nlin0110040/maw.tex
1: \documentclass{elsart}
2: \input{epsf}
3: \journal{Physica D}
4: \begin{document}
5: 
6: \begin{frontmatter}
7: 
8: \title{Nonlinear Analysis of the Eckhaus Instability:
9: Modulated Amplitude Waves and Phase Chaos with Non-zero Average Phase Gradient}
10: 
11: \author[ad1]{Lutz Brusch},
12: \author[ad2]{Alessandro Torcini} and
13: \author[ad1]{Markus B\"ar}
14: \address[ad1]{ Max-Planck-Institut f\"ur Physik
15:   komplexer Systeme, N\"othnitzer Stra{\ss}e 38, D-01187 Dresden, Germany}
16: \address[ad2]{Dipartimento di Energetica, Universit\'a di Firenze, 
17: via S. Marta 3 - I-50139 Firenze, Italy and
18: Istituto Nazionale di Ottica Applicata, L.go E. Fermi 6 - 
19: I-50125 Firenze, Italy}
20: 
21: \begin{abstract}
22: 
23: We analyze the Eckhaus instability of plane waves in the 
24: one-dimensional complex Ginzburg-Landau equation (CGLE) and 
25: describe the nonlinear effects arising 
26: in the Eckhaus unstable regime.
27: %
28: Modulated amplitude waves (MAWs) are quasi-periodic solutions of the CGLE 
29: that emerge near the Eckhaus instability of plane waves
30: and cease to exist due to saddle-node bifurcations (SN).
31: %
32: These MAWs can be characterized by their average phase gradient 
33: $\nu$ and by the spatial period $P$ of the periodic amplitude modulation. 
34: %
35: A numerical bifurcation analysis reveals the existence and stability 
36: properties of MAWs with arbitrary $\nu$ and $P$.
37: %
38: MAWs are found to be stable for large enough $\nu$ and intermediate
39: values of $P$. 
40: % 
41: For different parameter values they are unstable to splitting and
42: attractive interaction between subsequent extrema of the amplitude. 
43: %
44: Defects form from perturbed plane waves for parameter values above the 
45: SN of the corresponding MAWs. 
46: %
47: The break-down of phase chaos with average phase gradient 
48: $\nu \ne 0$ (``wound-up phase chaos'') is thus related to these SNs.
49: %
50: A lower bound for the break-down of wound-up phase chaos
51: is given by the necessary presence 
52: of SNs and an upper bound by  the absence of the splitting
53: instability of MAWs.
54: 
55: \end{abstract}
56: 
57: \begin{keyword}
58: 
59: 
60: Complex Ginzburg-Landau equation \sep
61: Coherent structures \sep
62: Modulated amplitude waves \sep
63: Phase chaos \sep
64: 
65: \PACS
66: 05.45.Jn \sep %extended chaos
67: 03.40.Kf \sep %Waves and wave propagation: general mathematical aspects 
68: 05.45.-a %Nonlinear dynamics and nonlinear dynamical systems
69: 
70: \end{keyword}
71: 
72: \end{frontmatter}
73: 
74: 
75: %_______________________________________________________________________________
76: \section{Introduction}
77: \label{intro}
78: 
79: The emergence of chaotic behaviour from ordered states in spatially
80: extended systems has been the subject of many recent experimental and
81: theoretical investigations  \cite{books,CH}.
82: %
83: Nonetheless, the mechanisms leading from stationary regimes to
84: chaotic (or spatially irregular) phases still pose many challenging
85: questions.
86: %
87: One of the most studied instabilities 
88: in extended oscillatory systems is the Eckhaus 
89: instability of plane waves \cite{eck}.
90: 
91: The occurrence of this instability has been experimentally 
92: observed in many quasi one-dimensional systems like
93: %
94: the oscillatory instability of a Rayleigh-B\'enard
95: convection pattern~\cite{janiaud},
96: hydrothermal waves \cite{h1,h2,garnier,CIAI4},
97: heated wire convection~\cite{wire},
98: sidewall convection~\cite{sidewall},
99: the Taylor-Dean system~\cite{dean} and
100: internal waves excited by the Marangoni effect~\cite{velarde}. 
101: %
102: The Eckhaus instability also plays an important role in 
103: the radial dynamics of spiral waves in the 
104: Belousov-Zhabotinsky reaction \cite{chem}.
105: 
106: \begin{figure}
107: \begin{center}
108:  \epsfxsize=0.7\hsize \mbox{\hspace*{-.06 \hsize} \epsffile{Fig/fig1.eps} }
109: \end{center}
110:  \caption[Phase diagram of the CGLE]
111:  {Phase diagram of the one-dimensional CGLE.
112:  The dotted curve indicates the Benjamin-Feir-Newell line.
113:  Plane waves undergo an Eckhaus instability at values of $c_1, c_3$ 
114:  below this curve depending on their wavenumber.
115:  Above the dashed curve the Eckhaus instability is supercritical whereas it is
116:  subcritical below \cite{janiaud}.
117:  MAWs and wound-up phase chaos with $\nu>0$ can be observed between the 
118:  dashed and the full curve. 
119:  Defect chaos can occur only above the full curve \cite{MAW1,MAW2} which 
120:  denotes the saddle-node bifurcation of MAWs with $\nu=0$ and $P\to\infty$.
121:  The vertical dot-dashed line indicates
122:  the cut of the parameter space at $c_1=3.5$ studied in this paper.
123:  }
124:  \label{c1c3}
125: \end{figure}
126: 
127: The complex Ginzburg-Landau equation (CGLE) \cite{CH,review} is the 
128:  appropriate amplitude equation to describe the slow dynamics
129: near a supercritical transition to unidirectional traveling waves. 
130: %
131: In one spatial dimension, the CGLE  reads:
132: \begin{equation} 
133:   \partial_{t} A = A + (1+ ic_1) \partial_{x}^{2} A - (1-i c_3) |A|^2 A
134:   ~, \label{cgle}
135: \end{equation}
136: where $c_1$ and $c_3$ are real coefficients and the field 
137: $A=A(x,t)= |A(x,t)| {\rm e}^{i \varphi (x,t)}$ 
138: has complex values.  
139: As exact solutions the CGLE admits plane waves of the form
140: $ A_q (x,t) = a_q \mbox{e}^{i (q x - \omega_q t)} $, where
141: $q$ indicates the wavenumber, $a_q=\sqrt{1-q^2}$ and 
142: $ \omega_q = - c_3 + q^2(c_1+c_3) $.
143: 
144: A linear stability analysis \cite{EckDiP} of these solutions can 
145: be performed by considering the perturbed solution
146: $ \tilde{A}_q (x,t) = (a_q+\delta a) \mbox{e}^{i (q x - \omega_q t)} $, where
147: $\delta a \propto \mbox{e}^{i k x} \mbox{e}^{\sigma(k) t}$.
148: The growth rates associated to the complex perturbation $\delta a$ is
149: \begin{eqnarray} 
150: \sigma(k) &=& -k^2 -2 i q c_1 k - (1-q^2) \nonumber  \\
151:          &&   \pm \sqrt{(1+c_3^2)(1-q^2)^2 -[c_1 k^2-2 i q k-c_3 (1-q^2)]^2}~.
152: \label{sigmak}
153: \end{eqnarray}
154: The plane waves become linearly unstable to 
155: long wavelength perturbations ($k\to 0$) for
156: $q = q_E \equiv \sqrt{(1-c_1 c_3)/(2 (1+c_3^2) + 1-c_1 c_3)}$.
157: This limit is called Eckhaus instability \cite{eck}.
158: Above the Benjamin-Feir-Newell (BFN) line $1-c_1 c_3 = 0$, 
159: all plane waves are unstable to homogeneous perturbations.
160: %
161: For a given $q \ge q_E$,
162: the corresponding plane wave is linearly unstable 
163: against perturbations with wavenumbers $k$ inside 
164: the interval $0 < |k| < k_c$. $k_{c}$ increases
165: for increasing values of the parameters $c_1$ and $c_3$ \cite{notevHB}.
166: 
167: As noticed in \cite{EckCI1,EckCI2}, the Eckhaus instability
168: is a convective instability. 
169: %
170: Thus, it is relevant in the systems with periodic boundary conditions
171: considered here,
172: while it would be suppressed in fixed boundary conditions , {\it
173: e. g.} zero-flux or Dirichlet boundary conditions. 
174: %
175: In the latter geometries, the absolute instability of the plane 
176: waves has to be computed. 
177: %
178: It occurs for sufficiently large $q$ or/and $c_1$,$c_3$ values 
179: inside the Eckhaus unstable range. 
180: 
181: 
182: Another interesting aspect of the Eckhaus instability is found 
183: when the nonlinearities of the CGLE are taken into account. 
184: A weakly nonlinear analysis \cite{janiaud} revealed that for
185: \begin{equation}
186: c_1^2 (1-6 c_3^2) + c_1 (2 c_3^3 + 16 c_3) - (8+c_3^2) > 0
187: \label{Ecksuper}
188: \end{equation}
189: the Eckhaus instability becomes supercritical, {\it i.e.}
190: the instabilities are saturated and the emerging quasi-periodic
191: solutions ({\it resp.} modulated amplitude waves) 
192: coexist with the unstable plane waves.
193: %
194: \begin{figure}
195: \begin{center}
196:  \epsfxsize=0.7\hsize \mbox{\hspace*{-.06 \hsize} \epsffile{Fig/fig2.eps} }
197: \end{center}
198:  \caption[Maximum conserved phase gradient $\nu_M (c_3)$]
199:  {Maximum (dots) conserved average phase gradient $\nu_M (c_3)$
200: for fixed $c_1=3.5$ 
201: obtained from numerical simulations (with system size $L=1,024-2,048$ and 
202: integration times $t\sim 10^5$) 
203:  for 50-70 different initial conditions 
204: (noise added to plane wave with wavenumber $q=\nu$). 
205:  For $\nu \le \nu_M$ no defects were present while above $\nu_M$ at least one
206:  initial condition caused defects. 
207:  The full curve denotes the Eckhaus instability of plane waves 
208:  that converges to the BFN line at $c_3=1/c_1$ for $\nu=0$.
209:  For $\nu$ above the diamonds regular states were observed after a transient 
210:  phase chaotic dynamics but below the diamonds most initial conditions led to persistent 
211:  spatio-temporal chaos \cite{torcini}.
212: L$_1$ denotes the lower bound for the occurrence of 
213: defect chaos in the thermodynamic limit as calculated in \cite{MAW1,MAW2}.
214:  }
215:  \label{FnuM}
216: \end{figure}
217: %
218: Numerical simulations 
219: \cite{janiaud,pumir,miguel,torcini,hager}
220: provided examples of such stable modulated amplitude waves (MAWs).
221: %
222: Stable MAWs have also been observed in experiments on 
223: surface-tension-driven hydrothermal waves \cite{h2} as well as 
224: on the Taylor-Dean system \cite{dean} and 
225: on internal waves excited by the Marangoni effect\cite{velarde}.
226: 
227: In addition, the CGLE exhibits two qualitatively
228: different spatiotemporal chaotic states known as phase chaos (when the 
229: modulus of the field $|A|$ is
230: bounded away from zero) and defect chaos (when the phase of $A$ displays
231: singularities where $|A|\!=\!0$) \cite{chao1,chate,saka,egolf}.
232: The subclass of MAWs with zero average phase gradient is important 
233: for understanding the transition from phase to defect chaos 
234: (see full curve in Fig.~\ref{c1c3}) \cite{MAW1,MAW2}.
235: In the phase chaos regime states with nonzero average phase gradient 
236: $\nu$ have a dynamics quite different from that at $\nu \sim 0$.
237: In particular, these states can be either chaotic or regular
238: depending on the initial conditions and on the parameters 
239: $c_1,c_3$ and $\nu$. 
240: %
241: In this paper we will focus on MAWs with $\nu \ne 0$ and on the 
242: dynamical regime associated to them, that is referred to as ``wound-up'' 
243: phase chaos \cite{miguel}.
244: It will be shown that MAWs and wound-up phase chaos exist between the 
245: dashed and the full curve in Fig.~\ref{c1c3}.
246: 
247: %ORGANIZATION___________________________________________________________________
248: 
249: In Section \ref{maw} 
250: the analysis of MAWs introduced in \cite{MAW1,MAW2} 
251: is extended to arbitrary values of the 
252: average phase gradients of the field.
253: The two parameter family of MAW solutions is parametrized by
254: the spatial period $P$ of the modulation and by 
255: the average phase gradient 
256: \begin{equation}
257: \nu := \frac{1}{P} \int_0^P dx \varphi_x ~.
258: \end{equation}
259: For plane wave solutions,  $\nu$ equals the wavenumber $q$.
260: %
261: In analogy, the phase gradient $\varphi_x$ is often called 
262: ``local wavenumber''.
263: %
264: A linear stability analysis will show that MAWs with $\nu \ne 0$ 
265: can be  stable even in infinitely  large systems.
266: %
267: In contrast,  MAWs with  $\nu=0$ are always unstable in
268: systems of large length $L \gg P$.
269: %
270: For systems with periodic boundary conditions the average
271: phase gradient of the whole system can only be changed, if a space-time defect
272: occurs : $|A(x,t)|$ drops to zero and $\varphi_x$ locally diverges at a
273: defect.
274: %
275: Persistent phase chaos with conserved $\nu\le\nu_M\ne0$ has been observed in
276: numerical simulations of the CGLE (\ref{cgle}) \cite{miguel,torcini}. 
277: %
278: The maximum conserved average phase gradient $\nu_M$ decreases as function of
279: the coefficients $c_1, c_3$ \cite{miguel,torcini} and vanishes at
280: the apparent transition from phase to defect chaos.
281: %
282: $\nu_M$ was therefore suggested \cite{torcini} as an order parameter for this
283: transition.
284: %
285: We extended the numerical determination of $\nu_M$ towards smaller $c_3$ and
286: report the corresponding data in Fig.~\ref{FnuM}.
287: %
288: 
289: In Section~\ref{nuM}, a nonlinear analysis of the Eckhaus instability
290: allows estimates for which parameter values defects occur.
291: %
292: Lower and upper bounds for the limit $\nu_M$ of wound-up phase chaos 
293: are derived from the existence and stability properties of the MAWs. 
294: %
295: For increasing $\nu$ the degree of chaoticity associated with the 
296: wound-up phase chaos decreases \cite{torcini}.
297: For large enough $\nu$ the dynamics can even become 
298: regular and stable quasi-periodic MAWs appear.
299: The diamonds in Fig.~\ref{FnuM} indicate this stability limit for
300: numerical simulations with fixed $c_1=3.5$.
301: The analysis in Subsection~\ref{nu1instab} will clarify this 
302: observation.
303: 
304: The large 
305: number of parameters ($c_1, c_3, \nu, P$) calls for restrictions.
306: %
307: We limit our analysis to fixed $c_1=3.5$ since most previous numerical work 
308: has been done at this value \cite{torcini,egolf}.
309: %
310: The results will be presented as projections of the 
311: $P$ direction onto the ($c_3 , \nu$) plane as well as in cuts through
312: the parameter space spanned by $c_3, \nu$ and $P$.
313: %
314: Additional investigations of the existence domains of MAWs revealed 
315: qualitatively similar results for fixed $c_1=0.4, 1.2, 2.1$ 
316: and $5$ and variable $c_3$ 
317: as well as for fixed $c_3=0.83$ and variable $c_1$. 
318: %
319: Two of these choices were studied by numerical simulations  in \cite{miguel}.
320: %
321: A similarity transformation maps coherent structures onto each other 
322: along curves $(c_1+c_3)/(1-c_1 c_3)=const$ in coefficient space
323: \cite{review}.
324: %
325: The parameters $\nu, P, \omega, v$ are transformed accordingly.
326: %
327: One can thereby extend the results presented here to other values of the 
328: coefficient $c_1$.
329: %
330: Section \ref{nu1exp} discusses possible observations of MAWs in 
331: experimental systems.
332: %
333: Finally, Section \ref{nu1disc} summarizes the main results.
334: 
335: %_______________________________________________________________________________
336: \section{Existence and stability of  modulated amplitude waves}
337: \label{maw}
338: 
339: \subsection{Coherent structure approach}
340: 
341: In analogy with the linear analysis of the Eckhaus instability 
342: of plane waves we make the following ansatz for saturated modulations
343: \begin{equation}
344: A(x,t)= a(z) \mbox{e}^{\displaystyle i \tilde{\phi} (z)} 
345: \mbox{e}^{\displaystyle i (q x - \tilde{\omega} t)}
346: \label{nu1ansatz0}
347: \end{equation}
348: and rewrite it as 
349: \begin{equation}
350: A(x,t)= a(z) \mbox{e}^{\displaystyle i \phi (z)} 
351: \mbox{e}^{\displaystyle i \omega t}~,
352: \label{nu1ansatz}
353: \end{equation}
354: where $a$ and $\phi$ are real-valued functions of $z\!:=\!x - vt$ and 
355: $\phi (z)=\tilde{\phi} (z)+q z, \omega=q v-\tilde{\omega}$.
356: Here $a(z)$ and $\phi(z)$ represent coherent structures \cite{saar}. 
357: Coherent structures have been studied extensively
358: \cite{torcini,MAW1,MAW2,saar,mvh} and play an important role in various regimes
359: of the CGLE \cite{janiaud,pumir,miguel,torcini,MAW1,MAW2,saar,mvh,mm}.
360: 
361: Substitution of ansatz~(\ref{nu1ansatz}) into the CGLE~(\ref{cgle}) 
362: yields the set of three coupled nonlinear ordinary differential equations
363: (ODEs) 
364: \begin{eqnarray}
365: a_z &=& b \nonumber \\ 
366: b_z &=& \psi^2 a- \gamma^{-1} [(1+c_1 \omega)a+v(b+c_1\psi a)-(1-c_1 c_3)a^3]
367: \label{ode} \\
368: \psi_z &=& -2b \psi/a+ \gamma^{-1} [c_1-\omega+v(c_1b/a-\psi)-(c_1+c_3) a^2]
369: \nonumber
370: \end{eqnarray}
371: where $b\!:=\!a_z, \psi\!:=\!\phi_z$ and $\gamma\!:=\!1+c_1^2$
372: \cite{note1}.
373: %
374: The continuation software {\sc AUTO97} \cite{auto97} is used to compute the
375: periodic orbits of the ODEs (\ref{ode}) that correspond to spatially periodic
376: functions $a(z), \phi(z)$. 
377: %
378: In order to choose a unique solution from the
379: continuous two-parameter family of periodic orbits we set the system size $L$ 
380: equal to the period $P$ of the periodic
381: orbit and fix its average phase gradient by $\nu = \frac{1}{L} \int_0^L \psi dz $~.
382: 
383: \begin{figure}
384: \begin{center}
385: \epsfxsize=0.9\hsize \mbox{\hspace*{-.06 \hsize} \epsffile{Fig/fig3.eps} }
386: \end{center}
387:  \caption[Bifurcation diagram for $\nu = 0.25, P=25.13, c_1=3.5$]
388:  {(a) Example of a bifurcation diagram showing the maximum of the modulus 
389:  for MAWs with $\nu = 0.25, c_1=3.5, P=2\pi/\nu=25.13$. 
390:  The plane wave that is stable (unstable) against modes of wavelength $P$ is
391:  represented by the thin full (dashed) line. The stable lower branch (unstable
392:  upper branch) of MAWs is denoted by the thick full (dashed) curve.
393:  HB denotes the Hopf bifurcation (square) of the plane wave
394:  solution whereas SN stands for the saddle-node bifurcation (triangle) that limits
395:  the existence of MAWs.
396: %
397:  Spatial portraits of (b) the modulus and (c) the phase gradient 
398:  are shown for a choice of
399:  solutions. The dotted line represents the plane wave whereas thin full and
400:  thick full curves give MAWs at locations labelled by A and SN in
401:  (a).
402:  The dashed curve denotes the saddle-type upper branch solution at C in (a).
403:  (d) shows the oscillation frequency $\tilde{\omega}=q v-\omega$ and
404:  (e) the velocity $v$ versus $c_3$.
405:  In (e) the dotted line denotes the group velocity
406:  ($k=0$) and the line below gives the velocity $v_c$ corresponding
407:  to the mode with finite wavelength $P$ \cite{notevHB}.
408:  }
409:  \label{q25L1n1bif}
410: \end{figure}
411: 
412: The continuation procedure starts from a fixed point $(a,b,\psi) =
413: (\sqrt{1-q^2},0,q)$ that corresponds to a plane wave solution.
414: % 
415: Varying $c_3$, a Hopf (HB) bifurcation (filled square in Fig. 
416: \ref{q25L1n1bif}) is detected in the ODEs 
417: where the mode with the smallest possible wave number 
418: $k_{HB}=2\pi/P $ destabilizes the plane wave. 
419: %
420: Continuing the resulting branch of MAWs the free parameters 
421: $\omega$ and $v$ are 
422: adjusted by the continuation algorithm. 
423: %
424: The continuation follows a unique branch of MAWs with $\nu=q$ and $P=L$.
425: %
426: Fig.~\ref{q25L1n1bif} shows examples of resulting bifurcation diagrams.
427: 
428: \begin{figure}
429: \begin{center}
430:  \epsfxsize=0.7\hsize \mbox{\hspace*{-.06 \hsize} \epsffile{Fig/fig4.eps} }
431: \end{center}
432:  \caption[Bifurcation diagram for subcritical Eckhaus instability]
433:  {Bifurcation diagram with exclusively upper branch MAWs for 
434:  $\nu=0.68, P=2\pi/\nu\approx9.24$.
435:  The solid line indicates stable and the dashed curves unstable
436:  solutions. The Hopf bifurcations are both subcritical.
437:  }
438:  \label{q68L1n1bif}
439: \end{figure}
440: 
441: %_______________________________________________________________________________
442: \subsection{Existence limits of MAWs}
443: \label{nu1bif}
444: 
445: Upon increasing of $c_3$ amplitude modulations grow and develop a 
446: localized depression $|A|_{min}$ where $\phi_x$ has a maximum 
447: (see Fig.~\ref{q25L1n1bif}b,c).
448: % 
449: As for $\nu=0$, these MAWs are called the {\it lower} branch 
450: in contrast to the coexisting {\it upper} branch MAWs.
451: The upper branch MAWs are always unstable,
452: while the MAWs of the lower branch can be stable in
453: appropriate parameter regions.
454: %
455: Examples of these lower branch MAWs have been obtained by 
456: numerical simulations earlier 
457: \cite{janiaud,pumir,miguel,torcini}. 
458: %
459: They have been analyzed in detail in \cite{hager}.
460: %
461: Numerical simulations can neither uncover unstable upper branch MAWs
462: nor elucidate the limits of existence of MAWs.
463: %
464: The bifurcation analysis presented here reveals that 
465: upper and lower branch meet and terminate in a saddle-node 
466: (SN) bifurcation (filled triangle in Fig.~\ref{q25L1n1bif} a,d,e ). 
467: %
468: Due to the SN bifurcation the upper branch MAWs always have at least one 
469: unstable eigenmode, see also  \cite{mvh,mm}.
470: %
471: The upper branch continues to negative $c_3$ and there connects to
472: another instability of the plane wave with identical $\nu$ and $ P$.
473: %
474: In the following we will concentrate on the lower branch MAWs.
475: 
476: For large $\nu$ and small $P$, the Hopf bifurcation is no longer supercritical
477: and an unstable branch emerges directly from the plane wave. 
478: %
479: This is in agreement with analytical predictions \cite{janiaud}.
480: %
481: Fig.~\ref{q68L1n1bif} shows an example which also includes the second HB
482: at negative $c_3$.
483: %
484: For $\nu=0$ the MAWs emerge stationary \cite{MAW1,MAW2} and acquire $v\ne0$ 
485: above a subsequent drift pitchfork bifurcation \cite{DwightDP}. 
486: %
487: In the present case $\nu\ne0$ the plane wave 
488: already breaks the reflection symmetry, the initial MAW
489: has a nonzero velocity and the drift pitchfork (DP) bifurcation 
490: (filled diamond) is unfolded. 
491: %
492: See Fig.~\ref{dpunfold} for an example at fixed $c_3=2$.
493: %
494: The branch emerging at the HB in Fig.~\ref{dpunfold}b represents the
495: MAWs as discussed above.
496: %
497: The second branch in Fig.~\ref{dpunfold}b emerges at the period doubling (PD) 
498: bifurcation (open square) of MAWs with half
499: the period. 
500: %
501: It always has unstable eigenmodes that drive the dynamics away 
502: from it to the coexisting MAWs of shorter period. 
503: %
504: Therefore this (upper) branch plays no
505: essential role and is not treated further.
506: 
507: \begin{figure}
508: \begin{center}
509:  \epsfxsize=0.7\hsize \mbox{\hspace*{-.06 \hsize} \epsffile{Fig/fig5.eps} }
510: \end{center}
511:  \caption[Unfolding of drift pitchfork bifurcation]
512:  {Bifurcation diagrams showing the velocity $v$ versus $c_1$. 
513:  (a) Branches with $v\ne 0$ emerge at the drift pitchfork bifurcation (DP) for
514:  $c_3=2, P=25, \nu=0$.
515:  (b) The bifurcation is unfolded for $\nu\ne0$, here $c_3=2, P=25, \nu=0.01$.
516:  An equivalent pair of branches exists for $\nu\to-\nu$ and $v\to-v$.
517:  }\label{dpunfold}
518: \end{figure}
519: 
520: %_______________________________________________________________________________
521: \subsubsection{Infinite system size}
522: 
523: We have analyzed the existence of lower branch MAWs in the entire parameter space
524: ($c_3, \nu, P$) at fixed $c_1=3.5$. The system size is assumed infinitely
525: large in order to allow for arbitrary periods $P$ of MAWs.
526: %
527: Fig.~\ref{cqexist} shows examples of existence
528: domains for $P=15, P=30$ and $P\to\infty$. 
529: %
530: We find that both HB and SN shift to
531: larger $c_3$ as the period $P$ is decreased. 
532: %
533: The same behavior has already been observed 
534: in the special  case $\nu=0$ \cite{MAW1,MAW2}. 
535: %
536: The dotted curve in Fig.~\ref{cqexist} indicates the
537: ``envelope'' of all SN bifurcations for MAWs of any
538: period and therefore is the upper boundary for the
539: existence domain of the MAWs.
540: 
541: 
542: %_______________________________________________________________________________
543: \subsubsection{Medium system size}
544: 
545: Experimental setups and numerical simulations are restricted to finite system
546: size $L$.
547: Often periodic boundary conditions (corresponding to an annular geometry) 
548: are used in order to study bulk effects of extended systems and to 
549: minimize boundary effects.
550: %
551: The periodic boundary conditions also restrict possible modes of
552: perturbations.
553: As described by Eq.~(\ref{sigmak}) the instability
554: threshold of plane wave solutions depends on the wavenumber $k$ of the
555: perturbation. Since in the studied range of coefficients the Eckhaus instability
556: is a long-wavelength instability the plane waves will be stabilized 
557: in small systems. 
558: %
559: The instability threshold is shifted to larger values
560: of the coefficients $c_1, c_3$ and can be computed from
561: Eq.~(\ref{sigmak}) setting $k=2\pi/L$.
562: 
563: \begin{figure}
564: \begin{center}
565:  \epsfxsize=0.7\hsize \mbox{\hspace*{-.06 \hsize} \epsffile{Fig/fig6.eps} }
566: \end{center}
567:  \caption[Existence domains of MAWs in $L\to\infty$]
568:  {Existence domains of MAWs with period $P$ projected onto the 
569:  ($c_3, \nu$) parameter plane.
570:  Thin curves denote the Eckhaus instability and HB which occur supercritical 
571:  (full curve) or subcritical (dashed) depending on $\nu$ and $P$.
572:  The thick curves give the SN for selected periods $P$.
573:  Three examples of existence domains for
574:  $P=15$ (right, dark shaded domain), 
575:  $P=30$ (middle, empty), 
576:  $P\to\infty$ (left, light shaded) are shown.
577:  The superposition of all existence domains is bounded by the dotted curve.
578:  }
579:  \label{cqexist}
580: \end{figure}
581: 
582: Clearly the selection of perturbations by periodic boundary conditions also
583: restricts possible MAWs. 
584: %
585: Their average phase gradient $\nu$ and the
586: period $P$ have to be consistent with the system size and this
587: renders the two-parameter family of MAWs discrete. 
588: %
589: It is thus convenient to parametrize MAWs by the average phase gradient
590: $\nu$ and the ratio $n$ of wavelength
591: \begin{equation}
592: n := \frac{P}{2\pi/\nu} ~.
593: \label{n}
594: \end{equation}
595: %
596: The ratio $n$ takes values of integer fractions where the nominator counts the
597: number of underlying wave length and the denominator the number of 
598: humps of the modulation. Hence this quantity is easily accessible in experiments.
599: %
600: The existence domains of MAWs with respective $n$ are 
601: presented in the ($c_3, \nu$) parameter plane in 
602: Fig.~\ref{mawn}.
603: 
604: \begin{figure}
605: \begin{center}
606:  \epsfxsize=0.9\hsize \mbox{\hspace*{-.06 \hsize} \epsffile{Fig/fig7.eps} }
607: \end{center}
608:  \caption[Existence domains of MAWs]{
609:  Existence domains of lower branch MAWs are denoted by shaded areas for
610:  (a) $n=20$, (b) $n=4$, (c) $n=2$, (d) $n=1$, (e) $n=1/2$ and (f) $n=1/4$. 
611:  They are limited by HB (solid curve) at small $c_3$ and by the SN 
612:  (dashed curve) at large $c_3$.
613:  Spatial profiles of coexisting MAWs at $\nu=0.05, c_3=0.5$ are shown for 
614:  (g) $n=1$, (h) $n=1/2$ and (i) $n=1/4$, corresponding to dots in (d-f).
615:  }
616:  \label{mawn}
617: \end{figure}
618: 
619: %_______________________________________________________________________________
620: \subsubsection{Small system size}
621: 
622: Here we focus on the extreme case. 
623: %
624: The shortest possible system 
625: with periodic boundary conditions only contains one wavelength of the 
626: plane wave, consequently its length $L$ is given by  $L=2\pi/\nu$. 
627: %
628: In \cite{torcini} the quantity $\nu_U$ was determined 
629: in analogy to $\nu_M$ for large systems. 
630: %
631: $\nu_U$ denotes the largest $\nu$ for which none of the random initial
632: conditions (different realizations of noise added to a plane wave) 
633: produced a defect. 
634: %
635: In the following these data (symbols in Fig.~\ref{L1}) are compared to the 
636: existence domains of MAWs.
637: 
638: \begin{figure}
639: \begin{center}
640:  \epsfxsize=0.9\hsize \mbox{\hspace*{-.06 \hsize} \epsffile{Fig/fig8.eps} }
641: \end{center}
642:  \caption[Existence domains of MAWs in $L=2\pi/\nu$]
643:  {For short system size $L=2\pi/\nu$ plane waves are stable for parameter choices
644:  inside the light shaded area.
645:  MAWs with a single hump ($P=L$) exist inside the dark shaded area bounded by 
646:  the supercritical HB to the left and the SN to the right.
647:  Thin curves give the limits of MAWs with two humps $P=L/2$ (dashed)
648:  and three humps $P=L/3$ (dotted). See the legend for the different cases.
649:  Symbols denote maximal $\nu=\nu_U$ that did not create defects but resulted in stable
650:  asymptotic states in simulations of the short system.
651:  Plane waves and single MAWs (squares), multi-hump MAWs with 2 humps 
652:  (triangles) and 3 humps (stars) have been observed at $\nu=\nu_U$.
653:  Data $\nu_U$ are taken from \cite{torcini}.
654:  } 
655:  \label{L1}
656: \end{figure}
657: 
658: Within the light shaded area in Fig.~\ref{L1} plane wave solutions with
659: wavenumber $\nu$ are stable in the short system. 
660: %
661: The stability area extends over the phase chaos and into the defect chaos 
662: region. 
663: %
664: We stress this result because from the 
665: experimental observation of stable plane waves one cannot necessarily 
666: infer that the dynamics of the system may be reproduced by the CGLE
667: with coefficients $c_1, c_3$ in the Benjamin-Feir stable region.
668: %
669: The dashed curve denotes a subcritical instability.
670: %
671: Only unstable upper branch MAWs exist to the left of this curve. 
672: %
673: For smaller $\nu$ the instability again
674: turns supercritical and stable lower branch MAWs exist
675: inside the dark shaded region. 
676: %
677: The thick full curve gives the SN bifurcation 
678: for MAWs with $P=L$. 
679: %
680: The thin curves show the respective limits of MAWs with shorter
681: period. 
682: %
683: Defects are expected beyond the SN \cite{MAW1,MAW2} and the subcritical
684: instability which well reproduces the data from numerical simulations
685: \cite{torcini} except at small $\nu$.
686: %
687: Simulations with 
688: $\nu \le \nu_U$ resulted in modulations with a single hump
689: (squares) or with two (triangles) or three (stars) humps of different size
690: \cite{torcini}.
691: %
692: The latter two are observed above the SN of MAWs with $P=L$. 
693: %
694: Here the initial conditions select
695: MAWs with shorter period which only coexist at small $\nu$. 
696: %
697: The SN with $P=L$ nevertheless gives a lower bound for the formation
698: of defects.  
699: 
700: %_______________________________________________________________________________
701: \subsection{Instabilities of MAWs}
702: \label{nu1instab}
703: 
704: In contrast to the case of MAWs with zero average phase gradients 
705: some MAWs with non-zero $\nu$ are stable even in very large systems. 
706: %
707: A linear stability analysis of MAWs as in \cite{MAW2} yields the spectrum of
708: eigenvalues as shown in Fig.~\ref{L_infty} for a typical example.
709: From Fig.~\ref{L_infty} we conclude that for this example the entire spectrum in 
710: the infinite system will be confined to the left half-plane.
711: %
712: Thus MAWs should be found in experiments, that can be well described 
713: by the CGLE for appropriate control parameters. 
714: %
715: In this section we present a detailed study of the stability
716: properties of MAWs. 
717: MAWs with a single hump per period $P$ will be called ``single MAWs''.
718: Their existence domains were studied in the previous section.
719: %
720: However, the effective interaction between adjacent periods of a single MAW 
721: can be repulsive or attractive (see Figs. \ref{nu1q25} and \ref{nu1c5}). 
722: %
723: Period doubling (PD) bifurcations (open squares) occur at the transitions
724: from repulsive to attractive interaction \cite{michal}.
725: % 
726: There, new branches of MAWs with longer period but many humps per period
727: emerge from the primary branch of single MAWs.
728: We will call these solutions ``multi-hump MAWs''. 
729: %
730: In their profile some humps gain more space and others are compressed in an
731: alternating fashion.
732: %
733: The new branches extend to larger $c_3$ than the corresponding single MAWs.
734: %
735: Fig.~\ref{L4pd} shows how these branches arrange in a system
736: with 4 interacting humps ($L=4*P$).
737: %
738: As long as the PD bifurcations are supercritical,
739: the multi-hump MAWs are stable. 
740: %
741: They represent the
742: saturated solution for attractive interaction between subsequent modulations.
743: For large systems a whole sequence of period doubling bifurcations will lead 
744: to multi-hump MAWs with an overall period equal to the system size. 
745: %
746: Hence they appear as an 
747: erratic spatial sequence of humps and depressions. 
748: %
749: This spatial sequence propagates in a coherent fashion.
750: %
751: We named these patterns multi-hump MAWs to emphasize the 
752: connection among the coherent structures.
753: 
754: \begin{figure}
755: \begin{center}
756:  \epsfxsize=0.7\hsize \mbox{\hspace*{-.06 \hsize} \epsffile{Fig/fig9.eps} }
757: \end{center}
758:  \caption[Stable MAW in $L\to\infty$]
759:  {
760:  (a) Spectrum of eigenvalues $\lambda$ of a stable lower branch MAW. 
761:  Parameters are $c_3=0.4, \nu=0.184, P=2\pi/\nu$.
762:  (b) Blow-up of the leading part of the spectrum.
763:  The dots correspond to system size $L=100*P=3415$ and have been calculated
764:  using the Bloch method \cite{MAW2,bloch}.
765:  } 
766:  \label{L_infty}
767: \end{figure}
768: 
769: Examples of these stable aperiodic patterns 
770: were already observed in numerical simulations. 
771: %
772: R. Montagne {\it et al.} \cite{miguel} denote this behavior as
773: ``frozen phase turbulence'' while
774: A. Torcini {\it et al.} \cite{torcini} use the term ``solutions of 
775: type $\beta$''. 
776: 
777: The observed coexistence of a large number of stable multi-hump MAWs results 
778: in a strong dependence of the final state on the initial 
779: conditions of the numerical simulation. 
780: %
781: Although each regular final configuration must be consistent with
782: a particular single or multi-hump MAW it is difficult to predict how 
783: the selected final patterns depend on the initial conditions. 
784: 
785: \begin{figure}
786: \begin{center}
787:  \epsfxsize=0.9\hsize \mbox{\hspace*{-.06 \hsize} \epsffile{Fig/fig10.eps} }
788: \end{center}
789:  \caption[Instability : multi-hump MAWs]
790:  {(a) Bifurcation diagram for $\nu=0.184, P=\pi/\nu$ and $L=4*P=68.3$ hence 4
791:  pulse-like modulations interact. 
792:  The maximum of the amplitude gradient is plotted since the
793:  interaction causes pulse shifts and the amplitude of single humps changes
794:  little. Thicker lines correspond to smaller overall period of the
795:  modulation. 
796:  Typical solutions are shown in (b)-(d) as indicated by arrows in (a).
797:  } 
798:  \label{L4pd}
799: \end{figure}
800: 
801: MAWs with large period $P$ undergo a ``splitting'' instability as in the limit
802: case $\nu=0$ \cite{MAW2}. Roughly, the spatial profiles of these MAWs consist of
803: a localized hump and a plane wave part. Since the extended plane wave is
804: linearly unstable the splitting instability is reminiscent of the Eckhaus
805: instability.
806: %
807: It creates more humps on the plateau of the unstable
808: MAW and reduces the period $P$ of MAWs on average.
809: 
810: Figs.~\ref{nu1q25} and \ref{nu1c5} represent cuts through the parameter space at
811: fixed $\nu=0.25$ and $c_3=0.5$, respectively. They show the typical arrangement
812: of stable and unstable parameter regions for single MAWs. 
813: Other examined cuts for
814: $c_3=0.1, 0.2, 0.3, 0.4, 0.6$ and $0.7$ qualitatively show the same order.
815: 
816: The cut through parameter space $c_3, \nu, P$ at $\nu=0.25$ is shown in
817: Fig.~\ref{nu1q25}. The HB (dashed curve in the figure) approaches the 
818: Eckhaus instability for 
819: $P\to\infty$ as the lower bound of the existence domain. 
820: %
821: From above the domain is limited by the SN (solid curve). 
822: %
823: For small $P$ (large $c_3$) the HB is subcritical and 
824: only unstable upper branch MAWs exist to the left.
825: %
826: In the infinite system MAWs are found to be linearly stable for
827: a broad range of parameters (dark shaded area). 
828: %
829: At low $P$ the interaction instability occurs (white area)
830: whereas at large $P$ the long plateau of the MAW is unstable to splitting (light
831: shaded area).
832: %
833: For $c_3<0.45$ most random initial conditions will evolve to stable MAWs. 
834: 
835: \begin{figure}
836: \begin{center}
837:  \epsfxsize=0.7\hsize \mbox{\hspace*{-.06 \hsize} \epsffile{Fig/fig11.eps} }
838: \end{center}
839:  \caption[Stability domains of MAWs for $\nu=0.25$]
840:  {Stability domain (dark area) of single MAWs for $\nu=0.25, L\to\infty$.
841:  MAWs exist between supercritical HB (dashed curve) and SN (full curve). 
842:  The tick marks at the right frame give the asymptotic values for $P\to\infty$. 
843:  The dot-dashed curve denotes the subcritical HB.
844:  MAWs are unstable to splitting within the light shaded domain at large $P$. 
845:  Within the white domain at small $P$ single MAWs are unstable to 
846:  interaction.
847:  } 
848:  \label{nu1q25}
849: \end{figure}
850: 
851: A cut perpendicular to the previous one is shown in Fig.~\ref{nu1c5}.
852: %
853: Curves and shadings have the same meaning as discussed above. 
854: %
855: Starting from random initial conditions at $\nu>0.1$ a transient
856: may again lead to a stable MAW with local 
857: periods $P$ inside the stable windows. 
858: %
859: At lower $\nu<0.1$ the probability
860: of approaching a stable configuration decreases 
861: since only a third of the previous stable $P$ intervals remains. 
862: %
863: Below $\nu=0.02$ no stable state can be prepared at all.
864: %
865: Instead one observes wound-up phase chaos 
866: with an associated maximal Lyapunov exponent that
867: increases for decreasing $\nu$.
868: %
869: As in the limit case $\nu=0$ (phase chaos), the dynamics is driven by the
870: attractive interaction and annihilation of 
871: localized modulations in competition with the splitting instability 
872: that produces new peaks in the modulations.
873: %
874: In particular for decreasing $\nu$, 
875: the splitting instability extends to shorter periods $P$
876: and significantly overlaps with the interaction
877: instability.
878: %
879: With the above arguments many results obtained by numerical simulations 
880: of the CGLE can be well interpreted. In particular it has been observed 
881: in \cite{torcini} for the same choice of parameters ($c_1=3.5$ and $c_3=0.5$) 
882: that the maximal Lyapunov exponent (averaged over many different initial
883: conditions) is positive for $\nu=0$ and decreases monotonously 
884: towards zero for increasing $\nu$. Above $\nu=0.09$ no chaotic solutions
885: have been observed.
886: 
887: \begin{figure}
888: \begin{center}
889:  \epsfxsize=0.7\hsize \mbox{\hspace*{-.06 \hsize} \epsffile{Fig/fig12.eps} }
890: \end{center}
891:  \caption[Stability domains of MAWs for $c_3=0.5$]
892:  {Stability domains (dark areas) of single MAWs for $c_3=0.5, L\to\infty$.
893:  Curves and instability domains (splitting=light shaded, interaction=white) have
894:  the same meaning as in Fig.~\ref{nu1q25}.
895:  } 
896:  \label{nu1c5}
897: \end{figure}
898: 
899: %_______________________________________________________________________________
900: \section{Defect formation in wound-up phase chaos}
901: \label{nuM}
902: 
903: In this section the formation of defects and the resulting change of the 
904: average phase gradient are studied. 
905: For $\nu\ne0$, the scenario of defect formation past the
906: saddle-node bifurcation of the relevant MAW is analogous to the
907: previously studied case $\nu = 0$  \cite{MAW1,MAW2}, see Subsection
908: \ref{nu1sn}.
909: %
910: In particular, the dependence of the final selected average phase gradient
911: $\nu_f$ on the initial value $\nu_i$ found in numerical simulations 
912: \cite{miguel} can be interpreted.
913: %
914: Subsection \ref{nuMok} is then devoted to the limit $\nu_M$ of wound-up phase
915: chaos. 
916: %
917: For a certain range of parameters the limit $\nu_M$ is reproduced by
918: means of the stability properties of MAWs at the saddle-node
919: bifurcation. 
920: %
921: These arguments work well for defect creation with $\nu > 0.1$. 
922: %
923: At smaller values of $\nu$ various instabilities (splitting and
924: interaction) of MAWs compete and a general statement is more
925: difficult, compare also \cite{MAW2}.
926: 
927: %_______________________________________________________________________________
928: \subsection{Beyond the saddle-node bifurcation}
929: \label{nu1sn}
930: 
931: The role of the SN bifurcation for the dynamics has been studied in
932: \cite{MAW1,MAW2} for the limit case $\nu=0$. For $\nu\ne0$ 
933: we find similar behavior. 
934: %
935: Fig.~\ref{nu1snsim} gives examples
936: for $\nu=0.25, P=2\pi/\nu$.
937: %
938: Perturbations of a plane wave lead to defects only above the SN, 
939: whereas below the SN such perturbations have to be very large to overcome the 
940: saddle-type upper branch MAW. 
941: %
942: 
943: There are no SNs for parameters below the SN corresponding to $P\to\infty$.
944: Thus, starting from random initial conditions defects may only form at 
945: parameters above the SN of $P\to\infty$. 
946: The SN of $P\to\infty$ represents a lower bound for defect formation.
947: 
948: For large systems the formation of defects depends on the local period of
949: initial perturbations in a similar way as for $\nu=0$ \cite{MAW1,MAW2}. 
950: The peak to peak distances of $\varphi_x(x,t)$ are used to determine local
951: periods $p$.
952: %%
953: In that context, the following explanation of defect 
954: formation has been proposed. Defects are observed in the
955: phase turbulent regime whenever local structures, similar to MAWs,
956: with spatial periods $p$ larger then $P_{SN}$ occur in the system.
957: Where $P_{SN}$ denotes the period of the MAWs at the SN
958: (that coincides with the maximal MAW-period)
959: for the considered choice of parameters.
960: %
961: For larger values of $\nu$ or $c_3$ the SN occurs for smaller $P_{SN}$ 
962: as shown in Figs.~\ref{nu1q25} and \ref{nu1c5}. 
963: %
964: Therefore, at larger $\nu, c_3$
965: local periods $p$ beyond the SN and
966: subsequent defect formation are more probable.
967: %
968: In contrast to the case $\nu=0$, there is only a short transient of 
969: phase chaos in the simulations with nonzero initial $\nu_i > \nu_M$.
970: %
971: The distribution of local periods $p$ of the perturbations is given by the
972: realization of the noise in the initial condition.
973: %
974: For local periods slightly above the relevant SN 
975: (as in Fig.~\ref{nu1snsim}e), 
976: the perturbation first increases to a modulation similar to MAWs and 
977: appears almost
978: saturated for some transient time until finally a defect appears. 
979: %
980: This transient of defect formation becomes shorter as the distance to
981: the SN grows.
982: %
983: If initial conditions are prepared with $\nu_i \gg \nu_M$, then some local
984: periods will be far beyond the corresponding SNs. 
985: %
986: The transients of defect formation are shorter in this case.
987: 
988: \begin{figure}
989: \begin{center}
990:  \epsfxsize=0.9\hsize \mbox{\hspace*{-.06 \hsize} \epsffile{Fig/fig13.eps} }
991: \end{center}
992:  \caption[Defect formation : dynamics along SN manifold]
993:  {(a) Bifurcation diagram as in Fig.~\ref{q25L1n1bif} showing the minimum modulus of the
994:  MAWs.
995:  (b)-(e) Numerical simulations illustrate the dynamics near the SN 
996:  corresponding to the arrows in (a).
997:  (b) Plane wave perturbed at one point and
998:  (c) unstable saddle-type MAW plus noise converge to the stable MAW.
999:  (d) Unstable saddle-type MAW plus a different realization of noise 
1000:  evolves to a defect that changes $\nu$ to 0.
1001:  (e) As (b) but beyond the SN which makes defect formation
1002:  possible for arbitrarily small perturbations of the plane wave.
1003:  Note the long living transient of a non-coherent modulation.
1004:  (b)-(d) are at $c_3=0.5$ below the SN and
1005:  (e) belongs to $c_3=0.55$ above the SN for $\nu=0.25,
1006:  P=L=2\pi/\nu=25.13$.
1007:  }
1008:  \label{nu1snsim}
1009: \end{figure}
1010: 
1011: %%
1012: These two observations suggest an interpretation of the curve $\nu(t)$
1013: representing the temporal evolution of $\nu$ during transients 
1014: with $\nu_i > \nu_M$ \cite{miguel}.
1015: The larger the initial $\nu_i$ is chosen the smaller is the final value $\nu_f$ .
1016: The time scales of competing processes have to be considered.
1017: Local defect formation will 
1018: not instantaneously effect distant spatial locations 
1019: along the system.
1020: Instead the local change of the
1021: average phase gradient $\nu$ via a defect will take a transient time to relax 
1022: over the entire system.
1023: For $\nu_i$ much larger than $\nu_M$, defect formation happens on a short 
1024: time scale and
1025: independently leads to defects at many different spatial locations
1026: before the relaxation of the decreased average phase gradient $\nu_f
1027: \ll \nu_M$ can stop defect formation.
1028: For $\nu_i$ just above $\nu_M$ defects form slowly and the reduced $\nu_f$ can relax the 
1029: phase gradient at distant locations before other defects occur.
1030: 
1031: Let us now verify if the mechanisms proposed to explain
1032: defect formation in the phase chaos regime for solutions
1033: with $\nu \approx 0$ still hold for the wound-up phase chaos regime.
1034: For $\nu \approx 0$, defects form if and only if the period $p$
1035: of local structures is larger than $P_{SN}$, where $P_{SN}$ is the 
1036: period for which a SN occurs at the chosen coefficients $c_1$ and $c_3$
1037: \cite{MAW1,MAW2}.
1038: We have considered two cuts in the parameter space at
1039: $c_3=0.5$ and $c_3=0.65$ and investigated the
1040: distributions of periods $p$ for solutions with average phase
1041: gradient $\nu \sim \nu_M$. In particular, 50 realizations
1042: of a system of length $L=512$ initialized as a plane wave
1043: with wavenumber $\nu$ plus noise in the amplitude and in the
1044: phase have been considered. Then these different initial
1045: conditions have been followed for an integration time
1046: $t=500,000-1,000,000$. The last part of the run has
1047: been examined in order to extract the length of the coherent
1048: structure with the maximal period $p_{max}$
1049: occurring during the evolution.
1050: 
1051: From these simulations we obtain the following: if defects
1052: occur then $p_{max} > P_{SN}$
1053: in all observed cases. However, it is not true that
1054: a defect is formed any time we observe $p_{max} > P_{SN}$.
1055: If we let the system relax for a long time
1056: ($t=500,000$) and measure $p_{max}$,
1057: then the number of initial conditions
1058: leading to a $p_{max} > P_{SN}$ without defect formation
1059: is noticeably reduced. For $c_3=0.65$  the
1060: maximal conserved phase gradient is $\nu_M \sim 0.1$ .
1061: In the simulations we do not observe a defect for
1062: $\nu=0.086$ and $\nu=0.098$ but in the first
1063: case only 2 \% of all runs show
1064: $p_{max} > P_{SN}$, while in the latter case this percentage
1065: increases to 8 \%. 
1066: Increasing $\nu$ the maximal period $p_{max}$ increases and
1067: more and more situations with $p_{max} > P_{SN}$ are found
1068: upon approach to $\nu_M$.
1069: 
1070: The differences to the $\nu=0$ case may be explained by the
1071: coexistence of chaotic and stable not-chaotic attractors.
1072: Depending on the initial conditions, the solution
1073: of the CGLE can evolve towards one or the other.
1074: Therefore the system may exhibit local structures similar to
1075: multi-hump MAWs that possess SN bifurcations at parameter values 
1076: larger than those of single MAWs (compare Fig.~\ref{L4pd}).
1077: In that case, some periods $p$ may even exceed $P_{SN}$.
1078: 
1079: %_______________________________________________________________________________
1080: \subsection{Limit of wound-up phase chaos}
1081: \label{nuMok}
1082: 
1083: For random initial conditions with $\nu$ in the narrow range between the SN 
1084: of $P\to\infty$ and the existence limit of MAWs (see Fig.~\ref{cqexist}),
1085: it depends on the specific realization of the noise
1086: whether a defect can form or a stable MAW results.
1087: %
1088: No defects form below the line $\nu_M (c_3)$.
1089: %
1090: In order to understand this observed limit $\nu_M (c_3)$ of wound-up phase chaos, 
1091: it is sufficient to consider the SNs of single MAWs.
1092: 
1093: Although initial conditions with large $P$ beyond a SN could lead to defects,
1094: this is often prevented by the action of the splitting instability. 
1095: %
1096: Then the period is decreased before a defect can form.
1097: %
1098: Following the SN curve in Figs.~\ref{nu1q25} and \ref{nu1c5} one encounters a
1099: transition from SNs with a splitting instability at large $P$ to SNs without
1100: this instability at short $P$.
1101: %
1102: Defect formation in wound-up phase chaos with $\nu > 0.1$ occurs for parameters
1103: where the splitting instability is {\it not} present near the SN, {\it
1104: i.e.} above the dotted line in Fig.~\ref{cqnuM}.
1105: 
1106: \begin{figure}
1107: \begin{center}
1108:  \epsfxsize=0.9\hsize \mbox{\hspace*{-.06 \hsize} \epsffile{Fig/fig14.eps} }
1109: \end{center}
1110:  \caption[Theoretical bounds for $\nu_M (c_3)$]
1111:  {Theoretical bounds for $\nu_M (c_3)$ (filled circles) : 
1112:  SN of $P\to\infty$ as lower bound (thick solid curve)
1113:  and the presence of splitting at SN as upper bound (dotted curve).
1114:  Other curves as in Fig.~\ref{FnuM}. 
1115:  See text for details.
1116:  } 
1117:  \label{cqnuM}
1118: \end{figure}
1119: 
1120: Fig.~\ref{cqnuM} summarizes the bounds found so far  for the limit
1121: of wound-up phase chaos.
1122: %
1123: The domain of stable plane waves at low $c_3$ is limited by the Eckhaus
1124: instability (thin curve). 
1125: %
1126: Within the shaded area only supercritical HBs of different period $P$ occur 
1127: but no SNs.
1128: %
1129: This area is limited by the lowest SN curve of $P\to\infty$ (thick
1130: solid curve). 
1131: %
1132: No defects do form from random initial conditions within the shaded
1133: area.
1134: %
1135: The dashed curve denotes the upper limit of the existence domain 
1136: of MAWs with any period.
1137: %
1138: Saddle-node bifurcations exist in the window between this dashed curve and 
1139: the thick solid curve for the SN with $P\to\infty$.
1140: %
1141: Defects will always form for any choice of initial conditions at parameters 
1142: above the dashed curve.
1143: %
1144: The dotted curve gives the transition from active (below) to inactive (above)
1145: splitting modes at the SN. 
1146: %
1147: This transition is computed by linear stability analysis along cuts like
1148: Figs.~\ref{nu1q25} and \ref{nu1c5}.
1149: %
1150: Splitting inhibits defect formation below this dotted curve.
1151: %
1152: Filled circles correspond to $\nu_M$ obtained from numerical simulations
1153: as in \cite{torcini}.
1154: Diamonds refer to the transition from chaotic (below) to non-chaotic (above) 
1155: asymptotic states.
1156: 
1157: The SN for $P\to\infty$ (thick solid) is a lower bound for defect
1158: formation which also holds in the limit $\nu=0$. 
1159: %
1160: The point L$_1$ marks the transition from phase to defect chaos
1161: studied earlier \cite{MAW1,MAW2}.
1162: 
1163: As long as the dynamics is regular ($\nu>0.1$), the upper bound for the limit
1164: (onset of splitting at SN, dotted curve) of wound-up phase chaos
1165: reproduces well the numerical observations. 
1166: %
1167: For chaotic states with $\nu<0.1$ defect formation eventually becomes possible 
1168: even despite the presence of the splitting instability. 
1169: This coincides with the increasing weight of the instability of single MAWs 
1170: to attractive interaction of subsequent amplitude peaks. 
1171: 
1172: %_______________________________________________________________________________
1173: \section{Experimental observations}
1174: \label{nu1exp}
1175: 
1176: A variety of experimental observations in
1177: quasi one-dimensional geometries can be well interpreted by MAWs.
1178: % 
1179: These systems shall also serve for testing further properties of MAWs
1180: \cite{mvhexp}.
1181:  
1182: ({\it i}) 
1183: For the oscillatory instability of Rayleigh-B\'enard convection patterns
1184: in an annular cell, B. Janiaud {\it et al.} report long living transients of 
1185: modulated waves that eventually cause defects \cite{janiaud}. 
1186: %
1187: The underlying Eckhaus instability was found to be subcritical. 
1188: %
1189: Then, we expect the modulation to grow roughly exponentially
1190: whereas the long transient of an almost saturated modulation is similar to
1191: the dynamics near the saddle-node bifurcation as shown in
1192: Fig.~\ref{nu1snsim}e.
1193: %
1194: In the latter case the Eckhaus instability is supercritical and
1195: stable MAWs may exist for nearby values of the experimental
1196: parameters, see also the discussion in \cite{janiaud2}.
1197: %
1198: In all cases a single pulse-shaped modulation with the period equal to the cell 
1199: length was present in the system and led to a defect. 
1200: %
1201: The excitation of several modulations per cell
1202: and thereby smaller period can stabilize the modulated pattern and provide more
1203: examples of the dynamics near the saddle-node bifurcation.
1204: 
1205: ({\it ii}) 
1206: Hydrothermal waves have been studied in ring-shaped cells \cite{h2} 
1207: as well as in linear cells \cite{h1,garnier,CIAI4}.
1208: %
1209: N. Mukolobwiez {\it et al.} report a supercritical Eckhaus instability,
1210: stable MAWs with the period equal to the cell length and defect formation 
1211: after a parameter change \cite{h2}. 
1212: %
1213: N. Garnier {\it et al.} observe modulated waves with both the wavenumber and the
1214: period of modulations being selected by one of the longitudinal
1215: boundaries \cite{garnier}.
1216: %
1217: Thermal or mechanical forcing at the boundary may yield more insight into the
1218: multistability of the two-parameter family of MAWs.
1219: 
1220: ({\it iii})
1221: In rotating Rayleigh-B\'enard convection Y. Liu {\it et al.} observe the
1222: subcritical Eckhaus instability of a traveling wave sidewall mode
1223: \cite{sidewall}. 
1224: %
1225: The authors suggest higher order corrections to the CGLE in
1226: order to explain the observed discrepancy between the linear group velocity and 
1227: the observed velocity of finite wavelength perturbations.
1228: %
1229: However, this difference already follows from the linear analysis 
1230: of the Eckhaus instability \cite{notevHB} and can be used as a test
1231: for the assumed coefficients of the CGLE.
1232: 
1233: ({\it iv})
1234: For the Taylor-Dean system I. Mutabazi {\it et al.} report a stable MAW that
1235: they called ``triplet state'' because of the length scale ratio of modulation and
1236: underlying wave \cite{dean}. 
1237: %
1238: They also observe the formation of defects. 
1239: %
1240: Clearly the triplet state is just one realization of the two-parameter family 
1241: of MAWs.
1242: 
1243: ({\it v})
1244: Finally we mention the heated wire convection \cite{wire}, internal waves
1245: \cite{velarde} and the oscillatory variant of the Belousov-Zhabotinsky reaction
1246: \cite{chem} where further investigations of the observed Eckhaus
1247: instabilities may also reveal MAWs.
1248: 
1249: %_______________________________________________________________________________
1250: \section{Discussion}
1251: \label{nu1disc}
1252: 
1253: 
1254: The bifurcation analysis of modulated amplitude waves (MAWs) 
1255: has been extended to nonzero average phase gradient ($\nu\ne0$).
1256: %
1257: Small amplitude MAWs (``lower branch''-MAWs) of specific spatial
1258: period  $P$ exist between a supercritical 
1259: Hopf bifurcation (HB) and a saddle-node (SN) bifurcation. 
1260: %
1261: The HB asymptotically reaches the Eckhaus instability from above as $P$ goes to
1262: infinity.
1263: %
1264: MAWs are a direct consequence of the Eckhaus instability
1265: of plane waves; they are obtained by  a computer-assisted nonlinear analysis
1266: of this instability.
1267: %
1268: We encounter SNs with decreasing values of $P$ as $c_1, c_3$ and $\nu$ are
1269: increased.
1270: %
1271: These SNs govern the formation of defects from random initial 
1272: conditions as well as many aspects of the evolution of wound-up phase chaos.
1273: %
1274: The SNs bound the existence region of MAWs in the
1275: Eckhaus unstable regime.
1276: 
1277: A linear stability analysis of MAWs revealed that
1278: they can be linearly stable even in systems of infinite size.
1279: %
1280: These domains are limited by the interaction instability at low and the
1281: splitting instability at high values of the spatial period $P$ of the MAW.
1282: %
1283: The competition of the two instabilities drives wound-up phase chaos and
1284: determines the degree of chaoticity of the dynamics.
1285: %
1286: %Wound-up phase chaos may be characterized as evolution along the 
1287: %manifold of MAW solutions as for $\nu=0$.
1288: %
1289: For fixed coefficients $c_1$ and $c_3$,
1290: the SN associated to $P\to\infty$ occurs at the lowest value of $\nu$.
1291: This SN establishes a ``lower'' bound for defect formation and 
1292: thereby for the limit $\nu_M$ of wound-up phase chaos.
1293: %
1294: The splitting instability can inhibit defect formation if the SN occurs at 
1295: large $P$ hence defects are created more frequently for parameters above a
1296: second curve where the splitting instability vanishes at the SN.
1297: Thereby an ``upper'' bound for the limit $\nu_M$ of wound-up phase chaos 
1298: is obtained.
1299: Earlier numerical observations on $\nu_M (c_3)$ are well reproduced for
1300: $\nu > 0.1$, respectively small $c_3$. 
1301: %
1302: For $\nu < 0.1$, the description of phase chaos
1303: relies on considerations similar to those already discussed in the 
1304: limit case $\nu=0$ \cite{MAW1,MAW2}.
1305: %
1306: Finally, several experimental observations were interpreted in terms of MAWs.
1307: 
1308: We acknowledge fruitful discussions with M. van Hecke, N. Garnier, F. Daviaud,
1309: W. van de Water and L. Kramer.
1310: AT was partially supported by the MURST-COFIN00 
1311: grant ``Chaos and localization in classical and quantum mechanics''.
1312: 
1313: %______________________________________________________________________________
1314: \begin{thebibliography}{00}
1315:   
1316: \bibitem{books} Y. Kuramoto, {\em Chemical Oscillations, Waves
1317: and Turbulence} (Springer, 1984, Berlin);
1318: P. Manneville, {\em Dissipative structures and Weak Turbulence}
1319: (Academic Press, 1990, San Diego);
1320: T. Bohr, M. H. Jensen, G. Paladin and A. Vulpiani, {\em Dynamical
1321: systems approach to turbulence} (Cambridge Univ. Press) (1998).
1322: 
1323: \bibitem{CH} M. C. Cross and P. C. Hohenberg, Rev. Mod. Phys. {\bf
1324: 65}, 851 (1993).
1325: 
1326: \bibitem{eck} W. Eckhaus, {\it Studies in Nonlinear Stability Theory}
1327: (Springer Verlag, New York, 1965).
1328: 
1329: \bibitem{janiaud} B. Janiaud, A. Pumir, D. Bensimon, V. Croquette, H.
1330:   Richter and L. Kramer, Physica D, {\bf 55}, 269 (1992).
1331: 
1332: \bibitem{h1} F. Daviaud and J. M. Vince, 
1333: Phys. Rev. E {\bf 48}, 4432 (1993).
1334: 
1335: \bibitem{h2} N. Mukolobwiez, A. Chiffaudel and F. Daviaud, 
1336: Phys. Rev. Lett. {\bf 80}, 4661 (1998).
1337: 
1338: \bibitem{garnier} N. Garnier, A. Chiffaudel and F. Daviaud, unpublished results. 
1339: 
1340: \bibitem{CIAI4} N. Garnier and A. Chiffaudel, 
1341: Phys. Rev. Lett. {\bf 86}, 75 (2001).
1342: 
1343: \bibitem{wire} J. M. Vince and M. Dubois, Physica D, {\bf 102}, 93
1344: (1997).
1345: 
1346: \bibitem{sidewall} Y. Liu and R. E. Ecke, Phys. Rev. Lett. {\bf 78},
1347: 4391 (1997); Phys. Rev. E {\bf 59}, 4091 (1999).
1348: 
1349: \bibitem{dean} I. Mutabazi, J. J. Hegesth, C. D. Andereck and J. E. Wesfreid,
1350: Phys. Rev. Lett. {\bf 64}, 1729 (1990);
1351: P. Bot, O. Cadot and I. Mutabazi, Phys. Rev. E {\bf 58}, 3089 (1998);
1352: P. Bot and I. Mutabazi, Eur. Phys. J. B {\bf 13}, 141 (2000).
1353: 
1354: \bibitem{velarde} 
1355: A. Wierschem, H. Linde and M. G. Velarde, Phys. Rev. E {\bf 62}, 6522
1356: (2000).
1357: 
1358: \bibitem{chem} Q. Ouyang and J. M. Flesselles, Nature {\bf 379}, 143
1359: (1996); Q. Ouyang, H. L. Swinney and G. Li, Phys. Rev. Lett. {\bf 84},
1360: 1047 (2000); L. Q. Zhou and Q. Ouyang, Phys. Rev. Lett. {\bf 85}, 1650
1361: (2000).
1362: 
1363: \bibitem{review} A recent and detailed review on the CGLE is
1364: the following : I.S. Aranson and L. Kramer, {\em The World
1365: of the Complex Ginzburg-Landau Equation}, to appear in Rev. Mod. Phys.
1366: (cond-mat/0106115).
1367: 
1368: \bibitem{EckDiP} J. T. Stuart and R. C. DiPrima, 
1369: Proc. R. Soc. London A {\bf 362}, 27 (1978).
1370: 
1371: \bibitem{notevHB}
1372: The perturbation with wavenumber $k_c$ of a plane wave $q$ 
1373: travels with a velocity $v_c = -\mbox{Im}[\sigma(k_c)]/k_c$.
1374: At the Eckhaus instability, in the infinite system, 
1375: $k_c\to 0$ and $v_c$ equals the group velocity 
1376: $v_g = \partial \omega_q /\partial q = 2 q (c_1+c_3)$.
1377: For finite system size $k_c > 0$ and $v_c \ne v_g$.
1378: 
1379: \bibitem{EckCI1} I. S. Aranson, L. Aranson, L. Kramer and A. Weber, 
1380: Phys. Rev. A {\bf 46}, R2992 (1992).
1381: 
1382: \bibitem{EckCI2} A. Weber, L. Kramer, I. S. Aranson and L. Aranson, 
1383: Physica D {\bf 61}, 279 (1992).
1384: 
1385: \bibitem{pumir}  A. Pumir, B. I. Shraiman, W. van Saarloos,
1386: P. C. Hohenberg,  H. Chat\'e and M. Holen, p. 173 in
1387: C. D. Andereck and F. Hayot (Eds.),``Ordered and Turbulent patterns in
1388: Taylor-Couette Flow'' (Plenum Press, New York, 1992)
1389: 
1390: \bibitem{miguel} R. Montagne, E. Hern\'andez-Garc\'{\i}a and M. San Miguel,
1391: Phys. Rev. Lett. {\bf 77}, 267 (1996); R. Montagne,
1392: E. Hern\'andez-Garc\'{\i}a, A. Amengual and M. San Miguel, Phys. Rev. E {\bf
1393: 55}, 151 (1997).
1394: 
1395: \bibitem{torcini} A. Torcini, Phys. Rev. Lett. {\bf 77}, 1047 (1996);
1396:  A. Torcini, H. Frauenkron and P. Grassberger, Phys.
1397:   Rev E {\bf 55}, 5073 (1997).
1398: 
1399: \bibitem{hager} G. Hager, {\it Quasiperiodische L\"osungen der eindimensionalen
1400: komplexen Ginzburg-Landau Gleichung}, Diploma Thesis, University of Bayreuth,
1401: Germany (1996).
1402: 
1403: \bibitem{chao1}  B. I. Shraiman,  A. Pumir,  W. van Saarloos,
1404: P. C. Hohenberg,  H. Chat\'e and M. Holen, Physica D {\bf 57}, 241
1405: (1992). 
1406: 
1407: \bibitem{chate} H. Chat\'e, Nonlinearity {\bf 7}, 185 (1994); p. 33 in
1408:  P. E. Cladis and Palffy-Muhoray (Eds.),``Spatio-Temporal Pattern Formation in
1409:  Nonequilibrium Complex Systems'' (Addison Wesley, Reading, 1995).
1410: 
1411: \bibitem{saka} H. Sakaguchi, Prog. Theor. Phys. {\bf 84}, 792 (1990).
1412: 
1413: \bibitem{egolf} D. A. Egolf and H. S. Greenside, 
1414:  Phys. Rev. Lett. {\bf 74}, 1751 (1995).
1415: 
1416: \bibitem{MAW1} L. Brusch, M. G. Zimmermann, M. van Hecke, M. B\"ar and
1417: A.  Torcini, Phys. Rev. Lett. {\bf85}, 86 (2000).
1418: 
1419: \bibitem{MAW2} L. Brusch, A.  Torcini, M. van Hecke, M. G. Zimmermann
1420: and M. B\"ar, to appear in Physica D (nlin.CD/0104029).
1421: 
1422: \bibitem{saar} W. van Saarloos and P. C. Hohenberg, Physica D {\bf
1423: 56}, 303 (1992); {\bf 69}, 209 (1993) [Errata].
1424: 
1425: \bibitem{mvh} M. van Hecke, Phys. Rev. Lett. {\bf 80}, 1896 (1998).
1426: 
1427: \bibitem{mm} M. van Hecke and M. Howard, Phys. Rev. Lett. {\bf 86},
1428: 2018 (2001).
1429: 
1430: \bibitem{note1} By substituting $\kappa\!:=\!a_z/a$ one reproduces the form of
1431: the ODEs used in \cite{saar} which is more appropriate for studies of fronts.
1432: 
1433: \bibitem{auto97} E. Doedel, A. Champneys, T. Fairgrieve, Y. 
1434: Kusnetsov, B. Sandstede and X. Wang, 
1435: {\sc AUTO97}:{\it Continuation and bifurcation software for ordinary
1436: differential equations} (Concordia University, Montreal, 1997).
1437: 
1438: \bibitem{DwightDP}
1439: M. Kness, L. Tuckerman and D. Barkley, Phys. Rev. A {\bf 46}, 
1440: 5054 (1992). 
1441: 
1442: \bibitem{bloch} 
1443: N. W. Ashcroft and N. D. Mermin, {\it Solid State Physics} (Holt, Rinehart and
1444: Winston, New York, 1976);
1445: P. Collet and J.-P. Eckmann, {\it Instabilities and Fronts in Extended Systems}
1446: (Princeton University Press, 1990).
1447: 
1448: \bibitem{michal} M. Or-Guil, I. G. Kevrekidis and M. B\"ar, Physica D {\bf
1449: 135}, 154 (2000).
1450: 
1451: \bibitem{mvhexp} M. van Hecke, submitted to Physica D (cond-mat/0110068).
1452: 
1453: \bibitem{janiaud2} B. Janiaud, E. Guyon, D. Bensimon, and V. Croquette, 
1454: in F. Busse and L. Kramer (Eds.),
1455: ``Nonlinear Evolution of Spatio-Temporal Structures in Dissipative Continuous
1456: Systems'', NATO ASI Series B, Vol.225 (Plenum, New York, 1990).
1457: 
1458: \end{thebibliography}
1459: 
1460: \end{document}
1461: 
1462: