1: %
2: %****************************************************************
3: %\documentstyle[aps,pre,eqsecnum,epsfig,amssymb,psfig]{revtex}
4: \documentclass[preprint,aps,pre,showpacs,eqsecnum,nobibnotes,amssymb,amsmath,floatfix]{revtex4}
5: %\documentclass[draft,aps,pre,showpacs,eqsecnum,nobibnotes,amssymb,amsmath,floatfix]{revtex4}
6:
7: %\documentclass[twocolumn,secnumarabic,amssymb,nobibnotes,aps,pre]{revtex4}
8: %\documentstyle[preprint,aps,pre,eqsecnum,epsfig,amssymb]{revtex}
9: %\documentstyle[twocolumn,aps,pre,eqsecnum,epsfig,amssymb]{revtex}
10: \usepackage{graphicx}
11: %\usepackage{docs}
12: \usepackage{fontenc}
13:
14:
15: \renewcommand{\imath}{{\rm i}}
16:
17:
18: \pagestyle{myheadings}
19: \markboth{\today}{\today}
20:
21: \setcounter{secnumdepth}{3} %determines which sect. units are numbered (default: 3)
22:
23:
24: %\bibliographystyle{prsty}
25: \bibliographystyle{apsrev} %%(?)
26: %%%\twocolumn[\hsize\textwidth\columnwidth\hsize\csname
27: %@twocolumnfalse\endcsname
28:
29: \begin{document}
30: %%%\texttt{graphics}
31:
32:
33: %****************************************************************
34: \title{Noise sensitivity of sub- and supercritically bifurcating patterns
35: with group velocities close to the convective-absolute instability\\}
36:
37: \author{A.~Szprynger}
38: \affiliation{Institute of Low Temperature and Structure Research,
39: Polish Academy of Sciences, POB 937, 51-112 Wroc\l aw, Poland}
40: \author{M.~L\"ucke}
41: \affiliation{Theoretische Physik, Universit\"at des Saarlandes,
42: Postfach 151150, D-66041 Saarbr\"ucken, Germany}
43: \date{\today}
44: %****************************************************************
45:
46: \begin{abstract}
47: The influence of small additive noise on structure formation
48: near a forwards and near an inverted bifurcation as described by a cubic and
49: quintic Ginzburg Landau amplitude equation, respectively, is studied numerically
50: for group velocities in the vicinity of the convective-absolute instability
51: where the deterministic front dynamics would empty the system.
52: \end{abstract}
53:
54: \pacs{PACS number(s): 47.20.Ky, 47.54.+r, 43.50.+y, 05.40.-a}
55: % 47.20.Ky : Nonlinearity (including bifurcation theory)
56: % 47.54.+r : Pattern selection; pattern formation
57: % 43.50.+y : Noise: its effects and control
58: % 05.40.-a : Fluctuation phenomena, random processes, noise, and Brownian motion
59: \maketitle
60: %\tableofcontents
61:
62: %%\narrowtext
63: %\clearpage
64: %%%%%%%%%%%%%%%%%%%%%%%%%%% Sec. I %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
65: \section{Introduction}
66: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
67:
68: The formation of macroscopic structures \cite{CH93} in systems that
69: are driven out of thermal equilibrium by an externally imposed generalized
70: stress are usually investigated by deterministic field
71: equations. However, under specific circumstances the influence of external
72: deterministic or stochastic perturbations
73: and of internal thermal noise on the pattern formation process should be taken into
74: account to achieve a more realistic and quantitative
75: description of experiments. One prominent example are the so-called noise
76: sustained structures
77: \cite{D85,BAC,Steinberg,MLK92,SR92,LR93,SBH94,Deissler94,NWS96,LS97,T97,SCMW97,CWS99}
78: in the convectively unstable parameter regime
79: \cite{BersBriggs,Huerre} in, e.g., the Taylor-Couette
80: \cite{BAC,Steinberg,LR93,SBH94,Deissler94},
81: the Rayleigh-B\'enard \cite{MLK92,SR92,T97} system, or nonlinear optics
82: \cite{SCMW97}. Further examples are certain open-flow instabilities , e.g., in
83: wakes and jets that are reviewed in \cite{Huerre}.
84:
85: The noise sustained structures
86: \cite{D85,BAC,Steinberg,MLK92,SR92,LR93,SBH94,Deissler94,NWS96,LS97,T97,SCMW97,CWS99}
87: arise when an externally imposed through-flow or an internally generated
88: group velocity $v$ is large enough to "blow"
89: the pattern out of the system according to the
90: deterministic field equations. In this driving regime one observes in
91: experiments \cite{BAC,Steinberg,SR92,T97,SCMW97} structures that are sustained by
92: sources that generate perturbations in the band of modes that are amplified
93: according to the supercritical deterministic growth dynamics in downstream
94: direction sufficiently far away from the inlet.
95:
96: The criterion \cite{BersBriggs,Huerre} at which $v$
97: the pattern is blown out of the system under deterministic laws which gave
98: the threshold for the appearance of the noise sustained, supercritically
99: bifurcating patterns in the above described experiments is a linear one. It was
100: nonlinearly extended by Chomaz \cite{Chomaz92} to the question of the
101: propagation direction of nonlinear
102: deterministic fronts in infinite systems that connect the unstructured state
103: to the finite-amplitude structured one.
104:
105: Here we study and compare the noise sensitivity of pattern forming systems in
106: which the above described fronts are linear or nonlinear ones. To that end we
107: investigate the cubic Ginzburg-Landau amplitude equation (GLE) for a supercritical
108: forwards bifurcation and the quintic GLE for a subcritical inverted bifurcation,
109: respectively, in one spatial dimension.
110:
111: We solve the GLE with additive stochastic forcing numerically. Our systems
112: are finite but sufficiently long to allow the establishment of a statistically
113: stationary large-amplitude bulk part -- provided the latter is possible with the
114: boundary condition of vanishing amplitude at the ends. We focus our attention to
115: parameters in the vicinity of the convective-absolute threshold at which the
116: fronts of the
117: deterministic GLE cease to propagate. And we investigate in particular the
118: statistical dynamics of phase and amplitude fluctuations in the front region.
119:
120: %\clearpage
121: %%%%%%%%%%%%%%%%%%%%%%%%%%% Sec. II %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
122: \section{SYSTEM}
123: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
124: We consider the stochastic, 1D Ginzburg-Landau equation
125: \begin{equation}
126: (\partial_{t}+v\partial_{x}) A\;=\;(\mu + \partial_{x}^2
127: + g_3 |A|^2 + g_5 |A|^4 ) A + \sigma \eta
128: \label{CGLE}
129: \end{equation}
130: for the complex amplitude
131: \begin{equation}
132: A\;=\;\Re A + \imath\, \Im A\;=\;R e^{\imath\,\Phi}
133: \label{A_ReIm}\end{equation}
134: depending on $x,t$. Here $\Re (\Im)$ denotes the real (imaginary) part
135: and $R=|A|$ is the modulus and $\Phi$ the phase of $A$.
136: The coefficients in (\ref{CGLE}) are taken as real for simplicity. We checked
137: however that taking into account the (small) imaginary parts, that appear
138: e.g.~in the case of transverse Rayleigh-B\'enard convection rolls propagating
139: downstream in a small externally imposed lateral through-flow \cite{MLK92} or in
140: the case of downstream propagating Taylor vortices \cite{BAC,RLM93} does not
141: change the major findings presented in this paper significantly.
142: We consider the group- or mean flow velocity $v\ge 0$ in positive $x$-direction
143: and the linear growth rate $\mu$ of $A$ as control parameters.
144:
145: We investigate
146: two fixed combinations of the nonlinear coefficients ($g_3,g_5$) that we refer
147: to in this paper as follows
148: \begin{subequations}
149: \begin{eqnarray}
150: g_3\;=\;-1\;, \; g_5\; = \;0\; &:& \quad \mbox{cubic GLE} \\
151: g_3\;=\; 1\;, \; g_5\; = \;-1\;&:& \quad \mbox{quintic GLE} \, .
152: \end{eqnarray}
153: \end{subequations}
154: The quantity $\sigma$ in (\ref{CGLE}) measures the real strength of the complex
155: stochastic force
156: \begin{equation}
157: \eta(x,t)\;=\;\Re\eta(x,t) + \imath\, \Im\eta(x,t)
158: \end{equation}
159: with statistically independent real and imaginary parts $\Re\eta$ and $\Im\eta$,
160: respectively. Both are Gaussian distributed with zero mean and
161: $\delta$-correlated such that
162: \begin{equation}
163: <\eta(x_1,t_1)[\eta(x_2,t_2)]^*>\;=\;2\delta(x_1-x_2)\delta(t_1-t_2) \;.
164: \label{f_corr}
165: \end{equation}
166: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Sec IIA %%%%%%%%%%%%%%%%%%%%%%%%
167: \subsection{Unforced homogeneous solution}
168: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
169: We are interested in the effect of small additive noise on
170: the spatio-temporal structure formation in large but finite or semiinfinite
171: systems. Nevertheless it is useful to briefly recall first the properties of the
172: most simple solutions of the unforced GLE in an infinite system. This shows what
173: one might expect to see in the bulk of a very large system far away from the
174: boundaries --- ignoring for the moment questions related to boundary induced
175: pattern selection processes.
176:
177: The GLE (\ref{CGLE}) shows for $\sigma$=0 a continuous family of traveling wave
178: (TW) solutions
179: \begin{equation}
180: \label{TWsolution}
181: A(x,t)= R e^{iq(x-vt)}
182: \end{equation}
183: with constant wave number $q$, frequency $\Omega=-qv$, and modulus $R$ given by
184: \begin{equation}
185: \mu -q^2 + g_3 R^2 + g_5 R^4 = 0 \, .
186: \end{equation}
187: This TW solution family bifurcates at the marginal stability curve, $\mu = q^2$,
188: of the $A$=0 solution out of the latter while the former becomes unstable there.
189: The critical values are $\mu_c=q_c=\Omega_c$=0. The bifurcation is nonhysteretic
190: and forwards in the cubic case
191: \begin{equation}
192: \label{Rcubic}
193: R^2=\mu -q^2
194: \end{equation}
195: and hysteretic, backwards in the quintic case
196: \begin{equation}
197: \label{Rquintic}
198: R^2=\frac{1}{2} \pm \sqrt{\mu -q^2 + \frac{1}{4}} \, .
199: \end{equation}
200: Here the lower sign refers to the lower unstable TW solution branch that exists
201: for $-\frac{1}{4} \leq \mu -q^2 \leq 0$. The upper TW solution branch identified
202: by the + sign in Eq.~(\ref{Rquintic}) exists beyond the saddle-node bifurcation
203: value $\mu = q^2 -\frac{1}{4}$. These TW solutions are stable for wave numbers
204: outside the Eckhaus unstable band \cite{BD92}.
205:
206: %%%%%%%%%%%%%%%%%%%%%%%%%%% Sec. IIB %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
207: \subsection{Convective-absolute instability}
208: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
209: The noise susceptibility of the pattern formation process described by the GLE
210: (\ref{CGLE}) changes significantly \cite{D85,Huerre} when crossing the parameter
211: combination of
212: $\mu, v$ shown in Fig.~\ref{figure:Maxwell} for the so called
213: convective-absolute instability \cite{BersBriggs}. This combination
214: \begin{eqnarray}
215: \mu_{c-a}=
216: \left\{ \begin{array}{ll} \frac{1}{4} v^2 &\mbox{cubic GLE} \\
217: \frac{3}{16}\left(v^2 + \frac{2}{\sqrt{3}} v - 1 \right)\quad & \mbox{quintic GLE}
218: \end{array}\right.
219: \label{mu_conv}
220: \end{eqnarray}
221: is marked by the front solution of the deterministic GLE with
222: $\sigma=0$ undergoing a reversal of the front propagating direction in an
223: infinite system. Consider a front that connects the basic state $A=0$ being
224: realized at $x\to-\infty$ to a homogeneous solution with $A\ne 0$ at
225: $x\to\infty$. For parameter values below (above) the respective curves in
226: Fig.~\ref{figure:Maxwell} this front moves to the right (left). Thus the basic
227: state $A=0$ (the homogeneous solution $A\ne 0$) expands to the right (left).
228: The region below (above) the respective curves in Fig.~\ref{figure:Maxwell}
229: where the basic state $A=0$ (the homogeneous state $A\ne 0$) invades the whole
230: system is called the convectively (absolutely) unstable region of the $A=0$
231: solution \cite{D85,Huerre}. Thus, the boundary (\ref{mu_conv}) is also called
232: the convective-absolute instability boundary.
233:
234: For the cubic GLE the boundary $\mu_{c-a}=v^2/4$ results from a linear analysis
235: \cite{D85}.
236: For the backwards bifurcating solution arising in the quintic GLE the respective
237: front that reverts its propagation direction is a nonlinear one \cite{SH92}. Note
238: that in the latter case the convective-absolute instability boundary
239: \cite{Chomaz92}
240: connects for $v\to 0$ to the so-called Maxwell point $\mu_{c-a}=\mu_M=-3/16$:
241: For this value the minima of the potential $V(A)=-\frac{\mu}{2}A^2 -
242: \frac{1}{4}A^4 + \frac{1}{6}A^6$ have equal height $V=0$.
243:
244: The boundary condition $A(x=0,t)=0$ that we apply in our simulations stops
245: any front
246: propagating to the left and it changes, i.e., it deforms the front profile when
247: the front is sufficiently close to the boundary at $x=0$. This can be seen in
248: Fig.~\ref{figure:R-kappa} for the example of the deterministic quintic GLE. There
249: the lines show the modulus profile $R$ and the spatial growth rate $\kappa=R'/R$
250: versus $x$ together with $\kappa$ versus $R$ obtained numerically for several
251: parameter values above the convective-absolute instability boundary. To
252: facilitate comparison of different cases we
253: introduce the reduced horizontal "distance"
254: \begin{equation}
255: \Delta\;=\;\frac{v}{v_{c-a}} - 1
256: \end{equation}
257: from the boundaries shown in Fig.~\ref{figure:Maxwell}. Here
258: \begin{eqnarray}
259: v_{c-a}(\mu)\;=\;
260: \left\{ \begin{array}{ll} 2 \sqrt{\mu} \quad &\mbox{cubic GLE} \\
261: \sqrt{\frac{4}{3} \left(1 + 4\mu\right)} - \sqrt{\frac{1}{3}}
262: \quad & \mbox{quintic GLE}
263: \end{array}\right. \label{v_c-a}
264: \end{eqnarray}
265: denotes the convective-absolute instability boundary (\ref{mu_conv}).
266:
267: The results that we present here were obtained for $\mu>0$, i.e., in a situation
268: where the basic state $A=0$ is unstable. For the backwards bifurcation in the
269: quintic GLE with negative growth rates $ -\frac{3}{16} < \mu < 0$ for which the
270: above cited potential has a minimum at $A=0$ the situation is more complicated
271: \cite{CWS99}: Not only does the establishment of the final front connecting the
272: inlet condition $A=0$ with a statistically stationary saturated bulk with
273: $|A|={\cal O}(1)$ depend sensitively on the initial condition [say, $A(x,t=0)=0$
274: versus $|A|={\cal O}(1)$] in the absolutely unstable regime, $\Delta < 0$. But
275: more importantly, in the convectively unstable regime, $\Delta > 0$, we found that
276: small
277: noise does not seem to be able to generate with the boundary condition $A(x=0)=0$
278: a noise sustained finite-amplitude structure with $\langle |A|^2\rangle$ of order
279: one when $\mu<0$: The deterministic front dynamics drives the large-amplitude part downstream
280: and eventually any finite system is filled only with small-amplitude fluctuations
281: of $A$ around the stable fixed point $A=0$ of the unforced system.
282:
283: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Sec IIC %%%%%%%%%%%%%%%%%%%%%%%%
284: \subsection{Noise strength}
285: \label{SEC:noise_strength}
286: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
287:
288: For the quintic GLE we choose the noise strength $\sigma=10^{-3}$. The noise
289: intensity $\sigma^2$ should be compared with the minimum of the potential
290: \begin{equation}\label{V_3-5}
291: V(A)=-\frac{\mu}{2}A^2 - g_3 \frac{1}{4}A^4 - g_5 \frac{1}{6}A^6 \, .
292: \end{equation}
293: For our quintic case ($g_3=1, g_5=-1$) the minimum at
294: $A^2=R_N^2=\frac{1}{2} + \sqrt{\frac{1}{4} + \mu}$ is
295: $V(R_N)=-\frac{1}{24}[1 + 6\mu + (1 + 4 \mu)^{3/2}]$. Thus, the noise
296: "temperature" $\sigma^2$ measured in units of $V(R_N)$ is $\sigma^2/|V(R_N)|=
297: 9.2 \, 10^{-6}$
298: for the control parameter $\mu=0.05$ that we have used in most of our
299: calculations.
300:
301: A rough estimate for an equivalent noise strength for the cubic GLE would be
302: to demand that the reduced noise "temperature" $\sigma^2/V(R_N)$ is in both
303: cases the same. This would
304: require for the cubic GLE at a common $\mu$ of, say, 0.05 that $\sigma$ is by
305: about a factor of 13 smaller than for the quintic GLE.
306:
307: However, basing the comparison on the requirement that $\sigma^2/ V(R_N)$
308: is the same for the cubic and quintic case one has to keep in mind that the
309: curvatures of $V$ around the states $A=0$ and $A=R_N$ which are connected by the
310: fronts remain different -- cf. Fig.~\ref{figure:V_3-5}. Since these curvatures
311: around $A=0$ ($A=R_N$) measure the growth (decay) rates of fluctuations around
312: the respective states it is useful to compare their ratios via a kind of
313: Ginzburg number $G=|V^{''}(0)|/V^{''}(R_N)$. One has $G_{cubic} = 1/2$
314: independent of $\mu$ and $g_3$ while
315: $G_{quintic} = \tilde \mu\left[ 1 + 4\tilde \mu
316: +\sqrt{1 + 4\tilde \mu} \right]^{-1}$ with $\tilde \mu = -g_5 \mu /g_3^2$. Thus
317: for $\mu=0.05$ and $g_3=1, g_5=-1$ one has
318: $G_{cubic} \simeq 23~G_{quintic}$. This largely explains the stronger noise
319: sensitivity of the cubic GLE for our parameters. In view of it we
320: investigated the whole range of $\sigma$ between $10^{-9}$ and $10^{-2}$
321: for the cubic GLE.
322:
323: The cubic GLE with additional (but very small) complex coefficients has
324: previously been investigated, e.g., for noise strengths of about
325: $\sigma=1.9\cdot10^{-6}$ in our units of eqs.~(\ref{CGLE}-\ref{f_corr}). The
326: corresponding noise "temperature" $\sigma^2/V(R_N)$ is about $10^{-8}$ for a
327: typical value
328: of, say, $\mu=0.035$ \cite{BAC}. This noise was found to fit the experimental
329: results on the noise sustained traveling Taylor vortices under statistically
330: stationary fronts in the convectively unstable regime
331: of open Taylor-Couette systems with axial through-flow \cite{BAC}.
332:
333: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Sec IID %%%%%%%%%%%%%%%%%%%%%%%%
334: \subsection{Numerical methods}
335: \label{SEC:num_methods}
336: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
337: Equation (\ref{CGLE}) was solved numerically with a forward-time, centered-space
338: method \cite{Pre94} subject to the boundary conditions
339: \begin{equation}
340: A(x=0,t)\;=\;0\;=\;A(x=L,t)
341: \end{equation}
342: on the complex amplitude. System sizes $L$ were chosen to be sufficiently large
343: to allow for the establishment of a saturated bulk amplitude. Typically,
344: a spatial step $dx= 0.4$ was used with a time step of $dt=0.072$. Calculations
345: were performed for sequences of the paramater $v$ at several values of the control
346: parameter $\mu$. Most of them
347: were done at $\mu=0.05$. The noise source $\eta$ was realized by Gaussian distributed
348: random numbers of unit variance that were divided by $\sqrt{dt dx}$
349: to ensure independence of the correlation functions of the discretization.
350: A test of different pseudo random number generators, namely, L'Ecuyer's method
351: with Bays-Durham shuffle \cite{Pre94}, \verb/ran3/ \cite{Pre94}, and the \verb/R250/
352: shift-register random number generator \cite{R250} gave similar results.
353:
354: After the simulations were started, a sufficiently long time depending on the
355: parameters, e.g., on the closeness to the convective-absolute threshold had to
356: be waited until the system relaxed into a statistically stationary state
357: with time independent averages. Thereafter time averages were evaluated over
358: several consecutive time intervals and finally averaged.
359: Within the forward-time integration method $A(x,t)$ remains uncorrelated with
360: $\eta(x',t)$ at the same time, $<f(\eta)g(A)>\; =\;<f(\eta)><g(A)>$, so that,
361: e.g., $<A\eta>\;=\;0$ as well as $<q\eta>\;=\;0$. But $<\Omega\eta>\;\ne\;0$. Here
362: the frequency $\Omega$ (wave number $q$) is defined as a forward-time
363: (centered-space) difference of the phase (\ref{Omegaq_def}).
364:
365:
366: %\clearpage
367: %%%%%%%%%%%%%%%%%%%%%%%%%%% Sec. III %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
368: \section{Results}
369: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
370: The influence of additive noise on the pattern formation process described by
371: the GLE (\ref{CGLE}) is described in this section.
372: %%%%%%%%%%%%%%%%%%%%%%%%%%% Sec. IIIA %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
373: \subsection{Growth length $\ell$}
374: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
375: In Fig.~\ref{figure:l-dl} we show how the growth length $\ell$ of the downstream
376: pattern occurring in the forced cubic GLE varies with noise strength $\sigma$.
377: Here $\ell$ is defined by the distance from $x=0$ at which the root-mean square
378: $\sqrt{<|A|^2>}$ of the fluctuating complex amplitude $A$ reaches half its bulk
379: value. In the absence of noise $\ell$ diverges at the convective-absolute
380: threshold
381: $v=v_{c-a}$ since there the deterministic pattern is blown out of the system.
382:
383: For finite $\sigma$ the solution with finite $A$ is noise-sustained in the
384: convectively unstable regime $\Delta>0$ \cite{D85}. In this regime $\ell$ is
385: far from the convective-absolute threshold well described by the relation
386: $\ell\sim - (1 + \sqrt{2\Delta})\ln{\sigma}$ following from a quasilinear
387: analysis of the cubic GLE \cite{LS97} presented here in an appendix.
388: However in the vicinity of the threshold
389: $\Delta\,=\,0$ the growth length $\ell$ obtained from the nonlinear GLE shows a
390: characteristic crossover to the behavior at $\Delta<0$.
391:
392: The noise influences also in this absolutely unstable regime, $\Delta<0$,
393: the finite amplitude solution at least close to threshold:
394: The curves $\ell(\Delta,\sigma)$ in Fig.~\ref{figure:l-dl}(a) break away from
395: the
396: dotted $\ell(\Delta,\sigma=0)$ reference growth length curve at negative
397: $\Delta$ values
398: that decrease with increasing $\sigma$, i.e., further and further away from the
399: convective-absolute threshold. The associated inflection points can be most
400: easily identified
401: by the maxima in $\partial\ell(\Delta,\sigma)/\partial\Delta$ shown
402: in Fig.~\ref{figure:l-dl}(b). These peak positions of
403: $\partial\ell/\partial\Delta$ vary with
404: $\sigma$ as shown in the inset of Fig.~\ref{figure:l-dl}(b). So the growth
405: length shows
406: for the cubic GLE a definite noise sensitivity also in the absolutely unstable
407: regime.
408:
409: This sensitivity is significantly smaller in the quintic GLE. This can be seen
410: by comparing the behavior of the growth length with the fluctuations of
411: the modulus
412: $R=|A|$, of the frequency, and of the wave-number (cf, Sec.~\ref{SEC:omega_q}).
413: To that end we show in Figs.~\ref{figure:l-s_qOm-cub} and ~\ref{figure:l-s_qOm-quin}
414: $\ell$ and $\partial\ell/\partial\Delta$ together with the inverse of the
415: standard deviations of the modulus
416: \begin{equation}
417: \label{sR_def}
418: s_R=\sqrt{<R^2> - <R>^2} \, ,
419: \end{equation}
420: of the frequency $s_{\Omega}$ (\ref{s_def}), and of the wave-number
421: $s_q$ (\ref{s_def}) at $\mu$=0.05
422: as functions of $\Delta$ for the cubic and quintic GLE, respectively.
423: The noise strengths $\sigma=2.5\cdot 10^{-5}$ and $10^{-3}$, respectively, used for
424: these figures are roughly equivalent based on the criterion described in
425: Sec.~\ref{SEC:noise_strength}. However, the potential minima in the cubic case are
426: broader than in the quintic case -- cf. Fig.~\ref{figure:V_3-5} -- and therefore
427: the modulus fluctuations in the
428: former are larger than those in the latter one. This can be seen by comparing the
429: reduced inverse $\sqrt{<R^2>}/s_R$ in the absolutely unstable regime, $\Delta<0$, of
430: Figs.~\ref{figure:l-s_qOm-cub}(c) and \ref{figure:l-s_qOm-quin}(c).
431:
432: The peak position of $\partial\ell/\partial\Delta$ coincides with the drop-off
433: in the inverse standard deviations $1/s$. For the cubic GLE
434: (Fig.~\ref{figure:l-s_qOm-cub}) it occurs
435: at $\Delta$=-0.049, thus being shifted significantly into the
436: absolutely unstable regime while that of the quintic GLE
437: (Fig.~\ref{figure:l-s_qOm-quin}) remains at $\Delta$=0.
438:
439: As an aside we mention that for the quintic GLE at a subcritical growth
440: parameter of, say, $\mu$=-0.05 the behavior of the growth length $\ell$ and
441: of $\partial\ell/\partial\Delta$ is for $\Delta<0$ similar to the one shown in
442: Fig.~\ref{figure:l-s_qOm-quin}(d) for $\mu$=0.05. For $\mu <0, \Delta > 0$ we
443: did not find a noise sustained large-amplitude solution.
444:
445: %%%%%%%%%%%%%%%%%%%%%%%%%%% Sec. IIIB %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
446: \subsection{Frequency and wave-number correlations}
447: \label{SEC:omega_q}
448: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
449: Previous investigations of the forced cubic GLE in the bulk
450: part of the solution at far downstream locations
451: $x\gg\ell$ showed for different but small noise strengths that frequency
452: fluctuations are in the absolutely unstable regime much smaller than in the
453: convectively unstable regime \cite{BAC}. In order to study this question of
454: the noise sensitivity in both regimes
455: we have investigated in more detail the frequency and wave-number
456: fluctuations at
457: $x=\ell/2,\ell$, and $2\ell$. The results are shown in Fig.~\ref{figure:l-s_qOm-cub}
458: for the cubic GLE and in Fig.~\ref{figure:l-s_qOm-quin} for the quintic GLE.
459: Before
460: we discuss them we first present some basic properties of the phase fluctuations
461: as described by the forced GLE (\ref{CGLE}).
462:
463: The phase fluctuations $\Phi$ of the complex amplitude (\ref{A_ReIm}) define
464: the frequency $\Omega$ and the wave number $q$
465: \begin{equation}
466: \Omega\;=\;\dot \Phi\;=\; \Im\left(\frac{\dot A}{A}\right)\;, \qquad %%{\cal I}m \frac{\dot A}{A}\;, \qquad
467: q\;=\;\Phi'\;=\;\Im\left( \frac{A'}{A}\right)\;,
468: \label{Omegaq_def} \end{equation} respectively. Here dot (prime) denotes temporal
469: (spatial) derivative. The growth rate $\kappa$ of the modulus is given by
470: \begin{equation} \kappa\;=\;\frac{R'}{R}\;=\;\Re\left( \frac{A'}{A}\right)\; .
471: \end{equation}
472: By means of Eq.~(\ref{CGLE}) the frequency can be expressed as
473: \begin{equation}
474: \Omega\;=\; (2\kappa -v) q + q' + \frac{\sigma}{R^2} \Im\left(\eta
475: A^*\right)\;.
476: \label{Omega_final}
477: \end{equation}
478: This relation holds for the cubic as well as for the quintic GLE with real
479: coefficients. By squaring and averaging Eq.~(\ref{Omega_final}) one
480: gets the correlation functions
481: \begin{eqnarray}
482: <\Omega^2> &+& v^2 <q^2> + 2v <\Omega q> +
483: <q^{\prime\; 2}> - 2 <\Omega q'> - 2 v <qq'>
484: \nonumber \\
485: &-& 4 v<\kappa q^2> + 4 <\kappa^2 q^2>
486: - 4 <\kappa \Omega q> + 4 <\kappa q q'>\nonumber \\\;&\simeq&\;
487: \frac{\sigma^2<|\eta|^2>}{2<R^2>} \; .
488: \label{Omega2_full}
489: \end{eqnarray}
490: On the r.h.s. we have used the fact that within our forward-time integration
491: method $A$ remains uncorrelated with $\eta$ at the same time and we have
492: approximated $<1/R^2>$ by $1/<R^2>$.
493:
494: Given that $<|\eta(t,x)|^2>=2 /dx dt$ in our finite difference
495: simulation it is convenient to scale all correlations in Eq.~(\ref{Omega2_full})
496: by the quantity
497: \begin{eqnarray}
498: \Sigma^2\;=\;\frac{\sigma^2}{R_N^2}\,\frac{1}{dx dt} \; , \qquad
499: R_N^2=\left\{ \begin{array}{ll} \mu &\mbox{cubic GLE} \\
500: \frac{1}{2} + \sqrt{\mu + \frac{1}{4}}\quad & \mbox{quintic GLE}
501: \end{array}\right.
502: \label{Sigma_def}
503: \end{eqnarray}
504: thereby removing the singularities from the reduced correlation functions. For
505: example one finds that
506: \begin{equation}
507: \frac{<(\Omega + v q - q')^2>}{\Sigma^2} \simeq \frac{R_N^2}{<R^2>}\;.
508: \label{Om2vsSigma}
509: \end{equation}
510: Here we have neglected the second line in Eq.~(\ref{Omega2_full}) since all
511: correlations in Eq.~(\ref{Omega2_full}) involving the growth rate $\kappa$ are
512: very small.
513:
514: $<\Omega^2>$ is typically two orders of magnitude larger than $<q^2>$ in the
515: absolutely unstable regime, $\Delta < 0$, -- cf.
516: Figs.~\ref{figure:l-s_qOm-cub} and \ref{figure:l-s_qOm-quin} discussed further
517: below. There the only contributions to
518: Eqs.~(\ref{Omega2_full},\ref{Om2vsSigma}) of the same order as $<\Omega^2>$ are
519: $<\Omega q'>$ and $<q'^2>$ -- all the other correlations can be neglected --
520: and furthermore $<\Omega q'> \simeq <q'^2>$. Thus,
521: \begin{equation}
522: <\Omega^2>\;\simeq\;\Sigma^2 + <q'^2>
523: \end{equation}
524: in the bulk part of the system with saturated amplitude where
525: $<R^2> \simeq R_N^2$.
526: However, in the convectively unstable regime, $\Delta > 0$, with much larger
527: phase fluctuations the situation is more complex. Here $<q^2>$ is larger than
528: $<\Omega^2>$ except for the upstream region where the reverse holds.
529:
530: In Fig.~\ref{figure:l-s_qOm-cub} and Fig.~\ref{figure:l-s_qOm-quin} we show
531: the inverse of the standard deviations
532: \begin{eqnarray} \label{s_def}
533: s_\Omega\;&=&\;\sqrt{<\Omega^2> - <\Omega>^2}\;, \qquad
534: s_q\;=\;\sqrt{<q^2> - <q>^2}\;,
535: \end{eqnarray}
536: reduced by $\Sigma$ (\ref{Sigma_def}) for the cubic and quintic GLE,
537: respectively, as functions of $\Delta$ for $x=\ell/2,\ell$, and $2\ell$.
538: For the parameters shown in Fig.~\ref{figure:l-s_qOm-cub} and
539: Fig.~\ref{figure:l-s_qOm-quin} the mean frequency $<\Omega>$ as well as the
540: mean wave number $<q>$ are negligible.
541: Plotting the inverse of $s_\Omega$, $s_q$, and $s_R$ allows to visualize the small
542: fluctuations in the absolutely unstable regime better than in a direct plot
543: of, say, $s_\Omega^2$. Such plots for $s_\Omega^2$ have been presented
544: previously for the small noise strengths occurring in Taylor-Couette
545: experiments \cite{BAC}. On the lower level of resolution inherent in this data
546: presentation these results show similar behavior as
547: ours. However, plotting $1/s_\Omega$ instead allows
548: to identify more clearly the crossover behavior
549: from the parameter regime with small fluctuations to the one with large ones.
550:
551: The $\Delta$-variations of $1/s_\Omega$, $1/s_q$, $1/s_R$, and of
552: $\partial\ell/\partial\Delta$ indicate that this transition is
553: shifted to negative $\Delta$, i.e.~into the absolutely unstable regime.
554: A similar result for the transition between deterministic and noise
555: sustained standing wave solutions of complex coupled cubic GLE's was deduced from
556: the behavior of the second moments of the frequency and wave-number power spectra
557: of the fluctuating amplitudes \cite{NWS96}: With decreasing $\mu$ the
558: correlation length defined via the time average of the second moment of the
559: Fourier spectrum of $A(k,t)$ begins to decrease towards values
560: characteristic for noise-sustained structures in the convectively unstable
561: regime clearly before $\mu_{c-a}$ is reached when noise is present. Similarly
562: the width of the frequency power spectrum starts to increase with decreasing
563: $\mu$ already above the convective-absolute threshold $\mu_{c-a}$ \cite{NWS96}.
564:
565: However, the variation of $1/s_\Omega$ with $\Delta$ shows for the cubic case in
566: Fig.~\ref{figure:l-s_qOm-cub} a broader crossover
567: interval between large frequency fluctuations in the convectively unstable
568: regime at $\Delta >0$ and small frequency fluctuations in the absolutely
569: unstable regime at $\Delta <0$ than the curves $1/s_q$ and $1/s_R$ for wave-number
570: and modulus fluctuations. The $\Delta$-value at which $1/s_q$ and $1/s_R$ drop
571: down towards zero
572: agrees quite well with the peak location of $\partial\ell/\partial\Delta$. The
573: latter moves with increasing noise strength further into the absolutely unstable
574: regime as shown, e.g., for the cubic GLE in the inset of
575: Fig.~\ref{figure:l-dl}(b).
576:
577: The variations of $s(\Delta)$ with $\Delta$ at different
578: downstream locations $x=\ell/2,\ell$, and $2\ell$ are similar to each other:
579: with $\Delta$ becoming more negative, i.e., further and further into
580: the absolutely unstable regime the fluctuations $s_\Omega$ and $s_q$ become
581: constant at levels that depend on the measuring location -- the closer to the
582: inlet where $R$
583: becomes smaller the larger are the fluctuations. This behavior is reflecting the
584: relation $s_\Omega \sim s_q \propto R^{-1}$ that can be read off directly from
585: Eq.~(\ref{Omega2_full}).
586:
587: The downstream reduction of the variance $s_\Omega$ of the frequency fluctuations
588: with increasing distance from the inlet and with increasing amplitude along the
589: front is shown in Fig.~\ref{R-s_om-x} for the quintic GLE. There we compare the
590: behavior
591: of $s_\Omega$ together with the front profiles of $\sqrt{\langle |A|^2 \rangle}$
592: in the absolutely and in the
593: convectively unstable regime close to the threshold $\Delta\,=\,0$ for $\mu=0.05$.
594:
595: %\clearpage
596: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
597: \section{Conclusion}
598: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
599: We have studied numerically the influence of small additive noise on pattern formation
600: near a forwards and near an inverted bifurcation as described by a cubic and
601: quintic GLE, respectively, when a finite group velocity $v$ can blow the
602: finite-amplitude part out of the system, i.e., in the vicinity of the
603: so-called convective-absolute instability at $\Delta=v/v_{c-a}(\mu)-1=0$. The
604: front that connects the inlet condition
605: $A(x=0)=0$ to the finite-amplitude downstream bulk part
606: $\langle |A|^2 \rangle \simeq R_N^2$ is for the cubic GLE
607: more sensitive to the applied noise strength than for the quintic case.
608: This is partly related to the different magnitudes of the curvatures of the
609: deterministic GLE potentials around the states $A=0$ and $A=R_N$: the resulting
610: growth enhancement
611: of fluctuations near $A=0$ is larger in the cubic than in the quintic case and
612: in addition the damping of fluctuations near $A=R_N$ is smaller in the cubic than
613: in the quintic case.
614:
615: In the cubic case the transition between the regimes of small and large
616: fluctuations of amplitude, frequency, and wave number is shifted to a negative
617: $\Delta$ into the absolutely unstable regime. Simultaneously the pattern growth
618: length $\ell(\Delta)$
619: has there a characteristic inflection point that shows up as a peak in
620: $\partial\ell/\partial\Delta$. In the quintic case all this occurs at the
621: unshifted convective-absolute threshold $\Delta=0$. Common to both cases is that the
622: fluctuations decrease along the front in both regimes with growing pattern
623: amplitude $\sqrt{<|A|^2>}$.
624:
625: For negative subcritical amplitude growth rates, $\mu < 0$, we did not find
626: noise-sustained,
627: large-amplitude, backwards bifurcating patterns when $\Delta$ is positive: the
628: nonlinear deterministic front dynamics of the quintic GLE blows any
629: large-amplitude part downstream away from the inlet where $A=0$
630: and eventually any finite system is filled only with small-amplitude fluctuations
631: of $A$ around the stable fixed point $A=0$ of the unforced system.
632:
633: \acknowledgments
634: Discussions with B. Neubert and his contributions to an early
635: stage of this research project are gratefully acknowledged.
636: One of us (A.~S.) acknowledges the hospitality of the Universit\"at
637: des Saarlandes.
638:
639: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
640: \begin{appendix}*
641: \section{} %{Noise dependence of the growth length}
642: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
643: Here we estimate the noise dependence of the downstream growth length $\ell$
644: of the nonlinear structure in the convectively
645: unstable regime of the cubic GLE where this structure is noise sustained.
646: To that end we approximate
647: $\ell$ by the length where the mean squared amplitude
648: $C_{lin}(x) = \langle |A_{lin}(x)|^2\rangle$ of the {\em linear} GLE has grown
649: from the inlet value $A(x=0)=0$ to, say, one half of the nonlinearly saturated
650: bulk value $\langle |A|^2\rangle \simeq \mu/2$. So we solve the equation
651: \begin{equation} \label{lin-cub}
652: C_{lin}(x=\ell)\;=\;\frac{1}{2} \mu\
653: \end{equation}
654: for $\ell$. Actually the linear solution may not hold there anymore. But as it
655: will become obvious
656: below the result is roughly independent of the coefficient chosen
657: in Eq.~(\ref{lin-cub}) so also smaller numbers than $\frac{1}{2}$ could be
658: chosen here for a characteristic growth length.
659:
660: We evaluate the equal-time correlation $C_{lin}(x)$ via the frequency integral
661: of the spectrum $C_{lin}(x,\omega)$ of the time-displaced autocorrelation
662: function of fluctuations of $A_{lin}$ at the same downstream position $x$.
663: For large downstream distances $x$ from the inlet this spectrum is given
664: by \cite{LS97}
665: \begin{equation} \label{C_asympt}
666: C_{lin}(x,\omega) \;=\; \frac{-\sigma^2}{2 | K_1^* - K_2 |^2 }
667: \left(\frac{1}{\Im K_1} + \frac{1}{\Im K_2} \right) e^{-2 \Im K_1 x}
668: \end{equation}
669: with
670: \begin{equation} \label{eigenvalues}
671: K_{\binom{1}{2}} \;=\; \pm i \sqrt{\mu_{c-a} - \mu -i\omega} -i \sqrt{\mu_{c-a}}\;.
672: \end{equation}
673: This spectrum (\ref{C_asympt}) is strongly peaked at the center, $\omega=0$, of
674: the band of
675: modes, $-2\sqrt{\mu\mu_{c-a}} < \omega < 2\sqrt{\mu\mu_{c-a}}$, that are
676: amplified in the convectively unstable regime. Thus, the aforementioned frequency
677: integral may be approximated by
678: \begin{eqnarray}
679: C_{lin}(x) = \int_{-\infty}^\infty\frac{d\omega}{2\pi} C_{lin}(x,\omega)
680: \sim \sqrt{\mu\mu_{c-a}} C_{lin}(x,\omega=0) =
681: \frac{\sigma^2}{4\sqrt{\mu}}\exp^{2 i K_1\left(\omega=0\right)x}\;.
682: \end{eqnarray}
683: The last equality follows from Eq.~(\ref{C_asympt}) at $\omega=0$.
684: Applying now the condition (\ref{lin-cub}) one obtains
685: \begin{equation}
686: \ell\;\sim\;\frac{1}{iK_1(\omega=0)} \ln{\frac{2^{1/2}\mu^{3/4}}{\sigma}}\;.
687: \end{equation}
688: Using $\mu_{c-a}/ \mu = (1+\Delta)^2$ in Eq.~(\ref{eigenvalues}) one sees that
689: $i K_1(\omega=0)=\sqrt{\mu}\left[1 - \sqrt{2\Delta} + {\cal O}(\Delta)\right]$
690: for $\Delta \ll 1$ so that finally at fixed $\mu$
691: \begin{equation}
692: \ell\;\sim\;-\left[1 + \sqrt{2\Delta} + {\cal O}(\Delta)\right]
693: \left(\ln{\sigma} + const \right)\; .
694: \end{equation}
695:
696: \end{appendix}
697:
698: \clearpage
699:
700: \begin{thebibliography}
701: \item
702: %------------------------------------------------------------------------------
703: \bibitem{CH93}
704: M.~C.~Cross and P.~C.~Hohenberg,
705: %{\it Pattern Formation Outside of Equilibrium},
706: Rev.~Mod.~Phys.~{\bf 65}, 851 (1993).
707: %------------------------------------------------------------------------------
708: \bibitem{D85}
709: R.~J.~Deissler,
710: %{\it Noise-Sustained Structure, Intermittency, and the Ginzburg-Landau Equation},
711: J.~Stat.~Phys.~{\bf 40}, 371 (1985).
712: %------------------------------------------------------------------------------
713: \bibitem{BAC}
714: K.~L.~Babcock, G.~Ahlers, and D.~S.~Cannell,
715: %{\it Noise-sustained structure in Taylor-Couette flow with through flow},
716: Phys.~Rev.~Lett.~{\bf 67}, 3388 (1991);
717: %{\it Noise amplification in open Taylor-Couette flow},
718: Phys.~Rev.~E {\bf 50}, 3670 (1994);
719: K.~L.~Babcock, D.~S.~Cannell, and G.~Ahlers,
720: %{\it Stability and noise in Taylor-Couette flow with through-flow},
721: Physica D {\bf 61}, 40 (1992).
722: %-----------------------------------------------------------------------------
723: \bibitem{Steinberg}
724: A.~Tsameret and V.~Steinberg,
725: Europhys.~Lett.~{\bf 14}, 331 (1991);
726: %{\it Noise modulated propagating patterns in a convectively unstable system},
727: Phys.~Rev.~Lett.~{\bf 67}, 3392 (1991);
728: %{\it Absolute and convective instabilities and noise-sustained structures in the
729: %Couette-Taylor system with an axial flow},
730: Phys.~Rev.~E {\bf 49}, 1291 (1994);
731: A.~Tsameret, G.~Goldner, and V.~Steinberg,
732: %{\it Experimental evaluation of the intrinsic noise in the Couette-Taylor
733: %system with an axial flow},
734: Phys.~Rev.~E {\bf 49}, 1309 (1994).
735: %------------------------------------------------------------------------------
736: \bibitem{MLK92}
737: H.~W.~M\"{u}ller, M.~L\"{u}cke, and M.~Kamps,
738: %{\it Transversal convection patterns in horizontal shear flow},
739: Phys.~Rev.~A {\bf 45}, 3714 (1992).
740: %-----------------------------------------------------------------------------
741: \bibitem{SR92}
742: W.~Sch\"opf and I.~Rehberg,
743: %{\it Amplification of Thermal Noise via Convective Instability in Binary-Fluid Mixtures},
744: Europhys.~Lett.~{\bf 17}, 321 (1992);
745: %{\it The Influence of Thermal Noise on the Onset of Traveling Wave Convection
746: %in Binary Fluid Mixtures: An Experimental Investigation},
747: J.~Fluid Mech.~{\bf 271}, 235 (1994).
748: %------------------------------------------------------------------------------
749: \bibitem{LR93}
750: M.~L\"ucke and A.~Recktenwald,
751: %{\it Amplification of Molecular Fluctuations into Macroscopic Vortices by
752: %Convective Instabilities},
753: Europhys.~Lett.~{\bf 22}, 559 (1993).
754: %------------------------------------------------------------------------------
755: \bibitem{SBH94}
756: J.~B.~Swift, K.~L.~Babcock, and P.~C.~Hohenberg,
757: %{\it Effects of Thermal Noise in Taylor-Couette Flow with
758: %Corotation and Axial Through-flow},
759: Physica A {\bf 204}, 625 (1994).
760: %------------------------------------------------------------------------------
761: \bibitem{Deissler94}
762: R.~J.~Deissler,
763: %{\it Thermally Sustained Structure in Convectively Unstable Systems},
764: Phys.~Rev.~E {\bf 49}, R31 (1994).
765: %------------------------------------------------------------------------------
766: \bibitem{NWS96}
767: M.Neufeld, D.~Walgraef, and M.~San Miguel,
768: %{\it Noise-sustained structures in coupled complex Ginzburg-Landau equations
769: %for a convectively unstable system},
770: Phys.~Rev.~E {\bf 54}, 6344 (1996).
771: %-----------------------------------------------------------------------------
772: \bibitem{LS97}
773: M.~L\"ucke and A.~Szprynger,
774: %{\it Noise Sustained Pattern Growth: Bulk Versus Boundary Effects},
775: Phys.~Rev.~E {\bf 55}, 5509 (1997).
776: %------------------------------------------------------------------------------
777: \bibitem{T97}
778: S.~P.~Trainoff,
779: %{Rayleigh-B\'enard Convection in the Presence of a Weak Lateral Flow},
780: PhD thesis, UCSB, 1997.
781: %------------------------------------------------------------------------------
782: \bibitem{SCMW97}
783: M.~Santagiustina, P.~Colet, M.~San Miguel, and D.~Walgraef,
784: %{\it Noise-Sustained Convective Structures in Nonlinear Optics},
785: Phys.~Rev.~Lett.~{\bf 79}, 3633 (1997).
786: %------------------------------------------------------------------------------
787: \bibitem{CWS99}
788: P.~Colet, D.~Walgraef, and M.~San Miguel,
789: %{\it Convective and Absolute Instabilities in the Subcritical Ginzburg-Landau
790: %Equation},
791: Eur.~Phys.~J.~B {\bf 11}, 517 (1999).
792: %-----------------------------------------------------------------------------
793: \bibitem{BersBriggs}
794: A.~Bers, in {\it Basic Plasma Physics I},
795: edited by A.~A.~Galeev and R.~N.~Sudan (North-Holland, New York, 1983);
796: R.~J.~Briggs, {\it Electron Stream Interaction with Plasmas}
797: (MIT Press, Cambridge, MA, 1964).
798: %-----------------------------------------------------------------------------
799: \bibitem{Huerre}
800: P.~Huerre and P.~A.~Monkewitz,
801: Annu.~Rev.~Fluid Mech.~{\bf 22}, 473 (1990);
802: J.~Fluid Mech.~{\bf 159}, 151 (1985);
803: P.~Huerre, in
804: {\it Instabilities and Nonequilibrium Structures}, edited
805: by E.~Tirapegui and D.~Villarroel (Reidel, Dordrecht, 1987), p.~141.
806: %-----------------------------------------------------------------------------
807: \bibitem{Chomaz92}
808: J.~M.~Chomaz,
809: %{\it Absolute and convective instabilities in nonlinear systems},
810: Phys.~Rev.~Lett.~{\bf 69}, 1931 (1992).
811: %-----------------------------------------------------------------------------
812: \bibitem{RLM93}
813: A.~Recktenwald, M.~L\"ucke, and H.~W.~M\"uller,
814: %{\it Taylor vortex formation in axial through-flow: Linear and weakly nonlinear analysis},
815: Phys.~Rev.~E {\bf 48}, 4444 (1993).
816: %------------------------------------------------------------------------------
817: \bibitem{BD92}
818: H.~Brand and R.~J.~Deissler,
819: %{\it Eckhaus and Benjamin-Feir instabilities near a weakly inverted bifurcation},
820: Phys.~Rev.~A {\bf 45}, 3732 (1992).
821: %------------------------------------------------------------------------------
822: \bibitem{SH92}
823: W.~van Saarloos and P.~C.~Hohenberg,
824: %{\it Fronts, Pulses, Sources and Sinks in Generalized Complex Ginzburg-Landau Equations},
825: Physica D {\bf 56}, 303 (1992).
826: %------------------------------------------------------------------------------
827: \bibitem{Pre94}
828: W.~H.~Press, S.~A.~Teukolsky, W.~T.~Vetterling and B.~P.~Flannery,
829: {\it Numerical Recipes in C},
830: Cambridge: Cambridge University Press (1994).
831: %------------------------------------------------------------------------------
832: \bibitem{R250}
833: S.~Kirkpatrick and E.~P.~Stoll,
834: J.~Comput.~Phys.~{\bf 40}, 517 (1981);
835: R.~C.~Tausworth, Math.~Comput.~{\bf 19} 201 (1965).
836: %------------------------------------------------------------------------------
837: \end{thebibliography}
838:
839:
840: \clearpage
841: %%%%%%%%%%%%%%%%%%%%%%%%%%%% Fig. 1 %%%%%%%%%%%%%%%%%%%%%%%%%%%%
842: \begin{figure}
843: %\includegraphics[scale=0.5]{figure1.eps}
844: \includegraphics[clip=true,width=12cm,angle=0]{figure1.eps}
845: \caption{\label{figure:Maxwell}
846: Convective-absolute instability boundaries (\ref{mu_conv}) for the unforced
847: cubic and quintic
848: GLE, respectively. For parameters below the respective curve front propagation
849: is such that in the absence of noise the $A$=0 state invades the $A\neq$0 state.
850: In the absolutely unstable parameter regime above the respective curve the
851: $A$=0 state recedes and the $A\neq$0 state expands (as long as the front is not
852: hindered by a boundary).}
853: \end{figure}
854: %%%%%%%%%%%%%%%%%%%%%%%%%%%% Fig. 2 %%%%%%%%%%%%%%%%%%%%%%%%%%%%
855: \begin{figure}
856: %\includegraphics[angle=0,scale=0.5]{figure2.eps}
857: \includegraphics[clip=true,width=9cm,angle=0]{figure2.eps}
858: \caption{\label{figure:R-kappa}
859: Deformation of the front solution $R(x)$ (a) of the
860: deterministic quintic GLE by the boundary condition $A(x=0,t)=0$ in the
861: absolutely unstable
862: regime for $\mu=0.05$ and $v=(1+\Delta)v_{c-a}$ as indicated.
863: The spatial growth rate $\kappa(x)=\partial_x ln{R(x)}$ (b) deviates from a
864: freely propagating front with wave number $q=0$ that would show \cite{SH92}
865: $\kappa_L=\frac{v}{2}+\sqrt{\frac{v^2}{4}-\mu}$ [thick dots in (b)] in the
866: small-amplitude "linear" part of the front.
867: (c) shows $\kappa$ versus $R$ in comparison with the prediction \cite{SH92}
868: $\kappa=\frac{1}{\sqrt{3}}\left(R_N^2 - R^2\right)$ for a
869: stationary front in an infinite system at $\Delta=0,\,\mu=0.05$ for which
870: $R_N^2$=1.048.
871: Thin dotted curves in (a)-(c) refer to a numerically obtained solution for
872: $\Delta=0$ at time $5\cdot 10^4$ which is not yet stationary. Here the profile is
873: still
874: moving to the right and in the absence of numerical "noise" we would expect this
875: transient to approach the $R\equiv 0$ basic state [c.f.~also Fig.~(c)].}
876: \end{figure}
877: %%%%%%%%%%%%%%%%%%%%%%%%%%%% >>>New<<< Fig. 3 %%%%%%%%%%%%%%%%%%%%%%%%%%%%
878: \begin{figure}
879: %\includegraphics[angle=0,scale=0.5]{figure3.eps}
880: \includegraphics[clip=true,width=12cm,angle=0]{figure3.eps}
881: \caption{\label{figure:V_3-5}
882: Reduced potentials V (\ref{V_3-5}) corresponding to the real cubic (dashed line)
883: and quintic (dotted line) GLE. For the cubic case the plot is independent of
884: $\mu, g_3$. For the quintic case it depends on the combination
885: $\tilde \mu = -g_5 \mu /g_3^2$. Here $\tilde \mu = 0.05$.}
886: \end{figure}
887: %%%%%%%%%%%%%%%%%%%%%%%%%%%% Fig. 4 %%%%%%%%%%%%%%%%%%%%%%%%%%%%
888: \begin{figure}
889: %\includegraphics[angle=0,scale=0.5]{figure4.eps}
890: \includegraphics[clip=true,width=14cm,angle=0]{figure4.eps}
891: \caption{\label{figure:l-dl}
892: Growth length $\ell$ (a) and its derivative $\partial\ell/\partial\Delta$ (b)
893: versus $\Delta$ near the convective-absolute threshold for the cubic GLE at
894: $\mu=0.05$ for
895: various noise strengths $\sigma$. Inset in (b) shows the variation of the peak of
896: $\partial\ell/\partial\Delta$ with $\sigma$.}
897: \end{figure}
898:
899: %%%%%%%%%%%%%%%%%%%%%%%%%%%% Fig. 5 %%%%%%%%%%%%%%%%%%%%%%%%%%%%
900: \begin{figure}
901: %\includegraphics[angle=0,scale=0.5]{figure5.eps}
902: \includegraphics[clip=true,width=13cm,angle=0]{figure5.eps}
903: \caption{\label{figure:l-s_qOm-cub}
904: Inverse of the standard deviations of frequency $s_\Omega$ (a),
905: wave number $s_q$ (b), and amplitude modulus $s_R$ (c) for the stochastic cubic
906: GLE. Results are
907: reduced by $\Sigma$ (\ref{Sigma_def}) or $\sqrt{<|A|^2>}$, respectively, and
908: plotted as functions of $\Delta$ for
909: three downstream locations $x=\ell/2$, $\ell$, and $2\ell$. (d) shows the growth
910: length $\ell$ of $\sqrt{<|A|^2>}$ together with its derivative
911: $\partial\ell/\partial\Delta$. Piecewise straight lines are guides to the eye.
912: Parameters are $\mu=0.05$ and $\sigma=2.5\cdot 10^{-5}$.}
913: \end{figure}
914: %%%%%%%%%%%%%%%%%%%%%%%%%%%% Fig. 6 %%%%%%%%%%%%%%%%%%%%%%%%%%%%
915: \begin{figure}
916: %\includegraphics[angle=0,scale=0.5]{figure6.eps}
917: \includegraphics[clip=true,width=13cm,angle=0]{figure6.eps}
918: \caption{\label{figure:l-s_qOm-quin}
919: Inverse of the standard deviations of frequency $s_\Omega$ (a),
920: wave number $s_q$ (b), and amplitude modulus $s_R$ (c) for the stochastic quintic
921: GLE. Results are
922: reduced by $\Sigma$ (\ref{Sigma_def}) or $\sqrt{<|A|^2>}$, respectively, and
923: plotted as functions of $\Delta$ for
924: three downstream locations $x=\ell/2$, $\ell$, and $2\ell$. (d) shows the growth
925: length $\ell$ of $\sqrt{<|A|^2>}$ together with its derivative
926: $\partial\ell/\partial\Delta$. Piecewise straight lines are guides to the eye.
927: Parameters are $\mu=0.05$ and $\sigma=10^{-3}$.}
928: \end{figure}
929: %%%%%%%%%%%%%%%%%%%%%%%%%%%% Fig. 7 %%%%%%%%%%%%%%%%%%%%%%%%%%%%
930: \begin{figure}
931: %\includegraphics[angle=0, scale=0.5]{figure7.eps}
932: \includegraphics[clip=true,width=13cm,angle=0]{figure7.eps}
933: \caption{\label{R-s_om-x}
934: Spatial variation of the standard deviation $s_\Omega$ of the frequency reduced by
935: $\Sigma$ (a) and of $\sqrt{<|A|^2>}$ (b) for the quintic GLE in the absolutely and
936: convectively regime at $\Delta = -0.1273$ and $\Delta = 0.0908$, respectively.
937: After the integration time of $T=5\cdot10^5$ used in this plot $s_\Omega$ was for
938: $\Delta = 0.0908$ not yet fully stationary in the growth region of
939: $\sqrt{<|A|^2>}$. Parameters are $\mu=0.05$ and $\sigma=10^{-3}$.}
940: \end{figure}
941: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
942:
943: \end{document}
944:
945: