2023 г. июль — август т. 78, вып.

4 (472)
УСПЕХИ МАТЕМАТИЧЕСКИХ НАУК

UDC 517.938 Dedicated to the memory of M. I. Vishik

on the occasion of his 100th birthday

Attractors. Then and now

S. V. Zelik
This survey is dedicated to the 100th anniversary of Mark Iosifovich Vishik
and is based on a number of mini-courses taught by the author at the Univer-
sity of Surrey (UK) and Lanzhou University (China). It discusses the classical
and modern results of the theory of attractors for dissipative PDEs, including
attractors for autonomous and non-autonomous equations, dynamical systems
in general topological spaces, various types of trajectory, pullback and ran-
dom attractors, exponential attractors, determining functionals and inertial
manifolds, as well as the dimension theory for the classes of attractors men-
tioned above. The theoretical results are illustrated by a number of clarifying
examples and counterexamples.
Bibliography: 248 titles.

Keywords: dissipative PDEs, attractors, inertial manifolds, finite-dimen-

sional reduction, determining functionals.
DOI: [Link]

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2. Attractors: basic theory and model examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3. Attractors: a unified approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4. Trajectory attractors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.1. ODEs and reaction-diffusion equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.2. Trajectory attractors for elliptic PDEs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.3. 3D Navier–Stokes system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.4. Connectedness of trajectory attractors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5. Attractors for non-autonomous problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.1. Pullback attractors: a general approach . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.2. Cocycles and random attractors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.3. Uniform attractors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
This work is partially supported by Moscow Center of Fundamental and Applied Mathematics,
Аgreement with the Ministry of Science and Higher Education of the Russian Federation, No.
075-15-2022-283.

⃝
c S. V. Zelik, 2023

53
54 S. V. ZELIK

6. Dimensions of the attractor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

6.1. Mané’s projection theorem and finite-dimensional reduction . . . . . . . 119
6.2. Upper bounds via the squeezing property: the autonomous case . . . 121
6.3. Upper bounds via the squeezing property: the non-autonomous case 124
6.4. Volume contraction and Lyapunov dimension. . . . . . . . . . . . . . . . . . . . . . 130
6.5. Lower bounds for the dimension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
7. Inertial manifolds and finite-dimensional reduction . . . . . . . . . . . . . . . . . . . . . . . 140
7.1. Spectral gap conditions and inertial manifolds . . . . . . . . . . . . . . . . . . . . . 141
7.2. Beyond the spectral gap conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
8. Exponential attractors and perturbation theory. . . . . . . . . . . . . . . . . . . . . . . . . . . 158
8.1. Exponential attractors via squeezing properties. . . . . . . . . . . . . . . . . . . . 164
8.2. Regular attractors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
9. Determining functionals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
10. Appendix. Function spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185

1. Introduction

One of the most surprising lessons of dynamical systems theory is that even rela-
tively simple and deterministic ordinary differential equations (ODEs) may demon-
strate a very complicated behaviour with strong random features (a so-called de-
terministic chaos). This phenomenon has intensively been studied starting from
the second part of the 20th century, a lot of prominent results have been obtained
in this direction, and a lot of methods for investigating it have been developed (such
as hyperbolic theory, Lyapunov exponents, homoclinic bifurcation theory, strange
attractors, and so on; see [62], [91], [116], [202], [210], and the references therein).
However, this phenomenon occurred to be much more complicated than it was
thought from the very beginning, so, despite many efforts made, the existing theory
is mainly limited to low-dimensional model examples (even in these examples, it is
established that, in some cases, the problem of a full description of the dynamics
is not algorithmically solvable) and we still do not have really effective methods for
studying the dynamical chaos in higher-dimensional systems of ODEs.
The situation becomes even more complicated when we deal with dynamical sys-
tems generated by partial differential equations (PDEs). For such systems, the ini-
tial phase space is infinite-dimensional (for example, L2 (Ω), where Ω is a domain
in Rd ), which causes a lot of extra difficulties. In addition, together with a tempo-
ral variable t and related temporal chaos, we now have spatial variables x ∈ Ω, so
the spatial chaos may naturally appear. Also, we may have an interaction between
the spatially and temporal chaotic modes which form the so-called spatio-temporal
chaos. As a result, a new type of purely infinite-dimensional dynamics with a prin-
cipally new level of complexity (unreachable in systems of ODEs studied in classical
dynamics) may arise; see [174], [221], [239], and the references therein.
Nevertheless, there exists a wide class of PDEs, namely, the class of dissipative
PDEs in bounded domains, where, despite the infinite-dimensionality of the ini-
tial phase space, the effective limit dynamics is finite-dimensional in a certain
 ATTRACTORS. THEN AND NOW 55

sense. In particular, this dynamics can be described using finitely many param-
eters (the “order parameters” in the terminology of I. Prigogine; see [187]), whose
evolution obeys a system of ODEs (which is called an inertial form of the initial
dissipative system).
As hinted by their name, dissipative systems consume energy (in contrast to con-
servative systems where the total energy is usually preserved), so, in order to have
non-trivial dynamics, the energy income should be taken into account (in other
words, the system under consideration should be open and should interact with
the external world). On the physical level of rigour, the rich and complicated dy-
namical structures (often referred to as dissipative structures following Prigogine)
arise as the result of interaction of the following three mechanisms:
1) energy decay, usually more essential in higher “Fourier modes”;
2) energy income, usually through the lower “Fourier modes”;
3) energy flow from lower to higher “modes” which is provided by the nonlinear-
ities.
Moreover, it is typical for dissipative PDEs in bounded domains that the number
of lower “modes” where the energy income is possible is finite. So, it is natural
to expect that these modes are, in a certain sense, dominating and the higher
modes are slaved by the lower ones. This supports the conjecture that the effec-
tive dynamics in such systems is finite-dimensional up to some transient behaviour
(where the unstable lower “Fourier modes” are treated as order parameters) and
somehow explains why the ideas and techniques of classical finite-dimensional the-
ory of dynamical systems are also effective for describing the dissipative dynamics
in such PDEs.
However, the above arguments are very non-rigorous from the mathematical
point of view, and it is extremely difficult/impossible to make them meaningful.
It is even unclear what are the “modes” in the above statements. Indeed, they
are rigorously defined as Fourier/spectral modes in the linear theory and a priori
make sense only for systems somehow close to linear ones. Moreover, such modes
are natural for the linearization near equilibria only and may not exist at all when
the linear equations with time-periodic coefficients are considered. Thus, despite
the common usage of modes and related length scales for highly nonlinear systems
in physics literature (for example, in turbulence, see [81] and the references therein),
the precise meaning of them is usually unclear.
That is the reason why the mathematical theory of dissipative PDEs is based
on related but different concepts which at least can be defined rigorously. Namely,
let a dissipative system be given by a semigroup S(t), t > 0, acting in a Hilbert,
Banach (or sometimes even Hausdorff topological) space Φ. Usually, in applica-
tions Φ is some infinite-dimensional function space where the initial data live (say,
Φ = L2 (Ω)) and S(t) is a solution operator of the PDE under consideration. Then
the dissipativity is often understood as the validity of the following dissipative es-
timate:
∥S(t)u0 ∥Φ 6 Q(∥u0 ∥Φ )e−αt + C∗ (1.1)
for appropriate positive constants C∗ and α and monotone increasing function Q
which are independent of u0 and t. Roughly speaking, this estimate shows that there
56 S. V. ZELIK

is a balance between energy injection and dissipation, so the energy cannot grow to
infinity. Moreover, for very high energy levels dissipation is dominating and the en-
ergy decays. As we have already mentioned, the class of dissipative PDEs is rather
wide and includes many physically relevant examples, such as the Navier–Stokes
equations, reaction-diffusion equations, damped wave equations, various pattern
formation equations, and so on (see, for example, [215] and the references therein
for more examples).
The central concept of the theory is a (global) attractor A. By definition, it is
a compact subset of the phase space Φ, which is invariant with respect to the semi-
group S(t) and attracts the images (under the map S(t)) of all bounded sets as
time tends to infinity. Thus, on the one hand the attractor A consists of the full
trajectories of the dynamical system under consideration and contains all of its
non-trivial dynamics. On the other hand, it is essentially smaller than the ini-
tial phase space. In particular, the compactness assumption shows that the higher
modes are indeed suppressed, so even the existence of a global attractor supports
somehow the conjecture of finite-dimensionality.
The ultimate goal of the present survey is to develop a unified approach to
the theory of attractors for dissipative dynamical systems generated by PDEs and
to clarify its role in the rigorous justification of the above-mentioned conjecture
of finite-dimensionality. We emphasize from the very beginning that, although
the study of such systems is also strongly based on the analysis of PDEs (for in-
stance, the well-posedness of the 3D Navier–Stokes problem is one of the most
challenging open problems in the theory of PDEs, and the absence of a reasonable
answer whether and in what sense this system is well posed clearly affects the at-
tractor theory for the Navier–Stokes equations), a more or less detailed exposition
of the modern methods of the analysis of PDEs is far beyond the scope of this
survey, where we concentrate mainly on various dynamical concepts of the theory
of attractors.
The systematic study of the dynamical properties of dissipative PDEs started, to
the best of our knowledge, from the pioneering papers of Foias and Prody [75] and
Ladyzhenskaya [147] and has been motivated by the dream to understand the tur-
bulence using the methods of the theory of dynamical systems for ODEs. In par-
ticular, the existence of an invariant measure for the Navier–Stokes system, as well
as finitely many determining modes, was established in [75] and the object which
coincides with the modern attractor for the Navier–Stokes problem was implicitly in-
troduced in [147]. We mention also [16], where the global attractors were introduced
with applications to delay differential equations. The theory of attractors has been
intensively developing during the last 50 years and many interesting and promis-
ing results have been obtained; see [12], [8], [38], [43], [44], [93], [149], [174], [192],
[215], [223], [243], and the references therein. In particular, one of the main results
of the attractor theory claims that, under the mild assumptions on the dissipative
system generated by an evolutionary PDE in a bounded domain, the corresponding
attractor A has a finite fractal dimension:

dimf (A, Φ) < ∞.

 ATTRACTORS. THEN AND NOW 57

Combining this fact with Mané’s projection theorem, we get a one-to-one home-
omorphic projection of the attractor A onto a finite-dimensional set A ⊂ RN .
Moreover, the reduced dynamics on A is governed by a system of ODEs. Thus,
the existence of a finite-dimensional attractor allows us to build up an inertial form
of the dissipative PDE under consideration and justify somehow the finite-dimen-
sionality conjecture; see [193], [243], and the references therein for more details.
However, this justification is not entirely satisfactory since the inertial form ob-
tained is only Hölder continuous, no matter how smooth the initial PDE is, so,
starting from a smooth dynamical system, we end up with non-smooth equations
where even the uniqueness of a solution may be lost. Any attempts to improve
the regularity of this reduction lead to extra restrictive assumptions on the initial
PDE which are hard to check and which are violated in many interesting exam-
ples. Moreover, as recent examples show, this problem is far from being technical
since, despite the finiteness of the fractal dimension of the global attractor and
the existence of a Hölder continuous inertial form, the associated limit dynamics
may remain, in a sense, infinite-dimensional and demonstrate features which are not
observable in classical dynamics generated by smooth ODEs (like limit cycles with
super-exponential rate of attraction, decaying travelling waves in Fourier space, and
so on); see [64], [133], [243], and the references therein. These examples allow us
to guess that the sharp borderline between finite and infinite-dimensional limiting
dynamics is more related to Lipschitz -continuous inertial forms and inertial man-
ifolds rather than to Hölder-continuous inertial forms and the fractal dimension
of the attractor.
Thus, despite many efforts made, the rigorous interpretation and justification
of the finite-dimensional reduction in dissipative systems remains a mystery and
is one of the most challenging problems in the modern theory of attractors. We
discuss below in more details some of the currently known approaches to tackle this
and related problems.
The survey is organized as follows.
In Section 2 we give a flavour of attractor theory by considering the simplest
low-dimensional examples, where the attractor can be found more or less explicitly,
and we demonstrate the difference between various types of attractors, their depen-
dence on parameters, their dimensions, and so on. We hope that this section will
help the reader with understanding the more general and abstract theory presented
in the next sections.
Section 3 is one of the central sections of the survey, where we develop the
attractor theory in general Hausdorff topological spaces, which, in turn, allows
us to build up a unified approach to different types of attractors. Namely, to
define an attractor, we need to specify first what objects will be attracted by this
attractor. We refer to the system of these objects as a bornology B on the phase space
of the problem, keeping in mind that in the standard theory a (global) attractor
usually attracts bounded subsets of the phase space. Next, we should specify in what
sense (in what topology) the attraction will hold. In other words, we need to specify
the topology on the phase space Φ. Surprisingly, a substantial attractor theory,
which is similar to the standard one, can be constructed in the general situation
58 S. V. ZELIK

where Φ is just a Hausdorff topological space, under some minimal assumptions on

the bornology B.
Although attractors in general topological spaces were studied before (see [38],
[86], [170], and the references therein), many results in Section 3 are hard/impossible
to find in the literature in the necessary generality, so we present more or less
detailed proofs for most of them.
In Section 4 we apply the unified approach presented above to the case where
the PDE under consideration does not possess the unique solvability property or this
uniqueness of a solution is not known yet. One of the most natural approaches to
tackle this type of problems (proposed by Chepyzhov and Vishik in [36], see also [38],
and by Sell [208]) is related to constructing the trajectory dynamical system asso-
ciated with the PDE under consideration and studying its attractors. Roughly
speaking, we replace the initial phase space Φ, where the problem is ill posed and
the corresponding solution semigroup can be defined as a multi-valued semigroup
only, by the new phase space K+ which consists of all positive semi-trajectories
of our PDE satisfying some nice properties. Then, if K+ is chosen in the proper
way, the semigroup Th , h ∈ R+ , of temporal shifts (defined by (Th u)(t) := u(t + h))
acts on K+ and (Th , K+ ) is exactly the trajectory dynamical system associated with
the problem under consideration.
Note that, in the case where the uniqueness property holds, this semigroup is
conjugate to the usual solution semigroup S(t) acting in the initial phase space Φ,
so the theory is consistent. The advantage of this approach is that the construction
of the trajectory dynamical system does not require the uniqueness property to
hold, so we may study its attractors in a usual way, avoiding the use of multi-valued
maps. The only problem is to define a bornology and topology on Φ in the proper
way. In relatively simple situations, both topology and bornology can be lifted from
the initial phase space, but in more complicated cases (like the 3D Navier–Stokes
equations) this does not work. Moreover, there are several alternative inequivalent
ways to do this, so the unified approach introduced in Section 3 works in full strength
here. The purpose of Section 4 is, in particular, to give the comparison of known
alternative approaches to the trajectory attractors using the example of the 3D
Navier–Stokes equations. To the best of our knowledge, this has never been done
before.
Section 5 extends the unified approach to attractors in the non-autonomous
case. We consider the most general case, where phase space for the dynamical pro-
cess U (t, τ ), t > τ , associated with the problem under consideration may depend
on time, so that U (t, τ ) : Φτ → Φt and {Φt }t∈R is a family of Hausdorff topological
spaces, and we use the pullback attraction property to define attractors. In this case
an attractor is understood as a time-dependent set A(t) ⊂ Φt , t ∈ R, the bornol-
ogy B (which is often referred to as a universe in this theory) also consists of time
dependent sets B(t) ⊂ Φt , t ∈ R, and the attraction property is pullback in time,
that is, if you fix t ∈ R and start from B(τ ) ∈ B, τ < t, then the image U (t, τ )B(τ )
is close to A(t) if t − τ is large enough (see Section 5 for the details). The particular
case where Φt are Banach spaces and B consists of uniformly (in time) bounded sets
was considered in [50]; the extension proposed allows us to treat from the unified
 ATTRACTORS. THEN AND NOW 59

point of view also the case of the bornology of tempered sets (which is important
for random attractors), as well as many other interesting examples.
This general theory covers, in particular, the case of cocycles (or skew-products).
We recall that a family of maps Sξ (t) : Φ → Φ, ξ ∈ Ψ, t > 0, is a cocycle over
a group T (h) : Ψ → Ψ, h ∈ R, acting on a topological space Ψ if

Sξ (0) = Id, Sξ (t + h) = ST (h)ξ (t) ◦ Sξ (h), t, h > 0.

These objects are natural for the theory of non-autonomous and random dynamical
systems. Roughly speaking, the underlying group T (h) : Ψ → Ψ describes the evo-
lution of the time-dependent symbol of the non-autonomous PDE in question under
time shifts and the Sξ (t) are the solution operators (from zero moment of time to
time t) of the PDE with the given symbol ξ ∈ Ψ; see [27], [35], [38], [96], [120],
[121], and the references therein. Recall that the dynamical process which corre-
sponds to the symbol ξ can be recovered from the cocycle using the simple formula
Uξ (t, τ ) = ST (τ )ξ (t−τ ), and exactly this relation allows us to extend the theory from
dynamical processes to cocycles. Note also that introducing a Borel probability mea-
sure on Ψ, which is invariant (and usually ergodic) with respect to the group T (h),
allows us to link this theory with the theory of random dynamical systems and their
attractors; see [3], [27], [52], [54], [58], [121], and the references therein.
In Section 5 we also discuss an alternative approach to attractors of non-autono-
mous dynamical systems, which is based on the reduction of the cocycle related to
the PDE under consideration to the autonomous semigroup acting on the extended
phase space Φ × Ψ. This approach leads to an object which is independent of time
and the rate of attraction to it is uniform in time as well as with respect to ξ ∈ Ψ;
for this reason it is referred to as a uniform attractor. We present here the clas-
sical results related to weak uniform attractors and to strong ones (in the case
of translation-compact external forces): see [37], [38] and the references therein,
as well as more recent results concerning strong uniform attractors in the case
of non-translation-compact external forces: see [244] and the references therein.
Section 6, which is the second central section of the survey, discusses the dimen-
sions of attractors. As we have already mentioned, the fact that the attractor A
of the PDE under consideration has a finite fractal dimension allows us to build
up an inertial form for this equation, and this is one of the possible ways to justify
the finite-dimensionality conjecture for the associated limit dynamics, so getting
realistic upper and lower bounds for this dimension is one of the most fashionable
branches of the attractor theory. This activity was originated by the seminal paper
of Mallet-Paret [165] (see also [148]), where a method for estimating the dimension
of negatively invariant sets, which is based on some smoothing/squeezing properties
for differences of solutions, was proposed. In a modern interpretation, the main re-
sult can be formulated as follows: let Φ and Φ1 be two Banach spaces such that Φ1
is compactly embedded in Φ, and let B be a bounded set in Φ1 which is negatively
invariant with respect to some map S. Assume that S enjoys the following kind
of squeezing property:

∥S(u1 ) − S(u2 )∥Φ1 6 κ∥u1 − u2 ∥Φ1 + L∥u1 − u2 ∥Φ ∀ u1 , u2 ∈ B (1.2)

60 S. V. ZELIK

for some κ ∈ [0, 1) and L > 0. Then the fractal dimension of A in Φ1 is finite
(see, for example, [71] or [46]). This result and various generalizations of it to
non-autonomous and random cases are presented in Section 6. Note that this ap-
proach has a tremendous number of applications in the modern theory of attractors
(see, for example, [46], [65], [129], [163], [164], [238], and the references therein);
in particular, most results concerning exponential attractors are strongly based on
this method; see the discussion below.
We also discuss there the volume contraction method for obtaining upper bounds
for the Hausdorff and fractal dimension of an attractor which was proposed by
Douady and Oesterlé [60] (see also [107] and [49] for generalizations to the infinite-
dimensional case). Roughly speaking, the key result here is that if you have a neg-
atively invariant (with respect to some smooth map S) compact set B of a Hilbert
space, such that the infinitesimal k-dimensional volumes in it are contracted by S,
then the dimension of B is less than k. In the case of the Hausdorff dimension this
result was established in [60] and [107] for the finite and infinite-dimensional cases,
respectively. The case of fractal dimension is more delicate and for a long time only
partial results with extra unnecessary assumptions were known (see [38], [49], and
the references therein). The breakthrough there was done by Hunt [105], where
exactly the same result was established for the fractal dimension in the finite-
dimensional case. This result was subsequently extended to the infinite-dimensional
case in [19] under some extra technical assumptions, which were finally removed
in [32]. The advantage of this method is that, combined with the use of the Lieb–
Thirring inequalities presented by Lieb in [155], it gives the best known upper
bounds for the fractal dimension of the attractor of the 2D Navier–Stokes equa-
tions: see, for example, [215], see also [79] and [32] for the best analytic bounds
for this dimension via the Lieb–Thirring inequalities. This makes the volume con-
traction method very popular in attractor theory (despite the fact that it also
has drawbacks and is not applicable in many cases, in particular, when the solu-
tion operators are not differentiable with respect to the initial data; see [179] and
the references therein). We give an exposition of this method in Section 6 too.
We also discuss lower bounds for the dimensions of attractors in that section.
Recall that the most widespread method to get such estimates is based on the fact
that an unstable manifold of any (hyperbolic) equilibrium always belongs to the at-
tractor, so its dimension cannot be smaller than the dimension of this unstable
manifold. This gives us lower bounds for the dimension of the attractor in terms
of the instability indices of equilibria (see [12], [215], and the references therein).
However, this is not enough for sharp lower bounds in some cases and there are
examples where the instability indices of all equilibria remain bounded, but the di-
mension of the attractor tends to infinity as the physical parameter of the system
under consideration tends to zero. For this reason we discuss also a more exotic but
promising alternative method (proposed by Turaev and Zelik [220]), which is based
on homoclinic bifurcation theory. Roughly speaking, this method relies on the fact
that, under some natural assumptions, an invariant torus of very high dimension
(which is restricted by the Lyapunov dimension of the corresponding equilibrium)
may bifurcate from a homoclinic orbit under an appropriate choice of the per-
turbation. This allows us to connect lower bounds with the Lyapunov dimension
 ATTRACTORS. THEN AND NOW 61

(similarly to the upper bounds obtained via the volume contraction method). This
is very important for weakly dissipative equations and gives us sharp upper and
lower bounds of the same order for some class of damped wave equations.
In Section 7, we discuss the theory of inertial manifolds for dissipative PDEs. Re-
call that, by definition, an inertial manifold is a smooth invariant finite-dimensional
submanifold of the phase space of the problem under consideration which is globally
stable and normally hyperbolic. The existence of such a manifold provides a per-
fect justification for the conjecture of finite dimensionality of limit dynamics. In-
deed, the restriction of our system to the inertial manifold gives the desired smooth
inertial form which is governed by a system of ODEs on the base of the mani-
fold. On the other hand, due to normal hyperbolicity we have the exponential
tracking (asymptotic phase) property, which guarantees that any other trajectory
of the PDE under consideration attracts exponentially to the corresponding trajec-
tory on the inertial manifold, so we do not lose any important information about
the limit dynamics after going over to the reduced inertial form on the manifold.
To the best of our knowledge, such an object was first constructed by Mané in [168]
in the case of a reaction-diffusion equation and became popular after [76], where
this result was extended to more general class of equations and was associated with
the dream to understand turbulence (even the attribute “inertial” introduced there
was motivated by the inertial scale in the conventional theory of turbulence; see,
for example, [81] and the references therein).
The classical construction of an inertial manifold requires the PDE under con-
sideration to satisfy rather restrictive conditions, which allow one to present the as-
sociated dynamical system as a slow-fast system and slave the fast modes to slow
ones using the standard methods of hyperbolic theory (see [76], [124], [175], [193],
[195], [199], [209], [215], [243], and the references therein). These conditions are
usually formulated in terms of spectral gap conditions on the leading linear part
of the equation under consideration. On the one hand, these conditions are not
satisfied, for instance, in the case of the 2D Navier–Stokes equations, so the exis-
tence or non-existence of an inertial manifold for the Navier–Stokes equations is one
of the most challenging open problems of the theory. On the other hand, it is also
known that the spectral gap conditions are sharp, at least on the level of abstract
functional models (for example, in the class of abstract semilinear parabolic equa-
tions; see [64], [243]) so a further progress in constructing inertial manifolds beyond
the spectral gap conditions is possible only by exploiting some special properties
of concrete classes of PDEs.
The first result in this direction was obtained by Mallet-Paret and Sell in [166],
where inertial manifolds were constructed in the case of scalar reaction-diffusion
equations in 3D with periodic boundary conditions, by means of the method of
spatial averaging; see also [167] where it was shown that this method does not work
in spaces of dimension higher than three. Taking into account the recent progress
in this area (for instance, the extension of methods of spatial averaging to the 3D
Cahn–Hilliard equation: see [131], various truncated or regularized versions of the
3D Navier–Stokes equations, see [82], [126], [128], the development of the method of
spatio-temporal averaging and its applications to the 3D complex Ginzburg–Landau
equation: see [127], [130]), we give a brief exposition of this method in Section 7.
62 S. V. ZELIK

An alternative method for constructing inertial manifolds is based on transform-

ing the initial PDE or embedding it into a new system of PDEs in such a way that
the obtained new system satisfies the spectral gap conditions. This method was orig-
inally related to an erroneous attempt of Kwak [144] to prove the existence of an in-
ertial manifold for the 2D Navier–Stokes equation; see also [145], [216], and see [134]
for a clarification of the nature of that error. For this reason, the whole method was
forgotten for a long time and was considered as suspicious and potentially erroneous.
The situation has changed recently after the works [126] and [132], where an appro-
priate modification of this method was been used to solve the long-standing open
problem about the existence of inertial manifolds for general 1D reaction-diffusion-
advection systems (see also [230]–[232] for some preliminary results in this direc-
tion). We include a brief exposition of this alternative method in Section 7.
We also discuss the smoothness of inertial manifolds. It is known that, in gen-
eral, even when the initial system is analytic, the inertial manifold related to this
PDE is only C 1+ε -smooth for some small ε > 0 and a further regularity of iner-
tial manifolds requires much stronger versions of spectral gap conditions, which are
not satisfied even in the simplest examples (see [40], [124]). This looks like a big
drawback of the theory since the regularity of an inertial manifold is important
from both the theoretical and applied points of view. Indeed, even the analysis
of simplest bifurcations requires a higher smoothness than C 1+ε (for example, for
an analysis of the Andronov–Hopf bifurcation, we need C 3 ; see, for instance, [210]).
On the other hand, the low regularity of the inertial manifold and the related in-
ertial form prevents us from using higher-order numerical methods. This problem
looked unsolvable for a long time, but as shown very recently (see [136]) it never-
theless can be overcome. Namely, by increasing the dimension of the manifold and
using a clever cut-off procedure (based on Whitney’s extension theorem) we may
kill resonances and other obstacles to the existence of smooth invariant manifolds
and get C k -smooth inertial manifolds for every finite k.
Section 8 of the survey is devoted to exponential attractors. These objects were
introduced in [63] as, in a certain sense, intermediate objects between usual attrac-
tors and inertial manifolds, in order to overcome key drawbacks of attractor theory.
The main of these drawbacks is exactly the slow rate of attraction and, which is even
more important, the fact that it is impossible, in a more or less general situation,
to control this rate of attraction in terms of physical parameters of the PDE under
consideration. This makes the attractor, in a sense, unobservable in experiments:
no matter how long we wait, we can never be sure that we are close to the attrac-
tor. The absence of such a control also leads to the sensitivity of the attractor to
perturbations.
Roughly speaking, the idea of the construction of an exponential attractor, is
to add some special points (for example, metastable states) to the usual attractor
in such a way that, on the one hand, the rate of attraction to the new object becomes
exponential and controllable and, on the other hand, the size of this object does
not grow too much: in particular, it should still have a finite fractal dimension, so
that Mané’s projection theorem still allows us to construct an inertial form on it.
Following [65], the modern theory of exponential attractors is based on the squeez-
ing property (1.2) and its various generalizations (see [9], [44], [65], [68], [69], [72],
 ATTRACTORS. THEN AND NOW 63

[84], [179], and the references therein). Thus, in contrast to inertial manifolds,
exponential attractors are as general as the usual finite-dimensional global attractors
(see also [67], [66] for infinite-dimensional exponential attractors in the case where
the dimension of a usual attractor is infinite).
We do not present the most general conditions of the form (1.2) which guarantee
the existence of exponential attractors (we refer the interested reader to the
surveys [67] and [179]). Instead, we discuss the impact of the theory of exponential
attractors on non-autonomous and random attractors. Our exposition follows
mainly [71] and [211] and is based on the straightforward extensions of (1.2) to
the non-autonomous case.
We recall that the theory of exponential attractors allows us to overcome the extra
drawbacks of the theory of usual attractors which arise when the non-autonomous
case is considered. Namely, again because of the non-controllable and non-uniform
rate of attraction to, say, pullback attractors, we lose in general attraction forward
in time. The situation is a bit better in the case of random attractors, where we
have attraction in probability forward in time, but the lack of the uniformity causes
a lot of difficulties, in particular, it does not allow us to establish the robustness
of random attractors in the random-deterministic limit (see, for example, [42], [54]
and also [24] and the references therein for the contemporary approach to this
problem). In contrast to this, non-autonomous exponential attractors can be
constructed in such a way that the rate of attraction is uniform and exponential
both forward and pullback in time and this remains true for random exponential
attractors as well (see Section 8 for the details).
Since the sensitivity of attractors with respect to perturbations is closely related
to the rate of attraction, we also include some elements of perturbation theory
of attractors to the survey, as well as a theory of so-called regular attractors
in the terminology of Babin and Vishik (see [10]). Such attractors are typical
for systems which possess a global Lyapunov function. Then, under the extra
generic assumption that the number of equilibria R is finite and all of them are
hyperbolic, the attractor consists of a finite union of finite-dimensional unstable
manifolds of these equilibria, and every complete trajectory lying on the attrac-
tor is a heteroclinic orbit between these equilibria. These attractors possess
a number of nice properties, and they have many similarities with exponential
attractors. In particular, the rate of attraction to them is exponential and they are
Hölder continuous with respect to perturbations (including non-autonomous ones);
see [10], [12], [26], [94], [95], [228] and the references therein.
In Section 9 we discuss an alternative approach to the problem of finite-dimen-
sional reduction (which was proposed in [75], see also [147], and was histor-
ically the first) related to determining functionals. By definition, a system
F := {F1 , . . . , FN } of continuous functionals Fi : Φ → R on the phase space Φ
is (asymptotically) determining for a semigroup S(t) : Φ → Φ if for any two
trajectories u1 (t) and u2 (t) of this semigroup, the convergence


lim Fi (u1 (t)) − Fi (u2 (t)) = 0, i = 1, . . . , N,
t→∞

implies that limt→∞ ∥u1 (t)−u2 (t)∥Φ = 0. Thus, the asymptotic behaviour of trajec-
tories of the system under consideration is determined by the behaviour of the finitely
64 S. V. ZELIK

many quantities ξi (t) := Fi (u(t)), i = 1, . . . , N . Note, however, that determining

functionals do not give a true finite-dimensional reduction since the quantities
ξi (t) do not obey a finite system of ODEs. Moreover, they typically satisfy some
kind of a system of ODEs with delay (and realize a Lyapunov–Schmidt reduc-
tion), so that the phase space of the reduced system of ODEs with delay remains
infinite-dimensional, and a true finite-dimensional reduction cannot actually be
constructed in this way. Nevertheless, this topic remains interesting and fashionable
(see [41], [46], [47], [77], [78], [184]) since determining functionals and the related
reduction to delay ODEs have many important applications. For instance, they
are used to establish the controllability of an initially infinite-dimensional system
by finitely many modes (see, for example, [5]), to verify the uniqueness of an in-
variant measure for random/stochasitc PDEs (see, for example, [142]), and in data
assimilation problems where the values of the functionals Fi (u(t)) are interpreted as
the results of observations and where the theory of determining functionals allows
us to build up new methods for recovering the trajectory u(t) from the results
of observations; see [5], [4], [184], and the references therein.
In our exposition we mainly follow the recent paper [114] and try to clarify
the nature of determining functionals as well as the nature of the minimum number
of such functionals (the so-called determining dimension dimdet (S(t)) of the system
under consideration). The trivial lower bound for this dimension is related to
the size of the set of equilibria R, namely its embedding dimension dimemb (R),
that is, the smallest N such that there is a continuous injective map F : R → RN .
Surprisingly, this lower bound is sharp and we have the two-sided estimate

dimemb (R) 6 dimdet (S(t)) 6 dimemb (R) + 1

(see [114] for the details). The proof of this result is based on the famous Takens
delay embedding theorem and its generalizations for Hölder continuous maps (see
[193], [203], [213], and the references therein).
In particular, in the generic case where the set R is finite, almost every continuous
function F : Φ → R is a determining functional for the PDE under consideration,
and such a determining functional can be chosen in the class of polynomial maps
of sufficiently high degree. Thus, in contrast to a widespread paradigm, the minimum
number of determining functionals describes the structure/size of the set of equi-
libria R of the system under consideration and is related neither to the complexity
of the corresponding dynamics on the attractor, nor to its dimension, nor even to
the dissipativity of the PDE under consideration. The corresponding examples,
illustrating this statement, are also presented in Section 9.
Finally, the function spaces most important for our purposes and their properties,
which are used throughout the survey, are collected in the appendix (see Section 10).
To conclude, we note that, because of the restricted size of this survey, it is not
possible to pay the proper attention to all important works in the area of attractors,
so the choice of material is somehow subjective and reflects the personal preferences
of the author, and many interesting areas (like PDEs with delay, attractors
in unbounded domains, approximate inertial manifolds, infinite-dimensional centre
manifolds and the corresponding hyperbolic theory, and so on) are out of this
survey. The author apologizes for the inconvenience caused.
 ATTRACTORS. THEN AND NOW 65

2. Attractors: basic theory and model examples

The aim of this section is to discuss briefly various concepts related to attractors
and illustrate them by simple examples. We assume here that we are given a phase
space Φ which will be a normed space for the moment (the more general situation
where Φ is a Hausdorff topological space will be considered in what follows) and
a semigroup S(t) : Φ → Φ satisfying

S(t + h) = S(t) ◦ S(h), t, h > 0, S(0) = Id . (2.1)

We refer to this semigroup S(t) acting on Φ as a dynamical system and denote it by

(S(t), Φ). Usually, the S(t) are the solution operators of an ODE or an evolutionary
PDE under consideration (which map the initial data to the corresponding solution
at time t). Thus, we assume implicitly that the corresponding initial value problem
is globally solvable and this solution is unique (although this concept may be used
even in cases without uniqueness; see Section 4 for more details). The phase space Φ
is often either some Sobolev space or its subspace endowed with an appropriate
topology; see examples below.
The key concept behind the attractor theory is the concept of an ω-limit set.
Definition 2.1. Let B ⊂ Φ be an arbitrary non-empty subset of Φ. Then
the ω-limit set of B is defined by
\ [ 
ω(B) := S(t)B , (2.2)
T >0 t>T Φ

where [V ]Φ stands for the closure of the set V in the topology of Φ. In the case
of metric spaces, we may give an equivalent, but sometimes more convenient,
sequential definition:
n o
ω(B) = u0 ∈ Φ | ∃un ∈ B, ∃tn → ∞ : u0 = lim S(tn )un . (2.3)
t→∞

In the general case where Φ is not metrizable, these two definitions may produce
different objects.
It is well known that, without further assumptions, an ω-limit set can easily
be empty or, even if it is occasionally not empty, it may not possess important
properties like invariance or/and attraction. In order to preserve them, we need
some kind of compactness, which is the central assumption of the attractor theory.
We summarize the properties of an ω-limit set in the following lemma.
Lemma 2.2. Let the semigroup S(t) be asymptotically compact on B , that is, for
any sequences un ∈ B and tn → ∞ the closure [S(tn )un ]Φ is a compact set in Φ.
Then the following assertions are valid.
1) The set ω(B) is not empty.
2) It attracts the images of the set B in the following sense: for every neighbourhood
O(ω(B)), there exists a moment of time T = T (O) such that

S(t)B ⊂ O(ω(B)), t > T. (2.4)

66 S. V. ZELIK

Assume in addition that the operators S(t) are continuous for every fixed t. Then
the ω-limit set is strictly invariant:

S(t)ω(B) = ω(B), t > 0. (2.5)

This lemma, which is proved in the next section, is the main building block
of attractor theory, and all versions of attractors (known to the author) use
the formula for an ω-limit set implicitly or explicitly. We also emphasize here that
the concept of an ω-limit set requires asymptotic compactness, so if we want to
have an attractor in our phase space, we need to fix the topology in Φ in such a way
that the assumption of asymptotic compactness holds.
The situation is much simpler in the finite-dimensional case Φ = Rn , where
the dissipative estimate:

∥S(t)u0 ∥Φ 6 Q(∥u0 ∥Φ )e−αt + C∗ (2.6)

(where α > 0, C∗ > 0, and Q is some monotone increasing function) is enough to

obtain asymptotic compactness. However, this estimate is not enough in infinite-di-
mensional case, so an extra work is required in order to verify this compactness.
The next principal question is what objects should be attracted by the attractor?
There are different answers to this question. First of all, there are local attractors
which are very natural for the modern theory of dynamical systems (see, for
example, [116]) and which attract only the trajectories starting from some (small)
neighbourhood of the attractor. A more advanced version is a Milnor attractor,
where the attraction property holds for the initial data up to a zero-measure set:
see [176], see also [48], [88], [100], [108], [202] for more delicate versions of attractors.
However, some global versions of attractors are traditionally preferable in the the-
ory of dissipative PDEs, which is partially explained by the ultimate goal of this
theory, to justify finite-dimensional reduction. The most widespread is a global
attractor, which attracts the images of all bounded sets (see [12], [93], [209], [215]).
Definition 2.3. A set A is a global attractor for the dynamical system S(t) :
Φ → Φ if
1) the set A is compact in Φ;
2) it is strictly invariant, that is, S(t)A = A for all t > 0;
3) the set A is an attracting set for the semigroup S(t), that is, for any bounded
B ⊂ Φ and any neighbourhood O(A) of the set A, there exists T = T (O, B) such
that
S(t)B ⊂ O(A)
for all t > T .
Thus, on the one hand a global attractor consists of the complete trajectories
of a dynamical system and contains all of the non-trivial dynamics (due to
the attraction property). On the other hand, due to the compactness property,
it is essentially smaller than the initial (usually infinite-dimensional) phase space,
so the existence of a global attractor gives us already some kind of reduction
of the degrees of freedom for the limit dynamics of the dissipative system under
consideration.
 ATTRACTORS. THEN AND NOW 67

The second possibility is a point attractor, which attracts single trajectories

only (not bounded sets of trajectories). Another possibility is a (Ψ, Φ)-attractor,
which was introduced in [12] and which attracts bounded sets in the space Ψ
in the topology of the space Φ. It may also be natural to consider the attraction
of only compact sets of initial data, and so on. The unified approach to these
attractors will be considered in the next section, and here we state only the simplest
and most popular version of the existence theorem for attractors (see [12], [215] for
more details).
Theorem 2.4. Let the operators S(t) be continuous for every fixed t > 0, and
let the semigroup S(t) possess a compact attracting set B in Φ, that is, for every
bounded set B and every neighbourhood O(B) there exists T = T (O, B) such that

S(t)B ⊂ O(B), t > T.

Then the semigroup S(t) possesses a global attractor A, which is a subset of B.

This theorem is an almost immediate corollary of the properties of ω-limit
sets formulated in Lemma 2.2. Indeed, the desired attractor can be found via
the following two equivalent formulae:
 [ 
A = ω(B) = ω(B) .
B is bounded Φ

However, the second expression is more general and works also for other types
of attractors, without the continuity assumption, and so on. Similar theorems also
hold for other types of attractors; one just needs to understand the attracting set
in the proper way, for example, as point-attracting, (Φ, Ψ)-attracting, and so on.
For instance, a point attractor can be found via
 [ 
Apoint = ω(u0 ) .
u0 ∈Φ Φ

We also mention an important property of global attractors, namely, a representation

formula:
A = K t=0 , (2.7)
where K is the set of all complete bounded orbits of the semigroup S(t):
n o
K := u : R → Φ | S(t)u(h) = u(t + h), t ∈ R+ , h ∈ R, sup ∥u(t)∥Φ 6 Cu .
t∈R

Formula (2.7) is actually one of the main technical tools to work with global
attractors. In particular, as we will see below, it is extremely useful in perturbation
theory, in estimates for the Hausdorff and fractal dimensions, and so on. However,
it may be not true for different types of attractors; for example, it fails in general
for point attractors or in the case where the operators S(t) are not continuous.
We defer the general theory of attractors to the next sections and turn to
examples which illustrate their basic properties.
Example 2.5. Let us start with the first-order scalar ODE
dy
= −y((y − 1)2 − ε), y t=0
= y0 ,
dt
68 S. V. ZELIK

where ε ∈ R is a parameter. First of all, multiplying this equation by y and using

Gronwall’s inequality, it is easy to see that for every value of ε there is a bounded
attracting (and even absorbing) set for the associated semigroup. Thus, we have
a global attractor Agl (ε) as well as a point attractor Apoint (ε), whose structure
depends on ε. For ε < 0 the zero equilibrium is globally exponentially stable, so we
have
Agl (ε) = Apoint (ε) = {0}.
√
For ε > 0 two extra equilibria z = 1 ± ε appear and the global attractor is
the closed interval containing these points:
 √ √ 
Agl (ε) = min{0, 1 − ε }, 1 + ε .

However, since any trajectory still stabilizes to one of these equilibria, the point
attractor consists of these three equilibria:
√
Apoint (ε) = {0, 1 ± ε }.

Even in this simplest example we can already see one of the major drawbacks
of global attractors, namely, they are not robust with respect to perturbations.
Indeed, at the bifurcation point ε = 0 we see a “jump” of the attractor from a single
point {0} to the whole interval [0, 1]. As we will see below, in general, global
attractors are only upper semicontinuous with respect to perturbations and lower
semicontinuity can be proved in exceptional cases only. In our example the point
attractor is also upper semicontinuous with respect to ε but, as next examples show,
even upper semicontinuity may be lost on the level of point attractors.

Example 2.6. More interesting things may happen when we consider equations
on the plane. To plot the phase portraits on a plane is more transparent than
to write down the explicit equations generating the corresponding dynamics, so
we describe below the dynamics by plotting the corresponding pictures. In this
example we consider two dynamical systems (see Fig. 1).
These systems have a saddle point at the origin and two foci, say at (1, 1) and
(−1, −1). In Fig. 1, (a), the foci are stable, so all trajectories converge to one of these
equilibria. Thus, Apoint consists of these equilibria, and the global attractor Agl
contains these equilibria together with two unstable separatrices of the saddle shown
by red lines in the picture. Thus, topologically, the global attractor is still a line
segment, and both the Hausdorff and fractal dimensions of it are equal to one.
In Fig. 1, (b), the foci become unstable via the Andronov–Hopf bifurcation and
two extra limit cycles are born (see, for example, [210]). In this case the point
attractor consists of three equilibria together with the newly born limit cycles. Now
the global attractor contains two extra discs bounded by limit cycles which are
shown by blue lines in the picture (together with the equilibria, limit cycles, and
two unstable separatrices of the saddles, shown by red colour). This follows, for
example, from the fact that the global attractor consists of all complete bounded
orbits. Both the Hausdorff and fractal dimensions are equal to two in this case.
Remarkable here is that the global attractor remains connected, but becomes not
linearly connected after the bifurcation.
 ATTRACTORS. THEN AND NOW 69

Figure 1. (a) The attractor before the Andronov–Hopf bifurcation. (b) The
attractor after the Andronov–Hopf bifurcation.

It is also interesting to look at the instant of bifurcation. At this moment,

the phase portrait remains topologically the same as in Fig. 1, (a), but the foci
dR
become degenerate. In a generic case the radius R of the spirals satisfies ∼ R3 ,
dt
so R ∼ t−1/2 ∼ ϕ−1/2 , where ϕ is an angle of the spiral. An elementary computation
shows that the fractal dimension of such a spiral is 4/3. Thus, since the unstable
separatrices of the saddle form such spirals, we have
4
dimf (Agl ) = ̸= dimH (Agl ) = 1 (2.8)
3
(the Hausdorff dimension is one since the attractor is a countable union of 1D
segments). This is the simplest mechanism which generates global attractors
with non-integer fractal dimension, as well as with inequal Hausdorff and fractal
dimensions. We also mention that the Lyapunov dimension of the attractor at
the instant of bifurcation is obviously equal to two.
Example 2.7. One more non-trivial example of 2D dynamics (introduced
in [114]) is given in Fig. 2.

Figure 2

This phase portrait consists of a saddle-node at the origin (0, 0) glued with
the disc {(x, y) ∈ R2 : x2 + (y − 1)2 6 1} filled by homoclinic orbits to the origin.
70 S. V. ZELIK

The key feature of this dynamical system is that the ω-limit set of any single
trajectory coincides with the origin, and therefore the point attractor Apoint = {0}
is trivial, but the global attractor

Agl = {(x, y) ∈ R2 : x2 + (y − 1)2 6 1}

is not trivial. Note that this phase portrait is extremely degenerate and an arbitrarily
small perturbation will destroy homoclinic orbits and produce a “big” limit cycle.
Thus, in contrast to the global attractor, the point attractor is in general not upper
semicontinuous with respect to perturbations, and this is one more reason why
the global attractor looks preferable.
Example 2.8. The situation becomes much more interesting when the dimension
of the phase space is larger than two, since we do not have the Poincaré–Bendixon
theorem any longer and more complicated structures than equilibria, limit cycles and
homo/heteroclinic connections between them may appear. One of the most popular
3D model examples here is the Lorenz attractor (see [156]), which illustrates
the possibility of a chaotic “unpredictable” behaviour in deterministic systems,
the so-called deterministic chaos. It is strongly believed that similar effects are
responsible for the complicated behaviour of more realistic systems (including PDEs)
arising, for example, in hydrodynamics, weather prediction, chemical reactions, and
so on.
The Lorenz system consists of the following equations:
dx dy dz
= σ(y − x), = x(ρ − z) − y, = xy − βz (2.9)
dt dt dt
and the standard choice of the parameters is σ = 10, β = 8/3, and ρ = 28. The
corresponding attractor is shown in Fig. 3 (the attractor may be essentially different
for other values of the parameters).

Figure 3. The Lorenz attractor: numeric simulation with standard parameters.

The structure of the Lorenz attractor is well understood nowadays, but this
theory is far beyond the scope of this survey, so we restrict ourselves to mentioning
a few of interesting properties of this object. Further details can be found
in [1], [87], [91], [92], [103], [118], [15], [189].
It is known that the Lorenz system is dissipative and possesses an absorbing
ellipsoid B ⊂ R3 , which is semi-invariant (see, for example, [215]) and, therefore
 ATTRACTORS. THEN AND NOW 71

the global attractor Agl of (2.9) exists. It is a fractal set whose Hausdorff and
fractal dimensions are strictly between 2 and 3:

2 < dimH (A) 6 dimf (A) < 3

(at least for the standard values of the parameters; see [183]). In particular,
the upper bounds follow from the volume contraction method and the explicit
formula for the Lyapunov dimension of this attractor:

2(σ + β + 1)
dimL (A) = 3 − p ,
σ+1+ (σ − 1)2 + 4σρ

which holds for all values of the parameters for which all three equilibria are
hyperbolic (see [151]).
It is worth emphasizing that this global attractor Agl does not coincide with
the classical Lorenz attractor Alor or geometric Lorenz attractor, whose existence
was established by Tucker using interval arithmetics and the computer assistant
proof (see [218]). Namely, he verified the existence of a compact semi-invariant
domain U in R3 (inside of the absorbing ellipsoid B). This domain looks like
a solid double torus which contains the zero saddle equilibrium C0 , but does
not contain the other two saddle-foci C± . The local attractor of this domain U
is exactly the Lorenz attractor Alor , which possesses a geometric description,
proposed in [1], [91], in terms of iterations of a 1D discontinuous Poincaré map, and
the main achievement of [218] was the rigorous numerical verification of the fact
that the Lorenz system with the classical values of the parameters satisfies all
the hypotheses stated in the geometric model. In particular, it follows from
the assumptions verified that Alor is topologically transitive, pseudo-hyperbolic,
possesses an invariant measure with nice properties (a Sinai–Ruelle–Bowen measure),
and the periodic orbits are dense in it.
On the other hand there is a strong numerical evidence that there are no complete
bounded trajectories belonging to the domain B \ U in the Lorenz system with
the standard parameters, although, to the best of our knowledge, this has not been
verified rigorously yet on the level of computer assistant proofs. If we believe in this
fact, then the relations between global, Lorenz, and point attractors for the Lorenz
system are given by

Apoint = Alor ∪ C+ ∪ C− , Agl = Alor ∪ M+ (C+ ) ∪ M+ (C− ),

where M+ (C± ) are the two-dimensional unstable manifolds of the saddle-foci C±

and all unions are disjoint.
Example 2.9. We now turn to examples of PDEs and start with the heat
equation in Rd :
∂t u = ∆x u, u t=0 = u0 ∈ Φ = L2 (Rd ). (2.10)
This problem is simple enough to be solved explicitly:

|x − y|2
Z  
1
u(t, x) = K(x, y)u0 (y) dy, K(x, y) = exp −
y∈Rd (4πt)d/2 4t
72 S. V. ZELIK

and from this formula we see that the solution semigroup S(t) : Φ → Φ is well
defined. Moreover, it is not difficult to prove that u(t) → 0 in Φ as t → ∞ for every
fixed u0 ∈ Φ. Thus, the point attractor exists and consists of the zero equilibrium:

Apoint = {0}.

On the other hand the global attractor Agl does not exist here. Indeed, since
the equation under consideration is linear, the existence of a global attractor implies
the exponential stability of the equilibrium, which is clearly not the case for (2.10).
There are several possibilities to extend the theory of global attractors to this
case based on different classes of “bounded” sets which must be attracted to it
or/and different topologies of the phase space. First of all, we may restrict our
attention to one-point sets, then we already have the attractor. However, doing
that, we lose a lot of information (clearly the attraction is actually stronger).
The second possibility would be to restrict the attraction property to compact
sets only. This is probably the most appropriate choice for equation (2.10) since we
have indeed the uniformity of attraction for compact sets only.
Other possibilities are related to the change of the topology of attraction. Indeed,
the attraction of all bounded sets of Φ = L2 (Rd ) holds if we endow the phase space Φ
with the topology induced by the embedding Φ ⊂ Φloc := L2loc (Rd ) (again, this is
a straightforward corollary of the explicit formula for solutions). This corresponds to
the locally compact attractor, which attracts bounded sets in the local topology only
and which is typical for PDEs in unbounded domains (see [179] and the references
therein for more details).
Alternatively, we may endow the space Φ = L2 (Rd ) with the weak topology.
Then all bounded sets are attracted in this topology. In all these cases the attractor
consists of the single point {0}.

Example 2.10. We complete this section by the model example of a nonlinear

heat equation in a bounded domain, which demonstrates the standard way how
the abstract result of Theorem 2.4 is used to verify the existence of an attractor for
nonlinear PDEs. Namely, we consider the following problem:

∂t u = ∂x2 u + au − u3 , x ∈ (0, π), u x=0,π

= 0, u t=0
= u0 (2.11)

in the interval Ω = (0, π) with Dirichlet boundary conditions. Here a > 0 is a fixed
parameter. First we recall the basic dissipative energy estimate in the phase space
Φ = L2 (Ω). In order to avoid technicalities, we give below only the formal derivation
of the required estimates without its justification (and even without formalizing
what is a weak solution of (2.11)). The skipped details can be found, for example,
in [12], [215].
Indeed, multiplying (2.11) by u and integrating with respect to x we have

1 d
∥u(t)∥2L2 + ∥∂x u(t)∥2L2 + ∥u(t)∥4L4 = a∥u(t)∥2L2 .
2 dt
 ATTRACTORS. THEN AND NOW 73

Using Poincaré’s inequality ∥∂x u∥2L2 > α∥u∥2L2 , together with the obvious estimate
a∥u∥2L2 6 ∥u∥4L4 + Ca2 , we arrive at

d
∥u(t)∥2L2 + α∥u(t)∥2L2 + ∥∇x u(t)∥2L2 6 Ca2 , (2.12)
dt
where C is a constant independent of a, and, after integration with respect to time,
we obtain the desired dissipative estimate in Φ:

∥u(t)∥2L2 6 ∥u(0)∥2L2 e−αt + C∗ a2 , (2.13)

where C∗ = C/α. The global existence of a solution is straightforward here

and can be verified, for example, by using the Galerkin approximations (see, for
instance, [12], [215] for more details). Let us verify uniqueness. Indeed, let u1 (t)
and u2 (t) be two solutions of (2.11), and let v(t) = u1 (t) − u2 (t). Then this function
solves

∂t v − ∂x2 v = av − (u21 + u1 u2 + u22 )v, v t=0

= u1 (0) − u2 (0). (2.14)

Multiplying this equation by v(t) and using that u21 + u1 u2 + u22 > 0, we have

d
∥v(t)∥2L2 + 2∥∂x v(t)∥2L2 6 2a∥v(t)∥2L2
dt
and
Z t
∥u1 (t) − u2 (t)∥2L2 + 2 e2a(t−s) ∥∂x v(s)∥2L2 ds 6 e2at ∥u1 (0) − u2 (0)∥2L2 , (2.15)
0

which proves uniqueness. Thus, we have constructed a solution semigroup

S(t) : Φ → Φ associated with equation (2.11). This semigroup possesses a dis-
sipative estimate (2.6) in the phase space Φ = L2 (Ω) and, for this reason, the set

B := {u0 ∈ Φ : ∥u0 ∥2Φ 6 2C∗ a2 }

is a bounded attracting (and even absorbing) set for this semigroup. However, this
is still not enough to verify the existence of a global attractor since the set B is
not compact in Φ. To overcome this difficulty we must use the parabolic smoothing
property. To obtain the desired smoothing property, we multiply (2.11) by t ∂x2 u
and integrate with respect to x. This gives
 
1 d 2 2 2 2 1
∥∂x u(t)∥2L2


t∥∂x u(t)∥L2 = −t∥∂x u(t)∥L2 − 3t∥u∂x u∥L2 + at +
2 dt 2
 
1
6 at + ∥∂x u(t)∥2L2 .
2

Integrating this estimate with respect to t ∈ [0, 1], we arrive at

Z 1
∥∂x u(1)∥2L2 6 (2a + 1) ∥∂x u(s)∥2L2 ds.
0
74 S. V. ZELIK

Finally, we estimate the right-hand side of this inequality by integrating esti-

mate (2.12) with respect to t ∈ [0, 1] and using the dissipative estimate (2.13). This
gives
∥∂x u(1)∥2L2 6 C(a + 1)∥u(0)∥2L2 ,
where the constant C is independent of the parameter a. Shifting time, we also see
that
∥∂x u(t + 1)∥2L2 6 C(a + 1)∥u(t)∥2L2 .
Combining this estimate with (2.13), we see that the set

B1 := u0 ∈ H01 (Ω) : ∥∂x u0 ∥2L2 6 2CC∗ a2 (a + 1)


(2.16)

is also an attracting (and even absorbing) set for the semigroup S(t). The difference
is that this set is compact in Φ. The continuity of the operators S(t) with
respect to the initial data follows from estimate (2.15). Thus, all the assumptions
of Theorem 2.4 are verified and the existence of a global attractor Agl ⊂ Φ for (2.11)
is established.
The next question is the structure of the global attractor Agl . The key role
in its further investigation is played by the global Lyapunov function. Indeed,
multiplying (2.11) by ∂t u and integrating with respect to x we derive that
 
d d 1 1 a
L(u(t)) := ∥∂x u(t)∥2L2 + ∥u(t)∥4L4 − ∥u(t)∥2L2 = −∥∂t u(t)∥2L2
dt dt 2 4 2

and see that the function u → L(u) is strictly decreasing along non-equilibrium
trajectories. This allows us to conclude that any trajectory of (2.11) tends to
the set R of equilibria as t → ∞. The equilibria u0 ∈ R solve a second-order scalar
ODE, and their structure can completely be understood. In particular, it is known
that new equilibria can bifurcate from the zero equilibrium only, and all non-zero
equilibria remain hyperbolic for all values of the parameter a. This fact shows that
√
there are exactly N = 2⌈ a ⌉ − 1 different equilibria in Agl for a > 1 (the case
a < 1 is not interesting since then we have a single globally exponentially stable zero
equilibrium). Thus, from this fact and the existence of a global Lyapunov function
we conclude that any trajectory of the equation under consideration stabilizes, as
t → ∞, to one of these equilibria:

Apoint = R = {u1 , u2 , . . . , uN }

and the global attractor Agl consists of the equilibria and heteroclinic orbits
connecting them. A more detailed analysis based on the Morse–Smale theory allows
us to show the robustness of the dynamics on the attractor for a ∈ (n2 , (n + 1)2 ),
n ∈ N, as well as to determine what equilibria are connected by heteroclinic orbits
and to compute the dimensions of the corresponding sets of heteroclinic orbits
(see [12], [73], [97] for more details).
To conclude we note that (2.11) is a rare exception in attractor theory, where
the structure of the global attractor can be understood fully. This is possible
thanks to three different nice properties of this equation. The first is the existence
of the global Lyapunov function L discussed above. The second is the existence
 ATTRACTORS. THEN AND NOW 75

of another type of (discrete) Lyapunov function, namely, the number Z(u(t)) of zeros
of the profile x → u(t, x) which is a non-increasing function of time. Exactly this
discrete Lyapunov function is responsible for the transversality of the stable and
unstable manifolds of equilibria and the Morse–Smale property. And the third
property is that the equation under consideration is order preserving with respect
to the cone of non-negative functions. This allows us to apply the Perron–Frobenius
theory and simplify essentially the analysis of the spectral properties of equilibria
(see [12], [29], [97], [138] for more details). In contrast to this, the structure
of the global attractor remains a mystery in a more or less general situation (say, for
the Navier–Stokes equations) and our knowledge is limited to the general facts that
it is compact, connected, consists of bounded trajectories, and has finite Hausdorff
and fractal dimensions.

3. Attractors: a unified approach

In this section we present a unified approach to various types of attractors. As

we have seen in the previous section, we need to specify two major things:
1) bornology: what sets will be attracted by the attractor (we will call such sets
“bounded” which explains the denomination bornology);
2) topology: in what sense attraction will hold.
We assume here that our phase space Φ is a Hausdorff topological space. On
the one hand, the main results of the theory discussed in the previous section can
easily be extended to general Hausdorff topological spaces and, on the other hand,
this extension makes the theory more convenient and more elegant, especially when
attraction in a weak or weak-star topology is considered.
Then we fix a family of sets B ⊂ 2Φ , which will be referred to as a bornology
(or a bounded structure) on the phase space Φ. Recall that usually the definition
of a bounded structure includes some assumptions on B, such as S stability with
respect to inclusions and finite unions, as well as the fact that Φ = B (see, for
B∈B
example, [102]). Since the attraction of a set B automatically implies the attraction
of all of its subsets and the same is true for finite unions, the first two assumptions
are actually not restrictive (we can always extend our bornology to satisfy them
without changing the attraction properties). However, the third assumption is
a big restriction since it excludes local attractors, so we prefer not to impose it.
We also do not impose the first two assumptions since they can be satisfied by
the straightforward extension mentioned above, so, in our theory, B is an arbitrary
family of non-empty sets.
Finally, we are given a semigroup S(t) : Φ → Φ acting on our phase space, which
is not assumed to be continuous, so we need some modifications in the theory. We
start with the standard definitions.

Definition 3.1. A set B ⊂ Φ is an absorbing (B-absorbing) set for the semi-

group S(t) if, for every B ∈ B, there exists time T = T (B) such that

S(t)B ⊂ B, t > T.
76 S. V. ZELIK

A set B is an attracting (B-attracting) set for the semigroup S(t) if, for every B ∈ B
and every neighbourhood O(B) of B, there exists T = T (B, O) such that

S(t)B ⊂ O(B), t > T.

We now turn to attractors. The key difference here is that the invariance
of an attractor A requires some kind of continuity of S(t) which we do not assume,
so we need to replace invariance by minimality.
Definition 3.2. A set A ⊂ Φ is an attractor (a B-attractor) of a semigroup S(t)
acting on the topological space Φ if the following assumptions are satisfied:
1) A is a compact set in Φ;
2) A is an attracting set (a B-attracting set) for the semigroup S(t);
3) A is a minimal set (with respect to inclusion) which satisfies properties 1)
and 2).
We are now ready to state the main result of this section, which is related to
the existence of an attractor.
Theorem 3.3. Let S(t) : Φ → Φ be a semigroup acting on a Hausdorff topological
space Φ endowed with a bornology B. Assume that S(t) possesses a compact
B-attracting set B. Then there exists a B-attractor A, which can be obtained as
follows:  [ 
A= ω(B) , (3.1)
B∈B Φ

where ω(B) is the ω-limit set of B defined by (2.2).

Proof. Although the proof looks more or less standard and various particular
cases of this theorem can be found in the literature (see [38], [170]), for the conve-
nience of the reader we give all details here. As usual, we need to verify the analogue
of (2.2) for our case.
Step 1: the set ω(B) is not empty for all non-empty B ∈ B. Let
[ 
KT := S(t)B , UT = Φ \ KT .
t>T Φ
T
Assume that ω(B) = KT = ∅. Then {UT }T >0 , is an open covering of Φ and,
T >0
in particular, an open covering of the compact set B. Since the UT are non-empty
and nested, there exists T > 0 such that B ⊂ UT , and therefore
[ 
KT ∩ B = S(t)B ∩ B = ∅.
t>T Φ

We claim that this contradicts the attraction property. Indeed, let x ∈ B be

S
arbitrary. Then there exists a neighbourhood Ux of x such that Ux ∩ S(t)B = ∅.
S t>T
Let U = Ux . Then on the one hand U is a neighbourhood of B and on the other
x∈B
[
U∩ S(t)B = ∅.
t>T
 ATTRACTORS. THEN AND NOW 77

Therefore, U ∩ S(t)B = ∅ for all t > T and B is not an attracting set for B. This
contradiction proves that ω(B) is not empty.
Step 2: ω(B) ⊂ B. Indeed, let x ∈ ω(B) \ B. Then, since Φ is Hausdorff and B
is compact, there exist neighbourhoods UB and Ux of the set B and the point x,
respectively, such that UB ∩ Ux = ∅. Then we know from the attraction property
that S(t)B ⊂ UB for all t > T . On theSother hand we know that x ∈ KT for
all T > 0. From this we conclude that S(t)B ∩ Ux ̸= ∅ for all T > 0. This
t>T
contradicts the fact that UB ∩ Ux = ∅ and proves the statement.
Step 3: ω(B) attracts B. Note that ω(B) is compact as a closed subset
of a compact set. Let U be an arbitrary neighbourhood of ω(B). Consider the
closed sets
CT := KT \ U = KT ∩ (Φ \ U ).
T T
T
Then CT = KT \ U = ∅ (since KT = ω(B) ⊂ U ). Thus, the sets
T >0 T >0 T >0
VT := Φ \ CT cover the compact set B. Extracting a finite covering and using
the fact that the sets VT are increasing, we get that B ⊂ VT for some T . Thus,
CT ∩ B = ∅. Since CT is closed and B is compact, there is a neighbourhood UB
of B such that CT ∩ UB = ∅. In particular, (S(t)B \ U ) ∩ UB = ∅ for all t > T
(since S(t)B ⊂ KT for t > T ). Since S(t)B ⊂ UB for a sufficiently large t due to
the attraction property, we must have S(t)B ⊂ U if t is large enough, which proves
the attraction property.
Step 4: the attractor. We define it by formula (3.1). Then, since ω(B) ⊂ B
for all B ∈ B, A is a closed subset of the compact set B, so it is compact in Φ.
The attraction property is also obvious since, by definition, A contains ω(B) for
every B ∈ B. Thus, we only need to check minimality. To this end, it is enough to
prove that ω(B) is a minimal set which attracts B. Indeed, let Ω(B) be another
compact attracting set for B. Then, arguing as at Step 2, we get that ω(B) ⊂ Ω(B)
and ω(B) is indeed minimal. This finishes the proof of Theorem 3.3.
Let us now discuss the invariance of the attractor A. This requires some kind
of continuity of the operators S(t).
Proposition 3.4. Let the assumptions of Theorem 3.3 hold, and let, in addition,
the operators S(t) : Φ → Φ be continuous for every fixed t. Then the attractor A is
strictly invariant with respect to S(t), that is,

S(t)A = A

for all t > 0.

Proof. First we prove that A ⊂ S(t)A. Indeed, S(t)A is compact as a continuous
image of a compact set. Let us prove that S(t)A is an attracting set. Let U be
a neighbourhood of S(t)A. Then, by continuity, V := S(t)−1 (U ) is a neighbourhood
of A. Due to the attraction property for A, for any B ∈ B there exists time
T = T (B, V ) such that S(h)B ⊂ V for all h > T . Then S(h)B ⊂ U for all h > T + t
and the attraction property is proved. Thus, by the minimality of A we have
the desired inclusion.
78 S. V. ZELIK

Let us prove the reverse inclusion. Indeed, for any B ∈ B, we have

\ [  \  [ 
S(t)ω(B) = S(t) S(h)B ⊂ S(t) S(h)B
T >0 h>T Φ T >0 h>T Φ
\ [  \ [ 
⊂ S(t) S(h)B = S(h)B = ω(B).
T >0 h>T Φ T >0 h>T +t Φ

Using again that S(t)[A]Φ ⊂ [S(t)A]Φ due to continuity, in combination with (3.1),
we conclude that S(t)A ⊂ A and finish the proof of the proposition.
The next example shows that the continuity assumption cannot be omitted.
Example 3.5. Let Φ := (−∞, 0]∪{1} with the standard topology and bornology
induced by the embedding in R, and let
(
e−t x, x < 0,
S(t)x :=
1, x > 0.
Then, obviously, A = {0, 1}. However, ω(A) = {1} ̸= A and S(t)A = {1} ̸= A.
This example shows that we cannot expect the strict invariance of the attractor
without continuity assumptions. Moreover, if B is a compact attracting set of S(t),
then in general we cannot expect that
A = ω(B). (3.2)
Indeed, in our case we may take B = A. In addition, this example shows that
the condition that the S(t) : B → Φ are continuous is not enough to have invariance
(again, the S(t) are continuous on B := A in our example) and we really need to
assume continuity on a larger set than B.
We now discuss the validity of the representation formula (2.7). We define
a bounded complete trajectory u : R → Φ as a full trajectory of S(t) such that
[
Bu := u(t) ∈ B,
t∈R

and we consider the set K of all complete bounded trajectories with respect to
a given bornology B.
It is immediate to see from the attraction property that any complete bounded
trajectory belongs to the attractor A, so we have
K t=0
⊂ A. (3.3)
However, the opposite embedding is not true in general and requires some extra
assumptions. In particular, we need some kind of continuity in order to get
the invariance of the attractor (see Example 3.5, where there are no complete
bounded trajectories passing through the point 0 ∈ A). This continuity is still not
enough to get the representation formula (2.7). For instance, if the bornology B
consists of one-point sets only, then complete bounded trajectories are equilibria
only, so (2.7) holds for the corresponding point attractor Apoint if and only if it
coincides with the set of equilibria (Apoint = R). The next proposition gives a useful
sufficient condition for the validity of (2.7).
 ATTRACTORS. THEN AND NOW 79

Proposition 3.6. Let Φ be a Hausdorff topological space with a bornology B which

is stable with respect to inclusions. Also let S(t) : Φ → Φ be a continuous semigroup
in Φ (that is, all maps S(t), t > 0, are continuous) which possesses a compact
bounded attracting set B. Then this semigroup possesses a B-attractor A ⊂ B,
which is strictly invariant and is generated by all complete bounded trajectories (that
is, (2.7) holds).
Proof. Indeed, the existence of the attractor A follows from Theorem 3.3. Its
strict invariance was proved in Proposition 3.4, the embedding (3.3) follows from
the attraction property, so we only need to verify that every point u0 ∈ A belongs to
some complete bounded trajectory. This trajectory can be constructed as follows:
for t > 0, we just define u(t) := S(t)u0 ∈ A. By the invariance established, we
have u(t) ∈ A for all t > 0. Let us now define u(−n) for n ∈ N using induction.
Indeed, by the strict invariance, there exists u(−1) ∈ A such that S(1)u(−1) = u0 ;
then there exists u(−2) ∈ A such that S(1)u(−2) = u(−1), and so on. Finally, for
t ∈ [−n, −n + 1], we set u(t) := S(t + n)u(−n). Then, by invariance u(t) ∈ A for
all t ∈ R and by construction u(0) = u0 . Since Bu ∈ A and A is bounded, using
the stability of B with respect to inclusion, we conclude that Bu ∈ B and u ∈ K.
This proves the proposition.
Remark 3.7. As we have seen, the continuity assumption for the semigroup S(t)
cannot be omitted. However, there are several possibilities to weaken it. One of them
is to assume that the graph of the map S(t) is closed in Φ × Φ for every fixed t > 0
(see [12], [185]). Another possibility is to verify this continuity not in the whole
phase space Φ, but on the absorbing set B0 only. Namely, it is not difficult to
verify that the strict invariance of the attractor remains true if S(t) : B0 → Φ is
continuous for every fixed t. This is especially useful in the case where Φ is endowed
with a weak or weak-star topology. Although this topology is not metrizable on
the whole space Φ, its restriction to a properly chosen absorbing set is very often
metrizable. This allows us to verify only the sequential continuity of the maps S(t).
This may be a great simplification since the topological continuity of the nonlinear
maps S(t) is much harder to check than sequential continuity (at least when we are
speaking about weak or weak-star topologies).
We also recall that in a general topological space compactness and sequential
compactness are different (even unrelated), so the two corresponding types of at-
tractors can be considered, topological and sequential ones, which may also be
completely different. We will not go into the details here (see [137] for these details),
but only mention two natural cases where they coincide. The first case is when we
have a metrizable absorbing set and the second, more interesting case is when Φ is
a Banach space endowed with the weak topology. Then compactness and sequential
compactness coincide due to the Eberlein–Smulian theory.
Remark 3.8. To conclude this section, we also mention that in many cases
the key dissipative estimates give us the existence of an absorbing set B, which is
‘bounded’ in Φ in a certain sense and for this reason is often a complete metrizable
space even when the original phase space Φ is not metrizable. Then the verification
of the existence of a compact attracting set is reduced to checking so-called
asymptotic compactness, namely, we need to check that, for any sequences tn → ∞
80 S. V. ZELIK

and un ∈ B, the sequence {S(tn )un }∞ n=1 is precompact. Actually, it is not difficult to
check that under the above assumptions this property is equivalent to the existence
of a compact attracting set, but the use of asymptotic compactness arguments
allows us to avoid constructing this compact attracting set explicitly, which is useful
in many applications.
At the moment there are many powerful methods for verifying asymptotic com-
pactness, for example, ones based on the Kuratowski measure of non-compactness,
or on the energy method, or on compensated compactness arguments. We refer
the interested reader to [44], [93], [181] (see also the references therein) for more
details.

4. Trajectory attractors

In this section we demonstrate how attractor theory can be applied to equations

without the uniqueness of solutions of the corresponding initial value problem.
One possible approach to handle the non-uniqueness problem is to consider semi-
groups of multivalued maps and extend the concept of an attractor to such
semigroups (see [11], [6], [13], [171]), however, there exists an alternative, more
elegant approach which was developed in [36] and [208] for the study of the 3D
Navier–Stokes equations and which we will use here (see also [225] for the case
of elliptic equations). Under this approach, one constructs a trajectory dynamical
system associated with the problem under consideration and then applies the usual
theory of attractors to this dynamical system. We illustrate below the main idea
by a number of examples.
4.1. ODEs and reaction-diffusion equations. We start with considering
a relatively simple case of ODEs and reaction-diffusion systems without uniqueness.
Example 4.1. Consider a system of ODEs

u′ + f (u) = 0, u = (u1 , . . . , uN ), u t=0

= u0 . (4.1)

We assume that the nonlinearity f is locally Lipschitz and satisfies the standard
dissipativity assumption:

f (u).u > −C + α|u|2 , u ∈ RN , (4.2)

for some positive constants C and α. (Here and in what follows u.v denotes the
standard dot product in RN .) Then problem (4.1) is locally uniquely solvable for
any u0 ∈ RN . Moreover, taking a scalar product of (4.1) with u we arrive at
the inequality
1 d
|u(t)|2 + α|u(t)|2 6 C
2 dt
and integrating this inequality, we arrive at the key dissipative estimate
C
|u(t)|2 6 |u0 |2 e−2αt + . (4.3)
α
This estimate gives us the global well-posedness of problem (4.1) for any u0 ∈ RN , and
therefore this problem generates a dissipative semigroup S(t) : Φ → Φ in the phase
 ATTRACTORS. THEN AND NOW 81

space Φ = RN . Note that this semigroup is not necessarily a group since

the solutions u(t) may blow up for negative times.
To construct an attractor for this semigroup we fix the standard bornology B
in RN which consists of usual bounded sets in RN . Then the ball
 
N 2 2C
B := u ∈ R : |u| 6
α

is a compact bounded absorbing set for the semigroup S(t). Moreover, the maps S(t)
are continuous with respect to the initial data. Thus, according to the general
theory, there exists a global attractor A ⊂ Φ.
To explain the idea of trajectory attractors, we consider the set K+ ⊂ Cloc (R+ , Φ)
which consists of all semi-trajectories u(t), t > 0, of the solution semigroup S(t)
constructed that start from all u0 ∈ Φ. Then the semigroup of time shifts T (h),
h > 0, acts on this set:

(T (h)u)(t) := u(t + h), t, h ∈ R+ , T (h)K+ ⊂ K+ . (4.4)

Thus, we have defined a dynamical system T (h) on K+ . This dynamical system is

called a trajectory dynamical system associated with equation (4.1), and the set K+
is referred to as the trajectory phase space.
Note that in the case where the uniqueness theorem holds, the trajectory
dynamical system constructed is topologically conjugate to the semigroup (S(t), Φ)
acting on the usual phase space Φ. Indeed, let S : Φ → K+ be the solution operator
u0 → u( · ) associated with equation (4.1). Then this map is one-to-one with
the inverse map defined via S−1 u := u(0), and it is a homeomorphism if we endow
the set K+ with the topology induced by the embedding in Cloc (R+ , Φ). Moreover,
we have the obvious relation

S(t) = S−1 ◦ T (t) ◦ S,

which shows that the trajectory dynamical system (T (h), K+ ) is equivalent to

the classical one (S(t), Φ). Finally, if we define a bornology Btr on K+ by

Btr = B ⊂ K+ : B t=0 is bounded in Φ , (4.5)

then we also have a one-to-one correspondence between the bounded sets in K+

and in Φ. Thus, since (S(t), Φ) possesses a global attractor A (as explained above),
the trajectory dynamical system (T (h), K+ ) also possesses an attractor Atr := SA,
which is called the trajectory attractor associated with equation (4.1). This attractor
is strictly invariant with respect to the shift semigroup and consists of complete
bounded trajectories:
Atr = K t>0 . (4.6)
Note also that, due to the dissipative estimate (4.3), the bornology Btr consists
of all bounded sets in Cloc (R, Φ) ∩ K+ (or even Cb (R, Φ) ∩ K+ ).
Up to the moment, the above constructions look as a tautology; however,
the construction of the trajectory dynamical system (T (h), K+ ) does not require
the uniqueness theorem to hold, so it can be extended to cases where the uniqueness
82 S. V. ZELIK

theorem is either violated or not known. For instance, in the example under
consideration we may relax the assumption of the Lipschitz continuity of f and
assume that f is only continuous, but satisfies the dissipativity assumption (4.2).
Then we still have the existence of a solution for any u0 ∈ Φ and the validity
of dissipative estimate (4.3), so we may construct the trajectory dynamical system
(T (h), K+ ) and the bornology Btr in the same way as before, and we may speak about
the (trajectory) attractor Atr of this system despite the fact that the uniqueness
theorem does not hold here in general and the classical dynamical system (S(t), Φ)
is not even properly defined.
Proposition 4.2. Let the function f be continuous and satisfy the dissipativity
assumption (4.2). Then the trajectory dynamical system (T (h), K+ ) endowed with
the topology of Cloc (R, Φ) and the bornology Btr described above possesses an attractor
Atr ⊂ K+ which is strictly invariant and consists of all complete bounded trajectories
of (4.1), that is, (4.6) holds.
Proof. In contrast to the case with uniqueness, we are now unable to get the re-
sult by lifting the global attractor A to the trajectory phase space (the dynamical
system (S(t), Φ) does not exist any longer), so we will use Proposition 3.6 instead.
Indeed, the shift maps T (h) are continuous in the topology of Cloc (R, Φ) and the dis-
sipative estimate (4.3) gives us a bounded absorbing set B for the semigroup T (h).
Namely, we may take
 
2C
B := u ∈ K+ : ∥u∥Cb (R,Φ) 6 .
α

We claim that this set is actually compact in K+ ⊂ Cloc (R+ , Φ). Indeed, from
equation (4.1), we conclude that u′ is uniformly bounded with respect to u ∈ B and
Arzelà’s theorem gives the desired compactness. Thus, we have proved the existence
of a trajectory attractor Atr ⊂ B, which is strictly invariant with respect to T (h)
and is generated by all complete bounded trajectories of (4.1) (being pedantic, we
have verified only that it is generated by all complete bounded trajectories of T (h),
but it is immediate to see that there is a one-to-one correspondence between such
trajectories and complete bounded solutions of (4.1)). The proposition is proved.
Example 4.3. We now consider (following mainly [227]) a bit more complicated
case of a reaction-diffusion equation in a bounded domain Ω ⊂ Rd :

∂t u = a∆x u − f (u), u ∂Ω
= 0, u t=0
= u0 . (4.7)

Here u = (u1 , . . . , uN ) is an unknown vector-valued function, ∆x is the Laplacian

with respect to the variable x = (x1 , . . . , xd ), a is a given diffusion matrix which
satisfies the condition a + a∗ > 0, and f is a given nonlinearity, which satisfies
a slightly stronger dissipativity assumption:

− C + α|u|p+1 6 f (u).u 6 C(1 + |u|p+1 ), (4.8)

where p > 0 is a given exponent. We say that the function

u ∈ L2loc R+ , W01,2 (Ω) ∩ Lp+1 p+1

loc R+ , L (Ω) =: Θloc
 ATTRACTORS. THEN AND NOW 83

is a weak solution of (4.7) if it satisfies the equation in the sense of distributions

(see, for example, [38]). It is important for us that from (4.7) we can see that for
any such solution

∂t u ∈ L2loc (R+ , W −1,2 (Ω)) + Lqloc (R+ , Lq (Ω)), (4.9)

where 1/q + 1/(p + 1) = 1. Therefore, the function t → ∥u(t)∥2L2 /2 is well defined

and absolutely continuous with respect to time and the energy identity
1 d
∥u(t)∥2L2 + (a∇x u(t), ∇x u(t)) + (f (u(t)), u(t)) = 0 (4.10)
2 dt
holds for almost all t > 0. In turn, this identity implies that u ∈ C([0, T ], L2 (Ω)),
and therefore the initial condition u t=0 = u0 is well defined (see [38], [227]).
It is not difficult to verify, using, for example, Galerkin approximations and
the energy identity (4.10), that weak solutions thus defined exist for every u0 ∈
L2 (Ω) := Φ and satisfy the dissipative estimate
Z t+1 Z t+1
∥u(t)∥2L2 + ∥∇x u(s)∥2L2 ds + ∥u(s)∥p+1 2
Lp+1 ds 6 C∥u0 ∥L2 e
−βt
+ C∗ (4.11)
t t

for some positive constants C, C∗ , and β (see [38], [227]).

Let us now turn to attractors. As before, we define the trajectory phase space
K+ ⊂ Θloc as the set of all weak solutions of problem (4.7). Then the semigroup
of time shifts T (h) acts on K+ , so the trajectory dynamical system (T (h), K+ ) is
well defined. As in the previous example, we lift the standard bornology of the phase
space Φ = L2 (Ω) to the set K+ , namely, B ∈ B if and only if B t=0 is bounded
in Φ. Then estimate (4.11) and the embedding

K+ ⊂ Cloc (R+ , Φ)

guarantee that the set

B := u ∈ K+ : ∥u∥2L2 (R

1,2 + ∥u∥Lp+1 (R+ ,Lp+1 ) 6 2C∗
b + ,W0 ) b

is a bounded absorbing set for the semigroup T (h). Moreover, since this set is
bounded in the reflexive locally convex space Θloc , it is precompact in the space Θw loc
(which is the space Θloc endowed with the weak topology) by the Banach–Alaoglu
theorem. The set B is also closed in Θw loc since it is precompact and sequentially
closed (the last fact is proved similarly to the proof of the existence of a weak
solution; see [38]), so it is compact by the Eberlein–Smulian theorem (actually, we
may avoid the use of the Eberlein–Smulian theory here since B is bounded in Θloc
and the weak topology on it is metrizable).
Thus, we have found a compact bounded absorbing set B for the trajectory
dynamical system (T (h), K+ ) endowed with the weak topology. Note also that
the shift semigroup T (h) is obviously continuous. Then, due to Proposition 3.6, there
exists a weak trajectory attractor Aw tr ⊂ K+ , which is the attractor of the trajectory
dynamical system (T (h), K+ ) in the weak topology of Θloc . It is strictly invariant
and is generated by all complete bounded solutions of equation (4.7).
84 S. V. ZELIK

We now prove that the trajectory attractor constructed is actually an attractor

in the strong topology of Θloc as well. To this end, we need an absorbing set which
will be compact in the strong topology as well. We claim that B1 := T (1)B is such
a set. To verify compactness, we use the energy method developed in [14], [227], [181].
This method is based on the fact that for many Banach spaces (including Hilbert
ones, the spaces Lp with 1 < p < ∞, and so on) the weak convergence ξn → ξ in
combination with the convergence of the norms imply strong convergence.
Let un ∈ B be a sequence of solutions. Then without loss of generality we may
assume that un → u ∈ K+ in the weak topology of Θloc . To verify the compactness
of T (1)B in the strong topology, it is sufficient to prove that un → u strongly
in L2 (T, T + 1; W01,2 ) and in Lp+1 (T, T + 1; Lp+1 ) for every T > 1. For simplicity we
consider only the case T = 1 (the general case is analogous). To do this we multiply
the energy identity (4.10) for un by t and integrate with respect to t ∈ [0, 2] to get
that
Z 2 Z 2
1
∥un (2)∥2Φ + t(a∇x un (t), ∇x un (t)) dt + t(f (un (t)), un (t)) dt
2 0 0
1 2
Z
= ∥un (t)∥2Φ dt. (4.12)
2 0

Using the boundedness of un in Θloc together with the control (4.9) of time
derivatives ∂t un , we conclude that un → u weakly in C(0, 2; Φ) and strongly
in L2 (0, 2; Φ). Thus, without loss of generality we also have the convergence un → u
almost everywhere. Passing to the subsequence if necessary, we extract from the last
equality that
Z 2 Z 2
1 2
lim ∥un (2)∥Φ + lim t(a∇x un (t), ∇x un (t)) dt + α lim t∥un (t)∥p+1
Lp+1 dt
2 n→∞ n→∞ 0 n→∞ 0
Z 2
1 2
Z
+ lim t(f (un (t)), un (t)) − αt∥un (t)∥p+1
Lp+1 dt = ∥u(t)∥2Φ dt. (4.13)
n→∞ 0 2 0

Using now the weak lower semicontinuity of convex functions and Fatou’s lemma
(in order to handle the term containing the nonlinearity f ) in combination with
identity (4.12) for the limit solution u(t), we conclude that
Z 2 Z 2
lim t(a∇x un (t), ∇x un (t)) dt = t(a∇x u(t), ∇x u(t)) dt
n→∞ 0 0

and Z 2 Z 2
lim t∥un (t)∥p+1
Lp+1 dt = t∥u(t)∥p+1
Lp+1 dt.
n→∞ 0 0

Taking weak convergence into account, we conclude that un → u strongly

in L2 (1, 2; W01,2 ) as well as in Lp+1 (1, 2; Lp+1 ). Thus, we have verified the compact-
ness of the set B1 in the strong topology of Θloc . In turn, this gives the existence
of a trajectory attractor Atr in the strong topology of Θloc which is generated by
all complete bounded trajectories of equation (4.7) and coincides with the weak
trajectory attractor constructed above.
 ATTRACTORS. THEN AND NOW 85

Remark 4.4. We recall that the possible non-uniqueness of solutions of (4.7) is

formally caused by the absence of the local Lipschitz continuity of the nonlinearity f .
However, in contrast to ODEs, adding this natural assumption would not change
the situation drastically due to possible blow up of smooth solutions which may
occur despite the dissipative energy estimate (4.11); see [21] for the case of the 3D
complex Ginzburg–Landau equation, and see [98], [15], [186] for different classes
of reaction-diffusion systems.
Remark 4.5. We emphasize another important property of system (4.7) which
does not hold for more general classes of PDEs, namely, the fact that every weak
solution of (4.7) satisfies the energy identity (4.10) and energy estimate (4.11).
In particular, this fact allows us to define the bornology in the trajectory phase
space K+ just by lifting bounded sets in the classical phase space Φ to the space
of trajectories (see (4.5)), which is not typical for the theory of trajectory attractors
(see examples below). As a result, we may define the generalized multi-valued
semigroup S(t) : Φ → 2Φ via

S(t)u0 := {u(t) ∈ Φ : u is a weak solution of (4.7) on the interval [0, t]} (4.14)

and consider the (generalized) global attractor Agl for this semigroup avoid-
ing the usage of any trajectory spaces. Of course, we will have the relation
Agl = Atr t=0 (see [38], [227], [224] for more details). Again, in a more general
situation, the straightforward formula (4.14) does not work and we cannot define
the associated multi-valued semigroup without using the trajectory phase space K+
and a non-trivial bornology on it. We also mention the concatenation property,
namely, if u1 (t), t ∈ [0, T1 ], and u2 (t), t ∈ [T1 , T2 ], are two weak solutions of (4.7)
and u1 (T1 ) = u2 (T1 ), then the compound function
(
u1 (t), t ∈ [0, T1 ],
u(t) =
u2 (t), t ∈ [T1 , T2 ],

is a weak solution of (4.7) on the interval t ∈ [0, T2 ]. This property also fails in more
complicated applications of the theory of trajectory attractors.
4.2. Trajectory attractors for elliptic PDEs. We continue by a more
simple and a bit artificial example, which demonstrates some important features
of trajectory attractors and gives us a toy example for trajectory attractors of elliptic
PDEs.
Example 4.6. Consider the following second-order ODE:

u′′ + γu′ + u3 − u = 0, u t=0

= u0 ∈ R. (4.15)

Of course, the solution of this problem is not unique since we have “forgotten” to
pose the initial condition for u′ t=0 . However, multiplying this equation by u′ , we
arrive at the following energy identity:

d u′2 u4 u2
 
+ − + γu′2 = 0 (4.16)
dt 2 4 2
86 S. V. ZELIK

which shows that all solutions of this equation are bounded in time. Moreover,
a bit more accurate arguments related to multiplication of this equation by u′ + αu
where α > 0 is properly chosen, give the dissipative estimate:

u′ (t)2 + u2 (t) 6 C(u(0)2 + u′ (0)2 )2 e−κt + C∗ , (4.17)

where C, κ, and C∗ are some positive constants which are independent of u.

1
Let us define the trajectory phase space K+ ⊂ Θloc := Cloc (R) as the set of all
solutions u(t), t > 0, of problem (4.15) which correspond to all u0 ∈ R. The
bornology B on it is defined as follows: B ⊂ K+ is bounded if B [0,1] is bounded
in C 1 [0, 1]. Then the dissipative estimate (4.17) can be rewritten in the following
form:
∥u∥C 1 [T,T +1] 6 C∥u∥2C 1 [0,1] e−κT + C∗
and, therefore, the set B := {u ∈ K+ : ∥u∥Cb1 (R+ ) 6 2C∗ } is a bounded absorbing set
for the trajectory dynamical system (T (h), K+ ). Moreover, expressing the second
derivative u′′ from the equation, we see that u′′ is also bounded if u ∈ B, so by
Arzelà’s theorem B is compact in Θloc . Thus, according to the general theory,
the trajectory dynamical system (T (h), K+ ) possesses an attractor Atr ⊂ K+ , which
is generated by all complete bounded solutions of (4.15).
The structure of the constructed attractor can easily be understood since we
have a global Lyapunov function (due to the identity (4.16)), so we have only
the equilibria u = 0, ±1 and heteroclinic orbits between them. Moreover, since
the map u → (u(0), u′ (0)) is one-to-one as a map from K+ to R2 , the attractor
obtained is actually two-dimensional and looks qualitatively like in Example 2.6
(Fig. 1, (a), up to a rotation through π/4). Thus, the trajectory approach allows us
to restore the “forgotten” initial condition on u′ t=0 and end up with the standard
attractor for this equation.
Remark 4.7. We see that in the previous example we are unable to obtain
a bornology on the trajectory phase space just by lifting the bounded sets from
the space R of the initial data (we will not have the dissipative estimate for such
“bounded” sets since u′ (0) will be out of control) and the concatenation property is
also lost in this case.
Actually, we may define a topology and a bornology on K+ in several equivalent
ways. For instance, we may take Θloc := Cloc (R+ ) and understand the first and
second derivative of u in the distributional sense (it is not difficult to show that such
2
distributional solutions are actually classical ones and a posteriori u ∈ Cloc (R+ )).
Bounded sets in K+ can also be defined alternatively as sets which are bounded
in the Fréchet space Θloc or in the Banach space Cb1 (R) (indeed, the dissipative
estimate (4.17) guarantees that in all these cases we have the same bornology
on K+ ). We mention here a slightly exotic construction of K+ and the associated
bornology proposed by Vishik and Chepyzhov to handle trajectory attractors for
damped wave equations and the 2D damped Euler equations (see [36], [38]). Namely,
let us define the trajectory space K+ as the set of all solutions u of (4.15) which
satisfy the following estimate:

∥u∥C 1 [T,T +1] 6 Cu e−κt + 2C∗ , T ∈ R+ ,

 ATTRACTORS. THEN AND NOW 87

where C∗ and κ are fixed (the same as in (4.17)) and the constant Cu may depend
on u. Note that the set K+ thus defined is shift invariant (so the trajectory
dynamical system (T (h), K+ ) is well posed), but a priori it may be smaller than
the set of all solutions of (4.15). However, for our model example, thanks to (4.17)
we know that a posteriori this is not true. The bornology on K+ is then naturally
defined as follows: B ∈ B if and only if Cu 6 CB < ∞ for all u ∈ B. It is also easy
to see that in our case this choice of B gives the same bounded sets as the previous
constructions.

Example 4.8. We now turn to a bit more interesting example of trajectory

attractors related to elliptic boundary problems in cylindrical domains. Namely,
consider the following elliptic boundary value problem:

a(∂t2 u+∆x u)+γ∂t u−f (u) = 0, u ∂Ω

= 0, u t=0
= u0 , t > 0, x ∈ Ω, (4.18)

where Ω is a bounded domain of Rd , u = (u1 (t, x), . . . , un (t, x)) is an unknown

vector-valued function, a = a∗ > 0 is a given diffusion matrix, γ ∈ R is a given
parameter, and f ∈ C(Rd , Rd ) is a given nonlinearity which satisfies the dissipativity
assumption
f (u).u > −C + α|u|2+ε (4.19)

for some positive C, α, and ε.

We emphasize that, originally, the PDE (4.18) is not an evolutionary one, but it
may appear, for instance, in the study of travelling wave solutions for evolutionary
PDEs. Then the parameter γ is naturally interpreted as a wave-speed (see [6], [25],
[173], [225], [229], and the references therein). One possible and rather popular
approach to study these equations is related to the interpretation of the variable
along the axis of the cylinder as time and apply the methods of the theory
of dynamical systems (centre manifolds, inertial manifolds, attractors, and so on;
see [6], [7], [119], [172] for more details).
Since the Cauchy problem is ill posed for elliptic equations, as a rule, we have
no uniqueness of solutions for problem (4.18) and, for this reason, it is natural
(following [225]) to use the trajectory approach. To this end we fix Φ := C0 (Ω) as
the space of initial data and consider the set K+ ⊂ Θloc := Cloc (R+ , Φ) of all (weak)
solutions u(t, x) which are defined in the whole of the semi-cylinder R+ × Ω. Note
from the very beginning that due to the interior regularity for elliptic equations,
any solution u ∈ W 2,p ((t, t + 1) × Ω) for all p < ∞ and the regularity of a solution
is restricted by the smoothness of f only.
The following lemma gives the crucial dissipative estimate for solutions of (4.18).

Lemma 4.9. Let u be a solution of problem (4.18) defined on the interval t ∈ [0, N ].
Then the following estimate holds:

∥u(t)∥Φ 6 Q(∥u(0)∥Φ )H(1 − t) + Q(∥u(N )∥Φ )H(t − N + 1) + C∗ , t ∈ [0, N ],

(4.20)
where C∗ and Q are a constant and a monotone increasing function, respectively,
which are independent of u, t, and N , and H(z) is the standard Heaviside function.
88 S. V. ZELIK

Sketch of the proof. Let w(t, x) := au(t, x).u(t, x). Then, taking the dot
product of equation (4.18) with u(t, x), we arrive at
1 2
(∂ w(t, x) + ∆x w(t, x)) = f (u(t, x)).u(t, x) + a∇t,x u(t, x).∇t,x u(t, x)
2 t
− γ∂t u(t, x).u(t, x)

and using the dissipativity condition on f and the positivity of the matrix a we get

∂t2 w(t, x) + ∆x w(t, x) + αw(t, x)1+ε/2 > −C.

Thus, due to the maximum/comparison principle it is enough to verify an analogue

of estimate (4.20) for the following ODE:

y ′′ (t) + αy(t)1+ε/2 = −C, y t=0

= ∥au(0).u(0)∥C , y t=N
= ∥au(N ).u(N )∥C .

But such an estimate for the ODE is straightforward and we leave it to the reader,
so the lemma is proved.
The estimate proved allows us to verify the existence of a solution for prob-
lem (4.18) for every u0 ∈ Φ. Indeed, to this end we first solve the corresponding
boundary value problem on the finite interval t ∈ [0, N ] with an extra boundary
condition uN t=N = 0 and then pass to the limit N → ∞. The possibility to do this
is guaranteed by this estimate (see [225] for more details). Moreover, passing to
the limit N → ∞ in (4.18), we get that any solution u ∈ K+ satisfies the estimate

∥u(t)∥Φ 6 Q(∥u0 ∥Φ )H(1 − t) + C∗ . (4.21)

In particular, any solution u ∈ K+ is bounded as t → ∞.

Thus, the trajectory dynamical system (T (h), K+ ) associated with equation (4.18)
is constructed. We endow the space K+ with the topology of Θloc and with
the bornology B lifted from the phase space Φ, namely, B ⊂ K+ is in B if B t=0 is
bounded in Φ. Then estimate (4.21) guarantees that the set

B := {u ∈ K+ : ∥u∥Cb (R+ ,Φ) 6 C∗ }

is a bounded absorbing set for the trajectory dynamical system defined above.
Moreover, due to the interior regularity estimate for elliptic equations, we know
that the set B1 := T (1)B is bounded in Cb1 (R+ × Ω) and, by Arzelà’s theorem,
it is compact in Θloc . Thus, a compact bounded absorbing set for the trajectory
dynamical system is constructed and the following result holds.
Proposition 4.10. Under the above assumptions problem (4.18) possesses a tra-
jectory attractor Atr ⊂ Θloc , which consists of all bounded solutions of (4.18) defined
on the whole of the cylinder R × Ω:

Atr = K t>0
.

Remark 4.11. Note that, at least for application to travelling waves, exactly
the set K of all bounded solutions of (4.18) on the whole cylinder consists of travelling
 ATTRACTORS. THEN AND NOW 89

waves and is the main object of interest. This set clearly contains the equilibria R
of (4.18) (trivial solutions which are independent of t) and it is an interesting
and important question whether or not it contains anything else (whether or not
non-trivial travelling waves exist). In the case of evolutionary equations we know
that the attractor is usually connected (see also [221], [123] and the material below
for the connectedness of attractors without uniqueness), and therefore we would
expect that K = ̸ R if R is disconnected. Surprisingly, this may be not true for
the attractors of elliptic equations. Indeed, consider equation (4.18) with Neumann
boundary conditions,Z γ = 0, and f (u) = u(u − 1)2 · · · (u − N )2 . Then, introducing
the function z(t) := au(t, x).u(t, x) dx, we get
Ω

z ′′ (t) = 2(f (u(t)), u(t)) + 2(a∇t,x u, ∇t,x u) > 0.

So the function z(t) is bounded and convex on R, and therefore it is a constant.
Since z ′′ (t) ≡ 0, we conclude that ∇t,x u ≡ 0 and u = const. Thus, Atr = R =
{0, 1, 2, . . . , N } is totally disconnected, and we do not have non-trivial travelling
waves.
Fortunately, the above example is an exception, rather than a rule. Indeed,
as proved in [74], we have K ̸= R under some extra mild assumptions, namely,
we need to have at least two non-degenerate (hyperbolic) equilibria to guarantee
the existence of non-trivial travelling waves. Note that in the above counterexample
we have exactly one non-degenerate equilibrium u = 0 and all others are degenerate.
The proof of this fact is based on the analogous result for the Galerkin system
of ODEs approximating (4.18) (where it can be established using the Conley index)
and the upper semicontinuity of the corresponding trajectory attractors.
Remark 4.12. We note that in the case of equation (4.18) the corresponding
bornology on the trajectory phase space is defined via lifting the bornology on
the usual phase space Φ, exactly as in the case of the reaction-diffusion system
considered above. Nevertheless, we do not have the concatenation property here and
cannot define the generalized multi-valued semigroup S(t) analogously to (4.14).
The reason is that we cannot guarantee that the solution u(t) of (4.18) which is
originally defined on the interval t ∈ [0, N ] can be continued to t > N (it may blow
up immediately for t > N ), so to define this semigroup properly, we need to use
only the solutions which are defined for all t > 0 (interpreting this assumption as
some kind of the second boundary condition at t = ∞); see [6]. Alternatively, it
can be proved that the projection


Π : K+ → Φ × Φ, Πu := u t=0 , ∂t u t=0 ,
is injective and therefore is a homeomorphism between K+ and G := ΠK+ ⊂ Φ2 .
Thus, we may define the equivalent dynamical system S(t) : G → G associated with
equation (4.18) and its trajectory dynamical system. Then we have its global
attractor Agl := ΠAtr . However, the structure of the set G remains unclear (and for
this reason such an approach does not give essential advantages in comparison with
the trajectory one). In addition, the semigroup is only Hölder-continuous (rather
than Lipschitz) in general, which, in turn, allows the attractors Agl and Atr to be
infinite-dimensional; see [173] for the corresponding example.
90 S. V. ZELIK

We also mention that the situation becomes much better in the case of so-called
fast travelling waves γ ≫ 1, where the uniqueness of a solution can be established
for problem (4.18). Then the theory becomes very similar to the standard situation
related, for example, to reaction-diffusion systems satisfying the assumptions
of the uniqueness theorem (see [25], [226]).
Remark 4.13. The trajectory approach is also applicable to elliptic boundary
value problems in non-cylindrical domains. Indeed, if an unbounded domain Ω ⊂ Rd
is invariant with respect to shifts in some fixed direction ⃗l ∈ Rd , that is,

T⃗l(h)Ω ⊂ Ω, h > 0, T⃗l(h)x := x + h⃗l,

then the corresponding semigroup of shifts (T⃗l(h)u)(x) := u(x + h⃗l) acts on an

appropriately defined set K+ of solutions of the elliptic boundary value problem
under consideration. Thus, we may interpret the direction ⃗l as the direction
of “time”, construct the associated trajectory dynamical system (T⃗l(h), K+ ), and
study its attractors. If in addition, the domain Ω satisfies
[
T⃗l(h)Ω = Rd ,
h60

then the trajectory attractor will be generated by all bounded solutions of the prob-
lem under consideration defined for all x ∈ Rd (see [237], [236] for the details).
There is also an interesting possibility to apply attractor theory to the case where
the domain Ω is not semi-invariant with respect to shifts in any direction (for
example, where Ω is an exterior domain). This possibility is based on the recently
developed theory of attractors for semigroups with multi-dimensional time (see [135]
for the details).
4.3. 3D Navier–Stokes system. In this subsection we consider various ap-
proaches to attractors of the 3D Navier–Stokes system:

∂t u + (u, ∇x )u + ∇x p = ν∆x u + g, div u = 0, u t=0

= u0 , (4.22)

in a bounded domain Ω ⊂ R3 with smooth boundary. Here u = (u1 , u2 , u3 ) is an

unknown velocity vector field, p is an unknown pressure,
3
X
(u, ∇x )u = ui ∂xi u,
i=1

ν > 0 is a given viscosity, and g is given external forces.

As usual, we denote by V := {ϕ ∈ C0∞ (Ω) : div ϕ = 0} the space of divergence-free
test functions. Then the phase space Φ = H and the space V are defined as the closure
of V in [L2 (Ω)]3 and [H 1 (Ω)]3 , respectively, and the space V −1 is the dual space
of V with respect to the duality generated by the standard inner product of H.
We also assume that u0 ∈ H and g ∈ V −1 (see, for example, [12], [215] for more
details).
By definition, a function u ∈ L∞ (R+ , H) ∩ L2b (R+ , V ) is a weak energy solution
of (4.22) if, for every test function φ ∈ C0∞ (R+ ×Ω) such that div φ = 0, the following
 ATTRACTORS. THEN AND NOW 91

identity holds:
Z Z


− (u(t), ∂t φ(t)) dt + (u(t), ∇x )u(t), φ(t) dt
R R
Z Z
+ ν (∇x u(t), ∇x φ(t)) dt = (g, u(t)) dt.
R R

It is well known that any such solution u belongs to C([0, T ], Hw ) for all T > 0 and,
4/3
moreover, ∂t u ∈ Lb (R+ , L4/3 (Ω)). (Here Hw denotes the space H endowed with
the weak topology.) In particular, the initial data u t=0 = u0 are well posed.
It is also well known that, for any u0 ∈ H and g ∈ V −1 , equation (4.22) possesses
at least one weak energy solution u which is, in addition, continuous at t = 0 as
a function taking values in H (with the strong topology) and satisfies the following
energy inequality:
1 d
∥u(t)∥2H + ν∥∇x u(t)∥2L2 6 (g, u(t)) (4.23)
2 dt
which is understood in the sense of distributions. By definition, an energy solution u
which has these extra properties is called a Leray–Hopf (LH) solution of (4.22)
(see [104], [152]). In turn, the energy inequality (4.23) is equivalent to the inequality
Z t Z t
∥u(t)∥2H + 2ν ∥∇x u(τ )∥2L2 dτ 6 ∥u(s)∥2H + 2 (g, u(τ )) dτ, (4.24)
s s

which holds for almost all s ∈ R+ and all t > s. The solution u(t) satisfying this
inequality is strongly continuous at t = 0 (that is, it is a Leray–Hopf solution) if
and only if the inequality holds for s = 0 (that is, s = 0 is not in the exceptional
set of measure zero); see [38], [194], [214] for more details. Note also that, applying
Gronwall’s inequality to (4.24), we get the dissipative energy estimate of the form
Z t
∥u(t)∥2H + ν e−β(t−τ ) ∥∇x u(τ )∥2L2 dτ 6 ∥u(s)∥2H − C∥g∥2V −1 e−β(t−s) + C∥g∥2V −1


0
(4.25)
which holds for almost all s > 0 and all t > s with some positive constants C and β
which are independent of t, s, g, and u.
We emphasize that the class of Leray–Hopf solutions is a priori not invariant
with respect to time shifts. Indeed, if u is a Leray–Hopf solution and T (s)u is a
time shift of it, then we cannot guarantee that (4.24) holds for s, and therefore
T (s)u may be discontinuous at t = 0. Important for us is the fact that T (s)u
is a Leray–Hopf solution for almost all s ∈ R+ . We also introduce the class
of generalized Leray–Hopf solutions as weak solutions which satisfy (4.23), but
which are not necessarily strongly continuous at t = 0. Obviously, such solutions
will be shift-invariant and T (s)u is a Leray–Hopf solution for almost every s ∈ R+ .
The facts mentioned above about the solutions of (4.22) give us a base for
attractor theory, but the construction of the corresponding trajectory attractor can
be realized in several alternative ways. We start with the most general scheme
proposed by Vishik and Chepyzhov; see [36] and [38] where it was applied to weakly
damped wave equations with fast growing nonlinearities as well as to the 2D damped
Euler equations.
92 S. V. ZELIK

VC
Example 4.14. Consider the set K+ of all weak energy solutions u of (4.22)
which correspond to all u0 ∈ H and satisfy the following estimate:
Z t
∥u(t)∥2H + ν e−β(t−s) ∥∇x u(s)∥2L2 ds 6 (Cu − C∥g∥2V −1 )e−βt + C∥g∥2V −1 , (4.26)
0

where the positive constants C and β are the same as in (4.25) and the constant Cu
depends on the solution u. In other words, a weak energy solution u belongs to
VC
the space K+ if and only if there exists a constant Cu depending on u such that
VC
inequality (4.26) is satisfied for all t > 0. Then, on the one hand, the set K+
is not empty and contains all Leray–Hopf solutions and, on the other hand, it is
VC VC
shift invariant: T (h)K+ ⊂ K+ , so the associated trajectory dynamical system
VC VC
(T (h), K+ ) is well defined. We endow the space K+ with the weak-star topology
∞ 2
of the space Θloc := Lloc (R+ , H) ∩ Lloc (R+ , V ) and with the bornology B discussed
VC
in Remark 4.7. Namely, B ⊂ K+ belongs to B if and only if supu∈B Cu =: CB < ∞.
Then, due to assumption (4.26), the set
 Z t 
B := u ∈ K+ VC
: ∥u(t)∥2H + ν e−β(t−s) ∥∇x u(s)∥2L2 ds 6 2C∥g∥2V −1 (4.27)
0

is a bounded absorbing set for the trajectory dynamical system constructed.

Moreover, since

L∞ (0, T ; H) ∩ L2 (0, T ; V ) = [L1 (0, T ; H) + L2 (0, T ; V −1 )]∗

VC
for all T , the set B is precompact in the space K+ endowed with the topology
w∗
of Θloc . Since the weak-star topology on B ⊂ Θloc is metrizable, to verify that B is
closed, it is sufficient to verify its sequential closedness. In turn, this can be done
by passing to the limit in (4.22), similarly to the proof of the existence of a weak
solution (see [38] for more details).
Thus, we have verified that B is a bounded compact absorbing set for the trajectory
VC
dynamical system (T (h), K+ ) and, according to the general theory, there exists
a trajectory attractor AVC tr ⊂ K VC
+ which is generated by all complete bounded weak
solutions of (4.22), that is, by all weak solutions which are defined for all t ∈ R and
satisfy the inequality
Z t
∥u(t)∥2H + e−β(t−s) ∥∇x u(s)∥2L2 ds 6 C∥g∥2V −1 . (4.28)
−∞

We denote the set of all such solutions by KVC .

Remark 4.15. The attractor AVC tr constructed above attracts, in particular, all
Leray–Hopf and generalized Leray–Hopf solutions, however, it is somehow “too
big” and contains a lot of non-physical solutions. In particular, it depends on
the choice of the constant C in (4.26). Indeed, from [20] we know that, given an
arbitrary sufficiently regular function E(t), there exists a weak energy solution u(t)
of problem (4.22) which satisfies the additional assumption ∥u(t)∥H = E(t), t > 0.
Based on this, we see that even the set KVC of complete bounded trajectories
depends on the choice of the constant C. For this reason it is interesting to discuss
alternative constructions which allow us to discard most non-physical solutions.
 ATTRACTORS. THEN AND NOW 93

gLH
Example 4.16. Consider the set K+ of all generalized Leray–Hopf solutions.
As we have already mentioned, this set is semi-invariant with respect to time shifts,
gLH
and therefore the corresponding trajectory dynamical system (T (h), K+ ) is well
gLH ∗
defined. The topology on K+ is naturally defined by the embedding in Θw loc as
in the previous case and we only need a bornology.
gLH
We say that B ⊂ K+ is bounded if B t∈[0,1] is a bounded set in L∞ (0, 1; H).
Then, due to estimate (4.25), we have
Z t
∥u(t)∥2H + ν e−β(t−τ ) ∥∇x u(τ )∥2L2 dτ 6 C∥u∥2L∞ (s,s+1;H) e−β(t−s) + C∥g∥2V −1
s
(4.29)
already for all t > s > 0, and therefore
 Z t 
gLH 2 −β(t−τ ) 2 2
B := u ∈ K+ : ∥u(t)∥H + ν e ∥∇x u(τ )∥L2 dτ 6 2C∥g∥V −1 (4.30)
0

is a bounded absorbing set for the trajectory dynamical system under consideration.
This set is precompact due to the Banach–Alaoglu theorem, and its compactness
follows from the standard fact that the weak limit of generalized Leray–Hopf
solutions is a generalized Leray–Hopf solution. Thus, according to the general theory,
the trajectory dynamical system under consideration possesses an attractor AgLH tr
which is generated by all bounded Leray–Hopf solutions of (4.22) defined for all
t ∈ R. We denote the set of such solutions by KLH .
Remark 4.17. The attractor AgLH tr constructed above looks as the most natural
trajectory attractor for the 3D Navier–Stokes equations and appears in slightly
different, but equivalent forms in many papers starting from the seminal works
of Vishik and Chepyzhov [36] and Sell [208]. For instance, one can replace
∗
the topology of Θw 2
loc by the strong topology of Lloc (R+ , H), avoiding the use
of weak topologies in definitions. Indeed, as is not difficult to show, using
the control of an appropriate norm of the time derivative ∂t u in equation (4.22)
∗
(see, for example, [208]), the topologies of L2loc (R+ , H) and Θw loc coincide on
the absorbing ball B. Moreover, due to the trick with an alternative generalization
of Leray–Hopf solutions proposed in [208], one can use bounded sets of L2loc (R+ , H)
to define a bornology on the trajectory phase space as well. However, we cannot use
the bornology generated by the sets of initial data bounded in H. As we have already
mentioned, estimate (4.25) may fail at s = 0 for generalized Leray–Hopf solutions,
and the boundedness of the set of u(0) does not imply that the corresponding set
of trajectories is bounded in Θloc .
Example 4.18. There is an alternative way, which allows us to construct
a trajectory attractor using the bornology related to bounded sets in H, namely,
LH
one may consider the set K+ of standard (not generalized) Leray–Hopf solutions
as a “trajectory phase space”. The problem here is that, as discussed before,
LH LH
T (h)K+ may be not a subset of K+ , so we need to deal with non-invariant sets
of trajectories. The extension of the attractor theory to this case has been developed
in [248]. We will not discuss this theory in more details since there is a simple
trick which allows us to reduce it to the general scheme considered in the previous
94 S. V. ZELIK

∗
section. Namely, consider the whole of the space Θw loc as the trajectory phase space
for problem (4.22). Then it is obvious that the semigroup T (h) of time shifts
∗
w∗
acts continuously on Θw loc , so the trajectory dynamical system (T (h), Θloc ) is well
defined.
Of course, at the moment this trajectory dynamical system looks as an abstract
nonsense since it is completely unrelated to the initial Navier–Stokes equation. But
we still have not introduced a bornology for this dynamical system and exactly
∗
the bornology will relate it with the original problem. Namely, a set B ⊂ Θw loc
belongs to the bornology B if
LH
1) B ⊂ K+ , that is, B consists of the Leray–Hopf solutions of the Navier–Stokes
problem;
2) the set B t=0 := {u(0), u ∈ B} is bounded in H.
Indeed, due to the dissipative estimate (4.25) and our choice of a bornology, we
see that the compact set (4.30) constructed above is an absorbing set for this new
trajectory dynamical system and, therefore, according to the general theory, we
gLH
have an attractor ALH LH
tr ⊂ B (and therefore Atr ⊂ K+ ). Since T (h) is continuous,
LH
this theory also guarantees that Atr is strictly invariant and therefore is generated
by some subset of complete bounded Leray–Hopf solutions:
gLH
ALH
tr ⊂ Atr = KLH t>0
.

Note that the general theory does not give us the coincidence of two attractors
since we cannot claim that the above absorbing set belongs to B (we do not know
LH
whether or not B ⊂ K+ ). But in our case it follows immediately from the fact that
LH LH
for every u ∈ K , we have T (h)u t>0 ∈ K+ for almost every h ∈ R. Thus,
gLH
ALH
tr = Atr

and considering the non-invariant spaces of trajectories does not bring anything new
to the theory of trajectory attractors for the 3D Navier–Stokes equations. We also
mention that this is a lucky exception in the theory of attractors of non-invariant sets
of trajectories that we are able to establish the last equality and clarify the structure
of the corresponding attractor. In more general situations we usually do not know
how this attractor is related to solutions of the equation under consideration and
what complete bounded trajectories generate it. This is actually the main drawback
of the theory.
Example 4.19. We now consider one more approach to trajectory attractors,
which was proposed in [241] (see also [39], [90], [179]) for the study of damped wave
equations with supercritical nonlinearities and which is based on approximations
of the original system by a system for which we have the uniqueness of solutions. We
restrict ourselves to the Galerkin approximations of the original 3D Navier–Stokes
system. Namely, let {en }∞ n=1 be the orthonormal (in H) system of eigenvectors
of the classical Stokes operator A := −Π∆x in Ω with Dirichlet boundary conditions,
and let PN be the corresponding orthoprojector onto the first N eigenvectors of A:
N
X
PN u := (u, en )en , HN := PN H.
n=1
 ATTRACTORS. THEN AND NOW 95

Then, given N ∈ N, the corresponding Galerkin approximation system reads

∂t uN + PN (uN , ∇x )uN = −νAuN + PN g,
N
X (4.31)
uN t=0
= u0N , uN = uiN (t)ei ∈ HN .
i=1

These equations are a smooth system of ODEs with respect to u1N (t), . . . , uN N (t).
In addition, any solution of this system satisfies exactly the same energy estimates
as the limiting Navier–Stokes system and, for this reason, we have the unique
global solvability of (4.31) as well as the uniform with respect to N dissipativity
estimate (4.25) for the solutions uN . Moreover, arguing in a standard way, we
conclude that any sequence uN of the Galerkin solutions such that the uN (0) are
∗
uniformly bounded in H has a subsequence converging in Θw loc to a generalized
Leray–Hopf solution u of the initial Navier–Stokes problem (4.22) (we do not assume
here that uN (0) converges strongly in H, so we cannot guarantee the continuity
of u(t) in H at t = 0). This is the standard way how Leray–Hopf solutions are
usually constructed; see [38], [194] for more details.
gal gLH
We now define the trajectory phase space K+ ⊂ K+ ⊂ Θloc as the set of all
generalized Leray–Hopf solutions of (4.22) which can be obtained as weak-star
limits of Galerkin solutions:
n o
gal
K+ := u ∈ Θloc : u = lim uNk , (4.32)
k→∞
∗
where the limit is taken in the topology of Θw
loc .
Since we do not assume that
gal
uNk (0) → u(0) strongly in H, the set K+ is invariant with respect to time shifts, and
gal
therefore the trajectory dynamical system (T (h), K+ ) is well defined. We endow
∗
this dynamical system with the topology of Θw loc , so, to speak about attractors,
gal
it remains to introduce a bornology on K+ . This requires some accuracy since
we want the weak-star limit of such solutions, belonging to a bounded set, to
be such a solution too, that is, it should be possible to obtain it as a weak-star
limit of Galerkin solutions. To this end, following [241], we introduce the so-called
M -functional:
gal
Mu (t) := inf lim inf ∥uNk (t)∥2H , u ∈ K+ . (4.33)
uNk ⇁u k→∞

The external infimum is taken over all subsequences of Galerkin solutions which
gal gal
converge to a given u ∈ K+ . By definition, the M -functional is well defined on K+
and the straightforward arguments show that
(a) MT (s)u (t) 6 Mu (t + s); (b) ∥u(t)∥2H 6 Mu (t);
Z t+s
(c) Mu (t + s) + ν ∥∇x u(τ )∥2L2 dτ 6 CMu (s)e−βt + C∥g∥2V −1 , t, s > 0.
s
(4.34)
The key property of the M -functional is stated in the lemma below.
gal ∗
Lemma 4.20. Let the sequence un ∈ K+ be such that un → u in Θw
loc . Then
gal
u ∈ K+ and
Mu (t) 6 lim inf Mun (t), t > 0.
n→∞
96 S. V. ZELIK

The proof of this lemma is based on the diagonal procedure and the fact
∗
that the topology of Θw loc is metrizable on bounded subsets of Θloc (see [241] for
the details).
We are now ready to define a bornology B for the trajectory dynamical system
gal gal
(T (h), K+ ). Namely, a set B ⊂ K+ is an element of B if and only if

sup Mu (0) < ∞.

u∈B

Then, according to Lemma 4.20 and the dissipative estimate (4.34), the set
gal
B := {u ∈ K+ : Mu (0) 6 2C∥g∥2V −1 }

is a compact and bounded absorbing set for the trajectory dynamical system
gal
(T (h), K+ ), so according to the general theory, we have the following result.
Theorem 4.21. Under the above assumptions, the trajectory dynamical system
gal
(T (h), K+ ) associated with the Navier–Stokes problem (4.22) possesses an attractor
gal gLH
Atr ⊂ Atr which is generated by all complete bounded trajectories of (4.22) which
can be obtained as a weak-star limit of the corresponding Galerkin approximations.
Namely,
Agal
tr = K
gal
t>0
,

where u ∈ Kgal if and only if there exist a sequence tk → −∞, a bounded in H

sequence of the initial data, and a sequence of the Galerkin solutions uNk (t), t > tk ,
such that u is a weak-star limit of uNk (see [241] for more details).
Remark 4.22. The key advantage of the approach related to approximations
and the M -functional is that we can use not just the energy estimates, but also
other types of estimates available on the level of approximations in order to
study the corresponding trajectory attractor. Indeed, this approach was originally
suggested in [241] for damped wave equations with supercritical nonlinearities,
in order to verify that any complete bounded trajectory u(t), t ∈ R, of such an
equation is actually smooth for t 6 Tu . This fact follows from the smoothness
of the corresponding set of equilibria and the gradient structure of the equation,
but its proof requires rather delicate estimates, which can be justified on the level
of Galerkin approximations only, so the analogous result is not known for other
types of trajectory attractors. As a drawback, we mention that the equality
AgLH
tr = Agal
tr is not known and the attractor obtained may a priori depend on
the way of approximation (for example, on the choice of the Galerkin base).
4.4. Connectedness of trajectory attractors. It is well known that global
attractors of evolutionary PDEs are usually connected (see [12], [215]). This is
based on the simple topological fact that an ω-limit set ω(B) of a connected set B is
connected if the corresponding dynamical system is continuous (this may not be true
under the closed graph assumption alone; see [185]). However, the situation with
trajectory attractors is more delicate since it is a priori not clear whether or not
the corresponding trajectory phase space is connected (and as we have seen above,
it may be not connected at least for the trajectory attractors of elliptic PDEs).
Nevertheless, in many cases the trajectory attractors related to evolutionary PDEs
 ATTRACTORS. THEN AND NOW 97

remain connected. We briefly discuss below this theory using the model example
of the 3D Navier–Stokes equations (an utilizing ideas from [122], [222]), although
the method has a general nature and works in many other cases. The key technical
tool here is the following simple lemma.
Lemma 4.23. Let Φ be a metric space, and let B be a compact set in it. Assume
that a family of connected sets γα ⊂ B, α ∈ A, where A is a topological space, is fixed
such that distΦ (γα , B) → 0 as α → α0 . Finally, let u1 , u2 ∈ B, and let there exist
sequences αn → α0 and uni ∈ γαn , i = 1, 2, such that uni → ui as n → ∞, i = 1, 2.
Then the points u1 and u2 belong to the same connected component of B. Here and
below distΦ (A, B) denotes the non-symmetric Hausdorff semi-distance between the
sets A and B in Φ.
Proof. Although the proof of the lemma is standard, for the convenience
of the reader we present it here. Indeed, assume that the statement is wrong, so
ui ∈ Bi where the Bi ⊂ B are closed, disjoint, and B = B1 ∪ B2 . Since B is compact,
there are disjoint ε-neighbourhoods Oε (Bi ) of them for some positive ε. Then, since
γα → B in the sense of upper semicontinuity, we have γαn ⊂ Oε (B1 ) ∪ Oε (B2 ) for
sufficiently large ns. On the other hand, since the ui are limit points, we must have
γαn ∩ Oε (Bi ) ̸= ∅, i = 1, 2, if n is large enough. This contradicts the connectedness
of γαn and proves the lemma.
This lemma will be used in the situation where all the γα are continuous curves
in Φ: γα : [a, b] → Φ and limα→α0 γα,a = u1 , limα→α0 γα,b = u2 for some u1 , u2 ∈ B.
If we succeed in finding such families of curves for any u1 , u2 ∈ B, this will imply
the desired connectedness of B. It is important to emphasize that, although
u1 , u2 ∈ B, the curves γα need not belong to the set B. In particular, in what follows
the set B will consist of trajectories of the dynamical system under consideration
and γα,s ∈ Φ will be generated by some cleverly chosen approximations for these
trajectories.
gal
We start with the simplest case of the trajectory dynamical system (T (h), K+ ).
gal
Proposition 4.24. Let (T (h), K+ ) be the trajectory dynamical system associated
with the Galerkin solutions of the Navier–Stokes equations that were constructed
gal
above. Then the set K+ , the absorbing set B, and the corresponding trajectory
gal ∗
attractor Atr are connected in the topology of Θw loc .

Proof. Fix a stationary point u of the Navier–Stokes problem under consid-

eration. It is well known that u exists and is regular enough (u ∈ H01 (Ω)), so
∗
PN u → u as N → ∞ strongly in H01 (Ω) and, in particular, in Θw loc . This stationary
solution can be obtained as a limit of the Galerkin approximations (4.31) starting
with u0N := PN u. This fact can easily be proved using weak-strong uniqueness
arguments. Thus, u ∈ B.
Now let u ∈ B be another Galerkin solution which belongs to the absorbing
set B, and let us prove that u and u belong to the same connected component of B.
By transitivity this will be enough to establish the connectedness of B and finish
the proof of the proposition.
By the definition of u there exists a sequence of Galerkin approximations uNk
∗
such that uNk → u in Φ := Θw loc as k → ∞. Let us construct a continuous
w∗
curve γk,s ⊂ Θloc , s ∈ [0, 1], by fixing suNk (0) + (1 − s)PNk u as the initial data
98 S. V. ZELIK

for the Galerkin system (4.31) and taking γk,s ( · ) ∈ Φ as a unique solution
of this problem. Then the continuity of this curve with respect to s follows
from the continuity of the system of ODEs (4.31) with respect to the initial data
and, by construction, limk→∞ γk,1 = u and limk→∞ γk,0 = u. Here we essentially
use that the whole sequence uN of Galerkin approximations to u (not only up
to a subsequence) is convergent, and this is the main reason why we take one
of the points in B as a stationary solution. Thus, to apply the previous lemma, we
only need to verify that distΦ (γk,s , B) → 0 as k → ∞. S
To show this we observe that by the Banach–Alaoglu theorem, γk,s is
k∈N,s∈[0,1]
∗
a precompact set in Θwloc and, by the definition of B any limit point of this set as
k → ∞ belongs to B. This gives the desired upper semicontinuity and allows us to
apply the lemma to verify that B is connected and finish the proof of the proposition.
We now turn to other types of trajectory attractors.
Proposition 4.25. The trajectory attractors AgLH
tr and AVC
tr are connected
w∗
in Θloc .
Sketch of the proof. The idea of the proof is similar: construct a family
∗
of continuous curves γk,s connecting different points of K+ in Θw
loc , pass to the limit
k → ∞, and use Lemma 4.23. However, we now do not have a canonical way to
approximate trajectories of our equation, so we need to proceed in a more accurate
way. The key method for constructing such curves was proposed in [122], [222] on
the level of multi-valued semigroups, so here we only indicate the straightforward
changes which should be made in order to adapt this method to the case of trajectory
attractors, leaving the details to the reader.
Note that the Galerkin approximations are no longer appropriate for our purposes,
so we consider an alternative, the Leray-α approximations:
∂t uα + (vα , ∇x )uα + ∇x p = ν∆x uα + g,
(4.35)
div uα = 0, vα = (1 − αA)−1 uα , uα t=0
= uα (0),
where α > 0 is a regularization parameter. It is well known that this problem
possesses a unique solution which depends continuously (in both weak and strong
topologies) on the initial data uα (0) ∈ H for every α > 0. Moreover, the solution uα
satisfies all the energy estimates stated above uniformly with respect to α and, as
∗
α → 0, we have convergence in Θw loc (up to extracting a subsequence) to generalized
Leray–Hopf solutions of the limit Navier–Stokes system (4.22). In addition, if
uα (0) → u(0) strongly in H, then the limit solution will be a Leray–Hopf solution
(see [34], [194] for more details).
gLH
Let us start with the trajectory dynamical system (T (h), K+ ). It is enough
∗
to prove that the absorbing set B defined by (4.30) is connected in Θw loc . In turn,
in order to verify this fact, it is enough to check that the restriction BT := B t∈[0,T ]
∗ ∗
is connected in Θw w
T = Θloc t∈[0,T ]
for any fixed T > 0.
∗
Now let u1 , u2 ∈ B. Given α > 0, we construct a continuous curve γα,s ∈ ΘwT ,
s ∈ [−T − 1, T ], as follows. For s ∈ [0, T ] we define
(
u1 (t), t 6 s,
γα,s (t) := (4.36)
uα,1 (t), t > s,
 ATTRACTORS. THEN AND NOW 99

where uα,1 (t) is a unique solution of the Leray-α approximation (4.35), where
uα,1 (s) = u1 (s). For s ∈ [−1, 0], γα,s (t) solves (4.35) for the initial data (s +
1)u1 (0) − su2 (0). Finally, for s ∈ [−T − 1, −1],
(
u2 (t), t 6 −s − 1,
γα,s (t) := (4.37)
uα,2 (t), t > −s − 1,

where uα,2 (t) solves (4.35) for the initial data uα,2 (−s − 1) = u2 (−s − 1). It is not
∗
difficult to see that γα,s is indeed a continuous curve in Θw T such that γα,T = u1
and γα,−T −1 = u2 .
∗
γα,s is precompact in Θw
S
As in the previous case, one can show that the set T
α,s
and all limit points as α → 0 are generalized Leray–Hopf solutions of the limiting
Navier–Stokes system and belong to B (to verify this we use essentially the fact that
the compound trajectories (4.36) and (4.37) satisfy the energy equality (4.23) on
the whole time interval t ∈ [0, T ]). Thus, by Lemma 4.23, u1 and u2 must belong
to the same connected component of BT and BT is connected.
VC
We now turn to the case of the trajectory dynamical system (T (h), K+ ). The
proof in this case is analogous: we use exactly the same continuous curves γα,s and
the only difference is that we need to check that the compound trajectories (4.36)
and (4.37) satisfy the energy estimate involved in the definition (4.27) uniformly
with respect to α. The last fact is an immediate corollary of (4.25) for uα . Thus,
the connectedness of AVC tr is verified and the proposition is proved.

Remark 4.26. Note that the concatenation property is not known for the tra-
gLH gal
jectory spaces K+ and K+ , nevertheless, the corresponding trajectory attractors
are connected. On the other hand, as we can see from the proof, the possibility
to construct a compound solutions by “gluing together” two different solutions is
crucial for this result. For instance, the concatenation property does not hold for two
generalized Leray–Hopf solutions, but it holds if the second solution is continuous at
the left-hand endpoint, and this is the key point of the proof. In contrast, in general
we are unable to extend a solution of an elliptic PDE considered in Example 4.8,
which is defined for t ∈ [0, T ], for t > T by solving the appropriate boundary value
problem with the “initial” condition at t = T , and this allows the trajectory phase
space to be disconnected.

5. Attractors for non-autonomous problems

In this section, we discuss attractor theory for the dynamical processes related
to dissipative PDEs which depend explicitly on time. At this moment there are
two major approaches to extend the concept of an attractor to non-autonomous
equations: the first treats such an attractor as a time-dependent set A(t), which leads
naturally to pullback attractors (or kernel sections in the terminology of Chepyzhov
and Vishik); see [27], [35], [52], [121], and the references therein; the second approach
is based on the reduction of a non-autonomous problem to an autonomous one
by an appropriate extension of the phase space and leads to uniform attractors
which remain independent of time: see [37], [38], [96]. We also mention that there
are several intermediate approaches (for example, the so-called forward attractors),
100 S. V. ZELIK

which are significantly more difficult to study and less popular. In addition, they
often somehow accumulate the drawbacks of both pullback and uniform attractors
and, for this reason, are not considered here (see [96], [123] for the details). We
start our exposition with pullback attractors.
5.1. Pullback attractors: a general approach. The aim of this subsection is
to give a natural extension of our main result, Theorem 3.3, to the non-autonomous
case. First we give a general definition of a dynamical process.
Definition 5.1. Let {Φτ }τ ∈R be a family of Hausdorff topological spaces. Then
a family of maps U (t, τ ) : Φτ → Φt , τ ∈ R, t > τ , is a dynamical process if

U (τ, τ ) = Id, U (t, s) = U (t, τ ) ◦ U (τ, s), t > τ > s. (5.1)

As in the autonomous case, in order to speak about attractors, we need to specify

what sets are “bounded”. In the non-autonomous case and under the pullback
approach, it is natural to consider bounded sets which also depend on time:
t → B(t) ∈ Φt , t ∈ R. Namely, we specify a bornology B as a system of such time
dependent sets: B = {B(t)}t∈R ∈ B. We impose a unique requirement on elements
B ∈ B: for any t ∈ R, B(t) ̸= ∅. We also mention that the bornology B is
often referred to as a “universe” in the theory of pullback attractors. However, we
prefer to keep the name of “bornology” to be consistent with our general theory for
the autonomous case.
Definition 5.2. Let U (t, τ ) be a dynamical process in Hausdorff topological
spaces Φt . A time-dependent set B = {B(t)}t∈R is a pullback absorbing set with
respect to the bornology B if for every B ∈ B and every t ∈ R there exists
T = T (B, t) > 0 such that

U (t, t − s)B(t − s) ⊂ B(t) ∀ s > T. (5.2)

Analogously, a time-dependent set B is an attracting set with respect to the bornol-

ogy B if for every B ∈ B, every t ∈ R, and every neighbourhood O(B(t)) of the set B(t)
in the topology of Φt , there exists T = T (B, t, O) such that

U (t, t − s)B(t − s) ⊂ O(B(t)) ∀ s > T. (5.3)

Remark 5.3. It may look a bit surprising, but exactly the pullback attraction
property is a natural generalization of the attraction property to the non-autonomous
case (at least if we treat the attractor as a time-dependent set). Roughly speaking,
if we fix a present moment of time t and start the evolution from a bounded
set B(t − s) sufficiently far back in the past, then its image U (t, t − s)B(t − s)
at present will be arbitrarily close to the attracting set B(t). In contrast to this,
the forward attraction property (that is, if you start from a bounded set B(t) at the
present moment of time, then the image U (t + s, t)B(t) is close to B(t + s) for s large
enough) is much more delicate and not convenient since it may fail for pullback
attractors (see Example 5.10 below). Moreover, the author is not aware about any
more or less general constructions which give a “forward” attractor with reasonably
good properties. Actually, exactly the lack of forward attraction is a key drawback
 ATTRACTORS. THEN AND NOW 101

of the theory of pullback attractors. There are two alternative ways to overcome
this drawback: one of them is to consider random external forces with some
ergodicity (then forward attraction holds in the sense of convergence in measure),
and the alternative is to consider non-autonomous exponential attractors where we
have uniform in time attraction; see the exposition below.

As in the autonomous case, the theory is based on an appropriate generalization

of ω-limit sets.

Definition 5.4. Let U (t, τ ) be a dynamical process in the Hausdorff topological

spaces Φt , t ∈ R, endowed with some bornology B. Then the pullback ω-limit set
of B ∈ B is defined by
\ [ 
ωB (t) := U (t, t − τ )B(t − τ ) . (5.4)
s>0 τ >s Φt

Definition 5.5. A time-dependent set A = {A(t)}t∈R is a pullback attractor

for a dynamical process U (t, τ ) in Hausdorff topological spaces Φt endowed with
some bornology B if
1) for any t ∈ R the set A(t) is compact in Φt ;
2) A is a pullback attracting set;
3) A is a minimal (with respect to inclusion) time-dependent set which has
properties 1) and 2).

The next theorem is an analogue of Theorem 3.3 and is the main result of this
subsection. We mention also that, in the case where Φt are Banach spaces and
the bornology B consists of uniformly bounded sets, this theorem was proved in [50].

Theorem 5.6. Let a dynamical process U (t, τ ), t > τ ∈ R, acting in Hausdorff

topological spaces {Φt }t∈R endowed with some bornology B, possess a compact
pullback attracting set B. Then the dynamical process U (t, τ ) possesses a pullback
attractor A(t), t ∈ R. Moreover, for all t ∈ R, we have A(t) ⊂ B(t).

Sketch of the proof. As in the autonomous case, the statement follows from
the corresponding properties of ω-limit sets. Namely, we need to verify that under
the assumptions of the theorem, for every B ∈ B, ωB (t) is a non-empty compact
subset of B(t) that pullback attracts the images of B and that is the minimal
compact set possessing this property. Then the desired attractor is defined via
the standard formula
 [ 
A(t) := ωB (t) , t ∈ R.
B∈B Φt

The verification of the above properties of pullback ω-limit sets repeats the proof
of Theorem 3.3 almost word by word and, for this reason, is left to the reader.

Analogously to the autonomous case, a pullback attractor is strictly invariant with

respect to the corresponding dynamical process if the maps U (t, τ ) are continuous.
102 S. V. ZELIK

Proposition 5.7. Let the assumptions of Theorem 5.6 hold, and let, in addition,
U (t, τ ) : Φτ → Φt be continuous for all fixed t > τ ∈ R. Then the pullback
attractor A(t) is strictly invariant:

A(t) = U (t, τ )A(τ ). (5.5)

The proof of this statement also repeats word by word the proof of Proposition 3.4
and thus is omitted.
Remark 5.8. We see that, analogously to the autonomous case, we have
the invariance of the pullback attractor if the maps U (t, τ ) are continuous and this
property remains true if we replace continuity by the assumption that the graphs
of all maps U (t, τ ) are closed. However, in contrast to the autonomous case, we
cannot in general replace the minimality assumption in Definition 5.5 by the strict
invariance (5.5) even if all maps U (t, τ ) are continuous. Indeed, as simplest examples
show (see, for instance, [27], [179] and Example 5.10 below), such a modification
of the definition may lead to the lack of uniqueness of a pullback attractor.
We now discuss the validity of the representation formula for pullback attractors.
We say that a complete trajectory u(t), t ∈ R, is bounded (B-bounded) if {u(t)}t∈R ∈
B and denote by K the set of all bounded complete trajectories of the dynamical
process U (t, τ ) (following Vishik and Chepyzhov, we refer to K as the B-kernel
of the dynamical process U (t, τ )). Then, obviously

K t=τ
⊂ A(τ )

if the pullback attractor exists. The next statement is a non-autonomous analogue

of Proposition 3.6.
Proposition 5.9. Let the assumptions of Theorem 5.6 hold, and let, in addition,
the maps U (t, τ ) be continuous for all fixed t > τ ∈ R and the compact absorbing
set B be bounded (B ∈ B). Assume also that the bornology B is stable with respect to
inclusions (if B ∈ B and B1 (t) ⊂ B(t) for all t ∈ R and B1 (t) ̸= ∅, then B1 ∈ B).
Then
A(τ ) = K t=τ , τ ∈ R, (5.6)
where K is the B-kernel of the dynamical process U (t, τ ).
The proof of this statement is similar to the proof of Proposition 3.6, and we
omit it.
Example 5.10. Consider the following ODE in Φt ≡ R:
(
−y, t < 0,
y ′ (t) = f (t, y), y t=τ = yτ , f (t, y) := 3
(5.7)
y − y , t > 0.

Also fix the standard bornology B as the bornology of uniformly bounded sets,
namely, B ∈ B if and only if B(t) ̸= ∅ and supt∈R ∥B∥ < ∞. Then, as is not
difficult to see, the solution operators U (t, τ ) of the ODE under consideration
generate a dynamical process in R which possesses a pullback attractor A(t) = {0}.
Note that this attractor does not possess the forward in time attraction property.
 ATTRACTORS. THEN AND NOW 103

Indeed, for t > 0, the equilibrium y = 0 is exponentially unstable and all close
trajectories run away from the neighbourhood of zero (and approach the interval
[−1, 1], which is a uniform attractor in this case). This example demonstrates
the key drawback of the theory of pullback attractors, which forces us to identify
the repelling point y = 0 with an “attractor”.
Fix two points a > 0 and b < 0 and consider the corresponding solutions a(t)
and b(t), t ∈ R, of equation (5.7) satisfying a(0) = a and b(0) = b. Then
the time-dependent set
A(t) := [b(t), a(t)]
has properties 1) and 2) in Definition 5.5 together with the strict invariance (5.5).
This example shows that the minimality property of pullback attractors cannot be
replaced by strict invariance without the risk to lose the uniqueness of a pullback
attractor.

5.2. Cocycles and random attractors. The abstract construction of a pull-

back attractor is somehow “too general” for practical applications, so it looks
reasonable to consider its particular cases. One of the most interesting special cases
is given by a cocycle, which, in particular, bridges the gap between the theories
of pullback and random attractors.
Definition 5.11. Let Φ be a Hausdorff topological space, Ψ be an arbitrary set,
and let T (h) : Ψ → Ψ, h ∈ R, be a group of operators acting on Ψ. Then the family
of maps Sξ (t) : Φ → Φ, ξ ∈ Ψ, t > 0, is a cocycle over the group T (h) if

1) Sξ (0) = Id; 2) Sξ (t + s) = ST (s)ξ (t) ◦ Sξ (s), t, s > 0, ξ ∈ Ψ. (5.8)

As is not difficult to check, any cocycle Sξ (t) generates a family of dynamical

processes on Φ depending on a parameter ξ ∈ Ψ via

Uξ (t, τ ) := ST (τ )ξ (t − τ ), t > τ ∈ R, ξ ∈ Ψ. (5.9)

This family satisfies the additional translation identity

Uξ (t + s, τ + s) = UT (s)ξ (t, τ ), ξ ∈ Ψ, t > τ ∈ R, s ∈ R. (5.10)

Vice versa, any family of dynamical processes Uξ (t, τ ) : Φ → Φ which satisfies

the translation identity (5.10) generates a cocycle Sξ (t) := Uξ (t, 0) in Φ. In
applications we are usually given a non-autonomous PDE

∂t u = A(u, ξ(t)), u t=τ

= uτ , (5.11)

where A( · , · ) is a nonlinear operator which we do not specify here and ξ = ξ(t)

accumulates all terms of the equation that depend explicitly on time, Ψ is an
appropriate space of time-dependent functions (for example, some shift-invariant
subspace of L2loc (R, H) where H is a Banach space) and T (h) : Ψ → Ψ, h ∈ R, is
the group of time shifts: (T (h)ξ)(t) = ξ(t + h). Then, if problem (5.11) is globally
well posed for all uτ ∈ Φ and all ξ ∈ Ψ, then the corresponding solution operators
Uξ (t, τ ) : Φ → Φ generate a translation-invariant family of dynamical processes,
104 S. V. ZELIK

and therefore also generate a cocycle over T (h) : Ψ → Ψ in the phase space Φ
(see [27], [38], [123] for more details).
In the case of cocycles it is natural to define a bornology B as a collection
of non-empty ξ-dependent sets {B(ξ)}ξ∈Ψ ∈ B. Then absorbing and attracting sets
will be also ξ-dependent sets. For instance, the set {B(ξ)}ξ∈Ψ is pullback attracting
if for every B ∈ B, every ξ ∈ Ψ, and every neighbourhood O(B(ξ)) of the set B(ξ)
in Φ, there exists τ = τ (ξ, B, O) such that

ST (−t)ξ (t)B(T (−t)ξ) ⊂ O(B(ξ)), t > τ.

Analogously, {A(ξ)}ξ∈Ψ is a pullback attractor for the cocycle Sξ (t) if

1) the set A(ξ) is compact for all ξ ∈ Ψ;
2) the set {A(ξ)}ξ∈Ψ is a pullback attracting set for Sξ (t);
3) it is the minimal (with respect to inclusion) set which has properties 1) and 2).
The next theorem gives the analogue of Theorem 5.6.
Theorem 5.12. Let Φ be a Hausdorff topological space, Ψ be a set, and let
Sξ (t) : Φ → Φ be a cocycle over the group T (h) : Ψ → Ψ, h ∈ R. Assume that this
cocycle possesses a compact pullback attracting set {B(ξ)}ξ∈Ψ . Then there exists
a pullback attractor {A(ξ)}ξ∈Ψ which is a compact subset of {B(ξ)}ξ∈Ψ .
Sketch of the proof. This statement is a straightforward corollary of Theo-
rem 5.6. Indeed, for every fixed ξ ∈ Ψ let us consider the corresponding dynamical
process Uξ (t, τ ) in the phase space Φ endowed with the bornology Bξ which consists
of time dependent sets Bξ (t) := B(T (t)ξ) for all B ∈ B. Then the dynamical pro-
cess Uξ (t, τ ) satisfies all the assumptions of Theorem 5.6, and therefore there exists
a pullback attractor Aξ (t), t ∈ R, for this process. Moreover, due to the translation
identity and the uniqueness of a pullback attractor, we have AT (h)ξ (t) = Aξ (t + h),
t, h ∈ R and ξ ∈ Ψ. Thus, the ξ-dependent set

A(ξ) := Aξ (0)

is well defined and gives the desired pullback attractor for the cocycle Sξ (t).
As in the previous case, if, in addition, the maps Sξ (t) are continuous for every
fixed t > 0 and ξ ∈ Ψ, then we have the strict invariance of the pullback attractor,
which now reads

Sξ (t)A(ξ) = A(T (t)ξ), ξ ∈ Ψ, t > 0.

Moreover, if, in addition, the compact attracting set B ∈ B and the bornology B is
stable with respect to inclusions, then we also have the representation formula

A(ξ) = Kξ t=0
,

where Kξ is the Bξ -kernel of the dynamical process Uξ (t, τ ).

As we have seen in Example 5.10, the forward attraction property may fail
even in the case of a relatively simple dependence of the symbols ξ( · ) ∈ Ψ on
time. The situation becomes much better if we assume some recurrence properties
of time-dependent external forces. Namely, assume that Ψ possesses an invariant
 ATTRACTORS. THEN AND NOW 105

(with respect to T (h)) probability measure µ, that is, (Ψ, F, µ), where F is
a σ-algebra, is a probability space and the maps T (h) are measure preserving.
Then it is also natural to assume that the maps ξ → Sξ (t) are measurable (or even
continuous) and for every B ∈ B the map ξ → B(ξ) is a measurable set-valued map
(the sets satisfying this property are called random sets, and the corresponding
cocycle is called a random dynamical system (RDS)). Moreover, in order to work
with random sets it is usually assumed that Φ is a Polish space (that is, it is
separable, metrizable, and complete). We also mention that it is natural for random
dynamical systems that all properties starting from (5.8) do not hold for all ξ ∈ Ψ,
but just for µ-almost all of them.
The detailed exposition of the theory of random dynamical systems is out
of scope of this survey; we refer the interested reader to [52], [51], [56], [55], [123],
and the references therein for more details, and restrict ourselves to mentioning
the random analogue of Theorem 5.12 and giving some examples.

Theorem 5.13. Let Sξ (t) : Φ → Φ be an RDS over the measure-preserving group

T (h) : Ψ → Ψ acting on the probability space Ψ, and let Φ be Polish. Assume
also that this RDS possesses a compact random attracting set B (that is, B(ξ)
is compact for almost all ξ ∈ Ψ) with respect to some bornology B which consists
of random sets. Then the RDS Sξ (t) possesses a pullback attractor which is a compact
random set in Φ. Moreover, if the maps Sξ (t) are continuous for almost all ξ , then
the attractor A(ω) is strictly invariant.

The proof of this theorem is reduced to Theorem 5.12 under the extra assumption
that the bornology B is separable, that is, there exists a countable bornology B0
such that for every B ∈ B there exists B0 ∈ B0 such that B ⊂ B0 (see [51]). In
the general case, when this property fails (for instance when the point random
attractors are considered), i is more delicate. In particular, it is not a priori
h S the proof
clear why the set ξ → ωB (ξ) , which contains a non-countable union, should
B∈B Φ
be a random set. Moreover, as shown in [56], this union can be significantly larger
than the desired random attractor A(ξ). In the example given there, the random
pullback attractor is a one-point set for almost all ξ ∈ Ψ and the above union
coincides with the whole space Φ almost surely. Nevertheless, as shown in [56],
the theorem remains true in the general setting as well (and even the measurability
of sets in B can be removed). We also mention the paper [206], where examples
of random attractors of a different type are given (in particular, examples where
the random attractor does not attract bounded sets forward in time).

Corollary 5.14. Under the assumptions of Theorem 5.13, the random attrac-
tor A(ξ) possesses the forward attraction property in the following sense: for every
B ∈ B and every ε > 0,


lim µ ξ ∈ Ψ : distΦ Sξ (t)B(ξ), A(T (t)ξ) > ε = 0, (5.12)
t→+∞

where distΦ (A, B) := supy∈A inf x∈B d(x, y) is a non-symmetric Hausdorff distance
in Φ.
106 S. V. ZELIK

Proof. Indeed, due to the pullback attraction property, for almost all ξ ∈ Ψ,
we have
lim distΦ (UT (−t)ξ (−t, 0)B(ξ), A(ξ)) = 0.
t→∞

Therefore, by Lebesgue’s dominated convergence theorem,

distΦ (Uξ (0, −t)B(T (−t)ξ), A(ξ))

Z
lim µ(dξ) = 0.
t→∞ ξ∈Ψ 1 + distΦ (UT (−t)ξ (−t, 0)B(ξ), A(ξ))

Since T (t) is measure preserving, the change of the independent variable ξ → T (t)η
in combination with the translation identity UT (t)η (0, −t) = Uη (t, 0) yields

distΦ (Uη (t, 0)B(η), A(T (t)η))

Z
lim µ(dη) = 0
t→∞ η∈Ψ 1 + distΦ (Uη (t, 0)B(η), A(T (t)η))

which gives the desired convergence in measure. The corollary is proved.

We illustrate the theory by few model examples; more examples can be found,
for instance, in [123], [206].
Example 5.15. Let us return to Example 5.10. It can be written in the form
of (5.11) with
(
3 0, t < 0,
A(y, ξ) := (2ξ − 1)y − ξy , ξ(t) := H(t) =
1, t > 0.

This non-autonomous dynamical system can be embedded in a cocycle, for example,

by introducing the hull

H(H) := {0} ∪ {1} ∪ {T (h)H, h ∈ R}

of the external forces H(t). We endow Ψ with the topology of L1loc (R), which makes
it a compact metric space, and T (h)Ψ = Ψ acts continuously on Ψ. After this,
we may consider the corresponding family of dynamical processes Uξ (t, τ ), ξ ∈ Ψ,
which satisfies the translation identity, so the cocycle associated with problem (5.7)
is defined.
The dynamical system (T (h), Ψ) carries exactly two ergodic invariant probability
measures: µ0 = δ(ξ) and µ1 = δ(ξ − 1). The supports of both of these measures
correspond to the autonomous equations with attractors Aµ0 = {0} and Aµ1 =
[−1, 1] both of which are forward “random” attractors.
Remark 5.16. We see that, although the measures constructed in the previous
example satisfy formally all of the assumptions of the theory, they are a posteriori
useless and do not give any additional information about the dynamical system
under consideration. This indicates a general problem of the theory of dynamical
systems and, in particular, ergodic theory. Namely, although the invariant measure
can be constructed on a general compact space using Prohorov’s theorem, this is
not enough to get reasonable information about the system under consideration,
and to do this we need some “good” properties of this measure (for instance, to have
a so-called “physical measure”), for example, its absolute continuity with respect to
 ATTRACTORS. THEN AND NOW 107

the Lebesgue measure if Ψ is a subset of Rn . Unfortunately, such a good measure

may not exist and it is not clear how to construct it in a more or less general
situation (see [116], [235] and the references therein for more details).

Example 5.17. Let us consider the damped nonlinear pendulum equation with
sign-changing dissipation:

y ′′ + γ(t)y ′ + y|y|p − αy = 0, ξy := {y, y ′ } t=τ

= ξτ , (5.13)

where γ ∈ L∞ (R) is a given damping coefficient, and p > 0 and α > 0 are fixed
parameters. We assume that γ ∈ Ψ, where Ψ is a compact set in L1loc (R) which is
strictly invariant with respect to time translations T (h) : Ψ → Ψ. Then the solution
operators Uγ (t, τ ) : Φ → Φ, Φ = R2 , satisfy the translation identity, and therefore
generate a cocycle Sγ (t) in Φ. Let us introduce the standard energy functional

1 ′ 2 1
Ey (t) := y (t) + |y(t)|p+2 .
2 p+2

It can be shown using some refined energy arguments that the following estimate
holds:
 Z t  
p+2
Ey (t) 6 CEy (s) exp − 2 γ(τ ) − κ dτ
s p+4
Z t  Z t  
p+2
+ Cκ exp − 2 γ(τ ) − κ dτ dm (5.14)
s m p+4

for every t > s ∈ R, every κ > 0, and every γ ∈ Ψ (see [30] where this estimate is
verified in the much more general setting of a hyperbolic PDE).
To get an RDS, we assume that there exists a Borel measure µ on the set Ψ
which is invariant with respect to shift operators T (h), h ∈ R, and is ergodic. We
also assume that the following dissipativity assumption holds:
Z Z 1
κ0 := γ(t) dt µ(dγ) > 0. (5.15)
γ∈Ψ 0

In order to specify the appropriate bornology B on Φ, we need to recall that

the function t → f (t), t ∈ R, is called tempered if, for every β > 0,

lim eβt |f (t)| = 0

t→−∞

(see [27], [120] for more details). We say that a random set {B(ξ)}ξ∈R is tempered
if the function t → ∥B(T (−t)ξ)∥Φ is tempered for µ-almost all ξ ∈ Φ. The standard
bornology which consists of all tempered sets {B(ξ)}ξ∈Ψ will be denoted by Btemp .
At the next step we use Birkhoff’s ergodic theorem in order to establish that there
exists a shift-invariant subset Ψerg ⊂ Ψ of full µ-measure such that
Z 0
1
lim γ(t) dt = κ0 > 0, γ ∈ Ψerg , (5.16)
T →∞ T −T
108 S. V. ZELIK

and the existence of this limit allows us to establish that the function
Z t  Z t  
p+2
bγ (t) := 2Cκ exp − 2 γ(τ ) − κ dτ dm (5.17)
−∞ m p+4

is well defined and tempered for all γ ∈ Ψerg if κ > 0 is small enough (the exponent κ
is fixed from now on in such a way that this condition is satisfied); see [30] for
the details. In turn, this fact, together with (5.14), allows us to verify that
the tempered random set

B(γ) := {ξy (0) ∈ Φ : Ey (0) 6 bγ (0)} (5.18)

is a compact pullback absorbing set for the RDS Sγ (t) : Φ → Φ associated with
equation (5.13); see [30]. Thus, according to the general theory, this RDS possesses
a tempered random attractor A(γ) ⊂ B(γ), γ ∈ Ψerg , which is strictly invariant and
possesses the standard representation formula:

A(γ) = Kγ t=0
, γ ∈ Ψerg , (5.19)

where the tempered kernel Kγ consists of all complete tempered trajectories

of equation (5.13).
As a natural example of a set Ψ and measure µ consider the set of random
piecewise constant dissipation rates

γl (t) := ln , t ∈ [n, n + 1), n ∈ Z,

where l = {ln }n∈Z ∈ Γ := {a, −b}Z . In other words, the dissipation rate γl (t)
may take only two values a and −b on the interval [n, n + 1), a, b > 0, which are
parameterized using the Bernoulli shift scheme Γ. We assume that the values a
and −b have probabilities q and 1 − q to appear, so we endow Γ with the standard
Bernoulli product measure µq . It is well known (see, for example, [116]) that
discrete Bernoulli shifts T (n) : Γ → Γ preserve this measure and that µq is ergodic
with respect to these shifts. Using the obvious fact that

T (n)γl = γT (n) , l ∈ Γ, n ∈ Z, (5.20)

we may lift the dynamical system (T (n), Γ) to the shift-invariant set Ψ ⊂ L∞ (R).
Namely, let
Ψ := {T (h)γl , h ∈ [0, 1], l ∈ Γ}.

Then Ψ is obviously shift invariant and is compact in L1loc (R). The commu-
tation (5.20) also allows us to lift the Bernoulli measure µq to the appropriate
shift-invariant measure µ̃q on Ψ. Finally, condition (5.15) now reads

κ0 = aq − b(1 − q) > 0

and gives us the sufficient condition for the existence of a random attractor A(γ)
(see [30] for the details).
 ATTRACTORS. THEN AND NOW 109

Remark 5.18. The above example demonstrates many typical features of attrac-
tor theory for random dynamical systems. First, we see the crucial role of Birkhoff’s
ergodic theorem or its probabilistic analogues in constructing random attractors.
Second, since the existence of a Birkhoff limit (5.16) is not guaranteed for all γ ∈ Ψ,
but only for almost all of them, the use of “µ-almost all” in all definitions related to
random attractors also looks unavoidable. In the exceptional measure-zero choices
of γ the corresponding dynamical process is simply not dissipative and does not
possess an attractor (for example, it will be so if we take γ(t) = −b < 0 for all
t ∈ R). We also mention that, at least in the example related to Bernoulli shifts,
the function bγ (t) is unbounded with probability one as t → ±∞. This explains
why it is important to develop the pullback attractor theory for non-autonomous
equations with unbounded in time coefficients (see also next example). The
bornology Btemp of tempered sets introduced above looks as an appropriate and
natural generalization of the bornology Bbound of uniformly bounded in time sets
(which is widely used in the deterministic case; see [12], [215]) to the dynamical
process related to random and stochastic dynamical systems; see [120], [123]. Note
however, that in many cases, the use of the bornology consisting of deterministic
(uniformly bounded) sets is enough to recover a random attractor in the proper
way; see [52], [56], and the references therein.

Our next example will be related to the simplest nonlinear stochastic ODE.

Example 5.19. Consider the following stochastic ODE:

y ′ + y 3 − y = εη ′ (t), y t=τ
= yτ , (5.21)

where η ′ (t) is a two-sided white noise, and ε > 0 is a small parameter.

Following the standard procedure we realize the two-sided Wiener process on
the space Ψ := C0 (R) of continuous functions η : R → R which are equal to zero
at t = 0 endowed with the locally compact topology and the standard Wiener
measure µ. Then the modified group of temporal shifts (T (h)η)(t) := η(t + h) − η(h)
acts on Ψ and the Wiener measure µ is invariant and ergodic (see, for example, [212]).
We present the solution y(t) = yη (t) in the form of yη (t) = εvη (t)+wη (t), where vη (t)
is the stationary Ornstein–Uhlenbeck process:
Z t
vη′ + vη = η ′ (t), that is, vη (t) = η(t) − η(s)es−t ds, (5.22)
−∞

and the remainder solves

wη′ − wη + (εvη (t) + wη )3 = 2εvη (t), wη t=0

= y0 − εvη (0). (5.23)

Thus, the random dynamical system Sη (t) : R → R over the group of shifts
T (h) : Ψ → Ψ is well defined and we may speak about its pullback random
attractor A(η). According to the general theory, to this end we only need to
construct a compact random absorbing set B(η) with respect to the bornology Btemp
of tempered random sets on Φ = R. In turn, this means that we need to construct
such an absorbing set for equation (5.23) for almost all η. This can be done in many
110 S. V. ZELIK

ways; for instance, multiplying the equation by wη (t) and using Young’s inequality,
one easily derives that
d 2
(w ) + wη2 6 C(1 + ε4 vη (t)4 )
dt η
for some deterministic constant C which is independent of ε. Therefore, for all
t − s > 0 we have
Z t
wη (t)2 6 wη (s)2 es−t + C eκ−t (1 + ε4 vη (κ)4 ) dκ. (5.24)
s

It only remains to note that the function

Z t
bη (t) := C eκ−t (1 + ε4 vη (κ)4 ) dκ
−∞

is tempered for almost all η, as also is the function vη (t) and, consequently
the random set
B(η) := vη (0) + y0 ∈ R : y02 6 bη (0)


is a compact tempered pullback absorbing set for the RDS Sη (t) constructed, which
is associated with equation (5.22). Thus, according to the general theory, this
system possesses a tempered pullback random attractor Aε (η) which is generated
by all complete tempered trajectories of (5.22) (see [53] for the details).
The RDS constructed possesses some remarkable properties. Namely, solving
the associated Kolmogorov–Fokker–Planck equation, we see that

exp{ε−1 (−y 4 /4 + y 2 /2)}

ν(dy) = Fε (y) dy, Fε (y) := R , (5.25)
R
exp{ε−1 (−y 4 /4 + y 2 /2)} dy

is an invariant (stationary) measure for the stochastic process under consideration.

Moreover, this measure is uniquely ergodic and mixing (see [53], [117] for the details).
Due to the ergodicity, the corresponding Lyapunov exponent is given by

1 3 I−3/4 (1/(8ε)) + I3/4 (1/(8ε))

Z
1
λ(ε) := (1 − 3y 2 )Fε (y) dy = − − 6 − < 0,
R 2 2 I−1/4 (1/(8ε)) + I1/4 (1/(8ε)) 2
(5.26)
where Iκ (z) is the modified Bessel function of the first kind and order κ.
Since the Lyapunov coefficient is negative for all ε > 0, the attractor Aε (η)
consists of a single point Aε (η) = {uε,η (0)} and the corresponding tempered
kernel Kε,η consists of a single exponentially stable trajectory uε,η (t) for almost
all η > 0. Thus, adding an arbitrary small additive noise destroys the pitchfork
instability and makes the random system exponentially stable. In particular,
the limit deterministic attractor A0 = [−1, 1] is not robust with respect to random
perturbations (see [53] for more details and a related discussion).
Remark 5.20. The stabilization effect of adding a small additive noise appears
in much more general situations, including stochastic reaction-diffusion equations,
and so on. The key ingredients here are: 1) the uniqueness and ergodicity
 ATTRACTORS. THEN AND NOW 111

of the invariant measure which is known in more or less general situation if

the noise is not too degenerate (see [117], [142], and the references therein);
2) the order-preserving structure which allows us to construct the maximal and
minimal solutions belonging to the attractor. If these two solutions coincide almost
surely, then the attractor consists of a single point as in the above example, and if
not, then every of them carries an invariant measure which contradicts uniqueness
(see [42] for the details). Note that this mechanism does not work if the limiting
deterministic dynamics is really chaotic and have positive Lyapunov exponents, for
instance, in the case of the Lorenz system (2.9) perturbed by a small additive noise,
so we expect non-trivial attractors and rich dynamics in the stochastic case as well.
Unfortunately, despite the solid numerical evidence, we failed to find a rigorous
mathematical proof of this fact.
5.3. Uniform attractors. We now turn to the alternative approach to attrac-
tors for non-autonomous equations, which is based on their reduction to autonomous
ones. We start with a cocycle Sξ (t) : Φ → Φ, over T (h) : Ψ → Ψ, where Φ and Ψ
are Hausdorff topological spaces, and the corresponding family Uξ (t, τ ), t > τ ,
ξ ∈ Ψ, of dynamical processes. Consider the extended phase space P := Φ × Ψ and
the associated extended dynamical system S(t), t > 0, on it defined by

S(t)(u0 , ξ) := (Sξ (t)u0 , T (t)ξ), u0 ∈ Φ, ξ ∈ Ψ. (5.27)

Indeed, it follows from the cocycle property that S(t) is a semigroup. Therefore, we
end up with the autonomous dynamical system acting on the extended phase space P
and may speak about its attractors. To this end, we need to fix a bornology B on
the space Φ (in contrast to the previous cases, B ∈ B are time-independent sets
B ⊂ Φ) and to define the extended bornology Bext on P via B ⊂ Bext if Π1 B ∈ B
where Π1 : P → Φ is a projection onto the first component of the Cartesian product.
Definition 5.21. A set B ⊂ Φ is a uniformly attracting set for the cocycle Sξ (t),
ξ ∈ Ψ, with respect to the bornology B if for any B ∈ B and any neighbourhood O(B)
of B in Φ, there exists T = T (O, B) such that

Sξ (t)B ⊂ O(B) ∀ ξ ∈ Ψ if t > T.

It is not difficult to see that if the set B is a uniformly attracting set for
the cocycle Sξ (t), then the set Bext := B × Ψ is an attracting set for the extended
dynamical system S(t) : P → P with respect to the bornology Bext . Vice versa,
if B is an attracting set for S(t), then B := Π1 B is a uniformly attracting set for
the cocycle Sξ (t), ξ ∈ Ψ.
Definition 5.22. Let Φ be a Hausdorff topological space and let Sξ (t) : Φ → Φ
be a cocycle over the dynamical system T (h) : Ψ → Ψ, h ∈ R. Then the set Aun ⊂ Φ
is a uniform attractor for this cocycle with respect to some bornology B if
1) Aun is a compact set in Φ;
2) Aun is a uniformly attracting set for the cocycle Sξ (t);
3) Aun is a minimal (by inclusion) set which has properties 1) and 2).
An analogue of the existence theorem for attractors in this case reads as follows.
112 S. V. ZELIK

Theorem 5.23. Let Φ and Ψ be Hausdorff topological spaces, and let Ψ be

a compact space. In addition, let Sξ (t) : Φ → Φ be a cocycle over the group
T (h) : Ψ → Ψ, h ∈ R, and let this cocycle possess a compact uniformly attracting
set B with respect to some bornology B on Φ. Then the extended semigroup
S(t) : P → P possesses an attractor A ⊂ B × Ψ. Moreover, its projection Aun := Π1 A
is a uniform attractor for the cocycle Sξ (t).
Assume, in addition, that the map (u0 , ξ) → (Sξ (t)u0 , T (t)ξ) is continuous for
every fixed t. Then the attractor A is strictly invariant with respect to S(t). If
also B ∈ B and B is stable with respect to inclusions, then Aun is generated by all
complete B-bounded trajectories of Sξ (t), that is,
[
Aun = Kξ t=0 , (5.28)
ξ∈Ψ

where Kξ is a B-kernel of the dynamical process Uξ (t, τ ). In particular, since

the pullback attractor Aξ,pb (τ ) of the dynamical process Uξ (t, τ ) is equal to Kξ t=τ ,
the relation [
Aun = Aξ,pb (0)
ξ∈Ψ
holds.
Sketch of the proof. All the statements of the theorem are straightforward
corollaries of our key result, Theorem 3.3, and Propositions 3.4 and 3.6. Indeed,
if B is a uniformly attracting set for the cocycle Sξ (t), then B := B × Ψ is a compact
attracting set for the extended semigroup S(t) : P → P. Thus, the existence
of the attractor A is verified. The fact that Aun = Π1 A is the desired uniform
attractor is also straightforward. Indeed, compactness and the attraction property
are obvious and we only need to check minimality. Let B1 be another uniformly
attracting set; then by the minimality of A, we have A ⊂ B1 × Ψ and Aun ⊂ B1 .
The remaining statements are immediate corollaries of Propositions 3.4 and 3.6.
Note that the definition of a uniform attractor given above does not require any
topology on the space Ψ. Moreover, its existence can be obtained on the basis of
the existence of a uniformly attracting set, namely, as in the autonomous case, we
may define the uniform ω-limit set
\ [ [ 
ωun (B) := Sξ (t)B (5.29)
T >0 ξ∈Ψ t>T Φ
h S i
and construct the uniform attractor via Aun := ωun (B) (see [38] for
B∈B Φ
the details). Then the compactness of Ψ (as well as the topology on it) is not
necessary to get a uniform attractor. In particular, we even may construct a uniform
attractor for a single dynamical process U (t, τ ) : Φ → Φ by introducing the group
of shifts T (h) : Ψ → Ψ acting on the space Ψ = R via T (h)ξ := ξ + h and by defining
the corresponding cocycle Sξ (t) := U (t + ξ, ξ). This corresponds to the trivial
reduction of a non-autonomous equation y ′ = f (t, y) to the autonomous system
(
ẏ = f (t, y),
ṫ = 1.
 ATTRACTORS. THEN AND NOW 113

The drawback of this approach is that it does not give any information on
the structure of the uniform attractor (for example, the representation formula (5.28)
fails without the compactness of Ψ) and it is not clear how to relate the attractor Aun
with the solutions of the PDE under consideration. Thus, if we want to have
the representation formula (5.28), we need to consider not only all time shifts
of our original equation, but also their limits in the appropriate topology. The next
proposition shows that, under natural assumptions, taking this closure does not
affect the size of the attractor.
Proposition 5.24. Let Φ and Ψ be two Hausdorff topological spaces, and let
Sξ (t) : Φ → Φ be a cocycle over a dynamical system T (h) : Ψ → Ψ. Assume also
that Ψ0 ⊂ Ψ is a dense set invariant with respect to T (h), h ∈ R, and that
the map ξ → Sξ (t)u0 is continuous for every fixed t and u0 ∈ Φ. Then the uniform
attractors AΨ and AΨ0 of Sξ (t), considered as the cocycles over T (h) : Ψ → Ψ and
T (h) : Ψ0 → Ψ0 , exist or do not exist simultaneously and coincide:

AΨ = AΨ0 . (5.30)

Sketch of the proof. Indeed, it is not difficult to show using continuity that
a set B is a compact uniformly attracting set for the cocycle Sξ (t), ξ ∈ Ψ, if and
only if it is a compact uniformly attracting set for Sξ (t), ξ ∈ Ψ0 , and this gives
the desired result (see, for example, [38]).
We see that in order to get the key representation formula (5.28), we need to take
a closure of time shifts of the dynamical process under consideration in a topology
which, on the one hand, makes the closure a compact Hausdorff topological space
and, on the other hand, preserves continuity. This is a non-trivial task, which does
not always have a positive solution (in that case the representation formula will be
lost); see, for example, [205]. We restrict ourselves to an important particular case
(well adapted to the study of equations with additive non-autonomous external
forces), where this problem possesses a more or less complete solution. Namely, we
assume that Φ is a separable reflexive Banach space and the elements of Ψ which
represent the non-autonomous external forces are functions ξ : R → H with values
in another separable reflexive Banach space H. Moreover, assume that we start
from a given external force ξ0 which satisfies

ξ0 ∈ Lpb (R, H) (5.31)

for some 1 < p < ∞. Then, due to the Banach–Alaoglu theorem, any bounded set
in Lploc (R, H) is precompact in the weak topology. Therefore, the hull

H(ξ0 ) := [T (h)ξ0 , h ∈ R]Lp,w

loc (R,H)
(5.32)

is a compact subset of Lp,w loc (R, H), so we may naturally take Ψ = H(ξ0 ). The
scheme works as follows: we start with equation (5.11) with a given non-autonomous
external force ξ0 ∈ Lpb (R, H) and consider the whole family of similar problems
generated by its shifts in time and appropriate limits. Namely, we consider the family

∂t u = A(u, ξ(t)), ξ ∈ Ψ := H(ξ0 ), u t=0

= u0 ∈ Φ. (5.33)
114 S. V. ZELIK

If these equations are uniquely solvable in an appropriate sense, then they define
a cocycle Sξ (t) : Φ → Φ over the group T (h) : Ψ → Ψ of time shifts. We fix
weak topologies on both spaces Φ and Ψ and let the bornology B on Φ consist
of all bounded subsets of the Banach space Φ. Since Φ is reflexive, any B ∈ B is
precompact in the weak topology, so if we find a uniformly absorbing/attracting
set B ∈ B for the cocycle Sξ (t) under consideration, then its closed convex hull will
be a compact bounded uniformly absorbing set. This, together with Theorem 5.23,
gives us the following result.
Corollary 5.25. Let the above assumptions hold, and let the cocycle Sξ (t) :
Φ → Φ over T (h) : Ψ → Ψ with Ψ = H(ξ0 ) possess a bounded uniformly absorbing
set. Assume also that the map (u0 , ξ) → Sξ (t)u0 is continuous for every fixed t > 0.
Then this cocycle possesses a uniform attractor Aw un in a weak topology of Φ, and
this attractors can be described as follows:
[
Awun = Kξ t=0 , (5.34)
ξ∈H(ξ0 )

where Kξ is a bounded kernel of the dynamical process Uξ (t, τ ).

Note that, as usual, the continuity assumption can be replaced by the closed
graph assumption.
We now turn to the case of strong attractors, so we want to fix a strong topology on
the Banach space Φ. Then, according to the general theory, the existence of a uniform
attractor will be guaranteed if we find a compact (in the strong topology) uniformly
attracting set for the cocycle Sξ (t) associated with (5.33). As elementary examples
show, assumption (5.31) is usually not enough to get this compactness, so some
extra assumptions are needed (see [244]). The most natural way would be to assume
that the hull H(ξ0 ) is compact not only in the weak topology of Lploc (R, H), but
also in the strong topology. The functions ξ0 ∈ Lpb (R, H) which satisfy this extra
assumption are called translation-compact (ξ0 ∈ Lptr-c (R, H)). Then Ψ is compact
in the strong topology as well and we can fix the strong topology in both Φ and Ψ
and get a complete analogue of Corollary 5.25 for the case of strong topology as
well. This scheme, which works for many important dissipative PDEs, was studied
in details in [38] (see also [36], [179]), so we do not give further details here. Instead,
we discuss a bit more delicate case where we take the strong topology on Φ, but
try to keep the weak topology on the hull Ψ = H(ξ0 ). The possibility to find
strong uniform attractors for non-translation-compact external forces was indicated
in [159] (see also [157], [158], [160]–[162], [240], [244]). Then we still have an analogue
of Corollary 5.25, since we take the strong and weak topologies on the first and
second components of the Cartesian product P := Φ × H(ξ0 ), so Theorem 5.23 is
still applicable if we have a compact uniformly attracting set for the associated
cocycle. Moreover, it is not difficult to show that in this case the uniform attractor
in the strong topology coincides with the already constructed Aw un :

Asun = Aw
un

and we may use (5.34) to describe the structure of the strong attractor Asun . Note
that we need not verify the continuity of the maps (ξ, u0 ) → Sξ (t)u0 ) in the weak
 ATTRACTORS. THEN AND NOW 115

topology on Ψ and the strong topology on Φ (which is usually not true) and
may check continuity (or the closedness of the graph) only in the weak topologies;
see [244] for more details.
To continue, we need to introduce some classes of external forces.
Definition 5.26. Let ξ0 ∈ Lpb (R, H), where H is a reflexive Banach space and
1 < p < ∞. We say that ξ0 is time-regular if there exists a sequence ξn ∈ Cb1 (R, H)
such that
lim ∥ξn − ξ0 ∥L2b (R,H) = 0. (5.35)
n→∞

We denote the class of such functions by L2t-reg (R, H).

Analogously, we say that ξ0 is space-regular, if there exist a sequence of finite-di-
mensional Banach spaces Hn ⊂ H and a sequence of functions ξn ∈ L2b (R, Hn ) such
that (5.35) holds. The class of such functions is denoted by Lpsp-reg (R, H).
The function ξ0 ∈ Lpb (R, H) is called normal if
Z t+τ
lim sup ∥ξ0 (s)∥pH ds = 0.
τ →0 t∈R t

The class of such functions is denoted by Lpnorm (R, H).

Remark 5.27. It was shown in [244] that

Lptr-c (R, H) = Lpt-reg (R, H) ∩ Lpsp-reg (R, H),

and both spaces on the right-hand side are strictly larger than the space
of translation-compact external forces. The machinery for obtaining uniform
attractors in a strong topology for various PDEs (including damped wave equations,
reaction-diffusion equations, and so on) with non-autonomous external forces from
these classes is also presented there.
Note that the assumption 1 < p < ∞ is crucial for the theory since it guarantees
the reflexivity of the space Lploc (R, H) and the possibility to use the Banach–Alaoglu
theorem. However, the case p = 1 is also interesting from the point of view
of applications, especially for wave equations where it is naturally related to
Strichartz’s estimates (see [18], [23], [115], [234]). The situation here is much more
delicate since, starting from a regular function ξ0 ∈ L1b (R, H), we may easily get
an H-valued measure when taking the closure in an appropriate topology. In turn,
this leads in a natural way to a dynamical process with trajectories wich are not
continuous in time (see [205]. The theory developed there is based on presenting
a function ξ0 ∈ L1b (R, H) as a regular Borel measure ξ0 ∈ Mb (R, H). Then, using
the fact that M (0, 1; H) = [C(0, 1; H)]∗ , we may endow the space Mb (R, H) with
the local w∗ -topology and consider the hull H(ξ0 ) in this topology. This is the way
how to restore the compactness of the hull, but as a price to pay, we may lose
the continuity of the map ξ → Sξ (t) (see [205] for more details).
We illustrate the theory by the example of the 2D Navier–Stokes system with
non-autonomous normal external forces.
Example 5.28. Consider the system

∂t u + (u, ∇x )u + ∇x p = ν∆x u + ξ0 (t), div u = 0, u t=0

= u0 , (5.36)
116 S. V. ZELIK

in a bounded 2D domain endowed with Dirichlet boundary conditions. We use

the notation of subsection 4.3 adapted to the 2D case. The initial data u0 are taken
from the phase space Φ which is the closure of the space V of divergence-free test
functions in the L2 -norm and the spaces V and V −1 are defined analogously, and
we assume that ξ0 ∈ L2b (R, V −1 ). Also we define a weak energy solution of (5.36)
exactly as in subsection 4.3.
It is also well known that, in contrast to the 3D case, in the 2D case, a weak
energy solution is unique and satisfies the energy identity
1 d
∥u(t)∥2Φ + ν∥∇x u(t)∥2L2 = (ξ0 (t), u(t)) (5.37)
2 dt
(see [12], [215] for the details). This identity gives us the dissipative estimate
Z t
2
∥u(t)∥Φ + ν e−β(t−s) ∥∇x u(s)∥2 ds 6 ∥u(0)∥2Φ e−βt + C∥ξ0 ∥2L2 (R,V −1 ) , (5.38)
b
0

for some positive constants β and C. We now consider the hull Ψ = H(ξ0 )
of the external force ξ0 in L2,w loc (R, H), and for every ξ ∈ Ψ we define a map
Sξ (t) : Φ → Φ as a solution operator at time t for problem (5.36), where ξ0 is replaced
by ξ. Obviously, Sξ (t) is a cocycle over T (h) : Ψ → Ψ, and the symbol space Ψ is
compact if we fix a weak topology on it. We also fix the standard bornology B which
consists of all bounded subsets of the Banach space Φ. Moreover, it is not difficult to
see that estimate (5.38) is uniform with respect to ξ ∈ H(ξ0 ) (see, for example, [38]),
so by the Banach–Alaoglu theorem the closed ball BR := {u0 ∈ Φ : ∥u0 ∥Φ 6 R} ∈ B
is a compact uniformly absorbing set for this cocycle if R is large enough. The
continuity of the map (ξ, u0 ) → Sξ (t)u0 in the chosen weak topologies on Φ and Ψ
is also straightforward and, due to Corollary 5.25, this cocycle possesses a uniform
attractor Aw un , which enjoys the representation formula (5.34).
We now turn to the case of strong topology in Φ. Since the existence of a uniform
attractor Aw un in the weak topology of Φ together with the representation formula
has already been established, we only need to find a compact (in the strong topology
of Φ) uniformly attracting (or even absorbing) set for this cocycle. We assume,
in addition, that ξ0 ∈ L2norm (R, V −1 ) and claim that the set
B := {Sξ (1)BR , ξ ∈ Ψ} (5.39)
is such a set. Indeed, from estimate (5.38) we see that B is a bounded uniformly
absorbing set, so we only need to verify compactness. To this end we use
the energy method. Namely, we consider arbitrary sequences ξn ∈ Ψ and un0 ∈ BR
and the sequence of the corresponding solutions un (t) := Sξn (t)un . Without loss
of generality we may assume that ξn → ξ and un0 → u0 in the weak topology and, due
to the weak continuity of Sξ (t), we conclude that un (1) ⇁ u(1) where u(t) = Sξ (t)u0 .
Thus, we only need to prove that un (1) → u(1) in the strong topology of Φ. In
turn, this will be proved once we have checked that ∥un (1)∥2Φ → ∥u(1)∥2Φ . To get
this convergence we rewrite the energy identity (5.37) in the form
d
(t∥un (t)∥2Φ ) + N (t∥un (t)∥2Φ ) + 2νt∥∇x un (t)∥2L2
dt
= (N t + 1)∥un (t)∥2Φ + 2t(un (t), ξn (t)),
 ATTRACTORS. THEN AND NOW 117

where N is an arbitrary positive number, multiply it by eN t , and integrate over

t ∈ [0, 1] to get the integral identity
Z 1
∥un (1)∥2Φ + 2ν e−N (1−t) t∥∇x un (s)∥2L2 ds
0
Z 1 Z 1
−N (1−t)
= e (N t + 1)∥un (t)∥2Φ ds + 2 e−N (1−t) t(un (t), ξn (t)) ds. (5.40)
0 0

We want to pass to the limit n → ∞ in (5.40). Using the compactness lemma

and arguing in a standard way (see, for example, [244]), we conclude that un → u
strongly in L2 (0, 1; Φ), so the passing to the limit in the first term on the right-hand
side is straightforward. To pass to the limit in the second term on the left-hand
side we use the weak lower semicontinuity of convex functions, so it only remains to
estimate the last term on the right-hand side. To this end, we use the key property
of normal functions, namely, that
Z t
lim sup sup e−N (t−s) ∥ξ(s)∥2V −1 ds = 0
N →∞ ξ∈H(ξ0 ) t∈R 0

(see, for example, [244]). Using this fact in combination with the uniform bound-
edness of un in L2 (0, 1; V ), we see that, for every ε > 0 there exists N = N (ε) such
that Z 1
e−N (1−t) (un (t), ξn (t)) dt 6 ε
0

and the same is true for the limit functions u and ξ. Now, passing to the limit
n → ∞ in (5.40), we get
Z 1
lim sup ∥un (1)∥2Φ + 2ν e−N (1−t) t∥∇x u(s)∥2L2 ds
n→∞ 0
Z 1
6 e−N (1−t) t(N t + 1)∥u(t)∥2Φ ds + 2ε (5.41)
0

and the comparison with the analogue of (5.40) for the limit functions u and ξ gives

∥u(1)∥2Φ 6 lim inf ∥un (1)∥2Φ 6 lim sup ∥un (1)∥2Φ 6 ∥u(1)∥2Φ + 4ε, (5.42)
n→∞ n→∞

where the first inequality is again a weak lower semicontinuity for a convex function.
Finally, passing to the limit ε → 0, we arrive at

lim ∥un (1)∥2Φ = ∥u(1)∥2Φ

n→∞

which finishes the proof that B is compact. A general theory now gives the existence
of a uniform attractor Asun in the strong topology of Φ and its coincidence with Aw
un .

Remark 5.29. It is straightforward to see that L2t-reg (R, H) ⊂ L2norm (R, H), so
the result obtained immediately gives the existence of a strong uniform attractor
in the case of time-regular external forces, but gives nothing for space-regular ones.
118 S. V. ZELIK

The concept of a normal function can be weakened (following [157], see also [244]),
namely, a function ξ0 ∈ Lpb (R, H) is weakly normal (ξ0 ∈ Lpw-norm (R, H)) if for
every ε > 0 there exist a finite-dimensional subspace Hε ⊂ H and a function
ξε ∈ Lpb (R, Hε ) such that
Z t+h
lim sup sup ∥ξ0 (s) − ξε (s)∥pH ds 6 ε.
h→0 t∈R t

Then, on the one hand

Lpt-reg (R, H) + Lpsp-reg (R, H) ⊂ Lpw-norm (R, H),

so the class of weakly regular functions includes both space- and time-regular
functions. On the other hand, the method presented in Example 5.28 can easily
be extended to the case of weakly normal external forces ξ0 ∈ L2w-norm (R, V −1 );
see [244] for the details. In particular, this gives a unified proof of the existence
of a strong uniform attractor for the 2D Navier–Stokes equations with space-regular
or time-regular external forces.
We however note that the class of normal external forces is mainly adapted
to parabolic equations and, in contrast to space or time regularity, the normality
of the external forces in an appropriate space is not sufficient to have a strong
uniform attractor for, say, damped wave equations (see [244]).

Remark 5.30. To conclude this section, we mention briefly that the general
results of Theorems 3.3 and 5.6 are also well adapted to developing the theory
of trajectory attractors for problems in the form (5.33) without the uniqueness
of solutions. The detailed exposition of this topic can be found in [38], so we just
explain schematically the main ideas. To this end, we introduce, for every ξ ∈ H(ξ0 ),
the corresponding set Kξ+ of solutions of (5.33) defined on a semi-interval R+ .
Analogously to the autonomous case, this may be the set of all weak solutions or
some subset of it consisting of some special solutions, but we need to satisfy the key
invariance assumption:

T (h) : Kξ+ ⊂ KT+(h)ξ , ξ ∈ H(ξ0 ), t > 0.

Then we define the set K+ := Kξ+ and consider the trajectory dynamical
S
ξ∈H(ξ0 )
system (T (h), K+ ), which gives a trajectory analogue of the extended dynamical
system (5.27). Then we may construct an attractor for this trajectory dynamical
system by verifying the assumptions of Theorem 3.3, and this gives us a uniform
trajectory attractor for problem (5.33) (see [38] for more details).
Alternatively, we may consider maps T (h) : Kξ+ → KT+(h)ξ , h > 0, as a cocycle
over the dynamical system T (h) : H(ξ0 ) → H(ξ0 ), h ∈ R. Then, fixing some
topologies on the spaces Kξ+ , we get a family of dynamical processes Uξ (t, τ ) :=
T (t − τ ) : KT+(τ )ξ → KT+(t)ξ and may use Theorem 5.6, our key result, in order to
construct the trajectory analogues of pullback attractors (see [246] for more details).
 ATTRACTORS. THEN AND NOW 119

6. Dimensions of the attractor

In this section we start to discuss the finite-dimensionality of attractors related

to dissipative PDEs. Note from the very beginning that attractors are usually not
regular, but fractal subsets of the phase space, so we need to use generalizations
of dimension which are suitable for fractal sets. Actually there are many such
generalizations like Lebesgue covering dimension, Hausdorff, fractal, Lyapunov, and
Assaud dimensions, and so on. In general, all of them may be different (see, for
instance, [64], [193] and also Example 2.6).
One of the main motivations to study the dimensions of attractors is to give
a rigorous justification of the heuristic idea that, despite the infinite-dimensionality
of the initial phase space, the limit dynamics of many important dissipative PDEs
is essentially finite-dimensional and can be described by means of finitely many
parameters (the order parameters in the terminology of Prigogine, see [187]) whose
evolution is governed by a system of ODEs. This finite-dimensional reduction would
allow us to reduce the study, say, of turbulence which is described by Navier–Stokes
equations to the study of a system of ODEs which can be further investigated
using the methods of classical dynamics. Unfortunately, despite many efforts in this
direction, the finite-dimensional reduction mentioned above remains a “mystery”
and the existing theory is still far from being complete (see [64], [243], [193], and
the references therein for more details).
6.1. Mané’s projection theorem and finite-dimensional reduction. In
this subsection we restrict ourselves to the best studied case of fractal dimension.
Definition 6.1. Let A be a (pre)compact set in a metric space Φ. Then, by
Hausdorff’s criterion, for any ε > 0 it can be covered by finitely many balls of radius ε
in Φ. Let Nε (A, Φ) be the smallest number of such balls. Then the Kolmogorov
entropy of A in Φ is the following quantity:

Hε (K, Φ) := log2 Nε (K, Φ),

where the base 2 of the logarithm comes from information theory (see [125] and
the references therein for more details). The (upper) fractal dimension of K in Φ
is defined as follows:
Hε (K, Φ)
dimf (K, Φ) := lim sup . (6.1)
ε→0 log2 (1/ε)

It is well known that dimf (K, Φ) equals n if K is an n-dimensional Lipschitz

manifold, but it may not be an integer if K has a fractal structure (for example,
dimf (K, [0, 1]) = (ln 2)/(ln 3) for the standard ternary Cantor set in [0, 1]). Roughly
speaking, Nε (K, Φ) ∼ (1/ε)κ if dimf (K, H) = κ. Also note that it can easily
occur in the case where Φ is infinite-dimensional that dimf (K, Φ) = ∞. This
simply means that Nε (K, Φ) has a stronger divergence rate as ε → 0 than (1/ε)κ .
In this case the problem of finding/estimating the fractal dimension transforms
naturally into the problem of finding the asymptotic behaviour of Nε (A, Φ) as ε → 0
(see [125], [217], [239], [179] for more details). We also mention that sometimes,
instead of covering the set A by ε-balls, coverings by sets of diameter less than or
120 S. V. ZELIK

equal to ε are used. Although this does not affect the value of the fractal dimension,
this may be more suitable for an estimate of it since the problem of whether or not
the centres of ε-balls belong to A disappears in this more general setting.
The applications of the fractal dimension to the above-mentioned finite-dimensio-
nal reduction problem are based on the following Mané projection theorem (see [169]
and also [106], [193], and the references therein).
Theorem 6.2. Let A be a compact subset of a Hilbert space H such that
dimf (A, H) 6 n for some n ∈ N. Then the orthoprojector PL onto a “generic” plane
L ⊂ H of dimension dim L > 2n + 1 is one-to-one on A.
Thus, since A is compact, PL : A → A := PL A ⊂ L is a homeomorphism. If A is
an attractor of a continuous dynamical system S(t) : H → H, then we may project
this semigroup onto the one acting on a finite-dimensional space L ∼
= R2n+1 :

S(t) : A → A, S(t) := PL ◦ S(t) ◦ PL−1 , (6.2)

and therefore the dynamics on A is indeed described by a continuous semigroup

acting on a compact subset of R2n+1 . This is exactly the way how we may realize
the desired finite-dimensional reduction on the basis on the finiteness of the fractal
dimension of the attractor and Mané’s projection theorem. Moreover, we may even
write out a system of ODEs on the order parameters y(t) := PL u(t) if u(t) ∈ A
solves a general PDE of the form

∂t u = A(u).

Namely, applying the projector PL to both sides of this equation, we formally get

d
y(t) = PL A(PL−1 y(t)) =: F(y(t)), y(t) ∈ A ⊂ L ∼
= R2n+1 . (6.3)
dt
This ODE is often referred to as an inertial form of the initial PDE. Thus, under
this approach, the fractal dimension of the attractor A is interpreted as the number
of effective degrees of freedom in the reduced inertial form. This, in turn, motivates
the great interest to various methods for obtaining upper and lower bounds for this
dimension (see [12], [179], [215], and the references therein).
Remark 6.3. There are many versions of Mané’s projection theorem, in partic-
ular, the Hölder Mané projection theorem, which allows us to establish the Hölder
continuity of the inverse map PL−1 : A → A. Moreover, under some extra conditions,
which are usually satisfied at least in the case of semilinear parabolic equations,
the Hölder exponent can be made arbitrarily close to 1 by increasing the dimension
of L. The fact that H is Hilbert is also not essential, and all these results remain
true in Banach spaces (see [193] and the references therein).
However, the finiteness of the fractal dimension is not enough to guarantee that
it is possible to find L in such a way that PL−1 is Lipschitz, so in general the inertial
form constructed in such a way has only a Hölder continuous vector field F, and this
is the key drawback of the described approach. Indeed, the Hölder continuity of F(y)
is not enough even to establish the uniqueness of solutions for the inertial form (6.3),
 ATTRACTORS. THEN AND NOW 121

so it is not clear how to select the “physically relevant” solutions of this system
without referring to the initial PDE. In addition, there are ever more examples
where the dynamics on the attractor A remains, in a sense, infinite-dimensional (for
example, admits super-exponentially stable limit cycles, decaying travelling waves
in the Fourier space, and so on) despite the finiteness of the fractal dimension. These
attractors cannot be embedded in any finite-dimensional Lipschitz or log-Lipschitz
submanifold of the phase space (see [64], [133], [243] and also Section 7 for more
details). Thus, despite the widely accepted paradigm, the finite-dimensional reduc-
tion based on Mané’s projection theorem does not look as an appropriate solution
of the problem and the fractal dimension of the attractor is not an appropriate tool
for estimating the effective number of degrees of freedom of the reduced system.
We will discuss the alternative methods in the next sections and the rest of this
section is devoted to upper and lower bounds for the fractal dimension.

6.2. Upper bounds via the squeezing property: the autonomous case.
Assume that the attractor A ⊂ Φ of the dynamical system under consideration
has already been constructed and the maps S(t) are continuous. Then, for every
t0 ∈ R+ , we have
S(A) = A, S := S(t0 ) (6.4)
and the key question is: under what assumptions on the map S can we guarantee
that A has a finite fractal dimension? The next theorem gives the simplest of such
assumptions.
Theorem 6.4. Let Φ and Φ1 be two Banach spaces, and let the embedding Φ1 ⊂ Φ
be compact. Also assume that A is a bounded subset of Φ1 and the map S : A → A
satisfies (6.4) and the following squeezing/smoothing property:

∥S(u1 ) − S(u2 )∥Φ1 6 L∥u1 − u2 ∥Φ , u1 , u2 ∈ Φ. (6.5)

Then the fractal dimension of A in Φ1 is finite and enjoys the following estimate:

dimf (A, Φ1 ) 6 H1/(4L) (Φ1 ,→ Φ), (6.6)

where H1/(4L) (Φ1 ,→ Φ) is the entropy of the unit ball of the space Φ1 considered as
a precompact set in Φ.
N (ε)
Proof. Indeed, assume that we have already constructed an ε-net {uk }k=1
of the set A in Φ1 (that is, the ε-balls of Φ1 with centres in uk ∈ A cover A). Let
us cover every of these balls by N balls of radius ε/(4L) in Φ. This is possible due
to the compactness of the embedding Φ1 ⊂ Φ. Moreover, the number of balls which
are necessary to cover every one of the balls does not exceed

N := Nε/(4L) (Bε (uk , Φ1 ), Φ) = N1/(4L) (B1 (0, Φ1 ), Φ).

Crucial for us is that this number is independent of ε and uk . This gives us a covering
of A by ε/(4L)-balls in Φ, and the number of these balls does not exceed N N (ε).
Moreover, increasing the radii of the balls by the factor of two, we may assume also
N N (ε)
that the centres {vk }k=1 of these balls belong to A.
122 S. V. ZELIK

Then, due to the invariance of A and condition (6.5), the ε/2-balls in Φ1

centred at {S(vk )} cover A. Thus, starting from an ε-covering of A which consists
of N (ε) elements, we end up with a new ε/2-covering, with the number of elements
N (ε/2) 6 N N (ε). Since A is bounded, we may start with some ε0 = R0 such
that A ⊂ BR0 (u0 , Φ1 ), and therefore N (ε0 ) = 1. Iterating the procedure described
above, we finally get the εn := R0 2−n -coverings which consist of

N (εn ) = N (R0 2−n ) 6 N n

elements. Thus,

Hε (A, Φ1 ) 6 n log2 N 6 nH1/(4L) (Φ1 ,→ Φ), R0 2−n 6 ε 6 R0 2−n+1

and this gives us the desired estimate for the fractal dimension of A.
Remark 6.5. To the best of our knowledge, the key idea used in the proof
of this theorem is due to Mallet-Paret (see [165], where he used it for estimating
the Hausdorff dimension under the extra assumption that S is a C 1 -map) and
the above theorem was proved by Ladyzhenskaya (see [148]). We present a more
or less complete proof here since it is simple and elegant on the one hand and, on
the other hand, all the known estimates of the fractal dimension of attractors are
based on similar iterative schemes.
Example 6.6. We return to Example 2.10 of the 1D semilinear parabolic
equation (2.11). The existence of solution semigroup S(t) associated with this
equation, as well as the existence of a global attractor A ⊂ H01 (0, π) ⊂ C[0, π]
for this semigroup were already verified in Example 2.10. Moreover, due to
the maximum/comparison principle we also know that any complete bounded
solution u(t), t ∈ R, of this equation satisfies

− a1/2 6 u(t, x) 6 a1/2 , t ∈ R, x ∈ [0, π]. (6.7)

Now let u1 (t) and u2 (t) be two trajectories belonging to the attractor A, and let
v(t) := u1 (t) − u2 (t). Then, multiplying (2.14) by t ∂x2 v, integrating by parts, and
using (6.7), we obtain

d
(t∥∂x v(t)∥2L2 ) − 2a(t∥∂x v(t)∥2L2 ) 6 Ca2 t∥v(t)∥2L2 + ∥∂x v(t)∥2L2 ,
dt
where C is independent of a. Integrating this inequality with respect to time and
using (2.15), we end up with
Z t
t∥∂x v(t)∥2L2 6 e2a(t−s) Ca2 s∥v(s)∥2L2 + ∥∂x v(s)∥2L2 ds 6 C1 (at + 1)eat ∥v(0)∥2L2


0

for some C1 > 0 which is independent of a. Fixing now t0 = a−1 and S := S(t0 ),
we prove that for any two points u1 (0), u2 (0) ∈ A, estimate (6.5) is satisfied with
Φ = L2 (0, π), Φ1 = H01 (0, π), and L = Ca1/2 for some C which is independent of a.
Thus,
dimf (A, H01 ) 6 H1/(4L) (H01 ,→ L2 ) 6 Ca1/2 , (6.8)
 ATTRACTORS. THEN AND NOW 123

where we have used the well-known result about the entropy of embeddings
of Sobolev spaces in bounded domains Ω ⊂ Rd , namely
 d/(s1 −s2 )
1
Hν (W s1 ,p1 ,→ W s2 ,p2 ) 6 C (6.9)
ν

(see [217]).
Remark 6.7. As we can see from Example 6.6, the squeezing property (6.5) is
a straightforward corollary of the parabolic smoothing property for the linear PDE
(equation of variations) for the difference of two solutions associated with the original
nonlinear PDE. Since such a smoothing property is typical for dissipative PDEs
(for non-parabolic equations it should be replaced by an appropriate asymptotic
smoothing property discussed below), this explains why, in many cases, we have
the finiteness of the fractal dimension for attractors of dissipative PDEs.
Note also that the above upper bound (6.8) is sharp with respect to a → ∞
(lower bounds of the same order in a are available). It is remarkable that, in order
to get reasonably sharp estimates, one should consider the squeezing property (6.5)
on a small time interval t0 → 0 as a → ∞. In contrast to this, estimates based on
volume contraction method usually work better for large t0 → ∞ (see [12], [215]).
We mention also that the key advantage of the squeezing property (6.5) is that,
in comparison with volume contraction, it does not require the semigroup S(t)
to be differentiable with respect to the initial data. This is crucial for example
for applications to singular or/and degenerate PDEs, where such differentiability
usually does not take place or is very difficult/impossible to verify. As a price to pay,
this method may give essentially worse estimates in comparison with the volume
contraction. For instance, for the 2D Navier–Stokes equation in a bounded domain,
the best known upper bound for the fractal dimension of the attractor reads

dimf (A, H) 6 Cν −2 ,

where ν is the kinematic viscosity (see [12], [215], and the references therein) but
not even bounds polynomial in ν −1 have been obtained via the squeezing property
so far.
We now state a natural analogue of the squeezing property (6.5) which, in par-
ticular, is suitable for non-parabolic equations.
Theorem 6.8. Let Φ and Φ1 be two Banach spaces, and let the embedding Φ1 ⊂ Φ
be compact. Assume also that A is a bounded subset of Φ1 and the map S : A → A
satisfies (6.4) and the following squeezing/smoothing property:

∥S(u1 ) − S(u2 )∥Φ1 6 κ∥u1 − u2 ∥Φ1 + L∥u1 − u2 ∥Φ , u1 , u2 ∈ Φ. (6.10)

Then the fractal dimension of A in Φ1 is finite and satisfies the following estimate:

H(1−κ)/(4L) (Φ1 ,→ Φ)
dimf (A, Φ1 ) 6 , (6.11)
log2 (2/(1 + κ))

where κ, 0 6 κ < 1, and L > 0 are some constants.

124 S. V. ZELIK

The proof of this theorem repeats almost word by word the arguments given
in the proof of the previous theorem. The only difference is that, to be able to work
with κ > 1/2, we need to use coverings by sets of diameter less than ε in the iteration
scheme (see [44] for more details).
We also mention that nowadays there are a huge amount of various formulations
of the squeezing property (which give the finiteness of the fractal dimension of an
invariant set) adapted to the concrete classes of problems. The detailed exposition
of them is out of scope of this survey, so we refer the interested reader to [68], [179],
and the references therein.
Example 6.9. Consider the degenerate version of the real Ginzburg–Landau
equation
∂t u = ∆x (u3 ) + u − u3 , u t=0 = u0 , u ∂Ω = 0 (6.12)
in a bounded smooth domain Ω of Rd . This equation generates a dissipative
semigroup S(t), say, in the space Φ = L1 (Ω) (the proof of this fact is based
on Kato’s inequality and the related multiplication of the equation by sgn(u),
see [45]). Moreover, this semigroup is globally Lipschitz in L1 (Ω) and, due to
Hölder continuity results for solutions of degenerate parabolic problems, it possesses
a compact absorbing set (with respect to the bornology of bounded sets in Φ) which
is bounded in the space C α (Ω) for some α > 0. Therefore, the attractor A exists
and is generated by all complete bounded trajectories which are Hölder continuous
in space and time; see [59], [69], [112].
However, as shown in [69], this attractor has infinite Hausdorff and fractal
dimensions:
dimH (A, Φ) = dimf (Aγ , Φ) = ∞.
Moreover, the infinite-dimensional family of complete bounded solutions can be
constructed almost explicitly on the basis of the finite propagation speed of compactly
supported solutions.
This infinite-dimensionality can be somehow explained by the fact that the formal
“linearization” of (6.12) on a degenerate solution u = 0 is unstable, so the energy
income in the system is possible in the “area” where the equation is degenerate.
If the areas where the equation is degenerate or singular and the ones where the energy
income may occur are separated, then, typically, the corresponding attractor has
a finite fractal dimension (see [69], [70], [177]–[180], [207] for the justification
of this heuristic principle for concrete classes of singular/degenerate equations). In
particular, if we consider a slightly modified version of equation (6.12),

∂t u = ∆x (u3 ) + 3u2 − 2u − u3 , u t=0

= u0 , u ∂Ω
= 0, (6.13)

then the formal linearization ∂t v = −2v on u = 0 is exponentially stable and an

appropriate version of the squeezing property gives us the finiteness of the fractal
dimension of the attractor A (see [69]).
6.3. Upper bounds via the squeezing property: the non-autonomous
case. The standard situation in this case is when we have a cocycle Sξ (t) : Φ → Φ,
ξ ∈ Ψ, over a dynamical system T (h) : Ψ → Ψ and an invariant set Aξ such that

Sξ (1)Aξ = AT (1)ξ , ξ ∈ Ψ.
 ATTRACTORS. THEN AND NOW 125

Instead of the time step t = 1, one may take an arbitrary t0 > 0, but for simplicity
we assume that t0 = 1 here. The non-autonomous analogue of the squeezing
property (6.10) reads
∥Sξ (1)u1 − Sξ (1)u2 ∥Φ1 6 κ∥u1 − u2 ∥Φ1 + L(ξ)∥u1 − u2 ∥Φ , (6.14)
which holds for all ξ ∈ Ψ and u1 , u2 ∈ Aξ . For simplicity we take a “deterministic”
value of κ ∈ [0, 1) (it is independent of ξ), but its generalization to the “random”
case κ = κ(ξ) is straightforward.
The estimate for the fractal dimension of Aξ can be obtained in the same way as
in Theorems 6.4 and 6.8. The only difference is that, instead of iterating a single
map S, we now need to iterate different maps from the cocycle and use the equality
ST (−1)ξ (1) ◦ · · · ◦ ST (n−1)ξ (1) ◦ ST (−n)ξ (1)AT (−n)ξ = Aξ
for all ξ ∈ Ψ and n ∈ N. This gives the following result (see [71], [211] for more
details).
Theorem 6.10. Let Φ1 ⊂ Φ be two embedded Banach spaces such that the em-
bedding is compact, and let Sξ (t) : Φ → Φ, ξ ∈ Ψ, be a cocycle over T (h) : Ψ → Ψ.
Assume that ξ → Aξ ⊂ Φ1 is a family of bounded strictly invariant sets which satisfy
the squeezing property (6.14) for some κ ∈ [0, 1). Then the fractal dimensions of Aξ
satisfies the following estimate:
1
dimf (Aξ , Φ1 ) 6
log2 (2/(1 + κ)) − lim supn→∞ n−1 log2 R(T (−n)ξ)
Pn
H(1−κ)/(4L(T (−k)ξ)) (Φ1 ,→ Φ)
× lim sup k=1 , (6.15)
n→∞ n
where R(ξ) := ∥Aξ ∥Φ1 and we assume that the right-hand side is infinite if
the denominator is negative.
The most straightforward application of this general theorem is related to
the “deterministic” (uniform) case, where the size R(ξ) of the attractors Aξ is
uniformly bounded: R(ξ) 6 R0 and the expanding factor L(ξ) 6 L is also uniformly
bounded. This is typical for non-autonomous equations with uniformly in time
bounded external forces considered in [71] (see also the references therein). Then
we have exactly the same estimate for the dimensions of Aξ as in the autonomous
case.
Corollary 6.11. Let the assumptions of Theorem 6.10 hold, and let, in addition,
R(ξ) 6 R0 and L(ξ) 6 L for all ξ ∈ Ψ. Then the fractal dimensions of the Aξ are
finite and satisfy estimate (6.10) uniformly with respect to ξ ∈ Ψ.
The applications of this general theorem to the random case are more delicate and
interesting. In this case we have an invariant ergodic measure µ for the dynamical
system T (h) : Ψ → Ψ and assume that the function ξ → Aξ is µ-measurable.
In addition, we assume that the expanding factor L(ξ) is also measurable and
the entropy of the embedding Φ1 ⊂ Φ has the estimate
 θ
1
Hν (Φ1 ,→ Φ) 6 C (6.16)
ν
126 S. V. ZELIK

for some positive C and θ. Thus assumption is not restrictive since in applications Φ
and Φ1 are usually Sobolev spaces, for which such an estimate holds. Then we have
the following result.
Corollary 6.12. Let the assumptions of Theorem 6.10 hold, and let, in addition,
the random structure mentioned above be introduced and (6.16) hold. Also assume
that the family of attractors Aξ is tempered in Φ1 and the expanding factor L(ξ) has
a finite θth moment: Z
E(Lθ ) := L(ξ)θ µ(dξ) < ∞. (6.17)
Ψ
Then, for almost all values ξ ∈ Ψ, the fractal dimension of Aξ is finite and satisfies
the estimate
−1
4θ

θ 2
dimf (Aξ , Φ1 ) 6 C E(L ) log 2 . (6.18)
(1 − κ)θ 1+κ

log2 R(T (−n)ξ)

Indeed, since Aξ is tempered, lim = 0. The second multiplier
n→∞ n
on the right-hand side of (6.15) is estimated using (6.16), (6.17), and Birkhoff’s
ergodic theorem.
Remark 6.13. It is typical for random dynamical systems that, in order to
have the finite-dimensionality of the attractor, we need to verify that some random
variable (which is responsible for the expansion rate of the distance between
two solutions or Lyapunov exponents) has a finite mean (see [54], [58], [211],
and the references therein). This condition is not trivial and is often the most
difficult to verify. On the other hand, as we will see in the next toy example
(suggested in [30]), the dissipation mechanism may be not strong enough to provide
the finite-dimensionality of a random attractor if this condition is violated.
Example 6.14. Let H = l2 (the space of square summable sequences), and
consider the RDS in H generated by the following equations:
d d
u1 + γ(t)u1 (t) = 1, uk + k 4 uk = u1 (t)uk − u3k , k = 2, 3, . . . , (6.19)
dt dt
where u = (u1 , u2 , . . . ) ∈ H and γ ∈ Ψ is exactly the Bernoulli process used
in Example 5.17. The first equation of this system models the energy evolution
of (5.13) and the other equations give some coupling of the first equation with
a parabolic PDE.
The system of ODEs (6.19) can be solved explicitly. In particular, if aq−(1−q)b >
0, then Z  Z t  t
u1 (t) = u1,γ (t) = exp − γ(l) dl ds (6.20)
−∞ s

is a unique tempered complete solution of the first equation (for almost all γ ∈ Ψ).
Moreover, it is not difficult to show that
Z
u1,γ (0) µ(dγ) = ∞ (6.21)
γ∈Ψ
 ATTRACTORS. THEN AND NOW 127

if
ln(qe−a + (1 − q)eb ) > 0 > −aq + (1 − q)b. (6.22)
In this case the following result holds.
Proposition 6.15. Let the exponents a, b > 0 and q ∈ (0, 1) satisfy (6.22). Then
the random attractor Aγ for system (6.19) in H has infinite Hausdorff and fractal
dimensions:
dimH (Aγ , H) = dimf (Aγ , H) = ∞ (6.23)
for almost all γ ∈ Ψ.
Sketch of the proof. The detailed proof of this result was given in [30].
Here we just discuss briefly the main ideas behind it. The existence of a random
tempered absorbing set for the first component u1 of system (6.19) follows from
the explicit formula for the solution. Now we assume that u1 (τ ) is already in this
absorbing set and get estimates for uk . To this end, multiplying the kth equation
by sgn(uk ) and taking the sum, after standard estimates, we get
∞  X∞ ∞
X 
d X 4 1
|uk | + k |uk | 6 C |u1 |3/2 6 C|u1 (t)|3/2 .
dt k2
k=2 k=2 k=2

Integrating this inequality and using that u1 (t) is tempered, we get a tempered
absorbing ball for u in l1 ⊂ H. To obtain a compact absorbing set it is enough
to use the parabolic smoothing property in a standard way. Thus, the existence
of a random attractor Aγ is verified.
Recall also that a random attractor consists of all complete tempered trajectories
of the system under consideration, so it is enough to find all such trajectories. The
first equation is linear and independent of the other equations, so such a trajectory
is unique and is given by (6.20). Thus, to find Kγ , we need to fix u1 (t) = u1,γ (t)
in the other equations of (6.20) and find the tempered attractor Akγ , k = 2, 3, . . . , for
every component of (6.20) separately. Then the desired attractor Aγ for the whole
system is presented as a product:
∞
O
Aγ = {u1,γ (0)} × Akγ . (6.24)
k=2

Moreover, since the attractor is always connected, every Akγ is a closed interval, so
to prove the desired infinite dimensionality, it is enough to prove that Akγ ̸= {0} for
all k. In other words, we need to find a non-zero tempered trajectory uk = uk (t)
for the equation
d
uk + k 4 uk = u1,γ (t)uk − u3k , (6.25)
dt
using that u1,γ (t) is tempered and has an infinite mean.
To this end, we note first that every solution of equation (6.25) is either tempered
as t → −∞ or blows up backward in time (this can easily be shown by comparison
using the fact that u1,γ (t) is tempered), so any complete solution uk (t), t ∈ R, is
automatically tempered.
128 S. V. ZELIK

We construct this solution by solving equation (6.25) explicitly. Namely,

 Z t  Z t  −1/2
4
uk,γ (t) := 2 exp 2 (k − u1,γ (l)) dl ds .
−∞ s

Indeed, the finiteness of the integral is guaranteed by (6.21) (see [30] for more
details). Thus, we have proved that Akγ ̸= {0} for all k and almost all γ and
the proposition is proved.
Remark 6.16. Actually, we have found explicitly the random attractor in the pre-
vious example:
O∞
Aγ = {u1,γ (0)} × [−uk,γ (0), uk,γ (0)].
k=2

This example shows that, in contradiction with the commonly accepted paradigm,
adding random terms may not only simplify the dynamics, but may also make it
essentially more complicated and even infinite-dimensional. We expect that this
phenomenon has a general nature and can be observed in more realistic equations.
We conclude this subsection by considering the applications of the dimension
estimate to uniform attractors. In this case we typically have a cocycle Sξ (t) :
Φ → Φ, ξ ∈ Ψ, over the dynamical system T (h) : Ψ → Ψ, a family of invariant sets
(pullback attractors) Aξ , ξ ∈ Ψ, and a union
[
Aun = Aξ , (6.26)
ξ∈Ψ

which is exactly the uniform attractor whose dimension we want to estimate. For
simplicity we assume that Ψ ⊂ Lpb (R, H) for some Banach space H and 1 6 p < ∞
is a hull of some translation-compact external force ξ0 ∈ Lpb (R, H), so we assume
that
Ψ := H(ξ0 ) b Lploc (R, H)
(see [38] for more details and more general exposition). It is also well known
that, despite the finiteness of the fractal dimension of every Aξ , the dimen-
sion of the union (6.26) is usually infinite, because of the infinite-dimensionality
of the hull H(ξ0 ). For this reason it looks natural (following Vishik and Chep-
yzhov) to study the Kolmogorov ε-entropy of Aun . To this end we need to add to
the squeezing property (6.10) also some kind of Lipschitz/Hölder continuity with
respect to ξ, namely, to assume that

∥Sξ1 (1)u1 −Sξ2 (1)u2 ∥Φ1 6 κ∥u1 −u2 ∥Φ1 +L∥u1 −u2 ∥Φ +K∥ξ1 −ξ2 ∥Lp (0,1;H) , (6.27)

where 0 6 κ < 1, L and K are independent of ξ1 , ξ2 ∈ Ψ, and u1 , u2 ∈ Aξ . Then

we get the following analogue of Theorem 6.8.
Theorem 6.17. Let Sξ (t) : Φ → Φ be a cocycle over a dynamical system
T (h) : Ψ → Ψ and Ψ = H(ξ0 ) be a hull of some function ξ0 ∈ Lpb (R, H) which
is translation-compact in this space. Assume also that there is a family of invariant
sets Aξ which satisfies the squeezing property (6.27), where Φ1 is some other Banach
 ATTRACTORS. THEN AND NOW 129

space compactly embedded in Φ, and also is uniformly bounded in the Banach space Φ1 .
Then the Kolmogorov entropy of the union (6.26) has the following estimate:

H(1−κ)/(4L) (Φ1 ,→ Φ) ε0
Hε (Aun , Φ1 ) 6 C0 + log2
log2 (2/(1 + κ)) ε
   
p ε0
+ Hε/K0 H(ξ0 ) t∈[0,L0 log (ε0 /ε)] , Lb 0, L0 log2 ,H ,
2 ε
(6.28)

where ε 6 ε0 and ε0 , C0 , K0 , L0 are some positive numbers which are independent

of ε.
The derivation of this estimate is similar to the proof of Theorem 6.4, so we omit
it here and refer the interested reader to [67] or [234] (see also [38] for the analogous
results obtained via the volume contraction method).
As we already mentioned, the dimension of Aun may be infinite due to the second
term on the right-hand side of (6.28) (see [38] and the references therein). However,
there are interesting particular cases where the dimension of a uniform attractor
remains finite; for example, it is so for periodic or quasi-periodic dynamical processes
as well as for dynamical processes stabilizing to autonomous ones as time tends
to ±∞ (see [38], [67], [234] for more examples). To extract this finite dimensionality
from the key estimate (6.28), we introduce the fractal dimension of the hull H(ξ0 )
as follows:
Hε (H(ξ0 )|t∈[0,log2 (1/ε)] , Lpb ([0, log2 (1/ε)], H))
dimf (H(ξ0 )) := lim sup . (6.29)
ε→0 log2 (1/ε)

Then estimate (6.28) reads

H(1−κ)/(4L) (Φ1 ,→ Φ)
dimf (Aun , Φ1 ) 6 + dimf (H(ξ0 )) (6.30)
log2 (2/(1 + κ))

(see [38] for details). The first term on the right-hand side can be interpreted as
(an upper bound for) the dimension of every Aξ and the second term is added due
to the union.
Remark 6.18. It is not difficult to see that dimf (H(ξ0 )) is indeed the standard
fractal dimension of the set H(ξ0 ) in the weighted norm of Lpe−|x| (R, H). Note that,
for a translation-compact ξ0 ∈ Lpb (R, H), the hull H(ξ0 ) is compact in the local
topology of Lploc (R, H) and, in order to speak about the dimension, we need to fix
a metric on H(ξ0 ). Estimate (6.28) hints that the weighted metric of Lpe−|x| (R, H)
is the most appropriate one for estimating the entropy of uniform attractors.
Note also that estimate (6.28) makes no sense if ξ0 is not translation-compact
(since the second term on the right-hand side will be infinite). However, there is
a natural generalization of this formula, which was proposed in [234] and allows
us to work with the classes of non-translation-compact external forces considered
in subsection 5.3.
It is worth mentioning that the result of Theorem 6.17 remains true with minor
changes if we replace Lipschitz continuity with respect to ξ in (6.27) by Hölder
130 S. V. ZELIK

continuity with an arbitrary exponent α ∈ (0, 1]. In contrast to this, one may
lose finite dimensionality (even in the autonomous case) if one replaces Lipschitz
continuity with respect to the initial data ui by Hölder or log-Lipschitz continuity
(see [173] for the counterexample related to attractors of elliptic PDEs).
6.4. Volume contraction and Lyapunov dimension. We now turn to
the scheme for estimating the dimension of the attractor, which is based on
volume contraction arguments. This scheme has become extremely popular due
to applications to the Navier–Stokes equations, where it gives the best estimates
available so far, which cannot be obtained using other methods (although this
scheme has significant drawbacks, for example, it requires the phase space to be
Hilbert and the corresponding dynamical system to be differentiable with respect
to the initial data, so it cannot replace other methods fully). Since there are
nice expositions of this material in the literature (see, for example, [12], [215], and
the references therein), we restrict ourselves to just a brief discussion.
We start with introducing the volume contraction factor. Let Φ be a separable
Hilbert space, and let ϕ1 , . . . , ϕd ∈ Φ. By definition, the wedge product ϕ1 ∧ · · · ∧ ϕd
is the d-linear antisymmetric form on Φ defined by

ϕ1 ∧ · · · ∧ ϕd (ψ1 , . . . , ψd ) := det (ϕi , ψj )di,j=1 ,

where ( · , · ) is the inner product in Φ. A d-linear form on Φ that is a wedge

product of d vectors of Φ is called decomposable. Let us denote by Λ̃d Φ the space
of d-linear antisymmetric forms which can be presented as a finite linear combination
of decomposable functionals. For two decomposable forms ϕ1 ∧ · · · ∧ ϕd and
ψ1 ∧ · · · ∧ ψd , their inner product is defined by

(ϕ1 ∧ · · · ∧ ϕd , ψ1 ∧ · · · ∧ ψd ) := det (ϕi , ψj )di,j=1 ,

and being extended to Λ̃d Φ by linearity, it defines an inner product on the space Λ̃d Φ
(see [215] for the details). Finally, the completion of Λ̃d Φ with respect to this
norm is called the dth exterior power of the space Φ and is denoted by Λd Φ. Note
that Λd Φ is the subspace of all continuous antisymmetric d-linear forms on Φ, and
this subspace is proper if d > 1 and dim Φ = ∞. Now let L ∈ L(Φ, Φ) be a linear
continuous operator on Φ. Then its dth exterior power Λd L is defined by

(Λd Lξ)(ψ1 , . . . , ψd ) := ξ(L∗ ψ1 , . . . , L∗ ψd ), ξ ∈ Λd Φ,

where L∗ is the operator adjoint to L. It is not difficult to check that Λd L : Λd Φ →

Λd Φ is a linear continuous operator and its norm does not exceed ∥L∥d . The
following lemma gives a geometric interpretation for the norm of this operator.
Lemma 6.19. The norm of the dth exterior power of L satisfies

∥(Lϕ1 ) ∧ · · · ∧ (Lϕd )∥Λd Φ vold (LΠ)

∥Λd L∥L(Λd Φ,Λd Φ) = sup = sup ,
ϕ1 ∧···∧ϕd ̸=0 ∥ϕ1 ∧ · · · ∧ ϕd ∥Λd Φ Πd ⊂Φ vold (Π)

where the last supremum is taken over all non-degenerate d-dimensional paral-
lelepipeds in Φ and vold stands for the d-dimensional Lebesgue measure.
 ATTRACTORS. THEN AND NOW 131

The proof of this lemma can be found, for example, in [215]. Thus, geomet-
rically, ∥Λd L∥ is the maximal expanding factor for d-dimensional volumes under
the action of the operator L.
Now assume that we are given a compact set A ⊂ Φ and a map S : A → A.
Assume that this map is uniformly Fréchet (quasi)differentiable on A, that is, there
exist linear continuous maps S ′ (u0 ) : Φ → Φ, u0 ∈ A, such that

∥S(u1 ) − S(u2 ) − S ′ (u1 )(u1 − u2 )∥Φ = o(∥u1 − u2 ∥Φ ) (6.31)

uniformly with respect to u1 , u2 ∈ A. Assume also that the map u0 → S ′ (u0 ) is con-
tinuous on A. Then we define an infinitesimal d-volume contraction factor ωd (S, A)
by
ωd (S, A) := sup ∥Λd S ′ (u0 )∥L(Λd Φ,Λd Φ) .
u0 ∈A

This definition can be extended also to non-integer values of s = s0 + α, α ∈ (0, 1),

via

ωd (S, A) := sup ∥Λd S ′ (u0 )∥1−α ∥Λd+1 S ′ (u0 )∥α


L(Λd Φ,Λd Φ) L(Λd+1 Φ,Λd+1 Φ) .
u0 ∈A

The main result of the theory is the following theorem.

Theorem 6.20. Let A be a compact set in a Hilbert space Φ which is strictly
invariant with respect to a map S : A → A. Assume also that the map S is uniformly
quasi-differentiable on A and the map u0 → S ′ (u0 ) is continuous. Finally, let
ωd (S, A) < 0 for some d ∈ R+ . Then

dimf (A, Φ) < d.

To the best of our knowledge, this theorem has been proved by Douady and
Oesterlé [60] in the finite-dimensional case for the Hausdorff dimension and has
been extended to the infinite-dimensional case in [107] (see also [49]). The case
of fractal dimension is more delicate and had initially been treated with some
rather annoying extra conditions (see [49]). These conditions were removed by
Hunt [105] in the finite-dimensional case (see also [19] for C 2 -diffeomorphisms in
the infinite-dimensional case). The result of the theorem in the form stated above
appeared in [32].
Remark 6.21. The proof of Theorem 6.20 follows the general scheme presented
in Theorem 6.4. Namely, we start with the covering of A by very small ε-balls and
obtain a better covering of the set S(A) using our assumptions on S. Then we
iterate this process in order to get an estimate for the dimension. Indeed, since ε
is small and S is smooth enough, the image S(Bε (u0 , Φ)) is close to the ellipsoid
S ′ (u0 )Bε (u0 , Φ), and by the assumptions of the theorem the d-dimensional volume
of this ellipsoid is strictly less than the d-dimensional volume of the initial ball. In
the case when the Hausdorff dimension is considered, it is more or less straightforward
to cover every such ellipsoid by smaller balls of different radii in such a way that
the corresponding d-dimensional Hausdorff measure is contracted at every step
of iterations, which gives the result (see [215] for the details). The difficulties
132 S. V. ZELIK

in the case of fractal dimension are related to the fact that, in contrast to the case
of conditions (6.5), different ellipsoids obtained as the images of ε-balls from
the initial covering may have considerably different “shapes” (only their volumes
are contracted), but we still need to cover them by balls of the same radius, say
ε/K. For this reason, much more delicate estimates are required.
Note that the quasi-differentiability condition is automatically satisfied if the map
S : Φ → Φ is C 1 -smooth. However, in applications it is often easier to verify
estimate (6.31) on the attractor only, since the attractor is usually more smooth.
This is the reason why quasi-differentiability is introduced. Note also that,
in contrast to the Fréchet derivative, the linear operator S ′ (u0 ) satisfying (6.31)
may be not unique.
It is worth mentioning that, in applications to PDEs, we normally use the theorem
for integer values of d only. However, the possibility to use non-integer ds is crucial
in many applications to ODEs, for example, to the Lorenz system.
Theorem 6.20 can be reformulated in a more elegant way using the fact that
the map S can be replaced by any iteration S n of it. Indeed, it is not difficult to
show using the chain rule that the sequence ln ωd (S n , A) is subadditive, so the limit

1/n
1/n
ω d (S, A) = lim ωd (S n , A) = inf ωd (S n , A)
n→∞ n∈N

exists for every d. In the case of continuous time and a semigroup S(t), we just
replace n ∈ N by t ∈ R+ and S n by S(t) throughout. Also recall that the uniform
Lyapunov dimension of the map S on A is defined by

dimL (S, A) := sup {ω d (S, A) > 0} (6.32)

d∈R+

(see [215], [242], and the references therein). Then Theorem 6.20 reads as follows.
Corollary 6.22. Let the assumptions of Theorem 6.20 hold. Then

dimf (A, Φ) 6 dimL (S, Φ). (6.33)

At the next step we discuss how to evaluate or estimate the Lyapunov dimension.
This can be done for dynamical systems with continuous time using analogues
of Liouville’s formula for the evolution of volumes. Namely, assume that, for
any v0 ∈ Φ and any u0 ∈ A the (quasi-)derivative v(t) := S ′ (t)(u0 )v0 of the map
S(t) : A → A solves the corresponding equation of variations:

∂t v(t) = L(u(t))v(t), v t=0

= v0 , u(t) := S(t)u0 , (6.34)

where L(u) : D → Φ are linear unbounded operators such that D is dense in Φ and
the operators L(u) are bounded above, that is,

(L(u)η, η) 6 C∥η∥2Φ , η ∈ D, u ∈ A,

where C is independent of η and C. The following lemma is a key technical tool

in estimating the Lyapunov dimension.
 ATTRACTORS. THEN AND NOW 133

Lemma 6.23. Under the above assumptions, the following identity holds:
1 d
∥v1 (t)∧· · ·∧vd (t)∥2Λd Φ = Tr(Q(t)◦L(u(t))◦Q(t))∥v1 (t)∧· · ·∧vd (t)∥2Λd Φ , (6.35)
2 dt
where the vi (t), i = 1, . . . , d, are solutions of the equation of variations (6.34), Q(t)
is the orthoprojector to d-dimensional subspace spanned by the vectors {vi (t)}di=1 ,
and Tr is the trace of a matrix in Rd .
This statement is the extension of the classical Liouville theorem to an
infinite-dimensional case (see, for example, [215] for the details). To proceed
further, we define the d-dimensional trace of the operator L(u(t)), namely,
d
X 


Trd L(u(t)) := sup (L(u(t))ψi , ψi ) : ψi ∈ D, (ψi , ψj ) = δij . (6.36)
i=1

Then identity (6.35), together with the obvious estimate

Tr(Q(t) ◦ L(u(t)) ◦ Q(t)) 6 Trd L(u(t)) ,

gives us the estimate

  Z T 

1/T 1
ωd (S(T ), A) 6 exp 2 sup Trd (L(u(t))) dt .
u0 ∈A T 0

This, in turn, gives a corollary of Theorem 6.20 which is suitable for applications
(see [38], [215] for more details).
Corollary 6.24. Let the above assumptions hold. Also let
 Z T 
1
q d := inf sup Trd (L(S(t)u0 )) dt . (6.37)
T >0 T u0 ∈A 0

Assume that q d < 0 for some d ∈ N. Then

dimf (A, Φ) 6 dimL (S(t), A) < d. (6.38)

Remark 6.25. Corollary 6.24 reduces the problem of estimating the fractal
dimension of the attractor to estimating the traces of linear operators related to
the corresponding equation of variations. This links the theory of attractors with
a classical topic of operator theory, namely, estimating the traces of various differ-
ential operators (see [80] and the references therein). In particular, in relatively
simple cases, such estimates can be obtained using the min-max principle, while
in more complicated cases (like the Navier–Stokes equations or other hydrodynam-
ical equations) an essential progress has been achieved using the Lieb–Thirring
inequalities, as well as other collective Sobolev inequalities (see [110], [109], [215],
and the references therein).
Note also that, although the definition of the Lyapunov dimension is independent
of the concrete choice of the equivalent inner product in the space Φ, a clever choice
of this inner product (which may depend on the point of the phase space), may
134 S. V. ZELIK

essentially improve the estimates; see, for example, [85] and [204] for estimates related
to the damped Schrödinger and hyperbolic Cahn–Hilliard equations, respectively,
as well as [151] for the calculation of the exact value of the Lyapunov dimension
of the Lorenz attractor. We will not go into further details here and restrict
ourselves to an example related to damped wave equations, which is important for
the discussion in the next subsection.
Example 6.26. Consider an abstract damped wave equation in a Hilbert space Φ:

∂t2 u + γ∂t u + Au = F (u, ∂t u), ξu t=0

= ξ0 , ξu (t) := {u(t), ∂t u(t)}, (6.39)

where A : D(A) → Φ is a positive self-adjoint linear operator with compact inverse,

γ > 0 is a given dissipation coefficient, and F is a given nonlinearity. We define
a scale Φs := D(As/2 ), s ∈ R, of Hilbert spaces associated with the operator A and
the associated scale of energy spaces E s := Φs+1 × Φs . For simplicity we assume
that F is smoothing and bounded, that is,

F ∈ Cb∞ (E −m , Φm ) ∀ m ∈ N. (6.40)

Then it is straightforward to prove that problem (6.39) is globally well posed

in the energy phase space E := E 0 , and the corresponding solution semigroup
S(t) : E → E is dissipative with respect to the standard bornology of bounded sets
in E and possesses an attractor A in E (see, for example, [220] for more details).
Our main task here is to estimate the fractal dimension of this attractor and
its dependence on the small parameter γ. Since we have made the extremely
strong assumptions (6.40) on the nonlinearity F , it is also immediate to verify that
the solution semigroup S(t) is C ∞ -smooth in E with respect to the initial data and
the Fréchet derivative ξv (t) := S ′ (ξ0 )(t)ξv0 is defined as a solution of the following
equation of variations:

∂t2 v + γ∂t v + Av = Fu′ (ξu (t))v + F∂′ t u (ξu (t))∂t v, ξv = ξ v0 ,

ξu (t) := S(t)ξu0 ,
t=0
(6.41)
so we only need to estimate the d-dimensional traces of the operator
 
0 1
L(ξu (t)) := (6.42)
−A + Fu′ −γ + F∂′ t u

in the space E. However, the volume contraction scheme does not work properly
for the original inner product in E induced by the Cartesian product and we need
to modify it, namely, we set
γ
({v, v ′ }, {w, w′ })γ := (Av, w) + (v ′ , w′ ) + ((v, w′ ) + (v ′ , w)), {v, v ′ }, {w, w′ } ∈ E.
2
Then this inner product is equivalent to the standard one if γ > 0 is small enough
and
γ
(Lξv , ξv )γ = − ∥v∥2Φ1 + ∥v ′ ∥2Φ + (F∂′ t u v ′ , v ′ ) + (Fu′ v, v ′ )


2
γ
+ (Fu′ v, v) + (F∂′ t u v ′ , v) − γ(v, v ′ ) .


2
 ATTRACTORS. THEN AND NOW 135

To estimate the terms containing F on the right-hand side, we use assumption (6.40).
For instance, for the first term we have

|(F∂′ t u v ′ , v ′ )| 6 ∥F∂′ t u v ′ ∥Φm ∥v ′ ∥Φ−m 6 C∥v ′ ∥2Φ−m = C(A−m v ′ , v ′ )

and estimating the remaining terms analogously, for sufficiently small γ > 0 we
arrive at the inequality
−γ/4 + Cm A−m
 
0
(L(ξu (t))ξv , ξv )γ 6 (Lm ξv , ξv )γ , Lm := ,
0 −γ/4 + Cm A−m

where m ∈ N is arbitrary and the constant Cm is independent of ξv = {v, v ′ } and ξu

(see [220]). Then, using the min-max principle, we finally arrive at
 d 
γ X
−m
Trd (L(ξu (t))) 6 Trd (Lm ) = 2 − d + Cm λn ,
4 n=1

where {λn }∞n=1 are the eigenvalues of the operator A enumerated in the non-decrea-
sing order. If we now assume that
∞
X
λ−m
n <∞
n=1

for some large m (which is always true if A is a uniformly elliptic differential

operator in a bounded domain), estimate (6.37) will show us that
γ
qd 6 − d + C
4
and then Corollary 6.24 gives us the final result that the fractal dimension
of the attractor A associated with equation (6.39) has the following upper bound:
C
dimf (A, E) 6 dimL (S(t), A) 6 , (6.43)
γ
where γ is small enough and C is independent of γ (see [220] for details).
6.5. Lower bounds for the dimension. Most lower bounds for the dimen-
sions of attractors which are available in the literature are based on estimates
for the instability index of an appropriately constructed equilibrium. Assume that
we are given a dissipative semigroup S(t) : Φ → Φ acting on a Banach space Φ with
the standard bornology B which consists of all bounded sets of Φ. Assume also
that u0 is an equilibrium of this semigroup, and consider the unstable set of this
equilibrium:
n o
M+ (u0 ) := v0 ∈ Φ : there exists u ∈ K such that u(0) = v0 , lim u(t) = u0 ,
t→−∞
(6.44)
where K is the set of complete bounded trajectories of S(t). Note that, due to
dissipativity, it is enough to construct a negative semi-trajectory u(t), t 6 0, with
the above properties. Then it follows from the definition of an attractor that

M+ (u0 ) ⊂ A (6.45)
136 S. V. ZELIK

if the attractor A exists. Thus, the fractal dimension of the attractor A is estimated
in terms of the dimension of the unstable set M+ (u0 ), which is usually a submanifold
itself or contains a submanifold of sufficiently large dimension.
Indeed, assume that S(t) is Fréchet differentiable near the equilibrium point u0
and the Fréchet derivative S ′ (u0 ) is given by the equation of variations (6.34) and
is uniformly continuous in a small neighbourhood of u0 . Assume also that the part
of the spectrum σ(L(u0 )) belonging to the positive semi-plane Re λ > 0 consists
of finitely many eigenvalues of finite algebraic multiplicity. Denote by N+ (u0 ) ∈ N
the number of all such eigenvalues, taking their multiplicities into account. Then
the standard theorem about unstable manifolds (see, for example, [12]) claims
that the set M+ (u0 ) contains an N+ (u0 )-dimensional local submanifold which is
generated by all complete trajectories which approach u0 exponentially fast as
t → −∞. Hence
dimf (A, Φ) > dimf (M+ (u0 ), Φ) > N+ (u0 ). (6.46)
Thus, obtaining lower bounds for the dimension of an attractor is also reduced
to a classical problem in spectral theory, namely, finding/estimating the number
of unstable eigenvalues of a given differential operator. Of course, the scheme
described above can be used not only in the case where u0 is an equilibrium. It may
be something more complicated, for example, a periodic or a quasi-periodic or even
a chaotic orbit. The theory of unstable manifolds for such objects is well developed
nowadays (see, for example, [12], [116], and the references therein). However,
such objects are almost never used in the literature to estimate the dimension of
an attractor since it is extremely difficult to find them and estimate explicitly their
instability index. An exception from this, which is based on the theory of homoclinic
bifurcations and, as believed, has a general nature, will be considered in examples
below.
We also mention that lower bounds for the Lyapunov dimension of an attractor
can be obtained via equilibria and the obvious inequality
dimL (S(t), A) > dimL (S(t), u0 ).
It is important that the Lyapunov dimension dimL (S(t), u0 ) can also be estimated via
the spectrum of the infinitesimal generator L(u0 ) of the linear semigroup S ′ (t)(u0 ).
To this end, we need the following definition.
Definition 6.27. Let Φ be a Hilbert space and L(u0 ) : D → Φ be a linear closed
bounded above operator in Φ. Also let µ∞ := sup Re σess (L(u0 )). Then there
is only a finite or countable number of eigenvalues λj of L(u0 ) in the half-plane
Re λ > µ∞ , which can be enumerated in the non-increasing order. Then we define
the Lyapunov exponents µn , n ∈ N, as follows: µn = Re λn if there exists at least n
eigenvalues satisfying Re λ > µ∞ , and µn = µ∞ otherwise. Then the Lyapunov
dimension of L(u0 ) is defined by
dimL (L(u0 )) := sup {µd > 0}, (6.47)
d∈R+
s
X
where µs = µn for integer s and µs = µ[s] + (s − [s])µ[s]+1 (see, for example, [215]
n=1
for more details).
 ATTRACTORS. THEN AND NOW 137

The next proposition gives us the desired estimate.

Proposition 6.28. Let the assumptions of Theorem 6.20 hold, and let u0 ∈ A
be an equilibrium. Then the following estimates hold:

1) N+ (u0 ) 6 dimf (A, Φ) 6 dimL (S(t), A),

(6.48)
2) dimL (L(u0 )) 6 dimL (S(t), u0 ) 6 dimL (S(t), A).

Indeed, only the left-hand side of the second inequality has not been proved yet.
Its proof follows in a straightforward way from definitions and the spectral mapping
theorem (see [45] for more details). Note that, in contrast to the finite-dimensional
case, this inequality may be strict if Φ is infinite-dimensional, since we have only
one-sided spectral mapping theorem for the essential spectrum.
We now turn to examples.
Example 6.29. We return to Example 2.10, which is the simplest example. We
have already proved that the fractal dimension of the corresponding attractor A
is finite in Φ = L2 (0, π) and satisfies the upper bound (6.28). We now apply
the volume contraction method and Proposition 6.28 to get more explicit upper
and lower bounds. First we note that the corresponding solution semigroup
is C ∞ -smooth here (see, for example, [12]), so all smoothness assumptions are
automatically satisfied, and we only need to get good upper and lower bounds for
the Lyapunov dimension and the instability indices.
The operator L(u(t)) for the equation of variations now reads L(u) = ∂x2 −3u2 +a,
and the eigenvalues of the operator −∂x2 with the Dirichlet boundary conditions are
λn = n2 . Therefore, by the min-max principle, for integer values of d we have
d
X d(d + 1)(2d + 1) d3
Trd (L(u)) 6 Trd (∂x2 + a) = − λn + ad = − + ad > − − ad.
n=1
6 3

We recall that u0 = 0 is an equilibrium and the first inequality in the above formula
is attained at this equilibrium. This gives us

dimL (S(t), A) = dimL (S(t), 0) = dimL (L(0)) = dimL (∂x2 + a), (6.49)

where the second equality is due to the fact that the spectral mapping theorem
holds for self-adjoint operators. Note that the right-hand side of equality (6.49) is
easy to compute explicitly, so the Lyapunov dimension of the attractor A has not
only been estimated, but can also be computed explicitly. We will not write out
this formula, but only mention that, for large a we have
√
dimH (A, Φ) 6 dimf (A, Φ) 6 dimL (S(t), A) ∼ 3a . (6.50)

Let us now look at lower bounds for Hausdorff and fractal dimensions of the at-
√
tractor A. We see that the instability index N+ (0) is of order a . Moreover,
since the equation under consideration possesses a global Lyapunov function,
the corresponding attractor is the union of the unstable sets of all equilibria, that is,
√
dimH (S(t), A) ∼ max N+ (u0 ) = N+ (0) ∼ a .
u0 ∈R
138 S. V. ZELIK

So the Hausdorff dimension of the attractor can also be explicitly computed. In

contrast to this, for the fractal dimension we only have the two-sided inequality
√ √
a ∼ N+ (0) ∼ dimH (A, Φ) 6 dimf (A, Φ) 6 dimL (S(t), A) = dimL (L(0)) ∼ 3a .
(6.51)
Note that even in this simplest example the fractal dimension may a priori be
larger than the Hausdorff one due to the non-trivial intersections of stable and
unstable manifolds of equilibria, so we basically do not have anything more than
the two-sided bounds (6.51).
Remark 6.30. Note that the positivity of the Lyapunov dimension dimL (S(t), A)
does not imply here the presence of a chaotic dynamics of the system under
consideration (in our example the system is gradient and the dynamics is trivial).
This should not be misleading, since in contrast to the classical dynamics, where
the Lyapunov exponents for individual trajectories (which exist for almost all
trajectories with respect to the “physical measure”) are usually considered (see,
for example, [116]), we consider here the so-called uniform Lyapunov exponents,
which may be positive even for the trivial gradient dynamics. Note also that
the coincidence of the Lyapunov dimension of the whole attractor with the dimension
of an equilibrium is not limited to gradient systems. For instance, exactly the same
happens (for some different reasons) in the case of the chaotic Lorenz attractor
considered in Example 2.8.
We notice also that, in the example considered above, the instability index N+ (u0 )
and the Lyapunov dimension L(u0 ) have√ the same asymptotic behaviour as a → ∞
(up to the non-essential multiplier 3 ), which allowed us to describe the sharp
asymptotic behaviour of dimf (A, Φ) as a → ∞. This is somehow typical for
parabolic equations but can be easily fail, for example, for hyperbolic equations.
Example 6.31. In particular, this is not true for the equation considered
in Example 6.26, to which we return now. Indeed, due to assumptions (6.40),
the eigenvalues νF of the linearized operator (6.42) at any equilibrium u0 are
asymptotically very close to the eigenvalues ν0 of the operator which corresponds
to F = 0, namely,
|νF,k − ν0,k | 6 CN k −N
for all N ∈ N (see [220]). The eigenvalues which correspond to F = 0 are easy to
compute: p
± −γ ± γ 2 − 4λk
ν0,k = .
2
Thus, N+ (u0 ) 6 CN γ 1/N (see [220] for any N ) and, therefore, we cannot get a lower
bound sharp in γ for (6.43) on the basis of the instability index of any equilibrium.
In contrast to this, if we look at the Lyapunov dimension of the equilibrium u0 ,
we see that the essential spectrum lies on the line Re λ = −γ/2 and, consequently,
µd > µ∞ > −γ/2 for all d. For this reason, if we find an equilibrium u0 in such
a way that it has an unstable eigenvalue with Re λ1 = O(1) > 0 as γ → 0, then we
get the estimate
C
dimL (L(u0 )) > .
γ
 ATTRACTORS. THEN AND NOW 139

Therefore, dimL (S(t), A) ∼ γ −1 , so we cannot obtain better upper bounds for

the fractal dimension:
C1 6 dimf (A, E) 6 C2 γ −1 . (6.52)
To overcome this difficulty and get sharp upper and lower bounds for the fractal
dimension of A, a new method for finding lower bounds was proposed in [220], which
is based on the homoclinic bifurcation result proved in [219]. Roughly speaking, this
result tells us that: if you are given a multi-dimensional system with an equilibrium
u0 = 0 and a homoclinic loop u(t) to it such that u(t) ∼ e∓λ as t → ±∞ for some
λ > 0 and you have sufficiently many “intermediate” eigenvalues of the linearization
matrix at u0 = 0:
− λ < Re λ1 6 · · · 6 Re λN < λ, (6.53)
then, making an arbitrarily smooth small perturbation of the system you may
produce an invariant torus T whose dimension is proportional to the Lyapunov
dimension of the equilibrium u0 = 0. In other words, the dimension of an invariant
manifold which can be born near a homoclinic loop is restricted by the Lyapunov
dimension of the origin of this loop only (of course, under some natural extra
assumptions on the loop).
Finally, if we embed such a construction in the attractor A of (6.39), then we
end up with the desired sharp lower bound:

dimf (A, Φ) > Cγ −1 .

Exactly this approach was realized in [220]. Namely, consider the following model
decoupled system of ODEs:

u′′0 = u0 − u30 , u′′n + γu′n + λn un = 0, n = 1, 2, . . . , (6.54)

where γ > 0 is small enough, and γ 2 − 4λn > 0. Then this system has a zero
equilibrium and a homoclinic loop {u0 (t), 0} to it. Moreover, u0 (t) ∼ e−|t| as
t → ±∞, and the remaining eigenvalues (which correspond to equations for un ,
n = 1, 2, . . . ) are p
± −γ ± γ 2 − 4λn
µn = .
2
Thus, Re µ± n = −γ/2 and conditions (6.53) are satisfied for λ = 1 if γ is small
enough. It was proved in [220] that for every m ∈ N, every ε > 0 and γ > 0 small
enough, there exists a perturbation Fi , ∥Fi ∥C m 6 ε, such that the perturbed system

u′′0 = u0 − u′0 + F0 (u0 , u′0 , u, ∂t u),

(6.55)
u′′n + γu′n + λn un = Fn (u0 , u′0 , u1 , u′1 , . . . ), n = 1, 2, . . . ,

possesses a Cγ −1 -dimensional invariant torus in a small neighbourhood of the un-

perturbed homoclinic loop. Moreover, only finitely many (n0 = Cγ −1 ) modes are
really perturbed and the perturbation depends on un , n 6 n0 , only.
It remains to note that equations (6.55) can easily be embedded in a system
of the form (6.39), where the C m -norm of the nonlinearity F is uniformly bounded
as γ → 0. This shows that the upper bounds dimf (A, E) ∼ Cγ −1 are indeed sharp.
140 S. V. ZELIK

Remark 6.32. Note that, in general, we need a large number of parameters

(of order proportional to dimL (S(t), u0 )) in order to be able to bear the torus and use
this scheme for constructing examples with large fractal dimension of the attractor.
We expect that this method will be helpful for other types of equations of hyperbolic
type, for example, for damped Euler equations and their various regularizations.

7. Inertial manifolds and finite-dimensional reduction

In the previous section we constructed a reduced system of ODEs (an inertial

form of the initial PDE), which captures the limit dynamics on the attractor A,
relying on the finiteness of its fractal dimension and Mané’s projection theorem
(see (6.3)). We also pointed out the main drawback of such a reduction, namely,
a drastic loss of smoothness. In this section we discuss an alternative approach to
the finite-dimensional reduction, which is based on the theory of centre manifolds
(or more generally, on the theory of normally-hyperbolic invariant manifolds),
which requires stronger assumptions on the dynamical system under consideration,
but gives a much more appropriate construction of the inertial form, which is
suitable for applications. To the best of our knowledge, this approach was proposed
by Mané [168] and has become very popular after the paper of Foias, Sell, and
Temam [76] (see also [2], [175], [195], [199], [243], and the references therein).
Following this approach, the main object of the theory is not an attractor,
but a globally stable finite-dimensional invariant submanifold of the initial phase
space, which is typically normally hyperbolic and, for this reason, exponentially
attractive. The reduced inertial form in this situation is nothing else than
the restriction of the initial PDE to this invariant manifold and is typically as
smooth as the manifold. Keeping in mind possible applications to hydrodynamics
and the hope to understand turbulence, this manifold has been referred to as the
inertial manifold in reference to the so-called inertial scale in the conventional
theory of turbulence (see, for example, [81] and the references therein). We start
with the formal definition of an inertial manifold.
Definition 7.1. Let S(t) : Φ → Φ be a dynamical system acting in a Banach
space Φ. Then a strictly invariant finite-dimensional Lipschitz submanifold M
of the phase space Φ is an inertial manifold for S(t) if it possesses an expo-
nential tracking property (an asymptotic phase) in the following form: for any
semi-trajectory u(t) = S(t)u0 , t > 0, of the dynamical system under consideration
there exists a semi-trajectory u(t) = S(t)u0 with u0 ∈ M such that

∥u(t) − u(t)∥Φ 6 Q(∥u0 ∥Φ )e−αt , t > 0, (7.1)

where the positive constant α and a monotone function Q are independent of u0

and t.
Remark 7.2. In applications an inertial manifold is often constructed as the
graph of a Lipschitz function. Namely, assume that Φ = Φ+ ⊕ Φ− is represented
as a direct sum of two Banach spaces (the corresponding projectors are denoted
by Π+ and Π− , respectively) where dim Φ+ = N . Then u(t) = Π+ u(t) + Π− u(t) =
u+ (t) + u− (t). The variables u+ (t) and u− (t) are treated as the “slow” and
 ATTRACTORS. THEN AND NOW 141

“fast” ones, respectively, and the manifold M slaves the fast variables u− (t) to
the slow ones u+ (t), that is, u− (t) = M (u+ (t)) for some at least Lipschitz function
M : Φ+ → Φ− , so that M is treated as the graph of this function:

M := {u+ + M (u+ ), u+ ∈ Φ+ }. (7.2)

We note from the very beginning that in applications one usually constructs
an inertial manifold not for the original PDE, but for a properly modified one
(the so-called “prepared” equation), whose solutions have the same asymptotic
behaviour, but it itself has a better structure, for example, its nonlinearity is
globally Lipschitz continuous. Since an inertial manifold is a kind of a (global)
centre manifold, it is not unique in general and we need to cut-off the nonlinearity
properly to restore the uniqueness and the possibility to find it using Banach’s
contraction theorem. In contrast to ODEs, this preparation procedure may be
very delicate: sometimes it is enough just to cut-off the nonlinearity outside an
appropriate absorbing set, thus making it globally Lipschitz (this usually works
when the spectral gap conditions are satisfied): see [76]; but in other cases, you
may need to embed your initial PDEs in a larger system of PDEs, do some
diffeomorphisms to restore the spectral gap conditions, change the leading-order
differential operator, and so on (see [133], [126], [131], [128], [166]). In addition, if
you want to improve the smoothness of an inertial manifold, you need to change
the vector field at points of the attractor as well (in an accurate way in order not
to affect the dynamics on the attractor, but kill the resonances); see [136]. We do
not go into further details here and will always assume in what follows that we have
already made some of these “preparations” and the nonlinearity is already globally
Lipschitz.

7.1. Spectral gap conditions and inertial manifolds. We start our ex-
position with the classical theory of inertial manifolds, which usually deals with
a semilinear equation of the form:

∂t u + Au = F (u), u t=0
= u0 ∈ Φ. (7.3)

We assume for simplicity that Φ is a real Hilbert space, A : D(A) → Φ is a positive

self-adjoint linear operator with compact inverse, and F : Φ → Φ is a given
nonlinearity, which is Lipschitz continuous with Lipschitz constant L:

∥F (u1 ) − F (u2 )∥Φ 6 L∥u1 − u2 ∥Φ , u1 , u2 ∈ Φ. (7.4)

From the Hilbert–Schmidt theorem we know that the operator A possesses a complete
basis of eigenvectors {ei }∞ ∞
i=1 with the corresponding eigenvalues {λi }i=1 , which we
enumerate in the non-decreasing order. The key result of the theory is the following
theorem.
Theorem 7.3. Let the nonlinearity F satisfy (7.4), and let N ∈ N be such that
the following spectral gap condition is satisfied:

λN +1 − λN > 2L. (7.5)

142 S. V. ZELIK

Then equation (7.3) possesses an N -dimensional inertial manifold which can be

represented as the graph of a Lipschitz-continuous function M : Φ+ → Φ− , where Φ+
is spanned by the first N eigenvectors of A and Φ− is the orthogonal complement of Φ+ .
Moreover, if F ∈ C 1+ε (Φ, Φ) for a sufficiently small ε > 0, then the corresponding
map M is also C 1+ε -smooth. We also mention that the attraction exponent α
in the definition of an inertial manifold can be chosen as α = (λN + λN +1 )/2.
Idea of the proof. The construction of the manifold is based on perturbation
arguments where the nonlinearity F is considered as a perturbation. Indeed,
for F ≡ 0 we have M = Φ+ , all trajectories on this manifold grow not faster
than e−θt as t → −∞, where θ ∈ (λN , λN +1 ), and this is a determining property
for the manifold M.
This observation hints at the way to construct the desired inertial manifold M,
namely, we need to solve the problem

∂t u + Au = F (u), t 6 0, Π+ u t=0
= u + ∈ Φ+ (7.6)

backward in time in the weighted space Leθt (R− , Φ), where θ ∈ (λN , λN +1 ) is chosen
in an optimal way (θ = (λN + λN +1 )/2 in our case). Then the desired map M
will be found as follows: M : u+ → Π− u(0) (see [243] for more details). Note that
the idea to use here the space L2 with respect to time belongs to Miklavcic [175].
The use of the space Ce√θt (R− , Φ), which appears more natural at first glance, leads

to the extra multiplier 2 on the right-hand side of (7.5).

Following [175], [243] (see also [76]), we solve equation (7.6) using Banach’s
contraction theorem, and the most important step here is to find the norm
of the solution operator Lθ : h → v for the following linear problem on the whole
line t ∈ R:
∂t v + Av = h(t), h ∈ L2eθt (R, Φ), (7.7)
in the weighted space L2eθt (R, H), namely, to verify that
 
1 1
∥Lθ ∥L(L θt ,L θt ) = max
2 2 , . (7.8)
e e θ − λN λN +1 − θ

The proof of this estimate is very elementary since the problem is reduced to
the analogous one for the scalar ODEs

vn′ + λn vn = hn (t),

which are the equations on the Fourier amplitudes for (7.7). An elementary
calculation shows that
1
∥Lθ,n ∥L(L2θt ,L2θt ) = (7.9)
e e |λn − θ|
and Parseval’s equality gives us the relation ∥Lθ ∥ = maxn∈N ∥Lθ,n ∥ (here we see
the advantage of using weighted L2 -spaces). This yields the desired estimate (7.8).
The rest of the proof is also straightforward. We just invert the linear part of (7.6)
(after an appropriate restriction to the negative semi-axis t 6 0) and apply Banach’s
contraction theorem. The optimal exponent θ = (λN + λN +1 )/2 gives us the value
2/(λN +1 − λN ) for the norm of Lθ , so the Lipschitz constant of the composition Lθ
 ATTRACTORS. THEN AND NOW 143

with F does not exceed 2L/(λN +1 −λN ) < 1 due to the spectral gap condition. Thus
the map constructed is indeed a contraction. The exponential tracking property
is also an almost immediate corollary of (7.8) and Banach’s contraction theorem
(see [243] for the missed details).
Remark 7.4. The spectral gap condition (7.5) can be generalized to the case
where the operator consumes smoothness. Namely, instead of (7.4) assume that
the operator F is globally Lipschitz as a map from Φ to Φ−s := D(A−s/2 ), for some
s ∈ (0, 2), with the same Lipschitz constant L. Then an analogue of (7.5) reads

λN +1 − λN
s/2 s/2
>L (7.10)
λN + λN +1

and this condition is sharp (moreover, s < 0 is also possible under some natural
extra assumptions); see [243]. In contrast to this, very little is known about sharp
spectral gap conditions in the case where the operator A is not self-adjoint. For
instance, in the model case of a coupled system of two equations in H = Φ × Φ
where the leading operator A has Jordan blocks:
 
1 1
A= A
0 1

and F : H → H is globally Lipschitz, the sharp spectral gap conditions read

(λN +1 − λN )2
q >L (7.11)
λN +1 + λN + 2 λ2N − λN λN +1 + λ2N +1

(see [134]). Such systems appear naturally after the so-called Kwak transform
applied to, say, the Navier–Stokes equations (see [144]). We see that (7.11) differ
drastically from the self-adjoint case (7.5) and are close to (7.10) for s = 1.
Thus, the presence of a Jordan block in the leading linear part of the equation
is somehow equivalent to consuming one unit of smoothness by the nonlinearity
(with self-adjoint linear part). Exactly this fact was overseen in Kwak’s erroneous
construction [144] (see also [111], [145], [216]) of an inertial manifold for the 2D
Navier–Stokes problem.
We mention here the paper [33] (see also [31] and [182] for weaker results), where
the hyperbolic relaxation

ε∂t2 u + ∂t u + Au = F (u), ε > 0,

of (7.3) is considered. It was shown there that the sharp spectral gap condition
for this problem coincides with (7.5) (where λn are the eigenvalues of A) and
is independent of ε for ε small enough. Inertial manifolds for elliptic boundary
problems in cylindrical domains were studied in [7] and [172]. We also mention that
the necessary conditions for the existence of inertial manifolds are often formulated
in the spirit of the theory of dynamical systems in terms of invariant cones. We do
not present the details here and refer the interested reader to [2], [195], [243] for
more details.
144 S. V. ZELIK

Let us discuss some examples where the classical theory is applicable (more
examples can be found in the survey [243], see also [215] and the references therein).
We start with the system of reaction-diffusion equations in a bounded domain Ω
of Rd :
∂t u = a∆x u − f (u), u = (u1 , . . . , um ), (7.12)
endowed with appropriate boundary conditions. We assume that the diffusion
matrix a is self-adjoint and positive definite and f is globally Lipschitz. Then
the operator A := −a∆x is self-adjoint and positive (non-negative in the case
of periodic or Neumann boundary conditions) in Φ = [L2 (Ω)]m and, due to Weyl’s
theorem, we have
λn ∼ Cm n2/d , (7.13)
so the validity of spectral gap conditions depends strongly on the dimension d.
When d = 1, we have infinitely many values of N such that λN +1 − λN > cN
for some positive c, so we have spectral gaps of any size, and therefore, for any
Lipschitz constant L of f , we may find N satisfying the spectral gap condition and
this guarantees the existence of an inertial manifold.
The case d = 2 is more interesting. In this case Weyl’s theorem can guarantee only
that λN +1 − λN > c for infinitely many values of N , so we may apply Theorem 7.3
only if L is small enough. However, we may still have spectral gaps of any size
in the spectrum of the Laplacian despite the relation λn ∼ Cn. For instance, for
Ω = [−π, π]2 with periodic boundary conditions, we have infinitely many values
of N satisfying
λN +1 − λN > C log λN
(see [190]), so an inertial manifold exists in the case of periodic boundary condi-
tions. The ideal situation from the point of view of inertial manifolds is the case
of the Laplace–Beltrami operator on the d-dimensional sphere Sd , where the in-
1/2
equality λN +1 − λN > cλN has infinitely many solutions in any space dimension d.
Thus, reaction-diffusion equations of the form (7.12) on a sphere always have
inertial manifolds in any space dimension. Note that, to the best of our knowledge,
the problem of the existence of spectral gaps of any size for the 2D Laplacian
in a bounded domain Ω ⊂ R2 is completely open. On the one hand, we do not
know any examples where such gaps do not exist and, on the other hand, we do not
know any reasonably general classes of domains that possess this property.
In contrast to this, in dimension three or higher spectral gaps of an arbitrary size
exist in very exceptional cases only (like sphere), so the classical inertial manifolds
theory is not very helpful for 3D reaction-diffusion equations. Nevertheless, spectral
gap conditions are still satisfied for higher-order equations like the Swift–Hohenberg
equation
∂t u + (∆x + 1)2 u = f (u),
since for the bi-Laplacian Weyl’s theorem gives λn ∼ cn4/3 , and this guarantees
the existence of spectral gaps of any size.
Now consider equations with nonlinearities decreasing the regularity. The classical
example here is the Kuramoto–Sivashinski equation in 1D:

∂t u + ∂x4 u − a∂x2 u + ∂x (u2 ) = 0, u t=0

= u0 , a > 0,
 ATTRACTORS. THEN AND NOW 145

in Ω = [−π, π] endowed with periodic Z π boundary conditions. Since we have

1
the conservation law here: ⟨u⟩ := u(x) dx, we must consider this equation
2π −π
in the phase space Φ := L2 (−π, π) ∩ {⟨u⟩ = 0}. It is well known (see [89], [215]) that
this equation generates a dissipative semigroup S(t) in Φ which possesses a smooth
2
a2

a
absorbing set. We set A := ∂x2 − + 1 and F (u) = + u − ∂x (u2 ). After
2 4u
an appropriate cut-off, the nonlinearity F becomes globally Lipschitz as a map
from H to H −1 (Ω) = D(A1/4 ), an therefore we need to check the spectral gap
condition (7.10) for s = 1/2. We know that λn ∼ cn4 , and therefore

λN +1 − λN
1/4 1/4
∼ c1 N 2
λN + λN +1

and we have spectral gaps of arbitrarily large size. Thus, this equation possesses an
inertial manifold.
One more example, which is interesting for what follows, is given by an 1D
system of reaction-diffusion advection equations:

∂t u − ∂x2 u + u = f (u, ∂x u), u = (u1 , . . . , um ), u t=0

= u0 , (7.14)

endowed with Dirichlet, Neumann, or periodic boundary conditions. We assume

that f is smooth and both fu′ and f∂′ x u are uniformly bounded. Here A = −∂x2 + 1
and the nonlinearity f consumes one unit of smoothness, so s = 1. Since λn ∼ cn2 ,
the corresponding spectral gap condition reads
λn+1 − λn
1/2 1/2
∼ c > L.
λn + λn+1

Thus, we have an inertial manifold via the classical theory if the Lipschitz constant L
of f is small enough. A slightly more accurate analysis shows (see [132], [168]) that
the size of fu′ is not essential and only L1 := supu,v∈R |fv′ (u, v)| should be small.
As we will see below, this observation is crucial for the recent theory of inertial
manifolds for these equations, which is discussed in the next subsection.
We now turn to the smoothness of inertial manifolds. First we note that,
in the case where M is the graph of a function M : Φ+ → Φ− and Φ+ is a spectral
subspace of the operator A, the inertial form (6.3) for equation (7.3) is significantly
simplified:

∂t u+ + Au+ = Π+ F (u+ + M (u+ )), u + ∈ Φ+ ∼

= RN , (7.15)

and we see that the regularity of the reduced equations is determined by the smooth-
ness of the map M . Theorem 7.3 guarantees that these reduced equations are
C 1+ε -smooth for some small ε > 0 if the spectral gap conditions are satisfied
and F is smooth. This regularity cannot be improved in general since, similarly to
the theory of centre manifolds (or, more generally, the theory of normally hyper-
bolic invariant manifolds), there are obstacles to a further regularity of M (see, for
instance, a counterexample of Sell [40] related to resonances). The nature of these
146 S. V. ZELIK

obstacles can be explained as follows. Let us formally differentiate equation (7.6)

in order to get an equation for the Fréchet derivative of the map M :

∂t v + Av = F ′ (u(t))v, Π+ v t=0
= ξ ∈ Φ+ , (7.16)

where M ′ (u+ )ξ := Π− v(0). Since ∥F ′ (u(t))∥ 6 L, the spectral gap conditions

still allow us to solve this equation uniquely in the space L2eθt (R− , Φ) and define
the map M ′ (u+ ). A slightly more accurate analysis shows that the map obtained is
Hölder continuous with small positive Hölder exponent ε > 0 and the function M
is C 1+ε (see, for example, [136], [243]).
The situation changes drastically when we differentiate equation (7.16) once more
and look at the second derivative:

∂t w + Aw − F ′ (u(t))w = F ′′ (u(t))[vξ (t), vη (t)] =: hξ,η (t), Π+ w t=0

= 0, (7.17)

where vξ and vη are the solutions of (7.16) with the initial data ξ and η, respectively.
The problem here is that vξ , vη ∈ L2eθt (R− , Φ) for some θ,

λN + L < θ < λN +1 − L (7.18)

(in any case, we can guarantee the existence of a solution of (7.16) on the
basis of Banach’s contraction theorem only for such values of θ). Therefore,
hξ,η ∈ L2e2θ (R− , Φ) (since we have the parabolic smoothing property, the product
of two solutions belonging to a weighted L2 is also in L2 with an appropriate
weight). But in order to solve (7.17), we need the exponent 2θ to satisfy (7.18),
and this is possible only if
λN +1 − 2λN > 3L. (7.19)
Analogously, if we want the inertial manifold to be C s -smooth for some s > 1, then
we need the following spectral gap:

λN +1 − sλN > (s + 1)L. (7.20)

This condition is actually sharp, and there is an example due to Sell [40] of an
equation of the form (7.3) which possesses a C 2−ε -smooth inertial manifold, but
does not possess any C 2 -smooth inertial manifold (see also [136] and the references
therein).
Note that there is a principal difference between conditions (7.5) and (7.19),
namely, the first condition requires the existence of gaps of arbitrarily large size
in the spectrum of A and can be satisfied at least for some elliptic operators
in bounded domains (for example, in the case of low space dimension). In contrast
to this, condition (7.19) requires exponentially big lacunas in the spectrum (for
example, λn = an with a > 2), which is difficult to expect in the case of elliptic
operators. In fact, we do not know any examples of such operators A for which
inequality (7.19) is solvable with respect to N for any Lipschitz constant L. Thus,
inertial manifolds constructed using Theorem 7.3 are never C 2 -smooth in general
if more or less realistic applications are considered. The only exception is the case
of local bifurcations and associated local centre manifolds, where λN ∼ 0 and L
 ATTRACTORS. THEN AND NOW 147

can be chosen arbitrarily small by decreasing the size of the neighbourhood. Then
condition (7.20) allows us to construct invariant manifolds of any finite smoothness.
Nevertheless, there is a possibility to overcome (at least partially) the smoothness
problem for inertial manifolds by increasing the dimension of the manifold and
cutting off the nonlinearity F , namely, the following result was proved in [136].

Theorem 7.5. Let the assumptions of Theorem 7.3 hold. Assume, in addition,
that F ∈ C ∞ (Φ, Φ) and the following stronger version of the spectral gap conditions
holds:
lim sup(λN +1 − λN ) = ∞. (7.21)
N →∞

Also let M1 be a C 1+ε -smooth inertial manifold which corresponds to the first value
of N which satisfies (7.5). Then for every m ∈ N and any small positive δ there
exists a C m -smooth modification Fm of the initial nonlinearity F such that
1) the manifold M1 remains an inertial manifold for the modified problem

∂t u + Au = Fm (u), u t=0
= u0 ; (7.22)

2) equation (7.22) possesses a C m -smooth inertial manifold Mm such that M1 is

a normally hyperbolic exponentially stable invariant submanifold of Mm ;
3) the nonlinearity Fm is δ-close to F in the C 1 -norm.

Idea of the proof. Thanks to (7.21), we have infinitely many spectral gaps
suitable for constructing inertial manifolds. The first of them is used to construct
the manifold M1 . The key idea is to use the second spectral gap to solve
equation (7.17) for the second derivative; then the third gap is used to find the third
derivative, and so on. Then, for every u ∈ M1 , the above procedure gives us an
m-jet which should correspond to the m-smooth extension Mm of the manifold M1 .
Such a jet can be constructed in many ways, and the most difficult part of the proof
is to fix it so that the compatibility conditions in Whitney’s extension theorem are
satisfied. Then we get the desired manifold Mm by this theorem, and after that
it is already not difficult to define the correction Fm of the initial nonlinearity F
in such a way that Mm is an invariant manifold for the modified system (7.22);
see [136] for the details.

Remark 7.6. Note that the construction in Theorem 7.5 replaces the require-
ment (7.20) that one “huge” spectral gap exists (which is almost never satisfied)
by the existence of many relatively small gaps, which is satisfied if we are able to
satisfy (7.5) for any Lipschitz constant L (so, this is not a big extra restriction).
Note also that this theorem allows us to interpret the C 1+ε -smooth inertial
form (7.15) (which is a system of ODEs in RN1 ) as the reduced equations on
the normally hyperbolic invariant manifold for the extended C m -smooth system
of ODEs in RNm , which is the inertial form related to the inertial manifold Mm .
Of course, such an extension does not exist for a general non-smooth system of ODEs
and is strongly related to the fact that the system of ODEs under consideration is
obtained as a reduction of the smooth infinite-dimensional system (7.3).
148 S. V. ZELIK

7.2. Beyond the spectral gap conditions. We now turn to the case where
the spectral gap condition (7.5) is violated. We start our exposition with the sharp-
ness of the assumptions of Theorem 7.3. Assume that (7.5) is not satisfied for some
fixed N ∈ N, namely, that λN +1 − λN < 2L. Consider the following linear version
of system (7.3):
d
un + λn un = 0, n ̸= N, N + 1,
dt (7.23)
d d
uN + λN uN = LuN +1 , uN +1 + λN +1 uN +1 = −LuN .
dt dt
Then for the corresponding F (u) = F u we have ∥F ∥L(Φ,Φ) = L, so L is indeed a Lip-
schitz constant for the perturbation F (u). Then the eigenvalues of the perturbed
system associated with the invariant subspace spanned by eN and eN +1 ,
s 2
± λN +1 + λN λN +1 − λN
µN = ± − L2 ,
2 2
are complex conjugate with non-zero imaginary part. For this reason we cannot
decouple uN (t) and uN +1 (t), and an invariant N -dimensional manifold (linear
subspace) with the base Φ+ cannot exist. The same example shows that we also do
not have a normally hyperbolic invariant subspace of dimension N if the spectral
gap condition is violated for N (see, for instance, [197], [243] for more details).
Now assume that (7.5) is violated for all N ∈ N, that is,
sup (λN +1 − λN ) < 2L. (7.24)
N ∈N

Then it is natural to find examples where we have no inertial manifold for any
dimension N . Such examples are often based on the following lemma (see [197], [243])
and its generalizations.
Lemma 7.7. Assume that the nonlinearity F in equation (7.3) belongs to C 1 ,
and let u± ∈ Φ be two equilibria of this equation. Let Lu± := −A + F ′ (u± ) be
the linearization of (7.3) at u = u± . Assume that the spectrum σ(Lu+ ) consists
of complex conjugate eigenvalues with non-zero imaginary parts and the spectrum
of σ(Lu− ) contains one simple positive real eigenvalue and the rest of it consists
of complex conjugate eigenvalues with non-zero imaginary parts. Then problem (7.3)
does not possess any finite-dimensional C 1 -smooth inertial manifold.
Sketch of the proof. Indeed, assume that such a manifold M exists.
Consider its tangent planes Tu± M at u = u± . Since the manifold is invariant,
the planes Tu± M are invariant with respect to Lu± . Therefore,


σ Lu± Tu ⊂ σ(Lu± ).
±

Since the equation is real-valued and the spectrum of Lu+ does not contain real
eigenvalues, we conclude that dim M = dim Tu+ M is even. On the other hand,
since the inertial manifold always contains an attractor and the attractor always
contains an unstable manifold, the direction of the simple real eigenvector of Lu−
must lie in Tu− M. Then the analogous arguments show that dim M = dim Tu− M
is odd. This contradiction proves that the inertial manifold M cannot exist.
 ATTRACTORS. THEN AND NOW 149

Corollary 7.8. Let λ1 < L, and let assumption (7.24) be satisfied. Then there
exists a smooth bounded and globally Lipschitz nonlinearity F with Lipschitz con-
stant L such that equation (7.3) does not possess any C 1 -smooth finite-dimensional in-
ertial manifold.
Sketch of the proof. According to the previous lemma, we only need to
construct F in equation (7.3) so that it possesses two equilibria u = u± such that
the linearizations near u+ and u− have the form

d d
v2n−1 + λ2n−1 v2n−1 = Lu2n , v2n + λ2n v2n = −Lu2n−1 , n ∈ N, (7.25)
dt dt
and
d
v1 + λ1 v1 = Lu1 ,
dt (7.26)
d d
v2n + λ2n v2n = Lu2n+1 , v2n+1 + λ2n+1 v2n+1 = −Lu2n , n∈N
dt dt
respectively. Then all assumptions of the previous corollary will be satisfied. The
construction of such a nonlinearity F is straightforward and we drop it here (see [243]
for the details).
Remark 7.9. Note that the fractal dimension of the global attractor A which
corresponds to equation (7.3) is always finite if F is bounded and globally Lips-
chitz. Thus, according to the common paradigm, the corresponding dynamics on
the attractor should be “finite-dimensional”. The result of Corollary 7.8 does not
essentially contradict this heuristic principle: it just tells us that the dimension
of the inertial “manifold” should be different in different parts of the phase space.
For instance, instead of an inertial manifold, one may try to consider an “iner-
tial CW-complex” with finite-dimensional dynamics on it. As we will see below,
this actually also does not work and the reduced dynamics may be essentially
infinite-dimensional despite the finiteness of the fractal dimension.
The further progress in understanding the obstacles to the existence of inertial
manifolds and their consequences is related to the recently discovered connections
of the theory of inertial manifolds with Floquet’s theory for infinite-dimensional
differential equations (see [64], [133], [139], [243], and the references therein).
Namely, consider the following linear time-periodic parabolic equation in a Hilbert
space Φ:
∂t v + Av = L(t)v, v t=0 = v0 , (7.27)
where L(t) is a linear operator T -periodic in time and satisfying ∥L(t)∥L(Φ,Φ) 6 L.
Roughly speaking, we fix the operator L(t) in such a way that (7.27) is close
to (7.25) and (7.26) on the first and second half-periods, respectively. Moreover,
this can be done in such a way that the period map P : u0 → u(t) has the form
of an infinite Jordan cell:

. . . , Pe2n = µ2n e2n−2 , ..., Pe2 = µ0 e1 ,

(7.28)
Pe1 = µ1 e3 , ..., Pe2n+1 = µ2n+1 e2n+3 , ...,
150 S. V. ZELIK

where µn ∼ e−α|n| for some positive α. In particular, the map P is compact and
σ(P) = {0}, so the corresponding equation (7.27) does not possess any Floquet
multipliers and any solution of it decays super-exponentially to zero:
2
∥v(t)∥Φ 6 C∥v0 ∥Φ e−αt /2
(7.29)

(see [64], [243] for more details). The next step is to realize the operator L(t)
in the form of L(u0 (t)), where u0 (t) is a time-periodic solution of an ODE which
is interpreted as an equation for the “zero-mode” of the abstract equation (7.3).
This will give us a super-exponentially attracting periodic orbit u(t) := {u0 (t), 0}
in the system of the form (7.3), which belongs to the attractor. It only remains to
modify the equations properly away from this periodic orbit in order to find another
periodic orbit u(t) belonging to the attractor and converging super-exponentially
fast to u(t):
2
∥u(t) − u(t)∥Φ 6 Ce−αt /2 , u, u ∈ A (7.30)
(see [64], [243] for more details). The presence of these two trajectories on
the attractor clearly excludes the existence of Lipschitz continuous inertial manifolds
of any finite dimension. Moreover, an appropriate modification of this construction
also excludes the possibility to embed the attractor A in any finite-dimensional
Lipschitz or even log-Lipschitz continuous (not necessarily invariant) submanifolds.
Namely, the following result was proved in [64].
Theorem 7.10. Let the assumptions of Corollary 7.8 be satisfied. Then there
exists a smooth and bounded nonlinearity F with global Lipschitz constant L such
that there exist two solutions u(t) and u(t) of equation (7.3) belonging to the attractor
and satisfying (7.30). Moreover, the corresponding attractor cannot be embedded
in any finite-dimensional log-Lipschitz submanifold of Φ. In particular, no inertial
manifold exists for this equation.
Remark 7.11. Since under the assumptions of Theorem 7.10 the fractal dimen-
sion of the attractor A is finite, thus due to Mané’s projection theorem we have
a Hölder-continuous inertial form on the attractor, as well as its embedding in
a Hölder-continuous submanifold of Φ. Moreover, the Hölder exponent can be made
as close to 1 as we want by increasing the dimension of the manifold (see [193] and
the references therein). Nevertheless, the super-exponential attraction of trajecto-
ries is not observed in the classical dynamics generated by smooth ODEs and can
hardly be interpreted as a finite-dimensional phenomenon. Thus, we see some kind
of infinite-dimensional limit dynamics on an attractor of finite fractal dimension.
This phenomenon is not properly understood yet, and it definitely deserves further
investigation.
Remark 7.12. Note that it is relatively easy to construct counterexam-
ples to Floquet’s theory on the level of abstract linear parabolic equations
of the form (7.27), for instance, by constructing the desired operator L(t)
in the Fourier base of the operator A. The situation becomes much more compli-
cated if we want (7.27) to be a true parabolic PDE or system of parabolic PDEs.
The existence of counterexamples to Floquet’s theory on the level of parabolic
PDEs in bounded domains was been an open problem till recently. This problem
 ATTRACTORS. THEN AND NOW 151

was affirmatively solved in [133], where the smooth space-time periodic m × m

matrices a(t, x) and b(t, x) were constructed for m > 4 in such a way that all
solutions of the corresponding linear reaction-diffusion-advection problem

∂t v − ∂x2 v = a(t, x)v + b(t, x) ∂x v, v t=0

= v0 , v = (v 1 , . . . , v m ),
x ∈ (−π, π),
(7.31)
endowed with periodic boundary conditions, decay super-exponentially in time:
3
∥v(t)∥Φ 6 ∥v0 ∥Φ e−αt , t > 0, α > 0.

Moreover, an example of a system of semilinear reaction-diffusion-advection

equations of the form

∂t u − ∂x2 u = f (u) + g(u) ∂x u, u t=0

= u0 , u = (u1 , . . . , um ),
x ∈ (−π, π),
(7.32)
with m > 8, periodic boundary conditions, and C ∞ -smooth functions f : Rm → Rm
and g : Rm → Mat(m × m), such that this system does not possess any
finite-dimensional inertial manifold and satisfies all the assertions of Theorem 7.10,
was given in [133] on the basis of the counterexample (7.31).
We now turn to examples where the inertial manifold still exists despite the fact
that the spectral gap condition is not satisfied. We have seen above that these
conditions are sharp on the level of abstract parabolic equations; however, there is still
a possibility to relax them if some specific subclass of such equations is considered.
For instance, the first such example was due to Sell and Mallet-Paret [166]
where the inertial manifold for a scalar semilinear heat equation on the 3D
torus Ω = [−π, π]3 was constructed. We return to this example below, but
prefer to consider first an alternative method related to finding an appropriate
transformations or/and embedding of the initial system into a larger one for which
the spectral gap conditions are satisfied. We illustrate this approach by an example
of the reaction-diffusion-advection system (7.32) endowed with Dirichlet boundary
conditions. Namely, let us make the change of the independent variable u = a(t, x)w,
where the matrix a(t, x) will be specified in what follows. Then we arrive at
the transformed equation

∂t w − ∂x2 w = {a−1 (2 ∂x a + g(aw)a) ∂x w}

+ {a−1 (∂x2 a − ∂t a + g(aw) ∂x a)w + a−1 f (aw)}
=: F1 (w) + F2 (w). (7.33)

We see that the operator F2 does not depend explicitly on ∂x w and this dependence
is presented in F1 only, so the naive idea would be to kill the term F1 by an
appropriate choice of the matrix a = a(u), for instance, by fixing it as a solution
of the matrix ODE
1 d
a + g(u)a = 0, a x=−π = Id .
2 dx
However, this naive idea will not work since the terms ∂t a and ∂x2 a will implicitly
depend on ∂x w and the operator F2 will be not bounded from Φ := H01 (−π, π) to Φ.
Fortunately, as we already mentioned in discussing equation (7.14), we need not kill
152 S. V. ZELIK

the operator F1 completely: it is sufficient to make its Lipschitz constant (as a map
from Φ to L2 (−π, π)) small enough and this hints at the suitable transformation:
we fix a = a(u) as a solution of the following problem:

1 d
a + g(PK u)a = 0, a x=−π
= Id, (7.34)
2 dt
where PK is some smoothing operator in x, for instance, we can fix it as the spectral
projector onto the first K eigenvectors of −∂x2 in Ω = (−π, π) with Dirichlet boundary
conditions and can set K to be large enough. It was shown in [132] that u = a(aw)w
thus defined is indeed a diffeomorphism of the phase space Φ, the Lipschitz norm
of F1 : Φ → L2 (−π, π) can be made arbitrarily small by choosing K large enough,
and the map F2 : Φ → Φ is globally Lipschitz continuous. Thus, an appropriate
version of the spectral gap conditions is satisfied, and we have the following result
(see [132] for the details).
Theorem 7.13. Let the functions f and g be smooth and have finite supports.
Then equation (7.32) endowed with Dirichlet boundary conditions possesses an
inertial manifold.
Remark 7.14. As we see from Remark 7.12 and Theorem 7.13, the existence or
non-existence of an inertial manifold for the reaction-diffusion-advection problems
depends strongly on the type of boundary conditions. This is related to the fact that
the transformation u = a(t, x)w we use preserves Dirichlet boundary conditions in
a natural way, but does not preserve Neumann or periodic conditions. In the case
of Neumann boundary conditions, there is a nice trick which allows us to overcome
this problem, namely, we use that v = ∂x u satisfies Dirichlet boundary conditions
and embed system (7.32) into the larger system

∂t u − ∂x u = f (u) + g(u)v, ∂x u x=±π

= 0,
∂t v − ∂x2 v = f ′ (u)v + g ′ (u)[v, v] + g(u) ∂x v, v x=±π
= 0.

Here we have only one dangerous term g(u)∂x v, and this term can be made
small by using the transformation of v-component only where Dirichlet boundary
conditions are preserved. This gives us an analogue of Theorem 7.13 in the case
of Neumann boundary conditions (see [133] for more details). Note that in the case
of periodic boundary conditions an analogue of this trick cannot exist since we have
a counterexample where an inertial manifold does not exist (see Remark 7.12). We
emphasize that this counterexample is constructed for systems with m > 8 only
and it is proved in [133] that, in the case of a scalar equation, that is, for m = 1,
we still have an inertial manifold.
We also mention that the general case of equation (7.14) can also be treated
in a similar way by differentiating the equation with respect to time and embedding
the problem into a larger system of equations of the form (7.32) (see [133] for more
details). Thus, the assumption of smallness of the derivative ∂ux f can be removed
in the case of Dirichlet or Neumann boundary conditions, but cannot be relaxed
in general for systems with periodic boundary conditions.
 ATTRACTORS. THEN AND NOW 153

Example 7.15. Consider a particular example of the forced Burgers equation:

∂t u − ν∂x2 u = ∂x (u2 ) + g(x), x ∈ (−π, π), u t=0

= u0 , (7.35)

endowed with Dirichlet boundary conditions (the Neumann and periodic cases are
similar, but we need to take care of the spatial mean value ⟨u⟩). We assume, say,
that g ∈ Φ := L2 (−π, π) and ν > 0. Then equation (7.35) is globally well posed and
dissipative in Φ. Indeed, multiplying (7.35) by u and integrating by parts (which
kills the nonlinear term) we end up with the energy identity

1 d
∥u(t)∥2Φ + ν∥∂x u(t)∥2L2 = (g, u(t)),
2 dt
which gives us the desired dissipativity (see [215] for more details). Moreover,
using the parabolic smoothing property we may construct an absorbing set for
the corresponding solution semigroup S(t) which is a bounded set in H 2 (−π, π) ⊂
C 1 [−π, π]. Thus, the semigroup S(t) possesses an attractor A in the phase space Φ
endowed with the standard bornology of bounded sets in Φ. In addition, we may
cut off the nonlinearity outside the absorbing set and write out the problem (7.35)
in the form of (7.32) with smooth nonlinearities with finite support. This gives us
the existence of an inertial manifold for this equation (see [132] for more details).
We note however that equation (7.35) is “too simple” to have any non-trivial
dynamics, and its attractor always consists of a single equilibrium:

A := {G}, νG′′ = (G2 )′ + g, G(±π) = 0. (7.36)

Moreover, this equation does not generate any dynamical instability, and every
trajectory of it is Lyapunov stable and even asymptotically stable. Indeed, let u1 (t)
and u2 (t) be two solutions of (7.35), and let v(t) := u1 (t) − u2 (t). Then v solves

∂t v − ∂x2 v = ∂x ((u1 + u2 )v), v t=0

= v0 .

Multiplying this equation by sgn(v(t)) and using Kato’s inequality (see [45]), we
end up with
d
∥v(t)∥L1 + ν|∂x v(−π)| + ν|∂x v(π)| 6 0. (7.37)
dt
Thus, the quantity ∥u1 (t) − u2 (t)∥L1 is non-increasing along trajectories. In partic-
ular, if we assume that u1 , u2 ∈ A, then we infer from (7.37) that

v x=±π
= ∂x u x=±π
=0

and Carleman-type estimates show us that v(t) ≡ 0 (see [200]). Since, by

the standard arguments, equation (7.35) possesses at least one equilibrium, we
see that this equilibrium is unique and (7.36) holds. Moreover, the standard
compactness arguments show that this equilibrium is exponentially stable, that is,

∥u1 (t) − G∥Φ 6 Cν ∥u1 (0) − G∥Φ e−αν t (7.38)

for some positive Cν and αν .

154 S. V. ZELIK

Thus, constructing inertial manifolds and studying the limit dynamics on the at-
tractor does not look very interesting for Burgers’ equation (7.35). (Although this
equation may still demonstrate a non-trivial intermediate behaviour, which may be
interesting from the hydrodynamical point of view (see [17], [141], and the references
therein), this phenomenon is more appropriate to study by using the concept of an
exponential attractor considered in the next section.) Alternatively, we may return
to Burgers’ original model of turbulence, where Burgers’ equation is coupled with
an ODE:
Z π
d
U + νU = P − u2 (t, x) dx, ∂t u − ν ∂x2 u + ∂x (u2 ) = U u (7.39)
dt −π

or to two-component Burgers equations

Z π
d
u2 (t, x) + v 2 (t, x) dx,


U + νU = P −
dt −π
∂t u − ν ∂x2 u + ∂x (u2 − v 2 ) = U (u − v),
∂t v − ν ∂x2 v − ∂x (2uv) = U (u + v),
(7.40)
where the function U (t) depends only on t and the components u and v are endowed
with Dirichlet boundary conditions (see [22] for the details). In contrast to the single
Burgers equation (7.35), we now have instability if the parameter P > 0 is large
enough (this parameter plays the role of external forces), so the attractor, as well as
the dynamics on it, becomes non-trivial. On the other hand, we still have an energy
identity and dissipativity for these equations, as well as the existence of a smooth
absorbing set. For instance, the energy identity for equation (7.40) reads

1 d 2
(U + ∥u∥2L2 + ∥v∥2L2 ) + ν(U 2 + ∥∂x u∥2L2 + ∥∂x v∥2L2 ) = P U.
2 dt
Moreover, the general theory of inertial manifolds for 1D reaction-diffusion-advection
problems is applicable here (we may make the terms containing ∂x u and ∂x v
small enough to satisfy the spectral gap conditions using the transformation
described above). This, in turn, allows us to construct inertial manifolds for both
equations (7.39) and (7.40). We observe that this result was stated in [111], however,
the proof there was based on an erroneous idea of Kwak, so the correct proof became
available only recently due to [132]. We also mention that some particular cases
of a general method proposed in [132] were considered in [230]–[232].

Remark 7.16. We recall that the 1D reaction-diffusion-advection equations can

be considered as simplified models for the 2D Navier–Stokes equations, so that
proving or disproving the existence of inertial manifolds for them was a long-
standing open problem of a great theoretical and practical interest. A number
of weaker results in this direction have been obtained. We mention here only
Romanov’s theory (see [196], [198]), which allowed us to construct Mané projectors
with Lipschitz inverses on attractors for some particular cases of these equations
(see [127], [197], [243]). A more or less complete answer to this question, which
was obtained in [133], [132] and was discussed above, is somehow unexpected
and surprising. Indeed, periodic boundary conditions are used in hydrodynamics
 ATTRACTORS. THEN AND NOW 155

mainly because they are “simpler” than more physical Dirichlet boundary condi-
tions. The situation with reaction-diffusion-advection equations hints that such an
approach may be essentially wrong and, in reality, periodic boundary conditions
may be the “most complicated” ones.
We conclude our exposition of inertial manifolds by considering the spatial aver-
aging method proposed by Sell and Mallet-Paret and its recent generalizations. We
recall that in order to construct an inertial manifold for equation (7.3), we need to
solve equation (7.6) backward in time in the properly chosen function space, and
the main step here is to solve the corresponding equation of variations (7.16). More-
over, looking at estimate (7.9) we see that the main contribution to the norm (7.8)
is made by the intermediate modes which correspond to the eigenvalues satisfying
λN − k < λn < λN +1 + k. Let Ik,N be the spectral orthoprojector onto these
modes. Then the most dangerous part of the operator F ′ (u(t)) in (7.16) is exactly
its intermediate part Ik,N ◦ F ′ (u(t)) ◦ Ik,N for the properly chosen k and N . The key
assumption of the method is that this part is close to a scalar operator.
Definition 7.17. We say that the nonlinearity F ∈ Cb1 (Φ, Φ) and the operator A
satisfy the spatial averaging condition if there exist a bounded measurable function
a : Φ → R and a positive number κ such that, for every ε > 0 and every k ∈ R+ ,
there exist infinitely many values of N ∈ N satisfying

∥Ik,N F ′ (u)Ik,N − a(u) Id ∥L(Φ,Φ) 6 ε, λN +1 − λN > κ. (7.41)

Then the function a(u) is referred to as a spatial averaging of the operator F ′ (u).
Theorem 7.18. Let the operator A : D(A) → Φ, A = A∗ > 0, with compact
inverse and nonlinearity F (u) satisfy the spatial averaging condition. Then equa-
tion (7.3) possesses an inertial manifold (actually, infinitely many of them) which
is the graph of a Lipschitz function over an appropriately chosen spectral subspace.
Moreover, if F and its spatial average are more regular, then the corresponding
inertial manifold is C 1+ε -smooth for some small ε > 0.
Idea of the proof. We only discuss how to solve the equation of varia-
tions (7.16); the full proof can be found in [166] (see also [30], [243]). Let us look at
the solution operator Lθ of equation (7.7) for θ = (λN +1 + λN )/2. Then by (7.9)
we have
1
∥Lθ ◦ (Id −Ik,N )∥L(Φ,Φ) 6 ,
k
and this part of the operator Lθ can be made small by fixing k large enough and
taking the standard Lyapunov metric

∥u∥2Φ,β := ∥(Id −Ik,N )u∥2Φ + β∥Ik,N u∥2Φ ,

where the parameter β is chosen properly, so that we only need to take care about
the intermediate modes (see [243] for the details). For these modes we use the spatial
averaging condition, which allows us to reduce the problem to equations of the form

d
vn + λn vn = a(u(t))vn + “small”.
dt
156 S. V. ZELIK

term a(u(t))vn 
Finally, the scalarZ can be removed by the change of independent vari-
s
ables vn (t) = exp a(u(s)) dx wn (t). It is important that this transformation
0
is the same for all (intermediate) modes. This reduces the problem to

d
wn + λn wn = “small”,
dt
and this problem can be solved using the fact that λN +1 − λN > κ.
This finishes the proof of solvability for the equation of variations. The
transition to the nonlinear case is a bit more complicated here since the weight
Z t 
exp a(u(s)) ds depends on the trajectory u(t), which is also unknown from
0
the very beginning. For this reason methods related to invariant cones are
traditionally used here (see [166], [243], [30] for the details), although this transition
can also be done by means of perturbation arguments and Banach’s contraction
theorem (see [126]).
In applications the operator A is usually the minus Laplacian with periodic
boundary conditions, mainly in the 3D case, say, in Φ = L2 (Ω), where Ω = [−π, π]3
(see [166], [146] for applications to irrational tori in 2D or/and equilateral triangles).
In this case the eigenvalues λn are naturally parameterized by points in the lattice Z3 :

λ⃗n = m2 + l2 + k 2 , ⃗n = (m, l, k) ∈ Z3 ,

and the number theory related to distributions of sums of squares comes into play
in a natural way. Namely, assume in addition that the operator F ′ (u) is pointwise
multiplication,
F ′ (u)v := ψ(x)v(x),
by a sufficiently smooth function ψ = ψu (x) (or is a sum of combinations of pointwise
multiplication and differentiation). Then, with respect to the Fourier basis,
the action of F ′ (u) becomes a convolution:
X
(F ′ (u)v)⃗n := ψ⃗n−⃗n1 v⃗n1 (7.42)
n1 ∈Z3
⃗

and we may use the following lemma to verify the spatial averaging condition.
Lemma 7.19. Let
k
CN := {⃗l ∈ Z3 : N − k 6 |⃗l|2 6 N + k}, Br := {⃗l ∈ Z3 : |⃗l| 6 r}. (7.43)

Then, for every k > 0 and r > 0, there exist infinitely many N ∈ N such that
k k
(CN − CN ) ∩ Br = {0}. (7.44)

The proof of this lemma is given in [166]. This lemma shows that the intermediate
part of the operator F ′ (u)v contains the leading term ψ0 v, and all other terms
contain ψ⃗n with |⃗n| > ρ (which are small if ψ is smooth enough). Thus, the spatial
 ATTRACTORS. THEN AND NOW 157

averaging condition is satisfied for a(u) = ψ0 = ⟨ψu ⟩, and this justifies the inter-
pretation of a(u) as a spatial averaging in the general theory (see [166], [243], and
the references therein for more details).
This scheme was proposed by Sell and Mallet-Paret [166] to construct an inertial
manifold for a scalar reaction-diffusion equation (7.12) in 3D with periodic boundary
conditions. Clearly, the spectral gap conditions are not satisfied there, and this was
historically the first example of an inertial manifold beyond the spectral gap con-
ditions. They also gave a counterexample showing that in 4D this method will not
work and the inertial manifold may not exist (at least in the class of normally hyper-
bolic inertial manifolds; see [167]). The corresponding counterexample for a system
of two reaction-diffusion equations in 3D (also with periodic boundary conditions)
where an inertial manifold does not exist (again, in the class of normally-hyperbolic
inertial manifolds) was given by Romanov [197]. The inertial manifolds for the
Cahn–Hilliard equations in 3D periodic domains were constructed in [131] by means
of a similar method.
As we have already mentioned, the spatial averaging method does not work
in general for systems of PDEs. Indeed, in this case a(u) will not be a scalar
operator, but a matrix operator, and this is not enough to solve the simplified
equation of variations for the intermediate modes. One exception is the case
where a(u) is a zero matrix: then everything should work similarly to the scalar
case. However, there was a non-trivial problem here related to a cut-off of the initial
equation transforming it into the so-called “prepared” form. Usually, the spatial
averaging method requires a very non-trivial cut-off procedure, and it was not
clear how to preserve the condition ⟨a(u)⟩ = 0 under this procedure. The progress
here is mainly related to Kostianko’s paper [126], where such a cut-off procedure
was proposed, which uses an appropriate modification of the nonlinearity F (u)
in the Fourier basis. In turn, this construction allowed us to use spatial averaging
method for inertial manifolds for some hydrodynamical problems. For instance,
the existence of an inertial manifold for the modified Leray-α model and hyperviscous
Navier–Stokes equations on a 3D torus was proved in [126] and [82], respectively,
and more general problems of the form
(
∂t u + (u, ∇x u) + (−∆x )1+γ u + ∇x p = g, u t=0 = u0 ,
(7.45)
div u = 0, u = (1 − α∆x )−γ u,

where γ, γ > 0 and γ + γ = 1/2, were considered in [30].

Let us now return to the case of systems where a(u) is not a scalar matrix and
consider the following version of a cross-diffusion system:

∂t u − (1 + iω)(∆x u − u) = f (u, u), u t=0

= u0 , (7.46)

where u(t, x) := Re u(t, x) + i Im u(t, x) is a complex-valued unknown function, u

is the complex conjugate, f is a given smooth function with finite support, and
ω ̸= 0 is a real cross-diffusion parameter. We consider this equation on the 3D
torus Ω = [−π, π]3 . The typical example of such an equation is the complex
Ginzburg–Landau equation

∂t u − (1 + iω)(∆x u − u) = ϕ(|u|2 )u. (7.47)

158 S. V. ZELIK

As usual, we assume that this equation possesses a smooth absorbing set, so without
loss of generality we may assume that ϕ is smooth and has a finite support.
Applying the method of spatial averaging to the equation of variations related to
problem (7.46), we end up with the following problem for the intermediate modes:

d
vn + (1 + iω)(λn + 1)vn = ⟨fu′ ⟩vn + ⟨fu′ ⟩v n + “small”. (7.48)
dt
We see that the first term on the right-hand side is a nice scalar operator, but
the second (⟨fu′ ⟩v n ) is not, so the sole spatial averaging is not sufficient for
constructing an inertial manifold (the corresponding example with ω = 0 was given
in [197]). The key idea here is (following [127]) to combine the spatial averaging
method with temporal averaging with respect to the rapid in time oscillations
generated by the large dispersion term iω(λn + 1)vn in equations (7.48). This
averaging kills (makes small) the non-scalar term ⟨fu′ ⟩v n in (7.48) and allows us to
complete the proof of the existence of an inertial manifold for equation (7.46) in
the case where ω ̸= 0 (see [127] and [130] for the missing details).

8. Exponential attractors and perturbation theory

We have seen that inertial manifolds gives us a perfect way to constructing

a finite-dimensional reduction for the limit dissipative dynamics. However, an
inertial manifold requires much more restrictive conditions to exist than an attractor.
In this section we discuss an intermediate concept between attractors and inertial
manifolds which was introduced in [63], namely, the concept of an exponential
attractor, which, on the one hand, is almost as common as usual attractors and,
on the other hand, allows us to overcome the most principal drawbacks of usual
(global) attractors. We start with an illustrative example.
Example 8.1. Consider the following 1D real Ginzburg–Landau equation with
non-zero boundary conditions:

∂t u = ν 2 ∂x2 u − u3 + u, u x=1
=u x=−1
= 1, u t=0
= u0 , (8.1)

where Ω = (−1, 1) and ν > 0 is a small parameter. Obviously, the solution semigroup
associated with this equation is dissipative and possesses a smooth absorbing set.
We claim that the attractor A of the corresponding solutions semigroup S(t)
in the phase space Φ := L2 (Ω) endowed with the standard bornology of bounded
sets (the bornology of all subsets of Φ is also possible here and gives rise to the same
attractor) consists of a single equilibrium:

A = {u ≡ 1}. (8.2)

Indeed, u = 1 is the unique equilibrium of this problem, no matter how small ν is.
This follows from multiplying the equation for the equilibria by u′x and integrating
with respect to x:
1
2
ν 2 u′x (x)2 = ν 2 u′x (−1)2 + u(x)2 − 1 .
2
 ATTRACTORS. THEN AND NOW 159

Thus, on the one hand, u′ (x)2 is non-decreasing and, on the other hand, u′ (1)2 =
u′ (−1)2 which is possible only if u(x) ≡ 1. Another observation is that equation (8.1)
possesses a global Lyapunov function:
 
d 1
ν 2 ∥u′x ∥2Φ − ((u2 − 1)2 , 1) = −2∥∂t u∥2Φ ,
dt 2

and therefore any trajectory must converge to the set of equilibria and (8.2) holds.
Thus, the limit dynamics is trivial for any ν > 0. It is just a single exponentially
stable equilibrium. Moreover, linearizing our equation near u = 1 we see that
the rate of attraction to the equilibrium is not slower than e−2t (and does not
become worse as ν → 0). So we may conclude that, for any (bounded) set B ⊂ Φ,

distΦ (S(t)B, A) 6 Ce−2t , (8.3)

where the constant C is actually independent of B, but may still depend on ν.

However, for small ν > 0, equation (8.1) possesses an interesting metastable
dynamics related to the slow evolution of multi-kinks. For instance, if you start
from a profile that is close to
 
1
u0 (x) = 1 − 2H − |x| ,
3

where H(x) is the standard Heaviside function, then the corresponding solution
is close to u0 (x) for an extremely long period of time: the life-span of this
“almost” equilibrium T is asymptotically e−1/ν (see [61], [245] for more details
about the evolution of multi-kinks via the centre manifold reduction). Thus, for
the constant C in (8.3), we have an estimate
c/ν
C > ee (8.4)

with some positive c which is independent of ν.

Thus, the behaviour of trajectories of (8.1) for small ν is described mainly
by the evolution of metastable states, and the attractor A becomes unobservable
in experiment and has only a limited theoretical interest.
Remark 8.2. The situation described above, where the limiting behaviour is
trivial, but there is a rich and interesting metastable dynamics, is somehow typical
for dissipative systems and is observed in many other situations, for instance
in the 1D Burgers equation (7.35) (see [17], [141], and the references therein).
A similar situation also arises for the stochastic equation (5.21) for small values
of ε > 0, where the corresponding random attractor is trivial (we even have the
uniform estimate (5.26) for the Lyapunov exponent), but it takes a very long time
(depending on ε in a way similar to (8.4)) to reach this attractor. To avoid such
problems, it is crucial to control the rate of attraction to the attractor under
consideration in terms of physical parameters. Unfortunately, there are no ways to
get this control on the level of usual (global) attractors in a more or less general
situation. On the other hand, we often have such a control for the rate of attraction
to an inertial manifold, which allows us to capture the metastable dynamics as
160 S. V. ZELIK

well. Capturing this dynamics in more general cases, where an inertial manifold
may not exist, is one of the main sources of motivation for the theory of exponential
attractors. Actually, in the examples discussed above, we have an exponential
rate of attraction to the corresponding global attractors, so the most important
advantage of the theory of exponential attractors (at least in these cases) is exactly
the ability to get a reasonable control (not like (8.4)) for the exponential rate of
attraction in terms of physical parameters.
The key idea of an exponential attractor is to add the metastable states to
the global attractor so that, on the one hand, to get an effective control of the rate
of attraction and, on the other hand, to preserve finite-dimensionality (in terms
of the fractal dimension), which allows us to use Mané’s projection theorem.
Definition 8.3. Let S(t) : Φ → Φ be a dynamical system acting on a metric
space Φ with a fixed bornology B. A set M is an exponential attractor for
the dynamical system S(t) if M satisfies the following conditions:
1) it is compact in Φ and has a finite fractal dimension:

dimf (M, Φ) < ∞; (8.5)

2) it is semi-invariant, that is, S(t)M ⊂ M for t > 0;

3) it attracts exponentially the images of B ∈ B:

distΦ (S(t)B, M) 6 QB e−αt (8.6)

for some positive constants α and QB .

We recall some basic facts of perturbation theory for global attractors before
discussing the constructions of exponential attractors. We start with upper
semicontinuity, which is based on the following fact from general topology.
Proposition 8.4. Let Φ and A be two Hausdorff topological spaces, and let A be
a compact set in Φ × A. Also let Π1 , Π2 be the projectors of A onto the first and
second component of the Cartesian product, and let Aα := Π1 Π−1 2 (α), α ∈ A. Then
the family of sets {Aα }α∈A is upper semicontinuous at every α0 ∈ A, namely, for
every neighbourhood O(Aα0 ) of the set Aα0 in Φ there exists a neighbourhood O(α0 )
in A such that
Aα ⊂ O(Aα0 ) ∀ α ∈ O(α0 ). (8.7)
The proof of this proposition can be found, for example, in [12] for the case where Φ
and A are metric spaces. The general case can be dealt with analogously. Note also
that in the metric case upper semicontinuity can be rewritten in an equivalent way
using the Hausdorff semi-distance:

lim distΦ (Aα , Aα0 ) = 0. (8.8)

α→α0
S
In applications, we usually take A := Aα × {α}, where the Aα are the attractors
α∈A
of dynamical systems Sα (t) : Φ → Φ depending on the parameter α ∈ A. Then,
to verify upper semicontinuity, we just need to consider an arbitrary sequence
 ATTRACTORS. THEN AND NOW 161

u0n ∈ Aαn , where αn → α0 , and extract from it a subsequence u0nk converging

to some u0α0 ∈ Aα0 . In turn, in order to check this property, we may use
the representation formula for Aαn , so we only need to consider a sequence
of complete bounded solutions uαn (t) ∈ Kαn and extract a subsequence uαn (t)
converging to uα0 (t) ∈ Kα0 . This is usually true under the minimal assumptions
on the dynamical system Sα (t). We emphasize that this method does not require
one to verify the closeness of individual semi-trajectories of the perturbed and
non-perturbed dynamical systems (which is often a much more difficult task,
especially in the case of singular perturbations) and is applicable, for example, to
trajectory attractors (see [12], [38], [179], and the references therein).
In contrast to upper semicontinuity, the lower semicontinuity of attractors Aα ,
α ∈ A, can easily be broken (see Example 2.5). Recall that in the metrizable case,
the lower semicontinuity of the family {Aα }α∈A reads

lim distΦ (Aα0 , Aα ) = 0, α0 ∈ A. (8.9)

α→α0

The continuity of the family of attractors {Aα }α∈A means that (8.8) and (8.9) hold
simultaneously, that is,

lim distsym
Φ (Aα , Aα0 ) = 0, α0 ∈ A, (8.10)
α→α0

where distsym
Φ (U, V ) := max{distΦ (U, V ), distΦ (V, U )} is the symmetric Hausdorff
distance between sets. The next standard proposition shows that the continuity
of attractors is determined by the rate of attraction.
Proposition 8.5. Let Φ and A be subsets of normed spaces, and let Sα (t) : Φ → Φ
be a family of dynamical systems on Φ satisfying

∥Sα1 (t)u0 − Sα2 (t)u0 ∥Φ 6 C∥α1 − α2 ∥A eKt (8.11)

for some positive constants C and K which are independent of αi ∈ A and u0 ∈ Φ.

Assume also that A is compact, the Sα (t) are continuous for every fixed t and α,
and, for every α ∈ A, the dynamical system Sα (t) possesses an attractor Aα with
respect to the bornology of all subsets of Φ. Then
1) the family {Aα }α∈A is uniformly continuous with respect to α ∈ A (that
is, (8.10) holds uniformly with respect to α0 ) if and only if

lim sup distΦ (Sα (t)Φ, Aα ) = 0; (8.12)

t→∞ α∈A

2) if, in addition, the following stronger version of (8.12) (exponential attraction)

holds:
distΦ (Sα (t)Φ, Aα ) 6 Qe−λt , (8.13)
where the positive constants Q and λ are independent of t and α, then the family
of attractors Aα is uniformly Hölder continuous:
λ/(K+λ)
distsym
Φ (Aα1 , Aα2 ) 6 (Q + C)∥α1 − α2 ∥A . (8.14)
162 S. V. ZELIK

Proof. We consider here only the second part of the statement; the first part
is more straightforward (see, for example, [12], [93], [101], [154]). Indeed, let
uα1 ∈ Aα1 . Then, by the invariance of the attractor, for any T > 0 there exists
vα1 ∈ Aα1 such that Sα1 (T )vα1 = uα1 . Consider the point uα1 := Sα2 (T )vα1 .
Then, due to (8.13), we have

distΦ (uα1 , Aα2 ) 6 Q e−λT

and, due to (8.11), we know that

∥uα1 − uα1 ∥Φ 6 C∥α1 − α2 ∥A eKT .

Therefore, using that uα1 ∈ Aα1 is arbitrary, we have

distΦ (Aα1 , Aα2 ) 6 Qe−λT + C∥α1 − α2 ∥A eKT .

1 1
Finally, fixing T = ln , we arrive at
λ + K ∥α1 − α2 ∥A

λ
distΦ (Aα1 , Aα2 ) 6 (Q + C)∥α1 − α2 ∥κA , κ := .
K +λ
Swapping α1 and α2 , we get the desired estimate (8.14).
Remark 8.6. Assumption (8.11) may look too restrictive since it requires
with respect to all u0 ∈ Φ, but it is actually only necessary for
uniformity S
u0 ∈ A := Aα . The uniform attraction property is also necessary for the set A
α∈A
only, so the assumption that the bornology on Φ consists of all subsets of Φ is
not restrictive at all. We may also formulate a natural non-uniform analogue
of Proposition 8.5 for the upper and lower semicontinuity of attractors Aα at
a single point α = α0 , as well as its analogues for the non-autonomous case (see [93]
for more details).
Crucial for us is that the continuity of attractors Aα is determined by the rate
of attraction, and even qualitative bounds for the closeness of the perturbed and
non-perturbed attractors are automatically obtained if the rate of attraction is
under control. Unfortunately, the problem of getting this control looks unsolvable
on the level of usual (global) attractors in a more or less general situation.
This makes attractors somehow unobservable, namely, no matter how long you
observe the system and how precise your measurements/simulations are, you cannot
guarantee that the reconstructed approximate attractor is close to the precise one,
and this is an extra source of motivation to consider exponential attractors. Note
also that the Hölder continuity (8.14) is the best what we can expect and it cannot
be improved to Lipschitz continuity even when ideal objects like Morse–Smale or
uniformly hyperbolic attractors are considered (see [116] and the references therein).
Remark 8.7. We also mention an interesting result, which shows that the lower
continuity of attractors Aα holds for “typical” values of the parameter α ∈ A
(see [101] and [233]). This result tells us that, given a family of dynamical
systems Sα (t) (acting on a separable, complete, and bounded metric space Φ) which
 ATTRACTORS. THEN AND NOW 163

depends continuously on both t and α ∈ A, where A is a compact metric space,

the corresponding attractors Aα (if exist) are upper and lower semicontinuous for
every α belonging to a residual subset of A (that is, for every α ∈ A away from
a countable union of nowhere dense sets). Moreover, given a Borel probability
measure µ on A, for every ε > 0 there exists a closed set Aε satisfying µ(A \ Aε ) 6 ε,
such that the family Aα of attractors is uniformly continuous on Aε in the sense
of the symmetric Hausdorff distance. Due to the first part of Proposition 8.5, this,
in turn, gives us a uniform rate of attraction to the attractors Aα with respect to
α ∈ Aε .
Indeed, consider the space BΦ of all closed non-empty subsets of Φ, with
the symmetric Hausdorff distance as a metric. Then it known that BΦ is a complete
separable metric space and the continuity of the family Aα can be interpreted as
the continuity of the map f : A → BΦ defined by f (α) := Aα . The idea of the proof
consists in approximating the function f (α) by fn (α) := Sα (n)Φ. It is not difficult
to see that the continuity assumptions imposed on Sα (t) imply that the functions fn
are continuous and the existence of the attractors Aα ensures that fn (α) → f (α)
pointwise. Thus, by Baire’s theorem, the set of discontinuities of f (α) is a countable
union of nowhere dense sets. The second statement is an immediate corollary
of Egorov’s theorem.
It is worth mentioning that these results cannot replace exponential attractors
since they neither give any reasonable way to find the set Aε explicitly or compute
it, nor any way to control the uniform rate of attraction to Aα for α ∈ Aε in terms
of the physical parameters.
The result of Proposition 8.5 can partially be extended to exponential attractors.
Proposition 8.8. Let the assumptions of Proposition 8.5 hold, and let the dy-
namical system Sα (t) possess exponential attractors Mα , α ∈ A which satisfy
the attraction property (8.13) uniformly with respect to α ∈ A. Then the following
estimate holds:
λ/(K+λ)
max{distΦ (Aα1 , Mα2 ), distΦ (Aα2 , Mα1 )} 6 (Q + C)∥α1 − α2 ∥A . (8.15)
The proof of this fact repeats word by word the arguments given in the proof
of Proposition 8.5, and for this reason it is omitted.
Remark 8.9. The robustness of exponential attractors with respect to pertur-
bations was stated in the original work [63] exactly in the form (8.15). The reason
why we have to replace Mα by Aα in (8.15) is that an exponential attractor is
only semi-invariant, so starting from uα ∈ Mα , we may fail to find vα in Mα (or
in a set satisfying the uniform attraction property) such that Sα (T )vα = uα . In
fact, we can do this if uα ∈ Sα (t)Mα for t large enough, so the following shifted
version of Hölder continuity holds:
max{distΦ (Sα1 (t)Mα1 , Mα2 ), distΦ (Sα2 (t)Aα2 , Mα1 )}
λ/(K+λ)
6 (Q + C)∥α1 − α2 ∥A , t > T, (8.16)
1 1
where T = T (α1 , α2 ) := ln (see [63]). It is remarkable that this
λ + K ∥α1 − α2 ∥A
shifted Hölder continuity holds for any choice of exponential attractors Mα with
164 S. V. ZELIK

the uniform attraction property. In contrast to this, if we want to have a full

analogue of estimate (8.14) for exponential attractors, we need to construct them
in a special way, taking care of the closeness of the sets Mα1 \ Sα1 (t)Mα1 and
Mα2 \ Sα2 (t)Mα2 for t 6 T (see [71], [179], and Theorem 8.14 below).
We state below the transitivity of exponential attraction, which is one of the key
tools in the theory of exponential attractors.
Proposition 8.10. Let Φ be a metric space and S(t) be a dynamical system on
it which is Lipschitz continuous:

d(S(t)u1 , S(t)u2 ) 6 CeKt d(u1 , u2 ), u1 , u2 ∈ Φ, t > 0, (8.17)

where the constants C and K are independent of t and ui ∈ Φ. Assume also that
there are three subsets Mi ⊂ Φ, i = 1, 2, 3, such that

distΦ (S(t)M1 , M2 ) 6 C1 e−λ1 t , distΦ (S(t)M2 , M3 ) 6 C2 e−λ2 t .

Then
′
distΦ (S(t)M1 , M3 ) 6 C ′ e−λ t , (8.18)
′ ′
where C = CC1 + C2 and λ = λ1 λ2 /(K + λ1 + λ2 ).
The proof of estimate (8.18) is based on arguments which are similar to
the derivation of (8.14) and can be found in [72].
8.1. Exponential attractors via squeezing properties. We now turn to
theorems which guarantee the existence of exponential attractors with nice prop-
erties. These theorems are based on various forms of the squeezing property and
are very close to the ones presented in subsection 6.2. Namely, the iterative
ε-nets constructed in order to estimate the fractal dimension of an attractor are
exactly the “metastable” states which we need to add to the attractor in order
to get exponential attraction. We demonstrate the main idea using an analogue
of Theorem 6.4 that was proved in [65] (see also [71] and [179] for more details).
Theorem 8.11. Let Φ and Φ1 be two Banach spaces and let the embedding
Φ1 ⊂ Φ be compact. Assume also that we are given a bounded subset B of Φ1 and
a map S : B → B satisfying the squeezing property

∥S(u1 ) − S(u2 )∥Φ1 6 L∥u1 − u2 ∥Φ , u1 , u2 ∈ B. (8.19)

Then the discrete dynamical system S(n) := S n , n ∈ N, possesses an exponential

attractor M with the following properties:
1) M is compact in Φ1 and is semi-invariant: SM ⊂ M;
2) its fractal dimension in Φ1 is finite and satisfies the estimate

dimf (M, Φ1 ) 6 H1/(4L) (Φ1 ,→ Φ); (8.20)

3) M attracts the set B exponentially and

distΦ1 (S(n)B, M) 6 R0 2−n , n ∈ N, (8.21)

where R0 is such that B ⊂ BR0 (0, Φ).

 ATTRACTORS. THEN AND NOW 165

Sketch of the proof. Indeed, arguing as in the proof of Theorem 6.4, for
every n ∈ N we construct an R0 2−n -net in Vn ⊂ S(n)B such that Vn ⊂ S(n)B
and #Vn 6 N n where N is the same as in the proof of Theorem 6.4. To preserve
semi-invariance we set E1 = V1 and En = Vn ∪ S(En−1 ) for n = 2, 3, . . . . Then
En ⊂ S(n)B, #En 6 CN n+1 , and S(En ) ⊂ En+1 . Let us finally define
[
M′ = En , M := [M′ ]Φ1 . (8.22)
n∈N

Then semi-invariance, as well as the exponential attraction (8.21), hold immediately

and we only need to check estimate (8.20) for the fractal dimension. Let εn = R0 2−n .
Then all the sets Ek for k > n are subsets of S(n)B, and therefore the εn -balls
Sn
centred at points of Vn cover them. So the set Ek is an εn -net of M, and
k=1
therefore
Pn
Hεn (M, Φ1 ) log2 C + log2 ( k=1 #En )
lim sup 6 lim
n→∞ log2 (1/εn ) n→∞ n − log2 R0
2 log2 C + (n + 2) log2 N
6 lim = log2 N,
n→∞ n − log2 R0

which gives the desired estimate (8.20) and finishes the proof of the theorem.
Now assume that we are given a continuous dynamical system S(t) in Φ such that
S = S(1) satisfies all the assumptions of Theorem 8.11. Then, first, we can construct
a discrete exponential attractor Md (that is, an attractor for the semigroup S(n),
n ∈ N) and after that can extend it to a continuous one via
 [ 
M := S(t)Md . (8.23)
t∈[1,2] Φ

Indeed, all the properties of an exponential attractor, except of the control of fractal
dimension, follows automatically from the analogous properties of Md , but to
control the fractal dimension in Φ, we need an extra Hölder continuity assumption:

α
∥S(t1 )u1 − S(t2 )u2 ∥Φ 6 C |t1 − t2 | + ∥u1 − u2 ∥Φ , t1 , t2 ∈ [1, 2], u1 , u2 ∈ B,
(8.24)
for some positive C and α. Then a transition from a discrete to a continuous
exponential attractor can increase the fractal dimension in Φ by the additive
factor α−1 at most. Moreover, using that S(1)M is also an exponential attractor
if M is, we may also get the finiteness of the fractal dimension in Φ1 (due to (8.19),
the fractal dimension of S(t)M in Φ1 is controlled by the fractal dimension of‘M
in Φ).
Remark 8.12. In applications B is usually a compact absorbing or an expo-
nentially attracting set of the semigroup S(t). Then, due to the transitivity
of exponential attraction, we conclude that the attractor M constructed attracts
not only the set B, but all bounded sets in the phase space Φ. Note also that
for semigroups S(t) generated by PDEs, the uniform Hölder continuity in time
166 S. V. ZELIK

stated in (8.24) typically holds only if B is more smooth than the initial phase
space Φ. This is not a problem when parabolic PDEs are considered since, due to
the instantaneous smoothing property, we may find a compact (and more regular)
absorbing set B. However, in more general cases (for example, for damped wave
equations), we only have an asymptotic smoothing property, so B must be a compact
exponentially attracting set and the use of the transitivity of exponential attraction
becomes unavoidable (see [72], [179], and the references therein).
At the next step, following [71] we consider an analogue of Theorem 6.8,
which also includes the Hölder continuity of exponential attractors with respect to
perturbations. To this end, we need the following definition.
Definition 8.13. Let Φ1 ⊂ Φ be two Banach spaces such that the embedding
is compact, and let a bounded subset B of Φ1 , ε > 0, κ ∈ [0, 1), and L > 0 be
given. Then a map S : Oε (B) → B belongs to the class Sε,κ,L (B) if the squeezing
property (6.10) is satisfied for all u1 , u2 ∈ Oε (B). The distance between two maps
S1 , S2 ∈ Sε,κ,δ (B) is defined as follows:

∥S1 − S2 ∥S := sup ∥S1 (u) − S2 (u)∥Φ1 .

u∈Oε (B)

Theorem 8.14. Let S ∈ Sε,δ,L (B), and let S(n) = S n be a discrete dynamical
system on B associated with this map. Then this semigroup possesses an exponential
attractor MS ⊂ B, which has the following properties:
1) MS is a compact semi-invariant set in B whose fractal dimension satisfies
an analogue of estimate (6.11);
2) the following exponential attraction property holds:
 n
1+κ
distΦ1 (S(n)B, MS ) 6 ε ; (8.25)
2

3) if S1 , S2 ∈ Sε,κ,L (B), then

distsym θ
Φ1 (MS1 , MS2 ) 6 C∥S1 − S1 ∥S (8.26)

for some positive C and θ depending only on ε, κ, L, and the spaces Φ and Φ1 .
 n
κ+1
Idea of the proof. The attractors MS can be constructed using the ε -
2
nets En = En (S) similarly to Theorem 8.11 (see also Theorem 6.8), so we only need
to explain how to obtain Hölder continuity (8.26). We actually need to estimate
the distance between En (S1 ) and MS2 for all n ∈ N. For large values of n we
use (8.16) and get the desired estimate without extra care. In contrast to this,
for relatively small values of n we need the extra assumption that the sets En are
constructed in such a way that

distsym n
Φ1 (En (S1 ), En (S2 )) 6 K ∥S1 − S2 ∥S (8.27)

for some K which is independent of n, S1 , and S2 . The details can be found in [71].
 ATTRACTORS. THEN AND NOW 167

Remark 8.15. Recall that, similarly to inertial manifolds, exponential attractors

are not unique, so the problem of choosing an “optimal” exponential attractor
becomes crucial for both theory and applications. The theorem stated above gives
us a single-valued Hölder continuous branch of the function S → MS for wide class
of nonlinear maps S ∈ Sε,κ,L (B).
We now turn to non-autonomous dynamical systems and start with the uniform
(deterministic) case. We say that a discrete cocycle Sξ (n) : Oε (B) → B, n ∈ N, over
a dynamical system T (n) : Ψ → Ψ belongs to the class Sε,κ,L (B) if Sξ (1) ∈ Sε,κ,L (B)
for all ξ ∈ Ψ. Then the non-autonomous analogue of Theorem 8.14 reads as follows.
Theorem 8.16. Let the cocycle Sξ (n) belong to the class Sε,κ,L (B), and let
Uξ (m, n), m > n, ξ ∈ Ψ, be the corresponding dynamical processes on B. Then there
exists a family MS (ξ), ξ ∈ Ψ, of compact sets in B (a non-autonomous exponential
attractor) which has the following properties.
1) Their fractal dimensions are finite and uniformly bounded:

dimf (MS (ξ), Φ1 ) 6 C, ξ ∈ Ψ.

2) They are semi-invariant: Sξ (1)MS (ξ) ⊂ MS (T (1)ξ), ξ ∈ Ψ, and the following

uniform exponential attraction property holds:

distΦ1 (Sξ (n)B, MS (T (n)ξ)) 6 Qe−αn , ξ ∈ Ψ, n ∈ N, (8.28)

for some positive constants Q and α depending only on Φ, Φ1 , B, ε, κ, and L.

3) For any two cocycles Sξ (n) and Ŝξ (n) over the same dynamical system
T (n) : Ψ → Ψ belonging to the class Sε,κ,L (B), the following uniform Hölder
continuity holds:

distsym
Φ1 (MS (ξ), MŜ (ξ)) 6 C sup{e
−βn
∥ST (−n)ξ (1) − ŜT (−n)ξ (1)∥θS }, (8.29)
n∈N

where the positive constants C , β , and θ depend only on Φ, Φ1 , B, ε, κ, and L.

The proof of this result is almost identical to the proof of Theorem 8.14 and is
given in [71].
Remark 8.17. As usual, the passage from discrete to continuous time requires
some uniform Hölder regularity with respect to time for the cocycle under con-
sideration (an analogue of estimate (8.24)). In this case the desired exponential
attractor MS (ξ) can be defined via
 [ 
MS (ξ) := Sξ (t)MdS (T (−t)ξ) . (8.30)
t∈[1,2] Φ1

This construction extends all properties of the discrete exponential attractor stated
in Theorem 8.16 to the case of continuous time (see [71] for the details). In particular,
if the cocycle Sξ (t) is autonomous, periodic, or almost-periodic in time, then the same
is true for the non-autonomous exponential attractors constructed. It also can be
checked that, under some further natural assumptions, the function t → MS (T (t)ξ)
is Hölder continuous in time.
168 S. V. ZELIK

As mentioned already, the assumption of Hölder continuity in time may be

rather restrictive, since it usually requires extra regularity in space (although
in most applications it is not a big problem due to the transitivity of exponential
attraction), so it would be interesting to relax it. The attempts to do so are related
to the use of a more straightforward and naive (than (8.30)) extension, namely,
the continuous attractor is defined via MS (t) := MS (T (t)ξ) for t = n ∈ Z and,
for t = n + s, where n ∈ Z and s ∈ [0, 1), we set MS (t) := ST (n)ξ (s)MdS (n).
In this case, we indeed have semi-invariance, exponential attraction, and uniform
bounds for the fractal dimension without Hölder continuity in time, but the object
obtained will no longer be compatible with the autonomous case (where it will give
an artificial time-periodic attractor), it will not be continuous in time (artificial
jumps at integer points), and so on. Since such an object can hardly be considered
as a satisfactory version of a non-autonomous exponential attractor, the problem
of removing/relaxing Hölder continuity in time remains open.
Remark 8.18. We emphasize that the non-autonomous exponential attrac-
tor MS (ξ), ξ ∈ Ψ, constructed in Theorem 8.16 is not just a pullback attractor: it
also attracts forward in time at a uniform exponential rate. This demonstrates one
of the main advantages of exponential attractors for non-autonomous equations,
namely, they allow us to settle the problem with forward attraction, which looks
unsolvable on the level of pullback attractors (see Example 5.10) and, in contrast
to uniform attractors, the object constructed remains finite-dimensional. For this
reason, the name of “pullback” exponential attractors, which is used by some authors
(see, for example, [28]), looks confusing for us, and we prefer to refer to them as
non-autonomous exponential attractors. Also, in contrast to uniform attractors,
such exponential attractors do not violate the causality principle. Indeed, esti-
mate (8.29) shows us that the exponential attractor MS (ξ) depends on ST (−n)ξ (1),
n ∈ N, only and does not depend on the future (ST (n)ξ (1) for n > 0). Moreover,
the impact of the past on the present attractor MS (ξ) decays exponentially rapidly
with respect to the time passed (in full accordance with our intuition). Note that
this property is also violated on the level of pullback attractors.
Uniform analogues of exponential attractors which are independent of time and
where finite-dimensionality is replaced by an appropriate estimate for the Kol-
mogorov ε-entropy also appear in the literature (see [66], [234], and the references
therein). Such constructions are usually based on straightforward generalizations
of Theorem 6.17, so we give no further details here. An alternative possibil-
ity, where infinite-dimensional exponential attractors may appear, is the theory
of dissipative PDEs in unbounded domains, where the global attractor is usu-
ally infinite-dimensional and we must use Kolmogorov entropy to control the size
of the attractor (see [67], [179], and the references therein).
Now we turn to the non-uniform (random) case where we have an ergodic
measure µ for the underlying dynamical system T (t) : Ψ → Ψ and where the set
B = B(ξ) ∈ Φ1 , as well as the squeezing factor L = L(ξ) in (6.14), are random
variables. Namely, similarly to Theorem 6.10 we say that a discrete measurable
cocycle Sξ (n) belongs to the class Sε,κ,L (B) if Sξ (1) : Oε (B(ξ)) → B(T (1)ξ), ξ ∈ Ψ,
and Sξ (1) satisfies the squeezing property (6.14) for all u1 , u2 ∈ Oε (B(ξ)). Then
the following result holds.
 ATTRACTORS. THEN AND NOW 169

Theorem 8.19. Let T (n) : Ψ → Ψ, n ∈ Z, be a dynamical system on a Polish

space Ψ which possesses a Borel ergodic probability measure µ. Let the cocycle Sξ (n)
belong to the class Sε,κ,L (B) for some deterministic constants ε > 0, κ ∈ [0, 1),
and a random constant L = L(ξ), let B(ξ) be a bounded random set in Φ1 , and
let the function ξ → ∥B(ξ)∥Φ1 be tempered. Also assume that (6.16) is satisfied
and E(Lθ ) < ∞. Then there exists a compact random set MS (ξ) ⊂ B(ξ) satisfying
the following conditions.
1) The sets MS (ξ) are semi-invariant: Sξ (n)MS (ξ) ⊂ MS (T (n)ξ) for all n ∈ N
and almost all ξ ∈ Ψ, and their fractal dimensions are bounded for almost all ξ ∈ Ψ
by a deterministic constant.
2) The uniform attraction property

distΦ1 Sξ (n)B(ξ), MS (T (n)ξ) 6 Ce−βn ,

n ∈ N, (8.31)

holds for almost every ξ ∈ Ψ for some deterministic positive constants C and β
which are independent of ξ and n.
3) Let Ŝξ (n) be another cocycle belonging to the class Sε,κ,L (B) such that

∥Sξ (1) − Ŝξ (1)∥S 6 K(ξ)δ (8.32)

for some deterministic δ and random K(ξ) such that E(K θ ) < ∞. Then there exist
a random variable P (ξ) which is finite almost everywhere and is independent of δ
and the concrete choice of Sξ and Ŝξ , and a deterministic constant γ > 0 such that

distsym γ
Φ1 (MS (ξ), MŜ (ξ)) 6 P (ξ)δ . (8.33)

The proof of this theorem in the particular case κ = 0 was given in [211]. The
general case is completely analogous and so is omitted.

Remark 8.20. We note that, similarly to the deterministic case, the random
attractor obtained is not just a pullback exponential attractor, but also a forward
attractor and the exponential rate of attraction is actually uniform. Moreover,
we have a deterministic rate of attraction to it, which can be controlled in terms
of the physical parameters of the system under consideration (as pointed out
in the original paper [211], where exponential attractors for a random dynamical
system were introduced). Some attempts to relax this uniform attraction rate and
allow the constants C and β in (8.31) to be random were also made subsequently
(see [247] and the references therein), but this did not lead to any essential
simplifications of the assumptions on the cocycle Sξ (n) which could compensate for
the drawbacks related to the loss of the uniform/deterministic attraction property.
For instance, the random attractor A(ξ) constructed in Example 5.19 becomes
“pullback exponential” if we allow the constant C = C(ξ) in the exponential
attraction property to be random, and then the control over the rate of attraction
and Hölder continuity with respect to ε → 0 are lost.
As in the deterministic case, passing from discrete to continuous exponential
attractors requires some Hölder continuity in time and a random version of the
transitivity of exponential attraction (see [211]).
170 S. V. ZELIK

We finally mention that, in the estimate for Hölder continuity (8.33), we

may only guarantee that the random variable P (ξ) is finite almost everywhere.
In order to verify that it has finite moments of some order, we need to control
the rate of convergence in Birkhoff’s ergodic theorem. This problem is rather
delicate and requires not only the finiteness of the moments of any order for L(ξ)
and K(ξ), but also some kind of exponential mixing (in applications to stochastic
equations we need the corresponding Ornstein–Uhlenbeck process to be exponentially
mixing). We do not give any more details here and address the interested reader
to [83], [99], [113], [142], [188], [211], and to the references therein.
We conclude this subsection by a model example of a reaction-diffusion system
perturbed by white noise.
Example 8.21. Let Ω ⊂ Rd be a bounded domain in Rd with smooth boundary.
Consider the following reaction-diffusion system in Ω:

∂t u = a∆x u − f (u) + δ ∂t η(t), u = (u1 , . . . , un ), u t=0

= u0 , u ∂Ω
= 0. (8.34)

It is assumed that a is a constant diffusion matrix satisfying the condition a+a∗ > 0,
the nonlinearity f ∈ C 2 (Rn , Rn ) satisfies dissipativity conditions of the form

1) f (u).u > −C + c|u|p+1 , 2) f ′ (u) > −K,

3) |f ′ (u)| 6 C(1 + |u|p−1 )
(8.35)
for some positive constants C, c, and K, and an exponent p such that
d+2
06p6 for d > 2.
d−2
Let {λi }∞
i=1 be the eigenvalues of the minus Laplacian in Ω enumerated in the non-
decreasing order and {ei }∞i=1 be the corresponding eigenvectors. We assume that
the two-sided Wiener process η(t) in a Hilbert space Φ := L2 (Ω) is given by
∞
X
η(t) := βi ηi (t)ei , (8.36)
i=1

where {ηi (t)}∞

i=1 are the standard independent scalar Wiener processes and the de-
terministic vectors βi ∈ Rn satisfy
∞
X
λ3i |βi |2 < ∞.
i=1

Finally, δ ∈ [0, 1] is a fixed parameter. Note that δ = 0 corresponds to the standard

deterministic reaction-diffusion system and positive δ ≪ 1 gives its stochastic
perturbation.
Following the standard scheme (see, for example, [57], [142], and the references
therein), we associate with equation (8.34) a solution cocycle Sξδ (t) : Φ → Φ
over the standard dynamical system on the canonical probability space (Ψ, F, µ).
Namely, analogously to Example 5.19, we consider the space C0 (R) endowed with
the locally compact topology, the scalar Wiener measure µ, and the dynamical
 ATTRACTORS. THEN AND NOW 171

system (θ(h)ξ)(t) := ξ(t + h) − ξ(h). Then Ψ is the space [C0 (R)]N endowed with
the Tychonoff topology, while T (h) : Ψ → Ψ and µ are the Cartesian products of the
semigroups θ(h) and the measures µ, respectively. It is well known that the measure µ
is ergodic (see, for example, [83]). The solution cocycle Sξδ (t) : Φ → Φ is constructed
similarly to Example 5.19, by subtracting the corresponding Ornstein–Uhlenbeck
process (see [211] for more details).
Moreover, it is straightforward to verify (again, similarly to Example 5.19) that
this cocycle possesses a tempered absorbing set Bξ in Φ1 := H01 (Ω) with respect to
the bornology of tempered random sets. In addition, this absorbing set is uniform
with respect to the bornology of bounded deterministic subsets of Φ. Thus, it is
enough to construct the random exponential attractors Mδ (ξ) for the sets B(ξ)
only, so we may use the result of Theorem 8.19 to construct these attractors. We
also note that the trajectories of the Wiener process are Hölder continuous in time
for almost all ξ ∈ Ψ, so passing from discrete to continuous attractors does not
cause any problems, and we only need to check the assumptions of Theorem 8.19
which was actually done in [211]. Thus, we have the following result, which was
proved in [211].
Theorem 8.22. Let the assumptions stated above hold. Then the solution
cocycle Sξδ (t) : Φ → Φ, δ ∈ [0, 1], associated with problem (8.34) possesses a random
family Mδ (ξ), δ ∈ [0, 1], of exponential attractors which satisfy the following
conditions.
1) They are semi-invariant and have finite fractal dimensions in Φ1 . These
dimensions are bounded by a deterministic constant independent of δ for almost all ξ .
2) They possess a uniform exponential attraction property for deterministic
bounded sets of Φ, that is, there is a positive constant α and a monotone function Q
such that for every bounded subset B of Φ we have
distΦ Sξδ (t)B, Mδ (T (t)ξ) 6 Q(∥B∥Φ )e−αt , t > 0,

for almost all ξ .

3) There exist an almost everywhere finite random variable P (ξ) and a positive
deterministic constant γ which are independent of δ such that
distsym δ1 δ2 β
Φ (M (ξ), M (ξ)) 6 P (ξ)|δ1 − δ2 | . (8.37)
Moreover, in the case δ = 0 the corresponding exponential attractor M0 (ξ) is
independent of ξ and is the standard exponential attractor for the limit autonomous
and deterministic reaction-diffusion system.
Remark 8.23. We see that properly constructed exponential attractors remain
robust with respect to white noise perturbations as well. This demonstrates an
advantage of exponential attractors in comparison to global ones, where the limit
deterministic attractor is not robust with respect to noise even in the ideal situation
of regular attractors considered below: see Example 5.19 (see also [53] and [24]
for the discussion related to stochastic bifurcation theory). We also mention that
the conditions posed on the stochastic reaction-diffusion system (8.34) are far from
being optimal and can be relaxed essentially, which is however beyond the scope
of this survey.
172 S. V. ZELIK

8.2. Regular attractors. To conclude this section, we consider briefly the class
of dynamical systems whose global attractors have an exponential rate of attraction,
so that they may also be considered as exponential ones. This usually happens
when the dynamical system possesses a global Lyapunov function and all equilibria
are hyperbolic. Then any complete trajectory on the attractor is a heteroclinic
orbit between these equilibria, and the attractor is a finite collection of smooth
finite-dimensional unstable submanifolds of equilibria. Following [10], such attrac-
tors are called regular and this is probably the only more or less general class
of attractors where we are able to understand their structure; see [12], [26], [94], [95]
for more details. In our exposition we mainly follow [228].
We restrict ourselves to considering only the case of discrete time (passing from
discrete to continuous time is straightforward; see [228] for the details), so we
assume that we are given a map S : Φ → Φ in a Banach space Φ and construct
a discrete semigroup via S(n) := S n in Φ. We make the following assumptions
about the map S.
Assumption A. The map S belongs to C 1 (Φ, Φ), and its Fréchet derivative S ′ (u)
is uniformly continuous on bounded subsets of Φ; S is injective and ker S ′ (u) = {0}
for all u ∈ Φ.
Assumption B. The set of equilibria R0 of the map S is finite: R0 :=
{u1 , . . . , uN } and every equilibrium is hyperbolic, the latter means that the spectrum
of S ′ (ui ) does not intersect with the unit circle for any ui ∈ R0 .
Assumption C. The semigroup S(n) : Φ → Φ possesses a global attractor A
in Φ endowed with the standard bornology of bounded subsets of Φ.
Assumption D. Any trajectory u(n) = S(n)u0 stabilizes to one of the equilibria
in R0 , and there are no homoclinic structures, that is, if u1 , . . . , uk ∈ l∞ (Z) are
complete bounded trajectories of S(n) such that

lim ∥ui (n) − vi ∥Φ = 0, lim ∥ui (n) − vi+1 ∥Φ = 0, i = 1, . . . , k,

n→−∞ n→+∞

for some equilibria v1 , . . . , vk+1 ∈ R0 , then all the vi are necessarily different.
In applications Assumption D, the key one, usually follows from the existence
of a global Lyapunov function. We recall that a continuous function L : Φ → R is
called a global Lyapunov functional if the function n → L(S(n)u0 ) is non-increasing
along trajectories and the equality L(Su0 ) = L(u0 ) implies that u0 ∈ R0 . Together
with Assumption C and the finiteness of the set R0 , this implies not only the validity
of Assumption D, but also the fact that any complete bounded trajectory is
a heteroclinic orbit between two equilibria (see [12], [228] for more details). In turn,
this gives the following description of the attractor A:
N
[
A= M+ (ui ), (8.38)
i=1

where M+ (ui ) is an unstable set of the equilibrium ui , that is, the set of all
initial data u0 for which there exists a complete trajectory u(n) converging to ui as
n → −∞.
 ATTRACTORS. THEN AND NOW 173

We now use Assumption B on hyperbolicity to verify that the sets M+ (ui ) are
actually finite-dimensional manifolds in Φ with dimensions equal to the instability
index ind(ui ) which is the algebraic number of eigenvalues of S ′ (u0 ) lying outside
the unit circle (it is easy to verify that a;; these numbers are finite due to the existence
of a compact global attractor). The proof of this fact can be carried put similarly to
our verification of the existence of inertial manifolds, so we do not present it here.
Note only that first we construct the manifolds locally in a small neighbourhood
of the equilibria and then extend them to global submanifolds using the assumption
of injectivity (see Assumption A) and the absence of homoclinic structures. This
gives us the fact that all M+ (ui ) are finite-dimensional submanifolds of Φ which
are diffeomorphic to Rind(ui ) (see [228] for the details).
Finally, we recall that, similarly to inertial manifolds, the manifolds M+ (ui )
possess an exponential tracking property, that is, any trajectory of S(n) is attracted
exponentially fast to some trajectory on M+ (ui ) until it occurs in a small neigh-
bourhood of ui . Moreover, due to Assumption D, any trajectory of S(n) spends
a finite time Tδ outside the δ-neighbourhood of R0 , and Tδ is uniform with respect
to all trajectories starting from a bounded set. These two facts give us exponential
attraction to A with the rate of convergence controlled by the number of equilibria
and their hyperbolicity constants (see [228] for details). We summarize the results
obtained in the following theorem.

Theorem 8.24. Let the map S : Φ → Φ satisfy Assumptions A–D. Then the at-
tractor A of the associated semigroup S(n) : Φ → Φ has the description (8.38),
where M+ (ui ) are the ind(ui )-dimensional unstable submanifolds of the equilib-
ria ui . Moreover, any trajectory belonging to the attractor is a heteroclinic orbit
between two equilibria, and the rate of attraction to A is exponential, that is, there
exist a positive constant α and a monotone function Q such that, for every bounded
subset B ⊂ Φ,

distΦ (S(n)B, A) 6 Q(∥B∥Φ )e−αn , n ∈ N. (8.39)

Remark 8.25. Note that, despite the exponential attraction rate (8.39) for
a regular attractor, it still makes sense to construct an exponential attractor
even in the case where the global attractor is regular. The problem here is that
the constant α and the function Q are still not controllable in terms of the physical
parameters of the dynamical system under consideration and we still may lose
important intermediate dynamics (see Example 8.1). Note also that, although
Assumption B is generic in a certain sense due to Sard’s theorem, the number
of equilibria and their hyperbolicity constants can only be found/estimated explicitly
in very exceptional cases, so even the attraction constant α in the exponential
attraction rate is usually “unobservable”.

We now turn to perturbation theory. To this end we assume that there is

a family of cocycles Sδ,ξ (n) : Φ → Φ for some dynamical system T (n) : Ψ → Ψ which
depends on a parameter δ ∈ [0, 1] (for simplicity we assume that the set Ψ and
the T (n) are independent of δ, although this is not essential). We assume that,
for δ = 0, the corresponding maps Sδ,ξ (n) are independent of ξ and S0 (1) satisfies
174 S. V. ZELIK

Assumptions A–D. Thus, in the limit, for δ = 0, we have a regular attractor. To

specify the perturbation, we need two more assumptions.
Assumption E. The maps Sδ,ξ (1) ∈ C 1 (Φ, Φ) and the corresponding Fréchet
′
derivatives Sδ,ξ (1) are uniformly continuous on bounded subsets of Φ (also uniformly
with respect to ξ and δ). Moreover, the maps Sδ,ξ (1) are injective for all ξ and δ
′
and ker Sδ,ξ (1) = {0} for all‘δ, ξ at any point of Φ. Finally, we assume that
′
∥Sδ,ξ (1)(v) − S0 (1)(v)∥Φ + ∥Sδ,ξ (1)(v) − S0′ (1)(v)∥L(Φ,Φ) 6 Cδ, (8.40)

where the constant C depends on the norm ∥v∥Φ only (and is independent of ξ
and δ).
Assumption F. The family of cocycles Sδ,ξ (n) possesses a compact uniformly
attracting set B ∈ Φ with respect to the standard bornology of bounded sets in Φ
(which is also uniform with respect to δ).
The perturbation theory of regular attractors is based on two relatively simple
observations.
1) Since the cocycle Sδ,ξ (n) is close to the limit semigroup S0 (n) and possesses
a compact uniformly attracting set, the properties of perturbed trajectories will be
close to the properties of non-perturbed ones. In particular, any perturbed trajectory
will visit some ε-neighbourhood of the set of equilibria R0 , where ε = ε(δ) tends
to zero as δ → 0. Moreover, due to the absence of homoclinic structures for
the limiting dynamical system, this trajectory is unable to visit an appropriate
neighbourhood of any ui ∈ R0 more than once and spends a uniformly bounded
time outside the ε-neighbourhood of R0 (see [228] for the details). This actually
reduces the analysis to local perturbation theory near the hyperbolic equilibria.
2) Since the limit equilibria ui ∈ R0 are hyperbolic, the saddle structure survives
under small non-autonomous perturbations. In particular, any equilibrium ui ∈ R0
generates an “equilibrium” ui (ξ) of the perturbed cocycle Sδ,ξ (n) if δ is small enough.
This equilibrium is uniquely determined by the condition

sup ∥ui (ξ) − ui ∥Φ 6 Cδ. (8.41)

ξ∈Ψ

The “equilibria” ui (ξ) we have constructed remain hyperbolic, and the corresponding
unstable manifolds M+ +
δ (ui , ξ) are close to the unstable manifold M0 (ui ) of the limit
problem as δ = 0 and possess the uniform exponential tracking property. In addition,
due to the injectivity property these manifolds (which are initially defined for small
neighbourhood of ui only) can be extended to global unstable submanifolds, which
will be diffeomorphic to Rind(ui ) . This allows us to establish an analogue of (8.38) for
the non-autonomous (pullback) attractor Aδ (ξ) for the perturbed cocycle Sδ,ξ (n):
N
[
Aδ (ξ) = M+
δ (ui , ξ). (8.42)
i=1

Moreover, we also have the property that any complete bounded trajectory is
a heteroclinic orbit between two “equilibria” ui (ξ), as well as uniform exponential
 ATTRACTORS. THEN AND NOW 175

attraction of bounded sets. Finally, arguing as in the proof of Proposition 8.5, we

establish uniform Hölder continuity of the following form:

distΦ (Aδ (ξ), A0 ) 6 Cδ κ (8.43)

for some positive κ and C which are independent of ξ and δ; see [228] for the details.
We summarize the results obtained here in the following theorem.
Theorem 8.26. Let the family of cocycles Sδ,ξ (n) : Φ → Φ over a dynamical sys-
tem T (h) : Ψ → Ψ depending on the parameter δ ∈ [0, 1] satisfy Assumptions A–F.
Then there exists δ0 > 0 such that, for every δ 6 δ0 , the non-autonomous (pullback)
attractor Aδ (ξ) of the cocycle Sδ,ξ (n) is a finite union of the unstable mani-
folds M+ δ (ui , ξ) of the corresponding perturbed “equilibria” ui (ξ), ξ ∈ Ψ, of the limit
hyperbolic equilibria ui ∈ R0 (that is, (8.42) holds). Moreover, any complete bounded
trajectory of the dynamical process Uδ,ξ (m, n) associated with the cocycle Sδ,ξ (n)
is a heteroclinic orbit between ui (T (n)ξ) and uj (T (n)ξ) for some i ̸= j . The rate
of attraction to the attractors Aδ (ξ) is uniform and exponential, that is, there exist
a positive constant α and a monotone function Q such that, for every bounded set
B ⊂ Φ and every m > n,

distΦ Uξ (m, n)B, Aδ (T (n)ξ) 6 Q(∥B∥Φ )e−α(m−n) ,

δ 6 δ0 , ξ ∈ Ψ.

Finally, the family of attractors Aδ (ξ) is Hölder continuous at δ = 0, that is, (8.43)
holds.
Remark 8.27. We see that the object obtained as a non-autonomous pertur-
bation of a regular attractor gives us an example of a non-autonomous exponential
attractor introduced before. In particular, it is not only a pullback, but also
a forward in time exponential attractor. This gives us an evidence that the defini-
tion of a non-autonomous exponential attractor that we use is consistent and natural
at least in the deterministic (uniform) case. Unfortunately, a reasonably general
analogue of Theorem 8.26 does not seem to exist in the random (non-uniform) case;
see Example 5.19. This is closely related to the well-known problem of developing
the theory of centre (stable/unstable) manifolds for random/stochastic dynamical
systems. The theory of random exponential attractors developed in [211] and
discussed above may be one of the possible ways to handle this problem.

9. Determining functionals

In this section we discuss an alternative approach to the justification of the finite-

dimensional reduction in dissipative PDEs, which is related to the concept of de-
termining functionals. This approach was introduced in [75] (see also [147]) for
the case of the 2D Navier–Stokes equations and Fourier modes and was extended
subsequently to many other classes of dissipative systems and various classes of de-
termining functionals (see [41], [44], [46], [47], [77], [78], and the references therein).
In our exposition we mainly follow the recent paper [114].
Definition 9.1. Let S(t) : Φ → Φ be a dynamical system acting on a Banach
space Φ. Then a finite system of continuous functionals F := {F1 , . . . , FN },
176 S. V. ZELIK

Fi : Φ → R, is called asymptotically determining if for any two trajectories u1 (t) :=

S(t)u1 and u2 (t) := S(t)u2 , the convergence


lim Fi (u1 (t)) − Fi (u2 (t)) = 0, i = 1, . . . , N,
t→∞

implies that limt→∞ ∥u1 (t) − u2 (t)∥Φ = 0.

Thus, if F is an asymptotically determining system, then the behaviour of any
trajectory u(t) as t → ∞ is determined by the behaviour of finitely many quantities
F1 (u(t)), . . . , FN (u(t)). At first glance, this may look as a kind of finite-dimensional
reduction, but a more detailed analysis shows that it is not the case in general
since, in contrast to inertial forms constructed via Mané’s projection theorem or
inertial manifolds, the quantities {Fi (u(t))}N i=1 are not obliged to satisfy a system
of ODEs. Actually, in many cases they satisfy a system of delay differential
equations, but the phase spaces of such systems remain infinite-dimensional, so
despite the widespread misunderstanding, determining functionals do not give
any finite-dimensional reduction and are actually responsible for the reduction to
a system of ODEs with delay. Nevertheless, such a reduction is also interesting from
the theoretical point of view and has non-trivial applications to many related areas,
for instance, to establishing the controllability of an originally infinite-dimensional
system by finitely many modes (see, for example, [5]), to verifying the uniqueness
of an invariant measure for random/stochastic PDEs (see, for example, [142]), and
so on. We also mention more recent but promising applications of determining
functionals to data assimilation problems where the values of the functionals
Fi (u(t)) are interpreted as the results of observations and the theory of determining
functionals allows us to build new methods for recovering the trajectory u(t) from
the results of observations (see [5], [4], [184], and the references therein).
In the case where the dynamical system under consideration possesses an attractor
(with respect to the standard bornology of bounded sets in Φ) we may use an
alternative, non-equivalent but closely related concept of separating functionals.
Definition 9.2. Let the dynamical system S(t) : Φ → Φ acting on a Banach
space Φ possess an attractor A with respect to the standard bornology. Then
a system F := {Fi }N i=1 is separating on the attractor if for any two complete
bounded trajectories u1 (t) and u2 (t) lying on the attractor the equality
Fi (u1 (t)) = Fi (u2 (t)), t ∈ R, i = 1, . . . , N,
implies that u1 (t) ≡ u2 (t).
It is not difficult to show that (under the mild extra assumption that the S(t)
are continuous for every t), any separating system F of continuous functionals
is automatically asymptotically determining (see, for example, [114]). The converse is
not true in general: the corresponding counterexample which is on Example 2.7
is also given in [114]. This simple observation is rather useful since usually
an attractor is more regular and the separation property is easier to verify.
In order to illustrate the standard approach to determining functionals, we
consider the model example of an abstract semilinear parabolic equation
∂t u + Au = F (u), u t=0
= u0 , (9.1)
 ATTRACTORS. THEN AND NOW 177

in a Hilbert space Φ, where A : D(A) → Φ is a self-adjoint linear positive op-

erator with compact inverse and F : Φ → Φ is globally Lipschitz in Φ with
Lipschitz constant L. Let {λi }∞
i=1 be the eigenvalues of the operator A enumerated
in the non-decreasing order and {ei }∞
i=1 be the corresponding eigenvectors.

Theorem 9.3. Let N ∈ N be such that λN +1 > L. Then the system of functionals
Fi (u) = (u, ei ), i = 1, . . . , N (which are the first N Fourier modes) is asymptotically
determining for the dynamical system generated by equation (9.1).
Proof. Indeed, let u1 (t) and u2 (t) be two solutions of equation (9.1) such that

PN (u1 (t) − u2 (t)) → 0 as t → ∞,

where PN is a spectral orthoprojector related to the first N Fourier modes, and let
QN = 1 − PN . Let

v(t) := PN (u1 (t) − u2 (t)) and w(t) := QN (u1 (t) − u2 (t)).

Then the last function solves the equation

∂t w + Aw = QN (F (u1 ) − F (u2 )). (9.2)

Multiplying this equation by w and using that F is globally Lipschitz, we end up

with
1 d
∥w(t)∥2Φ + λN +1 ∥w(t)∥2Φ 6 L∥w∥2Φ + L∥w∥Φ ∥v∥Φ .
2 dt
Using the assumption λN +1 > L, we arrive at
d
∥w(t)∥Φ + α∥w(t)∥Φ 6 L∥v(t)∥Φ
dt
for some positive α. Integrating this inequality and using that ∥v(t)∥Φ → 0 as
t → ∞, we conclude that also ∥w(t)∥Φ → 0, which finishes the proof of the theorem.
Remark 9.4. The result of the last theorem can be generalized in a straightfor-
ward way to more general classes of linear determining functionals. For instance, if
a system F := {F1 , . . . , FN } of linear continuous functionals satisfies the following
inequality:
N
X
(L + α)∥v∥2Φ 6 (Av, v) + C |Fn (v)|, v ∈ D(A1/2 ),
n=1

for some positive α and C, then taking two trajectories u1 (t) and u2 (t) and
v(t) := u1 (t)−u2 (t), multiplying the equation for v by v in Φ, and using the Lipschitz
continuity of F and the last assumption, we arrive at
N
1 d X
∥v(t)∥2Φ + α∥v(t)∥2Φ 6 C |Fn (v(t))|. (9.3)
2 dt n=1

Thus, the system F is indeed asymptotically determining. This allows us to

construct determining nodes, as well as much more general systems of linear
determining functionals (see [44] and the references therein for more details).
178 S. V. ZELIK

We also note that, due to (9.3), the complete trajectory u(t), t ∈ R, is determined
in a unique way by the values ξi (t) := Fi (u(t)), that is,

u(t) = F(ξ1 (t + ·), . . . , ξN (t + ·)), F : [C(−∞, 0; R)]N → Φ,

and we have the exponential decay of the delay kernel:

∥F(ξ) − F(ξ)∥Φ 6 C sup {eαs ∥ξ(−s) − ξ(−s)∥RN }.

s∈R+

Thus, turning back to the quantities ξn (t) := Fn (u(t)), we see that, in general, they
do not satisfy a system of ODEs in RN , but satisfy a system of retarded ODEs with
infinite delay and exponentially decaying delay kernel:
d
ξ(t) = G(ξ(t + ·)), ξ(t) ∈ RN , G : [C(−∞, 0; R)]N → RN .
dt
Thus, since the phase space for such a retarded system remains infinite-dimensional,
determining functionals do not provide any finite-dimensional reduction, but reduce
the initial PDE to a system of ODEs with delay (which is often referred to as the
Lyapunov–Schmidt reduction).
We now give a more precise look at the number of elements in the “optimal”
system of determining functionals. To this end, we give the following definition.
Definition 9.5. Let S(t) : Φ → Φ be a dynamical system in a Banach space Φ.
Then the determining dimension dimdet (S(t), Φ) is the minimum number N of con-
tinuous functionals F1 , . . . , FN : Φ → R such that F := {Fn }N
n=1 is asymptotically
determining.
Note that we do not require the functionals Fn to be linear. Although in
applications determining systems often consist of linear functionals, in the general
theory the requirement of linearity looks artificial when a nonlinear dynamical
system is considered. In addition, the use of nonlinear (for example, quadratic,
cubic, and so on) functionals simplifies the theory and makes it more elegant.
We recall that there exists a simple and natural lower bound for dimdet (S(t), Φ),
which is related to the embedding dimension of the set R of equilibria points.
Namely, by definition, dimemb (R, Φ) is the minimum number M ∈ N such that
there exists a continuous injective map F : R → RM . Then, obviously,

dimdet (S(t), Φ) > dimemb (R, Φ). (9.4)

Indeed, any asymptotically determining system must, in particular, distinguish

different equilibria. Note also that dimemb (R, Φ) is a topological invariant and can
be estimated using, for example, Lebesgue covering dimension.
Another interesting observation is that, due to Sard’s theorem, the set of equi-
libria R is generically finite (see [12], [193]); in particular, it is so for equation (9.1).
Therefore, in a generic situation, (9.4) gives us dimdet (S(t), Φ) > 1 for the lower
bound. One of the most surprising results of the theory is that this estimate is
sharp, that is, there exists a dense set of smooth functionals such that each of them
is asymptotically determining.
 ATTRACTORS. THEN AND NOW 179

Theorem 9.6. Let the set R of equilibria be finite. Then the determining
dimension of the solution semigroup S(t) : Φ → Φ associated with equation (9.1) is
equal to one. Moreover, there is a dense/prevalent set of polynomial functionals
on Φ such that each of them is a determining functional for S(t).

Idea of the proof. The proof is based on the Hölder continuous infinite-dimen-
sional version of the famous Takens delay embedding theorem (see [193], [203], [213],
and the references therein). To apply this theorem, we need to control the size
of the set R of equilibria (which can be done by our assumptions) and the size
of the sets of periodic orbits of period τ, 2τ, . . . , kτ , where k is large enough. This
control is attained using the fact that if τ ≪ 1, then all such periodic orbits
must be equilibria (see [193] for the details). On the other hand, the fractal
dimension of the attractor A is finite and we have the one-to-one Hölder-continuous
Mané projection of this attractor onto the finite-dimensional set A ⊂ RN with
the projected dynamical system S(t) on it.
Thus, by an appropriate version of Takens’s delay embedding theorem (see, for
example, [193]) there exists a prevalent set of continuous functionals F : Φ → R
(polynomials of sufficiently large degree are enough) such that the map

Fk : A → Rk ,


Fk (u0 ) := F(u0 ), F(S(τ )u0 ), . . . , F(S((k − 1)τ )u0 ) ,

is one-to-one on the attractor A if k is large enough and τ > 0 is small enough. Thus,
each of these functionals is separating trajectories on the attractor and therefore is
asymptotically determining (see [114], [193] for more details).

Remark 9.7. The result of Theorem 9.6 also gives a reduction to a delayed
ODE. Namely, as shown in [114], there exists a continuous function Θ : Rk → R
such that the observable ξ(t) := F(u(t)), where u(t) is any complete trajectory on
the attractor, solves the scalar ODE with finite delay:

d

ξ(t) = Θ ξ(t − τ ), . . . , ξ(t − kτ ) , (9.5)
dt
and if the values of ξ(t − τ ), ξ(t − 2τ ), . . . , ξ(t − kτ ) are known for some t ∈ R, then
the corresponding point u(t) ∈ A can be recovered using the attractor reconstruction
procedure provided by Takens’ delay embedding theorem.

The situation when the set R is not-finite looks similar.

Proposition 9.8. Let S(t) : Φ → Φ be a dynamical system associated with

equation (9.1), and let R be its set of equilibria. Then the determining dimension
of S(t) satisfies

dimemb (R, Φ) 6 dimdet (S(t), Φ) 6 dimemb (R, Φ) + 1. (9.6)

The proof of this proposition is analogous to the proof of Theorem 9.6 and is
presented in [114]. We expect that the left-hand inequality (9.6) is actually equality,
but we have not checked this so far. Moreover, an analogue of delay ODE (9.5) also
holds in this case (see [114]).
180 S. V. ZELIK

Remark 9.9. We see that the determining functionals are indeed responsible
for a reduction of the original PDE to ODEs with delay and are not related
to any kind of finite-dimensional reduction. In addition, the minimum number
of determining functionals (determining dimension) is related to the “size” of the set
of equilibria R of the dynamical system under consideration only (and is not related
to any dynamical properties of this dynamical system). Although these results are
stated for a semilinear parabolic system (9.1) only, their close analogues remain
true for much wider class of equations including the Navier–Stokes system, damped
wave equations, and so on.
We conclude this section by several illustrative examples. We start with the best
studied case of one spatial dimension.
Example 9.10. Consider the following 1D semilinear heat equation:

∂t u = ν∂x2 u − f (u) + g, x ∈ [0, π], ν > 0, (9.7)

endowed with the Dirichlet boundary conditions

u x=0
=u x=π
= 0.

Assume also that the function f ∈ C 1 (R, R) satisfies some dissipativity conditions,
say f (u)u > −C. Then equation (9.7) generates a dissipative dynamical system
in the phase space Φ = L2 (0, π). and this dynamical system possesses a global
attractor A which is bounded at least in C 2 ([0, π]); see, for example, [12].
Moreover, the equilibria R of this problem satisfy the second-order ODE

νu′′ (x) − f (u(x)) + g = 0, u(0) = u(π) = 0.

Thus, the map u → u′ (0) provides a homeomorphic (and even smooth) embedding
of R in R, so dimemb (R, Φ) = 1. Thus, we expect that dimdet (S(t), Φ) = 1
(or at most 2 in accordance with Proposition 9.8). The possible explicit form
of the determining functional is well known here: F(u) := u x=x , where x0 > 0 is
0
small enough (see [140]). Indeed, let u1 (t), u2 (t) ∈ A be two complete trajectories
of (9.7) lying on the attractor such that u1 (t, x0 ) ≡ u2 (t, x0 ). Then the function
v(t) = u1 (t) − u2 (t) solves the equation

∂t v = ν ∂x2 v − l(t)v, v x=0

=v x=x0
= 0, (9.8)

where Z 1
l(t) := f ′ (su1 (t) + (1 − s)u2 (t)) ds.
0

Since the attractor A is bounded in C[0, π], we know that l(t) is globally bounded
in C[0, π], that is, |l(t)|C[0,π] 6 L. Multiplying equation (9.8) by v, integrating
with respect to x ∈ [0, x0 ], and using the fact that the first eigenvalue of −∂x2 with
Dirichlet boundary conditions is (π/x0 )2 , we get
 2
1 d π
∥v(t)∥2L2 + ν ∥v(t)∥2L2 − L∥v(t)∥2L2 6 0.
2 dt x0
 ATTRACTORS. THEN AND NOW 181

Fixing x0 > 0 to be small enough so that ν(π/x0 )2 > L, applying Gronwall’s

inequality, and using that ∥v(t)∥L2 remains bounded as t → −∞, we conclude that
v(t) ≡ 0 for all t ∈ R and x ∈ [0, x0 ]. Thus, the trajectories u1 (t, x) and u2 (t, x)
coincide for all t and all x ∈ [0, x0 ]. Using now arguments related to logarithmic
convexity or Carleman-type estimates, which work for much more general class
of equations (see [200]), we conclude that the trajectories u1 and u2 coincide. Thus,
F is separating on the attractor and therefore is asymptotically determining. Being
pedantic, we need to note that the functional F(u) = u x=x is not defined on
0
the phase space H, but rather on the proper subspace C[0, π] of it (this is a typical
situation for determining nodes; see [184]). However, we have an instantaneous
Φ → C[0, π] smoothing property, thus if we start from u0 ∈ Φ, then the value F(u(t))
will be defined for all t > 0, so we just ignore this small inconsistency.

Example 9.11. Let us consider the same equation (9.7) on [0, π], but endowed
with periodic boundary conditions. In this case we do not have the condition
2 1
v(0) = 0, so the set R of equilibria

is naturally embedded in R , rather than in R ,
′
by the map u → u x=0 , u x=0 . Thus, we cannot expect that the determining
dimension is 1. Moreover, at least in the case where g = const equation (9.7)
possesses a spatial shift as a symmetry, and therefore each non-trivial equilibrium
generates a whole circle of equilibria. Since a circle cannot be homeomorphically
embedded in R1 , the determining dimension must be at least 2. We claim that it is
indeed 2 and determining functionals can be taken in the form

F1 (u) := u x=0
, F2 (u) := u x=x0
. (9.9)

Indeed, arguing exactly as in the previous example, we see that the system
F = {F1 , F2 } is asymptotically determining if x0 > 0 is small enough.

The next natural example shows that the determining dimension may be finite
and small even when the corresponding global attractor is infinite-dimensional.

Example 9.12. Consider the semilinear heat equation (9.7) on the whole line
x ∈ R. The natural phase space for this problem is the uniformly-local space
n o
Φ = L2b (R) := u ∈ L2loc (R) : ∥u∥L2b := sup ∥u∥L2 (x,x+1) < ∞ . (9.10)
x∈R

It is known that, under a natural dissipativity assumption on f ∈ C 1 (R) (for

example, f (u)u > −C + αu2 for α > 0), this equation generates a dissipative
dynamical system S(t) in Φ for every g ∈ L2b (R). Moreover, this dynamical system
possesses a locally compact global attractor A, that is, a strictly invariant set bounded
in Φ and compact in L2loc (R) which attracts bounded sets in Φ in the topology
of L2loc (R) (see [179] and the references therein). Note that, in contrast to the case
of bounded domains, compactness and the attraction property in Φ fail in general
in the case of unbounded domains. It is also known that, at least in the case where
equation (9.7) possesses a spatially homogeneous exponentially unstable equilibrium
(for example, in the case where g = 0 and f (u) = u3 − u), the fractal dimension of A
is infinite (actually, A contains submanifolds of any finite dimension); see [179].
182 S. V. ZELIK

Nevertheless, the system of linear functionals (9.9) remains determining for this
equation by exactly the same reasons as in Examples 9.10 and 9.11. Thus,

dimdet (S(t), Φ) = 2.

One functional is not determining in general for the reasons explained in Exam-
ple 9.11.
We now give an example of a non-dissipative and even conservative system with
determining dimension 1.
Example 9.13. Consider the following 1D wave equation:

∂t2 u = ∂x2 u, x ∈ (0, π), u x=0

=u x=π
= 0, ξu t=0
= ξ0 , (9.11)

where ξu (t) := {u(t), ∂t u(t)}. It is well known that problem (9.11) is well posed
in the energy phase space E := H01 (0, π) × L2 (0, π) and the energy identity holds:

∥∂t u(t)∥2L2 + ∥∂x u(t)∥2L2 = const . (9.12)

Moreover, the solution u(t) can be found explicitly in terms of sin-Fourier series:
∞
X
u(t) = (An cos(nt) + Bn sin(nt)) sin(nx), (9.13)
n=1

where An = (2/π)(u(0), sin(nx)) and Bn = (2/π)(u′ (0), sin(nx)). It is crucial for us

that (9.13) is an almost-periodic function of time with values in H01 . Now consider
a linear functional on Φ = L2 (0, π):
∞
X
Fu = (l(x), u) = ln un , (9.14)
n=1

where ln and un are the Fourier coefficients of l ∈ Φ and u, respectively. Then

∞
X
Fu(t) = [ln An cos(nt) + ln Bn sin(nt)] (9.15)
n=1

is a scalar almost-periodic function. Since the Fourier coefficients of an almost-perio-

dic function are uniquely determined by this function, we have

Fu(t) ≡ 0 if and only if ln An = ln Bn = 0

(see [153] for details). Thus, if we take a generic function l (for which ln ̸= 0 for all
n ∈ N), then Fu(t) ≡ 0 implies that An = Bn = 0, and therefore u(t) ≡ 0. Thus,
F is separating on the set of complete trajectories. It remains to note that, since
any trajectory of (9.11) is almost-periodic in E, the ω-limit set of any trajectory
exists and is compact in E. Then it is not difficult to check (see [114]) that F is
also asymptotically determining, so the determining dimension of this system is 1.
 ATTRACTORS. THEN AND NOW 183

Remark 9.14. The principal difference between the case of one spatial dimension
and the multi-dimensional case is that in the 1D case any equilibrium u0 ∈ R
of such a PDE solves a system of ODEs, so the dimension of R is restricted by
the order of this system. This allows us in many cases to get sharp estimates
for the determining dimension and even compute it explicitly. In particular,
Examples 9.10 and 9.11 clearly show that it is independent of the physical
parameters of the dynamical system, as well as of the dimension of the attractor.
In contrast to this, in the multi-dimensional case u0 ∈ R usually solves an elliptic
PDE and we do not have any good formulae for the dimension of R. The best we
can do in general is to use the obvious estimate

dimemb (R, Φ) 6 dimemb (A, Φ) 6 2 dimf (A, Φ) + 1.

Since the fractal dimension of an attractor A usually depends on physical parameters

(for example, on the Grashof number if the Navier–Stokes system is considered), this
may produce an illusion that the number of determining functionals is also related
to these parameters. Nevertheless, as Proposition 9.8 shows, it is still determined
by the size of the set of equilibria set and is not related to the complexity
of the dynamics on it. These arguments show also that, in contrast to the 1D case,
in the multi-dimensional case we really need to use some “generic” assumptions
in order to kill pathological equilibria and get a reasonable result.

10. Appendix. Function spaces

The aim of this appendix is to introduce and discuss various classes of function
spaces which are used throughout the survey. We start with Lebesgue and Sobolev
spaces.
Definition 10.1. Let Ω be a domain in Rd with a sufficiently smooth boundary.
We denote by Lp (Ω), 1 6 p 6 ∞, the Lebesgue space of functions whose pth power
is Lebesgue integrable. For any n ∈ N, we denote by W n,p (Ω) the Sobolev space
of distributions whose derivatives up to order n inclusive belong to the space Lp (Ω).
For non-integer positive s = n + α, n ∈ Z+ , α ∈ (0, 1), the space W s,p (Ω) is defined
s
as a Besov space Bp,p (Ω) via the following norm:
X Z Z |Dβ u(x) − Dβ u(y)|p
∥u∥pW s,p := ∥u∥pW n,p (Ω) + dx dy. (10.1)
Ω×Ω |x − y|d+pα
|β|=n

We also denote by W0s,p (Ω) the closure of C0∞ (Ω) in the metric of W s,p (Ω) and, for
negative values of s, we define W s,p (Ω) by duality
1 1
W s,p (Ω) := [W0−s,q (Ω)]∗ , + =1
p q
(see [217] and the references therein for more details). We also use the notation H s (Ω)
for the spaces W s,p (Ω) with p = 2.
Now let V be a Banach (or more generally, a locally convex) space, and let V ∗
be its dual space (the space of linear continuous functionals on V ). Then the weak
184 S. V. ZELIK

topology on V is defined by the following system of seminorms on V : pl (x) := |lx|,

l ∈ V ∗ . On the level of sequences this means that xn ⇁ x in V if and only if
lxn → lx for all l ∈ V ∗ . The weak-star topology on the dual space V ∗ is defined
analogously (see, for example, [191], [201], and the references therein).
The key result, which is widely used in the theory of attractors, is the Banach–
Alaoglu theorem, which claims that the closed unit ball in V ∗ is compact
in the weak-star topology (this result can be extended to locally convex or even
linear topological spaces if we replace the unit ball by the polar U 0 ∈ V ∗ of any
bounded set U ⊂ V ; see [191] for the details). We recall that a set U is bounded
in a linear topological space if it is absorbed by any neighbourhood of zero, and
the polar U 0 is defined by
n o
U 0 := l ∈ V ∗ : sup |lx| 6 1 .
x∈U

The analogous results hold for the weak topology if and only if the space V is reflexive
(V = V ∗∗ ). Most important for us is the fact that the Sobolev spaces W s,p (Ω)
are reflexive if and only if 1 < p < ∞ (see [217]). We also recall that compactness
and sequential compactness are different (unrelated) concepts in non-metrizable
topological spaces, so the Banach–Alaoglu theorem (which is based on Tychonoff’s
compactness theorem) does not give a sequential weak-star compactness of the unit
ball in the dual space. In order to get sequential compactness we need to assume
that the space V is either separable (then the unit ball in the dual space is
metrizable in the weak-star topology) or reflexive (then weak compactness and weak
sequential compactness coincide due to the Eberlein–Smulian theory); see [191] for
more details. We also recall that L∞ (Ω) is the dual space of L1 (Ω), so there is
a natural choice of a weak-star topology in L∞ (Ω) which gives the compactness
and sequential compactness of a unit ball. In contrast to this, the spaces C(Ω)
and L1 (Ω) are not dual to any Banach spaces, and there is no reasonable topology
in them that gives the compactness of their unit balls.
In order to consider evolutionary equations, we often use the spaces of func-
tions u(t) with values in some Banach space V . Here usually t ranging over R or a
subset of R is interpreted as time and V is an appropriately chosen Sobolev space.
For any [a, b] ⊂ R and any 1 6 p 6 ∞ we define Lp (a, b; V ) as the space of Bochner
measurable functions such that
Z b
∥u∥pLp (a,b;V ) := ∥u(t)∥pV dt < ∞.
a

The space of functions that are Bochner measurable and locally integrable to power p
will be denoted by Lploc (a, b; V ). Sometimes, when we need to emphasize the prop-
erties of functions near the finite endpoints a and b, we will write, say, Lploc ([a, b], V )
or Lploc ((a, b), V ). The time Sobolev and Besov spaces W s,p (a, b; V ) are defined
analogously to (10.1); see [150] and the references therein.
The uniformly local space Lpb (a, b; V ) is defined as the subset of Lploc (a, b; V ) for
which the following norm is finite:

∥u∥Lpb (a,b;V ) := sup ∥u∥Lp (s,s+1;V )

s∈R, [s,s+1]⊂[a,b]
 ATTRACTORS. THEN AND NOW 185

and in the case b − a < 1 we just set ∥u∥Lpb (a,b;V ) := ∥u∥Lp (a,b;V ) . The spaces
Wbs,p (a, b; V ) are defined analogously.
We also need some other properties of the locally convex (Fréchet) space
Lploc (R, V ). We assume that the space V is reflexive and separable. Then
[Lp (a, b; V )]∗ = Lq (a, b; V ∗ ) for 1/p + 1/q = 1, 1 6 p < ∞. In particular,
these spaces are reflexive and separable for any p, 1 < p < ∞, and therefore
the Banach–Alaoglu theorem gives us the weak sequential compactness of the unit
ball in Lp (a, b; V ). This, in turn, gives us the weak sequential compactness of any
bounded weakly closed subset of Lploc (R, V ). In particular, if this bounded set is
convex, then closedness in the strong topology is enough. For instance, the unit
ball in Lpb (R, V ) is weakly sequentially compact in Lploc (R, V ). We use this fact,
in particular, to define the hulls of translation bounded external forces.
Recall that Lploc (R, V ) with the weak topology can be obtained as the projective
limit of the spaces Lp (−n, n; V ) (endowed with the weak topology) as n → ∞.
In particular, un ⇁ u in Lploc (R, V ) if and only if for any n ∈ N and any
l ∈ Lq (−n, n; V ∗ ) we have the convergence
Z n Z n
⟨l(t), un (t)⟩ dt → ⟨l(t), u(t)⟩⟩ dt.
−n −n

The author thanks V. Chepyzhov, A. Ilyin, V. Kalantarov, A. Kostianko and

D. Turaev for many stimulating discussions.

References

[1] В. С. Афраймович, В. В. Быков, Л. П. Шильников, “О возникновении и струк-

туре аттрактора Лоренца”, Докл. АН СССР, 234:2 (1977), 336–339; англ. пер.:
V. S. Afraimovich, V. V. Bykov, L. P. Shil’nikov, “On the origin and structure
of the Lorenz attractor”, Soviet Phys. Dokl., 22 (1977), 253–255.
[2] M. Anikushin, “Frequency theorem for parabolic equations and its relation to inertial
manifolds theory”, J. Math. Anal. Appl., 505:1 (2022), 125454, 23 pp.
[3] L. Arnold, Random dynamical systems, Springer Monogr. Math., Springer-Verlag,
New York, 1998, xvi+586 pp.
[4] A. Azouani, E. Olson, E. S. Titi, “Continuous data assimilation using general inter-
polant observables”, J. Nonlinear Sci., 24:2 (2014), 277–304.
[5] A. Azouani, E. S. Titi, “Feedback control of nonlinear dissipative systems by finite
determining parameters – a reaction-diffusion paradigm”, Evol. Equ. Control Theory,
3:4 (2014), 579–594.
[6] А. В. Бабин, “Аттрактор обобщенной полугруппы, порождённой эллиптическим
уравнением в цилиндрической области”, Изв. РАН. Сер. матем., 58:2 (1994),
3–18; англ. пер.: A. V. Babin, “Attractor of the generalized semigroup generated
by an elliptic equation in a cylindrical domain”, Russian Acad. Sci. Izv. Math., 44:2
(1995), 207–223.
[7] A. V. Babin, “Inertial manifolds for travelling-wave solutions of reaction-diffusion
systems”, Comm. Pure Appl. Math., 48:2 (1995), 167–198.
[8] A. V. Babin, “Global attractors in PDE”, Handbook of dynamical systems, v. 1B,
Ch. 14, Elsevier B. V., Amsterdam, 2006, 983–1085.
[9] A. Babin, B. Nicolaenko, “Exponential attractors of reaction-diffusion systems in an
unbounded domain”, J. Dynam. Differential Equations, 7:4 (1995), 567–590.
186 S. V. ZELIK

[10] A. V. Babin, M. I. Vishik, “Regular attractors of semigroups and evolution equa-

tions”, J. Math. Pures Appl. (9), 62:4 (1983), 441–491.
[11] А. В. Бабин, М. И. Вишик, “Максимальные аттракторы полугрупп, соот-
ветствующих эволюционным дифференциальным уравнениям”, Матем. сб.,
126(168):3 (1985), 397–419; англ. пер.: A. V. Babin, M. I. Vishik, “Maximal
attractors of semigroups corresponding to evolution differential equations”, Math.
USSR-Sb., 54:2 (1986), 387–408.
[12] А. В. Бабин, М. И. Вишик, Аттракторы эволюционных уравнений, Наука, M.,
1989, 296 с.; англ. пер.: A. V. Babin, M. I. Vishik, Attractors of evolution equations,
Stud. Math. Appl., 25, North-Holland Publishing Co., Amsterdam, 1992, x+532 pp.
[13] J. M. Ball, “On the asymptotic behavior of generalized processes, with applications
to nonlinear evolution equations”, J. Differential Equations, 27:2 (1978), 224–265.
[14] J. M. Ball, “Global attractors for damped semilinear wave equations”, Discrete Con-
tin. Dyn. Syst., 10:1-2 (2004), 31–52.
[15] B. Bilgin, V. Kalantarov, S. Zelik, “Preventing blow up by convective terms in dis-
sipative PDE’s”, J. Math. Fluid Mech., 18:3 (2016), 463–479.
[16] J. E. Billotti, J. P. LaSalle, “Dissipative periodic processes”, Bull. Amer. Math. Soc.,
77 (1971), 1082–1088.
[17] А. Э. Бирюк, “Спектральные свойства решений уравнения Бюргерса с ма-
лой диссипацией”, Функц. анализ и его прил., 35:1 (2001), 1–15; англ. пер.:
A. È. Biryuk, “Spectral properties of solutions of the Burgers equation with small
dissipation”, Funct. Anal. Appl., 35:1 (2001), 1–12.
[18] M. D. Blair, H. F. Smith, C. D. Sogge, “Strichartz estimates for the wave equation
on manifolds with boundary”, Ann. Inst. H. Poincaré C Anal. Non Linéaire, 26:5
(2009), 1817–1829.
[19] M. A. Blinchevskaya, Yu. S. Ilyashenko, “Estimate for the entropy dimension
of the maximal attractor for k-contracting systems in an infinite-dimensional space”,
Russ. J. Math. Phys., 6:1 (1999), 20–26.
[20] T. Buckmaster, V. Vicol, “Nonuniqueness of weak solutions to the Navier–Stokes
equation”, Ann. of Math. (2), 189:1 (2019), 101–144.
[21] C. J. Budd, V. Rottschäfer, J. F. Williams, “Multibump, blow-up, self-similar solu-
tions of the complex Ginzburg–Landau equation”, SIAM J. Appl. Dyn. Syst., 4:3
(2005), 649–678.
[22] J. M. Burgers, “A mathematical model illustrating the theory of turbulence”, Ad-
vances in applied mechanics, v. 1, Academic Press, Inc., New York, N. Y., 1948,
171–199.
[23] N. Burq, G. Lebeau, F. Planchon, “Global existence for energy critical waves in 3-D
domains”, J. Amer. Math. Soc., 21:3 (2008), 831–845.
[24] M. Callaway, Thai Son Doan, J. S. W. Lamb, M. Rasmussen, “The dichotomy spec-
trum for random dynamical systems and pitchfork bifurcations with additive noise”,
Ann. Inst. Henri Poincaré Probab. Stat., 53:4 (2017), 1548–1574.
[25] A. Calsina, X. Mora, J. Sola-Morales, “The dynamical approach to elliptic problems
in cylindrical domains, and a study of their parabolic singular limit”, J. Differential
Equations, 102:2 (1993), 244–304.
[26] A. N. Carvalho, J. A. Langa, “An extension of the concept of gradient semi-
groups which is stable under perturbation”, J. Differential Equations, 246:7 (2009),
2646–2668.
[27] A. N. Carvalho, J. A. Langa, J. C. Robinson, Attractors for infinite-dimensional
non-autonomous dynamical systems, Appl. Math. Sci., 182, Springer, New York,
2013, xxxvi+409 pp.
 ATTRACTORS. THEN AND NOW 187

[28] A. N. de Carvalho, S. Sonner, “Pullback exponential attractors for evolution pro-

cesses in Banach spaces: theoretical results”, Commun. Pure Appl. Anal., 12:6
(2013), 3047–3071.
[29] N. Chafee, E. F. Infante, “A bifurcation problem for a nonlinear partial differential
equation of parabolic type”, Appl. Anal., 4 (1974), 17–37.
[30] Ч. Чанг, Д. Ли, Ч. Сун, С. В. Зелик, “Детерминистские и случайные аттракторы
волновых уравнений со знакопеременной диссипацией”, Изв. РАН. Сер. матем.,
87:1 (2023), 161–210.
[31] V. V. Chepyzhov, A. Yu. Goritsky, M. I. Vishik, “Integral manifolds and attractors
with exponential rate for nonautonomous hyperbolic equations with dissipation”,
Russ. J. Math. Phys., 12:1 (2005), 17–39.
[32] V. V. Chepyzhov, A. A. Ilyin, “On the fractal dimension of invariant sets: applications
to Navier–Stokes equations”, Discrete Contin. Dyn. Syst., 10:1-2 (2004), 117–135.
[33] V. V. Chepyzhov, A. Kostianko, S. Zelik, “Inertial manifolds for the hyperbolic re-
laxation of semilinear parabolic equations”, Discrete Contin. Dyn. Syst. Ser. B, 24:3
(2019), 1115–1142.
[34] V. V. Chepyzhov, E. S. Titi, M. I. Vishik, “On the convergence of solutions
of the Leray-α model to the trajectory attractor of the 3D Navier–Stokes system”,
Discrete Contin. Dyn. Syst., 17:3 (2007), 481–500.
[35] V. Chepyzhov, M. Vishik, “A Hausdorff dimension estimate for kernel sections
of non-autonomous evolution equations”, Indiana Univ. Math. J., 42:3 (1993),
1057–1076.
[36] V. V. Chepyzhov, M. I. Vishik, “Trajectory attractors for evolution equations”, C. R.
Acad. Sci. Paris Sér. I Math., 321:10 (1995), 1309–1314.
[37] V. V. Chepyzhov, M. I. Vishik, “Attractors of non-autonomous evolution equations
with translation-compact symbols”, Partial differential operators and mathematical
physics (Holzhau, 1994), Oper. Theory Adv. Appl., 78, Birkhäuser Verlag, Basel,
1995, 49–60.
[38] V. V. Chepyzhov, M. I. Vishik, Attractors for equations of mathematical physics,
Amer. Math. Soc. Colloq. Publ., 49, Amer. Math. Soc., Providence, RI, 2002,
xii+363 pp.
[39] V. V. Chepyzhov, M. I. Vishik, S. V. Zelik, “Strong trajectory attractors for dissipa-
tive Euler equations”, J. Math. Pures Appl. (9), 96:4 (2011), 395–407.
[40] Shui-Nee Chow, Kening Lu, G. R. Sell, “Smoothness of inertial manifolds”, J. Math.
Anal. Appl., 169:1 (1992), 283–312.
[41] И. Д. Чуешов, “Теория функционалов, однозначно определяющих асимптоти-
ческую динамику бесконечномерных диссипативных систем”, УМН, 53:4(322)
(1998), 77–124; англ. пер.: I. D. Chueshov, “Theory of functionals that uniquely
determine the asymptotic dynamics of infinite-dimensional dissipative systems”,
Russian Math. Surveys, 53:4 (1998), 731–776.
[42] I. Chueshov, Monotone random systems theory and applications, Lecture Notes
in Math., 1779, Springer-Verlag, Berlin, 2002, viii+234 pp.
[43] И. Д. Чуешов, Введение в теорию бесконечномерных диссипативных систем,
АКТА, Харьков, 1999, 436 с.; англ. пер.: I. D. Chueshov, Introduction to the theory
of infinite-dimensional dissipative systems, ACTA, Kharkiv, 2002, 418 pp.
[44] I. Chueshov, Dynamics of quasi-stable dissipative systems, Universitext, Springer,
Cham, 2015, xvii+390 pp.
[45] Ф. Клемент, Х. Хейманс, С. Ангенент, К. ван Дуйн, Б. де Пахтер, Однопа-
раметрические полугруппы, Мир, M., 1992, 352 с.; пер. с англ.: Ph. Clément,
H. J. A. M. Heijmans, S. Angenent, C. J. van Duijn, B. de Pagter, One-parameter
semigroups, CWI Monogr., 5, North-Holland Publishing Co., Amsterdam, 1987,
x+312 pp.
188 S. V. ZELIK

[46] B. Cockburn, D. A. Jones, E. S. Titi, “Determining degrees of freedom for nonlinear

dissipative equations”, C. R. Acad. Sci. Paris Sér. I Math., 321:5 (1995), 563–568.
[47] B. Cockburn, D. Jones, E. Titi, “Estimating the number of asymptotic de-
grees of freedom for nonlinear dissipative systems”, Math. Comp., 97:219 (1997),
1073–1087.
[48] C. Conley, Isolated invariant sets and the Morse index, CBMS Regional Conf. Ser.
in Math., 38, Amer. Math. Soc., Providence, RI, 1978, iii+89 pp.
[49] P. Constantin, C. Foias, “Global Lyapunov exponents, Kaplan–Yorke formulae and
the dimension of the attractors for 2D Navier–Stokes equations”, Comm. Pure Appl.
Math., 38:1 (1985), 1–27.
[50] M. Conti, V. Pata, R. Temam, “Attractors for processes on time-dependent spaces.
Applications to wave equations”, J. Differential Equations, 255:6 (2013), 1254–1277.
[51] H. Crauel, “Random point attractors versus random set attractors”, J. London Math.
Soc. (2), 63:2 (2001), 413–427.
[52] H. Crauel, F. Flandoli, “Attractors for random dynamical systems”, Probab. Theory
Related Fields, 100:3 (1994), 365–393.
[53] H. Crauel, F. Flandoli, “Additive noise destroys a pitchfork bifurcation”, J. Dynam.
Differential Equations, 10:2 (1998), 259–274.
[54] H. Crauel, F. Flandoli, “Hausdorff dimension of invariant sets for random dynamical
systems”, J. Dynam. Differential Equations, 10:3 (1998), 449–474.
[55] H. Crauel, P. E. Kloeden, “Nonautonomous and random attractors”, Jahresber.
Dtsch. Math.-Ver., 117:3 (2015), 173–206.
[56] H. Crauel, M. Scheutzow, “Minimal random attractors”, J. Differential Equations,
265:2 (2018), 702–718.
[57] G. Da Prato, J. Zabczyk, Stochastic equations in infinite dimensions, Encyclopedia
Math. Appl., 44, Cambridge Univ. Press, Cambridge, 1992, xviii+454 pp.
[58] A. Debussche, “Hausdorff dimension of a random invariant set”, J. Math. Pures
Appl. (9), 77:10 (1998), 967–988.
[59] E. DiBenedetto, Degenerate parabolic equations, Universitext, Springer-Verlag, New
York, 1993, xvi+387 pp.
[60] A. Douady, J. Oesterlé, “Dimension de Hausdorff des attracteurs”, C. R. Acad. Sci.
Paris Sér. A-B, 290:24 (1980), A1135–A1138.
[61] J.-P. Eckmann, J. Rougemont, “Coarsening by Ginzburg–Landau dynamics”, Comm.
Math. Phys., 199:2 (1998), 441–470.
[62] J.-P. Eckmann, D. Ruelle, “Ergodic theory of chaos and strange attractors”, Rev.
Modern Phys., 57:3 (1985), 617–656.
[63] A. Eden, C. Foias, B. Nicolaenko, R. Temam, Exponential attractors for dissipative
evolution equations, RAM Res. Appl. Math., 37, Masson, Paris; John Wiley & Sons,
Ltd., Chichester, 1994, viii+183 pp.
[64] А. Еден, С. В. Зелик, В. К. Калантаров, “Контрпримеры к регулярности про-
екций Мане в теории аттракторов”, УМН, 68:2(410) (2013), 3–32; англ. пер.:
A. Eden, S. V. Zelik, V. K. Kalantarov, “Counterexamples to regularity of Mañé
projections in the theory of attractors”, Russian Math. Surveys, 68:2 (2013),
199–226.
[65] M. Efendiev, A. Miranville, S. Zelik, “Exponential attractors for a nonlinear
reaction-diffusion system in R3 ”, C. R. Acad. Sci. Paris Sér. I Math., 330:8 (2000),
713–718.
[66] M. Efendiev, A. Miranville, S. Zelik, “Infinite dimensional exponential attractors
for a non-autonomous reaction-diffusion system”, Math. Nachr., 248–249:1 (2003),
72–96.
 ATTRACTORS. THEN AND NOW 189

[67] M. Efendiev, A. Miranville, S. Zelik, “Infinite-dimensional exponential attractors for

nonlinear reaction-diffusion systems in unbounded domains and their approxima-
tion”, Proc. R. Soc. Lond. Ser. A Math. Phys. Eng. Sci., 460:2044 (2004), 1107–1129.
[68] M. Efendiev, A. Miranville, S. Zelik, “Exponential attractors for evolution equa-
tions”, Evolution equations and asymptotic analysis of solutions, RIMS Kôkyûroku,
1436, RIMS, Kyoto Univ., Kyoto, 2005, 45–65.
[69] M. Efendiev, S. Zelik, “Finite- and infinite-dimensional attractors for porous media
equations”, Proc. Lond. Math. Soc. (3), 96:1 (2008), 51–77.
[70] M. Efendiev, S. Zelik, “Finite-dimensional attractors and exponential attractors for
degenerate doubly nonlinear equations”, Math. Methods Appl. Sci., 32:13 (2009),
1638–1668.
[71] M. Efendiev, S. Zelik, A. Miranville, “Exponential attractors and finite-dimensional
reduction for non-autonomous dynamical systems”, Proc. Roy. Soc. Edinburgh
Sect. A, 135:4 (2005), 703–730.
[72] P. Fabrie, C. Galusinski, A. Miranville, S. Zelik, “Uniform exponential attractors for
a singularly perturbed damped wave equation”, Discrete Contin. Dyn. Syst., 10:1-2
(2004), 211–238.
[73] B. Fiedler, C. Rocha, “Connectivity and design of planar global attractors of Sturm
type. III. Small and platonic examples”, J. Dynam. Differential Equations, 22:2
(2010), 121–162.
[74] B. Fiedler, A. Scheel, M. I. Vishik, “Large patterns of elliptic systems in infinite
cylinders”, J. Math. Pures Appl. (9), 77:9 (1998), 879–907.
[75] C. Foiaş, G. Prodi, “Sur le comportement global des solutions non-stationnaires des
équations de Navier–Stokes en dimension 2”, Rend. Sem. Mat. Univ. Padova, 39
(1967), 1–34.
[76] C. Foias, G. R. Sell, R. Temam, “Inertial manifolds for nonlinear evolutionary equa-
tions”, J. Differential Equations, 73:2 (1988), 309–353.
[77] C. Foias, R. Temam, “Determination of the solutions of the Navier–Stokes equations
by a set of nodal values”, Math. Comp., 43:167 (1984), 117–133.
[78] C. Foias, E. S. Titi, “Determining nodes, finite difference schemes and inertial man-
ifolds”, Nonlinearity, 4:1 (1991), 135–153.
[79] R. L. Frank, D. Hundertmark, M. Jex, Phan Thành Nam, “The Lieb–Thirring in-
equality revisited”, J. Eur. Math. Soc. (JEMS ), 23:8 (2021), 2583–2600.
[80] R. L. Frank, A. Laptev, T. Weidl, Schrödinger operators: eigenvalues and Lieb–
Thirring inequalities, Cambridge Stud. Adv. Math., 200, Cambridge Univ. Press,
Cambridge, 2023, xiii+507 pp.
[81] У. Фриш, Турбулентность. Наследие А. Н. Колмогорова, Фазис, М., 1998, 346 с.;
пер. с англ.: U. Frisch, Turbulence. The legacy of A. N. Kolmogorov, Cambridge
Univ. Press, Cambridge, 1995, xiv+296 pp.
[82] C. G. Gal, Yanqiu Guo, “Inertial manifolds for the hyperviscous Navier–Stokes equa-
tions”, J. Differential Equations, 265:9 (2018), 4335–4374.
[83] M. J. Garrido-Atienza, B. Schmalfuß, “Ergodicity of the infinite dimensional frac-
tional Brownian motion”, J. Dynam. Differential Equations, 23:3 (2011), 671–681.
[84] S. Gatti, A. Miranville, V. Pata, S. Zelik, “Continuous families of exponential attrac-
tors for singularly perturbed equations with memory”, Proc. Roy. Soc. Edinburgh
Sect. A, 140:2 (2010), 329–366.
[85] J.-M. Ghidaglia, “Finite-dimensional behavior for weakly damped driven Schrödin-
ger equations”, Ann. Inst. H. Poincaré Anal. Non Linéaire, 5:4 (1988), 365–405.
[86] A. Giraldo, M. A. Morón, F. R. Ruiz Del Portal, J. M. R. Sanjurjo, “Shape of global
attractors in topological spaces”, Nonlinear Anal., 60:5 (2005), 837–847.
190 S. V. ZELIK

[87] С. В. Гонченко, М. Н. Кайнов, А. О. Казаков, Д. В. Тураев, “О методах провер-

ки псевдогиперболичности странных аттракторов”, Известия вузов. ПНД, 29:1
(2021), 160–185.
[88] С. В. Гонченко, Д. В. Тураев, “О трех типах динамики и понятии аттракто-
ра”, Порядок и хаос в динамических системах, Сборник статей. К 80-летию со
дня рождения академика Дмитрия Викторовича Аносова, Труды МИАН, 297,
МАИК “Наука/Интерпериодика“, М., 2017, 133–157; англ. пер.: S. V. Gonchenko,
D. V. Turaev, “On three types of dynamics and the notion of attractor”, Proc. Steklov
Inst. Math., 297 (2017), 116–137.
[89] J. Goodman, “Stability of the Kuramoto–Sivashinsky and related systems”, Comm.
Pure Appl. Math., 47:3 (1994), 293–306.
[90] M. Grasselli, G. Schimperna, S. Zelik, “Trajectory and smooth attractors for Cahn–
Hilliard equations with inertial term”, Nonlinearity, 23:3 (2010), 707–737.
[91] J. Guckenheimer, P. Holmes, Nonlinear oscillations, dynamical systems, and bifur-
cations of vector fields, Appl. Math. Sci., 42, Springer-Verlag, New York, 1983,
xvi+453 pp.
[92] J. Guckenheimer, R. F. Williams, “Structural stability of Lorenz attractors”, Inst.
Hautes Études Sci. Publ. Math., 50 (1979), 59–72.
[93] J. K. Hale, Asymptotic behavior of dissipative systems, Math. Surveys Monogr., 25,
Amer. Math. Soc., Providence, RI, 1988, x+198 pp.
[94] J. K. Hale, “Stability and gradient dynamical systems”, Rev. Mat. Complut., 17:1
(2004), 7–57.
[95] J. K. Hale, G. Raugel, “Lower semicontinuity of attractors of gradient systems and
applications”, Ann. Mat. Pura Appl. (4), 154:1 (1989), 281–326.
[96] A. Haraux, Systèmes dynamiques dissipatifs et applications, Rech. Math. Appl., 17,
Masson, Paris, 1991, xii+132 pp.
[97] D. B. Henry, “Some infinite-dimensional Morse–Smale systems defined by parabolic
partial differential equations”, J. Differential Equations, 59:2 (1985), 165–205.
[98] M. A. Herrero, J. J. L. Velázquez, “A blow-up mechanism for a chemotaxis model”,
Ann. Scuola Norm. Sup. Pisa Cl. Sci. (4), 24:4 (1997), 633–683.
[99] C. Hipp, “Convergence rates of the strong law for stationary mixing sequences”,
Z. Wahrsch. Verw. Gebiete, 49:1 (1979), 49–62.
[100] M. W. Hirsch, H. L. Smith, Xiao-Qiang Zhao, “Chain transitivity, attractivity, and
strong repellors for semidynamical systems”, J. Dynam. Differential Equations, 13
(2001), 107–131.
[101] L. T. Hoang, E. J. Olson, J. C. Robinson, “On the continuity of global attractors”,
Proc. Amer. Math. Soc., 143:10 (2015), 4389–4395.
[102] H. Hogbe-Nlend, Bornologies and functional analysis. Introductory course on the the-
ory of duality topology-bornology and its use in functional analysis, North-Holland
Math. Stud., 26, Notas de Matemática, No. 62, North-Holland Publ. Co.,
Amsterdam–New York–Oxford, 1977, xii+144 pp.
[103] M. Holland, I. Melbourne, “Central limit theorems and invariance principles for
Lorenz attractors”, J. Lond. Math. Soc. (2), 76:2 (2007), 345–364.
[104] E. Hopf, “Über die Anfgangswertaufgabe für die hydrodynamischen Grundgleichun-
gen”, Math. Nachr., 4:1-6 (1950), 213–231.
[105] B. R. Hunt, “Maximum local Lyapunov dimension bounds the box dimension
of chaotic attractors”, Nonlinearity, 9:4 (1996), 845–852.
[106] B. R. Hunt, V. Yu. Kaloshin, “Regularity of embeddings of infinite-dimensional frac-
tal sets into finite-dimensional spaces”, Nonlinearity, 12:5 (1999), 1263–1275.
[107] Ю. С. Ильяшенко, “О размерности аттракторов k-сжимающих систем в бес-
конечномерном пространстве”, Вестн. Моск. ун-та. Сер. 1. Матем., мех.,
 ATTRACTORS. THEN AND NOW 191

1983, № 3, 52–59; англ. пер.: Yu. S. Il’yashenko, “On the dimension of attractors
of k-contracting systems in an infinite-dimensional space”, Moscow Univ. Math.
Bull., 38:3 (1983), 61–69.
[108] Ю. С. Ильяшенко, “Многомерные костистые аттракторы”, Функц. анализ и его
прил., 46:4 (2012), 1–13; англ. пер.: Yu. S. Ilyashenko, “Multidimensional bony
attractors”, Funct. Anal. Appl., 46:4 (2012), 239–248.
[109] A. Ilyin, A. Kostianko, S. Zelik, “Sharp upper and lower bounds of the attractor
dimension for 3D damped Euler–Bardina equations”, Phys. D, 432 (2022), 133156,
13 pp.
[110] A. A. Ilyin, S. Zelik, “Sharp dimension estimates of the attractor of the damped 2D
Euler–Bardina equations”, Partial differential equations, spectral theory, and math-
ematical physics. The Ari Laptev anniversary volume, EMS Ser. Congr. Rep., EMS
Press, Berlin, 2021, 209–229.
[111] N. Ishimura, I. Ohnishi, “Inertial manifolds for Burgers’ original model system of tur-
bulence”, Appl. Math. Lett., 7:3 (1994), 33–37.
[112] А. В. Иванов, “Квазилинейные вырождающиеся и неравномерно эллиптические
и параболические уравнения второго порядка”, Тр. МИАН СССР, 160, 1982,
3–285; англ. пер.: A. V. Ivanov, “Quasilinear degenerate and nonuniformly elliptic
and parabolic equations of second order”, Proc. Steklov Inst. Math., 160 (1984),
1–288.
[113] А. Г. Качуровский, “Скорости сходимости в эргодических теоремах”, УМН,
51:4(310) (1996), 73–124; англ. пер.: A. G. Kachurovskii, “The rate of convergence
in ergodic theorems”, Russian Math. Surveys, 51:4 (1996), 653–703.
[114] V. Kalantarov, A. Kostianko, S. Zelik, “Determining functionals and finite-dimen-
sional reduction for dissipative PDEs revisited”, J. Differential Equations, 345
(2023), 78–103.
[115] V. Kalantarov, A. Savostianov, S. Zelik, “Attractors for damped quintic wave equa-
tions in bounded domains”, Ann. Henri Poincaré, 17:9 (2016), 2555–2584.
[116] А. Б. Каток, Б. Хасселблат, Введение в современную теорию динамических
систем, Факториал, М., 1999, 768 с.; пер. с англ.: A. Katok, B. Hasselblatt,
Introduction to the modern theory of dynamical systems, Encyclopedia Math. Appl.,
54, Cambridge Univ. Press, Cambridge, 1995, xviii+802 pp.
[117] Р. З. Хасьминский, Устойчивость систем дифференциальных уравнений при
случайных возмущениях их параметров, Наука, М., 1969, 367 с.; англ. пер.:
R. Khasminskii, Stochastic stability of differential equations, 2nd ed., Stoch. Model.
Appl. Probab., 66, Springer, Heidelberg, 2012, xviii+339 pp.
[118] Y. Kifer, Random perturbations of dynamical systems, Progr. Probab. Statist., 16,
Birkhäuser Boston, Inc., Boston, MA, 1988, vi+294 pp.
[119] K. Kirchgässner, “Wave-solutions of reversible systems and applications”, J. Differ-
ential Equations, 45:1 (1982), 113–127.
[120] P. E. Kloeden, J. A. Langa, “Flattening, squeezing and the existence of random
attractors”, Proc. R. Soc. Lond. Ser. A Math. Phys. Eng. Sci., 463:2077 (2007),
163–181.
[121] P. E. Kloeden, M. Rasmussen, Nonautonomous dynamical systems, Math. Surveys
Monogr., 176, Amer. Math. Soc., Providence, RI, 2011, viii+264 pp.
[122] P. E. Kloeden, J. Valero, “The Kneser property of the weak solutions of the three
dimensional Navier–Stokes equations”, Discrete Contin. Dyn. Syst., 28:1 (2010),
161–179.
[123] P. E. Kloeden, Meihua Yang, An introduction to nonautonomous dynamical systems
and their attractors, Interdiscip. Math. Sci., 21, World Sci. Publ., Hackensack, NJ,
2021, xii+144 pp.
192 S. V. ZELIK

[124] N. Koksch, “Almost sharp conditions for the existence of smooth inertial manifolds”,
Proceedings of Equadiff 9, Conference on differential equations and their applications,
Masaryk Univ., Brno, 1998, 139–166.
[125] А. Н. Колмогоров, В. М. Тихомиров, “ε-энтропия и ε-емкость множеств в функ-
циональных пространствах”, УМН, 14:2(86) (1959), 3–86; англ. пер.: A. N. Kol-
mogorov, V. M. Tikhomirov, “ε-entropy and ε-capacity of sets in functional spaces”,
Selected works of A. N. Kolmogorov, v. III, Math. Appl. (Soviet Ser.), 27, Kluwer
Acad. Publ., Dordrecht, 1993, 86–170.
[126] A. Kostianko, “Inertial manifolds for the 3D modified-Leray-α model with periodic
boundary conditions”, J. Dynam. Differential Equations, 30:1 (2018), 1–24.
[127] A. Kostianko, “Bi-Lipschitz Mané projectors and finite-dimensional reduction for
complex Ginzburg–Landau equation”, Proc. A, 476:2239 (2020), 20200144, 14 pp.
[128] A. Kostianko, Xinhua Li, Chunyou Sun, S. Zelik, “Inertial manifolds via spatial
averaging revisited”, SIAM J. Math. Anal., 54:1 (2022), 268–305.
[129] A. Kostianko, Chunyou Sun, S. Zelik, “Reaction-diffusion systems with supercritical
nonlinearities revisited”, Math. Ann., 384:1-2 (2022), 1–45.
[130] A. Kostianko, Chunyou Sun, S. Zelik, “Inertial manifolds for 3D complex Ginzburg–
Landau equations with periodic boundary conditions”, Indiana Univ. Math. J. (to
appear).
[131] A. Kostianko, S. Zelik, “Inertial manifolds for the 3D Cahn–Hilliard equations with
periodic boundary conditions”, Commun. Pure Appl. Anal., 14:5 (2015), 2069–2094.
[132] A. Kostianko, S. Zelik, “Inertial manifolds for 1D reaction-diffusion-advection sys-
tems. Part I: Dirichlet and Neumann boundary conditions”, Commun. Pure Appl.
Anal., 16:6 (2017), 2357–2376.
[133] A. Kostianko, S. Zelik, “Inertial manifolds for 1D reaction-diffusion-advection sys-
tems. Part II: Periodic boundary conditions”, Commun. Pure Appl. Anal., 17:1
(2018), 285–317.
[134] A. Kostianko, S. Zelik, “Kwak transform and inertial manifolds revisited”, J. Dynam.
Differential Equations, 34:4 (2022), 2975–2995.
[135] A. Kostianko, S. Zelik, “Attractors for semigroups with multi-dimensional time and
PDEs in unbounded domains”, Turkish J. Math., 47:3 (2023), 988–1002.
[136] A. Kostianko, S. Zelik, “Smooth extensions of inertial manifolds”, Anal. PDE (to
appear).
[137] A. Kostianko, S. Zelik, Attractors in topological spaces revisited, preprint.
[138] I. N. Kostin, “Lower semicontinuity of a non-hyperbolic attractor”, J. London Math.
Soc. (2), 52:3 (1995), 568–582.
[139] P. Kuchment, Floquet theory for partial differential equations, Oper. Theory Adv.
Appl., 60, Birkhäuser Verlag, Basel, 1993, xiv+350 pp.
[140] I. Kukavica, “On the number of determining nodes for the Ginzburg–Landau equa-
tion”, Nonlinearity, 5:5 (1992), 997–1006.
[141] S. B. Kuksin, “Spectral properties of solutions for nonlinear PDEs in the turbulent
regime”, Geom. Funct. Anal., 9:1 (1999), 141–184.
[142] S. Kuksin, A. Shirikyan, Mathematics of two-dimensional turbulence, Cambridge
Tracts in Math., 194, Cambridge Univ. Press, Cambridge, 2012, xvi+320 pp.
[143] N. V. Kuznetsov, T. N. Mokaev, O. A. Kuznetsova, E. V. Kudryashova, “The Lorenz
system: hidden boundary of practical stability and the Lyapunov dimension”, Non-
linear Dynam., 102:2 (2020), 713–732.
[144] M. Kwak, “Finite-dimensional inertial forms for 2D Navier–Stokes equations”, Indi-
ana Univ. Math. J., 41:4 (1992), 927–981.
[145] M. Kwak, “Finite-dimensional description of convective reaction-diffusion equa-
tions”, J. Dynam. Differential Equations, 4:3 (1992), 515–543.
 ATTRACTORS. THEN AND NOW 193

[146] Hyukjin Kwean, “An extension of the principle of spatial averaging for inertial man-
ifolds”, J. Austral. Math. Soc. Ser. A, 66:1 (1999), 125–142.
[147] О. А. Ладыженская, “О динамической системе, порождаемой уравнениями На-
вье–Стокса”, Краевые задачи математической физики и смежные вопросы
теории функций. 6, Зап. науч. сем. ЛОМИ, 27, Изд-во “Наука”, Ленинград. отд.,
Л., 1972, 91–115; англ. пер.: O. A. Ladyzhenskaya, “A dynamical system generated
by the Navier–Stokes equations”, J. Soviet Math., 3:4 (1975), 458–479.
[148] О. А. Ладыженская, “О конечномерности ограниченных инвариантных мно-
жеств для системы Навье–Стокса и других диссипативных систем”, Краевые
задачи математической физики и смежные вопросы теории функций. 14, Зап.
науч. сем. ЛОМИ, 115, Изд-во “Наука”, Ленинград. отд., Л., 1982, 137–155;
англ. пер.: O. A. Ladyzhenskaya, “Finite-dimensionality of bounded invariant sets
for Navier–Stokes systems and other dissipative systems”, J. Soviet Math., 28:5
(1985), 714–726.
[149] O. A. Ladyzhenskaya, Attractors for semigroups and evolution equations, Reprint
of the 1991 ed., Cambridge Math. Lib., Cambridge Univ. Press, Cambridge, 2022,
xxviii+68 pp.
[150] P. G. Lemarié-Rieusset, Recent developments in the Navier–Stokes problem, Chap-
man & Hall/CRC Res. Notes Math., 431, Chapman & Hall/CRC, Boca Raton, FL,
2002, xiv+395 pp.
[151] G. A. Leonov, N. V. Kuznetsov, N. A. Korzhemanova, D. V. Kusakin, “Lyapunov di-
mension formula for the global attractor of the Lorenz system”, Commun. Nonlinear
Sci. Numer. Simul., 41 (2016), 84–103.
[152] J. Leray, “Sur le mouvement d’un liquide visqueux emplissant l’espace”, Acta Math.,
63:1 (1934), 193–248.
[153] Б. М. Левитан, В. В. Жиков, Почти периодические функции и дифференциаль-
ные уравнения, Изд-во Моск. ун-та, М., 1978, 204 с.; англ. пер.: B. M. Levitan,
V. V. Zhikov, Almost periodic functions and differential equations, Cambridge Univ.
Press, Cambridge–New York, 1982, xi+211 pp.
[154] Desheng Li, P. E. Kloeden, “Equi-attraction and the continuous dependence of at-
tractors on parameters”, Glasg. Math. J., 46:1 (2004), 131–141.
[155] E. H. Lieb, “On characteristic exponents in turbulence”, Comm. Math. Phys., 92:4
(1984), 473–480.
[156] E. N. Lorenz, “Deterministic nonperiodic flow”, J. Atmospheric Sci., 20:2 (1963),
130–141.
[157] Songsong Lu, “Attractors for nonautonomous 2D Navier–Stokes equations with less
regular normal forces”, J. Differential Equations, 230:1 (2006), 196–212.
[158] Songsong Lu, “Attractors for nonautonomous reaction-diffusion systems with sym-
bols without strong translation compactness”, Asymptot. Anal., 54:3-4 (2007),
197–210.
[159] Songsong Lu, Hongqing Wu, Chengkui Zhong, “Attractors for nonautonomous 2D
Navier–Stokes equations with normal external forces”, Discrete Contin. Dyn. Syst.,
13:3 (2005), 701–719.
[160] Shan Ma, Xiyou Cheng, Hongtao Li, “Attractors for non-autonomous wave equations
with a new class of external forces”, J. Math. Anal. Appl., 337:2 (2008), 808–820.
[161] Shan Ma, Chengkui Zhong, “The attractors for weakly damped non-autonomous
hyperbolic equations with a new class of external forces”, Discrete Contin. Dyn.
Syst., 18 (2007), 53–70.
[162] Shan Ma, Chengkui Zhong, Haitao Song, “Attractors for nonautonomous 2D Na-
vier–Stokes equations with less regular symbols”, Nonlinear Anal., 71:9 (2009),
4215–4222.
194 S. V. ZELIK

[163] J. Málek, J. Nečas, “A finite-dimensional attractor for three-dimensional flow of in-

compressible fluids”, J. Differential Equations, 127:2 (1996), 498–518.
[164] J. Málek, D. Pražák, “Large time behavior via the method of ℓ-trajectories”, J. Dif-
ferential Equations, 181:2 (2002), 243–279.
[165] J. Mallet-Paret, “Negatively invariant sets of compact maps and an extension
of a theorem of Cartwright”, J. Differential Equations, 22:2 (1976), 331–348.
[166] J. Mallet-Paret, G. Sell, “Inertial manifolds for reaction diffusion equations in higher
space dimensions”, J. Amer. Math. Soc., 1:4 (1988), 805–866.
[167] J. Mallet-Paret, G. Sell, Zhoude Shao, “Obstructions to the existence of normally
hyperbolic inertial manifolds”, Indiana Univ. Math. J., 42:3 (1993), 1027–1055.
[168] R. Mañé, “Reduction of semilinear parabolic equations to finite dimensional C 1
flows”, Geometry and topology, Proc. III Latin Amer. School of Math. (Inst. Mat.
Pura Aplicada CNPq, Rio de Janeiro, 1976), Lecture Notes in Math., 597, Springer,
Berlin, 1977, 361–378.
[169] R. Mañé, “On the dimension of the compact invariant sets of certain non-linear
maps”, Dynamical systems and turbulence, Warwick 1980 (Coventry, 1979/1980),
Lecture Notes in Math., 898, Springer, Berlin–New York, 1981, 230–242.
[170] A. Marzocchi, S. Z. Necca, “Attractors for dynamical systems in topological spaces”,
Discrete Contin. Dyn. Sys., 8:3 (2002), 585–597.
[171] V. S. Melnik, J. Valero, “On attractors of multivalued semi-flows and differential
inclusions”, Set-Valued Anal., 6:1 (1998), 83–111.
[172] A. Mielke, “Essential manifolds for an elliptic problem in an infinite strip”, J. Dif-
ferential Equations, 110:2 (1994), 322–355.
[173] А. Милке, С. В. Зелик, “Бесконечномерные траекторные аттракторы эллип-
тических краевых задач в цилиндрических областях”, УМН, 57:4(346) (2002),
119–150; англ. пер.: A. Mielke, S. V. Zelik, “Infinite-dimensional trajectory attrac-
tors of elliptic boundary-value problems in cylindrical domains”, Russian Math.
Surveys, 57:4 (2002), 753–784.
[174] A. Mielke, S. V. Zelik, “Infinite-dimensional hyperbolic sets and spatio-temporal
chaos in reaction diffusion systems in Rn ”, J. Dynam. Differential Equations, 19:2
(2007), 333–389.
[175] M. Miklavčič, “A sharp condition for existence of an inertial manifold”, J. Dynam.
Differential Equations, 3:3 (1991), 437–456.
[176] J. Milnor, “On the concept of attractor”, Comm. Math. Phys., 99:2 (1985), 177–195.
[177] A. Miranville, S. Zelik, “Exponential attractors for the Cahn–Hilliard equation with
dynamic boundary conditions”, Math. Methods Appl. Sci., 28:6 (2005), 709–735.
[178] A. Miranville, S. Zelik, “Finite-dimensionality of attractors of degenerate equations
of elliptic-parabolic type”, Nonlinearity, 20:8 (2007), 1773–1797.
[179] A. Miranville, S. Zelik, “Attractors for dissipative partial differential equations
in bounded and unbounded domains”, Handbook of differential equations: evolu-
tionary equations, v. IV, Handb. Differ. Equ., Elsevier/North-Holland, Amsterdam,
2008, 103–200.
[180] A. Miranville, S. Zelik, “Doubly nonlinear Cahn–Hilliard–Gurtin equations”,
Hokkaido Math. J., 38:2 (2009), 315–360.
[181] I. Moise, R. Rosa, Xiaoming Wang, “Attractors for non-compact semigroups via
energy equations”, Nonlinearity, 11:5 (1998), 1369–1393.
[182] X. Mora, J. Solà-Morales, “Inertial manifolds of damped semilinear wave equations”,
Attractors, inertial manifolds and their approximation (Marseille–Luminy, 1987),
RAIRO Modél. Math. Anal. Numér., 23:3 (1989), 489–505.
[183] C. G. T. Moreira, M. J. Pacifico, S. R. Ibarra, “Hausdorff dimension, Lagrange and
Markov dynamical spectra for geometric Lorenz attractors”, Bull. Amer. Math. Soc.
(N. S.), 57:2 (2020), 269–292.
 ATTRACTORS. THEN AND NOW 195

[184] E. Olson, E. S. Titi, “Determining modes and Grashof number in 2D turbulence:

a numerical case study”, Theor. Comput. Fluid Dyn., 22 (2008), 327–339.
[185] V. Pata, S. Zelik, “A result on the existence of global attractors for semigroups
of closed operators”, Commun. Pure Appl. Anal., 6:2 (2007), 481–486.
[186] M. Pierre, D. Schmidt, “Blowup in reaction-diffusion systems with dissipation
of mass”, SIAM Rev., 42:1 (2000), 93–106.
[187] I. Prigogine, Time, structure and fluctuations, Nobel lecture, 1977, 23 pp.,
[Link]
[188] E. Priola, Lihu Xu, J. Zabczyk, “Exponential mixing for some SPDEs with Lévy
noise”, Stoch. Dyn., 11:2-3 (2011), 521–534.
[189] D. Rand, “The topological classification of Lorenz attractors”, Math. Proc. Cam-
bridge Philos. Soc., 83:3 (1978), 451–460.
[190] I. Richards, “On the gaps between numbers which are sums of two squares”, Adv.
in Math., 46:1 (1982), 1–2.
[191] А. Робертсон, В. Робертсон, Топологические векторные пространства, Мир,
М., 1967, 257 с.; пер. с англ.: A. P. Robertson, W. J. Robertson, Topological vector
spaces, Cambridge Tracts in Math. and Math. Phys., 53, Cambridge Univ. Press,
New York, 1964, viii+158 pp.
[192] J. C. Robinson, Infinite-dimensional dynamical systems. An introduction to dissi-
pative parabolic PDEs and the theory of global attractors, Cambridge Texts Appl.
Math., Cambridge Univ. Press, Cambridge, 2001, xviii+461 pp.
[193] J. C. Robinson, Dimensions, embeddings, and attractors, Cambridge Tracts in Math.,
186, Cambridge Univ. Press, Cambridge, 2011, xii+205 pp.
[194] J. C. Robinson, J. L. Rodrigo, W. Sadowski, The three-dimensional Navier–Stokes
equations. Classical theory, Cambridge Stud. Adv. Math., 157, Cambridge Univ.
Press, Cambridge, 2016, xiv+471 pp.
[195] А. В. Романов, “Точные оценки размерности инерциальных многообразий для
нелинейных параболических уравнений”, Изв. РАН. Сер. матем., 57:4 (1993),
36–54; англ. пер.: A. V. Romanov, “Sharp estimates of the dimension of inertial
manifolds for nonlinear parabolic equations”, Russian Acad. Sci. Izv. Math., 43:1
(1994), 31–47.
[196] А. В. Романов, “О предельной динамике эволюционных уравнений”, УМН,
51:2(308) (1996), 173–174; англ. пер.: A. V. Romanov, “On the limit dynamics
of evolution equations”, Russian Math. Surveys, 51:2 (1996), 345–346.
[197] А. В. Романов, “Три контрпримера в теории инерциальных многообразий”, Ма-
тем. заметки, 68:3 (2000), 439–447; англ. пер.: A. V. Romanov, “Three counter-
examples in the theory of inertial manifolds”, Math. Notes, 68:3 (2000), 378–385.
[198] А. В. Романов, “Конечномерность динамики на аттракторе для нелинейных па-
раболических уравнений”, Изв. РАН. Сер. матем., 65:5 (2001), 129–152; англ.
пер.: A. V. Romanov, “Finite-dimensional dynamics on attractors of non-linear
parabolic equations”, Izv. Math., 65:5 (2001), 977–1001.
[199] R. Rosa, R. Temam, “Inertial manifolds and normal hyperbolicity”, Acta Appl.
Math., 45:1 (1996), 1–50.
[200] J. Le Rousseau, G. Lebeau, “On Carleman estimates for elliptic and parabolic op-
erators. Applications to unique continuation and control of parabolic equations”,
ESAIM Control Optim. Calc. Var., 18:3 (2012), 712–747.
[201] У. Рудин, Функциональный анализ, Мир, М., 1975, 443 с.; пер. с англ.: W. Rudin,
Functional analysis, 2nd ed., Internat. Ser. Pure Appl. Math., New York, 1991,
xviii+424 pp.
[202] D. Ruelle, “Small random perturbations of dynamical systems and the definition
of attractors”, Comm. Math. Phys., 82:1 (1981), 137–151.
196 S. V. ZELIK

[203] T. Sauer, J. A. Yorke, M. Casdagli, “Embedology”, J. Statist. Phys., 3-4 (1991),

579–616.
[204] A. Savostianov, S. Zelik, “Finite dimensionality of the attractor for the hyper-
bolic Cahn–Hilliard–Oono equation in R3 ”, Math. Methods Appl. Sci., 39:5 (2016),
1254–1267.
[205] А. К. Савостьянов, С. В. Зелик, “Равномерные аттракторы для волнового урав-
нения с нелинейностью пятой степени и мерой в качестве внешней силы”, УМН,
75:2(452) (2020), 61–132; англ. пер.: A. K. Savostianov, S. V. Zelik, “Uniform
attractors for measure-driven quintic wave equations”, Russian Math. Surveys, 75:2
(2020), 253–320.
[206] M. Scheutzow, “Comparison of various concepts of a random attractor: a case study”,
Arch. Math. (Basel ), 78:3 (2002), 233–240.
[207] A. Segatti, S. Zelik, “Finite-dimensional global and exponential attractors for
the reaction-diffusion problem with an obstacle potential”, Nonlinearity, 22:11
(2009), 2733–2760.
[208] G. R. Sell, “Global attractors for the three-dimensional Navier–Stokes equations”,
J. Dynam. Differential Equations, 8:1 (1996), 1–33.
[209] G. R. Sell, Yuncheng You, Dynamics of evolutionary equations, Appl. Math. Sci.,
143, Springer-Verlag, New York, 2002, xiv+670 pp.
[210] L. P. Shilnikov, A. L. Shilnikov, D. V. Turaev, L. O. Chua, Methods of qualitative
theory in nonlinear dynamics, Part II, World Sci. Ser. Nonlinear Sci. Ser. A Monogr.
Treatises, 5, World Sci. Publ., River Edge, NJ, 2001, i–xxiv and 393–957.
[211] A. Shirikyan, S. Zelik, “Exponential attractors for random dynamical systems and
applications”, Stoch. Partial Differ. Equ. Anal. Comput., 1:2 (2013), 241–281.
[212] D. W. Stroock, Probability theory. An analytic view, Cambridge Univ. Press, Cam-
bridge, 1993, xvi+512 pp.
[213] F. Takens, “Detecting strange attractors in turbulence”, Dynamical systems and
turbulence, Warwick 1980 (Coventry, 1979/1980), Lecture Notes in Math., 898,
Springer, Berlin–New York, 1981, 366–381.
[214] R. Temam, Navier–Stokes equations and nonlinear functional analysis, CBMS-NSF
Regional Conf. Ser. in Appl. Math., 41, Society for Industrial and Applied Mathe-
matics (SIAM), Philadelphia, PA, 1983, xii+122 pp.
[215] R. Temam, Infinite-dimensional dynamical systems in mechanics and physics, 2nd
ed., Appl. Math. Sci., 68, Springer-Verlag, New York, 1997, xxii+648 pp.
[216] R. Temam, Shou Hong Wang, “Inertial forms of Navier–Stokes equations on
the sphere”, J. Funct. Anal., 117:1 (1993), 215–242.
[217] Х. Трибель, Теория интерполяции, функциональные пространства, дифферен-
циальные операторы, Мир, М., 1980, 664 с.; пер. с англ.: H. Triebel, Interpolation
theory, function spaces, differential operators, North-Holland Math. Library, 18,
North-Holland Publishing Co., Amsterdam–New York, 1978, 528 pp.
[218] W. Tucker, “A rigorous ODE solver and Smale’s 14th problem”, Found. Comput.
Math., 2:1 (2002), 53–117.
[219] D. Turaev, “On dimension of non-local bifurcational problems”, Internat. J. Bifur.
Chaos Appl. Sci. Engrg., 6:5 (1996), 919–948.
[220] D. Turaev, S. Zelik, “Homoclinic bifurcations and dimension of the attractors for
damped nonlinear hyperbolic equations”, Nonlinearity, 16:6 (2003), 2163–2198.
[221] D. Turaev, S. Zelik, “Analytical proof of space-time chaos in Ginzburg–Landau equa-
tions”, Discrete Contin. Dyn. Syst., 28:4 (2010), 1713–1751.
[222] J. Valero, “On the Kneser property for some parabolic problems”, Topology Appl.,
153:5-6 (2005), 975–989.
 ATTRACTORS. THEN AND NOW 197

[223] M. I. Vishik, Asymptotic behaviour of solutions of evolutionary equations, Lezioni

Lincee, Cambridge Univ. Press, Cambridge, 1992, vi+155 pp.
[224] М. И. Вишик, В. В. Чепыжов, “Траекторные аттракторы уравнений математи-
ческой физики”, УМН, 66:4(400) (2011), 3–102; англ. пер.: M. I. Vishik, V. V. Che-
pyzhov, “Trajectory attractors of equations of mathematical physics”, Russian Math.
Surveys, 66:4 (2011), 637–731.
[225] М. И. Вишик, С. В. Зелик, “Траекторный аттрактор нелинейной эллиптической
системы в цилиндрической области”, Матем. сб., 187:12 (1996), 21–56; англ.
пер.: M. I. Vishik, S. V. Zelik, “The trajectory attractor of a non-linear elliptic system
in a cylindrical domain”, Sb. Math., 187:12 (1996), 1755–1789.
[226] М. И. Вишик, С. В. Зелик, “Регулярный аттрактор нелинейной эллиптической
системы в цилиндрической области”, Матем. сб., 190:6 (1999), 23–58; англ.
пер.: M. I. Vishik, S. V. Zelik, “Regular attractor for a non-linear elliptic system
in a cylindrical domain”, Sb. Math., 190:6 (1999), 803–834.
[227] М. И. Вишик, С. В. Зелик, В. В. Чепыжов, “Сильный траекторный аттрактор
диссипативной системы реакции-диффузии”, Докл. РАН, 435:2 (2010), 155–159;
англ. пер.: M. I. Vishik, S. V. Zelik, V. V. Chepyzhov, “Strong trajectory attractor
for a dissipative reaction-diffusion system”, Dokl. Math., 82:3 (2010), 869–873.
[228] М. И. Вишик, С. В. Зелик, В. В. Чепыжов, “Регулярные аттракторы и их неав-
тономные возмущения”, Матем. сб., 204:1 (2013), 3–46; англ. пер.: M. I. Vishik,
S. V. Zelik, V. V. Chepyzhov, “Regular attractors and nonautonomous perturbations
of them”, Sb. Math., 204:1 (2013), 1–42.
[229] A. I. Volpert, Vit. A. Volpert, Vl. A. Volpert, Traveling wave solutions of parabolic
systems, Transl. from the Russian manuscript, Transl. Math. Monogr., 140, Amer.
Math. Soc., Providence, RI, 1994, xii+448 pp.
[230] J. Vukadinovic, “Inertial manifolds for a Smoluchowski equation on a circle”, Non-
linearity, 21:7 (2008), 1533–1545.
[231] J. Vukadinovic, “Inertial manifolds for a Smoluchowski equation on the unit sphere”,
Comm. Math. Phys., 285:3 (2009), 975–990.
[232] J. Vukadinovic, “Global dissipativity and inertial manifolds for diffusive Burg-
ers equations with low-wavenumber instability”, Discrete Contin. Dyn. Syst., 29:1
(2011), 327–341.
[233] Xingxing Wang, Hongyong Cui, “On the residual continuity of global attractors”,
Mathematics, 10:9 (2022), 1444, 7 pp.
[234] Yangmin Xiong, A. Kostianko, Chunyou Sun, S. Zelik, Entropy estimates for uniform
attractors of dissipative PDEs with non translation-compact external forces, 2022,
32 pp., arXiv: 2202.05590.
[235] Lai-Sang Young, “What are SRB measures, and which dynamical systems have
them?”, J. Statist. Phys., 108:5-6 (2002), 733–754.
[236] С. В. Зелик, Траекторный аттрактор нелинейной эллиптической системы
в неограниченной области, Деп. в ВИНИТИ, № 3328-В 97, 1997.
[237] С. В. Зелик, “Траекторный аттрактор нелинейной эллиптической системы в не-
ограниченной области”, Матем. заметки, 63:1 (1998), 135–138; англ. пер.:
S. V. Zelik, “A trajectory attractor of a nonlinear elliptic system in an unbounded
domain”, Math. Notes, 63:1 (1998), 120–123.
[238] S. Zelik, “The attractor for a nonlinear reaction-diffusion system with a supercritical
nonlinearity and its dimension”, Rend. Accad. Naz. Sci. XL Mem. Mat. Appl. (5),
24 (2000), 1–25.
[239] С. В. Зелик, “Многопараметрические полугруппы и аттракторы уравнений реак-
ции-диффузии в Rn ”, Тр. ММО, 65, УРСС, М., 2004, 114–174; англ. пер.:
S. Zelik, “Multiparameter semigroups and attractors of reaction-diffusion equations
in Rn ”, Trans. Moscow Math. Soc., 65 (2004), 105–160.
198 S. V. ZELIK

[240] S. Zelik, “Asymptotic regularity of solutions of a nonautonomous damped wave

equation with a critical growth exponent”, Commun. Pure Appl. Anal., 3:4 (2004),
921–934.
[241] S. Zelik, “Asymptotic regularity of solutions of singularly perturbed damped wave
equations with supercritical nonlinearities”, Discrete Contin. Dyn. Syst., 11:2-3
(2004), 351–392.
[242] S. Zelik, “On the Lyapunov dimension of cascade systems”, Commun. Pure Appl.
Anal., 7:4 (2008), 971–985.
[243] S. Zelik, “Inertial manifolds and finite-dimensional reduction for dissipative PDEs”,
Proc. Roy. Soc. Edinburgh Sect. A, 144:6 (2014), 1245–1327.
[244] S. Zelik, “Strong uniform attractors for non-autonomous dissipative PDEs with
non translation-compact external forces”, Discrete Contin. Dyn. Syst. Ser. B, 20:3
(2015), 781–810.
[245] S. Zelik, A. Mielke, Multi-pulse evolution and space-time chaos in dissipative systems,
Mem. Amer. Math. Soc., 198, № 925, Amer. Math. Soc., Providence, RI, 2009,
vi+97 pp.
[246] Caidi Zhao, Shengfan Zhou, “Pullback trajectory attractors for evolution equa-
tions and application to 3D incompressible non-Newtonian fluid”, Nonlinearity, 21:8
(2008), 1691–1717.
[247] Shengfan Zhou, “Random exponential attractor for cocycle and application to non-
autonomous stochastic lattice systems with multiplicative white noise”, J. Differen-
tial Equations, 263:4 (2017), 2247–2279.
[248] В. Г. Звягин, С. К. Кондратьев, “Аттракторы уравнений неньютоновской гидро-
динамики”, УМН, 69:5(419) (2014), 81–156; англ. пер.: V. G. Zvyagin, S. K. Kond-
rat’ev, “Attractors of equations of non-Newtonian fluid dynamics”, Russian Math.
Surveys, 69:5 (2014), 845–913.

Sergey V. Zelik Поступила в редакцию

(Сергей Витальевич Зелик) 25.08.2022
Zhejiang Normal University,
Department of Mathematics, Zhejiang, P. R. China;
University of Surrey, Department of Mathematics,
Guildford, United Kingdom;
Keldysh Institute of Applied Mathematics
of Russian Academy of Sciences,
Moscow, Russia
E-mail : [Link]@[Link]