1: \documentstyle[aps,pre,epsfig]{revtex}
2: %\documentstyle[twocolumn,aps,prl]{revtex}
3: \def\pr#1#2#3{{ Phys. Rev.}, {\bf #1}, #2 (#3)}
4: \def\physd#1#2#3{{ Physica D}, {\bf #1}, #2 (#3)}
5: \def\prl#1#2#3{{ Phys. Rev. Lett.}, {\bf #1}, #2 (#3)}
6: \def\pla#1#2#3{{ Phys. Lett. A}, {\bf #1}, #2 (#3)}
7: \def\pra#1#2#3{{ Phys. Rev. A}, {\bf #1}, #2 (#3)}
8: \def\pre#1#2#3{{ Phys. Rev. E}, {\bf #1}, #2 (#3)}
9: \def\jpa#1#2#3{{ J. Phys. A}, {\bf #1}, #2 (#3)}
10: \def\jsp#1#2#3{{ J. Stat. Phys.}, {\bf #1}, #2 (#3)}
11: \def\ijbc#1#2#3{{ Int. J. Bifurcation and Chaos} {\bf #1}, #2 (#3)}
12: \def\e{$\epsilon$}
13: \def\etc{ etc.~}
14: \def\ie{i.e.}
15: \def\eg{e. g.}
16: \def\etl{$et ~al.$~}
17: \def\MLE{$\Lambda$}
18: \def\MLEN{\MLE$_N$}
19: \def\sna{SNA }
20: \def\R{$R$}
21: \def\h{$h$}
22: \def\beqr{\begin{eqnarray}}
23: \def\beq{\begin{equation}}
24: \def\eqn{\end{equation}\noindent}
25: \def\eqnr{\end{eqnarray}\noindent}
26: \def\btt#1{{\tt$\backslash$#1}}
27: \def\plaa#1{{ Phys. Lett. A}, (#1)}
28: \begin{document}
29:
30: \draft
31: \title{Intermittency transitions to strange nonchaotic attractors in
32: a quasiperiodically driven Duffing oscillator}
33: \author{A. Venkatesan and M. Lakshmanan}
34: \address{Center for Nonlinear Dynamics, Department of Physics,
35: Bharathidasan University, Tiruchirappalli - 620 024, India}
36: \author{A. Prasad and R. Ramaswamy}
37: \address{School of Physical Sciences, Jawaharlal Nehru University,
38: New Delhi - 110 067, India}
39: \date{\today}
40: \maketitle
41: \begin{abstract}
42: Different mechanisms for the creation of strange nonchaotic attractors
43: (SNAs) are studied in a two--frequency parametrically driven Duffing
44: oscillator. We focus on intermittency transitions in particular, and
45: show that SNAs in this system are created through quasiperiodic
46: saddle-node bifurcations (Type-I intermittency) as well as through a
47: quasiperiodic subharmonic bifurcation (Type-III intermittency). The
48: intermittent attractors are characterized via a number of Lyapunov
49: measures including the behavior of the largest nontrivial Lyapunov
50: exponent and its variance as well as through distributions of
51: finite--time Lyapunov exponents. These attractors are ubiquitous in
52: quasiperiodically driven systems; the regions of occurrence of various
53: SNAs are identified in a phase diagram of the Duffing system.
54: \end{abstract}
55:
56: \section{Introduction}
57:
58: Interest in the dynamics of quasiperiodically driven systems has grown
59: in recent years largely due to the existence of novel behavior such as
60: strange nonchaotic dynamics. The initial work of Grebogi \etl
61: \cite{R1} showed that with quasiperiodic forcing, nonlinear systems
62: could have strange nonchaotic attractors (SNAs), namely attractors with
63: a fractal geometry, but with non-positive Lyapunov exponent. Subsequent
64: studies have dealt with a number of important issues pertaining to
65: theoretical as well as experimental aspects \cite
66: {R2,R3,R4,R5,R6,R7a,R7b,R8,R9,R10,R11,R12,R13,R14,R15} of SNAs.
67:
68: While the existence of SNAs is firmly established, a question that
69: remains interesting is the mechanism or bifurcations through which
70: these are created from regular or chaotic attractors. To date a number
71: of different scenarios have been identified: these include torus
72: doubling to chaos via SNAs\cite{R3}, fractalization of torus\cite{R4},
73: the re-emergence of torus doubling sequence and the birth of SNAs
74: \cite{R5}, the occurrence of SNAs via blow out bifurcation\cite{R6}, the
75: appearance of SNAs through Type-I\ intermittent phenomenon\cite{R7a}, or
76: Type-III intermittency\cite{R8}, and so on
77: \cite{R9,R10,R11,R12,R13,R14,R15}.
78:
79: Scenarios for the formation of SNAs often have parallels in scenarios
80: for the formation of chaotic attractors. The most common route to SNAs
81: is the gradual ``fractalization'' of a torus \cite{R4} where an
82: amplitude or phase instability causes the collapse of the torus which
83: progressively gets more and more wrinkled as a parameter in the system
84: changes, eventually becoming a fractal attractor. This is also the
85: least well understood mechanism for the formation of SNAs since there
86: is no apparent bifurcation, unlike the crisis--like torus collision
87: mechanisms identified by Heagy and Hammel \cite{R3} and Feudel, Kurths
88: and Pikovsky \cite{R10}. In the former instance, a period--doubled
89: torus collides with its unstable parent, while in the latter, a stable
90: and unstable torus collide at a dense set of points, leading to SNAs.
91: The quasiperiodic analogue of a saddle--node bifurcation gives rise to
92: SNAs through the intermittent route \cite{R7a}, with the dynamics
93: exhibiting scaling behavior characteristic of Type--I intermittency
94: \cite{R22}. The quasiperiodic subharmonic bifurcation, on the other
95: hand, gives rise to Type--III intermittent SNAs \cite{R8}, when a torus
96: doubled attractor is interrupted by a subharmonic bifurcation,
97: resulting in the inhibition of torus doubling sequence.
98:
99: Our prime concern in the present work is to understand how a typical
100: nonlinear system, namely the Duffing oscillator, responds to a
101: quasiperiodic forcing and exhibits different dynamical transitions
102: involving SNAs. In particular, intermittencies through which SNAs are
103: formed have been investigated, besides the standard mechanisms like
104: fractalization and torus collision. Although some of these mechanisms
105: have already been identified in certain maps and continuous systems
106: \cite{R3,R4,R7a,R7b}, in order to generalize such dynamical transitions in
107: real physical systems, we undertake our investigation on a damped, two
108: frequency parametrically driven double--well Duffing
109: oscillator\cite{R16}
110: \begin{equation}
111: \ddot{x}+h\dot{x}-[1+A(R\cos t+\cos \Omega t)]x+x^3=0
112: \label{eq1}
113: \eqn
114: which is a well-suited model \cite{HD} for buckled beam oscillations.
115: The simplest experimental realization of the above equation, the
116: magnetoelastic ribbon has been extensively studied \cite{R16,R17}, and
117: is the first system where SNAs were observed \cite{R13} with $\Omega$
118: chosen to be an irrational number. The existence of several routes to
119: SNAs in Eq.~(\ref{eq1}) suggests that there may be experimental
120: realizations of different types of SNAs which are deserving of further
121: study; our analysis here is thus of some experimental relevance.
122:
123: The parametrically driven Duffing oscillator, Eq.~(\ref{eq1}), is
124: a rich dynamical system, possessing a variety of regular, strange
125: nonchaotic and chaotic
126: attractors. We concentrate on the intermittent transitions to SNAs and
127: the mechanism by which they arise in a range in $(R-h)$ parameter
128: space. (In addition to the intermittency routes mentioned above,
129: Yal\c{c}inkaya and Lai \cite{R6} have shown that on--off intermittency
130: can also be associated with SNAs created through a blowout bifurcation,
131: when a torus loses its transverse stability\cite{R6}. This bifurcation
132: does not occur in the present system in the range of parameters
133: studied.)
134:
135: SNAs represent dynamics which is intermediate between regular and
136: chaotic motion, and therefore they need to be distinguished from
137: chaotic attractors and quasiperiodic (torus) attractors. There are
138: several quantitative criteria that can be used to determine the strange
139: nonchaotic nature \cite{R1,R2,R3,R4,R5,R6,R7a,R7b,R8,R9,R10,R11,R12}. The
140: most direct characterization of SNAs is through the largest Lyapunov
141: exponent (which should be nonpositive) and the existence of a
142: nontrivial fractal structure. We employ these criteria in the present
143: work in order to identify strange nonchaotic behaviour: Lyapunov
144: exponents are calculated in the usual manner \cite {standard}, and the correlation
145: dimension (calculated through the standard Grassberger---Procaccia
146: algorithm \cite{grassberger}) is used to determine whether or
147: not the attractor is
148: fractal. In addition to the fact that Lyapunov exponents are negative
149: on SNAs, the variance of the Lyapunov exponents on SNAs is also large.
150: (Here, by variance we mean the fluctuations in the measured value of
151: finite--time Lyapunov exponents, as calculated from several
152: realizations of the dynamics; {see the Appendix}). This quantity
153: also shows characteristic changes across
154: transitions to SNAs (particularly when there are crises) \cite{R7b}.
155:
156: A host of other properties have been used to characterise SNAs, such as
157: the scaling of spectral features or the ``phase sensitivity'' \cite{R10}.
158: SNAs are characterised by specific signatures in their frequency
159: spectrum wherein they admit a power law relation: $N(\sigma) \sim
160: \sigma ^{- \alpha}, 1< \alpha < 2$, where the spectral distribution
161: function $N(\sigma)$ is defined as the number of peaks in the fourier
162: amplitude spectrum larger than some value $\sigma$. Another measure
163: \cite{R10} is based on the presence of a complicated path between the
164: real and imaginary Fourier amplitudes which reflects the fractal
165: geometry of SNAs.
166:
167: Finer distinctions among SNAs formed via different mechanisms can be
168: made by use of various measures, \eg~ the nature of the variation of
169: the Lyapunov exponents and its variance near the transition values of
170: the control parameter \cite{R7b}, the nature of the bursting and
171: scaling laws in the case of intermittent types SNAs \cite{R7a}, the
172: statistical properties of finite-time Lyapunov exponents \cite{trichy}
173: and so on.
174:
175: In the next section, the birth of SNAs associated with the mechanisms
176: mentioned above are discussed. In particular, the transition of a
177: two--frequency quasiperiodic attractor $\rightarrow $ torus doubled
178: orbit $\rightarrow $ SNA (through type-III intermittency) $\rightarrow
179: $ chaos and the transition from chaos $\rightarrow $ SNA (through
180: type-I intermittency) $\rightarrow $ torus, besides the standard ones,
181: are shown to be operative in Eq.~(\ref{eq1}). In Section-III, a number
182: of Lyapunov measures such as the behaviour of the largest Lyapunov
183: exponent, its fluctuations, the distribution of finite-time Lyapunov
184: exponents are used to characterize the transitions from two-frequency
185: quasiperiodicity to SNAs. Our results are summarized in Section IV.
186:
187: \section{ The Parametrically driven Duffing Oscillator}
188: For our analysis, Eq.~(1) can be rewritten as
189: \beqr
190: \dot{x}&=&y, \\
191: \dot{y}&=&-hy+[1+A(R\cos \phi +\cos \theta )]x-x^3,\label{eq2} \\
192: \dot{\phi} &=&1,~~~~\dot{\theta} = \Omega.
193: \eqnr
194: Note that the three equilibrium points of the system for $A=0$
195: correspond to $-x+x^3=0$, so that there are two stable fixed points
196: $x_{\pm }^s=\pm 1$ and an unstable fixed point at $x^u=0$.
197:
198: Fig.~1(a) is the overall phase diagram for the system with the
199: parameters $A=0.3, \Omega = {\sqrt{5}+1}/2,$ fixed, while
200: $R$ and $h$ are varied. The dynamical equations are integrated
201: numerically using a fourth-order Runge-Kutta algorithm with adaptive
202: step size. The dynamical states and transitions among them are
203: characterised through the Lyapunov exponents and its variance as well as
204: a number of different measures. The relevant details of the calculation
205: of Lyapunov exponents and their variance are given in the Appendix.
206:
207: There are a number of different regions where periodic, chaotic and
208: strange nonchaotic attractors can be found: Figs.~1(b), 1(c) and 1(d)
209: are blow--ups of regions W1, W2 and W3 respectively in Fig.~1(a). The
210: various features indicated in the phase diagram are summarized and the
211: dynamical transitions are elucidated in the following.
212:
213: The general features of the phase diagram fall into a familiar pattern.
214: Compared to the Duffing oscillator driven by a single frequency (the
215: case of $R = 0$), there are new chaotic regions C1, C2 and C3, and
216: bordering the chaotic regions, one has regions where the attractors are
217: strange and nonchaotic. The different regions where quasiperiodic
218: attractors can be found are also separated here by regions of chaotic
219: attractors and SNAs. In Fig.~1, the region marked TL contains
220: quasiperiodic attractors that oscillate about all the three equilibrium
221: points, $x^u, x^s_{\pm}$, while TS denotes a small quasiperiodic orbit
222: which oscillates around one of the stable fixed points alone,
223: depending on the initial conditions. The strip denoted D contains
224: interesting dynamical states, both chaotic attractors and SNAs, between
225: which there are a number of transitions.
226:
227: SNAs are found in a large number of regions, some of which are marked
228: GF1, GF2, GF3, HH1, HH2, S1 and S3 (based on the scenarios responsible
229: for their creation). It should also be pointed out that boundaries
230: separating different dynamical states are very uneven in this phase
231: diagram, which should be considered representative and schematic. In
232: order to illustrate the fine transitions that take place in certain
233: regions when parameter varies, we also present the Lyapunov spectrum as
234: a function of $h$ for fixed $R$ in Fig.~2. Fuller details are
235: discussed below.
236:
237: Periodic orbits of the forced system with $R =0$ become quasiperiodic
238: tori of the system with nonzero $R$. As $R$ is further increased, these
239: tori typically bifurcate via period doubling. Upon further increase of
240: the parameters, there can be further bifurcations or other
241: transformations of the torus attractors to SNAs.
242:
243: We first discuss the intermittency routes to SNA that can be observed
244: in this system.
245:
246: \subsection{Type~~III intermittency}
247:
248: In some regions in $(R-h)$ space the torus doubling sequence is tamed
249: due to subharmonic bifurcations which lead to the creation of SNAs
250: \cite {R8}. We find that a growth of the subharmonic amplitude begins
251: together with a decrease in the size of the fundamental amplitude; such
252: behaviour is characteristic of the so--called Type~III intermittency
253: \cite {R22,R23}. The transition from a period--doubled torus to
254: intermittent SNA takes place in the region marked S3 in Fig.~1. From
255: Figs.~1(a) and 1(c), it can be observed that initially the large
256: quasiperiodic orbit (TL) undergoes a torus doubling bifurcation to the
257: torus attractor denoted 2TL. One would then expect the doubling
258: sequence to continue as in the usual period-doubling cascade, but
259: instead, here the doubling is interrupted by the formation of an
260: intermittent SNA which then finally settles into the chaotic attractor
261: C2 as $h$ is increased.
262:
263: To understand the mechanism of the interruption of the doubling cascade
264: consider a specific parameter value of $R = 0.47$ while $h$ is varied
265: [see Figs.~2(a,b)]. For $h=0.08$, the attractor is a two-frequency
266: torus, but beyond this, further period doubling of doubled torus does
267: not take place. Instead, a new dynamical behaviour, namely
268: intermittent phenomenon starts appearing at $h=h_c$=0.088689. This
269: transition is clearly illustrated in Figs.~3 where we note in the (x,
270: $\phi$ mod $4\pi$)\footnote{The various bifurcations and different
271: routes to SNAs can be easily identified by displaying $\phi$ modulo
272: $4\pi$ instead of modulo $2\pi$ \cite{HD}.} plane that the amplitude of
273: one of the components increases while the amplitude of the other
274: component decreases when a transition from doubled torus to
275: intermittent phenomenon takes place. To exemplify this transition
276: further in the Fourier spectrum, it has been identified that the
277: amplitude of the subharmonic component (W2/2) increases while the
278: amplitude of fundamental component (W2) {\it decreases} during this
279: transition (see Figs.~3). This suggests that the birth of the
280: intermittent SNA is through a quasiperiodic analogue of the subharmonic
281: bifurcation. At this transition, the amplitude variation loses its
282: regularity (Figs.~3) and a burst appears in the regular phase. This
283: behaviour repeats as time increases as observed in the usual
284: intermittent scenario\cite{R22,R23}. Also, the duration of the laminar
285: phases (namely the quasiperiodic orbit) is random. At the intermittent
286: transition the distinctive signature is an abrupt change in the
287: Lyapunov exponent as well as in its variance as a function of $h$, as
288: shown in Fig.~4(a,b). This type of SNA occurs in the region
289: $0.088689<h<0.088962$. On further increase of the value of $h$ from $h=
290: 0.088963$, we find the emergence of a chaotic attractor (C2) shown in
291: Fig.~3(d), which though visibly similar to the SNA [see Fig.~3(c)] has a
292: positive Lyapunov exponent.
293:
294: To confirm further that the SNA attractor, Fig.~3(c), is associated
295: with standard type-III intermittent dynamics, we plot the frequency of
296: laminar periods of duration $\tau$, namely N($\tau$) in
297: Fig.~4(c). This obeys the scaling \cite{R23,R24}
298: \beq
299: N(\tau) \sim \left \{ {\exp (-4 \epsilon \tau) \over [1- \exp(-4
300: \epsilon \tau)]
301: } \right \} ^{0.5}.
302: \eqn
303: We find $ \epsilon=0.009 \pm 0.0003 $ to give a best fit for
304: the present data.
305:
306:
307: \subsection{Type~~I intermittency}
308:
309: On the right edge of the chaotic region C3, there is a transition from
310: a chaotic attractor to a SNA and then to a quasiperiodic torus, TS.
311: This transition proceeds via Type--I intermittency \cite{R7a} in the
312: region marked S1 in Fig.~1(a).
313:
314: Consider the specific parameter value $R=0.35$ and vary $h$; for
315: $h=0.1907$, the attractor is chaotic [Fig.~5(a)], and as $h$ is
316: increased to $h=0.190833$, the chaotic attractor transforms to an SNA
317: shown in Fig.~5(b). On increasing the value of $h$ further,
318: an intermittent transition from the SNA to a torus as shown in
319: Fig.~5(c) occurs at $h_c = 0.19088564\ldots$. At this transition, the
320: abrupt changes in the Lyapunov exponent as well as its variance
321: [Fig.~6(a,b)] shows the characteristic signature of the
322: intermittent route
323: to SNA as in the Type-III case discussed above. Here the SNA, hopping
324: between the two wells of the system, transforms to the small
325: quasiperiodic torus TS which oscillates in one of the
326: wells as in the case
327: of periodically driven Duffing oscillator \cite {R16}. Also, the plot
328: between the number of laminar periods N($\tau$) and the period length
329: $\tau$ [shown in Fig.~6(c)] indicates that after an initial steep decay
330: there is a slight hump and then a fall off to a small value of
331: N($\tau$). It also obeys the relation \cite {R23,R25}
332: \beq
333: N(\tau) \sim { \epsilon \over 2c} \left \{
334: \tau+ \tan \left [ \arctan \left ({c \over \sqrt {\epsilon \over
335: u}}\right ) - \tau \sqrt{\epsilon u} \right ] - \arctan \left ({c \over
336: \sqrt { \epsilon \over u}} \right ) \tau - \tau^2 \sqrt{\epsilon \over
337: u}
338: \right \}, \eqn where $c$ is the maximum value of $x(t)$, $u=5.0$
339: and $\epsilon=0.0003 \pm +0.00002.$
340:
341: \subsection{Other routes to SNAs}
342:
343: In addition to the intermittency routes discussed above, the Duffing
344: system has the usual scenarios of torus collision as well as
345: fractalization. The details are as follows.
346:
347: \subsubsection {Torus collision}
348:
349: Torus collisions---the route identified by Heagy and Hammel \cite{R3}--
350: are denoted HH in Fig.~1. In this scenario, a period$-$2$^n$ torus
351: attractor gets wrinkled and upon collision with its parent unstable
352: 2$^{n-1}$ torus, a 2$^{n-1}$-band SNA is formed. Such a route has been
353: identified in two different regions (HH1 and HH2) of the present
354: system; the resulting SNAs have somewhat different morphologies.
355:
356: To exemplify the nature of this transition, let us fix the parameter
357: \R~ at 0.3 while \h~ is varied. For \h=0.19, one finds the
358: presence of a small quasiperiodic attractor which is denoted as TS in
359: Fig.~1, but by \h = 0.185, the attractor undergoes a torus doubling
360: bifurcation and the corresponding region is denoted as 2TS in Fig.~1.
361: When \h~ is decreased further, the two strands of the doubled attractor
362: begin to wrinkle at \h $\approx$ 0.184 [Fig.~7(a)], and lose continuity
363: when \h $\approx$ 0.181, resulting in a fractal attractor as shown in
364: Fig.~7(b). The strange nonchaotic [the Lyapunov exponent is negative
365: as seen in Fig.~8(a)] attractor is thus born at the collision of a
366: stable period doubled torus 2TS and its unstable parent TS.
367: Furthermore, this is a one band SNA, as is clearly depicted in Figs.~7,
368: in which prior to the collision, the
369: dynamics in the (x, $\phi$ mod $2\pi$) plane corresponds to the two
370: branches [Fig.~7a(i)] while it corresponds to a simple curve in the (x,
371: $\phi$ mod 4$\pi$) plane [Fig.~7a(ii)]. However after the collision the
372: dynamics now goes over all of the attractor in both cases of (x, $\phi$
373: mod $2\pi$) [Fig.~7b(i)] and (x, $\phi$ mod $4\pi$) plane
374: [Fig.~7b(ii)].
375:
376: The Lyapunov exponent also shows distinct change in its behaviour for
377: this type of transition. Fig.~8(a) is a plot of the maximal Lyapunov
378: exponent as a function of \h~ for a fixed value of \R~ =0.3. In the
379: neighborhood of $h_c$, $\Lambda$ varies smoothly in the torus region
380: but in the SNA phase the variation is irregular and the crossover
381: between these two behaviours is abrupt. It is also possible to
382: identify the transition point from the examination of the variance in
383: $\Lambda$ [which increases significantly on the SNA as shown in Fig.~8(b)].
384:
385: Similar behaviour has also been observed in the region denoted
386: HH2, for $ 0.37 < R < 0.5$ and $0.11 < h < 0.16$, where the large
387: period-doubled torus (2TL) becomes wrinkled and then interacts with its
388: unstable parent torus leading to the creation of a one band SNA.
389:
390:
391: \subsubsection {Fractalization}
392:
393: The process of fractalization \cite{R4}, whereby a torus continuously
394: deforms and wrinkles to form a SNA can be seen in regions marked GF1--3
395: in Fig.~1. The qualitiative (geometric) structure of the attractors
396: remains more or less the same during the process, unlike the
397: intermittency routes or like the torus collision route discussed in the
398: previous subsections. In this route, a period$-2^n$ torus becomes
399: wrinkled and then the wrinkled attractor gradually loses its smoothness
400: and forms a $2^n-$band SNA.
401:
402: This transition can take place both as a function of increasing or
403: decreasing $h$ in different regions of the parameter space as can be
404: seen in the phase diagram. Consider the case $R= 0.47$, with varying
405: $h$. For $h=0.05$, the attractor is chaotic [the Lyapunov exponent
406: (see Fig.~2) is positive]. As $h$ is increased to $0.0558$, one obtains
407: the attractor shown in Fig.~9(a) which is morphologically similar to
408: the chaotic attractor, but as it has a negative Lyapunov exponent
409: (Fig.~2) it is in fact an SNA. On increasing \h~ to $0.065$, the
410: attractor loses its fractal character, becoming a (wrinkled) large
411: quasiperiodic orbit [see Fig.~9(b)]. The Lyapunov exponent ($\Lambda$)
412: and its variance ($\sigma$) do not show any specific signature as in
413: the case of the intermittent SNA or HH SNA [see Fig.~10(a)].
414:
415: There are a variety of transitions involving fractalization in the
416: phase diagram. On increasing $h$ to $0.072$, along $R=0.47$ in the GF2
417: region, the wrinkled attractor again loses its continuity and
418: ultimately approaches a fractal attrcator [see Figs.~9(c,d)], getting
419: considerably more wrinkled with increasing $h$. Beyond $h = 0.0725$,
420: the attractor becomes a doubled large quasiperiodic orbit with period
421: 2TL. This transition is illustrated in Figs.~7(c,d) (see also Fig.~2)
422: in which the 2TL orbit has two bifurcated strands of length 2$\pi$ for
423: $x>0$ which are actually a single strand of length 4$\pi$. However if
424: one looks in the reverse direction of $h$, one can note that as the 2TL
425: orbit loses its smoothness and develops a fractal nature, the dynamics
426: in the (x, $\phi$ modulo $2\pi$) plane and (x, $\phi$ modulo $4\pi$)
427: plane are geometrically the same as in the case of fractalization in
428: the GF1 region described above. The behaviour of the Lyapunov exponent
429: and its variance lack specific signatures at the torus to SNA
430: transitions as demonstrated in Fig.~10(b).
431:
432: Yet another transition to SNA occurs through gradual fractalization in
433: the region GF3 where a transition from small quasiperiodic attractor
434: (TS) to chaos occurs through torus doubling bifurcation [Figs.~1(a) and
435: 1(d)]. Here the period doubled orbit 2TS gets fractalized, leading to
436: the formation of the 2-band SNAs when $h$ is decreased from higher
437: values [see Fig.~1(d)] and the phenomenon is similar to what has been
438: discussed above.
439:
440: \subsection {Transition between different SNAs}
441:
442: Previous subsections have enumerated the several ways through which
443: SNAs are created from torus attractors. These processes include
444: transitions such as ...nT $\leftrightarrow$ $n$ band SNA
445: $\leftrightarrow$ chaos, ...$2^n$T $\leftrightarrow$ $2^{n-1}$ band SNA
446: $\leftrightarrow$ chaos, etc. One might also observe from the $(R-h)$
447: phase diagram, Figs.~1 that there are regions where transitions from
448: one type of SNA to another type can occur. Particularly, one may note
449: that transitions occur between (GF2 and S3), (S3 and HH2), (S1 and
450: HH1), and (HH1 and GF3). On closer scrutiny, we find that there exists
451: a narrow range of chaotic motion between the GF2 and S3 as well as the
452: S3 and HH2 regions, while 2TS/TS attractors intervene in between S1 and
453: HH1. That is, the transition between S3 and HH2 regions corresponds to:
454: .. $n$ band SNA $\leftrightarrow$ $n$ band chaos $\leftrightarrow$ $n$
455: band SNAs ... and between the regions GF2 and S3 corresponds to: ...
456: $n$ band SNAs $\leftrightarrow$ $n$ band chaos $\leftrightarrow$ ${n
457: \over 2}$ band chaos $\leftrightarrow$ ${n \over 2}$ band SNAs.
458: However in the transition region between GF3 and HH1, see Fig.~1(d),
459: there is band--merging, namely a transition from $n$--band SNA to
460: $n/2$--band SNA. This transition \cite{negi} is analogous to the
461: phenomenon of reverse bifurcation or band merging bifurcations
462: occurring in chaotic systems, though here the dynamics remains
463: nonchaotic and strange during the transition. Such phenomena have been
464: identified in the driven Henon and circle maps and also in the logistic
465: map \cite{R7b,R12,negi}. In the present case we find that as the
466: parameter passes through critical values, the torus doubled attractor
467: does not collide in the GF3 region when undergoing fractalization,
468: whereas it does collide in the HH1 region, thereby leading to a
469: transition from one type of SNA to another (See Fig.~11).
470:
471: \section{Probability distributions of Finite--time Lyapunov exponents}
472:
473: Owing to the underlying fractal structure, trajectories on a SNA can be
474: locally unstable: finite--time Lyapunov exponents can be positive
475: although the global (or asymptotic) Lyapunov exponent is nonpositive.
476:
477: As described in the Appendix, we obtain the distribution of finite
478: time Lyapunov exponents, $P(t,\lambda)$ and the variance, through
479: Eq.~(A3) or (A4).
480: For most ``typical'' chaotic motion, a general argument suggests that
481: the density $P(t,\lambda)$ [Eq.~(\ref{prob})], should be normally
482: distributed \cite{gbp,R21}.
483: On the other hand, for intermittent dynamics, for instance, the system
484: switches between periodic and chaotic states and shows long-range
485: temporal correlations. The probablity distribution of these
486: finite-time LEs is asymmetric since it arises from a superposition of
487: two independent densities, each of which is {\it separately} a Gaussian
488: distribution, centered on a distinct value of the average Lyapunov
489: exponent \cite{pre}. On intermittent SNAs, one of the components is a
490: torus with negative or zero Lyapunov exponent, while the other
491: component is chaotic, with average Lyapunov exponent being positive.
492: Finite stretches of intermittent dynamics involve both kinds of motion,
493: and therefore the density of finite--time LEs for intermittency usually
494: results in a stretched exponential tail \cite{pre} and in a
495: markedly asymmetric distribution for $P(t,\lambda)$.
496:
497: Although in the infinite time limit all distributions will go to a
498: $\delta-$ function (see Eq.~(\ref{limit})), for finite times,
499: $P(t,\lambda)$ can be very different for SNAs created through
500: different mechanisms \cite{R7b,trichy}. In particular, when the
501: dynamics is intermittent \cite{pre}, the exponential tails
502: in the distribution persist for long times.
503: Shown in Fig.~12(a) are
504: distributions $P(2048,\lambda)$ for the two intermittent transitions
505: discussed here.
506: Both the distributions have an asymmetric tail which
507: extends well into the locally chaotic ({\it i.e.} Lyapunov exponent
508: $\Lambda > $0) region, even for such a long time interval.
509: This correlates with enhanced fluctuations in
510: the Lyapunov exponent on intermittent SNAs. On SNAs formed by the
511: fractalization or torus collision mechanisms, in contrast, similar
512: densities are essentially gaussian \cite{trichy}; these are shown
513: for comparison in Fig.~12(b).
514:
515: In order to quantify the slow decay of the positive tail in the distribution
516: for intermittent SNAs, we
517: define the fraction of postive local Lyapunov exponents \cite{pre} as
518: \begin {equation}
519: F_+(t)=\int_0^\infty P(t,\lambda) d\lambda.
520: \end{equation}
521: Clearly, lim$_{t \rightarrow \infty}F_+(t)\rightarrow 0$.
522: Empirically, it has been
523: found \cite {trichy,pre} that on the intermittent SNA, this quantity
524: shows the large $t$ behaviour
525: \begin{equation}
526: F_{+}(t) \sim t^{-\beta};
527: \end{equation}
528: we find numerically that the exponents in both the
529: S1 and S3 regions are $\beta \approx 1$ \cite{expo}.
530: For the fractalized or Heagy-Hammel SNAs, the approach is exponentially
531: fast,
532: \begin{equation}
533: F_{+}(t) \sim \exp(-\gamma t).
534: \end{equation}
535:
536: Note that in the long--time limit, all distributions will collapse
537: to the Gaussian, so that for very long times, $t \to \infty$,
538: $F_+$ will decay exponentially on all SNAs. The exponents $\beta$
539: and $\gamma$ depend strongly on the system parameters.
540:
541: \section{Conclusion}
542:
543: On SNAs, as a consequence of the fractal geometry, stable and unstable
544: regions are interwoven in a complicated manner. Thus, although
545: trajectories with different initial conditions will eventually coincide
546: with each other \cite{synchro} since there are no positive Lyapunov
547: exponents, they do so in an intermittent fashion, unlike
548: the case of quasiperiodic attractors,
549: converging in the
550: locally stable regions and diverging in the locally unstable regions.
551:
552: In this paper we have described the transition from quasiperiodic
553: attractors to chaos through different types of SNAs in a protypical
554: example, namely the two-frequency parametrically driven Duffing
555: oscillator. We have demarcated the different regions in parameter space
556: where these different dynamical states can be located.
557: There are several mechanisms through which SNAs are formed
558: here, two of which appear to be quasiperiodic analogies of
559: intermittency transitions in unforced systems \cite{R22}. In addition
560: to these, we also find evidence for the torus collision route to SNA as
561: well as fractalization.
562:
563: To distinguish among the different mechanisms through which SNAs are
564: born, we examined the manner in which the maximal Lyapunov exponent and
565: its variance changes as a function of the parameters. In addition, we
566: have also examined the distribution of local Lyapunov exponents and
567: found that on different SNAs they have different characteristics. Our
568: analysis confirms that in the intermittent SNAs, the signature of the
569: transition is a discontinuous change in both the maximal Lyapunov
570: exponent and in the variance. Further, the two different intermittency
571: routes are distinguished by their different scaling behaviours.
572: The chaotic component on both types of
573: intermittent SNAs is long lived, giving, as a consequence, a slowly
574: decaying positive tail in P(N, $\lambda$) and a resulting power-law
575: decay for F$_+$(N).
576:
577: Since the driven Duffing
578: system can be experimentally realized in the driven magnetoelastic
579: ribbon the present work can help in identifying different conditions
580: under which SNAs can be found in a typical system. The study of
581: SNAs---from both theoretical and experimental points of view---is in
582: its initial stages and the present study may aid in the
583: realization of different bifurcation routes to SNA in experimental
584: systems.
585:
586: \section {Acknowledgment} This work forms the part of Department of
587: Science and Technology, Government of India research projects to
588: ML and to RR (SP-S2-E07-96). A. V.
589: wishes to acknowledge Council of Scientific and Industrial Research,
590: Government of India, for financial support.
591:
592: \appendix
593: \section{Finite-time Lyapunov exponents}
594: The largest Lyapunov exponent $\Lambda$ which measures the rate of separation
595: of nearby trajectories can be computed via a standard algorithm \cite{standard}.
596: This is an asymptotic quantity, and is a long
597: (in principle infinite) time average of the local rate of expansion in
598: phase space. The finite--time or local Lyapunov exponent $\lambda_i(t)$ is defined
599: in an analogus manner \cite{pre,abar} except that this is computed over a finite time
600: interval, $t$. The subscript $i$ indexes the segment (that is, in effect,
601: the initial conditions)
602: in which this exponent is evaluated. A given trajectory is thus divided into
603: segments of length $t$, and in each of these, the local Lyapunov
604: exponent $\lambda_i(t)$ is computed.
605:
606: As the finite--time Lyapunov exponents depend on the initial conditions,
607: we consider the probability distributions $P(t,\lambda)$, defined as
608: \beq
609: P(t, \lambda) d\lambda \equiv \mbox{ Probability that}
610: ~\lambda_i(t)
611: \mbox{~takes a value between} ~~\lambda \mbox{~and~} \lambda + d\lambda.
612: \label{prob}
613: \eqn
614: This distribution is a stationary quantity, and
615: is particularly useful in describing the structure and dynamics of
616: nonuniform attractors \cite{gbp,R21}. In the asymptotic limit $t
617: \to \infty$, this distribution will collapse to a $\delta$ function,
618: \beq
619: P(t,\lambda) \to \delta (\Lambda - \lambda).
620: \label{limit}
621: \eqn
622: The deviations from this limit for finite
623: times, and the asymptotics, namely the approach to the limit can be
624: very revealing of the underlying dynamics \cite{pre}.
625:
626: From this distribution, one can calculate the variance
627: of the Lyapunov exponent $\Lambda$ as
628: \beq
629: \sigma = \int_{ - \infty}^{\infty} (\Lambda - \lambda)^2 P(t,\lambda)
630: d \lambda.
631: \eqn
632: Equivalently, one can obtain the variance as
633: \beq
634: \sigma=\frac {1}{M} \sum_{i=1}^{M} (\Lambda-\lambda_i(t))^2,
635: \eqn
636: namely from a set of $M$ finite--time Lyapunov
637: exponents. In our numerical calculations in Sec. II, for instance,
638: we take $t=50$ and $M \cong 10^5$.
639: \begin{references}
640: \bibitem{R1} C. Grebogi, E. Ott, S. Pelikan and J. A.
641: Yorke, \physd{13}{261}{1984}; C. Grebogi, E. Ott, F.J. Romeiras, and
642: J. A. Yorke, \pra{36}{5365}{1987}.
643:
644: \bibitem{R2} F. J. Romeiras and E. Ott, \pra{35}{4404}{1987}; F. J.
645: Romeiras, A. Bondeson, E. Ott, T. M. Antonsen Jr. and C. Grebogi,
646: \physd{26}{277}{1987}; A. Bondeson, E. Ott and T. M. Antonsen Jr.,
647: \prl{55}{2103}{1985}; M. Ding, C. Grebogi and E. Ott,
648: \pra{39}{2593}{1989};
649: M. Ding and J. A. Scott Kelso, \ijbc{4}{553}{1994}.
650:
651: \bibitem{R3} J. F. Heagy and S. M. Hammel, \physd{70}{140}{1994}.
652:
653: \bibitem{R4} K. Kaneko, {Prog. Theor. Phys.} {\bf 71}, 140 (1984);
654: T. Nishikawa and K. Kaneko, \pre{54}{6114}{1996}.
655:
656: \bibitem{R5} A. Venkatesan and M. Lakshmanan, \pre{55}{4140}{1997};
657: \pre{58}{3008}{1998}.
658:
659: \bibitem{R6} T. Yal\c{c}inkaya and Y. C. Lai, \prl{77}{5040}{1996}.
660:
661: \bibitem{R7a} A. Prasad, V. Mehra and R. Ramaswamy,
662: \prl{79}{4127}{1997}.
663: \bibitem{R7b} A. Prasad, V. Mehra and R. Ramaswamy,
664: \pre{57}{1576}{1998}.
665:
666: \bibitem{R8} A. Venkatesan, K. Murali and M. Lakshmanan,
667: \plaa{1999}, in press.
668:
669: \bibitem{R9} Y. C. Lai, \pre{53}{57}{1996}; Y. C. Lai, U. Feudel and
670: C. Grebogi, \pre{54}{6114}{1996}.
671:
672: \bibitem{R10} A. S. Pikovsky and U. Feudel, \jpa{27}{5209}{1994}; A.
673: S. Pikovsky and U. Feudel, {\it Chaos} {\bf 5}, 253 (1995);
674: U. Feudel, J. Kurths and
675: A. S. Pikovsky, \physd{88}{176}{1995}; S. P. Kuznetsov, A. S. Pikovsky and
676: U. Feudel, \pre{51}{R1629}{1995}.
677:
678: \bibitem{R11} T. Kapitaniak and J. Wojewoda, {\it Attractors of
679: Quasiperiodically Forced Systems}, (World Scientific, Singapore, 1993);
680: T. Kapitaniak, \pre{47}{1408}{1993}; T. Kapitaniak and L. O. Chua,
681: \ijbc{7}{423}{1997}.
682:
683: \bibitem{R12} V. S. Anishchensko, T. K. Vadivasova, and O. Sosnovtseva,
684: \pre{53}{4451}{1996}; O. Sosnovtseva, U. Feudel, J. Kurths and A.
685: Pikovsky, \pla{218}{255}{1996}.
686:
687: \bibitem{R13} W. L. Ditto, M. L. Spano, H. T. Savage, S. N. Rauseo, J. F.
688: Heagy and E. Ott, \prl{65}{533}{1990}.
689:
690: \bibitem{R14} T. Zhou, F. Moss and A. Bulsara \pra{45}{5394}{1992}.
691:
692: \bibitem{R15} W. X. Ding, H. Deutsch, A. Dinklage and C. Wilke,
693: \pre{55}{3769}{1997}.
694:
695: \bibitem{R22} Y. Pomeau and P. Manneville, {Commun. Math. Phys.}
696: {\bf74}, 189 (1980).
697:
698: \bibitem{R16} M. Lakshmanan and K. Murali, {\it Chaos in Nonlinear
699: Oscillators :
700: Synchronization and Control}, (World Scientific, Singapore, 1996).
701:
702: \bibitem{HD} J. F. Heagy and W. L. Ditto, {J. Nonlinear Sci.} {\bf
703: 1}, 423 (1991).
704: \bibitem{R17} W. L. Ditto, S. Rauseo, R. Cawley, C. Grebogi,
705: G. H. Hsu, E. Kostelich, E. Ott, H. T. Savage, R. Segnam,
706: M. L. Spano and J. A. Yorke, \prl{63}{923}{1989};
707: H. T. Savage, W. L. Ditto, P. A. Braza, M. L. Spano
708: and W. C. Spring, {J. Appl. Phys.} {\bf 67}, 5619 (1990);
709: H. T. Savage and M. L. Spano, {J. Appl. Phys.} {\bf 53}, 8002
710: (1982);
711: H. T. Savage and C. Adler, {J. Magn. Mater.} {\bf 58}, 320 (1986).
712: \bibitem{standard}G. Benettin, L. Galgani, and J.-M. Strelcyn,
713: \pra{14}{2338}{1976}.
714: \bibitem{grassberger} P. Grassberger and I. Procaccia, \physd{9}{189}{1983}.
715: \bibitem{trichy} A. Prasad and R. Ramaswamy,
716: {\sl Finite time Lyapunov exponents on Strange Nonchaotic Attractors},
717: (LANL Bulletin Board, chao-dyn/9801021); in {\sl Nonlinear Dynamics:
718: Integrability and Chaos}, Ed. M. Daniel \etl, Narosa Publishing Co.,
719: New Delhi, 1999, to appear.
720: \bibitem{R23} H. G. Schuster, {\it Deterministic chaos},
721: (Physik$-$Verlag, Weinheim, 1984).
722:
723: \bibitem{R24}M. Dubois, M. A. Rubio and P. Berge, \prl{51}{1446}{1983}.
724: \bibitem{R25}J. E. Hirsch, B. A. Hubermann and D. J. Scalapino,
725: \pra{25}{519}{1982}.
726:
727: \bibitem{negi} S. S. Negi, A. Prasad, and R. Ramaswamy, to be published.
728:
729: \bibitem{gbp} P. Grassberger, R. Badii, and A. Politi, {J. Stat.
730: Phys.} {\bf 51}, 135 (1988).
731: \bibitem{R21} E. Ott, {\it Chaos in dynamical systems}, (Cambridge
732: University Press, Cambridge, 1994).
733: \bibitem{pre} A. Prasad and R. Ramaswamy, \pre{60}{2761}{1999}.
734:
735: \bibitem{expo} At $R=0.45$ and $h=0.100736782$ (in the S3 region) the
736: exponent $\beta$ is $1.14$ and in the S1 region, at $R=0.35$ and
737: $h=0.190865$, the exponent is $0.89$.
738: \bibitem{synchro} R. Ramaswamy, \pre{56}{7294}{1997}.
739:
740: \bibitem{abar} H.D.I. Abarbanel, R. Brown and M.B. Kennel, J. Nonlinear
741: Sci. {\bf 1}, 175 (1991).
742:
743:
744: \end{references}
745:
746:
747:
748: \begin{figure}
749: \caption{Phase diagram for the parametrically driven Duffing
750: oscillator system, Eqs.~(2)-(4), in the
751: $(R-h)$
752: parameter space. Here 2TS and 2TL correspond to torus doubled
753: attractors of small and large quasiperiodic orbits,
754: respectively. GF1, GF2 and GF3 correspond to the regions where the
755: process of gradual fractalization of torus occurs. HH1 and HH2
756: represent the regions where SNA is created through the
757: Heagy-Hammel route.
758: S1 and S3 denote regions where the SNA appears through Type-I and
759: Type-III intermittencies respectively. C1, C2 and C3 correspond to
760: chaotic attractors. Regions W1, W2 and W3 in Fig.~1(a) are expanded
761: in Figs.~1(b), 1(c) and 1(d), and the inset in Fig.~1(b)
762: illustrates the intertwining of chaotic (C1) and quasiperiodic
763: orbits (TL); the two levels curves correspond to the specific
764: values of the maximal Lyapunov exponent:
765: $\Lambda$=0 ($...$) and $\Lambda$=-0.005 ($---$).}
766: \end{figure}
767:
768: \begin{figure}
769: \caption{The behaviour of the Lyapunov exponent as a function of $h$
770: for (a) $R=0.47$, (b) $R=0.47$ and (c) $ R=0.4$.
771: Here the notation is the same as in Fig.~1
772: (Tr stands for transients).}
773: \end{figure}
774:
775: \begin{figure}
776: \caption{Projection of the two-frequency attractors of Eqs.~(2)-(4) for
777: R=0.47, with
778: the Poincar\'e plot in the (x, $\phi$) plane (i) $\phi$ modulo
779: 2$\pi$
780: and (ii) $\phi$ modulo 4$\pi$ (iii) power spectrum indicating
781: the transition from chaotic to quasiperiodic attractors via SNA through
782: type-III intermittent mechanism:
783: (a) two frequency doubled quasiperiodic orbit for $h = 0.075$; (b)
784: doubled torus for
785: $h=0.08$; (c) intermittent SNA for $h=0.088689$; (d) chaotic attractor
786: for $h
787: =0.088963$.}
788: \end{figure}
789:
790: \begin{figure}
791: \caption{Transition from doubled torus to SNA through type-III
792: intermittent
793: mechanism in the region S3: (a) the behaviour of the
794: Lyapunov exponent ($\Lambda$); (b) the variance ($\sigma$); (c)
795: Number of laminar periods $N(\tau)$ of
796: duration $\tau$ in the case of transition through type-III
797: intermittency.}
798: \end{figure}
799:
800: \begin{figure}
801: \caption{Projection of the two-frequency attractors of Eqs.~(2)-(4) for
802: R=0.35, with
803: the Poincar\'e plot in the (x, $\phi$) plane (i) $\phi$ modulo 2$\pi$
804: and (ii) $\phi$ modulo 4$\pi$ indicating
805: the transition from quasiperiodic to chaotic attractors via SNA through
806: type-I intermittency mechanism:
807: (a) chaotic attractor for $h=0.1907$; (b) intermittent SNA for
808: $h=0.190833$;
809: (c) torus for $h=0.19088564$.}
810: \end{figure}
811:
812: \begin{figure}
813: \caption{Transition from torus to SNA through type-III
814: intermittent
815: mechanism in the region S3: (a) the behaviour of the
816: Lyapunov exponent ($\Lambda$); (b) the variance ($\sigma$); (c)
817: Number of laminar periods $N(\tau)$ of
818: duration $\tau$ in the case of transition through type-I
819: intermittency.}
820: \end{figure}
821:
822: \begin{figure}
823: \caption{Projection of the two-frequency attractors of Eqs.~(2)-(4) for
824: $R=0.3$, with
825: the Poincar\'e plot in the (x, $\phi$) plane (i) $\phi$ modulo 2$\pi$
826: and (ii) $\phi$ modulo 4$\pi$ indicating
827: the transition from chaotic to quasiperiodic attractors via SNA through
828: Heagy- Hammel (HH1) mechanism:
829: (a) torus doubled attractor for $h=0.184$ ;
830: (b) strange nonchaotic attractor for $h = 0.181$.}
831: \end{figure}
832:
833: \begin{figure}
834: \caption{Transition from doubled torus to SNA through Heagy-Hammel
835: mechanism in the region HH1: (a) the behaviour of the
836: Lyapunov exponent ($\Lambda$); (b) the variance ($\sigma$).}
837: \end{figure}
838:
839: \begin {figure}
840: \caption{Projection of the two-frequency attractors of Eqs.~(2)-(4) for
841: $R=0.47$, with
842: the Poincar\'e plot in the (x, $\phi$) plane (i) $\phi$ modulo 2$\pi$
843: and (ii) $\phi$ modulo 4$\pi$ indicating
844: the transition from chaotic to quasiperiodic attractors via SNA through
845: fractalization (GF1 and GF2).
846: (a) strange nonchaotic attractor for $h = 0.0558$ (GF1); (b) wrinkled
847: attractor for $h=0.065$; (c) strange nonchaotic
848: attractor for $h = 0.072$ (GF2); (d) torus doubled attractor for $h=0.0729$.}
849: \end{figure}
850:
851: \begin{figure}
852: \caption{Transition from torus to SNA through Gradual fractalization
853: mechanism indicating (i) the behaviour of the
854: Lyapunov exponent ($\Lambda$); (ii) the variance ($\sigma$):
855: (a) in the region GF1; (b) in the region GF2.}
856: \end{figure}
857:
858: \begin{figure}
859: \caption{
860: Transition from gradual fractalization (GF3 region) type of SNA
861: to Heagy-Hammel (HH1 region) type of SNA for $h=0.1675$: (a) SNA (GF3
862: region) for $R=0.1955$;(b) SNA (GF3
863: region) for $R=0.197$; (c) SNA (HH1 region) for $R=0.1974$;
864: (d) SNA (HH1 region) for $R=0.203$.}
865: \end{figure}
866:
867: \begin{figure}
868: \caption{Distribution of finite--time Lyapunov exponents on
869: SNAs created through (a) Type--III (dashed line) and Type--I intermittency
870: (solid line), with $t= 2048$, and (b)
871: torus collision (the Heagy--Hammel route) (dotted line) and
872: fractalization (solid line).
873: The parameter values are those of
874: Figs.~3(c), 5(b), 7(b) and 11(a) respectively.}
875: \end{figure}
876: %\end{document}
877: \newpage
878: \centerline{\epsfig {figure=nfig1a.eps,width=.99\linewidth}}
879: \centerline{FIG. 1a (A. Venkatesan et al.) }
880: \centerline{\epsfig{figure= nfig1b.eps,width=.99\linewidth}}
881: \centerline{FIG. 1b (A. Venkatesan et al.) }
882: \centerline{\epsfig {figure= nfig1c.eps,width=.99\linewidth}}
883: \centerline{FIG. 1c (A. Venkatesan et al.) }
884: \centerline{\epsfig{figure= nfig1d.eps,width=.99\linewidth}}
885: \centerline{FIG. 1d (A. Venkatesan et al.) }
886: \newpage
887: \centerline{\epsfig {figure=nfig2.eps, width=.6\linewidth}}
888:
889: \centerline{FIG. 2 (A. Venkatesan et al.) }
890: \newpage
891: \centerline{\epsfig {figure=nfig3.eps, width=.9\linewidth}}
892: \centerline{FIG. 3 (A. Venkatesan et al.) }
893: \newpage
894: \centerline{\epsfig {figure=nfig4.eps, width=.5\linewidth}}
895: \centerline{FIG. 4 (A. Venkatesan et al.) }
896: \newpage
897: \centerline{\epsfig {figure= nfig5.eps, width=.9\linewidth}}
898: \centerline{FIG. 5 (A. Venkatesan et al.) }
899: \newpage
900: \centerline{\epsfig {figure= nfig6.eps, width=.5\linewidth}}
901: \centerline{FIG. 6 (A. Venkatesan et al.) }
902: \newpage
903: \centerline{\epsfig {figure= nfig7.eps, width=.8\linewidth}}
904: \centerline{FIG. 7 (A. Venkatesan et al.) }
905: \newpage
906: \centerline{\epsfig {figure= nfig8.eps, width=.7\linewidth}}
907: \centerline{FIG. 8 (A. Venkatesan et al.) }
908: \newpage
909: \centerline{\epsfig {figure= nfig9a.eps, width=.8\linewidth}}
910: \centerline{\epsfig {figure= nfig9b.eps, width=.8\linewidth}}
911: \centerline{FIG. 9 (A. Venkatesan et al.) }
912: \newpage
913: \centerline{\epsfig {figure= nfig10.eps, width=.99\linewidth}}
914: \centerline{FIG. 10 (A. Venkatesan et al.) }
915: \newpage
916: \centerline{\epsfig {figure=nfig11.eps, width=.8\linewidth}}
917: \centerline{FIG. 11 (A. Venkatesan et al.) }
918: \newpage
919: \centerline{\epsfig {figure=nfig12.eps, width=.8\linewidth}}
920: \centerline{FIG. 12 (A. Venkatesan et al.) }
921: \end{document}
922: