equation Ar −1 + Brα−1 = ν 2 N −2 .
c. Note that the statement of the point b., generally speaking, does not take place
if α > 1. This is proved by the example ẋ = −x + x3 .

Theorem 5.11 Let f ∈ C(R × E n , E n ) satisfies the condition of Lipschitz, H(f )

be compact and let exist A ∈ C(R, [E n ]) satisfying the following conditions:
dA(t)
1. Mf ⊆ MA ∩ MȦ (Ȧ(t) := );
∗

dt 
2. A(t) = A (t) and α|u| ≤ A(t)u, u ≤ β|u|2 u ∈ E n for all t ∈ R and
2

α, β > 0;


3. Ȧ(t)u, u ≤ 0 (u ∈ E n , t ∈ R);


4. there exists r > 0 such that Re A(t)u, f (t, u) ≤ −γ(|u|) for all u ∈ Ern t ∈ R+
where γ(s) > 0 as s ≥ r.

Then the equation (5.9) is dissipative.

Proof. According to Theorem 5.9 the function f is regular. Let (X, R+ , π),

(Y, R, σ), h be a non-autonomous dynamical system generated by the equation (5.9)
(see example 3.1) and g ∈ H(f ). There exists {tn } ∈ Mf such that g = lim ftk
k→+∞
and under the conditions of the theorem {Atk } is convergent. Denote its limit
by Bg . We define now on Xr := Ern × Y a function V by the following rule:
 Method of Lyapunov functions 185


V (v, g) = Bg (0)v, v for every v ∈ Ern and g ∈ H(f ). Note that Bg is defined
uniquely because Mf ⊆ MA . The continuity of V is proved in the same way that
in Theorems 5.7 and 5.8.
Let now |ϕ(τ, v, g)| ≥ r for all τ ∈ [0, t] (t > 0). Let us calculate the derivative
of V (ϕ(τ, v, g), gτ ) in τ ∈ R. For this note that from the inclusion Mf ⊆ MA ∩ MȦ
follows the equality
dV d

(ϕ(τ, v, g), gτ ) = B(τ )ϕ(τ, v, g), ϕ(τ, v, g) (5.26)
dt dτ



= Ḃ(τ )ϕ(τ, v, g), ϕ(τ, v, g) + 2Re B(τ )ϕ(τ, v, g), g(τ, ϕ(τ, v, g)) .

From the conditions 1.–4. of the theorem we obtain



Ḃ(t)v, v ≤ 0 and Re B(t)v, g(t, v) ≤ −γ(|v|) (5.27)

for all v ∈ Ern , B ∈ H(A), t ∈ R and g ∈ H(f ). From (5.26) and (5.27) follows
d
that V (ϕ(τ, v, g), gτ ) ≤ −2γ(|ϕ(τ, v, g)|), hence, V (ϕ(τ, v, g), gτ ) < V (v, g) for
dτ n
all v ∈ Er and g ∈ H(f ). To finish the proof it is sufficient to refer to Corollary 5.2
and Theorem 5.2. 

Corollary 5.4 If the equation (5.9) is τ -periodic and the conditions of Theorem
5.9 are fulfilled then the equation (5.9) has at least one τ -periodic solution.

The particular case of Corollary 5.4 is announced in the work [192].

Corollary 5.5 Let ai ∈ C(R, R). If H(ai ) is compact (i = 0, 2n + 1) and

a2n+1 (t) ≤ −α (α > 0) for all t ∈ R+ then the equation

v̇(t) = a0 (t) + a1 (t)v + · · · + a2n+1 v 2n+1 (5.28)

is dissipative.

Proof. The formulated statement follows from Theorem 5.9. For this it sufficient
to take as A(t) a unit operator and to notice that


A(t)v, f (t, v) = v(a0 (t) + a1 (t)v + · · · + a2n+1 (t)v 2n+1 ) =
 a2n (t) a0 (t)  α
v 2n+2 a2n+1 (t) + + · · · + 2n+1 ≤ − v 2n+2
v v 2
for all |v| ≥ r and t ∈ R and when r > 0 is big enough. 

Corollary 5.6 If functions ai (i ∈ 0, 2n + 1) are τ -periodic and a2n+1 (t) > 0 for
all t ∈ [0, τ [ then the equation (5.28) has at least one τ -periodic solution.
186 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Note that in the case when a2n+1 (t) < 0 (t ∈ [0, τ [) according to Corollary 5.6
the equation (5.28) is dissipative and, consequently, by Theorem 2.3 from [270] it
has at least one τ -periodic solution. The case a2n+1 (t) > 0 (t ∈ [0, τ [) is reduced
to the previous one by the replacement of t by −t.
In the case when a2n+1 ≡ 1 Corollary 5.6 coincides with the result established
in [270, p.126].

Corollary 5.7 Let f = (f1 , f2 , . . . , fn ) ∈ C(R × E n , E n ) be continuously dif-

∂fi
ferentiable. If ≤ −m < 0 (i ∈ 1, n), H(f ) is compact and functions
∂xi
fi (t, x1 , x2 , . . . , xi−1 , 0, xi+1 , . . . , xn ) are bounded, then the equation (5.9) is dissi-
pative.

Proof. From the equality

Zxi
∂fi
fi (t, x1 , x2 , . . . , xn ) = (t, x1 , x2 , . . . , xn )dxi
∂xi
0
+ fi (t, x1 , x2 , . . . , xi−1 , 0, xi+1 , . . . , xn ),

in the conditions of Corollary 5.7 follows that for all x ∈ Rn the inequality
xi fi (t, x1 , x2 , . . . , xn ) ≤ −mx2i + M |xi | holds, hence,
n
X m 2
hx, f (t, x)i = xi fi (t, x) ≤ −m|x|2 + M n|x| ≤ − |x|
i=1
2

2M n
for all t ∈ R and |x| ≥ . According to Theorem 5.6 the equation (5.9) is
m
dissipative. 

Note that in the case of periodicity of f Corollary 5.7 is well known [270]. When
f is not periodic the statement close to Corollary 5.7 is established in [264] with
some additional restrictions on f .

Theorem 5.12 Let the condition of uniqueness be fulfilled for every equation
(5.10) with g ∈ Σf := {fτ : τ ∈ R}. For the disiipaivity of the equation (5.9)
there is sufficient existence of a function V : R × Ern → R+ (r > 0) satisfying the
conditions:

(1) V satisfies the condition of Lipschitz;

n
(2) a(|u|) ≤ V (t, u) ≤ b(|u|) h (a, b ∈ A, Im a = Im b) wheni u ∈ Er ;
−1
(3) V̇f (t, u) := lim sup τ V (t + τ, u + τ f (t, u)) − V (t, u) ≤ −c(|u|) for u ∈ Enn ,
τ ↓0
where c : R+ → R+ is continuous and c(s) > 0 as s > r.
 Method of Lyapunov functions 187

Proof. In the same way as in Theorem 5.9 we can show that the solution ϕ(·, u, fτ )
of the equation (5.10) (g = fτ ) passing through u as t = 0 is extendable to the
right onto the whole semi-axis R+ . Assume Y := Σf and by (Y, R+ , σ) denote
the dynamical system defined by the following rule: π t (u, fτ ) := (ϕ(t, u, fτ ), fτ +t )
for all τ ∈ R, u ∈ E n and t ∈ R+ . Define on Xr := Ern × Y a function V
by the equality V(u, fτ ) := V (τ, u). Then V(π t (u, fτ )) = V(ϕ(t, u, fτ ), ft+τ ) =
V (t + τ, ϕ(t, u, fτ )) = V (s, ϕ(s − τ, u, fτ )) = V (s, x(s; u, τ )) (s = t + τ ) and,
consequently,
V̇π (u, fτ ) = lim sup t−1 [V(π t (u, fτ )) − V(u, fτ )]
t↓0
= lim sup(s − τ )−1 [V (s, x(s; u, τ )) − V (τ, x[τ ; u, τ ))]
s−τ ↓0
= V̇f (τ, x(τ ; u, τ )) = V̇f (τ, u) ≤ −c(|u|) (|u| ≥ r).

According to Theorem 5.3 the non-autonomous dynamical system (X, R+ , π),


(Y, R+ , σ), h (h = pr2 : X → Y ) is dissipative and, consequently, there exist R > 0
and l(u, τ ) > 0 such that |ϕ(t, u, fτ )| < R for all u ∈ E n , τ ∈ R and t ≥ l(u, τ ). As
it is in [137] we can show that the number l(u, τ ) > 0 can be chosen not depending
on τ ∈ R. Reasoning in the same way that in the proof of Lemma 3.1, we obtain
that the equation (5.9) is dissipative. 
Note that Theorem 5.12 makes precise the known theorem of T. Yoshizava
[137],[325]–[326].

5.3 Theorem of Barbashin–Krasovskii for non-autonomous dynamical

systems

The ideas and methods that we applied studying dissipative system in first two para-
graphs of Chapter 5 allow formulating and proving of a series of statements about
the stability of non-autonomous dynamical system and, in particular, generalizing
of the known theorem of Barbashin–Krasovskii [21] on asymptotic stability.
Let (X, h, Y ) be a finite-dimensional vectorial bundle fiber, h(X, T1 , π),
(Y, T2 , σ), hi be non-autonomous dynamical system, θy ∈ Xy := h−1 (y) be a null
element (|θy | = 0) and Θ := {θy | y ∈ Y } be a null section of the vectorial bun-
dle fiber (X, h, Y ). Below we will suppose that the null section Θ is invariant, i.e.
Θ ⊆ X is an invariant set of the dynamical system (X, T1 , π).

Definition 5.3 the null section Θ is called uniformly stable if for every ε > 0
there exists δ(ε) > 0 such that y ∈ Y, x ∈ Xy and |x| < δ implies |xt| < ε for all
t ≥ 0.
188 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Definition 5.4 If Θ is uniformly stable and lim |xt| = 0 for all x ∈ X, then
t→+∞
the null section is called globally uniformly asymptotically stable.

Theorem 5.13 Let Y be compact. For the null section Θ to be globally uniformly
asymptotically stable it is necessary and sufficient that there would exist a continuous
function V : X → R+ satisfying the following conditions:

1. V (x) ≥ a(|x|) for all x ∈ X, V (θy ) = 0 for all y ∈ Y and Im a = Im V, where

a ∈ A.
2. V (xt) ≤ V (x) for all x ∈ X and t ≥ 0.
3. the lines of level of V do not contain non-null ω-limit points of the dynamical
system (X, T, π).

Proof. Sufficiency. Let the conditions of the theory be satisfied. Show that the
null section Θ is uniformly stable. Suppose that it is not true. Then there exist
ε0 > 0, |xn | < δ, δn ↓ 0 and tn ≥ 0 such that

|xn tn | ≥ ε0 . (5.29)

On the other hand, 0 ≤ a(|xn tn |) ≤ V (xn tn ) ≤ V (xn ) → 0 as n → +∞ and,

consequently, |xn tn | → 0. The last contradicts to the equality (5.29).
Now let us show that lim |xt| = 0 for all x ∈ X. Really, if we suppose the
t→+∞
contrary then there exists x0 ∈ X (|x0 | 6= 0) such that lim sup |x0 t| > 0, i.e. there
t→+∞
exist ε0 > 0 and tn → +∞ for which

|x0 tn | ≥ ε0 . (5.30)

Note that Σ+x0 is relatively compact. In fact, a(|x0 t|) ≤ V (x0 t) ≤ V (x0 ) and,
consequently, |x0 t| ≤ a−1 (V (x0 )) for all t ≥ 0. So, the sequence {x0 tn } can be
considered convergent. Assume x̃ := lim x0 tn , then x̃ ∈ ωx0 . Reasoning in the
n→+∞
same way that in Theorem 5.1 we can show that there exists c ≥ 0 for which
V (x) = c for all x ∈ ωx0 . Since x̃ ∈ ωx0 and x̃ = lim x0 tn , from (5.30) follows
n→+∞
that |x̃| > 0, hence c = V (x̃) ≥ a(|x̃|) > 0. So, the lines of level of the function V
contain non-null ω–limit points. This contradicts to the conditions of the theorem.
The global uniform asymptotic stability of the null section is proved.
Necessity. Let the null section Θ be globally uniformly asymptotically stable.
Let V : X → R+ be the function, defined by the equality (5.1). Directly from the
definition of V follows that it satisfies to the conditions 1. and 2. of the theorem.
We will show that it is continuous. For that we note that under conditions of


Theorem the non-autonomous system (X, T1 , π), (Y, T2 , σ), h is dissipative. Now
we will show that J ⊆ Θ, where J is the center of Levinson of (X, T1 , π). Indeed, if
x ∈ J then according to Theorem 1.2 the semi-trajectory Σ+ x can be extended to the
 Method of Lyapunov functions 189

left, i.e. there exists a continuous mapping ϕ : S → J such that π t ϕ(s) = ϕ(t + s)
for all t ∈ T1 , s ∈ S and ϕ(0) = x. Since J is compact, the set of α–limit points
αϕx of the motion ϕ is not empty, compact and invariant. Obviously, αϕx ∩ Θ 6= ∅
and, consequently, there exists tn → −∞ such that |ϕ(tn )| → 0. Let ε > 0 and
δ(ε) > 0 be chosen out of the condition of the uniform stability of Θ. Then for n
big enough we have |ϕ(tn )| < δ and, consequently, |π t ϕ(tn )| = |ϕ(t + tn )| < ε for
all t ≥ 0. In particular, |x| = |ϕ(tn − tn )| < ε. And since ε is arbitrary, we have
|x| = 0, i.e. J ⊆ Θ.
We will show now that for every r > 0 the following equality takes place

lim sup |xt| = 0. (5.31)

t→+∞ |x|≤r

Suppose that it is not so; then there exist ε0 > 0, r0 > 0, |xn | < r0 and tn → +∞
such that

|xn tn | ≥ ε0 . (5.32)


In virtue of dissipativity of the system (X, T1 , π), (Y, T2 , σ), h the sequence {xn tn }
can be considered convergent. Assume x0 := lim xn tn and note that x0 ∈ J,
n→+∞
hence |x0 | = 0. The last contradicts to (5.32). So, the equality (5.31) takes place.
The continuity of V is established literally repeating the reasoning of Theorem 5.1
when instead of (5.2) we use the condition (5.31).
As for the 3rd condition, we verify it in the same way that in Theorem 5.1. The
theorem is proved. 

Note that Theorem 5.13 is valid if we replace X by some tubular neighborhood

Ur := {x ∈ X : |x| ≤ r} (r > 0) of the null section Θ.

Definition 5.5 The null section Θ is called uniformly asymptotically stable if it

is uniformly stable and there exists r > 0 such that lim |xt| = 0 for all x ∈ Ur .
t→+∞

If we slightly change the proof of Theorem 5.13, we can prove the following
theorem.

Theorem 5.14 Let Y be compact. For the null section Θ to be uniformly asymp-
totically stable it is necessary and sufficient that there would exist r > 0 and a
continuous function V : Ur → R+ satisfying the following conditions:

(1) V (x) ≥ a(|x|) for all x ∈ Ur , V (θy ) = 0 for all y ∈ Y ;

(2) V (xt) ≤ V (x) if xτ ∈ Ur for all τ ∈ [0, t];
(3) the lines of level of V do not contain non-null ω-limit points of the dynamical
system (X, T1 , π).
190 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Remark 5.6 As well as in Lemma 5.1 we can show that in Theorem 5.14 the
condition 3. can be replaced by the following one: the lines of level of the function
V do not positive semi-trajectories of the dynamical system (X, T1 , π).

At last note that the following theorem takes place.

Theorem 5.15 For the null section Θ would be uniformly stable it is necessary
and sufficient that there would exist r > 0 and a function V : Ur → R+ satisfying
the following conditions:

1. V (x) ≥ a(|x|) for all x ∈ Ur and lim V (x) = 0;

|x|→0
2. V (xt) ≤ V (x) if xτ ∈ Ur for all τ ∈ [0, t].

Proof. Sufficiency. Let us show that the null section is uniformly stable. Suppose
that it is not so. Then there exist ε0 > 0 (ε0 < r), δn ↓ 0, |xn | < δn and tn ≥ 0
such that xn τ ∈ Ur for all τ ∈ [0, tn ] and

|xn tn | ≥ ε0 .

Note that

a(ε0 ) ≤ a(|xn tn |) ≤ V (xn tn ) ≤ V (xn ). (5.33)

Passing to limit in (5.33) as n → +∞, we obtain a(ε0 ) ≤ 0 and, consequently,

ε0 = 0. The last contradicts to the choice of ε0 . So, the uniform stability of Θ is
proved.
Necessity. Let Θ be uniformly stable. For ε0 = 1 select δ0 = δ(ε0 ) > 0 (δ0 ≤ 1)
out of the condition of the uniform stability of Θ. Assume r := δ0 and define
V : Ur → R+ by the equality (5.1). It is easy to see that the function V is the one
in question, i.e. it satisfies the conditions 1. and 2. of Theorem 5.15. The theorem
is proved. 

Remark 5.7 By the conditions of Theorem 5.15 the function V , generally speak-
ing, is not continuous, unlike the case in Theorems 5.13 and 5.14.

From the theorems given in this section we can get the corresponding statements
for the equation (5.9). Let us previously bring the following definitions.
Let f (t, 0) ≡ 0 and the function f ∈ C(R × E n , E n ) be regular.

Definition 5.6 We will say that the null solution of the equation (5.9) is uniformly
stable if for every ε > 0 there exists δ(ε) > 0 such that g ∈ H(f ), |v| < δ implies
|ϕ(t, v, g)| < ε for all t ≥ 0.
 Method of Lyapunov functions 191

Definition 5.7 If the null solution of the equation (5.9) is uniformly stable and
there exists γ > 0 such that lim |ϕ(t, v, g)| = 0 for all g ∈ H(f ) and v ∈ B(0, γ),
t→+∞
then the null solution of the equation (5.9) is called uniformly asymptotically stable.

Definition 5.8 The null solution of (5.9) is globally uniformly asymptotically

stable if it is uniformly stable and |ϕ(t, v, g)| → 0 as t → +∞ for any g ∈ H(f ) and
v ∈ En.

Note that from the results of the works [8],[291] follows the equivalence of the
standard definition of the uniform stability (global uniform asymptotic stability)
and of the one given above.
From Theorems 5.13, 5.14 and Remark 5.6 follows the theorems below.

Theorem 5.16 Let H(f ) be compact and f (t, 0) ≡ 0. For the null solution of the
equation (5.9) to be globally uniformly stable it is necessary and sufficient that there
would exists a continuous function V : H(f ) × E n → R+ satisfying the following
conditions:
(1) V (g, v) ≥ a(|v|) for all v ∈ E n , V (g, 0) = 0, g ∈ H(f ) and Im a = Im V where
a ∈ A;
(2) V (gτ , ϕ(τ, v, g)) ≤ V (g, v) for all g ∈ H(f ), v ∈ E n and τ ≥ 0;
(3) whatever would be the function g ∈ H(f ), the lines of level of the function V do
not contain positive semi-trajectories of the equation (5.10) excepting the trivial
one.

Theorem 5.17 Let H(f ) be compact and f (t, 0) ≡ 0. For the null solution of the
equation (5.9) to be uniformly asymptotically stable it is necessary and sufficient
that there would exist a continuous function V : H(f ) × B[0, r0 ] → R+ satisfying
the following conditions:
(1) V (g, v) ≥ a(|v|) and V (g, 0) = 0 for all g ∈ H(f ) and x ∈ B[0, r0 ], where a ∈ A;
(2) V (gt , ϕ(t, v, g)) ≤ V (g, v) if ϕ(t, v, g) ∈ B[0, r0 ] for all τ ∈ [0, t];
(3) whatever would be the function g ∈ H(f ), the lines of level of the function V do
not contain positive semi-trajectories of the equation (5.10) excepting the trivial
one.

From Theorems 5.16 and 5.17 we can get the sufficient conditions of asymptotic
stability suitable for the applications.

Theorem 5.18 Let H(f ) be compact, f (t, 0) ≡ 0 and let there be a continuously
differentiable function V ∈ C(R × E n , R+ ) satisfying the conditions:
(1) a(|u|) ≤ V (t, u) ≤ b(|u|) (a, b ∈ A, Im a = Im b) for all t ∈ R and u ∈ E n ;
(2) Mf ⊆ MV ∩ M ∂V ∩ Mgradu V ;
∂t
192 Global Attractors of Non-autonomous Dissipative Dynamical Systems

∂

(3) V̇f (t, u) := V (t, u) + gradu V, f ≤ 0 (t ∈ R, u ∈ E n );
∂t
(4) whatever would be the function g ∈ H(f ), the lines of level of the function V do
not contain positive semi-trajectories of the equation (5.10) excepting the trivial
one.

Then the null solution of the equation (5.9) is uniformly asymptotically stable in
general.

Theorem 5.19 Let r > 0, H(f ) be compact, f (t, 0) ≡ 0 and let exist a con-
tinuously differentiable function V ∈ C(R × B[0, r], R+ ) satisfying the following
conditions:

(1) a(|u|) ≤ V (t, u) ≤ b(|u|) (a, b ∈ A) for all t ∈ R and x ∈ B[0, r];
(2) Mf ⊆ MV ∩ M ∂V ∩ Mgradu V ;
∂t
∂

(3) V̇f (t, u) := V (t, u) + gradu V (t, u), f (t, u) ≤ 0 (t ∈ R, u ∈ E n );
∂t
(4) whatever would be the function g ∈ H(f ), the lines of level of the function V do
not contain positive semi-trajectories of the equation (5.10) excepting the trivial
one.

Then the null solution of the equation (5.9) is uniformly asymptotically stable.

Proof. The proof of Theorems 5.18 and 5.19 is constructed after the same pattern
as that in Theorem 5.9. 

Statements similar to Theorem 5.19 have been obtained in the works [4],[193].

5.4 Equations with convergence

1. Let us give the criterions of the convergence of the nonlinear equation (5.9) re-
sulting from Theorems 2.7, 2.12 and 2.13, Corollary 2.2 and Theorems 2.16 and 2.17
by applying the last theorems to the non-autonomous dynamical system generated
by the equation (5.9).

Theorem 5.20 Let f ∈ C(R × E n , E n ) be recurrent w.r.t. t ∈ R uniformly w.r.t.

u on compacts from E n [300]. For the equation (5.9) to be convergent it is necessary
and sufficient that the following conditions would be satisfied:

(1) the equation (5.9) has at least one bounded on R+ solution;

(2) lim |ϕ(t, v1 , g) − ϕ(t, v2 , g)| = 0 for all g ∈ H(f ) and v1 , v2 ∈ E n ;
t→+∞
(3) for every ε > 0 and r > 0 there exists δ = δ(ε, r) > 0 such that |v 1 − v2 | < δ
(|v1 |, |v2 | < δ) implies |ϕ(t, v1 , g) − ϕ(t, v2 , g)| < ε for all t ≥ 0 and g ∈ H(f ).
 Method of Lyapunov functions 193

Note that in the case of almost periodicity of the function f Theorem 5.20
generalizes and refines the criterion of almost periodic convergence of V.I.Zubov
[336].

Theorem 5.21 Let f ∈ C(R × E n , E n ) be recurrent w.r.t. t ∈ R uniformly w.r.t.

u on compacts from E n . For the equation (5.9) to be convergent it is necessary and
sufficient that the following conditions would be fulfilled:

1) the equation (5.9) has at least one bounded on R+ solution;

2) whatever would be g ∈ H(f ) and v ∈ E n , the solution ϕ(·, v, g) of the equation
(5.10) is asymptotically stable.

In the case of periodicity of the function f Theorem 5.21 coincides with the
theorem of N.N.Krasovskii-V.A.Pliss [270].

Theorem 5.22 Let f ∈ C(R × E n , E n ) be recurrent w.r.t. t ∈ R uniformly w.r.t.

u on compacts from E n and the equation (5.9) be dissipative. For the equation (5.9)
to be convergent it is necessary and sufficient that there would exists a continuous
function V : H(f ) × E n × E n → R+ satisfying the following conditions:

1) V is positively defined, i.e. V (g, v1 , v2 ) = 0 (g ∈ H(f ) and v1 , v2 ∈ E n ) if and

only if v1 = v2 ;
2) V (gt , ϕ(t, v1 , g), ϕ(t, v2 , g)) ≤ V (g, v1 , v2 ) for all t ≥ 0, v1 , v2 ∈ E n and g ∈
H(f );
3) V (gt , ϕ(t, v1 , g), ϕ(t, v2 , g)) = V (g, v1 , v2 ) for all t ≥ 0 if and only if v1 , v2 ∈ E n .

In the case of the periodicity of the function f Theorem 5.22 coincides with
Theorems 7.2 and 7.3 from [270].

Theorem 5.23 Let f ∈ C(R × E n , E n ) be recurrent w.r.t. t ∈ R uniformly

w.r.t. u on compacts from E n and let the equation (5.9) be dissipative. For the
equation (5.9) to be convergent it is necessary and sufficient that there would exist a
continuous function V : H(f ) × E n × E n → R+ satisfying the following conditions:

(1) V is positively defined;

(2) V (gt , ϕ(t, v1 , g), ϕ(t, v2 , g)) < V (g, v1 , v2 ) for all t > 0, g ∈ H(f ) and v1 , v2 ∈
E n , v1 6= v2 ;

2. Let E be a Banach space with the norm | · | and [E] be the Banach space of
all linear bounded operators acting from E to E and equipped with the operational
norm.
Consider linear non-homogeneous equation

u̇ = A(t)u + f (t), (5.34)

194 Global Attractors of Non-autonomous Dissipative Dynamical Systems

where A ∈ C(R, [E]) and f ∈ C(R, E). Along with the equation (5.34) we will
consider the corresponding homogeneous equation

u̇ = A(t)u (5.35)

and denote by U (t, A) the Cauchy operator of equation (5.35).

In this paragraph we will suppose that the sets H(A) and H(f ) are compact
in C(R, [E]) and C(R, E) respectively, though some of the statements given below
take place even without this condition.

Lemma 5.2 For the equation (5.34) to be convergent it is necessary and sufficient
that there would exist positive numbers N and ν such that

kU (t, A)U −1 (τ, A)k ≤ N exp−ν(t−τ ) (5.36)

for all t ≥ τ (t, τ ∈ R).

Proof. The formulated statement follows from the definition of convergence and
from the fact that uniform asymptotic stability of the null solution of the equation
(5.35) is equivalent to the condition (5.36). 

Lemma 5.3 The mapping U : R × C(R, [E]) → [E] (U : (t, A) → U (t, A)) is
continuous w.r.t. A uniformly w.r.t. t ∈ R on compacts from R, i.e. for every
` > 0 lim max kU (t, An ) − U (t, A)k = 0 if An → A in C(R, [E]).
n→+∞ |t|≤`

Proof. Let An ⊂ C(R, [E]), An → A be uniform on compacts from R and ` > 0.

Then there exists positive number M (`) such that

max kAn (t)k ≤ M (`). (5.37)

|t|≤`

Since U (t, An ) is the solution of the system

(
U̇ (t, An ) = An (t)U (t, An )
U (0, An ) = IdE ,

where IdE is a unit operator in E, then we have

Z t
kU (t, An )k ≤ exp( kAn (s)kds) (t > t0 , t, t0 ∈ [−`, `]).
t0

From the last inequality follows that

max kU (t, An )k ≤ exp(2`M (`)) (n = 1, 2, . . . ).

|t|≤`
 Method of Lyapunov functions 195

For every n ∈ N the function Vn (t) := U (t, A) − U (t, An ) satisfies the system
(
V̇n (t) = A(t)Vn (t) + [A(t) − An (t)]U (t, An )
Vn (0) = 0.

From this follows that

Z t
Vn (t) = U (t, A) U −1 (τ, A)[An (τ ) − A(τ )]U (τ, An )dτ. (5.38)
0

Let K(`) := max{kU (t, A)k, kU −1(t, A)k}. From (5.37) and (5.38) follows the in-
|t|≤`
equality

max kVn (t)k ≤ K 2 (l)2` exp(2`M (`)) max kAn (t) − A(t)k. (5.39)
|t|≤` |t|≤`

Passing to limit in (5.39) as n → +∞, we obtain

lim max kU (t, A) − U (t, An )k = 0.

n→+∞ |t|≤l

The lemma is proved. 

Lemma 5.4 Let N > 0 and ν > 0 be such that there holds the inequality (5.36).
Then for every B ∈ H(A), t ≥ τ and τ ∈ R, the inequality

kU (t, B)U −1 (τ, B)k ≤ N exp(−ν(t − τ )), (5.40)

holds where U (t, B) is the Cauchy operator of the equation

ẏ = B(t)y (B ∈ H(A)).

Proof. Let N > 0 and ν > 0 be such that there takes place the inequality (5.36)
and let B ∈ H(A). Then there exists {τn } ⊂ R such that Aτn → B in C(R, [E]).
Note that U (t + τ, A) = U (t, Aτ )U (τ, A) for all t, τ ∈ R and A ∈ C(R, [E]), hence

U (t, Aτn )U −1 (τ, Aτn ) = U (t + τn , A)U −1 (t + τn , A). (5.41)

From what follows that

kU (t, Aτn )U −1 (τ, Aτn ) ≤ N exp(−ν(t − τ )) (t ≥ τ, n = 1, 2, . . . ).

Passing to limit in the last inequality as n → +∞ and taking into account Lemma
5.3 we will obtain the inequality (5.40). The lemma is proved. 

Lemma 5.5 Let the condition (5.36) be fulfilled. Then for every B ∈ H(A) and
p ≥ 1 by the equality
Z +∞
1
kvkB,p = { |U (t, B)v|p dt} p (5.42)
0
196 Global Attractors of Non-autonomous Dissipative Dynamical Systems

on E is defined a norm that is topologically equivalent to the old and moreover

a 1
kU (τ, B)v1 − U (τ, B)v2 kBτ ,p ≤ N ( ) p exp(−ντ )kv1 − v2 kB,p (5.43)
ν
for all τ ≥ 0.

Proof. From the inequality of Minkovskii follows that by the equality (5.42) is
defined some norm on E. Let us show that there exist positive number mp and Mp
that does not depend on B ∈ H(A) such that mp |v| ≤ kvkB,p ≤ Mp |v|. The first
inequality follows from (5.36) and from the inequality
Z +∞ Z +∞
kvkpB,p = p
|U (t, B)v| dt ≤ kU (t, B)kp dt|v|p ≤
0 0
Z +∞
p Np p
N exp(−νpt)dt|v|p = |v| .
0 νp

On the other hand since U (−τ, Bτ )U (τ, B)v = v for all τ ∈ R and v ∈ E, then

|v| = |U (−τ, Bτ )U (τ, B)v| ≤ kU (−τ, Bτ )k|U (τ, B)v| ≤

exp(a|τ |)|U (τ, B)v| (a := sup{kA(t)k : t ∈ R} < +∞)

and, consequently,
Z +∞ Z +∞
kvkpB,p = |U (t, B)v|p dt ≥ (exp(−at)|v|)p dt =
0 0
Z +∞
1
exp(−apt)dt|v|p = |v|p
0 ap
1
(mp = (ap)− p ). So,
1 1
(ap)− p |v| ≤ kvkB,p ≤ N (νp)− p |v|.

Now let us establish the inequality (5.43). For that we will note that
Z +∞
kU (τ, B)v1 − U (τ, B)v2 kpBτ ,p = |U (t, Bτ )[U (τ, B)v1 − U (τ, B)v2 ]|p dt
0
Z +∞ Z +∞
= |U (t, Bτ )U (τ, B)(v1 − v2 )|p dt = |U (t + τ, B)(v1 − v2 )|p dt
0 0
Z +∞ Z +∞
= |U (s, B)(v1 − v2 )|p ds ≤ N p exp(−νps)|v1 − v2 |p ds
τ τ
Np N p ap
= exp(−νpτ )|v1 − v2 |p ≤ exp(−νpτ )kv1 − v2 kpB,p .
νp νp

From the last inequality follows (5.43). The lemma is proved. 

 Method of Lyapunov functions 197

As it turned out, if the equation (5.34) is convergent then this property remains
valid under small nonlinear perturbations. The following theorem takes place.

Theorem 5.24 Let A, f and F be such that H(A), H(f ) and H(F ) are compact
in C(R, [E]), C(R, E) and C(R × E, E) respectively, and the function F satisfies
the condition of Lipschitz w.r.t. u ∈ E uniformly w.r.t. t ∈ R with the small
enough constant of Lipschitz (Lip(F ) < ν 2 (N a)−1 ). Then if the equation (5.34) is
convergent, the perturbed equation

u̇ = A(t)u + F (t, u) (5.44)

is convergent too.

Proof. Denote by H(A, f, F ) := {(Aτ , fτ , Fτ ) : τ ∈ R}. Let v ∈ B, (B, g, G) ∈

H(A, f, F ), τ ∈ R and ϕ(·, v, B, g, G) be the solution of equation

v̇ = B(t)v + g(t) + G(t, v), (5.45)

passing through the point v as t = 0.

According to the conditions of the theorem the equation (5.45) has a unique
solution defined on R and passing through the point v as τ = 0, whatever would
be (B, g, G) ∈ H(A, f, F ) and v ∈ E. To see that is sufficient to note that having
the conditions of Theorem 5.24 fulfilled we can apply Theorem 1.2 [132] (global
theorem of existence and uniqueness) to the equation (5.45).
Assume Y := H(A, f, F ) and by (Y, R, σ) denote the dynamical system of shifts
on Y . Let X := E × Y . Define on X a dynamical system by the following rule:
π τ (v; B, g, G) := (ϕ(τ, v, B, g, G); Bτ , gτ , Gτ ). By the conditions of Theorem 5.24
Y is compact and the triplet h(X, R, π), (Y, R, σ), hi, where h := pr2 : X → Y
is a non-autonomous dynamical system. Let us show that we can apply Theorem
2.15 and Corollary 2.6 to the constructed dynamical system. Really, the set Y
is compact under the conditions of the theorem. It is clear that the set of the
continuous sections Γ(H(A, f, F ), B × H(A, f, F )) is isomorphic to the set of all the
continuous mapping x : H(A, f, F ) → E (see paragraph 3.3) and, consequently, is
not empty. Define on X ×X ˙ a scalar non-negative function V by the following rule:
Z +∞
V ((v1 ; B, g, G), (v2 ; B, g, G)) := kv1 − v2 kB,1 = |U (t, B)(v1 − v2 )|dt
0

for every (B, g, G) ∈ H(A, f, F ) and v1 , v2 ∈ E. From Lemma 5.5 follows that
the constructed function satisfies the conditions 1.- 3. of Theorem 2.15. Let
ϕ(·, vi , B, g, G) be the solution of the equation (5.45) passing through the point
vi (i = 1, 2) as τ = 0. Then the identity

ϕ̇(τ, vi , B, g, G) ≡ B(τ )ϕ(τ, vi , B, g, G) + g(τ ) + G(τ, ϕ(τ, vi , B, g, G)).

198 Global Attractors of Non-autonomous Dissipative Dynamical Systems

holds. From this follows that

Z τ
ϕ(τ, vi , B, g, G) = U (τ, B)(vi + U −1 (s, B) (5.46)
0
[g(s) + G(s, ϕ(s, vi , B, g, G))]ds).

From (5.46) and Lemmas 5.4 and 5.5 follows

kϕ(τ, v1 , B, g, G) − ϕ(τ, v2 , B, g, G)kBτ =

Z +∞
|U (t, Bτ )(ϕ(τ, v1 , B, g, G) − ϕ(τ, v2 , B, g, G))|dt
0
Z +∞ Z τ
= |U (t, Bτ )U (τ, B)(v1 − v2 + U −1 (s, B) ×
0 0
[G(s, ϕ(s, v1 , B, g, G)) − G(s, ϕ(s, v2 , B, g, G))]ds|dt ≤
Z +∞
|U (t + τ, B)(v1 − v2 )|dt +
0
Z +∞ Z τ
| U (t + τ, B)U −1 (s, B) ×
0 0
[G(s, ϕ(s, v1 , B, g, G)) − G(s, ϕ(s, v2 , B, g, G))]ds|dt ≤
Z +∞ Z τ
N
exp(−ντ )|v1 − v2 | + N exp(−ν(t + τ − s))Lip(G) ×
ν 0 0
|ϕ(s, v1 , B, g, G) − ϕ(s, v2 , B, g, G)|ds dt =
N N
exp(−ντ )|v1 − v2 | + exp(−ντ )Lip(G) ×
Zν ν
τ
exp(νs|ϕ(s, v1 , B, g, G) − ϕ(s, v2 , B, g, G)|ds ≤
0
N Na
exp(−ντ )akv1 − v2 kB,1 + exp(−ντ )Lip(G) ×
Zν ν
τ
exp(νs)kϕ(s, v1 , B, g, G) − ϕ(s, v2 , B, g, G)kB,1 ds.
0

Let us multiply the both parts of the last inequality by exp(ντ ) and assuming
ϕ(τ ) := exp(νs)kϕ(s, v1 , B, g, G) − ϕ(s, v2 , B, g, G)kB,1 , we will obtain

Z τ
Na Na
ϕ(τ ) ≤ kv1 − v2 kB,1 + Lip(G) ϕ(τ )ds. (5.47)
ν ν 0

Applying to (5.47) the Gronwall-Bellman inequality we will obtain

Na Na
ϕ(τ ) ≤ kv1 − v2 kB,1 exp( Lip(G)τ ). (5.48)
ν ν
 Method of Lyapunov functions 199

Consequently,

kϕ(τ, v1 , B, g, G) − ϕ(τ, v2 , B, g, G)kB,1 ≤ (5.49)

Na Na
|v1 − v2 | exp(−(ν − Lip(G))τ ).
ν ν
From the inequality (5.49) we have

V ((ϕ(τ, v1 , B, g, G); Bτ , gτ , Gτ )), (ϕ(τ, v2 , B, g, G); Bτ , gτ , Gτ )) ≤

M exp(−λτ )V ((v1 ; B, g, G), (v2 ; B, g, G))

(here M = Nνa , λ = − Nνa Lip(G) + ν > 0) for all v1 , v2 ∈ E, τ ≥ 0 and (B, g, G) ∈

H(A, f, F ). So, all the conditions of Theorem 2.15 and of Corollary 2.6 are fulfilled.
Applying them to the constructed non-autonomous dynamical system we complete
the proof of the theorem. 

Corollary 5.8 Under the conditions of Theorem 5.24 if A, f and F are jointly
τ -periodic (almost periodic, recurrent) then the equation (5.44) is convergent and
every equation (5.45) has a unique τ -periodic (almost periodic, recurrent) globally
uniformly asymptotically stable solution.

Note, that the statements analogous to Lemmas 5.4 and 5.5 in the case when
the operator-function A(t) is stationary are proved in [132]. Corollary 5.8 in the
case of periodicity of A, f and F (E = Rn ) are established in [270, p.99].
3. Let H be a real or complex Hilbert space. Consider a differential equation

u̇ = f (t, u), (5.50)

where f ∈ C(R × H, H). Along with the equation (5.50) we will consider its H-class

v̇ = g(t, v) (g ∈ H(f )). (5.51)

There takes place

Theorem 5.25 Let f ∈ C(R × H, H) and let the set H(f ) be compact in C(R ×
H, H). If there exists a self-adjoint operator-function A ∈ C(R, [H]) satisfying the
following conditions:

1. Mf ⊂ MA ∩ MȦ ;
2. RehA(t)(u − v), f (t, u) − f (t, v)i ≤ −α|u − v| for all t ∈ R and u, v ∈ H (h·, ·i
is a scalar product in H, | · |2 = h·, ·i and α > 0 );
3. kȦ(t)k ≤ β for all t ∈ R (β ≥ 0);
4. hA(t)u, ui ≥ γ|u|2 for all t ∈ R u ∈ H and −2α + β := −λ < 0.
200 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Then the equation (5.50) is convergent, i.e.

a. whatever would be g ∈ H(f ) and v ∈ H, there exists a unique solution ϕ(·, v, g)
of the equation (5.51) passing through the point v as t = 0 and defined on R;
b. for every g ∈ H(f ) the equation (5.51) has a unique compact solution ψ g (t) =
ϕ(t, vg , g), defined on R;
c. there exist positive numbers N and ν (not depending on g ∈ H(f ) and v ∈ H)
such that

|ϕ(t, v, g) − ϕ(t, vg , g)| ≤ N exp(−νt)|v − vg | (5.52)

for all g ∈ H(f ), v ∈ H and t ∈ R+ .

Proof. Let us prove the formulated statement by several parts.

1. Since Mf ⊂ MA ∩ MȦ , then from [300] – [302] follows the existence of the
continuous mapping q : H(f ) → H(A) × H(Ȧ) satisfying the condition
˙ τ)
q(f ) = (A, Ȧ) and q(gτ ) = (q(g)τ , q(g)

for all g ∈ H(f ) and τ ∈ R.

2. Let us show that under the conditions of the theorem every function g ∈
H(f ) satisfies the conditions analogous to 1.–4. Really, let g ∈ H(f ) and q(g) =
(Bg , Ḃg ) ∈ H(A) × H(Ȧ). Then there exists {τn } ⊂ R such that g = lim fτn and
n→+∞
(Bg , Ḃg ) = ( lim Aτn , lim Ȧτn ). From the condition 2. of the theorem follows
n→+∞ n→+∞
that

RehA(t + τn )(u − v), f (t + τn , u)f (t + τn , v)i ≤ −α|u − v|2 (5.53)

for all t ∈ R, u v ∈ H (n=1,2,. . . ). Passing to limit in (5.53) as n → +∞ we will

obtain

RehBg (t)(u − v), g(t, u) − g(t, v)i ≤ −α|u − v|2 .

In the same way we can show that the operator-function Bg ∈ H(A) (g ∈ H(f ))
satisfies the conditions 3. and 4. with the same constants β and γ that has the
operator-function A ∈ C(R, [H]).
3. A slight modification of the reasoning in the work [267] (see also [329]) allows
to prove that if the conditions of the theorem are fulfilled then, whatever would
be u ∈ H, the equation (5.50) has a unique solution ϕ(·, u, f ) defined on R and
satisfying the condition ϕ(0, u, f ) = u. Taking into account that along with the
function f every other function g ∈ H(f ) satisfies the conditions of the theorem,
we make the conclusion that for every v ∈ H and g ∈ H(f ) the equation (5.51) has
a unique solution ϕ(·, v, g) defined on R and satisfying the condition ϕ(0, v, g) = v.
4. Assume Y := H(f ) and by (Y, R, σ) denote a dynamical system of shifts on
Y . Let X := H × H(f ). Define on X a dynamical system by the following rule:
 Method of Lyapunov functions 201

π((v, g), τ ) := (ϕ(τ, v, g), gτ ). Note that under the conditions of the theorem the
set Y is compact. The triplet h(X, R, π), (Y, R, σ), hi, where h := pr2 : X → Y ,
is a non-autonomous dynamical system. Let us show that we can apply to the
constructed dynamical system Theorem 2.15 and Corollary 2.6. Really, the set Y
is compact under the conditions of the theorem. The set of continuous sections
Γ(Y, X) is not empty. Define on X ×X ˙ a non-negative scalar function V by the
following rule:

V ((v1 , g), (v2 , g)) := hBg (0)(v1 − v2 ), v1 − v2 i

for every g ∈ H(f ) and v1 , v2 ∈ H (Bg = q(g)).

It is clear that the function V satisfies the conditions 2. and 3. of Theorem 2.15.
Really, according to the condition 4.

γ|v1 − v2 |2 ≤ hA(t)(v1 − v2 ), v1 − v2 i ≤ δ|v1 − v2 |2 (5.54)

for all v1 , v2 ∈ H and t ∈ R, where δ := sup{kA(t)k | t ∈ R} (δ < +∞, as Mf ⊂ MA

and H(f ) is compact). For g ∈ H(f ) there exists {τn } ⊂ R such that Aτn → Bg .
From (5.54) and the last correspondence follows that

γ|v1 − v2 |2 ≤ hBg (t)(v1 − v2 ), v1 − v2 i ≤ δ|v1 − v2 |2

for all g ∈ H(f ), t ∈ R and v1 , v2 ∈ H. In particular, we have

γ|v1 − v2 |2 ≤ V ((v1 , g), (v2 , g)) ≤ δ|v1 − v2 |2 (5.55)

for all g ∈ H(f ) and v1 , v2 ∈ H.

At last, let us show that the condition 4. of Lemma 2.13 is fulfilled too. Really,

V ((ϕ(τ, v1 , g), gτ ), (ϕ(τ, v2 , g), gτ )) =

hBgτ (0)(ϕ(τ, v1 , g) − ϕ(τ, v2 , g)), ϕ(τ, v1 , g) − ϕ(τ, v2 , g)i =
hBg (τ )(ϕ(τ, v1 , g) − ϕ(τ, v2 , g)), ϕ(τ, v1 , g) − ϕ(τ, v2 , g)i.

By direct calculation we can verify that

V̇ ((ϕ(τ, v1 , g), gτ ), (ϕ(τ, v2 , g), gτ )) =

hḂg (τ )(ϕ(τ, v1 , g) − ϕ(τ, v2 , g)), ϕ(τ, v1 , g) − ϕ(τ, v2 , g)i +
2RehBg (τ )(ϕ(τ, v1 , g) − ϕ(τ, v2 , g)), g(τ, ϕ(τ, v1 , g)) − g(τ, ϕ(τ, v2 , g))i.

From the last equality and (5.53) and also from the condition 3. of Theorem 5.25
follows the inequality

V̇ ((ϕ(τ, v1 , g), gτ ), (ϕ(τ, v2 , g), gτ )) ≤ (5.56)

2 2
β|ϕ(τ, v1 , g) − ϕ(τ, v2 , g)| − −2α|ϕ(τ, v1 , g) − ϕ(τ, v2 , g)| =
−λ|ϕ(τ, v1 , g) − ϕ(τ, v2 , g)|2 .
202 Global Attractors of Non-autonomous Dissipative Dynamical Systems

On the other hand, from the inequality (5.55) we obtain

V ((ϕ(τ, v1 , g), gτ ), (ϕ(τ, v2 , g), gτ )) ≤ δ|ϕ(τ, v1 , g) − ϕ(τ, v2 , g)|2 . (5.57)

From the inequality (5.56) and (5.57) we have

V̇ ((ϕ(τ, v1 , g), gτ ), (ϕ(τ, v2 , g), gτ )) ≤ −λδV ((ϕ(τ, v1 , g), gτ ), (ϕ(τ, v2 , g), gτ )).

Integrating the last inequality, we get the following inequality

V ((ϕ(τ, v1 , g), gτ ), (ϕ(τ, v2 , g), gτ )) ≤ exp(−λδτ )V ((v1 , g), (v2 , g))

for all g ∈ H(f ), v1 , v2 ∈ H and τ ∈ R+ . So, all the conditions of Theorem 2.15 are
fulfilled and applying it and Corollary 2.6 to the constructed dynamical system we
complete the proof of Theorem 5.25. 

Corollary 5.9 Under the conditions of Theorem 5.25 if the right hand side f
of the equation (5.50) is τ -periodic (almost periodic, recurrent), then the equation
(5.50) is convergent and every equation (5.51) has a unique τ -periodic (almost pe-
riodic, recurrent) solution complying with the estimation (5.52).

Remark 5.8 1. In the case when f is τ -periodic, Corollary 5.9 reinforces the
main result of the work [192] (in [192] under the same conditions on basis of the
principle of Lere-Showder there was proved only the existence of a unique τ -periodic
solution in the case when H = E n ).
2. If the right hand side f is almost periodic and the operator-function A(t)
does not depend on time (Ȧ(t) ≡ 0), then the statement of Corollary 5.9 coincides
with the main result of the work [267]. Also note, that the results of the work [267]
are achieved on basis of the series of delicate estimations. Our method is of purely
topological character.
3. Note, at last, that in the work [187] there are studied so-called α-contracting
non-autonomous dynamical systems (i.e. such systems that satisfy the condition:

ρ(x1 t, x2 t) ≤ αρ(x1 , x2 )

for every t ∈ R+ , x1 , x2 ∈ X and h(x1 ) = h(x2 ), where 0 ≤ α ≤ 1) in the relation

with the problem of stability and asymptotic stability in the sense of Poisson of the
solutions of differential equations with monotone right hand side.

In the case of finite-dimensional Hilbert space (H = E n ), when the function f

is recurrent and A(t) (Ȧ(t) ≡ 0) is stationary, Theorem 5.25 admits the following
amplification.

Theorem 5.26 Let f ∈ C(R × H, H) satisfy the condition of Lipschitz and be

recurrent w.r.t. t ∈ R uniformly w.r.t. u on compacts from H. If there exists a
 Method of Lyapunov functions 203

self-adjoint operator A ∈ [H] and a positively defined function c : R+ → R+ such

that

RehA(u − v), f (t, u) − f (t, v)i ≤ −c(|u − v|) (∀ u, v ∈ H) (5.58)

and c(r)r−1 → +∞ as r → +∞ (for example, c(r) = r 1+α , α > 0), then Equation
(5.50) is convergent.

Proof. In the same way that in Theorem 5.25 we establish that whatever would
be v ∈ H and g ∈ H(f ), the equation (5.51) has a unique solution ϕ(·, v, g) defined
on R and satisfying the condition ϕ(0, v, g) = v. Let us define a function V :
H(f ) × H → R+ by the following rule

V(g, v) := hAv, vi. (5.59)

From the equality (5.59) follows that

V̇(gτ , ϕ(τ, v, g)) = 2RehAϕ(τ, v, g), ϕ(τ, v, g)i. (5.60)

Since c(r)r−1 → +∞ as r → +∞, there exists r0 > 0 such that

c(r)r−1 > |f0 |kAk (5.61)

for all r > r0 , where |f0 | := sup{|f (t, 0)| : t ∈ R}. Let t > 0, v ∈ H and g ∈ H(f )
be such that |ϕ(τ, v, g)| ≥ r0 for all τ ∈ [0, t]. From (5.58), (5.60) and (5.61) follows
that V(gτ , ϕ(τ, v, g)) < V(g, v), and according to Theorem 5.7 the equation (5.50)
is dissipative.
Let us denote by V : H(f ) × H × H → R+ a function defined by the equation

V (g, u, v)) := hA(u − v), (u − v)i. (5.62)

By direct calculation we obtain

V̇ (gτ , ϕ(τ, u, g), ϕ(τ, v, g)) = 2RehA(ϕ(τ, u, g) − ϕ(τ, v, g)),

g(τ, ϕ(τ, u, g)) − g(τ, ϕ(τ, v, g))i. (5.63)

from (5.58) and (5.63) follows that V (gt , ϕ(t, u, g), ϕ(t, v, g)) < V (g, u, v) for all
t > 0, g ∈ H(f ) and u, v ∈ H (u 6= v). According to Theorem 2.17 the equation
(5.50) is convergent. 
204 Global Attractors of Non-autonomous Dissipative Dynamical Systems

5.5 Dissipativity and convergence of some equations of 2nd and

3rd order

1. Let us consider the system of two differential equation of the following kind
(
ẋ = y − F (x) + Q(x, y, t)
(5.64)
ẏ = −g(x),

where F and g are continuous functions defined on R and Q(x, y, t) is continuous

w.r.t. the set of variables, is bounded and uniformly continuous w.r.t. t ∈ R
uniformly w.r.t. (x, y) on compacts from R2 . As we know, we can reduce the
equation

ẍ + h(x)ẋ + g(x) = p(t) (5.65)

to the system (5.64) if we introduce a new variable y = ẋ + F (x) − g(t) where

Rx Rt
F (x) = 0 h(τ )dτ , g(t) = 0 p(s)ds. Along with the system (5.64) consider its
H-class
(
ẋ = y − F (x) + G(x, y, t)
(5.66)
ẏ = −g(x),

(G ∈ H(Q) = {Qτ : τ ∈ R}).

Theorem 5.27 Let the following conditions be satisfied:

(1) xg(x) > 0 for |x| ≥ 1;

R +∞ R0
(2) 0
g(x)dx = − −∞ g(x)dx = +∞;
(3) there exists constant N > 0 such that |Q(x, y, t)| ≤ N for |x| ≤ 1.
(4) [F (x) − Q(x, y, t)]sgn x ≥ 0 for |x| ≥ 1; besides for every G ∈ H(Q) there exists
θG > 0 such that F (x) − G(x, y, θG ) 6= 0 for |x| ≥ 1 and y ∈ R.
(5) when |x| ≥ 1 the inequality |F (x) − Q(x, y, t)| ≥ α(y) ≥ 0 is valid, where
α(y) on every interval is a measurable in the sense of Lebesgue function and
R +∞
−∞ α(y)dy > 0.

Then the system (5.64) is dissipative.

Proof. In the same way that in the work [160] we show that there exists η > 0
such that under certain τ every solution of the equation (5.66) gets in the rectangle
{(x, y) : |x| ≤ 1, |y| ≤ η} (τ depends on the system (5.66) and the according
solution). Now to complete the proof of the theorem it is sufficient to refer to
Theorem 5.8. 
Let us consider a more general equation

ẍ + h(x)ẋ + g(x) = R(x, ẋ, t).

 Method of Lyapunov functions 205

Rx
Assume y = ẋ + F (x) where F (x) = 0 h(τ )dτ . Then we obtain the following
system
(
ẋ = y − F (x)
(5.67)
ẏ = −g(x) + R(x, y, t)

Along with the system (5.67) consider its H-class

(
ẋ = y − F (x)
(5.68)
ẏ = −g(x) + G(x, y, t)

where G ∈ H(R).

Theorem 5.28 Let the functions h, g and R be continuous and R be bounded

and uniformly continuous w.r.t. t ∈ R uniformly w.r.t. (x, y) on compacts from R2 .
Suppose that in addition there exist constants L > N > 0 and K > 0 such that

(1) |R(x, y, t)| ≤ N for all x, y, t ∈ R;

(2) g(x)sgn x ≥ L for |x| ≥ 1;
(3) f (x) ≥ K for |x| ≥ 1.

Then the system (5.67) is dissipative.

Proof. As well as in the work [270] let us consider the following function V(x, y) =
Rx
y 2 − yF (x) + 12 F (x) + 2 0 g(τ )dτ . The derivative of this function taken in virtue
of the system (5.68) equals to

V̇(x, y, t) = −[y − F (x)]2 f (x) − g(x)F (x) + [2y − F (x)]G(x, y, t).

According to [270] there exists a > 0 such that for x2 + y 2 ≥ a2 V̇G (x, y, t) < 0 and
V(x, y) → +∞ for x2 + y 2 → +∞. According to Theorem 5.7 the system (5.67) is
dissipative. 

Consider the differential equation of the 3rd order

...
x + aẍ + bẋ + f (x) = p(x, ẋ, ẍ, t)

where a, b ∈ R. Assume y = ax + ẋ, z = bx + aẋ + ẍ. Then this equation is

equivalent to the system


 ẋ = y − ax
ẏ = z − bx (5.69)

 ż = −f (x) + p(x, y, z, t)
206 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Along with the system (5.69) consider its H-class



 ẋ = y − ax
ẏ = z − bx (5.70)

 ż = −f (x) + g(x, y, z, t)

where g ∈ H(p).

Theorem 5.29 Let the function f and p(x, y, z, t) be continuous and let the
following conditions be satisfied:
1. a, b > 0;
2. 0 < f (x)
x < ab for |x| ≥ 1;
3. lim |f (x)| = +∞;
|x|→+∞
4. lim |f (x) − abx| = +∞;
|x|→+∞
5. p is bounded and uniformly continuous w.r.t. t ∈ R uniformly w.r.t. (x, y, t) on
compacts from R3 ;
6. there exists A > 0 such that |p(x, y, z, t)| ≤ A for all x, y, z and t ∈ R.
Then the system (5.69) is dissipative.

Proof. To prove the formulated statement, following [270] we will consider the
function
Z x
1 2 2 1 2 b 2
V(x, y, z) = (a x − ay + z) + (z − bx) + y + a [f (τ ) − abτ ]dτ.
2 2 2 a

Under the conditions of Theorem 5.29 V(x, y, z) → +∞ for x2 + y 2 + z 2 → +∞

and the derivative of this function taken by virtue of the system (5.70) being to

V̇g (x, y, z, t) = −a(a2 x − ay + z)2

+[abx − f (x)][2(a2 x − ay + z) − bx][(a2 − b)x − ay + 2z]g(x, y, z, t).

In the same way that in [270] we can show that there exists a > 0 such that out
of the ball of radius a the function V decreases along the trajectories of (5.70) and,
according to Theorem 5.7, the system (5.69) is dissipative. 
2. Let us give two criterions of convergence of the equation (5.65) that is equiv-
alent to the system
(
ẋ = y
ẏ = −x − h(x)y + p(t),

or, respectively, to the system

(
ẋ = −F (x) + v
(5.72)
v̇ = −x + p(t)
 Method of Lyapunov functions 207

Rx
where v = ẋ + F (x) and F (x) := 0 h(t)dt. As usual, along with the system (5.72)
we consider its H-class
(
ẋ = −F (x) + v
(5.73)
v̇ = −x + g(t) (g ∈ H(p)).

Theorem 5.30 If h(x) > 0 (excepting maybe the discrete set of points), p(t) is
recurrent and F (+∞) = +∞ (or F (−∞) = −∞), then the equation (5.65) has a
unique bounded on R uniformly compatible and uniformly globally asymptotically
stable solution, i.e. (5.65) is convergent.

Proof. Reasoning in the same way that in [272] show that under the conditions of
the theorem there exists a closed contour that intersects all the solutions of every
system (5.73). According to Theorem 5.6 the system (5.72) is dissipative.
Let us show that the system (5.72) in fact is convergent. Let {xi (t), vi (t)}
(i = 1, 2) be two different solutions of the system (5.73). Construct the function

ϕ(t) = |x1 (t) − x2 (t)|2 + |v1 (t) − v2 (t)|2

and calculate its derivative. We will obtain:

ϕ̇(t) = −(x1 (t) − x2 (t))[F (x1 (t)) − F (x2 (t))]. (5.74)

Note that if ϕ̇(t) ≡ 0 then x1 (t) ≡ x2 (t) and, consequently, v1 (t) ≡ v2 (t). To finish
the proof of the theorem is sufficient to refer to Theorem 5.23. 

Theorem 5.31 If ψ1 (x) = h(x) − 2D (0 < D < 1) and ψ2 (x) = g(x) − x satisfy
the conditions:

(1) |ψ1 (x)| ≤ c1 for all x ∈ R;

(2) |ψ1 (x) − ψ2 (x)| ≤ c2 |x) − x| for all x̄, x ∈ R, where c1 + c2 = c < mD =
√
1 − DD,

then the system (5.72) has a unique bounded on R uniformly compatible, uniformly
globally asymptotically stable solution.

Proof. Let

{ϕ(t, xi , yi , g)} = {(ϕ1 (t, xi , yi , g), ϕ2 (t, xi , yi , g))

(i = 1, 2, g ∈ H(f )) are two solutions of the system (5.73) and

(
u(t) := ϕ1 (t, x1 , y1 , g) − ϕ1 (t, x2 , y2 , g)
v(t) := ϕ2 (t, x1 , y1 , g) − ϕ2 (t, x2 , y2 , g).
208 Global Attractors of Non-autonomous Dissipative Dynamical Systems

The functions u and v satisfy the following system of equations



 u̇ = v

 v̇ = −u − 2Dv − [ψ (ϕ (t, x , y , g)) − ψ (ϕ (t, x , y , g))]−
2 1 1 1 2 1 2 2

 [ψ 1 (ϕ 1 (t, x ,
1 1y , g))ϕ 2 (t, x ,
1 1y , g)−

 ψ1 (ϕ1 (t, x2 , y2 , g))ϕ2 (t, x2 , y2 , g)].

In the same way that in [272] we show that


 c c
 |u(t)| < (1 + m1 )α[exp(−(D − m )t) + exp(−Dt)]
c1 c
|v(t)| < (1 + c1 )(1 + m )α[exp(−(D − m )t) , (5.75)

 + exp(−Dt)]
where
1p 1√ p
α= |u(0)|2 + 2D|u(0)||v(0)| + |v(0)|2 ≤ 1 + D |u(0)|2 + |v(0)|2 .
m m
From (5.75) follows that
p c p
|u(t)|2 + |v(t)|2 ≤ N exp(−(D − )t) |u(0)|2 + |v(0)|2 ,
m
where N > 0 is some constant depending only on c1 , c2 and D. Now to complete
the proof of the theorem is sufficient to refer to Theorem 2.15 and Corollary 2.6.

Remark 5.9 All the statements given in this chapter in the periodical case coin-
cide with previously established results (see [160], [270], [272]).

5.6 Construction of Lyapunov function for homogeneous systems

Let (X, h, Y ) be a locally trivial Banach fibering, | · | : X → R+ is a norm on

(X, h, Y ) coordinated with the metric ρ on X.

Definition 5.9 The autonomous dynamical system (X, R+ , π) is said to be ho-

mogeneous of order m ∈ R+ , if for any x ∈ X, t ∈ R+ and λ > 0 the equality
π(t, λx) = λπ(λm−1 t, x) takes place.

Definition 5.10 h(X, T, π), ((Y, T, σ), hi is called homogeneous of order m = 1 if

the autonomous dynamical system (X, T, π) is homogeneous of order m = 1.

Theorem 5.32 For an autonomous homogeneous (of order m > 1 ) dynamical

system (X, R+ , π) the following assertions are equivalent:
1. there exist positive numbers ai and bi (i = 1, 2) such that
1 1
(a1 |x|1−m + b1 t) 1−m ≤ |π(t, x)| ≤ (a2 |x|1−m + b2 t) 1−m (5.76)

for all t ≥ 0 and x ∈ X;

 Method of Lyapunov functions 209

2. for all k > m − 1 there exists a continuous function V : X → R+ with the

following properties:
2.1. V (λx) = λk−m+1 V (x) for all λ ≥ 0 and x ∈ X;
2.2. α|x|k−m+1 ≤ V (x) ≤ β|x|k−m+1 for all x ∈ X, where α and β are certain
positive numbers;
d
2.3. Vπ0 (x) = −|x|k for all x ∈ X, where Vπ0 (x) = dt V (π(t, x)) |t=0 for T = R+
0
and Vπ (x) = V (π(1, x)) − V (x) for T = Z+ .

Proof. We will show that from 1. follows 2. Let a and b are positive numbers,
such that the inequality (5.76) takes place, then for each k > m − 1 we define the
function V : X → R+ by equality
Z +∞
V (x) = |π(t, x)|k dt. (5.77)
0

First of all we note that by equality (5.77) it is defined correctly the function
V : X → R+ because the integral, which figures in the second number of equation
(5.77) is convergent, moreover it is uniformly convergent w.r.t. x on every bounded
set from X. Indeed, since
k k
(a1 |x|1−m + b1 t) 1−m ≤ |π(t, x)|k ≤ (a2 |x|1−m + b2 t) 1−m , (5.78)

Z +∞ Z +∞
1−m k 1 k
|(a|x| + bt) 1−m dt = τ 1−m dτ (5.79)
0 b a|x|1−m

k
and 1−m < −1, then the integral (5.79) is convergent; moreover, the convergence
is uniform on every bounded set from X.
We will show that the function V , defined by equality (5.77), is our unknown
function. The continuity of V results from the continuity of mapping π : T×X → X
and uniform convergence of integral (5.79) w.r.t. x on every bounded set from X.
We now note that
Z +∞ Z +∞
V (λx) = |π(t, λx)|k dt = λk |π(λm−1 t, x)|k dt
0 0

Z +∞
k−m+1
=λ |π(τ, x)|k dt = λk−m+1 V (x)
0

for all λ > 0 and x ∈ X. It is not difficult to show that the function V is positive
definite. In virtue of (5.78) and (5.79) we have

α|x|k−m+1 ≤ V (x) ≤ β|x|k−m+1 (5.80)

210 Global Attractors of Non-autonomous Dissipative Dynamical Systems

for every x ∈ X, where

k−m+1 k−m+1
(m − 1)a1 1−m (m − 1)a2 1−m
α= and β = .
b1 (k − m + 1) b2 (k − m + 1)
From (5.79)-(5.80) results that the function V satisfies and the condition 2.2.
Finally, we note that
d
V (π(t, x)) = −|π(t, x)|k (5.81)
dt
and, consequently, Vπ0 (x) = −|x|k for all x ∈ X.
We will prove now that from condition 2. follows 1. In fact, we denote by
ψ(t) = V (π(t, x)), then in virtue of condition 2.3 we will have

ψ 0 (t) = −|π(t, x)|k (5.82)

for all t ≥ 0.
1
From the condition 2.2 we have |π(t, x)|k−m+1 ≥ β ψ(t) and, consequently,

1 k
ψ 0 (t) ≤ − k ψ(t) k−m+1
β k−m+1

for all t ≥ 0. If x 6= 0, then ψ(t) = V (π(t, x)) > 0 for all t ≥ 0, therefore
m−1 m−1 1 1
V (π(t, x)) ≤ (V − k−m+1 (x) + k t) 1−m (5.83)
k − m + 1 β k−m+1

for all x ∈ X and t ≥ 0.

From the condition 2.2 and the inequality (5.83) results that |π(t, x)| ≤
1
(a2 |x|1−m +b2 t) 1−m for all x ∈ X and t ≥ 0, where
m−1 m−1 k m−1
a2 = (αβ) k−m+1 and b2 = (α) k−m+1 (β) k−m+1 .
k−m+1
Analogously can be proved the second inequality. The theorem is completely
proved. 

Corollary 5.10 Let (X, R+ , π) be an autonomous homogeneous (of order m >

1 ) dynamical system (X, R+ , π) and the trivial motion of the dynamical system
(X, R+ , π) is uniform asymptotic stable, then for every number k > m − 1 there
exists a continuous function V : X → R+ which possesses properties 2.1-2.3 from
the theorem 5.32.

Proof. This assertion directly follows from Theorems 5.32 and 2.35. 

Theorem 5.33 Let a non-autonomous system h(X, T, π), ((Y, T, σ), hi be homo-
geneous of order m = 1. Then the conditions 1. and 2. are equivalent:
 Method of Lyapunov functions 211

1. there exist positive numbers Ni and νi (i = 1, 2) such that N1 e−ν1 t |x|≤

|π(t, x)| ≤N2 e−ν2 t |x| for all x ∈ X and t ≥ 0;
2. for each k > 0 there exists a continuous function V : X → R+ satisfying the
following conditions :
2.1. V (λx) = λk V (x) for all x ∈ X and λ > 0;
2.2. there exist positive numbers α ≥ 1 and β so that

α|x|k ≤ V (x) ≤ β|x|k (5.84)

for every x ∈ X;
2.3. Vπ0 (x) = −|x|k for all x ∈ X.

Proof. Let us show that under the conditions of theorem from assumption 1.
follows 2. First of all let us show that the function V , defined by equality (5.77) is
the unknown, in the case when T = R+ . We note that by equality (5.77) is correctly
defined the function V : X → R+ , such that

N1k e−ν1 kt |x|k ≤ |π(t, x)|k ≤ N2k e−ν2 kt |x|k (5.85)

for all t ≥ 0, x ∈ X and, consequently, the integral in second member of equality

(5.77) is convergent uniformly w.r.t. x on every bounded set from X. In particularly,
the function V is continuous w.r.t. x ∈ X.
From (5.77) and (5.85) follows that

α|x|k ≤ V (x) ≤ β|x|k (5.86)

for every x ∈ X, where α = N1k (ν1 k)−1 and β = N2k (ν2 k)−1 . From equality (5.79)
we have that V (λx) = λk V (x) for every x ∈ X and λ > 0, and from (5.81) follows
the equality Vπ0 (x) = −|x|k .
Let us show that the inverse implication holds too . We now suppose that
the conditions 2.1-2.3 of Theorem hold. One denote by ψ(t) = V (π(t, x)), then the
equality (5.82) holds for all t ≥ 0. From condition 2.2 follows that |π(t, x)|k ≥ β1 ψ(t)
and, consequently,
1
ψ 0 (t) ≤ − ψ(t) (5.87)
β
for all t ≥ 0. From (5.87) we have
1
V (π(t, x)) ≤ V (x)e− β t (5.88)

for any t ≥ 0 and x ∈ X. According to (5.86) and (5.88) we have

|π(t, x)| ≤ N2 e−ν2 t |x|

β k 1 1
for any t ≥ and x ∈ X, where N2 = ( α ) and ν2 = βk . Analogously can be
established the second inequality.
212 Global Attractors of Non-autonomous Dissipative Dynamical Systems

If T = Z+ , then we will define the function V : X → R+ by equality

+∞
X
V (x) = |π(n, x)|k . (5.89)
n=1

The series in second member of equality (5.89) is convergent uniformly w.r.t. x on

every bounded subset from X, because under the conditions of Theorem we have

|π(n, x)|k ≤ N k e−νkn |x|k (5.90)

for all n ∈ Z+ and x ∈ X and, consequently, the series (5.89) is majored by

geometric series with the denominator q = e−νk < 1. Thus the function V defined
by equality (5.89) is continuous.
From the homogeneity of (X, T, π) of order m = 1 and equality (5.89) follows
that V (λx) = λk V (x) for all λ > 0 and x ∈ X. According to (5.89) and (5.90) we
have

α|x|k ≤ V (x) ≤ β|x|k (5.91)

for any x ∈ X, where α = 1 and β = N k (1 − e−νk )−1 . Finally, from equality (5.89)
follows that Vπ0 (x) = V (π(1, x)) − V (x) = −|x|k for all x ∈ X.
We now will show that from the conditions 2.1–2.3 of Theorem (in the case, when
T = Z+ ) follows the estimation (5.84). In fact, we denote by ψ(n) = V (π(n, x)),
then from the conditions 2.2–2.3 we have
1
∆ψ(n) = ψ(n + 1) − ψ(n) = −|π(n, x)|k ≤ − ψ(n). (5.92)
β
We may assume without loss of generality that β > 1, then from (5.92) we have

V (π(n, x)) ≤ γ n V (x) (5.93)

for all x ∈ X and n ∈ Z+ , where γ = 1 − β −1 . From (5.91) and (5.93) follows

1
the inequality (5.84), if we suppose that N = ( βα ) k and ν = − k1 ln(1 − β −1 ). The
theorem is proved. 

Definition 5.11 The trivial section of fiber bundle (X, h, Y ) is called globally
uniformly asymptotically stable, if

(1) for arbitrary ε > 0 there is δ(ε) > 0 such that |π(t, x)| < ε for all t ≥ 0 and
|x| < δ;
(2) lim |π(t, x)| = 0 for all x ∈ X, moreover this equality holds uniformly in x on
t→+∞
the bounded subsets from X.

Corollary 5.11 Let h(X, T, π), ((Y, T, σ), hi be homogeneous of order m = 1.

Then the following conditions are equivalent:
 Method of Lyapunov functions 213

(1) the zero section of fibering (X, h, Y ) is uniformly asymptotically stable, i.e.
(a) for all ε > 0 there exists δ(ε) > 0 such that |π(t, x)| < ε for all t ≥ 0 and
|x| < δ;
(b) lim |π(t, x)| = 0 for all x ∈ X, moreover this equality holds uniformly
t→+∞
w.r.t. x on every bounded set from X;
(2) there exist positive numbers N and ν such that |π(t, x)| ≤ N e−νt |x| for all t ≥ 0
and x ∈ X;
(3) for every k > 0 there exists a continuous function V : X → R+ satisfying the
conditions 2.1-2.3 of Theorem 5.33.

Proof. This assertion follows from the Theorems 5.33 and 2.34. 

Remark 5.10 In the case, when the space X is finite dimensional Theorem 5.32
and Corollary 5.10 (for autonomous system with T = R+ ) generalize and refine
some results of V.I.Zubov (see, for example [337], Theorems 36 and 37).

5.7 Differentiable homogeneous systems

p
Let H be a Hilbert space with scalar product h·, ·i and the norm |·| = h·, ·i. We de-
note by C(E, B)(C 1 (E, B)) the space of all continuous ( differentiable continuous )
functions defined on the space E wit values in some Banach space B.
The function f : E × Ω → F is called homogeneous of order k w.r.t. variable
u ∈ E if the equality f (λx, ω) = λk f (x, ω) holds for all λ > 0, x ∈ E and ω ∈ Ω.

Lemma 5.6 The following assertions hold.

(1) Let Ω be a compact, f ∈ C(E × Ω, F ), f (0, ω) = 0 for all ω ∈ Ω, f (λu, ω) =

λm f (u, ω) (for every λ > 0and (u, ω) ∈ E × Ω)). Then there exists M > 0 such
that |f (u, ω)| ≤ M |u|m for all (u, ω) ∈ E × Ω.
(2) If the function f ∈ C 1 (E × Ω, F ) is homogeneous (of order m ≥ 1), then
the function Du f (·, ω) : E → L(E, F ) (ω ∈ Ω) will be homogeneous (of order
m−1), where L(E, F ) is the space of all linear continuous operators A : E → F ;
(3) The function f ∈ C 1 (E, F ) is homogeneous (of order m) if and only if f satisfies
the equation Df (x)x = mf (x) for all x ∈ E, where Df (x) is the derivative of
Frechet of function f ∈ C 1 (E) in the point x;

Proof. First of all we note that in conditions of Lemma 5.6 there exists δ0 > 0
such that |f (u, ω)| ≤ 1 for all |u| ≤ δ and ω ∈ Ω. If we suppose that it is not so,
then there exist δn → 0, |un | < δn and ωn ∈ Ω such that |f (un , ωn )| > 1. Since Ω is
compact we may assume that the sequence {ωn } is convergent. Let ωn → ω0 , then
214 Global Attractors of Non-autonomous Dissipative Dynamical Systems

according to continuity of f we have |f (0, ω0 )| ≥ 1. On the other hand f (0, ω0 ) = 0.

The obtained contradiction proves the required assertion.
Thus there exists δ > 0 such that |f (u, ω)| ≤ 1 for all |u| ≤ δ and ω ∈ Ω and,
consequently, we have

|u| uδ |u|m δu 1
|f (u, ω)| ≤ |f ( , ω)| = m |f ( , ω)| ≤ m |u|m .
δ |u| δ |u| δ

Let f ∈ C 1 (E, F ) and f be homogeneous (of order m), then

f (x + h) − f (x) = Df (x)h + r(x, h) (5.94)

and |r(x, h)| → 0 as |h| → 0. We now will replace x by λx in equality (5.94), then
we obtain

f (λx + h) − f (λx) = Df (λx) + r(λx, h)

and, consequently,

f (x + u) − f (x) = λ−m+1 Df (λx)h + λ−m r(λx, λu) (5.95)

for all λ > 0, where u = λ−1 h. From (5.95) follows, that λ−m+1 Df (λx) = Df (x),
i.e. Df (λx) = λm−1 Df (x) for any λ > 0.
Finally, let us prove the third assertion of Lemma. Let f ∈ C 1 (E, F ) be a
homogeneous function (of order m). Having differentiated the identity f (λx) =
λm f (x) w.r.t. λ and taking λ = 1 one obtain the identity Df (x)x = mx. Let now
the identity Df (x)x = mx take place. We denote by ϕ(λ) = λ−m f (λx)(λ > 0).
Thus Df (λx)λx = mf (λx), then

ϕ0 (λ) = −mλ−m+1 f (λx) + λ−m Df (λx)x =

−mλ−m−1 f (λx) + λ−m−1 mf (λx) = 0

for all λ > 0 and, consequently, ϕ(λ) = const for all λ > 0. Thus ϕ(1) = f (x), when
ϕ(λ) = f (x) for any λ > 0, i.e. f (λx) = λm f (x) for every λ > 0 and for all x ∈ E.
The lemma is proved. 

The function f ∈ C(E) is said to be regular, if for the differential equation

x0 = f (x) (5.96)

the conditions of existence, uniqueness on R+ and continuous dependence on initial

data are fulfilled, i.e. by equation (5.96) it is generated a semigroup dynamical
system (E, R+ , π), where π(t, x) is the solution of equation (5.96) with initial con-
dition π(0, x) = x.
 Method of Lyapunov functions 215

Theorem 5.34 Let f ∈ C 0 (E, F ) be regular and homogeneous (of order m > 1),
then the following conditions are equivalent:
1. the zero solution of equation (5.96) is uniform asymptotic stable;
2. for all sufficiently large k there exists a continuously differentiable function
V : E → R+ satisfying the following conditions:
2.1. V (λx) = λk−m+1 V (x) for all λ ≥ 0 and x ∈ E;
2.2. there exist positive numbers α and β such that α|x|k−m+1 ≤ V (x) ≤
β|x|k−m+1 for all x ∈ E;
2.3. Vπ0 (x) = DV (x)f (x) = −|x|k for any x ∈ E, where DV (x) is a derivative
of Frechet of function V in the point x.

Proof. We suppose that the zero solution of equation (5.96) is uniformly asymp-
totically stable. Denote by (E, R+ , π) a semigroup dynamical system, generated
by equation (5.96). Thus, f ∈ C 1 (E, F ) is homogeneous (of order m > 1), then
according to Theorem 3.4 [337] the equality π(t, λx) = λπ(λm−1 t, x) holds for all
x ∈ E, λ ≥ 0 and t ∈ R+ , i.e. a dynamical system (E, R+ , π) is homogeneous ( of
order m > 1 ). According to Theorem 5.32 and Corollary 5.10 by equality (5.95) it
is defined a continuous function V : E → R+ , satisfying the condition 2.1–2.3. of
Theorem 5.34. Let us show that under conditions of Theorem 5.34 function V will
be continuously differentiable and that the equality Vπ0 (x) = DV (x)x holds. To this
aim we will formally differentiate the equality (5.77) w.r.t. x, then we will have
Z +∞
DV (x)u = k|π(t, x)|k−2 RehDx π(t, x)u, π(t, x)idt. (5.97)
0

Now we will show that the integral figuring in the second hand of formula (5.97),
is uniformly convergent w.r.t. x and u on the every ball from E and, consequently,
the equality (5.97) really defines a derivative of function V . We note that operator-
function U (t, x) = Dx π(t, x) satisfies the operational equation

X 0 = A(t, x)X,

where A(t, x) = Dx f (π(t, x)), and initial condition U (0, x) = I (I−is identity
operator in E). According to Lemma 5.6 the function A(t, x) = Dx f (π(t, x)) is
homogeneous w.r.t. π(t, x) of order m − 1 and there exists a number M > 0 such
that

kA(t, x)k ≤ M |π(t, x)|m−1 (5.98)

for all x ∈ E and t ≥ 0. Since the zero solution of equation (5.96) is uniformly
asymptotically stable, in virtue of Corollary 5.10 there exist positive numbers a and
b such that
1
|π(t, x)| ≤ (a|x|1−m + bt) 1−m (5.99)
216 Global Attractors of Non-autonomous Dissipative Dynamical Systems

for any x ∈ E and t ≥ 0 and, consequently, from (5.98)– (5.99) we have

kA(t, x)k ≤ M (a|x|1−m + bt)−1 . (5.100)

Since
Rt
kA(τ,x)k,dτ
kU (t, x)k ≤ e 0 , (5.101)

from (5.100) and (5.101) we have

M M
kU (t, x)k ≤ (a|x|1−m + bt) b (a−1 |x|m−1 ) b (5.102)

for every t ≥ 0 and x ∈ E. By virtue of Inequality (5.102) we obtain

|π(t, x)|k−2 |RehDx π(t, x)u, π(t, x)i| ≤ |π(t, x)|k−1 kDx π(t, x)k|u| (5.103)

k−1 M M m
≤ (a|x|1−m + bt)− m−1 + b a− b |x|(m−1) b |u|

for all t ≥ 0 and x, u ∈ E. From estimation (5.103) follows that the integral in
second member of equality (5.97) is uniformly convergent w.r.t. x and u on the each
ball from E for sufficiently large k and, consequently, the function V , defined by
equality (5.77), under the conditions of Theorem 5.34 is continuously differentiable.
We now note, that
d  
 
Vπ0 (x) = V (π(t, x)) = DV (π(t, x))f (π(t, x)) = DV (x)f (x)
dt t=0 t=0

for all x ∈ E. On the other hand, according to Theorem 5.32 Vπ0 (x) = −|x|k . For
finishing the proof of the theorem it is sufficient to notice that in conformity with
Theorem 5.32 the inverse statement holds, i.e. conditions 2.1–2.3 of Theorem 5.34
imply 1. The theorem is completely proved. 

Now let f ∈ C 1 (E × F, E) and Φ ∈ C 1 (F, F ). Consider the differential system

( 0
u = f (u, ω)
(ω ∈ Ω), (5.104)
ω 0 = Φ(ω)
where Ω is a certain differentiable compact sub-manifold in F, f and Φ are regular
functions, i.e. for the differential system (5.104) and equation

ω 0 = Φ(ω) (5.105)

are fulfilled the conditions of existence, uniqueness and continuous dependence on

initial data on the R+ for (5.104) and on the R for (5.105). We denote by (Ω, R, σ)
a group dynamical system, generated by equation (5.105), then the system (5.104)
may be written in the form of non-autonomous equation

u0 = f (u, ωt), (ω ∈ Ω) (5.106)

 Method of Lyapunov functions 217

where ωt := σ(t, ω). Let ϕ(t, u, ω) be a solution of equation (5.106) satisfying

the initial condition ϕ(0, u, ω) = u, then it is not difficult to see that the triple
hE, ϕ, (Ω, R, σ)i is a cocycle over (Ω, R, σ) with the fiber E. It is easy to check that
if a function f is homogeneous of order m = 1 w.r.t. u ∈ E, then the equality
ϕ(t, λu, ω) = λϕ(t, u, ω), holds for all t ≥ 0, λ > 0, u ∈ E and ω ∈ Ω.

Theorem 5.35 Let the function f ∈ C 1 (E × F, F ) and Φ ∈ C 1 (F, F ) be regular,

Ω is compact and the function f is homogeneous (of order m = 1) w.r.t. variable
u ∈ E, then the following conditions are equivalent:

1. the trivial solution of equation (5.106) is uniform exponential stable, i.e. there
exist positive numbers N and ν such that

|ϕ(t, u, ω)| ≤ N e−νt |u| (5.107)

for all u ∈ E, t ≥ 0 and ω ∈ Ω;

2. for all number k > 1 there exists a continuously differentiable function V :
E × Ω → R+ satisfying the following conditions:
2.1. V (λu, ω) = λk V (u, ω) for all u ∈ Ω and λ > 0;
2.2. there exist positive numbers α and β such that α|u|k ≤ V (u, ω) ≤ β|u|k
for all u ∈ E and ω ∈ Ω;
2.3.
d 

Vπ0 (u, ω) := V (ϕ(t, u, ω), ωt) = Du V (u, ω)f (u, ω)+
dt t=0

Dω V (u, ω)Φ(ω) = −|u|k

for all u ∈ E and ω ∈ Ω, where Du V (respectivelyDω V ) is a partial

Frechet’s derivative of function V w.r.t. variable u ∈ E (respectively ω ∈
Ω).

Proof. Let h(X, R+ , π), (Ω, R, σ), hi be a non-autonomous dynamical system gen-
erated by differential system (5.104), where X = E × Ω, h = pr2 and π = (ϕ, σ).
The triple (X, h, Ω) is the trivial fiber bundle with the fiber E, moreover the norm
of element x = (u, ω) ∈ E × Ω is defined by equality |x| = |u|, where |u| is a norm of
element u in the space E. Then under the conditions of Theorem the zero section of
non-autonomous dynamical system, constructed above, will satisfy the conditions
of Theorem 5.35. Thus to finish the proof of Theorem 5.35 it is sufficient to show
that the function V : E × Ω → R+ from the theorem 5.33 under the conditions of
Theorem 5.35 will be continuously differentiable and satisfies the condition

Vπ0 (u, ω) = Du V (u, ω)f (u) + Dω V (u, ω)Φ(ω)

for all (u, ω) ∈ E × Ω.

218 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Let us show that a function V : E × Ω → R+ , defined by equality (5.77), i.e.

Z +∞
V (u, ω) = |ϕ(t, u, ω)|k dt (5.108)
0

for all (u, ω) ∈ E × Ω (such that |π(t, x)| = |(ϕ(t, u, ω), ωt)| = |ϕ(t, u, ω)|) is contin-
uously differentiable. To this aim we will formally differentiate the equality (5.108)
w.r.t. variable u ∈ E, thus we obtain

Du V (u, ω)v = (5.109)

Z +∞
k|ϕ(t, u, ω)|k−2 RehDu ϕ(t, u, ω)v, ϕ(t, u, ω)idt.
0

Let us show that integral in (5.109) is uniformly convergent w.r.t. ω ∈ Ω and

u, v on every bounded set from E and since by formula (5.109) is really defined a
derivative of Frechet of function V w.r.t. variable u ∈ E. First of all we note that
operator-function U (t, u, ω) := Du ϕ(t, u, ω) satisfies the operational equation

X 0 = B(t, u, ω)X, (5.110)

where B(t, u, ω) := Du f (ϕ(t, u, ω), ωt), and initial condition U (0, u, ω) = I (I

is the identity operator in E). According to Lemma 3.1 function B(t, u, ω) =
Du f (ϕ(t, u, ω), ωt) is homogeneous w.r.t. ϕ(t, u, ω) of order m = 1 and there exists
a number M > 0 such that

kB(t, u, ω)k ≤ M (5.111)

for all (u, ω) ∈ E × Ω and t ≥ 0. From the inequality (5.111) follows that
Rt
kB(τ,u,ω)kdτ
kU (t, u, ω)k ≤ e 0 ≤ eM t

for all t ≥ 0 and (u, ω) ∈ E × Ω and, consequently,

|ϕ(t, u, ω)|k−2 |RehDu ϕ(t, u, ω)v, ϕ(t, u, ω)i| (5.112)

k−1 k−1 −ν(k−1)t
|ϕ(t, u, ω)| kDu ϕ(t, u, ω)k|v| ≤ M |v|N e |u|

for any t ≥ 0, u, v ∈ E and ω ∈ Ω. From the inequality (5.112) follows that for k > 1
the integral in the second hand of (5.108) is convergent, moreover uniformly w.r.t.
ω ∈ Ω and u, v on the every bounded subset from E. Thus, the function V (u, ω),
defined by formula (5.108) is continuously differentiable w.r.t. variable u ∈ E and
by equality (5.109) it is defined its derivative of Fréchet w.r.t. variable u ∈ E.
Having formally differentiated the equality (5.108) w.r.t. variable ω ∈ F we will
obtain

Dω V (u, ω)w = (5.113)

Z +∞
k|ϕ(t, u, ω)|k−2 RehDω ϕ(t, u, ω)w, ϕ(t, u, ω)idt
0
 Method of Lyapunov functions 219

for all (u, ω) ∈ E × Ω. Let us show that integral in the second member of (5.113)
is uniformly convergent w.r.t. ω ∈ Ω and (u, w) ∈ E × F on every bounded subset
from E × F. First of all we note that from (5.107) we have
Z t
ϕ(t, u, ω) = u + f (ϕ(τ, u, ω), ωτ )dτ
0

for all t ≥ 0 and (u, ω) ∈ E × F and, consequently,

Z t
Dω ϕ(t, u, ω) = Du f (ϕ(τ, u, ω), ωτ )Dω ϕ(τ, u, ω)
0
+Dω f (ϕ(τ, u, ω), ωτ )DΦ(ωτ )dτ. (5.114)

We denote by V(t, u, ω) the operator-function Dω ϕ(t, u, ω), then from (5.114) follows
that V(t, u, ω) satisfies the operational equation

Y 0 = B(t, u, ω)Y + F(t, u, ω), (5.115)

where F(t, u, ω) = Dω f (ϕ(t, u, ω), ωt)DΦ(ωt), and initial condition

Y (0) = O (5.116)

(O is a null operator, acting from F into F ). From the equality (5.115) and the
condition (5.116) follows that
Z t
V(t, u, ω) = U (t, τ, u, ω)F(τ, u, ω)dτ (5.117)
0

for all t ≥ 0 and (u, ω) ∈ E × Ω, where U (t, τ, u, ω) is a solution of operational

equation (5.110), satisfying the initial condition U (τ, τ, u, ω) = U (τ, u, ω), therefore
the estimate
Rt
kB(τ,u,ω)kdτ
kU (t, τ, u, ω)k ≤ e τ ≤ eM (t−τ ) (5.118)

holds for any t ≥ τ (t, τ ∈ R+ ) and (u, ω) ∈ E × Ω.

Thus, a function f (u, ω) is homogeneous w.r.t. u ( of order m = 1), then
Dω f (u, ω) will be the same and, in virtue of Lemma 5.6 there exists a number
M > 0 such that

kDω f (u, ω)k ≤ M1 |u|

for all u ∈ E and ω ∈ Ω. Thus,

kF(t, u, ω)k = kDω f (ϕ(t, u, ω), ωt)DΦ(ωt)| ≤ L|ϕ(t, u, ω)| (5.119)

220 Global Attractors of Non-autonomous Dissipative Dynamical Systems

for any t ≥ 0 and (u, ω) ∈ E × Ω, where L = M1 max{kDΦ(ω)k : ω ∈ Ω}. From

(5.106),(5.117), (5.118) and (5.119) we have

|ϕ(t, u, ω)|k−2 |RehDω ϕ(t, u, ω)wi| ≤ |ϕ(t, u, ω)|k−1 kDω ϕ(t, u, ω)k|w|
Z t
≤ |ϕ(t, u, ω)|k−1 |w| kU (t, τ, u, ω)kkF(τ, u, ω)kdτ
0
Z t
≤ |ϕ(t, u, ω)|k−1 |w| eM (t−τ ) L|ϕ(τ, u, ω)|dτ ≤ N k−1 e−ν(k−1)t |u|k−1 |w|
0
Z t
−ντ (eM t − e−νt ) k
e M (t−τ )
LN e |u|dτ = N k Le−ν(k−1)t |u| |w|
0 M +ν
for every t ≥ 0, u ∈ E and ω ∈ Ω. From this follows, that for sufficiently large
k the integral (5.113) is uniformly convergent w.r.t. ω ∈ Ω and (u, w) ∈ E × F
on the every bounded subset from E × F . Therefore the function V (u, ω) defined
by formula (5.108) is continuously differentiable w.r.t. variable ω ∈ Ω and by
formula (5.113) it is defined its Frechet derivative w.r.t. variable ω ∈ Ω. So, the
function V : E × Ω → R+ is continuously differentiable w.r.t. variable u ∈ E and
variable ω ∈ Ω both and, consequently, it is continuously differentiable in the sense
of Frechet.
Let us note that
d
V ((ϕ(t, u, ω), ωt) = Du V (ϕ(t, u, ω), ωt)Du ϕ(t, u, ω) +
dt
d
Dω V (ϕ(t, u, ω), ωt) ωt = Du V (ϕ(t, u, ω), ωt)f (ϕ(t, u, ω), ωt) +
dt
Dω V (ϕ(t, u, ω), ωt)Φ(ωt)

for any t ≥ and (u, ω) ∈ E × Ω and, consequently,

d 

Vπ0 (u, ω) = V (ϕ(t, u, ω), ωt) = Du V (u, ω)f (u, ω) + Dω V (u, ω)Φ(u).
dt t=0

Thus, under the conditions of Theorem from 1. follows 2.

For finishing the proof of the theorem it is sufficient to remark that according
to Theorem 5.33 from the conditions 2.1–2.3 of Theorem follows 1. The theorem is
completely proved. 

Remark 5.11 For finite dimensional systems the theorem 5.34 generalizes and
refines the theorem 38 from [337].

5.8 Global attractors of quasi-homogeneous systems

Let E and F be finite-dimensional Banach spaces. Consider the differential equation

x0 = f (x), (5.120)
 Method of Lyapunov functions 221

where f ∈ C(E, E) is regular and homogeneous of order m ≥ 1. Along with

equation (5.120) consider also perturbed equation

x0 = f (x) + F (t, x) (5.121)

with the regular perturbation. Let us remember (see, for example [337], [224] and
[227]), that equation (5.121) is called quasi-homogeneous if

|F (t, x)|
lim sup = c, (5.122)
|x|→+∞ |x|m

where c is a certain nonnegative sufficiently small constant. Moreover, the limiting

relation (5.122) holds uniformly w.r.t. t ∈ R.
Along with equation (5.121) consider the family of equations

x0 = f (x) + G(t, x), (5.123)

where G ∈ H(f ) := {Fτ : τ ∈ R} and the bar denotes closure in C(R×E, E), C(R×
E, E) is equipped by topology of convergence on every compact from R × E. Let us
denote by (Y, R, σ) a dynamical system of translations on Y := H(F ), by ϕ(t, u, f +
G) the solution of equation (5.123), satisfying the condition ϕ(0, u, f + G) = u
and h(X, R+ , π), (Y, R, σ), hi (i.e. h := pr2 and π := (ϕ, σ)) a non-autonomous
dynamical system generated by equation (5.123).

Theorem 5.36 Let f ∈ C 1 (E, E) be regular, homogeneous of order m ≥ 1 and

the zero solution of equation (5.120) be uniformly asymptotically stable. If f + F ∈
C(R × E, E) is regular, equation (5.121) is quasi-homogeneous, then the following
assertions take place:

(1) a set IG = {u ∈ E| sup{|ϕ(t, u, f + G)| : t ∈ R} < +∞} is not empty, compact

and connected for each G ∈ H(F );
(2) ϕ(t, IG , f + G) = Iσ(t,f +G) for all t ∈ R+ and G ∈ H(F );
S
(3) a set I = {IG |G ∈ H(F )} is compact and connected;
(4) the equalities

lim β(ϕ(t, M, f + G−t ), IG ) = 0

t→+∞

and

lim β(ϕ(t, M, f + G), I) = 0

t→+∞

take place for all G ∈ H(F ) and bounded subset M ⊆ E.

Proof. Let h(X, R+ , π), (Y, R, σ), hi be a non-autonomous dynamical system, gen-
erated by equation (5.121). Thus, f ∈ C 1 (E, E) and the zero solution of equation
222 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(5.120) is uniformly asymptotically stable, then according to Theorem 5.34 by equal-

ity (5.77) it is defined a continuously differentiable function V : E → R+ , satisfying
the conditions 2.1–2.3 of Theorem 5.34. Let us define a function V : X → R+ in
the following way: V(x) = V (u) for all x = (u, G) ∈ E × H(F ). We note that

d 

Vπ0 (x) = V (ϕ(t, u, f + G)) (5.124)
dt t=0
d 

= DV (ϕ(t, u, f + G)) ϕ(t, u, f + G) = DV (u)(f (u) + G(0, u))
dt t=0
= DV (u)f (u) + DV (u)G(0, u) = −|u|k + DV (u)G(0, u).

In virtue of condition 2.1 of Theorem 5.34 the function V is homogeneous of order

k − m + 1, then according to Lemma 5.6 its Frechet derivative DV (u) will be
homogeneous of order k − m and, consequently, there exists a number L > 0 such
that

kDV (u)k ≤ L|u|k−m (5.125)

for all u ∈ E. From equality (5.122) follows that for all ε > 0 there exists r = r(ε) >
0 such that

|G(t, u)| ≤ (c + ε)|u|m (5.126)

for all |u| ≥ r, t ∈ R and G ∈ H(F ). From (5.125)–(5.126) we have

|DV (u)G(0, u)| ≤ L(c + ε)|u|k (5.127)

and, consequently, from (5.124),(5.127) we will obtain

Vπ0 (x) ≤ |u|k (−1 + L(c + ε)) = −γ|x|k

for each x ∈ Xr , where γ = 1 − l(c + ε) > 0 since c and ε are a sufficiently small
positive numbers. To finish the proof of Theorem it is sufficient to refer to Theorems
2.24- 2.25. 

Theorem 5.37 Let f ∈ C 1 (E, E), Φ ∈ C 1 (F, F ) and f + F ∈ C 1 (E × F, E) be

regular, Ω ⊆ F be a compact invariant set of dynamical system (5.105), the function
f be homogeneous (of order m > 1) and the zero solution of equation (5.105) be
uniformly asymptotically stable. If

|F (u, ω)| ≤ c|u|m

for all |u| ≥ r and ω ∈ Ω, where rand c are certain positive numbers, and moreover
c is sufficiently small, then the following assertions take place:
 Method of Lyapunov functions 223

(1) a set Iω = {u ∈ E| sup{|ϕ(t, u, ω)| : t ∈ R} < +∞} is not empty, compact and
connected for each ω ∈ Ω, where ϕ(t, u, ω) is a solution of equation

u0 = f (u) + F (u, ωt)

satisfying the initial condition ϕ(0, u, ω) = u;

(2) ϕ(t, Iω , ω) = Iσ(t,ω) for all t ∈ R+ and ω ∈ Ω;
S
(3) a set I = {Iω | ω ∈ Ω} is compact and connected;
(4) the equalities

lim β(ϕ(t, M, ω−t ), Iω ) = 0 (5.128)

t→+∞

and

lim β(ϕ(t, M, ω), I) = 0 (5.129)

t→+∞

take place for all ω ∈ Ω and bounded subset M ⊆ E.

Proof. The proof of formulated assertion is carried out using the same scheme as
in Theorem 5.36, therefore its proof may be omitted. 

Theorem 5.38 Let f ∈ C 1 (E × F, E), Φ ∈ C 1 (F, F ) and f + F ∈ C 1 (E × F, E)

be regular, Ω ⊆ F be a compact invariant set of dynamical system (5.105), function
f be homogeneous (of order m = 1) w.r.t. variable u ∈ E and the zero solution of
equation (5.106) be uniformly asymptotically stable.
If

|F (u, ω)| ≤ c|u| (5.130)

for all |u| ≥ r and ω ∈ Ω, where r and c are certain positive numbers, and moreover
c is sufficiently small. Then the following assertions take place:

(1) the set Iω = {u ∈ E| sup{|ϕ(t, u, ω)| : t ∈ R} < +∞} is not empty, compact
and connected for each ω ∈ Ω, where ϕ(t, u, ω) is a solution of equation

u0 = f (u, ωt) + F (u, ωt) (5.131)

satisfying the initial condition ϕ(0, u, ω) = u;

(2) ϕ(t, Iω , ω) = Iσ(t,ω) for all t ∈ R+ and ω ∈ Ω;
S
(3) the set I = {Iω |ω ∈ Ω} is compact and connected;
(4) the equalities (5.128) and (5.129) take place for every bounded subset M ⊆ E
and ω ∈ Ω.

Proof. Let h(X, R+ , π), (Y, R, σ), hi be a non-autonomous dynamical system, gen-
erated by equation (5.131). Since f ∈ C 1 (E × F, E) and the zero solution of equa-
tion (5.106) is uniformly asymptotically stable, then according to Theorem 5.35 by
224 Global Attractors of Non-autonomous Dissipative Dynamical Systems

equality
Z +∞
V (u, ω) = |ϕ(t, u, ω)|k dt
0

it is defined a continuously differentiable function V : X = E × Ω → R+ , satisfying

the conditions 2.1-2.3 of Theorem 5.33. Let us remark that
d  d

Vπ0 (u, ω) = V ((ϕ(t, u, ω), ωt) = {Du V (ϕ(t, u, ω), ωt) ϕ(t, u, ω)
dt t=0 dt
d 

+Dω V (ϕ(t, u, ω), ωt) ωt} = Du V (u, ω)[f (u, ω) + F (u, ω)]
dt t=0
+Dω V (u, ω)Φ(ω) = −|u|k + Du V (u, ω)F (u, ω). (5.132)

Since the function V (u, ω) is homogeneous of order k > 1 w.r.t. variable u ∈ E,

then in virtue of Lemma 5.6 there exists a number L > 0 such that

kDu V (u, ω)k ≤ L|u|k−1 (5.133)

for all u ∈ E and ω ∈ Ω. From (5.130) and (5.133) follows, that

|Du V (u, ω)F (u, ω)| ≤ cL|u|k (5.134)

for each |u| ≥ r and ω ∈ Ω and, consequently, according to (5.132), (5.134) we have

Vπ0 (u, ω) ≤ −|u|k + cL|u|k = −γ|u|k

for all ω ∈ Ω and |u| ≥ r, where γ = 1 − cL > 0 since a number c is sufficiently

small. For finishing the proof of the theorem it is sufficiently to refer to Theorem
2.24, 2.25 and 5.3. The theorem is proved. 

Remark 5.12 The dissipativity of equation (5.121) is established in [337]

(theorem 39 and corollary 1).
 Chapter 6

Dissipativity of some classes of equations

6.1 Difference equations

All the results about differential equations, which are presented in previous chap-
ters, for difference equations hold too, because they were formulated and proved
for general non-autonomous dynamical systems, both for dynamical systems with
continuous time and those with discrete time. Below we will give some of them that
we need to study periodical systems with impulse.
Consider a difference equation

u(k + 1) = Φ(k, u(k)) (Φ ∈ C(Z × E n , E n )). (6.1)

Denote by ϕ(·, u, Φ) a solution of the equation (6.1) passing through the point
u for k = 0. Suppose that Φ is p–periodic in k ∈ Z, where p ∈ Z. We set H(Φ) :=
{Φm = σ(Φ, m) : 0 ≤ m ≤ p − 1}, and denote by (Y, Z, σ) a dynamical system of
translations on Y := H(Φ) induced by (C(Z × E n , E n ), Z, σ). Let X := E n × Y and
define a mapping π : X × Z+ → X by the equality π((u, Φ̃), k) := (ϕ(k, u, Φ̃), Φ̃k ).
From general properties of solutions of the difference equation (6.1) follows that


(X, Z+ , π) is a semigroup dynamical system and (X, Z+ , π), (Y, Z, σ), h (h :=
pr2 : X → Y ) is a non-autonomous p–periodic dynamical system. Applying to the
above constructed non-autonomous dynamical system the results of section 4.6, we
get some assertions for the equation (6.1). Before formulating some statements of
this type we will define the notion of dissipativity for the equation (6.1).

Definition 6.1 By analogy with the case of differential equations, the difference
equation (6.1) is said to be dissipative, if there is a positive number r such that

lim sup |ϕ(k, v, Φ̃| < r

k→+∞

for all v ∈ E n and Φ̃ ∈ H(Φ).

We define a mapping P : E n → E n in the following way: P(v) := ϕ(p, v, Φ) and

denote by (E n , P ) the cascade generated by positive powers of P.

225
226 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Theorem 6.1 The equation (6.1) is dissipative if and only if the cascade (E n , P)
is dissipative.

Theorem 6.2 If the equation (6.1) is dissipative, then:

1. there exists h > 0 such that for all a > 0 there is k(a) ∈ Z+ for which
Pk B[0, a] ⊆ B[0, h] for all k ≥ k(a);
2. the mapping P has at least one fixed point u0 ∈ E n through which there passes
a p–periodic solution of (6.1).

Theorem 6.3 If the equation (6.1) is dissipative, then:

1. there is a nonempty compact set I ∈ E n which possesses the following proper-

ties:
1.a. P(I) = I;
1.b. for all ε > 0 there is δ(ε) > 0 such that ρ(u, I) < δ implies ρ(P k u, I) < ε
for all k ∈ Z+ ;
2. the set J := ∪{Jk := ϕ(k, I, Φ) : k ∈ Z+ } is nonempty, compact and possesses
the following properties:
2.a. the set J is invariant with respect to (6.1), i.e. v ∈ Jm implies
ϕ(k, v, Φm ) ∈ J for all k ∈ Z+ ;
2.b. Jk+p = Jk for all k ∈ Z+ ;
2.c. for all ε > 0 there exists δ(ε) > 0 such that ρ(ϕ(m, u, Φ), Jm ) < δ implies
ρ(ϕ(k, u, Φ), Jk ) < ε for all k ≥ m.

Theorem 6.4 Let the equation (6.1) be dissipative and the function Φ : Z×C n →
Cn be holomorphic in second variable, then the Levinson’s center of this equation
consists from a single p–periodic trajectory, i.e., the equation (6.1) is convergent.

Theorems 6.1 – 6.4 directly follow from the results of section 4.6. Concern-
ing Theorem 6.4 it is necessary to add that when the right hand side of (6.1) is
holomorphic, then the solution ϕ(·, z, Φ) (z ∈ Cn ) is holomorphic too.
Consider the quasi-linear equation

u(k + 1) = A(k)u(k) + F (k, u(k)), (6.2)

where A(k) is a non-stationary n × n–matrix and the function F ∈ C(Z × E n , E n )

satisfies some condition of ”smallness”.
Denote by U (k, A) the Cauchy’s matrix for linear equation

u(k + 1) = A(k)u(k).

Theorem 6.5 Let A(k) and F (k, u) be p-periodic in k ∈ Z+ . Suppose that the
following conditions hold:
 Dissipativity of some classes of equations 227

(1) there are positive numbers M and q < 1 such that

||U (k, A)U −1 (m, A)|| ≤ M q k−m (k ≥ m); (6.3)

(2) |F (k, u| ≤ C + D|u| (C ≥ 0, 0 ≤ D < (1 − q)M −1 ) for all u ∈ E n .

Then the equation (6.2) is dissipative.

Proof. Let ϕ(·, u, Φ) (Φ(k, u) := A(k)u + F (k, u)) be a solution of the equation
(6.2) passing through the point u ∈ E n for k = 0. According to the formula of
variation of constants (see, for example, [170]) we have

 k−1
X 
ϕ(k, u, Φ) = U (k, A) u + U −1 (m + 1, A)F (m, ϕ(m, u, Φ)) ,
m=0

and, consequently,

|ϕ(k, u, Φ)| ≤ M q k |u|
k−1
X 
+ q −m+1 (C + D|ϕ(m, u, Φ)|) . (6.4)
m=0

We set u(k) := q −k |ϕ(k, u, Φ)| and, taking into account (6.4), we obtain
k−1
X k−1
X
u(k) ≤ M |u| + CM q q −m + DM u(m). (6.5)
m=0 m=0

Denote the right hand side of the equality (6.5) by v(k). Note, that

CM DM DM CM −k
v(k + 1) − v(k) = q −k + u(k) ≤ v(k) + q ,
q q q q
and,hence,
 DM  CM −k
v(k + 1) ≤ 1 + v(k) + q .
q q
From this inequality we obtain
 DM k−1 CM 1 − q k−1
v(k) ≤ 1 + v(1) + .
q q 1−q
Therefore,
CM k−1
|ϕ(k, u, Φ)| ≤ (q + DM )k−1 qM |u| + (q − 1), (6.6)
q−1
228 Global Attractors of Non-autonomous Dissipative Dynamical Systems

because v(1) = M |u|. From (6.6) follows that

CM
lim |ϕ(k, u, Φ)| ≤ .
k→+∞ 1−q
The theorem is proved. 

Corollary 6.1 Under the conditions of Theorem 6.5 the equation (6.2) has at
least one p-periodic solution.

Proof. This statement follows from Theorems 6.2 and 6.5. 

6.2 Equations with impulse

Consider the following differential equation:

u̇ = f (t, u) (f ∈ C(R × E n , E n )). (6.7)

Suppose that f (t + τ, u) = f (t, u) (t ∈ R, u ∈ E n ) and f is regular, i.e. for the

equation

v̇ = g(t, v) (6.8)

the conditions of existence, uniqueness and continuability on R+ are fulfilled for

all g ∈ H(f ) := {fτ : τ ∈ [0, τ [}. Denote by ϕ(·, v, g) a solution of (6.8) passing
through the point v for t = 0. Note, that

ϕ(t + τ, v, g) = ϕ(t, ϕ(τ, v, g), gτ )

for all t, τ ∈ R+ , v ∈ E n and g ∈ H(f ).

Let {tk } ⊂ R (t0 = 0) and {sk } ⊂ E n be such that for certain p > 0 (p ∈
Z) tk+p − tk = τ and sk+p = sk for all k ∈ Z. It is well known [170] that
these conditions are necessary and sufficient for τ –periodicity of the distribution
P
+∞
s k δ tk .
k=−∞
Consider a nonlinear τ –periodic equation with impulse
+∞
X
u̇ = f (t, u) + sk δtk = F. (6.9)
k=−∞

It is known (see, for example, [170]) that, under the conditions above, the equa-
tion (6.9) has a unique generalized solution (which is piecewise continuous) passing
through the point u when t = 0 for every u ∈ E n . This solution we denote by
ϕ(·, u, F). Thus, lim ϕ(t, u, F) = u. In addition, on every segment ]tk , tk+1 [ the
t↓0
 Dissipativity of some classes of equations 229

equation (6.9) coincides with (6.7). Therefore, the equality

ϕ(t, u, F) = ϕ(t − tk , ck , ftk ) (6.10)

holds, where the sequence {ck } ⊂ E n satisfies the difference equation

ck+1 = ϕ(tk+1 − tk , ck , ftk ) + sk . (6.11)

The equation with impulse (6.9) is said to be dissipative if there exists r > 0
such that
e <r
lim sup |ϕ(t, v, F)| (6.12)
t→+∞

e ∈ H(F) := {Fs : s ∈ [0, τ [}, where Fs is an s-translation of the

for all v ∈ E n and F
distribution F.

Remark 6.1 For periodic equations the condition of dissipativity (6.12) can be
simplified because (6.12) is equivalent to the following relation:

lim sup |ϕ(t, u, F)| < r. (6.13)

t→+∞

Proof. In fact, (6.12) implies (6.13). Conversely. From the equality ϕ(t + s, u, F) =
ϕ(t, ϕ(s, u, F), Fs ) (if in the point s ∈ R the function ϕ(t, u, F) is discontinuous, then
ϕ(s, u, F) := ϕ(s + 0, u, F)) for all t, τ ∈ R+ . Therefore, we have
e = ϕ(t − τ + s, ϕ(τ − s, v, Fk ), F)
ϕ(t, v, F)
e = Fk .
for all t ≥ τ − s, where s ∈ [0, τ [ and F 

Lemma 6.1 The equation with impulse (6.9) is dissipative if and only if the
difference equation (6.11) is dissipative.

Proof. Let (6.9) be dissipative. Then there exists r > 0 such that the condition
(6.13) is fulfilled. If u ∈ E n , then

ϕ(k, u, Φ) = ϕ(tk , u, F)

and, hence,

lim sup |ϕ(k, u, Φ)| < r. (6.14)

k→+∞

Conversely. Let the equation (6.11) be dissipative. Then there is r > 0 such that
the condition (6.14) holds. Thus, for all u ∈ E n there exists k0 (u) ∈ Z+ such that
|ϕ(k, u, Φ)| < r for all k ≥ k0 (u). Let t ≥ tk0 and t ∈ [tk , tk+1 [ k ≥ k0 ). By (6.10),
ϕ(t, u, F) = ϕ(t − tk , ck , ftk ), where ck = ϕ(tk , u, F). We set

R0 := max{|ϕ(s, u, g)| : s ∈ [0, τ ], |u| ≤ r, g ∈ H(f )}

230 Global Attractors of Non-autonomous Dissipative Dynamical Systems

and note that |ϕ(t, u, F)| = |ϕ(t − tk , ck , ftk )| ≤ R0 for all t ≥ tk0 and, consequently,

lim sup |ϕ(t, u, F)| ≤ R0 . (6.16)

t→+∞

Moreover, the number R0 does not depend on u ∈ E n in the inequality (6.16). The
lemma is proved. 

Define a mapping P : E n → E n by the equality

P (u) := ϕ(τ, u, F) = ϕ(tp , u, F) = ϕ(p, u, Φ). (6.17)

From the equality (6.17) and general properties of solutions of difference equations
follows the continuity of the mapping P .

Theorem 6.6 The equation with impulse (6.9) is dissipative if and only if the
cascade (E n , P ), where P is defined by the equality (6.17), is dissipative.

Proof. This affirmation directly follows from Lemma 6.1 and Theorem 6.1. 

Corollary 6.2 If the equation (6.9) is dissipative, then there exists a nonempty
compact connected set I ⊂ E n possessing the following properties:

(1) P (I) = I, where P is a mapping of E n onto itself, defined by the formula (6.17);
(2) for an arbitrary ε > 0 there is δ(ε) > 0 such that ρ(u, I) < δ implies
ρ(P k (u), I) < ε for all k ∈ Z+ ;
(3) the equality lim ρ(P k (u), I) = 0 holds for all u ∈ E n .
k→+∞

Proof. Corollary 6.2 follows from Theorems 6.6 and 4.9. 

Corollary 6.3 If the equation (6.9) is dissipative, then:

(1) there is r > 0 and for all a > 0 there is k(a) ∈ Z+ such that P k B[0, a] ⊆ B[0, r]
for all k ≥ k(a);
(2) there is u0 ∈ I such that P (u0 ) = u0 and, consequently, ϕ(t, u0 , F) is a τ -
periodic solution of (6.9).

Consider a quasi-linear equation with impulse

+∞
X
u̇ = A(t)u + F (t, u) + s k δ tk . (6.18)
k=−∞

The following statement holds.

Theorem 6.7 Let A and F be τ -periodic in t ∈ R. The equation (6.18) is

dissipative, if the following conditions are fulfilled:
 Dissipativity of some classes of equations 231

(1) there are positive numbers N and ν such that

||U (t, A)U −1 (s, A)|| ≤ N e−ν(t−s) ,

where U (t, A) is a Cauchy matrix for the equation

u̇ = A(t)u.

(2) there exists a sufficiently small positive number ε0 such that |F (t, u)| ≤ C +
D|u| (u ∈ E n , C ≥ 0 and 0 ≤ D ≤ ε0 ).

Proof. Let f (t, u) := A(t)u + F (t, u). According to Lemma 6.1, to prove Theorem
6.7 it is sufficient to show that the difference equation (6.11) is dissipative. Let us
rewrite the equation (6.11) as follows:

e
ck+1 = A(k)c e
k + F (k, ck ), (6.19)

e
where A(k) := U (tk+1 − tk , Atk ) and

Fe(k, u) := ϕ(tk+1 − tk , u, ftk ) + sk − A(k)u.

e

e
Note, that under the conditions of Theorem 6.7 A(k) and Fe (k, u) are p–periodic in
k ∈ Z.
We will show that the trivial solution of the homogeneous equation

e
ck+1 = A(k)c k (6.20)

e
is uniformly asymptotically stable. In virtue of the periodicity of A(k) it is sufficient
to show that it is asymptotically stable. Note, that
k−1
Y
e =
U (k, A) U (tm+1 − tm , Atm ) = U (tk − tp−1 , Atp−1 ),
m=0

e ≤ N e−ν(tk −tp−1 ) . From this fact follows the asymptotic sta-

therefore, ||U (k, A)||
bility of the trivial solution of (6.20).
We will show now that |Fe(k, u)| ≤ C e + D(D)|u|
e e and
for all u ∈ E n , where C
e
D(D) e
are some positive constants, moreover, D(D) → 0 as D → 0. Indeed, since
ϕ(t, u, fl ) is a solution of the equation

u̇ = fl (t, u) = Al (t)u + Fl (t, u),

Rt

then ϕ(t, u, fl ) = U (t, Al ) u + o
U −1 (s, Al )Fl (, ϕ(s, u, fl ))ds and, consequently,
Z tk+1 −tk
Fe(k, u) = U (tk+1 − tk , Atk )U −1 (s, Atk )Ftk (s, ϕ(s, u, ftk ))ds + sk . (6.21)
0
232 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Using the same arguments as in Theorem 3.21, we can show that

CN
|ϕ(t, u, fl )| ≤ N |u|e(−ν+N D)t + (6.22)
ν − DN
for all t ≥ 0, l ∈ R and u ∈ E n (0 ≤ D ≤ νN −1 ). From (6.21) and (6.22) we have

Z −tk
tk+1

|Fe(k, u)| ≤ N e−ν(tk+1 −tk −s) (C + D|ϕ(s, u, fl )|)ds ≤

0

Z −tk
tk+1
 h i
−ν(tk+1 −tk ) CN
Ne Ceνs + Deνs + N |u|e−ν+N D)s ds
ν − DN
0

h CN i eν(tk+1 −tk ) − 1
= N e−ν(tk+1 −tk ) C +D +
ν − DN ν

eN D(tk+1 −tk ) − 1  NC  
+N |u|D = 1 − e−ν(tk+1 −tk ) +
ND ν − DN
 
e + D(D)|u|,
N |u| e(N D−ν)(tk+1 −tk ) − e−ν(tk+1 −tk ) ≤ C e

e := N C(ν − N D)−1 and

where C
h i
e
D(D) := N sup e(N D−ν)(tk+1 −tk ) − e−ν(tk+1 −tk ) . (6.23)
0≤k≤p−1

e
From (6.23) follows that D(D) → 0 as D → 0. Therefore, there is ε0 > 0, such
e 0 ) < M (1 − q) (see Theorem 6.5). According to Theorem 6.5, the
that D(ε −1

difference equation (6.19) is dissipative and, hence, the equation with impulse (6.18)
is dissipative too. The theorem is proved. 

6.3 Convergent periodic equations with impulse

Definition 6.2 Following [137], [270], the equation with impulse (6.9) we will
call convergent, if (6.9) has a unique τ –periodic solution ϕ(t, u, F) which is globally
uniformly asymptotically stable.

By analogy, it is possible to define the notion of convergence for the difference

equation (6.1). From this definition follows that convergent equations form that
part of dissipative equations which has the Levinson’s center consisting from a
single periodic trajectory (solution) of (6.9).
 Dissipativity of some classes of equations 233

Lemma 6.2 The τ -periodic equation with impulse (6.9) is convergent if and only
if (6.11) is convergent.

Proof. Let (6.9) be convergent. Then there is a unique τ -periodic solution

ϕ(t, u0 , F) of (6.9) which is globally uniformly asymptotically stable. Directly
from the corresponding definitions follows that ϕ(t, u0 , Φ) (Φ(k, u) := ϕ(tk+1 −
tk , u, ftk ) + sk ) is a unique p-periodic solution of (6.11) and is globally uniformly
asymptotically stable.
Conversely. Let (6.11) be convergent and ϕ(k, u0 , Φ) be its unique p-periodic
globally uniformly asymptotically stable solution. Obviously, ϕ(t, u0 , F) is a unique
τ -periodic solution of (6.9). Let k ∈ Z and t ∈ [tk , tk+1 [. According to (6.10),

|ϕ(t, u, F) − ϕ(t, u0 , F)| = |ϕ(t − tk , ck , fτk ) − ϕ(t − tk , c0k , ftk )|, (6.24)

where ck := ϕ(tk , u, F) and c0k := ϕ(tk , u0 , F). The mapping ϕ : [0, τ ] × K × H(f ) →
E n (ϕ : (s, u, g) → ϕ(s, u, g)) is uniformly continuous for every compact K ⊂ E n
and, hence, for all ε > 0 there is δ(ε) > 0 such that

|ϕ(s, u1 , g) − ϕ(s, u2 , g)| < ε, (6.25)

for all s ∈ [0, τ ], |u1 − u2 | < δ, u1 , u2 ∈ K and g ∈ H(f ).

From (6.24) and (6.25) and the global uniform asymptotic stability of
ϕ(k, u0 , Φ) = c0k follows an analogous property for ϕ(t, u0 , F). The lemma is proved.


Theorem 6.8 Let A(k) and F (k, u) be p–periodic in k ∈ Z and the following
conditions be fulfilled:

(1) there exist positive numbers M and q < 1 such that the inequality (6.3) holds;
(2) |F (k, u1 ) − F (k, u2 )| ≤ L|u1 − u2 | (0 ≤ L < M −1 (1 − q)) for all k ∈ Z and
u1 , u2 ∈ E n .

Then the equation (6.2) is convergent.

Proof. Let ϕ(k, u, Φ) (Φ(k, u) := A(k)u + F (k, u)) be a solution of (6.2) passing
through the point u for k = 0. In virtue of the formula of variation of constants we
have

 k−1
X 
ϕ(k, u, Φ) = U (k, A) u + U −1 (m + 1, A)F (m, ϕ(m, u, Φ)) .
m=0
234 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Consequently,

ϕ(k, u1 , Φ) − ϕ(k, u2 , Φ) = U (k, A) u1 − u2 +

k−1
X 
U −1 (m + 1, A)[F (m, ϕ(m, u, Φ)) − F (m, ϕ(m, u2 , Φ))] .
m=1

Thus,

|ϕ(k, u1 , Φ) − ϕ(k, u2 , Φ)| ≤ M q k |u1 − u2 |
k−1
X 
+Lq −1 q −m |ϕ(k, u1 , Φ) − ϕ(k, u2 , Φ)| . (6.26)
m=0

Let u(k) := |ϕ(k, u1 , Φ) − ϕ(k, u2 , Φ)|q −k . From (6.26) follows that

 k−1
X 
u(k) ≤ M |u1 − u2 | + Lq −1 u(m) . (6.27)
m=0

Denote by v(k) the right hand side of (6.27). The following inequality

v(k + 1) − v(k) = LM q −1 u(k) ≤ LM q −1 v(k). (6.28)

holds. From (6.28) we obtain

v(k) ≤ (1 + LM q −1 )k−1 v(1)

and, hence,

u(k) ≤ (1 + LM q −1 )M |u1 − u2 |, (6.29)

since v(1) = M |u1 − u2 |. From (6.29) we have

|ϕ(k, u1 , Φ) − Φ(k, u2 , Φ)| ≤ (q + LM )k−1 qM |u1 − u2 | (6.30)

for all u1 , u2 ∈ E n and k ∈ Z.

Let us note that under the conditions of Theorem 6.8 we can apply Theorem
6.5 and, consequently, the equation (6.2) admits at least one p-periodic solution
ϕ(k, u0 , Φ) (see also Corollary 6.1). From (6.30) follows the global asymptotic sta-
bility of the solution ϕ(k, u0 , Φ) and, according to the periodicity of (6.2), the
solution ϕ(k, u0 , Φ) will be globally uniformly asymptotically stable. The theorem
is proved. 

Theorem 6.9 Let A(t) and F (t, u) be τ -periodic in t ∈ R. The equation (6.18)
is convergent, if the following conditions are fulfilled:
 Dissipativity of some classes of equations 235

(1) there are positive numbers N and ν such that

||U (t, A)U −1 (s, A)|| ≤ N e−ν(t−s) (t ≥ s);

(2) there is a sufficiently small positive number ε0 such that

|F (t, u1 ) − F (t, u2 )| ≤ L|u1 − u2 |

(u1 , u2 ∈ E n , t ∈ R and 0 ≤ L ≤ ε0 ).

Proof. According to Lemma 6.2 to prove the convergence of (6.18) it is sufficient

to establish that under the conditions of Theorem 6.9 the difference equation (6.11)
possesses the same property. As well as in the proof of Theorem 6.7, we will present
the equation (6.11) in the form (6.19). Under the conditions of Theorem 6.9 the
trivial solution of (6.20) is uniformly asymptotically stable. Further, (6.21) implies
that
Z −tk
tk+1

|Fe (k, u1 ) − Fe(k, u2 )| = | U (tk+1 − tk , Atk )U −1 (s, Atk )×

0

[Ftk (s, ϕ(s, u1 , ftk ) − Ftk (s, ϕ(s, u2 , ftk ))]ds ≤ (6.31)

Z −tk
tk+1

N e−ν(tk+1 −tk −s) L|ϕ(s, u1 , ftk − ϕ(s, u2 , ftk )|ds.

0

Reasoning in the same way that in Theorem 5.23 we obtain

|ϕ(t, u1 , ftk − ϕ(t, u2 , ftk )| ≤ N |u1 − u2 |e−(ν−LN )t (6.32)

for all u1 , u2 ∈ E n and t ∈ R+ , where ftk (t, u) := Atk (t)u + Ftk (t, u). From (6.31)
and (6.32) we have

|Fe (k, u1 ) − Fe(k, u2 )| ≤ N e−ν(tk+1 −tk ) L ×

Z −tk
tk+1
2
e
eLN s N |u1 − u2 |ds ≤ L(L)|u 1 − u2 |,
0

where
h 2
i
e
L(L) := sup e(−ν+LN )(tk+1 −tk )
− e−ν(tk+1 −tk ) . (6.33)
0≤k≤p−1

e
From (6.33) follows that L(L) → 0 for L → 0. Therefore, there exists ε0 > 0 such
e −1
that L(L) < M (1 − q) for 0 ≤ L ≤ ε0 . By Theorem 6.8, the equation (6.20) is
convergent and to finish the proof of Theorem 6.9 it is sufficient to refer to Lemma
6.2. 
236 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Remark 6.2 a. A statement similar to Theorem 6.9 has also been proved in the
work [256].
b. The problem of the convergence and dissipativity of equations with impulse
are studied in the works [256], [115], [116], [256].

6.4 Asymptotic stability of linear functional differential equations

Using some ideas and methods developed while studying dissipative dynamical
systems we may receive a series of conditions which are equivalent to the asymp-
totic stability of linear non-autonomous dynamical systems with infinite dimensional
phase space. As an application of these result we may receive the corresponding
results for linear functional differential equations.
Let (X, h, Y ) be a vectorial fibering with the fiber E (E is a Banach space)
and | · | : X → R be a norm on X, compatible with the distance of X, i.e. | · | is
continuous and |x| = ρ(x, θy ), where x ∈ Xy , θy is the null element of the space Xy
and ρ is a distance on X.
Recall that a dynamical system (X, S+ , π) is called locally compact (locally
completely continuous) if for any x ∈ X there are δx > 0 and lx > 0 such that
π t B(x, δx ) (t ≥ lx ) is relatively compact.

Theorem 6.10 Let (X, S+ , π) be locally compact and Y be compact. Then the
following conditions are equivalent:

1. lim |xt| = 0 for all x ∈ X;

t→+∞
2. all the motions in (X, S+ , π) are relatively compact and (X, S+ , π) does not
admit nontrivial compact motions defined on S;
3. there are positive numbers N and ν such that |xt| ≤ N e−νt |x| for all x ∈ X
and t ∈ S+ .

Proof. From the equality lim |xt| = 0 follows that Σ+

x is relatively compact
t→+∞
and ωx ⊆ Θ := {θy : y ∈ JY , where θy is the null element of Xy }. Thus, the
dynamical system (X, S+ , π) is pointwise dissipative and, according to Theorem
1.10, it is compactly dissipative. Denote by JX the Levinson’s center of the
dynamical system (X, S+ , π). We will show that JX = Θ. Obviously, the set Θ
is compact and invariant (π t Θ = Θ for all t ∈ S+ ) and, consequently, Θ ⊆ JX .
From the last inclusion follows that h(JX ) = JY . Now let us show that JX = Θ. If
we suppose that it is not true, then JX \Θ 6= ∅ and, hence, there is x0 ∈ JX \Θ. Since
in the set JX all motions are continuable on S (see Theorem 1.6), then there exists
a continuous mapping ϕ : S → JX such that: ϕ(0) = x0 and π t ϕ(s) = ϕ(t + s)
for all s ∈ S and t ∈ S+ . On the other hand, by the linearity of the system


(X, S+ , π), (Y, S+ , σ), h along with the point x0 all the points λx0 (λ ∈ R) be-
 Dissipativity of some classes of equations 237

long to the set JX , because JX is a maximal compact invariant set in X. But

the inclusion λx0 ∈ J holds for all λ ∈ R if and only if x0 ∈ Θ. The obtained
contradiction shows that JX = Θ.
We will show that in (X, S+ , π) there is no nontrivial compact motions defined
on S. In fact, let x ∈ X be such point that there exists ϕ : S → X possessing the
following properties: ϕ(S) is relatively compact, π t ϕ(s) = ϕ(t + s) (t ∈ S+ , s ∈ S)
and ϕ(0) = x0 . Since JX is a maximal compact invariant set, then ϕ(S) ⊆ JX and,
in particular, x0 ∈ JX ⊆ Θ. Therefore, |x0 | = 0. Thus, we proved that 1. implies
2. The converse implication is evident.
We will show that 1. implies 3. Indeed, since under the conditions of Theorem
6.10 from 1. follows that (X, S+ , π) is pointwise dissipative, then, according to
Theorem 1.10, it is locally dissipative. As Y is compact and (X, S+ , π) is locally
dissipative, then there is δ > 0 such that

lim sup{|xt| : |x| < δ} = 0. (6.34)

t→+∞

Taking into consideration (6.34), by standard reasoning (see, for example, [208],
[275],and also Theorem 2.38) we may prove that there are N, ν > 0 such that
|xt| ≤ N e−νt |x| for all x ∈ X and t ≥ 0. Finally, 3. evidently implies 1. The
theorem is proved. 

Remark 6.3 a. The condition of local compactness in Theorem 6.10 is essential.

To confirm this statement it is sufficient to consider the example 1.8. In this example
all motions tend to zero as t → +∞, but this system is not exponentially stable. It
is not difficult to see that the dynamical system from the example 1.8 is not locally
compact.
b. Theorem 6.10 is also true if the space Y is just pseudo-metric.

A very important class of linear non-autonomous dynamical systems with infinite

dimensional phase space satisfying the condition of local compactness is the class
of linear non-autonomous functional differential equations [175]. Let us recall some
notions and denotations from [175]. Let r > 0, C([a, b], Rn ) be a Banach space of all
continuous functions ϕ : [a, b] → Rn equipped with the norm sup. If [a, b] = [−r, 0],
then we set C := C([−r, 0], Rn ). Let σ ∈ R, A ≥ 0 and u ∈ C([σ −r, σ +A], Rn ). We
will define ut ∈ C for all t ∈ [σ, σ + A] by the equality ut (θ) := u(t + θ), −r ≤ θ ≤ 0.
Denote by D = D(C, Rn ) a Banach space of all linear continuous operators acting
from C into Rn and endowed with operational norm. Consider a linear equation

u̇ = A(t, ut ), (6.35)

where A : R × C → Rn is continuous and linear with respect to the second variable,

i.e. A(t, ·) ∈ D for all t ∈ R. Let us set H(A) := {As : s ∈ R}, where As (t, ·) =
A(t + s, ·) and by bar we denote a closure in the compact-open topology on R × C.
238 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Along with the equation (6.35) let us consider the family of equations

v̇ = B(t, vt ), (6.36)

where B ∈ H(A). Let ϕ(·, φ, B) be a solution of (6.36) passing through the point
φ ∈ C for t = 0 defined for all t ≥ 0.
Let Y := H(A) and denote by (Y, R, σ) a dynamical system of translations on
H(A). Let X := C × Y , (X, R+ , π) be a dynamical system on X defined in the
following way: π((φ, B), τ ) := (ϕ(τ, φ, B), Bτ ). It is easy to see that the system


(X, R+ , σ), (Y, R, σ), h is linear. We remark a very important property of this
system.
The following statement holds.

Lemma 6.3 Let H(A) be compact. Then for any point x ∈ X := C × H(A) there
exist a neighborhood Ux of the point x and a positive number lx > 0 such that π t Ux
is relatively compact for all t ≥ lx , i.e. the dynamical system (X, R+ , π) is locally
compact.

Proof. This assertion follows from Lemmas 2.2.3 and 3.6.1 from [175] and from the
compactness of Y . 

Applying to the constructed non-autonomous dynamical system Theorem 6.10

(see also Remark 6.3) and taking into consideration Lemma 6.3, we will obtain the
following statement.

Theorem 6.11 Let H(A) be compact. Then the following statements are equiv-
alent:

(1) the trivial solution of (6.35) is uniformly exponentially stable, i.e. there are
positive numbers N and ν such that |ϕ(t, φ, B)| ≤ N e−νt |φ| for all B ∈ H(A)
and t ≥ 0;
(2) the trivial solution of (6.36) is uniformly exponentially stable for all B ∈ H(A).

6.5 Convergence of monotone evolutionary equations

Let H be a Hilbert space. Denote by Lploc (R, H) the space of all functions ϕ : R → H
that are locally integrable with the power p, that is
Z
|ϕ(t)|p dt < +∞ (p ≥ 1, k = 0, 1, 2 . . .).
|t|≤k
 Dissipativity of some classes of equations 239

Let us define a topology on Lploc (R, H) by the family of semi-norms | · |k :

 Z 1
p p
|ϕ|k := |ϕ(t)| dt (k = 0, 1, 2, . . .)
|t|≤k

and by (Lploc (R, H), R, σ) we denote the dynamical system of translations on

Lploc (R, H) (see [103]).

Definition 6.3 Recall [30] that an operator A : D(A) → H (D(A) ⊆ H) is called:

a. monotone, if RehAu1 − Au2 , u1 − u2 i ≥ 0 for all u1 , u2 ∈ D(A);

b. semi-continuous, if the function ϕ : R → R defined by the equality ϕ(λ) :=
hA(u + λv), wi is continuous;
c. uniformly monotone, if there exists α > 0 such that

RehAu − Av, u − vi ≥ α|u − v|2 (6.37)

for all u, v ∈ D(A).

Note, that the family of all monotone operators can be partially ordered with
respect to the inclusion of its graphs.

Definition 6.4 A monotone operator is called maximal, if it is maximal among

monotone operators.

Consider a differential equation

x0 + Ax = f (t), (6.38)

where f ∈ L1loc (R, H) and A is a maximal monotone operator with the domain
D(A). It is known [30] that for all x0 ∈ D(A) there exists a unique weak solution
ϕ(t, x0 , f ) of the equation (6.38) satisfying the condition ϕ(0, x0 , f ) = x0 and defined
on R+ . Let Y := H + (f ) and (Y, R+ , σ) be a dynamical system of translations on Y .
We denote by X := D(A) × Y and by (X, R+ , π) a dynamical system on X, where
π(t, hv, gi) := hϕ(t, v, g), gt i. Then the triple h(X, R+ , π), (Y, R+ , σ), hi(h = pr2 :
X → Y ) is the non-autonomous dynamical system [187] generated by the equation
(6.38). Applying the results obtained in Chapters 1-5 to the constructed in such
way non-autonomous dynamical system, we will obtain the following results for the
equation (6.38).

Theorem 6.12 Let f ∈ L1loc (R, H) and H(f ) be compact. If A is a maximal

monotone operator which is semi-continuous and uniformly monotone, then the
equation (6.38) is convergent, i.e. the equation (6.38) admits a unique compact on
240 Global Attractors of Non-autonomous Dissipative Dynamical Systems

R uniformly compatible solution which is globally asymptotically stable. In addition,

there exist positive numbers N and ν such that the equation

x0 + Ax = g(t) (6.39)

holds for all g ∈ H(f ) and admits a unique compact on R uniformly compatible
solution ϕ(t, vg , g) (g ∈ H(f )) such that

|ϕ(t, v, g) − ϕ(t, vg , g)| ≤ N eνt |v − vg | (6.40)

for all v ∈ D(A) and t ≥ 0.

Proof. Modifying the results from [187], we obtain that under the conditions of
Theorem 6.12 all the solutions of the equation (6.38) are compact on R+ . On the
other hand, by the condition (6.37) we have

|ϕ(t, x1 , f ) − ϕ(t, x2 , f )| ≤ e−αt |x1 − x2 | (6.41)

for all t ≥ 0 and x1 , x2 ∈ H. And, moreover, if g ∈ H(f ), then for solutions of the
equation (6.39) the following estimation

|ϕ(t, y1 , g) − ϕ(t, y2 , g)| ≤ e−αt |y1 − y2 | (6.42)

takes place for all t ≥ 0 and y1 , y2 ∈ H. The relation (6.42) implies the condi-
tion 4. of Theorem 2.15, if we take as V : X × X → R+ the following function
V ((v1 , g), (v2 , g)) := |v1 − v2 | (g ∈ H(f ), v1 , v2 ∈ D(A)). The theorem is proved.

Corollary 6.4 Let f ∈ L1loc (R, H) be stationary (τ −periodic, almost periodic,

recurrent). Then the equation (6.38) admits a unique stationary (τ −periodic, almost
periodic, recurrent) solution wich is globally asymptotically stable.

Finally, we will give an example of equations of the type (6.38).

Example 6.1 Consider the equation

∂2u ∂u
2
= ∆u − φ( ) + f (t) (6.43)
∂t ∂t
in the open set D ⊂ Rn with the boundary condition u|∂D = 0 on the boundary
∂D of D. Suppose that the function φ : R → R satisfies the conditions : φ(0) = 0
and 0 < c1 ≤ φ0 (ξ) ≤ c2 (ξ ∈ R). Then the equation (6.43) may be rewritten in the
following way:
(
∂u
∂t = v . (6.44)
∂v
∂t = ∆u − φ(v) + f (t)
 Dissipativity of some classes of equations 241

We denote by H := W01,2 (Ω) × L2 (Ω) and define on H a scalar product

Z
h(u, v), (u , v )i = [vv ∗ + ∆u∆u∗ + λuv ∗ + λu∗ v]dx,
∗ ∗

where λ is a certain positive constant depending only on c1 and c2 . It is possible

to verify (see, for example, [131]) that Theorem 6.12 can be applied to the system
(6.44), if the set H(f ) ⊂ L1loc (R, H) is compact.

Let I ⊆ R , H be a Hilbert space with scalar product h, i, D(I, R) be a space of

all infinitely differentiable functions ϕ : I → H with compact support and [H] be
an algebra of all linear bounded operators on H.
Consider the equation
Z
[hu(t), ϕ0 (t)i + hA(t)u(t), ϕ(t)i + hf (t), ϕ(t)i]dt = 0, (6.45)
R

where A ∈ C(R, [H]) and f ∈ C(R, H). A function u ∈ C(I, H) is called a solution
of the equation (6.45), if the equality (6.45) takes place for all ϕ ∈ D(I, H).
Let x ∈ H, ϕ(t, x, A, f ) be a solution of the equation (6.45) defined on R+ and
satisfying the condition ϕ(0, x, A, f ) = x and
Z
[hu(t), ϕ0 (t)i + hB(t)u(t), ϕ(t)i + hg(t), ϕ(t)i]dt = 0 (6.46)
R

be a family of equations, where (B, g) ∈ H(A, f ) := {(Aτ , fτ ) | τ ∈ R}. We will

suppose that the operator-function A ∈ C(R, [H]) is self-adjoint and negatively
defined, i.e. A(t) = −A1 (t) + iA2 (t) for all t ∈ R, where A1 (t) and A2 (t) are
self-adjoint and

hA1 (t)u, ui ≥ α|u|2 (6.47)

for all t ∈ R and u ∈ H, where α > 0.

Lemma 6.4 [5] We have

1 d 2
2 dt |ϕ(t, x, A, f )| = −hA1 (t)ϕ(t, x, A, f ), ϕ(t, x, A, f )i
+Rehf (t), ϕ(t, x, A, f )i (6.48)

for all t > 0.

Lemma 6.5 The following inequality

Z t
|ϕ(t, x, A, f )| ≤ |x| + |f (τ )|dτ (6.49)
0

holds for all t ≥ 0.

242 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Proof. In virtue of the equality (6.48) we have

1 d
|ϕ(t, x, A, f )|2 ≤ |f (t)||ϕ(t, x, A, f )|.
2 dt
p
Let v(t) := |ϕ(t, x, A, f )|2 . Then dvdt ≤ 2|f (t)| v(t) and, consequently,

p p Z t
v(t) − v(τ ) ≤ |f (s)|ds.
τ

From this follows the inequality (6.49). 

Lemma 6.6 Let l, r and β > 0, x0 ∈ H, A ∈ C(R, [H]) and f ∈ C(R, [H]). Then
there exists M = M (f, l, r, β, x0 ) > such that

|ϕ(t, x, B, g) − ϕ(t, x0 , A, f )| ≤ |x − x0 |+
Z t Z t
M kB(τ ) − A(τ )kdτ + |g(τ ) − f (τ )|dτ (6.50)
0 0

for all t ∈ [0, l] and x ∈ B[x0 , r] := {x | x ∈ H, |x − x0 | ≤ r}, if |g(t) − f (t)| ≤ β

and RehB(t)x, xi ≤ 0 for any t ∈ [0, l] and x ∈ H.

Proof. We denote by v(t) = ϕ(t, x, B, g) − ϕ(t, x0 , A, f ). Then

Z
[hv(t), ϕ0 (t)i + hA(t)v(t), ϕ(t)i+
R

hB(t) − A(t))v(t), ϕ(t)i + hg(t) − f (t), ϕ(t)i]dt = 0

for any ϕ ∈ D(R, H). By virtue of Lemma 6.4

1 d
|v(t)|2 = RehA(t)v(t), v(t)i
2 dt

+Re[h(B(t) − A(t))ϕ(t, x, B, f ), v(t)i + hg(t) − f (t), v(t)i]

and, according to Lemma 6.5, we have

Z t
|v(t)| ≤ |v(0)| + |(B(τ ) − A(τ ))ϕ(τ, x, B, g) + g(τ ) − f (τ )|dτ
0

Z t Z t
≤ |v(0)| + kB(τ ) − A(τ )k|ϕ(τ, x, B, g)|dτ + |g(τ ) − f (τ )|dτ. (6.51)
0 0

On the other hand, according to Lemma 6.5 for ϕ(t, x, B, g) we have

Z t
|ϕ(t, x, B, g)| ≤ |x| + |g(τ )|dτ ≤ |x0 | + r + βl
0
 Dissipativity of some classes of equations 243

+l max |f (t)| := M (f, l, r, β, x0 ). (6.52)

0≤t≤l

Taking into account the inequalities (6.51) and (6.52), we obtain (6.50). The lemma
is proved. 
Let X̄ = H × H + (A, f ) and denote by X the set of all hu, (B, g)i ∈ X̄ such that
through the point u ∈ H there passes a solution ϕ(t, u, B, g) of the equation (6.46)
defined on R+ .

Lemma 6.7 The set X ⊆ H × H(A, f ) is closed in H × H(A, f ).

Proof. Let hx, (B, g)i ∈ X̄. Then there exists a sequence hxk , (Bk , gk )i ∈ X such
that xk → x in the space H, Bk → B in C(R, [H]) and gk → g in C(R, H). Let l
and ε > 0 be such that

|xk − xm | < ε, |gk (t) − gm (t)| < ε and kBk (t) − Bm (t)khε (6.53)

for all t ∈ [0, l] and k, l ≥ k0 . Denote by r := sup{|xk | : k ∈ N}. Then according

to Lemma 6.6
Z t
|ϕ(t, xk , Bk , gk ) − ϕ(t, xm , Bm , gm )| ≤ |xk − xm | + M kBk (τ ) − Bm (τ )kdτ
0

Z t
+ |gk (τ ) − gm (τ )|dτ ≤ ε + M εl + εl (6.54)
0

for all t ∈ [0, l] and k, m ≥ k0 , where M is a positive constant which is not

dependent on r, l and g. Taking into account that the space C(R+ , H) is com-
plete and the inequality (6.54), we conclude that the sequence {ϕ(t, xk , Bk , gk )}
is convergent in C(R+ , H) and, according to the inequality (6.54), ϕ(t, x, B, g) =
lim ϕ(t, xk , Bk , gk ). The lemma is proved. 
k→+∞

Lemma 6.8 The mapping ϕ : R+ × X → H(ϕ : (t, hu, B, gi) → ϕ(t, u, B, g)) is
continuous.

Proof. Let tn → t, xk → x, Bk → B and gk → g. Then

|ϕ(t, xk , Bk , gk ) − ϕ(t, x, B, g)| ≤ |ϕ(t, xk , Bk , gk ) − ϕ(tk , x, B, g)|

+|ϕ(tk , x, B, g) − ϕ(t, x, B, g)| ≤ max |ϕ(t, xk , Bk , gk ) − ϕ(t, x, B, g)|

0≤t≤l

+|ϕ(tk , x, B, g) − ϕ(t, x, B, g)|. (6.55)

By virtue of the inequality (6.55) and Lemma 6.6 we obtain the necessary assertion.
The lemma is proved. 
244 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Lemma 6.9 For all (B, g) ∈ H + (A, f ) and x1 , x2 ∈ H we have

|ϕ(t, x1 , B, g) − ϕ(t, x2 , B, g)| ≤ e−αt |x1 − x2 |

for any t ∈ R+ .

Proof. If the operator-function A(t) is negatively defined, then every operator-

function B ∈ H + (A) is negatively defined too and RehB(t)u, ui ≥ α|u|2 (t ∈
R, u ∈ H), where α > 0 is the same constant as the one figuring in (6.47) for the
operator-function A(t). Let ω(t) := ϕ(t, x1 , B, g) − ϕ(t, x2 , B, g). Then, according
to Lemma 6.4, we have

1 d
|ω(t)|2 = RehB(t)ω(t), ω(t)i ≤ −α|ω(t)|2
2 dt
and, consequently, |ω(t)| ≤ |ω(0)|e−αt for all t ∈ R+ . 

Let us define on X a dynamical system in the following way: π(t, x) :=

π(hu, (b, g)i, t) = hϕ(t, u, B, g), (Bt , gt )i for all hu, (B, g)i ∈ X and R+ . Let
Y = H(A, f ) and (Y, R, σ) be a dynamical system of translations on Y and
h = pr2 : X → Y . Then the triple h(X, R+ , π), (Y, R, σ), hi is the non-autonomous
dynamical system generated by equation (6.45).

Definition 6.5 We will call the equation (6.45) convergent, if it admits a compact
solution on R that is globally asymptotically stable.

According to the results of Chapter 2 (see section 2.5), the equation (6.45) will
be convergent if and only if the non-autonomous dynamical system h(X, R+ , π), (Y,
R+ , σ), hi generated by the equation (6.45) is convergent.

Theorem 6.13 Let A ∈ C(R, [H]), f ∈ C(R, H) and H(A, f ) be compact. Then
the equation (6.45) is convergent.

Proof. According to the results from [59], [187], all the solutions of the equation
(6.45) are compact on R+ and by virtue of Lemma 6.9 every solution of the equation
(6.45) is globally asymptotically stable. 

6.6 Global attractors of non-autonomous Lorenz systems

This section is devoted to the study of non-autonomous Lorenz systems. The

problems are formulated and solved in the context of non-autonomous dynamical
systems. First, we prove that such systems admit a compact global attractor and
 Dissipativity of some classes of equations 245

characterize its structure. Then, we obtain conditions of convergence of the non-

autonomous Lorenz systems, under which all solutions approach a point attrac-
tor. Third, we derive a criterion for existence of almost periodic (periodic, quasi-
periodic) and recurrent solutions of the systems. Finally, we prove a global averaging
principle for non-autonomous Lorenz systems.
The following n-dimensional systems of differential equations are called systems
of hydrodynamic type or autonomous Lorenz systems([306]):

u0i = Σj,k bijk uj uk + Σj aij uj + fi , i = 1, 2, ..., n, (6.56)

where Σbijk ui uj uk is identically equal to zero, Σaij ui uj is negative definite, and

fi are constants. The well known three-dimensional Lorenz system for geophysical
flows or climate modelling [242] is a special case of this type of systems.
It is known that solutions of (6.56) imbed in some ellipsoid and do not leave it
later, i.e. the autonomous system (6.56) is dissipative, and hence admits a compact
global attractor. In the vector-matrix form the system (6.56) may be written as:

u0 = Au + B(u, u) + f, (6.57)

where A is a positive definite matrix and B : H × H → H (H is a n-dimensional

real or complex Euclidean space) is a bilinear form satisfying the identity

RehB(u, v), wi = −RehB(u, w), vi

for every u, v, w ∈ H.
When f is not a constant vector but a bounded function of time t, it is known
that the equation (6.57) also admits a compact global attractor [218].
Our aim is to study the non-autonomous version of the equation (6.57). Namely,
in this case, the matrix A, the bilinear form B, and the function f all depend on
time t. We will consider issues like compact global attractors, convergence, almost
periodic (including periodic and quasi-periodic) solutions and recurrent solutions,
and averaging principles.

6.6.1 Non-autonomous Lorenz systems

Let Ω be a compact metric space, R = (−∞, +∞), (Ω, R, σ) be a dynamical system
on Ω and H be a real or complex Hilbert space. We denote by L(H) (L2 (H))
the space of all linear (bilinear) operators on H. When W is some metric space,
C(Ω, W ) denotes the space of all continuous functions f : Ω → W , endowed with
the topology of uniform convergence.
Let us consider the non-autonomous Lorenz system

u0 = A(ωt)u + B(ωt)(u, u) + f (ωt), ω ∈ Ω, (6.58)

246 Global Attractors of Non-autonomous Dissipative Dynamical Systems

where ωt := σ(t, ω), A ∈ C(Ω, L(H)), B ∈ C(Ω, L2 (H)) and f ∈ C(Ω, H). Note
that when the autonomous Lorenz system (6.57) is perturbed by periodic, quasi-
periodic, almost periodic or recurrent forces, it can then be written as (6.58). More-
over, we assume that the following conditions are fulfilled:

(1) There exists α > 0 such that

RehA(ω)u, ui ≤ −α|u|2 (6.59)

for all ω ∈ Ω and u ∈ H, where | · | is a norm in H ;

(2)

RehB(ω)(u, v), wi = −RehB(ω)(u, w), vi (6.60)

for every u, v, w ∈ H and ω ∈ Ω .

Remark 6.4 a. It follows from (6.60) that

RehB(ω)(u, v), v)i = 0 (6.61)

for every u, v ∈ H and ω ∈ Ω.

b. From bilinearity and continuity, we obtain

|B(ω)(u, v)| ≤ CB |u||v| (6.62)

for all u, v ∈ H and ω ∈ Ω, where CB := sup{|B(ω)(u, v)| : ω ∈ Ω, u, v ∈ H, |u| ≤ 1

and |v| ≤ 1}.

Definition 6.6 We will call the system (6.58) with conditions (6.59) and (6.60)
a non-autonomous Lorenz system or a non-autonomous system of hydrodynamic
type.

We note that from the conditions (6.60) -(6.62) it follows that

|B(ω)(x1 , x1 ) − B(ω)(x2 , x2 )| ≤ CB (|x1 | + |x2 |)|x1 − x2 | (6.63)

for all x1 , x2 ∈ H and ω ∈ Ω.

Since the coefficients of (6.58) are locally Lipschitzian with respect to u ∈ H,
through every point x ∈ H passes a unique solution ϕ(t, x, ω) of equation (6.58) at
the initial moment t = 0. And this solution is defined on some interval [0, t(x,ω) ).
Let us note that

w0 (t) = 2Rehϕ0 (t, x, ω), ϕ(t, x, ω)i = 2RehA(ωt)ϕ(t, x, ω), ϕ(t, x, ω)i
+2RehB(ωt)(ϕ(t, x, ω), ϕ(t, x, ω)), ϕ(t, x, y)i + 2Rehf (ωt), ϕ(t, x, ω)i
= 2RehA(ωt)ϕ(t, x, ω), ϕ(t, x, ω)i + 2Rehf (ωt), ϕ(t, x, ω)i
≤ −2α|ϕ(t, x, ω)|2 + 2kf k|ϕ(t, x, ω)|, (6.64)
 Dissipativity of some classes of equations 247

where kf k := max{|f (ω)| : ω ∈ Ω} and w(t) := |ϕ(t, x, ω)|2 . Then

1
w0 ≤ −2αw + 2kf kw 2 .

Thus

w(t) ≤ v(t)

for all t ∈ [0, t(x,ω) ), where v(t) is a solution of the equation

1
v 0 = −2αv + 2||f ||v 2 ,

satisfying condition v(0) = w(0) = |x|2 . Hence

kf k −αt kf k 2
v(t) = [(|x| − )e + ]
α α
and consequently

kf k −αt kf k
|ϕ(t, x, ω)| ≤ (|x| − )e + (6.65)
α α
for all t ∈ [0, t(x,ω) ). It follows from the inequality (6.65) that the solution ϕ(t, x, ω)
is bounded and therefore it may be extended to a global solution on R+ = [0, +∞).
Thus we have proved the following theorem.

Theorem 6.14 Let the conditions (6.59) and (6.60) are fulfilled. Then the follow-
ing statements hold:

(i)

|ϕ(t, x, ω)| ≤ C(|x|),

kf k
for all t ≥ 0, ω ∈ Ω and x ∈ H, where C(r) = r if r ≥ r0 := α and C(r) = r0
if r ≤ r0 ;
(ii)

||f ||
lim sup sup{|ϕ(t, x, ω)| : |x| ≤ r, ω ∈ Ω} ≤
t→+∞ α

for every r > 0.

The item (i) in this Theorem means that the non-autonomous Lorenz flow is
bounded on bounded sets, while the item (ii) implies that the non-autonomous
Lorenz system is dissipative, i.e., it admits a bounded absorbing set.
248 Global Attractors of Non-autonomous Dissipative Dynamical Systems

6.6.2 Non-autonomous dissipative dynamical systems and their

attractors
Let Ω and W be two metric spaces and (Ω, R, σ) be an autonomous dynamical
system on Ω. Let us consider a continuous mapping ϕ : R+ × W × Ω → W satisfying
the following conditions:

ϕ(0, ·, ω) = idW ϕ(t + τ, x, ω) = ϕ(t, ϕ(τ, x, ω), ωτ )

for all t, τ ∈ R+ , ω ∈ Ω and x ∈ W . Here ωτ is the short notation for στ (ω) :=

σ(τ, ω). Such a mapping ϕ ( or more explicitly hW, ϕ, (Ω, R, σ)i) is called a cocycle
over (Ω, R, σ) with fiber W (see [6, 292]).

Example 6.2 Let E be a Banach space and C(R × E, E) be a space of all con-
tinuous functions F : R × E → E equipped by the compact–open topology. Let us
consider a parameterized differential equation
dx
= F (σt ω, x), ω ∈ Ω
dt
on a Banach space E with Ω := C(R × E, E), where σt ω := σ(t, ω). We will define
σt : Ω → Ω by σt ω(·, ·) = ω(t + ·, ·) for each t ∈ R and interpret ϕ(t, x, ω) as solution
of the initial value problem
d
x(t) = F (σt ω, x(t)), x(0) = x. (6.66)
dt
Under appropriate assumptions on F : Ω × E → E (or even F : R ×E → E with
ω(t) instead of σt ω in (6.66)) to ensure forward existence and uniqueness, then ϕ
is a cocycle on (C(R × E, E), R, σ) with fiber E. Note that (C(R × E, E), R, σ) is
a Bebutov’s dynamical system (see for example [32],[102],[292],[300].

Let ϕ be a cocycle on (Ω, R, σ) with the fiber E. Then the mapping π : R+ ×E×Ω
→ E × Ω defined by

π(t, x, ω) := (ϕ(t, x, ω), σt ω)

for all t ∈ R+ and (x, ω)∈ E × Ω forms a semi-group {π(t, ·, ·)}t∈R+ of mappings of
X := Ω × E into itself, thus a semi-dynamical system on the state space X, which
is called a skew-product flow [292]. The triplet h(X, R+ , π), (Ω, R, σ), hi (where
h := pr2 : X → Ω) is a non-autonomous dynamical system (see [33, 102]).

Definition 6.7 Recall that a cocycle ϕ over (Ω, R, σ) with the fiber W is called a
compact (bounded) dissipative cocycle, if there is a nonempty compact set K ⊆ W
such that

lim sup{β(U (t, ω)M, K)|ω ∈ Ω} = 0

t→+∞
 Dissipativity of some classes of equations 249

for any M ∈ C(W ) (respectively M ∈ B(W )), where C(W )( B(W )) denotes the
family of all compact (bounded) subsets of W , β is the semi-distance of Hausdorff
and U (t, ω) := ϕ(t, ·, ω). We can similarly define a compact or bounded dissipative
skew-product system.

Lemma 6.10 Let Ω be a compact metric space and hW, ϕ, (Ω, R, σ)i be a cocy-
cle over (Ω, R, σ) with the fiber W . In order for hW, ϕ, (Ω, R, σ)i to be compact
(bounded) dissipative, it is necessary and sufficient that the skew-product dynamical
system (X, R+ , π) is compact (bounded) dissipative.

Proof. This assertion follows from the corresponding definitions. 

Definition 6.8 The cocycle hW, ϕ, (Y, R, σi is called compact (asymptotically

compact) if the associated skew-product dynamical system (X, R+ , π) with X =
W × Y and π = (ϕ, σ) is compact (respectively asymptotically compact).

Let (X, R+ , π) be compact dissipative and K be a compact set, which attracts

all compact subsets of X. Let

J = Ω(K), (6.67)
T S
where Ω(K) = t≥0 τ ≥t π τ K. The set J defined by the equality (6.67) does not
depend on selection of the attracting set K, and is characterized only by the prop-
erties of the dynamical system (X, R+ , π) itself. The set J is called the Levinson’s
centre of the compact dissipative system (X, R+ , π).

Remark 6.5 Let hW, ϕ, (Ω, R, σi be a compact dissipative cocycle over (Ω, R, σi
with the fiber W and {Iω | ω ∈ Ω} is its compact global attarctor. Then the skew-
product dynamical system (X, R+ , π), generated by cocycle ϕ, is compact dissipative
too and the set J := ∪{Jω | ω ∈ Ω}, where Jω := Iω × {ω}, is the Levinson center
of (X, R+ , π).

Applying the general theorems about non-autonomous dissipative systems to

non-autonomous system constructed in the example 6.2, we will obtain series of
facts concerning the non-autonomous Lorenz system (6.58). In particular we have
the following results.

Theorem 6.15 Let Ω be a compact metric space, (Ω, R, σ) be a dynamical system

on Ω and the conditions (6.59) and (6.60) are fulfilled. If the cocycle ϕ generated
by non-autonomous Lorenz system (6.58) is asymptotically compact, then for every
ω ∈ Ω, there exists a non-empty compact connected set Iω ⊂ H such that the
following conditions hold:
250 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(1) The set I = ∪{Iω : ω ∈ Ω} is compact and connected in H, if the space Ω is

connected;
(2)

lim sup β(U (t, ω −t )M, I) = 0

t→+∞ ω∈Ω

for any bounded set M ⊂ H, where U (t, ω) := ϕ(t, ·, ω) and β is the semi-
distance of Hausdorff;
(3) U (t, ω)Iω = Iωt for all t ∈ R+ and ω ∈ Ω;
(4) Iω consists of those and only those points x ∈ H through which passes the
bounded solutions (on R) of the non-autonomous Lorenz system (6.58).

Proof. This assertion follows from Theorems 6.14, 2.24 and 2.25. 
This theorem states that ∪{Iω : ω ∈ Ω} is the compact global attractor of the
non-autonomous Lorenz system (6.58) and also characterizes the structure of the
sections Iω of the attractor.

Theorem 6.16 Under conditions of Theorem 6.15

kf k
|ϕ(t, x, ω)| ≤ (6.68)
α
for all t ∈ R, ω ∈ Ω and x ∈ Iω , where ϕ is the cocycle generated by Lorenz
non-autonomous system (6.58).
S
Proof. According to Remark 6.5 the set J := {Iω × {ω} : ω ∈ Ω} is a Levin-
son’s centre of dynamical system (X, R+ , π) and according to (6.67) for any point
(u0 , y0 ) = z ∈ J there exists tn → +∞, un ∈ H and ωn ∈ Ω such that the sequence
{un } is bounded, u0 = lim ϕ(tn , un , ωn ) and ω0 = lim ωn tn . From the inequal-
n→+∞ n→+∞
kf k
ity (6.65), it follows that |u0 | ≤ α , i.e. ϕ(t, x, ω) ∈ Iωt for all ω ∈ Ω and t ∈ R,
hence |ϕ(t, x, ω)| ≤ kfαk for any t ∈ R, x ∈ Iω and ω ∈ Ω. The theorem is proved.

Theorem 6.17 Let ϕ be the cocycle generated by the Lorenz non-autonomous

system (6.58). Under conditions of Theorem 6.15 and further assume that
α−2 CB kf k < 1. Then this cocycle ϕ is convergent, i.e. for any ω ∈ Ω the set
Iω contains a single point uω .

Proof. Let ω ∈ Ω and u1 , u2 ∈ Iω . We define ψ(t) := ϕ(t, u1 , ω) − ϕ(t, u2 , ω) and

w(t) := |ϕ(t, u1 , ω) − ϕ(t, u2 , ω)|2 .

According to Theorem 6.16, the function w(t) is bounded on R. On the other hand,
in view of (6.64) and (6.59), we have

w0 (t) ≤ −2αw(t) + 2RehB(ωt)(ψ(t), ϕ(t, u2 , ω)), ψ(t)i. (6.69)

 Dissipativity of some classes of equations 251

From the inequalities (6.63), (6.69) and Theorem 6.16, it follows that

kf k
w0 (t) ≤ −2αw(t) + 2CB w(t).
α
kf k
Hence, w(t) ≤ w(0)e−2(α−CB α )t , i.e.
kf k
|ϕ(t, u1 , ω) − ϕ(t, u2 , ω)| ≤ |u1 − u2 |e−(α− α CB )t

for all t ≥ 0, ω ∈ Ω and u1 , u2 ∈ Iω . In particular,

|u1 − u2 | ≤ |ϕ(t, ϕ(−t, u1 , σ(−t, ω)), ω) − (6.70)

−(α− kf k
α CB )t
ϕ(t, ϕ(−t, u2 , σ(−t, ω)), ω)|e

for all t ≥ 0, ω ∈ Ω and u1 , u2 ∈ Jω . Note that |ϕ(t, u1 , ω) − ϕ(t, u2 , ω)| is bounded

on R. Thus from (6.70) it follows that u1 = u2 , where ϕ(−t, x, ω) := uσ(−t,ω) for
all x ∈ Iω , t ≥ 0 and ω ∈ Ω. The theorem is proved. 

6.6.3 Almost periodic and recurrent solutions of non-autonomous

Lorenz systems
In this subsection, we discuss the problem of existence of almost periodic and recur-
rent solutions of non-autonomous Lorenz systems. Let T = R or R+ and (X, T, π)
be a dynamical system.

Definition 6.9 The solution ϕ(t, x, ω) of non-autonomous Lorenz system (6.58)

is called recurrent (almost periodic, quasi-periodic, periodic), if the point (x, ω) ∈
H×Ω is a recurrent (almost periodic, quasi-periodic, periodic) point of skew-product
dynamical system (X, R+ , π) (X := H × Ω and π := (ϕ, σ)).

We note (see, for example, [238]),[300] and [302]) that if ω ∈ Ω is a stationary

(τ -periodic, almost periodic, quasi periodic, recurrent) point of dynamical system
(Ω, R, σ) and h : Ω → X is a homomorphism of dynamical system (Ω, R, σ) onto
(X, R+ , π), then the point x = h(ω) is a stationary (τ -periodic, almost periodic,
quasi periodic, recurrent) point of the system (X, R+ , π).
Let X := H × Ω and π := (ϕ, σ), then mapping h : Ω → X is a homomorphism
of dynamical system (Ω, R, σ) onto (X, R+ , π) if and only if h(ω) = (u(ω), ω) for
all ω ∈ Ω, where u : Ω → H is a continuous mapping with the condition that
u(ωt) = ϕ(t, u(ω), ω) for all ω ∈ Ω and t ∈ R+ .

Theorem 6.18 Let Ω be a compact metric space, the cocycle ϕ, generated by the
non-autonomous Lorenz system (6.58), is asymptotic compact and the conditions
(6.59), (6.61)-(6.62) are fulfilled with kfαkC
2
B
< 1. Then the set Iω contains a unique
252 Global Attractors of Non-autonomous Dissipative Dynamical Systems

point xω (Iω = {xω }) for every ω ∈ Ω, the mapping u : Ω → H defined by u(ω) :=

xω is continuous and u(ωt) = ϕ(t, u(ω), ω) for all ω ∈ Ω and t ∈ R+ .

Proof. According to Theorem 6.17, it is sufficient to show that the mapping u : Ω →

H defined above is continuous. Let ω ∈ Ω, {ωn } ⊆ Ω and ωn → ω. Consider the
S
sequence {xn } := {xωn } ⊂ I := {Iω | ω ∈ Ω}. Since the set I is compact, then the
sequence {xn } is relatively compact. Let x0 be a limit point of this sequence, then
there is a subsequence {xkn } such that xkn → x0 . Let J be a Levinson center of the
skew-product dynamical system (X, R+ , π), generated by the cocycle ϕ. Note that
the point (xkn , ωkn ) ∈ Jωkn := Iωkn × {ωkn } ⊆ J and taking in the consideration
that J is compact we obtain that (x0 , ω) ∈ J. Thus (x0 , ω) ∈ Jω = Iω × {ω} and,
consequently, x0 ∈ Iω = {xω }, i.e. the relatively compact sequence {xn } has a
unique limit point xω . This means that the sequence {xn } converges to xω as
n → +∞. The theorem is proved. 

Corollary 6.5 Let Ω be a compact minimal (almost periodic minimal, quasi-

periodic minimal or periodic minimal) set of dynamical system (Ω, R, σ). Then
under the conditions of Theorem 6.18, the non-autonomous Lorenz system (6.58)
admits a compact global attractor I, and for all ω ∈ Ω, the section I ω of the attractor
contains a unique point xω through which passes a recurrent (almost periodic, quasi-
periodic, or periodic) solution of the equation (6.58).

Let H be a d-dimensional complex Euclidean space, i.e. H = Cd . Denote by

HC(Cd × Ω, Cd ) the space of all continuous functions f : Cd × Ω → Cd holomorphic
in z ∈ Cd and equipped with compact-open topology. Consider the differential
equation

dz
= f (z, σt ω), (ω ∈ Ω) (6.71)
dt

where f ∈ HC(Cd × Ω, Cd ). Let ϕ(t, ω, z) be the solution of equation (6.71) passing

through the point z at t = 0 and defined on R+ . The mapping ϕ : R+ ×Ω×Cd → Cd
has the following properties (see, for example, [47] and [186]):

a) ϕ(0, z, ω) = z for all z ∈ Cd .

b) ϕ(t + τ, z, ω) = ϕ(t, ϕ(τ, z, ω), στ ω) for all t, τ ∈ R+ , ω ∈ Ω and z ∈ Cd .
c) Mapping ϕ is continuous.
d) Mapping U (t, ω) := ϕ(t, ·, ω) : Cd → Cd is holomorphic for any t ∈ R+ and
ω ∈ Ω.

Definition 6.10 The cocycle hCd , ϕ, (Ω, T, θ)i is said to be C-analytic if the
mapping U (t, ω) : Cd → Cd is holomorphic for all t ∈ R+ and ω ∈ Ω.
 Dissipativity of some classes of equations 253

Example 6.3 Let (HC(R × Cd , Cd ), R, σ) be a dynamical system of translations

on HC(R×Cd , Cd ) (Bebutov’s dynamical system (see, for example, [102] and [300])).
Denote by F the mapping from Cd × HC(R × Cd , Cd ) to Cd defined by equality
F (z, f ) := f (0, z) for all z ∈ Cd and f ∈ HC(R × Cd , Cd ). Let Ω be the hull
H(f ) of given function f ∈ HC(R × Cd , Cd ), that is Ω = H(f ) := {fτ |τ ∈ R},
where fτ (t, z) := f (t + τ, z) for all t, τ ∈ R and z ∈ Cd . Denote the restriction of
(HC(R × Cd , Cd ), R, σ) on Ω by (Ω, R, σ). Then, under appropriate restriction on
the given function f ∈ HC(R × Cd , Cd ), the differential equation dz dt = f (z, t) =
F (z, σt f ) generates a C−analytic cocycle.

Theorem 6.19 Let H := Cd , Ω be a compact minimal set and the conditions

(6.59), (6.61)-(6.62) are fulfilled. Then the non-autonomous Lorenz system admits
a compact global attractor {Iω | ω ∈ Ω} and the set Iω contains a unique point
xω (Iω = {xω }) for every ω ∈ Ω, the mapping u : Ω → H defined by equality
u(ω) := xω is continuous and u(ωt) = ϕ(t, u(ω), ω) for all ω ∈ Ω and t ∈ R+ , where
ϕ is a cocycle generated by the non-autonomous Lorenz system.

Proof. We note that under the conditions of Theorem 6.19 the right-hand side
f (ω, z) := A(ω)z + B(ω)(z, z) + f (ω) is C-analytic because Dz f (ω, z)h = A(ω)h +
B(ω)(h, z) + B(ω)(z, h) for all ω ∈ Ω and z ∈ Cd , where Dz f (ω, z) is a derivative of
function f (ω, z) w.r.t. z ∈ Cd . Now our statement directly results from Theorems
6.15 and 3.2. The proof is complete. 

Corollary 6.6 Let Ω be a compact minimal (almost periodic minimal, quasi-

periodic minimal or periodic minimal) set of dynamical system (Ω, R, σ). Then
under the conditions of Theorem 6.19, the non-autonomous Lorenz system (6.58)
admits a compact global attractor I and for all ω ∈ Ω, the set I ω contains a unique
point xω through which passes a recurrent (almost periodic, quasi-periodic or peri-
odic) solution of equation (6.58).

6.6.4 Uniform averaging principle

Now we consider a uniform averaging principle for a general class of differential
equations. In the next subsection, we apply this averaging principle to the non-
autonomous Lorenz system (6.58).
Let C(R × H, H) be the space of all continuous functions f : R × H → H
equipped with compact open topology and let F ⊆ C(R × H, H). In Hilbert space
H (with the norm | · | induced by the scalar product) we will consider the family of
equations

x0 = εf (t, x), f ∈ F, (6.72)

254 Global Attractors of Non-autonomous Dissipative Dynamical Systems

containing a small parameter ε ∈ [0, ε0 ] (εo > 0).

We assume that on the set R+ × B[0, r], where B[0, r] := {x ∈ H | |x| ≤ r} is
a ball of radius r > 0 in H, the functions f ∈ F are uniformly bounded, i.e. there
exists a positive constant M such that

|f (t, x)| ≤ M (6.73)

for every f ∈ F, t ∈ R+ and x ∈ B[0, r], and satisfies the condition of Lipschitz

|f (t, x1 ) − f (t, x2 )| ≤ L|x1 − x2 | (x1 , x2 ∈ B[0, r]) (6.74)

with a constant L > 0 does not depending neither on t ∈ R+ nor on f ∈ F.

Furthermore, we assume that the mean value of f is uniform with respect to
(w.r.t.) f ∈ F and x ∈ B[0, r]
Z
1 T
f0 (x) := lim f (t, x)dt, (6.75)
T →+∞ T 0

i.e. for every ε > 0 there exists a l = l(ε) > 0 such that
Z
1 T
| f (t, x)dt − f0 (x)| < ε
T 0
for all T ≥ l(ε), x ∈ B[0, r] and f ∈ F, and the function f0 does not depend on
f ∈ F.

Lemma 6.11 The condition (6.75) holds if and only if there exists a decreasing
continuous function m : R+ → R+ , satisfying the condition m(t) → 0 as t → +∞,
such that
Z
1 T
| f (t, x)dt − f0 (x)| ≤ m(T ) (6.76)
T 0
for all T > 0, x ∈ B[0, r] and f ∈ F. The function m does not depend neither on
x ∈ B[0, r] nor on f ∈ F.

Proof. Denote by
Z T
1
k(T ) := sup | f (t, x)dt − f0 (x)|.
f ∈F ,x∈B[0,r] T 0

The mapping k possesses the following properties:

(1) 0 ≤ k(T ) ≤ 2M, where M := sup{|f (t, x)| : f ∈ F, |x| ≤ r};

(2) k(T ) → 0 as T → +∞.

Let

cn := sup k(T ),
T ≥n
 Dissipativity of some classes of equations 255

then co ≥ c1 ≥ ... ≥ cn ≥ ... and cn → 0 as n → +∞. Define now the function

m : R+ → R+ by the equality

m(t) := cn−1 + (t − n)(cn − cn−1 ) (n ≤ t ≤ n + 1, n = 0, 1, ...),

where c−1 := c0 + 1. The lemma is proved. 

Lemma 6.12 Let F ⊆ C(R × E, E) be a family of functions satisfying the con-

dition (6.75), then for every L > 0
Z τ
t
l(ε) := sup{| f ( , x)dt − f0 (x)| : 0 ≤ τ ≤ L, f ∈ F, |x| ≤ r} → 0
0 ε
as ε → 0.

Proof. According to Lemma 6.11 there exists a decreasing continuous function

m : R+ → R+ with the condition m(t) → 0 as t → +∞ and such that the inequality
(6.76) holds. Let ν ∈ (0, 1), then
Z τ
t
l(ε) ≤ sup{| f ( , x)dt| : 0 ≤ τ ≤ εν , f ∈ F, |x| ≤ r} +
0 ε
Z τ
t
sup{| f ( , x)dt| : εν ≤ τ ≤ L, f ∈ F, |x| ≤ r} =
0 ε
Z τε
ε
sup{τ | f (t, x)dt| : 0 ≤ τ ≤ εν , f ∈ F, |x| ≤ r} +
τ 0
Z τ
ε ε
sup{τ | f (t, x)dt| : εν ≤ τ ≤ L, f ∈ F, |x| ≤ r} ≤
τ 0
m(0)εν + Lm(εν−1 ) → 0

as ε → 0. The lemma is proved. 

Under the assumptions above, it is expedient to consider along with equation
(6.72) the averaged equation

x0 = εfo (x).

From (6.75) we see that the function f0 also satisfies the conditions (6.73) and
(6.74). Let ϕ(t, x) (0 ≤ t ≤ T0 ) be a solution of equation

y 0 = f0 (y). (6.77)

taking values in B[0, r] and passing through the point x at the initial moment
t = 0. Then, as can easily be seen, the function ϕ(t, x, ε) := ϕ(εt, x) is the solution
of equation (6.77) on the interval 0 ≤ t ≤ Tε0 . We will establish below a connection
between ϕ(t, x, ε) and the solution ϕ(t, x, f, ε) of equation (6.72) with the initial
condition ϕ(0, x, f, ε) = x.
More precisely, we will prove the following assertion.
256 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Theorem 6.20 Suppose that on R+ × B[0, r], the functions f ∈ F satisfy the
conditions (6.73)-(6.75). Then for any η > 0 there exists an ε > 0 (0 < ε < ε 0 )
such that the estimate
T0
|ϕ(t, x, f, ε) − ϕ(t, x, ε)| ≤ η (0 ≤ t ≤ )
ε
holds uniformly w.r.t. f ∈ F and x ∈ B[0, r].

Denote by K the family of all solutions x : [0, T0 ] → B[0, r] of the equation

(6.77). Let us prove an auxiliary assertion.

Lemma 6.13 Let F ⊆ C(R × H, H) be a family of functions satisfying the con-

ditions (6.73)-(6.75). Then the equality
Z τ Z τ
s
lim f ( , x(s))ds = f0 (x(s))ds (0 < τ ≤ T0 )
ε→0 0 ε 0

holds uniformly w.r.t. x ∈ K, τ ∈ [0, T0 ] and f ∈ F.

Proof. Observe that

Z τ
τ
lim f ( , x)dτ = τ f0 (x) (6.78)
ε→0 0 ε
or, equivalently,
Z τ
ε ε
lim f (t, x)dt = f0 (x) (6.79)
ε→0 τ 0

uniformly w.r.t. x ∈ K, τ ∈ [0, T0 ] and f ∈ F. In fact, according to Lemma 6.12

Z ε
ε τ
| f (t, x)dt − f0 (x)| → 0
τ 0
as ε → 0 uniformly w.r.t. x ∈ B[0, r], f ∈ F and τ ∈ [0, T0 ]. Let us note that
the equality (6.79) is equivalent to (6.75). From (6.78) it follows that for any
τ1 , τ2 ∈ [0, T0 ] we have
Z τ2 Z τ2
τ
lim f ( , x)dτ = f0 (x)dτ
ε→0 τ
1
ε τ1

uniformly w.r.t. x ∈ B[0, r] and f ∈ F. Hence for any 0 ≤ τ1 < τ2 < ...τn−1 < τn =
T0 , xk ∈ B[0, r] (k = 1, 2, ..., n), we conclude that
Xn Z τk n Z τk
X
lim f (τ, xk , ε)dτ = f0 (xk )dτ (6.80)
ε→0 τk−1 τk−1
1 1

uniformly w.r.t. x1 , x2 , ..., xn ∈ B[0, r] and f ∈ F.

 Dissipativity of some classes of equations 257

If we introduce the step functions x̃n (τ ) := x(τk ) (τk−1 ≤ τ ≤ τk ; τk − τk−1 =

1
n; k = 1, 2, ..., n and x ∈ K), then from the equality (6.80), we have the following
relation
Z τ Z τ
s
lim f ( , x̃n (s))ds = f0 (x̃n (s))ds. (6.81)
ε→0 0 ε 0

Under our assumption the family of functions K is equicontinuous on [0, T0 ] and,

consequently,

sup sup kx̃n (τ ) − x(τ )k → 0 (6.82)

x∈K 0≤τ ≤T0

as n → +∞. Using the condition of Lipschitz (6.74) for the family of functions F
we obtain the estimate
Z τ Z τ Z τ
s s s
| f ( , x(s))ds − f0 (x(s))ds| ≤ |f ( , x(s)) − f ( , x˜n (s))|ds+ (6.83)
0 ε 0 0 ε ε
Z τ Z τ
s
| [f ( , x˜n (s)) − f0 (x˜n (s))]ds| + |f0 (x(s)) − f0 (x˜n (s))|ds ≤
0 ε 0

Z τ
s
2LT0 sup sup |x̃n (τ ) − x(τ )| + | [f ( , x˜n (s)) − f0 (x˜n (s))]ds|.
x∈K 0≤τ ≤T0 0 ε
From (6.80) - (6.83) immediately we obtain the results in the lemma. 

Proof. Now we will prove Theorem 6.20. Denote by ψ(τ, x, f, ε) (respectively

ψ̄(τ, x)) a unique solution of equation
τ
x0 = f ( , x)
ε
(respectively (6.77)) passing through the point x ∈ B[0, r] at the moment τ = 0
and defined on [0, Tε0 ]. The functions ψ(τ, x, f, ε) and ψ̄(τ, x) satisfy the integral
equations
Z τ
s
ψ(τ, x, f, ε) = x + f ( , ψ(s, x, f, ε))ds
0 ε
and
Z τ
ψ̄(τ, x) = x + f0 (ψ̄(s, x))ds,
0

respectively. Using the condition of Lipschitz (6.74), we obtain the estimate

Z τ
s
|ψ(τ, x, f, ε) − ψ̄(τ, x)| = | [f ( , ψ(s, x, f, ε)) − f0 (ψ̄(s, x))]ds| ≤
0 ε
Z τ Z τ
s s s
|f ( , ψ(s, x, f, ε)) − f ( , ψ̄(s, x))|ds + | [f ( , ψ̄(s, x)) − f0 (ψ̄(s, x))]ds| ≤
0 ε ε 0 ε
258 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Z τ
L |ψ(s, x, f, ε) − ψ̄(s, x)|ds + c(ε),
0

where
Z τ
s
c(ε) := sup | [f0 , x(s)) − f0 (x(s))]ds|.
0≤τ ≤T0 ,x∈K 0 ε

According to the Gronwall-Bellman inequality (see, for example,[137] or [186]), we

can now conclude that

|ψ(τ, x, f, ε) − ψ̄(τ, x)| ≤ exp(2Lτ )c(ε)

and it remains only to note that by virtue of Lemma 6.13, c(ε) → 0 as ε → 0 and

|x(t, ε) − y(εt)| = |ψ(τ, x, f, ε) − ψ̄(τ, x)| ≤ exp(2Lτ )c(ε) = exp(2Lεt))c(ε)

for all t ∈ [0, Tε0 ]. The theorem is thus proved. 

In the next subsection, we will also need the following lemma.

Lemma 6.14 Let F be a transitive subset of C(R × H, H), i.e. there exists a
function g ∈ F such that F = H(g), where H(f ) if the hull of g. Then the following
two assertions are equivalent:

1. There exists f0 ∈ C(H, H) such that

Z
1 T
lim f (t, x)dt = f0 (x)
T →+∞ T 0

uniformly w.r.t. f ∈ F and x ∈ B[0, r];

2. There exists f0 ∈ C(H, H) such that
Z
1 t+T
lim g(τ, x)dτ = f0 (x)
T →+∞ T t

uniformly w.r.t. t ∈ R and x ∈ B[0, r].

Proof. It is evident that 1. implies 2. because gt ∈ F for all t ∈ R and, consequently,

Z Z
1 t+T 1 T
g(τ, x)dτ = g(t + τ, x)dτ → f0 (x)
T t T 0
as T → +∞ uniformly w.r.t. t ∈ R and x ∈ B[0, r].
Let now ε > 0 and f ∈ F = H(g), then there exists a sequence {tn } ⊂ R and
L(ε) > 0 such that gtn → f and
Z
1 T
| g(τ + tn , x)dτ − f0 (x)| < ε (6.84)
T 0
 Dissipativity of some classes of equations 259

for all T > L(ε). Passing to limit as n → +∞ in the inequality (6.84) we obtain
Z T
1
| f (τ, x)dτ − f0 (x)| ≤ ε
T 0

for all T > L(ε). From the latter inequality, the required statement immediately
follows. This proves the lemma. 

Remark 6.6 All the results of this subsection are true for arbitrary Banach space
too, not only for Hilbert space.

6.6.5 Global averaging principle for the non-autonomous Lorenz

systems
Now we consider a global averaging principle for the non-autonomous Lorenz
systems. Let Ω be a compact metric space and (Ω, R, σ) be a dynamical system on
Ω. We consider the ”perturbed” non-autonomous Lorenz equation
dx
= εA(ωt)x + εB(ωt)(x, x) + εf (ωt), (6.85)
dt
where ε ∈ [0, ε0 ] (ε0 > 0) is a small parameter. Suppose that the conditions (6.59)–
(6.62) are fulfilled and the following averaging values exist uniformly w.r.t. ω ∈ Ω:

ZT
1
A = lim A(ωt)dt, (6.86)
T →+∞ 2T
−T

ZT
1
B = lim B(ωt)dt, (6.87)
T →+∞ 2T
−T

and
ZT
1
f = lim f (ωt)dt. (6.88)
T →+∞ 2T
−T

Remark 6.7 The conditions (6.86) − (6.88) are fulfilled if a dynamical system
(Ω, R, σ) is strictly ergodic, i.e. there exists on Ω a unique invariant w.r.t. (Ω, R, σ)
measure µ.

Along with equation (6.85), we will also consider the averaged equation
dx
= εAx + εB(x, x) + εf. (6.89)
dt
260 Global Attractors of Non-autonomous Dissipative Dynamical Systems

If we introduce the ”slow time” τ := εt (ε > 0), then the equations (6.85) and (6.89)
can be written as
dx τ τ τ
= A(ω )x + B(ω )(x, x) + f (ω ) (6.90)
dτ ε ε ε
and
dx
= Ax + B(x, x) + f . (6.91)
dτ
Remark 6.8 a. From the conditions (6.61) and (6.87) it follows that

RehB(u, v), vi = 0 (6.92)

for all u, v ∈ H;
b. From the inequality (6.59) it follows that

RehAx, x)i ≤ −α|x|2 (6.93)

for all x ∈ H.

Theorem 6.21 Assume the conditions enumerated above are all satisfied. Then
for all T > 0 and ρ ≥ r0 := kfαk > 0, the solution for the non-autonomous Lorenz
equation (6.85) approaches the solution of the averaged Lorenz equation (6.89) in
the following sense:

max{|ϕ(t, x, ω, ε) − ϕ(t, x, ε)| : 0 ≤ t ≤ T /ε, |x| ≤ ρ, and ω ∈ Ω} → 0 (6.94)

as ε → 0, where ϕ(t, x, ω, ε) ( respectively ϕ(t, x, ε)) is a solution of equation (6.85)

(respectively (6.89)), passing through the point x at the initial moment t = 0.

Proof. According to Theorem 6.14, we have |ϕ(t, x, ω, ε)| ≤ ρ and |ϕ̄(t, x, ε)| ≤ ρ
for all t ≥ 0, |x| ≤ ρ, ω ∈ Ω and ε ∈ (0, ε0 ]. If we take F := {fω | ω ∈ Ω} ⊂
C(R × H, H), where fω (t, x) := A(ωt)x + B(ωt)(x, x) + f (ωt) for all t ∈ R and
x ∈ H, then the relation (6.94) follows from Theorem 6.20. This completes the
proof. 

Theorem 6.22 (Global averaging principle for non-autonomous Lorenz systems)

Let ϕε be a cocycle generated by the equation (6.85). Assume the conditions enumer-
ated above are all satisfied. If the cocycle ϕε ( ε ∈ [0, ε0 ]) is asymptotically compact,
then the following assertions hold:

(1) The averaged equation (6.91) admits a compact global attractor I ⊂ H;

(2) For every ε ∈ (0, ε0 ] the equation (6.85) has a compact global attractor {Iωε | ω ∈
Ω};
(3) The set I := ∪{I ε | ε ∈ [0, ε0 ]} is bounded, where I 0 := I and I ε := ∪{Iωε | ω ∈
Ω};
 Dissipativity of some classes of equations 261

(4)

lim sup β(Iωε , I) = 0

ε→0 ω∈Ω

and, in particular,

lim β(I ε , I) = 0.
ε→0

Proof. The first three statements of the theorem follow from Theorems 6.14, 6.15
and Remark 6.5. Now we will prove the fourth statement of the theorem. To this
end, we will use the same arguments as in [70, 199]. Let λ > 0 and B(I, λ) = {x ∈
H | ρ(x, I) < λ}. According to orbital stability of the set I (see, for example, [179,
Ch.I] or Theorem 1.6), for given λ there exists δ = δ(λ) > 0 (we may consider
δ(λ) < λ/2) such that

ϕ(t, B(I, δ)) ⊂ B(I, λ/2)

for all t ≥ 0. By virtue of boundedness of the set I = ∪{I ε | 0 ≤ ε ≤ ε0 } we may

choose ρ ≤ r0 such that I ⊂ B(0, ρ) = {x ∈ H | |x| < ρ}. Since I is a compact global
attractor of the system (6.91), then for the closed ball B[0, ρ] := {x ∈ H| |x| ≤ ρ}
and the number δ > 0 there exists T = T (ρ, δ) > 0 such that

ϕ(t, B[0, ρ]) ⊂ B(I, δ/2), t ≥ T. (6.95)

Let x ∈ B[0, ρ]. Then by virtue of Theorem 6.21 for the numbers ρ ≥ r0 and
T (ρ, δ) > 0 there exists µ = µ(ρ, δ) > 0 such that 0 < ε ≤ µ, m(ε) < λ/2 (see
(6.94)), i.e.

|ϕ(t, x, ω, ε) − ϕ(t, x)| < δ/2 (6.96)

for all x ∈ B[0, ρ], ω ∈ Ω, t ∈ [0, T /ε] and 0 < ε ≤ µ. According to (6.95)
we have ϕ(T /ε, x, ω, ε) ∈ B(I, δ/2). Thus, taking into account (6.96), we obtain
ϕ(T /ε, x, ω, ε) ∈ B(I, δ). Let us take the initial point x1 := ϕ(T /ε, x, ω, ε) and we
will repeat for this point the same reasoning as above. Taking into consideration
the equality ϕ(t, x1 , σ(T /ε, ω), ε) = ϕ(t + T /ε, x, ω, ε), we will have

|ϕ(t + T /ε, x, ω, ε) − ϕ(t, x1 )| < δ/2 (6.97)

for all t ∈ [0, T /ε], x ∈ B[0, ρ] and ω ∈ Ω, where x1 = ϕ(T /ε, x, ω, ε).
By the inequality (6.97) we obtain again x2 := ϕ(2T /ε, x, ω, ε) ∈ B(I, δ) and,
consequently,

ϕ(t + T /ε, x, ω, ε) ∈ B(I, λ/2 + δ/2) ⊂ B(I, λ).

262 Global Attractors of Non-autonomous Dissipative Dynamical Systems

If we continue this process and later (by virtue of uniformity w.r.t. |x| ≤ ρ and
ω ∈ Ω of the estimation (6.96) it is possible), we will obtain

ϕ(t, x, ω, ε) ∈ B(I, λ) (6.98)

for all t ≥ T /ε, x ∈ B[0, ρ], ω ∈ Ω and o ≤ ε ≤ µ and, consequently,

ϕ(t, x, σ(−t, ω), ε) ∈ B(I, λ)

for all t ≥ T /ε and |x| ≤ ρ. Since I = ∪{I ε | 0 ≤ ε ≤ ε0 } ⊆ B(0, ρ), then according
to Theorem 2.24
\ [
Iωε = ϕ(τ, B[0, ρ], σ(−τ, ω), ε).
t≥0 τ ≥t

Therefore, from (6.98) we have Iωε ⊂ B(I, λ) for all ω ∈ Ω and 0 < ε < µ. Note
that λ is arbitrarily chosen. Hence from the last inclusion we obtain the equality
(6.95). The theorem is proved. 
 Chapter 7

Upper semi-continuity of attractors

7.1 Introduction

The problem of upper semi-continuity of global attractors for small perturbations

is well studied (see, for example, [179] and references therein) for autonomous and
periodical dynamical systems. In the works [40] and [41] this problem was studied
for non-autonomous and random dynamical systems.
This chapter is devoted to a systematic study of the problem of upper semi-
continuity of compact global attractors and compact pullback attractors of abstract
non-autonomous dynamical systems for small perturbations. Several applications
of our results are given for different classes of evolutionary equations.
The chapter is organized as follows. In section 2 we study some general prop-
erties of maximal compact invariant sets of dynamical systems. In particular, we
prove that the compact global attractor and pullback attractor are maximal com-
pact invariant sets (Theorem 7.2).
Section 3 contains the main results about upper semi-continuity of compact
global attractors of abstract non-autonomous dynamical systems for small pertur-
bations (Lemmas 7.3, 7.4 and Theorems 7.3, 7.4, 7.5 and 7.6).
In section 4 we give conditions for connectedness and component connectedness
of global and pullback attractors (Theorem 7.7).
Section 5 is devoted to an application of our general results obtained in sections
2-4, to the study of different classes of non-autonomous differential equations (quasi-
homogeneous systems, monotone systems, non-autonomously perturbed systems,
non-autonomous 2D Navier-Stokes equations and quasi-linear functional-differential
equations).

7.2 Maximal compact invariant sets

Let W be a complete metric space, T = R or Z, Ω be a compact metric space,

(Ω, T, σ) be a group dynamical system on Ω and hW, ϕ, (Ω, T, σ)i be a cocycle with

263
264 Global Attractors of Non-autonomous Dissipative Dynamical Systems

fiber W , i.e. the mapping ϕ : T+ × W × Ω → W is continuous and possesses the

following properties: ϕ(0, x, ω) = x and ϕ(t + τ, x, ω) = ϕ(t, ϕ(τ, x, ω), ωt), where
ωt = σ(t, ω).
We denote by X = W × Ω, g = pr1 : X 7→ W, (X, T+ , π) a semi-group dynamical
system on X defined by the equality π = (ϕ, σ), i.e. π t x = (ϕ(t, u, ω), σ(t, ω)) for
every t ∈ T+ and x = (u, ω) ∈ X = W × Ω. Let h(X, T+ , π), (Ω, T, σ), hi be a
non-autonomous dynamical system, where h = pr2 : X 7→ Ω.

Definition 7.1 A family {Iω | ω ∈ Ω}(Iω ⊂ W ) of nonempty compact subsets of

W is called a maximal compact invariant set of cocycle ϕ, if the following conditions
are fulfilled:

(1) {Iω | ω ∈ Ω} is invariant, i.e. ϕ(t, Iω , ω) = Iωt for every ω ∈ Ω and t ∈ T+ ;

S
(2) I = {Iω | ω ∈ Ω} is relatively compact;
0
(3) {Iω | ω ∈ Ω} is maximal, i.e. if the family {Iω | ω ∈ Ω} is relatively compact
0
and invariant, then Iω ⊆ Iω for every ω ∈ Ω.

Lemma 7.1 The family {Iω | ω ∈ Ω} is invariant w.r.t. cocycle ϕ if and only
S
if the set J = {Jω | ω ∈ Ω}(Jω := Iω × {ω}) is invariant with respect to the
skew-product dynamical system (X, T+ , π).
S
Proof. Let the family {Iω | ω ∈ Ω} be invariant, J = {Jω |ω ∈ Ω} and Jω =
Iω × {ω}. Then
[ [
π t J = {π t Jω | ω ∈ Ω} = {(ϕ(t, Iω , ω), ωt) | ω ∈ Ω} (7.1)
[ [
= {Iωt × {ωt} | ω ∈ Ω} = {Jωt | ω ∈ Ω} = J

for all t ∈ T+ . From the equality (7.1) follows that the family {Iω | ω ∈ Ω} is
invariant w.r.t. cocycle ϕ if and only if a set J is invariant w.r.t. dynamical system
(X, T+ , π). 

Theorem 7.1 Let the family of sets {Iω | ω ∈ Ω} be maximal, compact and
invariant. Then it is closed.
S
Proof. We note that the set J = {Jω | ω ∈ Ω}(Jω := Iω × {ω}) is relatively
compact and according to Lemma 7.1 it is invariant. Let K = J, then K is compact.
We will show that K is invariant. If x ∈ K, then there exists {xn } ⊂ J such that
x = lim xn . Thus xn ∈ J = π t J for all t ∈ T+ , then for t ∈ T+ there exists
n→+∞
xn ∈ J such that xn = π t xn . Since J is relatively compact, then we can suppose that
the sequence {xn } is convergent. We denote by x = lim xn ,then x ∈ J, x = π t x
n→+∞
and, consequently, x ∈ π t J for all t ∈ T+ , i.e. J = π t J. Let I 0 = pr1 K, then we
S
have I 0 = {Iω0 | ω ∈ Ω}, where Iω0 = {u ∈ W | (u, ω) ∈ K} and Kω := Iω0 × {ω}.
 Upper semi-continuity of attractors 265

Since the set K is invariant, then according to Lemma 7.1 the set I 0 is also invariant
w.r.t. cocycle ϕ. The set I 0 is compact, because K is compact and pr1 : X 7→ W
is continuous. According to the maximality of the family {Iω | ω ∈ Ω} we have
Iω0 ⊆ Iω for every ω ∈ Ω and, consequently, I 0 ⊆ I. On the other hand I = pr1 J = I 0
and, consequently, I 0 = I. Thus the set I is compact. The theorem is proved. 

Denote by C(W ) the family of all compact subsets of W .

Definition 7.2 Recall that a family {Iω | ω ∈ Ω} (Iω ⊂ W ) of nonempty compact

subsets of W is called

− a compact pullback attractor of the cocycle ϕ, if the following conditions are

fulfilled:
S
a. I = {Iω | ω ∈ Ω} is relatively compact ;
b. I is invariant w.r.t. cocycle ϕ, i.e. ϕ(t, Iω , ω) = Iσ(t,ω) for all t ∈ T+ and
ω ∈ Ω;
c. for every ω ∈ Ω and K ∈ C(W )

lim β(ϕ(t, K, ω −t ), Iω ) = 0, (7.2)

t→+∞

where β(A, B) = sup{ρ(a, B) : a ∈ A} is a semi-distance of Hausdorff and

ω −t := σ(−t, ω).
− a compact global attractor, if the following conditions are fulfilled:
d. a family {Iω | ω ∈ Ω} is compact and invariant;
f. for every K ∈ C(W )

lim sup β(ϕ(t, K, ω), I) = 0, (7.3)

t→+∞ ω∈Ω
S
where I = {Iω | ω ∈ Ω}.

Theorem 7.2 A family {Iω | ω ∈ Ω} of nonempty compact subsets of W will

be maximal compact invariant set w.r.t. cocycle ϕ, if one of the following two
conditions is fulfilled :

a. {Iω | ω ∈ Ω} is a compact pullback attractor w.r.t. cocycle ϕ;

b. {Iω | ω ∈ Ω} is a compact global attractor w.r.t. cocycle ϕ.

Proof. a. Let the family {Iω | ω ∈ Ω} be a compact pullback attractor. If the

family {Iω0 | ω ∈ Ω} is a compact and invariant set of cocycle ϕ, then we have

β(Iω0 , Iω ) = β(ϕ(t, Iω0 −t , ω −t ), Iω ) ≤ β(ϕ(t, K, ω −t ), Iω ) → 0

S
as t → +∞, where K = {Iω0 | ω ∈ Ω}, and, consequently, Iω0 ⊆ Iω for every ω ∈ Ω,
i.e. {Iω | ω ∈ Ω} is maximal.
266 Global Attractors of Non-autonomous Dissipative Dynamical Systems

b. Let the family {Iω | ω ∈ Ω} be a compact global attractor w.r.t. cocycle

ϕ, then according to Theorem 2.24 it is a uniform compact pullback attractor and,
consequently, the family {Iω | ω ∈ Ω} is maximal compact invariant set of the
cocycle ϕ. 

Remark 7.1 The family {Iω | ω ∈ Ω} (Iω ⊂ W ) is a maximal compact invariant

S
w.r.t. cocycle ϕ if and only if the set J = {Jω | ω ∈ Ω}, where Jω = Iω × {ω}, is
a maximal compact invariant in the dynamical system (X, T, π).

7.3 Upper semi-continuity

Lemma 7.2 Let {Iω | ω ∈ Ω} be a maximal compact invariant set of cocycle

ϕ, then the function F : Ω 7→ C(W ), defined by equality F (ω) := Iω is upper
semi-continuous, i.e. for all ω0 ∈ Ω

β(F (ωk ), F (ω0 )) → 0,

if ρ(ωk , ω0 ) → 0.

Proof. Let ω0 ∈ Ω, ωk → ω0 and suppose there exists ε0 > 0 such that

β(F (ωk ), F (ω0 )) ≥ ε0 .

Then there exists xk ∈ Iωk such that

ρ(xk , Iω0 ) ≥ ε0 . (7.4)

As the set I is compact, without loss of generality we can suppose that the sequence
{xk } is convergent. Denote by x = lim xk , then by virtue of Theorem 7.1 the set
S k→+∞
I = {Iω | ω ∈ Ω} is compact and hence there exists ω0 ∈ Ω such that x ∈ Iω0 ⊂ I.
On the other hand, according to the inequality (7.4) x ∈ / Iω0 . This contradiction
shows that the function F is upper semi-continuous. 

Lemma 7.3 Let Λ be a compact metric space and ϕ : T+ × W × Λ × Ω 7→ W

verify the following conditions:

1. ϕ is continuous;
2. for every λ ∈ Λ the function ϕλ := ϕ(·, ·, λ, ·) : T+ × W × Ω 7→ W is a cocycle
on Ω with the fiber W ;
3. the cocycle ϕλ admits a pullback attractor {Iωλ | ω ∈ Ω} for every λ ∈ Λ;
S S
4. the set {I λ | λ ∈ Λ} is relatively compact, where I λ = {Iωλ | ω ∈ Ω},
then the equality

lim β(Iωλ , Iωλ00 ) = 0 (7.5)

λ→λ0 ,ω→ω0
 Upper semi-continuity of attractors 267

holds for every λ0 ∈ Λ and ω0 ∈ Ω and

lim β(Iλ , Iλ0 ) = 0 (7.6)

λ→λ0

for every λ0 ∈ Λ.

Proof. Let Y := Λ × Ω and µ : T × Y 7→ Y be the mapping defined by the

equality µ(t, (λ, ω)) := (λ, σ(t, ω)) for every t ∈ T, λ ∈ Λ and ω ∈ Ω. It is clear that
the triplet (Y, T, µ) is the group dynamical system on Y and ϕ : T+ × W × Y 7→
W (ϕ(t, x, (λ, ω)) := ϕ(t, x, λ, ω)) is the continuous cocycle on (Y, T, µ) with fiber
W . Under the conditions of Lemma 7.3 the cocycle ϕ admits a maximal compact
invariant set {Iy | y ∈ Y } (where Iy = I(λ,ω) = Iωλ ) because
[ [ [
{Iy | y ∈ Y } = {Iωλ | λ ∈ Λ, ω ∈ Ω} = {I λ | λ ∈ Λ}.

According to Lemma 7.2, the function F : Y 7→ C(W ), defined by the equality

F (λ, ω) := Iωλ , is upper semi-continuous and in particular the equality (7.6) holds.
We suppose that the equality (7.6) is not true, then there exist ε0 > 0, λ0 ∈
Λ, λk → λ0 , ωk ∈ Ω and xk ∈ Iωλkk such that

ρ(xk , Iλ0 ) ≥ ε0 . (7.7)

Without loss of generality we can suppose that ωk → ω0 , xk → x0 because the sets

S
Ω and {Iλ |λ ∈ Λ} are compact. According to the inequality (7.7) we have

ρ(x0 , Iλ0 ) ≥ ε0 .

On the other hand xk ∈ Iωλkk and from the equality (7.6) we have

x0 ∈ Iωλ00 ⊂ Iλ0

and, consequently,

ε0 ≤ ρ(x0 , Iλ0 ) ≤ β(Iωλ00 , Iλ0 ) = 0.

This contradiction shows that the equality (7.6) holds. 

Corollary 7.1 Under the conditions of Lemma 7.3 the equality

lim β(Iωλ , Iωλ0 ) = 0

λ→λ0

holds for each ω ∈ Ω.

Remark 7.2 The article [41] contains a statement close to Corollary 7.1 in the
case when the non-perturbed cocycle ϕλ0 is autonomous, i.e. the mapping ϕλ0 :
T+ × W × Ω → W does not depend of ω ∈ Ω .
268 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Lemma 7.4 Let the conditions of Lemma 7.3 and the following condition be ful-
filled:
5. for certain λ0 ∈ Λ the application F : Ω 7→ C(W ), defined by equality F (ω) :=
Iωλ0 is continuous, i.e. α(F (ω), F (ω0 )) → 0 if ω → ω0 for every ω0 ∈ Ω, where
α is the full metric of Hausdorff, i.e. α(A, B) := max{β(A, B), β(B, A)}.
Then

lim sup β(Iωλ , Iωλ0 ) = 0 . (7.8)

λ→λ0 ω∈Ω

Proof. Suppose that the equality (7.8) is not correct, then there exist ε0 > 0, λk →
λ0 , ωk ∈ Ω such that

β(Iωλkk , Iωλk0 ) ≥ ε0 . (7.9)

On the other hand we have

ε0 ≤ β(Iωλkk , Iωλk0 ) ≤ β(Iωλkk , Iωλ00 ) + β(Iωλ00 , Iωλk0 )

≤ β(Iωλkk , Iωλ00 ) + α(Iωλk0 , Iωλ00 ). (7.10)

According to Lemma 7.3 (see the equality (7.5))

lim β(Iωλkk , Iωλ00 ) = 0 . (7.11)

k→+∞

Under the condition 5. of Lemma 7.4 we have

lim α(Iωλk0 , Iωλ00 ) = 0. (7.12)

k→+∞

From (7.10)-(7.12) passing to the limit as k → +∞, we obtain ε0 ≤ 0. This

contradiction shows that the equality (7.8) holds. 

Definition 7.3 The family of cocycle {ϕλ }λ∈Λ is called collectively compact dis-
sipative (uniformly collectively compact dissipative), if there exists a nonempty
compact set K ⊆ W such that

lim sup{β(Uλ (t, ω)M, K)|ω ∈ Ω} = 0 ∀λ∈Λ (7.13)

t→+∞

(respectively lim sup{β(Uλ (t, ω)M, K) | ω ∈ Ω, λ ∈ Λ} = 0)

t→+∞

for all M ∈ C(W ), where Uλ (t, ω) := ϕλ (t, ·, ω).

Lemma 7.5 The following conditions are equivalent:

1. the family of cocycles {ϕλ }λ∈Λ is collectively compact dissipative;
2.2.a. every cocycle ϕλ (λ ∈ Λ) is compact dissipative;
S
2.b. the set {I λ | λ ∈ Λ} is compact.
 Upper semi-continuity of attractors 269

Proof. According to the equality (7.13) every cocycle ϕλ (λ ∈ Λ) is compact

S
dissipative and {I λ | λ ∈ Λ} ⊆ K.
S
Suppose that the conditions 2.a. and 2.b. hold. Let K := {I λ | λ ∈ Λ}, then
the equality (7.13) holds. 
Let {ϕλ }λ∈Λ be a family of cocycles on (Ω, T, σ) with fiber W and Ω̃ := Ω×Λ. On
Ω̃, we define a dynamical system (Ω̃, T, σ̃) by equality σ̃(t, (ω, λ)) := (σ(t, ω), λ) for
all t ∈ T, ω ∈ Ω and λ ∈ Λ. By family of cocycles {ϕλ }λ∈Λ is generated a cocycle ϕ̃
on (Ω̃, T, σ̃) with fiber W , defined in the following way: ϕ̃(t, w, (ω, λ)) := ϕλ (t, w, ω)
for all t ∈ T+ , w ∈ W, ω ∈ Ω and λ ∈ Λ.

Lemma 7.6 The following conditions are equivalent:

(1) the family of cocycles {ϕλ }λ∈Λ is uniformly collectively compact dissipative;
(2) the cocycle ϕ̃ is compact dissipative.

Proof. This assertion follows from the fact that

sup{β(Ũ (t, ω̃)M, K) | ω̃ ∈ Ω̃}

= sup{β(Uλ (t, ω)M, K) | ω ∈ Ω, λ ∈ Λ},

where Ũ (t, ω̃) := ϕ̃(t, ·, ω̃), and from the corresponding definitions. 

Theorem 7.3 Let Λ be a compact metric space and {ϕλ }λ∈Λ be a family of uni-
formly collectively compact dissipative cocycles on (Ω, T, σ) with fiber W , then the
following statements are true:
(1) every cocycle ϕλ (λ ∈ Λ) is compact dissipative;
(2) the family of compacts {Iωλ | ω ∈ Ω} = I λ is a compact global attractor of
cocycle ϕλ , where Iωλ := I(ω,λ) and I := {I(ω,λ) | (ω, λ) ∈ Ω̃} is a Levinson
center of cocycle ϕ̃;
S
(3) the set {I λ | λ ∈ Λ} is compact.

Proof. Consider the cocycle ϕ̃ generated by the family of cocycles {ϕ λ }λ∈Λ . Ac-
cording to Lemma 7.6 ϕ̃ is compact dissipative and in virtue of the Theorem 2.24
the following assertions take place:
1. Iω̃ = Ωω̃ (K) 6= ∅, is compact, Iω̃ ⊆ K and

lim β(Ũ (t, ω̃ −t )M, Iω̃ ) = 0 (7.14)

t→+∞

for every ω̃ ∈ Ω̃, where

\ [
Ωω̃ (K) = Ũ (τ, ω̃ −τ )K, (7.15)
t≥0 τ ≥t

ω̃ −τ := σ̃(−τ, ω̃) and K is a nonempty compact appearing in the equality (7.13);

270 Global Attractors of Non-autonomous Dissipative Dynamical Systems

2. Ũ (t, ω̃)Iω̃ = Iω̃t for all ω̃ ∈ Ω̃ and t ∈ T+ ;

S
3. the set I = {Iω̃ | ω̃ ∈ Ω̃} is compact.
To finish the proof we note that from the collective compact dissipativity of
the family of cocycles {ϕλ }λ∈Λ it follows that every cocycle ϕλ will be compact
dissipative. Let {Iωλ | ω ∈ Ω} = I λ be a Levinson center of the cocycle ϕλ , then
according to Theorem 2.24
\ [
Iωλ = Uλ (τ, ω −τ )K . (7.16)
t≥0 τ ≥t

From (7.15) and (7.16) it follows that Iωλ = Ωω̃ (K) = Iω̃ and, consequently, I λ =
S λ S S
{Iω | ω ∈ Ω} ⊆ {Iωλ | ω ∈ Ω, λ ∈ Λ} = I for all λ ∈ Λ. Thus {I λ | λ ∈ Λ} ⊆ I
and, consequently, it is compact. 

Definition 7.4 The family {(X, T+ , πλ )}λ∈Λ of autonomous dynamical systems is

called collectively ( uniformly collectively) asymptotic compact if for every bounded
positive invariant set M ⊆ X (i.e. πλ (t, M ) ⊆ M for all t ∈ T+ and λ ∈ Λ) there
exists a nonempty compact K such that

lim β(πλt M, K) = 0 ∀λ∈Λ (7.17)

t→+∞

( lim sup β(πλt M, K) = 0).

t→+∞ λ∈Λ

Definition 7.5 The bounded set K ⊂ X is called absorbing (uniformly absorb-

ing) for the family {(X, T+ , πλ )}λ∈Λ of autonomous dynamical systems if for any
bounded subset B ⊂ X there exists a number L = L(λ, B) > 0 (L = L(B) > 0)
such that πλt B ⊆ K for all t ≥ L(λ, B) (t ≥ L(B)) and λ ∈ Λ.

Theorem 7.4 Let Λ be a complete metric space. If the family {(X, T+ , πλ )}λ∈Λ
of autonomous dynamical systems admits an absorbing bounded set K ⊂ X and
is collectively asymptotic compact, then {(X, T+ , πλ )}λ∈Λ admits a global compact
attractor, i.e. there exists a nonempty compact set K ⊂ X such that

lim β(πλt B, K) = 0 (7.18)

t→+∞

for all λ ∈ Λ and bounded B ⊂ X.

Proof. Let the family {(X, T+ , πλ )}λ∈Λ of autonomous dynamical systems be

collectively asymptotic compact and a bounded M be its absorbing set. According
to Theorem 1.5 the nonempty set K = Ω(M ) is compact and the equality (7.18)
holds. The theorem is proved. 
 Upper semi-continuity of attractors 271

Theorem 7.5 Let Λ be a complete compact metric space. If the family

{(X, T+ , πλ )}λ∈Λ of autonomous dynamical systems admits a uniformly absorb-
ing bounded set K ⊂ X and it is uniformly collectively asymptotic compact, then
{(X, T+ , πλ )}λ∈Λ admits a uniform compact global attractor, i.e. there exists a
nonempty compact set K ⊂ X such that

lim sup β(πλt B, K) = 0 (7.19)

t→+∞ λ∈Λ

for all bounded B ⊂ X.

Proof. Consider the autonomous dynamical system (X̃, T+ , π̃) on X̃ := X × Λ

defined by equality π̃(t, (x, λ)) := (πλ (t, x), λ) for all t ∈ T+ , x ∈ X and λ ∈ Λ. We
note that under the conditions of Theorem 7.5 the bounded set K × Λ is absorbing
for dynamical system (X̃, T+ , π̃) if the set K is uniformly absorbing for the family
{(X, T+ , πλ )}λ∈Λ and (X̃, T+ , π̃) is asymptotically compact. According to Theorem
1.24 the dynamical system (X̃, T+ , π̃) admits a compact global attractor K̃ ⊂ X̃ :=
X × Λ. To finish the proof it is sufficient to note that the set K := pr1 K̃ ⊂ X is
compact and

sup β(πλt B, K) ≤ β(π̃λt B, K0 ) → 0

λ∈Λ

as t → +∞, where K0 := K × Λ ⊃ K̃, for all bounded subset B ⊂ X. 

Let ϕ be a cocycle on (Ω, T, σ) with fiber W and (X, T+ , π) be a skew-product

dynamical system, where X := W × Ω and π(t, (w, ω)) := (ϕ(t, w, ω), ωt) for all
t ∈ T+ , w ∈ W and ω ∈ Ω.

Definition 7.6 The cocycle ϕ is called asymptotically compact (a family

of cocycles {ϕλ }λ∈Λ is called collectively asymptotically compact) if a skew-
product dynamical system (X, T+ , π) (a family of skew-product dynamical systems
(X, T+ , πλ )λ∈Λ ) is asymptotically compact.

Theorem 7.6 Let Ω and Λ be compact metric spaces, W be a Banach space and
{ϕλ }λ∈Λ be a family of cocycles on (Ω, T, σ) with fiber W . If there exist r > 0 and
the function Vλ : W × Ω → R+ for all λ ∈ Λ, with the following properties:

(1) the family of cocycles {ϕλ }λ∈Λ is collectively asymptotically compact;

(2) the family of functions {Vλ }λ∈Λ is collectively bounded on bounded sets and for
every c ∈ R+ the sets {x ∈ Xr | Vλ (x) ≤ c} uniformly bounded;
(3) Vλ0 (w, ω) ≤ −c(|w|) for all w ∈ Wr = {w ∈ W | |w| ≥ r}, ω ∈ Ω
and λ ∈ Λ, where c : R+ → R+ is positive on [r, +∞), Vλ0 (w, ω) :=
lim sup t−1 [Vλ (ϕλ (t, w, ω), ωt) − Vλ (w, ω)] if T = R+ and Vλ0 (w, ω) :=
t→0+
Vλ (ϕλ (1, w, ω), σ(1, ω)) − Vλ (w, ω) if T = Z+ .
272 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Then every cocycle ϕλ (λ ∈ Λ) admits a uniform compact global attractor I λ (λ ∈ Λ)

S
and the set {I λ | λ ∈ Λ} is compact.

Proof. Let X := W × Ω and (X, T, πλ ) be a skew-product dynamical system,

generated by the cocycle ϕλ , then (X, h, Ω) , where h := pr2 : X → Ω, is a trivial
fiber bundle with fiber W . Under the conditions of Theorem 7.6 and according
to Theorem 5.2 the non-autonomous dynamical system h(X, T+ , πλ ), (Ω, T, σ), hi
admits a compact global attractor J λ and according to Theorem 2.24 the cocycle
ϕλ admits a compact global attractor I λ = {Iωλ | ω ∈ Ω}, where Iωλ := pr1 Jωλ and
T
Jωλ = pr2−1 (ω) J λ .
Let Ω̃ := Ω × Λ, (Ω̃, T, σ̃) be a dynamical system on Ω̃ defined by the
equality σ̃(t, (ω, λ)) := (σ(t, ω), λ) ( for all t ∈ T, ω ∈ Ω and λ ∈ Λ), X̃ :=
W × Ω̃ and (X̃, T+ , π̃) be an autonomous dynamical system defined by equality
π̃(t, (w, ω̃)) := (πλ (t, w), (ωt, λ)) for all ω̃ := (ω, λ) ∈ Ω̃ := Ω × Λ. Note that the
triplet (X̃, h, Ω), where h := pr2 : X̃ → Ω̃, is a trivial fiber bundle with fiber
W, h(X̃, T+ , π̃), (Ω̃, T, σ̃), hi is a non-autonomous dynamical system. The function
Ṽ : X̃r := Wr × Ω̃ → R+ , defined by the equality Ṽ (x̃) := Vλ (w, ω) for all
x̃ := (w, (ω, λ)) ∈ X̃r under the conditions of Theorem 7.6, all the conditions
of Theorem 5.3 hold and, consequently, the dynamical system (X̃, T+ , π̃) admits
a compact global attractor. To finish the proof of the theorem it is sufficiently to
note that if the dynamical system (X̃, T+ , π̃) admits a compact global attractor
˜ then the family of cocycles {ϕλ }λ∈Λ is uniformly collectively compact dissipa-
J,
S
tive and according to Theorem 7.3 the set I = {I λ | λ ∈ Λ} is compact, where
I λ = {Iωλ | ω ∈ Ω} is the compact global attractor of cocycle ϕλ . The theorem is
proved. 

7.4 Connectedness

Recall that the space W possesses the property (S) if for every compact K ∈ C(W )
there exists a compact connected set V ∈ C(W ) such that K ⊆ V .
If M ⊆ W , for each ω ∈ Ω, we write
\ [
Ωω (M ) = ϕ(τ, M, ω −τ ) .
t≥0 τ ≥t

Lemma 7.7 Suppose that the cocycle ϕ admits a compact pullback attractor
{Iω |ω ∈ Ω} , then the following assertions hold:

a. ∅ 6= Ωω (M ) ⊆ Iω for every M ∈ C(W ) and ω ∈ Ω;

b. the family {Ωω (M ) | ω ∈ Ω} is compact and invariant w.r.t. cocycle ϕ for every
M ∈ C(W );
 Upper semi-continuity of attractors 273

S
c. if I = {Iω | ω ∈ Ω} ⊆ M , then the following inclusion Iω ⊆ Ωω (M ) holds for
every ω ∈ Ω.

Proof. The first and second assertions follow from the definition of pullback at-
tractor and from the equalities (2.83)-(2.84).
Let I be a subset of M , then

Iω = ϕ(t, Iω−t , ω −t ) ⊆ ϕ(t, I, ω −t ) ⊆ ϕ(t, M, ω −t ) (7.20)

and according to the equality (2.83) we have Iω ⊆ Ωω (M ). 

Theorem 7.7 Let W possess the property (S) and let the cocycle ϕ admit a
compact pullback attractor {Iω | ω ∈ Ω}, then:

(1) the set Iω is connected for every ω ∈ Ω;

S
(2) if the space Ω is connected, then the set I = {Iω |ω ∈ Ω} also is connected.

Proof. 1. Since the equality (7.2) holds and the space W possesses the property
(S), then there exists a connected compact V ∈ C(W ) such that I ⊆ V and

lim β(ϕ(t, V, ω −t ), Iω ) = 0, (7.21)

t→+∞

for every ω ∈ Ω. We shall show that the set Iω is connected. If we suppose

F
that it is not true, then there are A1 , A2 6= ∅, closes and A1 A2 = Iω . Let
0 < ε0 < d(A1 , A2 ) and L = L(ε0 ) > 0 be such that
ε0
β(ϕ(t, V, ω −t ), Iω ) < (7.22)
3
for all t ≥ L(ε0 ). We note that the set ϕ(t, V, ω −t ) is connected and according to
the inclusion (7.20) and the inequality (7.22) the following condition
\ ε0 G ε0
ϕ(t, V, ω −t ) (W \ [B(A1 , ) B(A2 , )]) 6= ∅
3 3
is fulfilled for every t ≥ L(ε0 ) and ω ∈ Ω, where B(A, ε) := {u ∈ W | ρ(u, A) < ε}.
Then there exists tn → +∞ and un ∈ W such that
\ ε0 G ε0
un ∈ ϕ(tn , V, ω −tn ) (W \ [B(A1 , ) B(A2 , )]). (7.23)
3 3
According to the equality (7.21) it is possible to suppose that the sequence {u n } is
convergent. We denote by u := lim un , then from Lemma 2.15 follows that u ∈
n→+∞
Ωω (V ). Since I ⊆ V , then according to Lemma 2.15 we have u ∈ Ωω (V ) ⊆ Iω ⊆ I.
F
On the other hand according to (7.23) we have u ∈ / B(A1 , ε30 ) B(A2 , ε30 ). This
contradiction shows that the set Iω is connected.
2. Let the space Ω be connected. According to Lemma 7.3 the function F :
Ω 7→ C(W ), defined by equality F (ω) = Iω is upper semi-continuous and from the
274 Global Attractors of Non-autonomous Dissipative Dynamical Systems

S
corollary 1.8.13 [83] (see also Lemma 3.1 [148]) follows that the set I = {Iω | ω ∈
Ω} = F (Ω) is connected. 

Corollary 7.2 Let W be a metric space with the property (S) and let the cocycle
ϕ admit a compact global attractor {Iω | ω ∈ Ω}, then:

(1) the set Iω is connected for every ω ∈ Ω;

S
(2) if the space Ω is connected, then the set I = {Iω | ω ∈ Ω} also is connected.

Proof. This assertion follows from Theorems 7.2, 7.7 and Lemma 7.2. 

7.5 Applications

7.5.1 Quasi-homogeneous systems

Let E and G be two finite dimensional spaces.

Definition 7.7 The function f ∈ C(E × G, E) is called [94, 102] homogeneous of

order m with respect to variable u ∈ E if the equality f (λu, ω) = λm f (u, ω) holds
for all λ ≥ 0, u ∈ E and ω ∈ G.

Theorem 7.8 Let f ∈ C 1 (E), Φ ∈ C 1 (G), Ω ⊆ G be a compact invariant set of

dynamical system

ω 0 = Φ(ω), (7.24)

the function f be homogeneous (of order m > 1) and a zero solution of equation

u0 = f (u) (7.25)

be uniformly asymptotically stable. If F ∈ C 1 (E × G, E) and

|F (u, ω)| ≤ c|u|m

for all |u| ≥ r and ω ∈ Ω, where r and c are certain positive numbers, then there
exists a positive number λ0 such that for all λ ∈ Λ = [−λ0 , λ0 ] the following holds:

(1) a set Iωλ := {u ∈ E | sup{|ϕλ (t, u, ω)| : t ∈ R} < +∞} is not empty, compact
and connected for each ω ∈ Ω, where ϕλ (t, u, ω) is a unique solution of equation
u0 = f (u) + λF (u, ωt) satisfying the initial condition ϕλ (0, u, ω) = u;
(2) ϕλ (t, Iωλ , ω) = Iσ(t,ω)
λ
for all t ∈ R+ and ω ∈ Ω;
λ
S λ
(3) the set I = {Iω | ω ∈ Ω} is compact and connected;
(4) the equalities

lim β(ϕλ (t, M, ω−t ), Iωλ ) = 0 (7.26)

t→+∞
 Upper semi-continuity of attractors 275

and

lim β(ϕλ (t, M, ω), I λ ) = 0

t→+∞

take place for all λ ∈ Λ, ω ∈ Ω and bounded subset M ⊆ E.

S
(5) the set {I λ | λ ∈ Λ} is compact;
(6) the equality

lim sup β(Iωλ , 0) = 0

λ→0 ω∈Ω

holds.

Proof. Under the condition of Theorem 7.8 according to Theorem 5.34 by the
equality
Z +∞
V (u) = |π(t, u)|k dt,
0

where π(t, u) is a solution of equation (7.25) with condition π(0, u) = u, is defined a

continuously differentiable function V : E → R+ , verifying the following conditions:

a. V (µu) = µk−m+1 V (u) for all µ ≥ 0 and u ∈ E;

b. there exist positive numbers α and β such that α|u|k−m+1 ≤ V (u) ≤ β|u|k−m+1
for all u ∈ E;
c. V 0 (u) = DV (u)f (u) = −|u|k for all u ∈ E, where DV (u) is a derivative of
Frechet of function V in the point u.

Let us define a function V : X → R+ (X := E × Ω) in the following way:

V(u, ω) := V (u) for all (u, ω) ∈ X. Note that
d
V0 (u, ω) := V (ϕλ (t, u, ω))|t=0 = −|u|k + DV (u)λF (u, ω)
dt
and there exists λ0 > 0 such that the inequality

V0 (u, ω) ≤ −ν|u|k

holds for all ω ∈ Ω and |u| ≥ r, where ν = 1 − λ0 cL > 0 ( see the proof of Theorem
4.3 [102]).
For finishing the proof of the theorem it is sufficient to refer to Theorem 7.6 and
Lemma 7.4. 

Theorem 7.9 Let f ∈ C 1 (E × G, E), Φ ∈ C 1 (F ), Ω ⊆ G be a compact invariant

set of dynamical system (7.24), the function f be homogeneous (of order m = 1)
w.r.t. variable u ∈ E and a zero solution of equation

u0 = f (u, ωt) (ω ∈ Ω) (7.27)

276 Global Attractors of Non-autonomous Dissipative Dynamical Systems

be uniformly asymptotically stable. If |F (u, ω)| ≤ c|u| for all |u| ≥ r and ω ∈ Ω,
where r and c are certain positive numbers, then there exists a positive number λ 0
such that for all λ ∈ Λ = [−λ0 , λ0 ] the following assertions take place:

(1) a set Iωλ := {u ∈ E | sup{|ϕλ (t, u, ω)| : t ∈ R} < +∞} is not empty, compact
and connected for each ω ∈ Ω, where ϕλ (t, u, ω) is a unique solution of equation

u0 = f (u, ωt) + λF (u, ωt)

verifying the initial condition ϕλ (0, u, ω) = u;

(2) ϕλ (t, Iωλ , ω) = Iσ(t,ω)
λ
for all t ∈ R+ and ω ∈ Ω;
λ
S λ
(3) a set I := {Iω | ω ∈ Ω} is compact and connected;
(4) the equalities

lim β(ϕλ (t, M, ω−t ), Iωλ ) = 0

t→+∞

and

lim β(ϕλ (t, M, ω), I λ ) = 0 (7.28)

t→+∞

take place for all λ ∈ Λ, ω ∈ Ω and bounded subset M ⊆ E.

S
(5) the set {I λ | λ ∈ Λ} is compact;
(6) the following equality holds:

lim sup β(Iωλ , 0) = 0 .

λ→0 ω∈Ω

Proof. The proof of this assertion is similar to the proof of the theorem 7.8. 

7.5.2 Monotone systems

Let f ∈ C(E × Ω, E) be a function satisfying

Rehf (u1 , ω) − f (u2 , ω), u1 − u2 i ≤ −k|u1 − u2 |α (7.29)

for all ω ∈ Ω and u1 , u2 ∈ E (k > 0 and α ≥ 2).

Theorem 7.10 [316, 92] If the function f verifies the condition (7.29), then the
following statements are true:

(1) the set Iω = {u ∈ E | sup |ϕ(t, u, ω)| < +∞} contains a single point {ϕ(ω)}
t∈R
for all ω ∈ Ω, where ϕ(t, u, ω) is a solution of equation (7.26) with condition
ϕ(0, u, ω) = u;
(2) the inequalities

|ϕ(t, u, ω) − ϕ(ωt)| ≤ e−kt |u − ϕ(u)| (if α = 2),

 Upper semi-continuity of attractors 277

1
|ϕ(t, u, ω) − ϕ(ωt)| ≤ (|u − ϕ(u)|2−α + (α − 2)t) 2−α (if α > 2)

hold for all t ≥ 0, u ∈ E and ω ∈ Ω;

(3) the function γ : Ω → E, defined by equality γ(ω) = Iω is continuous and
γ(ωt) := ϕ(t, γ(ω), ω) for all t ≥ 0, u ∈ E and ω ∈ Ω.

Theorem 7.11 Let f ∈ C(E × Ω, E) be a function verifying the condition (7.29)

and F ∈ C(E × Ω, E) be a function with the condition

RehF (u1 , ω) − F (u2 , ω), u1 − u2 i ≤ L|u1 − u2 |α (7.30)

for all u1 , u2 ∈ E and ω ∈ Ω, where L is some positive number.

Then there exists a positive number λ0 such that for all |λ| ≤ λ0 the following
hold:

(1) the set Iωλ := {u ∈ E | sup |ϕλ (t, u, ω)| < +∞} contains a single point {ϕλ (ω)}
t∈R
for each ω ∈ Ω, where ϕλ (t, u, ω) is a unique solution of the equation

u0 = f (u, ωt) + λF (u, ωt) (ω ∈ Ω) (7.31)

satisfying the initial condition ϕλ (0, u, ω) = u;

(2) the function γλ : Ω → E defined by equality γλ (ω) = Iωλ is continuous and
γλ (ωt) = ϕλ (t, γλ (ω), ω) for all t ≥ 0, u ∈ E and ω ∈ Ω.
(3)

lim sup |γλ (ω) − γ0 (ω)| = 0.

λ→0 ω∈Ω

Proof. Let g := f + λF , then from (7.29)–(7.32) follows that

Rehg(u1 , ω) − g(u2 , ω), u1 − u2 i ≤ (−k + L|λ|)|u1 − u2 |α (7.32)

for all u1 , u2 ∈ E and ω ∈ Ω. From (7.32) follows that there exists λ0 > 0 such
that −k + L|λ| ≤ −k + Lλ0 < 0 for all |λ| ≤ λ0 and according to Theorem 7.10 the
assertions 1. and 2. of the theorem are true.
It is clear that for |λ| ≤ λ0 the cocycle ϕλ generated by the equation (7.31)
admits a compact global attractor I λ = {γλ (ω) | ω ∈ Ω}.
S
Now we will show that the set {I λ | λ ∈ Λ = [−λ0 , λ0 ]} is compact. Let
V (u) := 12 |u|2 , then

d
V 0 (u) = V (ϕλ (t, u, ω))|t=0 = Rehg(u, ω), ui (7.33)
dt
2
= Rehg(u, ω) − g(0, ω), ui + Rehg(0, ω), ui ≤ (−k + L|λ0 |)|u| + C|u|
C
= |u|2 (−k + L|λ0 | + α−1 ),
|u|
278 Global Attractors of Non-autonomous Dissipative Dynamical Systems

where C := max{|g(0, ω)| : ω ∈ Ω, λ ∈ Λ}. From the equality (7.33) follows that
there exists r > 0 such that for all |u| ≥ r

V 0 (u) ≤ −ν|u|2 , (7.34)

C
where ν = k − L|λ0 | − rα−1 > 0. Now to finish the proof of Theorem 7.10 it is
sufficient to refer to Theorem 7.6. The theorem is proved. 

7.5.3 Quasi-linear systems

Consider a non-autonomous quasi-linear system

u0 = A(ωt)u + λf (u, ωt) (ω ∈ Ω)

on E. Denote by [E] the space of all linear continuous operators acting onto E and
equipped with the operational norm.

Theorem 7.12 Let A ∈ C(Ω, [E]), f ∈ C(E × Ω, E) and let the following condi-
tions be fulfilled:

(1) there exists a positive constant α0 such that RehA(ω)u, ui ≤ −α0 |u|2 for all
u ∈ E and ω ∈ Ω;
(2) for any r > 0 there exists a positive constant L(r) such that

|f (u1 , ω) − f (u2 , ω)| ≤ L|u1 − u2 |

for all u1 , u2 ∈ B[0, r] := {u ∈ E | |u| ≤ r} and ω ∈ Ω.

Then there exists a positive constant λ0 such that for all λ ∈ Λ := [−λ0 , λ0 ] the
following statements are true:

(1) the set Iωλ := {u ∈ E | sup{|ϕλ (t, u, ω)| : t ∈ R} < +∞} is not empty, compact
and connected for each ω ∈ Ω, where ϕλ (t, u, ω) there is a unique solution of
equation

u0 = A(ωt)u + λF (u, ωt)

satisfying the initial condition ϕλ (0, u, ω) = u;

(2) ϕλ (t, Iωλ , ω) = Iσ(t,ω)
λ
for all t ∈ R+ and ω ∈ Ω;
λ
S λ
(3) the set I := {Iω | ω ∈ Ω} is compact and connected;
(4) the equalities

lim β(ϕλ (t, M, ω−t ), Iωλ ) = 0,

t→+∞

lim β(ϕλ (t, M, ω), I λ ) = 0

t→+∞

hold for all λ ∈ Λ, ω ∈ Ω and bounded subset M ⊆ E.

 Upper semi-continuity of attractors 279

S
(5) the set {I λ | λ ∈ Λ} is compact;
(6) the following equality is true

lim sup β(Iωλ , 0) = 0.

λ→0 ω∈Ω

Proof. Let λ0 be a positive number such that ν = α0 − λ0 α > 0, then the function
Fλ (u, ω) := A(ω)u + λf (u, ω) verifies the condition

RehFλ (u, ω), ui ≤ −ν|u|2 + λ0 β (7.35)

for all |λ| ≤ λ0 , ω ∈ Ω and u ∈ E .

From the inequality (7.35) follows (see, for example, [97, p.11]) that the inequal-
ity
λ0 β
|ϕλ (t, u, ω)|2 ≤ |u|2 e−2νt + (1 − e−2νt ) (7.36)
ν
holds for all t ∈ R+ , λ ∈ Λ and (u, ω) ∈ E × Ω and, consequently, the family of
cocycles {ϕλ }λ∈Λ admits a bounded absorbing set. Now to finish the proof of the
theorem it is sufficient to refer to Theorems 7.3, 7.5 and Lemma 7.4. 

7.5.4 Non-autonomously perturbed systems

Theorem 7.13 Suppose that f ∈ C(E) is uniformly Lipschitzian and an au-

tonomous system (7.24) has a global attractor I. Furthermore suppose that F ∈
C(E × Ω, E) is uniformly Lipschitz in u ∈ E and it is uniformly bounded on E × Ω,
i.e. sup{|F (u, ω)| : (u, ω) ∈ E × Ω} = K < +∞. Then there exists a positive
number λ0 > 0 such that for all λ ∈ Λ = [−λ0 , λ0 ] the following are true:

(1) the set Iωλ := {u ∈ E | sup{|ϕλ (t, u, ω)| : t ∈ R} < +∞} is not empty, compact
and connected for each ω ∈ Ω, where ϕλ (t, u, ω) there is a unique solution of
equation u0 = f (u) + λF (u, ωt) satisfying the initial condition ϕλ (0, u, ω) = u;
(2) ϕλ (t, Iωλ , ω) = Iσ(t,ω)
λ
for all t ∈ R+ and ω ∈ Ω;
λ
S λ
(3) the set I := {Iω | ω ∈ Ω} is compact and connected;
(4) the equalities

lim β(ϕλ (t, M, ω−t ), Iωλ ) = 0

t→+∞

and

lim β(ϕλ (t, M, ω), I λ ) = 0

t→+∞

take place for all λ ∈ Λ, ω ∈ Ω and bounded subset M ⊆ E.

S
(5) the set {I λ | λ ∈ Λ} is compact;
280 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(6) the following equality is true

lim sup β(Iωλ , I) = 0 .

λ→0 ω∈Ω

Proof. According to Theorem 22.5 from [327] (see also [217, 220]), under its con-
ditions there exists a continuous function V : E \ I → R+ with the following
properties:

a. V is uniformly Lipschitz in E, i.e. there exists a constant L > 0 such that

|V (u1 ) − V (u2 )| ≤ L|u1 − u2 | for all u1 , u2 ∈ E.
b. there exist continuous strictly increasing functions a, b : R+ → R+ with a(0) =
b(0) = 0 and 0 < a(r) < b(r) for all r > 0 such that a(β(u, I)) ≤ V (u) ≤
b(β(u, I)) for all u ∈ E, where β(u, I) := sup{ρ(u, x) | x ∈ I}.
c. there exists a constant c > 0 such that V 0 (u) ≤ −cV (u) for all u ∈ E \ I,
where V 0 (u) := lim sup t−1 [V (π(t, u)) − V (u)] and π(t, u) is a unique solution
t→0+
of equation (7.25) with initial condition π(0, u) = u.

Define a function V : X → R+ (X := E × Ω) in the following way: V(x) := V (u)

for all x = (u, ω) ∈ X. Note that

V0 (u, ω) = lim sup V (ϕλ (t, u, ω))|t=0 ≤ LKr − cV(u, ω)

t→0+

(see [220, p.11]) for all u ∈ E. Then there exist λ0 > 0 and r0 > 0 such that

V0 (u, ω) ≤ −LKλ0

for all |u| ≥ r0 and ω ∈ Ω.

To finish the proof of the theorem it is sufficient to refer to Theorem 7.6 and
Lemma 7.4. 

Remark 7.3 Similar theorem was proved in [220, Th.4.1] for the pullback attrac-
tors of non-autonomously perturbed systems.

7.5.5 Non-autonomous 2D Navier-Stokes equations

Let G ⊂ R2 be a bounded domain with C 2 smooth boundary,
(L2 (G))2
V := {u ∈ (Ẇ21 (G))2 , divu(x) = 0}, H := V ,

V 0 be the dual space of V, (Ẇ21 (G))2 denotes the Sobolev space of functions having
two components, and let π be the orthogonal projector from (L2 (G))2 onto H. The
operator F (u, v) := π(u, ∇)v has values in V 0 .
Let Ω be a compact complete metric space, (Ω, R, σ) be a dynamical system on
Ω, F ∈ C(V × Ω, V ) and satisfy the the following conditions:
 Upper semi-continuity of attractors 281

(i) |F(u1 , ω) − F(u2 , ω)| ≤ L|u1 − u2 | for all u1 , u2 ∈ V and ω ∈ Ω;

(ii) RehF(u, ω), ui ≤ M |u|2 + N for all u ∈ V and ω ∈ Ω, where L, M and N are
some positive constants.

Consider the perturbed 2D Navier-Stokes equation

u0 + νAu + B(u, u) = F(u, ωt) (ω ∈ Ω) (7.37)

on H, where B : V × V → V 0 is a bilinear continuous map and A is the extension

of −π∇ with zero boundary conditions on V and ν > 0. In particular, there exists
λ1 > 0 such that

hAu, ui ≥ |u|2V ≥ λ1 |u|2H

for any u ∈ V . According to [286],[311] by equation (7.37) is generated a cocycle

ϕ(t, u, ω) on (Ω, R, σ) with fiber H, where ϕ(t, u, ω) is a unique solution of equation
(7.37) with the condition ϕ(0, u, ω) = u.

Lemma 7.8 Under the conditions (i) and (ii) the following assertions holds:

(1) for any T > 0, ν > 0, ω ∈ Ω and any u ∈ H the equation (7.34) has a unique
solution ϕ(t, u, ω) with path in C([0, T ], H);
(2) the energy inequality holds

d |F(0, ωt)|2H
|ϕ(t, u, ω)|2H + νλ1 ϕ(t, u, ω)|2H ≤ + 2L|ϕ(t, u, ω)|2H (7.38)
dt νλ1

for all t ∈ [0, T ], u ∈ H and ω ∈ Ω;

(3) the mapping ϕ : R+ × H × Ω → H is continuous.

Proof. The assertions 1. and 2. follow from [111] (see also Lemma 3.1 [286]).
Now we will prove that the mapping ϕ : R+ × H × Ω → H is continuous. Let
t0 ∈ R+ , u0 ∈ H and ω0 ∈ Ω, then we have

|ϕ(t, u, ω) − ϕ(t0 , u0 , ω0 )|H (7.39)

≤ |ϕ(t, u, ω) − ϕ(t, u0 , ω0 )|H + |ϕ(t, u0 , ω0 ) − ϕ(t0 , u0 , ω0 )|H .

Denote by w(t) := ϕ(t, u, ω) − ϕ(t, u0 , ω0 ) and f (t) := F(ϕ(t, u, ω), ωt) −

F(ϕ(t, u0 , ω0 ), ω0 t), then the function w(t) verifies the following equation

dw
+ νAw + B(w, w) + B(w, u1 ) + B(u1 , w) = f (t), (7.40)
dt
where u1 := ϕ(t, u0 , ω0 ). Using the well-known identity hB(u, v), vi = 0 (∀v ∈ H),
282 Global Attractors of Non-autonomous Dissipative Dynamical Systems

where h·, ·i is the scalar product in H, we obtain

1 d
|w|2 = hẇ, wi = h−νAw − B(w, w) − B(w, u1 ) − B(u1 , w) (7.41)
2 dt H
+f (t), wi = −νhAw, wi − hB(w, w), wi − hB(w, u1 ), wi − hB(u1 , w), wi
+hf (t), wi = −νhAw, wi − hB(w, u1 ), wi + hf (t), wi.

Bearing in mind the inequality |u|2L4 ≤ |w|H |w|V (see [232]) we obtain

|hB(w, u1 ), wi| ≤ |w|2L4 |u1 |V

ν 1
≤ |w|H |w|V |u1 |V ≤ |w|2V + |w|2 |u1 |2V .
2 2ν H
Taking into account that |(f, w)| ≤ |f |H |w|H , we get from (7.41)
1 d νλ1 2 1
|w|2 ≤ −νλ1 |w|2V + |w|V + |w|2 |u1 |2V + |f |H |w|H . (7.42)
2 dt H 2 2νλ1 H
We remark that

|f (t)|H = |F(ϕ(t, u, ω), ωt) − F(ϕ(t, u0 , ω0 ), ω0 t)|H (7.43)

≤ L|ϕ(t, u0 , ω0 ) − ϕ(t, u0 , ω0 )| + |F(ϕ(t, u0 , ω0 ), ωt) − F(ϕ(t, u0 , ω0 ), ω0 t)|

and, consequently, from (7.41)-(7.43), we obtain

1 d 1 1 |f |2
|w|2H ≤ ( |u1 |2V + L + )|w|2H + . (7.44)
2 dt 2ν 2 2
From this differential inequality we deduce that
Z t
2 1 1

|w(t)|H ≤ exp ( |ϕ(τ, u0 , ω0 )|2V + L + )dτ |u − u0 |2H
0 2ν 2
Z t Z τ
1 1

+ exp − ( |ϕ(s, u0 , ω0 )|2V + L + )ds (7.45)
0 0 2ν 2
1  
× F(ϕ(τ, u0 , ω0 ), ωτ ) − F(ϕ(τ, u0 , ω0 ), ω0 τ )dτ.
2
Since F ∈ C(H × Ω, H), then

max |F(ϕ(t, u0 , ω0 ), ωt) − F(ϕ(t, u0 , ω0 ), ω0 t)| → 0

0≤t≤T

as ω → ω0 and, consequently, from (7.45) we obtain

max |w(t)| → 0. (7.46)

0≤t≤T

From (7.39) and (7.46) we obtain the continuity of mapping ϕ. The lemma is
proved. 
 Upper semi-continuity of attractors 283

Corollary 7.3 Under the conditions (i) and (ii) there exists a positive number
L0 < νλ2 1 such that if L < L0 , then the following inequality

|f |2
|ϕ(t, u, ω)|2 ≤ e(−νλ1 +2L0 )t |u|2 +
νλ1 (−2L0 + νλ1 )
holds for all t ≥ 0, u ∈ H and ω ∈ Ω, where |f | := max |F(0, ω)|.
ω∈Ω

This assertion follows from the second assertion of Lemma 7.8.

Theorem 7.14 There exists a positive number L0 > 0 such that the cocycle ϕ
generated by (7.37) admits a compact global attractor, if L ≤ L 0 .

Proof. According to Lemma 3.1 [286] there exists L0 > 0 (for example L0 < νλ2 1 )
such that the cocycle ϕ admits a bounded absorbing set if L < L0 . On the other
hand the cocycle ϕ is compact, i.e. the mapping ϕ(t, ·, ·) : V × Ω → V is completely
continuous for all t > 0. To finish the proof of the theorem it is sufficient to refer
to Theorem 2.24. 

Theorem 7.15 Under the conditions (i) and (ii) there exists a positive number
λ0 such that the following is true:

(1) the set Iωλ := {u ∈ H | sup |ϕλ (t, u, ω)| < +∞} is not empty, compact and
t∈R
connected for all ω ∈ Ω and λ ∈ Λ := [−λ0 , −λ0 ], where h ∈ H and ϕλ (t, u, ω)
is a unique solution of the equation

u0 + νAu + F (u, u) + h = λF(u, ωt) (ω ∈ Ω) (7.47)

satisfying the initial condition ϕλ (0, u, ω) = u;

(2) ϕλ (t, Iωλ , ω) = Iσ(t,ω)
λ
for all λ ∈ Λ, t ∈ R+ and ω ∈ Ω;
S
(3) the set I λ = {Iωλ | ω ∈ Ω} is compact and connected;
(4) the equalities

lim β(ϕλ (t, M, ω−t ), Iωλ ) = 0

t→+∞

and

lim β(ϕλ (t, M, ω), I λ ) = 0

t→+∞

take place for all λ ∈ Λ, ω ∈ Ω and bounded subset M ⊆ E.

S
(5) the set {I λ | λ ∈ Λ} is compact and connected;
(6) the equality

lim sup β(Iωλ , I) = 0

λ→0 ω∈Ω
284 Global Attractors of Non-autonomous Dissipative Dynamical Systems

holds, where I is a Levinson center for the equation

u0 + νAu + F (u, u) + h = 0.

νλ1
Proof. Let F̃ (u, ωt) := −h + λF(u, ωt) and λ0 < 2L , then for the equation

u0 + νAu + F (u, u) = λF̃ (u, ωt) (ω ∈ Ω)

the conditions of Theorem 7.14 are fulfilled. Let ϕλ be a cocycle generated by

equation (7.43), then according to Corollary 7.3 the family of cocycle {ϕ λ }λ∈Λ
admits a collectively absorbing bounded set. Since the imbedding V into H is
compact, to finish the proof of the theorem it is sufficient to refer to Theorem 7.6
and Lemma 7.4. The theorem is proved. 

7.5.6 Quasi-linear functional-differential equations

Let r > 0, C([a, b], Rn ) be the Banach space of continuous functions ν : [a, b] → Rn
with the sup-norm . If [a, b] := [−r, 0], then suppose C := C([−r, 0], Rn ). Let
σ ∈ R, A ≥ 0 and u ∈ C([σ − r, σ + A], Rn ). For any t ∈ [σ, σ + A] define ut ∈ C
by the equality ut (θ) = u(t + θ), −r ≤ θ ≤ 0. Let us define by C(Ω × C, Rn ) the
space of all continuous functions f : Ω × C → Rn , with compact-open topology and
let (Ω, R, σ) be a dynamical system on the compact metric space Ω. Consider the
equation

u0 = f (ωt, ut ) (ω ∈ Ω), (7.48)

where f ∈ C(Ω × C, Rn ). We will suppose that the function f is regular, that is for
any ω ∈ Ω and u ∈ C the equation (7.48) has a unique solution ϕ(t, u, ω) which is
defined on R+ = [0, +∞). Let X := C × Ω, and π : X × R+ → X be a dynamical
system on X defined by the following rule: π(τ, (u, ω)) := (ϕτ (u, ω), ωτ ), then the
triplet h(X, R+ , π), (Ω, R, σ), hi (h := pr2 : X → Ω) is a non-autonomous dynamical
system, where ϕτ (u, ω)(θ) := ϕ(τ + θ, u, ω).
From the general properties of solutions of (7.48) (see, for example [179]), we
have the following statement.

Theorem 7.16 The following statements are true:

(1) The non-autonomous dynamical system h(X, R+ , π), (Ω, R, σ), hi generated by
equation (7.45) is conditionally completely continuous;
(2) Let Ω be compact and the function f : Ω × C → Rn be bounded on Ω × B for
any bounded set B ⊂ C, then the non-autonomous dynamical system generated
by the equation (7.48) is conditionally completely continuous (in particular, it
is asymptotically compact).
 Upper semi-continuity of attractors 285

Denote by [C] the space of all linear continuous operators acting onto C and
equipped with the operational norm.

Theorem 7.17 [78] Let A ∈ C(Ω, [C]), f ∈ C(C × Ω, E) and let the following
conditions be fulfilled:

(1) a zero solution of equation

u0 = A(ωt)ut (7.49)

is uniformly asymptotically stable, i.e. there exist positive numbers N and ν

such that |ϕ0 (t, u, ω)| ≤ N e−νt |u| for all t ≥ 0, u ∈ C and ω ∈ Ω, where
ϕ0 (t, u, ω) is a solution of equation (7.49) with condition that ϕ0 (0, u, ω) = u;
(2) there exists a positive constant L such that

|f (u1 , ω) − f (u2 , ω)| ≤ L|u1 − u2 |

for all u1 , u2 ∈ C and ω ∈ Ω.

ν
Then there exists a positive constant ε0 (ε0 < N) such that
NM NM
|ϕ(t, u, ω)| ≤ + (N |u| − )e−(ν−N L)t
ν − NL ν − NL
for all t ≥ 0, L ∈ (0, ε0 ), u ∈ C and ω ∈ Ω, where ϕ(t, u, ω) is a unique solution of
the equation

u0 = A(ωt)ut + f (ut , ωt)

with the condition ϕ(0, u, ω) = u and M := max{|f (0, ω)| : ω ∈ Ω}.

Consider a non-autonomous quasi-linear system

u0 = A(ωt)ut + λf (ut , ωt) (ω ∈ Ω) (7.50)

on C.

Theorem 7.18 Let f ∈ C(C × Ω, E) and let the inequality

|f (u1 , ω) − f (u2 , ω)| ≤ L|u1 − u2 |

take place for all u1 , u2 ∈ C and ω ∈ Ω, where L is some positive number.

Then there exists a positive number λ0 such that for all λ ∈ Λ := [−λ0 , λ0 ] the
following statements are true:

(1) the set Iωλ := {u ∈ C | sup{|ϕλ (t, u, ω)| : t ∈ R} < +∞} is not empty, compact
and connected for each ω ∈ Ω,where ϕλ (t, u, ω) there is a unique solution of
equation (7.50) satisfying the initial condition ϕλ (0, u, ω) = u;
(2) ϕλ (t, Iωλ , ω) = Iσ(t,ω)
λ
for all t ∈ R+ and ω ∈ Ω;
286 Global Attractors of Non-autonomous Dissipative Dynamical Systems

S
(3) the set I λ = {Iωλ | ω ∈ Ω} is compact and connected;
(4) the equalities

lim β(ϕλ (t, M, ω−t ), Iωλ ) = 0,

t→+∞

lim β(ϕλ (t, M, ω), I λ ) = 0

t→+∞

hold for all λ ∈ Λ, ω ∈ Ω and bounded subset M ⊆ E.

S
(5) the set {I λ | λ ∈ Λ} is compact;
(6) the following equality holds

lim sup β(Iωλ , 0) = 0 .

λ→0 ω∈Ω

Proof. Let λ0 be a positive number such that ν = λ0 L < ν/N , then the function
Fλ (u, ω) := A(ω)u + λf (u, ω) satisfies the condition

|Fλ (u, ω)| ≤ ν|u| + M (7.51)

(with M := max |f (0, ω|) for all |λ| ≤ λ0 , ω ∈ Ω and u ∈ C. From the inequality
ω∈Ω
(7.51) and Theorem 7.16 follows that the family of cocycles {ϕλ }λ∈Λ admits a
bounded absorbing set. Now to finish the proof of theorem it is sufficient to refer
to Theorems 7.3, 7.5, 7.15 and Lemma 7.4. 
 Chapter 8

The relationship between pullback,

forward and global attractors

Non-autonomous dynamical systems can often be formulated in terms of a cocy-

cle mapping for the dynamics in the state space that is driven by an autonomous
dynamical system in what is called a parameter space. Traditionally the driv-
ing system is topological and the resulting cartesian product system forms an au-
tonomous semi–dynamical system that is known as a skew–product flow. Results on
global attractors for autonomous semi–dynamical systems can thus be adapted to
such non-autonomous dynamical systems via the associated skew–product flow [64,
75, 88, 83, 111, 185, 292, 318].
A new type of attractor, called a pullback attractor, was proposed and in-
vestigated for non-autonomous or random dynamical systems [97, 126, 153, 262,
284]. Essentially, it consists of a parametrized family of nonempty compact subsets
of the state space that are mapped onto each other by the cocycle mapping as the
parameter is changed by the underlying driving system. Pull back attraction to a
component subset for a fixed parameter value is achieved by starting progressively
earlier in time, that is, at parameter values that are carried forward to the fixed
value. The deeper reason for this procedure is that a cocycle can be interpreted
as a mapping between the fibers of a fiber bundle where the image fiber is fixed.
The kernels of a global attractor of the skew–product flows considered in [111] are
very similar. This differs from the more conventional forward convergence where
the parameter value of the limiting object also evolves with time, in which case the
parametrized family could be called a forward attractor.
Pullback attractors and forward attractors can, of course, be defined for
non-autonomous dynamical systems with a topological driving system [218, 220,
221]. In fact, when the driving system is the shift operator on the real line, forward
attraction to a time varying solution, say, is the same as the attraction in Lyapunov
asymptotic stability. The situation of a compact parameter space is dynamically
more interesting as the associated skew–product flow may then have a global at-
tractor. The relationship between the global attractor of the skew–product system
and the pullback and forward attractors of the cocycle system is investigated in

287
288 Global Attractors of Non-autonomous Dissipative Dynamical Systems

this chapter. We also note that forward attractors are stronger than global attrac-
tors if we suppose a compact set of non-autonomous perturbations. An example is
presented in which the cartesian product of the component subsets of a pullback
attractor is not a global attractor of the skew–product flow. This set is, however,
a maximal compact invariant subset of the skew–product flow. By a generalization
of some stability results of Zubov [336] it is asymptotically stable. Thus a pull-
back attractor always generates a local attractor of the skew–product system, but
this need not be a global attractor. If, however, the pullback attractor generates a
global attractor in the skew–product flow and if, in addition, its component subsets
depend lower continuously on the parameter, then the pullback attractor is also a
forward attractor. Several examples illustrating these results are presented in the
final section.

8.1 Pullback, forward and global attractors

A general non-autonomous dynamical system is defined here in terms of a cocycle

mapping ϕ on a state space W that is driven by an autonomous dynamical system
σ acting on a base space Ω, which will be called the parameter space. In particular,
let W be a complete metric space, let Ω be a compact metric space and let T, the
time set, be either R or Z.
Let (Ω, T, σ) be a dynamical system on the metric space Ω. Recall that the
triple hU, ϕ, (Ω, T, σ)i is called a cocycle (or non-autonomous dynamical system) on
the state space W , if the mapping ϕ : T+ ×W ×Ω → Ω is continuous, ϕ(0, u, ω) = u
and ϕ(t ∗ τ, u, ω) = ϕ(t, ϕ(τ, u, ω), ωτ ) fora all u ∈ W and t, τ ∈ T+ .
Let X be the cartesian product of W and Ω. Then the mapping π : T+ × X →
X defined by

π(t, (u, ω)) := (ϕ(t, u, ω), σt ω)

forms a semi–group on X over T+ [290, 291].

The autonomous semi–dynamical system (X, T+ , π) = (W × Ω, T+ , (ϕ, σ))
is called the skew–product dynamical system associated with the cocycle
hW, ϕ, (Ω, T, σ)i.
For example, let W be a Banach space and let the space C = C(R × W, W ) of
continuous functions f : R × W → W be equipped with the compact open topology.
Consider the autonomous dynamical system (C, R, σ), where σ is the shift operator
on C defined by σt f (·, ·) := f (· + t, ·) for all t ∈ T. Let Ω be the hull H(f ) of a
given functions f ∈ C, that is,
[
Ω = H(f ) := {f (· + t, ·)},
t∈R
 The relationship between pullback, forward and global attractors 289

and denote the restriction of (C, R, σ) to Ω by (Ω, R, σ). Let F : Ω × W → W

be the continuous mapping defined by F (ω, u) := ω(0, u) for ω ∈ Ω and u ∈ W .
Then, under appropriate restrictions on the given function f ∈ C (see Sell [290,
291]) defining Ω, the differential equation

u0 = ω(t, u) = F (σt ω, u) (8.1)

generates a cocycle hW, ϕ, (Ω, R, σ)i, where ϕ(t, u, ω) is the solution of (8.1) with
the initial value u at time t = 0.
Let β denote the Hausdorff semi-distance between two nonempty sets of a metric
space Y , that is,

β(A, B) = sup inf ρ(A, B),

a∈A b∈B

and let D(W ) be C(W ) or B(W ), where C(W ) (B(W )) is a classes of sets containing
the compact subsets (the bounded subsets) of the metric space W .
The definition of a global attractor for an autonomous semi–dynamical system
(X, T+ , π) is well known. Specifically, a nonempty compact subset A of X which is
π-invariant, that is, satisfies

π(t, A) = A for all t ∈ T+ , (8.2)

is called a global attractor for (X, T+ , π) with respect to D(X) if

lim ρ(π(t, D), A) = 0 (8.3)

t→∞

for every D ∈ D(X).

Conditions for the existence of such global attractors and examples can be found
in [20, 88, 179, 314, 318]. Of course, semi–dynamical systems need not be a skew–
product systems, but when they are, the following definition will be used.
Suppose that the skew–product dynamical system (X, T+ , π) = (W × Ω, T+ ,
(ϕ, σ)) has a global attractor A. Then we will call the set A the global attractor
with respect to D(W ) of the associated cocycle hW, ϕ, (Ω, T, σ)i.

Definition 8.1 The global attractor A with respect to D(W ) of the skew–product
dynamical system (X, T+ , π) = (W ×Ω, T+ , (ϕ, σ)) will be called the global attractor
with respect to D(W ) of the associated cocycle hW, ϕ, (Ω, T, σ)i.

Other types of attractors, in particular pullback attractors, that consist of a

family of nonempty compact subsets of the state space of the cocycle mapping have
been proposed for non-autonomous or random dynamical systems [126, 127, 218,
284, 285].
290 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Definition 8.2 Let I = {Iω | ω ∈ Ω} be a family of nonempty compact sets of

S
W for which ω∈Ω Iω ) is relatively compact and let I be ϕ–invariant with respect
to a cocycle hW, ϕ, (Ω, T, σ)i, that is, satisfies

ϕ(t, Iω , ω) = Iσt ω for all t ∈ T+ , ω ∈ Ω. (8.4)

Definition 8.3 The family I is called a pullback attractor of hW, ϕ, (Ω, T, σ)i
with respect to D(U ) if

lim β(ϕ(t, D, σ−t ω), Iω ) = 0 (8.5)

t→∞

for any D ∈ D(W ) and ω ∈ Ω, or a uniform pullback attractor if the convergence

(8.5) is uniform in ω ∈ Ω, that is, if

lim sup β(ϕ(t, D, σ−t ω), Iω ) = 0.

t→∞ ω∈Ω

Definition 8.4 The family I is called a forward attractor if the forward conver-
gence

lim β(ϕ(t, D, ω), Iσt ω ) = 0

t→∞

holds instead of the pullback convergence (8.5), or a uniform forward attractor if

this forward convergence is uniform in ω ∈ Ω, that is, if

lim sup β(ϕ(t, D, ω), A(σt ω)) = 0.

t→∞ ω∈Ω

The main task of this chapter is to investigate connections of different types of

attractors.
It follows directly from the definition that a pullback attractor is unique. Obvi-
ously, any uniform pullback attractor is also a uniform forward attractor, and vice
versa.
If I is a forward attractor for the cocycle hW, ϕ, (Ω, T, σ)i, then ([88] Lemma
4.2) the subset
[

J= Iω × {ω} (8.6)
ω∈Ω

of X is the global attractor for the skew–product dynamical system (X, T+ , π). A
weaker result holds when I is a pullback attractor. The inverse property is not true
in general.
Although we could formulate for weaker assumptions we will restrict ourselves
S
to the case that ω∈Ω Iω is compact and D(W ) consists of compact sets.
Our considerations can be embedded into the more general theory of pullback
attractors with a domain of attraction D consisting of family of sets D = {D(ω)} ω∈Ω
S
such that ω∈Ω D(ω) is relatively compact in W , see [285].
 The relationship between pullback, forward and global attractors 291

The following existence result for pullback attractors is adapted from [127, 211].

Theorem 8.1 Let hW, ϕ, (Ω, T, σ)i, with Ω compact be a cocycle and suppose
S
that there exists a family of nonempty sets C = {C(ω)}ω∈Ω , ω∈Ω C(ω) relatively
compact such that

lim β(ϕ(t, D, σ−t ω), C(ω)) = 0

t→∞

for any bounded subset D of W and any ω ∈ Ω. Then there exists a pullback
attractor.

A related result is given by Theorem 2.24: if the skew–product system (X, T+ , π)

has a global attractor A, then the cocycle hW, ϕ, (Ω, T, σ)i has a pullback attractor.
We now continue to derive properties of pullback attractors an skew–product
dynamical systems.

Lemma 8.1 If I is a pullback attractor of a cocycle hW, ϕ, (Ω, T, σ)i, where Ω

is compact, then the subset A of X defined by (8.6) is the maximal π-invariant
compact set of the associated skew–product dynamical system (X, T + , π).

Proof. The π-invariance follows from the ϕ-invariance of I via

[ [
π(t, J) = (ϕ(t, Iω , ω), σt ω) = (Iσt ω , σt ω) = J.
ω∈Ω ω∈Ω
S S
Now J ⊂ ω∈Ω Iω × Ω, where Ω is compact and ω∈Ω Iω is relatively compact, so
J is relatively compact. Hence B := J¯ is compact, from which it follows that

B(ω) := {u : (u, ω) ∈ B}

is a compact set in W for each ω ∈ Ω and that the set

[
B(ω) ⊂ pr1 B
ω∈Ω

is relatively compact. On the other hand, B is π-invariant since

π(t, B) = π(t, J¯) = π(t, J) = J¯ = B

for the continuous mapping π(t, ·). In addition, ϕ(t, B(ω), ω) = B(σt ω) holds, that
is, the B(ω) are ϕ-invariant, since
[ [
π(t, B) = (ϕ(t, B(p), p), σt p) = B = (B(σt p), σt p)
p∈P p∈P

and σt ω = σt ω̂ implies that ω = ω̂ for the homeomorphism σt . The construction

shows B(ω) ⊃ Iω . By the ϕ-invariance of the B(ω) and the pullback attraction
property it follows then that B(ω) = Iω such that A = B. Hence A is compact.
292 Global Attractors of Non-autonomous Dissipative Dynamical Systems

0
To prove that the compact invariant set J is maximal, let J be any other
0
compact invariant set the of skew–product dynamical system (X, T+ , π). Then J
0
= {I (ω)}ω∈Ω is a family of compact ϕ–invariant subsets of W and by pullback
attraction
0 0
β(Iω , Iω) = β(ϕ(t, I(σ−t ω) , σ−t ω), Iω )
≤ β(ϕ(t, K, σ−t ω), Iω ) → 0
0
as t → +∞, where K := bigcupω∈ΩIω0 is compact. Hence Iω ⊆ Iω for every ω ∈ Ω,
0
i.e. J ⊆ J, which means J is maximal for (X, T+ , π). 

Definition 8.5 A set–valued mapping M = {M (ω)}ω∈Ω for M (ω) ∈ W is called

upper semi–continuous if

lim β(M (ω), M (ω0 )) = 0 for any ω0 ∈ Ω.

ω→ω0

We call such a set–valued mapping lower semi–continuous if

lim β(M (ω0 ), M (ω)) = 0 for any ω0 ∈ Ω.

ω→ω0

Definition 8.6 M is called continuous if it is both upper and lower semi–

continuous.

Note that a set–valued mapping M is upper semi–continuous if and only if the

graph of is closed in W × Ω.
It follows straightforwardly from Lemma 8.1:

Corollary 8.1 The set valued mapping ω → Iω formed with the components sets
of a pullback attractor I = {Iω | ω ∈ Ω} of a cocycle hW, ϕ, (Ω, T, σ)i, where Ω is
compact, is upper semi–continuous.

The following example shows that, in general, a pullback attractor need not also
be a forward attractor nor form a global attractor of the associated skew–product
dynamical system.

Example 8.1 Let f be the function on R defined by

 2
1+t
f (t) = − , t ∈ R,
1 + t2
and let (Ω, R, σ) be the autonomous dynamical system Ω := H(f ), the hull of f in
C(R, R), with the shift operator σ. Note that
[
Ω = H(f ) = {f (· + h)} ∪ {0}.
h∈R
 The relationship between pullback, forward and global attractors 293

Finally, let E be the evaluation functional on C(R, R), that is E(ω) := ω(0) ∈ R.

Lemma 8.2 The functional

Z ∞ Z ∞
γ(ω) = − e−τ E(στ ω) dτ = − e−τ ω(τ ) dτ
0 0

is well defined and continuous on Ω, and the function of t ∈ R given by


Z ∞  1 : ω = σh f
γ(σt ω) = −e t −τ
e ω(τ ) dτ = 1 + (t + h)2
t 
0 : ω=0
is the unique solution of the differential equation

x0 = x + E(σt ω) = x + ω(t)

that exists and is bounded for all t ∈ R.

Proof. The proof is by straightforward calculation, so will be omitted. 

Consider now the non-autonomous differential equation

u0 = g(σt ω, u), (8.7)

where


−u − E(ω)u2 : 0 ≤ uγ(ω) ≤ 1, ω 6= 0


  
g(ω, u) := − 1 E(ω)
 1+ : 1 < u γ(ω), ω 6= 0 .

 γ(ω) γ(ω)


−u : 0 ≤ u, ω = 0
It is easily seen that this equation has a unique solution passing through any point
u ∈ W = R+ at time t = 0 defined on R. These solutions define a cocycle mapping
 u0

 t : 0 ≤ u0 γ(ω) ≤ 1, ω 6= 0

 e (1 − u 0 γ(ω)) + u0 γ(σt ω)
ϕ(t, u0 , ω) = u0 + 1 − 1 : 1 < u0 γ(ω), ω 6= 0 . (8.8)

 γ(σt ω) γ(ω)


e−t u0 : 0 ≤ u0 , ω = 0
According to the construction, the cocycle mapping ϕ admits as its only invariant
sets Iω = {0} for ω ∈ Ω. To see that the Iω = {0} form a pullback attractor, observe
that
 u0

 t : 0 ≤ u0 γ(ω) ≤ 1, ω 6= 0

 e (1 − u 0 γ(σ −t ω)) + u0 γ(ω)
ϕ(t, u0 , σ−t ω) = u0 + 1 − 1 : 1 < u0 γ(ω), ω 6= 0 .

 γ(ω) γ(σ−t ω)


e−t u0 : 0 ≤ u0 , ω = 0
294 Global Attractors of Non-autonomous Dissipative Dynamical Systems

In particular, note that t → γ(σt ω)−1 is a solution of the differential equation (8.7).
Since γ(σ−t ω)−1 tends to +∞ sub-exponentially fast for t → ∞, it follows that
1 −1 t
ϕ(t, u, σ−t ω) ≤ Le 2
2
for any u ∈ [0, L] for any L ≥ 0 and ω ∈ Ω provided t is sufficiently large. Conse-
quently I = {Iω | ω ∈ Ω} with Iω = {0} for all ω ∈ Ω is a pullback attractor for ϕ.
In view of (8.8), the stable set W s (J) := {x ∈ X| lim ρ(π t x, J) = 0} of J, that
t→+∞
is, the set of all points in X that are attracted to J by π, is given by

W s (J) = {(u, ω) : ω ∈ Ω, u ≥ 0, u γ(ω) < 1} 6= X.

Hence the cocycle mapping ϕ in this example admits a pullback attractor that is
neither a forward attractor for ϕ nor a global attractor of the associated skew–
product flow.

Other examples for different kinds of attractors are given by Scheutzow [283]
for random dynamical systems generated by one dimensional stochastic differen-
tial equations. However, these considerations are based on the theory of Markov
processes.

8.2 Asymptotic stability in α-condensing semi–dynamical systems

To continue to investigate general relations between pull back attractors and skew–
product flows we have to derive some results from the general stability theory. We
start with some definitions.
Let (X, T+ , π) be a semi–dynamical system. The ω– limit set of a set M is
defined to be
\ [
ω(M ) = π(t, M ).
τ ≥0 t≥τ

Definition 8.7 A set M is called Lyapunov stable if for any ε > 0 there exists
an δ > 0 such that π(t, B(M, δ)) ⊆ B(U, ε) for t ≥ 0.

Definition 8.8 M is called a local attractor if there exists a neighborhood

B(M, γ) (γ > 0) of M such that B(M, γ) ⊆ W s (M ).

Definition 8.9 A set M which is Lyapunov stable and a local attractor is called
asymptotically stable.

Note that any asymptotically stable compact set M also attracts compact sets
contained in U(M ).
 The relationship between pullback, forward and global attractors 295

Definition 8.10 Recall that a π–invariant compact set M is said to be locally

maximal if there exists a number δ > 0 such that any π–invariant compact set
contained in the open δ-neighborhood B(M, δ) of M is in fact contained in M .

Definition 8.11 An autonomous semi–dynamical system (X, T+ , π) is called

[179] α-condensing if π(t, B) is bounded and

α(π(t, B)) < α(B)

for all t > 0 for any bounded set B of X with α(B) > 0.

Remark 8.1 In the book [179] there are a lot class of dynamical systems which
possesses with property. For example: every d.s. on the finite-dimensional space,
every d.s. with compact π t (t > 0) or π t = m(t) + r(t)(m(t) : X → X is compact for
every t > 0 and r(t)x → 0 as t → +∞ uniformly w.r.t. x on every bounded set X).

Theorem 8.2 Let M be a locally maximal compact set for an α-condensing semi–
dynamical system (X, T+ , π). Then M is Lyapunov stable if and only if there exists
a δ > 0 such that

α γx ∩ M = ∅

for any entire trajectory γx through any x ∈ B(M, δ) \ M .

Proof. A proof of the necessity direction was given by Zubov in [336] Theorem
7 for a locally compact space X. This proof remains also true for a non locally
compact space under consideration here. Really, let M be a compact invariant set
for (X, T+ , π) stable in the positive direction. If we suppose that this assertion is
not true, then there exist x ∈ / M, γx and τn → −∞ such that ρ(γx (τn ), M ) → 0 as
n → ∞. Let 0 < ε < ρ(x, M ) and δ(ε) > 0 the corresponding positive number from
stability of set M , then for sufficiently large n we have ρ(γx (τn ), M ) < δ(ε) and,
consequently, ρ(π t γx (τn ), M ) < ε for all t ≥ 0. In particularly for t = −τn we have
ρ(x, M ) = ρ(π −τn γx (τn ), M ) < ε. The obtained contradiction prove our assertion.
For the sufficiency direction, consider first the case T+ = Z+ and let B(M, δ0 )
be a neighborhood such that M is locally maximal in B(M, δ0 ). Suppose that M
is not Lyapunov stable, but that the other condition of the theorem holds. Then
there exist an ε0 > 0 and sequences δn → 0, xn ∈ B(M, δn ), kn → ∞ such that
π(k, xn ) ∈ B(M, ε0 ) for 0 ≤ k ≤ kn − 1 and π(kn , xn ) 6∈ B(M, ε0 ). This ε0 has to
be chosen sufficiently small such that
δ0
β(π(1, B(M, ε0 )), M ) < .
2
Define A = {xn } and B = ∪n∈N {π(k, xn ) | 0 ≤ k ≤ kn − 1}. Then α(A) = 0(α
is the measure of non-compactness of Kuratowskii) since A is relatively compact.
296 Global Attractors of Non-autonomous Dissipative Dynamical Systems

In addition, π(1, B) ⊆ B(M, δ0 ), so π(1, B) is bounded. Suppose that B is not

relatively compact, so α(B) > 0. It follows by the properties of the measure of
non-compactness for the non relatively compact set B that

α(B) = α(A ∪ π(1, B) ∩ B) ≤ max(α(A), π(1, B)) = α(π(1, B)) < α(B)

which is a contradiction. This shows that B is relatively compact. Now γ̃x̃ is an

entire trajectory of the discrete–time semi–dynamical system above with γ̃ x̃ (0) = x̃
and γ̃x̃ (Z− ) ⊂ B̄. Thus the α limit set αγ̃x̃ is nonempty, compact and invariant. In
addition, αγ̃x̃ ⊂ B(M, ε0 ), hence αγ̃x̃ ⊂ M because M is a locally maximal invariant
compact set. On the other hand, γ̃x̃ (0) = x̃ ∈ B(M, ε0 ) \ M , so αγ̃x̃ ∩ M = ∅ holds
by the assumptions. This contradiction proves the sufficiency of the condition in
the discrete–time case.
Now let T+ = R+ and suppose that αγx ∩ M = ∅ where x 6∈ M holds for the
continuous–time system. Then it also holds for the restricted discrete–time system
generated by π1 := π(1, ·) because any entire trajectory γx of the restricted discrete–
time system can be extended to an entire trajectory of the continuous–time system
via

γx (t) = π(τ, γx (n)), n ∈ Z, t = n + τ, 0 < τ < 1.

Consequently, the set M is Lyapunov stable with respect to the restricted discrete–
time dynamical system generated by π1 . Since M is compact, for every ε > 0 there
exists a µ > 0 such that

ρ(π(t, x), M ) < ε for all t ∈ [0, 1], x ∈ B(M, µ).

In view of the first part of the proof above, there is a δ > 0 such that

ρ(π(n, x), M ) < min(µ, ε) for x ∈ B(M, δ) for n ∈ Z+ .

The Lyapunov stability of M for the continuous dynamical system (X, R+ , π) then
follows from the semi–group property of π. 
The next lemma will be needed to formulate the second main theorem of this
section. Asymptotic stability here means (locally) Lyapunov stability and attract-
ing.

Lemma 8.3 Let M be a compact subset of X that is positively invariant for a

semi–dynamical system (X, T+ , π). Then M is asymptotically stable if and only if
ω(M ) is locally maximal and asymptotically stable.

Proof. Suppose that M is asymptotically stable. Then there exists a closed posi-
tively invariant bounded neighborhood C of M contained in its stable set W s (M ).
The mapping π can be restricted to the complete metric space C to form a semi–
dynamical system (C, T+ , π). Since M is a locally attracting set it attracts compact
 The relationship between pullback, forward and global attractors 297

subsets of C. The assertion then follows by Theorems 2.4.2 and 3.4.2 in [179] be-
T
cause ω(M ) = t∈T+ π(t, M ).
Suppose instead that ω(M ) is asymptotically stable and locally maximal. Since
T
M is compact, ω(M ) = t≥0 π(t, M ). Hence there exist η > 0 and τ ∈ T+ such
that

π(τ, M ) ⊂ B(ω(M ), η) ⊂ W s (ω(M )).

Now π −1 (τ, B(ω(M ), η)), where π −1 denotes the pre–image of π(τ, ·) for fixed τ ,
is an open neighborhood of M and π(τ, π −1 (τ, B(ω(M ), η))) ⊂ W s (ω(M )). Hence
for any x ∈ π −1 (τ, B(ω(M ), η) ⊂ W s (ω(M )) we have that π(t, x) tends to ω(M )
as t → ∞, from which it follows that π(t, x) also tends to M because M ⊃ Ω(M ).
Then, if M were not Lyapunov stable, there would exist ε0 > 0, δn → 0, xn ∈
B(M, δn ) and tn → ∞ such that

ρ(π(tn , xn ), M ) ≥ ε0 . (8.9)

For sufficiently large n0 , the set {xn }n≥n0 would then be contained in the pre–image
π −1 (1, B(ω(M ), η). Since M is compact, so is the set {xn }n≥n0 . This set would
thus be attracted by ω(M ) ⊂ M , which contradicts (8.9) . 

Lemma 8.4 Let M be a compact subset of X that is a positively invariant set for
an asymptotically compact semi–dynamical system (X, T+ , π). Then the set M is
asymptotically stable if and only if ω(M ) is locally maximal and Lyapunov stable.

Proof. The necessity follows by Lemma 8.3. Suppose instead that ω(M ) is locally
maximal and Lyapunov stable. Then for any ε > 0 there exists a δ > 0 such that

π(t, B(ω(M ), δ) ⊂ B(ω(M ), ε) for all t ≥ 0.

By the assumption of asymptotic compactness, ω(B(ω(M ), δ)) is nonempty and

compact with

lim β(π(t, B(ω(M ), δ)), ω(B(ω(M ), δ))) = 0

t→∞

(see [179] Corollary 2.2.4.). Since ω(M ) is locally maximal, ω(B(ω(M ), δ)) ⊂ ω(M )
for sufficiently small δ > 0, which means that ω(M ) is asymptotically stable. The
conclusion then follows by Lemma 8.3. 

Corollary 8.2 Let (X, T+ , π) be asymptotically compact and let M be a compact

π–invariant set. Then M is asymptotically stable if and only if M is locally maximal
and Lyapunov stable.

Proof. Indeed, M = ω(M ) here, so just apply Lemma 8.4. 

298 Global Attractors of Non-autonomous Dissipative Dynamical Systems

The next theorem is a generalization to infinite dimensional spaces and α-

condensing systems of Theorem 8 of Zubov [336] characterizing the asymptotic
stability of a compact set.

Theorem 8.3 Let (X, T+ , π) be an α-condensing semi–dynamical system and let

M ⊂ X be a compact invariant set. Then the set M is asymptotically stable if and
only if
(i) M is locally maximal, and
(ii) there exists a δ > 0 such that αγx ∩ M = ∅ for any entire trajectory γx through
any x ∈ B(M, δ) \ M .

Proof. By Lemma 2.3.5 in [179] any α-condensing semi–dynamical system is asymp-

totically compact, so the assertion follows easily from Theorem 8.2 and Corollary
8.2. 

Definition 8.12 A cocycle hW, ϕ, (Ω, T, σ)i will be called α-condensing if the set
ϕ(t, B, Ω) is bounded and

α(ϕ(t, B, Ω)) < α(B)

for all t > 0 for any bounded subset B of W with α(B) > 0.

Lemma 8.5 Suppose that the cocycle hW, ϕ, (Ω, T, σ)i is α-condensing. Then the
associated skew–product flow (X, T+ , π) is also α-condensing.
S
Proof. Let M := ω∈Ω (Mω × {ω}) be a bounded set in X. Then M can be covered
by finitely many balls Mi ⊂ X, i = 1, · · · , n, of largest radius α(M ) + ε for an
arbitrary ε > 0. The sets pr1 Mi ⊂ W , i = 1, · · · , n, cover pr1 M . The sets Mi are
balls so α(pr1 Mi ) = α(Mi ) < α(M ) + ε for i = 1, · · · , n. It is easily seen that
[
π(t, M ) = {π(t, (Mω , ω))} =
ω∈Ω

[
{(ϕ(t, Mω , ω), σt ω)} ⊂ ϕ(t, pr1 M, Ω) × Ω.
ω∈Ω

Since ϕ is α-condensing, the set ϕ(t, pr1 M, Ω) is bounded. Hence

α(π(t, M )) ≤ α(ϕ(t, pr1 M, Ω) × Ω) (8.10)

≤ α(ϕ(t, pr1 M, Ω)) < α(pr1 M ) ≤ α(M ) for each t > 0.

The second inequality above is true by the compactness of Ω. Indeed, Ω can be

covered by finitely many open balls Ωi of arbitrarily small radius. Hence

α(ϕ(t, pr1 M, Ω) × Ω) ≤ max α(ϕ(t, pr1 M, Ω) × Ωi ) ≤ α(ϕ(t, pr1 M, Ω)) + ε

i
 The relationship between pullback, forward and global attractors 299

for arbitrarily small ε > 0. The conclusion of Lemma follows by (8.10). 

8.3 Uniform pullback attractors and global attractors

It was seen earlier that the set ∪ω∈Ω (Iω × {ω}) ⊂ X which was defined in terms
of the pullback attractor I = {Iω }ω∈Ω of a cocycle hW, ϕ, (Ω, T, σ)i is the maximal
π-invariant compact subset of the associated skew–product system (X, T+ , π), but
need not be a global attractor. However, this set is always a attractor under the
additional assumption that the cocycle ϕ is α-condensing.

Theorem 8.4 Let Ω be a compact space, hW, ϕ, (Ω, T, σ)i be an α-condensing

cocycle with a pullback attractor I = {Iω }ω∈Ω and define J = ∪ω∈Ω (Iω × {ω}).
Then

(i) the α-limit set αγx of any entire trajectory γx passing through x ∈ X \ J is
empty.
(ii) J is asymptotically stable with respect to π.

Proof. Suppose that there exists an entire trajectory γx through x = (u, ω) ∈

X \ J such that αγx 6= ∅. Then there exists a subsequence −τn → ∞ such that
S
γx (τn ) converges to a point in αγx . The set K = pr1 n∈N γx (τn ) is compact since
S
n∈N γx (τn ) is compact. Also I = {Iω | ω ∈ Ω} is a pullback attractor, so

lim β(ϕ(−τn , K, στn ω), Iω ) = 0

n→∞

from which it follows that u ∈ Iω . Hence (u, ω) ∈ J, which is a contradiction. This

proves the first assertion.
By Lemma 8.5 (X, T+ , π) is α-condensing. According to Lemma 8.1 J is a
maximal compact invariant set of (X, T+ , π) since I is a pullback attractor of the
cocycle ϕ. The second assertion then follows from Theorem 8.4 and from the first
assertion of this theorem. 

Remark 8.2 (i) The skew–product system in the example in Section 2 has only
a local attractor associated with the pullback attractor.
(ii) If in addition to the assumptions of Theorem 8.4 the stable set W s (J) of J
satisfies W s (J) = X, then J is in fact a global attractor (see Theorem 1.13).

Theorem 8.5 Suppose that Ω compact, hW, ϕ, (Ω, T, σ)i is a cocycle with a pull-
back attractor I = {Iω | ω ∈ Ω} and suppose that W s (J) = X, where J =
∪ω∈Ω (Iω × {ω}).
If the mapping ω → Iω is lower semi–continuous, then I is a uniform pullback
attractor, hence a uniform forward attractor.
300 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Proof. Suppose that the uniform convergence

lim sup β(ϕ(t, D, ω), Iσt ω ) = 0

t→∞ ω∈Ω

is not true for some D ∈ C(W ). Then there exist ε0 > 0, a set D0 ∈ C(W ) and
sequences tn → ∞, ωn ∈ Ω and un ∈ D0 such that:

ρ(ϕ(tn , un , ωn ), Iσtn ω ) ≥ ε0 . (8.11)

Now Ω is compact and J is a global attractor by Remark 8.2 (ii), so it can be

supposed that the sequences {ϕ(tn , un , ωn )} and {σtn ω} are convergent. Let ū =
limn→∞ ϕ(tn , un , ωn ) and ω̄ = limn→∞ σtn ωn . Then ū ∈ Iω̄ because x̄ = (ū, ω̄) ∈
J. On the other hand, by (8.11),

ε0 ≤ β(ϕ(tn , ωn , xn ), Iσtn ωn )
≤ β(ϕ(tn , ωn , xn ), Iω̄ ) + β(Iω̄ , Iσtn ωn ).

By the lower semi–continuity of ω → Iω it follows then that ū ∈

/ A(ω̄), which is a
contradiction. 

Remark 8.3 The example 8.1 shows that Theorem 8.5 is in general not true
without the assumption that W s (J) = X. In view of Corollary 8.1, the set valued
mapping ω → Iω will, in fact, then be continuous here.

8.4 Examples of uniform pullback attractors

Several examples illustrating the application of the above results, in particular of

Theorem 8.5, are now presented. More complicated examples will be discussed in
Chapter 12 (section 12.5).

8.4.1 Periodic driving systems

Consider a periodical dynamical system (Ω, T, σ), that is, for which there exists a
minimal positive number T such that σT ω = ω for any ω ∈ Ω.

Theorem 8.6 Suppose that a non-autonomous α-condensing dynamical system

hU, ϕ, (Ω, R, σ)i with a periodical dynamical system (Ω, R, σ) has a pullback attractor
I = {Iω }ω∈Ω . Then IJ is a uniform forward attractor for hW, ϕ, (Ω, R, σ)i.

Proof. Consider a sequence ωn → ω. By the periodicity of the driving system

there exists a sequence τn ∈ [0, T ] such that ωn = στn ω. By compactness, there is
a convergent subsequence (indexed here for convenience like the full one) τn → τ ∈
 The relationship between pullback, forward and global attractors 301

[0, T ]. Hence

ω = lim ωn = lim στn ω = στ ω

n→∞ n→∞

which means τ = 0 or T . Suppose that τ = T . Then

lim β(Iω , Iωn ) = lim β(Iω , ϕ(τn , Iω , ω))

n→∞ n→∞
= β(Iω , ϕ(T, Iω , ω)) = β(Iω , A(σT ω)) = 0

since ϕ is continuous and Iωn = Iστn ω = ϕ(τn , Iω , ω) by the ϕ–invariance of J.

Hence the set valued ω → Iω is lower semi–continuous.
Now ϕ(nT, u0 , ω) = ϕ(nT, u0 , σ−nT ω) since ω = σ−nT ω by the periodicity of
the driving system (Ω, R, σ). Hence from pullback convergence

lim ρ(ϕ(nT, u0 , ω), Iω ) = lim ρ(ϕ(nT, u0 , σ−nT ω), Iω ) = 0

n→∞ n→∞

for any (u0 , ω) ∈ U × P . On the other hand

sup ρ(ϕ(s + nT, u0 , ω), A(σs+nT ω))

s∈[0,T ]

= sup ρ(ϕ(s, ϕ(nT, u0 , ω), σnT ω), ϕ(s, A(σnT ω), σnT ω)) = 0
s∈[0,T ]

by the cocycle property of ϕ and the ϕ–invariance of J. Hence

lim ρ((ϕ(t, u0 , ω), σt ω), J) = 0,

t→∞

where J = ∪ω∈Ω (Iω × {ω}). This shows that W s (J) = X. The result then follows
by Theorem 8.5. 
Consider the 2–dimensional Navier-Stokes equation in the operator form
du
+ νAu + B(u) = f (t), u(0) = u0 ∈ H, (8.12)
dt
which can be interpreted as an evolution equation on the rigged space V ⊂ H ⊂
V 0 , where V and H are certain Banach spaces. In particular, here W = H, which
is in fact a Hilbert space, for the phase space. Then, from [311, Chapter 3],

Lemma 8.6 The 2–dimensional Navier-Stokes equation (8.12) has a unique solu-
tion u(·, u0 , f ) in C(0, T ; H) for each initial condition u0 ∈ H and forcing term f ∈
C(0, T ; H) for every T > 0. Moreover, u(t, u0 , f ) depends continuously on (t, u0 , f )
as a mapping from R+ × H × C(R, H) to H.

Now suppose that f is a periodic function in C(R, H) and define σt f (·) :=

S
f (· + t). Then Ω := t∈R σt f is a compact subset of C(R, H). By Lemma 8.6 the
mapping (t, u0 , ω) → ϕ(t, u0 , ω) from R+ × H × C(R, H) → H defined by ϕ(t, u0 , ω)
= u(t, u0 , ω) is continuous and forms a cocycle mapping with respect to σ on Ω.
302 Global Attractors of Non-autonomous Dissipative Dynamical Systems

By [311] Theorem III.3.10 the mapping ϕ is completely continuous and hence α-

condensing.

Lemma 8.7 The cocycle hH, ϕ, (Ω, R, σ)i generated by the Navier-Stokes equation
(8.12) with periodic forcing term in C(R, H) has a pullback attractor.

Proof. The solution of the Navier-Stokes equation satisfies an energy inequality

Z t Z
1 t
|u(t)|2H + λ1 ν |u(τ )|2H dτ ≤ |u0 |2H + |ω(τ )|2V 0 dτ
0 ν 0
where λ1 is the smallest eigenvalue of A. It follows that the balls B[0, R(ω)] in H
with center zero and square radius
Z
1 0 νλ1 τ
R2 (ω) = e |ω(τ )|2V 0 dτ
ν −∞

is a pullback attracting family of sets in the sense of Theorem 8.1. In particu-

lar, C(ω) := ϕ(1, B[0, R(σ−1 ω)], σ−1 ω) satisfies all of the required properties of
Theorem 8.1 because ϕ(1, ·, ω) is completely continuous. 

This theorem and Theorem 8.6 give

Theorem 8.7 The cocycle hH, ϕ,(Ω, R, σ)i generated by the Navier-Stokes equa-
tion (8.12) with periodic forcing term in C(R, H) has a uniform pullback attractor,
which is also a uniform forward attractor.

Remark 8.4 See [149] for a related result involving a different type of non-
autonomous attractor.

8.4.2 Pullback attractors with singleton component sets

Now pullback attractors with singleton component sets, that is with

Iω = {a(ω)}, a(ω) ∈ W,

will be considered.

Lemma 8.8 Let hW, ϕ, (Ω, T, σ)i be a cocycle, J = {Iω }ω∈Ω be a pullback attrac-
tor with singleton component sets. Then the mapping ω → Iω is continuous.

Proof. This follows from Corollary 8.1 since the upper semi–continuity of a set
valued mapping ω → Iω reduces to continuity when the Iω are single point sets. 

It follows straightforwardly from this lemma and Theorem 8.5 that

 The relationship between pullback, forward and global attractors 303

Theorem 8.8 Suppose that Ω is compact and the cocycle hW, ϕ, (Ω, T, σ)i has a
pullback attractor I = {Iω }ω∈Ω with singleton component sets which generates a
global attractor J = ∪ω∈Ω Iω × {ω}. Then J is a uniform pullback attractor and,
hence, also a uniform forward attractor.

The previous theorem can be applied to differential equations on a Hilbert space

(H, h·, ·i) of the form

u0 = F (σt ω, u) (8.13)

where F ∈ C(Ω × H, H) is uniformly dissipative, that is, there exist ν ≥ 2, α > 0

hF (ω, u1 ) − F (ω, u2 ), u1 − u2 i ≤ −α|u1 − u2 |ν (8.14)

for any u1 , u2 ∈ H and ω ∈ Ω.

Theorem 8.9 [92] The cocycle hH, ϕ, (Ω, T, σ)i generated by (8.13) has a uniform
pullback attractor that consists of singleton component subsets.

For example, the equation

u0 = F (σt ω, x) = −u|u| + g(σt ω)

with g ∈ C(Ω, R) satisfies

1 1
hW1 − u2 , F (σt ω, u1 ) − F (σt ω, u2 )i ≤ − |u1 − u2 |2 (|u1 | + |u2 |) ≤ − |u1 − u2 |3 ,
2 2
which is condition (8.14) with α = 12 and ν = 3.
The above considerations apply also to nonlinear non-autonomous partial dif-
ferential equations with a uniform dissipative structure, such as the dissipative
quasi–geostrophic equations

ut + J(ψ, u) + βψx = ν∆u − ru + f (x, y, t) , (8.15)

with relative vorticity u(x, y, t) = ∆ψ(x, y, t), where J(f, g) = fx gy − fy gx is the

Jacobian operator. This equation can be supplemented by homogeneous Dirichlet
boundary conditions for both ψ and u

ψ(x, y, t) = 0, u(x, y, t) = 0 on ∂D , (8.16)

and an initial condition,

u(x, y, 0) = u0 (x, y) on D,

where D is an arbitrary bounded planar domain with area |D| and piecewise smooth
boundary. Let W be the Hilbert space L2 (D) with norm | · |.
304 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Theorem 8.10 Assume that

 
r πν 1 |D|
+ > β +1
2 |D| 2 π
and that the wind forcing f (x, y, t) is temporally periodic (quasi-periodic, almost
periodic, recurrent) with its L2 (D)–norm bounded uniformly in time t ∈ R by
r    23
πr r πν 1 |D|
||f (·, ·, t)|| ≤ + − β +1 .
3|D| 2 |D| 2 π
Then the dissipative quasigeostrophic model (8.15)–(8.16) has a unique temporally
periodic (quasi-periodic, almost periodic, recurrent) solution that exists for all time
t ∈ R.

The proof in [141] (for the almost periodic case) involves explicitly constructing
a uniform pullback and forward absorbing ball in L2 (D) for the vorticity ω, hence
implying the existence of a uniform pullback attractor as well as a global attractor
for the associate skew–product flow system for which the component sets are sin-
gleton sets. The parameter set Ω here is the hull of the forcing term f in L2 (D) and
a completely continuous cocycle ϕ(t, u0 , ω) = ω(t, u0 , ω) with respect to the shift
operator σ on Ω that is continuous in all variables.

8.4.3 Distal dynamical systems

A function γ(u,ω) : R → W represents an entire trajectory γ(u,ω) of a cocycle hW, ϕ,
(Ω, T, σ)i if γ(u,ω) (0) = u ∈ W and ϕ(t, γ(u,ω) (τ ), στ ω) = γ(u,ω) (t + τ ) for t ≥ 0 and
τ ∈ R.

Definition 8.13 A cocycle is called distal on T− if

inf ρ γ(u1 ,ω) (t), γ(u2 ,ω) (t) > 0
t∈T−

for any entire trajectories γ(u1 ,ω) and γ(u2 ,ω) with u1 6= u2 ∈ W and any ω ∈ Ω.

Definition 8.14 A cocycle ϕ is said to be uniformly Lyapunov stable if for any

ε > 0 there exists a δ = δ(ε) > 0 such that

ρ (ϕ(t, u1 , ω), ϕ(t, u2 , ω)) < δ

for all u1 , u2 ∈ W with ρ (u1 , u2 ) < ε, ω ∈ Ω and t ≥ 0.

Recall that an autonomous dynamical system (Ω, T, σ) is called minimal if Ω

does not contain proper compact subsets which are σ–invariant.
The following lemma is due to Furstenberg [156] (see also [32, Chapter 3] or
[238, Chapter 7] Proposition 4).
 The relationship between pullback, forward and global attractors 305

Lemma 8.9 Suppose that a cocycle hW, ϕ, (Ω, T, σ)i is distal on T− and that
(Ω, T, σ) is minimal. In addition suppose that a compact subset J of X is π–
invariant with respect to the skew–product system (X, T+ , π). Then the mapping ω
→ Iω := {u ∈ W : (u, ω) ∈ J} is continuous.

The following theorem gives the existence of uniform forward attractors.

Theorem 8.11 Suppose that the cocycle hW, ϕ, (Ω, T, σ)i is uniformly Lyapunov
stable and that the skew–product system (X, T+ , π) has a global attractor J =
∪ω∈Ω Iω × {ω}. Then I = {Iω }ω∈Ω is a uniform forward attractor for hW, ϕ,
(Ω, T, σ)i.

Proof. Suppose that the cocycle hW, ϕ, (Ω, T, σ)i is not distal. Then there is a ω0
∈ Ω, a sequence tn → ∞ and entire trajectories γ(u1 ,ω0 ) , γ(u2 ,ω0 ) with u1 6= u2 such
that


lim ρ γ(u1 ,ω0 ) (−tn ), γ(u2 ,ω0 ) (−tn ) = 0.
n→∞

Let ε = ρ (u1 , u2 ) > 0 and choose δ = δ(ε) > 0 from the uniformly Lyapunov
stability property. Then


ρ γ(u1 ,ω0 ) (−tn ), γ(u2 ,ω0 ) (−tn ) < δ

for sufficiently large n. Hence

ρ ϕ(t, γ(u1 ,ω0 ) (−tn ), σ−tn ω0 ), ϕ(t, γ(u2 ,ω0 ) (−tn ), σ−tn ω0 ) < ε

for t ≥ 0 and, in particular,

ε = ρ(u1 , u2 ) = ρ(ϕ(tn , γ(u1 ,ω0 ) (−tn ), σ−tn ω0 ),

ϕ(tn , γ(u2 ,ω0 ) (−tn ), σ−tn ω0 )) < ε

for t = tn , which is a contradiction. The cocycle is thus distal, so ω → Iω is

continuous by Lemma 8.9. The result then follows from Theorem 8.5 since {Iω | ω ∈
Ω} generates a pullback attractor. 
This theorem will now be applied to the non-autonomous differential equation
(8.13) on a Hilbert space H, which is assumed to generate a cocycle ϕ that is
continuous on T+ × Ω × H and asymptotically compact.

Theorem 8.12 Suppose that F ∈ C(H × Ω, H) satisfies the dissipativity condi-

tions

hF (u1 , ω) − F (u2 , ω), u1 − u2 i ≤ 0 (8.17)

hF (u, ω), ui ≤ −µ(|u|) (8.18)

306 Global Attractors of Non-autonomous Dissipative Dynamical Systems

for u1 , u2 , u ∈ H and ω ∈ Ω, where µ : [R, ∞) → R+ \ {0}. Suppose also that

(8.13) generates a cocycle ϕ that is continuous and asymptotically compact. Finally,
suppose that (Ω, T, σ) is a minimal dynamical system.
Then the cocycle hH, ϕ, (Ω, T+ , σ)i has a uniform pullback attractor.

Proof. It follows by the chain rule applied to |u|2 for a solution of (8.13) that
|ϕ(t, u, ω)| < |u| for |u| > R, t > 0 and ω ∈ Ω. Hence, according to Theorem
5.5, the non-autonomous dynamical system h(X, T+ , π), (Ω, T+ , σ), hi (where X :=
H × Ω, h := pr2 and π := (ϕ, σ)) has a global attractor. On the other hand, by
(8.17),

|ϕ(t, u1 , ω) − ϕ(t, u2 , ω)| ≤ |u1 − u2 |

for t ≥ 0, ω ∈ Ω and u1 , u2 ∈ H. Theorem 8.11 then gives the result. 

The above theorem holds for a differential equation (8.13) on H = R with


 −(u + 1) + g(ω) : u < −1


F (ω, u) = g(ω) : |u| ≤ 1




−(u − 1) + g(ω) : u > 1.
where g ∈ C(Ω, R).
 Chapter 9

Pullback attractors of C-analytic systems

One of the most studied classes of nonlinear ODEs is the class of C - analytic
differential equations, i.e. the equations

dz
= f (t, z), (9.1)
dt

where the right hand side f is a holomorphic function with respect to complex
variable z ∈ Cd . Let ϕ(t, f, z) be a unique solution of equation (9.1) with initial
condition ϕ(0, f, z) = z and be defined on R+ . In virtue of fundamental theory
of ODEs with holomorphic right hand side (see, for example [122] and [186]) the
mapping ϕ possesses the following properties:

1. ϕ(0, f, z) = z.
2. ϕ(t + τ, f, z) = ϕ(t, fτ , ϕ(τ, f, z)) for every t, τ ∈ R+ and z ∈ Cd , where fτ is a
τ - translation of function f .
3. ϕ is continuous.
4. ϕ(t, f, ·) : Cd → Cd is holomorphic for every t and f .

The properties 1.-4. will be the basis of our research of abstract C−analytic
non-autonomous dynamical system.
The dissipative periodic equation (9.1) was studied by I.L.Zinchenko [334] and
he proved that in this case the equation (9.1) admits a unique periodic globally
uniformly asymptotically stable solution. This result was generalized for almost
periodic equations (9.1) by D.N.Cheban [62] and [84] (see also the Chapter 3). He
studied this problem within the framework of general C- analytic non-autonomous
dynamical systems.
In this chapter we study the structure of global pullback attractors of general
C- analytic cocycles with noncompact base (in terminology of equation (9.1): the
right hand side f is unbounded with respect to time t ∈ R).
This chapter is organized as follows. In section 1 we introduce the class of C−
analytic cocycle.

307
308 Global Attractors of Non-autonomous Dissipative Dynamical Systems

In section 2 we establish some general facts about non-autonomous dynamical

systems. We introduce the semigroup Eω+ , Eω− and Eω acting on the fiber Xω of
stratification (X, h, Ω). These semigroups are sub-semigroups of Ellis semigroup
(in the case of compact base Ω) and play an important role in the study of non-
autonomous dynamical system.
Section 3 is devoted to positively uniformly stable cocycles. For this class of
cocycles we prove that on every compact invariant set the corresponding cocycle
can be prolonged uniquely in the negative direction.
In section 4 we study the structure of compact global pullback attractor of C−
analytic cocycles with compact base. The main result in this section is Theorem 9.4
which states that for considered class of cocycles the pullback attractor {I ω | ω ∈ Ω}
is trivial, i.e. the section Iω contains a single point.
Section 5 is devoted to study of the uniform dissipative cocycles with noncompact
base. For this class of cocycles we prove the triviality of its global pullback attractor
(see Theorem 9.7).
In section 6 we introduce the class of cocycles possessing the property of dissi-
pativity (non-uniform) with non-compact base. The main result in this section is
Theorem 9.8 which describes the structure of compact pullback attractor of men-
tioned class of cocycles. In particular its triviality is proved.
Section 7 is devoted to application of our general results, obtained in sections
3-6 to study of differential equations (ODEs, Caratheodory equations with almost
periodic coefficients, almost periodic ODEs with impulse).
This paper is organized as follows. In section 1 we introduce the class of C−
analytic cocycle.
In section 2 we establish some general facts about non-autonomous dynamical
systems. We introduce the semigroup Eω+ , Eω− and Eω acting on the fiber Xω of
stratification (X, h, Ω). These semigroups are sub-semigroups of Ellis semigroup
(in the case of compact base Ω) and play an important role in the study of non-
autonomous dynamical system.
Section 3 is devoted to positively uniformly stable cocycles. For this class of
cocycles we prove that on every compact invariant set the corresponding cocycle
can be prolonged uniquely in the negative direction.
In section 4 we study the structure of compact global pullback attractor of C−
analytic cocycles with compact base. The main result in this section is Theorem 9.4
which states that for considered class of cocycles the pullback attractor {I ω | ω ∈ Ω}
is trivial, i.e. the section Iω contains a single point.
Section 5 is devoted to study of the uniform dissipative cocycles with noncompact
base. For this class of cocycles we prove the triviality of its global pullback attractor
(see Theorem 9.7).
In section 6 we introduce the class of cocycles possessing the property of dissi-
pativity (non-uniform) with non-compact base. The main result in this section is
 Pullback attractors of C-analytic systems 309

Theorem 9.8 which describes the structure of compact pullback attractor of men-
tioned class of cocycles. In particular its triviality is proved.
Section 7 is devoted to application of our general results, obtained in sections
3-6 to study of differential equations (ODEs, Caratheodory equations with almost
periodic coefficients, almost periodic ODEs with impulse).

9.1 C-analytic cocycles

Let Ω be a complete metric space, let T, the time set, be either R or Z, T+ =

{t ∈ T | t ≥ 0} (T− = {t ∈ T | t ≤ 0}), let (Ω, T, σ) be an autonomous
two-sided dynamical system on Ω and E d be a d-dimensional real (Rd ) or complex
(Cd ) Euclidean space with the norm | · |.
Denote by HC(Cd ×Ω, Cd ) the space of all the continuous functions f : Cd ×Ω →
C holomorphic in z ∈ Cd and equipped by compact-open topology. Consider the
d

differential equation

dz
= f (z, σt ω), (ω ∈ Ω) (9.2)
dt

where f ∈ HC(Cd × Ω, Cd ). Let ϕ(t, z, ω) be the solution of equation (9.2) passing

through the point z for t = 0 and defined on R+ . The mapping ϕ : R+ ×Cd ×Ω → Cd
has the following properties (see, for example, [122] and [186]):

a) ϕ(0, z, ω) = z for all z ∈ Cd .

b) ϕ(t + τ, z, ω) = ϕ(t, ϕ(τ, z, ω), στ ω) for all t, τ ∈ R+ , ω ∈ Ω and z ∈ Cd .
c) the mapping ϕ is continuous.
d) the mapping ϕ(t, ω) := ϕ(t, ·, ω) : Cd → Cd is holomorphic for any t ∈ R+ and
ω ∈ Ω.

Definition 9.1 The cocycle hCd , ϕ, (Ω, T, σ)i is called C-analytic if the mapping
ϕ(t, ω) : Cd → Cd is holomorphic for all t ∈ T+ and ω ∈ Ω.

Example 9.1 Let (HC(R × Cd , Cd ), R, σ) be a dynamical system of translations

on HC(R × Cd , Cd ) (Bebutov’s dynamical system (see, for example, [300])). Denote
by F the mapping from Cd × HC(R × Cd , Cd ) to Cd defined by equality F (z, f ) :=
f (0, z) for all z ∈ Cd and f ∈ HC(R × Cd , Cd ). Let Ω be the hull H(f ) of a given
function f ∈ HC(R × Cd , Cd ), that is Ω = H(f ) := {fτ |τ ∈ R}, where fτ (t, z) :=
f (t+τ, z) for all t, τ ∈ R and z ∈ Cd . Denote the restriction of (HC(R×Cd , Cd ), R, σ)
on Ω by (Ω, R, σ). Then, under appropriate restriction on the given function f ∈
HC(R × Cd , Cd ) defining Ω, the differential equation dz dt = f (t, z) = F (z, σt f )
generates a C−analytic cocycle.
310 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Remark 9.1 Analogously as above every difference equation with holomorphic

right hand side generates a C-analytic cocycle with discrete time Z + .

9.2 Some general facts about non-autonomous dynamical systems

Definition 9.2 The point ω ∈ Ω is called (see, for example, [302] and [304])
positively (negatively) stable in the sense of Poisson if there exists a sequence t n →
+∞ (tn → −∞ respectively) such that σtn ω → ω. If the point ω is Poisson stable
in both directions, in this case it is called Poisson stable.

Denote by Nω = {{tn } | σtn ω → ω}, N+

ω := {{tn } ∈ Nω | tn → +∞} and
−
Nω := {{tn } ∈ Nω |tn → −∞}.

Definition 9.3 (Conditional compactness). Let (X, h, Ω) be a fiber space, i.e. X

and Ω be two metric spaces and h : X → Ω be a homomorphism from X into Ω.
The subset M ⊆ X is said to be conditionally relatively compact, if the pre-image
T
h−1 (Ω0 ) M of every relatively compact subset Ω0 ⊆ Ω is a relatively compact
T
subset of X, in particularly Mω := h−1 (ω) M is relatively compact for every ω.
The set M is called conditionally compact if it is closed and conditionally relatively
compact.

Example 9.2 Let K be a compact space, X := K × Ω, h = pr2 : X → Ω, then

the triplet (X, h, Ω) be a fiber space, the space X is conditionally compact, but not
compact.

Let h(X, T+ , π), (Ω, T, σ), hi be a non-autonomous dynamical system and ω ∈ Ω

be a positively Poisson stable point. Denote by

Eω+ := {ξ| ∃{tn } ∈ N+

ω such that π tn |Xω → ξ},

where Xω := {x ∈ X| h(x) = ω} and → means the pointwise convergence.

Lemma 9.1 Let ω ∈ Ω be a positively Poisson stable point, h(X, T+ , π),

(Ω, T, σ), hi be a non-autonomous dynamical system and X be a conditionally com-
pact space, then Eω+ is a nonempty compact sub-semigroup of the semigroup XωXω
(w.r.t. composition of mappings).

Proof. Let {tn } ∈ N+

ω , then σtn ω → ω and, consequently, the set
[
Q := {π tn (Xω )|n ∈ N}

is compact, because X is conditionally compact. Thus {π tn |Xω } ⊆ QXω and ac-

cording to Tyhonov’s Theorem this sequence is relatively compact. Let ξ be a limit
point of {π tn |Xω }, then ξ ∈ Eω+ and, consequently, Eω+ 6= ∅.
 Pullback attractors of C-analytic systems 311

We note that Eω+ ⊆ XωXω and, consequently, Eω+ is relatively compact. Let now
ξ1 , ξ2 ∈ Eω+ , we will prove that ξ1 · ξ2 ∈ Eω+ . Since ξ1 , ξ2 ∈ Eω+ , then there are two
sequences {tin } ∈ N+ ω (i = 1, 2) such that
i
π t n |X ω → ξ i (i = 1, 2).

Denote by ξ := ξ1 ·ξ2 ∈ XωXω ⊆ QXω , then we have π tn ·ξ2 → ξ1 ·ξ2 = ξ as n → +∞.

Let Uξ ⊂ XωXω be an arbitrary open neighborhood of point ξ in XωXω , then from
relation (9.3) results that there exists a number n1 (ξ) ∈ N such that π tn · ξ2 ∈ Uξ
for all n ≥ n1 (ξ). Now we fix n ≥ n1 (ξ), then there exist an open neighborhood
1
Uπt1n ·ξ2 ⊂ Uξ of point π tn · ξ2 ∈ QXω and a number mn ∈ N such that
1 2
π tn · π tm |Xω ∈ Uπt1n · ξ2

for any n ≥ n1 (ξ) and m ≥ mn (ξ) and, consequently,

1 2
π t n · π t m |X ω ∈ U ξ
1 2
for any n ≥ n1 (ξ) and m ≥ mn (ξ). Thus from sequence {π tn +tm |Xω } it is possible to
1 2 1 2
extract a subsequence {π tnk +tmk |Xω } (t1nk +t2mk → +∞) such that π tnk +tmk |Xω →
ξ and, consequently, ξ = ξ1 · ξ2 ∈ Eω+ . The Lemma is proved. 

Corollary 9.1 Let ω ∈ Ω be a negatively Poisson stable point, h(X, T, π),

(Ω, T, σ), hi be a two-sided non-autonomous dynamical system and X be a con-
ditionally compact space, then Eω− = {ξ| ∃{tn } ∈ N−
ω such that π tn |Xω → ξ} is
Xω
a nonempty compact sub-semigroup of semigroup Xω .

Proof. This assertion follows from Lemma 9.1. 

Lemma 9.2 Let ω ∈ Ω be a two-sided Poisson stable point, h(X, T, π), (Ω, T, σ), hi
be a two-sided non-autonomous dynamical system and X be a conditionally com-
pact space, then Eω = {ξ| ∃{tn } ∈ Nω such that π tn |Xω → ξ} is a nonempty
compact sub-semigroup of the semigroup XωXω .

Proof. This assertion can be proved using the same type of arguments as well as
in the proof of Lemma 9.1 and therefore we omit the details. 

Corollary 9.2 Under the conditions of Lemma 9.2 Eω+ and Eω− are two nonempty
sub-semigroups of the semigroup Eω .

Lemma 9.3 Under the conditions of Lemma 9.2 the following assertions hold:
T
(1) if ξ1 ∈ Eω− and ξ2 ∈ Eω+ , then ξ1 · ξ2 ∈ Eω− Eω+ .
T
(2) Eω− Eω+ is a sub-semigroup of the semigroup Eω− , Eω+ and Eω .
312 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(3) Eω− · Eω ⊆ Eω− and Eω+ · Eω ⊆ Eω+ , where A1 · A2 := {ξ1 · ξ2 |ξi ∈ Ai (i = 1, 2)}
and Ai ⊆ Eω .
(4) if at least one of the sub-semigroups Eω− or Eω+ is a group, then Eω− = Eω+ = Eω .

Proof. Let ξ1 ∈ Eω− and ξ2 ∈ Eω+ , then there are t1n → −∞ and t2n → +∞ such that
i
σtin ω → ω and π tn |Xω → ξi (i = 1, 2). Using the same type arguments as well as in
the proof of Lemma 9.1 we may choose the subsequence {t1nk +t2mk } ⊂ {t1n +t2m } with
1 2
the following properties: a) t1nk +t2mk ≥ k or t1nk +t2mk ≤ −k and b) π tnk +tmk → ξ1 ·ξ2 ,
T
i.e. ξ1 · ξ2 ∈ Eω+ Eω− .
The second statement follows from the first one.
Let ξ1 ∈ Eω+ (Eω− , respectively) and ξ2 ∈ Eω , then there exist two sequences
i
tn → +∞ (or −∞, respectively) and t2n such that π tn → ξi (i = 1, 2). Then we
1

may choose a subsequence {t1nk + t2mk } with the following properties:

1 2
a)t1nk + t2mk ≥ k (≤ −k, respectively) and b)π tnk +tmk → ξ1 · ξ2 ,

and consequently, ξ1 · ξ2 ∈ Eω+ (Eω− , respectively).

Finally, let Eω− be a subgroup of the semigroup Eω . According to the third
statement of Lemma 9.3 Eω− · Eω ⊆ Eω− . Since Eω− is a nonempty compact invariant
set w.r.t. Eω , then in Eω− exists a compact minimal subset I ⊂ Eω− , i.e. I 6= ∅,
compact and u · Eω = I for every u ∈ I. Let now u ∈ I be an idempotent element of
right ideal I of semigroup Eω , then u is an unit element (u(x) = x ∀x ∈ Xω ) of I
because I ⊆ Eω− and Eω− according to conditions of Lemma 9.3 is a subgroup of the
semigroup Eω . Thus we have Eω = u · Eω = I ⊆ Eω− and, consequently, Eω− = Eω .
Analogously Eω+ = Eω . The theorem is proved. 

Lemma 9.4 Let ω ∈ Ω be a two-sided Poisson stable point, h(X, T, π), (Ω, T, σ), hi
be a two-sided non-autonomous dynamical system and X be a conditionally compact
space and

inf ρ(x1 tn , x2 tn ) > 0 (9.3)

n∈N

for all {tn } ∈ N−

ω and x1 , x2 ∈ Xω (x1 6= x2 ), then Eω− is a subgroup of the
semigroup Eω .

Proof. Indeed, if u ∈ Eω− is an arbitrary idempotent element of Eω− , then u2 = u

and there exists a sequence {tn } ∈ N− ω such that π
tn
→ u. According to (9.3) we
have u(x1 ) 6= u(x2 ) for all x1 6= x2 (x1 , x2 ∈ Xω ). On the other hand u2 (x) = u(x)
for all x ∈ Xω and, consequently, u(x) = x for all x ∈ Xω . Thus every idempotent
of semigroup Eω− is an unit element of Eω (in particular Eω− ) and, consequently,
Eω− is a group (see, for example [32]). 
 Pullback attractors of C-analytic systems 313

Lemma 9.5 Let ω ∈ Ω be a two-sided Poisson stable point, h(X, T, π), (Ω, T, σ), hi
be a two-sided non-autonomous dynamical system and X be a conditionally compact
space and the condition (9.3) holds for all {tn } ∈ N− ω and x1 , x2 ∈ Xω (x1 6= x2 ),
then inequality (9.3) is fulfilled for any {tn } ∈ N+
ω and x1 , x2 ∈ Xω (x1 6= x2 ).

Proof. According to Lemma 9.3 under the conditions of Lemma 9.5 we have
Eω− = Eω+ = Eω and Eω is a group. Suppose that for some sequence {tn } ∈ N+
ω

inf ρ(x1 tn , x2 tn ) = 0. (9.4)

n∈N

Since σtn ω → ω and the space X is conditionally compact, the sequence {π tn |Xω }
S
is relatively compact in QXω , where Q := {π tn (Xω )|n ∈ N} and, consequently,
we may assume that it is convergent. Let ξ := lim π tn |Xω , then from equality
n→+∞
(9.4) results that ξ(x1 ) = ξ(x2 ) (x1 6= x2 ), but ξ ∈ Eω and Eω is a group and,
consequently, ξ is a one-to-one mapping. The obtained contradiction proves our
assertion. 

Lemma 9.6 Let ω ∈ Ω be a positively Poisson stable point, h(X, T+ , π),

(Ω, T, σ), hi be a non-autonomous dynamical system, generated by cocycle ϕ ,
S t
pr1 ( π Xω ) be relatively compact and
t≥0

\[ \
Aω (Xω ) := ( π τ Xω ) Xω ,
t≥0

then for any x ∈ Aω (Xω ) there exists an entire trajectory of dynamical system
(X, T+ , π) passing through point x for t = 0 and pr1 (γ(T)) (γ(T) := {γ(t)|t ∈ T})
is relatively compact.

Proof. Let x ∈ Aω (Xω ), then there are {tn } ∈ Nω and xn ∈ Xω such that x =
lim π tn xn , σtn ω → ω and tn → +∞. We consider the sequence {γn } ⊂ C(T, M ),
n →∞
S t
where M := π Xω , defined by equality
t≥0

γn (t) = π t+tn xn , if t ≥ −tn and γn (t) = xn for t ≤ tn .

Now we will prove that the sequence {γn } is equicontinuous on every segment
[−l, l] ⊂ T. If we suppose that it is not true, then there exist ε0 , l0 > 0, tin ∈ [−l0 , l0 ]
and δn → 0 (δn > 0) such that

|t1n − t2n | ≤ δn and ρ(γn (t1n ), γn (t2n )) ≥ ε0 . (9.5)

We may suppose that tin → t0 (i = 1, 2). From (9.5) we obtain

1 2
ε0 ≤ ρ(γn (t1n ), γn (t2n )) = ρ(π tn +l0 (π tn −l0 xn ), π tn +l0 (π tn −l0 xn )) (9.6)
314 Global Attractors of Non-autonomous Dissipative Dynamical Systems

for sufficiently large n (tn ≥ l0 ). Note that the sequence {π tn −l0 xn } is rela-
tively compact and h(π tn −l0 xn ) = σtn −l0 h(xn ) = σtn −l0 ω → σ−l0 ω. Let x̄ =
lim π tn −l0 xn , then passing to limit in the inequality (9.6) we obtain ε0 ≤ 0. The
n→∞
obtained contradiction proves our assertion.
Now taking into account the conditional compactness of set K we can affirm
that {γn } is a relatively compact sequence of C(T, M ). Let γ be a limit point of
sequence {γn }, then there exists a subsequence {γkn } such that γ(t) = lim γkn (t)
n→∞
uniformly on every segment [−l, l] ⊂ T. In particular γ ∈ C(T, M ). We note that
π t γ(s) = lim π t γkn (s) = lim γkn (s + t) = γ(s + t) for all t ∈ T+ and s ∈ T.
n→∞ n→∞
Finally, we see that γ(0) = lim γkn (0) = lim π tkn xkn = x, i.e. γ is an entire
n→∞ n→∞
trajectory of dynamical system (X, T+ , π) passing through point x. The Lemma is
completely proved. 

9.3 Positively uniformly stable cocycles

Let E d be a d−dimensional real (Rd ) or complex (Cd ) Euclidean space with the
norm | · |, ρ be the distance generated by this norm, Ω be a metric space and the
triplet hE d , ϕ, (Ω, T, σ)i be a cocycle on the state space E d .

Theorem 9.1 Let hE d , ϕ, (Ω, T, σ)i be a cocycle with the following properties:

(1) It admits a conditionally relatively compact invariant set {I ω | ω ∈ Ω} (i.e.

S
{Iω | ω ∈ Ω0 } is relatively compact subset of E d for any relatively compact
subset Ω0 of Ω).
(2) The cocycle ϕ is positively uniformly stable on {Iω | ω ∈ Ω}.
S
Then all motions on J := {Jω | ω ∈ Ω} (Jω := Iω × {ω}) may be continued
uniquely to the left and define on J a two-sided dynamical system (J, T, π), i.e.
the skew-product system (X, T+ , π) generates on J a two-sided dynamical system
(J, T, π).

Proof. First step: we will prove that the set J ⊂ X is distal in the negative direction
w.r.t. the non-autonomous dynamical system h(X, T+ , π), (Ω, T, σ), hi, i.e. for all
ω ∈ Ω and u1 , u2 ∈ Iω (u1 6= u2 ) the following inequality holds

inf ρ(γ1 (t), γ2 (t)) > 0 (9.7)

t≤0

for all γi ∈ Φ(ui ,ω) (i = 1, 2), where by Φ(u,ω) it is denoted the family of all the entire
trajectories of (X, T+ , π) passing through point (u, ω) and belonging to J. If it is
not true, then there exist ω0 ∈ Ω, u0i ∈ Iω0 (u01 6= u02 ), γi0 ∈ Φ(u0i ,ω0 ) (i = 1, 2) and
 Pullback attractors of C-analytic systems 315

−tn → −∞ such that

ρ(γ10 (−tn ), γ20 (−tn )) → 0 (9.8)

as n → ∞. Let ε := ρ(u01 , u02 ) > 0 and δ = δ(ε) > 0 be chosen from positively
uniformly stability of cocycle ϕ on family of compact subsets {Iω | ω ∈ Ω}, then for
sufficiently large n from (9.8) we have ρ(γ10 (−tn ), γ20 (−tn )) < δ and, consequently,
ε = ρ(u01 , u02 ) = ρ(π tn γ10 (−tn ), π tn γ20 (−tn )) < ε. The obtained contradiction proves
our assertion.
Second step: we will prove that for any ω ∈ Ω and u ∈ Iω the set Φ(u,ω) contains
S
only one entire trajectory of (X, T+ , π) belonging to J. Let Φ := {Φ(u,ω) | (u, ω) ∈
J} ⊂ C(T, X), where C(T, X) is a space of all the continuous functions f : T → X
equipped with compact-open topology and (C(T, X), T, λ) is Bebutov’s dynamical
system (dynamical system of translations (see, for example, [292, 300])). It is easy
to verify that Φ is a closed and invariant subset of dynamical system (C(T, X), T, λ)
and, consequently, induces on the set Φ the dynamical system (Φ, T, λ). Let H be
a mapping from Φ into Ω, defined by equality H(γ) := h(γ(0)), then it is possible
to verify (see [83]) that the triplet h(Φ, T, λ), (Ω, T, σ), Hi is a non-autonomous
dynamical system. Now we will show that this non-autonomous dynamical system
is distal on the negative direction, i.e.

inf ρ(γ1t , γ2t ) > 0

t≤0

for all γ1 , γ2 ∈ H −1 (ω) (γ1 6= γ2 ) and ω ∈ Ω. Indeed, otherwise there exist

ω0 , γ1 , γ2 ∈ H −1 (ω0 ) (γ1 6= γ2 ) and tn → +∞ such that ρ(γ1−tn , γ2−tn ) → 0 (where
γ τ := σ(γ, τ ), i.e. γ τ (s) := γ(τ + t) for all s ∈ T) as n → ∞ and, consequently,

|γ1 (−tn ) − γ2 (−tn )| ≤ ρ(γ1−tn , γ2−tn ) → 0. (9.9)

Since γ1 6= γ2 , then there exists t0 ∈ T such that γ1 (t0 ) 6= γ2 (t0 ). Let γ̃i (t) :=
γi (t + t0 ) for all t ∈ T, then γ̃i ∈ Φω0 and from inequality (9.9) we have

|γ̃1 (−tn ) − γ̃2 (−tn )| → 0. (9.10)

as n → ∞, −tn −t0 → −∞. Thus we found ω0 := h(γi (t0 )) and ui := pr1 γi (t0 ) (i =
1, 2), u1 , u2 ∈ Iω0 (u1 6= u2 ) and the entire trajectories γ˜i ∈ Φ(ω,ui ) (i = 1, 2)
such that γ˜1 and γ˜2 are proximal (see (9.10)). But (9.10) and (9.7) are con-
tradictory. Thus the negative distality of the non-autonomous dynamical system
h(Φ, T, σ), (Ω, T, σ), Hi is proved.
Now we can prove that for any ω ∈ Ω and u ∈ Iω the set Φ(ω,u) contains a
unique entire trajectory. In fact, if it is not true, then there exists (ω0 , u0 ) ∈ Ω × E d
and two different trajectories γ1 , γ2 ∈ Φ(ω0 ,u0 ) (γ1 6= γ2 ). In virtue of above γ1 and
316 Global Attractors of Non-autonomous Dissipative Dynamical Systems

γ2 are negatively distal with respect to h(Φ, T, σ), (Ω, T, σ), Hi, i.e.

α(γ1 , γ2 ) := inf ρ(γ1t , γ2t ) > 0

t≤0

and, consequently, ρ(γ1 (t), γ2 (t)) ≥ α(γ1 , γ2 ) > 0 for all t ≥ 0. In particular
γ1 (0) 6= γ2 (0). The obtained contradiction proves our statement.
Third step: let now π̃ be a mapping from T × J into J defined by equality

π̃(t, x) = π(t, x) if t ≤ 0 and γx (t) if t<0

for all x ∈ J, where γx is a unique entire trajectory of the dynamical system

(X, T+ , π) passing through point x and belonging to J. To prove that (J, T, π̃) is
a two-sided dynamical system on J it is sufficient to verify the continuity of the
mapping π̃. Let x ∈ J, t ∈ T−, xn → x and tn → t, then there is a l0 > 0 such
that tn ∈ [−l0 , l0 ] and, consequently,

ρ(π̃(tn , xn ), π̃(t, x)) = ρ(π tn +l0 γxn (−l0 ), π t+l0 γx (−l0 )) ≤ (9.11)

ρ(π tn +l0 γxn (−l0 ), π tn +l0 γx (−l0 )) + ρ(π tn +l0 γx (−l0 ), π t+l0 γx (−l0 )).

Reasoning as in the proof of Lemma 9.2 it is possible to establish that the

sequence {γxn } is relatively compact in C(T, J) and that every limit point of this
sequence γ ∈ Φ and γ(0) = x. Taking into account the result of the second step we
claim that γxn → γx uniformly on every segment [−l, l] ⊂ T(l > 0). In particular,
γxn (−l0 ) → γx (−l0 ). Passing now to limit in inequality (9.11) when n → ∞ we
obtain the continuity of mapping π̃ in the point (t, x). The theorem is completely
proved. 

Remark 9.2 Theorem 9.1 is true and in the case if we replace the condition 2.
by the following: for arbitrary ε > 0 there exist two positive numbers δ(ε) and L(ε)
such that

ρ(ϕ(t, ω, u1 ), ϕ(t, ω, u2 )) < ε (9.12)

for all ω ∈ Ω, t ≥ L(ε) and u1 , u2 ∈ Iω with condition ρ(u1 , u2 ) < δ.

9.4 The compact global pullback attractors of C-analytic cocycles with

compact base

In this section we suppose that hCd , ϕ, (Ω, T, σ)i is a C−analytic cocycle and Ω is
a compact space.

Theorem 9.2 Let hCd , ϕ, (Ω, T, σ)i be a C−analytic cocycle admitting a compact
global pullback attractor {Iω | ω ∈ Ω}, then:
 Pullback attractors of C-analytic systems 317

S
(1) The compact invariant set J = {Jω | ω ∈ Ω} of the skew-product dynamical
system (X, T+ , π) (X := Cd × Ω, π := (ϕ, σ)) is asymptotically stable.
(2) There exists a positive number δ0 such that the cocycle ϕ is positively uniformly
S
stable on the compact set B[I, δ] := {B[Iω , δ] | ω ∈ Ω}, where B[Iω , δ] :=
{z ∈ Cd | ρ(z, Iω ) ≤ δ}, for all 0 < δ < δ0 .
(3) The skew-product dynamical system (X, T+ , π) generates on J a group
dynamical system (J, T, π).

Proof. Denote by X := Cd × Ω and by (X, T+ , π) the skew-product dynamical

S
system. Then under the conditions of the theorem the set J = {Jω |ω ∈ Ω} is
a nonempty compact invariant set and according to Theorem 4.1 [100] is asymp-
totically stable with respect to (X, T+ , π). In particular there exists a δ0 > 0 such
that the set B[J, δ0 ] := {x ∈ X | ρ(x, J) ≤ δ0 } is positively invariant. Since Ω
is compact and π t (u, ω) := (ϕ(t, u, ω), σt ω), then there exists a positive number
C = C(δ0 ) such that |ϕ(t, u, ω)| ≤ C for all ω ∈ Ω and u ∈ B[Iω , δ0 ]. Taking into
account the connectedness of set Iω (see, for example [126] and also Theorem 2.25)
according to Cauchy’s Theorem for all δ < δ0 there exists a positive number L(δ)
such that

|ϕ(t, ω, u1 ) − ϕ(t, ω, u2 )| ≤ L(δ)|u1 − u2 | (9.13)

for all ω ∈ Ω, t ∈ R+ and u1 , u2 ∈ B[Iω , δ]. It is easy to see that from inequality
(9.13) results the positively uniformly stability of set B[I, δ] for every 0 < δ < δ 0 .
S
Particularly the set I := {Iω | ω ∈ Ω} will be positively uniformly stable and to
finish the proof of Theorem it is sufficiently to apply Theorem 9.1 to our situation
for the skew-product system (X, T+ , π). The Theorem is completely proved. 

Definition 9.4 The cocycle hEd , ϕ, (Ω, T, σ)i is called linear (see, for example,
[6],[33] and [292]) if the mapping ϕ(t, ω) : E d → E d is linear for every t ∈ T+ and
ω ∈ Ω.

Theorem 9.3 Let hCd , ϕ, (Ω, T, σ)i be a linear cocycle, then the following condi-
tions are equivalent:

(1) 1. lim |ϕ(t, ω, u)| = 0 for all u ∈ E d and ω ∈ Ω.

t→+∞
(2) 2. There exist positive numbers N, ν such that |ϕ(t, ω, u)| ≤ N exp (−νt)|u| for
all t ∈ T+ , ω ∈ Ω and u ∈ E d .

Proof. This statement follows from Theorems 1.10 and 2.38. 

Theorem 9.4 Let hCd , ϕ, (Ω, T, σ)i be a C−analytic cocycle admitting a compact
pullback attractor {Iω | ω ∈ Ω}, and let every point ω ∈ Ω be positively Poisson
stable. Then the following assertions hold:
318 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(1) For every ω ∈ Ω the set Iω consists of a unique point ν(ω).

(2) ν(σt ω) = ϕ(t, ν(ω), ω) for all ω ∈ Ω and t ∈ T+ .
(3) The mapping ω → γ(ω) is continuous, where γ := (ν, IdΩ ).
(4) Every point γ(ω) is positively Poisson stable.
(5) The continuous invariant section ν is uniformly asymptotically stable, i.e.
(a) for arbitrary ε > 0 there exists δ(ε) > 0 such that ρ(z, ν(ω)) < δ implies
ρ(ϕ(t, z, ω), ν(σt ω)) < ε for all t ≥ 0 and ω ∈ Ω.
(b) There exists δ0 > 0 such that

lim ρ(ϕ(t, z, ω), ν(σt ω)) = 0

t→+∞

for all ω ∈ Ω and z with the condition ρ(z, ν(ω)) ≤ δ0 .

Proof. Under the conditions of Theorem 9.3 there exists a positive number δ0 such
that the set M := B[J, δ0 ] is a compact and positively invariant set of skew-product
S
system (X, T+ , π) (see the proof of Theorem 9.2), where B[I, δ0 ] := {B[Iω , δ0 ] ×
{ω} | ω ∈ Ω}. Denote by E = E(M, T+ , π) the Ellis semigroup of the dynamical
system (M, T+ , π), Eω := {ξ ∈ E | ξMω ⊆ Mω }, where Mω := {(u, ω)| (u, ω) ∈
M } and Eω+ := {ξ ∈ Eω | ∃{tn } ∈ N+ tn
ω such that π |Mω → ξ}. According to
Theorem 9.2 and Lemma 9.1 Eω+ is a nonempty compact sub-semigroup of the Ellis
semigroup E. Note that every mapping ξ ∈ Eω+ , which maps B[Iω , δ0 ] into Iω ,
is holomorphic because, according to Theorem 9.2, the convergence π tn |Mω → ξ
is uniform on Mω . Consider an idempotent v ∈ Eω+ , then v(v(u)) = v(u) for all
u ∈ Mω and, consequently, v(p) = p for every p ∈ v(Mω ) = v(Iω ). Since v is
holomorphic and v(Iω ) is a compact connected set, then [167] v(Iω ) contains only
one point ν(ω). On the other hand we have v(v(u)) = v(u) for all u ∈ Mω , i.e.
v(ν(ω)) = v(u). Thus there exists a sequence tn → +∞ such that

|ϕ(tn , u, ω) − ϕ(tn , ν(u), ω)| = 0 (9.14)

for all u ∈ Mω . Taking into account the positively uniformly stability of cocycle ϕ
from (9.14) we obtain the equality

|ϕ(t, u, ω) − ϕ(t, ν(u), ω)| = 0 (9.15)

for all u ∈ B(Iω , δ0 ) := {u ∈ Cd | ρ(u, Iω ) < δ0 }. Now we will prove that Iω = {ν(ω)}
for every ω ∈ Ω. Let 0 < δ < δ0 , u ∈ Iω and h ∈ Cd with condition |h| < δ, then
according to equality (9.15) we have

lim sup |ϕ(t, ω, u + h) − ϕ(t, ω, u)| = 0 (9.16)

t→+∞ |h|≤δ

for all ω ∈ Ω and u ∈ Iω . In virtue of Cauchy’s formula (see [42] and also [85])

U (t, (u, ω))w = (9.17)

 Pullback attractors of C-analytic systems 319

Z Z d
1 ϕ(t, ω, u + v) − ϕ(t, ω, u) X wk
·... · dv1 · ... · dvd ,
(2πi)d v1 · ... · vd vk
k=1
|v1 |= 2δ |vd |= 2δ

where U (t, (u, ω)) := ∂ϕ(t,ω,u)

∂u for all (u, ω) ∈ M and t ∈ T+ . From (9.16) and (9.17)
it follows that lim kU (t, (u, ω))k = 0 for all (u, ω) ∈ M. According to Theorem
t→+∞
9.3 there exist positive numbers N and α such that

kU (t, (u, ω))k ≤ N exp (−αt) (9.18)

for any (u, ω) ∈ M . Let now u1 , u2 ∈ Iω and ψ : [0, 1] → B(Iω , δ) be a continuously

differentiable function with properties: ψ(0) = u1 and ψ(1) = u2 . Consider the
function ∆(s) := ϕ(t, ψ(s), ω), then according to Lagrange’s formula we have

∆(1) − ∆(0) = ∆0 (τ ), (9.19)

where 0 < τ < 1. Hence from (9.18) and (9.19) we have

|ϕ(t, u1 , ω) − ϕ(t, u2 , ω)| ≤ N1 exp (−αt)|u1 − u2 | (9.20)

for all t ∈ T+ , ω ∈ Ω and u1 , u2 ∈ Mω , where N1 := N · m and m := max |ψ 0 (s)|.

0≤s≤1
To finish the proof of theorem it is sufficient to remark that according to Theorem
9.1 on set J there is defined a two-sided dynamical system (J, T, π) and, in partic-
ular, through every point u ∈ Iω passes a unique entire trajectory of cocycle ϕ, i.e.
the function ϕ(t, ω, u) (u ∈ Iω and ω ∈ Ω) is defined on T. If u1 6= u2 (u1 , u2 ∈ Iω ),
then from (9.20) follows that

|ϕ(−t, ω, u1 ) − ϕ(−t, ω, u2 )| ≥ N1 exp (αt)|u1 − u2 |

for all t ∈ T+ . But the trajectories ϕ(t, ω, ui ) ∈ Iσt ω (i = 1, 2; t ∈ T) are bounded

on T. The obtained contradiction proves our assertion. Thus we have Iω = {ν(ω)}.
Now it is easy to see that the mapping ω → γ(ω) is continuous, where γ =
(ν, IdΩ ), and π t γ(ω) = γ(σt ω) for all ω ∈ Ω and t ∈ T+ and, consequently, ν(σt ω) =
ϕ(t, ω, ν(ω)) for all ω ∈ Ω and t ∈ T+ .
Next we will note that every point γ(ω) is positively Poisson stable. Indeed,
let {tn } ∈ Nω , then π tn γ(ω) = γ(σtn ω) → γ(ω) and, consequently, {tn } ∈ Nγ(ω) .
Finally, the uniformly asymptotically stability of continuous and invariant section
ν results from Theorem 9.2. The theorem is completely proved. 

9.5 The uniform dissipative cocycles with noncompact base

Let Ω be a complete metric space (generally speaking noncompact), hE d , ϕ,

(Ω, T, σ)i be a cocycle on the state space E d and (X, T+ , π) be the correspond-
ing skew-product dynamical system, where X := E d × Ω and π := (ϕ, σ).
320 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Definition 9.5 The cocycle hE d , ϕ, (Ω, T, σ)i is said to be dissipative if for any
ω ∈ Ω there is a positive number rω such that

lim sup |ϕ(t, ω, u)| < rω

t→+∞

for all ω ∈ Ω and u ∈ E d , i.e. for all u ∈ E d and ω ∈ Ω there exists a positive
number L(ω, u) such that |ϕ(t, ω, u)| < rω for all t ≥ L(ω, u).

Definition 9.6 The cocycle hE d , ϕ, (Ω, T, σ)i is said to be uniformly dissipative

if there exists a positive number r (r is not depend upon ω ∈ Ω) such that for any
R > 0 there is a positive number L(R) such that |ϕ(t, ω, u)| < r for all ω ∈ Ω and
|u| ≤ R and t ≥ L(R).

Theorem 9.5 ([100, 127, 217]) Let hE d , ϕ, (Ω, T, σ)i be an uniformly dissipative
cocycle, then it admits a compact global pullback attractor {I ω | ω ∈ Ω} with |u| ≤ r
for all u ∈ Iω and ω ∈ Ω, where r is the positive number in definition 9.6.

Theorem 9.6 Let hCd , ϕ, (Ω, T, σ)i be a C−analytic uniformly dissipative cocycle,
then the following statements hold:

i) The cocycle ϕ admits a compact global pullback attractor {I ω | ω ∈ Ω} with

|u| ≤ r for all u ∈ Iω and ω ∈ Ω, where r is the positive number in condition
(9.23).
ii) For any R > 0 there exist positive constants C = C(R) and L(R) such that

ρ(ϕ(t, ω, u1 ), ϕ(t, ω, u2 )) ≤ Cρ(u1 , u2 ) (9.21)

for all t ≥ L(R), ω ∈ Ω and u1 , u2 ∈ Cd with the condition |ui | ≤ R (i = 1, 2).

iii) For arbitrary ε > 0 there exist L(ε) > 0 and δ(ε) > 0 such that

|ϕ(t, ω, u + h) − ϕ(t, ω, u))| < ε

for all t ≥ L(ε), u ∈ Iω , ω ∈ Ω and |h| < δ.

iv) The set J of (X, T+ , π) is negatively distal, i.e.

inf ρ(γ1 (t), γ2 (t)) > 0,

t≤0

where γi (i=1,2) is a entire trajectory passing through point (ui , ω) ∈ J, u1 6= u2

and γi (S) ⊆ J.
v) On the set J there is defined a two-sided dynamical system (J, T, π) generated
by skew-product system (X, T+ , π).

Proof. The first assertion of theorem results from Theorem 9.5. Let now R > 0
and R0 > R, according to uniformly dissipativity of cocycle ϕ there exists L(R 0 ) > 0
such that |ϕ(t, ω, u)| < r for all t ≥ L(R0 ), ω ∈ Ω and |u| ≤ R0 . In virtue of Cauchy’s
 Pullback attractors of C-analytic systems 321

formula for R < R0 there is a constant C(R) > 0 such that | ∂ϕ

∂u (t, u, ω)| ≤ C(R) for
all t ≥ L(R), ω ∈ Ω and |u| ≤ R and, consequently the inequality (9.21) holds.
The third assertion we will prove by method of contradiction. If it is not true,
then there exist ε0 > 0, δn → 0 (δn > 0), |hn | < δn (hn ∈ Cd ), tn ≥ n, ωn ∈ Ω
and un ∈ Iωn such that

|ϕ(tn , ωn , un + hn ) − ϕ(tn , ωn , un ))| ≥ ε0

Let now R > r and C(R), L(R) be positive constants figuring in the inequality
(9.24), then we have the following inequality

ε0 ≤ |ϕ(tn , ωn , un + hn ) − ϕ(tn , ωn , un ))| ≤ C(R)|hn | ≤ C(R)δn . (9.22)

Passing to limit in the inequality (9.22) as n → ∞ we obtain ε0 ≤ 0. The obtained

contradiction proves our assertion.
The fourth and fifth statements follow from Theorem 9.1 (see also Remark 9.2)
because from condition iii) results that for arbitrary ε > 0 there exist two positive
constants δ(ε) and L(ε) satisfying the inequality (9.12). The Theorem is completely
proved. 

Theorem 9.7 Let hCd , ϕ, (Ω, T, σ)i be a C−analytic uniformly dissipative cocycle
and every point ω ∈ Ω be positively Poisson stable, then:

1. The set Iω consists of only one point ν(ω) for every ω.

2. The mapping ω → γ(ω) is continuous, where γ := (ν, IdΩ ).
3. ν(σt ω) = ϕ(t, ν(ω), ω) for all ω ∈ Ω and t ∈ T+ .
4. lim ρ(ϕ(t, σ−t ω)z, ν(ω)) = 0 for every ω ∈ Ω uniformly with respect to z in
t→+∞
compact subsets of Cd .
5. Every point γ(ω) is positively Poisson stable.
6. lim ρ(ϕ(t, z, ω), ν(σt ω)) = 0 for all ω ∈ Ω and z ∈ Cd , i.e. every positive
t→+∞
semi trajectory ϕ(t, ω, z) is asymptotically Poisson stable in positive direction.

Proof. Let hCd , ϕ, (Ω, T, σ)i be a C−analytic uniformly dissipative cocycle, then
according to Theorem 9.6 this cocycle has the properties i)-v). Let {Iω | ω ∈ Ω}
be the compact global pullback attractor of cocycle ϕ and let (X, T+ , π) be the
skew-product dynamical system. Denote by

Eω+ := {ξ| ∃{tn } ∈ N+ tn

ω , π |Xω → ξ}.

Since the cocycle ϕ possesses the property ii), then the pointwise convergence
π tn |Xω → ξ coincides with uniform convergence on every compact subset of
Xω := Cd × {ω} and, consequently, every mapping ξ ∈ Eω+ is holomorphic. As well
as in Lemma 9.1 it is possible to show that Eω+ is a nonempty compact semigroup
w.r.t. composition of mappings. Consider the idempotent element v of semigroup
322 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Eω+ . We will show that v(Xω ) ⊆ Iω . Indeed, v ∈ Eω+ and, consequently, there exists
a sequence {tn } ∈ N+ tn
ω such that v = lim π |Xω . Let x̄ ∈ v(Xω ), i.e. x̄ = v(x)
n→∞
for some x ∈ Xω . This means that x̄ = lim π tn x. According to Lemma 9.6 there
n→∞
exists an entire trajectory γ of the skew-product system (X, T+ , π) passing through
the point x̄ for t = 0 and γ(T) := {γ(t) | t ∈ T} is conditionally relatively compact.
Taking into account that J is a maximal invariant set of (X, T+ , π) with relatively
compact pr1 J we have x̄ ∈ Jω , i.e. v(Xω ) ⊆ Jω . Since Xω = Cd × {ω} and v is
holomorphic by virtue of Liouville’s Theorem the holomorphic function v is a con-
stant, i.e. there exists γ(ω) ∈ Jω such that v(Xω ) = {γ(ω)}. We note that v 2 = v
and, consequently, v(v(x)) = v(x) for all x ∈ Xω , i.e. v(γ(ω)) = v(x). Thus, there
exists a sequence tn → +∞ such that

lim ρ(π tn γ(ω), π tn x) = 0. (9.23)

t→+∞

Taking into consideration the property ii) of cocycle ϕ we obtain from (9.23) the
equality

lim ρ(ϕ(t, z, ω), ϕ(t, ν(ω), ω)) = 0 (9.24)

n→∞

for all z ∈ Cd , where γ := (ν, IdΩ ).

Now we will show that there exists δ0 > 0 such that for arbitrary ε > 0 there is
L(ε) > 0 with the property

|ϕ(t, ω, u + h) − ϕ(t, ω, u)| < ε (9.25)

for all (u, ω) ∈ J, t ≥ L(ε) and uniformly w.r.t. |h| ≤ δ0 . If it is not true, then there
are δ → +0, ε0 > 0, |hn | ≤ δn , ωn ∈ Ω, un ∈ Iωn and tn ≥ n such that

|ϕ(tn , ωn , un + hn ) − ϕ(tn , ωn , un )| ≥ ε0 .

On the other hand, according to property ii), there exists C(R) > 0 (R > sup δn )
n∈N
such that for sufficiently large n we have

ε0 ≤ |ϕ(tn , ωn , un + hn ) − ϕ(tn , ωn , un )| ≤ C(R)δn . (9.26)

Taking into account that δn → 0 from (9.26) it follows that ε0 ≤ 0. The obtained
contradiction prove our assertion.
From equality (9.17) and inequality (9.25) it follows that

lim kU (t, (u, ω))k = 0 (9.27)

t→+∞

uniformly with respect to (u, ω) ∈ B[J, δ0 ]. Denote by

m(t) := sup{kU (t, (u, ω))k |(u, ω) ∈ B[J, δ0 ]},

then
 Pullback attractors of C-analytic systems 323

a) m(t) → 0 as t → +∞.
b) ∃L > 0 such that m is bounded on [L, +∞).
c) m(t + τ ) ≤ m(t)m(τ ) for all t, τ ≥ L.

From a)-c) it follows (see, for example, [91]) that there exist N, α > 0 such that
m(t) ≤ N exp (−αt) for all t ≥ L and, consequently, from (9.27) we have

kU (t, (u, ω))k ≤ N exp (−αt)

for all (u, ω) ∈ B[J, δ0 ] and t ≥ L. Using the same arguments as well as as in the
proof of Theorem 9.4 we conclude that Iω = {ν(ω)} for all ω ∈ Ω.
Now we will prove that the mapping ω → γ(ω) is continuous. Let ωn → ω and
consider the sequence {γ(ωn )} ⊂ J. Since J is conditionally compact, this sequence
is relatively compact. Let x̄ be a limit point of {γ(ωn )}, then it is easy to see
that x̄ ∈ Jω = {γ(ω)} and, consequently, γ(ω) is a unique limit point of relatively
compact sequence {γ(ωn )}. Hence γ(ωn ) → γ(ω).
The equality ν(σt ω) = ϕ(t, ν(ω), ω) follows from invariance of J and from equal-
ity Jω = {γ(ω)} for all ω ∈ Ω, taking into account that ν(ω) = pr1 γ(ω).
The equality 4. follows from equality Jω = {γ(ω)} = {(ν(ω), ω)} and from the
fact that {ν(ω) | ω ∈ Ω} is a compact global pullback attractor of cocycle ϕ.
The stability in the sense of Poisson in the positive direction of point γ(ω) follows
from the continuity of γ and the equality π t γ(ω) = γ(σt ω) for all t ∈ T+ and ω ∈ Ω.
The sixth assertion follows from (9.24) and the equality ϕ(t, ν(ω), ω) = ν(σt ω)
for all t ∈ T+ and ω ∈ Ω. The Theorem is completely proved. 

9.6 The compact and local dissipative cocycles with noncompact base

Definition 9.7 The cocycle hE d , ϕ, (Ω, T, σ)i is said to be compact dissipative

if for any nonempty compact Ω0 ⊂ Ω there is a positive number rΩ0 such that for
arbitrary R > 0 there exists a positive number L(R, Ω0 ) with the following property

|ϕ(t, ω, u)| < rΩ0 (9.28)

for all ω ∈ Ω0 , |u| ≤ R and t ≥ L(R, Ω0 ).

Definition 9.8 The cocycle hE d , ϕ, (Ω, T, σ)i is said to be locally dissipative if

for any ω ∈ Ω and R > 0 there are positive numbers rω , δω and L(R, ω) such that

|ϕ(t, ω̃, u)| < rΩ0 (9.29)

for all ω̃ ∈ B(ω, δω ) := {ω̃ ∈ Ω | ρ(ω̃, ω) < δω }, |u| ≤ R and t ≥ L(R, ω).

Lemma 9.7 Every locally dissipative cocycle ϕ is compact dissipative.

324 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Proof. Suppose that ϕ is locally dissipative and let Ω0 ⊂ Ω be a nonempty compact

set. According to locally dissipativity of ϕ for every ω ∈ Ω, and R > 0 there exist
rω , L(R, ω) > 0 and δω > 0 such that the inequality (9.29) holds. Considering the
S
open covering {B(ω, δω ) | ω ∈ Ω0 } of compact set Ω0 , we may extract the finite
S
sub covering {B(ωi , δωi ) | i = 1, k}. Let L(R, Ω0 ) := max{L(R, ωi ) | i = 1, k},
then it is clear that inequality (9.28) holds for all ω ∈ Ω0 , |u| ≤ R and t ≥ L(R, Ω).
The Lemma is proved. 

Remark 9.3 Compact dissipativity, generally speaking, does not imply locally
dissipativity.

Lemma 9.8 Let hE d , ϕ, (Ω, T, σ)i be a compact dissipative C−analytic cocycle.

Then for any nonempty compact Ω0 ⊂ Ω and R > 0 there exist L(Ω0 , R) and
C = C(Ω0 , R) > 0 such that

ρ(ϕ(t, u1 , ω), ϕ(t, u2 , ω)) ≤ C(Ω0 , R)ρ(u1 , u2 ) (9.30)

for any ω ∈ Ω0 , |ui | ≤ R (i = 1, 2) and t ≥ L(Ω0 , R).

Proof. Let R > 0 and R0 > R, then according to the compact dissipativity of
cocycle ϕ for nonempty compact Ω0 ⊂ Ω and R0 > 0 there exist rΩ0 > 0 and
L(Ω0 , R0 ) > 0 such that |ϕ(t, u, ω)| < rΩ0 for all ω ∈ Ω, |u| ≤ R0 and t ≥ L(Ω0 , R0 ).
In view of Cauchy‘s formula for R < R0 there exists a constant C = C(R, Ω) >
0 such that | ∂ϕ 0 0 0 0
∂u (t, u, ω)| ≤ C(R, Ω ) for all t ≥ L(R , Ω ), ω ∈ Ω and |u| ≤ R
and, consequently, the inequality (9.30) holds for |u1 |, |u2 | ≤ R, ω ∈ Ω0 and t ≥
L(Ω0 , R) := inf{L(Ω0 , R0 ) | R0 > R}. 

Lemma 9.9 Let hE d , ϕ, (Ω, T, σ)i be a compact dissipative C−analytic cocycle,

and γ1 , γ2 are two entire bounded trajectories passing through point (u1 , ω) and
(u2 , ω) for t = 0 respectively, then the following assertions hold:

(1) If ω ∈ Ω is negatively Poisson stable and {tn } ∈ N−

ω , then

inf ρ(γ1 (tn ), γ2 (tn )) > 0 (9.31)

n∈N

if u1 6= u2 .
(2) If ω ∈ Ω is positively Poisson stable and {tn } ∈ N+
ω , then the equality

inf ρ(ϕ(tn , ω, u1 ), ϕ(tn , ω, u2 )) = 0 (9.32)

n∈N

implies

lim ρ(ϕ(t, ω, u1 ), ϕ(t, ω, u2 )) = 0. (9.33)

t→+∞
 Pullback attractors of C-analytic systems 325

Proof. Let ω ∈ Ω be a negative Poisson stable point, {tn } ∈ N− ω and u1 6= u2 . If the

equality (9.31) is not true, then ρ(γ1 (tn ), γ2 (tn )) → 0 as n → ∞. Denote by R :=
sup max{|γ1 (t)|, |γ2 (t)|} > 0, Ω0 := {σtn ω | n ∈ N} ⊂ Ω and let C(Ω0 , R), L(Ω0 , R)
t∈T
be the corresponding constants figuring in the inequality (9.30), then we obtain

ρ(u1 , u2 ) = ρ(ϕ(−tn , γ1 (tn ), σtn ω), ϕ(−tn , γ2 (tn ), σtn ω)) (9.34)
0
≤ C(Ω , R)ρ(γ1 (tn ), γ2 (tn ))

for sufficiently large n (−tn ≥ L(Ω0 , R)). Passing to limit in the inequality (9.34)
as n → ∞ we have ρ(u1 , u2 ) ≤ 0, but u1 6= u2 . The obtained contradiction prove
the first statement of Lemma 9.9 .
Let now ω ∈ Ω be a positive Poisson stable point and {tn } ∈ N+ ω such that the
inequality (9.33) holds, then for arbitrary ε > 0 there exists n0 = n0 (ε) such that
ε
ρ(ϕ(t, u1 , ω), ϕ(t, u2 , ω)) < (9.35)
2C(Ω0 , R)
and, consequently, according to Lemma 9.8 we obtain

ρ(ϕ(t, u1 , ω), ϕ(t, u2 , ω)) = ρ(ϕ(t − tn , ϕ( tn , u1 , ω), σtn ω), (9.36)

0
ϕ(t − tn , ϕ(tn , u2 , ω), σtn ω) ≤ C(Ω , R)ρ(ϕ(tn , u1 , ω), ϕ(tn , u2 , ω))

for all t ≥ tn + L(Ω0 , R), where R := sup{max{|ϕ(t, u1 , ω)|, |ϕ(t, u2 , ω)|} | t ∈ T+ }.

Denote by L(ε) := tn0 (ε) + L(Ω, R), then from inequalities (9.35) and (9.36) we
obtain

ρ(ϕ(t, u1 , ω), ϕ(t, u2 , ω)) < ε

for all t ≥ L(ε) and, consequently, (9.33) holds. The lemma is proved. 

Theorem 9.8 Let hE d , ϕ, (Ω, T, σ)i be a compact dissipative C−analytic cocycle

admitting a compact pullback attractor {Iω | ω ∈ Ω} and every point ω ∈ Ω be
two-sided Poisson stable, then the following assertions hold:

1. The set Iω consists of only one point ν(ω), i.e. Iω = {ν(ω)} for every ω ∈ Ω.
2. The mapping ω → γ(ω) is continuous, where γ = (ν, IdΩ ).
3. ν(σt ω) = ϕ(t, ν(ω), ω) for all ω ∈ Ω and t ∈ T + .
4. The point γ(ω) is Poisson’s stable for all ω ∈ Ω.
5. lim β(ϕ(t, σ−t ω)K, ν(ω)) = 0 for all compact subsets K of Cd , where β(A, B)
t→+∞
is the semi-distance of Hausdorff between A and B.
6. lim |ϕ(t, ω, z) − ν(σt ω)| = 0 for all ω ∈ Ω and z ∈ Cd .
t→+∞

Proof. To prove the first assertion of Theorem 9.8 we consider the non-autonomous
dynamical system h(Φ, T, λ), (Ω, T, σ), Hi constructed in the proof of Theorem 9.1.
Using the same type of argument as in Theorem 9.1 and taking into consideration
326 Global Attractors of Non-autonomous Dissipative Dynamical Systems

the Lemma 9.9 we may state that the system h(Φ, T, λ), (Ω, T, σ), Hi possesses the
following properties:

a) Φ is conditionally compact invariant set.

b) for every ω ∈ Ω, γ1 , γ2 ∈ ϕω (γ1 6= γ2 ) and {tn } ∈ N−
ω holds

inf ρ(γ1tn , γ2tn ) > 0,

n∈N

where γ t is a t-translation of γ, i.e. γ t (s) := γ(t + s) for all s ∈ T.

Then according to Lemmas 9.3–9.5 we have that Eω− = Eω+ = Eω is a group.

Particularly there are two sequences t1n → +∞ and t2n → −∞ such that
i
lim γ tn = γ (i = 1, 2). (9.37)
n→∞

for all γ ∈ Φω , i.e. every entire trajectory γ of global pullback attractor {Iω | ω ∈ Ω}
is two-sided Poisson stable. On the other hand according to Lemma 9.9 we have

lim ρ(γ1 (t), γ2 (t)) = 0. (9.38)

t→+∞

From (9.37) and (9.38) we obtain

ρ(γ1 (t), γ2 (t)) = lim ρ(γ1 (t + t1n ), γ2 (t + t1n )) = 0

n→∞

for all t ∈ T, i.e. γ1 = γ2 . Thus there exists a unique entire trajectory γ̃ω ∈ Φω , i.e.
Φω = {γ̃ω } and, consequently, Iω = {γ̃ω (0)} := {γ(ω)}, where γ(ω) := γ̃ω (0).
The proof of item 2.– 6. of Theorem 9.8 uses the same type of arguments as in
Theorem 9.7. The Theorem is proved. 

9.7 Applications

9.7.1 ODEs
Consider the differential equation
dz
= f (t, z), (9.39)
dt
where f ∈ CH(R × Cd , Cd ) and the family of equations
dz
= g(t, z), (9.40)
dt
where g ∈ H(f ) := {fτ |τ ∈ R} and fτ is a τ −translation of function f w.r.t.
variable t, i.e. fτ (t, z) := f (t + τ, z) for all t ∈ R and z ∈ Cd . Denote by ϕ(t, z, g)
the solution of equation (9.40) with the initial condition ϕ(0, z, g) = z, then ϕ is a
C−analytic cocycle on Cd .
 Pullback attractors of C-analytic systems 327

Definition 9.9 The equation (9.39) is called dissipative if there exists a positive
number r such that lim sup |ϕ(t, z, g)| < r for all z ∈ Cd and g ∈ H(f ).
t→+∞

Definition 9.10 The function f ∈ CH(R × Cd , Cd ) is called positively (neg-

atively) Poisson stable in t ∈ R uniformly w.r.t. z on compact subsets of Cd
[292],[300] if there exists tn → +∞ (tn → −∞, respectively) such that f (t+tn, z) →
f (t, z) as n → ∞ uniformly on every compact subsets of R × Cd .

Theorem 9.9 Suppose that the following conditions hold:

(1) The set H(f ) ⊂ CH(R × Cd , Cd ) is compact.

(2) Every function g ∈ H(f ) is positively Poisson stable (in this case function f is
called [300],[304] quasi recurrent).
(3) The equation (9.39) is dissipative.

Then every equation (9.40) admits a unique bounded on R and positively Poisson
stable solution. This solution is globally uniformly asymptotically stable.

Proof. We consider the dynamical system of translations (Bebutov’s system

[292],[300]) (CH(R × Cd , Cd ), R, σ). Since H(f ) is invariant and closed subset of
CH(R × Cd , Cd ), then on set H(f ) it is induced a dynamical system (H(f ), R, σ).
Let Ω := H(f ), then on space Cd it is defined a C−analytic cocycle hCd , ϕ, (Ω, R, σ)i,
generated by equation (9.39). According to the general properties of equation (9.39)
with holomorphic right hand side f , the cocycle ϕ will be C−analytic (see, for
example, [122]). To finish the proof of this theorem it is sufficient to apply Theorem
9.4. 

Definition 9.11 (Bohr’s almost periodic function) The function f ∈ CH(R ×

Cd , Cd ) is called almost periodic (in the sense of Bohr) in t ∈ R uniformly w.r.t. z
on compact subsets of Cd [238, 292, 300] if for every ε > 0 and nonempty compact
subset K ⊂ Cd the set

T(ε, f, K) := {τ ∈ R| max |f (t + τ, z) − f (t, z)| < ε}

z∈K

is relatively dense on R, i.e. there is a number l = l(ε, f, K) > 0 such that

\
T(ε, f, K) [a, a + l] 6= ∅

for all a ∈ R.

Definition 9.12 The equation (9.39) is called pullback dissipative if for every
g ∈ H(f ) there exists a positive number rg such that for all R > 0

lim sup |ϕ(t, g−t , z)| < rg

t→+∞
328 Global Attractors of Non-autonomous Dissipative Dynamical Systems

uniformly w.r.t. |z| ≤ R.

Theorem 9.10 Let f ∈ CH(R × Cd , Cd ) be an almost periodic function in t ∈ R

uniformly w.r.t. z on compact subsets of Cd and the equation (9.39) be pullback
dissipative, then every equation (9.40) admits a unique bounded solution ν g (t), which
is almost periodic and satisfies the following conditions:

(1) νg (t) is uniformly asymptotically stable (locally).

(2) lim sup |ϕ(t, ω−t )z − νg (0)| = 0.
t→+∞ |z|≤R

Proof. The proof of this theorem uses the same type of argument as the proof of
Theorem 9.9 and is based on the Theorem 9.3. 

9.7.2 Caratheodory differential equations

Consider now the equation (9.39) with right hand side f satisfying the conditions
of Caratheodory (see, for example,[292] ) and holomorphic w.r.t. variable z ∈ Cd .
The space of all the Carateodory functions we denote by CH(R × Cd , Cd ). The
topology on this space is defined by family of semi-norm (see [292])
Z n
pn,m (f ) := max |f (t, z)|dt.
−n |z|≤m

This space is metrizable and on CH(R×Cd , Cd ) the dynamical system of translations

(CH(R × Cd , Cd ), R, σ) can be defined.
Using the standard arguments for ODEs (see,for example, [122] and [186]) one
can prove that every equation (9.40) admits a unique solution ϕ(t, z, g) with initial
condition ϕ(0, z, g) = z and supplementary the mapping ϕ(t, g) := ϕ(t, ·, g) : Cd →
Cd is holomorphic. Thus if the solutions ϕ(t, z, g) are defined on R+ , the mapping
ϕ : R+ × Cd × H(f ) → Cd defines a C−analytic cocycle on Cd with the base
H(f ), where H(f ) := {fτ |τ ∈ R} and the bar denotes the closure in the space
CH(R × Cd , Cd ). Hence we may apply the general results from sections 1.–6. to
cocycle ϕ, generated by equation (9.39) with Caratheodory’s right hand side, and
we will obtain some results for this type of equations. For example the following
assertion holds.

Theorem 9.11 Let f ∈ CH(R × Cd , Cd ) be an almost periodic function in t ∈ R

(in the sense of Stepanov [238]) uniformly w.r.t. z on compact subsets of Cd , i.e.
for every ε > 0 and compact subset K ⊂ Cd the set
Z 1
T(ε, f, K) := {τ ∈ R| max |f (t + τ + s, z) − f (t + s, z)|ds < ε}
0 x∈K
 Pullback attractors of C-analytic systems 329

is relatively dense on R. Suppose that the equation (9.39) is dissipative, then every
equation (9.40) admits a unique almost periodic (in the sense of Bohr) solution
νg (t) which is globally uniformly asymptotically stable.

9.7.3 ODEs with impulses

Let {tn }n∈Z be a sequence of real numbers, inf{tn+1 − tn | n ∈ Z} > 0, p : R → Cd
be a continuously differentiable function on every interval (tn , tn+1 ), continuous to
the right in every point t = tn , almost periodic in the sense of Stepanov and
X
p0 (t) = s n δ tn ,
n∈Z

where sn := p(tn + 0) − p(tn − 0) (i.e. the function p is piecewise constant). More

information about the function described above can be found in the books [170] and
[279].
Consider the equation with impulses
dz X
= f (t, z) + s n δ tn (9.41)
dt
n∈Z

or equivalently
dz
= f (t, z) + p0 (t). (9.42)
dt
At the same time we consider the family of equations
dz
= g(t, z) + q 0 (t), (9.43)
dt

where (g, q) ∈ H(f, p) := {(fτ , pτ )|τ ∈ R} and by the bar we denote the closure in
the product-space CH(R × Cd , Cd ) × C(R, Cd ).
Denote by ϕ(t, z, g, q) the unique solution of equation (9.43) (see [170] and [279])
satisfying the initial condition ϕ(0, z, g, q) = z. This solution is continuous on every
interval (tn , tn+1 ) and continuous to the right in every point t = tn (see [170] and
[279]).

Definition 9.13 The equation (9.41) is called dissipative if there exists a number
r > 0 such that lim sup |ϕ(t, z, g, q)| < r for every (g, q) ∈ H(f, p) and z ∈ Cd .
t→+∞

Using the transformation w := z + q(t) we can transform the equation (9.43)

into the equation
dw
= g(t, w + q(t)). (9.44)
dt
330 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Remark 9.4 Denote by ϕ̃(t, z, g, q) the cocycle defined by the family of equations
(9.44), then it is clear that the cocycle ϕ generated by (9.42) is dissipative if and
only if the cocycle ϕ̃ generated by (9.44) is dissipative.

Theorem 9.12 Let f ∈ CH(R × Cd , Cd ) be a Bohr’s almost periodic function in

t ∈ R uniformly with respect to z on every compacts subsets of Cd , locally Lipshitz
in z uniformly w.r.t. t ∈ T and p ∈ C(R, Cd ) be a Stepanov almost periodic function
bounded on R. If the equation (9.42) is dissipative, then for every (g, q) ∈ H(f, p)
the equation (9.43) admits a unique Stepanov almost periodic solution and this so-
lution is globally uniformly asymptotically stable .

Proof. Let ϕ(t, z, g, q) be the cocycle generated by equation (9.42) and let
ϕ̃(t, w, g, q) be the cocycle generated by equation (9.44). Then we have the following
equality

ϕ(t, z, g, q) = q(t) + ϕ̃(t, z − q(0), g, q). (9.45)

Under the conditions of Theorem 9.12 the cocycle ϕ̃ is dissipative, C−analytic and
the right hand side of equation (9.44) under the conditions of Theorem 9.12 is
Stepanov almost periodic in t ∈ R uniformly on every compact of Cd w.r.t. z. To
finish the proof of this theorem we apply the Theorem 9.11 to the equation (9.44)
and take into consideration the relation (9.45). The theorem is proved. 
 Chapter 10

Pullback attractors under discretization

A defining characteristic of an autonomous dynamical system is its dependence

on the time that has elapsed only and not on the absolute time itself. Conse-
quently, limiting objects, such as attractors, actually exist for all time as invari-
ant sets under the evolution of the autonomous system. Although such concepts
can also be used for general non-autonomous systems, where the absolute start-
ing time is as important as the time elapsed since starting, they are often too
restrictive and exclude many interesting types of dynamical behaviour. A simple
example is that of an asymptotically stable solution that is neither constant nor
periodic. What are the limiting or attracting points here? What are the corre-
sponding invariant sets, and, important for numerical considerations, how can one
assure convergence to a particular point in such an invariant set? The forwards
running convergence of an asymptotically stable solution is of little direct use in
constructing the limiting solution since this solution may itself be changing with
increasing time. An alternative is to use pullback convergence, that is to hold fixed
an absolute time instant and to consider the limiting values at this time instant of
other solutions that start progressively early in absolute time. The limiting sets now
depend on absolute time and are invariant under the evolution of the system, that
is, are carried forward onto each other as time increases. This idea was introduced
several years ago in the context of random dynamical systems [126, 153, 154, 284],
which are intrinsically non-autonomous, but had been used already in the 1960s by
M.Krasnoselskii [223] to establish the existence of solutions of deterministic systems
that are bounded over the entire time axis. It has also been applied recently [217,
219] to investigate variable time-step (hence non-autonomous) numerical approxi-
mations of global attractors of autonomous systems governed by dissipative ordinary
differential equations.
Another key idea in [126, 153, 154, 284] is to formulate the non-autonomous dy-
namics on Rd in terms of a cocycle mapping ϕ that is driven by an underlying
autonomous system σ on some parameter set Ω. At its simplest, Ω is just the abso-
lute time set R and σ is the shift operator that essentially resets the starting time

331
332 Global Attractors of Non-autonomous Dissipative Dynamical Systems

to the current absolute time value. More useful is to consider for Ω a function space
of admissible vector fields as proposed by G. R. Sell [292] or as a probability sample
space as in [126, 153, 154, 284], where the current parameter value takes the role of
absolute time and is adjusted by σ with the passage of time. The advantage here
is, in the first case at least, that the parameter space can be topologized (often as
a compact space) and the product system (σ, ϕ) is an autonomous semi–dynamical
system known as skew product flow on the new product state space Ω × Rd . The
extensive theory of autonomous dynamical systems can then be applied to such
skew product flows, in particular concepts such as invariant sets, limit sets and at-
tractors, but just how these manifest themselves in terms of the original dynamics
on the original state space Rd and what relationship, if any, they have with pullback
convergence need to be clarified.
In this chapter we investigate the effect of time discretization on the pullback
attractor of a non-autonomous ordinary differential equation for which the vector
fields depend on a parameter that varies in time rather than depending directly on
time itself. The parameter space is assumed to be compact so the skew product
flow formalism as well as cocycle formalism also applies and the vector fields have a
strong dissipative structure that implies the existence of a compact set that absorbs
all compact sets under the resulting non-autonomous dynamics. The numerical
scheme considered is a general 1–step scheme such as the Euler scheme with variable
time-steps. Our main result is to show that the numerical scheme interpreted as
a discrete time non-autonomous dynamical system, hence discrete time cocycle
mapping and skew product flow on an extended parameter space, also possesses a
cocycle attractor and that its component subsets converge upper semi–continuously
to those of the cocycle attractor of the original system governed by the differential
equation. This is a non-autonomous analogue of a result of P. E. Kloeden and
J. Lorenz [216] on the discretization of an autonomous attractor; see also [179,
308]. We will also see that the corresponding skew product flow systems have
global attractors with the cocycle attractor component sets as their cross-sectional
sets in the original state space Rd . Finally, we investigate the periodicity and almost
periodicity of the discretized pullback attractor when the parameter dynamics in
the ordinary differential equation is periodic or almost periodic and the pullback
attractor consists of singleton valued component sets, i.e. the pullback attractor is
a single trajectory.
The chapter is organized as follows. Pullback attractors, cocycles and skew prod-
uct flows are defined in Section 1 and a theorem is stated, summarizing results from
the literature on the relationship between pullback attractors and global attractors
of skew product flows. The class of non-autonomous differential equations and the
corresponding variable time-step 1–step schemes to be considered are introduced in
Section 2 and their cocycle formalism is then established in Section 3. The main
result, Theorem 10.2, is formulated and proved in Section 4. Section 5 is devoted
 Pullback attractors under discretization 333

to periodic and almost periodic behaviour when the pullback attractor is a single
trajectory. Finally, the section 6 contains the proof of a lemma used earlier, which
compares the cocycle mappings of the original continuous time system and of the
discrete time numerical systems.

10.1 Non-autonomous dynamical systems and pullback attractors

Consider an autonomous dynamical system on a metric space Ω described by a

group σ = {σt }t∈T of mappings of Ω into itself, where the time set T is either Z
(discrete time) or R (continuous time).
Let W be a complete metric space and consider a continuous mapping ϕ :
T+ × W × Ω → W satisfying the properties

ϕ(0, ·, ω) = idW , ϕ(τ + t, u, ω) = ϕ(τ, ϕ(t, u, ω), σt ω)

for all t, τ ∈ T+ , ω ∈ Ω and u ∈ W . The mapping ϕ is called a (continuous) cocycle

on W with respect to σ. Then the mapping π : T+ × W × Ω → W × Ω defined by

π(t, u, ω) := (ϕ(t, u, ω), σt ω)

for all t ∈ T+ , (u, ω) ∈ Ω × W forms an autonomous semi–dynamical system on the

state space W × Ω, i.e. the set of mappings {π(t, ·, ·)}t∈T+ of W × Ω into itself is a
semi–group, which is called a continuous skew product flow [292].
The usual concept of a global attractor for the autonomous semi–dynamical
system π on the state space Ω × W can be used here. It is the maximal nonempty
compact subset A of Ω × W which is π–invariant, that is

π(t, A) = A for all t ∈ T+ ,

and attracts all compact subsets of Ω × W , that is

lim β (π(t, D), A) = 0 for all D ∈ C(W × Ω),

t→∞

where C(W × Ω) is the space of all nonempty compact subsets of Ω × W and β is

the Hausdorff semi-distance on C(W × Ω).
Another type of attractor, called a pullback attractor, consists of subsets of
the original state space W , which is advantageous for discretizations of the non-
autonomous system.

Theorem 10.1 Let ϕ be a continuous cocycle on W with respect to a group σ

of continuous mappings on Ω and let π = (ϕ, σ) be the corresponding skew product
flow on W × Ω. In addition, suppose that there is a nonempty compact subset B of
W and for every D ∈ C(W ) there exists a T (D) ∈ T+ , which is independent f ω ∈
334 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Ω, such that

ϕ(t, D, ω) ⊂ B for all t > T (D). (10.1)

Then

1. there exists a unique pullback attractor I = {Iω | ω ∈ Ω} of the cocycle ϕ on

W , where
\ [
Iω = ϕ (t, B, σ−t ω). (10.2)
τ ∈T+ t>τ

2. there exists a global compact attractor A of the autonomous semi–dynamical

system π on W × Ω, where
\ [
A= π (t, B × Ω).
τ ∈T+ t>τ

3.
[
J= Iω × {ω}.
ω∈Ω

See Crauel and Flandoli [126] and Schmalfuß [284] for the proof of Assertion 1. and
Cheban and Fakeeh [83] and Hale [179] for the proof of Assertion 2. Assertion 3.
has been proved by Cheban [88].

Remark 10.1 Assertion 1. true remains under weaker conditions. For instance,
the sets D in the absorbing condition (10.1) could be parameter dependent, the
parameter space Ω need not be compact nor the mappings σt continuous (which is
the situation in random dynamical systems, see Arnold [6]). Note that the validity of
Assertion 3. for non-uniform absorbing times, a situation which occurs in important
applications, remains open. See [97] for a systematic investigation of the relationship
between the pullback and forwards attractors of the cocycle system and the global
attractor of the associated skew product flow.

10.2 Non-autonomous quasi-linear differential equation

We consider a non-autonomous quasi-linear differential equation

u̇ = A(ω)u + f (u, ωt) (10.3)

on Rd where ω ∈ Ω, on which there exists a group of mappings σt : Ω → Ω for all t

∈ R, where ωt := ωt. A solution x(t) = ϕ(t, u0 , ω) of (10.3) with initial value x(0)
 Pullback attractors under discretization 335

= u0 satisfies the equation

dϕ
(t, u0 , ω) = A(σt ω)ϕ(t, u0 , ω) + f (σt ω, ϕ(t, u0 , ω)).
dt
Our assumptions are

D1. Ω is a compact metric space and (t, ω) 7→ σt ω is continuous.

D2. ω 7→ A(ω) is continuous and satisfies

(A(ω)u, u) ≤ −α(ω) |u|2

for all (u, ω) ∈ Rd × Ω.

D3. (u, ω) 7→ f (u, ω) is continuous, is locally Lipschitz in u uniformly in ω ∈ Ω

and satisfies

(f (u, ω), u) ≤ a(ω) |u|2 + c(ω)

for all (u, ω) ∈ Rd × Ω.

D4. α(ω) > 0, α(ω) − a(ω) ≥ α0 > 0 and c(ω) ≤ c0 < ∞ for all ω ∈ Ω.

Remark 10.2 . The mapping f in D3 denotes the remaining part of the dif-
ferential equation. It may also contain linear terms that has not been included in
linear part with the matrix A(ω). Though seemingly superfluous, it is sometimes
convenient to distinguish the matrix operator A(ω) in this way (especially in infinite
dimensional generalizations, which are not considered here).

Remark 10.3 By the above continuity and the compactness of Ω we have the
finite uniform upper bounds

A0 := sup kA(ω)k, FR := sup |F (u, ω)|.

ω∈Ω ω∈Ω,|u|≤R

We also consider a variable time-step one-step explicit numerical scheme cor-

responding to the differential equation (10.3), such as the Euler scheme, which we
write as

un+1 = un + hn F (hn , σtn ω, un ) (10.4)

Pn−1 Pn
where t0 = 0 and tn = j=0 hj , t−n = − j=1 h−j for n ≥ 1 for {hj }n∈Z a given
two sided sequence of positive terms and F : [0, 1] × Ω × Rd → Rd is the increment
function. We make the following assumptions:
336 Global Attractors of Non-autonomous Dissipative Dynamical Systems

N1. F is continuous in all of its variables and locally Lipschitz in u uniformly in

(h, p) ∈ [0, 1] × Ω;

N2. The numerical scheme (10.4) satisfies a local discretization error estimate of
the form

|ϕ(h0 , u0 , ω) − u1 | ≤ h0 µR (h0 ), |u0 | ≤ R,

for each R > 0, where µR (h) > 0 for h > 0 and µR (h) → 0 as h → 0+.

N3. F satisfies the consistency condition

F (0, u, ω) = A(ω)u + f (u, ω)

for all (u, ω) ∈ Rd × Ω;

N4. F satisfies the Lipschitz consistency condition

|(F (h, u, ω) − A(ω)u − f (u, ω)) −

(F (h, v, ω) − A(ω)v − f (v, ω))| ≤ µ̄R (h)|u − v|

for all |u|, |v| ≤ R uniformly in ω ∈ Ω, where µ̄R (h) > 0 for h > 0 and µ̄R (h) → 0
as h → 0+.

For example, F (h, u, ω) ≡ A(ω)u + f (u, ω) for the Euler scheme applied to the
differential equation (10.3). Note that for one-step order schemes such as the Euler
and Runge–Kutta schemes, µ(h) is typically of the form KR hp for some integer p
≥ 1. In our case here this would require the differentiability of F in ω as well as u
and of σt in t, which is too restrictive for certain applications.

Our main result (see Theorem 10.2 below) is that non-autonomous dynamical
systems generated by the differential equation (10.3) and the numerical scheme
(10.4) both have pullback and global attractors, and that the numerical attractors
converge upper semi–continuously to the corresponding attractors of the differential
equation as the step size goes to zero. We will apply Theorem 10.1 to establish the
existence of such attractors, but first we need to show in what sense the numerical
scheme (10.4) generates a discrete time cocycle mapping and skew product flow.
An example
We consider the 3–dimensional Lorenz system with time dependent coefficients

u˙1 = −p(t)u1 + p(t)u2

u˙2 = r(t)u1 − u2 − u1 u3
u˙3 = −b(t)u3 + u1 u2
 Pullback attractors under discretization 337

(see Temam [314], Chapter I.2.3). Assuming that p and r are differentiable functions
of R into itself, we can rewrite this system after the transformation u3 := u3 −r(t)−
p(t) in the form (10.3) with the matrix
 
−p(0) p(0) 0
 
 
A(ω) =  −p(0) −1 0 
 
0 0 −1

and the mapping

 
0
 
 
f (u, ω) =  −u1 u3 ,
 
u1 u2 − σ̇(0) − ṙ(0) − b(0)(r(0) + σ(0)) − (b(0) − 1)u3

where u = (u1 , u2 , u3 ) and ω := ω(·) = (b(·), r(·), σ(·)) ∈ C(R, R) × C 1 (R, R) ×

C 1 (R, R).
Suppose that b is almost periodic in C(R, R) and that σ, r are almost periodic
in C 1 (R, R) with

0 < σmin := inf σ(t), 1 < inf b(t).

t∈R t∈R

Then define σt for each t ∈ R by σt p(·) := ω(· + t). Finally, define the parameter
set Ω to be the closed hull [292]
[
P := σt (b(·), r(·), p(·)),
t∈R

which is a compact subset of C(R, R) × C 1 (R, R) × C 1 (R, R).

We need to check that the assumptions D1–D4 are satisfied by the differential
equation with this matrix A and function f . The first assumption D1 follows
straighforwardly and the second D2 holds with α(ω) ≡ α0 := min(σmin , 1) > 0
with the assumptions on p and r ensuing the continuity of A. These assumptions
on p and r also ensure the continuity of f needed in D3 and the estimate is given
by

(f (u, ω), u) = (−ṗ(0) − ṙ(0))u3 − b(0)(r(0) + p(0))u3 − (b(0) − 1)u23

(ṗ(0) + ṙ(0))2 b2 (0)(r(0) + p(0))2

≤ + =: c(ω)
2(b(0) − 1) 2(b(0) − 1)

with a(ω) ≡ 0. Finally, it is obvious from these definitions of the constants that
D4 is also satisfied, in particular with c0 := supω∈Ω c(ω) < ∞.
338 Global Attractors of Non-autonomous Dissipative Dynamical Systems

The Euler scheme for this differential equation also satisfies assumptions N1–
N4. Since F (h, u, ω) ≡ A(ω)u + f (u, ω) here, assumptions N3 and N4 hold triv-
ially, while assumption N1 follows from D1–D3 above. Assumption N2 also follows
from D1–D3 with the proof being almost the same as in the first part of the proof
in the Appendix. Note that if b is also assumed to be continuously differentiable,
then we have the usual second order local discretization error here. One can show
that assumptions N1–N4 are also satisfied by higher order schemes such as Runge–
Kutta schemes, but the details are not as straightforward or clean as for the Euler
scheme.

10.3 Cocycle property

The solution ϕ(t, u0 , ω) of the differential equation (10.3) satisfies assumptions N1–
N4 the initial condition

ϕ(0, u0 , ω) = u0 for all (u0 , ω) ∈ Rd × Ω

and the cocycle property

ϕ(s + t, u0 , ω) = ϕ(s, ϕ(t, u0 , ω), σt p) for all s, t ∈ R+ , (u0 , ω) ∈ Rd × Ω

with respect to the autonomous dynamical system generated by the group {σt }t∈R
on Ω. (Existence of such solutions for all t ∈ R+ is assured by that of an absorbing
set to be established in the proof of Theorem 10.2 below). By our assumptions the
mapping (t, u, ω) 7→ ϕ(t, u, σt ω) is continuous. Moreover, the mapping ϕ := (ϕ, σ)
defined on R+ × Rd × Ω

ϕ(t, u, ω) := (ϕ(t, u, ω), σt ω) for all (t, u0 , ω) ∈ R+ × Rd × Ω

generates an autonomous semi-dynamical system, that is a skew product flow, on

the state space X := Rd × Ω.
The situation is somewhat more complicated for the discrete time system gen-
erated by the numerical scheme with variable time steps. For this we will restrict
the choice of admissible step-size sequences. For each δ > 0, we define H δ to be the
set of all two sided sequences {hn }n∈Z satisfying
1
δ ≤ hn ≤ δ (10.5)
2
for each n ∈ Z (the particular factor 1/2 here is chosen just for convenience). The
set Hδ is compact metric space with the metric
  ∞
X  
 (1) 
ρHδ h(1) , h(2) = 2−|n| hn − h(2)
n .
n=−∞
 Pullback attractors under discretization 339

We then consider the shift operator σ̃ : Hδ → Hδ defined by σ̃h = σ̃{hn }n∈Z

:= {hn+1 }n∈Z . The operator σ̃ is a homeomorphism with respect to the above
metric on Hδ and its group of iterates {σ̃n }n∈Z forms a discrete time autonomous


dynamical system on the compact metric space Hδ , ρHδ . Finally, for a given
Pn−1
sequence {hn }n∈Z we set t0 = 0 and define tn = tn (h) := j=0 hj and t−n =
Pn
t−n (h) := − j=1 h−j for n ≥ 1.
Now we introduce the parameter space Qδ := Hδ × Ω for a fixed δ > 0 and we
use the following lemma to introduce a discrete time autonomous dynamical system
Θ = {Θn }n∈Z on Qδ .

Lemma 10.1 The mappings Θn : Qδ → Qδ , n ∈ Z, generated by iteration of

Θ0 := idQδ , Θ1 (h, p) := (σ̃1 h, σh0 p) , Θ−1 (h, p) := σ̃−1 h, σ−h−1 p

form a group of continuous mappings on Hδ × P .

Proof. We first show that Θ1 Θ−1 = Θ−1 Θ1 = Θ0 = idQδ . Indeed we have

Θ1 Θ−1 (h, p) = Θ1 σ̃−1 h, σh−1 p = σ̃1 σ̃−1 h, σh−1 σ−h−1 p = (h, p)

Θ−1 Θ1 (h, p) = Θ−1 (σ̃1 h, σh0 p) = (σ̃−1 σ̃1 h, σ−h0 σh0 p) = (h, p).

The continuity of the Θn mappings follows from the facts that (t, p) 7→ σt p is
continuous, σ̃ is a homeomorphism and the composition and cartesian products of
continuous mappings are continuous. 
d δ d
We define a mapping ψ : Z+ × R × Q → R by

ψ(0, u0 , q) := u0 , ψ(n, u0 , q) = ψ(n, u0 , (h, p)) := un n ≥ 1,

where un is the nth iterate of the numerical scheme (10.4) with initial value u0 ∈
Rd , initial parameter ω ∈ Ω and step-size sequence h ∈ Hδ . These mappings are
continuous on Rd × Qδ . They also satisfy a cocycle property with respect to Θ. 

Lemma 10.2 ψ is a discrete time cocycle over (Qδ , Z, Θ) Rd with fiber Rd .

Proof. We write the numerical scheme (10.4) for a given (h, p) and u0 as


un+1 = un + hn F hn , σtn (h) p, un , n ∈ Z+ ,
Pn−1
where t0 (h) = 0 and tn (h) = j=0 hj for n ≥ 1. Hence, in terms of the ψ mapping
we have

ψ(n + 1, u0 , (h, p)) = ψ(n, u0 , (h, p)) + hn F (hn , σtn (h) p, ψ(n, u0 , (h, p)))
340 Global Attractors of Non-autonomous Dissipative Dynamical Systems

in general and


ψ(1, u0 , (h, p)) = u0 + h0 F h0 , σt0 (h) p, u0 )

for n = 0 since by definition ψ(0, u0 , (h, p)) = u0 , which gives the identity property.
The cocycle property is established as follows. Let n ≥ 0. Then

ψ(1 + n, u0 , q) = ψ(1 + n, u0 , (h, p))

= ψ(n, u0 , (h, p)) + hn F hn , σtn (h) p, ψ(n, u0 , (h, p))


= ψ(n, u0 , (h, p)) + (σ̃n h)0 F (σ̃n h)0 , σt0 (σ̃n h) p, ψ(n, u0 , (h, p))

= ψ (1, p, Θn (h), ψ(n, u0 , (h, p))) = ψ (1, ψ(n, u0 , q), Θn q) ,

that is u1+n = = ψ (1, un , Θn q). Iterating this n times, we obtain

un = ψ (1, u0 , Θn−1 q) ◦ · · · ◦ ψ (1, u0 , Θ0 q) .

Similarly with m ≥ 2 we have

ψ(m + n, u0 , q) = um+n = ψ (1, ·, Θm+n−1 q) ◦ · · · ◦ ψ (1, ·, Θn q) ◦

◦ψ (1, ·, Θn−1 q) ◦ · · · ◦ ψ (1, u0 , Θ0 q)

= ψ (1, ·, Θm+n−1 q) ◦ · · · ◦ ψ (1, un , Θn q)

= ψ (1, ·, Θm−1 Θn q) ◦ · · · ◦ ψ (1, un , Θ0 Θn q)

= ψ (m, un , Θn q) = ψ (m, ψ (n, q, u0 ) , Θn q) ,

which is the desired cocycle property.

Finally we define Ψ := (ψ, Θ) and observe that the mappings Ψ(n, ·.·) are con-
tinuous on Rd × Qδ for n ∈ Z+ . 

Remark 10.4 The mappings ψ, Θ and Ψ are defined in the same way for each
δ > 0, so we do not index them with δ.

From the cocycle and group properties we obtain

Lemma 10.3 Ψ = (ψ, Θ) is a discrete time autonomous semi–dynamical system

on the state space Rd × Qδ .
 Pullback attractors under discretization 341

10.4 Main result

Our main result is the to establish the existence of pullback attractors for the
non-autonomous dynamical systems (NDS) generated by the differential equation
and numerical scheme and to show that the components of the numerical pullback
attractor are upper semi–continuous in their parameter and converge upper semi–
continuously to the corresponding components of the differential equation’s pullback
attractor. Global attractors also exist for the corresponding skew product flows and
converge upper semi–continuously in an appropriate sense.

Theorem 10.2 Let Assumptions D1–D4 and N1–N3 hold. Then the continuous
time NDS (ϕ, σ) generated by the differential equation (10.3) has a pullback attractor
I = {Iω | ω ∈ Ω} and the discrete time NDS (Ψ, Θ) generated by the numerical
scheme (10.4) has a pullback attractor I δ = {Iqδ | q ∈ Qδ }, provided the maximal
step-size δ is sufficiently small, such that the set-valued mappings ω 7→ I ω and (p, h)
δ
7→ I(p,h) are upper semi–continuous with respect to the Hausdorff semi-distance and
satisfy
 
δ
lim sup β I(ω,h) , Iω = 0 for each ω ∈ Ω.
δ→0+ h∈Hδ

Moreover, the corresponding skew product flows have global attractors J and J δ ,
respectively, of the form
[ [
J= Iω × {ω}, Jδ = Iqδ × {q},
ω∈Ω q∈Qδ

which satisfy


lim β PrRd ×Ω J δ , J = 0.
δ→0+

Proof. The proof of the convergence assertions in Theorem 10.2 will follow imme-
diately from an application of Theorem 10.1 after
p it has been shown that the ball
B[0; R0 ] in Rd with centre 0 and radius R0 := 3 c0 /α0 is a forwards absorbing set
uniformly in all parameters for both the continuous time and discrete time cocycle
systems under consideration. The convergence assertions require additional work.

10.4.1 Existence of an absorbing set

Write x(t) for the solution ϕ(t, u0 , ω) of (10.3), so

dx
(t) = A(σt ω)x(t) + f (σt ω, x(t)).
dt
342 Global Attractors of Non-autonomous Dissipative Dynamical Systems

The following estimate will also be used to construct an absorbing set.

 
d dx
|x(t)|2 = 2 (t), x(t) = 2 (A(σt p)x(t) + f (σt p, x(t)), x(t)) (10.6)
dt dt

= 2 (A(σt p)x(t)) + 2 (f (σt p, x(t)), x(t))

≤ −2α(σt p) |x(t)|2 + 2a(σt p) |x(t)|2 + 2c(σt p)

≤ −2 (α(σt p) − a(σt p)) |x(t)|2 + 2c0 ≤ −2α0 |x(t)|2 + 2c0

and so

c0

|x(t)|2 ≤ |u0 |2 e−2α0 t + 1 − e−2α0 t .
α0
p
This implies that the ball B[0; R0 ] with radius R0 = 3 c0 /α0 is a forwards absorb-
ing and positively invariant set for all solutions of the differential equation (10.3)
uniformly in ω ∈ Ω. Note for later purposes that the ball B[0; 2R0 /3] is also posi-
tively invariant for the differential equation.
The proof that B[0; R0 ] is a uniform forwards absorbing set for the numerical
scheme (10.4) is more complicated. First we show that the inequality (10.6) implies
r
−α0 t c0

|x(t)| ≤ |u0 |e + 1 − e−α0 t . (10.7)
α0
q
c0
as long as |x(t)| ≥ 13 R0 = α0 . To see this note that (10.6) can be rewritten as

d c0
|x(t)|2 ≤ −2α0 |x(t)|2 + 2c0 ≤ −2α0 |x(t)|2 + 2 q |x(t)|
dt c0
α0

and hence as

d √
|x(t)| ≤ −α0 |x(t)| + c0 α0
dt

as long as |x(t)| ≥ 13 R0 .

We note also by continuity that there exists a T = T (c0 , α0 ) > 0 such that |x(t)|
= |ϕ(t, u0 , ω)| ≥ 31 R0 for all t ∈ [0, T ], u0 with |u0 | ≥ 32 R0 and ω ∈ Ω.

We fix an R  R0 and let KR be the constant in the local discretization

error estimate for the ball B[0; R]. Then from (10.7) with x(h) = ϕ(h, u0 , ω) and
 Pullback attractors under discretization 343

Assumption N2 we have

|u1 | ≤ |ϕ(h, u0 , ω)| + |ϕ(h, u0 , ω) − u1 |

r
c0

≤ |u0 |e−α0 h + 1 − e−α0 h + KR hµR (h) (10.8)
α0

for u0 ∈ A[R, 2R0 /3] := B[0; R] \ B[0; 2R0 /3] and h ∈ (0, T ]. Now let δ0 ∈ (0, 1] ∩
(0, T ] be such that

1 KR hµR (h) 1
KR hµR (h) ≤ R0 , −α h
≤ R − R0
3 1−e 0 3

for h ∈ (0, δ0 ]. Then from (10.8) for u0 ∈ A[R, 2R0 /3] and h ∈ (0, δ0 ] we have
r
c0

|u1 | ≤ |u0 |e−α0 h + 1 − e−α0 h + KR hµR (h)
α0
 
1
1

≤ Re−α0 h + R0 1 − e−α0 h + R − R0 1 − e−α0 h ≤ R,
3 3

so u1 ∈ B[0; R]. In addition, for u0 ∈ B[0; 2R0 /3] we have x(h) ∈ B[0; 2R0 /3], so

2 2 1
|u1 | ≤ |ϕ(h, u0 , ω)| + |ϕ(h, u0 , ω) − u1 | ≤ R0 + KR hµR (h) ≤ R0 + R0 = R0
3 3 3

for h ∈ (0, δ0 ], so u1 ∈ B[0; R] here too. Hence, the ball B[0; R] is positively invariant
for the numerical scheme for u0 ∈ B[0; R] and h ∈ (0, δ0 ]. We can thus apply the
inequality (10.8) iteratively as long as the un ∈ A[R, 2R0 /3], that is we have
r
−α0 h c0

|un+1 | ≤ |un |e + 1 − e−α0 h + KR hµR (h) (10.9)
α0

when u0 , u1 , . . ., un ∈ A[R, 2R0 /3] and h ∈ (0, δ0 ].

Now further restrict the step-size so that

r
KR hµR (h) 1 1 1 1 c0
−α h
≤ R0 − R0 = R0 −
1−e 0 2 3 2 2 α0

for all h ∈ (0, δ1 ], where δ1 ∈ (0, δ0 ). Using a similar argument as above for the ball
B[0; R], we can show that the ball B[0; R0 ] is positively invariant for the numerical
scheme with step-sizes h ∈ (0, δ1 ].
344 Global Attractors of Non-autonomous Dissipative Dynamical Systems

To show that the ball B[0; R0 ] is absorbing, we further restrict the step-size so
that
1
1
1 + e−α0 h ≤ e− 4 α0 h
2

for all h ∈ (0, δ2 ], where δ2 ∈ (0, δ1 ]. Then, if u0 ∈ A[R, R0 ], from inequality (10.9)
we have
r
−α0 h c0

|u1 | ≤ |u0 |e + 1 − e−α0 h + KR hµR (h)
α0
r  r 
−α0 h c0
1 c0

≤ |u0 |e + 1 − e−α0 h + R0 − 1 − e−α0 h
α0 2 α0

1

≤ |u0 |e−α0 h + |u0 | 1 − e−α0 h
2
1
1
≤ 1 + e−α0 h |u0 | ≤ e− 4 α0 h |u0 |.
2

In particular, when u0 , u1 , . . ., uj ∈ B[0; R] \ B[0; R0 ] we can iterate the last

inequality to obtain
1
|uj | ≤ e−j 8 α0 δ |u0 |

if we use variable step-sizes with δ/2 ≤ hj δ with δ ∈ (0, δ2 ]. Obviously, there

exists a finite integer JR (u0 ) such that |uj | ≤ R0 for all j ≥ JR (u0 ), that is the ball
B[0; R0 ] is absorbing for the numerical scheme for all step-size sequences h ∈ H δ
with δ ∈ (0, δ2 ]. Note that this holds uniformly in ω ∈ Ω.

10.4.2 Upper semi–continuity of the pullback attractor component

sets
Let ψ(n, x, q) denote the numerical trajectory and Θ the shift operator on Qδ . The
mappings q 7→ Θq and (x, q) 7→ ψ(n, x, q) for each integer positive n are continuous.
The absorbing set B = B[0; R0 ] is compact absorbing set and forwards in-
variant uniformly in q ∈ Qδ . Hence, by the cocycle property, the compact sets
ψ(n, B, Θ−n q) are nested with increasing n and the pullback attractor has compo-
nent sets defined by
\
Iqδ := ψ(n, B, Θ−n q)
n≥0
 Pullback attractors under discretization 345

This means, in particular, that β(ψ(n, B, Θ−n q), Iqδ ) → 0 as n → ∞. Let ε > 0 and
pick n0 = n0 (ε, q) > 0 so that

ψ(n0 , B, Θ−n0 q) ⊂ B Iqδ , ε ,

where B Iqδ , ε is the ball of radius ε about Iqδ .
Now the compact set-valued mappings q 7→ ψ(n, B, Θ−n q) are continuous in q
with respect to the Hausdorff semi-distance for each fixed n. Fix ε > 0 and pick n0
= n0 (ε, q̄) from above. Then there exists δ(ε, q̄) = δ(ε, n0 (ε, q̄)) > 0 such that

β (ψ(n0 , B, Θ−n0 q), ψ(n0 , B, Θ−n0 q̄)) < ε

for all q ∈ Qδ with ρ(q, q̄) < δ(ε, q̄). In particular,

ψ(n0 , B, Θ−n0 q) ⊂ B (ψ(n0 , B, Θ−n0 q̄), ε)

for all q ∈ Qδ with ρQδ (q, q̄) < δ(ε, q̄). Hence we have

Iqδ ⊂ ψ(n0 , B, Θ−n0 q) ⊂ B (ψ(n0 , B, Θ−n0 q̄), ε)

⊂ B B Iq̄δ , ε , ε = B Iq̄δ , 2ε

that is Iqδ ⊂ B Iq̄δ , 2ε or equivalently β Iqδ , Iq̄δ < 2ε for all q ∈ Qδ with ρ(q, q̄) <
δ(ε, q̄). This means the mapping q 7→ Iqδ is upper semi–continuous.
The proof for the mapping ω 7→ Aω is essentially the same.

10.4.3 Upper semi–continuous convergence of the discretized

pullback attractors
We will now prove the upper semi–continuous convergence of the discretized pull-
back attractor component sets to their continuous time counterparts. For the proof
we need the following lemma on the convergence of the numerical trajectories to the
corresponding continuous time trajectory with convergence of the maximum step
size to zero. Its proof is given in the appendix.

Lemma 10.4 For fixed t > 0 and h = {hn }n∈Z ∈ Hδ for some δ > 0, let N (t, h)
be the positive integer such that

h−1 + h−2 + · · · + h−N (t,h) ≤ t < h−1 + h−2 + · · · + h−N (t,h)−1

346 Global Attractors of Non-autonomous Dissipative Dynamical Systems

and consider a sequence (of step-size sequences) with hm ∈ Hδm , where δm → 0 as

m → ∞. Then


ψ N (t, hm ), um , Θ−N (t,hm ) (hm , ωm ) → ϕ(t, u0 , σ−t ω) as m→∞

for any sequence um → u0 ∈ Rd and ωm → ω ∈ Ω.

We suppose that the upper semi–continuous convergence assertion of Theorem

10.2 is not true. Then there exists an ε0 > 0 and sub-sequences (for convenience
δm
we use the original index) hm in Hδm and am ∈ I(ω,h m ) for m ∈ N such that

dist am , I(ω,0) ≥ ε0 . (10.10)
δ
Note that
p the component sets I(ω,h) and Aq are contained in the ball B[0; R0 ] where
δm
R0 = 3 c0 /α0 for each ω ∈ Ω. Then am ∈ A(p,hm ) ⊂ B[0; R0 ] for each for m ∈
N and B[0; R0 ] is compact, so there exists a convergent subsequence (again we use
the original index) am → a∗ ∈ B[0; R0 ]. Thus we have


dist a∗ , I(ω,0) ≥ ε0 .

Now choose t > 0 sufficiently large so that by pullback attraction

1
dist ϕ (t, σ−t p, B[0; R0 ]) , I(ω,0) < ε0 . (10.11)
2

By the invariance property of a pullback attractor there exist bm ∈ AδΘm−N (t,hm ) (ω,hm )
such that


ψ N (t, hm ), Θ−N (t,hm ) (p, hm ), bm = am . (10.12)

Since the AδΘm−N (t,hm ) (ω,hm ) ⊂ B[0; R0 ] for each for m ∈ N, there exists a convergent
subsequence (once again we use the original index) bm → b∗ ∈ B[0; R0 ]. By (10.12)
and Lemma 10.4 we have

ϕ (t, σ−t ω, b∗ ) = a∗ ,

which contradicts (10.10) with respect (10.11). This contradiction proves the upper
semi–continuous convergence of the numerical pullback attractor components.

10.4.4 Upper semi–continuous convergence of the discretized

global attractors
Let J ⊆ Rd × Ω be the global attractor of the continuous time skew prod-
uct flow dynamical system π(t, u, ω) = (ϕ(t, u, ω), σt ω) and J δ ⊂ Hδ × Rd × Ω
the global attractor of the discrete time semi–dynamical system π δ (n, h, u, ω) =
 Pullback attractors under discretization 347

(Θn (h, ω), ψ δ (n, x, (h, ω)) based on the numerical scheme, where we include the su-
perscript δ on π δ and ψ δ for emphasis. Then J δ converges to A as δ → 0 uniformly
in the sense that



lim sup β pδ , xδ , J = 0. (10.13)
δ→0 (hδ ,xδ ,ω δ )∈J δ

Suppose that (10.13) is not true. Then there exist sequences δn →0 and


hδn , xδn , ω δn ⊂ J δn and an ε0 > 0 such that



β xδ n , ω δ n , A ≥ ε 0 .

Let B be a absorbing compact set in Rd that is independent of δ. Since J δ ⊂

Hδ × B × Ω and Hδ × B × Ω is compact, we can select a subsequence (we use the
same index for convenience) such that ω δn → ω 0 and xδn → x0 as n → ∞. Hence



β x0 , ω 0 , J ≥ ε 0 . (10.14)

On the other hand, since B is compact and the absorption is uniform in ω ∈ Ω,

there exists a t > 0 such that
1
β (π(t, B, Ω), J) < ε0 .
2
In addition, by invariance, π δ (j, J δ ) = J δ for all j ∈ N. Now choose for j the
δn
integer
 N (t, h ), where N (t, h) is defined in Lemma 10.4. Then we can find a
ĥδn , ω̂ δn , x̂δn ∈ Hδ × B × Ω such that
 

π δn N (t, hδn ), ĥδn , x̂δn , ω̂ δn = hδn , xδn , ω δn ;
 
indeed we can define ĥδn , ω̂ δn by Θ−N (t,hδn ) (hδn , ω δn ). By a similar compactness
argument, there is subsequence of this subsequence (again we use the original index)



such that x̂δn , p̂δn → x̂0 , ω̂ 0 ∈ B × Ω. By Lemma 10.4 we then have
  

xδn = ψ δn N (t, hδn ), ĥδn , x̂δn , ω̂ δn → ϕ t, x̂0 , ω̂ 0 = x0

while
 



ΘN (t,hδn ) ĥδn , ω̂ δn = hδn , ω δn → 0, ω 0 = 0, σt ω̂ 0 .

Combining these results we have

 



π δn N (t, hδn ), ĥδn , x̂δn , ω̂ δn → 0, π t, x̂0 , ω̂ 0 = 0, x0 , ω 0



and hence β x0 , ω 0 , J < 12 ε0 , which contradicts (10.14). Hence the original
assertion (10.13) must be true. This completes the proof of the main theorem,
Theorem 10.2. 
348 Global Attractors of Non-autonomous Dissipative Dynamical Systems

10.5 Singleton set-valued pullback attractor case

Let u1 (t) and u2 (t) be two solutions of the differential equation (10.3) with the
same initial parameter ω but different initial values in the positively absorbing ball
B[0; R0 ] and write

∆(t) = u1 (t) − u2 (t), ∆f,ω (t) = f (σt ω, u1 (t)) − f (σt ω, u2 (t)).

Let L0 be the local Lipschitz constant of f in B[0; R0 ], which by assumption is

uniform in ω ∈ Ω, so

|f (ω, u1 (t)) − f (ω, u2 (t))| ≤ L0 |u1 (t) − u2 (t)|

or |∆f,ω (t)| ≤ L0 |∆(t)| in B[0; R0 ]. We assume that L0 ≤ α0 /2. Then similarly to

earlier (but now we do not use the inner product inequality on the f )
 
d d
|∆(t)|2 = 2 ∆(t), ∆(t)
dt dt

= 2 (A(σt ω)∆(t), ∆(t)) + 2 (∆f,ω (t), ∆(t))

≤ −2α0 |∆(t)|2 + 2|∆f,ω (t)||∆(t)|

≤ −2α0 |∆(t)|2 + 2L0 |∆(t)|2

≤ −2 (α0 − L0 ) |∆(t)|2 ≤ −α0 |∆(t)|2

and so

|∆(t)| ≤ |∆(0)|e−(α0 /2)t ,

which means the solution operator u 7→ φ(t, ω, u) is a contraction mapping on the

ball B[0; R0 ] for each t > 0 and ω ∈ Ω.
Now consider the numerical scheme. Let

y1 = u1 + hF (h, ω, u1 ), y2 = u2 + hF (h, ω, u2 )

where h ∈ [δ/2, δ] and u1 , u2 ∈ B[0; R0 ]. Write

∆x = u1 − u2 , ∆y = y1 − y2 , ∆F (h) = F (h, ω, u1 ) − F (h, ω, u2 ),

so |∆F | ≤ L00 |∆x| in B[0; R0 ], where L00 is the local Lipschitz constant of F in u
on B[0; R0 ], uniformly in (h, ω) ∈ [0, 1] × Ω. We assume the Lipschitz consistency
condition N4 here, so

|∆F (h) − A∆x − ∆f,ω (h)| ≤ µ̄R (h)|∆x|

 Pullback attractors under discretization 349

where we omit the parameter ω for convenience. Thus we have

|∆y|2 = (∆y, ∆y) = (∆x + h∆F (h), ∆x + h∆F (h))

= (∆x, ∆x) + 2h(∆F (h), ∆x) + h2 (∆F (h), ∆F (h))

= |∆x|2 + 2h(A∆x + ∆f,ω (h), ∆x) + h2 |∆F (h)|2

+2h(∆F (h) − A∆x − ∆f,ω (h), ∆x)

≤ |∆x|2 − 2hα0 |∆x|2 + 2h|∆f,ω (h)| |∆x| + h2 L200 |∆x|2

+2h|∆F (h) − A∆x − ∆f,ω (h)||∆x|

≤ |∆x|2 (1 − 2hα0 ) + 2hL0 |∆x|2 + h2 L200 |∆x|2

+2hµ̄R (h)|∆x|2


≤ |∆x|2 1 − 2h(α0 − L0 ) + h2 L200 + 2hµ̄R (h)


≤ |∆x|2 1 − hα0 + h2 L200 + 2hµ̄R (h) ,

where we have used the assumption that L0 ≤ α0 /2. By further restricting h from
above we can assure that

|∆y|2 ≤ |∆x|2 (1 − hα0 /2) ≤ |∆x|2 γ(δα0 /4)

for h ∈ [δ/2, δ] and δ sufficiently small. This means that the numerical solution
satisfies the contractive condition
 
ψ(n, u0 , (h̄, ω)) − ψ(n, x̄0 , (h̄, ω)) ≤ |u0 − x̄0 |γ n
0

for all u0 , x̄0 ∈ B[0; R0 ], ω ∈ Ω and step-size sequence h̄ ∈ Hδ , where γ0 :=

p
γ(δα0 /4).
From the Contraction Mapping Principal we conclude that the original and
numerical pullback attractors each consist of a single trajectory. The continuity
of their component elements with respect to the parameter follows from Theorem
10.2 and the fact that upper semi–continuity there reduces to continuity for the
singleton set-valued mappings.

Theorem 10.3 Let the Assumptions D1–D4 and N1–N4 hold and suppose
that 2L0 ≤ α0 and that δ is sufficiently small. Then the pullback attractors of
Theorem 10.2 consist of singleton component sets, that is Iω = {a∗ (ω)} and I(h,ω) δ

= {a∗δ (h, ω)}, where the mappings ω 7→ a∗ (ω) and (h, ω) 7→ a∗δ (h, ω) are continuous.
350 Global Attractors of Non-autonomous Dissipative Dynamical Systems

These singleton valued pullback attractor–trajectories inherit the periodicity or

almost periodicity of the differential equation and of the differential equation and
step-size sequence, respectively. This is formulated in the following theorem, the
proof of which will be presented in the remainder of this section. The periodic case
is straightforward, while the almost periodic case is considerably more complicated
and requires the introduction of appropriate definitions and a number of auxiliary
results.
Recall that the set A ⊂ P is called minimal with respect to a dynamical system
(Ω, R, σ) if it is nonempty, closed and invariant and if no proper subset of A has
these properties.

Theorem 10.4 Suppose that the assumptions of Theorem 10.3 hold and that Ω
is minimal. Then the singleton valued pullback attractor–trajectory I ω = {a∗ (ω)} is
periodic (resp., almost periodic) if ω ∈ Ω is periodic (resp., almost periodic), whereas
δ
the numerical singleton valued pullback attractor–trajectory I (h,ω) = {a∗δ (h, ω)} is
δ
periodic (resp., almost periodic) if q = (h, ω) ∈ Q is periodic (resp., almost peri-
odic).

Definition 10.1 A sequence h = {hn }n∈Z is m–periodic if hn+m = hn for all n

∈ Z or, equivalently, if σ̃m h = h, where m is the smallest integer for which these
equalities hold.

Recall that we have defined a time sequence {tn (h)}n∈Z by t0 (h) = 0, tn (h) :=
Pn−1 Pn
j=0 hj and t−n (h) := − j=1 h−j for n ≥ 1 corresponding to a given sequence
h = {hn }n∈Z .

Lemma 10.5 Let h ∈ Hδ be m–periodic and let ω ∈ Ω be τ –periodic with respect

to σ, that is with στ ω = ω where τ ∈ R+ . Then the point (h, ω) ∈ Qδ = Hδ × Ω is
periodic with respect to Θ = (σ̃, σ) if and only if tm (h)/τ is rational.

Proof. Suppose that tm (h)/τ = k/l for some k, l ∈ N. Then ltm (h) = kτ and
Θlm (h, ω) = (h, σltm (h) ω) = (h, σkτ ω) = (h, ω). On the other hand, suppose that
Θk (h, ω) = (h, ω) for some k ∈ N. Then (σ̃k h, σtk (h) ω) = (h, ω), which implies that
k = l1 m and tk (h) = l1 tm (h) = l1 tm (h) = l2 τ where l1 , l2 ∈ N. Hence tm (h)/τ =
l1 /l2 . 

The following result can be found in [32, 304].

Theorem 10.5 Let (X, T, π) be a dynamical system on a compact metric space

(X, ρ). Then a point x ∈ X is almost periodic if and only if for every ε > 0 there
 Pullback attractors under discretization 351

exists a δ = δ(ε) > 0 such that

ρ (π(t + t1 , x), π(t + t2 , x)) < ε, for all t ∈ T,

whenever ρ (π(t1 , x), π(t2 , x)) < δ.

Definition 10.2 A sequence {cn } in R is said to be almost periodic if the function

ϕ : Z → R defined by ϕ(n) := cn for n∈ Z is almost periodic.

Definition 10.3 A sequence {τn } in R will be called regular if it has the form

τn = an + cn , for all n ∈ Z,

where a ∈ R is a constant and{cn } is an almost periodic sequence; see Samoilenko

and Trofimchuk [280, 281].

Theorem 10.6 Suppose that the time sequence {tn (h)}n∈Z corresponding to h
∈ Hδ is regular. In addition, suppose that dynamical system (Ω, R, σ) is minimal
and almost periodic, that is Ω is minimal and every point ω ∈ Ω is almost periodic.
Then the point (h, ω) ∈ Qδ is almost periodic for the dynamical system (Qδ , Z, Θ).

Proof. Since the point ω ∈ Ω is almost periodic for the dynamical system (Ω, R, σ),
then by Theorem 10.5 the point ω ∈ Ω will be almost periodic relatively to the
discrete time dynamical system (Ω, Z, σ (a) ), where σ (a) = {σan }n∈Z and tn (h) =
an + cn is the regularity representation of {tn (h)}n∈Z . We apply Theorem 10.5 to
the dynamical system (Ω, R, σ). Given ε > 0, let δ(ε) ∈ (0, ε/3) be such that

ε
ρ(σt p1 , σt p2 ) < (10.15)
3

for every t ∈ R and ω1 , ω2 ∈ Ω with ρ(ω1 , ω2 ) < δ. Then we use uniform continuity
on the compact space Ω: given the above δ(ε) > 0, let γ(ε) ∈ (0, δ(ε)) be such that

ρ(σs ω, ω) < δ (10.16)

for every ω ∈ Ω and s ∈ R with |s| ≤ γ.

Now for this γ(ε) > 0 we denote by Mγ(ε) the relatively dense subset of Z subset
for which

ρ(σ̃n+m h, σ̃n h) < γ(ε), |cn+m − cn | < γ(ε), ρ(σa(n+m) ω, σan ω) < γ(ε) (10.17)
352 Global Attractors of Non-autonomous Dissipative Dynamical Systems

for all m ∈ Mε , n ∈ Z and ω ∈ Ω. From (10.15)–(10.17) we have

ρ(Θn+m (h, ω), Θn (h, ω)) = ρ(σ̃n+m h, σ̃n h) + ρ(σtn+m (h) ω, σtn (h) ω)
= ρ(σ̃n+m h, σ̃n h) + ρ(σa(n+m)+cn+m ω, σan+cn ω) < ρ(σ̃n+m h, σ̃n h)
+ρ(σa(n+m) (σcn+m ω), σa(n+m) (σcn ω)) + ρ(σa(n+m) (σcn ω), σan (σcn ω))
ε ε ε ε
< γ(ε) + + γ(ε) < + + = ε
3 3 3 3

for all n ∈ Z and m ∈ Mγ(ε) . Hence the point (h, p) ∈ Qδ is almost periodic for
the dynamical system (Qδ , Z, Θ). 

Corollary 10.1 Let h ∈ Hδ be m–periodic and let ω ∈ Ω be almost periodic for

the dynamical system (Ω, R, σ). Then the point (h, ω) ∈ Qδ is almost periodic for
the dynamical system (Qδ , Z, Θ). In particular, if tm (h)/τ is irrational, then point
(h, p) is almost periodic, but not periodic.

As the final step in our proof of Theorem 10.4, we need the following lemma,
which we prove directly here noting that the result also follows from Theorems 1
and 2 in [276].

Lemma 10.6 Suppose that the assumptions of Theorem 10.3 hold and let
{a∗δ (q)}q∈Qδ denote the singleton valued pullback attractor for the numerical scheme
(10.4). Then the function n 7→ a∗δ (Θn q) for n ∈ Z is periodic (resp., almost pe-
riodic) if the point q is periodic (resp., almost periodic) for the dynamical system
(Qδ , Z, Θ).

Proof. Let q ∈ Qδ be m–periodic, that is Θm q = q. Then a∗δ (Θn+m q) = a∗δ (Θn Θm q)

= a∗δ (Θn q) for every n ∈ Z. Hence n 7→ a∗δ (Θn q) is periodic.
The function a∗δ : Qδ → Rd defined by q 7→ a∗δ (q) for each q ∈ Qδ is continuous,
hence uniformly continuous, on the compact space Qδ . That is, for every ε > 0
there exists a δ(ε) > 0 such that |a∗δ (q1 ) − a∗δ (q2 )| < ε whenever ρQδ (q1 , q2 ) < δ.
Now let the point q be almost periodic and for δ = δ(ε) > 0 denote by Mδ the
relatively dense subset of Z such that ρ(Θn+m q, Θn q) < δ for all m ∈ Mδ and n ∈
Z. From this and the uniform continuity we have

|a∗δ (Θn+m q) − a∗δ (Θn q)| < ε

for all n ∈ Z and m ∈ Mδ(ε) . Hence n 7→ a∗δ (Θn q) is almost periodic. 

In conclusion, we can restate the assertions of Theorem 10.4 in more detail as

follows.
 Pullback attractors under discretization 353

Corollary 10.2 Suppose that the assumptions of Theorem 10.3 hold and that Ω
is minimal. In addition, let {a∗δ (h, ω)}(h,ω)∈Qδ denote the singleton valued pullback
attractor for the numerical scheme (10.4).
1. Let h ∈ Hδ be m–periodic and ω ∈ Ω be τ –periodic. Then n 7→ a∗δ (Θn (h, ω)} is
periodic if tm (h)/τ is rational and almost periodic if tm (h)/τ is irrational.
2. Let the time sequence {tn (h)}n∈Z corresponding to h ∈ Hδ be regular and let the
point ω ∈ Ω be almost periodic. Then n 7→ a∗δ (Θn (h, ω)} is almost periodic.

10.6 Appendix: Proof of Lemma 10.4

Consider an autonomous dynamical system on a compact metric space Ω described

by a group σ = {σt }t∈R of mappings of Ω into itself such that the mapping (t, ω)
7→ σt ω is continuous. Consider also an ordinary differential equation

u̇ = g(u, ωt) (ω ∈ Ω)

on Rd with a unique solution ϕ(t, u0 , ω) satisfying the initial value problem

d
ϕ(t, u0 , ω) = g(ϕ(t, u0 , ω), σt ω), x(0, u0 , ω) = u0 .
dt

We assume that (u, ω) 7→ g(u, ω) is continuous on Rd × Ω and locally Lipschitz in

u uniformly in ω, that is for each R > 0 there exists an LR such that

|g(x, ω) − g(y, ω)| ≤ LR |x − y|, ∀x, y ∈ B[0; R].

In particular, then for a sequence of times tn and time steps hn = tn+1 − tn this
gives in integral equation form
Z tn+1
ϕ(tn+1 , u0 , ω) = ϕ(tn , u0 , ω) + g(στ ω, ϕ(τ, u0 , ω)) dτ.
tn

In future we just write x(t) for this solution. By the Mean Value Theorem there
exists τn ∈ [0, 1] such that

ϕ(tn+1 , u0 , ω) = ϕ(tn , u0 , ω) + x(tn+1 ) = x(tn ) +

hn g(σtn +τn hn ω, ϕ(tn + τn hn , u0 , ω)).

The corresponding higher order scheme solution is

un+1 = un + hn F (hn , σtn ω, un ),

354 Global Attractors of Non-autonomous Dissipative Dynamical Systems

where the increment function F (h, u, ω) is continuous and satisfies the consistency
condition

F (0, u, ω) = g(u, ω), ∀ x, ω.

Let x(t) := ϕ(t, u0 , ω), then

x(tn+1 ) − un+1 = x(tn ) − un +

hn [g(x(tn + τn hn ), σtn +τn hn ω) − F (hn , σtn ω, un )]

so the global discretization error En := |x(tn ) − un | is estimated by

En+1 ≤ En + hn |g(x(tn + τn hn ), σtn +τn hn ω) − F (hn , σtn ω, un )|

≤ En + hn |g(x(tn + τn hn ), σtn +τn hn ω) − g(x(tn ), σtn +τn hn ω)|

+hn |g(x(tn ), σtn +τn hn ω) − g(x(tn ), σtn ω)|
+hn |g(x(tn ), σtn ω) − g(un , σtn ω)|
+hn |g(un , σtn ω) − F (hn , σtn ω, un )|

≤ En + hn LR |x(tn + τn hn ) − x(tn )| + hn ωg (∆n ω; R)

+hn LR |x(tn ) − un | + hn ωF (hn ω; R)
Z 
 tn +τn hn 
 
= (1 + hn LR )En + hn LR  g(x(s), σs ω) ds
 tn 
+hn ωg (∆n ω; R) + hn ωF (hn ω; R)

≤ (1 + hn LR )En + h2n LR MR + hn ωg (∆n ω; R) + hn ωF (hn ω; R)

where MR := maxω∈Ω,x∈B[0;R] |g(u, ω)| and ωg (δ; R) is the modulus of continuity

of g(σ· ω, x) uniformly in ω ∈ Ω and u ∈ B[0; R] and ωF (hn ω; R) is the modulus of
continuity of F (·, u, ω) uniformly in ω ∈ Ω and u ∈ B[0; R] that is

ωg (δ; R) := sup sup |g(σt ω, x) − g(ω, x)|

0≤t≤δ ω∈Ω, x∈B[0;R]

and

ωF (δ; R) := sup sup |F (h, u, ω) − F (0, u, ω)|.

0≤h≤δ ω∈Ω, x∈B[0;R]

Here ωg (δ; R) → 0 and ωF (δ; R) → 0 as δ → 0.

Now we consider an interval [0, T ] and restrict to step-sizes hn ∈ [δ/2, δ] for
Pn
some δ > 0. Note then that tn+1 = j=0 hj then satisfies nδ/2 ≤ tn ≤ nδ with tn
≤ T , which means nδ ≤ 2T for these choices of n. The above difference inequality
 Pullback attractors under discretization 355

thus satisfies

En+1 ≤ (1 + LR δ)En + δ (ωg (δ; R) + ωF (δ; R) + LR MR δ)

and hence with E0 = 0 yields

(1 + LR δ)n − 1
En ≤ δ (ωg (δ; R) + ωF (δ; R) + LR MR δ)
(1 + LR δ) − 1

1 LR nδ
≤ (ωg (δ; R) + ωF (δ; R) + LR MR δ) e
LR

1 2LR T
≤ (ωg (δ; R) + ωF (δ; R) + LR MR δ) e ,
LR
that is
1 2LR T
|x(tn ) − un | ≤ (ωg (δ; R) + ωF (δ; R) + LR MR δ) e .
LR
Hence for t ∈ (tn , tn+1 ), we have

|x(t) − un | ≤ |x(t) − x(tn )| + |x(tn ) − un | ≤

Z t 
  1 2LR T
 g(σs ω, x(s)) ds + (ωg (δ; R) + ωF (δ; R) + LR MR δ) e ≤
 LR
tn
1 2LR T
δMR + (ωg (δ; R) + ωF (δ; R) + LR MR δ) e .
LR
Let us now consider variable parameters and initial values. let ωj → ω in Ω
and u0j → u0 in Rd . Let ϕ(t, u0 , ω) and un (u0 , ω), etc, denote the corresponding
solutions. By continuity in initial conditions and parameters uniformly on a compact
time interval [0, T ], we have

ϕ(t, u0j , ωj ) → ϕ(t, u0 , ω),

as j → ∞ for t ∈ [0, T ].
Combining all of these partial results for t ∈ (tn , tn+1 ) we obtain

|un (u0j , ωj ) − x(t, u0 , ω)| ≤ |un (u0j , ωj ) − x(t, u0j , ωj )| +

|x(t, u0j , ωj ) − x(t, u0 , ω)| ≤ δMR + ((ωg (δ; R) + ωF (δ; R) +
1 2LR T
LR MR δ) e + |x(t, u0j , ωj ) − x(t, u0 , ω)|
LR
which converges to zero as the maximum step-size δ converges to zero and j tends
to ∞.
356 Global Attractors of Non-autonomous Dissipative Dynamical Systems
 Chapter 11

Global attractors of non-autonomous

Navier-Stokes equations

This chapter is devoted to the study of non-autonomous Navier-Stokes equations.

It is proved that such systems admit compact global attractors. This problem is
formulated and solved in the terms of general non-autonomous dynamical systems.
We give conditions of convergence of non-autonomous Navier-Stokes equations . A
test of existence of almost periodic (quasi periodic, recurrent, pseudo recurrent)
solutions of non-autonomous Navier-Stokes equations is given. We prove the global
averaging principle for non-autonomous Navier-Stokes equations.
We consider the two-dimensional Navier-Stokes system
2
X
u0 + q(t) ui ∂i u = ν∆u − ∇p + φ(t)
i=1
div u = 0, u|∂D = 0, (11.1)

where D is an open bounded set with boundary ∂D ∈ C 2 . This equation can be

written in the following form

u0 + Au + B(t)(u, u) = f (t) (11.2)

on the corresponding Sobolev’s space E, where −A is a Stokes operator, B(t) is a

bilinear form satisfying the identity

RehB(t)(u, v), wi = −RehB(t)(u, w), vi (11.3)

for all t ∈ R and u, v, w ∈ E, and f is forcing term.

In the work [112],[140],[199],[200] there is studied a non-stationary equation
(11.2), when f is a function of time t ∈ R. It is shown that the equation with
compact f (in particularly, almost periodic) admits a compact global attractor and
also for small nonlinear (bilinear) term it was proved the existence a unique almost
periodic (quasi periodic, periodic) solution of equation (11.2) if the forcing term f
is almost periodic (quasi periodic, periodic).
The aim of the chapter is to study the equation (11.2) in the case, when the the
bilinear operator B, and the function f are non-stationary. The conditions under

357
358 Global Attractors of Non-autonomous Dissipative Dynamical Systems

which a non-stationary equation of type (11.2) admits a compact global attractor

are indicated.
The theorem of ”partial” averaging on finite interval for ordinary differential
equations it was proved in the work [183]. The works [140],[199] and [200] are
devoted to generalization of method of averaging for dissipative partial differential
equations. We prove the theorem of ”partial” averaging for non-autonomous Navier-
Stokes equation (11.2) (i.e. the bilinear form and forcing term are non-stationaries).
This chapter is organized as follows:
In Section 1 we introduce a class of non-autonomous Navier-Stokes equations
and establish its dissipativity (Theorem 11.4).
In Section 2 we prove that non-autonomous Navier-Stokes equations admit a
compact global attractor (Theorem 11.6).
Section 3 is devoted to study of the problem of existence of almost periodic
(quasi periodic, recurrent, pseudo recurrent) solutions of non-autonomous Navier-
Stokes equations (Corollaries 11.3) and we give the conditions of convergence of this
equations (Theorem 11.8).
In Section 4 we prove the uniform averaging principle for the non-autonomous
Navier-Stokes equations on the finite segment (Theorem 11.9).
Section 5 is devoted to prove the global averaging principle for non-autonomous
Navier-Stokes equations on the semi-axis (Theorems 11.10,11.11 and 11.12).

11.1 Non-autonomous Navier-Stokes equations

Some results from the theory of semigroups of linear operators [188] and PDEs
[199], [294],[314] are collected below.
A closed operator A with domain D(A) that is dense in a Banach space X is
called a sectorial operator if for some a ∈ R and ϕ ∈ (0, π2 ) the sector

Sa,ϕ := {λ ∈ C, π ≥ |arg(λ − a)| ≥ ϕ} (11.4)

is contained in the resolvent set and for λ ∈ Sa,ϕ

c
k(λI − A)−1 kX→X ≤ . (11.5)
|λ − a| + 1

For a sectorial operator A the analytic semigroup of linear bounded operators

in X is defined and denoted by e−At , t ≥ 0.
Let A be a sectorial operator with Reσ(A) > 0. For α ∈ (0, 1) we define fractional
powers of A as follows:
Z ∞
1
Aα := (A−α )−1 , where A−α := tα−1 e−At dt.
Γ(α) 0
 Global attractors of non-autonomous Navier-Stokes equations 359

The corresponding domains D(Aα ) are Banach spaces with norm given by

| · |α := | · |D(Aα ) = |Aα · |.

Theorem 11.1 The following estimates are valid:

(1)

ke−At kX→X ≤ Ce−at , t ≥ 0, (11.6)

(2)

kAα e−At kX→X ≤ Cα t−α e−at , t > 0. (11.7)

Let Ω be a compact metric space, R = (−∞, +∞), (Ω, R, σ) be a dynamical

system on Ω, E be a real or complex Hilbert space, L(E) be the space of all linear
continuous operators on E, L2 (E) be the space of all bilinear continuous operators
B : E × E → F and C(Ω, W ) be a space of all continuous functions f : Ω → W (W
is some metric space), endowed with the topology of uniform convergence. Let us
consider the equation

u0 + Au + B(ωt)(u, u) = f (ωt), (11.8)

(ω ∈ Ω) where ωt := σ(t, ω), B ∈ C(Ω, L2 (E)), f ∈ C(Ω, E) and A is a linear

operator.
Below we will use some notions, denotations and results from [200]. Let Hilbert
spaces E, F, X satisfy E ⊂ F ; E, F, X ⊂ E, each embedding being dense and
continuous.
Operator A. We further suppose that the linear operator A is densely defined in
E and such that the linear equation

u0 + Au = 0 (11.9)

generates the c0 -semigroup of linear bounded operators

e−At : E → E, ϕ(t, x) := e−At x,

which for t > 0 can be extended to the linear bounded operators from F to E
satisfying the following estimates

ke−At kE→E ≤ Ke−at , (11.10)

ke−At kF →E ≤ Kt−α1 e−at , 0 ≤ α1 < 1, (11.11)

kAe−At kF →E ≤ Kt−α2 e−at , 0 ≤ α2 < 2. (11.12)

360 Global Attractors of Non-autonomous Dissipative Dynamical Systems

We also suppose that the following condition is satisfied

AeAt = eAt A, (11.13)

in the sense of L(F, E) := {A : F → E |A is linear and bounded } equipped with

the operationel norm.
Bilinear operator B. Denote by L2 (E, F ) the space of all bilinear continuous
operators B : E × E → F with the norm

kBk := sup{|B(u, v)|F : |u|E ≤ 1, |v|E ≤ 1}.

Let C(Ω, L2 (E, F )) be a space of all continuous mappings B : Ω → L2 (E, F ) and

CB := sup{|B(ω)(u, v)|F : ω ∈ Ω, |u|E ≤ 1, |v|E ≤ 1},

then the mapping F : Ω × E → F (F (ω, u) := B(ω)(u, u)) satisfies the following

inequality

|B(ω)(u1 , u1 ) − B(ω)(u2 , u2 )|F ≤ CB (|u1 |E + |u2 |E )|u1 − u2 |E (11.14)

for all u1 , u2 ∈ E.
From the inequality (11.14) follows that on every ball B[0, R] := {u ∈ E :
|u|E ≤ R} we have

|B(ω)(u1 , u1 ) − B(ω)(u2 , u2 )|F ≤ 2CB R|u1 − u2 |E (11.15)

for all u1 , u2 ∈ E.

Remark 11.1 The space of all the bilinear continuous operators C(Ω, L 2 (E, F ))
is a Banach space with the norm kBk := CB .

Function f . The external force f : Ω → X is continuous, i.e. f ∈ C(Ω, X).

Operators e−At . The operators e−At (t > 0) can be extended to the linear
bounded operators from X to E satisfying the estimates

ke−At kX→E ≤ Kt−β1 e−at , 0 ≤ β1 < 1, (11.16)

kAe−At kX→E ≤ Kt−β2 e−at , 0 ≤ β2 < 2, (11.17)

and the equation (11.13), this time in the sense of L(X, E) .

We suppose that the following conditions are fulfilled:

(1) there exists α > 0 such that

RehAu, ui ≥ α|u|2E (11.18)

for all u ∈ E; ;
 Global attractors of non-autonomous Navier-Stokes equations 361

(2)

RehB(ω)(u, v), wi = −RehB(ω)(u, w), vi (11.19)

for every u, v, w ∈ E and ω ∈ Ω .

Remark 11.2 a. It follows from (11.19) that

RehB(ω)(u, v), v)i = 0 (11.20)

for every u, v ∈ E and ω ∈ Ω.

b.

|B(ω)(u, v)|F ≤ CB |u|E |v|E (11.21)

for all u, v ∈ E and ω ∈ Ω, where CB := sup{|B(ω)(u, v)|F : ω ∈ Ω, u, v ∈

E, |u|E ≤ 1, and |v|E ≤ 1}.

The equation (11.8) with conditions (11.18) and (11.19) is called a non-
autonomous Navier-Stokes equation. We will consider the mild solutions of the
equation (11.8), i.e. u ∈ C([0, T ], E) and satisfy the following integral equation
Z t
u(t) = e−At x + e−A(t−s) (−B(ωs)(u(s), u(s)) + f (ωs))ds. (11.22)
0

Theorem 11.2 Let x0 ∈ E, r > 0 and the conditions (11.10),(11.11) and (11.18)
are fulfilled, then there exist positive numbers δ = δ(x0 , r) and T = T (x0 , r) such
that the equation (11.22) admits a unique solution ϕ(t, x, ω) (x ∈ B[x0 , δ] := {x ∈
E | |x − x0 | ≤ δ}) defined on the interval [0, T ] with the conditions: ϕ(0, x, ω) = x,
|ϕ(t, x, ω) − x0 | ≤ r for all t ∈ [0, T ] and the mapping ϕ : [0, T ] × B[x0 , δ] × Ω →
E ((t, x, ω) → ϕ(t, x, ω)) is continuous.

Proof. Let x0 ∈ E, r > 0, δ > 0 and T > 0. We consider a space Cx0 ,r,δ,T of all
continuous functions ψ : [0, T ] × B[x0 , δ] × Ω → B[x0 , r] equipped with the distance

d(ψ1 , ψ2 ) := sup{|ψ1 (t, x, ω) − ψ(t, x, ω)|E : 0 ≤ t ≤ T, x ∈ B[x0 , δ], ω ∈ Ω}

is a complete metric space.

We define the operator Φ acting onto Cx0 ,r,δ,T by the equality
Z t
−At
(Φψ)(t, x, ω) = e x+ e−A(t−s) (−B(ωs)(ψ(s, x, ω), ψ(s, x, ω)) + f (ωs))ds.
0
362 Global Attractors of Non-autonomous Dissipative Dynamical Systems

There exist δ1 = δ1 (x0 , r) > 0 and T1 = T1 (x0 , r) > 0 such that ΦCx0 ,r,δ,T ⊆
Cx0 ,r,δ,T for all δ ∈ (0, δ1 ] and T ∈ (0, T1 ]. In fact,

|(Φψ)(t, x, ω) − x0 |E ≤ |e−At x − x0 |E +
Z t Z t
| e−A(t−s) B(ωs)(ψ(s, x, ω), ψ(s, x, ω)))ds|E + | e−A(t−s) f (ωs)ds|E ≤
0 0
Z t
m(δ, T ) + Ke−a(t−s) (t − s)−α1 |ψ(s, x, ω)|2E ds +
0
Z t
T 1−α1
Ke−a(t−s) (t − s)−β1 kf kds ≤ m(δ, T ) + K(|x0 |E + r)2
0 1 − α1
1−β1
T
+Kkf k := d1 (x0 , r, δ, T ) → 0
1 − β1
as δ + T → 0, where m(δ, T ) := sup{|e−tA x − x0 |E : t ∈ [0, T ], x ∈ B[x0 , r]}
and kf k := sup{|f (ω)|X : ω ∈ Ω}. Thus there exist δ1 = δ1 (x0 , r) > 0 and
T1 = T1 (x0 , r) > 0 such that d1 (x0 , r, δ, T ) ≤ r for all δ ∈ (0, δ1 ] and T ∈ (0, T1 ].
Let now ψ1 , ψ2 ∈ Cx0 ,r,δ,T , then

|(Φψ1 )(t, x, ω)) − (Φψ2 )(t, x, ω))|E =

Z t
| [B(ωs)(ψ1 (s, x, ω), ψ1 (s, x, ω)) − B(ωs)(ψ2 (s, x, ω), ψ2 (s, x, ω))]|E ≤
0
2CB (|x0 |E + r)T d(ψ1 , ψ2 )

and, consequently, d(Φψ1 , Φψ2 ) ≤ L(x0 , r, T )d(ψ1 , ψ2 ), where L(x0 , r, T ) :=

2CB (|x0 | + r)T → 0 as T → 0. Thus there exists T2 = T2 (x0 , r) > 0 such
that L(x0 , r, T ) < 1 for all T ∈ (0, T2 ]. Denote by δ(x0 , r) := δ1 (x0 , r) and
T (x0 , r) := min(T1 (x0 , r), T2 (x0 , r)), then the mapping Φ : Cx0 ,r,δ,T → Cx0 ,r,δ,T
is a contraction and, consequently, there exists a unique function ϕ ∈ Cx0 ,r,δ,T
satisfying the equation (11.22) on the interval [0, T ]. The theorem is proved. 

Remark 11.3 The theorem 11.2 is true and for the equation

u0 + Au = F(ωt, u)

if the continuous function F : Ω × E → F satisfies the following conditions:

(1)

sup{|F(ω, 0)|E : ω ∈ Ω} < ∞

(Ω ,generally speaking, is not compact);

(2) F is locally Lipschitz, i.e. for every r > 0 there exists L(r) > 0 such that

|F(ω, u1 ) − F(ω, u2 )|F ≤ L(r)|u1 − u2 |E

for all u1 , u2 ∈ E with condition: |ui |E ≤ r (i = 1, 2).

 Global attractors of non-autonomous Navier-Stokes equations 363

Theorem 11.3 Let K be a family of solutions of equation (11.22) satisfying the

following condition: there exists a positive constant M such that |x(t)| D(A) ≤ M
for all t ∈ R+ (|x|D(A) := |Ax|E ). If there exists C˜B > 0 such that

|B(ω)(u, v)|F ≤ C˜B |u|D(A) |v|D(A)

for all u, v ∈ D(A) , then this family of functions is uniform equicontinuous on

R+ , i.e. for every ε > 0 there exists δ(ε) > 0 such that |t1 − t2 | < δ implies
|x(t1 ) − x(t2 )| < ε for all t1 , t2 ∈ R+ and x ∈ K.

Proof. Let ψ ∈ K and x := ψ(0), then ψ(t) = ϕ(t, x, ω) for all t ∈ R+ and we have
Z t
−At
|ϕ(t, x, ω) − x|E ≤ |e x − x|E + | e−A(t−τ ) (−B(ωs)(ϕ(s, x, ω),
0
Z t
ϕ(s, x, ω)) + f (ωs))ds|E ≤ e−as s−α1 |x|D(A) ds +
0
Z t Z t
−as −α1
e (t − s) 2
Cb |ϕ(s, x, ω)|D(A) ds + e−a(t−s) (t − s)−β1 kf kds ≤
0 0
t1−α1 t1−α1 t1−β1
M + CB M 2 + kf k .
1 − α1 1 − α1 1 − β1
From the last inequality we obtain

sup{|ϕ(t, x, ω) − x|E | |x|D(A) ≤ M, ω ∈ Ω} → 0

as t → 0 and, consequently,

|ϕ(t2 , x, ω) − ϕ(t1 , x, ω)|E = |ϕ(t2 − t1 , ϕ(t1 , x, ω), ωt1 ) − ϕ(t1 , x, ω)|E ≤

sup{|ϕ(t2 − t1 , x, ω) − x|E : |x|D(A) ≤ M, ω ∈ Ω} → 0

as t2 − t1 → 0. The theorem is proved. 

Example 11.1 We consider the two-dimensional Navier-Stokes system

2
X
u0 + q(t) ui ∂i u = ν∆u − ∇p + φ(t) (11.23)
i=1
div u = 0, u|∂D = 0,

where D is an open bounded set with smooth boundary ∂D ∈ C 2 .

The functional setting of the problem is well known [228],[311] . We denote
by H and V the closure of the linear space {u ∈ C0∞ (D)2 , div u = 0} in L2 (D)2
and H01 (D)2 , respectively. Denote by P the corresponding orthogonal projection
P : L2 (D)2 → H. We further set
2
X
A := −νP ∆, B(t)(u, v) := q(t)P ( ui ∂i v).
i=1
364 Global Attractors of Non-autonomous Dissipative Dynamical Systems

The Stokes operator A is self-adjoint positive with domain D(A) dense in H.

The inverse operator is compact. We define the Hilbert spaces D(Aα ), α ∈ (0, 1] as
the domains of the powers of A in the standard way. Furthermore, V := D(A1/2 ),
and |u|D(A1/2 ) = |∇u|.
Applying P we write (11.23) as the evolution equation of the following form

u0 + Au + B(t)(u, u) = F(t), F(t) := P φ(t). (11.24)

Let F ∈C(R,H) (X := H) and B ∈C(R,L2 (H,D(A−δ )) (F := D(A−δ )). De-

note by Y := C(R, H)×C(R,L2 (H,D(A−δ ))) and (Y, R, σ) a dynamical system of
translations (Bebutov’s dynamical system, see for example, [300],[302] and [292]).
Let Ω := H(B, F) = {(Bτ , Fτ )| τ ∈ R}, where Bτ (t) := B(t + τ ) ( respectively
Fτ (t) := F(t + τ )) for all t ∈ R, by bar we denote a closure in the compact-open
topology and (Ω, R, σ) be a dynamical system of translations on Ω.
Along with the equation (11.24) we consider its H-class

u0 + Au + B̃(t)(u, u) = F̃(t), (11.25)

where (B̃, F̃) ∈ H(B, F). Let B : Ω → L2 (H, D(A−δ )) (respectively f : Ω → H) be

a mapping defined by equality

˜ (f (ω) = f (B̃, F̃) := F(0)),

B(ω) = B(B̃, F̃) := B(0) ˜

where ω = (B̃, F̃) ∈ Ω, then the equation (11.24) and its H-class can be written in
the form (11.21).
We now set in the notation above E := D(A1/2 ), X := H, F := D(A−δ ) and see
that (11.10)-(11.12), (11.18) and (11.16)-(11.17) are valid with α 1 = 1/2 + δ, β1 =
1/2, β2 = 3/2.

According to Theorem 11.2 through every point x ∈ H passes a unique solution

ϕ(t, x, ω) of equation (11.8) at the initial moment t = 0. And this solution is defined
on some interval [0, t(x,ω) ). Let us note, that

w0 (t) = 2Rehϕ0 (t, x, ω), ϕ(t, x, ω)i = 2RehA(ωt)ϕ(t, x, ω), ϕ(t, x, ω)i
+2RehB(ωt)(ϕ(t, x, ω), ϕ(t, x, ω)), ϕ(t, x, y)i + 2Rehf (ωt), ϕ(t, x, ω)i
= 2RehA(ωt)ϕ(t, x, ω), ϕ(t, x, ω)i + 2Rehf (ωt), ϕ(t, x, ω)i
≤ −2α|ϕ(t, x, ω)|2E + 2kf k|ϕ(t, x, ω)|E , (11.26)

where kf k := max{|f (ω)|X : ω ∈ Ω} and w(t) = |ϕ(t, x, ω)|2E . Then

1
w0 ≤ −2αw + 2kf kw 2 (11.27)
 Global attractors of non-autonomous Navier-Stokes equations 365

and consequently

w(t) ≤ v(t) (11.28)

for all t ∈ [0, t(x,ω) ), where v(t) is an upper solution of equation

1
v 0 = −2αv + 2||f ||v 2 , (11.29)

satisfying condition v(0) = w(0) = |x|2 . Thus

kf k −αt kf k 2
v(t) = [(|x|E − )e + ] (11.30)
α α
and consequently
kf k −αt kf k
|ϕ(t, x, ω)|E ≤ (|x|E − )e + (11.31)
α α
for all t ∈ [0, t(x,ω) ). It follows from the inequality (11.27) that solution ϕ(t, x, ω) is
bounded and therefore it may be prolonged on R+ = [0, +∞).
Thus we have proved the following theorem.

Theorem 11.4 Let the conditions (11.18) and (11.19) are fulfilled. Then the
following statements hold:

(i) Every solution ϕ(t, x, ω) of non-autonomous Navier-Stokes equation (11.8) is

bounded and therefore it may be prolonged on R+ .
(ii)

|ϕ(t, x, ω)|E ≤ C(|x|E ), (11.32)

kf k
for all t ≥ 0, ω ∈ Ω and x ∈ E, where C(r) = r if r ≥ r0 := α and
C(r) = r0 if r ≤ r0 ;
(iii)
||f ||
lim sup sup{|ϕ(t, x, ω)|E : |x|E ≤ r, ω ∈ Ω} ≤ (11.33)
t→+∞ α
for every r > 0.

Lemma 11.1 Under the conditions of Theorem 11.4 we have

Z t+l
r2 r
|ϕ(τ, x, ω)|2E dτ ≤ + lkf k := M (r) (11.34)
t 2α α
for all t ≥ 0 and r ≥ r0 .

Proof. From the equality (11.27) after integration in t between t and t+l we obtain
Z t+l
2α |ϕ(τ, x, ω)|2E dτ ≤ |ϕ(t, x, ω)|2E + 2rlkf k (11.35)
t
366 Global Attractors of Non-autonomous Dissipative Dynamical Systems

and,consequently,
Z t+l
r2 r
|ϕ(τ, x, ω)|2E dτ ≤ + lkf k := M (r). (11.36)
t 2α α


Lemma 11.2 ([314, Ch.3]) (The Uniform Gronwall Lemma). Let g, h, y, be

three positive locally integrable functions on ]t0 , ∞[ such that y 0 is locally integrable
on ]t0 , ∞[, and which satisfy

y 0 ≤ gy + h for t ≥ t0 ,
Z t+l Z t+l Z t+l
g(s)ds ≤ a1 , h(s)ds ≤ a2 , y(s)ds ≤ a3 for t ≥ t0 ,
t t t

where l, a1 , a2 , a3 , are positive constants. Then

a3
y(t + l) ≤ ( + a2 )ea1 ∀ t ≥ t0 .
l

Theorem 11.5 Under the conditions of Theorem 11.4 if

|hB(ω)(u, v), wi| ≤ C|u|1/2 |Au|1/2 |v|1/2 |w| (11.37)

∀ u ∈ D(A), v ∈ V, w ∈ H,

then

|ϕ(t, x, ω)|D(A) ≤ K(r) ∀|x| ≤ r (r ≥ r0 ) (11.38)

for all t ≥ 0 and ω ∈ Ω, where K(r) is some positive constant depending only on r.

Proof. Since

1 d
hAϕ(t, x, ω), ϕ(t, x, ω)i = |ϕ(t, x, ω)|2E (11.39)
2 dt
by taking the scalar product of (11.8) with Au we find

1 d
|ϕ(t, x, ω)|2E + |Aϕ(t, x, ω)|2E + (11.40)
2 dt
hB(ϕ(t, x, ω), ϕ(t, x, ω)), Aϕ(t, x, ω)i = hf (ωt), Aϕ(t, x, ω)i.

Taking into account the inequality

|hf (ωt), Aϕ(t, x, ω)i| ≤ |f (ωt)|E |Aϕ(t, x, ω)|E ≤

1
|Aϕ(t, x, ω)|2 + kf k2 (11.41)
4
 Global attractors of non-autonomous Navier-Stokes equations 367

and using (11.39) and the Young inequality we obtain

|hB(ω)(ϕ(t, x, ω), ϕ(t, x, ω)), Aϕ(t, x, ω)i| ≤

c1 |ϕ(t, x, ω)|1/2 kϕ(t, x, ω)k|Aϕ(t, x, ω)|3/2 ≤ (11.42)
1
|Aϕ(t, x, ω)|2 + c01 |ϕ(t, x, ω)|2 kϕ(t, x, ω)k4 .
4
Hence
1 d 1
kϕ(t, x, ω)k2 + |Aϕ(t, x, ω)|2 ≤ kf k2 + |Aϕ(t, x, ω)|2 +
2 dt 2
c01 |ϕ(t, x, ω)|2 kϕ(t, x, ω)k4 . (11.43)

From this inequality according to Gronwal lemma we can prove that

|ϕ(t, x, ω)|D(A) is uniformly (w.r.t. x and ω) bounded on interval [0, l]. Applying
the uniform Gronwal lemma with g, h, y replaced by

c01 |ϕ(t, x, ω)|2 kϕ(t, x, ω)k2 , kf k2 , kϕ(t, x, ω)k2 (11.44)

we obtain that kϕ(t, x, ω)k2 is bounded on [l, ∞[ and, consequently, it is bounded

on [0, ∞[ uniformly w.r.t. kxk ≤ r and ω ∈ Ω. The theorem is proved. 

11.2 Attractors of non-autonomous dynamical systems

Lemma 11.3 The non-autonomous Navier-Stokes equation (11.8) generates a

cocycle ϕ ( or more explicit hE, ϕ, (Ω, R, σ)i), where ϕ(t, x, ω) is a unique solution
of equation (11.8) defined on R+ with the initial condition ϕ(0, x, ω) = x.

Proof. In fact, according to Theorems 11.2 the mapping ϕ : ×E × Ω →

E ((t, x, ω) → ϕ(t, x, ω)) is continuous and in view of uniqueness of solution
ϕ(t, x, ω) we have the following identity: ϕ(t + τ, x, ω) = ϕ(t, ϕ(τ, x, ω), ωτ ) for
all t, τ ∈ R+ , x ∈ E and ω ∈ Ω, where ωτ := σ(τ, ω). 

Example 11.2 Let E be a Banach space and C(R × E, E) be a space of all

continuous functions F : R × E → E equipped with the compact-open topology.
Let us consider a parameterized differential equation

dx
+ Ax = F (σt ω, x) (ω ∈ Ω)
dt
on a Banach space E with Ω := C(R × E, E), where σt ω := σ(t, ω) and the linear
operator A is densely defined in E and such that the linear equation

u0 + Au = 0
368 Global Attractors of Non-autonomous Dissipative Dynamical Systems

generates the c0 -semigroup of linear bounded operators

e−At : E → E, ϕ(t, x) := e−At x.

We will define σt : Ω → Ω by σt ω(·, ·) = ω(t + ·, ·) for each t ∈ R and interpret

ϕ(t, x, ω) as mild solution of the initial value problem
d
x(t) + Ax = F (σt ω, x(t)), x(0) = x. (11.45)
dt
Under appropriate assumptions on F : Ω × E → E (or even F : R ×E → E with
ω(t) instead of σt ω in (11.45)) to ensure forwards existence and uniqueness, then
ϕ is a cocycle on (C(R × E, E), R, σ) with fiber E, where (C(R × E, E), R, σ) is a
Bebutov’s dynamical system (see for example [32],[102],[292]) [300].

Definition 11.1 Recall that the cocycle hW, ϕ, (Y, R, σ)i is called compact
(asymptotically compact) if the skew-product dynamical system (X, R+ , π) (X =
W × Y, π = (ϕ, σ)) is compact (respectively asymptotic compact).

Let (X, R+ , π) be compact dissipative and K be a compact set, which attracts

all compact subsets of X. Suppose

J = Ω(K), (11.46)
T S
where Ω(K) = t≥0 τ ≥t π τ K. The set J is called the Levinson’s center of the
compact dissipative system (X, R+ , π).
Applying general theorems about non-autonomous dissipative systems (see
Chapter 2) to non-autonomous system constructed in the example 11.2, we will
obtain series of facts concerning the equation (11.8). In particular, from Theorems
11.4, 2.19 and 2.20 follows the theorem below.

Theorem 11.6 Let Ω be a compact metric space, (Ω, R, σ) be a dynamical system

on Ω and the conditions (11.18) and (11.19) are fulfilled. If the cocycle ϕ generated
by non-autonomous Navier-Stokes equation is asymptotically compact, then for ev-
ery ω ∈ Ω there exists a non-empty compact and connected Iω ⊂ E such that the
following conditions hold:

(1) the set I = ∪{Iω : ω ∈ Ω} is compact and connected in E;

(2) I is connected if Ω is connected;
(3)

lim sup β(U (t, ω −t )M, I) = 0

t→+∞ ω∈Ω

for any bounded set M ⊂ E, where U (t, ω) = ϕ(t, ·, ω) and β is the semi-
distance of Hausdorff;
(4) U (t, ω)Iω = Iωt for all t ∈ R+ and ω ∈ Ω;
 Global attractors of non-autonomous Navier-Stokes equations 369

(5) Iω consists of those and only those points x ∈ E through which passes the
solution of the equation (11.8) bounded on R.

Theorem 11.7 Under conditions of Theorem 11.6

kf k
|ϕ(t, x, ω)| ≤
α
for all t ∈ R, ω ∈ Ω and x ∈ Iω , where ϕ is a cocycle, generated by non-autonomous
Navier-Stokes equation.

S
Proof. The set J = {Iω × {ω} : ω ∈ Ω} is a Levinson’s center of dynamical
system (X, R+ , π) and according to (11.46) for any point (u0 , y0 ) = z ∈ J there
exists tn → +∞, un ∈ E and ωn ∈ Ω such that the sequence {un } is bounded,
u0 = lim ϕ(tn , un , ωn ) and ω0 = lim ωn tn . From the inequality (11.27) follows
n→+∞ n→+∞
that |u0 | ≤ kfαk , i.e. ϕ(t, x, ω) ∈ Iωt for all ω ∈ Ω and t ≥ 0, hence |ϕ(t, x, ω)| ≤ kf k
α
for any t ∈ R, x ∈ Iω and ω ∈ Ω. The theorem is proved. 

11.3 Almost periodic and recurrent solutions of non-autonomous

Navier-Stokes equations

Definition 11.2 An autonomous dynamical system (Ω, T, σ) is said to be pseudo

recurrent if the following conditions are fulfilled:

a) Ω is compact;
b) (Ω, T, σ) is transitive, i.e. there exists a point ω0 ∈ Ω such that Ω =
{ω0 t | t ∈ T};
c) every point ω ∈ Ω is stable in the sense of Poisson, i.e.

Nω = {{tn } | ωtn → ω and |tn | → +∞} 6= ∅.

Lemma 11.4 ([105]) Let h(X, T1 , π), (Ω, T2 , σ), hi be a non-autonomous

dynamical system and the following conditions are fulfilled:

1) (Ω, T2 , σ) is pseudo recurrent;

2) γ ∈ C(Ω, X) is an invariant section of the homomorphism h : X → Ω, i.e.
h(γ(ω)) = ω and γ(σ(t, ω)) = π(t, γ(ω)) for all ω ∈ Ω and t ∈ T2 .

Then the autonomous dynamical system (γ(Ω), T2 , π) is pseudo recurrent too.

Definition 11.3 The solution ϕ(t, x, ω) of non-autonomous Navier-Stokes equa-

tion (11.8) is called recurrent (pseudo recurrent, almost periodic, quasi periodic),
if the point (x, ω) ∈ H × Ω is a recurrent (pseudo recurrent, almost periodic, quasi
370 Global Attractors of Non-autonomous Dissipative Dynamical Systems

periodic) point of skew-product dynamical system (X, R+ , π) (X = H × Ω and

π = (ϕ, σ)).

We note (see, for example, [238] and [300],[302]) that if ω ∈ Ω is a stationary

(τ -periodic, almost periodic, quasi periodic, recurrent) point of dynamical system
(Ω, R, σ) and h : Ω → X is a homomorphism of dynamical system (Ω, R, σ) onto
(X, R+ , π), then the point x = h(ω) is a stationary (τ -periodic, almost periodic,
quasi periodic, recurrent) point of the system (X, R+ , π).
Let X := H × Ω and π := (ϕ, σ), then mapping h : Ω → X is a homomorphism
of dynamical system (Ω, R, σ) onto (X, R+ , π) if and only if h(ω) = (u(ω), ω) for
all ω ∈ Ω, where u : Ω → H is a continuous mapping with the condition that
u(ωt) = ϕ(t, u(ω), ω) for all ω ∈ Ω and t ∈ R+ .
The following affirmations hold:

Lemma 11.5 Let Ω be a compact metric space, A be a linear operator densely

defined in E such that the equation

x0 + Ax = 0

generates a c0 -semigroup {U (t)}t≥0 . If the condition (11.18) is fulfilled, then

kU (t)k ≤ exp(−αt)

for all t ∈ R+ , where U (t) is a Cauchy operator of equation (11.48).

Proof. Let ϕ(t, x) := U (t)x, then according to the inequality 11.18 we have
d
|ϕ(t, x)|2 ≤ −2α|ϕ(t, x)|2
dt
and, consequently,|ϕ(t, x)| ≤ exp(−αt)|x| for all x ∈ H and t ∈ R+ . Thus we have
|U (t)x| ≤ exp(−αt)|x|, therefore kU (t)k ≤ exp(−αt) for all t ∈ R+ . 

Lemma 11.6 Suppose that the condition (11.18) is fulfilled. Then for every func-
tion f ∈ C(Ω, H) there exists a unique function γ ∈ C(Ω, H) defined by equality
Z 0
γ(ω) = U (−τ )f (ωτ )dτ
−∞

such that

γ(ωt) = ϕ(t, γ(ω), ω) (11.48)

for every ω ∈ Ω and t ∈ R+ , where ϕ(t, x, ω) is a solution of equation

u0 = Au + f (ωt)
 Global attractors of non-autonomous Navier-Stokes equations 371

with the initial condition ϕ(0, x, ω) = x and the following inequality

1
kγk ≤ kf k
α
takes place.

Proof. The formulated statement results from Lemma 11.5 and Proposition 7.33
from [47]. 

Lemma 11.7 Let Ω be a compact metric space, the cocycle ϕ, generated by the
non-autonomous Navier-Stokes equation (11.8) and α−2 kf kCB < 1, then the follow-
ing inequality
kf k
|ϕ(t, x1 , ω) − ϕ(t, x2 , ω)| ≤ e−(α−CB α )t |x1 − x2 |
kf k
(x1 , x2 ∈ B[0, r0 ], r0 := , ω ∈ Ω and t ∈ R+ )
α
takes place.

Proof. Let r0 := kfαk and x1 , x2 ∈ B[0, r0 ] := {x ∈ E : |x| ≤ r0 }. According to

Theorem 11.4 we have |ϕ(t, xi , ω)| ≤ r0 for all t ≥ 0, ω ∈ Ω and i = 1, 2. Denote by
ψ(t) := ϕ(t, x1 , ω) − ϕ(t, x2 , ω), then we obtain
d
|ψ(t)|2 = 2RehAψ(t), ψ(t)i + 2RehB(ωt)(ψ(t), ϕ(t, x2 , ω)), ψ(t)i ≤
dt
−2α|ψ(t)|2 + 2CB |ϕ(t, x2 , ω)||ψ(t)|2 ≤
kf k kf k
−2α|ψ(t)|2 + 2CB |ψ(t)|2 = −2(α − CB )|ψ(t)|2
α α
and, consequently,
kf k
|ψ(t)|2 ≤ e−2(α−CB α )t |ψ(0)|2 .

Thus we have
kf k
|ϕ(t, x1 , ω) − ϕ(t, x2 , ω)| ≤ e−(α−CB α )t |x1 − x2 |
(x1 , x2 ∈ B[0, r0 ], ω ∈ Ω and t ∈ R+ )

for all x1 , x2 ∈ B[0, r0 ], ω ∈ Ω and t ∈ R+ . The Lemma is proved. 

Theorem 11.8 Let Ω be a compact metric space, ϕ be the cocycle, generated by

the non-autonomous Navier-Stokes equation (11.8) and kfαkC
2
B
< 1. Then there
exists a function γ ∈ C(Ω, B[0, r0 ]) such that:

a.

γ(ωt) = ϕ(t, γ(ω), ω) (11.49)

372 Global Attractors of Non-autonomous Dissipative Dynamical Systems

for every ω ∈ Ω and t ∈ R+ , where ϕ(t, x, ω) is a solution of equation (11.8)

with the initial condition ϕ(0, x, ω) = x;
b.
kf k
kγk ≤ ;
α
(11.50)

c.
kf k
|ϕ(t, x, ω) − γ(ωt)| ≤ e−(α−CB α )t |x − γ(ω)| (11.51)

for all x ∈ E, ω ∈ Ω and t ∈ R+ , where kγk := sup{|γ(ω)| : ω ∈ Ω}.

Proof. Let Γ := C(Ω, B[0, r0 ]) (C(Ω, E)) be a space all the continuous functions
f : Ω → B[0, r0 ] (respectively f : Ω → E ) equipped with the distance

d(f1 , f2 ) = max{|f1 (ω) − f2 (ω)| : ω ∈ Ω} .

Then (Γ, d) ( respectively (C(Ω, E), d)) is a complete metric space.

Let t ∈ R+ . We define the mapping S t : Γ → C(Ω, E) by the equality

(S t ν)(ω) := U (t, ω −t )ν(ω −t )

for all ω ∈ Ω, where ω −t := σ(−t, ω) and U (t, ω) := ϕ(t, ·, ω). According to Theorem
11.4 we have S t (Γ) ⊆ Γ for all t ∈ R+ . It easy to see that the family of mappings
{S t | t ∈ R+ } possesses the following properties:

(1)

S 0 = IdΓ

and
(2)

S t+τ = S t S τ

for all t, τ ∈ R+ .
Thus {S t | t ∈ R+ } forms a commutative semigroup with identity element. Now
we will show that the mapping S t (t > 0) is a contraction. In fact, let ν1 , ν2 ∈ Γ,
then we have

(S t ν1 )(ω) − (S t ν2 )(ω) = U (t, ω −t )ν1 (ω −t ) − U (t, ω −t )ν2 (ω −t ). (11.52)

From the lemma 11.7 and the equality (11.52) it follows that
kf k
d(S t ν1 , S t ν2 ) ≤ e−(α−CB α )t d(ν1 , ν2 )
 Global attractors of non-autonomous Navier-Stokes equations 373

for all t ∈ R+ and, consequently, there exists a unique common fixed point γ ∈ Γ,
i.e. S t γ = γ for all t ∈ R+ . In particularly

U (t, ω −t )γ(ω −t ) = γ(ω)

for all t ∈ R+ and ω ∈ Ω. From this equality follows that

γ(ωt) = U (t, ω)γ(ω) = ϕ(t, γ(ω), ω)

for all t ∈ R+ and ω ∈ Ω.

Let x ∈ E, ϕ(t, x, ω) be a unique solution of equation (11.8) with the initial
condition ϕ(0, x, ω) = x and γ ∈ Γ the function with the property (11.49). Denote
by ψ(t) := ϕ(t, x, ω) − γ(ωt), then we have
d
|ψ(t)|2 = 2RehAψ(t), ψ(t)i + 2RehB(ωt)(ψ(t), γ(ωt)), ψ(t)i ≤
dt
−2α|ψ(t)|2 + 2CB |γ(ωt)||ψ(t)|2 ≤ −2α|ψ(t)|2 +
kf k kf k
2CB |ψ(t)|2 = −2(α − CB )|ψ(t)|2
α α
and, consequently,
kf k
|ψ(t)|2 ≤ e−2(α−CB α )t |ψ(0)|2 .

Thus we have
kf k
|ϕ(t, x, ω) − γ(ωt)| ≤ e−(α−CB α )t |x − γ(ω)|

for all x ∈ E, ω ∈ Ω and t ∈ R+ . The theorem is proved. 

Corollary 11.1 Under the conditions of Theorem 11.8 there exists a unique func-
tion γ ∈ C(Ω, E) such that

γ(ωt) = ϕ(t, γ(ω), ω) (11.53)

for all t ∈ R+ and ω ∈ Ω.

Proof. Let γ ∈ C(Ω, E) be a function satisfying the equality (11.53) and γ ∈ Γ =

C(Ω, B[0, r0 ]) the function from Theorem 11.8. Since γ(ωt) = ϕ(t, γ(ω), ω) for all
t ∈ R+ and ω ∈ Ω, then according to the inequality (11.51) we have
kf k
|γ(ωt) − γ(ωt)| ≤ e−(α−CB α )t |γ(ω) − γ(ω)| (11.54)

for all t ∈ R+ and ω ∈ Ω. In particularly, from (11.54) we obtain

kf k
|γ(ω) − γ(ω)| ≤ | ≤ e−(α−CB α )t kγ − γk (11.55)
374 Global Attractors of Non-autonomous Dissipative Dynamical Systems

for all t ∈ R+ and ω ∈ Ω, where ω −t := σ(−t, ω) and kγ −γk := max{|γ(ω)−γ(ω)| :

ω ∈ Ω}. Passing to the limit in the inequality (11.55) we obtain γ(ω) = γ(ω) for all
ω ∈ Ω. 

Corollary 11.2 Under the conditions of Theorem 11.8 the equation (11.8) admits
a compact global attractor {Iω : ω ∈ Ω} and Iω = {γ(ω)} for all ω ∈ Ω, where
γ ∈ Γ is a function from Theorem 11.8.

Corollary 11.3 Let Ω be a compact minimal set containing only the periodic
(quasi periodic, almost periodic, recurrent, pseudo recurrent) motions, then under
conditions of Theorem 11.8 the non-autonomous Navier-Stokes equation (11.8))
admits a unique periodic (quasi periodic, almost periodic, recurrent, pseudo recur-
rent) solution γ(ωt) and every other solution of this equation is asymptotic peri-
odic (asymptotic quasi periodic, asymptotic almost periodic, asymptotic recurrent,
asymptotic pseudo recurrent).

Proof. Let γ ∈ Γ be a function from Theorem 11.8, then according this theorem
we have ϕ(t, γ(ω), ω) = γ(ωt) for all t ∈ R+ and ω ∈ Ω and, consequently, the
solution ϕ(t, γ(ω), ω) is periodic (quasi periodic, almost periodic, recurrent). Let
ϕ(t, x, ω) be a arbitrary solution of equation (11.8), then taking into consideration
the inequality (11.51) we conclude that ϕ(t, x, ω) is asymptotic periodic (asymp-
totic quasi periodic, asymptotic almost periodic, asymptotic recurrent, asymptotic
pseudo recurrent). 

11.4 Uniform averaging for a finite interval

We shall be dealing with the non-autonomous Navier-Stokes equation

u0 + εAu + εB(ωt)(u, u) = εf (ωt), (11.56)

where ε ∈ [0, ε0 ], A is linear and B(ω) is a bilinear operator, f is a forcing term.

Existence of partial averaged. Below we will suppose that B(ω) = B0 (ω) +
B1 (ω) (B0 , B1 ∈ C(Ω, L2 (E, F )) for all ω ∈ Ω and the average of B1 (ω) is equal to
0, that is,
Z
1 t
lim B1 (ωτ )dτ = 0 (11.57)
t→∞ t 0

uniformly with respect to ω ∈ Ω.

Remark 11.4 1. The condition (11.57) is fulfilled if a dynamical system (Ω, R, σ)

is strictly ergodic, i.e. on Ω exists a unique invariant w.r.t. (Ω, R, σ) measure µ.
 Global attractors of non-autonomous Navier-Stokes equations 375

2. According to Lemma 5.1 from [104] the equality (11.57) takes place if and
only if there exists a positive continuous on R+ function k with lim k(t) = 0 such
t→∞
that
Z
1 t
k B1 (ωτ )dτ kL2 (E,F ) ≤ k(t) (11.58)
t 0
for all ω ∈ Ω and t ∈ R+ .

The bilinear operator B0 ∈ C(Ω, L2 (E, F )) satisfies the condition

RehB0 (ω)(u, v), wi = −RehB0 (ω)(u, w), vi (11.59)

for all u, v ∈ E and w ∈ F.

The forcing term f (ω) = f0 (ω) + f1 (ω) for all ω ∈ Ω (f0 , f1 ∈ C(Ω, X)) and
the function f1 has the average which is equal to 0, i.e. there exists a positive
continuous on R+ function k1 with lim k1 (t) = 0 such that
t→∞
Z t
1
| k1 (ωτ )dτ |X ≤ k1 (t) (11.60)
t 0

for all ω ∈ Ω and t ∈ R+ .

Along with equation (11.56) we consider also the partial averaged equation

u0 + εAx + εB0 (ωt)(u, u) = εf0 (ωt). (11.61)

If we introduce the ”slow time” τ := εt (ε > 0), then the equations (11.56) and
(11.61) can be written in the following way
τ τ
u0 + Au + B(ω )(u, u) = f (ω ) (11.62)
ε ε
and
τ τ
ū0 + Aū + B0 (ω )(ū, ū) = f0 (ω ). (11.63)
ε ε
We will consider the mild solutions u(t) and ū(t) of the equations (11.62) and
(11.63), i.e. u, ū ∈ C([0, T ], E) and satisfy the following integral equations
Z τ
s s
u(τ ) = e−Aτ x + e−A(τ −s) (−B(ω )(u(s), u(s)) + f (ω ))ds, (11.64)
0 ε ε
and
Z τ
−Aτ s s
ū(τ ) = e x+ e−A(τ −s) (−B0 (ω )(ū(s), ū(s)) + f0 (ω ))ds. (11.65)
0 ε ε
Denote by ϕ(τ, x, ω, ε) (ϕ̄(τ, x, ω, ε)) a unique solution of equation (11.64) (re-
spectively (11.65)). According to Theorem 11.4 the cocycle ϕ(·, ·, ·, ε) (ϕ̄(·, ·, ·, ε)),
generated by equation (11.64) (respectively (11.65)), has an absorbing ball
B[0, R0 ] (B[0, R̄0 ]) in E, where R0 := kfαk (R̄0 := kfα0 k ). This means that for
376 Global Attractors of Non-autonomous Dissipative Dynamical Systems

every ball B[0, R] (respectively B[0, R̄]) there exists a positive number L = L(R)
(respectively L̄ = L̄(R̄)) such that

U (t, ω, ε)B[0, R] ⊆ B[0, R0 ] (11.66)

(Ū (t, ω, ε)B[0, R̄] ⊆ B[0, R̄0 ]) (11.67)

for all t ≥ L (t ≥ L̄), ε ∈ [0, ε0 ] and ω ∈ Ω, where U (t, ω, ε) := ϕ(t, ·, ω, ε)

Ū (t, ω, ε) := ϕ̄(t, ·, ω, ε).
According to Theorem 11.4 the cocycle ϕ(·, ·, ·, ε) (ϕ̄(·, ·, ·, ε)) is uniformly
bounded for t ≥ 0, that is, for every ball B[0, R1 ] (B[0, R̄1 ]) there exists a ball
B[0, R2 ] (B[0, R̄2 ]) such that

U (t, ω, ε)B[0, R1 ] ⊆ B[0, R2 ] (11.68)

(Ū (t, ω, ε)B[0, R̄1 ] ⊆ B[0, R̄2 ]) (11.69)

for all t ≥ 0, ε ∈ [0, ε0 ] and ω ∈ Ω.

Let C(R×E, E) be the space of all continuous functions f : R×E → E equipped
with compact open topology and let F ⊆ C(R × E, E).

Theorem 11.9 Let L > 0 be arbitrary but fixed. If ϕ(0, x, ω, ε) = ϕ̄(0, x, ω, ε) =

x ∈ B[0, R0 ], that is, the initial points coincide and belong to the absorbing ball of
equation (11.63) and the condition (11.32) is fulfilled, then the following relation
takes place

sup{|ϕ(t, x, ω, ε) − ϕ̄(t, x, ω, ε)|E : 0 ≤ t ≤ L, |x|E ≤ R0 , ω ∈ Ω} → 0 (11.70)

as ε → 0.

Proof. The proof below goes along the same lines as the proofs of the correspond-
ing results from [140],[199] and [200]. We set v(t) := ϕ(t, x, ω, ε) − ϕ̄(t, x, ω, ε).
Subtracting the equation (11.64) from the equation (11.65), we obtain
Z t
s
v(t) = e(t−s)A (−B(ω )(v(s), ϕ(s, x, ω, ε)) −
0 ε
Z t
s s
B(ω )(ϕ̄(s, x, ω, ε), v(s)))ds − e(t−s)A B1 (ω )(ϕ̄(s, x, ω, ε),
ε 0 ε
Z t
(t−s)A
ϕ̄(s, x, ω, ε))ds + e f1 (ωs)ds (11.71)
0
 Global attractors of non-autonomous Navier-Stokes equations 377

According to Theorem 11.4 |ϕ(t, x, ω, ε)|, |ϕ̄(t, x, ω, ε)| ≤ r0 for all t ≥ 0, where
r0 := max{ kfαk , kfα1 k }. In view of (11.71) v(t) satisfies the inequality
Z t
s
|v(t)|E ≤ | e(t−s)A (B(ω )(v(s), ϕ(s, x, ω, ε)) +
0 ε
Z t
s s
B(ω )(ϕ̄(s, x, ω, ε), v(s)))ds|E + | e(t−s)A B1 (ω )(ϕ̄(s, x, ω, ε),
ε 0 ε
Z t
s
ϕ̄(s, x, ω, ε))ds|E + | e(t−s)A f1 (ω )ds|E , (11.72)
0 ε

for all t ∈ [0, L]. By (11.11) and (11.12) we see that the first term on the right-hand
side of (11.72) is less than
Z t
2r0 KCB e−a(t−s) (t − s)α1 |v(s)|E ds . (11.73)
0

We now show that the sum of the second and third terms in (11.72) tends to
0 as ε → 0 uniformly w.r.t. t ∈ [0, L], |x| ≤ R0 and ω ∈ Ω. Before estimating the
second term, we must integrate it by parts. We have
Z t
s
e−(t−s)A B1 (ω )(ϕ̄(s, x, ω), ϕ̄(s, x, ω)ds =
0 ε
Z t
s
e−tA B1 (ω )(ϕ̄(s, x, ω), ϕ̄(s, x, ω)ds + (11.74)
0 ε
Z t Z s
τ
Ae−(t−s)A B1 (ω )(ϕ̄(τ, x, ω), ϕ̄(τ, x, ω)dτ ds.
0 0 ε

Hence
Z t
s
| e−(t−s)A B1 (ω )(ϕ̄(s, x, ω), ϕ̄(s, x, ω)ds|E ≤
0 ε
Z t
s
|e−tA B1 (ω )(ϕ̄(s, x, ω), ϕ̄(s, x, ω)ds|E + (11.75)
0 ε
Z t Z s
τ
| Ae−(t−s)A B1 (ω )(ϕ̄(τ, x, ω), ϕ̄(τ, x, ω)dτ ds ≤
0 0 ε
1−α1
L
e−aL k2 (ε) + k2 (ε) := c(ε),
1 − α1
where
Z t
s
k2 (ε) := sup{| B1 (ω )(ϕ̄(s, x, ω), ϕ̄(s, x, ω)ds|E :
0 ε
ω ∈ Ω, |x|E ≤ R0 , 0 ≤ t ≤ L}.
(11.76)
378 Global Attractors of Non-autonomous Dissipative Dynamical Systems

According to Lemma 6.13 k2 (ε) → 0 as ε → 0, because according to Theorems

11.5 and 11.3 the family of functions K := {ϕ̄(·, x, ω) : |x|E ≤ R0 , ω ∈ Ω} is
equicontinuous on R+ .
We begin with the third term. Integrating by part in s and taking into account
the inequalities (11.11)-(11.12), (11.16)-(11.17) and (11.59) we find
Z t Z t
s s
| e(t−s)A f1 (ω )ds|E = |eAt f1 (ω )ds +
0 ε 0 ε
Z t Z s Z t
(t−s)A τ s
Ae f1 (ω )dτ ds|E ≤ keAτ kX→E | f1 (ω )ds|X +
0 t ε 0 ε
Z t Z s
s
kAeA(t−s) kX→E | f1 (ω )ds|X
0 t ε
Z t
t t−s
≤ Kt1−β1 e−at k1 ( ) + K(t − s)1−β2 e−a(t−s) k1 ( )ds. (11.77)
ε 0 ε
Let α ∈ [0, 1), ν ∈ (0, 1) and β ∈ [0, 2). Since
t t
tα k1 ( ) ≤ sup tα k1 ( ) + sup tα k1 (t) ≤
ε 0≤t≤ε ν ε εν ≤t≤L

εαν k1 (0) + Lα k1 (εν−1 ) → 0 as ε → 0

and
Z t Z εν Z t
s s s
s1−β k1 ( )ds = s1−β k1 ( )ds + s1−β k1 ( )ds (11.78)
0 ε 0 ε εν ε
εν(2−β) (t2−β − ε ) ν(2−β)
≤ k1 (0) + k(εν−1 )
2−β 2−β
εν(2−β) (L2−β − εν(2−β) )
≤ k1 (0) + k(εν−1 ) as ε → 0
2−β 2−β
uniformly w.r.t. t ∈ [0, L] and ω ∈ Ω, then
Z t
s
sup | e(t−s)A f1 (ω )ds|E → 0 as ε → 0. (11.79)
0≤t≤L 0 ε

Thus,
Z t
|v(t)|E ≤ C(ε) + D (t − s)−α1 |v(s)|E ds (11.80)
0

for all t ∈ [0, L], where C(ε) := Ke−aL (L1−β1 + L1−β2 )k1 ( Lε ) + c(ε) → 0 as ε → 0
and D := 2R0 CB K.
We now use the known inequality [188, Ch.7]. If
Z t
u(t) ≤ a + b (t − s)β−1 u(s)ds, 0 < β ≤ 1, (11.81)
0
 Global attractors of non-autonomous Navier-Stokes equations 379

then

u(t) ≤ aGβ ([bΓ(β)]1/β t), (11.82)

where Gα (x) is a monotone function, while Γ(β) is a gamma function.

In our case we have

|v(t)|E ≤ C(ε)Gβ ([bΓ(β)]1/β t) ≤ (11.83)

1/β
C(ε)Gβ ([bΓ(β)] L) := d(ε) → 0 (β := 1 − α1 ∈ (0, 1])

as ε → 0 uniformly w.r.t. ω ∈ Ω, x ∈ B[0, R0 ] and t ∈ [0, L] for every L > 0. The

theorem is proved. 

11.5 The global averaging principle for Navier-Stokes equations

Let Ω be a compact metric space, (Ω, R, σ) be a dynamical system on Ω, E be a

Banach space and hE, ϕ, (Ω, R, σ)i be a cocycle on (Ω, R, σ) with fiber E.

Definition 11.4 A family of nonempty compact sets { Iω | ω ∈ Ω} (Iω ⊂ E)

is called a local attractor (local forward attractor) if the followings conditions are
fulfilled:

(1)
[
I= {Iω : ω ∈ Ω}

is compact;
(2)
λ0
ϕλ0 (t, Iωλ0 , ω) = Iσ(t,ω)

for all t ∈ R+ and ω ∈ Ω;

(3) there exists R0 > 0 such that I ⊂ B(0, R0 ) := {x ∈ E| |x| < R0 } and

lim sup β(ϕ(t, B[0, R0 ], ω), I) = 0

t→∞ ω∈Ω

(respectively lim sup β(ϕ(t, B[0, R0 ], ω), Iωt ) = 0)

t→∞ ω∈Ω

Theorem 11.10 Let Λ be a compact metric space, E be a banach space and

ϕλ (λ ∈ Λ) be a cocycle on (Ω, R, σ) with fiber E. Suppose that the following
conditions are fulfilled:

(1) the cocycle φλ0 admits a local forward attractor,

380 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(2) the following relation takes place

mL (λ) := sup{|ϕλ (t, x, ω) − ϕλ0 (t, x, ω)| :

0 ≤ t ≤ L, ω ∈ Ω, |x| ≤ R0 } → 0 (11.84)

as λ → λ0 for every positive number L;

(3) every cocycle ϕλ is asymptotically compact.

Then the next statements are valid:

a. there exists a positive number µ such that for all λ ∈ B[λ0 , µ] := {λ ∈ Λ :

ρ(λ, λ0 ) ≤ µ} the cocycle ϕλ admits in B[0, R0 ] a forward attractor {Iω : ω ∈
Ω};
b.

sup β(Iωλ , Iωλ0 ) → 0

ω∈Ω

as λ → λ0 .

Proof. Let ρ > 0 be an arbitrary small number such that B[I λ0 , ρ] ⊂ B[0, R0 ]. We
choose L = L( 3ρ ) according to the condition
ρ
ϕλ0 (t, B[0, R0 ], ω) ⊂ B[Iωλ0 , ]
3
for all ω ∈ Ω and t ≥ L( ρ3 ). Now we choose ε0 = ε0 (L) so that m(λ) < ρ3 for all
λ ∈ B[λ0 , ε0 ].
λ0 ρ
Let t1 := L, then we have ϕλ0 (t1 , x, ω) ∈ B[Iωt , ] and ϕλ (t1 , x, ω) ∈
1 3
B[0, R0 ]. We take the point x1 := ϕλ (t1 , x, ω) as the initial point and we con-
sider ϕλ (t, x1 , ωt1 ) on the segment [0, L],

ϕλ0 (t, x1 , ωt1 ); ϕλ (t, x1 , ωt1 ) = ϕλ (t, ϕλ (t1 , x, ω), ωt1 ) = ϕλ (t + t1 , x, ω).

On this segment ϕλ (t, x1 , ωt1 ) and ϕλ0 (t, x1 , ωt1 ) will diverge by the value less than
ρ λ0 ρ 2ρ
3 . Since ϕλ0 (t, x1 , ωt1 ) ∈ B[Iωt1 , 3 ], we get ϕλ (2t1 , x, ω) ∈ B[Iω2t1 , 3 ].
If we take the point x2 := ϕλ (2t1 , x, ω) as the initial one, then we see that the
situation is similar to that occurred at the previous step.
λ0
Repeating this process, we arrive at a conclusion that ϕλ (t, x, ω) ∈ B[Iωt , ρ] ⊂
ρ
B(0, R0 ) for all t ≥ L( 3 ) and ω ∈ Ω. Since the cocycle ϕλ is asymptotic compact
S
then it admits a forward attractor {Iλ : ω ∈ Ω} such that I λ := {Iωλ : ω ∈
Ω} ⊆ B[I λ0 , ρ] and, consequently, β(I λ , I λ0 ) → 0 as λ → λ0 .
λ0 ρ
Below we proved the inclusion ϕλ (t, B[0, R0 ], ω) ⊆ B[Iωt , 3 ] for all t ≥ L and
ω ∈ Ω and, consequently, we obtain
ρ
ϕλ (t, B[0, R0 ], ωt ) ⊆ B[Iωλ0 , ] (11.85)
3
 Global attractors of non-autonomous Navier-Stokes equations 381

for all t ≥ L and ω ∈ Ω. Taking onto consideration that

\ [
Iωλ = ϕλ (τ, B[0, R0 ], σ(−τ, ω)) (11.86)
t≥0 τ ≥t

from (11.85) and (11.86) it follows that Iωλ ⊆ B[Iωλ0 , ρ] for all ω ∈ Ω and λ ∈ B[λ0 , ε0 ]
and, consequently, sup β(Iωλ , Iωλ0 ) → 0 as λ → λ0 . The theorem is proved. 
ω∈Ω

Remark 11.5 1. The second condition of Theorem 11.10 is fulfilled, for example,
if the space E is finite-dimensional and the mapping ϕ : R+ × E × Ω × Λ → E,
defined by the equality ϕ(t, x, ω, λ) := ϕλ (t, x, ω), is continuous.
In fact, if we suppose that it is not true, then there exist L0 > 0, λk → λ0 , xk ∈
B[0, R0 ], tl ∈ [0, L0 ] and ωk ∈ Ω such that

|ϕλk (tk , xk , ωk ) − ϕλ0 (tk , xk , ωk )| ≥ ε0 > 0. (11.87)

Since the sets B[0, R0 ], Ω and [0, L0 ] are compacts, we can suppose that the se-
quences {xk }, {tk } and {ωk } are convergent. Denote by t0 := lim tk , x0 := lim xk
k→∞ k→∞
and ω0 := lim ωk . Passing to limit in the equality (11.87) and taking into account
k→∞
the continuity of the mapping ϕ we obtain 0 ≥ ε0 . The obtained contradiction prove
our statement.
2. Under the conditions of Theorem 11.10 if we suppose that the cocycle ϕ λ0
admits a compact global forward attractor {Iωλ0 : ω ∈ Ω}, i.e.

lim sup β(ϕλ0 (t, B[0, R], ω), Iωt ) = 0

t→∞ ω∈Ω

for every R > 0, then should be naturally to hope that for the λ sufficiently close to
λ0 the cocycle ϕλ also will admits a compact global forward attractor {I ωλ : ω ∈ Ω}
in the small neighborhood of I λ0 . Unfortunately, generally speaking, this assertion
is not true.
In fact, let ϕ0 be a cocycle (dynamical system) generated by the equation x0 =
−x and ϕλ be a cocycle generated by the equation x0 = −x + λx3 (λ > 0). It is
clear that the cocycle ϕ0 (ϕλ ) admits a compact global attractor I 0 = {0} (I λ =
[−λ−1/2 , λ−1/2 ]). In the small neighborhood of the attractor I 0 = {0} the cocycle
ϕλ ( for small λ) admits a local (but not global) attractor I λ = {0}.

Theorem 11.11 Let Λ be a compact metric space, (Ω, R, σ) be a dynamical system

on the compact metric space Ω, E be a Banach space and ϕλ (λ ∈ Λ) be a cocycle
on (Ω, R, σ) with fiber E. Suppose that the following conditions are fulfilled:

(1) the cocycle ϕλ0 admits a compact global forward attractor;

(2) the following relation takes place

mL (λ) := sup{|ϕλ (t, x, ω) − ϕλ0 (t, x, ω)| :

382 Global Attractors of Non-autonomous Dissipative Dynamical Systems

0 ≤ t ≤ L, ω ∈ Ω, |x| ≤ R0 } → 0

as λ → λ0 for every positive number L;

(3) every cocycle ϕλ admits a compact global attractor {Iωλ : ω ∈ Ω};
S
(4) the set I := {I λ : λ ∈ Λ} is bounded in E.

Then the following equality

lim sup β(Iωλ , Iωλ0 ) = 0

λ→λ0 ω∈Ω

is fulfilled and, in particularly,

lim β(I λ , I λ0 ) = 0.
λ→λ0

Proof. Suppose that the conditions of the theorem are fulfilled. According to the
condition 4. there exists a positive number R0 such that I ⊂ B(0, R0 ). Reasoning
as in Theorem 11.10 for all ρ > 0 we will find a L = L( 3ρ ) > 0 and δ0 = δ0 (ρ) > 0
such that
λ0
ϕλ (t, Iωλ , ω) ⊆ B[Iωt , ρ]

for all t ≥ L and ω ∈ Ω and, consequently,

Iωλ = ϕλ (t, Iσ(−t,ω) , σ(−t, ω)) ⊆ B[Iωλ0 , ρ]

for all ω ∈ Ω and ρ(λ, λ0 ) < δ0 . The theorem is proved. 

Theorem 11.12 Let ε ∈ (0, ε0 ), Ω be compact and connected and ϕε ( ϕ̄ε ) be a

cocycle generated by the equation (11.56) (respectively by the equation (11.61)).
Suppose that the following conditions are fulfilled:

(1) B(ω) := B0 (ω) + B1 (ω) (ω ∈ Ω), B0 , B1 ∈ C(Ω, L2 (E, F ));

(2) the bilinear forms B and B0 satisfy the condition (11.19);
Rt
(3) the average of B1 (ω) is equal to 0, i.e. lim 1t 0 B1 (ωs)ds = 0 uniformly w.r.t.
t→∞
ω ∈ Ω;
(4) the bilinear form B0 satisfies the condition (11.32);
(5) the cocycles ϕε and ϕ̄ε (ε ∈ (0, ε0 ])) are asymptotically compact .

Then the following statements are true:

a. for every ε ∈ (0, ε0 ]) and ω ∈ Ω the set Iωε := {x ∈ E : the solution ϕε (t, x, ω)
of equation (11.57) is defined and bounded on R} (respectively I¯ωε := {x ∈
E | the solution ϕ̄ε (t, x, ω) of equation (11.61) is defined and bounded on R})
is nonempty, compact and connected;
b. the cocycle ϕε (ϕ̄ε ) admits a compact global attractor {Iωε : ω ∈ Ω} (respectively
{I¯ωε : ω ∈ Ω});
 Global attractors of non-autonomous Navier-Stokes equations 383

c. the set I ε (respectively I¯ε ) is compact and connected;

S S
d. the set I := {I ε : ε ∈ [0, ε0 ]} (respectively I¯ := {I¯ε : ε ∈ [0, ε0 ]},
S S
where I¯ε := {I¯ωε : ω ∈ Ω}) is compact, where I ε := {Iωε : ω ∈ Ω},
S
I 0 = I¯0 := {I¯ω0 : ω ∈ Ω} and {I¯ω0 : ω ∈ Ω} is a compact global attractor of
equation (11.61), when ε = 1;
e. lim β(I ε , I¯0 ) = 0, where β is a semi-distance of Hausdorff;
ε→0
f. If a dynamical system (Ω, R, σ) is periodic, i.e. there exists ω0 ∈ Ω such that
ω0 τ = ω0 and Ω = {ω0 t : t ∈ [0, τ )}, then

lim sup {β(Iωε , Iω0 )} = 0.

ε→0 ω∈Ω

Proof. Let ε ∈ (0, ε0 ), Ω be compact and connected and ϕε ( ϕ̄ε ) be a cocycle

generated by the equation (11.56) (respectively by the equation (11.61)), then we
have

ϕε (t, x, ω) = ϕ(εt, x, ω, ε) (11.88)

(respectively ϕ̄ε (t, x, ω) = ϕ̄(εt, x, ω, ε)

for all t ∈ R+ , x ∈ E and ω ∈ Ω, where ϕ(·, ·, ·, ε) (respectively ϕ̄(·, ·, ·, ε)) is a

cocycle generated by the equation (11.62) (respectively (11.63)). From the equality
(11.88) it follows that {Iωε : ω ∈ Ω} (respectively {I¯ωε : ω ∈ Ω},) is a compact
global attractor of the equation (11.62) (respectively (11.63)). Now to finish the
proof of theorem it is sufficient to apply Theorems 11.9, 11.11, 7.7 and Lemmas 7.3,
7.4. The theorem is proved. 
384 Global Attractors of Non-autonomous Dissipative Dynamical Systems
 Chapter 12

Global attractors of V -monotone

dynamical systems

Differential equations with monotone right hand side make one of the most studied
classes of nonlinear equations (see, for example, [30], [185], [240], [316], [333] and
the literature quoted there).
The problem of existence of almost periodic solutions of monotone nonlinear
almost periodic equation has been studied by many authors (see [113], [114], [184],
[189], [238], [316], [333] and others).
The purpose of this chapter is to study the global attractors of general V - mono-
tone non-autonomous dynamical systems and their applications to different classes
of differential equations (ODEs, ODEs with impulse, some classes of evolutionary
partial differential equations).
For autonomous equations an analogous problem was studied before (see, for
example, [20], [43],[314]), but for non-autonomous dynamical systems this problem
is considered for the first time in our book.

12.1 Global attractors of V -monotone NDS

Lemma 12.1 Let (X, S+ , π) be a compactly k-dissipative dynamical system with

the Levinson’s center J. Then the dynamical system (X × X, S+ , π × π) (where
π×π(t, (x1 , x2 )) := (π(t, x1 ), π(t, x2 )) for all t ∈ S and x1 , x2 ∈ X) is also compactly
k-dissipative and J × J is its Levinson’s center.

Proof. Let (X, S+ , π) be a compactly k-dissipative dynamical system. Then there

exists a nonempty compact K ⊆ X such that

lim β(π t M, K) = 0 (12.1)

t→+∞

for every M ∈ C(X). Let K 0 := K × K and M 0 ∈ C(X × X). Then from the
equality (12.1) follows that

lim β((π × π)t M 0 , K 0 ) = 0

t→+∞

385
386 Global Attractors of Non-autonomous Dissipative Dynamical Systems

and, consequently, (X × X, S+ , π × π) is compactly k-dissipative and the set J 0 :=

Ω(K 0 ) is its Levinson’s center. We note that Ω(K × K) ⊆ Ω(K) × Ω(K) and,
consequently J 0 ⊆ J × J, because J = Ω(K). On the other hand, the set J × J is a
compact invariant set for the dynamical system (X ×X, S+ , π×π) and, consequently,
J × J ⊆ J 0 because J 0 is a maximal compact invariant set in (X × X, S+ , π × π).
The Lemma is proved. 

Let Ω be a compact topological space , (E, h, Ω) be locally trivial Banach strat-

ification [29] and | · | be a norm on (E, h, Ω) co-ordinate with the metric ρ on E
(that is ρ(x1 , x2 ) = |x1 − x2 | for any x1 , x2 ∈ X such that h(x1 ) = h(x2 )).
Recall that a non-autonomous dynamical system h(X,T+ ,π), (Ω,S,σ),hi is said
to be uniformly stable in positive direction on compacts of X [62], if for arbitrary
ε > 0 and K ⊆ X there is δ = δ(ε, K) > 0 such that the inequality ρ(x1 , x2 ) <
δ (h(x1 ) = h(x2 )) implies that ρ(π t x1 , π t x2 ) < ε for t ∈ T+ .
Denote by X ×X ˙ = {(x1 , x2 ) ∈ X × X | h(x1 ) = h(x2 ) }.

Definition 12.1 The non-autonomous dynamical system h(X, S+ , π), (Ω, S, σ),
hi is called (see [113], [114] and [333], [238]) V -monotone, if there exists a function
˙ → R+ with the following properties:
V : X ×X

a. V is continuous.
b. V is positively defined, i.e. V (x1 , x2 ) = 0 if and only if x1 = x2 .
˙ and t ∈ S+ .
c. V (x1 t, x2 t) ≤ V (x1 , x2 ) for all (x1 , x2 ) ∈ X ×X

Theorem 12.1 Let h(X, S+ , π), (Ω, S, σ), hi be V -monotone and compactly dissi-
pative, then it is uniformly stable in positive direction on compacts from X.

Proof. Let h(X, S+ , π), (Ω, S, σ), hi be a V -monotone non-autonomous dynamical

system and let it be not uniformly stable in positive direction on compacts from X.
Then there are ε0 > 0, a sequence {tn } ⊆ S+ (tn → +∞ as n → +∞), a sequence
δn → 0 (δn > 0), a compact K0 ⊆ X and sequences {xin } ⊆ K0 (i = 1, 2) such that

ρ(x1n , x2n ) < δn and ρ(x1n tn , x2n tn ) ≥ ε0 (12.2)

for all n ∈ N. Since the dynamical system h(X, S+ , π), (Ω, S, σ), hi is com-
pactly dissipative, then we may suppose without loss of generality that the se-
quences {xin } (i = 1, 2) and {xin tn } (i = 1, 2) are convergent. We denote by
xi = lim xin (i = 1, 2) and x̄i = lim xin tn (i = 1, 2). According to the in-
n→+∞ n→+∞
equality (12.2) we obtain x1 = x2 and x̄1 6= x̄2 . On the other hand, in view of V -
monotonicity of h(X, S+ , π), (Ω, S, σ), hi, we have

V (x1n tn , x2n tn ) ≤ V (x1n , x2n ) (12.3)

 Global attractors of V -monotone dynamical systems 387

for all n ∈ N. Passing to limit in (12.3) as n → +∞, we obtain the equality x̄1 = x̄2
which contradicts to the inequality (12.2). This contradiction proves Theorem 12.1.


Corollary 12.1 Let h(X, S+ , π), (Ω, S, σ), hi be a V -monotone compactly dissipa-
tive non-autonomous dynamical system and Ω be minimal. Then:

1. J is uniformly orbitally stable in positive direction, i.e. for ε > 0 there is

δ(ε) > 0 such that the inequality ρ(x, Jh(x) ) < δ implies that ρ(π t x, Jh(πt x) ) < ε
for t ≥ 0;
2. J is an attractor of compact sets from X, i.e. for ε > 0 and a compact K ⊆ X
there is L(ε, K) > 0 such that π t Kω ⊆ B̃(Jσt ω , ε) for ω ∈ Ω and t ≥ L(ε, K),
where B̃(M, ε) := {x ∈ X | ρ(x, Mh(x) ) < ε};
3. all motions on J can be continued to the left and J is bilaterally distal;
T
4. Jω := Xω J for ω ∈ Ω and is a connected set if Xω is connected, and for
distinct ω1 and ω2 the sets Jω1 and Jω2 are homeomorphic;
5. J is formed of recurrent trajectories, and every two arbitrary points x 1 , x2 ∈
Jω (ω ∈ Ω) are mutually recurrent.

Theorem 12.2 Let h(X, S+ , π), (Ω, S, σ), hi be a V -monotone compactly dissipa-
tive non-autonomous dynamical system, Ω be a minimal set and J be its Levinson’s
center. Then

V (x1 t, x2 t) = V (x1 , x2 ) (12.4)

for all x1 , x2 ∈ J such that h(x1 ) = h(x2 ).

Proof. Let x1 , x2 ∈ J be such that h(x1 ) = h(x2 ). Then according to Corollary 12.1
x1 , x2 ∈ Jω (h(x1 ) = h(x2 ) = ω ∈ Ω) are mutually recurrent and, consequently,
the function φ ∈ C(S, R) defined by the equality φ(t) := V (x1 t, x2 t) (t ∈ S) is
recurrent. On the other hand, since h(X, S+ , π), (Ω, S, σ), hi is V -monotone, we
obtain φ(t2 ) ≤ φ(t1 ) for all t2 ≥ t1 (t1 , t2 ∈ S). As the function φ is recurrent and
monotone, then it is a constant. In fact, if we suppose that there exists t2 > t1
such that φ(t2 ) < φ(t1 ), then in virtue of the recurrence of the function φ there
exists a sequence {sn } such that sn → +∞ as n → +∞ and φ(sn ) → φ(t1 ).
On the other hand, φ(sn ) < φ(t2 ) for a sufficiently large n and, consequently,
φ(t1 ) ≤ φ(t2 ) < ϕ(t1 ). The obtained contradiction proves Theorem 12.2. 

Corollary 12.2 Under the conditions of Theorem 12.2, if a non-autonomous

dynamical system h(X, S+ , π), (Ω, S, σ), hi is strictly monotone, i.e. V (x1 t, x2 t) <
T
˙
V (x1 , x2 ) for all t > 0 and (x1 , x2 ) ∈ X ×X (x1 6= x2 ), then Jω := J Xω consists
of a single point for all ω ∈ Ω.
388 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Theorem 12.3 Let h(X, S+ , π), (Ω, S, σ), hi be a V -monotone compactly dissipa-
tive non-autonomous dynamical system with the compact minimal base Ω and J be
its Levinson’s center. Then for every point x ∈ Xy there exists a unique recurrent
point p ∈ Jω such that

lim ρ(xt, pt) = 0, (12.5)

t→+∞

i.e. every trajectory of this system is asymptotically recurrent.

Proof. Let ω ∈ Ω and x ∈ Xω be an arbitrary point. Since h(X,S+ ,π),(Ω,S, σ),hi is

compactly dissipative, then the semi-trajectory {xt | t ∈ S+ } is relatively compact.
T
According to Lemma 22.29 from [32], there exists a recurrent point p ∈ ωx Xω
and a sequence tn → +∞ such that

lim ρ(xtn , ptn ) = 0. (12.6)

n→+∞

According to Theorem 12.1, the non-autonomous dynamical system h(X, S+ , π),

(Ω, T,σ), hi is uniformly stable in positive direction on compact K := {xt | t ∈ S+ }.
Let ε > 0 and δ = δ(ε, K) > 0 be chosen out of the uniform stability of K. Then
from equality (12.5) we have ρ(xtn , ptn ) < δ(ε, K) for all sufficiently large n and,
consequently,

ρ(x(tn + t), p(tn + t)) < ε

for all t ≥ 0 or ρ(xt, pt) < ε for every t ≥ tn , i.e. the equality (12.5) holds.
T
Thus, we found a recurrent point p ∈ ωx Xω satisfying the equality (12.5). It
is clear that p ∈ Jω . Now we will prove that point p with the property (12.5) is
unique. Let us suppose that p1 , p2 ∈ Jω and

lim ρ(xt, pi t) = 0 (12.7)

t→+∞

for i = 1, 2. Then according to Corollary 12.1 the points p1 and p2 are mutually
recurrent. In particular, there exists a sequence tn → +∞ such that

lim pi tn = pi (i = 1, 2). (12.8)

n→+∞

On the other hand, from the equality (12.7) we have

lim ρ(p1 tn , p2 tn ) = 0. (12.9)

n→+∞

From (12.8) and (12.9) follows that p1 = p2 . The theorem is completely proved. 

Corollary 12.3 Under the conditions of Theorem 12.3 the following assertions
hold:
 Global attractors of V -monotone dynamical systems 389

(1) ω-limit set (which is denoted as ωx ) of every point x ∈ X is a compact minimal

set.
T
(2) if x1 , x2 ∈ Xω ( ω ∈ Ω), then ωx1 = ωx2 or ωx1 ωx2 = ∅.

12.2 On the structure of Levinson center of V -monotone NDS

Definition 12.2 (X, ρ) is called [189] a metric space with segments, if for any
x1 , x2 ∈ X and α ∈ [0, 1] the intersection of B[x1 , αr] (a closed ball with its center
at x and radius αr, where r = ρ(x1 , x2 )) and B[x2 , (1 − α)r] consists of a unique
element S(α, x1 , x2 ).

Definition 12.3 A metric space (X, ρ) is called [189] strictly-convex, if (X, ρ) is

a metric space with segments and for any x1 , x2 , x3 ∈ X, x2 6= x3 and α ∈ (0, 1)
there holds the inequality ρ(x1 , S(a, x2 , x3 )) < max{ρ(x1 , x2 ), ρ(x1 , x3 )}.

Definition 12.4 Let X be a strictly metric-convex space. A subset M of X is

said to be metric-convex, if S(α, x1 , x2 ) ∈ M for any α ∈ (0, 1) and x1 , x2 ∈ M .

We note that every convex closed subset X of the Hilbert space H equipped
with the metric ρ(x1 , x2 ) = |x1 − x2 | is strictly metric-convex.
Let x ∈ X. Denote by Φx the family of all entire trajectories of the dynamical
system (X, S+ , π) passing through the point x for t = 0, i.e. γ ∈ Φx if and only if
γ : S → X is a continuous mapping with the properties: γ(0) = x and π t γ(τ ) =
γ(t + τ ) for all t ∈ S+ and τ ∈ S.

Theorem 12.4 Let h(X, S+ , π), (Ω, S, Θ), hi be a V -monotone compactly dissipa-
tive non-autonomous dynamical system, J be its Levinson center and the following
conditions be held:
˙
1. V (x1 , x2 ) = V (x2 , x1 ) for all (x1 , x2 ) ∈ X ×X.
2. V (x1 , x2 ) ≤ V (x1 , x3 ) + V (x3 , x2 ) for all x1 , x2 , x3 ∈ X with the condition
h(x1 ) = h(x2 ) = h(x3 ).
3. the space (Xω , Vω ) is strictly metric-convex for all ω ∈ Ω, where Xω :=
h−1 (ω) = {x ∈ X|h(x) = ω} (ω ∈ Ω) and Vω := V |Xω ×Xω .

If γxi ∈ Φxi ( i = 1, 2) and x1 , x2 ∈ Jω (ω ∈ Ω), then the function γ : S → X (γ(t) =

S(α, γx1 (t), γx2 (t)) for all t ∈ S) defines an entire trajectory of the dynamical system
(X, S+ , π).

Proof. Let ω ∈ Ω, α ∈ [0, 1] and x1 , x2 ∈ Jω . We denote by x := S(α, x1 , x2 ). Let

γxi ∈ Φxi (i = 1, 2). Consider the function γ : S → J defined by the equality

γ(t) := S(α, γx1 (t), γx2 (t)) (12.10)

390 Global Attractors of Non-autonomous Dissipative Dynamical Systems

for all t ∈ S. We will show that γ is an entire trajectory of (X, S+ , π) with the
condition γ(0) = x. In fact, according to Theorem 12.2

V (γ1 (t), γ2 (t)) = V (x1 , x2 ) = d

for all t ∈ S. Since V (γ1 (t), γ(t)) = αd for all t ∈ S and h(X, S+ , π), (Y, S, Θ), hi is
V -monotone, we have

V (γ1 (t + τ ), π t γ(τ )) = V (π t γ1 (τ ), π t γ(τ )) ≤ V (γ1 (τ ), γ(τ )) ≤ αd

and

V (γ2 (t + τ ), π t γ(τ )) = V (π t γ2 (τ ), π t γ(τ )) ≤ V (γ2 (τ ), γ(τ )) ≤ (1 − α)d

and, consequently,

π t γ(τ ) ∈ S(α, γ1 (t + τ ), γ2 (t + τ )).

So, π t γ(τ ) = γ(t + τ ) for all τ ∈ S and t ∈ S+ . Now we will prove that the function
γ is continuous. It is clear that γ is continuous on S+ . Let t0 ∈ S, t0 ≤ 0 and
t = t0 + h (|h| < δ, δ > 0). Then we have

ρ(γ(t0 + h), γ(t0 )) = ρ(π t0 +|t0 |+δ+h γ(−|t0 | − δ),

π t0 +|t0 |+δ γ(−|t0 | − δ)). (12.11)

Passing to limit in (12.11) as h → 0, we obtain lim γ(t0 + h) = γ(t0 ). 

h→0

We denote by K = {a ∈ C(S+ , R+ ) | a(0) = 0, a is strictly increasing}.

Theorem 12.5 Under the conditions of Theorem 12.4, if, in addition, a non-
autonomous dynamical system h(X, S+ , π), (Ω, S, Θ), hi is boundedly k-dissipative
and there exists a function a ∈ K with the property lim a(t) = +∞ such that
t→+∞
˙
a(ρ(x1 , x2 )) ≤ V (x1 , x2 ) for all (x1 , x2 ) ∈ X ×X, then Jω will be metric-convex for
T
all ω ∈ Ω, where Jω := J Xω and J is the Levinson’s center of (X, S+ , π).

Proof. Let ω ∈ Ω, x1 , x2 ∈ Jω , α ∈ (0, 1) and x := S(α, x1 , x2 ). We will show

that x ∈ Jω . Since J is invariant, there exist entire trajectories γ1 and γ2 such
that γi (0) = xi (i = 1, 2) and γi (t) ∈ Jθt ω for all t ∈ S. Consider the function
γ : S → J defined by the equality (12.10). According to Theorem 12.4, the mapping
γ defines an antire trajectory of the dynamical system (X, S+ , π) . We note that the
trajectory γ is bounded. In fact, by the construction of γ we have V (γ1 (t), γ(t)) ≤
αd, and according to the conditions of Theorem 12.5 we obtain a(ρ(γ(t), γ 1 (t))) ≤
V (γ(t), γ1 (t)) ≤ αd. and, consequently, ρ(γ(t), γ1 (t)) ≤ a−1 (αd) for all t ∈ S. Thus,
the trajectory γ is bounded and, taking into account that the set γ(S) is invariant
and the Levinson’s center J attracts every bounded set from X, we have x ∈ Jω .
The theorem is proved. 
 Global attractors of V -monotone dynamical systems 391

12.3 Almost periodic solutions of V -monotone systems

Let (X, ρ) be a metric space.

Definition 12.5 Recall that a function ϕ : S → X is called almost periodic (in

the sense of Bohr), if for every ε > 0 there exists a relatively dense subset A ε of R
such that

ρ(ϕ(t + τ ), ϕ(t)) < ε

for all t ∈ R and τ ∈ Aε .

Lemma 12.2 [238] Let Ω be a compact minimal set and M ⊆ X be a compact

invariant set of (X, S+ , π). If the non-autonomous dynamical system h(X, S+ , π),
(Ω, S, σ), hi is distal on M in negative direction, then the mapping ω 7−→ Mω :=
T
M Xω is continuous with respect to the Hausdorff metric.

Lemma 12.3 Let M ⊆ X be a compact invariant set of (X, S+ , π). If the non-
autonomous dynamical system h(X, S+ , π), (Ω, S, σ), hi is uniformly stable in posi-
tive direction on compacts from X, then h(X, S+ , π), (Ω, S, σ), hi is distal in negative
direction on the invariant set M .

Proof. We will prove that under the conditions of Lemma 12.3 a compact invariant
set M is distal in negative direction. Really, if we suppose the contrary, then there
exist ω0 ∈ Ω, tn ≥ 0, x1 6= x2 (x1 , x2 ∈ M and h(x1 ) = h(x2 )) and γxi ∈ Φxi (i =
1, 2) such that

ρ(γx1 (−tn ), γx2 (−tn )) → 0

as n → +∞. Let ε0 = ρ(x1 , x2 ) > 0 and δ0 = δ(ε0 , M ) > 0 is chosen from the
condition of the uniform stability in positive direction of h(X, S+ , π), (Ω, S, σ), hi on
M . Then for sufficiently large n we have

ρ(γx1 (−tn ), γx2 (−tn )) < δ0

and, consequently,

ρ(π t γx1 (−tn ), π t γx2 (−tn )) < ε0

for all t ≥ 0. In particular, for t = −tn we have

ρ(x1 , x2 ) = ρ(π tn γx1 (−tn ), π tn γx2 (−tn )) < ε0 = ρ(x1 , x2 ).

The obtained contradiction proves the necessary statement. 

Corollary 12.4 Under the conditions of Lemma 12.3, if Ω is a compact minimal

set, then the mapping ω 7−→ Jω is continuous with respect to the Hausdorff metric.
392 Global Attractors of Non-autonomous Dissipative Dynamical Systems

This assertion follows from Lemmas 12.2 and 12.3.

Lemma 12.4 Let (M, ρ) be a compact strictly metric-convex space and E be a

compact sub-semigroup of isometrics of the semigroup M M (i.e. E ⊆ M M and
ρ(ξx1 , ξx2 ) = ρ(x1 , x2 ) for all x1 , x2 ∈ M ). Then there exists a common fixed point
x̄ ∈ M of E, i.e. ξ(x̄) = x̄ for all ξ ∈ E.

Proof. Let x0 be an arbitrary point of M . We denote by `(x) = sup ρ(ξ(x), x0 )

ξ∈E
and

`0 = inf `(x). (12.12)

x∈M

Let {xn } ⊆ M be a minimizing sequence for (12.12), i.e.

1
`0 ≤ `(xn ) ≤ `0 + (12.13)
n
for all n ∈ N. Since the set M is compact, we may suppose that the sequence {xn }
is convergent. Let x0 = lim xn . Then passing to limit in the inequality (12.13),
n→+∞
we obtain `(x0 ) ≤ `0 . On the other hand, `(x) ≥ `0 for all x ∈ M and, consequently,
`(x0 ) = `0 . We put M 0 = {x ∈ M | `(x) = `0 }. Then M 0 6= ∅. The set M 0 is
invariant with respect to the semigroup E, i.e. ξ(M 0 ) ⊆ M 0 for all ξ ∈ E. In fact,
let η ∈ E and x0 ∈ M 0 . Then

`(η(x0 )) = sup ρ(ξ(η(x0 )), x0 ) ≤ sup ρ(ξ(x0 ), x0 ) = `(x0 ) = `0

η∈E ξ∈E

and, consequently, `(η(x0 )) = `0 . We will show now that the set M 0 contains a
single point. Really, if we suppose the contrary, then there exist x1 , x2 ∈ M 0 (x1 6=
x2 ). We consider x = S( 12 , x1 , x2 ), which is an element from M , because M is
strictly metric-convex. Since the semigroup E is compact, then under the conditions
of Lemma 12.4 there exists ξ ∈ E such that `0 ≤ `(x) = ρ(ξ(x), x0 ). On the
other hand, according to the strict metric-convexity of M we have ρ(ξ(x), x 0 ) <
max{ρ(ξ(x1 ), x0 ), ρ(ξ(x2 ), x0 )} ≤ `0 and, consequently, `(x) < `0 . The obtained
contradiction proves that M 0 contains a unique point x̄. Taking into account the
invariance of the set M 0 with respect to the semigroup E, we have ξ(x̄) = x̄ for all
ξ ∈ E. The Lemma is proved. 

Theorem 12.6 Let h(X, S+ , π), (Ω, S, σ), hi be a V -monotone bounded k-

dissipative NDS, J be its Levinson’s center and the following conditions be held:
˙
1. V (x1 , x2 ) = V (x2 , x1 ) for all (x1 , x2 ) ∈ X ×X.
2. V (x1 , x2 ) ≤ V (x1 , x3 ) + V (x3 , x2 ) for all x1 , x2 , x3 ∈ X with the condition
h(x1 ) = h(x2 ) = h(x3 ).
 Global attractors of V -monotone dynamical systems 393

3. the space (Xω , Vω ) is strictly metric-convex for all ω ∈ Ω, where Xω :=

h−1 (ω) := {x ∈ X | h(x) = ω } (ω ∈ Ω) and Vω := V |Xω ×Xω .

Then the set-valued mapping ω → Jω admits at least one continuous invariant

section, i.e. there exists a continuous mapping ν : Ω → J with the properties:
h(ν(ω)) = ω and ν(σ(t, y)) = π(t, ν(ω)) for all t ∈ S and ω ∈ Ω.

Proof. According to Corollary 12.1, under the conditions of Theorem 12.6 the
semi-group dynamical system (X, S+ , π) defines on J a group dynamical system
(J, S, π). Let ω0 ∈ Ω be an arbitrary point of Ω , J be the Levinson’s center of
the dynamical system (X, S+ , π) and E = E(J, S, π) be the Ellis semigroup of the
dynamical system (J, S, π), i.e. E(J, S, π) = {π t | t ∈ S }, where by bar we denote
the close in J J (J J is equipped with the topology of Tihonoff). We denote by
Eω0 := {ξ ∈ E | ξ(Jω0 ) ⊆ Jω0 }. Then under the conditions of Theorem 12.2,
Eω0 6= ∅ is a compact sub-semigroup of Ellis of the semigroup E. According to
Theorem 12.2, we have V (ξ(x1 ), ξ(x2 )) = V (x1 , x2 ) for all (x1 , x2 ) ∈ Jω0 × Jω0
and consequently, under the conditions of Theorem 12.6 we have a strictly metric-
convex (with respect to the complete metric Vω0 := V |Jω0 × Jω0 ) compact set
Jω0 and a compact semigroup of isometrics Eω acting on Jω0 . Applying Lemma
12.4, we obtain a common fixed point x̄ω0 ∈ Jω0 . We denote by Σ := H(x̄ω0 ) =
{x̄ω0 t | t ∈ S }. It is clear that Σ is a compactly invariant set of (J, S, π). Obviously,
T T
Σω0 := Σ Jω0 = {xω0 }. We will prove that Σω := Σ Jω contains a single point
x̄ω . It is evident that Σω 6= ∅ for all ω ∈ Ω, because Ω and Σ are compactly
invariant sets and Ω is minimal. Now we will prove that Σω contains exactly one
point. If we suppose the contrary, then there exist ω ∈ Ω and x1 , x2 ∈ Σω such
that x1 6= x2 . Since the set Ω is minimal, then there exists a sequence {tn } → −∞
such that ωtn → ω0 as n → +∞. Taking into consideration the compactness of Σ,
we may suppose that the sequences {xi tn } (i = 1, 2) are convergent. We denote
by x0i = lim xi tn . It is clear that x0i ∈ Σω0 and, consequently, x01 = x02 . On the
n→+∞
other hand, according to Corollary 12.3 the dynamical system (J, S, π) is distal in
negative direction and, consequently, x01 6= x02 . The obtained contradiction proves
our statement. Thus, we have a compactly invariant set Σ ⊆ J with the following
T
property Σω = Σ Jω = {xω } for all ω ∈ Ω. Now we define a mapping ν : Ω → J
by the following equality: ν(ω) = xω for all ω ∈ Ω. It is easy to verify that ν is
a continuous and invariant section of the set-valued mapping h−1 . The theorem is
proved. 

Corollary 12.5 Under the conditions of Theorem 12.6 the Levinson’s center of
the dynamical system (X, S+ , π) contains at least one stationary (τ (τ > 0) - pe-
riodic, quasi-periodic, almost periodic) point, if the minimal set Ω consists of a
stationary (τ (τ > 0) - periodic, quasi-periodic, almost periodic) point.
394 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Remark 12.1 a. In the proof of Lemma 12.4 we used the well known Favard
method of minimax (see, for example, [151] and [238]).
b. In the second part of the proof of Theorem 12.6 we used some ideas of
Zhikov [238] (Ch.VII) and Shcherbakov [302] (Ch.III).

12.4 Pullback attractors of V -monotone NDS

Let E be a Banach space and hE, ϕ, (Ω, S, σ), hi be a cocycle on the state space E.
Denote by X := E ×Ω, (X, S+ , π) a skew-product dynamical system, i.e. π = (ϕ, σ)
and h(X, S+ , π), (Ω, S, σ), hi is a non-autonomous dynamical system generated by
the cocycle ϕ, where h := pr2 : X → Ω.

Definition 12.6 The cocycle ϕ is called V -monotone (see [113], [114], [238],
[333]), if there exists a continuous function V : E × E × Ω → R+ with the following
properties:

a. V(u1 , u2 , ω) ≥ 0 for all ω ∈ Ω and u1 , u2 ∈ Ω.

b. V(u1 , u2 , ω) = 0 if and only if u1 = u2 .
c. V(ϕ(t, ω, u1 ), ϕ(t, ω, u2 ), σt ω) ≤ V(u1 , u2 , ω) for all u1 , u2 ∈ E, ω ∈ Ω and t ∈
S+ .

Lemma 12.5 The cocycle ϕ is V −monotone if and only if the non-autonomous

dynamical system h(X, S+ , π), (Ω, S, σ), hi generated by the cocycle ϕ is V -
monotone.

˙ → R+
Proof. Let ϕ be a V -monotone cocycle. Then we define a mapping V : X ×X
by the equality

V ((ω, u1 ), (ω, u2 )) := V(u1 , u2 , ω) (12.14)

˙
for all (ω, u1 ), (ω, u2 ) ∈ X ×X, where V : E ×E ×Y → R+ is the function figuring in
the definition of V -monotone cocycle ϕ and X ×X ˙ := {(x1 , x2 ) ∈ X × X | h(x1 ) =
h(x2 ) }.
From the equality (12.14) and a.-c. follows that V possesses the following prop-
erties:
˙ and V (x1 , x2 ) = 0 if and only if x1 = x2 .
i) V (x1 , x2 ) ≥ 0 for all (x1 , x2 ) ∈ X ×X
ii) V is continuous.
˙ and t ∈ S+ .
iii) V (x1 t, x2 t) ≤ V (x1 , x2 ) for all (x1 , x2 ) ∈ X ×X

The inverse statement is proved by analogy. 

Theorem 12.7 Let ϕ be a V −monotone cocycle admitting a compact pullback

attractor {Iω | ω ∈ Ω } and let exist a function a ∈ K with the property a(r) → +∞
 Global attractors of V -monotone dynamical systems 395

as r → +∞ such that V(u1 , u2 , ω) ≥ a(ρ(u1 , u2 )) for all u1 , u2 ∈ E and ω ∈ Ω. If

the NDS h(X, S+ , π), (Ω, S, σ), hi generated by the cocycle ϕ is α-condensing, then
it admits a compact global attractor J and Jω = Iω × {ω} for every ω ∈ Ω, where
T
Jω := h−1 J.

Proof. First step: we will prove that under the conditions of Theorem 12.7 all
trajectory of the dynamical system (X, S+ , π) are positively compact. Really, let
x ∈ X (x := (u, ω)) and q := (p, ω) ∈ Jω := Iω × {ω} be an arbitrary point. Then
in virtue of the V −monotonicity of the cocycle ϕ we have

V(ϕ(t, u, ω), ϕ(t, p, ω), σt ω) ≤ V(u, p, ω) (12.15)

for all t ∈ S+ . From the inequality (12.15) results that {ϕ(t, ω, u) | t ∈ S+ } is

bounded. If we suppose that it is not true, then there exists a sequence tn → +∞
such that

lim ρ(ϕ(tn , u, ω), ϕ(tn , p, ω)) = +∞. (12.16)

n→+∞

Under the conditions of Theorem 12.7 from (12.15) and (12.16) follows

V(u, p, ω) ≥ V(ϕ(tn , u, ω), ϕ(tn , p, ω), σtn ω) ≥

a(ρ((ϕ(tn , u, ω), ϕ(tn , p, ω)) → +∞

as n → +∞. The obtained contradiction proves our assertion.

Taking into consideration the asymptotic compactness of the dynamical
system (X, S+ , π) and the boundedness of the positively invariant set M :=
{(ϕ(t, ω, u), σt ω) | t ∈ S+ }, we conclude that it is relatively compact. Thus, under
the conditions of Theorem 12.7 every positive semi-trajectory is relatively compact.
Second step: since ϕ admits a compact global pullback attractor, then according
to Theorems 8.4 and 7.2 the set J is asymptotically stable and it is the maximal
compact invariant set of X. We will show that W s (J) = X. Really, let x ∈ X be
an arbitrary point. Then its ω-limit set ωx is nonempty, compact and invariant,
because {xt | t ∈ S+ } is relatively compact. Taking into account the maximality
of J, we have ωx ⊆ J for all x ∈ X and, consequently, W s (J) = X, i.e. J is the
compact global attractor of (X, S+ , π). The theorem is proved. 

12.5 Applications

12.5.1 Finite-dimensional systems

Denote by Rn a real n−dimensional Euclidean space with the scalar product h, i
and the norm | · | generated by the scalar product. Let [Rn ] be a space of all the
linear mappings A : Rn → Rn equipped with the operational norm.
396 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Theorem 12.8 Let Ω be a compact minimal set, F ∈ C(Ω × Rn , Rn ), W ∈

C(Ω, [Rn ]) and the following conditions be held:

1. the matrix-function W is positively defined, i.e. hW (ω)u, ui ∈ R for all ω ∈

Ω, u ∈ Rn and there exists a positive constant a such that hW (ω)u, ui ≥ a|u|2
for all ω ∈ Ω and u ∈ Rn .
2. the function t → W (σt ω) is differentiable for every ω ∈ Ω and Ẇ (ω) ∈
d
C(Ω, [Rn ]), where Ẇ (ω) := dt W (σt ω)|t=0 .
3. hẆ (ω)(u−v)+W (ω)(F (ω, u)−F (ω, v)), u−vi ≤ 0 for all ω ∈ Ω and u, v ∈ Rn .
4. there exist a positive constant r and a function c : [r, +∞) → (0, +∞) such that
hẆ (ω)u + W (ω)F (ω, u), ui ≤ −c(|u|) for all |u| > r.

Then the equation

u0 = F (u, σt ω) (12.17)

generates a cocycle ϕ on Rn which admits a compact global attractor I = {Iω | ω ∈

Ω} with the following properties:

a. Iω is a nonempty, compact and convex subset of Rn for every ω ∈ Ω.

S
b. I = {Iω | ω ∈ Ω} is connected.
c. the mapping ω → Iω is continuous with respect to the metric of Hausdorff.
d. I = {Iω | ω ∈ Ω} is invariant, i.e. ϕ(t, ω, Iω ) = Iσt ω for all ω ∈ Ω and t ∈ R+ .
e. lim β(ϕ(t, σt ω)M, Iω ) = 0 for all M ∈ C(Rn ) and ω ∈ Ω.
t→+∞
f. lim sup{β(ϕ(t, M, σt ω), I) | ω ∈ Ω } = 0 for any M ∈ C(Rn ), where I =
St→+∞
{Iω | ω ∈ Ω }.
g. I = {Iω | ω ∈ Ω } is a uniform forward attractor ,i.e.

lim sup β(ϕ(t, ω)M, Iσt ω ) = 0

t→+∞ ω∈Ω

for any M ∈ C(Rn ).

h. the equation (12.17) admits at least one stationary (τ −periodic, quasi-periodic,
almost periodic) solution, if the point ω ∈ Ω is stationary (τ −periodic, quasi-
periodic, almost periodic).

Proof. Since the function F ∈ C(Ω × Rn , Rn ), then, according to the theorem of

Peano (see, for example, [186]), the equation (12.17) admits at least one solution
u(t) (t ∈ [0, tu ), tu > 0) with the condition u(0) = x for every x ∈ Rn . We will
show that under the conditions of Theorem 12.8 this solution is unique. In fact,
let ui (t) (i = 1, 2) be two solutions of the equation (12.17) defined on [0, t0 ) with
the condition ui (0) = x (i = 1, 2). We consider the function hW (ωt)u1 (t), u2 (t)i.
 Global attractors of V -monotone dynamical systems 397

According to the conditions 1. and 3. of Theorem 12.8 we have

a|u1 (t) − u2 (t)|2 ≤ 0

for all t ∈ [0, t0 ) and, consequently, u1 (t) = u2 (t) for all t ∈ [0, t0 ).
Now we will prove that every solution of the equation (12.17) is defined on R + .
Let u ∈ Rn and ϕ(t, u, ω) be the unique solution of the equation (12.17) defined
on [0, t(u,ω) ). To prove that t(u,ω) = +∞ it is sufficient to show that the solution
ϕ(t, u, ω) is bounded on [0, t(u,ω) ). We denote by b := max kW (ω)k and T1 (u, ω) =
ω∈Ω
{t ∈ [0, t(u,ω) ) | |ϕ(t, u, ω)| ≤ r } and T2 (u, ω) = [0, t(u,ω) ) \ T1 (u, ω). It is clear that
S
the set T2 (u, ω) is open and, consequently, T2 (u, ω) = {(tα , tβ ) | β = β(α) }. For
α
all t ∈ T2 (u, ω) there exists α such that t ∈ (tα , tβ ) , |ϕ(tα , u, ω)| = |ϕ(tβ , u, ω)| and
|ϕ(t, u, ω)| > r. Consequently,

a|ϕ(t, u, ω)|2 ≤ hW (σt ω)ϕ(t, u, ω), ϕ(t, u, ω)i ≤ (12.18)

hW (ωtα )ϕ(tα , u, ω), ϕ(tα , u, ω)i ≤ br2 .

q
From the inequality (12.18) follows that |ϕ(t, u, ω)| ≤ ab r and, consequently,
q
we obtain sup{|ϕ(t, u, ω)| | t ∈ [0, t(u,ω) ) } ≤ r0 := max{r, ab r}. Thus, the
equation (12.17) defines a cocycle ϕ on Rn .
Let X = Rn × Ω, (X, R+ , π) be a skew-product dynamical system and
h(X, R+ , π), (Ω, R, σ), hi be the non-autonomous dynamical system generated by
the equation (12.17). Denote by V : X → R+ the function defined by the equal-
ity V(u, ω) := hW (ω)u, ui for all (u, ω) ∈ X := Rn × Ω. If |ϕ(t, u, ω)| > r for all
t ∈ (t1 , t2 ) ⊂ R+ , then
d
V(σt ω, ϕ(t, u, ω)) = hẆ (σt ω)ϕ(t, u, ω), ϕ(t, u, ω)i+
dt

hW (σω )F (σt ω, ϕ(t, u, ω)), ϕ(t, u, ω)i ≤ −c(|ϕ(t, u, ω)|) < 0.

In view of Theorem 5.3 the non-autonomous dynamical system h(X, R+ , π),

(Ω, R, σ), hi admits a compact global attractor.
Let V : X ×X˙ → R+ be the function defined by the equality V (ω, u1 , u2 ) :=
hW (ω)(u1 − u2 ), u1 − u2 i. Then
d
V (σt ω, ϕ(t, u1 , ω), ϕ(t, u2 , ω)) = hẆ (σt ω)(ϕ(t, u1 , ω) −
dt
ϕ(t, u2 , ω)), ϕ(t, u1 , ω) − ϕ(t, u2 , ω)i +
hW (σt ω)(F (σω, ϕ(t, u1 , ω)) −
F (σω, ϕ(t, u2 , ω))), ϕ(t, u1 , ω) − ϕ(t, u2 , ω)i ≤ 0
398 Global Attractors of Non-autonomous Dissipative Dynamical Systems

and, consequently, V (σt ω, ϕ(t, u1 , ω), ϕ(t, u2 , ω)) ≤ V (ω, u1 , u2 ) for all ω ∈
Ω, u1 , u2 ∈ Rn and t ∈ R+ . It is easy to verify that the function V satisfies
all the conditions of Theorems 12.4, 12.5 and the statements a.-h. follow from
Corollary 12.1, Theorems 12.4, 12.5 and Lemma 12.4. The theorem is proved. 

Example 12.1 As an example that illustrates this theorem we can consider the
following equation

u0 = g(u) + f (σt ω),

where f ∈ C(Ω, R) and



 (u + 1)2 , u < −1


g(u) = 0 , |u| ≤ 1




−(u − 1)2 ) , u > 1.

Example 12.2 Let us consider the equation

x00 + p(x)x0 + ax = f (σt ω),

where p ∈ C(R, R), f ∈ C(Ω, R) and a is a positive number. Denote by y :=

Rx
x0 + F (x), where F (x) := 0 p(s)ds. Then we obtain the system

 x0 = y − F (x)
(12.19)
 y 0 = −ax + f (σ ω).
t

Theorem 12.9 Suppose the following conditions hold:

1. p(x) ≥ 0 for all x ∈ R.
2. there exist positive numbers r and k such that p(x) ≥ k for all |x| ≥ r.
Then the non-autonomous dynamical system generated by (12.19) is compactly
dissipative and V −monotone.

Proof. Let X := Ω × R and h(X, R+ , π), (Ω, R, σ), hi be the non-autonomous

dynamical system generated by (12.19). We define the function V : X → R+ by
the equality
1
V (ω, x, y) := y 2 − yF (x) + F 2 (x) + ax2 .
2
Then
d
V (π t (ω, x, y))|t=0 = −p(x)[y − F (x)]2 − axF (x) + (2y − F (x))f (ω).
dt
 Global attractors of V -monotone dynamical systems 399

According to [270] (see the proof of Theorem 12.1.2), there exists R > 0 such
d
that dt V (π t (ω, x, y))|t=0 < 0 for all x2 + y 2 ≥ R2 and V (ω, x, y) → +∞ as
2 2
x + y → +∞. In view of Theorem 5.3 the non-autonomous dynamical system
h(X, R+ , π), (Ω, R, σ), hi is compactly dissipative.
Let V : X ×X ˙ → R+ be the function defined by the equality
V ((u1 , ω), (u2 , ω)) := hu1 − u2 ), u1 − u2 i. Then
d
V (σt ω, ϕ(t, u1 , ω), ϕ(t, u2 , ω)) = −(x1 (t) − x2 (t))[F (x1 (t)) − F (x2 (t))] ≤ 0
dt
for all t ∈ R, where xi (t) = pr1 ϕ(t, ui , ω) (i = 1, 2), and, consequently,
V (π t (u1 , ω), π t (u2 , ω)) ≤ V ((u1 , ω), (u2 , ω)) for all t ∈ R+ . To finish the proof
it is sufficient to refer to Theorem 12.8. The theorem is proved. 

12.5.2 Caratheodory’s differential equations

Let us consider now the equation (12.17) with the right hand side f satisfying the
conditions of Caratheodory (see, for example,[292]). The space of all Caratheodory‘s
functions we denote by C(R × Rn , Rn ). Topology on this space is defined by the
family of semi-norms (see [292])
Z k
dk,m (f ) := max |f (t, x)|dt.
−k |x|≤m

This space is metrizable, and on C(R × Rn , Rn ) there can be defined a dynamical

system of translations (C(R × Rn , Rn ), R, σ).
We consider the equation
dx
= f (t, x), (12.20)
dt
where f ∈ C(R × Rn , Rn ), and the family of equations
dx
= g(t, x), (12.21)
dt
where g ∈ H(f ) := {fτ | τ ∈ R }, and fτ is a τ −translation of the function f w.r.t.
the variable t, i.e. fτ (t, x) := f (t + τ, x) for all t ∈ R and x ∈ Rn and by bar is
denoted the closure in the space C(R × Rn , Rn ). Denote by ϕ(t, x, g) the solution of
equation the (12.21) with the initial condition ϕ(0, g, x) = x. Then ϕ is a cocycle on
Rn (see, for example [292]) with the base H(f ). Hence, we may apply the general
results from section 1-4 to the cocycle ϕ generated by the equation (12.20) with a
Caratheodory‘s right hand side and obtain some results for this type of equations.
For instance, the following assertion holds.

Theorem 12.10 Let f ∈ C(R × Rn , Rn ) be an almost periodic function in t ∈ R

(in the sense of Stepanoff [238]) uniformly w.r.t. x on compacts from Rn , i.e. for
400 Global Attractors of Non-autonomous Dissipative Dynamical Systems

every ε > 0 and compact K ⊂ Rn the set

Z 1
T(ε, f, K) := {τ ∈ R | max |f (t + τ + s, x) − f (t + s, x)|ds < ε }
0 x∈K

is relatively dense on R. Suppose that

1. hf (t, x1 ) − f (t, x2 ), x1 − x2 i ≤ 0 for all t ∈ R and x1 , x2 ∈ Rn .

2. there exists a positive constant r and a function c : [r, +∞) → (0, +∞) such
that hf (t, u), ui ≤ −c(|u|) for all |u| > r.

Then on Rn the equation (12.20) generates a cocycle ϕ which admits a compact

global attractor I = {Ig | g ∈ H(f ) } with the following properties:

a.Ig is a nonempty, compact and convex subset of Rn for every g ∈ H(f ).

S
b.I = {Ig | g ∈ H(f ) } is connected.
c.the mapping g → Ig is continuous with respect to the metric of Hausdorff.
d.I = {Ig | g ∈ H(f )} is invariant, i.e. ϕ(t, g, Ig ) = Iσt g for all g ∈ H(f ) and
t ∈ R+ .
e. lim β(ϕ(t, σ−t g)M, Ig ) = 0 for all M ∈ C(Rn ) and g ∈ H(f ).
t→+∞
f. lim sup{β(ϕ(t, σ−t g)M, I) | g ∈ H(f ) } = 0 for any M ∈ C(Rn ), where
t→+∞
S
I = {Ig | g ∈ H(f ) }.
g. I = {Ig | g ∈ H(f ) } is a uniform forward attractor, i.e.

lim sup β(ϕ(t, g)M, Iσt g ) = 0

t→+∞ g∈H(f )

for any M ∈ C(Rn ).

h. the equation (12.20) admits at least one stationary (τ −periodic, quasi-periodic,
almost periodic) solution, if the function f ∈ C(R × Rn , Rn ) is stationary
(τ −periodic, quasi-periodic, almost periodic) in t ∈ R uniformly w.r.t. x on
compacts from Rn .

Proof. This theorem is proved using the same arguments that we used in the proof
of Theorem 12.8. 

12.5.3 ODEs with impulse

Let {tk }k∈Z be a two-sided sequence of real numbers, p : R → Rn be a continuously
differentiable on every interval (tk , tk+1 ) function, continuous to the right in every
point t = tk , bounded on R, almost periodic in the sense of Stepanoff and
X
p0 (t) = s k δ tk ,
k∈Z

where sk := p(tk + 0) − p(tk − 0).

 Global attractors of V -monotone dynamical systems 401

Consider the equation with impulse

dx X
= f (t, x) + s k δ tk
dt
k∈Z

or, what is equivalent,

dx
= f (t, x) + p0 (t) (12.22)
dt
and paralelly consider the family of equations
dx
= g(t, x) + q 0 (t), (12.23)
dt
where (g, q) ∈ H(f, p) := {(fτ , pτ )|τ ∈ R} and by bar we denote the closure in the
product-space C(R × Rn , Rn ) × C(R, Rn ).
Denote by ϕ(t, x, g, q) the unique solution of the equation (12.23) (see [170] and
[279]) satisfying the initial condition ϕ(0, x, g, q) = x. This solution is continuous
on every interval (tk , tk+1 ) and continuous to the right in every point t = tk (see
[170] and [279]).
By the transformation

x := y + q(t) (12.24)

we can bring the equation (12.23) to the equation

dy
= g(t, y + q(t)). (12.25)
dt
Theorem 12.11 Let f ∈ C(R × Rn , Rn ) be a Bohr’s almost periodic function in
t ∈ R uniformly with respect to x on every compact from Rn and p ∈ C(R, Rn ) be a
Stepanoff ’s almost periodic function. Suppose that hf (t, x1 ) − f (t, x2 ), x1 − x2 i ≤ 0
for all t ∈ R and that there exist positive numbers α, L1 , L2 and r such that

hf (t, x), xi ≤ −L1 |x|α+1 and |f (t, x)| ≤ L2 |x|α

for all t ∈ R and |x| > r (x ∈ Rn ).

Then the equation (12.22) admits a compact global attractor {I (g,q) | (g, q) ∈
H(f, p) } with the following properties:

a. I(g,q) is a nonempty, compact and convex subset of Rn for every (g, q) ∈ H(f, p).
S
b. I = {I(g,q) | (g, q) ∈ H(f, p) } is disconnected.
c. I = {I(g,q) | (g, q) ∈ H(f, p) } is invariant, i.e. ϕ(t, I(g,q) , g, q) = Iσt (g,q) for all
(g, q) ∈ H(f, p) and t ∈ R+ .
d. lim β(ϕ(t, σt (g, q))M, I(g,q) ) = 0 for all M ∈ C(Rn ) and (g, q) ∈ H(f, g).
t→+∞
e. lim sup{β(ϕ(t, σ−t (g, q))M, I) | (g, q) ∈ H(f, p) } = 0 for any M ∈ C(Rn ),
t→+∞ S
where I = {I(g,q) | (g, q) ∈ H(f, p) }.
402 Global Attractors of Non-autonomous Dissipative Dynamical Systems

f. I = {Ig | g ∈ H(f ) } is a uniform forward attractor ,i.e.

lim sup β(ϕ(t, g, q)M, Iσt (g,q) ) = 0

t→+∞ (g,q)∈H(f,p)

for any M ∈ C(Rn ).

g. the equation (12.22) admits at least one stationary (τ - periodic, quasi-periodic,
almost periodic in the sense of Stepanoff ) solution, if the function (f, p) ∈
C(R × Rn , Rn ) × C(R, Rn ) is stationary (τ - periodic, quasi-periodic, almost
periodic in t ∈ R) uniformly w.r.t. x on compacts of Rn .

Proof. Let ϕ(t, x, g, q) be the cocycle generated by the family of equations (12.23)
and ϕ̃(t, y, g, q) be the cocycle generated by the family of equations (12.25). Then
we have the following equality

ϕ(t, x, g, q) = q(t) + ϕ̃(t, x − q(0), g, q)). (12.26)

We will show that it is possible to apply Theorem12.10 to the equation

dy
= f (t, y + p(t)).
dt
Really,

hf (t, y1 + p(t)) − f (t, y2 + p(t)), y1 − y2 i

for all t ∈ R and y1 , y2 ∈ Rn , and

hf (t, y + p(t)), yi = hf (t, y + p(t)), y + p(t)i − hf (t, y + p(t)), p(t)i ≤ (12.27)

−L1 |y + p(t)|α+1 + L2 kpk|y + p(t)|α

for all t ∈ R and |y + p(t)| > r, where kpk := sup{|p(t)| | t ∈ R }. Taking into
account the fact that the function p is bounded on R, from (12.27) we obtain the
existence of positive numbers R (that sufficiently large) and L01 , L02 such that

hf (t, y + p(t)), yi ≤ −L01 |y|α+1 and hf (t, y + p(t))i ≤ L02 |y|α

for all t ∈ R and |y| > R. To finish the proof of the theorem it is sufficient to apply
Theorem 12.10 and take into consideration the relations (12.24) and (12.26). The
theorem is proved. 

12.5.4 Evolution equations with monotone operators

p
Let H be a real Hilbert space with the inner product h, i, | · | = h, i and E be a
reflexive Banach space contained in H algebraically and topologically. Furthermore,
let E be dense in H, and here H can be identified with a subspace of the dual E 0
of E and h, i can be extended by continuity to E 0 × E.
 Global attractors of V -monotone dynamical systems 403

We consider the initial value problem

u0 (t) + Au(t) = f (σt ω) (12.28)

u(0) = u, (12.29)

where A : E → E 0 is bounded (generally nonlinear),

|Au|E 0 ≤ C|u|p−1
E + K, u ∈ E, p > 1,

coercive,

hAu, ui ≥ a|u|pE , u ∈ E, a > 0,

monotone,

hAu1 − Au2 , u1 − u2 i ≥ 0, u1 , u2 ∈ E,

and semi-continuous (see [240]).

A nonlinear ”elliptic” operator
Xn
∂ ∂u
Au = − φ( ) in D ⊂ Rn
i=1
∂x i ∂x i

u = 0 on ∂D,

where D is a bounded domain in Rn , φ(·), is an increasing function satisfying

n
X
φ|[−1,1] = 0, c|ξ|p ≤ ξi φ(ξi ) ≤ C|ξ|p ( for all |ξ| ≥ 2 ),
i=1
0
and provides an example with H = L2 (D), E = W01,p (D), E 0 = W −1,p (D), p0 =
p
p−1 .
The following result is established in [240] (Ch.2 and Ch.4). If x ∈ H and
p
f ∈ C(Ω, E 0 ), p0 = p−1 , then there exists a unique solution ϕ ∈ C(R+ , H) of
(12.28) – (12.29).
We denote by ϕ(·, u, ω) the unique solution of (12.28) and (12.29). According
to [187], ϕ(·, u, ω) is a continuous cocycle on H.

Theorem 12.12 Suppose that the operator A satisfies the conditions above and
the cocycle ϕ generated by the equation (12.28) is asymptotically compact. Then the
equation (12.28) admits a compact global attractor I = {I ω | ω ∈ Ω } possessing the
following properties:

a. Iω is a nonempty, compact and convex subset of H for every ω ∈ Ω.

S
b. I = {Iω | ω ∈ Ω } is connected.
c. the mapping ω → Iω is continuous with respect to the metric of Hausdorff.
404 Global Attractors of Non-autonomous Dissipative Dynamical Systems

d. I = {Iω | ω ∈ Ω } is invariant, i.e. ϕ(t, Iω , ω) = Iσt ω for all ω ∈ Ω and t ∈ R+ .

e. lim β(ϕ(t, σ−t ω)M, Iω ) = 0 for all M ∈ C(H) and ω ∈ Ω.
t→+∞
f. lim sup{β(ϕ(t, σt ω)M, I)| ω ∈ Ω } = 0 for any M ∈ C(H), where I =
t→+∞
S
{Iω | ω ∈ Ω }.
g. I = {Iω | ω ∈ Ω } is a uniform forward attractor ,i.e.

lim sup β(ϕ(t, ω)M, Iσt ω ) = 0

t→+∞ ω∈Ω

for any M ∈ C(H).

h. the equation (12.28) admits at least one stationary (τ - periodic, quasi-periodic,
almost periodic) solution, if the point ω ∈ Ω is stationary (τ - periodic, quasi-
periodic, almost periodic).

Proof. Let X := H × Ω, (X, R+ , π) be a skew-product dynamical system and

h(X, R+ , π), (Ω, R, σ), hi be a non-autonomous dynamical system generated by the
equation (12.28). Denote by V → R+ the function defined by the equality V(ω, u) :=
1
hu, ui for all (u, ω) ∈ X = H × Ω. If |ϕ(t, u, ω)| > r := ( kfak ) p−1 (where kf k :=
max{|f (ω)| | ω ∈ Ω }) for all t ∈ (t1 , t2 ) ⊂ R+ , then

d
V(σt ω, ϕ(t, u, ω)) = hA(ϕ(t, u, ω)), ϕ(t, u, ω)i
dt
≤ −a|ϕ(t, u, ω)|p + kf k|ϕ(t, u, ω)| < 0.

In view of Theorem 5.3 the non-autonomous dynamical system h(X, R+ , π),

(Ω, R, σ), hi admits a compact global attractor.
Let V : X ×X ˙ → R+ be a function defined by the equality V (ω, u1 , u2 ) :=
hu1 − u2 , u1 − u2 i. Then

d
V (σt ω, ϕ(t, u1 , ω), ϕ(t, u2 , ω))
dt
d
= 2h (ϕ(t, u1 , ω) − ϕ(t, u2 , ω)), ϕ(t, u1 , ω) − ϕ(t, u2 , ω)i
dt
= 2hA(ϕ(t, ω, u1 )) − A(ϕ(t, u2 , ω)), ϕ(t, u1 , ω) − ϕ(t, u2 , ω)i ≤ 0

and, consequently, V (σt ω, ϕ(t, u1 , ω), ϕ(t, u2 , ω)) ≤ V (ω, u1 , u2 ) for all ω ∈
Ω, u1 , u2 ∈ H and t ∈ R+ . It is easy to verify that the function V satisfies all the
conditions of Theorems 12.4, 12.5 and the statements a.-h. follow from Theorems
12.4, 12.5, 12.6 and Corollary 12.5. The theorem is proved. 

Remark 12.2 If the injection of E into H is compact, then the cocycle ϕ gener-
ated by the equation (12.28), evidently, is asymptotically compact.
 Global attractors of V -monotone dynamical systems 405

Example 12.3 A typical example of the equation of the type (12.28) is the
equation
X ∂n
∂ ∂u
u= φ( ) + f (σt ω), u|∂D = 0 (12.30)
∂t i=1
∂xi ∂xi
Pn ∂ ∂u
with ”nonlinear Laplacian” Au = i=1 ∂xi φ( ∂xi ), where φ(·) is an increasing func-
tion satisfying the condition
n
X
c|ξ|p ≤ ξi φ(ξi ) ≤ C|ξ|p
i=1

for all |ξ| ≥ 2 and φ|[−1,1] = 0, provides an example with H = L2 (D), E =

0
W01,p (D), E 0 = W −1,p (D), p0 = p−1 p
. It is possible to verify (see, for example,
[240],[30] and [20]) that the ”nonlinear Laplacian” satisfies all the conditions of
Theorem 12.12 and, consequently, (12.30) admits a compact global attractors with
the properties a.-h. We note that the attractor of the equation (12.30) is not trivial,
i.e. the set Iω is not single point set, at least for certain ω ∈ Ω.
Pn ∂ ∂u
Remark 12.3 If the operator A := i=1 ∂xi φ( ∂xi ) is uniformly elliptic, i.e.
P n
c|ξ|p ≤ i=1 ξi φ(ξi ) ≤ C|ξ|p (for all ξ ∈ Rn ), then the set Iω is the single point
set for all ω ∈ Ω (for autonomous systems see [314], Ch.III), because in this case
the non-autonomous dynamical system generated by the equation (12.30) is strictly
monotone.
406 Global Attractors of Non-autonomous Dissipative Dynamical Systems
 Chapter 13

Linear almost periodic dynamical systems

13.1 Bounded motions of linear systems

Assume that X and Y are complete metric spaces, R (Z) is the group of real
numbers (integers), S = R or S = Z, S+ = {s ∈ S | s ≥ 0}, S− = {s ∈ S | s ≤ 0},
and T = S+ , S− or S. Let (X, T, π) ((Y, S, σ)) be a semigroup (group) dynamical
system on X(Y ).
Recall that a triple h(X, T, π), (Y, S, σ), hi, where h is a homomorphism of
(X, T, π) onto (Y, S, σ), is called [69] a non-autonomous dynamical system.

Definition 13.1 h(X, T, π), (Y, S, σ), hi is said [32, 333] to be distal on S+ in
the fiber Xy := {x ∈ X | h(x) = y} if inf{ρ(x1 t, x2 t) | t ∈ S+ } > 0 for all
x1 , x2 ∈ Xy , x1 6= x2 .

For group non-autonomous dynamical systems the distalness on S− and S on

the fiber Xy can be defined likewise.

Definition 13.2 A non-autonomous system is said to be distal on S+ (S− , S) if it

is distal in every fiber Xy , y ∈ Y .

Assume that (Xi , T, πi ) is a dynamical system on Xi , i = 1, . . . , k; let X :=

X1 × ... × Xk , and let π := (π1 , ..., k) : X × T → X be defined by the formula

π(x, t) := (π1 (x1 , t), ..., πk (xk , t))

for all t ∈ T and x := (x1 , ..., xk ) ∈ X.

Definition 13.3 The dynamical system (X, T, π), where X := X1 × ... × Xk and
π := (π1 , ..., πk ), is called the direct product of the dynamical systems (Xi , T, πi ), i =
1, ..., k and denoted by (X1 , T, π1 )×, ..., (Xk , T, πk ).

If Xi = X, i = 1, ..., k, and πi = π, i = 1, ..., k, then

(X, T, π) × (X, T, π) × ... × (X, T, π) := (X k , T, π).

407
408 Global Attractors of Non-autonomous Dissipative Dynamical Systems

The direct product of group dynamical systems is defined likewise.

Definition 13.4 The points x1 , ..., xk ∈ X are said to be jointly recurrent if the
point (x1 , ..., xk ) ∈ X k is recurrent in the dynamical system (X k , T, π).

Lemma 13.1 [32, 333]. The following assertions hold.

(1) Assume that X is compact and (Y, S, σ) is minimal. If the group non-
autonomous dynamical system h(X, S, π), (Y, S, σ), hi is distal on S+ (S− ), then
it is distal on S.
(2) Assume that X is compact, (Y, S, σ) is minimal, and y ∈ Y . Then the following
conditions are equivalent :
(a) the group non-autonomous system h(X, S, π), (Y, S, σ), hi is distal on S in
the fiber Xy ;
(b) for any points x1 , ..., xk ∈ X, where k is any positive integer ≥ 2, the point
(x1 , ..., xk ) ∈ X k is recurrent in (X k , S, π).

Let C(S, X) be the space of all continuous maps ϕ : S → X equipped with the
compact-open topology and let (C(S, X), S, σ) be the dynamical system of transla-
tions (shifts) on C(S, X). Let d be a metric on C(S, X) consistent with its topology
(for example, the Bebutov metric).

Lemma 13.2 [282]. Let (X, S+ , π) be a semigroup dynamical system and assume
that for any t ∈ S+ the map π t : X → X is a homeomorphism and π̃ is the map of
X × S toX defined by the equality

(
π(x, t), (x, t) ∈ X × S+ ,
π̃(x, t) := −t −1
(π ) (x), (x, t) ∈ X × S− .
Then the triple (X, S, π̃) is a group dynamical system.

Lemma 13.3 [32]. Assume that h(X, S+ , π), (Y, S, σ), hi is a non-autonomous
dynamical system, Y is compact, (Y, S, σ) is minimal, and x0 ∈ Xy has a relatively
compact semi-trajectory {x0 t | t ∈ S+ }. Then one can find a recurrent point x ∈
ωx0 (x ∈ Xy ) and a sequence tn → +∞ such that ρ(x0 tn , xtn ) → 0 as n → +∞.

Recall that a dynamical system (X, S, π) is said to be asymptotically compact if

for any bounded positively invariant set B ⊆ X there is a non-empty compact set
K ⊆ X such that

lim sup{ρ(xt, K) | x ∈ B} = 0.
t→+∞

Remark 13.1 (i) Assume that x ∈ X is such that {xt | t ∈ S+ } is bounded and
(X, S+ , π) is asymptotically compact. Then {xt | t ∈ S+ } is relatively compact.
 Linear almost periodic dynamical systems 409

(ii) Let M ⊆ X be bounded and invariant. Then M is relatively compact if the

dynamical system (X, S+ , π) is asymptotically compact. In particular, if x ∈ X and
γ ∈ Φx is such that γ(S) is bounded, then γ(S) is relatively compact, where Φ x is a
set of all entire trajectories of (X, S+ , π) passing through x at t = 0.

If X := E×Y, π := (ϕ, σ), that is, π((u, y), t) := (ϕ(t, x, y), σ(t, y)) for all (u, y) ∈
E×Y and t ∈ S, then the non-autonomous dynamical system h(X, T, π), (Y, S, σ), hi,
where h := pr2 : X → Y, is called [275] a skew product over (Y, S, σ) with the fiber
E.
Let (X, h, Y ) be a locally trivial Banach fiber bundle [28].
Recall that a non-autonomous dynamical system h(X, T, π), (Y, S, σ), hi is said
[275], [33] to be linear if the map π t : Xy → Xyt is linear for every t ∈ T and y ∈ Y .
If h(X, T, π), (Y, S, σ), hi is a skew product over (Y, S, σ) with the fiber E, then
it is linear if and only if E is a Banach space and the map ϕ(t, ·, y) : E → E is
linear for every y ∈ Y and t ∈ T.
Throughout the rest of this section we assume that Y is compact, the dynamical
system (Y, S, σ) is minimal, X = E × Y, E is a Banach space with the norm | · |, the
non-autonomous dynamical system h(X, T, π), (Y, S, σ), hi is linear, π = (ϕ, σ), and
h = pr2 .
Let F ⊆ E × Y be a closed vectorial subset of the trivial fiber bundle (E ×
Y, pr2 , Y ) that is positively invariant relative to (X, T, π). We put

B+ = {(x, y) ∈ F | sup |ϕ(t, x, y)| : t ∈ S+ < +∞}.

The set B− is defined likewise. If h(X, S+ , π), (Y, S, σ), hi is a semigroup non-
autonomous dynamical system, then B is the set of all points of F with the follow-
ing property: there is an entire trajectory of the dynamical system (F, S+ , π)
T
bounded on S that passes through this point. We put B+ y := B+ Xy and
T
By := B Xy , y ∈ Y.

Theorem 13.1 The following conditions are equivalent :

(i) there is an M > 0 such that

|ϕ(t, x, y)| ≤ M |x| (13.1)

for all (x, y) ∈ B+ (B− , B) and t ∈ S+ (S− , S);

(ii) B+ (B− , B) is closed in F.

Proof. We prove this theorem in the case when T = S+ . In the case when T = S−
or T = S it can be proved in a similar way. We claim that (i) implies (ii). Assume
that (x, y) ∈ B+ . Then there is an (xn , yn ) ∈ B+ such that (xn , yn ) → (x, y) as
n → +∞. By condition (i), the inequality

|ϕ(t, xn , yn )| ≤ M |xn | (13.2)

410 Global Attractors of Non-autonomous Dissipative Dynamical Systems

is valid for all n = 1, 2, ... and t ∈ S+ . Passing to the limit in (13.2) as n → +∞, we
obtain that |ϕ(t, x, y)| ≤ M |x| for all t ∈ S+ , that is, (x, y) ∈ B+ .
Now we claim that (ii) implies (i). Let B+ be closed and let y ∈ Y. We put

d(y) = sup{|ϕ(t, x, y)| : t ∈ S+ , (x, y) ∈ B+ , |x| ≤ 1}. (13.3)

We claim that the function d : Y → R+ defined by formula (13.3) is lower semi-

continuous. Assume the contrary. Then there are ε > 0, y ∈ Y, and yn → y such
that

lim d(yn ) = d(y) − ε. (13.4)

n→+∞

Formula (13.3) implies that there are |xn | ≤ 1 and {tn } ⊆ S+ ((xn , y) ∈ B+ ) such
that

d(y) = lim |ϕ(tn , xn , y)|.

n→+∞

Therefore, for ε > 0 one can find a k such that the inequality
ε
||ϕ(tn , xn , y)| − d(y)| < (13.5)
4
is valid for all n ≥ k. Since the map ϕ(tk , xk , ·) : Y → X is continuous, there is an
n = n(k) such that
ε
|ϕ(tk , xk , yn ) − ϕ(tk , xk , y)| < (13.6)
4
for all n ≥ n(k). Inequalities (13.5) and (13.6) imply that
ε
|d(y) − |(tk , xk , yn )|| < (13.7)
4
for all n ≥ n(k). Hence,
ε
d(y) − d(yn ) ≤ (13.8)
2
for all n ≥ n(k). On the other hand, formula (13.4) implies that
ε
d(y) − d(yn ) ≥ 3 (13.9)
4
if n is sufficiently large.
Inequality (13.9) contradicts (13.8). This contradiction proves that d : Y → R +
is lower semi-continuous. Hence, this function has a set of points of continuity
D ⊆ Y of the type Gδ . Let p ∈ D, then there exists positive numbers δp and Mp
such that d(y) ≤ Mp for all

y ∈ B[p, δp ] := {y ∈ Y | ρ(y, p) ≤ δp } ⊆ Y.

Since Y is minimal, there are negative numbers t1 , ..., tm such that Y =

Sm
[ ]
i=1 σ(ti , B[p, δp ]) (see 238, p.134 ). We put L := max{|ti | : i = 1, ..., m}.
 Linear almost periodic dynamical systems 411

We claim that the family of operators {π t | t ∈ [0, L]} is uniformly continuous, that
is, for any ε > 0 there is a δ(ε) > 0 such that |x| ≤ δ implies that |xt| ≤ ε for all
t ∈ [0, L]. Assume the contrary. Then there are ε0 > 0, δn ↓ 0, |xn | < δn , yn ∈ Y,
and tn ∈ [0, L] such that

|ϕ(tn , xn , yn )| ≥ ε0 . (13.10)

Since Y and [0, L] are compact, we can assume that the sequences {yn } and {tn }
are convergent. Put y0 = lim yn and t0 := lim tn . Passing to the limit in
n→+∞ n→+∞
(13.10) as n → +∞, we obtain 0 ≥ ε0 . The last inequality contradicts the choice of
ε0 . This contradiction proves the above assertion.
If α > 0 is such that |ϕ(t, x, y)| ≤ 1 for all t ∈ [0, L], |x| ≤ α, and y ∈ Y, then

|ϕ(t, x, y)| ≤ α−1 |x| (13.11)

for all t ∈ [0, L], x ∈ E, and y ∈ Y. Assume that q ∈ Y, y ∈ B[p, δp ], and ti are such
that q = yti . Then

|ϕ(t, x, q)| = |ϕ(t, x, yti )| = |ϕ(t + ti , ϕ(−ti , x, yti ), y)| =

ϕ(−ti , x, yti )
|ϕ(−ti , x, yti )||ϕ(t + ti , , y)|. (13.12)
|ϕ(−ti , x, yti )|

The set B+ is positively invariant and contains (x, q). Therefore, B+ con-
tains π −ti (x, q), where π −ti (x, q) = (ϕ(−ti , x, q), σ(−ti , q)). Hence, π −ti (x, q) =
(ϕ(−ti , x, q), σ(−ti , q)) ∈ B+
y and

ϕ(−ti , x, yti )
|ϕ(t + ti , , y)| ≤ Mp (13.13)
|ϕ(−ti , x, yti )|, y)|

for all t ≥ L. On the other hand, inequality (13.11) implies that

|ϕ(−ti , x, yti )| ≤ α−1 |x| (13.14)

for all x ∈ E. Formulas (13.12)-(13.14) imply that |ϕ(t, x, q)| ≤ M |x| for all t ∈ S+
and x ∈ E, where M = α−1 max{1, Mp }. This completes the proof of the theorem. 

Lemma 13.4 Assume that h(X, S+ , π), (Y, S, σ), hi is a linear non-autonomous
dynamical system, (X, S+ , π) is asymptotic compact, and there is an M > 0 such
that

|γ(t)| ≤ M |x| (13.15)

S
for all γ ∈ Φ(x,y) , (x, y) ∈ B, and t ∈ S. Then the set Φ0 := {Φ(x,y) | (x, y) ∈
B, |x| ≤ 1} is relatively compact in C(S, X).
412 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Proof. Consider the set K := {γ(t) | t ∈ S, γ ∈ Φ0 } ⊆ B. It is obvious that K

is invariant. By (13.15), it is bounded. Since (X, S+ , π) is asymptotic compact,
K is relatively compact (see remark 13.1). We claim that the family of functions
Φ0 ⊆ C(S, X) is equicontinuous. Assume the contrary. Then there are ε0 > 0, δn ↓
0, {tin }(i = 1, 2), and {γn } ⊆ Φ0 such that |t1n − t2n | < δn and

|γn (t1n ) − γn (t2n )| ≥ ε0 . (13.16)

Without loss of generality we can assume that t2 − n > t1n . Then inequality
(13.16) implies that

|π τn xn − xn | ≥ ε0 (13.17)

for all n = 1, 2, ..., where τn := t2n − t1n and xn := γ( t1n ). Since {xn } ⊆ K, we can
assume that the sequence {xn } converges. Let x0 = lim xn . Passing to the limit
n→+∞
in inequality (13.17) as n → +∞, we obtain the inequality 0 ≥ ε0 , which contradicts
the choice of ε0 . To complete the proof of the lemma, it is sufficient to apply the
Arzela–Ascoli- theorem. 

Theorem 13.2 Assume that h(X, S+ , π), (Y, S, σ), hi is a linear non-autonomous
dynamical system, (X, S+ , π) is asymptotic compact, and there is an M > 0 such
that inequality (13.15) is valid for all γ ∈ Φ(x,y) , (x, y) ∈ B, and t ∈ S. Then the
following assertions hold:

(i) Any two different entire trajectories γ1 and γ2 (h(γ1 (0)) = h(γ2 (0))) are jointly
recurrent;
(ii) For any (x, y) ∈ B the set Φ(x,y) consists of a single entire recurrent trajectory;
(iii) B is closed in F ;
(iv) (X, S+ , π) induces a group dynamical system (B, S, π) on B;
(v) For any y ∈ Y the set By is finite-dimensional and dim By does not depend
on y ∈ Y .

Proof. Assume that (x, y) ∈ B, and let γ ∈ Φ(x,y) be bounded on S. By Lemma

13.4, the set H(γ) := {γτ | τ ∈ S} is compact in C(S, X), since γτ ∈ Φγ(τ ) ,
where γτ is the shift of γ by τ and the bar denotes closure in C(S, X). Con-
sider the group non-autonomous dynamical system h(H(γ), S, λ), (Y, S, σ), hi, where
(H(γ), S, λ) is the dynamical system of shifts on H(γ) induced by the Bebutov’s
system (C(S, X), S, σ) and µ : H(γ) → Y is the map defined by the equality
µ(ψ) = h(ψ(0)). Under the conditions of Theorem 13.2 (see also inequality (13.15))
the non-autonomous dynamical system h(H(γ), S, λ), (Y, S, σ), µi is negatively dis-
tal, that is, inf{dλ(t, γ1 ), λ(t, γ2 )) : t ∈ S− } > 0 for any γ1 , γ2 ∈ H(γ) such that
γ1 6= γ2 and µ(ψ1 ) = µ(ψ2 ). By Lemma 1 in [238, p.104], h(H(γ), S, λ), (Y, S, σ), hi
is distal on S. Therefore, γ1 and γ2 are jointly recurrent. In particular, they are
 Linear almost periodic dynamical systems 413

recurrent. Moreover, by Lemma 13.1, any two entire trajectories γ1 ∈ Φ(x,y) and
γ2 ∈ Φ(x,y) are jointly recurrent.
We claim that for any (x, y) ∈ B the set Φ(x,y) consists of a single entire recurrent
trajectory. Assume the contrary. Then there are γ1 , γ2 ∈ Φ(x,y) such that γ1 6= γ2 .
Putting γ(t) := γ1 (t)−γ2 (t), we obtain a recurrent function γ 6= 0 such that γ(t) = 0
for all t ∈ S+ , which is impossible.
Now let (x, y) ∈ B̄. Then there is an (xn , yn ) ∈ B such that (xn , yn ) → (x, y).
Let γn be the (unique) entire trajectory of (F, S+ , π) bounded on S and satisfying
the condition γn (0) = (xn , yn ). By Lemma 13.4, we can assume that the sequence
{γn } converges in C(S, X). Inequality (13.15) implies that γ = lim γn is an entire
n→+∞
trajectory of (F, S+ , π) bounded on S. Moreover, γ(0) = (x, y). Thus, γ ∈ Φ(x,y) ,
whence (x, y) ∈ B.
Since B is closed and invariant, (X, S+ , π) induces a dynamical system (B, S+ , π)
on B, and π t B = B for all t ∈ S+ . By assertion (ii) of the theorem, π t : B → B (t ∈
S+ ) is a one-to-one map and (π t )−1 (b) = γb (−t) for all t ∈ S+ and b ∈ B, where
{γb } = Φb . Lemma 13.4 implies that π t : B → B is a homeomorphism. To prove
assertion (iv), it is sufficient to apply Lemma 13.2.
We now prove the last assertion of the theorem. Since (X, S+ , π) is asymptotic
compact, K := {(x, y) | (x, y) ∈ B, |x| ≤ 1, y ∈ Y } is compact. Therefore, every
linear subspace By of Xy := E × {y} is finite-dimensional. Let y ∈ Y and let
x1 , ..., xk ∈ By be a basis in By . Then Lemma 13.1 implies that the points x1 , ..., xk
are jointly recurrent. Let q be an arbitrary point of Y. Since Y is minimal, there is
a sequence {tn } ⊂ S such that ytn → q and π(tn , xi ) → ξi (i = 1, ..., k) as n → +∞,
where ξ1 , ..., ξk ∈ Bq and the points ξ1 , ..., ξk are jointly recurrent.
We claim that ξ1 , ..., ξk are linearly independent. Assume the contrary. Then
P
there are constants c1 , ..., ck such that c1 ξ1 + ... + ck ξk = 0 and ki=1 |ci | 6= 0. Since
(ξ1 , ..., ξk ) ∈ H(x1 , ..., xk ) and (x1 , ..., xk ) is recurrent, there is a {τn } ⊂ S such that
qτn → y and π(ξi , τn ) → xi (i = 1, ..., k) as n → +∞. Therefore,

c1 x1 + ... + ck xk = lim π(τn , c1 ξ1 + ... + ck ξk ) = 0.

n→+∞

The last relation contradicts the choice of x1 , ..., xk . Thus, dimBq ≥ dim By for all
q ∈ Y. Since Y is minimal, the reverse inequality also holds. Hence, dim Bq = dimBy
for all q ∈ Y, which completes the proof of the theorem. 

Theorem 13.3 Let h(X, S+ , π), (Y, S, σ), hi be a linear non-autonomous

dynamical system and assume that (X, S+ , π) is asymptotic compact. Then the
following conditions are equivalent :

(i) there is an M > 0 such that (13.15) is valid for all γ ∈ Φ(x,y) , (x, y) ∈ B, and
t ∈ S;
(ii) B is closed in F.
414 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Proof. By Theorem 13.2, the theorem will be proved if we prove that (ii) implies
(i). By Theorem 13.1, there is an M > 0 such that (13.1) is valid for all (x, y) ∈ B
and t ∈ S+ . Since B is invariant, for any (x, y) ∈ B and γ ∈ Φ(x,y) the inequality

|γ(t)| ≥ M −1 |x| (13.18)

is valid for all t ∈ S− . Repeating the arguments used in Lemma 13.4, we can show
that H(γ) := {γτ | τ ∈ S} is compact in C(S, X). Consider the group dynamical
system h(H(γ), S, λ), (Y, S, σ), µi (see the proof of Theorem 13.2). Inequality (13.18)
implies that the non-autonomous system (H(γ), S, λ), (Y, S, σ), µi, is distal on S− .
By Lemma 13.1, γ is recurrent. Therefore,

sup{|γ(t)| : t ∈ S} = sup{|γ(t)| : t ∈ S+ } ≤ M |x|

for all γ ∈ Φ(x,y) and (x, y) ∈ B. Hence, condition (i) holds and the theorem is
proved. 

Theorem 13.4 Let h(X, S+ , π), (Y, S, σ), hi be a linear non-autonomous

dynamical system. Assume that (X, S+ , π) is asymptotic compact and B+ is closed.
Then

(i) B is closed, and

(ii) for any (x, y) ∈ B+ there is a recurrent point (x, y) ∈ B such that
lim |ϕ(t, x0 , y) − ϕ(t, x, y)| = 0.
t→+∞

Proof. Assume that B+ is closed. By Theorem 13.1, there is an M > 0 such

that (13.1) is valid for all (x, y) ∈ B+ and t ∈ S+ . Assume that (x, y) ∈ B ⊆ B+ .
Then inequality (13.1) implies that (13.18) is valid for all t ∈ S− and γ ∈ Φ(x,y) .
Repeating the arguments used in the proof of Theorem 13.3, we obtain (13.15) for
all γ ∈ Φ(x,y) , (x, y) ∈ B and t ∈ S. To complete the proof of the first assertion of
the theorem, it is sufficient to apply Theorem 13.2.
Now we prove the second assertion of the theorem. Let (x0 , y) ∈ B+ . Since
(X, S+ , π) is asymptotic compact, the semi-trajectory {π t (x0 , y) | t ∈ S+ } of the
point (x0 , y) is relatively compact. Therefore, ω(x0 ,y) 6= ∅ is compact and invariant.
By Lemma 13.3, one can find a recurrent point (x, y) ∈ ω(x0 ,y) and a sequence
tn → +∞ such that

lim |ϕ(tn , x0 , y) − ϕ(tn , x, y)| = 0. (13.19)

n→+∞

Inequality (13.15) implies that

|ϕ(t, x0 , y) − ϕ(t, x, y)| ≤ M |ϕ(tn , x0 , y) − ϕ(tn , x, y)| (13.20)

for all t ≥ tn . Formulas (13.19) and (13.20) imply the desired assertion, which
completes the proof of the theorem. 
 Linear almost periodic dynamical systems 415

Remark 13.2 The second assertion of Theorem 13.3 remains true even if we do
not assume that B+ is closed.

We conclude this section with a condition under which a linear non-autonomous

system is asymptotic compact.
Let P : X → X be a projection of the vector bundle, that is, Py := P |Xy
projection in Xy for every y ∈ Y. Then P is said to be completely continuous if
P (M ) is relatively compact for any bounded set M ⊆ X.

Lemma 13.5 Let h(X, S+ , π), (Y, S, σ), hi be a linear non-autonomous dynamical
system. Assume that the maps π t := π(t, ·) : X → X can be represented as sums
π(t, x) := π1 (t, x) + π2 (t, x) for all t ∈ S+ and x ∈ X, and that the following
conditions hold.
(i) |π1 (t, x)| ≤ m(t, r) for all t ∈ S+ and x ∈ X, where m : R2+ → R+ and for
every r ≥ 0 the function m(t, r) tends to zero as t → +∞.
(ii) The maps π2 (t, ·) : X → X (t > 0) are conditionally completely continuous,
that is, π2 (t, A) is relatively compact for any t > 0 and any bounded positively
invariant set A ⊆ X.
Then the dynamical system (X, S+ , π) is asymptotically compact.

Proof. Let A ⊆ X be a bounded positively invariant set. Then A = Σ+ (A) :=

{π t A | t ∈ S+ }. Since Y is compact and (X, h, Y ) is local trivial, there is an r > 0
such that A ⊂ {x ∈ X : |x| ≤ r}. We claim that for any {xk } ⊂ A and tn → +∞
the sequence {xk tk } is relatively compact. We can cover M := {xk tk } with a finite
ε-net for any ε > 0. Assume that ε > 0 and l > 0 are such that m(l, r) < 2ε . We
S
represent M as the union M1 M2 , where M1 := {xk tk }kk=1 1
, M2 := {xk tk }∞k=k1+1
l +
and k1 := max{k | tk ≤ l}. The set M2 is a subset of π (Σ (A)) whose elements
can be represented in the form π1 (t, x) + π2 (t, x) (x ∈ Σ+ (A)). Since π2 (l, Σ+ (A)) is
relatively compact, we can cover it with a finite ( ε2 )-net. For any y ∈ π1 (l, Σ+ (A))
there is an x ∈ Σ+ (A) such that y = π1 (l, x) and |y| = |π1 (l, x)| ≤ m(l, r) <
ε ε
2 . Therefore, the zero section Θ of the vector bundle (X, h, Y ) is an ( 2 )-net of
π1 (l, Σ+ (A)). Since Y is compact and (X, h, Y ) is local trivial, the zero section Θ is
S
compact. Hence, M2 and M are covered by the ε-net θ M1 . Since Θ is compact
and the space Y is complete, M = {xk tk }∞ k=1 is relatively compact. We complete
the proof by applying Lemma 1.3. 

Corollary 13.1 Let h(X, S+ , π), (Y, S, σ), hi be a linear non-autonomous

dynamical system and let P : X → X be a completely continuous projection.
Assume that there are positive numbers N and ν such that |π t Q(x)| ≤ N e−νt |x|
for all x ∈ X and t ∈ S+ , where Q : X → X and Qy := Q|Xy = Iy − Py for all
y ∈ Y (Iy := idXy ). Then (X, S+ , π) is asymptotic compact.
416 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Proof. To deduce this corollary from Lemma 13.5, it is sufficient to observe that
π(t, x) = π1 (t, x) + π2 (t, x) for all x ∈ X and t ∈ S+ , where π1 (t, x) := π t Q(x) and
π2 (t, x) := π t P (x). Under the hypotheses of Corollary 13.1, for every t > 0 the map
π2 (t, ·) := π t P is completely continuous and |π1 (t, x)| ≤ N e−νt |x|. Hence, Lemma
13.5 is applicable. 

Remark 13.3 In the proofs of Lemma 13.5 and Corollary 13.1 we used only the
compactness of Y . Hence, these statements are also valid in the case when the
dynamical system (Y, S, σ) is not minimal.

13.2 Bounded solutions of linear equations

Let [E] be the Banach space of all bounded linear operators that act on a Banach
space E equipped with the operator norm. Let Λ be a complete metric space of
closed linear operators that act on E (for example, Λ = [E] or Λ = {A0 + B | B ∈
[E]}, where A0 is a closed operator that acts on E). Let C(R, Λ) be the space of all
continuous operator-valued functions A : R → Λ equipped with the compact-open
topology and let (C(R, Λ), R, σ) be the dynamical system of shifts on C(R, Λ).
Linear ordinary differential equations. Let Λ = [E] and consider the dif-
ferential equation

u0 = A(t)u, (13.21)

where A ∈ C(R, [E]). Consider the H-class of equation (13.21), that is, the family
of equations

v 0 = B(t)v, (13.22)

with B ∈ H(A) := {Aτ | τ ∈ R}, Aτ (t) = A(t+τ ), and t ∈ R, where the bar denotes
closure in C(R, [E]). Let ϕ(t, v, B) be the solution of equation (13.22) that satisfies
the condition ϕ(0, v, B) = v.
We put Y := H(A) and denote the dynamical system of shifts on H(A) by
(Y, R, σ). We put X := E × Y and define a dynamical system on X by set-
ting π(t, (v, B)) := (ϕ(t, v, B), Bt ) for all (v, B) ∈ E × Y and t ∈ R. Then
h(X, R, π), (Y, R, σ), hi is a linear group non-autonomous dynamical system, where
h := pr2 : X → Y. Applying the results of section 13.1 to this system, we obtain
the following assertions.

Lemma 13.6 [51, 60]

(i) The map (t, u, A) 7→ ϕ(t, u, A) of R × E × C(R, [E]) to E is continuous, and

(ii) the map A 7→ U (·, A) of C(R, [E]) to C(R, [E]) is continuous, where U (t, A) is
the Cauchy’s operator [132] of equation (13.21).
 Linear almost periodic dynamical systems 417

Theorem 13.5 Assume that A ∈ C(R, [E]) is recurrent (that is, H(A) is a com-
pact minimal set of (C(R, [E]), R, σ)). Then the following conditions are equivalent:
(a) the set

B+ (B− , B) = {(v, B) ∈ E × H(A) | sup |ϕ(t, v, B)| < +∞} (13.23)

t∈R+ (R− ,R)

is closed in E × H(A), and

(b) there is a positive number M such that

|ϕ(t, v, B)| ≤ M |v| (13.24)

for all t ∈ R+ (R− , R) and (v, B) ∈ B+ (B− , B).

Corollary 13.2 Let A ∈ C(R, [E]) be recurrent. Then the following assertions
are equivalent :
(i) all solutions of all equations (13.22) are bounded on R+ (R− , R), and
(ii) there is an M > 0 such that (13.24) is valid for all v ∈ E, B ∈ H(A), and
t ∈ R+ (R− , R).

Theorem 13.6 Assume that A ∈ C(R, [E]) is recurrent, the linear non-
autonomous dynamical system generated by equation (13.21) is asymptotic compact,
and all solutions of all equations (13.22) are bounded on R+ . Then
(i) there is an M > 0 such that (13.24) is valid for all t ∈ R+ , v ∈ E and
B ∈ H(A),
(ii) the set B defined by formula (13.23) is closed in E × H(A),
(iii) all solutions of all equations (13.22) bounded on R are recurrent,
(iv) for any B ∈ H(A) equation (13.22) has only a finite number nB of solutions
that are linearly independent and bounded on R, and nB = nA for all B ∈
H(A),
(v) for any v0 ∈ E and B ∈ H(A) there is a (v, B) ∈ B such that

lim |ϕ(t, v0 , B) − ϕ(t, v, B)| = 0,

t→+∞

that is, any solution of any equation (13.22) is asymptotic recurrent.

We now formulate some sufficient conditions for the asymptotic compactness of

the linear non-autonomous dynamical system generated by equation (13.21).

Theorem 13.7 Let A ∈ C(R, [E]), A(t) = A1 (t)+A2 (t) for all t ∈ R, and assume
that H(Ai ), i = 1, 2, are compact and the following conditions hold.
(i) The zero solution of the equation

u0 = A1 (t)u (13.25)
418 Global Attractors of Non-autonomous Dissipative Dynamical Systems

is uniformly asymptotically stable, that is, there are positive numbers N and ν
such that

kU (t, A1 )U −1 (τ, A1 )k ≤ N e−ν(t−τ ) (13.26)

for all t ≥ τ (t, τ ∈ R), where U (t, A1 ) is the Cauchy’s operator of the equation
(13.25).
(ii) The family of operators {A2 (t) | t > 0} is uniformly completely continuous, that
is, for any bounded set A ∈ E the set {A2 (t)A | t > 0} is relatively compact.

Then the linear non-autonomous dynamical system generated by equation

(13.21) is asymptotic compact.

Proof. Let B ∈ H(A). Then there are {tn } ⊂ R and Bi ∈ H(Ai ), i = 1, 2, such
that B(t) = B1 (t) + B2 (t) and Bi (t) = lim Ai (t + tn ). Note that
n→+∞
Z t
ϕ(t, v, B) = U (t, B1 )v + U (t, B1 )U − 1(τ, B1 )B2 (τ )ϕ(τ, v, B)dτ. (13.27)
0

By Lemma 13.6,

U (t, Bi ) = lim U (t, Aitn ), Aitn (t) := Ai (t + tn ),

n→+∞

and the equality

U (t, A1tn )U − 1(τ, A1tn ) = U (t + tn , A1 )U − 1(τ + tn , A1 )

and inequality (13.26) imply that

kU (t, B1 )U − 1(τ, B1 )k ≤ N e−ν(t−τ ) (13.28)

for all t ≥ τ and B1 ∈ H(A1 ). By Lemma 13.5, Theorem 13.7 will be proved if we
can prove that the set
Z t
{ U (t, B1 )U − 1(τ, B1 )B2 (τ )ϕ(τ, v, B)dτ | (v, B) ∈ A}
0

is relatively compact for every t > 0 and every bounded positively invariant set
A ⊆ E × Y. We put

KA := {B2 (t)ϕ(t, v, B) | t ∈ R+ , (v, B) ∈ A}.

Then
Z t
U (t, B1 )U − 1(τ, B1 )B2 (τ )ϕ(τ, v, B)dτ (13.29)
0
∈ t · conv{U (t, B1 )U − 1(τ, B1 )w | 0 ≤ τ ≤ t, B1 ∈∈ H(A1 ), w ∈ KA }.
 Linear almost periodic dynamical systems 419

Since H(A1 ), H(A), and KA are compact sets, formula (13.29), condition (ii)
S
of Theorem 13.7, and Lemma 13.6 imply that {U (t, B1 )U − 1(τ, B1 )w | 0 ≤ τ ≤
t, B1 ∈ H(A1 ), w ∈ KA } is compact, which completes the proof of the theorem. 

Theorem 13.8 Let H(A) be compact and assume that there is a finite-
dimensional projection P ∈ [E] such that

(i) the family of projections {P (t) | t ∈ R}, where P (t) := U (t, A)P U −1 (t, A), is
relatively compact in [E], and
(ii) there are positive numbers N and ν such that

kU (t, A)QU −1 (τ, A)k ≤ N e−ν(t−τ )

for all t ≥ τ, where Q := I − P.

Then the linear non-autonomous dynamical system generated by equation

(13.21) is asymptotically compact.

Proof. Since the family of projections {P (t) | t ∈ R} is relatively compact,

we can assume that the sequence {P (tn )} converges. Let P (B) := lim P (tn ).
n→+∞
We claim that the family H := {P (t) | t ∈ R} is uniformly completely continu-
ous, where the bar denotes closure in [E]. Indeed, let A be a bounded subset of
E, {xn } ⊆ {QA | Q ∈ H}, and εn ↓ 0. Then there are tn ∈ R and vn ∈ A such
that ρ(xn , P (tn )vn ) ≤ εn . Since the sequence {P (tn )} is relatively compact, we can
assume that it converges. Let L := lim P (tn ). Then L is completely continuous,
n→+∞
which implies that the sequence {x0n } := {Lvn } is relatively compact. Note that

ρ(xn , x0n ) ≤ ρ(xn , P (tn )vn ) + ρ(P (tn )vn , Lvn ) ≤ εn + kP (tn ) − Lk|vn |,

which implies that ρ(xn , x0n ) → 0 as n → +∞. Hence, {xn } is relatively compact.
Assume that B ∈ H(A) and {tn } ⊂ R are such that

B = lim Atn , P (B) := lim P (Atn ),

n→+∞ n→+∞

where P (Atn ) := U (tn , A)P U −1 (tn , A). The assertions proved above imply that
the family {P (B) | B ∈ H(A)} is uniformly completely continuous. Note that
Q(B) = lim Q(Atn ), where Q(B) := I − P (B) and Q(Atn ) := I − P (Atn ).
n→+∞
Moreover, condition (ii) of Theorem 13.8 implies that

kU (t, Atn )Q(A)U −1 (τ, Atn )k ≤ N e−ν(t−τ ) (13.30)

for all t ≥ τ. Passing to the limit in (13.30) as n → +∞ and taking Lemma 13.6
into account, we obtain that

kU (t, B)Q(B)U −1 (τ, B)k ≤ N e−ν(t−τ )

420 Global Attractors of Non-autonomous Dissipative Dynamical Systems

for all t ≥ τ and B ∈ H(A). We complete the proof of the theorem by observing
that U (t, B)Q(B) + U (t, B)P (B) = U (t, B) and applying Lemma 13.5. 

Linear functional-differential equations. Let r > 0, C([a, b], Rn ) be the

Banach space of all continuous functions ϕ : [a, b] → Rn with the norm sup . For
[a, b] := [−r, 0] we put C := C([−r, 0], Rn ). Let c ∈ R, a ≥ 0, and u ∈ C([c − r, c +
a], Rn ). We define ut ∈ C for any t ∈ [c, c + a] by the relation ut (θ) := u(t + θ), −r ≥
θ ≥ 0. Let A = A(C, Rn ) be the Banach space of all linear operators that act from
C → Rn equipped with the operator norm, let C(R, A) be the space of all operator-
valued functions A : R → A with the compact-open topology, and let (C(R, A), R, σ)
be the dynamical system of shifts on C(R, A). Let H(A) := {Aτ | τ ∈ R}, where
Aτ is the shift of the operator-valued function A by τ and the bar denotes closure
in C(R, A).
Consider the linear functional-differential equation with delay

u0 = A(t)ut (13.31)

along with the family of equations

v 0 = B(t)vt , (13.32)

where B ∈ H(A). Let ϕ(t, v, B) be the solution of equation (13.32) satisfying the
condition ϕ(0, v, B) = v and defined for all t ≥ 0. Let Y := H(A) and denote
the dynamical system of shifts on H(A) by (Y, R, σ). Let X := C × Y and let
π := (ϕ, σ) be the dynamical system on X defined by the equality π(τ, (v, B)) :=
(ϕ(τ, v, B), Bτ ). The non-autonomous system h(X, R+ , π), (Y, R, ), hi(h := pr2 :
X → Y ) is linear.

Lemma 13.7 Let H(A) be compact in C(R, A). Then the linear non-autonomous
dynamical system h(X, R+ , π), (Y, R, σ), hi generated by equation (13.31) is com-
pletely continuous, that is, for any bounded set A ⊆ X there is an l = l(A) > 0
such that π l A is relatively compact.

Proof. This follows from general properties of solutions of linear functional- differ-
ential equations with delay (see, for example, [175], Lemmas 2.2.3 and 3.6.1) since
Y = H(A) is compact. 

Applying the results obtained in section 13.1 to the linear non-autonomous

dynamical system generated by equation (13.31), we obtain the following assertions.

Theorem 13.9 Let A ∈ C(R, A) be recurrent. Then the following conditions are
equivalent :

(i) all solutions of all equations (13.32) are bounded on R+ ,

 Linear almost periodic dynamical systems 421

(ii) there is a positive number M such that |ϕ(t, v, B)| ≤ M |v| for all t ≥ 0, v ∈ C,
and B ∈ H(A).

Theorem 13.10 Let A ∈ C(R, A) be recurrent, and assume that all solutions of
all equations (13.32) are bounded on R+ . Then

(i) the set of all the solutions of all equations (13.32) that are bounded on R is
closed in C(R, C) × H(A),
(ii) all the solutions of all equations (13.32) that are bounded on R are recurrent,
(iii) for any B ∈ H(A) equation (13.32) has only a finite number nB of solutions
that are linearly independent and bounded on R, and nB = nA for all B ∈
H(A),
(iv) all solutions of all equations (13.32) are asymptotic recurrent.

Now consider the neutral functional-differential equation

d
Dut = A(t)ut , (13.33)
dt
where A ∈ C(R, A) and the operator D ∈ A is atomic at zero [175, p.67]. Like
(13.31), equation (13.33) generates a linear non-autonomous dynamical system
h(X, R+ , π), (Y, R, σ), hi, where X := C × Y, Y := H(A), and π := (ϕ, σ).

Lemma 13.8 Let H(A) be compact, and assume that the operator D is sta-
ble, that is, the zero solution of the homogeneous difference equation Dy t = 0 is
uniformly asymptotically stable [175]. Then the linear non-autonomous dynamical
system (X, R+ , π), (Y, R, σ), hi generated by equation (13.33) is asymptotically com-
pact.

Proof. This follows from Theorem 12.3.2 and Lemma 12.4.1 in [175], since Y =
H(A) is compact. 

Therefore, Theorems 13.9 and 13.10 hold for equations (13.33).

Linear partial differential equations. Consider the differential equation
(13.21) with unbounded coefficients. Let A ∈ C(R, Λ), where Λ is a complete metric
space of linear closed operators that act on E (for example, Λ := {A0 +B | B ∈ [E]},
where A0 is a closed operator that acts on E). Consider the H-class (13.22) of
equation (13.21), where B ∈ H(A). We assume that the following conditions are
fulfilled for equation (13.21) and its H-class:

(i) for any v ∈ E and B ∈ H(A) equation (13.22) has precisely one solution
ϕ(0, v, B) that is defined on R+ and satisfies the condition ϕ(0, v, B) = v;
(ii) the map ϕ : (t, v, B) → ϕ(t, v, B) is continuous in the topology of R+ × E ×
C(R, A);
422 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(iii) for every t ∈ R+ the map U (t, ·) : H(A) → [E] is continuous, where U (t, B)
is the Cauchy’s operator of equation (13.22), that is, U (t, B)v := ϕ(t, v, B) for
all t ∈ R+ and v ∈ E.

Under the above assumptions equation (13.21) generates a linear non-

autonomous dynamical system to which the results obtained in section 13.1 can
be applied. Therefore, Theorems 13.5 and 13.6 hold in this case.
In conclusion we consider a partial differential equation that satisfies conditions
(i)-(iii).

Example 13.1 A closed linear operator A : D(A) → E with dense domain D(A)
is said [188] to be sectorial if one can find a φ ∈ (0, π2 ), an M ≥ 1, and a real number
a such that the sector

Sa,φ := {λ | |arg (λ − a)| ≤ π, λ 6= a}

lies in the resolvent set ρ(A) of A and k(λI − A)−1 k ≤ M |λ − a|−1 for all λ ∈ Sa,φ .
An important class of sectorial operators is formed by elliptic operators [188], [202].
Consider the differential equation

u0 = (A1 + A2 (t))u, (13.34)

where A1 is a sectorial operator that does not depend on t ∈ R, and A2 ∈ C(R, [E]).

The results of [238], [188] imply that equation (13.34) satisfies conditions (i)-(iii).
Therefore, analogies of Theorems 13.5 and 13.6 hold for equation (13.34).

Remark 13.4 Statements similar to Theorems 13.5 and 13.6 hold for difference
equations and can be deduced from the results of section 13.1 by applying these
results to linear non-autonomous dynamical systems with discrete time generated by
the corresponding difference equations.

13.3 Finite-dimensional systems

Throughout this section we assume that the Banach space E is finite-dimensional

and its norm | · | is induced by the scalar product h·, ·i, that is, | · |2 := h·, ·i. We
propose several conditions for equations (13.21) in a finite-dimensional space that
are equivalent to the uniform bistability of the zero solution of equation (13.21), and
prove that the uniform Lyapunov stability of the zero solution of equation (13.21)
implies that there is a frame of solutions of equation (13.21) bounded on R whose
Gram determinant is separated from zero.
Let x1 , ..., xk be a set of vectors in E. Let us recall [23] that the Gram de-
terminant Γ(x1 , ..., xk ) of the vectors x1 , ..., xk is defined to be the determinant
 Linear almost periodic dynamical systems 423

|hxi , xj i|ki,j=1 . The Gram determinant of the vectors x1 , ..., xk is no-negative, and
equals zero only if the vectors x1 , ..., xk are linearly dependent.

Theorem 13.11 The following assertions are equivalent:

(i) there is an M > 0 such that (13.23) is valid for all t ∈ R and (x, y) ∈ B;
(ii) B is closed;
(iii) B is a subbundle of F , that is, B is closed and there is a k such that dimB y = k
for all y ∈ Y ;
(iv) all the motions in F that are non-trivial and bounded on R are separated from
zero, that is, inf{ |ϕ(t, x, y)| : t ∈ R} > 0 for any (x, y) ∈ B such that x 6= 0;
(v) one can find a y0 ∈ Y and a basis ξ1 , ..., ξk ∈ By0 such that

inf Γ(ξ1 , ..., ξk ; t) := α > 0, (13.35)

t∈R

where ξi = (xi , y0 ), i = 1, ..., k and Γ(ξ1 , ..., ξk ; t) is the Gram determinant of

the vectors π(t, ξi ) ∈ E × Y, i = 1, ..., k.

Proof. Assertions (i) and (ii) are equivalent by Theorem 13.1. By Theorems 13.1
and 13.2, (ii) implies (iii) (the reverse implication is obvious). The equivalence of
conditions (iii) and (iv) follows from [33], Theorem 8.22.
Assume that (iv) is fulfilled, let y0 ∈ Y, and let ξ1 , ..., ξk ∈ By be a basis in B.
We claim that (13.35) holds. Assume the contrary. Then one can find a {tn } ⊂ R
such that |tn | → +∞ and Γ(ξ1 , ..., ξk ; tn ) → 0 as n → +∞. Since all non-zero
motions in F are separated from zero, Lemma 13.1 implies that ξ1 , ..., ξk and y
are jointly recurrent. Without loss of generality, we can assume that the sequences
{π(tn , ξi )}, i = 1, ..., k and {σ(tn , y)} are convergent.
Let

ηi := lim π(tn , ξi ), i = 1, ..., k, q := lim σ(tn , y0 ).

n→+∞ n→+∞

Then

Γ(η1 , ..., ηk ) = lim Γ(π(tn , ξ1 ), ..., π(tn , ξk )) = 0.

n→+∞

Hence, η1 , ..., ηk are linearly dependent. Repeating the above argument, we deduce
from the last fact that ξ1 , ..., ξk are linearly dependent, which contradicts their
choice. Hence, (iv) implies (v).
We claim that (v) implies (iv). First, we prove that for any q ∈ Y one can find
a basis η1 , ..., ηk such that Γ(η1 , ..., ηk ; t) ≥ α > 0. Indeed, since Y is minimal, there
is a sequence {tn } ⊂ R such that σ(tn , y0 ) → q. Since ξ1 , ..., k ∈ By0 , we can assume
that the sequences {π(tn , ξi )}, i = 1, ..., k are convergent. Let ηi := lim π(tn , ξi ).
n→+∞
424 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Then η1 , ..., ηk ∈ Bq and

Γ(η1 , ..., ηk ; t) = lim Γ(ξ1 , ..., ξk ; t + tn ) (13.36)

n→+∞

for all t ∈ R. Therefore, η1 , ..., ηk are linearly independent. Hence, nq := dimBq ≥

ny0 := dimBy0 . Since Y is minimal, the reverse inequality also holds. Therefore,
nq = ny0 for all q ∈ Y. This implies that η1 , ..., ηk is a basis in Bq . Thus, we
have shown that for any q ∈ Y there is a basis η1 , ..., ηk ∈ Bq that satisfies condition
(13.36). For any q ∈ Y and (x, q) ∈ Bq we have inf{ |ϕ(t, x, q)| : t ∈ R} > 0. Indeed,
if we assume the contrary, then there are (x0 , q) ∈ Bq , x0 6= 0, and |tn | → +∞ such
that

|ϕ(tn , x0 , y)| → 0 (13.37)

as n → +∞. Let η10 , ..., ηk0 be a basis in Bq , and let η10 = (x0 , q). Then (13.37) implies
that

Γ(η10 tn , ..., ηk0 tn ) → 0. (13.38)

Since η1 , ..., ηk and , η10 , ..., ηk0 are two bases in Bq , there is a non-degenerate linear
transformation that transforms the first basis into the second, and vice versa. For-
mula (13.38) implies that a similar relation holds for the basis η1 , ..., ηk , which
contradicts inequality (13.36). This contradiction completes the proof of the impli-
cation (v)=⇒ (iv). The theorem is proved. 
Applying Theorem 13.11 to the linear non-autonomous dynamical system gen-
erated by equation (13.21), we obtain the following assertion.

Theorem 13.12 Let A ∈ C(R, [E]) be recurrent. Then the following conditions
are equivalent:

(a) the set B := {(u, B) ∈ E ×H(A) | sup |ϕ(t, u, B)| < +∞} is closed in E ×H(A);
t∈R
(b) there is a positive number M such that

|ϕ(t, u, B)| ≤ M |u| (13.39)

for all t ∈ R and (u, B) ∈ B;

(c) B is closed and all fibres BB have the same dimension, that is, all the equations
(13.22) have the same number of solutions that are linearly independent and
bounded on R;
(d) all non-trivial solutions of all equations in the H-class (13.22) bounded on R
are separated from zero, that is,

inf |ϕ(t, u, B)| > 0 (13.40)

t∈R

for all (u, B) ∈ B, u 6= 0;

 Linear almost periodic dynamical systems 425

(e) there is a basis ϕ1 , ..., ϕk (k = dimBA and BA := {(u, A) | (u, A) ∈ B}) that
consists of solutions of equation (13.21) bounded on R and satisfies the condition
Γ(ϕ1 , ..., ϕk ; t) > 0 for all t ∈ R, where Γ(ϕ1 , ..., k; t) := |hϕi (t), ϕj (t)i|ni,j=1 is
the Gram determinant of ϕ1 , ..., ϕk .

Corollary 13.3 Let A ∈ C(R, [E]) be recurrent. Then the following conditions
are equivalent:

(i) all solutions of all equations from the H-class (13.22) are bounded on R;
(ii) there is an M > 0 such that |ϕ(t, u, B)| ≤ M |u| for all t ∈ R and u ∈ E;
(iii) all non-zero solutions of all equations from the H-class (13.22) are bounded
on R and separated from zero, that is, (13.40) is valid for all (u, B) ∈ B such
that u 6= 0;
(iv) there are positive numbers C and such that kU (t, A)k ≤ C for all t ∈ R and
inf{|det U (t, A)| : t ∈ R} = α, where U (t, A) is the Cauchy operator of
equation (13.21).

Proof. This follows from Theorem 13.12, since Γ(ϕ1 , ..., ϕn ; t) = |det U (t, A)|2 ,
where ϕ1 , ..., ϕn are the column-vectors of the matrix U (t, A). 

Remark 13.5 The equivalence of conditions (i) and (ii) was established in [39],
[204]. The implication (iv)=⇒ (i) sharpens a result in [204].

Theorem 13.13 Assume that all solutions of all equations from the H-class
(13.22) are bounded on R+ . Then there is an M > 0 such that |ϕ(t, u, B)| ≤ M |u|
for all t ∈ R and (u, B) ∈ B, that is, B is closed.

Proof. This follows from Theorems 13.1 and 13.4. 

In conclusion we consider examples that illustrate the above results.

Example 13.2 Let a ∈ C(R, R) be the Bohr almost periodic function defined by
the equality
∞
X 1 t
a(t) := 3/2
sin , (13.41)
(2k + 1) 2k + 1
k=0

and let
Z t ∞
X 2 t
h(t) := a(s)ds = 1/2
sin2 .
0 (2k + 1) 2(2k + 1)
k=0

Note that a(t + tn ) → −a(t) uniformly on R, where tn := (2n + 1)!!. Therefore,

−a ∈ H(a) := {aτ | τ ∈ R}. Using the inequality | sin t| ≥ 21 |t| with |t| ≤ 1, we
426 Global Attractors of Non-autonomous Dissipative Dynamical Systems

obtain that
∞
X X
1 t t2 1
h(t) = 1/2
sin2 ≥
(2k + 1) 2(2k + 1) 8 (2k + 1)5/2
k=0 k≥ 21 ( |t|
2 −1)
Z
t2 ds t2 23/2 1
≥ 5/2
= = √ |t|1/2 → +∞
8 |s|≥ 21 ( |t|
2 −1)
(2s + 1) 24|t|3/2 6 2
as |t| → +∞. This implies that the module of all non-zero solutions of the equation

x0 = a(t)x (13.42)

tend to +∞ as |t| → +∞, whereas those of the equation

y 0 = b(t)y, (13.43)

with b := −a ∈ H(a) tend to zero.

The above example is a slight modification of the well-known example of Favard

(see [137, p.435] or [150]). Our case differs from Favard’s example in that the
solutions of equation (13.43) are not only bounded on R, but they tend to zero as
|t| → +∞. Thus, a non-zero solution of equation (13.43) is asymptotically stable,
but the zero solution of equation (13.42) is not, even though a ∈ H(b).

Example 13.3 Assume that a ∈ C(R, R) is defined by the formula a(t) := −1 +

1
sin t 3 , t ∈ R. For equation (13.42) the sets

B+ := {(x, b) | x ∈ R, b ∈ H(a)} = R × H(a),

B := {(0, b) | b ∈ H(a)} ∪ {(x, θ) | x ∈ R},

where θ ∈ C(R, R) is the function identically equal to zero, are closed. Thus, the
recurrence of A in Theorem 13.12 (Theorem 13.13) is a sufficient condition, but this
condition is not necessary.

13.4 Relationship between different types of stability

In 1962 W. Hahn [168] posed the problem of whether asymptotic stability implies
uniform stability for linear equation

x0 = A(t)x (x ∈ Rn ) (13.44)

with almost periodic coefficients. C. C. Conley and R. K. Miller [123] gave a negative
answer to this by constructing a scalar equation x0 = a(t)x with the property that
every solution ϕ(t, x, a) → 0 as t → +∞, but the null solution is not uniformly
stable (see also [89]). From the results of R. J. Sacker and G. R. Sell [275] and
I. U. Bronshteyn [33, p.141] follows, that for linear system (13.44) with recurrent
 Linear almost periodic dynamical systems 427

(in particular, almost periodic) matrix from asymptotic stability of null solution of
system (13.44) and all system

x0 = B(t)x, (13.45)

where B ∈ H(A) := {Aτ : τ ∈ R}, Aτ is the translation of matrix A on τ and by

bar is denoted the closure in the topology of uniform convergence uniform on every
compact from R, follows the uniform stability of null solution of system (13.44).
Finally we note that from results of author [65] follows the validity of above men-
tioned result for arbitrary system (13.44) with compact matrix (i.e. when H(A) is
compact). Below we study the relationship between the asymptotic stability and
uniform stability of null solution of system (13.44) in the arbitrary Banach space.
Our main result is that for linear system (13.44) with recurrent coefficients in
the arbitrary Banach space the following statement takes place: if the null solution
of equation (13.44) and all the equation (13.45) are asymptotically stable, then the
null solution of equation (13.44) is uniformly stable.
From the theorem of Banach-Steinhauss follows that point dissipativity and
compact dissipativity are equivalent for autonomous linear system. One example of
linear autonomous dynamical system which is compact dissipative, but is not local
dissipative, is constructed in the section 1.6 (see example1.8).

Theorem 13.14 Let h(X, S+ , π), (Y, S, σ), hi be a linear non-autonomous

dynamical system and the following conditions hold:

1. Y is compact and minimal ( i.e. Y = H(y) := {yt : t ∈ S} for all y ∈ Y );

2. for any x ∈ X there exists Cx ≥ 0 such that

|xt| ≤ Cx (13.46)

for all t ∈ S+ ;
3. the mapping y 7→ kπyt k is continuous, where kπyt k is a norm of linear operator
πyt := π t |Xy , for every t ∈ S+ .

Then there exists M ≥ 0 such that the inequality

|π t x| ≤ M |x| (13.47)

takes place for all t ∈ S+ and x ∈ X.

Proof. From condition 2. and theorem of Banach-Steinhauss follows the uniform

boundedness of the family of linear operators {πyt : t ∈ S+ } for every y ∈ Y , i.e.
for any y ∈ Y there exists My ≥ 0 such that kπyt k ≤ My for all t ∈ S+ . We put

d(y) := sup{kπyt k : t ∈ S+ } (13.48)

428 Global Attractors of Non-autonomous Dissipative Dynamical Systems

and claim that the function d : Y → R+ , defined by equality (13.48) be lower-

semicontinuous, i.e. lim inf d(yn ) ≥ d(y) for all y ∈ Y and {yn } → y. Suppose that
yn →y
it is not true, then there exist y ∈ Y, {yn } and ε > 0 such that

lim inf d(yn ) = d(y) − ε. (13.49)

yn →y

From the equality (13.48) follows d(y) = lim kπytn k for some sequence {tn } ⊆ S+
n→+∞
and, consequently, there exists k such that
ε
|kπytn k − d(y)| < (13.50)
4
for all n ≥ k. According to continuity of mapping y 7→ kπyt k there exists n(k) such
that
ε
|kπytkn k − kπytk k| < (13.51)
4
for all n ≥ n(k). From (13.50) and (13.51) follows
ε
|d(y) − kπytkn k| < (13.52)
2
for all n ≥ n(k). From inequality (13.52) results
ε
|d(y) − d(yn )| ≤ (13.53)
2
for all n ≥ n(k). Inequality (13.53) contradicts (13.49). This contradiction proves
that d : Y → R+ is lower semi-continuous. Hence, this function has a set of points
of continuity D ⊂ Y of the type Gδ . Let p ∈ D, then there exist a positive numbers
δp and Mp such that d(y) ≤ Mp for all y ∈ B[p, δp ] = {y ∈ Y | ρ(y, p) ≤ δp } ⊂ Y.
Since Y is minimal, there are negative numbers t1 , t2 , ..., tm such that Y =
Sm
i=1 σ(S[p, δp ], ti ) (see [238, p.134]). We put L := max{ti |i = 1, 2, ..., m}.
Assume that m ∈ Y, y ∈ B[p, δp ] and ti are such that m = yti . Then
−ti
|xt| = |πyt+ti (πyt i
(x))| ≤ Mp C|x| (13.54)

for all x ∈ X and t ≥ L, where

C := max{max{kπy−ti k |y ∈ Y }, i = 1, 2, ..., m}.

We claim that the family of operators {π t : t ∈ [0, L]} is uniformly continuous,

that is, for any ε > 0 there is a δ(ε) > 0 such that |x| ≤ δ implies |xt| ≤ ε for all
t ∈ [0, L]. Assume the contrary. Then there are ε0 > 0, δn → 0 (δn > 0), |xn | < δn
and tn ∈ [0, L] such that

|xn tn | ≥ ε0 . (13.55)

Since (X, h, Y ) is locally trivial Banach fiber bundle and Y is compact, then the
zero section Θ = {θy : y ∈ Y } of (X, h, Y ) is compact and, consequently, we can
 Linear almost periodic dynamical systems 429

assume that the sequences {xn } and {tn } are convergent. Put x0 = lim xn and
n→+∞
t0 = lim tn , then x0 = θy0 (y0 = h(x0 )). Passing to the limit in (13.55) as
n→+∞
n → +∞, we obtain 0 = |x0 t0 | ≥ ε0 . The last inequality contradicts the choice of
ε0 . This contradiction proves the above assertion. If γ > 0 is such that |π t x| ≤ 1
for all |x| ≤ γ and t ∈ [0, L], then
1
|xt| ≤ |x| (13.4.12)
γ

for all t ∈ [0, L] and x ∈ X. We put M := max{γ −1 , Mp C}, then from (13.54) and
(13.4.12) follows the inequality (13.47) for all t ≥ 0 and x ∈ X. The theorem is
proved. 

Remark 13.6 a. If the fiber bundle (X, h, Y ) is finite-dimensional, then the

condition 3. of Theorem 13.14 holds.
b. Let X := E × Y, where E is a Banach space and π := (ϕ, σ), i.e. π t x :=
(ϕ(t, u, y), σ t y) for all t ∈ S+ and x := (u, y) ∈ X = E × Y. Then the condition
3. of Theorem 13.14 holds, if for every t ∈ S+ the mapping U (t, ·) : Y → [E] is
continuous, where U (t, y)u = ϕ(t, u, y) for any (t, u, y) ∈ S+ × E × Y and [E] is
a Banach space of all continuous operators acting onto E and equipped with the
operational norm.

Theorem 13.15 Let h(X, S+ , π), (Y, S, σ), hi be a linear non-autonomous

dynamical system, Y be a compact minimal set and the mapping y 7→ kπyt k be
continuous for each t ∈ S+ . Then from point dissipativity of h(X, S+ , π), (Y, S, σ), hi
follows its compact dissipativity.

Proof. Assume that the conditions of Theorem 13.15 are fulfilled and the non-
autonomous dynamical system h(X, S+ , π), (Y, S, σ), hi is point dissipative, then ac-
cording to Theorem 2.36 for every x ∈ X there exists a constant Cx ≥ 0 such that
the inequality (13.46) takes place for all x ∈ X and t ∈ S+ . Next to complete the
proof of Theorem 13.15, it is sufficient to refer to Theorems 13.14 and 2.37. 

Linear ordinary differential equations.

Let Λ be a complete metric space of linear operators that act on Banach space
E and C(R, Λ) be a space of all continuous operator-functions A : R → Λ equipped
with open-compact topology and (C(R, Λ), R, σ) be a dynamical system of shifts on
C(R, Λ).
Let Λ = [E] and consider the linear differential equation

u0 = A(t)u , (13.56)

where A ∈ C(R, Λ). Along with equation (13.56), we shall also consider its H−class,
430 Global Attractors of Non-autonomous Dissipative Dynamical Systems

that is, the family of equations

v 0 = B(t)v , (13.57)

where B ∈ H(A) := {Aτ : τ ∈ R}, Aτ (t) = A(t + τ ) (t ∈ R) and the bar denotes
closure in C(R, Λ). Let ϕ(t, u, B) be the solution of equation (13.57) that satis-
fies the condition ϕ(0, v, B) = v. We put Y := H(A) and denote the dynamical
system of shifts on H(A) by (Y, R, σ), then the triple h(X, R+ , π), (Y, R, σ), hi is a
linear non-autonomous dynamical system, where X := E × Y, π := (ϕ, σ) ( i.e.
π(τ, (v, B)) := (ϕ(τ, v, B), Bτ ) and h := pr2 : X → Y. Applying Theorem 13.15 to
this system, we obtain the following assertion.

Theorem 13.16 Let A ∈ C(R, Λ) be recurrent (i.e. H(A) is compact minimal

set of (C(R, Λ), R, σ) ) and the zero solution of equation (13.56) and all equation
(13.57) are asymptotically stable, i.e. lim |ϕ(t, v, B)| = 0 for all v ∈ E and
t→+∞
B ∈ H(A). Then the zero solution of equation (13.56) is uniformly stable, i.e. there
exists M ≥ 0 such that |ϕ(t, v, B)| ≤ M |v| for all t ≥ 0, v ∈ E and B ∈ H(A).

Proof. According to Lemma 2 [60] the mapping B 7→ ϕ(t, ·, B) from H(A) into [E]
is continuous for all t ∈ R. To finish the proof of the theorem it suffices to refer to
Theorem 13.14. 
Linear Partial differential equations. Let Λ be some complete metric
space of linear closed operators acting into Banach space E ( for example Λ =
{A0 + B|B ∈ [E]}, where A0 is a closed operator that acts on E). We assume that
the following conditions are fulfilled for equation (13.56) and its H− class (13.57):
a. for any v ∈ E and B ∈ H(A) equation (13.57) has exactly one solution that is
defined on R+ and satisfies the condition ϕ(0, v, B) = v;
b. the mapping ϕ : (t, v, B) → ϕ(t, v, B) is continuous in the topology of R+ ×
E × C(R; Λ);
c. for every t ∈ R+ the mapping U (t, ·) : H(A) → [E] is continuous, where U (t, ·)
is the Cauchy’s operator of equation (13.57), i.e. U (t, B)v := ϕ(t, v, B) (t ∈
R+ , v ∈ E and ∈H(A) ).
Under the above assumptions the equation (13.56) generates a linear non-
autonomous dynamical system h(X, R+ , π), (Y, R, σ), hi, where X = E × Y, π =
(ϕ, σ) and h := pr2 : X → Y. Applying Theorem 13.15 to this system, we will
obtain the analogue of Theorem 13.16 for different classes of partial differential
equations.
We will consider an example of partial differential equation which satisfies the
above conditions a.-c. Let H be a Hilbert space with a scalar product h·, ·i =
| · |2 , D(R+ , H) be a set of all infinite differentiable and finite into R+ functions with
values into H.
 Linear almost periodic dynamical systems 431

Denote by (C(R, [H]), R, σ) a dynamical system of shifts on C(R, [H]). Consider

the equation
Z
[hu(t), ϕ0 (t)i + hA(t)u(t), ϕ(t)i]dt = 0, (13.58)
R+

along with the family of equations

Z
[hu(t), ϕ0 (t)i + hB(t)u(t), ϕ(t)i]dt = 0, (13.59)
R+

where B ∈ H(A) := {Aτ |τ ∈ R}, Aτ (t) := (t + τ ) and the bar denotes closure in
C(R, [H]).
The function u ∈ C(R+ , H) is called a solution of equation (13.58), if (13.58)
takes place for all ϕ ∈ D(R+ , H).
Assume that the operator A(t) is self-adjoint. Let (H(A), R, σ) be a dynamical
system of shifts on H(A), ϕ(t, v, B) be a solution of equation (13.59) with condition
ϕ(0, v, B) = v, X̃ = H × H(A), X be a set of all the points hu, Bi ∈ X̃ such that
through point u ∈ H passes a solution ϕ(t, u, A) of equation (13.58) defined on
R+ . According to Lemma 2.21 [92] the set X is closed in X̃. In virtue of Lemma
2.22 [92] the triple (X, R+ , π) is a dynamical system on X ( where π := (ϕ, σ))
and h(X, R+ , π), (Y, R, σ), hi is a linear non-autonomous dynamical system, where
h := pr2 : X → Y = H(A). Applying the results from [5] it is possible to show that
for every t the mapping B 7→ U (t, B) ( where U (t, B)v := ϕ(t, v, B)) from H(A)
into [H] is continuous and, consequently, for this system is applicable Theorem
13.14. Thus the following assertion takes place.

Theorem 13.17 Let A ∈ C(R, [H]) be recurrent and the zero solution of equation
(13.56) and all equation (13.57) are asymptotically stable, i.e. lim |ϕ(t, v, B)| = 0
t→+∞
for all v ∈ E and B ∈ H(A). Then the zero solution of equation (13.56) is uniformly
stable, i.e. there exists M ≥ 0 such that |ϕ(t, v, B)| ≤ M |v| for all t ≥ 0, v ∈ H and
B ∈ H(A).

We will give the example of limit problem reducing to equation of type (13.58).
Let Ω be a bounded domain in Rn , Γ be frontier of Ω, Q = R+ × Ω and S = R+ × Γ.
Consider in Q the first initial limit problem for equation
∂u
= L(t)u (u|t=0 = ϕ, u|S = 0), (13.60)
∂t
where
X n
∂ ∂u
L(t)u := (aij (t, x) ) − a(t, x)u
i,j=1
∂x i ∂x j
432 Global Attractors of Non-autonomous Dissipative Dynamical Systems

The operator A(t) according to theorem of Riesz is defined by equality

Z X n
∂u ∂ϕ
hA(t)u, ϕi = − [ aij (t, x) + a(t, x)uϕ]dx.
i,j=1
∂xj ∂xi
Ω

If aij (t, x) = aji (t, x) and the functions aij (t, x) and a(t, x) are recurrent (almost
periodic) with respect to t ∈ R uniformly with respect to x ∈ Ω, then for equation
(13.60) it is applicable Theorem 13.17, if H = Ẇ21 (Ω)
Linear functional-differential equations. Denote by A = A(C, Rn ) a
Banach space of all linear continuous operators acting from C := C([−r, 0], R n ) into
Rn , equipped by operational norm. Consider the equation

u0 = A(t)ut , (13.61)

where A ∈ C(R, A). We put H(A) := {Aτ : τ ∈ R}, Aτ (t) := A(t + τ ) and the bar
denotes closure in the topology of uniform convergence on every compact from R.
Along with equation (13.61) we also consider the family of equations

u0 = B(t)ut , (13.62)

where B ∈ H(A). Let ϕt (v, B) be a solution of equation (13.62) with condition

ϕ0 (v, B) = v, defined on R+ . We put Y := H(A) and denote by (Y, R, σ) a
dynamical system of shifts on H(A). Let X := C × Y and π := (ϕ, σ) a dynamical
system on X, defined by equality π(t, (v, B)) := (ϕτ (v, B), Bτ ). A non-autonomous
dynamical system h(X, R+ , π), (Y, R, σ), hi (h := pr2 : X → Y ) is linear. The
following assertion takes place.

Lemma 13.9 Let H(A) be compact in C(R, A), then the non-autonomous
dynamical system h(X, R+ , π), (Y, R, σ), hi generated by equation (13.61) is com-
pletely continuous.

Proof. Let B ⊂ C = C[−r, 0] be a bounded set and t ≥ r. According to the

continuity of mapping ϕ : R+ × C × H(A) 7→ C and compactness of H(A) there
exists a positive number M such that |ϕτ (v, B)| ≤ M and |B(τ )ϕτ (v, B)| ≤ M
for all τ ∈ [0, t], B ∈ H(A) and v ∈ B, and, consequently, |ϕ̇(τ, v, B)| ≤ M for
all τ ∈ [0, t], B ∈ H(A) and v ∈ B, i.e. the family of functions {ϕt (v, B) : B ∈
H(A), v ∈ B} ( for t ≥ r) is uniformly continuous on [−r, 0]. Therefore this family
of functions is relatively compact. The Lemma is proved. 

Theorem 13.18 Let H(A) be compact. Then the following assertion are equiva-
lent:

(1) for any B ∈ H(A) the zero solution of equation (13.62) is asymptotically stable,
i.e. lim |ϕt (v, B)| = 0 for all v ∈ C and B ∈ H(A);
t→+∞
 Linear almost periodic dynamical systems 433

(2) the zero solution of equation (13.61) is uniformly asymptotically stable, i.e.
there are the positive numbers N and ν such that |ϕt (v, B)| ≤ N e−νt |v| for all
t ≥ 0, v ∈ C and B ∈ H(A).

Proof. Let h(X, R+ , π), (Y, R, σ), hi be a linear non-autonomous dynamical system,
generated by equation (13.61). According to Lemma 13.9 this system is completely
continuous and to finish the proof it is sufficiently to refer to Theorems 2.38 and
6.10. 

Consider the neutral functional differential equation

d
Dxt = A(t)xt , (13.63)
dt
where A ∈ C(R, A) and D ∈ A is non-atomic at zero operator [175, p.67]. As well
as in the case of equation (13.61), the equation (13.63) generates a linear dynamical
system h(X, R+ , π), (Y, R, σ), hi, where X := C × Y, Y := H(A) and π := (ϕ, σ).
The following statement takes place.

Lemma 13.10 Let H(A) be compact and the operator D is stable, i.e. the zero
solution of homogeneous difference equation Dyt = 0 is uniformly asymptotically
stable. Then a linear non-autonomous dynamical system h(X, R+ , π), (Y, R, σ), hi,
generated by equation (13.63), is asymptotically compact.

Proof. According to [179] (see, p.119, formula (5.18)) the mapping ϕt (·, B) : C → C
can be written as

ϕt (·, B) = St (·) + Ut (·, B)

for all B ∈ H(A), where Ut (·, B) is conditionally completely continuous for t ≥ r

and there exist constants N > 0, ν > 0 such that kSt k ≤ N e−νt (t ≥ 0). To finish
the proof of Lemma 13.10 it is sufficiently to refer to Theorem 2.22. 

Theorem 13.19 Let A ∈ C(R, A) be recurrent (i.e. H(A) is compact minimal in

the dynamical system of shifts (C(R, A), R, σ) ) and D is stable, then the following
assertions are equivalent:

(1) the zero solution of equation (13.61) and all equation

d
Dxt = B(t)xt , (13.64)
dt
where B ∈ H(A), are asymptotically stable, i.e. lim |ϕt (v, B)| = 0 for all
t→+∞
v ∈ C and B ∈ H(A) (ϕt (v, B) is a solution of equation (13.64) with condition
ϕ0 (v, B) = v);
434 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(2) the zero solution of equation (13.63) is uniformly exponentially stable, i.e. there
are a positive numbers N and ν such that |ϕt (v, B)| ≤ N e−νt |v| for all t ≥ 0, v ∈
C and B ∈ H(A).

Proof. Let h(X, R+ , π), (Y, R, σ), hi be a linear non-autonomous dynamical system,
generated by equation (13.63). According to Lemma 13.10 this system is asymp-
totically compact. To finish the proof of Theorem 13.19 it is sufficiently to refer to
Theorems 2.38 and 1.25. The theorem is proved. 

13.5 Linear α-condensing systems

Let A(t) be a continuous n×n matrix-function and H(A) be the family of all matrix-
functions B = lim Atn , where {tn } ⊂ R, Atn (t) = A(tn + t) and the convergence
n→+∞
Atn → B is uniform on every compact subset of R. The following result is well
known.

Theorem 13.20 [275, 33, 65] Let A be a bounded and uniformly continuous
matrix-function on R, then the following conditions are equivalent:

1. The trivial solution of equation

x0 = A(t)x (13.65)

is uniformly exponentially stable.

2. The trivial solution of equation (13.65) is uniformly asymptotically stable.
3. The trivial solution of equation (13.65) and every equation

y 0 = B(t)y (B ∈ H(A)) (13.66)

is asymptotically stable.

For equations in infinite-dimensional spaces conditions 1., 2., and 3. are not
equivalent; see example 1.8 and also [81, 121, 249]. However, in the general infinite-
dimensional case condition 1. implies condition 2., and condition 2. implies condi-
tion 3.
Linear non-autonomous dynamical systems satisfying one of the conditions 1.,
or 2., or 3. are studied in [81]; see also Chapter 2 (section 2.11). We will show that
if the operator corresponding to the the Cauchy’s problem for (13.65) satisfies some
compactness condition, then condition 3 implies condition 1 (see Theorems 13.23
and 13.24).
For recurrent (almost periodic) systems this result is made precise in Theorems
13.25 and 13.26. Applications of this result to different classes of linear evolution
equations (ordinary linear differential equations in a Banach space, retarded and
 Linear almost periodic dynamical systems 435

neutral functional differential equations, some classes of evolution partial differential

equations) are given.
Assume that X and Y are complete metric spaces, R is the set of real numbers,
Z is the set of integer numbers, S = R or Z, S+ = {t ∈ S : t ≥ 0} and S− = {t ∈
S|t ≤ 0}.
Recall that a dynamical system (X, S+ , π) is said to be conditionally β-
condensing [179] if there exists t0 > 0 such that β(π t0 B) < β(B) for all bounded sets
B in X with β(B) > 0. The dynamical system (X, S+ , π) is said to be β-condensing
if it is conditionally β-condensing and the set π t0 B is bounded for all bounded sets
B ⊆ X.
According to Lemma 2.3.5 in [179, p.15] and Lemma 3.3 in [82] the conditional
condensing dynamical system (X, S+ , π) is asymptotically compact.
Let E be a metric space, X := E × Y , A ⊂ X, and Ay := {x ∈ A : pr2 x = y}.
Then A = ∪{Ay : y ∈ Y }. Let Ãy := pr1 Ay and à = ∪{Ãy : y ∈ Y }. Note that if
the space Y is compact, then a set A ⊂ X is bounded in X if and only if the set Ã
is bounded in E.

Lemma 13.11 The equality α(A) = α(Ã) takes place for all bounded sets A ⊂ X,
where α(A) and α(Ã) are the Kuratowskii’s measure of non-compactness for the sets
A ⊂ X and à ⊂ E.

Proof. Let ε > 0 and A be a bounded subset in X, then there are sets A1 , A2 , . . . , An
such that A = ∪{Ai : i = 1, 2, . . . , n} and diam Ai < α(A) + ε. Note that à =
∪{Ãi : i = 1, 2, . . . , n} and diam Ãi ≤ diam Ai < α(A) + ε, and consequently,
α(Ã) ≤ α(A).
Let ε be a positive constant, A be a bounded set in X, Ã = ∪{Ãk : k =
1, 2, . . . , m} and diam Ãk < α(Ã) + ε. Since Y is compact, there are sets
Y1 , Y2 , . . . , Y` such that Y1 ∪ Y2 ∪ · · · ∪ Y` = Y and diam Yj < ε (j = 1, 2, . . . `).
Let Ai = pr1−1 (Ãi ) ∩ A, and

Aij = pr2−1 (pr2 (pr1−1 (Ãi ) ∩ A) ∩ Yj ) ∩ Ai .

Note that Aij ⊆ Ãi × Yj , and that

diam Aij ≤ diam Ãi + diam Yj < α(Ã) + ε + ε = α(A) + 2ε .

Since A = ∪{Aij : i = 1, 2, . . . , n, j = 1, 2, . . . , `}, it follows that α(A) ≤ α(Ã) and

α(A) = α(Ã). which concludes the present proof. 

Definition 13.5 A cocycle hE, ϕ, (Y, ,σ)i is called conditionally α-condensing

if there exists t0 > 0 such that for any bounded set B ⊆ E the inequality
α(ϕ(t0 , B, Y )) < α(B) holds if α(B) > 0. The cocycle ϕ is called α-condensing if
436 Global Attractors of Non-autonomous Dissipative Dynamical Systems

it is a conditional α-condensing cocycle and the set ϕ(t0 , B, Y ) = ∪{ϕ(t0 , u, Y )|u ∈

B, y ∈ Y } is bounded for all bounded set B ⊆ E.

Definition 13.6 A cocycle ϕ is called conditional α-contraction of order k ∈

[0, 1), if there exists t0 > 0 such that for any bounded set B ⊆ E for
which ϕ(t0 , B, Y ) = ∪{ϕ(t0 , u, Y )|u ∈ B, y ∈ Y } is bounded the inequality
α(ϕ(t0 , B, Y )) ≤ kα(B) holds. The cocycle ϕ is called α-contraction if it is a condi-
tional α-contraction cocycle and the set ϕ(t0 , B, Y ) = ∪{ϕ(t0 , u, Y )|u ∈ B, y ∈ Y }
is bounded for all bounded sets B ⊆ E.

Lemma 13.12 Let Y be compact and the cocycle ϕ be α-condensing. Then

the skew-product dynamical system (X, S+ , π), generated by the cocycle ϕ, is α-
condensing too.

Proof. Let A ⊂ X be a bounded subset, t0 > 0 and α(A) > 0, then

π(t0 , A) = ∪{π(t0 , Ay |y ∈ Y }
= ∪{(ϕ(t0 , Ay , y), yt)|y ∈ Y } ⊆ ϕ(t0 , Ã, Y ) × Y. (13.67)

Since A is bounded, Ã is also bounded in E and according to the condition of

the lemma the set ϕ(t0 , Ã, Y ) is bounded and, consequently, π(t0 , A) is bounded.
According to Lemma 13.11 and (13.67) we have

α(π(t0 , A)) = α(∪{(ϕ(t0 , Ay , y), yt0 )|y ∈ Y }) ≤ α(ϕ(t0 , Ã, Y )) < α(Ã) = α(A).

The lemma is proved. 

Theorem 13.21 Let E be a Banach space, ϕ be a cocycle on (Y, S, σ) with fiber

E and the following conditions be fulfilled:

(1) ϕ(t, u, y) = ψ(t, u, y) + γ(t, u, y) for all t ∈ S+ , u ∈ E and y ∈ Y.

(2) There exists a function m : R2+ → R+ satisfying the condition m(t, r) → 0 as
t → +∞ (for every r > 0) such that |ψ(t, u1 , y) − ψ(t, u2 , y)| ≤ m(t, r)|u1 − u2 |
for all t ∈ S+ , u1 , u2 ∈ B[0, r] and y ∈ Y .
(3) γ(t, A, Y ) is compact for all bounded A ⊂ X and t > 0.

Then the cocycle ϕ is an α-contraction.

Proof. Let ε > 0 and A be a bounded set in E, then there are sets A1 , A2 , . . . , An
such that A = ∪{Ai : i = 1, 2, . . . , n} and diam Ai < α(A) + ε for i = 1, 2, . . . , n.
Since Y is compact, then there are a sets Y1 , Y2 , . . . , Ym such that Y1 ∪Y2 ∪· · ·∪Ym =
Y with condition diam Yj < ε for all j = 1, 2, . . . , m.
 Linear almost periodic dynamical systems 437

Let r := diam A and t0 be a positive number such that m(t0 , r) < 1. We note
that

ϕ(t0 , A, Y ) ⊆ ψ(t0 , A, Y ) + γ(t0 , A, Y )

= ∪{ψ(t0 , Ai , Yj )|i = 1, 2, . . . , n; j = 1, 2, .., m} + γ(t0 , A, Y ). (13.68)

According to the conditions of Theorem 13.21, α(γ(t0 , A, Y )) = 0 and

diam ψ(t0 , Ai , y) ≤ m(t0 , r) diam Ai

for all y ∈ Y . Thus we have

|ψ(t0 , u1 , y1 ) − ψ(t0 , u2 , y2 )| ≤ |ψ(t0 , u1 , y1 ) − ψ(t0 , u2 , y1 )|

+|ψ(t0 , u2 , y1 ) − ψ(t0 , u2 , y2 )| (13.69)

and, consequently,

diam ψ(t0 , Ai , y) ≤ m(t0 , r) diam Ai + diam ψ(t0 , u2 , Yj ) for all y ∈ Yj . (13.70)

Since Y is compact, from (13.69)-(13.70) follows the inequality

diam ψ(t0 , Ai , Yj ) ≤ m(t0 , r) diam Ai ≤ m(t0 , r)(α(A) + ε)

and, consequently, α(ϕ(t0 , A, Y )) ≤ m(t0 , r)α(A). The theorem is proved. 

13.6 Exponential stable systems

Theorem 13.22 Let h(X, S+ , π), (Y, S, σ), hi be a linear non-autonomous

dynamical system, Y be a compact set. Then the following conditions are equiv-
alent:

1. The non-autonomous dynamical system h(X, S+ , π), (Y, S, σ), hi is uniformly ex-
ponentially stable, i.e. there exist two positive constants N and ν such that
|π(t, x)| ≤ N e−νt |x| for all t ∈ S+ and x ∈ X.
2. kπ t k → 0 as t → +∞, where kπ t k = sup{|π t x| : x ∈ X, |x| ≤ 1}.
3. The non-autonomous dynamical system h(X, S+ , π), (Y, S, σ), hi is locally dissi-
pative.

Proof. According to Theorem 2.38, conditions 1. and 3. are equivalent. Now we

will prove that the conditions 1. and 2. are equivalent. It is clear that from 1.
follows 2. According to condition 2. there exists L > 0 such that

kπ t k ≤ 1 (13.71)

for all t ≥ L. We claim that the family of operators {π t : t ∈ [0, L]} is uniformly
continuous, that is, for any ε > 0 there is a δ(ε) > 0 such that |x| ≤ δ implies
438 Global Attractors of Non-autonomous Dissipative Dynamical Systems

|xt| ≤ ε for all t ∈ [0, L]. On the contrary, assume that there are ε0 > 0, δn > 0
with δn → 0, |xn | < δn and tn ∈ [0, L] such that

|xn tn | ≥ ε0 . (13.72)

Since (X, h, Y ) is a locally trivial Banach fiber bundle and Y is compact, the zero
section Θ := {θy : y ∈ Y, θy ∈ Xy , |θy | = 0} of (X, h, Y ) is compact and, con-
sequently, we can assume that the sequences {xn } and {tn } are convergent. Put
x0 = lim xn and t0 = lim tn , then x0 = θy0 (y0 = h(x0 )). Passing to the
n→+∞ n→+∞
limit in (13.72) as n → +∞, we obtain 0 = |x0 t0 | ≥ ε0 . This last inequality contra-
dicts the choice of ε0 , and hence proves the above assertion. If γ > 0 is such that
|π t x| ≤ 1 for all |x| ≤ γ and t ∈ [0, L], then
1
|xt| ≤ |x| (13.73)
γ
for all t ∈ [0, L] and x ∈ X. We put M := max{γ −1 , 1}, then from (13.71) and
(13.73) follows

kπ t k ≤ M (13.74)

for all t ≥ 0 and x ∈ X. Consider the function m(t) := kπ t k. We note that

m(t + τ ) ≤ m(t)m(τ ) for all t, τ ∈ S+ and m(t) ≤ M for all t ∈ S+ and m(t) → 0 as
t → +∞. According to Lemma 2.19 there exist positive numbers N and ν such that
m(t) ≤ N e−νt for all t ∈ S+ . Thus |π(t, x)| ≤ kπ t k|x| ≤ N e−νt |x| for all t ∈ S+ and
x ∈ X. The theorem is proved. 
Let B := {x ∈ X : ∃γ ∈ Φx such that sup |γ(t)| < +∞}.
t∈S

Theorem 13.23 Let h(X, S+ , π), (Y, S, σ), hi be a linear non-autonomous

dynamical system, Y be compact and (X, S+ , π) be conditionally α-condensing.
Then the following assertions are equivalent:

(1) The non-autonomous dynamical system h(X, S+ , π), (Y, S, σ), hi is point dissi-
pative and this system doesn’t admit non-trivial bounded trajectories on S, i.e.
B ⊆ Θ = {θy : y ∈ Y, θy ∈ Xy , |θy | = 0}.
(2) The non-autonomous dynamical system h(X, S+ , π), (Y, S, σ), hi is uniformly ex-
ponentially stable.

Proof. Denote by Θ = {θy : y ∈ Y, θy ∈ Xy , |θy | = 0} the zero section of the vector

fibering (X, h, Y ). Since (X, h, Y ) is locally trivial and Y is compact, the zero section
Θ is compact and an invariant set of the dynamical system (X, S+ , π). Taking
into account that the dynamical system (X, S+ , π) is conditionally α-condensing,
according to Theorem 2.4.8 [179] the set Θ is orbitally stable and in particular there
exists a positive constant N such that |xt| ≤ N |x| for all t ∈ S+ and x ∈ X. By
 Linear almost periodic dynamical systems 439

virtue of Theorem 2.38 the dynamical system (X, S+ , π) is compact dissipative and
according to Theorem 2.38 (X, S+ , π) is local dissipative. It follows from Theorem
13.22 that (X, S+ , π) is uniformly exponentially stable.
Let now the non-autonomous dynamical system h(X, S+ , π), (Y, S, σ), hi be uni-
formly exponentially stable, then according to Theorem 13.22 it is locally dissipa-
tive. Let J be its Levinson’s centre (i.e. maximal compact invariant set of dynamical
system (X, S+ , π)) . We note that according to the linearity of non-autonomous
dynamical system h(X, S+ , π), (Y, S, σ), hi we have J = Θ. Let ϕ be a entire
bounded trajectory of dynamical system (X, S+ , π) . Since the non-autonomous
dynamical system h(X, S+ , π), (Y, S, σ), hi is conditionally α-condensing, in partic-
ularly, it is asymptotically compact and the set M = ϕ(S) is relatively compact. In
fact, the set M is invariant Ω(M ) = M and in view of Lemma 3.3 [82] the set M is
relatively compact. We note that ϕ(S) ⊆ J = Θ because J is the maximal compact
invariant set of (X, S+ , π). The theorem is proved. 

Remark 13.7 Theorem A in [277] implies a version of Theorem 13.23 under

slightly stronger assumptions.

Theorem 13.24 Let h(X, S+ , π), (Y, S, σ), hi be a linear non-autonomous

dynamical system, Y be compact and (X, S+ , π) be completely continuous, i.e. for
any bounded set A ⊆ X there exists a positive number ` such that π ` (A) is relatively
compact. Then the following assertions are equivalent:

1. The non-autonomous dynamical system h(X, S+ , π), (Y, S, σ), hi is uniformly ex-
ponentially stable.
2. lim |π t x| = 0 for all x ∈ X.
t→+∞

Proof. It is clear that condition 1 implies 2. Now we will show that condition 1
follows from 2. According to Theorem 2.38 the non-autonomous dynamical system
h(X, S+ , π), (Y, S, σ), hi is point dissipative. Since the dynamical system (X, S+ , π)
is completely continuous, by virtue of Theorem 2.38 the dynamical system (X, S+ , π)
is locally dissipative. To prove the theorem it is sufficient to refer to Theorem 13.22.


13.7 Linear system with a minimal base

In this section we study a linear system h(X,S+ ,π), (Y,S,σ),hi with compact minimal
base (Y, S, σ).

Theorem 13.25 Let h(X, S+ , π), (Y, S, σ), hi be a linear non-autonomous

dynamical system and the following conditions hold:

(1) Y is compact and minimal.

440 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(2) The dynamical system (X, S+ , π) is asymptotically compact.

(3) The mapping y 7→ kπyt k is continuous, where kπyt k is the norm of the linear
operator πyt := π t |Xy , for every t ∈ S+ or (X, S+ , π) is a skew-product dynamical
system.

Then the non-autonomous dynamical system h(X, S+ , π), (Y, S, σ), hi is uni-
formly exponentially stable if and only if

lim |π t x| = 0 (13.75)
t→+∞

for all x ∈ X.

Proof. It is clear that from uniform exponential stability of the non-autonomous

dynamical system h(X, S+ , π), (Y, S, σ), hi follows the equality (13.75).
From condition (13.75) and minimality of (Y, S, σ) by virtue of Theorem 13.14 it
follows that for the non-autonomous dynamical system h(X, S+ , π), (Y, S, σ), hi the
inequality

|π(t, x)| ≤ M |x| (13.76)

holds for all t ∈ S+ and x ∈ X. According to Theorem 2.38 this system is com-
pact dissipative. Since the dynamical system (X, S+ , π) is asymptotically com-
pact, to finish the proof of the Theorem it is sufficient to remark that accord-
ing to Theorem 2.13 [86] every compact dissipative and asymptotically compact
dynamical system is local dissipative. Thus a linear non-autonomous dynamical
system h(X, S+ , π), (Y, S, σ), hi is local dissipative and, consequently, it is uniform
exponentially stable. The theorem is proved. 

Theorem 13.26 Suppose that the following conditions are satisfied:

(1) A dynamical system (Y, S, σ)is compact and minimal.

(2) A linear non-autonomous dynamical system h(X, S+ , π), (Y, S, σ), hi is genera-
ted by cocycle ϕ (i.e. X = E × Y , π = (ϕ, σ) and h = pr2 : X 7→ Y ).
(3) The dynamical system (X, S+ , π) is conditionally α-condensing.
(4) There exists a positive number M such that |ϕ(t, u, y)| ≤ M |u| for all t ∈ Y
and t ∈ S+ .

Then there are two vectorial positively invariant sub-fiberings (X 0 , h, Y ) and

s
(X , h, Y ) of (X, h, Y ) such that:

a. Xy = Xy0 + Xys and Xy0 ∩ Xys = θy for all y ∈ Y , where θy = (0, y) ∈ X = E × Y

and 0 is the zero in the Banach space E.
b. The vector sub-fibering (X 0 , h, Y ) is finite dimensional, invariant (i.e. π t X 0 =
X 0 for all t ∈ S+ ) and every trajectory of the dynamical system (X, S+ , π)
belonging to X 0 is recurrent.
 Linear almost periodic dynamical systems 441

c. There exist two positive numbers N and ν such that |ϕ(t, u, y)| ≤ N e−νt |u| for
all (u, y) ∈ X s and t ∈ S+ .

Proof. Let X 0 := B, then according to Theorem 13.2, statement b. holds. Denote

by Py the projection of Xy := h−1 (y) to By := B∩h−1 (y), then Py (u, y) = (P(y)u, y)
for all u ∈ E, P 2 (y) = P(y) and the mapping P : Y → [E] ( y 7→ P(y)) is continuous,
where by [E] denotes the set of all linear continuous operators acting on E. Now we
set Xys := Q(y)Xy and X s := ∪{Xys : y ∈ Y }, where Q(y) := IdE − P(y). We will
show that X s is closed in X. In fact, let {xn } = {(un , yn )} ⊆ X s and x0 = (u0 , y0 ) =
lim xn . Note that Py0 (x0 ) = (P(y0 )u0 , y0 ) = ( lim P(yn )un , y0 ) = (0, y0 ) = θy0
n→∞ n→∞
and, consequently, x0 ∈ Xys0 ⊆ X s .
Let (X s , S+ , π) be the dynamical system induced by (X, S+ , π). It is clear
that under the conditions of Theorem 13.26 the dynamical system (X s , S+ , π) is
asymptotically compact and every positive semi-trajectory is relatively compact
and, consequently, lim |π(t, x)| = 0 for all x ∈ X s because the dynamical system
t→∞
(X s , S+ , π) doesn’t have a non-trivial entire trajectory bounded on S. In fact, if
we suppose that it is not true, then there exist x0 = (u0 , y0 ) and tn → +∞ such
that: |u0 | 6= 0, lim π(tn , x) = x0 and through point x0 pass a non-trivial entire
n→+∞
trajectory bounded on S. This contradiction proves the necessary assertion. Thus
we can apply Theorem 2.38 according which there exist two positive constants N
and ν such that |ϕ(t, u, y)| ≤ N e−νt |u| for all (u, y) ∈ X s and t ∈ S+ . The theorem
is proved. 

13.8 Some classes of uniformly exponentially stable equations

Let Λ be a complete metric space of linear operators that act on a Banach space E
and C(R, Λ) be the space of all continuous operator-functions A : R → Λ equipped
with the open-compact topology and (C(R, Λ), R, σ) be the dynamical system of
shifts on C(R, Λ).
Linear ordinary differential equations. Let Λ = [E] and consider the linear
differential equation

u0 = A(t)u , (13.77)

where A ∈ C(R, Λ). Along with equation (13.77), we shall also consider its H-class,
that is, the family of equations

v 0 = B(t)v , (13.78)

where B ∈ H(A) := {Aτ : τ ∈ R}, Aτ (t) = A(t + τ ) (t ∈ R), and the bar denotes
closure in C(R, Λ). Let ϕ(t, u, B) be the solution of equation (13.78) that satis-
fies the condition ϕ(0, u, B) = u. We put Y := H(A) and denote the dynamical
442 Global Attractors of Non-autonomous Dissipative Dynamical Systems

system of shifts on H(A) by (Y, R, σ). Then the triple h(X, R+ , π), (Y, R, σ), hi is
a linear non-autonomous dynamical system, where X := E × Y , π := (ϕ, σ); i.e.,
π((v, B), τ ) := (ϕ(τ, v, B), Bτ ) and h := pr2 : X → Y ).

Lemma 13.13 [51, 60]

(i) The mapping (t, u, A) 7→ ϕ(t, u, A) of R × E × C(R, [E]) to E is continuous,

and
(ii) the mapping A 7→ U (·, A) of C(R, [E]) to C(R, [E]) is continuous, where U (·, A)
is the Cauchy’s operator [132] of equation (13.77).

Theorem 13.27 Let A ∈ C(R, Λ) be compact (i.e. H(A) is a compact set of

(C(R, Λ), R, σ) ), then the following conditions are equivalent:

1. The trivial solution of equation (13.77) is uniformly exponentially stable, i.e.

there exist positive numbers N and ν such that kU (t, A)U (τ, A)−1 k ≤ N e−(t−τ )
for all t ≥ τ.
2. There exist positive numbers N and ν such that kU (t, B)U (τ, B)−1 k ≤ N e−(t−τ )
for all t ≥ τ and B ∈ H(A).
3. lim sup{kU (t, B)k : B ∈ H(A)} = 0.
t→+∞

Proof. Applying Theorem 13.22 to the non-autonomous system h(X,R+ ,π),

(Y,R,σ), hi, generated by equation (13.77), we obtain the equivalence of conditions
2. and 3. According to Lemma 3 [60] conditions 1. and 2. are equivalent. The
theorem is proved. 

Theorem 13.28 Let A ∈ C(R, Λ) be recurrent with respect to t ∈ S (i.e. H(A)

is a compact and minimal set of (C(R, Λ), R, σ) ), the non-autonomous system
h(X, R+ , π), (Y, R, σ), hi generated by equation (13.77) is asymptotically compact.
Then the following conditions are equivalent:

1. The trivial solution of equation (13.77) is uniformly exponentially stable, i.e.

there exist positive numbers N and ν such that kU (t, A)U (τ, A)−1 k ≤ N e−(t−τ )
for all t ≥ τ.
2. lim sup |ϕ(t, u, B)| = 0 for every u ∈ E and B ∈ H(A).
t→+∞

Proof. According to Lemma 13.13 the mapping U (t, ·) : [E] → [E] is continuous
and, consequently, the mapping B 7→ kU (t, B)k is also continuous for every t ∈ S.
Now applying Theorem 13.22 to non-autonomous system h(X, R+ , π), (Y, R, σ), hi
generated by equation (13.77), we obtain the equivalence of conditions 1. and 2.
The theorem is proved. 
 Linear almost periodic dynamical systems 443

We now formulate some sufficient conditions for the α− condensedness (in par-
ticular, asymptotic compactness) of the linear non-autonomous dynamical system
generated by equation (13.77).

Theorem 13.29 Let A ∈ C(R, [E]), A(t) = A1 (t) + A2 (t) for all t ∈ R, and
assume that H(Ai ) (i = 1, 2) are compact and the following conditions hold:

(i) The zero solution of the equation

u0 = A1 (t)u (13.79)

is uniformly asymptotically stable, that is, there are positive numbers N and ν
such that

kU (t, A1 )U −1 (τ, A1 )k ≤ N e−ν(t−τ ) (13.80)

for all t ≥ τ (t, τ ∈ R), where U (t, A1 ) is the Cauchy’s operator of equation
(13.79).
(ii) The family of operators {A2 (t) : t > 0} is uniformly completely continuous, that
is, for any bounded set A ⊂ E the set {A2 (t)A : t > 0} is relatively compact.

Then the linear non-autonomous dynamical system generated by equation

(13.77) is an α-contraction.

Proof. First of all we note that the set ϕ(t, A, Y ) is bounded for every t > 0
and bounded set A ⊆ E. Let B ∈ H(A). Then there are {tn } ⊂ S such that
B(t) = B1 (t) + B2 (t) and Bi (t) = lim Ai (t + tn ). Note that
t→+∞
Z t
ϕ(t, v, B) = U (t, B1 )v + U (t, B1 )U −1 (τ, B1 )B2 (τ )ϕ(τ, v, B)dτ.
0

By Lemma 13.13,

U (t, Bi ) = lim U (t, Aitn ), Aitn (t) = Ai (t + tn ),

t→+∞

and the equality

U (t, A1tn )U −1 (τ, A1tn ) = U (t + tn , A1 )U −1 (τ + tn , A1 )

and inequality (13.80) imply that

kU (t, B1 )U −1 (τ, B1 )k ≤ N e−ν(t−τ ) (13.81)

for all t ≥ τ and B1 ∈ H(A1 ). Theorem 13.29 will be proved if we can prove that
the set
Z t
{ U (t, B1 )U −1 (τ, B1 )B2 (τ )ϕ(τ, v, B)dτ : (v, B) ∈ A}
0
444 Global Attractors of Non-autonomous Dissipative Dynamical Systems

is relatively compact for every t > 0 and every bounded positively invariant set
A ⊆ E × Y . We put
t
KA = {B2 (τ )ϕ(τ, v, B) : τ ∈ [0, t], (v, B) ∈ A}
t
and we note that the set KA is compact. Really, the set ϕ([0, t], A) = ∪{ϕ(τ, v, B) :
τ ∈ [0, t], (v, B) ∈ A} is bounded because |ϕ(τ, v, B)| ≤ eM t r for all τ ∈ [0, t]
and (v, B) ∈ A, where r = sup{|v| : ∃B ∈ H(A), such that (v, B) ∈ A} and
M = sup{kA(t)k : t ∈ S}. Then
Z t
U (t, B1 )U −1 (τ, B1 )B2 (τ )ϕ(τ, v, B)dτ (13.82)
0
∈ t conv{U (t, B1 )U −1 (τ, B1 )w : 0 ≤ τ ≤ t, B1 ∈ H(A1 ), w ∈ KA
t
}.
t
Since H(A1 ), H(A), and KA are compact sets, then formula (13.82), condition (ii)
of Theorem 13.29, and Lemma 13.13 imply that {U (t, B1 )U −1 (τ, B1 )w : 0 ≤ τ ≤
t
t, B1 ∈ H(A1 ), w ∈ KA } is compact, which completes the proof of the theorem. 

Theorem 13.30 Let H(A) be compact and assume that there is a finite-
dimensional projection P ∈ [E] such that

(i) the family of projections {P (t) : t ∈ R}, where P (t) := U (t, A)P U −1 (t, A), is
relatively compact in [E], and
(ii) there are positive numbers N and ν such that

kU (t, A)QU −1 (τ, A)k ≤ N e−ν(t−τ )

for all t ≥ τ, where Q := I − P .

Then the linear non-autonomous dynamical system generated by equation

(13.77) is an α-contraction.

Proof. Since the family of projections P (t) = U (t, A)P U −1 (t, A) is relatively
compact in [E], the family H = {P (t) : t ∈ R} is uniformly completely continuous,
where the bar denotes closure in [E]. Indeed, let A be a bounded subset of E,
{xn } ⊆ {QA : Q ∈ H}, and εn ↓ 0. Then there are tn ∈ R and vn ∈ A such
that |xn − P (tn )vn | ≤ εn . Since the sequence {P (tn )} is relatively compact, we can
assume that it converges. Let L := lim P (tn ). Then L is completely continuous,
n→+∞
which implies that the sequence {x0n } = {Lvn } is relatively compact. Note that

|xn − x0n | ≤ |xn − P (tn )vn | + |P (tn )vn − Lvn | ≤ εn + kP (tn ) − Lk|vn |,

which implies that |xn − x0n | → 0 as n → +∞. Hence, {xn } is relatively compact.
Assume that B ∈ H(A) and {tn } ⊂ R are such that

B = lim Atn , P (B) = lim P (Atn ),

n→+∞ n→+∞
 Linear almost periodic dynamical systems 445

where P (Atn ) = U (tn , A)P U −1 (tn , A). The assertions proved above imply that the
family {P (B) : B ∈ H(A)} is uniformly completely continuous. Note that Q(B) =
lim Q(Atn ), where Q(B) := I − P (B) and Q(Atn ) = I − P (Atn ). Moreover,
n→+∞
condition (ii) of Theorem 13.30 implies that

kU (t, Atn )QU −1 (τ, Atn )k ≤ N e−ν(t−τ ) (13.83)

for all t ≥ τ . Passing to the limit in (13.83) as n → +∞ and taking into account
Lemma 13.13, we obtain that

kU (t, B)QU −1 (τ, B)k ≤ N e−ν(t−τ )

for all t ≥ τ and B ∈ H(A). We complete the proof of the theorem by observing
that U (t, B)Q(B) + U (t, B)P (B) = U (t, B) and applying Theorem 13.21. 

Theorem 13.31 Suppose that the following conditions are satisfied:

(1) The operator-function A ∈ C(R, [E]) is recurrent with respect to t ∈ R.
(2) The linear non-autonomous dynamical system h(X, S+ , π), (Y, S, σ), hi gener-
ated by equation (13.77) is conditionally α-condensing.
(3) the trivial solution of equation (13.77) is uniformly stable in the positive direc-
tion, i.e. there exist a positive number M such that

kU (t, A)U −1 (τ, A)k ≤ M (13.84)

for all t ≥ τ .
Then there are two vectorial positively invariant sub-fibers (X 0 , h, Y ) and
(X s , h, Y ) of (X, h, Y ) such that:
a. Xy = Xy0 + Xys and Xy0 ∩ Xys = θy for all y ∈ Y , where θy = (0, y) ∈ X = E × Y
and 0 is the zero in the Banach space E.
b. The vectorial sub-fiber (X 0 , h, Y ) is finite dimensional, invariant (i.e. π t X 0 =
X 0 for all t ∈ S+ ) and every trajectory of a dynamical system (X, S+ , π) be-
longing to X 0 is recurrent.
c. There exist two positive numbers N and ν such that |ϕ(t, u, B)| ≤ N e−νt |u| for
all (u, B) ∈ X s and t ∈ S+ , where ϕ(t, u, B) := U (t, B)u.

Proof. Assume that B ∈ H(A) and {tn } ⊂ R are such that B = lim Atn , then
n→+∞
condition (3) of Theorem 13.31 implies that

kU (t, Atn )U −1 (τ, Atn )k ≤ N (13.85)

for all t ≥ τ . Passing to the limit in (13.85) as n → +∞ and taking into account
Lemma 13.13, we obtain that

kU (t, B)U −1 (τ, B)k ≤ N

446 Global Attractors of Non-autonomous Dissipative Dynamical Systems

for all t ≥ τ and B ∈ H(A) and, consequently,

kU (t, B)k ≤ N

for all t ≥ 0 and B ∈ H(A). Now to finish the proof of Theorem 13.31 it is
sufficiently to refer on Theorem 13.26. 

Linear partial differential equations. Let Λ be some complete metric space

of linear closed operators acting into Banach space E ( for example Λ = {A 0 +
C|C ∈ [E]}, where A0 is a closed operator that acts on E). We assume that the
following conditions are fulfilled for equation (13.77) and its H− class (13.78):

a. for any v ∈ E and B ∈ H(A) equation (13.78) has exactly one mild solution
ϕ(t, v, B) (i.e. ϕ(·, v, B) is continuous, defined on R+ and satisfies of equation
Z t
ϕ(t, v, B) = U (t, B)v + U (t − τ, B)ϕ(τ, v, B)dτ. (13.86)
0

and the condition ϕ(0, v, B) = v;

b. the mapping ϕ : (t, v, B) → ϕ(t, v, B) is continuous in the topology of R+ ×
E × C(R; Λ);
c. for every t ∈ R+ the mapping U (t, ·) : H(A) → [E] is continuous, where U (t, ·)
is the Cauchy’s operator of equation (13.78), i.e. U (t, B)v := ϕ(t, v, B) (t ∈
R+ , v ∈ E and B ∈ H(A) ).

Under the above assumptions the equation (13.77) generates a linear non-
autonomous dynamical system h(X, R+ , π), (Y, R, σ), hi, where X := E × Y, π =
(ϕ, σ) and h := pr2 : X → Y. Applying the results from the sections 13.6 and 13.7
to this system, we will obtain the analogous assertions for different classes of partial
differential equations.
We will consider examples of partial differential equations which satisfy the
above conditions a.-c.

Example 13.4 Consider the differential equation

u0 = (A1 + A2 (t))u, (13.87)

where A1 is a sectorial operator that does not depend on t ∈ R, and A2 ∈ C(R, [E]).
The results of [188], [238] imply that equation (13.87) satisfies conditions a.-c.

Example 13.5 Let H be a Hilbert space with a scalar product h·, ·i = | · |2 ,

D(R+ , H) be the set of all infinite differentiable, bounded functions on R+ with
values into H.
 Linear almost periodic dynamical systems 447

Denote by (C(R, [H]), R, σ) a dynamical system of shifts on C(R, [H]). Consider

the equation
Z
[hu(t), ϕ0 (t)i + hA(t)u(t), ϕ(t)i]dt = 0, (13.88)
R+

along with the family of equations

Z
[hu(t), ϕ0 (t)i + hB(t)u(t), ϕ(t)i]dt = 0, (13.89)
R+

where B ∈ H(A) := {Aτ |τ ∈ R}, Aτ (t) := (t + τ ) and the bar denotes closure in
C(R, [H]).
The function u ∈ C(R+ , H) is called a solution of equation (13.88), if (13.88)
takes place for all ϕ ∈ D(R+ , H).
Let (H(A), R, σ) be a dynamical system of shifts on H(A), ϕ(t, v, B) be a solution
of (13.89) with condition ϕ(0, v, B) = v, X̃ := H ×H(A), X be a set of all the points
hu, Bi ∈ X̃ such that through point u ∈ H passes a solution ϕ(t, u, A) of equation
(13.88) defined on R+ . According to Lemma 2.21 in [92] the set X is closed in
X̃. In virtue of Lemma 2.22 in [92] the triple (X, R+ , π) is a dynamical system on
X (where π := (ϕ, σ) ) and h(X, R+ , π), (Y, R, σ), hi is a linear non-autonomous
dynamical system, where h := pr2 : X → Y := H(A).

Applying the results from [5] it is possible to show that for every t the mapping
B 7→ U (t, B) (where U (t, B)v := ϕ(t, v, B)) from H(A) into [H] is continuous
and, consequently, for this system Theorem 13.22 is applicable. Thus the following
assertion takes place.

Theorem 13.32 Let A ∈ C(R, [H]) be compact, then the following assertion hold:

(1) The trivial solution of equation (13.77) is uniformly exponentially stable, i.e.
there exist positive numbers N and ν such that kU (t, B)k ≤ N e−νt for all t ≥ 0
and B ∈ H(A).
(2) lim sup{kU (t, B)k : B ∈ H(A)} = 0.
t→+∞

Proof. This statement follows directly from Theorem 13.22. 

Linear functional-differential equations. Consider the equation

u0 = A(t)ut , (13.90)

where A ∈ C(R, A). We put H(A) := {Aτ : τ ∈ R}, Aτ (t) = A(t + τ ), where the
bar denotes closure in the topology of uniform convergence on every compact of R.
448 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Along with equation (13.90) we also consider the family of equations

u0 = B(t)ut , (13.91)

where B ∈ H(A). Let ϕt (v, B) be a solution of equation (13.91) with condition

ϕ0 (v, B) = v defined on R+ . We put Y := H(A) and denote by (Y, R, σ) the
dynamical system of shifts on H(A). Let X := C × Y and π := (ϕ, σ) the dynamical
system on X defined by the equality π(τ, (v, B)) := (ϕτ (v, B), Bτ ). The non-
autonomous dynamical system h(X, R+ , π), (Y, R, σ), hi (h := pr2 : X → Y ) is
linear. The following assertion takes place.

Theorem 13.33 Let H(A) be compact. Then the following assertions are equiv-
alent:

(1) For any B ∈ H(A) the zero solution of equation (13.91) is asymptotically stable,
i.e. lim |ϕt (v, B)| = 0 for all v ∈ C and B ∈ H(A)
t→+∞
(2) The zero solution of equation (13.90) is uniformly exponentially stable, i.e.
there are positive numbers N and ν such that |ϕt (v, B)| ≤ N e−νt |v| for all
t ≥ 0, v ∈ C and B ∈ H(A).

Proof. Let h(X, R+ , π), (Y, R, σ), hi be the linear non-autonomous dynamical
system, generated by equation (13.90). According to Lemma 13.9 this system is
completely continuous and to finish the proof it is sufficient to refer to Theorem
13.21. 

Consider the neutral functional differential equation

d
Dxt = A(t)xt , (13.92)
dt
where A ∈ C(R, A) and D ∈ A is an operator non-atomic at zero [19, p.67]. As
well as in the case of equation (13.90), the equation (13.92) generates a linear
non-autonomous dynamical system h(X, R+ , π), (Y, R, σ), hi, where X = C × Y ,
Y = H(A) and π = (ϕ, σ). The following statement takes place.

Lemma 13.14 Let H(A) be compact and the operator D is stable; i.e., the zero
solution of the homogeneous difference equation Dyt = 0 is uniformly asymptotically
stable. Then the linear non-autonomous dynamical system h(X, R+ , π), (Y, R, σ), hi,
generated by equation (13.92), is conditionally α-condensing.

Proof. According to [20, p.119, formula (5.18)] the mapping ϕt (·, B) : C → C can
be written as

ϕt (·, B) = St (·) + Ut (·, B)

 Linear almost periodic dynamical systems 449

for all B ∈ H(A), where Ut (·, B) is conditionally completely continuous for t ≥ r.

Also there exist positive constants N, ν such that kSt k ≤ N e−νt (t ≥ 0). Then the
proof is completed by referring to Theorem 13.23. 

Theorem 13.34 Let A ∈ C(R, A) be recurrent (i.e. H(A) is compact minimal

set in the dynamical system of shifts (C(R, A), R, σ) ) and D is stable, then the
following assertions are equivalent:

(1) The zero solution of equation (13.90) and all equations

d
Dxt = B(t)xt , (13.93)
dt
where B ∈ H(A), is asymptotically stable, i.e. lim |ϕt (v, B)| = 0 for all v ∈ C
t→+∞
and B ∈ H(A) (ϕt (v, B) is the solution of equation (13.93) with condition
ϕ0 (v, B) = v).
(2) The zero solution of equation (13.92) is uniformly exponentially stable, i.e.
there are positive numbers N and ν such that |ϕt (v, B)| ≤ N e−νt |v| for all
t ≥ 0, v ∈ C and B ∈ H(A).
(3) All solutions of all equations (13.93) are bounded on R+ and they don’t have
non-trivial solutions bounded on R .

Proof. Let h(X, R+ , π), (Y, R, σ), hi be the linear non-autonomous dynamical
system, generated by equation (13.92). According to Lemma 13.14 this system
is conditionally α condensing. To finish the proof of Theorem 13.34 it is sufficient
to refer to Theorems 13.23 and 13.26. 

Theorem 13.35 Let A ∈ C(R, A) be recurrent (i.e. H(A) is compact minimal

in the dynamical system of shifts (C(R, A), R, σ) ), D is stable, and all solutions of
all equations (13.93) are bounded on R+ . Let h(X, S+ , π), (Y, S, σ), hi be the linear
non-autonomous dynamical system generated by equation (13.92), then there are
two positively invariant vector sub-fiberings (X 0 , h, Y ) and (X s , h, Y ) of (X, h, Y )
such that:

a. Xy = Xy0 + Xys and Xy0 ∩ Xys = θy for all y ∈ Y , where θy = (0, y) ∈ X = E × Y

and 0 is the zero in the Banach space E.
b. The vector subfibering (X 0 , h, Y ) is finite dimensional, invariant (i.e.π t X 0 =
X 0 for all t ∈ S+ ) and every trajectory of a dynamical system (X, S+ , π) be-
longing to X 0 is recurrent.
c. There exist two positive numbers N and ν such that |ϕt (v, B)| ≤ N e−νt |v| for
all (v, B) ∈ X s , where ϕt (v, B) := U (t, B)v and U (t, B) is the Cauchy operator
of equation (13.93).
450 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Proof. Let h(X, S+ , π), (Y, S, σ), hi be the linear non-autonomous dynamical system
generated by equation (13.92). In virtue of Lemma 13.14 this non-autonomous
dynamical system is conditionally α-condensing. According to Theorem 13.1 there
exists a positive number M such that |ϕt (v, B)| ≤ M |v| for all t ≥ 0, v ∈ C and
B ∈ H(A). To finish the proof of Theorem 13.35 it is sufficient to refer to Theorem
13.26. 

13.9 Linear periodic systems

This section is devoted to the study of linear periodic dynamical systems, possessing
the property of uniform exponential stability. It is proved that if the Cauchy’s oper-
ator of these systems possesses a certain compactness property, then the asymptotic
stability implies the uniform exponential stability. We also give the application to
different classes of linear evolution equations, such as ordinary linear differential
equations in the space of Banach, retarded and neutral functional differential equa-
tions, some classes of evolution partial differential equations.
Let A(t) be a τ -periodic continuous n × n matrix-function. It is well-known that
the following three conditions are equivalent:

1. The trivial solution of equation

u0 = A(t)u (13.94)

is uniformly exponentially stable.

2. The trivial solution of equation (13.94) is uniformly asymptotically stable.
3. The trivial solution of equation (13.94) is asymptotically stable.

For equations in infinite-dimensional spaces the statements 1.-3. are not equiv-
alent, as shown by the examples in [121, 249].
It is clear that in general for the infinite-dimensional case condition 1. implies
2. and 2. implies 3. In this section we show that if the Cauchy operator of equation
(13.94) satisfies some compactness condition, then the condition 3. implies 1. (see
Theorem 13.37 below).
Applications to different classes of linear evolution equations (ordinary linear
differential equations in a Banach space, retarded and neutral functional-differential
equations, some classes of evolutionary partial differential equations) are given.
The exponential dichotomy of asymptotically compact cocycles was studied by
R. Sacker and G. Sell [277]. The general case was studied by C. Chicone and Yu.
Latushkin [117] (see also their references), Yu. Latushkin and R. Schnaubelt [236],
and many other authors.
 Linear almost periodic dynamical systems 451

13.9.1 Exponential stable linear periodic dynamical systems

Let X and Y be complete metric spaces, (X, h, Y ) be a locally trivial Banach fiber
bundle over Y [29], [E] be a Banach space of the all linear continuous operators
acting onto Banach space E with the operator norm and U : S+ × Y 7→ [E] be
a mapping with properties: U (0, y) = I, U (t + τ, y) = U (t, σ(τ, y))U (τ, y) for all
y ∈ Y and t, τ ∈ S+ and the mapping ϕ(·, u, ·) : S+ × Y → E (ϕ(t, u, y) := U (t, y)u)
is continuous for every u ∈ E.

Definition 13.7 A triplet h[E], U, (Y, S, σ)i is called a c0 − cocycle on (Y, S, σ)

with fiber [E].

Lemma 13.15 Let h[E], U, (Y, S, σ)i be a c0 − cocycle on (Y, S, σ) with fiber [E]
and Y be a compact, then the following assertions hold:
(1) For every ` > 0 there exists a positive number M (`) such that kU (t, y)k ≤ M (`)
for all t ∈ [0, `] and y ∈ Y ;
(2) The mapping ϕ : S+ × E × Y 7→ E (ϕ(t, u, y) = U (t, y)u) is continuous;
(3) There exist positive numbers N and ν such that kU (t, y)k ≤ N eνt for all t ∈ S+
and y ∈ Y .

Proof. Let ` > 0 and u ∈ E, then there exists a positive number M (`, u) such that
|U (t, y)u| ≤ M (`, u) for all (t, y) ∈ [0, `]×Y because the mapping (t, y) → U (t, y)u is
continuous. According to principle of uniformly boundedness there exists a positive
number M (`) such that kU (t, y)k ≤ M (`) for all (t, y) ∈ [0, `] × Y .
Let now (t0 , u0 , y0 ) ∈ S+ × E × Y and tn → t0 , un → u0 and yn → y0 , then we
have

|ϕ(tn , un , yn ) − ϕ(t0 , u0 , y0 )|

≤ |ϕ(tn , un , yn ) − ϕ(tn , u0 , yn )| + |ϕ(tn , u0 , yn ) − ϕ(t0 , u0 , y0 )|

≤ kU (tn , yn )(un − u0 )k + |(U (tn , yn ) − U (t0 , y0 ))u0 |. (13.95)

In view of first statement of Lemma 13.15 there exists the positive number M such
that

kU (tn , yn )k ≤ M (13.96)

for all n ∈ N. From inequalities (13.95) and (13.96) follows the continuity of
mapping ϕ : S+ × E × Y → E (ϕ(t, u, y) = U (t, y)u).
Denote by a := sup{kU (t, y)k : (t, y) ∈ [0, 1] × Y } and let t ∈ S+ , t = n + τ (n ∈
N, τ ∈ [0, 1)), then we obtain

kU (t, y)k ≤ kU (n, yτ )kkU (τ, y)k ≤ an+1 ≤ N eνt

for all t ∈ S+ and y ∈ Y , where N := a and ν := ln a. 

452 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Lemma 13.16 [121, Chapter 9] Let m : R+ 7→ R+ be a positive and continuous

function. If there exists a positive constant M such that m(t + s) ≤ M m(t) for all
R +∞
s ∈ [0, 1] and t ∈ S+ , then 0 m(t)dt < +∞ implies m(t) → 0 as t → +∞.

Theorem 13.36 Let h[E], U, (Y, S, σ)i be the C0 − cocycle on (Y, S, σ) with fiber
[E] and (Y, S, σ) be a periodical dynamical system (i.e. there are y0 ∈ Y and τ ∈
S (τ > 0) such that Y = {y0 t : 0 ≤ t < τ }). Then the following conditions are
equivalent:

(i)

lim kU (t, y0 )k = 0. (13.97)

t→+∞

(ii) There exist positive constants N and ν such that for all t ∈ S+ and y ∈ Y ,

kU (t, y)k ≤ N e−νt . (13.98)

(iii) There exists p ≥ 1 such that for all u ∈ E,

Z +∞
|U (t, y0 )u|p dt < +∞ . (13.99)
0

Proof. We remark that from equality (13.97) follows the condition

lim sup kU (s + nτ, y0 )k = 0. (13.100)

n→+∞ 0≤s≤τ

In fact, by virtue of Lemma 13.15 there exists a positive constant M such that

kU (s, y)k ≤ M (13.101)

for all s ∈ [0, τ ] and y ∈ Y . Therefore,

kU (s + nτ, y0 )k = kU (s, y0 )U (nτ, y0 )k ≤ M kU (nτ, y0 )k (13.102)

for all 0 ≤ s ≤ τ . Consequently, from (13.97) and (13.102) results the condition
(13.100).
We will show that under the condition (13.100) the equality

lim sup kU (t, y)k = 0 (13.103)

t→+∞ y∈Y

holds. In fact, let y ∈ Y then there exists a number s ∈ [0, τ ) such that y = y0 s
and, consequently, for t ∈ S+ (t = nτ + t̄, t̄ ∈ [0, τ )) we obtain
 Linear almost periodic dynamical systems 453

kU (t, y)k = kU (t, y0 s)k = kU (nτ + t̄, y0 s)k

= kU ((n − 1)τ + t̄ + s, y0 τ )U (τ − s, y0 s)k

≤ M max{ sup kU ((n − 1)τ + s, y0 )k, sup kU (nτ + s, y0 )k}. (13.104)

0≤s≤τ 0≤s≤τ

From (13.100) and (13.104) results the equality (13.103). For finishing the proof
that (i) implies (ii) is sufficient to apply Theorem 13.22.
The fact that (ii) implies (iii) is obvious. Now we prove that (iii) implies (i).
Indeed, let u ∈ E and we consider the function m(t) = |U (t, y0 )u|p (t ≥ 0). We
note that
m(t + s) =|U (t + s, y0 )u|p = |U (s, y0 t)U (t, y0 )u|p
≤kU (s, y0 t)kp |U (t, y0 )u|p ≤ M p m(t)

for all t ∈ S+ and s ∈ [0, 1], where M := sup kU (s, y)k. By Lemma 13.16
0≤s≤1,y∈Y
m(t) → 0 as t → +∞ and, consequently,

lim |U (t, y0 )u|p = 0 (13.105)

t→+∞

for all u ∈ E. Let now y ∈ Y , then there exists s ∈ [0, τ ) such that y = y0 s and for
t ≥ τ − s we have

U (t, y)u = U (t, y0 s)u = U (t − τ + s, y0 )U (τ − s, y0 s)u. (13.106)

From equalities (13.105) and (13.106),

lim |U (t, y)u|p = 0 (13.107)

t→+∞

for all u ∈ E and y ∈ Y. According to Theorem 13.1 there exists a positive number
M such that kU (t, y)k ≤ M for all t ∈ S+ and y ∈ Y . Let t > 0 and u ∈ E, then
we obtain
Z t
t|U (t, y0 )u|p = |U (t, y0 )u|p ds ≤ (13.108)
0
Z t
|U (t − s, y0 s)|p |U (s, y0 )u|p ds
0
≤
Z t Z +∞
Mp |U (s, y0 )u|p ds ≤ M p |U (s, y0 )u|p ds = Cu
0 0

for all t ≥ 0. By virtue of principle of uniformly boundedness there exists a positive

number C such that

tkU (t, y0 )kp ≤ C

454 Global Attractors of Non-autonomous Dissipative Dynamical Systems

for all t > 0 and, consequently

1 1
kU (t, y0 )k ≤ C p t− p → 0

as t → +∞. This completes the present proof. 

Remark 13.8 a. Theorem 13.36 (the equivalence of assertions (ii) and (iii)) is
a variant of the Datko-Pazy theorem (see [121],[133] and [134]) for the cocycle over
periodic dynamical systems.
b. Periodic, almost periodic and asymptotic almost periodic mild solutions of
inhomogeneous periodic Cauchy’s problems considered recently by C. J. K. Batty,
W.Hutter and F. Räbiger [22] and W. Hutter [191].

Definition 13.8 The operator U (τ, y0 ) is called operator of monodromy for τ -

periodic cocycle U (t, y). The number 0 6= λ ∈ C is called multiplicator of operator
of monodromy U (τ, y0 ) if there exists u0 ∈ E (u0 6= 0) such that U (τ, y0 )u0 = λu0
(or, what is the same, U (t + τ, y0 )u0 = λU (t, y0 )u0 for all t ∈ S+ ).

Remark 13.9 a. Condition (13.97) and the equality

lim kU (nτ, y0 )k = 0. (13.109)

n→+∞

are equivalent. We show that (13.109) implies (13.97) as follows. Let now t = nτ +
s, 0 ≤ s < τ , then U (t, y0 ) = U (s + nτ, y0 ) = U (s, y0 )U (nτ, y0 ) and, consequently,

kU (t, y0 )k ≤ max kU (s, y0 )kkU (nτ, y0 )k. (13.110)

0≤s≤τ

From conditions (13.109) and (13.110) results (13.97).

b. Condition (13.98) and the inequality

kU (t, y0 )k ≤ N1 e−ν1 t (∀t ∈ S+ ) (13.111)

are equivalent, where N1 and ν1 are some positive constants. Indeed, from (13.111),
taking into account (13.106), we obtain (13.98).
c. Condition (13.109) is satisfied if and only if σ(U (τ, y0 )) ⊂ D = {z ∈ C :
|z| < 1}, where σ(U (τ, y0 )) is a spectrum of operator of monodromy U (τ, y0 ). In
fact, from (13.98) results that rU (τ,y0 ) := lim sup(kU (nτ, y0 ))k)1/n ≤ e−ν < 1,
n→+∞
because U n (τ, y0 ) = U (nτ, y0 ). If γ = rU (τ,y0 ) < 1, then for all ε > 0 there exists
a n(ε) ∈ N such that (kU (nτ, y0 )k)1/n ≤ γ + ε for all n ≥ n(ε) and, consequently,
kU (nτ, y0 )k ≤ (γ + ε)n for all n ≥ n(ε). Thus kU (nτ, y0 )k → 0 as n → +∞.

Theorem 13.37 Let h[E], U, (Y, S, σ)i be a c0 − cocycle on (Y, S, σ) with fiber [E],
(Y, S, σ) be a periodic dynamical system and U (τ, y0 ) be asymptotically compact (i.e.
if kn → +∞ (kn ∈ N), the sequences {un } ⊆ E and {U (kn τ, y0 )un } are bounded;
 Linear almost periodic dynamical systems 455

then the sequence {U (kn τ, y0 )un } is relatively compact). Then the following condi-
tions are equivalent

1. Equality (13.97) holds.

2. For all u ∈ E,

lim |U (t, y0 )u| = 0 . (13.112)

t→+∞

Proof. It is evidently that 1. implies 2. Now, under the conditions of Theorem

13.37 the mapping P := U (τ, y0 ) : E 7→ E is asymptotically compact because P n =
U (nτ, y0 ). From condition (13.112) according to uniform boundedness principle it
follows that there is a positive constant M such that kP n k ≤ M for all n ∈ Z+ and,
consequently, the set B = ∪{P n x : |x| ≤ 1, n ∈ Z+ } is bounded and P (B) ⊂ B.
Since the mapping P is asymptotically compact in virtue of Lemma 1.3 the set
\ [
ω(B) = P m (B)
n≥0 m≥n

is nonempty, compact, and invariant and ω(B) attracts B.

Now we will prove that lim kP n k = 0. If we suppose the contrary, then there
n→+∞
are ε0 > 0, {xn }(|xn | ≤ 1) and nk → +∞({nk } ⊂ Z+ ) such that

|P nk xk | ≥ ε0 . (13.113)

Since P is asymptotically compact without loss of generality we can suppose that the
sequence {P nk xk } is convergent. Let x̄ := lim P nk xk , then x̄ ∈ ω(B) and from
k→+∞
(13.113) we have |x̄| ≥ ε0 > 0. According to the invariance of the set ω(B) there
exists a beside sequence {wn }n∈Z ⊂ ω(B) such that: w0 = x̄ and P (wn ) = wn+1
for all n ∈ Z. We note that

inf |wn | = 0. (13.114)

n∈Z−

Suppose that it is not true, then there is a positive number ` such that

|wn | ≥ ` (13.115)

for all n ∈ Z− . Let p := lim wnk and {zn } ⊆ αw0 , where

k→+∞
\ [
αw 0 = wm ,
n≤0 m≤n

be a beside sequence such that z0 = p and P (zn ) = zn+1 for all n ∈ Z. From the
inequality (13.115) results that |zn | ≥ ` for all n ∈ Z. On the other hand in view
of (13.112) lim |wn | = lim |P n w0 | = 0. The obtained contradiction proves the
n→+∞ n→+∞
equality (13.114).
456 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Let now nr → −∞ and |wnr | → 0, then w0 = P −nr wnr for all r ∈ N and,
consequently, |w0 | = 0 because |w0 | ≤ kP −nr k|wnr | ≤ M |wnr |. On the other hand
|w0 | = |x̄| ≥ ε0 > 0. The obtained contradiction finishes the proof of our assertion.
The Theorem is proved. 

Remark 13.10 C.Buşe wrote several papers [36],[37]and [38] on evolution pe-
riodic processes that are in the spirit of the current section. In particularly, in
[38] it is proved that a trivial solution of equation u0 (t) = A(t)u(t) with p−
periodic coefficients on a separable Hilbert space H is uniformly exponentially
stable if the mild solution uµx of a well-posed inhomogeneous Cauchy’s problem
u0 (t) = A(t)u(t) + eiµt x(t ≥ 0), µ ∈ R, u(0) = 0 satisfies the following condition
sup sup |uµx (t)| < +∞, ∀x ∈ H.
µ∈R t>0

13.9.2 Some classes of linear uniformly exponentially stable

periodic differential equations
Let Λ be the complete metric space of linear operators that act on Banach space E
and C(R, Λ) be the space of all continuous operator-functions A : R → Λ equipped
with open-compact topology and (C(R, Λ), R, σ) be the dynamical system of shifts
on C(R, Λ).
Linear ordinary differential equations. Let Λ = [E] and consider the
linear differential equation

u0 = A(t)u , (13.116)

where A ∈ C(R, Λ). Along with equation (13.116), we shall also consider its
H−class, that is, the family of equations

v 0 = B(t)v , (13.117)

where B ∈ H(A) := {As : s ∈ R}, As (t) := A(t + s) (t ∈ R) and the bar denotes
closure in C(R, Λ). Let ϕ(t, u, B) be the solution of equation (13.117) that satisfies
the condition ϕ(0, v, B) = v. We put Y := H(A) and denote the dynamical system
of shifts on H(A) by (Y, R, σ), then the triple h[E], U, (Y, R, σ)i is the linear cocycle
on (Y, R, σ), where U (t, B) := ϕ(t, ·, B) for all t ∈ R and B ∈ Y .
According to Lemma 13.13

(i) The mapping (t, u, A) 7→ ϕ(t, u, A) of R × E × C(R, [E]) to E is continuous, and

(ii) the mapping U : A → U (·, A) of C(R, [E]) to C(R, [E]) is continuous, where
U (·, A) is the Cauchy’s operator [93] of equation (13.116).

Theorem 13.38 Let A ∈ C(R, Λ) be τ -periodic (i.e. A(t + τ ) = A(t) for all
t ∈ R), then the following conditions are equivalent:
 Linear almost periodic dynamical systems 457

1. The trivial solution of (13.116) is uniformly exponentially stable, i.e. there exist
positive numbers N and ν such that kU (t, A)U (τ, A)−1 k ≤ N e−ν(t−τ ) for all
t ≥ τ.
2. There exist positive numbers N and ν such that kU (t, B)U (τ, B)−1 k ≤
N e−ν(t−τ ) for all t ≥ τ and B ∈ H(A) = {As : s ∈ [0, τ )}.
3. lim kU (t, A)k = 0.
t→+∞
R +∞
4. There exists p ≥ 1 such that 0 |U (t, A)u|p dt < +∞ for all u ∈ E.

Proof. Applying Theorem 13.36 to the cocycle h[E], U, (Y, R, σ)i, generated by
equation (13.116) we obtain the equivalence of conditions 2., 3. and 4. According
to Lemma 3 [60] the conditions 1. and 2. are equivalent. The theorem is proved.


Theorem 13.39 Let A ∈ C(R, Λ) be τ -periodic and U (τ, A) be asymptotically

compact, then the following conditions are equivalent:

(1) The trivial solution of equation (13.116) is uniformly exponentially stable.

(2) lim |U (t, A)u| = 0 for every u ∈ E.
t→+∞

Proof. Applying Theorem 13.37 to non-autonomous system h(X, R+ , π),

(Y, R, σ), hi generated by equation (13.116), we obtain the equivalence of condi-
tions 1. and 2. The theorem is proved. 

Linear partial differential equations. Let Λ be some complete metric

space of linear closed operators acting into a Banach space E (for example Λ =
{A0 + B|B ∈ [E]}, where A0 is a closed operator that acts on E). We assume that
the following conditions are fulfilled for equation (13.116) and its H-class (13.117):

(a) for any v ∈ E and B ∈ H(A) equation (13.117) has exactly one mild solution
defined on R+ and satisfies the condition ϕ(0, v, B) = v;
(b) the mapping ϕ : (t, v, B) → ϕ(t, v, B) is continuous in the topology of R+ ×
E × C(R; Λ);

Under the assumptions above, (13.116) generates a linear cocycle h[E], U, (Y, R, σ)i,
where U (t, B) := ϕ(t, ·, B).
Applying the results from subsections 13.9.1 and 13.9.2 to this cocycle, we will
obtain the analogous assertions for different classes of partial differential equations.
We will consider examples of partial differential equations which satisfy the
above conditions a. and b.
Consider the differential equation

u0 = (A0 + A(t))u , (13.118)

458 Global Attractors of Non-autonomous Dissipative Dynamical Systems

where A0 is a sectorial operator that does not depend on t ∈ R, and A ∈ C(R, [E]).
The results of [122] imply that equation (13.118) satisfies conditions a. and b.
Under the assumptions above, (13.118) generates a linear cocycle h[E], U,
(Y, R, σ)i, where Y := H(A) and U (t, B) := ϕ(t, ·, B). Applying the results from
subsections 13.9.1 and 13.9.2 to this system, we will obtain the following results.

Theorem 13.40 Let A0 - be the sectorial operator and A ∈ C(R, Λ) be τ -periodic,

then the following conditions are equivalent:

(1) The trivial solution of equation (13.118) is uniformly exponentially stable, i.e.
there exist positive numbers N and ν such that kU (t, A0 + A)U (τ, A0 + A)−1 k ≤
N e−ν(t−τ ) for all t ≥ τ.
(2) There exist positive numbers N and ν such that kU (t, A0 +B)U (τ, A0 +B)−1 k ≤
N e−ν(t−τ ) for all t ≥ τ and B ∈ H(A).
(3) lim kU (t, A0 + A)k = 0.
t→+∞
R +∞
(4) There exists p ≥ 1 such that 0 |U (t, A0 + A)u|p dt < +∞ for all u ∈ E.
(5) σ(U (τ, A0 + A)) ⊂ D.

Theorem 13.41 Let A0 - be the sectorial operator with compact resolvant and
A ∈ C(R, Λ) be τ - periodic, then the following conditions are equivalent:

1. The trivial solution of equation (13.118) is uniformly exponentially stable.

2. lim |U (t, A0 + A)u| = 0 for every u ∈ E.
t→+∞
3. |λ| < 1 for every multiplicator λ of operator of monodromy U (τ, A 0 + A).

Proof. Since the sectorial operator A0 admits a compact resolvant, then in view of
Lemma 7.2.2 [122] the operator U (τ, A0 + A) is compact and, consequently (see,for
example [328, p.391-396]), every 0 6= λ ∈ σ(U (τ, A0 + A)) is a multiplicator for
operator of monodromy U (τ, A0 + A). Applying Theorem 13.40 (see also Remark
13.8) to linear cocycle h[E], U, (Y, R, σ)i generated by equation (13.118), we obtain
the equivalence of conditions 1., 2. and 3. The theorem is proved. 

Linear functional-differential equations. Let A = A(C, Rn ) be the Banach

space of all linear continuous operators acting from C into Rn , equipped by operator
norm. Consider the equation

u0 = A(t)ut , (13.119)

where A ∈ C(R, A). We put H(A) := {Aτ : τ ∈ R}, Aτ (t) := A(t + τ ) and the bar
denotes the closure in the topology of uniform convergence on compacts of R.
Along with equation (13.119) we also consider the family of equations

u0 = B(t)ut , (13.120)
 Linear almost periodic dynamical systems 459

where B ∈ H(A). Let ϕt (v, B) be a solution of equation (13.120) with condition

ϕ0 (v, B) = v defined on R+ . We put Y := H(A) and denote by (Y, R, σ) the
dynamical system of shifts on H(A). Let C := C([−r, 0], Rn ), X := C × Y and
π := (ϕ, σ) the dynamical system on X, defined by the equality π(τ, (v, B)) :=
(ϕτ (v, B), Bτ ). The non-autonomous dynamical system h(X, R+ , π), (Y, R, σ), hi
(h := pr2 : X → Y ) is linear. The following assertion takes place.

Lemma 13.17 [93] Let H(A) be compact in C(R, A), then the non-autonomous
dynamical system h(X,R+ ,π),(Y,R,σ),hi generated by equation (13.119) is com-
pletely continuous, i.e. for every bounded set A ⊂ X there exists a positive number
` such that π ` A is relatively compact.

Theorem 13.42 Let A be τ -periodic. Then the following assertions are equiva-
lent:

(1) The trivial solution of equation (13.119) is uniformly exponentially stable.

(2) lim |U (t, A)u| = 0 for every u ∈ E.
t→+∞
(3) |λ| < 1 for every multiplicator λ of operator of monodromy U (τ, A).

Proof. Let h(X, R+ , π), (Y, R, σ), hi be the linear non-autonomous dynamical
system, generated by equation (13.119). According to Lemma 13.17 this system
is completely continuous and, consequently, there exists a number k ∈ N such that
U k (τ, y0 ) = U (kτ, y0 ) is relatively compact. By virtue of theory of Riesz-Schauder
(see for example [328, p.391-395]) every 0 6= λ ∈ σ(U (τ, A)) is a multiplicator of op-
erator of monodromy U (τ, A). To finish the proof it is sufficient to refer to Theorems
13.36, 13.37 and Remark 13.8. 

Consider the neutral functional differential equation

d
Dut = A(t)ut , (13.121)
dt
where A ∈ C(R, A) and D ∈ A is non-atomic at zero operator [175, p.67]. As well
as in the case of equation (13.119), the equation (13.121) generates a linear non-
autonomous dynamical system h(X, R+ , π), (Y, R, σ), hi, where X = C × Y, Y ; =
H(A) and π := (ϕ, σ). The following statement holds.

Theorem 13.43 Let A ∈ C(R, A) be τ -periodic and D is stable, then the following
assertions are equivalent:

(1) The trivial solution of equation (13.121) is uniformly exponentially stable;

(2) lim |U (t, A)u| = 0 for every u ∈ E;
t→+∞
(3) |λ| < 1 for every multiplier λ of operator of monodromy U (τ, A).
460 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Proof. Let h(X, R+ , π), (Y, R, σ), hi be the linear non-autonomous dynamical
system, generated by equation (13.121). According to Lemma 13.10 this system
is asymptotically compact. According to results of [175, Chapter 12] every 0 6= λ ∈
σ(U (τ, y0 )) is a multiplier of operator of monodromy U (τ, y0 ). To finish the proof
of Theorem 13.43 it is sufficient to refer to Theorems 13.36, 13.37 and Remark 13.8.
The theorem is proved. 

Remark 13.11 The equivalence of conditions 1. and 3. in Theorem 13.41

(Theorem 13.42, Theorem 13.43) was proved in [188, p.219] (resp. in [188, p.233],
[175, p.365]).
 Chapter 14

Triangular maps

Chapter 14 is devoted to the study of quasi-linear triangular maps: chaos, almost

periodic and recurrent solutions, integral manifolds, chaotic sets etc. This problem
is formulated and solved in the framework of non-autonomous dynamical systems
with discrete time. We prove that such systems admit an invariant continuous
section (an invariant manifold). Then, we obtain the conditions of the existence of
a compact global attractor and characterize its structure. We give a criterion for the
existence of almost periodic and recurrent solutions of the quasi-linear triangular
maps. Finally, we prove that quasi-linear maps with chaotic base admits a chaotic
compact invariant set.
This chapter is organized as follows.
In Section 1 we establish the relation between triangular maps and non-
autonomous dynamical systems with discrete time.
In Section 2 we study linear non-autonomous dynamical systems with discrete
time and prove that they admit a unique continuous invariant section – invariant
manifold (Theorem 14.2).
Section 3 is devoted to the study of the existence of invariant sections of quasi-
linear non-autonomous dynamical systems with discrete time (Theorems 14.3 and
14.4).
In Section 4 we prove the existence of compact global attractors of quasi-linear
dynamical systems (Theorems 14.5 and 14.7) and give the description of the struc-
ture of these attractors (Theorems 14.9 and 14.10).
Section 5 is devoted to the study of almost periodic and recurrent solutions of
quasi-linear difference equations (Theorem 14.13 and 14.14).
In section 6 we introduce the notion of pseudo-recurrent solution and prove
that quasi-linear dynamical systems with pseudo-recurrent base admit a pseudo-
recurrent solution (Theorem 14.15).
Section 7 is devoted to the study of chaos in triangular maps and non-
autonomous dynamical systems with discrete time (Theorem 14.17).

461
462 Global Attractors of Non-autonomous Dissipative Dynamical Systems

14.1 Triangular maps and non-autonomous dynamical systems

Let W and Ω be two complete metric spaces and denote by X := W ×Ω its cartesian
product. Recall (see, for example, [214, 222]) that a continuous map F : X → X is
called triangular if there are two continuous maps f : W × Ω → W and g : Ω → Ω
such that F = (f, g), i.e. F (x) = F (u, ω) = (f (u, ω), g(ω)) for all x =: (u, ω) ∈ X.
Consider a system of difference equations
(
un+1 = f (ωn , un )
(14.1)
ωn+1 = g(ωn ),

for all n ∈ Z+ , where Z+ is the set of all non-negative integer numbers.

Along with system (14.1) we consider the family of equations

un+1 = f (g n ω, un ) (ω ∈ Ω), (14.2)

which is equivalent to system (14.1). Let ϕ(n, u, ω) be a solution of equation (14.2)

passing through the point u ∈ W for n = 0. It is easy to verify that the map
ϕ : Z+ × W × Ω → W ((n, u, ω) 7→ ϕ(n, u, ω) ) satisfies the following conditions:

(1) ϕ(0, u, ω) = u for all u ∈ W ;

(2) ϕ(n + m, u, ω) = ϕ(n, ϕ(m, u, ω), σ(m, ω)) for all n, m ∈ Z+ , u ∈ W and ω ∈ Ω,
where σ(n, ω) := g n ω;
(3) the map ϕ : Z+ × W × Ω → W is continuous.

Denote by (Ω, Z+ , σ) the semi-group dynamical system generated by positive

powers of the map g : Ω → Ω, i.e. σ(n, ω) := g n ω for all n ∈ Z+ and ω ∈ Ω.

Definition 14.1 Recall [102] that a triplet hW, ϕ, (Ω, Z+ , σ)i (or briefly ϕ) is
called a cocycle (or non-autonomous dynamical system) over the dynamical system
(Ω, Z+ , σ) with fiber W .

Thus, the reasoning above shows that every triangular map generates a cocycle
and, obviously, vice versa. Taking into consideration this remark we can study
triangular maps in the framework of non-autonomous dynamical systems (cocycles)
with discrete time.

Definition 14.2 A map γ : Z → Ω (respectively α : Z → X, where X := W × Ω)

is called an entire trajectory of the dynamical system (Ω, Z+ , σ) (respectively, of the
skew-product dynamical system (X, Z+ , π), where π := (ϕ, σ) and hW, ϕ, (Ω, Z+ , σ)i
is a cocycle over (Ω, Z+ , σ) with fiber W ) passing through the point ω ∈ Ω (respec-
tively, x := (u, ω) ∈ X), if γ(0) = ω (resp. α(0) = x) and γ(n + m) = σ(m, γ(n))
(resp. α(n + m) = π(m, α(n)) for all n ∈ Z and m ∈ Z+ .
 Triangular maps 463

Denote by Φω the set of all the entire trajectories of the semi-group dynamical
system (Ω, Z+ , σ) passing through the point ω ∈ Ω at the initial moment n = 0 and
S
Φ := {Φω | ω ∈ Ω}.

Definition 14.3 A map ν : Z → W is called an entire trajectory of the cocycle

hW, ϕ, (Ω, Z+ , σ)i passing through the point u ∈ W if there are ω ∈ Ω and γω ∈ Φω
such that the map α : Z → X (X := W × Ω) defined by by the equality α := (ν, γω )
is an entire trajectory of the skew-product dynamical system (X, Z+ , π), i.e. α ∈ Φx
and x := (u, ω).

14.2 Linear non-autonomous dynamical systems

Let Ω be a compact metric space and (Ω, Z+ , σ) be a semi-group dynamical system

on Ω with discrete time.

Definition 14.4 Recall that a subset A ⊆ Ω is called invariant (positively invari-

ant, negatively invariant) if σ n A = A (σ n A ⊆ A, A ⊆ σ n A) for all n ∈ Z+ , where
σ n := σ(n, ·) : Ω → Ω.

Below we will suppose that the set Ω is invariant, i.e. σ n Ω = Ω for all n ∈ Z+ .
Let E be a finite-dimensional Banach space with the norm | · | and W be a complete
metric space. Denote by L(E) the space of all linear continuous operators on E
and by C(Ω, W ) the space of all the continuous functions f : Ω → W endowed with
the compact-open topology, i.e. the uniform convergence on compact subsets in Ω.
The results of this section will be used in the next sections.
Consider a linear equation

un+1 = A(σ n ω)un (ω ∈ Ω, σ n ω := σ(n, ω)) (14.3)

and an in-homogeneous equation

un+1 = A(σ n ω)un + f (σ n ω), (14.4)

where A ∈ C(Ω, L(E)) and f ∈ C(Ω, E).

Definition 14.5 Recall that a linear bounded operator P : E → E is called a

projection, if P 2 = P, where P 2 := P ◦ P .

Definition 14.6 Let U (n, ω) be the operator of Cauchy (a solution operator)

of linear equation (14.3). Following [117] we will say that equation (14.3) has an
exponential dichotomy on Ω, if there exists a continuous projection valued function
P : Ω → L(E) satisfying:

(1) P (σ n ω)U (n, ω) = U (n, ω)P (ω);

464 Global Attractors of Non-autonomous Dissipative Dynamical Systems

(2) UQ (n, ω) is inversible as an operator from ImQ(ω) to ImQ(σ n ω), where

UQ (n, ω) := U (n, ω)Q(ω);
(3) there exist constants 0 < ν < 1 and N > 0 such that

kUP (n, ω)k ≤ N q n and kUQ (n, ω)−1 k ≤ N q n

for all ω ∈ Ω and n ∈ Z+ , where UP (n, ω) := U (n, ω)P (ω).

Let ω ∈ Ω and γω ∈ Φω . Consider a difference equation

un+1 = A(γω (n))un + f (γω (n)), (14.5)

and the corresponding homogeneous linear equation

un+1 = A(γω (n))un (ω ∈ Ω). (14.6)

Let (X, ρ) be a metric space with distance ρ. Denote by C(Z, X) the space of
all the functions f : Z → X equipped with a pointwise topology. This topology can
be metrized. For example, by the equality
+∞
X 1 dn (f1 , d2 )
d(f1 , f2 ) := n 1 + d (f , d )
,
1
2 n 1 2

where dn (f1 , d2 ) := max{ρ(f1 (k), f2 (k)) | k ∈ [−n, n]}, there is defined a distance
on C(Z, X) which generates the pointwise topology.
If x ∈ X and A, B ⊆ X, then denote by ρ(x, A) := inf{ρ(x, a) | a ∈ A} and
β(A, B) := sup{ρ(a, B) | a ∈ A} the semi-distance of Hausdorff.

Theorem 14.1 Suppose that linear equation (14.3) has an exponential dichotomy
on Ω. Then for f ∈ C(Ω, E) the following statements hold:

(1) the set Iω := {u ∈ E | ∃γω ∈ Φω such that equation (14.5) admits a bounded
solution ψω defined on Z with the initial condition ψω (0) = u} is nonempty and
compact;
(2) ϕ(n, Iω , ω) = Iσ(n,ω) for all n ∈ Z+ and ω ∈ Ω, where ϕ(n, u, ω) is a solu-
tion of equation (14.4) with the condition ϕ(0, u, ω) = u and ϕ(n, M, ω) :=
{ϕ(n, u, ω) | u ∈ M };
(3) the map ω → Iω is upper-semicontinuous, i.e.

lim β(Iω , Iω0 ) = 0

ω→ω0

for every ω0 ∈ Ω, where β is the semi-distance of Hausdorff;

S
(4) the set I := {Iω | ω ∈ Ω} is compact.

Proof. Let ω ∈ Ω. Since Ω is compact and invariant, the set Φω 6= ∅. We fix

γω ∈ Φω . Under the conditions of Theorem 14.1 equation (14.6) has an exponential
 Triangular maps 465

dichotomy on Ω with the same constants N and q that in equation (14.3). Then
equation (14.5) admits the unique solution νω : Z → E with the condition

1+q 1+q
kνω k∞ ≤ N kf k∞ ≤ N kf k, (14.7)
1−q 1−q

where kf k := sup{|f (ω)| | ω ∈ Ω} and kνω k∞ := sup{|νω (n)| | n ∈ Z} (see, for

example, [188, 120]). Thus, the set Iω is not empty. From the continuity of the
function ϕ : Z+ × E × Ω → E and inequality (14.7) follows that the set Iω is closed,
bounded and
1+q
|u| ≤ N kf k
1−q

for all u ∈ Iω and ω ∈ Ω.

The second statement of the theorem follows from the equality Sh (Φω ) = Φσ(h,ω)
(h ∈ Z), where Sh γω is an h-translation of the trajectory γω , i.e. Sh γω (n) :=
γω (n + h) for all n ∈ Z.
We will prove now the third affirmation. Let ω0 ∈ Ω, ωk → ω0 , uk ∈ Iωk and
uk → u. To prove our statement it is sufficient to show that u ∈ Iω0 . Since uk ∈ Iωk ,
there is a trajectory γωk ∈ Φωk such that γωk converges to γω0 ∈ Φω0 in C(Z, Ω)
and the equation

un+1 = A(γωk (n))un + f (γωk (n)) (14.8)

has a solution νωk with the initial condition νωk (0) = uk and satisfying inequality
(14.7), i.e.

1+q 1+q
|νωk (n)| ≤ N kf k∞ ≤ N kf k (14.9)
1−q 1−q

for all n ∈ Z and k ∈ N. It is clear that the sequence {νωk (n) } converges for every
n ∈ Z. By Tihonoff’s theorem the sequences {νωk } ⊂ C(Z, E) is relatively compact.
From equality (14.8) and inequality (14.9) follows that every limit point of the
sequence {νωk } is a bounded on Z solution of the equation

un+1 = A(γω0 (n))un + f (γω0 (n)). (14.10)

Taking into account that equation (14.10) admits a unique bounded on Z solution,
we obtain the convergence of the sequence {νωk } in the space C(Z, E). We put
ν0 := lim νωk . It is easy to see that ν0 (0) = u and, consequently, u ∈ Iω0 .
k→+∞
To prove the fourth affirmation it is sufficient to remark that for every ω ∈ Ω
the set Iω is compact, the map F : ω → Iω (F (ω) := Iω ) is upper-semicontinuous
S
and, consequently, the set I := {Iω | ω ∈ Ω} = F (Ω) is compact. The theorem is
completely proved. 
466 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Lemma 14.1 Let ω ∈ Ω and γωl ∈ Φω (l = 1, 2). Under the conditions of Theorem
14.1 we have

|νω1 (n) − νω2 (n)| ≤ N q n |u1 − u2 | (14.11)

for all n ∈ N, where νωl is a bounded on Z solution of the equation

un+1 = A(γωl (n))un + f (γωl (n)) (14.12)

(l = 1, 2).

Proof. Denote ν(n) := νω1 (n) − νω2 (n), then the sequence ν is a bounded on Z+
solution of equation (14.6), because γω1 (n) = γω2 (n) = σ(n, ω) for all n ∈ Z+ .
Since equation (14.3) (and, consequently, equation (14.6) too) has an exponential
dichotomy, we have |ν(n)| ≤ N q n |ν(0)| for all n ∈ Z+ . Taking into consideration
that ν(0) = u1 − u2 , we obtain the required statement. The lemma is proved. 

Lemma 14.2 Under the conditions of Theorem 14.1 for every ω ∈ Ω the set I ω
contains a single point uω .

Proof. Suppose that the statement of the lemma is not true. Then there exists
ω0 ∈ Ω such that Iω0 contains at least two points u1 , u2 (u1 6= u2 ). Let νi (i = 1, 2) be
a bounded on Z solution of equation (14.10) with the condition νωi (0) = ui (i = 1, 2).
According to inequality (14.11), we have

|νω1 (−n) − νω2 (−n)| ≥ N q −n |u1 − u2 | (14.13)

for all n ∈ N. On the other hand, since νωl (l = 1, 2) is a bounded on Z solution of

equation (14.12), we have

sup |νω1 (n) − νω2 (n)| < +∞. (14.14)

n∈Z

Inequalities (14.13) and (14.14) are contradictory. The obtained contradiction

proves our affirmation. The lemma is proved. 

Theorem 14.2 Under the condition of Theorem 14.1 there exists a unique con-
tinuous function ν : Ω → E satisfying the following conditions:

a. the equality

ν(σ(n, ω)) = ϕ(n, ν(ω), ω)

holds for all n ∈ Z+ and ω ∈ Ω, where ϕ(n, u, ω) is the unique solution of

equation (14.4) with the initial condition ϕ(0, u, ω) = u.
 Triangular maps 467

b.
1+q
kνk ≤ N kf k.
1−q

Proof. This affirmations follows directly from Theorem 14.1 and Lemmas 14.1 and
14.2. 

Definition 14.7 A dynamical system (Ω, Z+ , σ) is said to be invertible, if the

map f := σ(1, ·) : Ω → Ω is invertible and, consequently, on Ω there is defined a
group dynamical system (Ω, Z, σ), where π(−n, ·) := (π(n, ·)−1 .

Remark 14.1 In case the dynamical system (Ω, Z+ , σ) is invertible, Theorem

14.2 is well known (see, for example, [122]).

14.3 Quasi-linear non-autonomous dynamical systems

Let us consider the following quasi-linear equation

un+1 = A(σ n ω)un + f (σ n ω) + F (σ n ω, un ), (14.15)

where A ∈ C(Ω, L(E)), f ∈ C(Ω, E) and F ∈ C(Ω × E, E).

1−q
Theorem 14.3 Assume that there exist positive numbers L < L0 := N (1+q) and
r < r0 := ε0 N1−q
(1+q)
(1 − 1+q −1
N L0 1−q ) such that

kF (ω, x1 ) − F (ω, x2 )k ≤ Lkx1 − x2 k

for all ω ∈ Ω and x1 , x2 ∈ B[Q, r] = {x ∈ E | ρ(x, Q) ≤ r}, where Q := ν(Ω),

ν ∈ C(Ω, E) is the unique function from C(Ω, E) figuring in Theorem 14.2 and
ε0 = max kF (ω, ν(ω))k. Then there exists a function w ∈ C(Ω, B[Q, r]) such that
ω∈Ω

w(σ n ω) = ψ(n, w(ω), ω)

for all n ∈ Z+ and ω ∈ Ω, where ψ(·, u, ω) is the unique solution of quasi-linear

equation (14.15) with the initial condition ψ(0, u, ω) = u.

Proof. Let u := v + ν(σ n ω). Then from equation (14.15) we get

vn+1 = A(σ n ω)vn + F (σ n ω, v + ν(σ n ω)).

468 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Let 0 < r < r0 and α ∈ C(Ω, B[Q, r]), where B[Q, r] := {u ∈ E | ρ(u, Q) ≤ r}.
Consider the equation

vn+1 = A(σ n ω)vn + f˜(σ n ω),

where f˜(ω) := F (ω, α(ω) + ν(ω)) for all ω ∈ Ω. According to Theorem 14.2, there
exists a unique function να ∈ C(Ω, E) with properties a. and b. In virtue of
Theorem 14.2, we have
1+q
kνα k ≤ N max kF (ω, α(ω) + ν(ω))k
1 − q ω∈Ω
1+q 1+q
≤N max |F (ω, α(ω) + ν(ω)) − F (ω, γ(ω))| + N max |F (ω, γ(ω))|
1 − q ω∈Ω 1 − q ω∈Ω
1+q 1+q 1+q 2N
≤N Lkαk + N ε0 ≤ N Lr + ε0
1−q 1−q 1−q ν
1+q 1+q
≤N L0 r0 + N ε0 = r 0
1−q 1−q
and, consequently, F(C(Ω, B[Q, r0 ])) ⊆ C(Ω, B[Q, r0 ]), where F : C(Ω, B[Q, r])) →
C(Ω, E) is the map defined by the equality F(α) := να .
Now we will show that the map F : C(Ω, B[Q, r0 ]) → C(Ω, B[Q, r0 ]) is a Lips-
chitzian one. In fact, according to Theorem 14.2 we have
1+q
kF(α1 ) − F(α2 )k ≤ N max |F (ω, α1 (ω) + ν(ω)) − F (ω, α2 (ω) + ν(ω))|
1 − q ω∈Ω
1+q
≤N L max |α1 (ω) − α2 (ω)|.
1 − q ω∈Ω
1+q 1+q
We note that N 1−q L ≤ N 1−q L0 < 1. Thus, the map F is a contraction
and, consequently, by Banach fixed point theorem, there exists a unique function
α ∈ C(Ω, B[Q, r0 ]) such that F(α) = α. To finish the proof of the theorem it is
sufficient to put w := γ + α. 

We now consider a perturbed quasi-linear equation

un+1 = A(σ n ω)un + f (σ n ω) + λF (σ n ω, un ), (14.16)

where λ ∈ [−λ0 , λ0 ] (λ0 > 0) is a small parameter.

Theorem 14.4 Assume that there exist positive numbers r and L such that

|F (ω, u1 ) − F (ω, u2 )| ≤ L|u1 − u2 |

for all ω ∈ Ω and u1 , u2 ∈ B[Q, r]. Then for λ small enough there exists a unique
function νλ ∈ C(Ω, B[Q, r]) such that

νλ (σ n ω) = ψλ (n, νλ (ω), ω)
 Triangular maps 469

for all n ∈ Z+ and ω ∈ Ω, where ψλ (·, u, ω) is the unique solution of equation

(14.16) with the initial condition ψλ (0, u, ω) = u. Moreover,

lim max |νλ (ω) − ν(ω)| = 0, (14.17)

λ→0 ω∈Ω

where ν ∈ C(ω, E) is defined in Theorem 14.2.

Proof. We can prove the existence of νλ by a slight modification of the proof of

Theorem 14.3.
To prove (14.17) we note that

|F (ω, νλ (ω))| ≤ |F (ω, νλ (ω)) − F (ω, ν(ω))| + |F (ω, ν(ω))| ≤ Lr + ε0 .

Denote ∆λ (ω) := νλ (ω) − ν(ω) (ω ∈ Ω). It is easy to see that the sequence
{∆λ (σ n ω)}n∈Z is the unique bounded on Z solution of the equation

zn+1 = A(σ n ω)zn + λF (σ n ω, νλ (σ n (ω))

and, consequently,
1+q
|νλ (σ n ω) − ν(σ n ω)| = |∆(σ n ω)|N (14.18)
1−q
1+q
× sup |λF (σ n ω, νλ (σ n ω)| ≤ |λ|N (Lr + ε0 )
n∈Z 1−q
for all ω ∈ Ω, n ∈ Z and λ ∈ [−λ0 , λ0 ]. Thus, from inequality (14.18) we obtain
1+q
|νλ (ω) − ν(ω)| ≤ |λ|N (Lr + ε0 ) (14.19)
1−q
for all ω ∈ Ω and λ ∈ [−λ0 , λ0 ]. Passing to limit in inequality (14.19) as λ → 0, we
obtain (14.17). 

14.4 Global attractors of quasi-linear triangular systems

Consider a difference equation

un+1 = f (σ n ω, un ) (ω ∈ Ω). (14.20)

Denote by ϕ(n, u, ω) a unique solution of equation (14.20) with the initial condition
ϕ(0, u, ω) = u.

Definition 14.8 Equation (14.20) is said to be dissipative, if there exists a posi-

tive number r such that

lim sup |ϕ(n, u, ω)| ≤ r

n→+∞

for all u ∈ E and ω ∈ Ω.

470 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Consider a quasi-linear equation

un+1 = A(σ n ω)uk + F (σ n ω, uk ), (14.21)

where A ∈ C(Ω, [E]) and the function F ∈ C(Ω × E, E) satisfies to ”the condition
of smallness”.
Denote by U (k, ω) the Cauchy’s matrix for the linear equation

un+1 = A(σ n ω)uk .

Theorem 14.5 Suppose that the following conditions hold:

(1) there are positive numbers N and q < 1 such that

kU (n, ω)k ≤ M q n (n ∈ Z+ ); (14.22)

(2) |F (ω, u)| ≤ C + D|u| (C ≥ 0, 0 ≤ D < (1 − q)N −1 ) for all u ∈ E and ω ∈ Ω.

Then equation (14.21) is dissipative.

Proof. Let ϕ(·, u, ω) be the solution of equation (14.21) passing through the point
u ∈ E for n = 0. According to the formula of the variation of constants (see, for
example, [170] and [188]) we have
n−1
X
ϕ(n, u, ω) = U (k, ω)u + U (n − m − 1, ω)F (σ m ω, ϕ(m, u, ω)),
m=0

and, consequently,
n−1
X
|ϕ(n, u, ω)| ≤ N q n |u| + q n−m−1 (C + D|ϕ(m, u, ω)|). (14.23)
m=0

We set u(n) := q −n |ϕ(n, u, ω)| and, taking into account (14.23), obtain
n−1
X n−1
X
u(n) ≤ N |u| + CN q −1 q −m + DN q −1 u(m). (14.24)
m=0 m=0

Denote the right hand side of inequality (14.24) by v(n). Note, that
CN DN DN CN −n
v(n + 1) − v(n) = q −n + u(n) ≤ v(n) + q ,
q q q q
and,hence,
 DN  CN −n
v(n + 1) ≤ 1 + v(n) + q .
q q
From this inequality we obtain
 DN n−1 CN 1 − q n−1
v(n) ≤ 1 + v(1) + .
q q 1−q
 Triangular maps 471

Therefore,
CN n−1
|ϕ(n, u, ω)| ≤ (q + DN )n−1 qN |u| + (q − 1), (14.25)
q−1
because v(1) = N |u|. From (14.25) follows that
CN
lim sup |ϕ(n, u, ω)| ≤
k→+∞ 1−q
for all u ∈ E and ω ∈ Ω. The theorem is proved. 

Let hE, ϕ, (Ω, Z+ , σ)i be a cocycle over (Ω, Z+ , σ) with the fiber E.

Definition 14.9 Recall that a family {Iω | ω ∈ Ω}(Iω ⊂ E) of nonempty com-

pact subsets is called a compact global attractor of the cocycle ϕ, if the following
conditions are fulfilled:
S
(1) the set {Iω | ω ∈ Ω} is relatively compact;
(2) the family I := {Iω | ω ∈ Ω} is invariant with respect to the cocycle ϕ, i.e.
ϕ(n, Iω , ω) = Iσn ω for all n ∈ Z+ and ω ∈ Ω;
(3) the equality

lim sup β(ϕ(n, K, ω), I) = 0

n→+∞ ω∈Ω

takes place for every K ∈ C(E), where C(E) is a family of all compacts from
E.

Definition 14.10 A cocycle hE, ϕ, (Ω, T, σ)i is said to be dissipative, if there

exists a positive number r such that

lim sup |ϕ(n, u, ω)| ≤ r

n→+∞

for all u ∈ E and ω ∈ Ω.

Theorem 14.6 ([102]) Every dissipative cocycle hE, ϕ, (Ω, T, σ)i admits a com-
pact global attractor {Iω | ω ∈ Ω} such that:

(1) the component Iω (ω ∈ Ω) is connected;

S
(2) the set I = {Iω | ω ∈ Ω} is connected, if the space Ω also is.

Theorem 14.7 Under the conditions of Theorem 14.5, equation (14.21) (i.e. the
cocycle ϕ generated by equation (14.21)) admits a compact global attractor {I ω | ω ∈
Ω} with the following properties:

(1) the component Iω (ω ∈ Ω) is connected;

S
(2) the set I = {Iω | ω ∈ Ω} is connected, if the space Ω also is.
472 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Proof. This statement follows directly from Theorems 14.5 and 14.6. 

Definition 14.11 Recall that a cocycle hE, ϕ, (Ω, T, σ)i is said to be convergent,
if it admits a compact global attractor {Iω | ω ∈ Ω} such that every component Iω
(ω ∈ Ω) contains a single point uω , i.e. Iω = {uω } (ω ∈ Ω).

Theorem 14.8 [102, Ch.2] Let hE, ϕ, (Ω, T, σ)i be a cocycle admitting a compact
global attractor I = {Iω | ω ∈ Ω}. If

lim max |ϕ(n, u1 , ω) − ϕ(n, u2 , ω)| = 0 (14.26)

n→+∞ |u1 |,|u2 |≤r,ω∈Ω

for every r > 0, then the set Iω contains a single point for every ω ∈ Ω.

Theorem 14.9 Let A ∈ C(Ω, [E]) and F ∈ C(Ω × E, E) and the following con-
ditions be fulfilled:

(1) there exist positive numbers N and q < 1 such that inequality (14.22) holds;
(2) |F (ω, u1 ) − F (ω, u2 )| ≤ L|u1 − u2 | (0 ≤ L < N −1 (1 − q)) for all ω ∈ Ω and
u1 , u2 ∈ E.

Then equation (14.21) is convergent, i.e. the cocycle generated by this equation
is convergent.

Proof. First step. We will prove that under the conditions of Theorem 14.9 equation
(14.21) admits a compact global attractor I = {Iω | ω ∈ Ω}. In fact,

|F (ω, u)| ≤ |F (ω, 0)| + L|u| ≤ C + L|u|

for all u ∈ E, where C := max{|F (ω, 0)| | ω ∈ Ω}. According to Theorems 14.5 and
14.6, the equation (14.21) admits a compact global attractor I = {I ω | ω ∈ Ω}.
Second step. Let ϕ be the cocycle generated by equation (14.21). In virtue of
the formula of the variation of constants, we have
n−1
X
ϕ(n, u, ω) = U (n, ω)u + U (n − m − 1, ω)F (σ m ω, ϕ(m, u, ω)).
m=0

Consequently,

ϕ(n, u1 , ω) − ϕ(n, u2 , ω) = U (n, ω)(u1 − u2 )+

n−1
X
U (n − m − 1, A)[F (σ m ω, ϕ(m, u, ω)) − F (σ m ω, ϕ(m, u2 , ω)].
m=1
 Triangular maps 473

Thus,

|ϕ(n, u1 , ω) − ϕ(n, u2 , ω)| ≤ N q n (|u1 − u2 |

n−1
X
+Lq −1 q −m |ϕ(m, u1 , ω) − ϕ(m, u2 , ω)|). (14.27)
m=0

Let u(n) := |ϕ(n, u1 , ω) − ϕ(n, u2 , ω)|q −n . From (14.27) follows that

 n−1
X 
u(n) ≤ N |u1 − u2 | + Lq −1 u(m) . (14.28)
m=0

Denote by v(n) the right hand side of (14.28). The following inequality

v(n + 1) − v(n) = LN q −1 u(n) ≤ LN q −1 v(n). (14.29)

holds. From (14.29) we obtain

v(n) ≤ (1 + LN q −1 )n−1 v(1)

and, since v(1) = N |u1 − u2 |, we get

u(n) ≤ (1 + LN q −1 )N |u1 − u2 |. (14.30)

From (14.30) we have

|ϕ(n, u1 , ω) − ϕ(n, u2 , ω)| ≤ (q + LM )n−1 qN |u1 − u2 |

for all u1 , u2 ∈ E and ω ∈ Ω.

To finish the proof of Theorem 14.9 it is sufficient to refer to Theorem 14.8. The
theorem is proved. 

Remark 14.2 Simple examples show that under the conditions of Theorem 14.7
the compact global attractor {Iω | ω ∈ Ω}, generally speaking, is not trivial, i.e. the
component set Iω contains more than one point. This statement can be illustrated
2un
by the following example: un+ = 12 un + 1+u 2 .
n

Theorem 14.10 Let A ∈ C(Ω, [E]), F ∈ C(Ω×E, E) and the following conditions
be fulfilled:

(1) there are positive numbers N and q < 1 such that

kU (n, ω)k ≤ N q n (n ≥ 0, ω ∈ Ω);

1−q
(2) |F (ω, u)| ≤ M + ε|u| ( u ∈ E, ω ∈ Ω ) and 0 ≤ ε ≤ ε0 < N (1+q) ;
1+q
(3) the restriction F0 of the function F on Ω × B[0, r0 ], where r0 := N M 1−q (1 −
1+q −1
ε0 N 1−q ) , satisfies the condition of Lipschitz with the Lipschitz’s constant
Lip(F0 ) < N1−q
(1+q) .
474 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Then the following statements hold:

a. equation (14.21) admits a compact global attractor I = {I ω | ω ∈ Ω};

b. there exists a continuous function ν : Ω → E satisfying the following conditions:
b1. the equality

ν(σ(n, ω)) = ϕ(n, ν(ω), ω)

holds for all n ∈ Z+ and ω ∈ Ω, where ϕ(n, u, ω) is the unique solution of

equation (14.21) with the initial condition ϕ(0, u, ω) = u.
b2.

ν(ω) ∈ Iω for all ω ∈ Ω.

Proof. The first statement follows directly from Theorems 14.5 and 14.7.
We will prove now the second affirmation. Define an operator F :
C(Ω, B[0, r0 ]) → C(Ω, E) in the following way. Let α ∈ C(Ω, B[0, r0 ]), then under
the conditions of Theorem 14.10 we have f˜(·) := F (·, α(·)) ∈ C(Ω, B[0, r0 ]). Accord-
ing to Theorem 14.2, there exists a unique function να ∈ C(Ω, E) with properties
a. and b. In virtue of Theorem 14.2, we have
1+q
kνα k ≤ N max |F (ω, α(ω))|
1 − q ω∈Ω
1+q 1+q
≤N (M + ε0 kαk) ≤ N (M + ε0 r0 ) ≤ r0
1−q 1−q

and, consequently, F(C(Ω, B[Q, r0 ])) ⊆ C(Ω, B[Q, r0 ]), where F : C(Ω, B[Q, r])) →
C(Ω, E) is the map defined by the equality F(α) := να .
Now we will show that the map F : C(Ω, B[Q, r0 ]) → C(Ω, B[Q, r0 ]) is Lips-
chitzian. In fact, according to Theorem 14.2 we have
1+q
kF(α1 ) − F(α2 )k ≤ N max |F (ω, α1 (ω)) − F (ω, α2 (ω)|
1 − q ω∈Ω
1+q
≤N L max |α1 (ω) − α2 (ω)|.
1 − q ω∈Ω
1+q 1+q
We note that N 1−q L ≤ N 1−q L0 < 1. Thus, the map F is a contraction and,
consequently, in virtue of the Banach’s fixed point theorem there exists a unique
function w ∈ C(Ω, B[Q, r0 ]) such that F(w) = w. It is easy to see that w satisfies
the first condition. To finish the proof of the theorem it is sufficient to remark that
S
the set A := {Aω | ω ∈ Ω}, where Aω := {w(ω)} × {ω}, is a compact invariant
set in the skew-product dynamical system (X, Z+ , π) (X := E × Ω and π := (ϕ, σ))
S
and J := {Jω | ω ∈ Ω} is a maximal compact invariant set of this system. Thus,
we have A ⊂ J and, consequently, w(ω) ∈ Iω for all ω ∈ Ω. 
 Triangular maps 475

Remark 14.3 Under the conditions of Theorem 14.10, dissipativity does not re-
duce to convergence.

We will give an example which confirms this statement.

Example 14.1 Let k ∈ ( 12 , 1). Consider a scalar equation

un+1 = kun + αF (un ) , (14.31)

where
(
x2
2 for |x| ≤ 10
F (x) :=
50 + 10[1 − exp(10 − |x|)] for |x| > 10

for k < 1 < k + 5α. We can take r0 := 2(1−k) α , then it is easy to check that for
equation (14.31) all the conditions of Theorem 14.10 are fulfilled for chosen α, k,
and F . In addition, its Levinson’s center (compact global attractor) contains at
least two fixed points (in fact, there are 3 of them) and, consequently, equation
(14.31) is not convergent.

14.5 Almost periodic and recurrent solutions

Let (X, Z+ , π) be a dynamical system, x ∈ X, m ∈ Z+ , m > 0, ε > 0.

Denote Mx = {{tn } | {xtn } is convergent}.

Theorem 14.11 ([300], [302]) Let (X, Z+ , π) and (Y, Z+ , σ) be two dynamical
systems. Assume that h : X → Y is a homomorphism of (X, Z+ , π) onto (Y, Z+ , σ).
If a point x ∈ X is stationary (m-periodic, almost periodic, recurrent), then the point
y := h(x) is also stationary (m-periodic, almost periodic, recurrent) and M x ⊂ My .

Definition 14.12 A solution ϕ(n, u, ω) of equation (14.20) is said to be sta-

tionary (m-periodic, almost periodic, recurrent), if the point x := (u, ω) ∈ X :=
E × Ω is a stationary (m-periodic, almost periodic, recurrent) point of the skew-
product dynamical system (X, Z+ , π), where π := (ϕ, σ), i.e. π(n, (u, ω)) :=
(ϕ(n, u, ω), σ(n, ω)) for all n ∈ Z+ and (u, ω) ∈ E × Ω.

Lemma 14.3 Suppose that u ∈ C(Ω, E) satisfies the condition

u(σ(n, ω)) = ϕ(n, u(ω), ω) (14.32)

for all n ∈ Z+ and ω ∈ Ω. Then the map h : Ω → X, defined by

h(ω) := (u(ω), ω) (14.33)

for all ω ∈ Ω, is a homomorphism of (Ω, Z+ , σ) onto (X, Z+ , π).

476 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Proof. This assertion follows from equalities (14.32) and (14.33) 

Remark 14.4 A function u ∈ C(Ω, E) with property (14.32) is called a con-

tinuous invariant section (or an integral manifold) of non-autonomous difference
equation (14.20).

Theorem 14.12 If a function u ∈ C(Ω, E) satisfies condition (14.32) and a

point ω ∈ Ω is stationary (m-periodic, almost periodic, recurrent), then the solution
ϕ(n, u(ω), ω) of equation (14.20) also is stationary (m-periodic, almost periodic,
recurrent).

Proof. This statement follows from Theorem 14.11 and Lemma 14.3. 
Using Theorem 14.12 and the results from sections 3 and 4we obtain some crite-
rions of the existence of periodic (almost periodic, recurrent) solutions of equation
(14.15). For example, the following statements hold.

Theorem 14.13 Let (Ω, Z+ , π) be a dynamical system and Ω consists of m-

periodic (almost periodic, recurrent) points. Then under the conditions of Theorem
14.3 the equation (14.15) admits an invariant integral manifold consisting from m-
periodic (almost periodic, recurrent) solutions.

Theorem 14.14 Let (Ω, Z+ , π) be a dynamical system and Ω consists of m-

periodic (almost periodic, recurrent) points. Under the conditions of Theorem 14.4
for a sufficiently small λ the equation (14.16) admits a unique invariant mani-
fold νλ consisting from m-periodic (almost periodic, recurrent) solutions of equation
(14.16).

Example 14.2 Consider the equation

un+1 = f (n, un ) (14.34)

where f ∈ C(Z+ × E, E); here C(Z+ × E, E) is the space of all continuous functions
Z+ × E → E) equipped with a compact-open topology. This topology can be
metrized. For example, by the equality
+∞
X 1 dn (f1 , d2 )
d(f1 , f2 ) := n 1 + d (f , d )
,
1
2 n 1 2

where dn (f1 , d2 ) := max{ρ(f1 (k, u), f2 (k, u)) | k ∈ [0, n], |u| ≤ n}, there is defined
a distance on C(Z+ × E, E) which generates the topology of pointwise convergence
with respect to n ∈ Z+ uniformly with respect to u on every compact from E.
Along with equation (14.34), we will consider the H-class of equation (14.34)

vn+1 = g(n, vn ) (g ∈ H(f )), (14.35)

 Triangular maps 477

where H(f ) = {fm | m ∈ Z+ } and the over bar denotes the closure in C(Z+ ×E, E),
and fm (n, u) = f (n + m, u) for all n ∈ Z+ and u ∈ E. Denote by (C(Z+ ×
E, E), Z+ , σ) the dynamical system of translations. Here σ(m, g) := gm for all
m ∈ Z+ and g ∈ C(Z+ × E, E).
Let Ω be the hull H(f ) of a given function f ∈ C(Z+ × E, E) and denote the
restriction of (C(Z+ × E, E), R, σ) on Ω by (Ω, Z+ , σ). Let F : Ω × E → E be a
continuous map defined by F (g, u) = g(0, u) for g ∈ Ω and u ∈ E. Then equation
(14.35) can be rewritten in such form:

un+1 = F (σ n ω, un ),

where ω := g and σ n ω := gn .

Definition 14.13 The function f ∈ C(Z+ × E, E) is said to be periodic (almost

periodic, recurrent), if f ∈ C(Z+ × E, E) is a periodic (almost periodic, recurrent)
point of the dynamical system of translations (C(Z+ × E, E), Z+ , σ).

If the function f ∈ C(Z+ × E, E) is periodic (almost periodic, recurrent), then

the set Ω := H(f ) is the compact minimal set of the dynamical system (C(Z+ ×
E, E), Z+ , σ) consisting from periodic (almost periodic, recurrent) points.

14.6 Pseudo recurrent solutions

Definition 14.14 Recall that an autonomous dynamical system (Ω, Z+ , σ) is said

to be pseudo recurrent, if the following conditions are fulfilled:

a) Ω is compact;
b) (Ω, Z+ , σ) is transitive, i.e. there exists a point ω0 ∈ Ω such that Ω =
{σ n ω0 | n ∈ Z+ };
c) every point ω ∈ Ω is stable in the sense of Poisson, i.e.

Nω = {{nk } | σ nk ω → ω and nk → +∞} 6= ∅.

Lemma 14.4 Let h(X, Z+ , π), (Ω, Z+ , σ), hi be a non-autonomous dynamical

system and the following conditions be fulfilled:

1) (Ω, Z+ , σ) is pseudo recurrent;

2) γ ∈ C(Ω, X) is an invariant section of the homomorphism h : X → Ω.

Then the autonomous dynamical system (γ(Ω), Z+ , π) is pseudo recurrent too.

Proof. It is evident that the space γ(Ω) is compact, because Ω is compact and
γ ∈ C(Ω, X). We note that on the space γ(Ω) we have the dynamical system
(γ(Ω), Z+ , π
b) defined by the homomorphism γ : Ω → γ(Ω). Namely, π b n γ(ω) :=
478 Global Attractors of Non-autonomous Dissipative Dynamical Systems

γ(σ n ω) for all n ∈ Z+ and ω ∈ Ω. Hence, πbn γ(ω) = π n γ(ω) for all n ∈ Z+ and
ω ∈ Ω. Now we will show that γ(Ω) = {π n γ(ω0 ) | n ∈ Z+ }. Really, let x ∈ γ(Ω).
Then there exists a unique point ω ∈ Ω such that x = γ(ω). Let {nk } ⊂ Z+ be a
sequence such that σ nk ω0 → ω. Then x = γ(ω) = lim γ(σ nk ω0 ) = lim π nk γ(ω)
k→+∞ k→+∞
and, consequently, γ(Ω) ⊂ {π n γ(ω0 ) | n ∈ Z+ }. The inverse inclusion is trivial.
Thus, γ(Ω) = {π n γ(ω0 ) | n ∈ Z+ }. To finish the proof of the lemma it is sufficient
to note that Nω ⊆ Nγ(ω) for every point ω ∈ Ω and, consequently, every point γ(ω)
is stable in the sense of Poisson. The lemma is proved. 

Lemma 14.4 implies that under the conditions of Theorem 14.2 (respectively of
Th. 14.3, 14.5 or Th. 14.9) equation (14.32) or equation (14.15)) admits a pseudo
recurrent integral manifold.
Therefore, we have the following result.

Theorem 14.15 Assume the driving dynamical system (Ω, Z+ , σ) is pseudo re-
current, and assume the conditions of Theorem 14.2 (respectively of Th. 14.3, 14.5
or Th. 14.9) are satisfied. Then equation (14.32) (respectively, equation (14.34) or
equation (14.15)) admits a pseudo recurrent integral manifold.

14.7 Chaos in triangular maps

Let (X, ρ) be a metric space and (X, Z+ , π) be a discrete dynamical system gener-
ated by positive powers of the map f : X → X, i.e. π(n, x) = f n x for all x ∈ X
and n ∈ Z+ , where f n := f n−1 ◦ f.

Definition 14.15 A subset M ⊆ X is called transitive, if there exists a point

x0 ∈ X such that H(x0 ) := {π(n, x0 ) | n ∈ Z+ } = M.

Definition 14.16 {p, q} ⊆ X is called a Li-Yorke pair, if simultaneously

lim inf ρ(π(n, p), π(n, q)) = 0 and lim sup ρ(π(n, p), π(n, q)) > 0.
n→+∞ n→+∞

Definition 14.17 A set M ⊆ X is called scrambled, if any pair of distinct points

{p, q} ⊆ M is a Li-Yorke pair.

Definition 14.18 A dynamical system (X, Z+ , π) is said to be chaotic, if X

contains an uncountable subset M satisfying the following conditions:

(1) the set M is transitive;

(2) M is scrambled;
(3) P (M ) = M, where P (M ) := {x ∈ M | Nx 6= ∅} (i.e. x ∈ P (M ), if x is stable
in the sense of Poisson) and by bar we denote the closure in X.
 Bibliography 479

Theorem 14.16 Let (X, Z+ , π) and (Ω, Z+ , σ) be two dynamical systems and ν :
X → Ω be a homeomorphism of (Ω, Z+ , σ) onto (X, Z+ , π). Assume that (Ω, Z+ , σ)
is chaotic. Then the dynamical system (X, Z+ , π) is chaotic too.

Proof. Let (Ω, Z+ , σ) be chaotic and ν : X → Ω be a homeomorphism of (Ω, Z+ , σ)

onto (X, Z+ , π). Then there exists a subset A ⊆ Ω satisfying all the conditions of
Definition 14.17. Denote M := ν(A). Then the set M is transitive and P (M ) = M
(see the proof of Lemma 14.4). Note that the set M := ν(A) is uncountable,
because the homeomorphism ν : Ω → X transforms different points to different
points. Finally, to finish the proof of Theorem 14.16 it is sufficient to note that the
homeomorphism ν : Ω → X maps a Li-Yorke pair in (Ω, Z+ , σ) to a Li-Yorke pair
in (X, Z+ , π). The theorem is proved. 
Using Theorem 14.16 and the results from sections 3 and 4 we will obtain some
criterions of the existence of chaotic sets for triangular maps. For example the
following statements are held.

Theorem 14.17 Let (Ω, Z+ , π) be a chaotic dynamical system. Then under the
conditions of Theorem 14.3 equation (14.15) (respectively triangular map) admits a
compact invariant chaotic set.
480 Global Attractors of Non-autonomous Dissipative Dynamical Systems
 Bibliography

[1] Abergel F., Existence and Finite Dimensionality of the Global Attracor for
Evolutin Equations and Unbounded Domains. Journal of Differential Equa-
tions, 83(11):85–108, 1990.
[2] Alekseev V. M., Symbolic Dynamics. Naukova Dumka, Kiev, 1986. The 11th
Mathematical School.
[3] Alekseev V. M. and Yakobson M. V., Symbolic Dinamics and Hyperbolic Dy-
namic Sistems. In R. Bowen. Methods of Symbolic Dynamics, 196–240. M.,
1979.
[4] Andreev A. S., On Asymptotical Stability and Unstability of Zero Solution of
Nonautonomous Systems. Prikladnaya Matematika i Mehanika., 48(2):1225–
232, 1984.
[5] Ararktsyan B. G., Asymptotic Almost Periodic Solutions of certain Linear
Evolutionary Equations. Matematicheskii Sbornik, 133(175), 1(5):3–10, 1988.
[6] Arnold L., Random Dynamical Systems. Springer-Verlag, 1998.
[7] Arnold V. I., Additional Chapters of Theory of Ordinary Differential Equa-
tions. Nauka, Moscow, 1978.
[8] Artstein Z., Uniform Asymptotic Stability via the Limiting Equations. Journal
of Differential Equations, 27(2):172–189, 1978.
[9] Atrashenok P. V., Some Questions of Stability of Motion. Vestnik LGU,
Matematika., (8):79–106, 1954.
[10] Aulbach B. and Garay B. M., Discretization of Semilinear Equations with an
Exponential Dichotomy. Computers Math. Applic. 28 (1994), 23–35.
[11] Auslander J. and Seifert P., Prolongation Stability in Dynamical Systems.
Ann. Inst. Fourier, Grenoble, 14(2):237–268, 1964.
[12] Babin A. V. and Vishik M. I., Attractors of Evolutionary Equations with
Partial Derivatives and Estimation of Their Dimensionality. Uspekhi Matem-
aticheskih Nauk, 38(4(232)):133-187, 1983.
[13] Babin A. V. and Vishik M. I., Upper and Lower Estimation of Dimensional-
ity of Attractors of Evolutionary Partial Equations. Sibirskii Matematicheskii

481
482 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Zhurnal., 24(5):15–30, 1983.

[14] Babin A. V. and Vishik M. I., Attractors of Parabolic Equations and Navier–
Stocks Systems and Estimation of Their Dimensionality. In collection. Ob-
schaya Teoriya Granichnyh Zadach. Kiev, 1983. pages 14–25.
[15] Babin A. V. and Vishik M. I., Regular Attractor of Semigroups and Evolu-
tional Equations. Journal Math. Pures et Appl., 62:172–189, 1983.
[16] Babin A. V. and Vishik M. I., Unstable Invariant Sets of Semi-groups of
Nonlinear Operators and Their Perturbations. Uspekhi Matematicheskih Nauk,
41(4(250)):3–34, 1986.
[17] Babin A. V. and Vishik M. I., On Unstable Sets of Evolutionary Equations
in Neighbourhood of Critical Points of Stationary Curve. Izvestiya AN SSSR,
Matematika., 51(1):44–78, 1987.
[18] Babin A. V. and Vishik M. I., Semi-Groups Depending on Parameter, their
Attractors and Asymptotical Behaviour. In collection Globalnyi Analiz i Ne-
lineinye Uravneniya. Iz-vo VGU, 1988.
[19] Babin A. V. and Vishik M. I., Attractors of Parabolic and Hyperbolic Equa-
tions, Character of their Compactness and Attractions to Them. Vestnik
MGU, Series I, Matematika, Mehanika, (3):71–73, 1988.
[20] Babin A. V. and Vishik M. I., Attractors of Evolutionary Equations. Nauka,
Moscow, 1989; English translation, North–Holland, Amsterdam, 1992.
[21] Barbashin E. A., Introduction in Theory of Stability. Nauka, Moscow, 1967.
[22] Batty C. J. K., Hutter W. and Räbiger F., Almost Periodicity of Mild So-
lutions of Inhomogeneous Periodic Cauchy Problems. Journal of Differential
Equations, 156(1999), no.2, pp.309–327
[23] Bellman R., Introduction to Matrix Analysis. McGraw-Hill, New York-
Toronto-London 1960
[24] Bhatia N. P. and Szegö G. P., Local Semi-Dynamical Systems. Lecture Notes
in Mathematics, v.80. Springer, Berlin–Heidelberg–New York, 1969.
[25] Bhatia N. P. and Szegö G. P., Stability Theory of Dynamical Systems. Lecture
Notes in Mathematics. Springer, Berlin–Heidelberg–New York, 1970.
[26] Billotti J. E. and LaSalle J. P., Dissipative Periodic Process. Bull. Amer.
Math. Soc., 77(6):1082–1088, 1971.
[27] Bogolyubov N. N. and Mitropolsky Yu. A., Asymptotic Methods in the Theory
of Non-Linear Oscilations. Fizmatgiz, Moscow 1963; English transl., Gordon
and Dreach, New York 1962.
[28] Bourbaki N., Espaces Vectoriels Topologiques. Hermann, Paris, 1955.
[29] Bourbaki N., Variétés Différentielles et Analitiques (Fascicule de résultats).
Herman, Paris, 1971.
[30] Brezis H., Operateurs Maximaux Monotones et Semigroupes de Contractions
dans les Espaces de Hilbert. Math.Studies, v.5. North Holland, 1973.
[31] Bronstein I. U. and Gerko A. I., On Inclusion of Some Topological Semi-
 Bibliography 483

groups of Transformations in Topological Groups of Transformations. Izvestiya

AN Moldavskoi SSR, seriya fiz.-tehn. i matem. nauk, 3:18-24, 1970.
[32] Bronsteyn I. U., Extensions of Minimal Transformation Group. Noordhoff,
1979.
[33] Bronstein I. U., Nonautonomous Dynamical Systems. Stiintsa, Chişinău, 1984.
[34] Bronstein I. U. and Kopanskii A. Ya., Smooth Invariant Manifolds and Normal
Forms. World Scientific, Rive Edge, NJ, 1994.
[35] Browder F. E., Some New Asymptotic Fixed Point Theorems. Proc. Acad.
Sci., 71:2734–2735, 1974.
[36] Buşe C., Asymptotic Stability and Perron Condition for Periodic Evolution
Families on the Half Line. Preprint in Evolution Equations and Semigroups
(http://math.lsu.edu/tiger/evolve/archiv/98-00013ps)
[37] Buşe C. and Giurgiulescu M., A New Proof for a Barbashin’s Theorem
in the Periodic Case. Preprint in Evolution Equations and Semigroups
(http://math.lsu.edu/tiger/evolve/archiv/99-00099ps)
[38] Buşe C., An Exponential Stability Criterion for q-Periodic Evolution Fam-
ilies on Hilbert Spaces. Preprint in Evolution Equations and Semigroups
(http://math.lsu.edu/tiger/evolve/archiv/00-00001ps)
[39] Cameron R. H., Almost Periodic Properties of Bounded Solutions of Linear
Differential Equation with Almost Periodic Coefficients. J. Math. Phys, 15:73–
81, 1936.
[40] Caraballo T., Langa J. A. and Robinson J., Upper Semicontinuity of Attrac-
tors for Small Random Perturbations of Dynamical Systems. Comm. in Partial
Differential Equations, 23:1557-1581, 1998.
[41] Caraballo T. and Langa J. A., On the Upper Semicontinuity of Cocycle
Attractors for Nonautonomous and Random Dynamical Systems. Dynamics
of Continuous, Discrete and Impulsive Systems, 10(2003), no. 4, pp.491-514.
[42] Cartan H., Elementary Theory of Analytic Functions of one and Several Com-
plex Variables [Russian translation]. Mir, Moscow, 1963.
[43] Carvalho A. N., Cholewa J. W. and Dlotko T., Global Attractors for
Problems with Monotone Operators. Boll. Unione Mat.Ital. Sez.B Ar-
tic.Ric.Mat.(8)2(1999), no.3, pp.693-706.
[44] Cartwrigth M. L. and Littlewood J. E., On Non-Linear Differential Equations
of the Second Order, I. The Equation ÿ − k(1 − y 2 )ẏ + y = bµk cos(µt + α)k
for Large k. J. Math. Soc., 20:180–189, 1945.
[45] Cartwrigth M. L. and Littlewood J. E., On Non-Linear Differential Equations
of the Second Order, II. The Equation ÿ − k(1 − y 2 )ẏ + y = bµk cos(µt + α)k
for Large k and its Generalizations. Acta Math., 97:3–4, 1957.
[46] Cartwrigth M. L. and Littlewood J. E., On Non-Linear Differential Equations
of the Second Order, III. The General Equation. Acta Math., 98:1–2, 1957.
[47] Cerons S. S. and Lopes O., α-Contractions and Attractors for Dissipative
484 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Semilinear Hiperbolic Equations and Systems. Annali di Matematica Pura ad

Applicata, CLX(4):193–206, 1991.
[48] Cheban D. N., Poisson Asymptotic Stability of the Solutions of Operational
Equations. Differentsial’nye Uravneniya, 13(8):978–983, 1977.
[49] Cheban D. N., Comparability of the Points of Dynamical Systems by Char-
acter of Recurrence in the Limit. Matematicheskie Nauki, Ştiinţa, Chişinău,
1:66-71, 1977.
[50] Cheban D. N., Linear Differential Equations Satisfying the Condition of
Favard. In Proceedings of the 5th All-Union Conference on the Qualitative
Theory of Differential Equations, pp.181–182, Chişinău, 1979.
[51] Cheban D. N., Theory of Linear Differential Equations (Selected Topics).
Ştiincta, Chişinău, 1980.
[52] Cheban D. N., Stability of the Levinson Center of Nonautonomous Dissipative
Dynamical Systems. Differentsial’nye Uravneniya, 20(11):2016–2018, 1984.
[53] Cheban D. N., Nonautonomous Dissipative Dynamical Systems. Differ-
entsial’nye Uravneniya, 21(5):913–914, 1985.
[54] Cheban D. N., C-Analytic Dissipative Dynamical Systems. In Proceedings of
the 3rd International Conference of Differential Equations and Applications,
p.141, Roose (Bulgaria), 1985.
[55] Cheban D. N., Dissipative Dynamical Systems. In Proceedings of the Extended
Sessions of Seminar of the I. N. Vekua Institute of Applied Mathematics, v.
1, no. 3, pp.154–157, Tbilisi, 1985.
[56] Cheban D. N., Lyapunov’s Second Method in the Theory of Stability of
Dynamical Systems. In Mat. Issled. No. 80, Differentsial’nye Uravneniya i
Dynamicheskie Sistemy, pp.139–147. Ştiinţa, Chişinău, 1985.
[57] Cheban D. N., Quasiperiodic Solutions of Dissipative Systems with Quasiperi-
odic Coefficients. Differentsial’nye Uravneniya, 22(2):267–278, 1986.
[58] Cheban D. N., Nonautonomous Dissipative Dynamical Systems. Dokl. Akad.
Nauk SSSR, 286(1986), pp.136-143. Translation in Soviet Math. Dokl,
33(1):207–210, 1986.
[59] Cheban D. N., A criterion for Convergence of Nonlinear Systems in Hilbert
Space. In Mat. Issled., No. 88, Differentsial’nye Uravneniya i ih Invarianty,
pp. 136–143. Ştiinţa, Chişinău, 1986.
[60] Cheban D. N., A criterion for Convergence of Nonlinear Systems with Respect
to First Approximation. In Mat. Issled., No. 88, Differentsial’nye Uravneniya
i ih Invarianty, pp. 144–150. Ştiinţa, Chişinău, 1986.
[61] Cheban D. N., Nonautonomous Dynamical Systems. Method of Lyapunov’s
Functions. In Proceedings of the 6th All-Union Conference of Qualitative
Theory of Differential Equations, pages 197–198, Irkutsk, 1986.
[62] Cheban D. N., C-Analytic Dissipative Dynamical Systems. Differentsial’nye
Uravneniya, 22(11):1915–1922, 1986.
 Bibliography 485

[63] Cheban D. N., On Nonautonomous Dissipative Dynamical Systems. Uspekhi

Matematicheskih Nauk, 41(4(250)):169, 1986.
[64] Cheban D. N., Nonautonomous Dissipative Dynamical Systems. Method of
Lyapunov’s Functions. Differentsial’nye Uravneniya, 23(3):464–474, 1987.
[65] Cheban D. N., Boundedness, Dissipativity and Almost Periodicity of the So-
lutions of Linear and Quasi-Linear Systems of Differential Equations. In Mat.
Issled. No. 92, Dinamicheskie Sistemy i Kraevye Zadachi, pages 143–159.
Ştiinţa, Chişinău, 1987.
[66] Cheban D. N., Nonautonomous Dissipative Dynamical systems. Method of
Lyapunov’s Functions. In Problems in the Qualitative Theory of Differential
Equations, ”Nauka”, Sibirsk. Otdel., Novosibirsk, 1988, pp. 56–64.
[67] Cheban D. N., Impulse and Difference Dissipative Systems with Periodic Co-
efficients. In Issledovaniya po Differentsialnym Uravneniyam i Matematich-
eskomu Analizu, pages 127–142. Ştiinţa, Chişinău, 1988.
[68] Cheban D. N., On the Structure of Levinson’s Centre of Dissipative Dynamical
Systems. Differentsial’nye Uravneniya, 24(6):1086, 1988.
[69] Cheban D. N., On the Structure of Levinson’s Centre of Dissipative Dynamical
Systems. Differentsial’nye Uravneniya, 24(9):1564–1576, 1988; Translation in
Differential Equation, 24 (1989), no.9, pp. 1031–1040.
[70] Cheban D. N., Principle of Averaging on Semi-Axis for Dissipative Systems. In
Dinamicheskie Sistemy i Uravneniya Matematicheskoi Fiziki, pages 149–161.
Ştiinţa, Chişinău, 1988.
[71] Cheban D. N., Nonautonomous Dissipative Dynamical Systems with Hyper-
bolic Subset of Centre (One-Dimensional Case). Mat. Zametki, 45(6):93–98,
1989; translation in Math. Notes 45 (1989), no. 5-6, pp.493–498.
[72] Cheban D. N., On One Problem of J.Hale. In Proceedings of the 7th All-
Union Conference of Qualitative Theory of Differential Equations, pp. 227,
Riga, 1989.
[73] Cheban D. N., Some Problems of Theory of Dissipative Dynamical Systems,
I. In Differentsial’nye Uravneniya i Matematicheskaya Fizika, pages 106:147–
163. Ştiinţa, Chişinău, 1989.
[74] Cheban D. N., On One Problem of J.Hale. Mat. Zametki, 46(1):120–121, 1989.
[75] Cheban D. N., Nonautonomous Dynamical Systems with Convergence. Dif-
ferentsial’nye Uravneniya, 25(9):1633–1635, 1989.
[76] Cheban D. N., On the Structure of Levinson’s Centre of Dissipative Dynamical
System with Condition of Hyperbolicity on Closure of the Set of Recurrent
Motions. Differentsial’nye Uravneniya, 26(5):913-914, 1990.
[77] Cheban D. N., On the Structure of Levinson’s Centre of Dissipative Dynamical
System with Condition of Hyperbolicity on Closure of the Set of Recurrent
Motions. Bulletin of Academy of Sciences of Republic of Moldova, Mathemat-
ics, (2):34–43, 1990.
486 Global Attractors of Non-autonomous Dissipative Dynamical Systems

[78] Cheban D. N., Dissipative Functional Differential Equations. Bulletin of

Academy of Sciences of Republic of Moldova. Mathematics, (2(5)):3–12, 1991.
[79] Cheban D. N., Nonautonomous Dissipative Dynamical Systems. Thesis of Dr.
of Sc., Institute of Mathematics, Minsk, 1991.
[80] Cheban D. N., Some Problems of Theory of Dissipative Dynamical Systems,
I. In Mat.Issled. No. 124, Funktsionalnaye Metody v Teorii Differentsial’nyh
Uravnenii, pp.106–122. Ştiinţa, Chişinău, 1992.
[81] Cheban D. N., Locally Dissipative Dynamical Systems and Some Their Appli-
cations. Bulletin of Academy of Sciences of Republic of Moldova. Mathematics,
(1):7–14, 1992.
[82] Cheban D. N., Global Attractors of Infinite-dimensional Dynamical Systems,
I. Bulletin of Academy of Sciences of Republic of Moldova, Mathematics,
(2(15)):12–21, 1994.
[83] Cheban D. N. and Fakeeh D. S., Global Attractors of Disperse Dynamical
Systems. Sigma, Chişinău, 1994.
[84] Cheban D. N., On the Structure of Compact Asymptotic Stable Invari-
ant Set of C-analytic Almost periodic Systems. Differential Equations, v.31,
No.12, pp.1995-1998, 1995 [Translated from Differentsial‘nye Uravneniya,
v.31, No.12, pp.2025-2028, 1995].
[85] Cheban D. N., On the Conversibility of Theorem of Lyapunov about a Stability
Asymptotic by First Approximation. Mat, Zametki 57(1995), no.1, 139–142;
Translation in Mathematical Notes, v.57, no.1-2, 1995, 100–102.
[86] Cheban D. N., Global Attractors of Infinite-Dimensional Dynamical Systems,
II. Bulletin of Academy of Sciences of Republic of Moldova, Mathematics,
1(17):28–37, 1995.
[87] Cheban D. N. and Fakeeh D. S., Global Attractors of Infinite-Dimensional
Dynamical Systems, III. Bulletin of Academy of Sciences of Republic of
Moldova, Mathematics, 2-3(18-19):3–13, 1995.
[88] Cheban D. N., Global Attractors of Infinite-Dimensional Nonautonomous
Dynamical Systems,I. Bulletin of Academy of Sciences of Republic of Moldova.
Mathematics. 1997, N3 (25), pp. 42-55
[89] Cheban D. N., Bounded Solutions of Linear Almost Periodic Differential Equa-
tions. Izv. Ross. Akad. Nauk, Ser. Mat. 62(1998), no.3, 155-174. Translation
in Izvestiya: Mathematics, 3(62):581–600, 1998.
[90] Cheban D. N., Global Attractors of Infinite-Dimensional Nonautonomous
Dynamical Systems, II. Bulletin of Academy of Sciences of Republic of
Moldova, Mathematics, (2(27)):111–126, 1998.
[91] Cheban D. N., The Asymptotics of Solutions of Infinite Dimensional Homoge-
neous Dynamical Systems. Mat. Zametki 63 (1998), No.1, 115-126; Translation
in Mathematical Notes, 1998. v. 63, No.1 , pp.115-126.
[92] Cheban D. N., The Global Attractors of Nonautonomous Dynamical Systems
 Bibliography 487

and Almost Periodic Limit Regimes of some Classes of Evolution Equations.

Anale Fac. de Mat. si Inform., Chişinău, v.1, 1999, pp.1-26.
[93] Cheban D. N., Relations Between the Different Type of Stability of the Linear
Almost Periodical Systems in the Banach Space. Electron. J. Diff. Eqns.
v.1999 (1999), No.46, pp.1-9.
[94] Cheban D. N., Global Attractors of Quasi-Homogeneous Nonautonomous
Dynamical Systems. Proceedings of the International Conference on
Dynamical Systems and Differential Equations. May 18-21, 2000, Kennesaw,
USA. pp.96-101.
[95] Cheban D. N., Uniform Exponential Stability of Linear Almost Periodic
Systems in Banach Space. Electron. J. Diff. Eqns., v.2000 (2000), No.29, pp.1-
18
[96] Cheban D. N., Kloeden P. E. and Schmalfuss B., Pullbac Attractors under
Discretization. Proceeding EQUADIFF99. Berlin 1999, vol.2 (Edited by B.
Fiedler, K. Groger and J. Sprekels). World Scientific 2000, pp.1024-1029.
[97] Cheban D. N., Schmalfuss B. and Kloeden P. E., Pullback Attractors in Dis-
sipative Nonautonomous Differential Equations under Discretization. Journal
of Dynamics and Differential Equations, v.13, No.1, 2001, pp. 185-213.
[98] Cheban D. N., Global Pullback Attractors of C-Analytic Nonautonomous
Dynamical Systems. Stochastics and Dynamics, v.1, No.4, 2001, pp.511-536.
[99] Cheban D. N., Global Attractors of Quasi-Homogeneous Nonautonomous
Dynamical Systems. Electron. J. Diff. Eqns., v.2002(2002), No.10, pp.1-19.
[100] Cheban D. N., Kloeden P. E. and Schmalfuss B., Relation Between Pullback
and Global Attractors of Nonautonomous Dynamical Systems. Nonlinear Dy-
namics and Systems Theory, v.2, No.2(2002), pp.9-28.
[101] Cheban D. N., Upper Semicontinuity of Attractors of Non-autonomous
Dynamical Systems for Small Perturbations. Electron. J. Diff. Eqns.,
v.2002(2002), No.42, pp.1-21.
[102] Cheban D. N., Global Attractors of Nonautonomous Dynamical Systems.
Chişinău, University Press of USM, 2002 (in Russian).
[103] Cheban D. N., Asymptotic Almost Periodic Solutions of Differential Equa-
tions. Chişinău, University Press of USM, 2002.(in Russian)
[104] Cheban D. N. and Duan J., Recurrent Motions and Global Attractors of
Nonautonomous Lorenz Systems. Dynamical Systems: An International Jour-
nal, v.19, No.1, 2004, pp.41-59.
[105] Cheban D. N., Duan J. and Gherco A. I., Generalization of the Second Bo-
golyubov’s Theorem for Non-Almost Periodic Systems. Nonlinear Analysis:
Real World Applications, v.4, No.4, 2003, pp.599-613.
[106] Cheban D. N., Kloeden P. E. and Schmalfuss B., Global Attractors of V-
monotone Nonautonomous Dynamical Systems. Bulletin of Academy of sci-
ences of Republic of Moldova. Mathematics, 2003. No.1(41), pp.47-57.
488 Global Attractors of Non-autonomous Dissipative Dynamical Systems

[107] Cheban D.N. and Duan J., Almost Periodic Solutions and Global Attractors
of Non-Atonomous Navier-Stokes Equations. Journal of Dynamics and Differ-
ential Equations, v.16, No.1, 2004, pp.
[108] Cheban D.N. and Mammana C., Invariant manifolds, global attractors and
almost periodic solutions of non-autonomous difference equations. Nonlinear
Analysis, series A v.56, No.4, 2004, pp.465-484.
[109] Chepyzhev V. V., Unbounded Attractors of Some Parabolic Systems of Differ-
ential Equations and Bound of their Dimensionality. Dolady Akademii Nauk
SSSR, 301(1):46-49, 1988.
[110] Chepyzhov V. V. and Vishik M. I., A Hausdorff Dimension Estimate for Ker-
nel Sections of Non-autonomous Evolutions Equations. Indian Univ. Math. J,
42(3):1057–1076, 1993.
[111] Chepyzhov V. V. and Vishik M. I., Attractors of Non-Autonomous Dynamical
Systems and their Dimension. J. Math. Pures Appl., 73 (1994), 279–333.
[112] Chepyzhov V. V. and Vishik M. I., Attractors for Equations of Mathematical
Physics. Amer. Math. Soc., Providence, RI, 2002.
[113] Cheresiz V. M., V -Monotone Systems and Almost Periodic Solutions. Sibirskii
Matematicheskii Zhurnal. 1972, v.13, No.4, pp.921–932.
[114] Cheresiz V. M., Uniformly V - Monotone Systems. Almost Periodic Solutions.
Sibirskii Matematicheskii Zhurnal. 1972, v.13, No.5, pp.11107–1123.
[115] Chernikova O. S., On Dissipativity of Systems of Differential Equations with
Impulse Effect. Ukrainskii Matematicheskii Zhurnal, 35(5):656–660, 1983.
[116] Chernikova O. S., On Boundedness of Solutions of Differential Equations with
Impulse Effect. Ukrainskii Matematicheskii Zhurnal, 38(1):124–127, 1986.
[117] Chicone C. and Latushkin Yu., Evolution Semigroups in Dynamicals Systems
and Differential Equations. Amer. Math. Soc., Providence, RI, 1999.
[118] Chueshov I. D., Finite-Dimensionality of Attractor in Some Problems of Non-
linear Theory of Envelopes. Matematicheskii Sbornik, 133(4):419–428, 1989.
[119] Chueshov I. D., Strong Solutions and Attractor of System of Karman’s Equa-
tions. Matematicheskii Sbornik, 181(1):25–36, 1990.
[120] Chueshov I. D., Introduction to the Theory of Infinite-Dimensional Dissipative
Systems. Acta Scientific Publishing Hous, Kharkov, 2002.
[121] Clement Ph, Heijmans H. J. A. M., Angenent S., C.V. van Duijn and B.
de Pagter, One-Parameter Semigroups. CWI Monograph 5. North-Holland,
Amsterdam.New York.Oxford.Tokyo, 1987
[122] Coddington E. A. and Levinson N., Theory of Ordinary Differential Equations.
McGraw Hill, New York, 1955.
[123] Conley C. C. and Miller R. K., Asymptotic Stability without Uniform Sta-
bility: Almost Periodic Coefficients. Journal of Differential Equations, 1:333–
336, 1965.
[124] Cooperman G. D., α-Condensing Maps and Dissipative Systems. PhD thesis,
 Bibliography 489

Brown University, 1978.

[125] Corduneanu C., Systems Differentiels Admettant des Solutions Bornées. C.R.
Acad. Sci., Ser.A–B., 245:21–24, 1957.
[126] Crauel H. and Flandoli F., Attractors for Random Dynamical Systems. Prob.
Theor. Related Fields, 100 (1994), 365-393.
[127] Crauel H., Debussche A. and Flandoli F., Random Attractors. Journal of
Dynamics and Differential Equations, 9 (1997), 307–341.
[128] Dafermos C. M., An Invariance Principle for Compact Processes. Journal of
Differential Equations, (9):239–252, 1971.
[129] Dafermos C. M., Uniform Processes and Semicontinuous Lyapunov Fonction-
als. Journal of Differential Equations, (11):401–415, 1972.
[130] Dafermos C. M., Semiflows Associated with Compact and Uniform Processes.
Math. Syst. Theory, 8(2):142–149, 1974.
[131] Dafermos C. M., Almost Periodic Process and Almost Periodic Solutions of
Evolution Equations. In Dynamical Systems. Proceedings of a University of
Florida International Symposium, pp.43–58, 1977.
[132] Daletskii Yu. L. and Krein M. G., Stability of Solutions of Differential Equa-
tions in Banach Space. Moscow, ”Nauka”, 1970. [English transl., Amer. Math.
Soc., Providence, RI 1974.]
[133] Datko R., Extending a Theorem of A. M. Lyapunov to Hilbert Space. J. Math.
Anal. Appl. 32 (1970), pp. 610-616
[134] Datko R., Uniform Asymptotic Stability of Evolutionary Processes in a Ba-
nach Space. SIAM, J. Math. Anal. 3 (1973), pp. 428-445
[135] Demidovich B. P., On Dissipativity of Certain Nonlinear Systems of Differen-
tial Equations, I. Vestnik MGU, (6):19–27, 1961.
[136] Demidovich B. P., On Dissipativity of Certain Nonlinear Systems of Differen-
tial Equations, II. Vestnik MGU, (1):3–8, 1962.
[137] Demidovich B. P., Lektsii po Matematicheskoi Teorii Ustoichivosti. Moscow,
”Nauka”, 1967.
[138] Denjoy A., Sur les Courbes Definiés par les Equations Differentielles a la
Surface du Tore. Journal Math. Pures et Appll (Ser.9), (11):333–375, 1932.
[139] Deyseach L. G. and Sell G. R., On the Existence of Almost Periodic Motions.
Michigan Mathematic Journal, 12(1):87–95, 1965.
[140] Dymnikov V. P. and Filatov A. N., Mathematics of Climate Modeling.
Birkhaüser, Boston, MA, 1997.
[141] Duan J. and Kloeden P. E., Dissipative Quasigeostrophic Motion under Tem-
porally Almost Periodic Forcing. J. Math. Anal.Applns. 236 (1999), 74-85.
[142] Engel K. J. and Nagel R., One-Parameter Semigroups for Linear Evolution
Equations. Springer-Verlag, Berlin, 1999.
[143] Eskman J. P., Roads to Turbulence in Dissipative Dynamical Systems.
Rev.Mod.Phys, 53:643–654, 1981.
490 Global Attractors of Non-autonomous Dissipative Dynamical Systems

[144] Fakeeh D. S., Levinson’s Center of Disperse Dissipative Dynamical Systems.

Izvestiya AN SSRM, (3):55–59, 1990.
[145] Fakeeh D. S., On Structure of Levinson’s Center of Disperse Dynamical
Systems. Izvestiya AN SSRM, (1):62–67, 1991.
[146] Fakeeh D. S., Analogue of Levinson–Pliss’ Theorem for Differential Inclu-
sions. In collection Funktsionalnye Metody v Teorii Differentsialnyh Uravnenii,
Matematicheskie Issledovaniya. 124:100–105, Chişinău, 1992.
[147] Fakeeh D. S. and Cheban D. N., Connectedness of Levinson’s Center of Com-
pact Dissipative Dynamical System without Uniqueness. Izvestiya AN RM,
Matematika, (1):15–22, 1993.
[148] Fang S., Global Attractor for General Non-Autonomous Dynamical Systems.
Nonlinear World, 2 (1995), pp.191-216.
[149] Fang S., Finite Dimensonal Behaviour of Periodic and Asymptotically Periodic
Processes. Nonlinear Analysis TMA, 28 (11) (1997), 1785–1797.
[150] Favard J., Sur les Equations Differentielles a Coefficients Presque-Periodiques.
Acta Math. 51 (1927), pp.31-81.
[151] Favard J., Leçons sur Fonction Presque-Periodiques. Paris, Gauthier-Villars,
1933.
[152] Fink A. M. and Fredericson P. O., Ultimate Boundedness Does not Imply
Almost Periodicity. Journal of Differential Equations, (9):280–284, 1971.
[153] Flandoli F. and Schmalfuß B., Random Attractors for the Stochastic 3D
Navier–Stokes Equation with Multiplicative White Noise, Stochastics and
Stochastics Reports, 59 (1996), 21–45.
[154] Flandoli F. and Schmalfuß B., Weak Solutions and Attractors of the 3D Navier
Stokes Equation with Nonregular Force, Journal of Dynamics and Differential
Equations, v. 11, No. 2(1999), pp. 355–398.
[155] Foias C., Sell G. R. and Temam R., Inertial Manifolds of Nonlinear Evolu-
tionary Equations. Journal of Differential Equations, 73(4):309–353, 1988.
[156] Furstenberg H., The Structure of Distal Flows. Amer. J. Math., 85 (1963),
477–515.
[157] Fusco G. and Oliva M., Dissipative Systems with Constraints. Journal of Dif-
ferential Equations, 63:362–388, 1986.
[158] Gerko A. I., Extensions of Topological Semi-Groups of Transformations. Uni-
versity Press of USM, Chişinău, 2001.
[159] Gerstein V. M. and Krasnoselskii M. A., Structure of Set of Solution of Dis-
sipative Equations. Doklady Akademii Nauk SSSR, 183(2):267–269, 1968.
[160] Gerstein V. M., On Dissipativity of One Two-Dimensional System. Differ-
entsial’nye Uravneniya, 5(8):1438–1444, 1969.
[161] Gerstein V. M., Note on Dissipative Flows. Trudy Matematicheskogo Fakultete
Voronezhskogo Gosuniversiteta, (1):26–34, 1970.
[162] Gerstein V. M., On Theory of Dissipative Differential Equations in Banach
 Bibliography 491

Space. Funktsionalnyi Analiz i Ego Prilozheniya, 4(3):99–100, 1970.

[163] Gerstein V. M., On Nonperiodical Dissipative Systems. Trudy Matematich-
eskogo Fakulteta Voronezhskogo Gosuniversiteta, (6):12–20, 1972.
[164] Ghidaglia J. M. and Temam R., Attractor for Damped Nonlinear Hiperbolic
Equations. J. Pures et Appl., 66:273–319, 1987.
[165] Glavan V. A., About One Problem of Johnson–Sell. Discontinuous Dynamical
Systems. Reports of the Scientific Conference, p.9, Ivano–Frankovsk, Ukraine,
1990.
[166] Gobbino M. and Sardella M., On the Connectedness of Attractors for
Dynamical Systems. Journal of Differential equations, 133(1):1-14, 1997.
[167] Gunning C. R. and Rossi H., Analytic Functions of Several Complex Variables.
Prentice–Hall, Englewood Cliffs, New Jersey, 1965.
[168] Hahn W., The Present State of Lyapunov’s Method. In ”Nonlinear Problems”
(R.E.Langer, ed.) University of Wisconsin,1959.
[169] Hahn w., Stability of Motion. Springer-Verlag, Berlin, 1967.
[170] Halanay A. and Wexler D., Teoria Calitativă a Sistemelor cu Impulsuri. Bu-
cureşti, 1968.
[171] Hale J. K., LaSalle J. P. and Slemrod M., Theory of a General Class of Dissi-
pative Processes. Journal Math. Appl., 39:177–191, 1972.
[172] Hale J. K., α-Contraction and Differential Equations. In Actes de la Confer-
ence Internationalle “Equa–Difs–73”, Actualites Scientifiques et Industrielles
1361, pp.16–41, Bruxelles et Louvain, 1973. Hermann.
[173] Hale J. K. and Lopes O. F., Fixed Point Theorems and Dissipative Process.
Journal of Differential Equations, 13:391–402, 1973.
[174] Hale J. K., Some Recent Results on Dissipative Processes. Lecture Notes in
Mathematics, 1980, v. 799, pp. 152–172.
[175] Hale J. K., Theory of Functional-Differential Equations. Mir, Moscow, 1984.
[176] Hale J. K., Asymptotic Behavior and Dynamics in Infinite Dimensions. In
Nonlinear Differ. Equations, pages 1–42. Boston e. á, 1985.
[177] Hale J. K., Asymptotically Smooth Semigroups and Applications. Lecture
Notes in Math., (1248):85–93, 1987.
[178] Hale J. K. and Raugel G., Upper Semicontinuity for a Singulary Perturbed
Hyperbolic Equation. Journal of Differential Equations, 73(2):197–214, 1988.
[179] Hale J. K., Asymptotic Behaviour of Dissipative Systems. Amer. Math. Soc.,
Providence, RI, 1988.
[180] Hale J. K. and Raugel G., Lower Semicontinuity of Attractors of Gradient
Systems and Applications. Ann. Math. Pura Appl., 4(154):281–326, 1989.
[181] Hale J. K., Asymptotic Behavior of Dissipative Systems. Bulletin of the Amer-
ican Mathematical Society, (1):175–183, 1990.
[182] Halmosh P., A Hilbert Space Problem Book. D. Van Nostrand Company, Inc.,
Toronto–London, 1967.
492 Global Attractors of Non-autonomous Dissipative Dynamical Systems

[183] Hapaev M. M., Averaging in Stability Theory. Nauka, Moscow (1986); English
transl. Kluwer Dordrech (1992).
[184] Haraux A., Attractors of Asymptotically Compact Processes and Applica-
tions to Nonlinear Partial Differential Equations. Commun. in Partial Differ.
Equat., 13(11):1383–1414, 1988.
[185] Haraux A., Systémes Dynamiques Dissipativs et Applications. Masson, Paris–
Milan–Barselona–Rome, 1991.
[186] Hartman P., Ordinary Differential Equations. Birkhauser, Boston–Basel–
Stuttgart, 1982.
[187] Hassani N., Systems Dynamiques Nonautonomes Contractants et leur Appli-
cations. Theése de magister. Algerie, USTHB, 1983.
[188] Henry D., Geometric Theory of Semilinear Parabolic Equations. Lecture Notes
in Mathematics, No.840, Springer-Verlag, New York 1981.
[189] Hitoshi I., On the Existence of Almost Periodic Complete Trajectories for
Contractive Almost Periodic Processes. Journal of Differential Equations,
43(1):66–72, 1982.
[190] Husemoller D., Fibre Bundles. Springer–Verlag, Berlin–Heidelberg–New York,
1994.
[191] Hutter W., Hyperbolicity of Almost Periodic Evolution Families.
Preprint (http://michelangelo.mathematik.uni-tuebingen.de/tuebingerichte/
1999.html)
[192] Iftrode V., A-Dissipativity for Periodic Solutions in Ordinary Differential
Equations. Rev. Roum. Pures et Appl., 23(6):513–521, 1983.
[193] Ignatiev A. O., Some Generalizations of Barbashin–Krasovskii’s Theorems.
Matematicheskaya Fizika, 34:19–22, 1983.
[194] Ilyashenko Yu. S., Weakly Contractive Systems and Attractors of Galerkian
Approksimations of Navier–Stocks’ Equation. Uspehi Matematicheskih Nauk,
36(3):243–244, 1981.
[195] Ilyashenko Yu. S., Weakly Contractive Dynamical Systems and Attractors.
Uspehi Matematicheskih Nauk, 37(1):166, 1982.
[196] Ilyashenko Yu. S. and Chetaev A. I., On Dimensionality of One Class of Dis-
sipative Systems. Prikladnaya Matematika i Mehanika, 46(3):374–381, 1982.
[197] Ilyashenko Yu. S., On Dimensionality of Attractors k-Contractive Systems in
Infinite-Dimensional Space. Vestnik MGU, Series I, Matematika, Mehanika,
3:52–59, 1983.
[198] Ilyashenko Yu. S., The Concept of Minimal Attractor and Maximal Attractor
of Partial Differential Equations of the Kuramoto–Sivashinsky Type. Chaos,
1(2):168–173, 1991.
[199] Ilyin A. A., Averaging Principle for Dissipative Dynamical System with
Rapidly Oscillating Right-Hand Sides. Mathematicheskii Sbornik, 187(1996),
No.5, 15-58: English Transl. in Sbornik: Mathematics, 187(1996).
 Bibliography 493

[200] Ilyin A. A., Global Averaging of Dissipative Dynamical System. Rendiconti

Academia Nazionale delle Scidetta dli XL. Memorie di Matematica e Appli-
cazioni. 116(1998), v.XXII, fasc.1, pp.165-191.
[201] Izé A. F. and J.G.Reis J. G., Contributions of Stability of Neutral Functional
Differential Equations. Journal of Differential Equations, 29:58–65, 1978.
[202] Izé A. F., On a Topological Method for the Analysis of the Asymptotic Be-
havior of Dynamical Systems and Processes. Complex Analysis, Functional
Analysis and Approximation Theory, North-Holland 1986, pp. 109-128.
[203] Johnson R. A., Ergodic Theory and Linear Differential Equations Systems.
Journal of Differential Equations, 28:23–34, 1978.
[204] Johnson R. A., On a Floquet Theory for Almost Periodic Two-Dimensional
Linear Systems. Journal of Differential Equations, 37:184–205, 1980.
[205] Jones G.S., The existence of Critical Points in Generalized Dynamical
Systems. In Springer-Verlag, editor, Lecture Notes in Math., v.60, pp.7–19,
1968. Seminaire on Differential Equations and Dynamical Systems.
[206] Kapitanskii L. V. and Kostin I. N., Attractors of Nonlinear Evolutionary Equa-
tions and Their Approximation. Algebra i Analiz, 2(1):114–140, 1990.
[207] Kartan A., Basic Theory of Analitic Functions of One and Several Complex
Variables. Mir, Moscow, 1963.
[208] Kato J. and Nakajima F., On Sacker–Sell’s Theorem for a Linear Skew Flow.
Tôhoku Math. Journ., 28:79–88, 1976.
[209] Kato J., Martynyuk A. A. and Shestakov A. A., Stability of Motion of Nonau-
tonomous Systems (Method of Limiting Equations). Gordon and Breach, Lux-
embourg, 1996.
[210] Katok A. B., Dynamical Systems with Hyperbolic Structure. Kiev, 1972, pages
125–211. 9th Mathematical School.
[211] Keller H. and Schmalfuß B., Random Attractors for Stochastic Sine Gordon
Equation via Transformation into Random Equations. Institut für Dynamische
System, Universität Bremen, 1999.
[212] Keller H. and Schmalfuß B., Attractors for Stochastic Differential Equations
with Nontrivial Noise. Izvestiya Akad. Nauk. RM., 1(26):43–54, 1998.
[213] Kelley J., General Topology. D. Van Norstand Company Inc., Toronto–New
York–London, 1957.
[214] Kloeden P. E., On Sharkowsky’s Cycle Coexistence Ordering, Bull. Austr.
Math. Soc. 20 (1979), 171-177.
[215] Kolyada S., On Dynamics of Triangular Maps of Square, Ergodic Theory and
Dynamical Systems, 12(1992), p.749-768.
[216] Kloeden P. E. and Lorenz J., Stable Attracting Sets in Dynamical Systems
and in their One-Step Discretizations. SIAM J. Numer. Analysis 23 (1986),
986-995.
[217] Kloeden P. E. and Schmalfuß B., Lyapunov Functions and Attractors
494 Global Attractors of Non-autonomous Dissipative Dynamical Systems

under Variable Time-Step Discretization. Discrete and Continuous Dynamical

Systems, 2(2):163–172, 1996.
[218] Kloeden P. E. and Schmalfuß B., Cocycle Attractors of Variable Time Step
Discretizations of Lorenzian Systems. J. Difference Eqns. Applns., 3 (1997),
125–145.
[219] Kloeden P. E. and Schmalfuss B., Nonautonomous Systems, Cocycle Attrac-
tors and Variable Time–Step Discretization. Numer. Algorithms 14 (1997),
141–152.
[220] Kloeden P. E. and Stonier D. J., Cocycle Attractors in Nonautonomously
Perturbed Differential Equations. Dynamics of Continuous, Discrete and Im-
pulsive Systems. 4(1998),211-226.
[221] Kloeden P. E. and Kozyakin V. S., The Inflation of Nonautonomous Systems
and their Pullback Attractors. Transactions of the Russian Academy of Nat-
ural Sciences, Series MMMIU. 4, No. 1-2, (2000), 144-169.
[222] Kolyada S., On Dynamics of Triangular Maps of Square, Ergodic Theory and
Dynamical Systems, 12(1992), p.749-768.
[223] Krasnosel’skii M. A., The Operator of Translation along Trajectories of Dif-
ferential Equations. Translations of Mathematical Monographs, Volume 19.
American Math. Soc., Providence, R.I., 1968.
[224] Krasnosel’skii M. A. and Zabreiko P. P., Geometrical Methodes of Nonlinear
Analyses. Nauka, Moscow, 1975
[225] Kuratovskii K., Topology. Mir, Moscow, 1966 2 volumes, v. 1.
[226] Kuratovskii K., Topology. Mir, Moscow, 1969 2 volumes, v. 2.
[227] Ladis N. N., Asymptotics of the Solutions of Quasihomogeneous Systems.
Differential Equations, 9(12):2257-2260, 1973.
[228] Ladyzhenskaya O. A., The Mathematical Theory of Viscous Incompressible
Flows. Nauka, Moscow, (1970); English transl. Gordon and Breach, New York,
1969.
[229] Ladyzhenskaya O. A., On Dynamical System Generated by Navier–Stocks’
Equations. In Zapiski Nauchnyh Seminarov LOMI, 27:911–115, 1972.
[230] Ladyzhenskaya O. A., On Finite-Dimensionality of Invariant Sets for Navier–
Stocks’ System and Other Dissipative Systems. In Zapiski Nauchnyh Semi-
narov LOMI, 112:137–155, 1982.
[231] Ladyzhenskaya O. A., On Attractors of Nonlinear Evolutionary Problems with
Dissipation. In Zapiski Nauchnyh Seminarov LOMI, 152:72–85, 1986.
[232] Ladyzhenskaya O. A., On the Determination of Minimal Global Attractors
for Navier–Stocks’ Equations and Other Equations with Partial Derivatives.
Uspekhi Mat. Nauk, 42(6(258)):25–60, 1987.
[233] Ladyzhenskaya O.A., Some Additions and Concretizations to My Works on
Theory of Attractors for Abstract Semi-groups. In Zapiski Nauchnyh Semi-
narov LOMI, 182:102–112, 1990.
 Bibliography 495

[234] Ladyzhenskaya O. A., Attractors for Semigroups and Evolution Equations.

Lizioni Lincei, Cambridge Univ. Press, Cambridge, New-York, 1991.
[235] LaSalle J. P., Dissipative Systems. In Proccedings of the Conference on Ordi-
nary Differential Equations, pp.165–174, Washington, D.C., 1971. New York.
[236] Latushkin Yu. and Schnaubelt R., Evolution Semigroups, Translation Alge-
bras, and Exponential Dichotomy of Cocycles. Journal of Differential Equa-
tions 1999, v.159, pp.321-369
[237] Levinson N., Transformation Theory of Nonlinear Differential Equations of
the Second Order. Ann. Math., 45(4):723–737, 1944.
[238] Levitan B. M. and Zhikov V. V., Almost Periodic Functions and Differential
Equations. Cambridge Univ. Press, Cambridge, 1982.
[239] Lihtenberg A. and Liberman N., Regular and Stochastic Dynamics. Mir,
Moscow, 1984.
[240] Loins J. L., Quelques Methodes de Résolution des Problèmes aux Limites non
Linéaires. Dunod, Paris, 1969.
[241] Liubich Yu. I., Note on Stability of Complex Dynamical Systems. Izvestiya
VUZov, Matematika, (10(257)):49–50, 1983.
[242] Lorenz E. N., Deterministic Nonperiodic Flow. Journal of the Atmospheric
Sciences. 1962, 20, pp.130-141.
[243] Mallet-Paret J., Morse Decomposition for Delay-Differential Equations. Jour-
nal of Differential Equations, 72(2):270–315, 1988.
[244] Mallet-Paret J. and Sell G. R., Inertial Manifolds for Reaction Diffusion Equa-
tions in Higher Space Dimentions. Journal Amer. Math. Soc., (1):805–866,
1998.
[245] Manfredi Bianca, Dissipativity, quasiperiodicity and recurrence. Riv. Mat.
Univ. Parma (4) 10 (1984), pp.449–458.
[246] Massat P., Stability and Fixed Points of Point-Dissipative Systems. Journal
of Differential Equations, 40:217–231, 1981.
[247] Massat P., On Obtaininig Ultimate Boundedness for α- Contraction Volterra
and Functional-Differential Equations. In Lecture Notes in Pure and Appl.
Math., pp.273–280. Dekker, New York, 1983.
[248] Massat P., Attractivity Properties of α-Contractions. Journal of Differential
Equations, 48:326–333, 1983.
[249] Massera J. L. and Schäffer J. J., Linear Differential Equations and Functional
Analysis, I. Ann. of Math., 1958, v.67, No.2, pp. 517-573
[250] Massera J. L. and Shäffer J. J., Linear Differential Equations and Function
Spaces. Academic Press, New York–London, 1966.
[251] Matrosov V. M., Principle of Comparison with Vector Function of Lyapunov.
Differentsial’nye Uravneniya, 5(12):2129–2143, 1969.
[252] Miller R. K., Almost Periodic Differential Equations as Dynamical System
with Applications to the Existence of Almost Periodic Solutions. Journal of
496 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Differential Equations, 1:337–345, 1965.

[253] Millionshchikov V. M., Recurrent and Almost Periodic Limit Trajectories of
Nonautonomous Systems of Differential Equations. Doklady Akademii Nauk
SSSR, 161(1):43–44, 1965.
[254] Millionshchikov V. M., On Recurrent and Almost Periodic Limit Solutions of
Nonautonomous Systems. Differentsial’nye Uravneniya, 4(9):1555–1559, 1968.
[255] Millionshchikov V. M., Linear Systems of Ordinary Differential Equations.
Differentsial’nye Uravneniya, 7(3):387–390, 1971.
[256] Mitropolskii Yu. A., Perestiuk N. A. and Chernikova O. S., Convergence of
Systems of Differential Equations with Impulse Effect. Doklady Akademii Nauk
SSSR, A(11):11–15, 1983.
[257] Mitropolskii Yu. A. and Kulik V. L., Bounded Solutions of Nonlinear Systems
of Differential Equations. Ukrainskii Matematicheskii Zhurnal, 36(6):720–729,
1984.
[258] Nemytskii V. V. and Stepanov V. V., Qualitative Theory of Differential Equa-
tions. Nauka, Moscow, 1949.
[259] Nemytskii V. V., On Some Methods of Global Qualitative Research of Multi-
Dimensional Autonomous Systems. Trudy Moskovskogo Matematicheskogo
Obschestva, 5:455–482, 1956.
[260] Nemytskii V. V., Some Modern Problems of Qualitative Theory of Ordinary
Differential Equations. Uspekhi Matematicheskih Nauk, 22(4):3–36, 1965.
[261] Nitecki Z., Differentiable Dynamics. The MIT Press, Cambridge–
Massachusetts–London, 1971.
[262] Ochs G., Weak Random Attractors. Technical Report, Institut für Dynamische
Systeme, Universität Bremen, 1999.
[263] Opial Z., Sur une Equation Diffirentielle Presque-Periodique sans Solution
Presque-Periodique. Bull. Acad. Polon. Sci., Ser. Math. Astron. Phys., 9:673–
676, 1961.
[264] Pavel N., On Dissipative Systems. Bolletino Un. Mat. Ital., 4(5):701–707,
1971.
[265] Pavel N., A Generalization of Ultimately Bounded Systems. An. Sti. Uni.
“Al.I.Cuza”, Iaşi. Sect. I.a, Mat., 18:81–86, 1972.
[266] Pavel N., On the Boundedness of Solutions of Systems of Differential Equa-
tions. Tohoku Math. Journal, 24:21–32, 1972.
[267] Perov A.I. and Trubnikov Yu. V., Monotone Differential Equations, I. Differ-
entsial’nye Uravneniya, 10(5):804–815, 1974.
[268] Perov A.I. and Trubnikov Yu. V., Monotone Differential Equations, II. Dif-
ferentsial’nye Uravneniya, 12(7):1223–1237, 1976.
[269] Piliugin S. Yu., Structure of Attracting Sets of Hyperbolic Systems of Differ-
ential Equations. Thesis of Dr. of Sc., Leningrad, 1983.
[270] Pliss V. A. Nonlocal Problems in the Theory of Oscilations. Academic Press,
 Bibliography 497

1966.
[271] Pliss V. A., Integral Sets of Periodic Systems of Differential Equations. Nauka,
Moscow, 1977.
[272] Reissig R., Sansone G. and Conti R., Qualitative Theorie Nichthlinearer Dif-
ferentialgleichungen. Edizioni Cremonese, Roma, 1963.
[273] Rouche N., Habets P. and Laloy M., Stability Theory by Lyapunov’s Direct
Method. Springer–Verlag, New York–Heidelberg–Berlin, 1977.
[274] Rudin U., Functional Analysis. Mir, Moscow, 1975.
[275] Sacker R. J. and Sell G. R., Existence of Dichotomies and Invariant Splittings
for Linear Differential Systems, I. Journal of Differential Equations, 15:429–
458, 1974.
[276] Sacker R. J. and Sell G. R., Lifting Properties in Skew–Product Flows with
Applications to Differential Equations. Memoirs oft he American Math. Soc.,
v.190, Providence, R.I., 1977.
[277] Sacker R. J. and Sell G. R., Dichotomies for Linear Evoltionary Equations in
Banach Spaces. Journal of Differential Equations 1994, v.113, pp.17-67
[278] Sadovskii B. N., Limit Compact and Condensing Operators. Uspechi Matem.
Nauk, 27(1(163)):81–146, 1972.
[279] Samoilenko A. M. and Perestiuk N. A., Impulsive Differential Equations.
World Scientific Publishing Co., Inc., River Edge, NY, 1995.
[280] Samoilenko A. M., and Trofimuchiuk S. I., Unbounded Functions with Almost
Periodic Differences. Ukrainian Math. J., 43 (1991), 1409-1413.
[281] Samoilenko A. M. and Trofimuchiuk S. I., On the Space of Piecewise Con-
tinuous Almost Periodic Functions and Almost Periodic Sets on the Line.
Ukrainian Math. J.,43 (1991), 1613–1619.
[282] Saperstone S. N., Semidynamical Systems in Infinite Dimensional Spaces.
Springer-Verlag, New York-Heidelberg-Berlin 1981.
[283] Scheutzow M., Comparison of Various Concepts of a Random Atractor: A
Case Study. Technical report, Technische Universität Berlin, 2000.
[284] Schmalfuß B., Backward Cocycles and Attractors of Stochastic Differen-
tial Equations. in International Seminar on Applied Mathematics–Nonlinear
Dynamics: Attractor Approximation and Global Behaviour, V. Reitmann,
T. Riedrich, and N. Koksch (eds.), TU Dresden, 1992, pp 185–192.
[285] Schmalfuß B., Attractors for the Non-Autonomous Dynamical Systems. In
K. Gröger B. Fiedler and J. Sprekels, editors, Proceedings EQUADIFF99,
pages 684–690. World Scientific, 2000.
[286] Schmalfuss B., Attractors for the Non-Autonomous Navier-Stokes Equation
(to appear).
[287] Schwartz L., Analyse Mathématique, v. 1. Hermann, 1967.
[288] Schwartz L., Analyse Mathématique, v. 2. Hermann, 1967.
[289] Seifert G., Almost Periodic Solutions for Almost Periodic System of Ordinary
498 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Differential Equations. Journal of Differential Equations, 2:305–319, 1966.

[290] Sell G. R., Non-Autonomous Differential Equations and Topological Dynam-
ics, I. The basic theory. Trans. Amer. Math. Soc., 127:241–262, 1967.
[291] Sell G. R., Non-Autonomous Differential Equations and Topological Dynam-
ics, I. Limiting equations. Trans. Amer. Math. Soc., 127:263–283, 1967.
[292] Sell G. R., Lectures on Topological Dynamics and Differential Equations, vol-
ume 2 of Van Nostrand Reinhold math. studies. Van Nostrand–Reinbold, Lon-
don, 1971.
[293] Sell G. R., W. Shen and Y. Yi, Topological Dynamics and Differential Equa-
tions, Contemporary Mathematics, volume 215, 1998, pp.279-297.
[294] Sell G. R. and You Y., Dynamics of Evolutionary Equations. Springer-Verlag,
New York, 2002.
[295] Shabat B. V., Introduction into Complex Analysis. Moscow, ”Nauka”, 1969.
[296] Sharkovskii A. N., Attractors of Some Nonlinear Boundary Problems. Pro-
ceedings of 6th All-Union Conference on Qualitative Theory of Differential
Equations, pp.205, Irkutsk, 1986.
[297] Sharkovskii A. N, Maistrenko Yu. L. and Romanenko E. Yu., Difference Equa-
tions and Their Applications. Naukova Dumka, Kiev, 1986.
[298] Shchennikov V. N., Research of Convergence in Non-Autonomous Differen-
tial System with the Help of Vector Function of Lyapunov. Differentsial’nye
Uravneniya, 19(11):1902–1907, 1983.
[299] Shchennikov V. N., Convergence of Complicated Systems of Differential Equa-
tions. Differentsial’nye Uravneniya, 19(11):1568–1571, 1984.
[300] Shcherbakov B. A., Topologic Dynamics and Poisson Stability of Solutions of
Differential Equations. Ştiinţa, Chişinău, 1972.
[301] Shcherbakov B.A. and Cheban D. N., Asymptotically Poisson Stable Motions
of Dynamical Systems and Comparability of Their Reccurence in Limit. Dif-
ferentsial’nye Uravneniya, 13(5):898–906, 1977.
[302] Shcherbakov B. A., Poisson Stability of Motions of Dynamical Systems and
Solutions of Differential Equations. Ştiinţa, Chişinău, 1985.
[303] Shestakov A. A., Generalized Lyapunov’s Method for Abstract Semi-
Dynamical Processes, I. Semi-dynamical Processes as Semi-Dynamical
Systems. Localisation of Limite Set of Autonomous and Asymptoti-
caly Autonomous Semi-Dynamical Processes. Differentsial’nye Uravneniya,
22(9):1475–1490, 1986.
[304] Sibirsky K. S., Introduction to Topological Dynamics. Noordhoff, Leyden,
1975.
[305] Sibirskii K. S. and Shube A. S., Semidynamical Systems. Stiintsa, Kishinev
1987. (Russian)
[306] Sinai Yu. G. and Shilnikov L. P. (eds.), Strange Attractors (Collection of Ar-
ticles). Mir, Moscow, 1981.
 Bibliography 499

[307] Skovronski J. and Ziemba S., The Problem of Vibrations of Nonautonomic

Systems with Strong Nonlinearity. Arch. Mech. Stosowanej, 10(4):517–523,
1958.
[308] Stuart A. M. and Humphries A. R., Numerical Analysis and Dynamical
Systems. Cambridge University Press, Cambridge 1996.
[309] Talpalaru P., Sur les Systems Dissipatifs. Ann. Stiint. Univ. Iasi, Sec. I.a,
13(1):43–47, 1967.
[310] Tchéban D. N., Systemes Dynamiques Dissipatifs. In II-Rencontre Nationale
sur les Equations Differentielles Ordinaires. Resumes des Exposes, pages 40–
41, Alger, 1983.
[311] Temam R., Navier–Stokes Equation–Theory and Numerical Analysis. North-
Holland, Amsterdam, 1979.
[312] Temam R., Induced Trajectories and Approximate Inertial Manifolds. Math.
Modelling. Numer. Anal., 23:541–561, 1989.
[313] Temam R., Inertial Manifolds. Math. Intelligencer, 12(4):68–74, 1990.
[314] Temam R., Infinite–Dimensional Dynamical Systems in Mechanics and
Physics. 2nd edition, Springer Verlag, New York,1997.
[315] Titi E. S., On Approximate-Inertial Manifolds to the Navier–Stokes Equations.
Journal Math. Anal. Appl., 149:540–557, 1990.
[316] Trubnikov Yu. V. and Perov A. I., The Differential Equations with Monotone
Nonlinearity. Nauka i Tehnika. Minsk, 1986 (in Russian).
[317] Vishik M. I. and Chepyzhev V. V., Attractors of Non-Autonomous Dynamical
Systems. Matematicheskie Zametki, 51(6):141–143, 1992.
[318] Vishik M. I., Asymptotic Behaviour of Solutions of Evolutionary Equations.
Cambridge Univ. Press, Cambridge, 1992.
[319] Vuillermot Pierre-A., Attracteurs Presque-Priodiques pour une Classe
d’Equations Paraboliques non Linaires du Type Raction-Diffusion sur R N .
(French. English summary) [Almost-Periodic Attractors for a Class of Non-
linear Parabolic Reaction-Diffusion Equations in RN ] C. R. Acad. Sci. Paris
Sr. I Math. 311 (1990), no. 10, pp.583–588.
[320] Vuillermot Pierre-A., Almost-Periodic Attractors for a Class of Non-
Autonomous Reaction-Diffusion Equations on R N .I. Global Stabilization Pro-
cesses. J. Differential Equations 94 (1991), no. 2, 228–253.
[321] Vuillermot Pierre-A., Almost-Periodic Attractors for a Class of Non-
Autonomous Reaction-Diffusion Equations on R N . II. Codimension-One Sta-
ble Manifolds. Differential Integral Equations 5 (1992), no. 3, 693–720.
[322] Vuillermot Pierre-A., Global Exponential Attractors for a Class of Almost-
Periodic Parabolic Equations in RN . Proc. Amer. Math. Soc. 116 (1992), no.
3, 775–782.
[323] Vuillermot Pierre-A., Almost-Periodic Attractors for a Class of Non-
Autonomous Reaction-Diffusion Equations on R N . III. Center Curves and
500 Global Attractors of Non-autonomous Dissipative Dynamical Systems

Liapounov Stability. Nonlinear Anal. 22 (1994), no. 5, 533–559.

[324] Xavier M., Finite-Dimentional Attracting Manifolds in Reaction-Diffusion
Equations. Contemp. Math., 17:540–557, 1983.
[325] Yoshizawa T., Note on the Boundedness and the Ultimate Boundedness of
Solutions x0 = f (t, x). Memoirs of the College of Science. University of Kyoto,
Serie A, 29(3):275–291, 1955.
[326] Yoshizawa T., Lyapunov’s Function and Boundedness of Solutions. Funkcialaj
Ekvacioj, 2:95–142, 1959.
[327] Yoshizawa T., Stability Theory by Lyapunov’s Second Method. The Mathe-
matical Series of Japan. Tokyo, 1966.
[328] Yosida K., Functional Analysis. Springer–Verlag, Berlin, 1995.
[329] Zadorozhnyi V. G., On Nonlocal Solutions of V -Dissipative Differential Equa-
tions. Ukrainskii Matematicheskii Zhurnal, 35(3):303–308, 1983.
[330] Zhang S., Functional Differential Equations and Topological Dynamics. Chi-
nese Science Bulletin, 35(17):1441–1448, 1990.
[331] Zhikov V. V., On Problem of Existence of Almost Periodic Solutions of Differ-
ential and Operator Equations. Nauchnye Trudy VVPI, Matematika, (8):94–
188, 1969.
[332] Zhikov V. V., On Stability and Unstability of Levinson’s centre. Differ-
entsial’nye Uravneniya, 8(12):2167–2170, 1972.
[333] Zhikov V. V., Monotonicity in the Theory of Almost Periodic Solutions of Non-
Linear operator Equations. Mat. Sbornik 90 (1973), 214–228; English transl.,
Math. USSR-Sb. 19 (1974), 209-223.
[334] Zinchenko I. L., On Structure of Asymptotically Stable Invariant Sets of Pe-
riodical Complex Analytic Systems of Differential Equations. Vestnik LGU,
Matematika, Mehanika, Astronomiya, 1(1):23–25, 1980.
[335] Zinchenko I. L., Theorem on Convergence of Complex Analitic Systems. Dif-
ferential Equations, 25(10):1809–1810, 1989.
[336] Zubov V. I., The Methods of A. M. Lyapunov and Their Application. Noord-
hoof, Groningen, 1964.
[337] Zubov V. I., Stability of Motion. Vyshaya Shkola, Moscow, 1973.
[338] Zubov V. I., Theory of Oscillations. Nauka, Moscow, 1979.
 Index

c0 − cocycle, 451 asymptotically almost periodic, 74

D+ (M ), 18 asymptotically compact, 31
Eω− , 311 asymptotically recurrent, 74
Eω+ , 310 asymptotically stationary, 74
H-class, 7, 135 attracting, 3
J + (M ), 18 autonomous Lorenz systems, 245
V - monotone, 386
Vπ0 (x), 174 base of extension, 6
V −monotone cocycle, 394 Bebutov’s dynamical system, 5
α-condensing, 295, 435 Bohr’s almost periodic function, 327
α-condensing cocycle, 299 bounded, 36
α-contraction, 436 bounded k (b)-dissipativity, 11
α-limit, 2 bounded dissipative, 10
λ-condensing, 35
λ-contraction, 33 cascade (discrete flow), 1
C-analytic dissipative dynamical system, Cauchy’s matrix, 226
124 center of Levinson, 11, 12
C-analytic system, 123 chain recurrent, 155
K, 390 chaotic, 478
Mϕ , 146 cocycle, 7
N+ω , 310 collectively ( uniformly collectively)
N−ω , 310 asymptotic compact, 270
Nω , 310 collectively compact dissipative, 268
M-dissipative, 10 compact k (b)-dissipative, 11
Mx , 475 compact k (b)-dissipativity, 11
ω-limit, 2 compact bounded, 36
ε > 0 shift (almost period) of point compact dissipative, 10
x ∈ X, 2 compact dissipative cocycle, 249
r-cycle, 156 comparable in limit, 74
A, 173 completely continuous, 30
1-cycle, 156 condensing, 34
condition (C), 53
absorbing (uniformly absorbing), 270 condition (A), 70
almost recurrent (almost periodic), 2 condition of Ladyzhenskaya, 30
asymptotic stable, 3 conditional α-contraction, 436
asymptotically τ -periodic, 74 conditionally α-condensing, 435

501
502 Global Attractors of Non-autonomous Dissipative Dynamical Systems

conditionally relatively compact, 310 locally maximal, 295

connectedness, 38
continuous section, 67 maximal compact invariant set of cocycle,
convergent, 70 264
maximal monotone operator, 239
direct product of the dynamical systems, measure of non-compactness of
407 Kuratowsky, 9
dissipativity of the equation, 120 metric space with segments, 389
distal, 304 minimal set, 2
dynamical system, 1 monotone, 239
dynamical system of translations, 5 monotone operator, 239
motion, 1
exponential dichotomy, 463 multiplicator, 454
extension, 6
non-autonomous dynamical system, 6
factor of dynamical system, 6 non-autonomous dynamical system,
flow, 1 generated by cocycle, 7
forward attractor, 290 non-autonomous Lorenz, 245
non-autonomous Navier-Stokes equation,
generalized homoclinic contour, 156 361
global asymptotic stable, 3 nonlinear elliptic operator, 403
global attractor, 29
global attractor of the cocycle, 93 operator of monodromy, 454
globally asymptotically stable in the sense orbital stable, 3
of Lyapunov-Barbashin, 54 orbitally stable with respect to the
group (semi-group) system, 1 non-autonomous system, 54

homogeneous, 101 Poincare transformation, 169

homomorphism, 6 point k (b)-dissipative, 11
hyperbolic, 157 point k(b)-dissipativity, 11
hyperbolic structure, 157 point dissipative, 10
Poisson stable, 2
indecomposable, 2, 38 positively invariant (negatively invariant,
invariant section, 67 invariant), 1
invariant with respect to a cocycle, 290 power-law asymptotic, 107
isomorphism, 6 Problem of J. Hale, 33
projection, 463
jointly recurrent, 408 property (S), 43, 272
pseudo recurrent, 369, 477
linear non-autonomous dynamical system, pullback attractor, 93
113 pullback dissipative, 327
linear non-homogeneous system, 115
local k (b)-dissipativity, 11 quasi-minimal, 165
local attractor, 294, 379
local condition of Lipschitz, 179 recurrent, 2
local forward attractor, 379 relation of partial order, 156
locally bounded, 35
locally compact, 73 sectorial operator, 358
locally completely continuous, 16 semi-continuous, 239
locally dissipative, 10 shifts dynamical system, 5
 Index 503

skew-product dynamical system, 7 uniform pullback attractor, 93

stable (unstable) manifold, 3 uniform stable in the positive direction, 61
stable in the sense of Poisson in the uniformly collectively compact dissipative,
negative direction, 2 268
stable in the sense of Poisson in the uniformly compatible, 146
positive direction, 2 uniformly dissipative, 121, 320
stationary (τ -periodic, τ > 0, τ ∈ T) uniformly monotone, 239
point, 2 uniformly stable in the positive direction,
Stepanov almost periodic function, 330 79
strictly metric-convex space, 389 uniformly stable in the sense of Lagrange,
systems of hydrodynamic type, 245 44
upper semi–continuous, 292
trajectory, 1
transitive, 369, 478 weak attractor, 16
triangular map, 462 weakly b-dissipative, 30
tubular neighborhood, 189 weakly dissipative, 16
weakly regular, 146
uniform attracting, 3 whole trajectory, 3
uniform forward attractor, 290