1: %\documentstyle[aps,psfig,prl,preprint]{revtex}

2: \documentstyle[aps,aps10,psfig,prl,multicol]{revtex}

3: \def\be{\begin{equation}}

4: \def\ee{\end{equation}}

5: \def\bea{\begin{eqnarray}}

6: \def\eea{\end{eqnarray}}

7: %\bibliographystyle{unsrt}

8: 

9: \begin{document}

10: \title{Experimental Evidence of Time Delay Induced Death in Coupled

11: Limit Cycle Oscillators}

12: \author{D.~V.~Ramana Reddy, A. Sen, and

13: G.~L.~Johnston\cite{GLJadd}}

14: \address{Institute for Plasma Research, Bhat, Gandhinagar 382428, India}

15: %\date{\today}

16: \maketitle

17: 

18: \begin{abstract}

19: Experimental observations of time delay induced amplitude death in

20: a pair of coupled nonlinear electronic circuits that are

21: individually capable of exhibiting limit cycle oscillations are

22: described. In particular, the existence of multiply connected

23: {\it death islands} in the parameter space of the coupling

24: strength and the time delay parameter for coupled identical

25: oscillators is established. The existence of such regions was

26: predicted earlier on theoretical grounds in [Phys. Rev.  Lett.

27: {\bf 80}, 5109 (1998); Physica {\bf 129D}, 15 (1999)]. The

28: experiments also reveal the occurrence of multiple frequency

29: states, frequency suppression of oscillations with increased time

30: delay and the onset of both

31: in-phase and anti-phase collective oscillations.\\

32: 

33: %\pacs{PACS numbers:  05.45.Xt, 87.10.+e}

34: PACS numbers:  05.45.Xt, 87.10.+e

35: \pacs{}

36: 

37: \end{abstract}

38: 

39: %\vspace{0.5cm}

40: \begin{multicols}{2}

41: Coupled limit cycle oscillator models have been extensively

42: studied in recent years because of the useful insights they

43: provide into the collective behaviour of many physical, chemical

44: and biological systems \cite{KN:87,HGEK:97,CE:89,CS:93a}. One of

45: the simplest such models, the so-called Kuramoto model

46: \cite{KN:87}, which retains only the phase information of each

47: oscillator and is valid in the limit of weak mutual couplings,

48: displays a spontaneous transition to a synchronized collective

49: state of a single frequency when the coupling strength exceeds a

50: critical value. Similar collective behaviour is observed

51: in many natural systems, such as the synchronous flashing of

52: fireflies, the phase locking of cardiac pacemaker

53: cells, and the collective chirping of crickets \cite{CS:93a}.

54: Phase locking has also been demonstrated experimentally in arrays

55: of coupled nonlinear electronic circuits \cite{BL:96}. When the

56: coupling becomes stronger, amplitude effects become important and

57: give rise to other interesting collective states, such as that of

58: {\it amplitude death} in which the various oscillators pull each

59: other off their periodic states and collapse to a state of zero

60: amplitude. The condition for such a state to occur is for the

61: oscillators to have a broad dispersion in their

62: natural frequencies and for the coupling strength to exceed a

63: threshold value. Thus, as has been pointed out in a number of

64: theoretical studies \cite{AEK:90}, a collection of identical

65: limit cycle oscillators cannot display {\it amplitude death}.

66: However, more recent investigations \cite{RSJ:98} indicate that the

67: presence of finite propagation time delays in the coupling

68: removes this restriction and predict the possibility of inducing

69: the death state even in a system of two coupled identical limit

70: cycle oscillators. Time delay is ubiquitous in most physical

71: systems due to finite propagation speeds of signals, finite

72: chemical reaction times, finite response times of synapses, etc.,

73: and its influence on the collective dynamics of coupled systems

74: can have wide-ranging implications \cite{Strog:98}. It is

75: important therefore to establish the

76: experimental feasibility of such a death phenomenon.

77: 

78: In this Letter we present experimental observations on time delay

79: induced death in two coupled nonlinear circuits that are

80: individually capable of exhibiting limit cycle oscillations. A

81: specially designed digital delay line gives precise control over

82: the delay time in the coupling and permits us to explore a large

83: area of parameter space. Observations on the phenomenon of {\it

84: amplitude death} have recently been reported \cite{HFRPO:00} for

85: a pair of optothermal oscillators that are thermally coupled and

86: for which the occurrence of death for strong couplings and its

87: relation to Hopf bifurcations of the uncoupled and coupled

88: systems have been experimentally verified. However, as the

89: authors themselves point out, their experimental results are not

90: conclusive about the role of delay in the death phenomenon and in

91: particular they have not investigated the phenomenon of death

92: islands or, indeed, that of the multiplicity of death islands

93: predicted by the theory of time delay induced death

94: \cite{RSJ:98}. In our experiments we provide clear evidence of time delay

95: induced death islands and their multiple connectedness in the

96: parameter space defined by the coupling strength, time delay and

97: frequency, for the case of two coupled identical oscillators. We

98: also find the existence of multiple frequency states, frequency

99: suppression of oscillations with increased time delay and the

100: onset of both in-phase and anti-phase collective

101: oscillations. 

102: 

103: Our experimental system, schematically shown in

104: Fig.~\ref{FIG:ckt}, consists of two nonlinear $LCR$ circuits

105: coupled through a digital delay line (DDL). The individual oscillator

106: circuits are a variant of the so-called Chua circuit, which has

107: been widely employed in a number of nonlinear dynamical studies

108: \cite{Chua:93}. The nonlinear resistive element $R_{N}$ has the

109: typical \(V-I\) characteristic, with negative resistance around

110: the origin and positive slopes away from the origin as shown in

111: Fig.~\ref{FIG:gv}. This enables each individual oscillator

112: circuit in the uncoupled state to sustain limit cycle

113: oscillations with a characteristic frequency of $1/\sqrt{LC}$.

114: The two channel DDL taps the input signal at

115: a certain rate using an analog to digital converter (in this case

116: an ADC0809 with a conversion rate of $100 \mu $s) and stores it

117: in a random access memory (RAM) bank. The stored bits are read

118: using a logical circuit and converted back to an analog signal to

119: be fed back into the circuit. The coupling between the

120: oscillators is linear, resistive and proportional to the

121: difference in the signal strengths of the two oscillators with a

122: time delay. The coupling strength is varied by changing the

123: resistances ($R_K$) that join the two oscillators. By using the

124: current summation rules at nodes $V_{1}$ and $V_{2}$, a

125: theoretical description of the circuit can be given as, 

126: %

127: \be

128: \label{EQN:model1} \ddot{V}_{i} + g(V_i)~\dot{V}_{i} +

129: \omega_{i}^2 ~V_{i} = K_{i} [\dot{V}_{j}(t-\tau) -

130: \dot{V}_{i}(t)], 

131: \ee

132: %

133: where $i,j=1,2, i \ne j$ and $g(V_i) =(1 /

134: C_{i}) ( - 0.95 + 0.66 V_i^2 + 322.0 V_i^4 ) \times 10^{-5}$

135: Amp/F, the nonlinear damping factor, is obtained from an

136: empirical fit of the \(V-I\) characteristics of the nonlinear

137: resistive element as shown in Fig.~\ref{FIG:gv}.

138: The frequencies

139: are $\omega_j = 1 /\sqrt{L_j C_j}$ and the coupling strengths

140: $K_j = 1 / (C_j R_{K_j})$. For our experiments on identical

141: oscillators we have fixed $C_{1}= C_{2} = 0.1 \mu F$ and varied

142: $L = L_{1} = L_{2}$ and $R = R_{K_1} = R_{K_2}$ such that 

143: $\omega = \omega_{1} =\omega_{2}$ spans the 

144: range of $100 s^{-1}$ to $1000 s^{-1}$. 

145: %

146: %-------------

147: % figure 1

148: %-------------

149: \begin{figure}

150: \narrowtext

151: \centerline{\psfig{file=ckt2_f.ps,width=7.5cm,height=4cm}}

152: \caption{The circuit diagram of two delay coupled limit cycle

153: oscillators consisting of a $C$, $L$ and a nonlinear resistance

154: $R_N$ connected in parallel. The OP-AMPs are buffer amplifiers

155: and ideally draw no input currents. The digital delay line

156: outputs $V_{1,2}^{\tau} = V_{1,2}(t-\tau)$.} 

157: \label{FIG:ckt}

158: \end{figure}

159: %

160: %

161: %-------------

162: % figure 2

163: %-------------

164: \begin{figure}

165: \centerline{\psfig{file=vi_non_f.ps,width=6cm,height=4.5cm}}

166: \caption{The V-I characteristics of the nonlinear component

167: $R_{N}$. The continuous line is a polynomial fit of the

168: experimental points.} \label{FIG:gv}

169: \end{figure}

170: %

171: %

172: The coupling strength varies in the range of $10 s^{-1}$ to $10^6 s^{-1}$.

173: Examples of our experimental

174: results for the time evolution of the oscillator voltages as a

175: function of the delay parameter are shown in

176: Fig.~\ref{FIG:5.v1inanti} for $K= 1000 s^{-1}$ and $\omega = 837

177: s^{-1}$. The identical oscillators acquire an in-phase locked

178: state (for $\tau = 0.514$ ms), death (for $\tau = 2$ ms) and an

179: anti-phase locked state (for $\tau = 4.428$ ms). A series of such

180: observations are recorded for a fixed frequency and at different

181: fixed values of $K$ by varying the time delay parameter. The

182: output voltage $V$ of the oscillator is monitored on the

183: oscilloscope up to the point where it shows a sudden drop in

184: value by several orders of magnitude, i.e., nearly to zero. The

185: value of $\tau$ is further increased up to the point where the

186: signal from the oscillator again suddenly revives to a large

187: value. An entire parameter space in \(K - \tau \) for amplitude

188: death is thereby explored.

189: In Fig.~\ref{FIG:isle} the solid squares are the experimental data

190: points marking the loci of the death regions. They clearly reveal

191: closed {\it death island} regions in the plane of the coupling

192: strength, $K$, and the time delay, $\tau$, for the common

193: frequency of $\omega = 837 s^{-1}$. 

194: %

195: %

196: %-------------

197: % figure 3

198: %-------------

199: \begin{figure}

200: \centerline{\psfig{file=wform.ps,width=7.5cm,height=11cm}}

201: \caption{Oscilloscope traces of the temporal

202: behavior of the voltages of the two delay coupled oscillators for

203: different delay times with $K = 1000 s^{-1}$ and  $\omega = 837

204: s^{-1}$. Time is marked in units of $5$ ms along the horizontal

205: axis. The oscillators show (top panel) in-phase locking for $\tau

206: = 0.514 $ ms, (middle panel) amplitude death for $\tau = 2.0$

207: ms, and (bottom panel) anti-phase locking for $\tau = 4.428$ ms.}

208: \label{FIG:5.v1inanti}

209: \end{figure}

210: %

211: %

212: %-------------

213: % figure 4

214: %-------------

215: \begin{figure}

216: \centerline{\psfig{file=dis.ps,width=8cm,height=6cm}}

217: \caption{Multiply connected death islands in \(K-\tau\) space for

218: $\omega = 837 s^{-1}$. Solid squares mark the experimental points

219: while the solid and dashed lines are theoretical curves obtained

220: from model equation (1).} \label{FIG:isle}

221: \end{figure}

222: %

223: %

224: The solid and dashed lines

225: are analytic curves obtained from a linear stability analysis of

226: the origin for the model equation (\ref{EQN:model1}).

227: These curves are derived in a standard manner \cite{RSJ:98}

228: from the characteristic equation for the eigenvalue $\lambda$, namely,

229: $\lambda^2 (\lambda-\frac{a_1}{C} + K)^2 + 2 \omega^2

230: (\lambda-\frac{a_1}{C} + K) \lambda + \omega^4 - K^2 \lambda^2

231: e^{-2\lambda \tau} = 0$, and are given by,

232: \bea

233: \label{EQN:DI1}

234: \tau_1(n,K)

235: &=& \frac{(n-1)\pi+\cos^{-1}(1-\frac{a_1}{C K})}

236:                     {-\frac{A}{2} + \sqrt{(\frac{A}{2})^2+\omega^2}}, \\

237: \label{EQN:DI2}

238: \tau_2(n,K) &=&

239: \frac{n\pi-\cos^{-1}(1-\frac{a_1}{C K})}

240:                     {\frac{A}{2} + \sqrt{(\frac{A}{2})^2+\omega^2}},

241: \eea

242: %%%%%%%%%%%%%%%

243: where $A = \sqrt{\frac{2 a_1}{C} K - (\frac{a_1}{C})^2}$,

244: $a_1=0.95 \times 10^{-5}$ and $n = 1, 2, \ldots$. The region

245: enclosed by the intersection of $\tau_1(1,K)$ and $\tau_2(1,K)$

246: is the primary death island and that enclosed between

247: $\tau_1(2,K)$ and $\tau_2(2,K)$ is the secondary (higher order)

248: death island. We observe that the experimental observations not

249: only display topological similarity to the analytic predictions

250: but are also in good quantitative agreement with them. The

251: deviations observed are primarily due to two reasons. First, the

252: analytic domains are sensitive to the values of the coefficients

253: used in fitting the \(V-I\) characteristic of the nonlinear

254: resistive element. The relative downward shift and larger width

255: of the experimental islands compared to the analytic region are

256: due to these estimation errors.

257: Second, the deviation at very high

258: values of $K$ (and hence low $R_{K}$) is related to limitations

259: on the maximum current that can flow through $R_{K_j}$ for the

260: rated biased  voltages of our experimental circuit. The existence

261: of the higher order island is particularly significant since it

262: confirms the theoretical finding of multiple connectivity of the

263: stability region of the origin \cite{RSJ:98}. 

264: %

265: %-------------

266: % figure 5

267: %-------------

268: \begin{figure}

269: \centerline{\psfig{file=wtau.ps,width=7cm,height=5cm}}

270: \caption{Amplitude death region in the plane of $(\omega,\tau)$

271: for a fixed coupling strength of $K = 500 s^{-1}$.}

272: \label{FIG:ftau}

273: \end{figure}

274: %

275: %

276: In Fig.~\ref{FIG:ftau}, the experimental death region is shown at a

277: constant coupling strength $K = 500 s^{-1}$ for different values

278: of the frequency. The death region ceases to exist below a

279: certain threshold frequency, in accordance with theoretical

280: predictions \cite{RSJ:98}.

281: On either side of this region, there

282: are phase locked states exhibiting in-phase locked oscillations

283: for small time delays and anti-phase locked oscillations for

284: longer time delays.

285: 

286: Another important characteristic of time delay systems in general

287: \cite{NSK:91,KPR:97}, and delay coupled limit cycle oscillators

288: in particular \cite{RSJ:98}, is the existence of multiple phase

289: locked states. When the oscillators are identical, these phase

290: locked states coincide with in-phase (phase difference, $\phi =

291: 0$) and anti-phase ($\phi = \pi$) solutions. In

292: Fig.~\ref{FIG:coexsup}(a), the domains of existence of these

293: solutions in $\tau$ space are indicated. 

294: Here $I_n$ represent the in-phase solutions and $A_n$ represent

295: the anti-phase solutions. These domains have been determined

296: experimentally. For a fixed $\tau$ the various existing multiple

297: states can be accessed with slight perturbations to the system,

298: i.e., by slightly changing the initial conditions. We have also

299: observed hysteretic behaviour between the various branches.

300: In Fig.~\ref{FIG:coexsup}(b), the frequencies of the first few

301: multiple states are plotted for increasing values of $\tau$. Note

302: that the collective states can be much 

303: higher than the individual oscillator frequencies and that 

304: %

305: %

306: %-------------

307: % figure 6

308: %-------------

309: \begin{figure}

310: \centerline{\psfig{file=coexsup.ps,width=8cm,height=4.5cm,angle=270}}

311: \caption{(a) Coexistence of in-phase and anti-phase locked

312: states, and (b) suppression of the phase locked states as $\tau$

313: is increased for $K= 1000 s^{-1}$ and $\omega = 837 s^{-1}$.}

314: \label{FIG:coexsup}

315: \end{figure}

316: %

317: %

318: \noindent

319: their magnitude decreases with

320: increasing values of time delay. Such a phenomenon of time delay

321: induced frequency suppression has been discussed earlier in the

322: context of phase only oscillators \cite{NSK:91}.\\

323: 

324: In conclusion, we have experimentally demonstrated for the first

325: time the existence of delay induced amplitude death in a pair of

326: coupled identical nonlinear oscillators \cite{RSJ:98}. We have

327: also demonstrated the coexistence of in-phase and anti-phase

328: states with multiple frequencies and the suppression of the common

329: frequency with time delay. The observation of multiply connected

330: death islands (for identical oscillators) in the parameter space

331: of the coupling strength and time delay is our most important

332: result. This result not only bears topological

333: resemblance to earlier model calculations \cite{RSJ:98}, 

334: but also shows good quantitative agreement with analytic estimates

335: obtained from a theoretical model of the experimental circuits.

336: It should be noted that our experimental system of coupled

337: nonlinear electronic circuits (and its theoretical model) is

338: capable of richer nonlinear behaviour and goes beyond the simple

339: normal form model used in previous theoretical studies

340: \cite{RSJ:98}. In this sense the results on delay induced 

341: amplitude death can be

342: said to have a broader validity and therefore wider applicability

343: to physically realistic systems.\\

344: 

345: We now briefly discuss a few interesting directions for future

346: experiments that are relevant to our present results as

347: well as to practical applications. A natural next step is to

348: couple more than two oscillators to explore some of the other

349: theoretical predictions discussed in \cite{RSJ:98}. The principal

350: technical challenge and the major expensive elements for such an

351: endeavor are the construction of additional delay lines with

352: enhanced number of channels. Such an effort is presently

353: underway.  Another interesting possibility is to move away from

354: the limit cycle regime and investigate the effect of time delay

355: on the synchronization of coupled chaotic oscillators -  an area

356: of much current scientific interest \cite{RPK:96}. In principle

357: this should be possible with our present experimental system

358: since the basic Chua circuits are capable of displaying chaotic

359: oscillations. It would be necessary however to employ a faster

360: delay line. Time delay could facilitate phase synchronization in

361: view of its significant influence on phenomena like phase slips

362: \cite{RSJ:00}. A final area of practical interest is the stability

363: of these limit cycle collective states to external noise. Past

364: numerical studies indicate that time delay induced higher

365: frequency states can exhibit metastable behaviour and decay to the

366: lowest periodic state in the presence of sufficient external noise

367: \cite{NSK:91}. We do not observe this behaviour since by design

368: our system has very low intrinsic (thermal) noise in order to

369: facilitate access to the higher frequency states. It would be

370: interesting to introduce a suitably designed noise source in the

371: individual oscillator circuits to experimentally test these

372: theoretical predictions including the influence of noise on the

373: threshold for onset of synchronization \cite{Shi:85}.

374: 

375: 

376: We acknowledge the important contribution of H.~S. Mazumdar 

377: in the design, development and construction of the digital delay line.

378: 

379: \begin{thebibliography}{10}

380: 

381: \bibitem[*]{GLJadd}

382: Present Address: EduTron Corp., 5 Cox Rd., Winchester, MA 01890.

383: 

384: \bibitem{KN:87}

385: Y.~Kuramoto and I.~Nishikawa,

386: \newblock J. Stat. Phys. {\bf 49}, 569 (1987).

387: 

388: \bibitem{HGEK:97}

389: A.~Hohl, A.~Gavrielides, T.~Erneux, and V.~Kovanis,

390: \newblock Phys. Rev. Lett. {\bf 78}, 4745 (1997);

391: J.~Benford, H.~Sze, W.~Woo, R.~R. Smith, and B.~Harteneck, Phys.

392: Rev. Lett. {\bf 62}, 969 (1989).

393: 

394: \bibitem{CE:89}

395: M.~F. Crowley and I.~R. Epstein,

396: \newblock J. Phys. Chem. {\bf 93}, 2496 (1989);

397: M.~Yoshimoto, K.~Yoshikawa, and Y.~Mori, Phys. Rev. E {\bf 47},

398: 864 (1993).

399: 

400: \bibitem{CS:93a}

401: J.~J. Collins and I.~N. Stewart,

402: \newblock J. Nonlinear Science {\bf 3}, 341 (1993);

403: M.~A. Branham and M.~D. Greenfield, Nature {\bf 381}, 745 (1996);

404: M. Kawato and R. Suzuki, J. Theor. Biol. {\bf 86}, 547 (1980); E.

405: Sismondo, Science {\bf 249}, 55 (1990)

406: 

407: \bibitem{BL:96} A.~A. Brailove and P.S. Linsay, Int. J.

408: Bifurcation Chaos {\bf 6}, 1211 (1996).

409: 

410: \bibitem{AEK:90}

411: D.~G. Aronson, G.~B. Ermentrout, and N. Kopell, Physica

412: (Amsterdam) {\bf 41D}, 403 (1990); P.~C. Matthews and S.~H.

413: Strogatz, Phys. Rev. Lett. {\bf 65}, (1990); P.~C. Matthews, R.~E.

414: Mirollo, and S.~H. Strogatz, Physica D {\bf 52}, 293 (1991).

415: 

416: \bibitem{RSJ:98}

417: D.~V.~Ramana Reddy, A.~Sen, and G.~L. Johnston,

418: \newblock Phys. Rev. Lett. {\bf 80}, 5109 (1998);

419: D.~V.~Ramana Reddy, A.~Sen, and G.~L. Johnston, Physica (Amsterdam) {\bf

420: 129D}, 15 (1999).

421: 

422: \bibitem{Strog:98}

423: S.~H. Strogatz, Nature (London) {\bf 394}, 316 (1998).

424: 

425: \bibitem{HFRPO:00}

426: R. Herrero, M. Figueras, J. Rius, F. Pi, and G. Orriols,

427: \newblock Phys. Rev. Lett. {\bf 84}, 5312 (2000).

428: 

429: \bibitem{Chua:93}

430: M. Komuro, R. Tokunaga, T. Matsumoto, L.O. Chua, and A. Hotta,

431: \newblock Int. J. Bifurcation and Chaos {\bf 1}, 139 (1991), and

432: references therein; {\it Chua's Circuit: A Paradigm for Chaos},

433: edited by R.~N. Madan (World Scientific, Singapore, 1993).

434: 

435: \bibitem{NSK:91}

436: E.~Niebur, H.~G. Schuster, and D.~Kammen,

437: \newblock Phys. Rev. Lett. {\bf 67}, 2753 (1991).

438: 

439: \bibitem{KPR:97}

440: S. Kim, S.~H. Park, and C.~S. Ryu,

441: \newblock Phys. Rev. Lett. {\bf 79}, 2911 (1997);

442: M.~K.~S. Yeung and S.~H. Strogatz, Phys. Rev. Lett. {\bf 82}, 648

443: (1999).

444: 

445: \bibitem{RPK:96} M.~G. Rosenblum, A. S. Pikovsky, and J. Kurths,

446: \newblock Phys. Rev. Lett. {\bf 76} (1996) 1804.

447: 

448: \bibitem{RSJ:00} D.~V.~Ramana Reddy, A.~Sen, and G.~L. Johnston,

449: \newblock Physica (Amsterdam) {\bf 144D} (2000) 336.

450: 

451: \bibitem{Shi:85} M. Shiino, Phys. Lett. A {\bf 111}, 396 (1985).

452: 

453: \end{thebibliography}

454: 

455: \end{multicols}

456: 

457: \end{document}

458: 

459: