Available online at www.sciencedirect.

com

Physica A 335 (2004) 365 – 370

www.elsevier.com/locate/physa

Synchronization in coupled map lattices with

small-world delayed interactions
Chunguang Lia;∗ , Shaowen Lia; b , Xiaofeng Liaoa , Juebang Yua
a Institute
of Electronic Systems, College of Electronic Engineering, University of Electronic Science and
Technology of China, Chengdu 610054, PR China
b Department of Mathematics, Southwestern University of Finance and Economics, Chengdu 610000,

PR China

Received 19 August 2003; received in revised form 11 November 2003

Abstract
We study the synchronization of coupled map lattices with small-world interactions and cou-
pling delays in this paper. A stability criterion for the synchronized state is derived. Numerical
examples are presented to demonstrate the correctness of our theoretical result.
c 2003 Elsevier B.V. All rights reserved.

PACS: 05.45.−a

Keywords: Small world network; Synchronization; Delay

1. Introduction

Coupled map lattices (CML) were introduced by Kaneko (for a list of references see
Ref. [1]) as models to understand spatially extended nonlinear systems using building
blocks that are well understood. Over the last two decades CML have attracted much
attention [1,2]. Up to date, most of the existing work on CML assumes that the coupling
con?guration is completely regular, while a few studies address the randomly coupled
CML. However, interactions in many biological, technological and social systems are
neither completely regular nor completely random. To interpolate between these two
extremes, Watts and Strogatz [3] introduced the interesting concept of small-world
(SW) networks. The so-called SW networks have intermediate connectivity properties
but exhibit a high degree of clustering as in the regular networks and a small average

∗ Corresponding author.
E-mail address: [email protected] (C. Li).

c 2003 Elsevier B.V. All rights reserved.

0378-4371/$ - see front matter 
doi:10.1016/j.physa.2003.11.018
366 C. Li et al. / Physica A 335 (2004) 365 – 370

distance between vertices as in the random networks. In Ref. [4], the authors investigate
the synchronization of CML with SW interactions.
On the other hand, the inclusion of time delays into lattices is natural in the realistic
consideration of ?nite transmission of interactions. The synchronization of globally
coupled map with delayed interactions has recently been studied in Ref. [5].
In this paper, we investigate the synchronization in CML with SW interactions and
delayed coupling. We consider the following one-dimensional array of N maps, coupled
as
N


xi (t + 1) = (1 − )f(xi (t)) + aij f(xj (t − )) ; (1)
Ni
j=1

where f(x) is 1-D map,  ¿ 0 is the coupling strength, Ni is the number of signals
the ith map receives. is the time delay constant, which is a nonnegative integer (for
simplicity, we assume that all delays are the same in the lattice). The matrix A = {aij }
encodes the connection topology: if map j sends a signal to i, aij =1; otherwise, aij =0.
N
aii = 0 for all i, and Ni = j=1 aij . If aij = aji , the matrix de?nes an undirected lattice,
otherwise it describes a directed lattice, in which the vertices are the maps. The delays
can be viewed as arising from the ?nite speeds of transmission and spreading as well
as congestions. The normalized prefactor 1=Ni in the coupling means that each vertex
is inJuenced equally by all of the others that send signals to it.
The original SW model introduced in Ref. [3] can be described as follows. Take a
one-dimensional lattice of N vertices arranged in a ring with connections only between
nearest neighbors. We “rewire” each connection with probability p. Rewiring in this
context means reconnecting randomly the whole lattice, with the constraint that no two
diNerent vertices can have more than one connection in between, and no vertex can
have a connection with itself.
Note, however, that there is a possibility for the SW model to be broken into un-
connected clusters. This problem can be resolved by a slight modi?cation of the SW
model, suggested by Newman and Watts (NW) [6]. In the NW model, we do not break
any connection between any two nearest neighbors. Instead, we add with probability
p a connection between each unconnected pair of vertices. Likewise, we do not allow
a vertex to be coupled to another vertex more than once, or coupling of a vertex with
itself. For p = 0, it reduces to the originally nearest-neighbor coupled lattice; for p = 1,
it becomes a globally coupled lattice. From a coupling-matrix point of view, CML (1)
with small-world connections evolves according to the rule that, in the nearest-neighbor
coupling matrix A, if Aij = 0, we set Aij = 1 with probability p.

2. Stability of synchronized state

In this section, we analyze the stability of the synchronized state. For simplicity, we
let = 1 in this section. The cases with other values of can be analyzed similarly.
Clearly, the synchronized state s(t) is governed by
s(t + 1) = (1 − )f(s(t)) + f(s(t − 1)) : (2)
 C. Li et al. / Physica A 335 (2004) 365 – 370 367

Linearizing (1) around the synchronized state s(t) in (2), we have

N


ei (t + 1) = (1 − )f
(s(t))ei (t) + f (s(t − 1)) aij ej (t − 1) ; (3)
Ni
j=1

where ei (t) denotes the ith map’s deviations from s(t). By letting e(t)=[e1 (t), e2 (t); : : : ;
eN (t)], Eq. (3) can be recast as

e(t + 1) = (1 − )f
(s(t))e(t) + f
(s(t − 1))e(t − 1)G T ; (4)

where G = {Gij }N ×N , with Gij = aij =Ni . Clearly, Gii = 0 for all i, and the sum of the
entries in each row of G is equal to 1. Let  be the eigenvalue of G. Denote the
corresponding eigenvector by d and let u(t) = e(t)d. Then

u(t + 1) = (1 − )f
(s(t))u(t) + f
(s(t − 1))u(t − 1) : (5)

The above equation can be rewritten as

    
u(t + 1) (1 − )f
(s(t)) f
(s(t − 1)) u(t)
= : (6)
u(t) 1 0 u(t − 1)

It is obvious that the synchronized state (2) is (locally) stable if and only if all eigen-
values of the matrix of the error dynamics (6) are within the unit circle on the complex
plane (including t → ∞). The characteristic equation of the error dynamics (6) is

2 − (1 − )f
(s(t)) − f
(s(t − 1)) = 0 : (7)

Instead of calculating the solution, we use the Schur Stability Test (see Appendix A
in Ref. [7] or [8]) to determine the stability conditions. From the Schur Stability Test,
the stability conditions of the synchronized state are
|f
(s(t − 1))| ¡ 1 ;
(8)
|(1 − )f
(s(t))(1 + f
(s(t − 1)))| ¡ |1 − 2 ||2 f
(s(t − 1))2 | :

Although we can only calculate the eigenvalues of the connection matrix numerically
in SW networks, we can bound their location by using the Gerschgorin’s circle theorem
[9]. Applying Gerschgorin’s circle theorem to matrix G, we ?nd that all the eigenvalues
lie in, or on the unit circle, since Gii = 0 for all i and the sum of the entries in each
row of G is equal to 1. That is, || 6 1.
If the lattice is undirected, that is A is symmetric, then  is real valued. From
(8) and || 6 1, we know that the following condition guarantees the stability of the
synchronized state:

|f
(s(t − 1))| + |(1 − )f
(s(t))| ¡ 1 : (9)

which is easy to be veri?ed and is independent of the lattice coupling con?gurations,

although it maybe more conservative than (8).
368 C. Li et al. / Physica A 335 (2004) 365 – 370

We also analyze the stability conditions of the synchronized states of networks with-
out and with longer interaction delays using similar methods. All the stability conditions
depend on the form of the function f. When considering general function f, no explicit
relationship among these stability conditions is found.

3. Numerical simulations

We consider some numerical examples in this section. We choose the individual

map to be the logistic map, f(x) = ax(1 − x). The eigenvalue spectrum of the coupling
matrix of SW CML with diNerent pN can be determined similarly as that in Ref. [4].
For simplicity, we consider undirected CML in this section, that is the coupling matrix
A is symmetric. Firstly, we consider the case a = 3:02. In this case, the logistic map
evolve in a limit cycle of period 2. From (9), we know that if the coupling strength
 = 0:5, then the synchronized state is stable for any topology of the CML. We show
the curve of xk (t) − x1 (t) for k = 2; : : : ; N in Fig. 1 for diNerent values of p and N .
From these ?gures, we know that the synchronized state is stable and is independent
of the lattice topology, which is consistent with our theoretical result.
Next, we study the behavior of the mean ?eld on SW lattices for a given number of
N
average nonlocal connections pN . We study the mean ?eld h(t) = (1=N ) k=1 xk (t)
as a function of p. We choose a = 3:2 and  = 0:1 in these simulations. As in Ref. [4],
we study this behavior for two values pN = 1 and 2, i.e., the cases of one nonlocal
coupling per site and two nonlocal coupling per site on average. Figs. 2(a) and (b)
show the “bifurcation diagram” for the mean ?eld as a function of the probability
p for pN = 1 and 2. There are diNerences in behaviors of individual con?gurations
due to randomness. We have shown representative con?gurations. We can see that
the behavior of the mean ?eld h(t) is diNerent for diNerent p in both the cases. The
spread of the values of h(t) becomes wide when the p is increased. We try to study
the variance of the mean ?eld 2 = h(t)2  − h(t)2 , and we use the mean value

Fig. 1. The curve of xk (t) − x1 (t): (a) p = 0:1, N = 100; (b) p = 0:5, N = 200.
 C. Li et al. / Physica A 335 (2004) 365 – 370 369

Fig. 2. The “bifurcation diagram” of the mean ?eld h(t) as a function of p: (a) pN = 1; (b) pN = 2.

Fig. 3. The variance of the mean ?eld as a function of p.

of variance over 20 con?gurations 2  to quantify the uncorrelateness of individual

elements. Fig. 3 shows 2  as a function of p for pN = 1 and 2. As we can see that
there is a increscent trend of 2  when p is increased for both pN = 1 and 2, which
is consistent with Fig. 2.

4. Conclusions

In summary, we have investigated the synchronization of coupled map lattices with

SW interactions and coupling delays. A stability criterion for the synchronized state
is derived. Numerical examples have demonstrated the correctness of our theoretical
result. The behavior of mean ?eld as a function of p is also studied.
370 C. Li et al. / Physica A 335 (2004) 365 – 370

Acknowledgements

This work was supported by the National Natural Science Foundation of China under
Grant 60271019, and the Youth Science and Technology Foundation of UESTC under
Grant YF020207.

References

[1] K. Kaneko (Ed.), Theory and Applications of Coupled Map Lattices, Wiley, New York, 1993.
[2] K. Kaneko, Phys. Rev. Lett. 65 (1990) 1391;
P.M. Gade, H. Cerdeira, R. Ramaswamy, Phys. Rev. E 52 (1995) 2478;
D.H. Zanette, Europhys. Lett. 45 (2000) 424;
A.L. Gelover-Santiago, R. Lima, G. Matinez-Mekler, Physica A 283 (2000) 131.
[3] D.J. Watts, S.H. Strogatz, Nature 393 (1998) 440.
[4] P.M. Gade, C.K. Hu, Phys. Rev. E 62 (2000) 6409;
C. Li, X. Liao, J. Yu, Physica A, in press.
[5] C. Masoller, A.C. Marti, D.H. Zanette, Physica A 325 (2003) 186.
[6] M.E.J. Newman, D.J. Watts, Phys. Rev. E 60 (1999) 7332.
[7] K. Konishi, H. Kokame, Physica D 127 (1999) 1.
[8] S.P. Bhattacharyya, H. Chapellat, L.H. Keel, Robust Control: The Parametric Approach, Prentice-Hall,
Englewood CliNs, NJ, 1995.
[9] H. Minc, Nonnegative Matrices, Wiley, New York, 1988.