1: \documentclass{elsart}

2: \usepackage[english]{babel}

3: \usepackage{graphicx,subfigure}

4: \usepackage{amsfonts,amsmath}

5: \usepackage{comment}

6: \graphicspath{{./Figsnew2/}}

7: \newcommand{\im}{\mathrm{i}}

8: \newcommand{\CC}{\mathcal C}

9: \newcommand{\co}{\cos(\psi_0)}

10: \newcommand{\so}{\sin(\psi_0)}

11: \begin{document}

12: \begin{frontmatter}

13: \title{Birhythmicity, Synchronization, and Turbulence in an Oscillatory

14: System with Nonlocal Inertial Coupling}

15: \author{Vanessa Casagrande\thanksref{kuramoto}}

16: and

17: \author{Alexander S. Mikhailov\corauthref{cor}\thanksref{kuramoto}}

18: \address{Abteilung Physikalische Chemie,\\

19: Fritz-Haber-Institut der Max-Planck-Gesellschaft,\\

20: Faradayweg 4-6, 14195 Berlin, Germany}

21: \corauth[cor]{Corresponding author}

22: \thanks[kuramoto]{Dedicated to Prof. Y. Kuramoto on the occasion of 

23: his retirement}

24: \ead{mikhailov@fhi-berlin.mpg.de}

25: \begin{abstract}

26: We consider a model where a population of diffusively coupled 

27: limit-cycle oscillators, described by the complex Ginzburg-Landau 

28: equation, interacts nonlocally via an inertial field. 

29: For sufficiently high intensity of nonlocal inertial coupling, the 

30: system exhibits birhythmicity with two oscillation modes at largely 

31: different frequencies. Stability of uniform oscillations in the 

32: birhythmic region is analyzed by means of the phase

33: dynamics approximation. Numerical simulations show that, depending on 

34: its parameters, the system has irregular intermittent regimes with 

35: local bursts of synchronization or desynchronization. 

36: \end{abstract}

37: \end{frontmatter}

38: 

39: 

40: \section{Introduction}

41: Beginning with the pioneering contributions by Kuramoto \cite{kurbook} 

42: and Winfree \cite{winfree}, studies of synchronization and 

43: spatiotemporal chaos (turbulence) in populations of coupled active

44: oscillators have developed into a classical field of research. 

45: In chemical systems, each oscillator 

46: represents a reaction element and coupling between such elements is 

47: usually due to diffusion of reactants in the system. This coupling 

48: extends only within a short diffusion length and is therefore local. 

49: The complex Ginzburg-Landau equation is the canonical model for 

50: oscillatory systems with local coupling

51: near a supercritical Hopf bifurcation. In surface chemical reactions

52: \cite{ertl}, reaction elements are however additionally interacting

53: through the gas phase where instantaneous complete mixing occurs. 

54: As a result, global coupling between chemical oscillators arises 

55: \cite{global}.

56: Moreover, global delayed feedbacks through the gas phase could be

57: artificially introduced to control turbulence in surface reactions 

58: \cite{batt96,batt97,kawamura,science,bertram1,bertram2}.

59: A special class of systems are arrays of active oscillators that do not

60: directly interact one with another, but are all coupled to a single

61: diffusing field (in biochemistry, an example of such a system would be

62: provided by arrays of allosteric enzymes interacting through a chemical

63: messenger \cite{enzymes}). 

64: Adiabatically eliminating this field, models with nonlocal

65: coupling between the oscillators, described by a finite-range integral,

66: were derived \cite{kur95,kur98}. 

67: Investigations of nonlocally coupled oscillator

68: arrays have revealed that such form of coupling does not always 

69: synchronize neighbouring oscillators and discontinuous distributions 

70: with scaleless fractal structure can develop \cite{kur97,kur98}. 

71: Spatiotemporal chaos can develop in such systems even in the 

72: Benjamin-Feir stable regime \cite{batt00}

73: and spiral waves with phase-randomized cores can exist here 

74: \cite{shima,chapter9}.

75: 

76: Furthermore, arrays with both local and nonlocal coupling between

77: oscillators are possible. This situation is characteristic, for example,

78: for surface chemical reactions. In such systems, diffusion provides local

79: coupling between neighbouring surface oscillators, whereas much faster

80: heat conduction is responsible for nonlocal coupling between them. 

81: 

82: Recently, Tanaka and Kuramoto \cite{tanaka} have shown how, in the 

83: vicinity of a supercritical Hopf bifurcation, the description of 

84: arrays of nonlocally coupled oscillators can be reduced to the complex 

85: Ginzburg-Landau equation with nonlocal coupling. Because of the critical

86: slowing down near the bifurcation point, the coupling is effectively 

87: instantaneous in the reduced description.

88: 

89: In realistic systems, which are not too close to the Hopf bifurcation, one

90: can however also encounter the opposite situation, with a very slow

91: inertial field giving rise to nonlocal coupling. The aim of our study is

92: to analyze what new effects, primarily due to slow nonlocal coupling, are

93: possible in this class of systems. For our investigations, we have chosen

94: an abstract model of the complex Ginzburg-Landau equation interacting with

95: an additional slow field, so that the coupling is nonlocal both with

96: respect to space and time. We expect that the behaviour found in this

97: general model would be typical for a broad class of realistic systems. 

98: 

99: The principal effect of coupling inertiality is that birhythmicity,

100: leading to chaotic intermittency, can develop in such systems. In

101: contrast to the birhythmicity in reaction-diffusion systems near a 

102: pitchfork-Hopf bifurcation (see \cite{michael01,michael02}), the two 

103: oscillatory states 

104: are characterized here by very different frequencies and oscillation 

105: amplitudes. For the rapid oscillatory state, the additional field 

106: responsible for the inertial long-range coupling is almost absent

107: and only diffusive local coupling between the oscillators is 

108: important. In slow oscillations, the elements are however entrained by 

109: the long-range inertial field.

110: 

111: Depending on the parameters, two new kinds of intermittency are found

112: here. When rapid oscillations are modulationally unstable and give rise to

113: turbulence, spatial regions occupied by slow entrained oscillations

114: spontaneously develop and die out in the medium. They can be interpreted

115: as bursts of synchronization on a turbulent background. On the other hand,

116: relatively small, irregularly evolving islands filled with rapid

117: turbulence can persist on the background of slow almost uniform

118: oscillations. They can be therefore described as bursts of

119: desynchronization on the background of regular slow oscillations.  Our

120: analysis of such phenomena is based on the phase dynamics approximation,

121: derived for the birhythmic system. It is complemented by 1D and 2D

122: numerical simulations.

123: 

124: \section{The Model}

125: 

126: The investigated model describes a system of diffusively coupled active

127: oscillators coupled to an additional inertial diffusive field. 

128: Introducing the local complex oscillation amplitude $\eta (x,t)$ and 

129: denoting as $z(x,t)$ the additional complex-valued diffusive field, 

130: we have therefore a system of two equations 

131: \begin{subequations}

132: \begin{align}

133: &\dot\eta&=&\quad(1+\im\omega)\eta-(1+\im\alpha)|\eta|^2\eta+

134: (1+\im\beta)\nabla^2\eta+K(z-\eta)\label{model:one}\\

135: &\tau\dot z&=&\quad\eta-z+l^2\nabla ^2 z\label{model:two}

136: \end{align}\label{model}

137: \end{subequations}

138: The additional last term in the complex Ginzburg-Landau equation

139: (\ref{model:one}) takes into account coupling of the oscillatory 

140: subsystem with the field $z$ whose evolution obeys equation 

141: (\ref{model:two}); $K$ is the respective coupling constant. 

142: The equations are brought into a dimensionless form by choosing

143: the characteristic diffusion length in the oscillatory subsystem as the

144: length unit and taking the characteristic relaxation time scale of the

145: oscillators as the time unit. The parameters $\tau$ and $l$ determine

146: characteristic time and length scales of the additional field $z$. 

147: We assume that this field is inertial ($\tau \gg 1$) and slowly 

148: varying in space ($l\gg 1$).

149: 

150: The linear equation (\ref{model:two}) can easily be solved as 

151: \begin{equation}

152: z(x,t)=\int_{-\infty }^{\infty }\int_{0}^{t}G(x-x^{\prime },t-t^{\prime

153: })\eta (x^{\prime },t^{\prime })dx^{\prime }dt^{\prime }.

154: \end{equation}

155: where the kernel is given by 

156: \begin{equation}

157: G(x,t)=\frac{1}{2\tau \sqrt{\pi Dt}}\exp \left( -\frac{x^{2}}{4Dt}-

158: \frac{t}{\tau }\right) .  \label{kernel}

159: \end{equation}

160: Substituting this into equation (\ref{model:one}), we obtain an 

161: equivalent integro-differential form of the considered model, 

162: \begin{eqnarray}

163: \dot\eta &=&(1+\im\omega)\eta-(1+\im\alpha)|\eta|^{2}\eta 

164: +(1+\im\beta)\nabla^2\eta  \nonumber\\

165: &&+K\int_{-\infty }^{\infty }\int_{0}^{t}G(x-x^{\prime },t-t^{\prime

166: })\left[ \eta (x^{\prime },t^{\prime })-\eta (x,t)\right] dx^{\prime

167: }dt^{\prime } \label{integralform}

168: \end{eqnarray}

169: We see that, besides of diffusion, the model also includes an 

170: additional coupling, nonlocal both with respect to space and time.

171: 

172: Note that very close to a supercritical Hopf bifurcation all processes

173: become fast as compared to the relaxation time scale of individual

174: oscillators, because this time is inversely proportional to the 

175: distance from the bifurcation point (critical slowing down). 

176: However, it is known

177: that the complex Ginzburg-Landau equation yields qualitatively correct

178: description even relatively far from the bifurcation. 

179: Equations (\ref{model}) and (\ref{integralform}) can be viewed as 

180: providing a simple model of an oscillating system coupled to an 

181: inertial diffusive field.

182: 

183: 

184: \section{Birhythmicity}

185: 

186: The system described by equations (\ref{model}) is birhythmic,

187: i.e. it can have two different oscillatory uniform states. 

188: Assuming that

189: $\eta (t)=\rho e^{-\im\gamma t}$ and $z(t)=re^{-\im\gamma t}$ and

190: substituting this into equations (\ref{model:one}) and 

191: (\ref{model:two}), we obtain a

192: cubic equation for the oscillation frequency $\gamma $, 

193: \begin{equation}

194: \tau ^{2}\gamma ^{3}+\tau ^{2}(\omega -\alpha +\alpha K)

195: \gamma ^{2}+(1+\tau

196: K)\gamma +\omega -\alpha =0.  \label{3rd_order}

197: \end{equation}

198: 

199: When the frequency $\gamma$ is known, the oscillation amplitude 

200: $\rho$ of the field $\eta$ is given by 

201: \begin{equation}

202: \rho ^{2}=1-\frac{K\tau ^{2}\gamma ^{2}}{1+\tau ^{2}\gamma ^{2}}.

203: \label{ampl}

204: \end{equation}

205: The respective oscillation amplitude of the field $z$ is 

206: \begin{equation}

207: r=\frac{\rho}{\sqrt{1+\tau ^{2}\gamma ^{2}}}.

208: \end{equation}

209: Depending on its parameters, equation (\ref{3rd_order}) can have 

210: either one or three real roots. The three roots correspond to three 

211: possible modes of uniform oscillations with frequencies 

212: $\gamma_{1,2,3}$, such that $\gamma_{1}<\gamma_{2}<\gamma_{3}$. 

213: It can be checked that oscillations with the

214: middle frequency $\gamma_{2}$ are always unstable. 

215: In contrast to this,

216: uniform oscillations with frequencies $\gamma_{1}$ and $\gamma_{3}$ 

217: are possible (but may still be unstable with respect to nonuniform

218: perturbations, see the discussion below). 

219: 

220: Figure \ref{ak_bif} shows the region in the parameter planes 

221: ($\alpha-\omega,K$)  and  ($\tau,K$) where birhythmicity exists. 

222: The birhythmicity is possible only for sufficiently strong coupling

223: $K$. Moreover, it develops only if the 

224: characteristic time $\tau$ of the field $z$ is sufficiently large

225: (Fig. \ref{tk_bif}).

226: 

227: \begin{figure}

228: \begin{center}

229: \subfigure[]{\scalebox{0.4}[0.4]

230: {\label{ak_bif}\includegraphics*{delta_border_alpha.eps}}}

231: \subfigure[]{\scalebox{0.4}[0.4]

232: {\label{tk_bif}\includegraphics*{delta_border_tau2.eps}}}

233: \caption{Birhythimicity regions (gray) for the model (\ref{model}) 

234: in the parameter

235: planes $(\alpha-\omega ,K)$ and $(\tau ,K).$ The fixed parameters are 

236: $\tau=10$ in the first diagram and $\omega =2,\alpha =2.5$ in the 

237: second diagram.

238: The boundaries of the displayed regions do not depend on the parameter 

239: $\beta$ of the model.}

240: \label{bif}

241: \end{center}

242: \end{figure}

243: 

244: Numerical solutions of equation (\ref{3rd_order}) are displayed in Fig.

245: \ref{bif_diag}. If the parameter $\alpha$ is kept constant and the

246: coupling strength $K$ is varied (Fig. \ref{bif_k}), birhythmicity 

247: is found inside the interval $0.4<K<1$. 

248: When $K\ge 1,$ the system has a stable stationary state coexisting

249: with oscillations. The dependence of the effective oscillation 

250: frequency $\gamma$ on the difference $\alpha-\omega$ for $K=0.5$ is 

251: shown in Fig.\ref{bif_a}. 

252: Two stable limit cycles coexist within the interval $-0.1<\alpha-\omega

253: <0.8$.

254: \begin{figure}

255: \begin{center}

256: \subfigure[]{\scalebox{.45}[.45]{\label{bif_k}

257: \includegraphics*{bifurcation_k.eps}}}

258: \subfigure[]{\scalebox{.45}[.45]{\label{bif_a}

259: \includegraphics*{bifurcation_a.eps}}}

260: \subfigure[]{\scalebox{.45}[.45]{\label{bif_t}

261: \includegraphics*{bifurcation_tau.eps}}}

262: \caption{Bifurcation diagrams. Frequencies $\gamma$ of uniform 

263: oscillations, together with the respective amplitudes $\rho =|\eta|$ 

264: and $r=|z|$ as functions of $K$, $\alpha -\omega $ and $\tau $ for

265: slow (bold line), rapid (dashed) and absolutely unstable (dotted) 

266: uniform oscillation modes. The parameters are (a) $\omega =2,

267: \alpha =2.5,\tau =10$ (b) $\omega =2,K=0.5,\tau =10,$ (c) 

268: $\omega =2,\alpha =2.5,K=0.5.$}

269: \label{bif_diag}

270: \end{center}

271: \end{figure}

272: 

273: The difference between the two limit cycles becomes clear if we compare

274: the amplitudes of oscillations of the fields $\eta$ and $z$ which are 

275: presented below in the same figures. As we have already observed, 

276: at $K=0$ only one oscillating mode is present, as we are in this limit 

277: reduced to the standard CGLE. 

278: Thus, we can expect that the branch starting at $K=0$ (which is the

279: upper one in the bifurcation diagrams of Fig. \ref{bif_diag})

280: would have the strongest similarities to the CGLE, even though for

281: finite $K$ it would be affected by the coupling to the second field. 

282: This branch is characterized by a large amplitude of the field $\eta$, 

283: which never gets significantly smaller than 1. 

284: The corresponding frequency is quite small, starting from $\gamma =0.5$

285: at $K=0$ and slowly decreasing as $K$ increases.

286: Below, we refer to this attractor as corresponding to the {\em slow\/} 

287: limit cycle.

288: 

289: In contrast to this, oscillations in the lower branch have a smaller

290: amplitude $\rho$, which decreases for higher coupling strengths $K$ and

291: eventually vanishes at $K=1$. Note that $\gamma$ is negative for this 

292: mode and therefore such oscillations are counter-rotating with respect 

293: to the upper branch. The frequency varies significantly for this mode. 

294: Inside the birhythmicity interval, the frequency of such {\em rapid\/} 

295: limit cycle is always much larger (in its magnitude) than the frequency

296: of the slow mode. A characteristic feature of the rapid limit cycle is 

297: that the amplitude $r$ of the field $z$ is very small here. 

298: This is because the oscillations of the

299: field $\eta$ are so rapid here that the inertial field $z$ cannot 

300: respond to them. Hence, the coupling term $Kz$ in the system 

301: (\ref{model}) is almost negligible for the rapid limit cycle so that 

302: the nonlocal coupling is ineffective and oscillators interact only via 

303: the local diffusional coupling in this mode.

304: 

305: \section{The Phase Dynamics Approximation}

306: 

307: Above, uniform oscillations in the model have been discussed. 

308: Now, we want to consider the properties of {\em almost\/} uniform 

309: oscillations, characterized by small phase gradients. 

310: Because the system is invariant with

311: respect to the uniform shift of all phases, the evolution of phase

312: distributions with small phase gradients should be slow, in contrast 

313: to fast local relaxation of the amplitude perturbations. 

314: Therefore, a reduced description, known as the phase dynamics 

315: approximation, can be constructed in such cases \cite{kurbook}. 

316: When birhythmicity is present, the coefficients

317: of the phase dynamics equation are different for the two oscillation 

318: modes.

319: 

320: The derivation of the phase dynamics equation for this system is 

321: lengthy and has been performed by using a computer program for 

322: algebraic calculations ({\em Mathematica\/}). 

323: Below in this section we show the scheme of derivation

324: and discuss the results. The complete derivation is presented in 

325: Appendix A.

326: 

327: We define the phases ($\phi$ and $\varphi$) and amplitudes 

328: ($\rho$ and $r$) of complex fields $\eta$ and $z$ as 

329: $\eta =\rho e^{\im\phi }$, $z=re^{\im\varphi}$. 

330: It is convenient to use instead of $\phi $ and $\varphi $

331: the variables $\psi =\phi-\varphi $ and $\Theta =\phi+\varphi$. 

332: Note that the phase sum $\Theta$ is the slow variable in this system. 

333: The phases of the two fields are rigidly correlated in the uniform 

334: state and their difference $\psi $ undergoes rapid relaxation, 

335: similar to oscillation amplitudes $\rho$ and $z$.

336: 

337: In terms of the new variables, the model (\ref{model}) takes the form 

338: \begin{subequations}

339: \begin{eqnarray}

340: \dot\rho&=&\rho-\rho^3+K r \cos\psi-K\rho+\rho\left

341: (-\displaystyle\frac{\nabla\Theta^2}{4}+\displaystyle\frac{1}{2}

342: \beta\nabla^2\Theta\right)\\

343: \dot r&=&\displaystyle\frac{\rho}{\tau}\cos{\psi}-

344: \displaystyle\frac{r}{\tau}-\displaystyle\frac{l^2}{4\tau}r

345: \nabla\Theta^2\\

346: \dot\psi&=&-\omega+\alpha\rho^2-\left(\displaystyle\frac{k r}{\rho}

347: +\displaystyle\frac{\rho}{\tau r}\right)\sin\psi

348: +\displaystyle\frac{\beta}{4}\nabla\Theta^2

349: +\displaystyle\frac{1}{2}\left(1-

350: \displaystyle\frac{l^2}{\tau}\right)\nabla^2\Theta\\

351: \dot\Theta&=&-\omega+\alpha\rho^2-\left(\displaystyle\frac{k r}{\rho}

352: -\displaystyle\frac{\rho}{\tau r}\right)\sin\psi+\displaystyle

353: \frac{\beta}{4}\nabla\Theta^2+\displaystyle\frac{1}{2}\left(

354: 1+\displaystyle\frac{l^2}{\tau}\right)\nabla^2\Theta

355: \end{eqnarray}\label{four}

356: \end{subequations}

357: Suppose now that the birhythmic system is close to one of its two

358: oscillatory states, characterized by some amplitudes $\rho_0$ and 

359: $r_0$, and the relative phase $\psi_0$, whereas the phase sum 

360: $\Theta$ remains arbitrary. 

361: If the phase sum $\Theta$ is slowly varying in space,

362: the variables $\rho$, $r$ and $\psi $ would only slightly deviate from

363: their equilibrium values. 

364: Therefore, we can introduce small perturbations

365: of such variables, $\rho =\rho_0+\delta \rho$, $r=r_0+\delta r$, 

366: $\psi=\psi_0+\delta \psi $, and linearize the equations for with 

367: respect to such perturbations. 

368: The linearized evolution equation for the phase $\Theta $ is 

369: \begin{eqnarray}

370: \dot\Theta&=&-\omega+\alpha\rho_0^2-\left(\frac{Kr_0}{\rho_0}-

371: \frac{\rho_0}{\tau r_0}\right)\sin\psi_0\nonumber\\

372: &+&\delta\rho\left[2\alpha\rho_0+\left(\frac{Kr_0}{\rho_0^2}

373: +\frac{1}{\tau r_0}\right)\sin\psi_0\right]\nonumber\\

374: &+&\delta r\left(-\frac{K}{\rho_0}-\frac{\rho_0}{\tau r_0^2}\right)

375: \sin \psi_0+\delta\psi\left(-\frac{Kr_0}{\rho_0}

376: +\frac{\rho_0}{\tau r_0}\right) \cos\psi_0\nonumber\\

377: &+&\frac{\beta}{4}\nabla\Theta^2+\frac{1}{2}\left(1

378: +\frac{l^2}{\tau}\right)\nabla^2\Theta

379: \label{sum}

380: \end{eqnarray}

381: The linearized equations for the variables $\delta \rho ,\delta r,

382: \delta\psi$ can be solved in the adiabatic approximation, because 

383: these fast variables are adjusting to slow variation of $\Theta$. 

384: Substituting the result into (\ref{sum}), a closed evolution equation 

385: for the phase variable $\Theta $ is derived. 

386: This equation has the form:

387: \begin{equation}

388: \dot\Theta=C_0+C_1(\nabla\Theta)^2+C_2\nabla^2\Theta

389: \end{equation}

390: The analytical expressions for the coefficients $C_0,$ $C_1$

391: are given in Appendix A. Note that inside the

392: birhythmic region there are two uniform oscillation modes with 

393: different $\rho_0$, $r_0$, and $\psi _0$, and thus with 

394: different values of these coefficients.

395: 

396: The most important coefficient is $C_2$ because its sign controls

397: stability of uniform oscillations with respect to phase modulation 

398: (the Benjamin-Feir instability). 

399: If $C_2>0$, uniform oscillations are stable,

400: if $C_2<0$ they are unstable. We have computed this coefficient as

401: functions of several model parameters, by using the obtained analytical

402: expressions. The computed dependences are shown in Fig. \ref{coeff2}. 

403: Two different branches correspond to the two different limit cycles. 

404: The upper curves in Fig. \ref{coeff2}(a,b,d) are for the slow mode, 

405: whereas the lower curves are for the

406: rapid oscillations. In Fig. \ref{coeff2}(c), the upper curves 

407: correspond to the slow mode for larger values of the parameter $l$.

408: \begin{figure}

409: \begin{center}

410: \subfigure[]{\label{coeff2:k}\scalebox{.27}[.27]

411: {\includegraphics*{k_stability.eps}}}

412: \hspace{0.5cm}

413: \subfigure[]{\label{coeff2:a}\scalebox{.27}[.27]

414: {\includegraphics*{alpha_stability.eps}}}\\

415: \subfigure[]{\label{coeff2:l}\scalebox{.27}[.27]

416: {\includegraphics*{l_stability.eps}}}

417: \hspace{0.5cm}

418: \subfigure[]{\label{coeff2:t}\scalebox{.27}[.27]

419: {\includegraphics*{t_stability.eps}}}

420: \end{center}

421: \caption{

422: The coefficient $C_2$ of the phase dynamics equations as functions

423: of $K,\alpha ,l$ and $\tau $ for several different values of the 

424: parameter $\beta $. The upper branches in (a,b,d) and for the larger 

425: values of $l$ in (c) correspond to slow oscillations, the lower 

426: branches are for the rapid oscillations. The oscillations are unstable 

427: if $C_2<0$. The parameters are (a) $\alpha =2.5,$ $l=10$, $\tau =10,$

428: (b) $K=0.5,l=10,\tau =10,$ (c) $\alpha =2.5,$ $K=0.5,$ $\tau =10,$ 

429: (d) $\alpha =2.5,$ $K=0.5,$ $l=10;$ for all curves $\omega =2.$

430: }

431: \label{coeff2}

432: \end{figure}

433: 

434: Examining these plots, several observations can be made: Slow mode

435: oscillations are always stable inside the birhythmic region, even when 

436: the system without coupling is Benjamin-Feir unstable 

437: (i.e., $1+\alpha \beta <0$ and therefore $C_2<0$ at $K=0$). 

438: In contrast to this, rapid oscillations are always unstable when the 

439: condition $1+\alpha \beta <0$ is satisfied.

440: Moreover, they are unstable near the boundary of the birhythmic region,

441: at which they first appear, even when this condition is violated. 

442: Increasing the diffusion length $l$ of the field $z$ favors instability

443: of rapid uniform oscillations.

444: 

445: \section{Numerical Simulations}

446: 

447: To investigate nonlinear spatiotemporal dynamics in the considered 

448: model, simulations of equations (\ref{model:one}) and (\ref{model:two})

449: have been performed. For numerical integration of these equations, 

450: the fourth-order Runge-Kutta algorithm has been used. 

451: The mesh size for space discretization and the time

452: step have been chosen to optimize the computational time for each 

453: parameter choice. Since the diffusion length and the diffusion constant

454: of the oscillatory field have both been chosen to be equal to unity, 

455: $\Delta x$ varied between 0.3 and 0.5, while $\Delta t$ could vary 

456: between $(\Delta x)^{2}/5$ and $(\Delta x)^{2}/2$. 

457: Both one- and two-dimensional systems were

458: investigated. No-flux boundary conditions were employed. 

459: Initial conditions varied depending on a particular simulation.

460: 

461: To display the results of one-dimensional simulations, space-time 

462: diagrams have been constructed. 

463: In such diagrams, time always runs along the

464: horizontal axis and the vertical axis is corresponding to the spatial

465: coordinate. In two dimensions, snapshots at some time moments and 

466: space-time diagrams showing evolution of the pattern along some linear 

467: cross-section are presented. The patterns are shown by using a gray 

468: scale where the white color encodes the smallest and the black color 

469: the largest values of the displayed variable.

470: 

471: First, the behavior of fronts separating spatial regions with two 

472: different oscillatory states in the birhythmic system has been studied 

473: (Fig. \ref{front}). 

474: The front travels towards the slow oscillating region, in such a way 

475: that the system ends up with uniform oscillations of the rapid mode. 

476: Its propagation is characterized by periodic appearance of amplitude 

477: defects which occur when the phase difference between the two 

478: oscillating regions equals $2\pi $ (see Fig \ref{front:a25}). 

479: The front moves at a constant velocity which depends on the

480: parameters of the system. Fig. \ref{front_vel_a} 

481: shows the dependence of the front velocity

482: on the parameter $\alpha $ for several values of the coupling strength 

483: $K$. In Fig. \ref{front_vel_dom}, the dependence of the front 

484: propagation velocity on the frequency difference 

485: $\Delta \gamma =\gamma _{3}-\gamma _{1}$ of two uniform

486: oscillatory modes is presented. 

487: The front always moves faster when this difference is increased.

488: \begin{figure}

489: \begin{center}

490: \subfigure[Re$(\eta)$]{\label{front:re25}\includegraphics*

491: [width=3cm,height=5cm, angle=-90]{front_real_k05_al25.eps}}

492: \hspace{0.5cm}

493: \subfigure[$|\eta|$]{\label{front:a25}\includegraphics*

494: [width=3cm,height=5cm,angle=-90]{front_ampl_k05_al25.eps}}

495: \caption{

496: Spatiotemporal diagrams showing front propagation in the birhythmic

497: system with $\omega =2,\alpha =2.5,\beta =1,K=0.5,l=10,$ and 

498: $\tau =10.$ 

499: The total system length is $L=120$ and the total displayed time 

500: interval is $T=300$. 

501: Bright regions correspond to small values of displayed variables.

502: Time runs from left to right along the horizontal axis, the spatial

503: coordinate is varied in vertical direction.}

504: \label{front}

505: \end{center}

506: \end{figure}

507: 

508: \begin{figure}

509: \begin{center}

510: \subfigure[]{\label{front_vel_a}{\scalebox{0.25}[0.25]

511: {\includegraphics*{frontvelocity_alpha.eps}}}}

512: \hspace{0.5cm}

513: \subfigure[]{\label{front_vel_dom}{\scalebox{0.25}[0.25]

514: {\includegraphics*{frontvelocity_dom.eps}}}}

515: \caption{Front velocity as functions of the parameter $\alpha $ and

516: difference $\Delta \gamma $ of oscillation frequencies in the two modes

517: in the birhythmic regime. The fixed parameters are $\omega =2,\beta =1,

518: l=10,$ and $\tau =10.$}

519: \label{front_vel}

520: \end{center}

521: \end{figure}

522: 

523: As follows from the stability analysis of uniform oscillatory states 

524: (see, e.g., Fig. \ref{coeff2:a}), 

525: rapid oscillations should be unstable with respect to phase

526: modulation near the birhythmicity boundary and complex spatiotemporal

527: regimes can be expected there. We have performed a series of numerical

528: simulations, exploring pattern formation in the system within the 

529: intervals $2.55<\alpha <2.8$ and $-1<\beta <2.5$, while keeping fixed 

530: the other parameters ($\omega =2,K=0.5,l=10,$ and $\tau =10$). 

531: Each simulation started from independently chosen, completely random 

532: initial conditions.

533: 

534: Figure \ref{phasediagram} is a schematic summary of these 

535: numerical investigations. The solid

536: line is the stability boundary of rapid oscillations ($C_2=0$); rapid

537: oscillations are unstable below this boundary. Slow oscillations are 

538: always stable in the considered region. Simulations have been performed

539: for a set of parameter values indicated by symbols in this diagram.

540: \begin{figure}

541: \begin{center}

542: \scalebox{.4}[.4]{\includegraphics*{phasediagram_new.eps}}

543: \caption{

544: Schematic phase diagram. Numerical simulations of the

545: one-dimensional model were performed at the values of $\alpha $ and 

546: $\beta $ indicated by symbols in this diagram. 

547: Depending on the observed properties of patterns, the diagram is 

548: divided into regions 1 to 6. The circles indicate the values of these 

549: parameters used to produce a typical patterns

550: for the corresponding region, displayed in Fig. \ref{patterns}. 

551: The solid line shows the stability boundary

552: of rapid uniform oscillations, given by the condition $C_2=0$. Other

553: parameters are $\omega =2,K=0.5,l=10,$ and $\tau =10$.}

554: \label{phasediagram}

555: \end{center}

556: \end{figure}

557: 

558: Comparing properties of observed patterns, we can qualitatively divide 

559: them into several groups occupying different regions in the parameter 

560: plane. Each group is marked by its own symbol and the respective 

561: regions in Fig. \ref{phasediagram} are

562: numbered. Examples of typical spatiotemporal patterns, observed in

563: one-dimensional simulations within each region, are displayed in 

564: Fig. \ref{patterns}.

565: These simulations were all performed for a system of the total length 

566: $L=200$ and for a total time varying between 500 to 3000 time units. 

567: To clearly present the observed patterns, only parts of their entire 

568: evolution are however displayed.

569: \begin{figure}

570: \begin{tabular}{cccc}

571: & Re$(\eta)$ & $|\eta|$ & $|z|$\\

572: \begin{minipage}[b]{1ex}

573: \textbf{1}\\\mbox{ }\\\mbox{ }

574: \end{minipage}

575: &{\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

576: {patt1re.eps}} 

577: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

578: {patt1ae.eps}} 

579: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

580: {patt1az.eps}}\\

581: \begin{minipage}[b]{1ex}

582: \textbf{2}\\\mbox{ }\\\mbox{ }

583: \end{minipage}

584: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

585: {patt2re.eps}} 

586: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

587: {patt2ae.eps}} 

588: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

589: {patt2az.eps}}\\

590: \begin{minipage}[b]{1ex}

591: \textbf{3}\\\mbox{ }\\\mbox{ }

592: \end{minipage}

593: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

594: {patt3re.eps}} 

595: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

596: {patt3ae.eps}} 

597: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

598: {patt3az.eps}}\\

599: \begin{minipage}[b]{1ex}

600: \textbf{4}\\\mbox{ }\\\mbox{ }

601: \end{minipage}

602: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

603: {patt4re.eps}} 

604: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

605: {patt4ae.eps}} 

606: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

607: {patt4az.eps}}\\

608: \begin{minipage}[b]{1ex}

609: \textbf{5}\\\mbox{ }\\\mbox{ }

610: \end{minipage}

611: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

612: {patt5re.eps}} 

613: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

614: {patt5ae.eps}} 

615: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

616: {patt5az.eps}} \\

617: \begin{minipage}[b]{1ex}

618: \textbf{6}\\\mbox{ }\\\mbox{ }

619: \end{minipage}

620: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

621: {patt6re2.eps}} 

622: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

623: {patt6ae2.eps}} 

624: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

625: {patt6az2.eps}}\\

626: \begin{minipage}[b]{1ex}

627: \textbf{7}\\\mbox{ }\\\mbox{ }

628: \end{minipage}

629: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

630: {patt7re.eps}} 

631: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

632: {patt7ae.eps}} 

633: & {\includegraphics*[width=2.5cm,height=4cm,angle=-90,origin=rB]

634: {patt7az.eps}} \\

635: \end{tabular}

636: \caption{Spatiotemporal diagrams displaying evolution of 

637: Re($\eta$), $\rho=|\eta| $ and $r=|z| $ in typical patterns 

638: observed in the regions 1-6 for the one-dimensional system. 

639: The respective parameter values are given in Fig. \ref{phasediagram}. 

640: The displayed space and time intervals are (1) 

641: $L=100,$ $T=166,$ (2) $L=100,$ $T=166,$ (3) $L=100,$ $T=250,$ 

642: (4) $L=200,$ $T=500,$ (5) $L=100,$ $T=250,$ (6) $L=200,$ $T=500,$ 

643: and (7) $L=200,$ $T=250.$

644: The contrast level is adjusted individually in each plot to ensure best

645: visualization of the pattern.}

646: \label{patterns}

647: \end{figure}

648: 

649: In region 1, the slow limit cycle is Benjamin-Feir stable, while the 

650: rapid limit cycle is unstable. The system is found here in the regime 

651: of fully developed amplitude turbulence characterized by creation of 

652: multiple amplitude defects. Increasing parameter $\beta$, we enter a 

653: region 2 closer to the stability boundary where turbulence becomes 

654: intermittent. We observed the emergence of larger groups of 

655: synchronized oscillators which become able to reach the rapid limit 

656: cycle and perform harmonical oscillations for several periods. 

657: Inside the small parameter region 3 near the stability boundary,

658: the frequency and the amplitude of oscillations correspond almost 

659: everywhere to those of the rapid limit cycle. 

660: However, the system does not undergo

661: complete synchronization. Lines of amplitude defects travelling 

662: through the system act as wave sources here.

663: 

664: In regions 5 and 7, both lying above the stability boundary of rapid

665: oscillations, uniform oscillations are observed starting from random 

666: initial conditions. 

667: The oscillations are rapid inside region 5 and slow inside

668: region 7. These two domains are separated by regions 4 and 6 in the

669: parameter plane, where competition between two oscillation mode takes 

670: place leading to complex spatiotemporal regimes.

671: 

672: The patterns inside region 4 are characterized by a background of rapid

673: chaotic oscillations, with small amplitude and numerous amplitude 

674: defects.

675: On this highly desynchronized background, {\em bursts of 

676: synchronization\/} emerge. Such bursts consist of large groups of 

677: elements which suddenly start to oscillate together with a large 

678: amplitude and a small frequency corresponding to the stable slow 

679: limit cycle. 

680: However, these groups do not keep synchronized over a long time: after 

681: less than one oscillation period the turbulence overwhelms again. 

682: As already noticed in our discussion of

683: birhythmicity, the amplitude of the coupling field $z$ is very small 

684: in the rapid limit cycle, while in the slow limit cycle it gets a 

685: larger value comparable to the amplitude of $\eta $. Analyzing patterns

686: in region 4, it can be seen that during the synchronization bursts, 

687: the amplitude $|z|$ is indeed much larger than in the turbulent 

688: background. Note that such patterns

689: with synchronization bursts have also been observed for some parameter

690: values below the stability boundary of rapid uniform oscillations.

691: 

692: Region 6 lies near region 7, where slow uniform oscillations are 

693: winning the competition. 

694: Here, the patterns can be described as exhibiting {\em bursts

695: of desynchronization\/} on the background of slow uniform oscillations.

696: Inside such bursts, the coupling field $z$ is strongly decreased in 

697: amplitude and only short-range diffusive coupling between oscillators 

698: is effective.

699: 

700: To further illustrate this regime, phase portraits of the patterns in 

701: region 6 were constructed at different time moments. Fig. 

702: \ref{phaseportr_snap} shows several subsequent snapshots from the 

703: recorded video. 

704: Each point in a phase portrait represents a state of one of the 

705: oscillators. The spatial information (i.e., spatial distances between 

706: the displayed oscillators) is lost in this

707: representation, but the motions performed by the oscillator population 

708: are seen more clearly. The two circles indicate the two coexisting 

709: limit cycles.

710: The cycle with a smaller oscillation amplitude is rapid, whereas the 

711: cycle with a larger amplitude is slow.

712: \begin{figure}

713: \begin{center}

714: \scalebox{0.5}[0.5]{\includegraphics*{phaseportr_snap_ax2.eps}}

715: \caption{Selected snapshots of the phase portraits for the pattern 6 in

716: Fig. \ref{patterns}. 

717: The circles show two different limits cycles in the birhythmic system.}

718: \label{phaseportr_snap}

719: \end{center}

720: \end{figure}

721: 

722: We can see from these snapshots that the system is usually divided into

723: two groups of oscillators, occupying the two coexisting attractors 

724: (limit cycles). Some oscillators are found in the vicinity of the 

725: origin $\eta =0$.

726: They correspond to amplitude defects generated at the border between 

727: spatial regions with different oscillation frequencies (bright dots in 

728: the spatiotemporal plot of $|\eta|$ for pattern 6 in Fig. 

729: \ref{patterns}). Such defects are not present at all times, and the 

730: respective points are not, for example, seen in the phase portraits at 

731: $t=16.7$ and $t=281.7$. 

732: The oscillators sitting on the rapid limit cycle with the smaller 

733: amplitude belong to the desynchronization bursts.

734: 

735: For selected parameter values, two-dimensional simulations of the 

736: system have additionally been performed. 

737: Figure \ref{patt3_2d} displays the behavior of a

738: two-dimensional system at the same parameters as for the pattern 3 in 

739: Fig. \ref{patterns}.

740: The space-time diagrams in the right panels show the pattern 

741: development along a horizontal cross section. Starting from random

742: initial conditions, a population of rotating spiral waves develops. 

743: After a transient, a stable configuration of rotating spirals is 

744: reached in two dimensions.

745: 

746: Figure \ref{patt4_2d} shows two-dimensional patterns corresponding 

747: to synchronization bursts (pattern 4 in Fig. \ref{patterns}). 

748: Inside such a burst, occupying for example the

749: left central region in the left panels, the coupling field $z$ is 

750: increased in magnitude. The space-time diagrams (right panels) reveal 

751: that the synchronized regions with slow oscillations have only 

752: relatively short lifetimes and are replaced by irregular rapid 

753: oscillations. However, they appear again and again in the course of 

754: time.

755: \begin{figure}

756: \begin{center}

757: \begin{tabular}{ccc}

758: \begin{minipage}[b]{2em}

759: Re$(\eta)$\\\mbox{ }\\\mbox{ }

760: \end{minipage}

761: &{\includegraphics[width=2.5cm]{patt3re2d.eps}}

762: & {\includegraphics[width=2.5cm,height=6cm,angle=-90,origin=rB]

763: {patt3_cut_re.eps}}\\

764: \begin{minipage}[b]{2em}

765: $|\eta|$\\\mbox{ }\\\mbox{ }

766: \end{minipage}

767: &{\includegraphics[width=2.5cm]{patt3ae2d.eps}}

768: & {\includegraphics[width=2.5cm,height=6cm,angle=-90,origin=rB]

769: {patt3_cut_ae.eps}}\\

770: \begin{minipage}[b]{2em}

771: $|z|$ \\\mbox{ }\\\mbox{ }

772: \end{minipage}

773: &{\includegraphics[width=2.5cm]{patt3az2d.eps}}

774: & {\includegraphics[width=2.5cm,height=6cm,angle=-90,origin=rB]

775: {patt3_cut_az.eps}}\\

776: \end{tabular}

777: \caption{

778:  Multiple spiral waves in a two-dimensional system of size $120\times

779: 120;$ the same parameter values as for the pattern 3 in

780:  Fig. \ref{patterns}. The right

781: panels are space-time plots showing evolution of the pattern along one

782: horizontal cross section; the displayed time interval is $T=480$}

783: \label{patt3_2d}

784: \end{center}

785: \end{figure}

786: \begin{figure}

787: \begin{center}

788: \begin{tabular}{ccc}

789: \begin{minipage}[b]{2em}

790: Re$(\eta)$\\\mbox{ }\\\mbox{ }

791: \end{minipage}

792: &{\includegraphics*[width=2.5cm]{patt4re2d.eps}}

793: & {\includegraphics*[width=2.5cm,height=6cm,angle=-90,origin=rB]

794: {patt4_cut_re.eps}}\\

795: \begin{minipage}[b]{2em}

796: $|\eta|$\\\mbox{ }\\\mbox{ }

797: \end{minipage}

798: &{\includegraphics*[width=2.5cm]{patt4ae2d.eps}}

799: & {\includegraphics*[width=2.5cm,height=6cm,angle=-90,origin=rB]

800: {patt4_cut_ae.eps}}\\

801: \begin{minipage}[b]{2em}

802: $|z|$ \\\mbox{ }\\\mbox{ }

803: \end{minipage}

804: &{\includegraphics*[width=2.5cm]{patt4az2d.eps}}

805: & {\includegraphics*[width=2.5cm,height=6cm,angle=-90,origin=rB]

806: {patt4_cut_az.eps}}\\

807: \end{tabular}

808:  \caption{Bursts of synchronization in a two-dimensional system of size

809: $120\times 120;$ the same parameter values as for the pattern 4 in 

810: Fig. \ref{patterns}.

811: The right panels are space-time plots showing evolution of the pattern 

812: along one horizontal cross-section; $T=480$.}

813: \label{patt4_2d}

814: \end{center}

815: \end{figure}

816: 

817: A behavior, which can be described as bursts of desynchronization, was 

818: found in our two-dimensional simulations even outside of the 

819: birhythmicity region, where only slow uniform oscillations are possible

820: (Fig. \ref{des_bursts_first} and \ref{des_bursts_second}). 

821: We started here

822: with the initial condition, which is commonly used to generate rotating

823: spirals. A rotating spiral was indeed first formed. However, some

824: instability has then developed beginning from the its central region, 

825: where the oscillation amplitude $z$ was decreased (see Fig. 

826: \ref{des_bursts_first}).

827: The development of this instability has resulted 

828: (Fig. \ref{des_bursts_second}) in complete destruction of the spiral 

829: and the appearance of relatively small chaotic domains on the 

830: background of almost uniform slow oscillations. 

831: \begin{figure}

832: \begin{center}

833: \begin{tabular}{cccc}

834: \includegraphics*[width=2.5cm]{des_bursts1.eps}

835: &\includegraphics*[width=2.5cm]{des_bursts2.eps}

836: &\includegraphics*[width=2.5cm]{des_bursts3.eps}

837: &\includegraphics*[width=2.5cm]{des_bursts4.eps}\\

838: \includegraphics*[width=2.5cm]{des_bursts5.eps}

839: &\includegraphics*[width=2.5cm]{des_bursts6.eps}

840: &\includegraphics*[width=2.5cm]{des_bursts7.eps}

841: &\includegraphics*[width=2.5cm]{des_bursts8.eps}

842: \end{tabular}

843: \end{center}

844: \caption{Instability of a rotating spiral. Simulation for a 

845: two-dimensional system of size $120\times 120$ with parameters 

846: $\omega =2,\alpha =3,\beta =1,K=0.4,l=10,$ and $\tau =10$. 

847: Subsequent snapshots of the field Re($\eta$) at times 

848: $T$ = 2.8, 10.0, 15.2, 37.2, 96.8, 119.6, 130.0, and 139.2 are 

849: presented.}

850: \label{des_bursts_first}

851: \end{figure}

852: 

853: \begin{figure}

854: \begin{center}

855: \begin{tabular}{ccc}

856: \begin{minipage}[b]{2em}

857: Re$(\eta)$\\\mbox{ }\\\mbox{ }

858: \end{minipage}

859: &{\includegraphics*[width=2.5cm]{des_bursts_r_e.eps}}

860: & {\includegraphics*[width=2.5cm,height=6cm,angle=-90,origin=rB]

861: {des_bursts_cut_re.eps}}\\

862: \begin{minipage}[b]{2em}

863: $|\eta|$\\\mbox{ }\\\mbox{ }

864: \end{minipage}

865: &{\includegraphics*[width=2.5cm]{des_bursts_a_e.eps}}

866: & {\includegraphics*[width=2.5cm,height=6cm,angle=-90,origin=rB]

867: {des_bursts_cut_ae.eps}}\\

868: \begin{minipage}[b]{2em}

869: $|z|$ \\\mbox{ }\\\mbox{ }

870: \end{minipage}

871: &{\includegraphics*[width=2.5cm]{des_bursts_a_z.eps}}

872: & {\includegraphics*[width=2.5cm,height=6cm,angle=-90,origin=rB]

873: {des_bursts_cut_az.eps}}\\

874: \end{tabular}

875:  \caption{

876: Bursts of desynchronization in a two-dimensional system of size 

877: $120\times 120;$ continuation of the simulation presented in Fig. 

878: \ref{des_bursts_first}. 

879: The parameters are $\omega =2,\alpha =3,\beta =1,K=0.4,l=10,$ and 

880: $\tau =10$; $T=720$}

881: \label{des_bursts_second}

882: \end{center}

883: \end{figure}

884: 

885: Typical snapshots of spatial distributions of the variables Re$(\eta)$,

886: $|\eta|$ and $|z|$ in this pattern are displayed in the left panels 

887: in Fig. \ref{des_bursts_second}. 

888: Inside the domains, the coupling field $z$ is

889: reduced in magnitude and these small spatial regions are filled with

890: irregular rapid variations of the complex field $\eta$. Thus, they can

891: be classified as desynchronization bursts. The space-time diagrams 

892: through one horizontal cross section of the medium (right panels in 

893: Fig. \ref{des_bursts_second}) show

894: that such domains can spread or shrink, and travel through the medium. 

895: The edges of these structures are marked by the appearance of amplitude

896: defects where $|\eta|$ is close to zero.

897: 

898: Remarkably, if we keep fixed all other parameter values, but eliminate

899: diffusional coupling between the oscillators (the Laplacian term in 

900: equation (\ref{model:one})), evolution starting from the same initial 

901: conditions leads to

902: spiral waves with phase-randomized cores, similar to those which were

903: previously considered \cite{shima,chapter9}. Thus, the inclusion of

904: diffusive coupling has a strong effect on pattern formation in the 

905: system.

906: Though patterns resembling spiral waves with phase-randomized core are

907: indeed initially developing in the diffusively coupled case, they are

908: unstable and, after a transient, lead to the development of 

909: intermittent spatiotemporal regimes with desynchronization bursts.

910: 

911: \section{Discussion}

912: 

913: Our study can be viewed as an extension of a series of detailed

914: investigations of systems with nonlocal coupling between oscillators

915: performed by Y. Kuramoto with his coworkers. We have chosen the model 

916: (\ref{model}), which belongs to the previously discussed class, and

917: focused our analysis on the effects of coupling inertiality and 

918: diffusion in this model. 

919: The principal effect of the inertiality in nonlocal coupling

920: is that the system becomes birhythmic and has uniform slow and rapid

921: oscillations as two different coexisting attractors.

922: 

923: For rapid oscillations, the inertial field $z$ responsible for nonlocal

924: coupling is not excited and the system is essentially reduced to an 

925: array of only diffusively coupled oscillators with amplitudes $\eta$. 

926: In contrast to this, the nonlocal coupling field $z$ is involved in 

927: slow oscillations and nonlocal coupling between oscillators is in 

928: operation under these conditions.

929: 

930: Our stability analysis based on the phase dynamics equation has shown 

931: that, though slow uniform oscillations are always stable in the 

932: considered model, rapid uniform oscillations may become modulationally 

933: unstable near the birhythmicity boundary. 

934: Traveling fronts, separating spatial regions with two different 

935: oscillation modes, have been seen in our numerical simulations.

936: Note that the fronts separating oscillating and non-oscillating regions

937: have been recently observed in the vicinity of a subcritical 

938: Hopf bifurcation \cite{coullet04}.

939: 

940: Irregular intermittent spatiotemporal regimes of two different kinds 

941: have been numerically observed. In the patterns with synchronization 

942: bursts, the background is occupied by rapid chaotic oscillations. 

943: On this background, short-living small domains with slow and rather 

944: uniform oscillations spontaneously develop. 

945: Inside such domains, the nonlocal coupling field is

946: much larger in magnitude than for the rapidly oscillating background. 

947: The characteristic size of such domains was comparable to the 

948: nonlocality radius in the model. This kind of intermittent turbulence 

949: is to some extent similar to the behavior observed for a system near 

950: the Andronov homoclinic bifurcation, exhibiting bistability  between 

951: a stationary and an oscillatory state \cite{coullet97,coullet98}. 

952: However, the developing domains are filled here not by a stationary

953: state, but by slow oscillations.

954: 

955: In the spatiotemporal patterns with desynchronization bursts, the 

956: background is occupied by slow oscillations which are almost uniform. 

957: On this background, a cascade of desynchronization bursts develops. 

958: Each burst represents a localized turbulent spot filled with rapid 

959: irregular oscillations. 

960: Inside it, the nonlocal coupling field is reduced in

961: magnitude. Such patterns correspond to the regimes of intermittent 

962: localized turbulence previously seen for the complex Ginzburg-Landau 

963: equation with global coupling \cite{batt97}, in the experiments with 

964: global delayed feedback in

965: surface chemical reactions \cite{science} and in the respective

966: realistic simulations \cite{bertram1}. In some cases, they

967: develop in the situations when rotating spiral waves with 

968: phase-randomized cores have been seen \cite{shima,chapter9} in systems 

969: where only nonlocal coupling was present.

970: Birhythmicity as a cause of weak turbulence has recently been

971: investigated in a biological model of glycolitic oscillations

972: \cite{batt04}.

973: 

974: Stimulating discussions with Prof. Y. Kuramoto are gratefully 

975: acknowledged.

976: We thank M. Stich for the discussion of questions related to

977: birhythmicity.

978: 

979: \section*{Appendix A}

980: Starting from system (\ref{four}) we linearize the first three 

981: equations

982: around the values $\rho_0, r_0, \psi_0$ of the uniform oscillations. 

983: The system that we obtain can be written as

984: \begin{subequations}

985: \begin{eqnarray}

986: \dot{\delta\rho}&=&a_1\delta\rho+b_1\delta r+c_1\delta\psi

987: +d_1\nabla\Theta^2+e_1\nabla^2\Theta\\

988: \dot{\delta r}&=&a_2\delta\rho+b_2\delta r+c_2\delta\psi

989: +d_2\nabla\Theta^2\\

990: \dot{\delta\psi}&=&a_3\delta\rho+b_3\delta r+c_3\delta\psi

991: +d_3\nabla\Theta^2+e_3\nabla^2\Theta\\

992: \dot\Theta&=&a_{4}\delta\rho+b_{4}\delta r

993: +c_{4}\delta\psi+d_{4}\nabla\Theta^2+e_{4}\nabla^2\Theta+f_4

994: \label{linear}

995: \end{eqnarray}

996: \end{subequations}

997: where the coefficients are given in the following table

998: \begin{center}

999: \begin{tabular}{|l||l|}

1000: \hline

1001: $a_1 = 1-3 \rho_0^2-K$ 

1002: & $a_3 = 2\alpha\rho_0+\left[\frac{K r_0}{\rho_0^2}-r_0\frac{1}{\tau}

1003: \right]\so$\\

1004: \hline

1005: $b_1 = K\co$

1006: & $b_3 = \so\left[-\frac{K}{\rho_0}+\frac{\rho_0}{\tau r_0^2}\right]$\\

1007: \hline

1008: $c_1 = -K r_0\so$

1009: & $c_3 = \co\left[-K \frac{r_0}{\rho_0}+\frac{\rho_0}{\tau r_0}

1010: \right]$\\

1011: \hline

1012: $d_1 = -\frac{\rho_0}{4}$

1013: & $d_3 = -l^2\frac{r_0}{4\tau}$\\

1014: \hline

1015: $e_1 = \frac{\zeta}{2}\rho_0\beta$

1016: & $e_3 = \frac{1}{2}\left(1-\frac{l^2}{\tau}\right)$\\

1017: \hline\hline

1018: $a_2 = \frac{1}{\tau}\co$

1019: & $a_4 = 2\alpha\rho_0+\left[K\frac{r_0}{\rho_0^2}+\frac{1}

1020: {\tau\rho_0}\right]\so$\\

1021: \hline

1022: $b_2 = -\frac{1}{\tau}$

1023: & $b_4 = \so\left[-\frac{K}{\rho_0}-\frac{\rho_0}{\tau r_0^2}\right]$\\

1024: \hline

1025: $c_2 = -\frac{\rho_0}{\tau}\so$

1026: & $c_4 = \co\left[-K\frac{r_0}{\rho_0}+\frac{\rho_0}{\tau r_0}

1027: \right]$\\

1028: \hline

1029: $d_2 = -l^2\frac{r_0}{4\tau}$

1030: & $d_{4} = \frac{\beta}{4}$\\

1031: \hline

1032: $e_2 = 0$

1033: & $e_4 = \frac{1}{2}\left(1+\frac{l^2}{\tau}\right)$\\

1034: \hline\hline

1035: \multicolumn{2}{|c|}{$f_4 = -\omega+\alpha\rho_0^2+

1036: \left[-K\frac{r_0}{R_0}+\frac{\rho_0}{\tau r_0}\right]\so$} \\

1037: \hline

1038: \end{tabular}

1039: \end{center}

1040: 

1041: Now, since we are in the approximation where $\rho, r, \psi$ adjust 

1042: adiabatically to $\Theta$, we can assume 

1043: $\dot{\delta\rho}=\dot{\delta r}=\dot{\delta\psi}=0$, so that we get 

1044: from (\ref{linear})

1045: \begin{eqnarray}

1046: \delta \rho &=& \frac{b_3 c_2 d_1-b_2 c_3 d_1-b_3 c_1 d_2+b_1 c_3 d_2

1047: +b_2 c_1 d_3-b_1 c_2 d_3}

1048: {-a_3 b_2 c_1 + a_2 b_3 c_1 + a_3 b_1 c_2-a_1 b_3 c_2 + a_2 b_1 c_3 

1049: - a_1 b_2 c_3}\nabla\Theta^2\nonumber\\

1050: & &+\frac{b_3 c_2 e_1-b_2c_3e_1+b_2c_1e_3-b_1c_2e_3}

1051: {-a_3 b_2 c_1 + a_2 b_3 c_1 + a_3 b_1 c_2-a_1 b_3 c_2 + a_2 b_1 c_3 

1052: - a_1 b_2 c_3}\nabla^2\Theta\\

1053: \delta r &=& \frac{a_3c_2d_1-a_2c_3d_1-a_3c_1d_2+a_1c_3d_2+a_2c_1d_3

1054: -a_1c_2d_3}

1055: {a_3b_2c_1-a_2b_3c_1-a_3b_1c_2+a_1b_3c_2+a_2b_1c_3-a_1b_2c_3}

1056: \nabla\Theta^2\nonumber\\

1057: & &+\frac{a_3c_2e_1-a_2c_3e_1+a_2c_1e_3-a_1c_2e_3}

1058: {a_3b_2c_1-a_2b_3c_1-a_3b_1c_2+a_1b_3c_2+a_2b_1c_3-a_1b_2c_3}

1059: \nabla^2\Theta\\

1060: \delta\psi &=& \frac{a_3 b_2 d_1-a_2b_3d_1-a_3b_1d_2+a_1b_3d_2

1061: +a_2b_1d_3-a_1b_2d_3}

1062: {-a_3 b_2 c_1+a_2b_3c_1+a_3b_1c_2-a_1 b_3 c_2 -a_2 b_1 c_3 

1063: + a_1 b_2 c_3}\nabla\Theta^2\nonumber\\

1064: & &+\frac{a_3 b_2 e_1-a_2 b_3 e_1+a_2 b_1 e_3-a_1 b_2 e_3}

1065: {-a_3 b_2 c_1+a_2b_3c_1+a_3b_1c_2-a_1 b_3 c_2 -a_2 b_1 c_3 

1066: + a_1 b_2 c_3}\nabla^2\Theta.

1067: \end{eqnarray}

1068: These expressions can be put into the equation for $\Theta$ to get

1069: \begin{eqnarray}

1070: \dot\Theta=C_0+C_1(\nabla\Theta)^2+C_2\nabla^2\Theta

1071: \end{eqnarray}

1072: where

1073: \begin{eqnarray}

1074: C_0&=&f_4\\

1075: C_1&=&d_{4}+\left[c_4(a_3b_2d_1-a_2b_3d_1-a_3b_1d_2+a_1b_3d_2

1076: +a_2b_1d_3-a_1b_2d_3)\right.\nonumber\\

1077: & &-b_{4}(a_3c_2d_1-a_2c_3d_1-a_3c_1d_2+a_1c_3d_2+a_2c_1d_3

1078: -a_1c_2d_3)\nonumber\\

1079: & &\left.+a_{4}(b_3c_2d_1-b_2c_3d_1-b_3c_1d_2+b_1c_3d_2+b_2c_1d_3

1080: -b_1c_2d_3)\right]/\nonumber\\

1081: & &(-a_3b_2c_1+a_2b_3c_1+a_3b_1c_2-a_1b_3c_2-a_2b_1c_3+a_1b_2c_3)\\

1082: C_2&=&e_{4}+\left[c_{4}(a_3b_2e_1-a_2b_3e_1+a_2b_1e_3-a_1b_2e_3)

1083: \right.\nonumber\\

1084: & &-b_{4}(a_3c_2e_1-a_2c_3e_1+a_2c_1e_3-a_1c_2e_3)\nonumber\\

1085: & &\left.+a_{4}(b_3c_2e_1-b_2c_3e_1+b_2c_1e_3-b_1c_2e_3)\right]/

1086: \nonumber\\

1087: & &(-a_3b_2c_1+a_2b_3c_1+a_3b_1c_2-a_1b_3c_2-a_2b_1c_3+a_1b_2c_3)

1088: \end{eqnarray}

1089: 

1090: % REFERENCES

1091: \begin{thebibliography}{9}

1092: 

1093: \bibitem{kurbook}

1094: Y. Kuramoto,

1095: {\em Chemical Oscillations, Waves and Turbulence\/}, 

1096: (Springer, New York, 1984).

1097: 

1098: \bibitem{winfree}

1099: A. T. Winfree,

1100: {\em J. Theor. Biol. \/} \textbf{16} (1967) 15.

1101: 

1102: \bibitem{ertl}

1103: R. Imbihl and G. Ertl,

1104: {\em Chem. Rev.\/} \textbf{95} (1995) 697.

1105: 

1106: \bibitem{global}

1107: G. Veser, F. Mertens, A. S. Mikhailov, and R. Imbihl, 

1108: {\em Phys. Rev. Lett.\/} \textbf{71} (1993) 935-939. 

1109: 

1110: \bibitem{enzymes}

1111: A. S. Mikhailov and B. Hess, 

1112: {\em J. Phys. Chem.\/} \textbf{100} (1996) 19059--19065.

1113: 

1114: \bibitem{kur95}

1115: Y. Kuramoto,

1116: {\em Prog. of Theor. Phys.\/} \textbf{94} (1995) 321--330.

1117: 

1118: \bibitem{kur96}

1119: Y. Kuramoto and H. Nakao,

1120: {\em Phys. Rev. Lett.\/} \textbf{76} (1996) 4352--4355.

1121: 

1122: \bibitem{kur97}

1123: Y. Kuramoto and H. Nakao,

1124: {\em Physica D\/} \textbf{103} (1997) 294--313.

1125: 

1126: \bibitem{kur98}

1127: Y. Kuramoto, D. Battogtokh and H. Nakao,

1128: {\em Phys. Rev. Lett.\/}\textbf{81} (1998) 3543--3546

1129: 

1130: \bibitem{batt99}

1131: D. Battogtokh,

1132: {\em Prog. of Theor. Phys.\/} \textbf{102} (1999) 947--952.

1133: 

1134: \bibitem{kur00}

1135: Y. Kuramoto, H. Nakao and D. Battogtokh,

1136: {\em Physica A\/} \textbf{288} (2000) 244--264.

1137: 

1138: \bibitem{batt00}

1139: D. Battogtokh and Y. Kuramoto,

1140: {\em Phys. Rev. E\/} \textbf{61} (2000) 3227--3230.

1141: 

1142: \bibitem{batt02}

1143: D. Battogtokh,

1144: {\em Phys. Lett. A\/} \textbf{299} (2002) 558--564.

1145: 

1146: \bibitem{chapter9}

1147: Y. Kuramoto,

1148: in {\em Nonlinear Dynamics and Chaos: Where do we go from here?\/}, 

1149: edited by S.J. Hogan, A.R. Champneys, B. Krauskopf, M.d. Bernardo, 

1150: R.E. Wilson, H.M. Osinga, and M.E. Homer (Institute of Physics, 

1151: Bristol, 2002) Chapter: Reduction methods applied to nonlocally 

1152: coupled oscillator systems 800--819.

1153: 

1154: \bibitem{shima}

1155: S. Shima and Y. Kuramoto,

1156: {\em Phys. Rev. E} \textbf{69} (2004) 036213.

1157: 

1158: \bibitem{tanaka}

1159: D. Tanaka and Y. Kuramoto,

1160: {\em Phys. Rev. E\/} \textbf{68} (2003) 026219.

1161: 

1162: \bibitem{batt96}

1163: D. Battogtokh and A.S. Mikhailov,

1164: {\em Physica D\/} \textbf{90} (1996) 84.

1165: 

1166: \bibitem{batt97}

1167: D. Battogtokh, A.S. Mikhailov, and A. Preusser

1168: {\em Physica D\/} \textbf{106} (1997) 327.

1169: 

1170: \bibitem{kawamura}

1171: Y. Kawamura and Y. Kuramoto, 

1172: {\em Phys. Rev. E\/} \textbf{69} (2004) 016202. 

1173: 

1174: \bibitem{michael01}

1175: M. Stich, M. Ipsen and A. Mikhailov,

1176: {\em Phys. Rev. Lett.} \textbf{86} (2001) 4406.

1177: 

1178: \bibitem{michael02}

1179: M. Stich, M. Ipsen and A. Mikhailov,

1180: {\em Physica D} \textbf{171} (2002) 19.

1181: 

1182: \bibitem{coullet97}

1183: M. Argentina and P. Coullet,

1184: {\em Phys. Rev. E\/} \textbf{56} (1997) R2359-R2362.

1185: 

1186: \bibitem{coullet98}

1187: M. Argentina and P. Coullet, 

1188: {\em Physica A\/} \textbf{257} (1998) 45--60.

1189: 

1190: \bibitem{coullet04}

1191: P. Coullet and L. Kramer,

1192: {\em Chaos} \textbf{14} (2004) 244.

1193: 

1194: \bibitem{batt04}

1195: D. Battogtokh and J.J. Tyson,

1196: {\em Phys. Rev. E\/} \textbf{70} (2004) 026212.

1197: 

1198: \bibitem{science}

1199: M. Kim, M. Bertram, M. Pollmann, A. von Oertzen, A.S. Mikhailov, 

1200: H.H. Rotermund, and G. Ertl,

1201: {\em Science\/} \textbf{292} (2001) 1357.

1202: 

1203: \bibitem{bertram1}

1204: M. Bertram and A. S. Mikhailov, 

1205: {\em Phys. Rev. E\/} \textbf{67} (2003) 036207.

1206: 

1207: \bibitem{bertram2}

1208: M. Bertram, C. Beta, M. Pollmann, A. S. Mikhailov, H. H. Rotermund, 

1209: and G. Ertl, 

1210: {\em Phys. Rev. E\/} \textbf{67} (2003) 036208.

1211: 

1212: \end{thebibliography}

1213: 

1214: \end{document}

1215: