1: \documentclass{elsart}
2: %\documentclass[doublespacing]{elsart}
3: %
4: %......packages
5: \usepackage{graphicx}
6: \usepackage{amssymb}
7: \usepackage{rotate}
8: \usepackage{rotating}
9:
10: \bibliographystyle{elsart-num}
11:
12: \newcommand{\imag}{\mbox{i}}
13: \newcommand{\iimag}{\mbox{\scriptsize i}}
14: \newcommand{\expo}{\mbox{e}}
15: \newcommand{\g}[1]{\mbox{#1}}
16:
17: \begin{document}
18:
19: \begin{frontmatter}
20: \title{Controlling spatiotemporal chaos\\
21: in oscillatory reaction-diffusion systems\\
22: by time-delay autosynchronization}
23:
24: \author{C. Beta},
25: \author{A. S. Mikhailov\corauthref{cor1}},
26: \corauth[cor1]{Corresponding author.
27: Tel.: +49-30-8413-5122;
28: fax: +49-30-8413-5106.
29: \ead{mikhailov@fhi-berlin.mpg.de}}
30:
31: \address{Abteilung Physikalische Chemie,\\
32: Fritz-Haber-Institut der Max-Plack-Gesellschaft,\\
33: Faradayweg 4--6, 14195 Berlin, Germany}
34:
35: \begin{abstract}
36: Diffusion-induced turbulence in spatially extended oscillatory
37: media near a supercritical Hopf bifurcation can be
38: controlled by applying global time-delay autosynchronization.
39: We consider the complex Ginzburg-Landau equation in the Benjamin-Feir
40: unstable regime and
41: analytically investigate the stability of uniform oscillations
42: depending on the feedback parameters.
43: We show that a noninvasive stabilization of uniform oscillations
44: is not possible in this type of systems.
45: The synchronization diagram in the plane spanned by the
46: feedback parameters is derived.
47: Numerical simulations confirm the analytical results and
48: give additional information on the spatiotemporal dynamics of the
49: system close to complete synchronization.
50: \end{abstract}
51:
52: \begin{keyword}
53: % keywords here, in the form: keyword \sep keyword
54: Reaction-diffusion systems \sep Turbulence \sep Feedback
55:
56: % PACS codes here, in the form: \PACS code \sep code
57: \PACS 05.45.+b \sep 82.40.Bj \sep 82.20.Wt
58: % we use PACS-1995 classification as required in the
59: % authors guideline
60: \end{keyword}
61: \end{frontmatter}
62: %
63: %---------------------------------------------------------------------
64: \section{Introduction}
65: \label{sc:introduction}
66: %
67: Pattern formation in spatially extended systems far from thermal
68: equilibrium has attracted much attention in recent years and became
69: a field of active research
70: \cite{cross-hohenberg-1993,mikhailov-book1}.
71: A large class of these systems is described by equations of the
72: reaction-diffusion type.
73: In such systems, a broad variety of complex
74: spatiotemporal patterns has been observed, ranging from
75: travelling pulses and target patterns to rotating spiral waves and
76: spatiotemporal chaos.
77: Important examples are chemical reaction-diffusion
78: systems such as the Belousov-Zhabotinsky reaction
79: or catalytic surface chemical reactions
80: \cite{zaikin-zhabotinsky-1970,winfree-1972,imbihl-ertl-1995},
81: as well as semiconductor and biological systems
82: \cite{schoell-book-2001,benjacob-cohen-levine-2000}.
83:
84: Many systems show oscillatory dynamics of their individual elements.
85: Close to the onset of oscillations, the equations describing
86: any oscillatory system can be reduced
87: to a simple universal equation for the complex oscillation amplitude, the
88: {\itshape complex Ginzburg-Landau equation} (CGLE)
89: \cite{kuramoto-book,aranson-kramer-2002}.
90: For a distributed system with diffusive coupling, the CGLE corresponds to
91: the normal form of a supercritical Hopf bifurcation.
92: Although the CGLE is strictly valid only
93: in the immediate vicinity of the oscillation onset, it has been found in
94: many cases that its qualitative predictions also hold within a wider range
95: near the bifurcation point.
96: The CGLE thus provides a general framework for studying common features
97: of the dynamical behaviour in oscillatory reaction-diffusion systems.
98: If oscillations are synchronized, uniform oscillations can be
99: established in the medium or regular spatiotemporal patterns
100: can emerge.
101:
102: If the Benjamin-Feir criterion is met, synchronous oscillations become
103: unstable and a turbulent state develops
104: where two types of chaotic behaviour can be distinguished.
105: In the state of phase turbulence,
106: the local oscillation phase exhibits weak irregular fluctuations while
107: the real amplitude remains saturated showing only small variations
108: \cite{kuramoto-book}.
109: Amplitude or defect turbulence
110: is characterized by strong fluctuations of both phase and amplitude
111: which are due to the presence of defects, the locations with a vanishing
112: amplitude
113: \cite{coullet-gil-lega-1989}.
114:
115: Over the past decade,
116: problems of chaos control became to play a central role in the
117: studies of chaotic dynamics
118: \cite{schuster-book-1999}.
119: Inspired by the pioneering work of Ott, Grebogi, and Yorke (OGY)
120: \cite{ott-grebogi-yorke-1990},
121: control of chaos has been studied in the context of many different
122: dynamical systems
123: \cite{ditto-rauseo-spano-1990,carroll-et-al-1992,petrov-et-al-1993}
124: However, the OGY method is designed to control chaos in
125: low-dimensional dynamical systems.
126: It requires continuous extensive analysis of the system dynamics
127: which is virtually impossible to carry out in the case of fully developed
128: high-dimensional turbulence in spatially extended systems.
129: A different empirical control method based on implementing a delayed
130: feedback loop was proposed by Pyragas
131: and is known as {\itshape time-delay autosynchronization}
132: (TDAS)~\cite{pyragas-1992,socolar-sukow-gauthier-1994}.
133: In this method, a feedback signal $F$ is applied to the system that is
134: proportional to the
135: difference between the delayed value of a given system variable $u$ and
136: its instantaneous value,
137: $F \sim u(t-\tau) - u(t)$.
138: This approach is easily extended to spatially distributed experimental
139: systems that often do not allow individual addressing of their local
140: elements.
141: Then, the global feedback signal is generated from the integral
142: value of $u$ over all system elements.
143: In recent studies, the TDAS method has been used in a number of
144: different dynamical systems
145: \cite{pierre-bonhomme-atipo-1996,franceschini-bose-schoell-1999,parmananda-et-al-1999}
146: and was modified and improved considerably
147: \cite{beck-et-al-2002,baba-et-al-2002,just-et-al-2003}.
148: For the CGLE, the action of local TDAS has been previously
149: considered
150: \cite{harrington-socolar-2001,montgomery-silber-2003pre}.
151:
152: The present work was motivated by our experimental study of catalytic CO
153: oxidation under TDAS
154: \cite{beta-et-al-2003}.
155: The catalytic CO oxidation on a platinum (110) single crystal surface
156: constitutes the most thoroughly studied example among a class of simple
157: heterogeneous catalytic reactions for which reaction-induced
158: spatiotemporal pattern formation has been reported
159: \cite{imbihl-ertl-1995}.
160: It displays a particularly rich variety of spatiotemporal concentration
161: patterns
162: \cite{jakubith-et-al-1990}
163: that can be modeled numerically using a simple model of three coupled
164: partial differential equations
165: \cite{krischer-eiswirth-ertl-1992,baer-et-al-1992}.
166: The effect of global delayed feedback on pattern formation in catalytic
167: CO oxidation was studied previously using a different time-delay feedback
168: scheme
169: \cite{kim-et-al-2001,bertram-et-al-2003,bertram-mikhailov-2003}.
170: The same feedback scheme was also investigated in detail in the context
171: of the complex Ginzburg-Landau equation
172: \cite{battogtokh-mikhailov-1996,battogtokh-preusser-mikhailov-1997}.
173: Our work on suppression of chemical turbulence in catalytic CO
174: oxidation using the TDAS scheme
175: \cite{beta-et-al-2003}
176: was focused on investigating the question of invasiveness.
177: We found in the experiment that, although we could reduce the magnitude
178: of the feedback signal considerably, we were not able to reach the ideal
179: limit of a vanishing feedback signal in the state of control.
180: In the present work, we use the complex Ginzburg-Landau equation as a
181: general model for spatially extended oscillatory systems to study common
182: aspects of control of diffusion-induced chemical turbulence by TDAS.
183:
184: The paper is organized as follows. In the next section, the physical
185: problem and the corresponding equations will be introduced.
186: In Section \ref{sc:stability}, we analyze the stability of uniform
187: oscillations in the presence of TDAS and derive a synchronization
188: diagram.
189: Numerical simulations of the complex Ginzburg-Landau
190: equation with TDAS are described in Section
191: \ref{sc:numerics}.
192: The paper ends with conclusions and a discussion of the obtained results.
193: %
194: %---------------------------------------------------------------------
195: \section{Formulation of the problem}
196: \label{sc:equation}
197: %
198: In absence of feedback, the dynamics of a chemical
199: reaction-diffusion system can be generally described by a set of
200: coupled partial differential equations
201: %
202: \begin{equation}
203: \label{eq:rds}
204: \mathbf{\dot{u}}=\mathbf{f}(\mathbf{u}\,,\mathbf{p})+
205: \mathbf{D}\nabla^2\,\mathbf{u}\, ,
206: \end{equation}
207: %
208: where $\mathbf{u}=(u_1,u_2,...,u_N)$ represents concentrations of
209: reacting species and $\mathbf{D}$ is their diffusion matrix.
210: The set of nonlinear functions $\mathbf{f}(\mathbf{u},\mathbf{p})$ of
211: the concentrations $\mathbf{u}$ accounts for the reaction part of the
212: dynamics and depends on a number of parameters
213: $\mathbf{p}=(p_1,p_2,...,p_M)$, such as rate constants and external
214: conditions.
215:
216: Suppose that the dynamics of system (\ref{eq:rds}) is such that a
217: supercritical Hopf bifurcation occurs when crossing a certain threshold
218: $\mathbf{p}=\mathbf{p}_0$ in the parameter space.
219: Close to the onset of oscillations, system (\ref{eq:rds}) can be
220: transformed to a simple description of the dynamics in terms of a complex
221: oscillation amplitude $\eta(x,t)$.
222: This transformation is based on retaining only the leading critical
223: modes that govern the dynamics close to the bifurcation point.
224: It yields as equation of motion for the amplitude $\eta(x,t)$, the complex
225: Ginzburg-Landau equation
226: \cite{kuramoto-book},
227: %
228: \begin{equation}
229: \label{eq:cgle}
230: \dot{\eta}=(1-\imag\omega)\eta-(1+\imag\beta)|\eta|^2\eta+
231: (1+\imag\varepsilon)\nabla^2\eta \, ,
232: \end{equation}
233: %
234: where the parameters $\omega$, $\beta$, and $\varepsilon$ denote the
235: linear oscillation frequency, the nonlinear frequency
236: shift, and the linear dispersion coefficient, respectively.
237: When the condition $1+\varepsilon\beta<0$ of the Benjamin-Feir
238: instability is satisfied, uniform oscillations are unstable and
239: turbulence spontaneously develops in the system.
240: To control the turbulence, global feedback can be introduced.
241: In the presence of global time-delayed feedback, the considered
242: system is described by the equation
243: %
244: \begin{equation}
245: \label{eq:cglefeed}
246: \dot{\eta}=(1-\imag\omega)\eta-(1+\imag\beta)|\eta|^2\eta+
247: (1+\imag\varepsilon)\nabla^2\eta+F(t) \, ,
248: \end{equation}
249: %
250: where the feedback term is given by
251: %
252: \begin{equation}
253: \label{eq:fterm}
254: F(t)=\mu \, \expo^{\iimag\chi}(\bar{\eta}(t-\tau)-\bar{\eta}(t)) \, ,
255: \end{equation}
256: %
257: and
258: %
259: \begin{equation}
260: \label{eq:avampl}
261: \bar{\eta}(t)=\frac{1}{L}\int_0^L\eta(x,t)\g{d}x \, .
262: \end{equation}
263: %
264: Here, $\mu$ is the feedback intensity factor, $\tau$ is the delay time,
265: and $\chi$ is a phase shift in the application of the control force.
266:
267: Previously,
268: we showed experimentally~\cite{beta-et-al-2003} that diffusion-induced
269: turbulence could be
270: efficiently suppressed in an oscillatory chemical reaction-diffusion
271: system using TDAS.
272: It was found that the invasiveness of the control scheme could be significantly
273: reduced for an optimized choice of feedback parameters.
274: Figure~\ref{fg:exp} presents the period of uniform oscillations in the state
275: of control and the feedback magnitude as a function of the delay time
276: in the experiments with CO oxidation on Pt(110).
277: Although the feedback magnitude was significantly reduced for a delay
278: close to $\tau=6$ s, the optimal case of a vanishingly small feedback signal
279: could not be reached.
280: Near the point, for which the period $T$ of stabilized uniform
281: oscillations becomes equal to the delay time $\tau$,
282: an instability was found.
283: As seen in Fig.~\ref{fg:exp}, the system avoids a state corresponding
284: to the intersection point with the line for which $T=\tau$.
285: The stability of this state, for which the feedback vanishes, has been
286: earlier investigated~\cite{beta-et-al-2003} for the uniform system
287: in terms of a phase dynamics equation of a single oscillator.
288: In this article, we extend the analysis to the theoretical study of
289: a spatially extended system.
290:
291: Since our present work was motivated by the experimental results in
292: \cite{beta-et-al-2003}, we restrict our investigation to the case of a
293: globally applied control signal generated from the averaged complex
294: amplitude.
295: This case is relevant in many experimental situations where a controlled
296: manipulation of the individual system elements is not possible.
297: Note, however, that a different behaviour can be expected for a
298: space dependent application of the control scheme.
299: %
300: %---------------------------------------------------------------------
301: \section{Linear stability analysis}
302: \label{sc:stability}
303: %
304: We choose a superposition of a homogeneous mode
305: $H$ with a small spatially inhomogeneous perturbation
306: of arbitrary non-zero wave number
307: $\kappa$ as an ansatz for the complex oscillation amplitude,
308: %
309: \begin{equation}
310: \label{eq:ansatz_ampl}
311: \eta(x,t)=H(t)+A_+(t)e^{\iimag\kappa x}+A_-(t)e^{-\iimag\kappa x} \, .
312: \end{equation}
313: %
314: Substituting expression (\ref{eq:ansatz_ampl}) into Eq.
315: (\ref{eq:cglefeed}), we can separate, for small $A_{\pm}$,
316: homogeneous contributions from the spatially
317: inhomogeneous terms.
318: In this way, we find an equation for the homogeneous mode which is
319: decoupled from the nonuniform contributions,
320: %
321: \begin{equation}
322: \label{eq:h}
323: \dot{H}=(1-\imag\omega)H-(1+\imag\beta)|H|^2H+
324: \mu e^{\iimag\chi}(H(t-\tau)-H(t)) \, ,
325: \end{equation}
326: %
327: and a pair of coupled equations for the wave amplitudes $A_{\pm}$,
328: %
329: \begin{eqnarray}
330: \dot{A}_+&=(1-\imag\omega)A_+-(1+\imag\varepsilon)\kappa^2A_+-
331: 2(1+&\imag\beta)|H|^2A_+-(1+\imag\beta)H^2A_-^*,\label{eq:a+all} \\
332: \dot{A}_-^*&=(1+\imag\omega)A_-^*-(1-\imag\varepsilon)\kappa^2A_-^*-
333: 2(1-&\imag\beta)|H|^2A_-^*-(1-\imag\beta)H^{*2}A_+.\label{eq:a-all}
334: \end{eqnarray}
335: %
336: Let us first discuss the dynamics of the uniform mode
337: described by Eq.~(\ref{eq:h}).
338: Substituting $H=\rho_0\,e^{-\iimag\Omega t}$ into
339: Eq.~(\ref{eq:h}) we obtain the amplitude of uniform
340: oscillations in the presence of TDAS,
341: %
342: \begin{equation}
343: \label{eq:rho}
344: \rho_0=\sqrt{1+\mu(\cos(\chi+\Omega\tau)-\cos\chi)} \, ,
345: \end{equation}
346: %
347: and derive the equation
348: %
349: \begin{equation}
350: \label{eq:omega}
351: \Omega=\omega+\beta+\mu\beta(\cos(\chi+\Omega\tau)-\cos\chi)
352: -\mu(\sin(\chi+\Omega\tau)-\sin\chi)
353: \end{equation}
354: %
355: determining the frequency $\Omega$ of uniform oscillations.
356: Note that without the feedback ($\mu=0$) the system shows uniform
357: oscillations with the frequency $\Omega_0=\omega+\beta$.
358:
359: Figure~\ref{fg:period} shows the results of numerical integration
360: of Eq.~(\ref{eq:h}).
361: The oscillation period $T=2\pi/\Omega$ is shown as a function of the
362: delay time $\tau$ for weak and strong feedbacks.
363: We see that the state with $T=\tau$ is stable for a weak
364: feedback and becomes unstable for a high feedback intensity.
365: The strong feedback case [Fig.~\ref{fg:period}(b)] shows a
366: good qualitative agreement with the experimental data in
367: Fig.~\ref{fg:exp}.
368:
369: The stability of the state with $T=\tau$ can be understood in
370: terms of the bifurcation diagram presented in Fig.~\ref{fg:bifur}.
371: This diagram is constructed by solving Eq.~(\ref{eq:omega})
372: for a delay time $\tau$ equal to the period $T_0=2\pi/(\omega+\beta)$
373: of oscillations in the nonperturbed uniform system.
374: Besides a solution with a vanishing feedback term, for which
375: $\Omega=\Omega_0$, other solutions with a non-zero feedback and
376: $\Omega\ne\Omega_0$ are yielded by Eq.~(\ref{eq:omega}).
377: Linear stability analysis shows that the solution with
378: $\Omega=\Omega_0$ and a vanishing feedback is stable for small $\mu$ and
379: becomes unstable beyond some critical feedback intensity $\mu_0$.
380: These results are similar to those obtained by
381: using the phase dynamics approximation for a single
382: oscillator in the presence of TDAS~\cite{beta-et-al-2003}.
383:
384: We now turn to the stability analysis of uniform oscillations
385: with respect to spatially inhomogeneous perturbations.
386: The solution of the linear equations (\ref{eq:a+all}) and (\ref{eq:a-all})
387: can be sought as
388: $A_+=A_+^0e^{-i\Omega t}e^{\lambda t}$ and
389: $A_-^*=A_-^{*0}e^{i\Omega t}e^{\lambda t}$.
390: Substituting this into (\ref{eq:a+all}) and (\ref{eq:a-all}),
391: we obtain the following expression
392: %
393: \begin{equation}
394: \label{eq:lambda}
395: \lambda_{1,2}=1-\kappa^2-2\rho_0^2\pm\sqrt{(1+\beta^2)\rho_0^4
396: -(\Omega-\omega-\varepsilon\kappa^2-2\beta\rho_0^2)^2} \, ,
397: \end{equation}
398: %
399: where $\rho_0$ is given by Eq.~(\ref{eq:rho}) and $\Omega$ is
400: a solution of Eq.~(\ref{eq:omega}).
401: Uniform oscillations are stable with respect to the growth of
402: spatially nonuniform modes if Re $\lambda_{1,2}<0$ for all
403: wavenumbers $\kappa$.
404: The instability boundary is therefore determined by the
405: conditions
406: %
407: \begin{eqnarray}
408: \mbox{Re}\,\lambda(\mu_c,\kappa_c)&=&0 \, , \label{eq:cond1} \\
409: \frac{\partial}{\partial \kappa}\mbox{Re}\,
410: \lambda(\mu_c,\kappa_c)&=&0 \, . \label{eq:cond2}
411: \end{eqnarray}
412: %
413: Here, Eq. (\ref{eq:cond2}) accounts for the fact that the
414: wave number of the first unstable mode corresponds to a maximum in
415: $\lambda$ plotted as a function of $\kappa$.
416: The equations (\ref{eq:cond1}) and (\ref{eq:cond2})
417: can be solved numerically, taking into account
418: Eq. (\ref{eq:omega}) yielding the dependence of the frequency $\Omega$
419: of the uniform mode on the feedback intensity $\mu$ and the
420: delay time $\tau$.
421:
422: We are particularly interested in the stability of uniform
423: oscillations with $\Omega=\Omega_0$ and $\tau=T_0$.
424: Inserting this into the
425: general expression (\ref{eq:lambda}) for $\lambda$, we find
426: %
427: \begin{equation}
428: \label{eq:lambdaspecial}
429: \lambda_{1,2}=-\kappa^2-1\pm
430: \sqrt{-\varepsilon^2\kappa^4-2\beta\varepsilon\kappa^2+1} \, ,
431: \end{equation}
432: %
433: which is independent of $\mu$.
434: Therefore, this state is always unstable when the Benjamin-Feir
435: condition $1+\varepsilon\beta<0$ is fulfilled.
436: All inhomogeneous modes with a wavenumber less than
437: %
438: \begin{equation}
439: \label{eq:kappaspecial}
440: \kappa=\sqrt{-\frac{2(1+\varepsilon\beta)}{1+\varepsilon^2}}
441: \end{equation}
442: %
443: are growing, no matter how $\mu$ is chosen.
444: Thus, for $1+\varepsilon\beta<0$ the solution with
445: $\Omega=\omega+\beta$ will be always unstable so that a noninvasive
446: stabilization of uniform oscillations with TDAS is not possible in this
447: type of system.
448:
449: For the other solutions with $\Omega\ne\Omega_0$ at $\tau=T_0$, presented in
450: Fig.~\ref{fg:bifur}, the feedback signal is nonzero.
451: In this case, we can determine $\mu_c$ from the general
452: expression (\ref{eq:lambda}) for $\lambda$ by numerically solving
453: equations (\ref{eq:cond1}) and (\ref{eq:cond2}).
454: Figure~\ref{fg:mueps} shows the critical feedback intensity $\mu_c$ as
455: a function of the dispersion parameter $\varepsilon$ for
456: $\tau=T_0$ and the other parameters chosen as in Fig.~\ref{fg:bifur}.
457: In the Benjamin-Feir unstable regime, $\mu_c$ lies above the bifurcation
458: point $\mu_0$ at which the solution with $\Omega=\Omega_0$
459: becomes unstable, cf. Fig. \ref{fg:bifur}.
460: As the system approaches the Benjamin-Feir line with decreasing
461: $\varepsilon$, the critical feedback intensity necessary to stabilize uniform
462: oscillations decreases and finally converges towards $\mu_0$.
463:
464: From the conditions (\ref{eq:cond1}) and (\ref{eq:cond2}), we can
465: also determine the critical feedback intensity $\mu_c$ as a function
466: of the delay time $\tau$.
467: The resulting synchronization diagram in the plane spanned by the feedback
468: parameters $\tau$ and $\mu$ is displayed in Fig.~\ref{fg:diagram}(a).
469: The curve for the critical feedback intensity $\mu_c$ divides the
470: plane into a shaded region, where uniform oscillations are linearly stable
471: with respect to small perturbations of arbitrary wavenumber, and a region
472: where uniform osciallations are unstable.
473: The characteristic feature of this diagram is the repeated
474: appearance of cusps.
475: They are observed whenever $\tau$ becomes equal to an integer multiple
476: of the period of the unperturbed uniform system,
477: $\tau=k\,2\pi/(\omega+\beta),k=1,2,3,...$ .
478: Figure~\ref{fg:diagram_large} shows an extension of the top part of
479: Fig.~\ref{fg:diagram} towards large delay times.
480: With increasing $\tau$, the cusps get less pronounced and the
481: boundary converges to a flat line at $\mu\approx0.16$.
482:
483: According to this analytically derived synchronization diagram, stability
484: of uniform oscillations can also be maintained by applying global feedbacks
485: with very large delay times.
486: Moreover, the critical feedback strength, needed to maintain
487: synchronization, does not depend on the delay in the limit
488: $\tau \rightarrow \infty$.
489: To understand this result, we note
490: that the feedback signal $F(t)$ for $\tau \rightarrow \infty$
491: is given by
492: %
493: \begin{equation}
494: \label{eq:taulimit}
495: F(t)=\mu e^{i\chi }\left[\rho_{0}e^{-i(\Omega t+\phi_{0})}
496: -\bar{\eta}(t)\right]
497: \end{equation}
498: %
499: where $\phi_{0}$ is a constant phase shift.
500: The first term here corresponds to the state
501: $\bar{\eta}(t-\tau)$ at $\tau \rightarrow \infty$
502: which is essentially the initial state of the system.
503: When destabilization of initally uniform oscillations is considered to
504: determine the stability boundary, this initial state represents uniform
505: oscillations with frequency $\Omega$ and amplitude $\rho_{0}$.
506: Substituting this expression for $F(t)$ into Eq.~(\ref{eq:cglefeed}),
507: we see that then a situation with external periodic forcing is effectively
508: realized.
509: The critical value $\mu_{c}$ corresponds in this case to the
510: minimum forcing intensity needed to maintain uniform oscillations in
511: the system.
512:
513: In Fig.~\ref{fg:diagram}(b,c), the critical
514: wavenumber and the frequency are shown, respectively, as functions of
515: the delay time along the lower part ABC of the stability boundary.
516: The two curves are of similar shape:
517: they display a decrease for increasing delay time interrupted
518: by discontinuous jumps.
519: These discontinuities occur at the locations where the cusps are
520: found in the stability boundary in Fig.~\ref{fg:diagram}(a).
521:
522: In Fig.~\ref{fg:lambdak}, we show the real part of $\lambda$ as a
523: function of $\kappa$ at three different points on the stability curve
524: in the ($\mu,\tau$) plane.
525: The three cases correspond to what has been found earlier for
526: a different global delayed feedback scheme in the
527: CGLE~\cite{battogtokh-mikhailov-1996}.
528: On the branch AB (and similarly also on the branches BC and to the
529: right of C), the first unstable modes occur with a wavenumber
530: $\kappa_0\ne0$ where $\mbox{Im}(\lambda)=0$, see
531: Fig.~\ref{fg:lambdak}(b,c).
532: Thus, if we cross this branch of the stability boundary by reducing the
533: feedback intensity below the critical value $\mu_c$, uniform
534: oscillations will become unstable and standing waves with
535: wavenumber $\kappa_0$ will emerge (cf. Sec.~\ref{sc:numerics}).
536: A different situation is encountered on the branch reaching
537: from A upwards, see Fig.~\ref{fg:lambdak}(a).
538: Here, the instability will occur by periodic spatiotemporal
539: modulations of uniform oscillations, since
540: $\mbox{Im}(\lambda)\ne0$ and the most unstable modes will have
541: wavenumbers close to $\kappa_0=0$.
542: %
543: %---------------------------------------------------------------------
544: \section{Numerical simulations}
545: \label{sc:numerics}
546: %
547: All simulations were carried out for a one-dimensional system of
548: length $L=128$ on an equidistant grid with $\Delta x=0.32$
549: (corresponding to a total number of 400 grid points).
550: We use a second-order finite-difference scheme for the
551: discretization of the Laplacian operator and impose periodic
552: boundary conditions.
553: An explicit Euler scheme with a fixed time step of $\Delta t=0.001$
554: was employed for integration.
555: As initial condition, a uniform state superposed with a small
556: spatially inhomogeneous perturbation was chosen.
557: The set of parameters is as in the previous section:
558: $\varepsilon=2$, $\beta=-1.4$, $\omega=2\pi-\beta\approx7.68$,
559: and $\chi=\pi/2$.
560: The choice of the feedback parameters $\mu$ and $\tau$ is different
561: for the different simulations and specified below.
562:
563: In the first series of simulations, we systematically scan the parameters
564: $\mu$ and $\tau$ in steps of $\Delta\mu=\Delta\tau=0.1$ between
565: 0 and 2.5, respectively, to verify the shape of the stability domain
566: in the ($\mu, \tau$) plane.
567: When starting from uniform initial conditions,
568: the stability diagram displayed in Fig.~\ref{fg:diagram}(a) is nicely
569: reproduced after transients.
570: Asymptotic states are uniform inside the domain of stability and nonuniform
571: outside this region.
572: These nonuniform states are, however, of different types.
573: Far from the region of stability, we asymptotically reach a fully
574: developed state of defect-mediated turbulence.
575: As the stability boundary is approached, phase turbulence and
576: standing wave patterns are observed close to the branches
577: AB, BC, and to the right of C.
578:
579: Figure~\ref{fg:regular} shows the results of a more detailed scan
580: of $\mu$ and $\tau$ in the vicinity of the branch AB of the
581: synchronization diagram in Fig.~\ref{fg:diagram}(a).
582: Again, we start from uniform initial conditions.
583: The feedback parameters $\mu$ and $\tau$ are changed in
584: steps of $\Delta\mu=\Delta\tau=0.025$ between $\tau=0.25\,...\,0.975$
585: and $\mu=0\,...\,0.7$.
586: Simulations that, after transients, resulted in a regular nonuniform
587: spatiotemporal state are marked by bold dots at the corresponding
588: ($\mu, \tau$) coordinates
589: (simulations leading to a uniformly oscillating or turbulent
590: asymptotic states are not shown).
591: Obviously, the parameter region, where regular spatiotemporal patterns
592: occur, constitutes a slightly asymmetric prolongation of the
593: tongue-shaped stability domain for uniform oscillations towards
594: smaller feedback intensities.
595: From the result of the first coarse parameter scan we conjecture, that
596: the regions where spatiotemporal patterns occur look qualitatively similar
597: at the other tongue-shaped branches of the stability curve.
598:
599: When starting from turbulent initial conditions, the stability
600: boundary for uniform oscillations is moved towards larger
601: feedback intensities in the vicinity of the cusps.
602: The results of numerical simulations are summarized in
603: Fig.~\ref{fg:desyn}.
604: Here, open circles indicate a turbulent asymptotic state and bold
605: circles again denote the appearance of a regular nonuniform wave pattern.
606: Inside the area where no symbols are shown, simulations
607: converge towards uniform oscillations.
608: For comparison, the analytically derived synchronization diagram from
609: Fig.~\ref{fg:diagram}(a) is also displayed.
610:
611: To get a better idea of how the spatiotemporal dynamics of the system
612: changes in the
613: parameter range between turbulence and uniform oscillations, we show in
614: Fig. \ref{fg:patterns} a series of spacetime plots of asymptotic
615: dynamical states reached for different feedback intensities at a fixed delay
616: time of $\tau=0.5$.
617: The amplitude $|\eta|$ is plotted in a grey scale color coding and
618: the feedback intensity $\mu$ is increased in Fig.~\ref{fg:patterns} from
619: (a) to (e).
620: In absence of feedback (Fig.~\ref{fg:patterns}(a), $\mu=0$)
621: and for small feedback intensities (Fig.~\ref{fg:patterns}(b),
622: $\mu=0.05$), an irregular state of defect-mediated turbulence is
623: observed.
624: However, the number of defects is smaller in the presence of a weak
625: feedback and the
626: time evolution shows intervalls, where almost no defects are seen.
627: If the feedback intensity is increased
628: (Fig.~\ref{fg:patterns}(c), $\mu=0.07$), defects are no
629: longer observed and the system displays a disordered state of phase
630: turbulence.
631: For even stronger feedback, breathing
632: (Fig.~\ref{fg:patterns}(d), $\mu=0.1$) and stationary
633: standing wave patterns
634: (Fig.~\ref{fg:patterns}(e), $\mu=0.15$) can be observed.
635: For still larger feedbacks, $\mu>0.2$ (not shown), uniform oscillations
636: take place.
637:
638: To study the behaviour of the
639: system for very large delays, a series of simulations with the delays
640: $\tau$ varying from 0.5 to 29.5 in steps of $\Delta\tau=1$ has further
641: been performed.
642: The feedback intensity $\mu$ was varied for each choice of $\tau$
643: from 0.1 to 0.3 in the steps of $\Delta\mu =0.1$.
644: The initial conditions in these
645: simulations represented the state of amplitude turbulence reached by
646: the system if $\mu=0$.
647: These numerical investigations have revealed
648: that synchronization is possible for all chosen delays, when the
649: feedback intensity exceeds a critical level.
650: In another series of experiments, the feedback intensity was fixed
651: at $\mu=1$ and the delay time $\tau$ was increased from 0 to 50 in
652: steps of 0.1.
653: Here, the transient time was determined for each of the simulations by
654: looking at the convergence of the statistical variance of the
655: amplitude $\rho$.
656: It was found that for small delays the transient time increases
657: proportionally to the delay $\tau$.
658: For large delays, the transient time undergoes saturation at a level
659: independent of $\tau$, but depending on the intensity of the applied
660: feedback.
661:
662: For very long delays $\tau$, the
663: component $\bar{\eta}(t-\tau)$ in the feedback signal $F(t)$ at
664: time $t$ corresponds to the initial state of amplitude turbulence.
665: Thus, the global delayed feedback scheme in the considered
666: limit effectively represents global forcing of the system with an
667: external chaotic signal corresponding to the initial turbulent state in
668: absence of feedback.
669: To verify this conjecture, special numerical simulations have been
670: performed.
671: In these simulations, the feedback
672: signal was generated by replacing $\bar{\eta}(t-\tau)$ with the
673: average complex oscillation amplitude of the
674: same system without the feedback.
675: We have found that, by applying such chaotic external forcing,
676: synchronization of uniform oscillations can also be achieved.
677: %
678: %---------------------------------------------------------------------
679: \section{Discussion}
680: \label{sc:discussion}
681: %
682: Motivated by our experiments~\cite{beta-et-al-2003} on control of chemical
683: turbulence in the CO oxidation
684: reaction on Pt(110), we have performed a detailed study of the effects of
685: time-delay autosynchronization on uniform oscillations in a general
686: model described by the complex Ginzburg-Landau equation.
687: We have seen
688: that, like in the experiments, the noninvasive stabilization of uniform
689: oscillations by this method is not possible, though the required
690: magnitude of the feedback signal can be significantly reduced by using
691: an optimal delay time.
692: The obtained synchronization diagram exhibits a
693: series of cusps at the values of delay times equal to integer multiples
694: of the oscillation period of the unperturbed uniform system.
695: This feature seems to be common for various oscillatory systems with delayed
696: coupling and has also been found in the case of the Kuramoto model of
697: phase oscillators with a delayed global coupling~\cite{yeung-strogatz-1999}.
698: Near the
699: synchronization boundary, the formation of standing
700: waves and a state of modified intermittent turbulence were
701: observed.
702: These effects are similar to those previously discussed for a
703: different feedback scheme~\cite{bertram-mikhailov-2003,battogtokh-mikhailov-1996,battogtokh-preusser-mikhailov-1997}.
704: In contrast to our previous studies, the action of feedbacks with long
705: delay times has also been investigated.
706: We have shown that such feedbacks are also capable
707: of synchronizing oscillations and this effect can essentially be
708: explained as suppression of turbulence by external noisy forcing.
709: %
710: %---------------------------------------------------------------------
711: %
712: %\bibliography{sources}
713: \begin{thebibliography}{10}
714: \expandafter\ifx\csname url\endcsname\relax
715: \def\url#1{\texttt{#1}}\fi
716: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi
717:
718: \bibitem{cross-hohenberg-1993}
719: M.~C. Cross, P.~C. Hohenberg, Pattern formation outside of equilibrium, Rev.
720: Mod. Phys. 65 (1993) 851--1112.
721:
722: \bibitem{mikhailov-book1}
723: A.~S. Mikhailov, Foundations of Synergetics I, Distributed Active Systems,
724: Springer, Berlin, 1994.
725:
726: \bibitem{zaikin-zhabotinsky-1970}
727: A.~Zaikin, A.~Zhabotinsky, Concentration wave propagation in two--dimensional
728: liquid--phase self--oscillating system, Nature 225 (1970) 535--537.
729:
730: \bibitem{winfree-1972}
731: A.~Winfree, Spiral waves of chemical activity, Science 175 (1972) 634--636.
732:
733: \bibitem{imbihl-ertl-1995}
734: R.~Imbihl, G.~Ertl, Oscillatory kinetics in heterogeneous catalysis, Chem. Rev.
735: 95 (1995) 697--733.
736:
737: \bibitem{schoell-book-2001}
738: E.~Sch{\"o}ll, Nonlinear Spatio-Temporal Dynamics and Chaos in Semiconductors,
739: Cambridge University Press, Cambridge, 2001.
740:
741: \bibitem{benjacob-cohen-levine-2000}
742: E.~Ben-Jacob, I.~Cohen, H.~Levine, Cooperative self--organization of
743: microorganisms, Adv. Phys. 49~(4) (2000) 395--554.
744:
745: \bibitem{kuramoto-book}
746: Y.~Kuramoto, Chemical Oscillations, Waves, and Turbulence, Springer, Berlin,
747: 1984.
748:
749: \bibitem{aranson-kramer-2002}
750: I.~S. Aranson, L.~Kramer, The world of the complex {G}inzburg-{L}andau
751: equation, Rev. Mod. Phys. 74 (2002) 99--143.
752:
753: \bibitem{coullet-gil-lega-1989}
754: P.~Coullet, L.~Gil, J.~Lega, Defect-mediated turbulence, Phys. Rev. Lett. 62
755: (1989) 1619--1622.
756:
757: \bibitem{schuster-book-1999}
758: H.~Schuster, Handbook of Chaos Control, Wiley-VCH, Weinheim, 1999.
759:
760: \bibitem{ott-grebogi-yorke-1990}
761: E.~Ott, C.~Grebogi, J.~A. Yorke, Controlling chaos, Phys. Rev. Lett. 64~(11)
762: (1990) 1196--1199.
763:
764: \bibitem{ditto-rauseo-spano-1990}
765: W.~L. Ditto, S.~N. Rauseo, M.~L. Spano, Experimental control of chaos, Phys.
766: Rev. Lett. 65 (1990) 3211--3214.
767:
768: \bibitem{carroll-et-al-1992}
769: T.~L. Carroll, I.~Triandaf, I.~Schwartz, L.~Pecora, Tracking unstable orbits in
770: an experiment, Phys. Rev. A 46 (1992) 6189--6192.
771:
772: \bibitem{petrov-et-al-1993}
773: V.~Petrov, V.~G\'{a}sp\'{a}r, J.~Masere, K.~Showalter, Controlling chaos in the
774: {B}elousov-{Z}habotinsky reaction, Nature 361 (1993) 240--243.
775:
776: \bibitem{pyragas-1992}
777: K.~Pyragas, Continuous control of chaos by self--controlling feedback, Phys.
778: Lett. A 170 (1992) 421--428.
779:
780: \bibitem{socolar-sukow-gauthier-1994}
781: J.~E.~S. Socolar, D.~W. Sukow, D.~J. Gauthier, Stabilizing unstable periodic
782: orbits in fast dynamical systems, Phys. Rev. E 50~(4) (1994) 3245--3248.
783:
784: \bibitem{pierre-bonhomme-atipo-1996}
785: T.~Pierre, G.~Bonhomme, A.~Atipo, Controlling the chaotic regime of nonlinear
786: ionization waves using the time--delay autosynchronization method, Phys. Rev.
787: Lett. 76~(13) (1996) 2290--2293.
788:
789: \bibitem{franceschini-bose-schoell-1999}
790: G.~Franceschini, S.~Bose, E.~Sch{\"o}ll, Control of chaotic spatiotemporal
791: spiking by time--delay autosynchronization, Phys. Rev. E 60~(5) (1999)
792: 5426--5434.
793:
794: \bibitem{parmananda-et-al-1999}
795: P.~Parmananda, R.~Madrigal, M.~Rivera, L.~Nyikos, I.~Kiss, V.~G\'{a}sp\'{a}r,
796: Stabilization of unstable steady states and periodic orbits in an
797: electrochemical system using delayed--feedback control, Phys. Rev. E 59~(5)
798: (1999) 5266--5271.
799:
800: \bibitem{beck-et-al-2002}
801: O.~Beck, A.~Amann, E.~Sch{\"o}ll, J.~Socolar, W.~Just, Comparison of
802: time-delayed feedback schemes for spatiotemporal control of chaos in a
803: reaction-diffusion system with global coupling, Phys. Rev. E 66 (2002)
804: 016213.
805:
806: \bibitem{baba-et-al-2002}
807: N.~Baba, A.~Amann, E.~Sch{\"o}ll, Giant improvement of time--delayed feedback
808: control by spatio--temporal filtering, Phys. Rev. Lett. 89~(7) (2002) 074101.
809:
810: \bibitem{just-et-al-2003}
811: W.~Just, S.~Popovich, A.~Amann, N.~Baba, E.~Sch{ö}ll, Improvement of
812: time-delayed feedback control by periodic modulation: analytical theory of
813: {Floquet} mode control scheme, Phys. Rev. E 67 (2003) 026222.
814:
815: \bibitem{harrington-socolar-2001}
816: I.~Harrington, J.~E.~S. Socolar, Limitation on stabilizing plane waves via
817: time-delay feedback, Phys. Rev. E 64 (2001) 056206.
818:
819: \bibitem{montgomery-silber-2003pre}
820: K.~A. Montgomery, M.~Silber, Feedback control of traveling wave solutions in
821: the complex {G}inzburg-{L}andau equation, Nonlinearity, {s}ubmitted (cf.
822: http://arxiv.org/abs/nlin.PS/0308021).
823:
824: \bibitem{beta-et-al-2003}
825: C.~Beta, M.~Bertram, A.~S. Mikhailov, H.~H. Rotermund, G.~Ertl, Controlling
826: turbulence in a surface chemical reaction by time-delay autosynchronization,
827: Phys. Rev. E 67 (2003) 046224.
828:
829: \bibitem{jakubith-et-al-1990}
830: S.~Jakubith, H.~H. Rotermund, W.~Engel, A.~v.~Oertzen, G.~Ertl, Spatiotemporal
831: concentration patterns in a surface reaction: Propagating and standing waves,
832: rotating spirals, and turbulence, Phys. Rev. Lett. 65~(24) (1990) 3013--3016.
833:
834: \bibitem{krischer-eiswirth-ertl-1992}
835: K.~Krischer, M.~Eiswirth, G.~Ertl, Oscillatory {CO} oxidation on {P}t(110):
836: Modeling of temporal self--organization, J. Chem. Phys. 96~(12) (1992)
837: 9161--9172.
838:
839: \bibitem{baer-et-al-1992}
840: M.~B{\"a}r, M.~Eiswirth, H.~Rotermund, G.~Ertl, Solitary--wave phenomena in an
841: excitable surface reaction, Phys. Rev. Lett. 69~(6) (1992) 945--948.
842:
843: \bibitem{kim-et-al-2001}
844: M.~Kim, M.~Bertram, M.~Pollmann, A.~v.~Oertzen, A.~S. Mikhailov, H.~H.
845: Rotermund, G.~Ertl, Controlling chemical turbulence by global delayed
846: feedback: Pattern formation in catalytic {CO} oxidation on {P}t(110), Science
847: 292 (2001) 1357--1360.
848:
849: \bibitem{bertram-et-al-2003}
850: M.~Bertram, C.~Beta, M.~Pollmann, A.~S. Mikhailov, H.~H. Rotermund, G.~Ertl,
851: Pattern formation on the edge of chaos: Experiments with {CO} oxidation on a
852: {P}t(110) surface under global delayed feedback, Phys. Rev. E 67 (2003)
853: 036208.
854:
855: \bibitem{bertram-mikhailov-2003}
856: M.~Bertram, A.~S. Mikhailov, Pattern formation on the edge of chaos:
857: Mathematical modeling of {CO} oxidation on a {P}t(110) surface under global
858: delayed feedback, Phys. Rev. E 67 (2003) 036207.
859:
860: \bibitem{battogtokh-mikhailov-1996}
861: D.~Battogtokh, A.~S. Mikhailov, Controlling turbulence in the complex
862: {G}inzburg--{L}andau equation, Physica D 90 (1996) 84--95.
863:
864: \bibitem{battogtokh-preusser-mikhailov-1997}
865: D.~Battogtokh, A.~Preusser, A.~S. Mikhailov, Controlling turbulence in the
866: complex {G}inzburg--{L}andau equation {II}. {T}wo--dimensional systems,
867: Physica D 106 (1997) 327--362.
868:
869: \bibitem{yeung-strogatz-1999}
870: M.~K. Yeung, S.~H. Strogatz, Time delay in the {K}uramoto model of coupled
871: oscillators, Phys. Rev. Lett. 83 (1999) 648--651.
872:
873: \end{thebibliography}
874: %
875: %---------------------------------------------------------------------
876: \newpage
877: \section{Figures}
878: \label{sc:fig}
879: %
880: \begin{figure}[htbh]
881: \begin{center}
882: \includegraphics[width=10cm]{exp.eps}
883: \end{center}
884: \caption{Experimental results on the application of TDAS to the
885: catalytic CO oxidation on Pt(110).
886: Period $T$ of homogeneous oscillations (black squares) and feedback
887: magnitude $M$ (open circles) in dependence on the delay time $\tau$.
888: Reproduced from Ref.~\cite{beta-et-al-2003}.
889: }
890: \label{fg:exp}
891: \end{figure}
892: \vspace{10mm}
893: %
894: \begin{figure}[htbh]
895: \setlength{\unitlength}{1mm}
896: \begin{center}
897: \begin{picture}(160,60)
898: \put(65,45){a}
899: \put(145,45){b}
900: \put(80,0){\includegraphics[width=7.5cm]{hyst.eps}}
901: \put(0,0){\includegraphics[width=7.5cm]{period.eps}}
902: \end{picture}
903: \end{center}
904: \caption{Results of numerical integration of~Eq.(\ref{eq:h}).
905: The oscillation period is plotted as a function of the delay time
906: for (a) low ($\mu=0.5$) and (b) high ($\mu=1.0$) feedback intensities.
907: The parameters are
908: $\beta=-1.4$,
909: $\omega=2\pi-\beta\approx7.68$, and
910: $\chi=\pi/2$.}
911: \label{fg:period}
912: \end{figure}
913: %
914: \begin{figure}[htbh]
915: \begin{center}
916: \includegraphics[width=7cm,angle=-90]{bifur.ps}
917: \end{center}
918: \caption{Bifurcation diagram for $\tau=T_0=1$.
919: The parameters are as in Fig. \ref{fg:period}.
920: Dotted lines denote unstable branches.
921: The uniform solution with $\Omega=\Omega_0=2\pi$ and
922: a vanishing feedback signal undergoes a transcritical
923: bifurcation at $\mu=\mu_0$ and becomes unstable.}
924: \label{fg:bifur}
925: \end{figure}
926: %
927: \begin{figure}
928: \begin{center}
929: \includegraphics[width=10cm]{mu_vs_eps.eps}
930: \end{center}
931: \caption{Critical feedback intensity $\mu_c$ for $\tau=T_0=1$
932: as a function of
933: the dispersion coefficient $\varepsilon$ with $\beta=-1.4$.
934: The line $\mu=\mu_0$ denotes the feedback intensity for which the
935: transcritical bifurcation occurs in the uniform system (cf.
936: Fig.~\ref{fg:bifur}).
937: The other parameters are as in Fig.~\ref{fg:period}.}
938: \label{fg:mueps}
939: \end{figure}
940: %
941: \begin{figure}[htbh]
942: \setlength{\unitlength}{1mm}
943: \begin{center}
944: \begin{picture}(110,110)
945: \put(2,57){\includegraphics[width=10cm]{diagram.eps}}
946: \put(91,108){a}
947: \put(91,56.5){b}
948: \put(91,29){c}
949: \put(19,81){A}
950: \put(45,87){B}
951: \put(78,78){C}
952: \put(0,89){$\mu_c$}
953: \put(0,49){$\kappa_c$}
954: \put(2,30){\includegraphics[width=10cm]{k_vs_tau.eps}}
955: \put(0,22){$\Omega_c$}
956: \put(6.4,3){\includegraphics[width=9.53cm]{omega_vs_tau.eps}}
957: \put(54,0){$\tau$}
958: \end{picture}
959: \end{center}
960: \caption{(a) Synchronization diagram.
961: Uniform oscillations are stable inside the shaded region.
962: (b,c) The dependences of the critical wavenumber $\kappa_c$ and the
963: critical frequency $\Omega_c$ on the delay time $\tau$.
964: The parameters are
965: $\varepsilon=2$,
966: $\beta=-1.4$,
967: $\omega=2\pi-\beta\approx7.68$, and
968: $\chi=\pi/2$.
969: }
970: \label{fg:diagram}
971: \end{figure}
972: %
973: \begin{figure}
974: \begin{center}
975: \includegraphics[width=10cm]{diagram_large.eps}
976: \end{center}
977: \caption{Extended synchronization diagram.
978: The same parameters as in Fig.~\ref{fg:diagram}.}
979: \label{fg:diagram_large}
980: \end{figure}
981: %
982: \begin{figure}[htbh]
983: \setlength{\unitlength}{1mm}
984: \hspace{-1.9cm}
985: \begin{picture}(120,60)
986: \put(50,43){a}
987: \put(110,43){b}
988: \put(170,43){c}
989: \put(0,0){\includegraphics[width=8cm]{lambda1.eps}}
990: \put(60,0){\includegraphics[width=8cm]{lambda2.eps}}
991: \put(120,0){\includegraphics[width=8cm]{lambda3.eps}}
992: \end{picture}
993: \caption{Growth rate $\mbox{Re}\,\lambda$ of spatially nonuniform
994: modes as a function of wavenumber $\kappa$ at three different points
995: on the stability curve displayed in Fig.~\ref{fg:diagram}(a).
996: The parameters are $\tau=0.29$, $\mu=1.29$ (a),
997: $\tau=0.36$, $\mu=0.4$ (b), and
998: $\tau=0.7$, $\mu=0.16$ (c);
999: other parameters as in Fig. \ref{fg:diagram}.
1000: }
1001: \label{fg:lambdak}
1002: \end{figure}
1003: %
1004: \begin{figure}
1005: \begin{center}
1006: \includegraphics[width=10cm]{regular.eps}
1007: \end{center}
1008: \caption{Regular spatiotemporal patterns at the border of complete
1009: synchronization for uniform initial conditions.
1010: Section of the ($\mu,\tau$) plane with the analytical result for the
1011: AB branch of the stability curve (cf. Fig \ref{fg:diagram}) plotted
1012: as a solid line.
1013: Bold dots mark the parameters where regular spatiotemporal patters
1014: are observed.}
1015: \label{fg:regular}
1016: \end{figure}
1017: %
1018: \begin{figure}
1019: \begin{center}
1020: \includegraphics[width=10cm]{desyn.eps}
1021: \end{center}
1022: \caption{Summary of numerical simulations starting from turbulent
1023: initial conditions.
1024: Asymptotic states are shown in the ($\mu,\tau$) plane.
1025: Open circles denote amplitude turbulence and bold circles indicate
1026: regular nonuniform wave patterns.
1027: Symbols are omitted in the areas where computations converge to
1028: uniform oscillations.
1029: The solid line shows the analytically derived stability boundary
1030: for uniform initial conditions.
1031: The parameters are as in Fig.~\ref{fg:diagram}.}
1032: \label{fg:desyn}
1033: \end{figure}
1034: %
1035: \begin{figure}
1036: \setlength{\unitlength}{1mm}
1037: \begin{picture}(140,140)
1038: \put(132,121){(a)}
1039: \put(0,108){\frame{\includegraphics[width=13cm,height=2.5cm]{amplitude0000small.eps}}}
1040: \put(132,94){(b)}
1041: \put(0,81){\frame{\includegraphics[width=13cm,height=2.5cm]{amplitude0051small.eps}}}
1042: \put(132,67){(c)}
1043: \put(0,54){\frame{\includegraphics[width=13cm,height=2.5cm]{amplitude0070small.eps}}}
1044: \put(132,40){(d)}
1045: \put(0,27){\frame{\includegraphics[width=13cm,height=2.5cm]{amplitude0100small.eps}}}
1046: \put(132,13){(e)}
1047: \put(0,0){\frame{\includegraphics[width=13cm,height=2.5cm]{amplitude0150small.eps}}}
1048: \put(-12,41){space}
1049: \put(-3,-3){\vector(1,0){40}}
1050: \put(40,-4){time}
1051: \put(-3,-3){\vector(0,1){40}}
1052: \end{picture}
1053: \vspace{1mm}
1054: \caption{Spatiotemporal patterns in the parameter region between turbulence
1055: and complete synchronization.
1056: The amplitude $|\eta|$ is displayed in a grey scale coding, where
1057: black (white) denotes low (high) values of the real amplitude.
1058: The delay time is kept constant, $\tau=0.5$ s, and the feedback intensity
1059: increases from top to bottom, (a) $\mu=0$, (b) $\mu=0.05$, (c) $\mu=0.07$,
1060: (d) $\mu=0.1$, and (e) $\mu=0.15$.}
1061: \label{fg:patterns}
1062: \end{figure}
1063: %
1064: %---------------------------------------------------------------------
1065: \end{document}
1066: %---------------------------------------------------------------------
1067: % this should be the end
1068: