1: %\documentstyle[twocolumn,prl,aps,epsfig]{revtex}
2: \documentstyle[pre,aps,epsfig]{revtex}
3: %\tightenlines
4: \begin{document}
5:
6: \draft
7:
8: % for two column activate the line below...
9: %\twocolumn[\hsize\textwidth\columnwidth\hsize\csname@twocolumnfalse\endcsname
10: \title{
11: Interruption of torus doubling bifurcation and genesis of strange nonchaotic
12: attractors in a quasiperiodically forced map : Mechanisms and their
13: characterizations}
14:
15: \author{A. Venkatesan and M. Lakshmanan}
16: \address{Centre for Nonlinear dynamics,\\ Department of Physics,\\
17: Bharathidasan University,\\Tiruchirapalli 620 024, INDIA }
18:
19: \maketitle
20:
21: \begin{abstract}
22:
23: A simple quasiperiodically forced one-dimensional cubic map is shown to
24: exhibit very many types of routes to chaos via strange nonchaotic
25: attractors (SNAs) with reference to a two-parameter $(A-f)$ space. The
26: routes include transitions to chaos via SNAs from both one frequency
27: torus and period doubled torus. In the former case, we identify the
28: fractalization and type I intermittency routes. In the latter case, we
29: point out that atleast four distinct routes through which the
30: truncation of torus doubling bifurcation and the birth of SNAs take
31: place in this model. In particular, the formation of SNAs through
32: Heagy-Hammel, fractalization and type--III intermittent mechanisms are
33: described. In addition, it has been found that in this system there
34: are some regions in the parameter space where a novel dynamics
35: involving a sudden expansion of the attractor which tames the growth of
36: period-doubling bifurcation takes place, giving birth to SNA. The SNAs
37: created through different mechanisms are characterized by the behaviour
38: of the Lyapunov exponents and their variance, by the estimation of
39: phase sensitivity exponent as well as through the distribution of
40: finite-time Lyapunov exponents.
41:
42: \end{abstract}
43: \pacs{PACS number(s): 05.45.+b}
44: %\tightenlines
45: %\narrowtext
46: %\vskip1pc]
47: \section{ INTRODUCTION}
48:
49: Torus doubling bifurcation (geometrically similar to period doubling
50: bifurcation) as a universal route to chaos is turning out to be one of
51: the leading topics of research in the study of quasiperiodically
52: forced chaotic dynamical systems during the past few years \cite
53: {R1,R2,R3,R4,R5,R6}. The existence of such an exotic bifurcation in
54: several experimental situations and theoretical models indicates the
55: importance of this bifurcation in improving our understanding of the
56: qualitative and quantitative behaviours of dynamical systems \cite
57: {R1,R2,R3,R4,R5,R6,R7,R8,R9,R9a,R10}. A very common observation is
58: that such systems do not undergo an infinite sequence of doubling
59: bifurcations as in the case of the lower dimensional systems; instead,
60: the truncation of torus doubling begins when the doubled torus becomes
61: extremely wrinkled and then gets destroyed. Such a destroyed torus is
62: a geometrically strange (fractal dimensional) object in the phase
63: space, a property that usually corresponds to that of a chaotic
64: attractor. However, it would not exhibit sensitivity to initial
65: conditions asymptotically (for example, Lyapunov exponents are
66: nonpositive) and hence is not chaotic and so it is a strange nonchaotic
67: attractor(SNA)
68: \cite{R11,R12,R12a,R12b,R12c,R12d,R13,R13a,R13b,R14,R16,R16a,R17,R18,R19,R20,R21,R22}.
69: Actually the existence of SNA was first identified by Grebogi {\it et
70: al.} \cite {R11} in their work on the transition from two-frequency
71: torus to chaos via SNA. Later on, it was found that these attractors
72: can arise in physically relevant situations such as a quasiperiodically
73: forced pendulum \cite{R12a,R16}, quantum particles in quasiperiodic
74: potentials \cite{R12a}, biological oscillators \cite{R12b},
75: Duffing-type oscillators \cite{R12c,R12d,R13}, velocity-dependent
76: potential systems \cite{R10}, electronic circuits \cite{R13a,R13b}, and
77: in certain maps \cite {R14,R16,R16a,R17,R18,R19} in different
78: transitions to SNA including the torus doubling bifurcation and the
79: birth of SNAs. Also, the existence of torus doubling truncation and
80: the appearance of SNA was confirmed by an experiment consisting of a
81: quasiperiodically forced, buckled magnetoelastic ribbon \cite{R20}.
82: Besides this experiment, the exotic strange nonchaotic attractors were
83: studied in analog simulations of a multistable potential \cite{R21},
84: and in neon glow discharge experiment \cite {R22} through different
85: transitions to SNAs. The existence of SNAs in such physically relevant
86: systems has naturally motivated further intense investigations on the
87: nature and occurrence of them.
88:
89:
90: A question of intense further interest in the topic is the way in
91: which the truncation of period doubling occurs to give birth to SNAs.
92: In particular, it has been found that the birth of SNAs often occurs
93: due to the collision of a period doubled torus with its unstable
94: parent so that a period 2$^k-$torus gives rise to a 2$^{k-1}-$band SNA
95: \cite {R8} or a gradual fractalization of torus, in which a period
96: 2$^k-$torus approaches a 2$^k-$band SNA \cite {R9}. Recently, the
97: present authors have identified that the torus doubling sequence is
98: tamed due to a subharmonic bifurcation (subcritical period-doubling
99: bifurcation) leading to the creation of SNAs. In addition, this
100: transition has been shown to exhibit type-III intermittent
101: characteristic scaling \cite {R12d,R13a}. Apart from the creation of
102: SNAs due to the collapse of the tori, the authors have also shown that
103: there are some regions of the system parameters where the torus
104: doubling sequence is truncated by a merging bifurcation leading to the
105: formation of a torus bubble \cite {R10}, reminiscent of period bubbles
106: in low dimensional systems. Also, using the renormalization group
107: approach, Kuznetsov, Feudal and Pikovsky have revealed scaling properties
108: both for the critical attractor and for the parameter plane topography
109: near the terminal point of the torus doubling bifurcation \cite{R9a} in
110: connection with the above mentioned collision scenario.
111:
112:
113: Besides the creation of SNAs through the above mentioned truncation of
114: torus doubling bifurcation, several other mechanisms have also been
115: studied in the literature for the birth of SNAs. The most common one is
116: gradual fractalization of a torus where an amplitude or phase
117: instability causes the collapse of the torus \cite{R9}. This is in fact
118: one of the least understood mechanisms for the formation of SNAs since
119: there is no apparent bifurcation unlike the torus collision mechanism
120: identified by Feudel, Kurths and Pikovsky where a stable torus and an
121: unstable torus collide at a dense set of points, leading to the
122: creation of SNAs \cite{R16a}. Prasad, Mehra and Ramaswamy \cite{R14}
123: have shown that a quasiperiodic analogue of a saddle-node bifurcation
124: gives rise to SNAs through the intermittent route with the dynamics
125: exhibiting scaling behaviour characteristic of type--I intermittency.
126: Yalcinkaya and Lai have described that an on-off intermittency can be
127: associated with SNA creation through a blow-out bifurcation when a
128: torus loses its transverse stability \cite{R13}. Other than the above
129: said scenarios, a number of other quasiperiodic routes to SNAs have
130: been described in the literature \cite{R12,R18}. They include the
131: existence of SNAs in the transition from two-frequency to
132: three-frequency quasiperiodicity \cite{R12}, transition from
133: three-frequency to chaos via SNA and transition to chaos via strange
134: nonchaotic trajectories on torus \cite{R18}.
135:
136: Considering particularly the different routes discussed above for the
137: inhibition of torus doubling sequence and the birth of SNAs, we note
138: that they have so far been identified essentially in $\it{different}$
139: dynamical systems. However, it is important to study the truncation of
140: the torus doubling bifurcations and the appearance of SNAs in a single
141: system in order to understand the mechanisms and their characteristic
142: features clearly. In this connection, we consider a simple model in the
143: form of an one-dimensional cubic map,
144: $$x_{i+1}= - A x_i+x_i^3, \eqno(1) $$
145: which is quite analogous to the typical Duffing oscillator
146: \cite{R23,R24}. The existence of different dynamical features of this
147: system has been studied in refs.\cite{R24,R25}. In the present work, we
148: investigate the dynamics of (1) with the addition of a constant bias,
149: $$x_{i+1}= Q - Ax_i+x_i^3, \eqno(2) $$
150: and also subject to an additional quasiperiodic forcing,
151: $$x_{i+1}= Q +f \cos (2 \pi \theta_i) - A x_i+x_i^3, \eqno(3) $$
152: $$\theta_{i+1}=\theta_i+ \omega (\text{mod} 1),$$
153: and show that the
154: latter is a rich dynamical system in comparison with the former,
155: possessing a vast number of
156: regular, strange nonchaotic and chaotic attractors in a two-parameter
157: $(A-f)$ space for a fixed $Q$. In particular, we focus our attention
158: mainly on the
159: truncation of torus-doubling bifurcations leading to the birth of SNAs and the
160: mechanisms by which they arise in a range of the two-parameter $(A-f)$
161: space, besides pointing out the standard transitions to chaos via SNAs
162: from one frequency torus. A variety of transitions from truncated
163: doubled torus to SNAs could be identified, characterized and
164: distinguished in this system.
165:
166: To start with, we show that the system (2) undergoes one or more period
167: doublings but it need not complete the entire Feigenbaum cascade, and
168: that it may be possible to have only a finite number of period
169: doublings, followed by, for example, undoubling or other bifurcations
170: in the presence of constant bias, as was shown by Bier and Bountis in
171: different systems \cite {R25}. The possibility of such a different
172: remerging bifurcation phenomenon in the torus doubling sequence is
173: reported in the present case, when the system (2) is subjected to
174: quasiperiodic forcing as in model (3). As the system (3) possesses
175: more than one control parameter and remains invariant under reflection
176: symmetry, the remergence is likely to occur as in the absence of
177: quasiperiodic forcing in the system (2). Our numerical study shows
178: that for fixed value of $Q$ in some regions of the $(A-f)$ parameter
179: space a torus doubled orbit emerges and remerges from a single torus
180: orbit at two different parameter values of $f$ to form a single torus
181: bubble. Such a remerging bifurcation can retard the growth of the
182: torus doubled bifurcations and the development of the associated universal
183: route to chaos further. However, the nature of remerging torus doubled
184: bifurcation or, more specifically, the torus bubbling ensures the
185: existence of different routes for the creation of SNAs when the full
186: range of parameters is explored. To illustrate such possibilities in
187: the present system in the two-parameter $(A-f)$ space, we first
188: enumerate three standard types of routes to SNA namely, (1)
189: Heagy-Hammel ($-$ collision of period doubled torus with its unstable
190: parent), (2) gradual fractalization ($-$the amplitude or phase
191: instability), and (3) type--III intermittency ($-$the subharmonic
192: instability) routes through which the truncation of torus doubling
193: bifurcation occurs leading to the birth of SNAs within the torus
194: bubble region.
195:
196: In addition, we identify that in some cross sections of the
197: $(A-f)$ parameter space, particularly within the torus bubble region,
198: the period doubling bifurcation phenomenon still persists in the
199: destroyed torus, eventhough the actual doubling of the torus
200: itself has been terminated. However, we show that {\bf the novel
201: dynamics involved in this transition is a sudden expansion in the
202: attractor}. This transition seems to look like the interior crisis which
203: occurs in low dimensional chaotic systems \cite {R26}. We also
204: demonstrate the occurrence of SNAs through gradual fractalization and
205: type--I intermittent nature, during the transition from one-frequency
206: quasiperiodicity to chaos that exists outside the torus bubble region.
207:
208: In all our studies the transitions to different SNAs at different
209: parts of the border in the $(A-f)$ parameter plane and
210: characterization of them is carried out on the
211: basis of specific quantities such as Lyapunov exponents and their
212: variance as well as finite-time Lyapunov exponents, dimensions, power
213: spectral measures and phase sensitivity exponents. Brief details of
214: these characterizing quantities are given in the Appendix A. In Sec. II we
215: describe the phenomenon of the remerging of Feigenbaum tree in the
216: absence of quasiperiodic forcing in the map (2). The existence of remerging
217: torus doubling is pointed out in Sec. III. Various transitions
218: to SNAs through the truncated doubled torus are demonstrated in Sec. IV. In
219: particular, the truncation of torus doubling bifurcation and the birth of SNAs
220: through torus collision, fractalization and type--III intermittent mechanisms
221: have been explained. Further, a sudden expansion of the attractor causing the truncation
222: of torus doubling bifurcation and the genesis of SNA is also demonstrated.
223: In Sec. V, the transition from one-frequency torus to SNA through
224: type--I intermittent as well as fractalization mechanisms is described.
225: In Sec. VI the transitions between different SNAs are discuused.
226: In
227: Sec. VII we address the issue of distinguishing among SNAs formed by
228: different routes through the use of finite-time Lyapunov exponents. Finally,
229: in Sec. VIII, the results are summarized.
230:
231:
232: \section{Remerging of Feigenbaum trees in the absence of quasiperiodic forcing}
233:
234: To start with, we consider the system (2) and numerically iterate it by
235: varying the values of $A$ and $Q$. For any $Q$ value and low $A$
236: values, the system (2) exhibits periodic oscillations with period 1T.
237: As $A$ increases, a bifurcation occurs and the stable period T orbit
238: transits into a stable period 2T bubble, as shown in Fig.~1(a). For
239: example, when the value of $Q$ exceeds a certain critical value
240: $Q$=$-$0.99 for a fixed $A$, $A$=1.5, a transition from period T orbit
241: to period 2T orbit occurs on increasing $Q$, essentially due to period
242: doubling bifurcation. Then the period 2T attractor merges and forms a
243: period T attractor when the value of $Q$ increases to $Q$=0.99 at the
244: same fixed $A$. At even higher values of $A$, $A$=1.7, the primary
245: period 2T bubble bifurcates into secondary period 4T bubbles, as shown
246: in Fig.~1(b). This bubble develops into further bubbles as $A$ gets larger,
247: until an infinitely branched Feigenbaum tree leading to the onset of
248: chaos finally appears, as shown in Fig.~1(c) for $A$=1.8.
249:
250: Bier and Bountis showed that such a remerging of Feigenbaum trees is
251: quite common in certain models possessing a kind of reflection
252: symmetry property coupled with more than one parameter \cite {R25}.
253: Further, they added that the formation of primary period 2T bubble is
254: seen to lead to higher order bubbles and the development of the
255: associated universal route to chaos in these systems. It is also
256: stated in the literature that the reversal of period doubling occurs
257: when the system possesses a positive Schwarzian derivative at the
258: bifurcation point \cite {R27,R28}. This is true for the present case
259: that we study. However, there are some counter examples as pointed out
260: by Nusse and Yorke \cite {R28} to show that the positivity of
261: Schwarzian derivative is not a sufficient condition to rule out the
262: period halving bifurcations.
263:
264: In the present paper, our aim is to investigate the effect of a
265: quasiperiodic forcing on the system (2) as given by Eq.(3).
266: In particular, we point out that with the addition of the quasiperiodic
267: forcing for a fixed $Q$, the dynamics is dominated by quasiperiodic
268: attractors and transitions to chaos via strange nonchaotic attractors
269: (SNAs) along different routes in contrast to the type of attractors
270: shown in Fig.~1. For this purpose we also work out a two parameter $(A-f)$
271: phase diagram (Fig.~2) to identify the changes in the dynamics.
272:
273:
274: \section{SNAs in the quasiperiodically forced cubic map}
275:
276: Now we consider the dynamics of the quasiperiodically driven map (3)
277: and numerically iterate it with the value of the parameter $\omega$
278: fixed at $\omega = {{\sqrt{5} - 1} \over 2}$ and by varying the
279: values of $A$ and $f$ for different fixed values of $Q$. The results
280: are then summarized in a suitable two-parameter $(A-f)$ phase diagram
281: for each fixed value of $Q$. Various dynamical behaviours $-$
282: quasiperiodic, strange nonchaotic, and chaotic attractors $-$ have been
283: identified by characterizing the attractors by quantities such as
284: Lyapunov exponents and their variance as well as finite-time Lyapunov
285: exponents, dimensions, power spectral measures and phase sensitivity
286: exponents (for details, see in Appendix A).
287:
288: In the absence of the external forcing $(f=0)$, from Fig.~1, we can
289: easily check that for fixed $Q$ and for given $A$ the dynamics
290: corresponds to periodic or chaotic attractors. For instance, for $Q=0$
291: and for any values of $A$, the system admits a period-2 solution.
292: Similarly, for $Q=0.25$ and $A=1.8$, it is a period-4 orbit, while for
293: $Q=0.5$ and $A=1.8$, it is a chaotic orbit. We now include the effect
294: of quasiperiodic forcing $(f \ne 0)$ and analyse the dynamics involving
295: torus, period doubled torus and chaos via SNAs. A very clear picture of
296: the various types of transitions becomes available for the case $Q=0$
297: in the region $f \in (-0.8,0.8)$ and $A \in (0.8,2.4)$, while similar
298: structure arises in a larger region for other values of $Q$.
299: Consequently, we present in the following results for $Q=0$ only in
300: the form of the phase diagram in Fig.~2. The various features indicated
301: in Fig.~2 are summarised and the dynamical transitions are discussed in
302: the following.
303:
304: The general features of the phase diagram fall into a very interesting
305: pattern. It can be observed from Fig.~2 that the dynamics is symmetric
306: about $f=$0. Therefore, in the following we present the details for the
307: right half of $f=0$ line only. The features are exactly similar in the
308: left half of the $f=0$ line. There are two chaotic regions $C1$ and
309: $C2$. Bordering these chaotic regions, one has the regions where the
310: attractors are strange and nonchaotic. Such SNAs are found to appear in
311: a large number of regions under various mechanisms, some of which are
312: marked GF1, GF2, GF3 \& GF4, HH, IC, S1 \& S3. Besides the strange
313: nonchaotic and chaotic attractors in the phase diagram Fig.~2, one can
314: also observe different regions where quasiperiodic attractors can be
315: found. In Fig.~2, such regions are marked as 1T and 2T, corresponding
316: to the quasiperiodic attractors of period-1 and period-2 respectively.
317: Fuller details are given below.
318:
319: For low $A$ and any $f$ value, the system exhibits quasiperiodic
320: oscillations denoted by 1T in Fig.~2. On increasing the value of $A$
321: further, the fascinating phenomenon of torus bubble appears within a
322: range of values of $f$. To be more specific,the parameter $A$ is for
323: example fixed at $A$=1.1 and then $f$ is varied. For $f$=$-$0.3, the
324: attractor is a quasiperiodic one (1T). As $f$ is increased to
325: $f$=$-$0.18, the attractor undergoes torus-doubling bifurcation and the
326: corresponding orbit is denoted as 2T in Fig.~2. As $f$ is increased
327: further, one then expects that the doubled attractor continues the
328: doubling sequence as in the case of the generic period-doubling
329: phenomenon. Instead, in the present case, the doubled attractor begins
330: to merge into that of a single attractor at $f$= 0.18, leading to the
331: formation of a torus bubble reminiscent of period bubbles in low
332: dimensional systems as in the previous section. On refixing the
333: parameter $A$ at higher values, one finds that there are two prominent
334: regions of chaotic oscillations C1 and C2 as shown in Fig.~2. The
335: chaotic region C1 exists outside the torus bubble region. That is, it
336: essentially occurs for larger A values, $A>$1.2 and $f>$0.6. On the
337: other hand, the region C2 emerges within the torus bubble region. That
338: is, it appears predominantly for even larger values of $A$, $A>$1.549
339: and $f$ lying between, $-$0.8$<f<$0.8. We have identified two
340: interesting dynamical transitions from one frequency quasiperiodicity
341: to chaos via SNAs outside the torus bubble region. They are (1)
342: gradual fractalization of the torus leading to creation of SNA (GF4),
343: and (2) type--I intermittent route leading to the birth of SNA (S1).
344: On the other hand there exist atleast four types of transitions to
345: chaos via SNAs within the torus bubble where the doubling of torus is
346: interrupted, namely (1) Heagy-Hammel (HH), (2) fractalization
347: (GF1, GF2 \& GF3), (3) type--III intermittent (S3) and (4) doubling of
348: destroyed tori routes through which the torus doubling bifurcation is
349: truncated and the birth of strange nonchaotic attractor takes place.
350: The details for each of the regions are given in the following sections.
351:
352: \section{Dynamics within the torus bubble}
353:
354: In this section, we will describe each one of the four types of transition
355: to chaos via
356: SNAs within the torus bubble region in detail.
357:
358: \subsection{Heagy-Hammel route}
359:
360: The first of the routes which we encounter is the Heagy-Hammel route in
361: which a period$-$ 2$^n$ torus gets wrinkled and upon collision with its
362: unstable parent period$-$ 2$^{n-1}$ torus bifurcates into an SNA \cite
363: {R8}. Such a route has been identified in the region C2 within the
364: range of $A$ values, 1.549$<A<$2.183, and $f$ values, 0.39$<f<$0.8.
365: That is the doubling bifurcation is truncated due to the collision of
366: the doubled torus with its unstable parent on increasing the value of
367: A, in the range 1.549$<A<$2.183, for a fixed $f$ value (0.39$<f<$0.8).
368: It is denoted as HH in Fig.~2. For example, let us fix the parameter
369: $f$ at $f$=0.7 and vary $A$. For $A$=1.8, the attractor is a
370: quasiperiodic one, as denoted by 1T in Fig.~2. As $A$ is increased to
371: $A$=1.876, the attractor undergoes torus doubling bifurcation and the
372: corresponding periodic orbit is denoted as 2T in Fig.~2. In the
373: generic case, the period doubling occurs in an infinite sequence until
374: the accumulation point is reached, beyond which chaotic behaviour
375: appears. However, with tori, in the present case, the truncation of the
376: torus doubling begins when the two strands of the 2T attractor become
377: extremely wrinkled. For example, when the value of $A$ is increased to
378: $A$=1.8868, the attractor becomes wrinkled as shown in Fig.~3(a). At
379: this transition, the strands are seen to come closer to the unstable
380: period 1T orbit and lose their continuities when the strands of torus
381: doubled orbit collide with unstable parent and ultimately result in a
382: fractal attractor as shown in Fig.~3(b) when $A$ is increased to
383: $A$=1.88697. At such a value, the attractor, Fig.~3(b), possesses a
384: geometrically strange property but does not exhibit any sensitivity to
385: initial conditions (the maximal Lyapunov exponent is negative as seen
386: in Fig.~4(a)) and so it is indeed a strange nonchaotic attractor. At
387: this transition, the two branches of the wrinkled attractor collide and
388: form a one band SNA. This kind of transition is similar to the
389: attractor merging crisis occuring in chaotic systems \cite {R26}. As
390: $A$ is increased further to $A$=1.8878, the attractor has eventually a
391: positive Lyapunov exponent and hence it corresponds to a chaotic
392: attractor (C2).
393:
394: Now we examine the Lyapunov exponent for the transition from torus to
395: SNA. Fig.~4(a) is a plot of the maximal Lyapunov exponent as a
396: function of $A$ for $f$=0.7. When we examine this in a
397: sufficiently small neighbourhood of the critical value $A_{HH}$=1.88697,
398: the transition is clearly revealed by the Lyapunov exponent which
399: varies smoothly in the torus region ($A<A_{HH}$) while it varies
400: irregularly in the SNA region ($A>A_{HH}$). It is also possible to
401: identify this transition point by examining the variance of the Lyapunov
402: exponent, as shown in Fig.~4(b) in which the fluctuation is small in the
403: torus region while it is large in the SNA region.
404:
405: In addition, in order to distinguish the quasiperiodic attractor and
406: the strange nonchaotic attractor, we may examine the attractor with
407: reference to the phase $\theta$ of the external force. The details of
408: this analysis are given in Appendix A. From Eq.(A.4), one infers that
409: the function $\Gamma_N$ grows infinitely for a SNA with some relation
410: such as $\Gamma_N = N^{\mu}$, where $\mu$ is a positive quantity which
411: characterises the SNA and we may call it the phase sensitivity
412: exponent. For the present case, it is $\mu$=0.98. However, in the case
413: of a chaotic attractor, it grows exponentially with N (see Fig.
414: 4(c)).
415:
416:
417:
418: \subsection{Fractalization route}
419:
420: The second of the routes is the gradual fractalization route where a
421: torus gets increasingly wrinkled and then transits to a SNA without
422: interaction (as against the previous case) with a nearby unstable orbit
423: as we change the system parameter. In this route, a period$-$2$^n$
424: torus becomes wrinkled and then the wrinkled attractor gradually loses
425: its smoothness and forms a 2$^n-$band SNA as we change the system
426: parameter. Such a phenomenon has been identified in the present system
427: in three different regions indicated as GF1, GF2 and GF3 in Fig. 2. To
428: exemplify this nature of transition, we fix the parameter $f$ at
429: $f$=0.1 and vary $A$ in the GF3 region. For $A$=1.0, the system
430: exhibits quasiperiodic oscillation of period-1T. The attractor
431: undergoes a torus doubling bifurcation as $A$ is increased to
432: $A$=1.06. Increasing the $A$ value further, a second period doubling
433: of the doubled torus does not take place. Instead, oscillations of the
434: doubled torus in the amplitude direction starts to appear at $A$=2.165
435: as shown in Fig.~5(a). As $A$ is increased further to $A$=2.167, the
436: oscillatory behaviour of the torus approaches a fractal nature
437: gradually. At such values, the nature of the attractor is strange [see
438: Fig.~5(b)] eventhough the largest Lyapunov exponent in Fig.~6(a)
439: remains negative. Such a phenomenon is essentially a gradual
440: fractalization of the doubled torus as was shown by Nishikawa and
441: Kaneko in their route to chaos via SNA \cite {R9}. In this route, there
442: is no collision involved among the orbits and therefore the Lyapunov
443: exponent increases only slowly as shown in Fig.~6(a) and there are no
444: significant changes in its variance [see Fig.~ 6(b)]. Further, the
445: phase sensitivity function $\Gamma_N$ grows unboundedly with the
446: power-law relation $\Gamma_N= N^{\mu}$, $\mu$=0.83 in the SNA region,
447: while they are bounded in the torus region [see Fig. 6(c)]. At even
448: higher values of $A$, $A$=2.17, the system exhibits chaotic
449: oscillations (C2). The quantity $\Gamma_N$ grows exponentially with N
450: for the chaotic attractors.
451:
452:
453:
454: \subsection{Type III intermittent route}
455:
456: The third one of the routes that is predominant in this system within
457: the torus doubled region is an intermittent route in which the torus
458: doubling bifurcation is tamed due to subharmonic bifurcations leading
459: to the creation of SNA. Such a phenomenon has been identified within
460: the range of $f$ values 0.33$<f<$0.41 and on increasing the value $A$,
461: 1.81$<A<$2.18, for a fixed $f$. To illustrate this transition, let us
462: fix the parameter $f$ at $f$=0.35 and vary $A$. For $A$=1.0, the
463: attractor is quasiperiodic attractor. As $A$ is increased to $A$=1.28,
464: the attractor undergoes torus doubling bifurcation. On increasing the
465: value of $A$ further, $A$=2.13, the attractor starts to wrinkle. On
466: further increase of $A$ value to $A$=2.135, the attractor becomes
467: extremely wrinkled and has several sharp bends, as shown in Fig.
468: 7(a). It has been observed in lower dimensional chaotic systems
469: \cite{R28,R29} that when the system undergoes subcritical period
470: doubling bifurcation, the dynamical behaviour exhibits type III
471: intermittent motion. In a similar manner,one finds
472: that the wrinkled attractor undergoes a quasiperiodic analogue of
473: subcritical period doubling bifurcation, on increasing the value of $A$
474: further to $A$=2.14. The corresponding intermittent motion is shown in
475: Fig.~7(b). The emergence of such intermittent dynamical behaviour has
476: been found in different continuous systems by the present authors and
477: their collaborators through the intermittent route to chaos via SNA
478: \cite {R12d} in which it was shown that during the transition from
479: torus doubled attractor to SNA, a growth of subharmonic amplitude
480: begins together with a decrease in the size of the fundamental
481: amplitude. At the critical parameter value, the intermittent attractor
482: loses its smoothness and becomes strange. The attractor shown in
483: Fig.~7(b) is nothing but a strange nonchaotic one as the Lyapunov
484: exponent turns out to be negative [Fig.~8(a)]. When examining the
485: Lyapunov exponent at this transition, it has been observed in Fig.~8(a)
486: that the Lyapunov exponent shows an abrupt change with a power-law
487: dependence on the parameter on the SNA side of the transition and the
488: variance shows a remarkable and abrupt increase at the transition point
489: as shown in Fig.~8(b). Further, the phase sensitivity function
490: $\Gamma_N$ is bounded for the torus region while it is unboundedly
491: changing with a power-law variation with $N$ for SNA region [Fig.~8(c)]
492: with $\mu =0.85$. On increasing the value of $A$ further to $A$=2.153,
493: we find the emergence of chaotic attractor (C2) where the quantity
494: $\Gamma_N$ grows exponentially with N [Fig.~8(c)].
495:
496: In the HH case, the points on the SNA are distributed over the entire
497: region enclosed by the wrinkled bounding torus, while in the GF case
498: the points on the SNA are distributed mainly on the boundary of the
499: torus. Interestingly in the present case shown in Fig.~7(b), most of
500: the points of the SNA remain within the wrinkled torus with sporadic
501: large deviations. The dynamics at this transition obviously involves a
502: kind of intermittency. Such an intermittency transition could be
503: characterised by the scaling behaviour. The laminar phase in this case
504: is the torus while the burst phase is the nonchaotic attractor. In
505: order to calculate the associated scaling constant, we coevolve the
506: trajectories for two different values of $A$, namely $A_c$ and another
507: nearby value of $A_c$, while keeping identical initial conditions
508: ($x_i$,$\theta_i$) as well as the same parameter value $f$. As the
509: angular coordinate $\theta_i$ remains identical, the difference in
510: x$_i$ allows one to compute the average laminar length between the
511: bursts and it fits with the scaling form $$ <l> = (A_c-A)^{-\alpha}.
512: \eqno(4) $$ The numerical value obtained for the attractor shown in
513: Fig. 7(b) is $\alpha
514: \sim$ 1.1 [see Fig.~9(a)]. To confirm further that the SNA attractor
515: [Fig.~7(b)] is associated with intermittent dynamics, we plot the
516: frequency of laminar periods of duration $\tau$, namely N($\tau$) in
517: Fig.~9(b). It obeys the scaling \cite {R30} law, $$ N(\tau) \sim \left
518: \{ {\exp (-4 \epsilon \tau) \over [1- \exp(-4 \epsilon \tau)]
519: } \right \} ^{0.5}. \eqno(5) $$ We find that
520: $\epsilon$=0.007$\pm$0.0002 to give a best fit for the present data.
521: These characteristic studies suggest that the intermittency is of type
522: III as discussed by Pomeau and Manneville in low dimensional systems
523: \cite {R29,R30}.
524:
525: \subsection{Crisis-induced intermittency}
526:
527: In the previous subsections, we have seen that the period doubling
528: bifurcation of a torus has been truncated by the destruction of it
529: leading to the emergence of an SNA in certain regions of the $(A-f)$
530: parameter space. Further, we observe that in the present system in
531: some cross sections of the $(A-f)$ parameter space the period doubling
532: phenomenon still persists in the destroyed torus eventhough the actual
533: doubling phenomenon has been truncated. But in the present case, it is
534: observed that the doubling of destroyed torus involves a kind of sudden
535: widening of the attractor similar to the crisis phenomenon that occurs
536: in chaotic systems. Such a phenomenon has been observed in the present
537: model in a range of $f$ values, 0.13$<f<$0.24, and for a narrow range of
538: $A$ values, 2.12$<A<$2.14. It is denoted as IC in Fig.~2. For
539: example, let us choose $f$=0.2 and vary the value of $A$. For $A$=0.8,
540: the attractor is a quasiperiodic one. As $A$ is increased to
541: $A$=1.18, the system undergoes torus doubling bifurcation. On
542: increasing the value of $A$ further to $A$=2.138, the attractor begins
543: to wrinkle as shown in Fig. 10(a). On increasing the value of $A$
544: further, say, to $A$=2.1387 the wrinkled attractor undergoes torus doubling
545: bifurcation and the corresponding orbit is shown in Fig.~10(b). It is
546: also seen from Figs.~10(b) \&(c) that when A is slightly larger than
547: $A_{IC}=2.1387$ the orbit on the attractor spends long stretches of
548: time in the region to which the attractor was confined before the
549: crisis. At the end of these long stretches the orbit bursts out of the
550: old region and bounces around in the new region made available to it
551: by the crisis. It then returns to the old region for another stretch of
552: time, followed by a burst and so on. This kind of widening of
553: attractor usually occurs in the chaotic systems at a crisis \cite
554: {R26}. However, in the present case, we have shown such a possibility
555: in a quasiperiodically forced system which also tames the growth of
556: torus doubled cascade, giving birth to SNAs. The variation of the
557: Lyapunov exponent at such transition does not follow a uniform pattern
558: in contrast to the case of low-dimensional chaotic systems exhibiting
559: crisis phenomenon [see Figs.~11 (a) \& (b)]. In addition, the phase
560: sensitivity function $\Gamma_N$ grows with $N$ with a kind power-law
561: relations for SNA while it is bounded for the torus regions [see
562: Fig.~11 (c)]. On further increase of the value of $A$ to $A$=2.143,
563: the system exhibits chaotic oscillations (C2) and the quantity
564: $\Gamma_N$ grows exponentially with $N$.
565:
566: \section{Dynamics outside the torus bubble}
567:
568: Two additional interesting transitions take place outside the torus
569: bubble region, namely (1) gradual fractalization of the torus leading
570: to the creation of SNA, and (2) type--I intermittent route leading to the
571: birth of SNA. The details are as follows.
572:
573: \subsection{Fractalization route}
574:
575: The first one is the gradual fractalization route which is the same as
576: studied in the previous subsection IV B, but the only difference now is
577: that here a transition from a one-frequency torus (1T) to chaos via
578: SNA is realized through gradual fractalization process instead of the
579: transition from the 2T torus discussed above. Such a phenomenon is
580: identified in the lower side of the C1 region (GF4). Specifically, within
581: the region of $f$,0.58$<f<$0.8, on increasing the value of $A$ to the region
582: 1.211$<A<$1.569 for a
583: fixed $f$ the SNA is created through gradual fractalization route. At
584: first, a transition from one-frequency torus to a wrinkled attractor
585: takes place on increasing $A$, $A$=1.25 for $f$=0.7 as shown in Fig.12(a). The
586: wrinkled attractor loses its continuity considerably as $A$ is
587: increased further and then finally ends up with fractal phenomenon at
588: $A$=1.265 (see fig. 12(b)). It is very obvious from these transitions
589: that the torus gradually loses its smoothness and ultimately
590: approaches the fractal behaviour via SNA before the onset of chaos as
591: the parameter $A$ increases to $A$=1.3. In addition, it is observed that
592: there is no apparent interaction among the orbits. The property has been confirmed
593: through the calculation of maximal Lyapunov exponent, its variance and phase sensitivity
594: studies as in the period-doubled regions.
595:
596: \subsection{Type--I intermittent route}
597:
598: A different type of intermittent route, namely type--I \cite {R14}, via
599: SNA is also observed at the upper region of C1. Within the range of
600: values $f$,0.58$<f<$0.8, on increasing the value of $A$ in the range 1.5
601: $<A<$2.0, a transition from the chaotic attractor (C1) to SNA takes
602: place at first and then the SNA is eventually replaced by a
603: one-frequency quasiperiodic orbit through a quasiperiodic analogue of
604: saddle-node bifurcation. At this transition, the dynamics is found to
605: be again intermittent but of different type.
606:
607: To understand more about this phenomenon, let us consider the specific
608: parameter value $f$= 0.7 and vary $A$. For $A$=1.80165, the attractor is a
609: chaotic one (C1). As $A$ is increased to $A$=1.801685, the chaotic
610: attractor transits to SNA as shown in Fig. 13(a). On increasing the value
611: of $A$ further, an intermittent transition from the SNA to a torus as
612: shown in Fig. 13(b) occurs at $A$=1.8017 . At this transition, abrupt
613: changes in the Lyapunov exponent as well as its variance shows the
614: characteristic signature of intermittent route (indicated in Fig. 14 (a)
615: $\&$ (b)) to SNA as in the type III case. In addition, again
616: the quantity $\Gamma_N$ grows with $N$ with a power law relation for SNA while
617: it is bounded for the torus regions (see Fig. 14 c). However, in the chaotic
618: region, the quantity $\Gamma_N$ grows exponentially
619: with N.
620:
621: Further, the plots of laminar
622: length $<l>$ as a function of the derived bifurcation parameter
623: $\epsilon =A-A_c$, where A$_c$ is a critical parameter for the
624: occurrence of intermittent transition, for this attractor reveals a
625: power-law relationship of the form
626: $$
627: <l> = \epsilon^{-\beta}, \eqno(6)
628: $$
629: with
630: an estimated value of $\beta \sim $ 0.53 (fig. 15(a)).
631: Also, the plot
632: between the number of laminar periods N($\tau$) and the period length
633: $\tau$ [shown in Fig.~15(b)] indicates that after an initial steep decay
634: there is an increase to a large value of
635: N($\tau$). It also obeys the relation
636: $$
637: N(\tau) \sim { \epsilon \over 2c} \left \{
638: \tau+ \tan \left [ \arctan \left ({c \over \sqrt {\epsilon \over
639: u}}\right ) - \tau \sqrt{\epsilon u} \right ] - \arctan \left ({c \over
640: \sqrt { \epsilon \over u}} \right ) \tau - \tau^2 \sqrt{\epsilon \over
641: u}
642: \right \}, \eqno(7)
643: $$
644: where $c$ is the maximum value of $x(t)$, $u=0.9$
645: and $\epsilon=0.0005 \pm 0.00003.$
646: The above analysis
647: confirms that such an attractor is associated with the standard
648: intermittent dynamics of type I described in ref. \cite {R29,R30}.
649:
650: \section{Transition between different SNAs}
651:
652: In the previous sections, we have enumerated the several ways through which
653: SNAs are created from torus attractors. One might observe from the $(A-f)$
654: phase diagram, Fig.~2, that there are several regions where transitions from
655: one type of SNA to another type occurs along the borders separating
656: quasiperiodic and chaotic attractors. Particularly,
657: transitions occur between GF3 and IC, IC and GF2, GF2 and
658: S3, S3 and GF1, GF1 and HH, and, S1 and GF4. On closer scrutiny,
659: we find that there exists
660: a very narrow range of parameters between different regions of SNAs where
661: chaotic motion occurs. That is the SNA of one region transits to chaos before
662: exhibiting
663: a different nature of SNA in the next region. However, we refrain from giving
664: finer details as they do not seem to be of significance.
665:
666:
667:
668:
669:
670: \section{\protect\smallskip Signatures of finite time Lyapunov
671: exponents at the transition to Different SNAs}
672:
673: Recently, it has been noted by Prasad, Mehra and Ramaswamy \cite {R14}
674: that a typical trajectory on a SNA\ actually possesses positive
675: Lyapunov exponents in finite time intervals, although the asymptotic
676: exponent is negative. As a consequence, one observes the different
677: characteristics of the SNAs born through different mechanisms through a
678: study of the differences in the distribution of finite-time exponents
679: P(N,$\lambda $) \cite{R14}. For each of the cases, the distribution
680: can be obtained by taking a long trajectory and dividing into segments
681: of length N, from which the local Lyapunov exponent can be calculated.
682: In the limit of large N, this distribution will collapse to a $\delta$
683: function $P(N,\lambda) \rightarrow \delta (\Delta - \lambda)$. The
684: deviations from$-$ and the approach to $-$ the limit can be very
685: different for SNAs created through different mechanisms. Figs.~16
686: illustrate the distributions for $P(50,\delta)$ across the five
687: different transitions discussed in the present study. A common feature
688: of these cases is that $P(N,\lambda)$ is strongly peaked about the
689: Lyapunov exponent when the attractor is a torus, but on the SNA, the
690: distribution picks up a tail which extends into the local Lyapunov
691: exponent $\lambda > $0 region. This tail is directly correlated with
692: enhanced fluctuation in the Lyapunov exponent on SNAs. On the
693: fractalised SNA and doubling of SNA, the distribution shifts
694: continuously to larger Lyapunov exponents, but the shape remains the
695: same for torus regions as well as SNA regions while on the HH and
696: intermittent SNA, the actual shapes of the distribution on the torus
697: and the SNA are very different. One remarkable feature of intermittent
698: SNAs is that the positive tail in the distribution decays very
699: slowly.
700:
701: To quantify further on the distribution of finite-time Lyapunov
702: exponents, let us consider, for example, the fraction of exponents lying
703: above $\lambda$=0, $F_{+}(N)$ vs $N$ for the different SNAs. It has
704: been found that except for the intermittent SNA (both for type III and
705: I), for which $F_{+}(N) \sim N^{-\beta}$, this quantity decays
706: exponentially , $F_{+}(N) \sim exp (- \gamma N)$ for all other
707: transitions, with the exponents $\beta$ and $\gamma$ dependent strongly
708: on the parameters of the system. For the specific SNAs corresponding to
709: the parameters reported in the previous section these quantities take
710: the following values: For type III and type I, the $\beta$ values are
711: 0.38 and 0.71. However, for HH, GF and IC, the $\beta$ and $\alpha$
712: values are 0.27 \& 0.32, 0.31 \& 017 and 0.24 \& 0.34, respectively.
713:
714:
715:
716: \section{Conclusion}
717:
718: In this paper we have described the creation of SNAs through various
719: routes and mechanisms in a protypical example, namely the
720: quasiperiodically driven cubic map. These are summarised in Table I.
721: Torus doubling bifurcations are not mandatory for the creation of SNAs.
722: However, they are merely a convenient agent in setting the stage for
723: the appearance of SNAs. There are atleast four different mechanisms,
724: namely Heagy-Hammel, gradual fractilazation, type--III and and
725: crisis-induced intermittency through which the truncation of torus
726: doubling and the birth of SNAs occur. The truncation of torus doubling
727: and the genesis of SNA through crisis-induced intermittency is a new
728: one and also it is entirely different from the interior crisis
729: mechanism for appearence of SNAs as found by Witt {\it et al.} \cite{R33} We
730: have further observed atleast two different ways, namely type--I
731: intermittency and gradual fractalization, through which SNAs are formed
732: when a transition from one-frequency torus to chaos takes place. All
733: these phenomena have been identified in a two parameter $(A-f)$ phase
734: diagram.
735: To distinguish among the different mechanisms through which SNAs are
736: born, we have examined the manner in which the maximal Lyapunov
737: exponent and its variance change as a function of the parameters. In
738: addition, we have also examined the distribution of local Lyapunov
739: exponents and found that on different SNAs they have different
740: characteristics. The analysis confirms that in the intermittent SNAs,
741: the signature of the transition is a discontinuous change in both the
742: maximal Lyapunov exponent and in the variance. The chaotic component
743: on the intermittent SNA is long lived. As a consequence, a slow
744: positive tail in the P(N,$\lambda$) and a resulting power-law decay for
745: F$_+$(N) can be identified. For the other SNAs, the resulting
746: exponential decay for F$_+$(N) has been identified.
747:
748: \section*{Acknowledgment}
749:
750: This work forms of a Department of Science and Technology, Government
751: of India research project. A. V. wishes to acknowledge Council of
752: Scientific and Industrial Research, Government of India, for financial
753: support.
754:
755:
756:
757: \appendix
758: \section{Characterizations of SNA}
759: \subsection{ Phase sensitivity exponent}
760: In order to distinguish the smooth and the fractal torus (SNA and chaos), we
761: examine the attractor with reference to the phase $\theta$ of the external force.
762: Eventhough, no exponential divergence of orbits exists for both the smooth
763: and the fractal torus (SNA), they are different from each other in terms of the phase
764: sensitivity. Pikovsky and Feudal \cite {R16a,R17} have shown that how two points on the SNA
765: which have close $\theta$ values separate from each other by introducing the following phase
766: sensitivity exponent: To appreciate this, we note that the absolute value of the first
767: derivative of the orbit$| {\partial x_n \over \partial \theta_n} | $ fluctuates with time
768: and some times has large bursts.To see this, one can proceed as follows. An aribitarary large burst can appear when the
769: system is iterated for infinite time steps. By differentiating (3), one
770: obtains
771: $${\partial x_{n+1} \over \partial {\theta}} = -2\pi Q sin(2 \pi {\theta})-(A-3x_n^{2})
772: {\partial x_n \over \partial {\theta}}. \eqno(A.1) $$
773: So, starting from a suitable initial derivative $| {\partial x_0 \over \partial {\theta}} | $,
774: one can obtain derivatives at all points of the trajectory
775: $${\partial x_{N} \over \partial {\theta}} =S_N= \sum_{k=1}^{N} \left \{-2\pi Q sin(2 \pi {\theta}_{k-1})
776: \left [\prod_{i=0}^{N-k-1}-(A-3x_{k+i}^{2}) \right ] \right \}
777: + \prod_{i=0}^{N-1} \left (-(A-3x_{i}^2){\partial x_0 \over \partial {\theta}} \right ), \eqno(A.2) $$
778: with the condition that for $k=N$,
779: $$\prod_{i=0}^{N-k-1}\left [-(A-3x_{k+i}^{2}) \right ]=1. \nonumber$$ Naturally,
780: in the case of a smooth attractor if one iterates (A.1) and (A.2) starting from
781: aribitrary values of $x$ and $ {\partial x \over \partial \theta} $ and for large $N$, they
782: converge to the attractor and its derivative, respectively. Thus, partial sums
783: $S_N$ computed from (A.2) are bounded by the maximum derivative $ {\partial x \over \partial \theta}$
784: along the attractor. But in the case of fractal attractor, the attractor is nonsmooth and the
785: derivative $ {\partial x \over \partial \theta}$ does not exist, so the consideration above is no longer
786: valid. These can be illustrated by calculating the partial sums $S_N$ as given by (A.2). It has been found \cite{R16a,R17}
787: that the behviour of the sum can be
788: very intermittent. The key observation is that these sums are quite large and practically unbounded. Hence,
789: we plot the maximum of $|S_N|$,
790: $$\gamma_N(x,\theta)=max|S_N|. \eqno (A.3) $$
791: The value of $\gamma_N$ grows with $N$, which means that arbitararily large values of $|S_N|$ apppear. From this
792: it follows immediately that the attractor can not have finite derivative with respect to the external phase if the
793: attractor is nonsmooth. Consequently, the assumption of a finite derivative is inconsistent with the relation (A.2),
794: where the second term on the RHS is exponentially small and the first term on the RHS can be arbitrary large. Thus,
795: calculating the partial sums (A.2), we can distinguish strange (sums are unbounded) and nonstrange (sums are bounded)
796: attractor.
797:
798: The growth rate of the partial sums with time represents a degree of
799: strangeness of the attractor, and can be used as a quantitaive
800: characteristic of SNAs. For this purpose, we require a quantity which
801: is independent of a particular trajectory while it represents average
802: properties of the attractor. The appropriate quantity seems \cite{R16a,R17} to be the
803: minimum value of $\gamma_N(x,\theta)$ with respect to randomly chosen
804: initial points $(x,\theta)$:
805: $$\Gamma_N=min \gamma_N(x,\theta). \eqno
806: (A.4) $$
807: It allows a more reliable inference about whether the attractor is
808: nonsmooth. One can also infer from (A.4) that $\Gamma_N$ grows
809: infinitely for a SNA with some relations as $\Gamma_N = N^{\mu}$, where
810: $\mu$ is a postive quantity which characterises the SNA and we call it
811: the phase sensitivity exponent. However, in the case of a chaotic
812: attractor, it grows exponentially with $N$.
813:
814: \subsection{Variation of finite--time Lyapunov exponents}
815:
816: As the finite--time or local Lyapunov exponents ($\lambda_i$, $i=1,2,\cdots,M$)
817: depend on the initial conditions,
818: it will be relevant to consider the variance of the average Lyapunov
819: exponent $\Lambda $ about the $\lambda_i$'s, $
820: i=1,2,\cdots,M$.It is defined \cite{R12d,R14,R26,R32} as
821: $$
822: \sigma=\frac {1}{M} \sum_{i=1}^{M} (\Lambda-\lambda_i(N))^2. \eqno(A.5) $$
823:
824: In all our numerical calculations, we take $N=50$ and $M \cong 10^5$.
825:
826: The variation of the local Lyapunov exponents in a fixed time
827: interval $t$ can also be discussed by
828: examining the probability distributions $P(t,\lambda)$ for the exponents.
829: In fact, $P(t,\lambda)$ corresponds to counting the normalized number of times any one of the
830: $\lambda$ appears for fixed time $t$. That is,
831: the distribution of local Lyapunov exponents, which is a stationary quantity,
832: is defined as \cite{R26}
833: $$
834: P(t, \lambda) d\lambda \equiv \mbox{ Probability that}
835: ~\lambda(t)
836: \mbox{~takes a value between} ~~\lambda \mbox{~and~} \lambda + d\lambda. \eqno(A.6)
837: \label{prob}
838: $$
839: This is particularly useful in describing the structure and dynamics of
840: nonuniform attractors. In the asymptotic limit $t
841: \to \infty$, this distribution will collapse to a $\delta$ function,
842: $$P(t,\lambda) \to \delta (\Lambda - \lambda).$$
843: The deviations from this limit for finite
844: times, and the asymptotics, namely the approach to the limit can be
845: very revealing of the underlying dynamics \cite{R14}.
846:
847: One can also
848: calculate the arithmetic mean of all the distributions and obtain the variance
849: of the Lyapunov exponent $\Lambda$ as
850: $$
851: \sigma = \int_{ - \infty}^{\infty} (\Lambda - \lambda)^2 P(t,\lambda)
852: d \lambda. \eqno(A.7)
853: $$
854: Dividing the total length of the orbit into $M$ bins as before and defining
855: the local Lyapunov exponents as $\lambda_i$, replacing $P(t,\lambda_i)$ by $\frac {\delta(\Lambda-\lambda_i)}
856: {M}$, the above equation of variance goes over to the
857: form given by Eq.~(A.5).
858:
859:
860: \subsection{ Power spectrum analysis}
861:
862: To quantify the changes in the power spectrum (obtained using Fast Fourier Transform (FFT)
863: technique), one can compute the so called spectral distribution function $N ( \sigma) $,
864: defined to be the number of peaks in the Fourier amplitude spectrum larger than
865: some value say $ \sigma $. Scaling relations have been predicted for $N ( \sigma) $
866: in the case of two and three frequency quasiperiodic attractors and strange nonchaotic attractors.
867: These scaling relations are $N(\sigma) \sim \ln {1 \over \sigma}$,$N(\sigma) \sim \ln^2 {\sigma},$
868: and $N(\sigma) \sim \sigma^{-\beta} $, respectively, corresponding to the two, three frequency
869: quasiperiodic and strange nonchaotic attractors. In the work of Romeiras and Ott \cite{R12},
870: the power law exponent was found empirically to lie with in the range
871: $1 < \beta < 2$ for the strange nonchaotic attractor.
872:
873: \subsection{ Dimensions}
874: To quantify geometric properties of attractors, several methods have been used to
875: compute the dimension of the attractors. Among them, we have used the
876: correlation dimension (introduced by Grassberger and Procaccia \cite{R31}) in our present
877: study, which may be computed
878: from the correlation function C(R) defined as
879: $$ C(R) =\lim_{N \to \infty} \left [ {1 \over N^2} \sum_{i,j=1}^N H(R- \mid x_i -x_j \mid) \right ], \nonumber $$
880: \noindent
881: where $x_i$ and $x_j$ are points on the attractor, H(y) is the Heaviside function
882: (1 if y $\geq$ 0 and 0 if y $ < $ 0), N is the number of points randomly chosen
883: from the entire data set. The Heaviside function simply counts the number of points
884: within the radius R of the point denoted by $x_i$ and C(R) gives the average fraction of
885: points. Now the correlation dimension is defined by the variations of C(R) with R:
886: $$ C(R) \sim R^d \hskip 20pt \mbox{as} \hskip 20pt R \rightarrow 0. \nonumber $$
887: Therefore the correlation dimension (d) is the slope of a graph of log C(R) versus log R.
888: Once one obtains the dimensions of the attractors, it will be easy to quantify strange
889: property of the attractors. In all our studies, we have verified that the SNAs have noninteger
890: correlation dimensions.
891:
892: \begin{references}
893:
894: \bibitem{R1} K. Kaneko, {\it {Collapse of tori and genesis of chaos in dissipative
895: systems}}, (World Scientific, Singapore), (1986).
896:
897: \bibitem{R2} V. Franceschini, Physica D6, 285 (1983).
898:
899: \bibitem{R3} A. Aren\'eodo, P.H. Coullet and E. A. Speigel, Phys. Lett
900: 94A, 1 (1983).
901:
902: \bibitem{R4}J. M. Flesselles, V. Croquette and S. Jucquois, Phys. Rev.
903: Letts 72, 2871 (1994).
904:
905: \bibitem{R5} K. E. Mckell, D. Broomhead, R. Jones and D.T.J. Hurle,
906: Euro. Phys. Lett. 12, 513 (1990).
907:
908: \bibitem{R6}M. R. Basset and J. L. Hudson, Physica D35, 289 (1989).
909:
910: \bibitem{R7} J.E. Kim, H.K. Park and H.T Moon, Phys. Rev. E55, 3948
911: (1997).
912:
913: \bibitem{R8} J. F. Heagy and S. M. Hammel, Physica 70D, 140 (1994).
914:
915: \bibitem{R9} K. Kaneko, Prog. Theor. Phys. 71, 140 (1994); T. Nishikawa and
916: K. Kaneko, Phys. Rev. E54, 6114 (1996).
917:
918: \bibitem{R9a} S. Kuznetsov, U. Feudal and A. Pikovsky, Phys. Rev. E57, 1585 (1998).
919:
920: \bibitem{R10} A. Venkatesan and M. Lakshmanan, Phys. Rev E55, 4140
921: (1997), Phys. Rev. E58, 3008 (1998).
922:
923: \bibitem{R11} C. Grebogi, E. Ott, S.Pelikan and J.A. Yorke, Physica
924: 13D, 261 (1984); C. Grebogi, E. Ott, F.J. Romeiras, and J. A. Yorke,
925: Phys. Rev. A 36, 5365 (1987).
926:
927: \bibitem{R12}
928: M. Ding, C.Grebogi and E. Ott, Phys.
929: Rev A39, 2593 (1989);F.J. Romeiras and E. Ott, Phys. Rev A35, 4404 (1987).
930:
931: \bibitem{R12a}
932: F.J. Romeiras, A. Bonderson, E. Ott, T.M. Andonsen jr. and C. Grebogi,
933: Physica 26D, 277 (1987); A. Bonderson, E. Ott and T. M. Andonsen jr.,
934: Phys. Rev. Lett 55, 2103 (1985).
935:
936: \bibitem{R12b} M. Ding and J.A. Scott Relso, Int. J. Bifurcation
937: and Chaos 4, 553 (1994).
938:
939: \bibitem{R12c} J.F. Heagy and W.L. Ditto, J. Nonlinear Sci 1, 423 (1991).
940:
941: \bibitem{R12d}
942: A. Venkatesan, M. Lakshmanan, A. Prasad and R. Ramaswamy, Phys. Rev. E61,
943: 3641 (2000).
944:
945: \bibitem{R13} T. Yalcinkaya and Y. C. Lai, Phys. Rev. Lett. 77, 5040
946: (1996); T. Yalcinkaya and Y. C. Lai, Phys. Rev E56, 1623 (1997).
947:
948:
949: \bibitem{R13a} A. Venkatesan, K. Murali and M. Lakshmanan, Phys.Lett A259, 246
950: (1999).
951:
952: \bibitem{R13b}Z. Liu and Z. Zhua, Int. J. Bifurcation, 6, 1383 (1996);
953: Z. Zhua and Z. Liu, Int. J. Bifurcation, 7, 227 (1997); T.~Yang ang
954: K.~Bilimgut, Phys. Lett A236, 494 (1997).
955:
956: \bibitem{R14} A. Prasad, V.Mehra and R. Ramaswamy, Phys. Rev. Lett 79,
957: 4127 (1997), Phys. Rev E57, 1576 (1998).
958:
959: \bibitem{R16} Y.C. Lai, Phys. Rev E53, 57 (1996); Y. C. Lai, U. Feudel
960: and C. Grebogi, Phys. Rev E54, 6114 (1996).
961:
962: \bibitem{R16a}
963: A. S. Pikovsky and U. Feudal, Chaos 5, 253 (1995); U. Feudal, J.
964: Kurths and A.S. Pikovsky, Physica 88D, 176 (1995).
965:
966: \bibitem{R17} A. S. Pikovsky and U. Feudal, J. Phys. A27, 5209 (1994);
967: S. P. Kuznetsov,
968: A.S. Pikovsky and U. Feudel, Phys. Rev. E51, R1629 (1995).
969:
970: \bibitem{R18} T. Kapitaniak and J. Wojewoda, {\it Attractors of
971: Quasiperiodically Forced Systems}, (World Scientific, Singapore, 1993);
972: T. Kapitaniak, Phys. Rev. E47, 1408 (1993); T. Kapitaniak and L. O.
973: Chua, Int. J. Bifurcation and Chaos 7, 423 (1997).
974:
975: \bibitem{R19} V.S. Anishchensko, T.K. Vadivasova, and O. Sosnovtseva,
976: Phys. Rev. E 53, 4451 (1996), O. Sosnovtseva, U. Feudel, J. Kurths and
977: A. Pikovsky, Phys. Lett A218, 255 (1996).
978:
979: \bibitem{R20} W. L. Ditto, M. L. Spano, H.T. Savage, S.N. Rauseo,
980: J.F. Heagy and E.Ott, Phys. Rev. Lett 65, 533 (1990).
981:
982: \bibitem{R21} T. Zhou, F. Moss and A. Bulsara Phys. Rev. A45, 5394
983: (1992).
984:
985: \bibitem{R22} W.X. Ding, H. Deutsch, A. Dingklage and C. Wilke, Phys.
986: Rev. E55, 3769 (1997).
987:
988: \bibitem{R23} M. Lakshmanan and K. Murali, Chaos in Nonlinear
989: Oscillators : Synchronization and Controlling, (World Scientific,
990: Singapore, 1996).
991:
992: \bibitem{R24} J. Guckenheimer and P. Holmes, Nonlinear Oscillations,
993: Dynamical Systems, and Bifurcation of Vector Fields (Springer-Verlag,
994: New York, 1983).
995:
996: \bibitem{R25} M. Bier and T. Bountis, Phys. Lett A104, 239 (1983).
997:
998: \bibitem{R26} E. Ott, Chaos in dynamical systems, (Cambridge University
999: Press, Cambridge, 1994).
1000:
1001: \bibitem{R27} A. Jackson, Perspectives of Nonlinear Dynamics (Cambridge
1002: University Press, New York, 1991).
1003:
1004: \bibitem{R28}H. E. Nusse and J. Yorke, Phys. Lett. A127, 328 (1988).
1005:
1006: \bibitem{R29} Y. Pomeau and P. Manneville, Commun. Math. Phys. 74, 189 (1980).
1007:
1008: \bibitem{R30} H. G. Schuster, Deterministic chaos, (Physik$-$Verlag,
1009: Weinheim, 1984).
1010:
1011: \bibitem{R33}A. Witt, U. Feudel and A. Pikovsky, Physica D109, 180 (1997).
1012:
1013: \bibitem{R32}G. Benettin, L. Galgani, and J.-M. Strelcyn,
1014: Phys. Rev. A14, 2338 (1976).
1015:
1016: \bibitem{R31} P. Grassberger and I. Procaccia, Physica D9, 189, (1983).
1017:
1018: \end{references}
1019:
1020:
1021:
1022: %\end{document}
1023:
1024: %\end{document}
1025: %\newpage
1026: \begin{figure}
1027: \centerline{\epsfig {figure=nfig1.eps,width=\linewidth}}
1028: \caption{Bifurcation diagram for the map
1029: (2) in the $(x,Q)$ plane: (a) The primary bubble at $A$=1.5, (b) Period-2 and period-4
1030: bubbles at $A$=1.7, (c) Period doubling route to chaos and inverse period doubling at $A$=1.8.}
1031: \end{figure}
1032: %\end{document}
1033: \begin{figure}
1034: \centerline{\epsfig {figure=nfig2.eps,width=\linewidth}}
1035: \caption{Phase diagram for quasiperiodically forced cubic map, Eq.~(3),
1036: in the$(A-f)$
1037: parameter space for $Q=0$: Here 1T and 2T correspond to torus of period one and
1038: of two respectively. GF1, GF2, GF3 and GF4 correspond to the regions
1039: where the process of gradual fractalization of torus occurs. HH
1040: represents the regions where SNA is created through the Heagy-Hammel
1041: route. S1 and S3 denote regions where the SNA appears through Type-I
1042: and Type-III intermittencies, respectively. IC denotes region where
1043: the SNA is created through crisis-induced intermittency. C1 and C2 correspond to
1044: chaotic attractors.
1045: }
1046: \end{figure}
1047: \begin{figure}
1048: \centerline{\epsfig {figure=nfig3.eps, width=\linewidth}}
1049: \caption{Projection of the attractors of Eqs.~(3) for
1050: $f$=0.7 in the $(x,\theta)$ plane indicating
1051: the transition from quasiperiodic attractor to chaotic attractor via SNA through
1052: Heagy- Hammel mechanism:
1053: (a) wrinkled attractor (period 2T) for $A$ = 1.8868; (b)
1054: SNA at $A$=1.88697.}
1055: \end{figure}
1056: \begin{figure}
1057: %\newpage
1058: \centerline{\epsfig {figure=nfig4.eps, width=0.5\linewidth}}
1059: \caption{Transition from doubled torus to SNA through Heagy- Hammel
1060: mechanism in the region HH: (a) the behaviour of the
1061: Lyapunov exponent ($\Lambda$); (b) the variance ($\sigma$); (c) plot of phase
1062: sensitivity function $\Gamma_N$ vs $N$ (dotted line corresponds to torus for $A$=1.83,
1063: dashed line belongs to SNA for $A$=1.88697 and solid line represents chaos for $A$=1.8878).}
1064: \end{figure}
1065: \begin{figure}
1066: \centerline{\epsfig {figure=nfig5.eps, width=\linewidth}}
1067: \caption{Projection of the attractors of Eqs.~(3) for
1068: $f$=0.1 in the $(x,\theta)$ plane indicating
1069: the transition from quasiperiodic attractor to chaotic attractor via SNA through
1070: gradual fractalization mechanism:
1071: (a) wrinkled attractor (period 2T) for $A$ = 2.165; (b)
1072: SNA at $A$=2.167.}
1073: \end{figure}
1074: %\newpage
1075: \begin{figure}
1076: \centerline{\epsfig {figure=nfig6.eps, width=0.5\linewidth}}
1077: \caption{Transition from doubled torus to SNA through gradual fractalization
1078: mechanism in the region GF3: (a) the behaviour of the
1079: Lyapunov exponent ($\Lambda$); (b) the variance ($\sigma$); (c) plot of phase
1080: sensitivity function $\Gamma_N$ vs $N$ (dotted line corresponds to torus for $A$=1.85,
1081: dashed line belongs to SNA for $A$=2.167 and solid line represents chaos for $A$=2.17).}
1082: \end{figure}
1083: %\newpage
1084: \begin{figure}
1085: \centerline{\epsfig {figure=nfig7.eps, width=\linewidth}}
1086: \caption{Projection of the attractors of Eqs.~(3) for
1087: $f$=0.35 in the $(x,\theta)$ plane indicating
1088: the transition from quasiperiodic attractor to chaotic attractor via SNA through
1089: type--III intermittent mechanism:
1090: (a) wrinkled attractor (period 2T) for $A$ = 2.135; (b)
1091: SNA at $A$=2.14.}
1092: \end{figure}
1093: %\newpage
1094: \begin{figure}
1095: \centerline{\epsfig {figure=nfig8.eps, width=0.4\linewidth}}
1096: \vskip 10pt
1097: \caption{Transition from doubled torus to SNA through type--III intermittent
1098: mechanism in the region S3: (a) the behaviour of the
1099: Lyapunov exponent ($\Lambda$); (b) the variance ($\sigma$);(c) plot of phase
1100: sensitivity function $\Gamma_N$ vs $N$ (dotted line corresponds to torus for $A$=1.83,
1101: dashed line belongs to SNA for $A$= 2.14 and solid line represents chaos for $A$=2.15).}
1102: \end{figure}
1103: %\newpage
1104: \begin{figure}
1105: \centerline{\epsfig {figure=nfig9.eps, width=\linewidth}}
1106: \caption { (a) Average laminar length ($<l>$) vs $(A-A_c)$ at $f$=0.35;
1107: (b) Number of laminar periods $N(\tau)$ of duration $\tau$
1108: in the case of transition through type-III intermittency.}
1109: \end{figure}
1110: %\newpage
1111: \begin{figure}
1112: \centerline{\epsfig {figure=nfig10.eps, width=0.8\linewidth}}
1113: \caption{Projection of the attractors of Eqs.~(3) for
1114: $f$=0.2 (i) in the $(x,\theta)$ plane; (ii) in the $(x,i)$ plane indicating
1115: the transition from quasiperiodic attractor to chaotic attractor via SNA through
1116: crisis-induced intermittent mechanism:
1117: (a) wrinkled attractor (period 2T) for $A$ = 2.138; (b)
1118: SNA at $A$=2.1387; (c) SNA at $A$=2.1388.}
1119: \end{figure}
1120: \begin{figure}
1121: \centerline{\epsfig {figure=nfig11.eps, width=.4\linewidth}}
1122: \caption{Transition from doubled torus to SNA through crisis-induced
1123: intermittency
1124: mechanism in the region IC: (a) the behaviour of the
1125: Lyapunov exponent ($\Lambda$); (b) the variance ($\sigma$);(c) plot of phase
1126: sensitivity function $\Gamma_N$ vs $N$ (dotted line corresponds to torus for $A$=1.83,
1127: dashed line belongs to SNA for $A$=2.1388 and solid line represents chaos for $A$=2.139).}
1128: \end{figure}
1129: %\newpage
1130: \begin{figure}
1131: \centerline{\epsfig {figure=nfig12.eps, width=\linewidth}}
1132: \caption{Projection of the attractors of Eqs.~(3) for
1133: $f$=0.7 in the $(x,\theta)$ plane indicating
1134: the transition from one-frequency quasiperiodic attractor to chaotic attractor via SNA through
1135: gradual fractalization mechanism:
1136: (a) wrinkled attractor (Period 1T) for $A$ = 1.25; (b)
1137: SNA at $A$=1.265.}
1138: \end{figure}
1139: %\newpage
1140: \begin{figure}
1141: \centerline{\epsfig {figure=nfig13.eps, width=\linewidth}}
1142: \caption{Projection of the attractors of Eqs.~(3) for
1143: $f$=0.7 in the $(x,\theta)$ plane indicating
1144: the transition from one-frequency quasiperiodic attractor to chaotic attractor via SNA through
1145: Type-I intermittency mechanism:
1146: (a) intermittent SNA for $A$ = 1.801685; (b)
1147: torus at $A$=1.8017.}
1148: \end{figure}
1149: %\newpage
1150: \begin{figure}
1151: \centerline{\epsfig {figure=nfig14.eps, width=.4\linewidth}}
1152: \caption{Transition from intermittent SNA to torus through type I intermittent
1153: mechanism in the region S1: (a) the behaviour of the
1154: Lyapunov exponent ($\Lambda$); (b) the variance ($\sigma$);
1155: (c) plot of phase
1156: sensitivity function $\Gamma_N$ vs $N$ (dotted line corresponds to torus for $A$=1.8017,
1157: dashed line belongs to SNA for $A$=1.801685 and solid line represents to chaos for $A$=1.8015).}
1158: \end{figure}
1159: \begin{figure}
1160: \centerline{\epsfig {figure=nfig15.eps, width=\linewidth}}
1161: \caption {(a) Average laminar length between ($<l>$) vs $(A-A_c)$ at $f$=0.7;
1162: (b) Number of laminar periods $N(\tau)$ of duration $\tau$
1163: in the case of transition through type-I intermittency.}
1164: \end{figure}
1165: \begin{figure}
1166: \centerline{\epsfig {figure=nfig16.eps, width=\linewidth}}
1167: \caption{Distribution of finite--time Lyapunov exponents on
1168: SNAs created through (a) Heagy-Hammel, (b) gradual fractalization, (c) type III intermittency
1169: and (d) crisis-induced intermittency (e) type I intermittency.
1170: Solid and dashed lines correspond to SNA and torus distributions.}
1171: \end{figure}
1172: \newpage
1173:
1174: \begin{table}
1175: Table I Routes and mechanisms of the onset of various SNAs in the
1176: quasiperiodically forced cubic map
1177: \begin{tabular}{l|l|l|l|l|l|l}
1178: {Type of Route} & {Mechansim} & \multicolumn{4}{c|}{Characteristic properties} & { Figures } \\
1179: \hline
1180: \multicolumn{2}{c|}{} & {Lyapunov}&{Variance $\sigma$} &{The fraction of } & {Scaling law} & \\
1181: \multicolumn{2}{l|}{} & {exponent $\lambda$} & & {positive valued} & {$<l> $} &\\
1182: \multicolumn{2}{l|}{}& & &{finite-time }& {$\sim $} & \\
1183: \multicolumn{2}{l|}{}& & & {Lyapunov}&{$({A_c-A})^{\alpha}$} & \\
1184: \multicolumn{2}{l|}{}& & & {exponent} & & \\
1185: \multicolumn{2}{l|}{}& & & {($F_+{(N,\lambda)}$)} & & \\
1186: \hline
1187: \multicolumn{7}{l}{ A. Interruption of torus doubling } \\
1188: \hline
1189: {1. Heagy-}& {Collision between a period- }&{Irregular in SNA} &{Small in torus}&{Decays} & & { Figs.~3}\\
1190: {Hammel [8]}&{doubled torus and its}&{region \& smooth } & {\& large in SNA} & {exponentially} & &{\& 4} \\
1191: &{unstable parent}& {in torus} & & & & \\
1192: \hline
1193: {2. Gradual }& {Torus gets increasingly } &{Increases slowly } &{No significant}&{Decays}& &{ Figs.~5}\\
1194: {fractalization [9]}&{wrinkled and transforms} &{during the transi} &{changes}&{exponentially} & &{\& 6} \\
1195: &{into a SNA without any} &{-tion from torus} & & & & \\
1196: &{interaction with a nearby} & {to SNA} & & & & \\
1197: &{unstable periodic orbit} & & & & \\
1198: \hline
1199: {3. Type--III} & {During the transition from} & {Abrupt change } & {Abrupt increase} &{Power-law} & {$\alpha \sim 1.1$}&{Figs.~7,8} \\
1200: {intermittency} &{torus doubled attractor to} & {during the} & {at the transi} &{variation} &{(see also} & {\& 9}\\
1201: {[15]}&{SNA, a growth of subhar} &{transition from} &{-tion point} & &{eq.(5))} \\
1202: &{-monic amplitude begins } & {torus to SNA} & & & \\
1203: &{together with a decrease } & & & & &\\
1204: &{in the size of the } & & & & \\
1205: &{fundamental amplitude} & & & & \\
1206: \hline
1207: {4. Crisis-induced}&{Doubling of destroyed torus} &{Does not follow} & {Irregular variation} & {Decays} & & {Figs.~10,} \\
1208: {intermittency} & {involves a kind of sudden }&{uniform pattern} &{in SNA region} & {exponentially} & &{\& 11} \\
1209: &{widening of the attractor} & & & & \\
1210: \hline
1211: \multicolumn{7}{l}{B. Transition from one-frequency torus} \\
1212: \hline
1213: {1. Gradual} & {Torus gets increasingly} & {Increases slowly } & {No significant} & {Decays} & &{Fig.~12} \\
1214: {fractalization} & {wrinkled and transforms} & {during the transition } & {change} & {exponentially} & & \\
1215: {[9]} & {into a SNA} & {from torus to SNA} & & & & \\
1216: \hline
1217: {4. Type--I} & {Torus is eventually } & {Abrupt change } & {Abrupt increase} &{Power-law} & {$\alpha \sim
1218: 0.55$}&{Figs.~13,} \\
1219: {intermittency} &{replaced by SNA } & {during the} & {at the transi} &{variation} &{(see also} &{14,\& 15} \\
1220: {[14]}&{through an analog } &{transition from} &{-tion point} & &{eq.(7))} &\\
1221: &{of the saddle-node } & {torus to SNA} & & & \\
1222: &{bifurcation } & & & & & \\
1223: \end{tabular}
1224: \end{table}
1225: \end{document}
1226:
1227:
1228:
1229:
1230:
1231:
1232:
1233:
1234:
1235:
1236:
1237: