1: \documentclass[twocolumn,prl,aps,showpacs]{revtex4}
2: \usepackage{graphicx}
3: \def\pla{Phys. Lett. A }
4: \begin{document}
5:
6: \title{Comment on ``Intermittent Synchronization in a Pair of Coupled
7: Chaotic Pendula"}
8: \author{P.~Muruganandam}
9: \author{S.~Parthasarathy}
10: \author{M.~Lakshmanan}
11: \affiliation{Centre for Nonlinear Dynamics, Department of Physics,
12: Bharathidasan University, Tiruchirapalli 620 024, India}
13:
14: \pacs{05.45.+b}
15: \maketitle
16:
17: In a recent Letter, Baker, Blackburn and Smith~\cite{baker} reported
18: that permanent synchronization of a pair of unidirectionally coupled
19: identical chaotic pendula does not occur except as a numerical artifact
20: arising from finite computational precisions. Baker {\em et al.} showed
21: that the synchronization of a pair of unidirectionally coupled pendula
22: is always {\em intermittent}, for any value of coupling coefficient
23: $c$, by using their numerical and analytical tests. This was also found
24: to be true in the case of coupled Duffing oscillators~\cite{baker}.
25: However, on careful numerical analysis using Conditional Lyapunov
26: Exponents (CLEs), we find that there exists some specific range of $c$
27: values for which persistent synchronization can occur for these
28: systems.
29:
30: The main point of this comment is to clarify the conditions under which
31: coupled chaotic systems can exhibit permanent synchronization.
32: According to Pecora and Carroll\cite{pecora}, the master and slave
33: systems will perfectly synchronize only if the sub-Lyapunov exponents
34: or CLEs are all negative. The two analytic criteria mentioned in
35: Ref.\ \cite{baker} for asymptotic stability are, ({\em i}) the largest
36: eigenvalue of the Jacobian matrix corresponding to the flow evaluated
37: on the synchronization manifold must always be negative and ({\em ii})
38: existence of a suitable Lyapunov function. However,
39: it is not possible, in general, to show analytically the above two
40: criteria are uniformly obeyed for all $c$ values for most of the
41: chaotic systems, as shown in the case of coupled pendula (cf.
42: Eqs.\ (9) and (11) of Ref.\ \cite{baker}). In such cases, it is always
43: possible to compute the Lyapunov exponents of the coupled system for a
44: range of $c$ values and show that the CLEs take large negative values
45: for permanent synchronization.
46:
47: We have computed the Lyapunov exponents for the coupled pendula and the
48: Duffing oscillators using the standard Wolf {\it et al.} algorithm.
49: For numerical simulations we have used the same parameter values as in
50: Ref.\ \cite{baker} and used $5000$ drive cycles for calculations after
51: leaving $5000$ drive cycles as transient. Figure \ref{fig1}(a) shows
52: the variation of CLE as a function of $c$ for the coupled pendula and
53: it takes negative values for $0.796 \le c < 1.0$ only. The CLE value
54: for $c=0.79$ is $+0.00320$ ($\approx 10^{-3}$), for which
55: synchronization can not occur and hence the observed intermittent
56: synchronization (cf. Fig.\ 1 in Ref.\ \cite{baker}) is a computer
57: artifact. However, we find that for $c=0.932$, the CLE value becomes
58: the lowest ($-0.06825$) and is a large negative value for which
59: intermittent synchronization is absent as shown in Fig.\ \ref{fig1}(b).
60: This has been verified upto $2\times 10^6$ drive cycles with the
61: addition of tiny noise levels showing that permanent synchronization
62: does occur. The same phenomenon persists over a range of $c$ values
63: close to $0.932$.
64: \begin{figure}[!ht]
65: \begin{center}
66: \includegraphics[width=\linewidth]{fig1}
67: \end{center}
68: \caption{
69: (a) Variation of the Conditional Lyapunov Exponent (CLE) as a function
70: of $c$, (b) synchronization error $\eta$ (given by Eq.\ (5) in Ref.~[1])
71: versus time in drive cycles for coupled pendula; (c) CLE versus $c$ and
72: (d) $\eta$ versus drive cycles for coupled Duffing oscillators.
73: }
74: \label{fig1}
75: \end{figure}
76:
77: Similar arguments hold good for the coupled Duffing oscillator also and
78: the results are shown in Figs.\ \ref{fig1}(c) and (d). The CLE takes
79: small negative values close to zero ($\approx 10^{-3}$) for
80: $c\in(1.48,1.64)$ where synchronization is intermittent and numerical
81: artifact arises in this range also. The CLE value for $c=1.5$ is
82: $-0.00562$ and hence the observed hard locking becomes intermittent
83: with the addition of noise in the $8$-th digit (cf. Fig.~4 in
84: Ref.~\cite{baker}). In addition, there exists another range of $c$
85: values, $2.74\le c\le 3.84$ where the CLEs are negative (cf.
86: Fig.\ \ref{fig1}(c)). We find that for $c=3.06$, the CLE takes the
87: lowest value of $-0.05967$ (large negative value) for which persistent
88: synchronization occurs and the intermittent synchronization is absent
89: (cf. Fig.\ \ref{fig1}(d)). Again the same phenomenon persists over
90: a range of $c$ values close to $3.06$.
91:
92: In summary, the main aim of our comment is to emphasize that the
93: conditional Lyapunov exponents play an important role in distinguishing
94: between intermittent and persistent synchronization, when it is not
95: possible to show analytically the criteria for asymptotic stability are
96: uniformly obeyed. We find that intermittent synchronization can occur
97: when CLEs are very small positive or negative values close to zero
98: while persistent synchronization occurs when CLEs become sufficiently
99: large negative values.
100:
101: This work was supported by DST, Govt. of India (M.L. and S.P.),
102: DAE-NBHM (M.L.) and CSIR (P.M.).
103:
104: \begin{thebibliography}{99}
105: \bibitem{baker}
106: G.~L.~Baker, J.~A.~Blackburn and H.~J.~T.~Smith, \prl
107: {\bf 81}, 554 (1998)
108: \bibitem{pecora}
109: L.~M.~Pecora and T.~L.~Carroll, \prl
110: {\bf 64}, 821 (1990)
111: \end{thebibliography}
112:
113: \end{document}
114:
115:
116: