Chaos synchronization: a review

Rosário D. Laureano, Diana A. Mendes, Manuel Alberto M. Ferreira

ISCTE - Instituto Universitário de Lisboa

ABSTRACT: This article provides a self-contained comprehensive review of the

arXiv:2412.20649v1 [nlin.CD] 30 Dec 2024

phenomenon of synchronization in dynamical systems, with a particular focus on chaotic

systems in both continuous-time and discrete-time contexts.
Synchronization, initially observed by Christiaan Huygens in 1665, has evolved from
the study of periodic signals to encompass chaotic systems, where the sensitive depen-
dence on initial conditions poses unique challenges. This review pointed to both the-
oretical foundations and contributions (concepts and methods) and practical insights,
reinforcing the relevance of chaos synchronization in physics, biology, engineering, social
sciences, economics and communication systems. The study investigates various coup-
ling schemes, such as unidirectional and bidirectional coupling, and presents stability
criteria under different configurations.
In a very concise way, some ongoing research carried out by the authors is also in-
dicated, using Lorenz, Rössler and hyperchaotic Rössler systems, and quadratic maps
as case studies with parameter values that lead to chaotic behaviour. Special atten-
tion is given to the stability of synchronized states and the role of multi-stability and
bifurcations, and its implications to loss of synchronization. We highlight the role of
Lyapunov exponents, eigenvalues, and Lyapunov functions in guaranteeing local and
global stability of the synchronized state.
We aim to contribute to a broader understanding of chaos synchronization and its
practical applications in diverse fields of knowledge. This text shed light on the control
and stability of coupled chaotic systems, offering new perspectives on the synchronization
of non-identical systems and the emergence of complex synchronization dynamics.
Keywords: Chaos synchronization, synchronization error, synchronization manifold,
coupling schemes, asymmetric coupling, Lyapunov exponents, Lyapunov function
The first observation of synchronization was described by the Dutch physicist Chris-
tiaan Huygens in 1665 [1]. In this case, the phenomenon was evidenced by the equality
of periods in mutually coupled clocks. Later, synchronization of periodic signals was
detected in many other dynamical processes, which became common in other scientific
areas and the subject of diverse applications in engineering. Given this original dis-
covery by Huygens, the synchronization phenomenon is, in the classical sense of the
term, associated with the ability of self-oscillatory dissipative systems to adjust their

1
behaviours to follow a global periodic motion. It should be noted that the study of
dynamical systems for a long time was based on examples of differential equations with
regular solutions. If these solutions remained in a bounded region of the phase space,
then they would adjust to one of two types of behaviour, a stable equilibrium point or
a periodic or quasi-periodic oscillation.
Nowadays, the term "synchronization" is used in a broader sense, including chaotic
systems, both discrete and continuous. Numerical analysis of chaotic populations present
in nature, even heterogeneous ones, reveals that synchronization is possible and frequent
but is just an example.

1 Chaotic behaviour
Chaos belongs to the vast field of the theory of non-linear oscillations, whose significant
development began in the past century. Although in the 1920s and 1930s, Andronov pro-
posed a classification of non-linear behaviours (1933), and Van der Pol experimentally
detected noise-like oscillations in electronic circuits (1927), the experience that drove
the consideration of chaotic behaviour is attributed to Lorenz [2]. In 1961, working
on a simplified model of atmospheric transfer with three simple non-linear differential
equations, he numerically observed that making very small changes to the initial condi-
tions had a huge effect on his solutions. This was evidence of one of the fundamental
properties of chaotic dynamics, which would later be termed sensitive dependence on
initial conditions. It should be noted that this property had already been investigated
from a topological perspective by Poincaré and described in his monograph Science and
Method in 1903.
The sensitivity to initial conditions implies that an infinitesimal perturbation in the
initial conditions of a dynamical system, in either discrete or continuous time, leads to
exponential divergence of initially closed orbits over time. For many years, this property
made chaos undesirable as it reduces the predictability of a chaotic system over long
periods. However, the scientific community gradually became aware of a third type of
behaviour in dynamics: chaotic behaviour. Some experiments, whose anomalous results
had previously been explained in terms of experimental error or introduced noise, were
re-evaluated and explained in terms of chaos, which then became the subject of rigorous
mathematical study. A measure of sensitivity to initial conditions can be quantified by
the Lyapunov exponents. However, the topological entropy and the metric entropy are
quantities that, like the Lyapunov exponents, also measure the complexity of a system’s
dynamics.
In the mid-1970s, the term "deterministic chaos" was introduced by Li and Yorke [3]
in a famous article titled Period three implies chaos. This article presents the study of
possible periods in periodic points of real continuous maps defined on an interval and
proves that if a map has a periodic orbit of period three, then it is chaotic. However,
this result can be considered a corollary of a stronger result by Sharkowski [4] presented
in a 1964 article written in Russian. This result, concerning the existence of periodic
points for continuous maps, is based on a certain order of natural numbers now known

2
as the Sharkowski order. Although there is no unique formal definition of the term
"deterministic chaos", chaotic behaviour can be defined as an observable pattern that
appears irregular and unpredictable over large time scales.
Many non-linear dynamical systems are dependent on control parameters in a certain
parametric space. For application purposes, it is crucial to understand the evolution of
the qualitative behaviour of elements of a family of dynamical systems as the parameters
vary. The qualitative change in the behaviour of dynamical systems induced by the
variation of one of its parameters is now known as bifurcation. The Poincaré’s work —
to describe the separation of equilibrium solutions in a family of differential equations —
is pioneering in the importance given to the qualitative/topological changes of dynamical
behaviour and in the introduction of the term "bifurcation". However, there is still no
consensus on the rigorous meaning of this term. It should be noted that bifurcation
theory has its roots in classical mathematics, as seen in Euler’s work in the 18th century.
To clarify the possible "routes to chaos", the article [5] by May, published in 1976 in the
journal Nature, is highly relevant, as it describes the period-doubling bifurcation in the
logistic map.
Geometrically, an attractor can be a single point, a curve, a manifold, or even a
set with a fractal structure known as a strange attractor. The presence of a chaotic
attractor in phase space, which typically has an infinite and dense set of unstable periodic
orbits embedded within it, implies that it is impossible to determine the position of the
system on the attractor over time, even knowing its position on this same attractor
at an earlier moment. The unstable periodic orbits embedded in a chaotic attractor
constitute a dynamical invariant of the system, and their systematic investigation allows
for the characterization and estimation of many other dynamical invariants, such as
the natural invariant measure, the spectrum of Lyapunov exponents, and the fractal
dimensions. The relationship between the trajectories of the chaotic attractor and the
unstable periodic orbits embedded within it has been explored in detail for hyperbolic
dynamical systems, for which the separation into stable and unstable invariant subspaces
is consistent (robust) under the evolution of dynamics [6], and for non-hyperbolic systems
with homoclinic tangencies. The infinite set of unstable periodic orbits embedded in a
chaotic attractor located on some symmetric invariant manifold plays a fundamental role
in the destabilization mechanism of that attractor, as it is responsible for the dynamics
of phenomena such as the riddling of the basin of attraction and the bubbling of the
chaotic attractor ([7],[8]).
It may happen that, for different regions of the system’s parametric space, different
attractors (multiple attractors) exist, and as such, even the qualitative asymptotic be-
haviour of the system depends significantly on the initial conditions established. Even
in systems with symmetry, the phenomenon of multi-stability can occur, where the co-
existence of more than one attractor for a given set of parameters is observed. In this
case, the qualitative asymptotic behaviour of the system cannot be predicted, as for each
initial condition, it is not known a priori in which attractor the system’s dynamics will
eventually stabilize.

3
2 Chaos synchronization
Chaotic behaviour can be observed in natural systems, in experimental settings, and
in computational models from many scientific fields, revealing itself as a robust phe-
nomenon. The capacity of chaotic dynamics to amplify small perturbations enhances its
use in achieving specific desired states in a chaotic system, without substantially altering
its main dynamical properties and with a small energy expenditure. The developments
in control theory and chaos synchronization theory are a consequence of this. These two
processes (phenomena) have common roots. In both processes, the underlying idea is to
intervene in a non-linear chaotic system by choosing parametric regions to perturb or
by applying external control to restrict its movement to an accessible subspace of the
phase space. In chaos control, the aim is to shift the system to an accessible invariant
set where movements are more regular, while in chaos synchronization, the goal is to
obtain an invariant set in the phase space of the coupled system - the synchronization
subspace - where synchronous movement takes place. Although the concept of chaos
synchronization has generally evolved autonomously, there is a recent trend to unify the
study of chaos control and chaos synchronization under the same framework.
The dynamics of a system exhibit chaotic behaviour when it never repeats itself,
and even if initial conditions are correlated by proximity, the corresponding trajecto-
ries quickly become uncorrelated. As such, the possibility of achieving synchronization
of chaotic systems through appropriate coupling may seem impossible. Intuitively, we
might say that the terms "chaos" and "synchronization" are mutually exclusive. Even
when tracking the evolution of two identical chaotic systems that start from nearby
points in phase space but are not exactly identical, we find that they diverge in their be-
haviours, although both retain the same chaotic attractor pattern. Two chaotic processes
are observed without any mutual correlation (independent).
On the other hand, the possibility of qualitative/topological changes in the dynamical
behaviour of elements in a family of dynamical systems, indicated by possible bifurcation
points, complicates the synchronization of non-identical systems, even if they differ only
by small parametric mismatches. An infinitesimal deviation in any of the parameters
can lead to qualitatively distinct dynamics.
Moreover, two already synchronized trajectories may lose the stability of synchroni-
zation due to the influence of external noise. However, due to the ergodic property of
chaotic trajectories, after a finite transient period, they will once again become close
and may resynchronize. In this sense, the method of synchronization is robust against
small external noise.
Studies conducted over the last three decades have shown that the intuitive idea that
chaotic systems cannot behave in synchrony is false. In fact, although it is impossible
to exactly reproduce identical initial conditions and equal parameters, there are sets of
coupled chaotic oscillators in which the attractive effect of a sufficiently strong coupling
can cancel/compensate for the tendency of trajectories to diverge due to chaotic dy-
namics. As a result, it is possible to achieve total synchronization in chaotic systems as
long as they are coupled by an appropriate dissipative coupling. Synchronizing chaotic
systems, therefore, means coupling them in such a way as to "force" them to follow the

4
same trajectory on the chaotic attractor by applying small perturbations between the
systems.
In the synchronization of two chaotic systems, it is possible to consider two coupling
configurations: unidirectional (or directional) and bidirectional (mutual or global). In
unidirectional coupling, one of the systems, referred to as the driver or master, influences
the other, referred to as the response or slave, without the latter providing feedback to
the former. Bidirectional coupling implies mutual interaction between the systems.
Coupled chaotic dynamical systems are constructed from simple, low-dimensional ele-
ments to form new, more complex organizations, ensuring that the dominant charac-
teristics of the underlying individual components are preserved. This "cumulative"
construction can also be used to create a new system whose behaviour is more flexi-
ble and/or richer than that of its components, but whose analysis and control remain
accessible. The concept that several systems, when interacting non-linearly, collectively
give rise to new dynamics that cannot be attributed to the individual components is
known as emergence.
It is not possible to predict in advance the consequences of a coupling. The introduc-
tion of coupling in chaotic systems can drastically change the qualitative properties of
dynamics. It can stabilize into periodic behaviour, it can occasionally produce hidden
correlations between the elements (even when the dynamics seem turbulent), or it can
induce the synchronization of a subset of the dynamical variables. The effectiveness of a
coupling between systems of equal dimension depends, first of all, on the analysis of the
difference between the respective coordinates of the variables of the systems involved,
which can be referred to as the synchronization error. In an optimal situation, coupling
chaotic systems leads to their asymptotic synchronization, where the synchronization
error converges to zero over time. However, a less demanding form of synchronization
was recently introduced by Stefański and Kapitaniak [9], where only stabilization of the
synchronization error below a constant of less than one is expected. This is known as
practical synchronization in the sense of Kapitaniak. It is noted that chaotic dynamics
introduce new degrees of freedom into sets of coupled systems. However, when two or
more chaotic oscillators are coupled and synchronization is achieved, the number of dy-
namical degrees of freedom for the coupled system generally decreases. Nevertheless, the
specificities of chaotic behaviour make it impossible to directly apply methods developed
for the synchronization of periodic oscillations to the synchronization of chaotic systems.

2.1 Identical synchronization

The synchronization of chaos began in the mid-1980s with the studies of Kaneko [10]
and Afraimovich et al. [11] on the coupling of identical discrete and continuous systems,
respectively, evolving from different initial conditions. It is worth noting, in the same
decade, the works of Fujisaka and Yamada ([12],[13]), who introduced the study of trans-
verse Lyapunov exponents in the coupled system, emphasizing how the dynamics can
change with the coupling, and the works of Pikovsky ([14],[15]) and Afraimovich et al.
[11], who presented many of the fundamental concepts in chaos synchronization. It was
only in the 1990s, with the work of Pecora and Carroll ([16],[17],[18]), that chaos syn-

5
chronization was consolidated as an autonomous research topic, alongside the rigorous
establishment of chaos control theory by Ott, Grebogy and Yorke [19]. The aforemen-
tioned articles immediately attracted significant attention from the scientific community
and triggered the development of several synchronization methods.
In communication systems, especially those involving signals whose future behaviour
is difficult (or impossible) to predict during transmission, some form of synchronization
between the transmitter and receiver is evidently useful. This fact motivated the work
of Pecora and Carroll ([16],[17],[18]), where a method of synchronization is established
by coupling two identical chaotic dynamical systems through transmission of a driving
subsystem. This subsystem acts as a common chaotic signal between them. Thus, the
method of Pecora and Carroll, also known as complete replacement, suggests how a
synchronous chaotic state can be used as a driver in communication. Therefore, given
a chaotic system, synchronization by this method requires decomposing the system to
obtain a suitable driving subsystem. Usually, various combinations of a subset of state
variables are tested to identify a stable driving subsystem. It may seem counterintuitive
that a non-dissipative system could lead to synchronization, but in a multidimensional
volume-preserving dynamical system, there must be at least one contracting direction
so that phase-space volumes are conserved, which allows for the selection of a stable
subsystem. Given the possibility of synchronizing two chaotic systems, it is necessary
to determine the conditions under which such synchronization is stable. In addition to
establishing the coupling mechanism, which is relatively straightforward but deceptively
simple, Pecora and Carroll ([16],[17]) also provide the first answers to this question.
Although in complete replacement there is only a finite number of possible decom-
positions of a chaotic system, the core idea of Pecora and Carroll to decompose chaotic
systems to obtain a driving subsystem has led to other, more general methods of syn-
chronization that do not require decomposing the original chaotic system into two stable
subsystems. Kocarev and Parlitz ([20],[21]) proposed the active-passive decomposition
method by decomposing the driver system into two components: an active component
and a passive one. A scalar signal from the driver system is transmitted to the response
system, and this can be a function of an information signal. Different replicas of the
passive component synchronize when driven by the same active component. According
to Stefański and Kapitaniak [22], the chaotic trajectory of a system can synchronize with
the chaotic trajectory of an identical system by adding a linear damping term propor-
tional to the difference between the corresponding state variables of the two systems.
This term acts as a control signal applied only to the response system in the form of
negative feedback, and as such the coupling is referred to as negative feedback control.
The application of this method to different experimental models by authors such as Lai
and Grebogy [23], Kapitaniak [24], John and Amritkar [25], Anishchenko et al. [26],
Ding and Ott [27], Pyragas [28] and Wu and Chua [29] shows that it is effective when
the coupled system has a single attractor.
According to Fujisaka and Yamada ([12], [13]), a natural way to introduce bidirectional
coupling between two identical chaotic systems is to add symmetric linear coupling terms
to the expressions that define them. This coupling mechanism, which can be total or
partial, is called linear diffusive coupling. A study by Stefański [30] shows that the

6
properties of exponential divergence and convergence in total coupling make it possible
to estimate the largest Lyapunov exponent of any chaotic dynamical system, a possibility
that is particularly useful in non-smooth systems, where the estimation of Lyapunov
exponents is not straightforward.
As discussed, there are several coupling configurations between identical chaotic sys-
tems so that they become synchronized. In this regime, known as identical synchroni-
zation, the chaotic trajectories of the coupled identical chaotic systems coincide exactly
over time due to the strong interaction between them. Each of the systems maintains
its chaotic behaviour, but when the symmetric synchronous state is achieved - that is,
when the evolution of their state vectors coincides - the dynamics of the coupled system
are restricted to a synchronization hyperplane in the phase space.
In the coupling of discrete chaotic systems, parameters are considered that control
the coupling strength between the systems, as in linear diffusive coupling and negative
feedback control coupling of continuous systems. Depending on the structure of the
coupling terms involving these parameters, several coupling schemes in discrete time
are distinguished: external or internal dissipative coupling, linear diffusive coupling,
quadratic coupling terms, or bilinear coupling. The stability results for the chaotic
synchronous state depend on the coupling parameters considered.

2.2 Generalized synchronization

Although the phenomenon of chaos synchronization was detected from examples in-
volving identical chaotic systems, the behaviour of these systems in synchrony represents
only a sample of the abundance of different types of interdependent behaviour that can
be detected in coupled chaotic systems. While the identical synchronization regime is
the most common and has the largest number of theoretical results, many studies have
revealed that it is also possible to synchronize non-identical systems, defined by evolu-
tion laws that differ only by small parametric mismatches or that are even distinct (and
may even differ in dimension). In this case, the synchronization is called generalized
synchronization, either in a bidirectional or even unidirectional configuration. Many
of the coupling schemes between non-identical chaotic systems are extensions of those
known for identical systems.
The first mathematical definition of synchronized chaos in a generalized sense was
made by Afraimovich et al. [11] in 1986 and is based on the existence of a homeo-
morphism between systems with parametric mismatches that relates the projections of
the synchronized chaotic trajectories onto subspaces of the phase spaces of the drive
and response systems. However, the term "generalized synchronization" was only intro-
duced by Rulkov et al. [31] in 1995. Referring to the synchronization of unidirectionally
coupled periodic systems, the central idea in this article, as well as in Kocarev and
Parlitz’s [32] article from 1996, is to take the ability to predict the current state of the
response system from the chaotic information measured in the drive system as a defini-
tion of generalized synchronization. Predictability points to the existence and stability
of a chaotic attractor in the coupled system.
Thus, most detection methods for generalized synchronization consider it to be re-

7
presented by the existence of a functional relationship between the systems, the more
regular, the better, whose graph contains an invariant and stable manifold, called the
synchronization manifold, within which the chaotic attractor related to synchronization
is embedded. This synchronization regime is weaker than identical synchronization,
since the persistent (robust) and stable dependence between the state vectors of each
system is not necessarily the identity of states. However, in general, the absence of in-
trinsic symmetry in the coupled system makes it difficult to obtain a stable synchronous
state. It is noted that the functional relationship does not necessarily have to be valid
throughout the phase space of the coupled systems but only on the invariant manifold.
It should be noted that the original definition of Afraimovich et al. [11], although it
allows for a set of analytical results on the stability of synchronous states, is not par-
ticularly satisfactory as it does not refer to the attractor nature of the synchronization
set and requires the verification of conditions whose validity in real experiments cannot
always be demonstrated. On the other hand, the definition presented by Rulkov et al.
[31] encompasses situations in physics, biology, and economics, in which chaos synchro-
nization has been detected, where the requirement of a homeomorphism between the
projections was not met. However, for systems with invertible dynamics, Rulkov et al.’s
definition [31] is equivalent to the existence of a continuous function between the states
of the systems when they are on the synchronized chaotic attractor.
Studies show that the response system is asymptotically stable whenever there is
a function that transforms each trajectory on the attractor of the drive system into
a trajectory on the attractor of the response system, known as the synchronization
function. In this case, the synchronized chaotic trajectories are located on a stable
synchronization manifold. Based on the equivalence between generalized synchronization
in the coupled system and the asymptotic stability of the response system, Abarbanel
et al. [33] established a criterion for detecting generalized synchronization, called the
auxiliary system approach.
Generalized synchronization includes identical synchronization as in a particular case,
in which the functional relationship is the identity function, and the synchronization
manifold is a hyperplane. However, while this is easily visualized in the representation
of the difference between the coordinates of the two coupled systems, the detection of
generalized synchronization does not follow a simple method, especially when analy-
zing information obtained experimentally. Except in special cases of coupling between
systems with small parametric mismatches, it is rarely possible to provide explicit for-
mulas for the synchronization function or to have a trivial synchronization manifold in
phase space. In general, the synchronized chaotic oscillations are different from those
generated by the uncoupled chaotic systems. Therefore, the analogy between the syn-
chronized chaotic attractor and the chaotic attractors of the uncoupled systems cannot
be considered a requirement to define generalized synchronization.
The asymptotic stability of the response system does not guarantee that the synchro-
nization function is continuous, nor even the existence of a synchronization function
between the systems. Experimental situations have been observed in which the response
system is asymptotically stable, but the chaotic attractor of the coupled system has a
complex structure, and the synchronization function is not differentiable. The dynamics

8
on the synchronization manifold are generally quite complex, due to the lack of sym-
metry in the coupled system or the non-invertibility of the drive system. Contrary to
what is observed in identical synchronization, where the trajectories are attracted to the
symmetry plane and the synchronization manifold is trivial, many real systems exhibit
synchronization subspaces with non-trivial geometric structures inherent to the coupled
system: roughness, cusps, or bands, which may coexist in the same system and have
a detrimental effect on the detection of synchronization. Various existing methods for
detecting generalized synchronization, presented in [31], [34], and [35], are hindered by
the presence of such structures.
The occurrence of roughness is generally caused by the existence of invariant sets
embedded in the synchronization subspace in which the synchronization function has
different degrees of Hölder regularity. The different Hölder exponents, given by the
modulus of the ratio between the Lyapunov exponent relative to the transverse con-
tracting direction and the smallest negative Lyapunov exponent of the driver, depend
on the intensity of the contraction rate in the transverse direction to the synchronization
subspace [36].
The presence of cusps typically results from the existence of critical points in the
attractor of the driver system defined by a smooth, non-invertible map. In the vicinity
of a critical point, where the Jacobian matrix is singular, orbits of the driver system may
exist along which the contraction is arbitrarily large, and the synchronization subspace
is typically non-differentiable near it.
If the driver system is non-invertible, the synchronization function may not be con-
tinuous or may not even exist as a function, as there are typical states of the driver that
have more than one pre-image. In the synchronization subspace, several bands may then
occur, in which case the synchronization function is generally replaced by a multivalued
relationship between the coupled systems, although the response remains asymptotically
stable. In this case, it is not possible to predict a response state from that of the driver.
The detection of the characteristics of generalized synchronization based on experi-
mental information strongly relies on the continuity of the synchronization function and,
in general, also requires a certain degree of smoothness. In this case, the functions in the
coupled system are unknown, and when generalized synchronization is stable, an attempt
can be made, in relatively simple cases, to approximate the synchronization function us-
ing numerical methods. However, if deviations from the functional dependence between
the systems occur, it will never be clear whether these are due to the loss of synchro-
nization in the coupled system or to the inaccuracy of the function considered. Thus,
alternative definitions have emerged that differ in the regularity properties imposed on
the synchronization function and have yielded different results in the detection of gener-
alized synchronization in experimental information. Recently, diffeomorphic properties
have been required by authors such as Abarbanel et al. [33], Pyragas [37], and Hunt et
al. [36]. Pyragas [37] further distinguishes between two types of generalized synchro-
nization: strong synchronization in the case of a smooth synchronization function and
weak synchronization otherwise.

9
2.3 Local and global stability
For generalized synchronization to have practical (applicative) interest, it must persist
under arbitrarily small perturbations, whether of the coupling or of the dynamics of the
component systems.
As in the case of identical synchronization, the stability of the synchronization ma-
nifold can be local - guaranteed by the negativity of the Lyapunov exponents that
characterize the perturbations transverse to the synchronization manifold (transversal
or conditional Lyapunov exponents) [37] and/or by the study of the eigenvalues of the
linearization of the coupled system - or global, guaranteed by the existence of a Lyapunov
function (Lyapunov’s direct method) [32].
The local stability results for synchronization do not guarantee that it will hold when
the coupled system is started from another initial condition. To investigate the possibi-
lity of stable synchronization in the coupled system, the choice of the initial condition
is not irrelevant when there is more than one attractor in the phase space. In contrast
to Lyapunov’s direct method, the study of transversal Lyapunov exponents is quite
straightforward and can be easily applied, even in very complex systems. However,
Stefański and Kapitaniak [9] have pointed out that, in practice, the negativity of Lya-
punov exponents does not always guarantee that there are no unstable invariant sets in
the synchronization manifold that could cause a loss of synchronization when noise or
small parametric mismatches are present. Note that, unlike what happens in unidirec-
tional coupling, in bidirectional coupling, the Lyapunov exponents of one of the systems
are not the same as the exponents that characterize the transverse perturbations. In
unidirectional coupling, the behaviour of the coupled system on the synchronization ma-
nifold is controlled solely by the dynamics of the driver system. When synchronization
is lost, the driver system ceases to have full control over the behaviour of the response
system, and small perturbations in the response system will grow. Although the process
of loss of synchronization is similar to that observed in identical synchronization, identi-
fying bifurcations of the bubbling or blowout type may be hindered by the complexity
of the synchronization subspace. By continuously differentiating between the driver and
the response as the coupling strength decreases, Barreto et al. [38] propose a method
that allows the problem to be addressed using a decomposition based on the identific-
ation of unstable periodic orbits of the driver system. The creation and evolution of
a complicated set of orbits that develop outside the synchronization manifold, known
as the emergent set, are described. A critical transition point for this process is also
identified.

3 Some explorations on chaos synchronization

From this essential support, we are applying identical and generalized synchronization
regimes, in various coupling schemes, to the Lorenz and Rössler systems and logistic
maps, using parameter values that lead to chaotic behaviour. These studies will be made
available in publications that are in progress, but we can give a brief description now.

10
The results obtained provide rigorous criteria for establishing local and global stability
of synchronized states through methods that involve Lyapunov exponents, eigenvalues,
and Lyapunov functions. This approach not only broadens our understanding of exis-
ting coupling mechanisms, but also proposes novel coupling schemes, such as partial
replacements in the Lorenz and Rössler systems, and innovative asymmetric couplings
in quadratic discrete systems. The study of stability transitions and loss of synchroniza-
tion through bubbling and blowout bifurcations further illustrates the delicate balance
required to maintain synchronization in chaotic systems.

3.1 In continuous chaotic dynamical systems

Our research focuses on the possibility of asymptotically synchronizing pairs of continu-
ous chaotic systems, either by unidirectional or bidirectional coupling, using Lorenz,
Rössler and hyperchaotic Rössler systems.
In unidirectional coupling, we explore coupling schemes by total or partial replace-
ment, active-passive decomposition, negative feedback control, and singular value de-
composition, indicating their respective stability criteria. We are investigating the at-
tainment of stable synchronization when some of the coupling schemes are combined
with partial replacement, particularly in the non-linear terms of the second system. As
far as we know, this possibility has been little explored. Although in some schemes only
local stability of the synchronous state is concluded, we explore coupling schemes where
global stability is guaranteed. The global stability conditions result from the direct
Lyapunov method, taking a different approach from the usual one.
We also address linear bidirectional diffusive coupling and describe, in terms of the
coupling parameter, the various stability transitions of the chaotic attractor embedded in
the synchronization manifold, considering the bifurcations characterized by the equations
defining the coupled system.
Some results of asymptotic synchronization stability are obtained. To investigate
the local asymptotic stability, we calculate the eigenvalues of the linearized equation of
the coupled system or study the transverse Lyapunov exponents. By constructing an
appropriate Lyapunov function whose derivative can be bounded, we manage to exhibit
conditions dependent on the coupling parameter(s) that guarantee global asymptotic
stability. These conditions result from the possibility of bounding the derivative of the
Lyapunov function through constants that limit the dynamical variables of the systems.

3.2 In discrete chaotic dynamical systems

We are studying a non-linear coupling scheme that naturally arises from the family of
analytical complex quadratic maps. The decomposition into real and imaginary parts of
each map results in a bidirectionally coupled system of two unidimensional real quadratic
maps with distinct control parameters. The resulting coupling term is proportional to
the square of the difference between the dynamic variables of the component systems. To
the best of our knowledge, this coupling has not been studied so far. It is an asymmetric
coupling between real quadratic maps, which presents an additional challenge.

11
 In the study carried out, we take parameter values that lead to chaotic behaviour in
the component maps. The dynamics of coupled bidimensional discrete chaotic systems
are studied, obtained through asymmetric coupling. When practical synchronization is
not achieved, but the difference between the variables of the two systems is bounded, a
control technique was applied to the coupled system to optimize the results obtained.
The control technique used extends the well-known Ott-Grebogy-Yorke chaos control
method [19], through a small perturbation of the coupling parameter.
In addition, we explore some variants of the original coupling. We obtain stable
identical and generalized synchronization when we consider some variants of the original
coupling, favouring the absence of symmetry in the coupling. Two of these constitute
a generalization to the use of distinct coupling parameters. We continue to consider a
single coupling parameter as we aim to investigate the advantages of symmetry in the
coupled system. This is still a bidirectional coupling between identical systems that, to
our knowledge, has also not been studied. We also explore the study of a generalization
of the coupling to the case of distinct coupling parameters in each of the component
systems: the case of identical and non-identical maps.
For all couplings, we analyze the dynamics of the coupled system and, subsequently,
investigate the local asymptotic stability of synchronization, based on the eigenvalues of
the linearized equation of the transverse system and the estimation of the corresponding
Lyapunov exponents. By analyzing the difference between the dynamical variables of
the systems, we establish some results that guarantee their stable synchronization. The
computational approaches presented, in addition to confirming these results, shed light
on the necessary study in each coupling scheme to determine the parameter values that
lead to stable synchronization.

4 Conclusion
This review study provides a self-contained approach to the preliminary, but funda-
mental, concepts and results in synchronization of dynamical systems, with emphasis in
chaotic dynamical systems, and establishes the terminology necessary for the research
work carried out. By tracing the evolution from classical concepts to modern gener-
alizations, this article reveals how the phenomenon of synchronization extends beyond
periodic oscillations to the realm of chaotic behaviour with profound theoretical and
practical implications. The identification and characterization of identical and gener-
alized synchronization regimes have demonstrated the richness and complexity of the
behaviour of coupled systems. We have provided an overview of the key types of syn-
chronization, identical and generalized synchronization, and discussed their theoretical
underpinnings. The stability analysis of coupled systems was presented using tools
such as Lyapunov exponents, eigenvalues, and Lyapunov functions, allowing local and
global stability criteria to be established. We highlighted the challenges posed by multi-
stability, bifurcations, and the loss of synchronization through bubbling and blowout
transitions. Special attention was paid to the role of chaotic attractors, which serve as
a foundation for the control and predictability of synchronized states.

12
 There are several reasons that motivated the choice of this research topic. The phe-
nomenon of chaos synchronization is, from the outset, interesting due to its high applic-
ation potential. This potential is transversal to knowledge areas as distinct as physics,
biology, engineering, or economics. We find it particularly stimulating to study a phe-
nomenon that requires the adjustment of dynamical behaviours to obtain coincident
chaotic motion, which is possible precisely in chaotic dynamical systems where sensitive
dependence on initial conditions is one of the characteristics. The existence of certain
analogies between synchronization and chaos control and the possibility of applying
chaos control techniques as a way to optimize the results of synchronization constitute
yet another motivating factor for choosing the topic. Moreover, we do not overlook
the fact that the phenomenon of chaos synchronization is quite recent in the non-linear
theory of dynamical systems and continues to generate significant interest in the sci-
entific community. However, despite the efforts made to investigate this phenomenon,
many questions remain open.
From secure communication systems to control applications in biological and engin-
eering networks, synchronization has the potential to become a powerful tool in system
design. The insights into multi-stability, attractor co-existence, and the role of chaotic
trajectories in synchronization loss provide fertile ground for future research, particularly
in applications where robustness against parameter mismatches and noise is crucial. The
diversity of approaches, including analytical techniques, computational experiments, and
theoretical generalizations, demonstrates that synchronization is a versatile and evolving
research area. By bridging the gap between classical synchronization theory and modern
chaos theory, this work contributes to a broader understanding of complex systems. It
opens avenues for new studies on the dynamics of non-identical systems, the interplay
between chaos and control, and the effects of coupling asymmetries on global stability.
The proposed techniques and methodologies can inspire further exploration, promoting
the design of more efficient coupling protocols, and enhancing the predictability and
stability of synchronized states in real-world applications.
This review serves as a resource for researchers and practitioners interested in under-
standing and harnessing synchronization in diverse applications, from secure commu-
nications to the control of biological and engineering networks. The insights presented
pave the way for future investigations into the synchronization of non-identical systems,
the impact of asymmetries in coupling, and the development of robust synchronization
protocols that resist parameter mismatches and noise. The concepts discussed here have
significant potential for real-world applications, underscoring the continued relevance of
synchronization in both the theoretical and applied sciences.

References
[1] Ch. Huygens (1673), Horoloquium Oscillatorium, Apud. F. Muguet, Paris.

13
 [2] E.N. Lorenz (1963), Deterministic non-periodic flows, J. Atmos. Sci. 20, 130-
141.

[3] T. Li e J.A. Yorke (1975), Period three implies chaos, Amer. Math. Monthly
82, 985-992.

[4] A. N. Sharkovkii (1964), Coexistence of cycles of a continuous map of a line

into itself, Ukrainian Math. J. 16, 61-71.

[5] +R.M. May (1976), Simple mathematical models with very complicated dy-
namics, Nature 261, 459.

[6] +J.-P. Eckmann e D. Ruelle (1985), Ergodic theory of chaos and strange at-
tractors, Rev. Mod. Phys. 57, 617.

[7] Y. Nagai e Y.-C. Lai (1997), Periodic-orbit theory of the blowout bifurcation,
Phys. Rev. E 56, 4031-4041.

[8] ++ Y.-C. Lai (1996), (XXXXXX), Phys. Rev. E 53, R4267.

[9] A. Stefański e T. Kapitaniak (2003), Estimation of the dominant Lyapunov

exponent of non-smooth systems on the basis of maps synchronization, Chaos,
Solitons and Fractals 15, 233-244.

[10] K. Kaneko (1986), Lyapunov analysis and information flow in coupled map
lattices, Physica D 23, 436.

[11] V.S. Afraimovich, N.N. Verichev e M.I. Rabinovich (1986), Stochastic synchro-
nization of oscillations in dissipative systems, Radiophys. Quantum Electron.
29 (9), 795-803.

[12] T. Yamada e H. Fujisaka (1983), Stability theory of synchronized motion in

coupled-oscillator systems, Prog. Theoret. Phys. 69(1), 32-47.

[13] T. Yamada e H. Fujisaka (1983), Stability theory of synchronized motion in

coupled-oscillator systems. II, Prog. Theoret. Phys. 70 (5), 1240-1248.

[14] A.S. Pikovsky (1984), On the interaction of strange attractors, Z. Physik B 55

(2), 149-154.

[15] A.S. Pikovsky (1984), Synchronization and stochastization of the ensemble of

autogenerators by external noise, Radiophys. Quant. Electron. 27 (5), 576-581.

[16] L.M. Pecora e T.L. Carroll (1990), Synchronization in chaotic systems, Phys.
Rev. Lett. 64 (8), 821-824.

[17] L.M. Pecora e T.L. Carroll (1991), Driving systems with chaotic signals, Phys.
Rev. A 44 (4), 2374-2383.

14
[18] T.L. Carroll e L.M. Pecora (1991), Synchronizing chaotic circuits, IEEE Trans.
Circuits Systems 38 (4), 453-456.

[19] E. Ott, C. Grebogy e J.A. Yorke (1990), Controlling chaos, Phys. Rev. Lett.
64 (11), 1196-1199.

[20] L. Kocarev e U. Parlitz (1995), General approach for chaotic synchronization

with applications to communication, Phys. Rev. Lett. 74 (25), 5028-5031.

[21] U. Parlitz (1996), Estimating model parameters from time series by autosyn-
chronization, Phys. Rev. Lett. 76 (8), 1232-1235.

[22] A. Stefański e T. Kapitaniak (2003), Synchronization of two chaotic oscillators

via a negative feedback mechanism, Int. J. Solids Structures 40, 5175-5185.

[23] Y.-C. Lai e C. Grebogy (1993), Synchronization of chaotic trajectories using

control, Phys. Rev. E 47(4), 2357-2360.

[24] T. Kapitaniak (1994), Synchronization of chaos using continuous control, Phys.

Rev. E 50, 1642-1644.

[25] J.K. John e R.E. Amritkar (1994), Synchronization of unstable orbits using
adaptive control, Phys. Rev. E 49 (6), 4843-4848.

[26] + V.S. Anishchenko, T.E. Vadivasova, D.E. Posnov e M.A. Safonova (1991),
Forced and mutual synchronization of chaos, Sov. J. Commun. Technol. Elec-
tron. 36, 21(ou 23 ?).

[27] M. Ding e E. Ott (1994), Enhancing synchronism of chaotic systems, Phys. Rev.
E 49 (2), 945-948.

[28] K. Pyragas (1992), Continuous control of chaos by self-controlling feedback,

Phys. Lett. A 170 (6), 421-428.

[29] C.W. Wu e L.O. Chua (1994), A unified framework for synchronization and
control of dynamical systems, Int. J. Bifur. Chaos 4 (4), 979-998.

[30] A. Stefański (2000), Estimation of the largest Lyapunov exponent in systems

with impacts, Chaos, Solitons and Fractals 11, 2443-2451.

[31] N.F. Rulkov, M.M. Sushchik, L.S. Tsimring e H.D.I. Abarbanel (1995), Gener-
alized synchronization of chaos in directionally coupled chaotic systems, Phys.
Rev. E 51 (2), 980-994.

[32] L. Kocarev e U. Parlitz (1996), Generalized synchronization, predictability, and

equivalence of unidirectionally coupled dynamical systems, Phys. Rev. Lett. 76
(11), 1816-1819.

15
[33] H.D.I. Abarbanel, N.F. Rulkov e M.M. Sushchik (1996), Generalized synchro-
nization of chaos: The auxiliary system approach, Phys. Rev. E 53, 4528-4535.

[34] + L.M. Pecora, T.L. Carroll e J.F. Heagy (1995), Statistics for mathematical
properties of maps between time-series embeddings, Phys. Rev. E 52 (2), 3420.

[35] + P. So, E. Barreto, K. Josić, E. Sander e S. Schiff (2002), Limits to the

experimental detection of nonlinear synchrony, Phys. Rev. E 65 (4), R46225.

[36] B.R. Hunt, E. Ott e J.A. Yorke (1997), Differentiable generalized synchroniza-
tion of chaos, Phys. Rev. E 55 (4), 4029-4034.

[37] K. Pyragas (1996), Weak and strong synchronization of chaos, Phys. Rev. E
54, R4508-11.

[38] E. Barreto, P. So, B.J. Gluckman e S.J. Schiff (2000), From generalized syn-
chrony to topological decoherence: Emergent sets in coupled chaotic systems,
Phys. Rev. Lett. 84 (8), 1689-1692.

16