nlin0505028/p.tex
1:                
2:  %\documentclass[preprint, aps, pre, preprintnumbers,amsmath,amssymb,showpacs]{revtex4}
3: \documentclass[preprint, aps, pre, eqsecnum,amsmath,amssymb,showpacs]{revtex4}              
4: %\documentclass[draft,twocolumn, aps, pre, eqsecnum,amsmath, amssymb,showpacs]{revtex4}
5: %\documentclass[draft,twocolumn, aps, pre, eqsecnum,amsmath, amssymb,showpacs]{revtex4}
6: 
7:  \usepackage[final]{graphicx}  % to include figures
8:  %\usepackage[draft]{graphicx}  % includes only space for figures
9: 
10:  \renewcommand{\arraystretch}{1.5}  %faktor 1.5 fuer tabellenzeilenabstand                                             
11: 
12:  \newcommand{\Nabla}{\mbox{\bf\boldmath $\nabla$}}                              
13:  \newcommand{\ve}[1]{\stackrel{\rightharpoonup}{#1}}                                                             
14:  \renewcommand{\vec}[1]{{\bf #1}}                                                
15:  \newcommand{\Vxi}{\mbox{\bf\boldmath $\xi$}}
16:  \newcommand{\Vphi}{\mbox{\boldmath $\varphi$}}
17:  \newcommand{\Vzeta}{\mbox{\boldmath $\zeta$}}                   
18:  \newcommand{\VPsi}{\mbox{\boldmath $\psi$}}
19:  \newcommand{\Overrightarrow}[1]{\stackrel{\textstyle\rightarrow}{#1}}
20: 
21:  % info und seite in kopfzeile
22:  \pagestyle{myheadings}                                                          
23:  % datum in kopfzeile. falls 'twoside': {leftside}{rightside}
24:  \markboth{\today}{\today}                                                       
25: 
26:  \bibliographystyle{prsty}                                                       
27: 
28:  \begin{document}
29: 
30:  \title{\bf 
31:   Traveling Wave Fronts and Localized Traveling Wave Convection in Binary Fluid 
32:   Mixtures}                                
33: 
34:  \author{D.~Jung and M.~L\"ucke}                                   
35:  \affiliation{Institut f\"ur Theoretische Physik, Universit\"at des Saarlandes,      
36:  Postfach 151150, \\ D-66041 Saarbr\"ucken, Germany}                              
37:    
38:  \date{\today}                                                                   
39: 
40:  \begin{abstract}
41:   Nonlinear fronts between spatially extended traveling wave convection (TW) 
42:   and quiescent fluid and spatially localized traveling waves (LTWs) are 
43:   investigated in quantitative detail in the bistable regime of binary fluid 
44:   mixtures heated from below. 
45:   A finite-difference method is used to solve the full hydrodynamic field 
46:   equations in a vertical cross section of the layer perpendicular
47:   to the convection roll axes. Results are presented for ethanol-water
48:   parameters with several strongly negative separation ratios where TW solutions
49:   bifurcate subcritically. Fronts and LTWs
50:   are compared with each other and similarities and differences are elucidated.
51:   Phase propagation out of the
52:   quiescent fluid into the convective structure entails a unique selection of
53:   the latter while fronts and interfaces where the phase moves into the
54:   quiescent state behave differently. Interpretations of various experimental
55:   observations are suggested. 
56:  \end{abstract}
57: 
58: 
59:  %\pacs{PACS number(s): 47.20.-k, 47.10.+g, 47.27.Te, 47.54.+r}
60:   \pacs{47.20.-k, 47.54.+r, 44.27.+g, 47.20.Ky}
61:  %****************************************************************
62:  % 47.20.-k : Hydrodynamic stability
63:  % 47.10.+g : General theory (of fluid dynamics)
64:  % 47.27.Te   Convection and heat transfer
65:  % 47.54.+r   Pattern selection; pattern formation  
66:  % 44.27.+g   Forced convection 
67:  % 47.20.Ky   Nonlinearity (including bifurcation theory) 
68:  % 05.45.-a   Nonlinear dynamics and nonlinear dynamical systems
69:  % 05.65.+b   Self-organized systems
70:  %****************************************************************
71:  \maketitle                                                                  
72:  %\tableofcontents
73:  %\newpage
74:  % skip 2/6inch
75:  \vskip2pc
76: 
77:  %im gegenstz zu \widetext (z.b. bei formeln)
78:  %halbe zeitenbreite f"ur spaltentext
79:  %\narrowtext
80:  
81: %---------------------Sec. I ------------------------------------------ 
82: \section{INTRODUCTION}
83: %----------------------------------------------------------
84: 
85: Many nonlinear dissipative systems that are driven sufficiently far away from 
86:  thermal equilibrium show selforganization out of an unstructured state: A 
87:  structured one can appear that is characterized by a (spatially extended) 
88: pattern which retains some of the symmetries of the system \cite{CH93}.
89:  Convection in binary miscible fluids like ethanol-water, $^3$He 
90: $-^4$He, or various gas mixtures is an example of such systems. It shows rich 
91: and interesting pattern formation behavior and it is paradigmatic for problems 
92: related to instabilities, bifurcations, and
93: selforganization with complex spatiotemporal behavior.
94: 
95: Compared to convection in one-component 
96: fluids like, e.g., pure water the spatiotemporal properties are far more 
97: complex. The reason is that concentration variations 
98: which are generated via thermodiffusion, i.e., the Soret effect by 
99: externally imposed and by internal temperature gradients influence the buoyancy,
100: i.e., the driving force for convective flow. The latter in turn mixes by
101: advectively redistributing concentration. This nonlinear advection
102: gets in developed convective flow typically much larger than the smoothening 
103: by linear diffusion --- P\'eclet numbers measuring the strength
104: of advective concentration transport relative to diffusion are easily 
105: $\cal O$(1000). Thus, the concentration balance is strongly nonlinear 
106: giving rise to strong variations of the concentration field and to
107: boundary layer behavior. In contrast to that, momentum and 
108: heat balances remain weakly nonlinear close to onset as in pure fluids implying
109: only smooth and basically harmonic variations of velocity and temperature fields
110: as of the critical modes. 
111: 
112: Without the thermodiffusive Soret coupling between temperature and
113: concentration any initial concentration deviation from the mean diffuses away 
114: and influences no longer the balances of the other fields. 
115: Hence, the feedback interplay between ({\em i}) the Soret generated 
116: concentration variations, ({\em ii}) the resulting modified buoyancy, and
117: ({\em iii}) the 
118: strongly nonlinear advective transport and mixing causes binary mixture 
119: convection to be rather complex with respect to its spatiotemporal
120: properties and its bifurcation behavior.
121: Take for example the case of negative Soret coupling, $\psi < 0$, 
122: between temperature and concentration fields \cite{coupling} when the lighter
123: component migrates to the colder regions thereby stabilizing the density 
124: stratification in the quiescent, laterally homogeneous conductive fluid state.
125: Then the above 
126: described feedback interplay generates oscillations. In fact the buoyancy 
127: difference in regions
128: with different concentrations was identified already in \cite{WKPS85} as the 
129: cause for traveling wave convection.
130: 
131: Oscillatory convection appears in the form of the
132: transient growth of convection at supercritical heating, in spatially
133: extended  
134: nonlinear traveling wave (TW) and standing wave solutions that branch 
135: in general subcritically
136: out of the conductive state via a common Hopf bifurcation, in spatially
137: localized traveling wave (LTW) states, and in various types of fronts. TW and LTW convection 
138: has been studied experimentally and theoretically  for some time
139: \cite{CH93,Moses87Heinrichs87,Behringer9091,Winkler92,Kolodner94,Platten96,Kaplan94,Surko91,
140:       Aegert01,Ning97b,Batiste01,Jung02}.
141: The transient oscillatory growth of convection was investigated by
142: numerical simulations \cite{FL}. Nonlinear standing wave solutions were obtained 
143: only recently \cite{MJL04,JML04}. Freely propagating convection fronts that
144: connect subcritically bifurcating nonlinear TW convection with the {\em stable}
145: quiescent fluid do not seem to have been investigated in detail beyond some
146: first preliminary results \cite{Bensimon90,Kolodner92Fronts,Bu99}. Here we determine such fronts in
147: quantitative detail and compare their properties with those of LTWs.
148: 
149:  In narrow rectangular and annular channels convection occurs in 
150: the form of rolls with axes oriented perpendicular to the long sidewalls 
151: \cite{Surko87,CH93}. These stuctures can efficiently be described in the two
152: dimensional vertical $x-z$ cross section in the middle of the channel 
153: perpendicular to the roll axes ignoring variations in axis direction. 
154: Furthermore, these convection structures have relevant phase gradients only in  
155: $x$-direction thus causing effectively one dimensional patterns \cite{AlBa04}. 
156: 
157: When comparing 
158: experiments with analytical calculations or numerical simulations
159: performed under the above described conditions it is useful to do that on the
160: basis of reduced Rayleigh numbers, $r=R/R^0_c$, with $R^0_c$ being the critical
161: one for onset of pure fluid convection for the respective experiment, analytical
162: method, or numerical method. This significantly reduces the dependence 
163: of, say, the bifurcation diagrams of convective states on the specific
164: geometry of the respective set-up. In laterally unbounded systems the analytical
165: value for $R^0_c$ is 1707.762.
166: 
167: %---------------------------------------------------------
168: {\em Localized traveling waves.} \hspace{0.3cm}
169: %---------------------------------------------------------
170: For weak negative Soret coupling one has observed in experiments a competition 
171: between homogeneous laterally extended TW convection and so-called dispersive 
172: chaos with an irregular repetitive formation and collapse of spatially localized
173: TW pulses \cite{Kolodner90Glazier91,Kaplan94}. During the pulse formation their 
174: drift velocities can drop abruptly 
175: to about a tenth of the initial group velocity \cite{Kolodner90Glazier91}.
176: We consider this to be a characteristic signal that the lateral redistribution
177: of concentration over the pulse \cite{Jung02} becomes important and that the 
178: strongly nonlinear dynamics sets in. For more negative $\psi \lesssim -0.06$ the 
179: collapse is in general less dramatic. There, convection is dominated 
180: by isolated strongly peaked localized states. Eventually, at $\psi\simeq -0.07$ 
181: a regime is reached where stable LTWs coexist near onset with extended 
182: TWs \cite{Barten95I,Barten95II,Luecke98,Moses87Heinrichs87,Kolodner88,
183: Kolodner91a,Kolodner94}.
184: Increasing the Soret coupling strength further to more negative $\psi$  
185: the band ($r^{LTW}_{min},r^{LTW}_{max}$) of Rayleigh numbers in which stable 
186: LTWs exist increases monotonically while shifting upwards as a whole ---  
187: $r^{LTW}_{max}(\psi)$ grows stronger than $r^{LTW}_{min}(\psi)$. Simultaneosly,
188: the lower band limit for the existence of extended TW states 
189: $r^{TW}_{min}(\psi)$, i.e., the lowest saddle-node of TWs increases even steeper
190: so that eventually for $\psi \lesssim -0.4$ the complete LTW band comes to lie 
191: below the existence range of TWs, $r^{LTW}_{max} \leq r^{TW}_{min}$
192: \cite{Jung02}.
193: 
194: LTWs consist of slowly drifting, spatially confined convective regions that are 
195: embedded in the quiescent fluid.
196: These intriguing structures have been investigated in experiments
197: \cite{Moses87Heinrichs87,Kolodner88,Bensimon90,Niemela90,Behringer9091,
198: Kolodner9091,Steinberg91,Kolodner91a,
199: Surko91,Kolodner91c,Kolodner91d,Kolodner93II,Kolodner94} and numerical simulations
200: \cite{Barten91,Barten95II,Luecke98,Jung02}. A discussion of
201: various theoretical models aiming at their explanation is contained in 
202: Sec.~\ref{SEC:LTWmodels}. Roll vortices grow in a LTW structure out
203: of the quiescent fluid at one end, travel with spatially varying phase velocity 
204: $v_p(x)$ to the other end, and decay there back into the basic state. The two 
205: interfaces to conduction and with it the whole convective region move with 
206: constant, uniquely selected drift velocity $v_d$. The latter is a 
207: function of $r, \psi$ with magnitude much smaller than the phase velocities.
208: Also the oscillation frequency of the LTW is uniquely selected; it is constant 
209: in space and time in the frame that is comoving with its drift velocity. 
210: And finally, the length $l(r,\psi)$ of the convective region of stable LTWs and 
211: their spatial stucture are uniquely selected. This length grows with increasing
212: heating $r$. 
213: 
214: A central role for the stable existence of LTWs plays a large-scale
215: mean concentration current. Extending over the whole LTW it redistributes
216: concentration and thereby changes the buoyancy in a decisive way \cite{Luecke98}.
217: This effect 
218: can sustain LTWs even at low $r$ where no extended TWs exist \cite{Jung02}.
219: 
220: %-----------------------------------------------------------------
221: {\em Blinking states in rectangular channels.} \hspace{0.3cm}
222: %-----------------------------------------------------------------
223: The LTW confinement of convection occurring in translationally invariant annular 
224: channels is obviously an inherent process of the hydrodynamic balances.
225: But one has also observed end-wall-assisted or at least end-wall-modified 
226: confinement of convection close to the ends of rectangular channels.
227: The weakly nonlinear varieties of such a confinement can largely be understood
228: in terms of the convective behavior
229: of TW packets, their reflection properties at the end walls, and the 
230: destructive interaction 
231: between left and right traveling patterns \cite{Cross868889,Brand86,Fineberg90}.
232: These effects give rise near onset to a wide range of weakly nonlinear and 
233: effectively low dimensional spatiotemporal behavior that depends sensitively
234: on the specific experimental set-up like, e.g., the end-wall boundary
235: conditions and the system length  
236: \cite{Kolodner8889,Fineberg88Steinberg89,Bestehorn89,Kolodner93,Batiste01}.
237: While the linear eigenmodes of such sytems ('linear counterpropagating 
238: waves' or 'chevrons') \cite{Kolodner86,Surko87,Kolodner87,Batiste01} 
239: are laterally symmetric or antisymmetric
240: localization sets in via a temporal amplitude modulation. Thereby
241: convection is alternatingly weakened and enhanced in the left and the right part
242: of the system part giving rise to a 'blinking' state  
243: \cite{Surko87,Kolodner8889,Fineberg88Steinberg89,Steinberg91,Kolodner93}.
244: The so-called 'chaotic blinking' states \cite{Kolodner8889,
245: Fineberg88Steinberg89,Steinberg91,Kolodner93} seem to be the 
246: analogue of the 'chaotic dispersive' pulse formation in annular containers 
247: \cite{Steinberg91,Kaplan94}. Also 'blinking' modes with different 
248: frequencies at
249: both ends of the channel were observed \cite{Fineberg88Steinberg89,Steinberg91}.
250: But their possible relation to a large-scale mean concentration variation 
251: \cite{Moses86} produced by nonlinear propagating waves in a finite cell has not 
252: been discussed.
253: 
254: %--------------------------------------------------------
255: {\em Wall-attached structures.} \hspace{0.3cm}
256: %--------------------------------------------------------
257: At larger $r$ one has observed wall-attached TW structures with amplitudes
258: confined to the vicinity of one or both end walls. These wall-attached 
259: convective patches 
260: \cite{Moses87Heinrichs87,Kolodner8889,Fineberg88Steinberg89,Kolodner90,Kolodner91b,Yahata91,
261: Ning96,Ning97a} are closely related to free LTWs \cite{Niemela90}. 
262: They are strongly nonlinear as indicated by their low frequency 
263: \cite{Kolodner8889,Fineberg88Steinberg89}.
264: Moreover, their spatial structure and their region of existence is largely
265: unaffected by the details
266: of the lateral boundaries or by the container length in contrast to the linear
267: and weakly nonlinear behavior described above \cite{Fineberg88Steinberg89}. The 
268: more extensive wall-attached structures show some similarities with front-like
269: states. Note, however, that here the source or the sink of the propagating rolls
270: is pinned near a wall and the interface to the quiescent fluid in the bulk of the
271: channel does not move \cite{Kolodner90}.
272: 
273: %-------------------------------------------------------- 
274: {\em Our numerical simulations.} \hspace{0.3cm}
275: %--------------------------------------------------------
276: Our numerical simulations have been performed in order to elucidate in
277: quantitative detail the properties of relaxed nonlinear TW convection structures
278: that contain an interface (or two of them) to the
279: quiescent fluid as an integrated structural element. We compare
280: for a wide range of Soret coupling strengths front states and LTW states
281: showing what they have in common and how they differ. We
282: focus our interest to those parameters where the quiescent conductive state of
283: the fluid is stable and where the solutions describing spatially extended,
284: laterally periodic TW convection bifurcate subcritically out of it.
285: 
286: The system we have in mind is a binary fluid layer of thickness $d$
287: which is bounded by two solid horizontal plates perpendicular to 
288: the gravitational acceleration $\vec{g}$.
289: The fluid might be a mixture of water with the lighter component ethanol at a
290: mean concentration $\overline{C}$.
291: It is heated from below. The temperatures at the plates are 
292:  $\overline{T} \pm \Delta T/2$.
293: The variation of the fluid density $\rho$ due to temperature and
294: concentration variations is governed by the linear thermal and solutal
295: expansion coefficients
296: $ \alpha = - \frac{1}{\rho}\frac{\partial\rho}{\partial {T}} $ and
297: $ \beta = - \frac{1}{\rho}\frac{\partial\rho}{\partial {C}} $, respectively.
298: Both are positive for ethanol-water.  The solutal diffusivity
299: of the binary  mixture is $D$, its thermal diffusivity is $\kappa$, and its
300: viscosity is $\nu$.
301: Length and time is scaled by $d$ and $d^2/\kappa$, respectively, so that
302: velocity is measured in units of $\kappa/d$.
303: Temperatures are reduced by the vertical temperature
304: difference $\Delta T$ across the layer and concentrations
305: by $\frac{\alpha}{\beta}\Delta T$.
306: The scale for the pressure is given by $\frac{\rho\kappa^2}{d^2}$.
307: 
308: Then, the balance equations for mass, momentum, heat, and concentration
309:   \cite{LL66,Platten84} read in Oberbeck--Boussinesq approximation \cite{Barten95I}
310:  \begin{subequations}
311:  \label{eq:baleqs}
312:  \begin{eqnarray}
313:  \Nabla \cdot \vec{u} = 0 \label{eq:baleqmass}\\
314:  \partial_t\, {\bf u}   =    - {\bf \mbox{\boldmath $\nabla$} }\,
315: ({\bf u : u}  +  p
316: - \sigma \, {\bf \mbox{\boldmath $\nabla :  $}\,\, u }) +
317: {\bf B} \,\,\,;\,\,\,  
318: {\bf B}    =    \sigma\, R\, (\delta T + \delta C) {\bf e}_z \label{eq:baleqveloc}\\ 
319:  \partial_t\delta T  =  
320:   - \Nabla \cdot \left[ \vec{u}\delta T - \Nabla \delta T\right]\label{eq:baleqheat}\\
321:  \partial_t \delta C  = 
322:   - \Nabla \cdot \left[ \vec{u} \delta C - L\Nabla\left(\delta C -\psi \delta T\right) \right]\ .
323:   \label{eq:baleqconc}
324:  \end{eqnarray}
325:  \end{subequations}
326: Here, $\delta T$ and $\delta C$ denote deviations of
327: the temperature and concentration fields, respectively, from their 
328: mean $\overline{T}$ and $\overline{C}$ and $\bf B$ is the buoyancy.
329: The Dufour effect \cite{HLL92,HL95}
330: that provides a coupling of concentration gradients into
331: the heat current and a change of the thermal diffusivity
332: is discarded in (\ref{eq:baleqheat}) since it is relevant
333: only in few binary gas mixtures \cite{LA96} and possibly in
334: liquids near the liquid--vapor critical point \cite{LLT83}.
335: 
336: Besides the Rayleigh number $R=\frac{\alpha g d^3}{\nu \kappa}\Delta T$
337: measuring the thermal driving of the fluid three additional
338: numbers enter into the field equations: the Prandtl number
339: $\sigma=\nu/\kappa$, the Lewis number $L=D/\kappa$, and the separation ratio 
340: $\psi=-\frac{\beta}{\alpha}\frac{k_T}{\overline{T}}= 
341: - S_T\overline{C}(1-\overline{C})\frac{\beta}{\alpha}$. 
342: Here $k_T = \overline{T}\,\overline{C}(1-\overline{C})S_T$ is the thermodiffusion coefficient
343: \cite{LL66} and $S_T$ the Soret coefficient. They measure changes
344: of concentration fluctuations due to temperature gradients in the fluid.
345: $\psi$ characterizes the sign and the strength of the Soret effect. 
346: Negative Soret coupling $\psi$ (i.e., positive $S_T$ for mixtures like ethanol
347: water with positive $\alpha$ and $\beta$) induces concentration gradients of 
348: the lighter component  that are antiparallel to temperature gradients. 
349: In this situation, the buoyancy induced by solutal changes in density is opposed
350: to the thermal buoyancy.
351: 
352: When the gradient of the total buoyancy exceeds a threshold, convection sets in
353:  --- typically
354: in the form of straight rolls. For sufficiently negative $\psi$ the primary 
355: instability is oscillatory. Ignoring field
356: variations along the roll axes we describe here 2D convection in an
357: $x$--$z$ plane perpendicular to the roll axes with a velocity field
358: \begin{equation}
359: \vec{u}(x,z,t) = u(x,z,t)\,\vec{e}_x + w(x,z,t)\,\vec{e}_z \, .
360: \end{equation}
361: 
362: To find the time-dependent solutions of the above partial differential equations
363: subject to realistic horizontal boundary conditions \cite{Barten95I}
364: we performed numerical simulations with a modification of 
365: the SOLA code that is based on the MAC method \cite{MAC-SOLA,PT83}. 
366: This is a finite-difference method of second order in space formulated on 
367: staggered grids for the different fields. The Poisson equation for the pressure 
368: field that results from taking the divergence of (\ref{eq:baleqveloc}) was solved
369: iteratively with the artificial compressibility   
370: method \cite{PT83} by incorporating a multi-grid technique. 
371:   
372: Throughout this paper we consider mixtures with $L=0.01$, $\sigma=10$, 
373: and various negative values of $\psi$ that are easily accessible with 
374: ethanol-water experiments. The paper is organized as follows: In 
375: Sec.~\ref{Sec:Fronts} we first describe our methods for characterizing the
376: various convective states. Then we present results for the two different types of
377: TW front states that can arise in laterally homogeneous mirror symmetric systems
378: with either the phase propagating out of the quiescent fluid or into it. Also
379: transient two-front structures are discussed. Sec.~\ref{SEC:LTW} deals with LTW
380: states and their relation to fronts. The transient dynamics towards the selected
381: LTW, the stabilization via front repulsion, the difference between long and short
382: LTWs, and a critical appraisal of LTW models are topics covered here. In 
383: Sec.~\ref{SEC:Compare-exp} we present a comparison with experiments and a
384: discussion. The last section contains a conclusion.
385: 
386: %\clearpage
387:  %-------------- Sec. II --------------------------------
388:  \section{Fronts} \label{Sec:Fronts} 
389:  %-------------------------------------------------------
390: 
391: Here we discuss front solutions where part of the system is occupied by the 
392: quiescent fluid while the other one shows fully developed, saturated, strongly 
393: nonlinear TW convection with laterally homogeneous amplitude. 
394: Strictly speaking these two states are realized only in the two opposing limits
395: of $x\to \pm \infty$.
396: We focus our investigation of fronts on parameters where the quiescent fluid 
397: state is stable and where the TW solutions bifurcate subcritically out of the
398: conductive state. Then, any {\it linear} growing and spreading of 
399: infinitesimal, localized convective perturbations in the
400: quiescent fluid which could possibly dominate the low amplitude behavior 
401: of fronts as in the case of an unstable zero amplitude state 
402: \cite{vSaarloos9092,vSaarloos03} is absent.
403: 
404: Little general is known about pattern forming fronts in real bistable systems
405: \cite{CH93}. Most of the research activities were centered on fronts in the 
406: quintic Ginzburg-Landau equation \cite{vSaarloos9092,vHvSHo93,CoWaSMi99,CoCh01,
407: SzLu03,Rabaud03}. 
408: One can expect that the front properties are fixed by a strongly 
409: nonlinear eigenvalue problem 
410: describing a heteroclinic orbit between the two involved states.
411: Some of these front solutions will be unstable. There might be also  
412: multistable coexistence of fronts so that depending on initial conditions and 
413: on the history of the (control) parameters different fronts could finally be 
414: realized. We call a front uniquely selected when our numerical simulations 
415: indicated that different formation processes ended in the same front for a fixed
416: parameter combination.  
417: 
418: Fronts can be classified into coherent and incoherent ones \cite{vSaarloos03}.
419: We focus here on the first kind which in our system are characterized as 
420: strictly time periodic 
421: states in a frame that is comoving with the front's velocity $v_F$.  
422: Such a front state being monochromatic is a global nonlinear mode. Its 
423: frequency is an eigenvalue, i.e., a global constant in space and time so that 
424: the convection oscillations have everywhere the same period.
425: 
426: Thus, we do not consider here, e.g., complex large scale or chaotic 
427: spatiotemporal interface behavior. The coherent 
428: fronts of the various hydrodynamic fields and quantities in this paper have 
429: a smooth and basically monotonous profile which connects  
430: the quiescent fluid with the nonlinear saturated extended TW. The transition 
431: region between conduction and convection that is characterized by large 
432: amplitude variations is quite short and consists typically 
433: only of about 3 -4 convection rolls. We call this transition region also
434: the interface between conduction and convection.
435: 
436: If the selected front pattern is incompatible with any stable bulk structure 
437: there are two possibilities:
438: ({\it i}) A perturbation in the unstable convection bulk grows and expands 
439: towards 
440: the interface. This would destroy front coherence and could lead to more 
441: complex large scale variations, perhaps chaotic spatiotemporal behavior.
442: ({\it ii}) The interface region is only convectively unstable against
443: perturbations of the bulk nonlinear TW. Then initially localized
444: perturbations would be advected out of every finite system and could not  
445: reach the unperturbed interface region of the front.
446: 
447: %-------------- Sec. II A ----------------------------
448:  \subsection{Methods of characterization} \label{SEC:FRONT-charact}
449: %---------------------------------------------------
450: 
451: %-------------- Sec. II A1 ----------------------------
452:  \subsubsection{Definitions} \label{SEC:FRONT-def}
453: %---------------------------------------------------
454: 
455: We call a front to be of type $+$ when its envelope grows at $x=-\infty$ out 
456: of the basic quiescent state.
457: Otherwise it is a $-$front. Then the amplitude falls to zero at $x=+\infty$ 
458: \cite{Buechel00}. The phase of the convection pattern in a front state of type
459: $+$ can either propagate to the left or to the right and similarly for the 
460: $-$front state. Hence, one would have 
461: to discuss four front states separately. However, because of the invariance 
462: of the system under $x \to -x$  
463: a $+$front state with positive (negative) phase velocity $v_p$ is the mirror
464: image of the $-$front state with negative (positive) $v_p$. Therefore, it
465: suffices to consider only the front states that consist of roll vortices 
466: traveling, say, in positive $x$-direction and to use the superscript $+$ or
467: $-$ to identify the properties of the front in question in a unique way. So,
468: the phase velocities of all oscillatory convective structures investigated in 
469: this paper are positive. We call the direction of positive $x$ into which
470: the phase propagates also 'downstream' and the opposite one 'upstream'.
471: 
472: So, to sum up our notation: In a $+$front state the quiescent fluid is located
473: 'upstream' and a source of phase with the latter propagating out of the
474: conductive state into convection. In a $-$front the quiescent fluid is located 
475: in 'downstream' direction and a sink since phase moves out of convection 
476: into conduction. 
477: 
478: Fig.~\ref{Fig:frontpics} shows fronts of each type. Under the $+$front (left
479: half of Fig.~\ref{Fig:frontpics}) convection rolls grow out of the
480: quiescent fluid and saturate in a 'downstream' bulk TW. On the other hand,
481: a $-$front (right half of Fig.~\ref{Fig:frontpics}) annihilates
482: roll vortices. In this process their phase velocity is accelerated [cf. the 
483: increase in the lateral profile of $v_p(x)$ in Fig.~\ref{Fig:frontpics}(f)] 
484: and they are stretched horizontally. 
485: 
486: It is clear from Fig.~\ref{Fig:frontpics} that the quiescent (convecting) region
487: expands into the convecting (quiescent) one when the velocity $v^+_F$ of the 
488: $+$front is positive (negative) and vice versa for the $-$front.
489: 
490: %-------------- Sec. II A2 ----------------------------
491:  \subsubsection{Mixing number}  \label{SEC:FRONT-mix}
492: %---------------------------------------------------
493: 
494: In order to monitor how well the fluid is mixed along the front we always
495: determined for the relaxed front states the mixing number 
496: \begin{equation} \label{Eq:M(x)}
497: M(x)= 
498: \left[< \overline{(\delta C)^2} > /  \overline{ (\delta C_{cond})^2}\right]^{1/2}
499: \end{equation}
500: as a function of lateral position $x$. It basically
501: measures the mean square of the deviations $\delta C(x,z,t)$ of the concentration
502: field from its global mean: the overbars imply a vertical average and the
503: brackets a temporal average at the specific horizontal location $x$ in the
504: frame comoving with the front velocity $v_F$.
505: The subscript {\it cond} denotes the reference quiescent 
506: conductive state with its linear vertical concentation variation. The mixing
507: number is defined such that $M=0$ 
508: in a perfectly mixed fluid and $M=1$ in the quiescent state.
509: 
510: In laterally extended TWs $\omega$ and with it $v_p$ increase when the 
511: concentration variations become larger \cite{Barten95I,Luecke98}. In fact, there
512: is a
513: universal scaling relation between $M$ and $\omega$ \cite{HoBuLu97} which shows
514: that $M$ and $v_p$ are linearly related to each other.
515: This relation also holds for the bulk part of front states far away from the
516: interface where the convection is TW-like with only slow spatial amplitude 
517: variation (Fig.~\ref{Fig:frontpics}).
518: 
519: %-------------- Sec. II A3 ----------------------------
520:  \subsubsection{Concentration current}  \label{SEC:FRONT-current}
521: %---------------------------------------------------
522: 
523: The phase shift between the concentration and velocity waves in the TW-like 
524: bulk of the front states sustains as in extended TW states a mean lateral
525: concentration current $<\vec{J}>(x,z)$ \cite{Linz88,Barten89,Barten95I,Luecke98}:
526: \begin{equation} \label{Eq:J}
527: <\vec{J}>= < \vec{u} \delta C - L \Nabla (\delta C - \psi \delta T) >
528: \end{equation}
529: where $\vec{u}$ is the velocity field and $\delta T$ the temperature deviation 
530: from the global mean. Again, the brackets imply a temporal average in the 
531: frame that is comoving with the front velocity. The Lewis number $L=0.01$ 
532: being rather small in our simulations
533: implies that $<\vec{J}>$ is dominated by the advective contribution except in
534: those boundary regions in which  $\vec{u}$ becomes small. 
535: 
536: The vertical variation of $<\vec{J}>$ is such that positive (negative)
537: $\delta C$ is transported in phase direction in the upper (lower) half of the 
538: layer. This transport causes a large-scale concentration redistribution in a 
539: front state between its TW bulk and its interface to the quiescent fluid and it 
540: is responsible for the different characteristic structures of the interfaces in
541: a $+$ and a $-$front as we will see further below.
542: 
543: %-------------- Sec. II A4 ----------------------------
544:  \subsubsection{Preparation and lateral boundary conditions}
545:  \label{SEC:FRONT-prep}
546: %---------------------------------------------------
547: 
548: We simulated systems containing up to 160 rolls.
549: The initial state was prepared by filling one half of the sytem with a nonlinear 
550: TW that was previously generated with periodic boundaries to have some fixed 
551: wavelength $\lambda$. The other half contained the stable temperature and 
552: concentration distribution of the pure quiescent basic state.
553: 
554: To simulate $+$fronts in infinite systems that connect to developed TW convection 
555: with some wavelength $\lambda$ far away from the interface 
556: between conduction and convection we imposed at the 'downstream' boundary $x=L$ 
557: of our computation domain the periodicity condition $f(L)=f(L-\lambda)$. 
558: For the case of $-$fronts we found that imposing the analogous condition at the
559: 'upstream' boundary of the developed TW part at $x=-L$ typically will introduce 
560: perturbations that can grow in 'downstream'
561: direction for example when the TW region is Eckhaus unstable.
562: The different aspects of the stability of $+$ and $-$fronts are discussed 
563: further below in the paper.
564:  
565: After a relaxation time of typically 100 to 200 
566: vertical thermal diffusion times 
567: we then could observe under certain conditions a coherent front state connecting 
568: a quiescent region of the system to a TW with asymptotic wavelength $\lambda$.
569: Here the fact that the frequency $\omega$ of such a coherent front state is 
570: constant in space and time in the  
571: frame that is comoving with the front velocity $v_F$ proved to be a good 
572: relaxation criterion to effectively determine whether such a state had been
573: obtained. 
574: 
575: 
576: %-------------- Sec. II B ----------------------------
577:  \subsection{$+$Fronts}  \label{SEC:+front}
578: %---------------------------------------------------
579: 
580: %-------------- Sec. II B1 ----------------------------
581:  \subsubsection{Structure and dynamics}  
582: %---------------------------------------------------
583: 
584: As soon as the growing convection rolls in a $+$front have become sufficiently 
585: nonlinear, i.e., when their lateral flow velocity $u$ has grown up to about 
586: their phase
587: velocity $v_p$ [e.g., close to the vertical arrow in Fig.~\ref{Fig:frontpics}(c)]
588: they start to alternatingly suck in positive ('blue') and negative ('red') 
589: $\delta C$ from the top and bottom concentration boundary layers, respectively. 
590: It is transported away into the well mixing convection bulk and replaced at the
591: interface location by neutral ('yellow/green') $\delta C$. Note that increasing
592: $u$ beyond $v_p$ causes the appearence of closed streamlines of the velocity
593: field in the frame comoving with the phase velocity of a traveling roll 
594: \cite{Linz88,Luecke98,JML04}. These closed streamlines regions
595:  are responsible for the characteristic roll
596:  structure of the $C$ field in Fig.~\ref{Fig:frontpics}(a): Positive (negative) 
597:  $\delta C$ is collected from the top (bottom) boundary
598:  layers and transported within the homogeneously mixed closed streamline 
599:  regions in phase direction while mean concentration, $\delta C \simeq 0$, is
600:  advected along the meandering "green-yellow stripe" in 
601:  Fig.~\ref{Fig:frontpics}(a) to the left \cite{Jung02}. The mean concentration 
602:  current   $<\vec{J}>$ resulting from this complicated concentration
603:  redistribution is shown in Fig.~\ref{Fig:frontpics}(i). All in all, mean
604:  concentration is accumulated (depleted) at the $+$ ($-$) front interface. 
605:  
606: The concentration redistribution reduces at the interface of the $+$front the
607: Soret-induced solutal 
608: stabilization that occurs to the left of it   
609: as a result of the large conductive vertical concentration gradient: at the
610: interface one can observe a minimal mixing number [Fig.~\ref{Fig:frontpics}(e)] 
611: and with it a buoyancy overshoot [Fig.~\ref{Fig:frontpics}(g)] which is 
612: sufficiently large to sustain local convection
613: growth there and cause even invasion of convection into the quiescent region whenever
614: $v_F^+<0$. With the fluid being well mixed there, i.e., with $M$ being small 
615: the local phase velocity is also small there -- in fact the minimum of $v_p(x)$
616: in Fig.~\ref{Fig:frontpics}(e) lies close to the one in $M$. 
617: 
618: Since the strongly stable quiescent fluid to the left of the $+$front 
619: prohibits a well developed advectively mixing 
620: front tail the reduction of $\delta C$ variations there is driven primarily by 
621: diffusion. 
622: The latter having a characteristic time scale given by  $L=0.01$ explains why
623: the front velocities are much smaller than the fast phase velocity.
624: 
625: When $r$ is increased $v_F^+$ tends to become (more) negative: 
626: convection to the right of the $+$interface can now, with increased heating, 
627: better invade the quiescent fluid to the left of it and thus 
628: $\partial_r v_F^+(r,\psi) < 0$. Similarly, when $\psi$ is increased, i.e.,
629: when the convection suppressing Soret effect is diminished the expansion of 
630: TW convection is favoured and thus $\partial_{\psi} v_F^+(r,\psi) < 0$.  
631: 
632: Moving along the $+$front in Fig.~\ref{Fig:frontpics} to the right 
633: from the interface towards the asymptotic TW state at large $x$ there develops 
634: an equilibrium  
635: between the $\delta C$ feed-in from the boundary layers at the plates and the
636: amount of advective mixing: The concentration contrast between two
637: neighboring rolls increases on the way towards the TW bulk. 
638: With it the phase speed $v_p(x)$, the wavelength 
639: $\lambda(x)=2\pi v_p(x)/\omega$, and the lateral concentration
640: current $<\vec{J}>$ grow monotonously up to their asymptotic TW values. This 
641: growth extends laterally over 
642: a wide interval which itself increases when the Soret coupling becomes stronger.
643:  
644: We found that the minimal wavelength in a $+$front state is located at the
645: interface and -- more remarkably -- that it is  about $\lambda_{min}\sim 1.4$ 
646: for {\it all $r$ and $\psi$} that we have simulated. We have no real 
647: quantitative explanation for this strong universal selection of the local
648: wavelength at the interface.
649: Intuitively the growing rolls are squeezed in the region with the negative
650: lateral gradient of $M$. The squeezing is relaxed when the rolls begin to 
651: absorb high concentration contrasts from the plate layers which increases 
652: $v_p$ again [arrow in Fig.~\ref{Fig:frontpics}(c)].
653: 
654: It is interesting to note that the mean concentration current $<\vec{J}>$ of TWs
655: becomes maximal close to the TW saddle node, i.e., where the asymptotic TW 
656: parts of our front states are located. 
657: Finally we mention that the front states do not sustain a measurable lateral 
658: meanflow; the quiescent fluid prohibits that.
659: On the other hand, extended TWs in laterally periodic systems show in general a
660: Reynolds stress-induced meanflow of the order $10^{-3}$ 
661: \cite{Linz88,Barten95I,Luecke98}. But it goes through zero just 
662: near the TW saddle node.
663: 
664: %-------------- Sec. II B2 ----------------------------
665:  \subsubsection{Bifurcation properties}  \label{SEC:bifprop}
666: %---------------------------------------------------
667: 
668: In Figs.~\ref{Fig:Psi25v+om+k-r} - \ref{Fig:3DPsi40om-k-r} we show the
669: bifurcation properties of fronts in comparison with LTWs and laterally periodic
670: TW states. We use front velocities and frequencies being temporally and 
671: spatially constant as order parameters to characterize all of the aforementioned 
672: oscillatory states. In addition we also consider the local wave numbers of 
673: front states and of LTWs in the bulk spatial regions where $\lambda(x)$ has 
674: reached a plateau, i.e., sufficiently away from any interface to conduction.
675: 
676: Figs.~\ref{Fig:Psi25v+om+k-r} and \ref{Fig:3Psisv+l+w-r} show that the front
677: velocities of $+$ and $-$fronts vary quite differently as a function of $r$.
678: The former decrease linearly with growing $r$ and the latter increase, albeit
679: not linearly. Thus, there is a crossing at $r^F_{eq}$ where $v_F^+$ becomes 
680: equal to $v_F^-$, so that both fronts move with the same velocity. At this
681: Rayleigh number the length $l$ of the LTWs diverges, i.e., 
682: $r^{LTW}_\infty = r^F_{eq}$. There, and strictly speaking only there, this 
683: limiting LTW can be seen as a state consisting of two fronts. 
684: 
685: The frequency and bulk wave number selected by a $+$front and of a very long 
686: LTW are close to those of the respective, laterally extended saddle-node 
687: TW (Figs.~\ref{Fig:Psi25v+om+k-r}, \ref{Fig:4Psisom+r-k}, \ref{Fig:3DPsi40om-k-r}).   
688: A somewhat hand-waving explanation for the selection of the saddle-node
689: frequency is as follows: With ({\it i}) convection growing out of conduction in a $+$
690: front, with ({\it ii}) small-amplitude extended TW perturbations of the latter 
691: oscillating according to a purely linear balance with the large Hopf frequency, and 
692: with ({\it iii}) the tendency to decrease 
693: $\omega$ with
694: growing convection amplitude the saddle-node frequency is the first, i.e., the
695: largest possible eigenfrequency of the full nonlinear front problem to allow 
696: for a stable TW region away from the interface.
697: 
698: A stable front state that has a TW bulk part extending laterally to infinity 
699: with frequency $\omega$ and wave number $k$ cannot be realized at $r$-values 
700: that lie below the saddle-node curve of laterally
701: extended TWs, cf. the curve marked $r_s^{TW}$ in the $k-r$ plane of 
702: Fig.~\ref{Fig:3DPsi40om-k-r}. Thus, the lowest Rayleigh number $r^F_{min}$ for
703: the existence of fronts is $r_{min}^{TW} = r_s^{TW}(k \simeq \pi)$, i.e., the 
704: location of the
705: tip of the nose-shaped TW bifurcation surface like the grey surface in 
706: Fig.~\ref{Fig:3DPsi40om-k-r}. Ahead of this nose one
707: cannot realize front states because at such locations there are no 
708: TWs to which the interface from conduction could connect. 
709: 
710: The TW bulk parts of our $+$fronts are practically saddle-node TWs that have  
711: bulk wave numbers on the saddle-node curve $r_s^{TW}(k)$.
712: Furthermore, it is interesting to note that they are on the {\em large}-$k$ 
713: branch of 
714: $r_s^{TW}(k)$ --- the big plusses in Fig.~\ref{Fig:3DPsi40om-k-r} marking the
715: bulk values of the front states lie all above $k \simeq \pi$. In fact, in all 
716: our simulations we did not find front states with bulk wave numbers smaller than 
717: $\pi$. This value marks for all $\psi$ that we investigated the tip of the 
718: nose-shaped TW bifurcation surface like the grey surface in 
719: Fig.~\ref{Fig:3DPsi40om-k-r}.
720: 
721: In contrast to fronts, however, LTWs of 
722: {\em finite} length $l$ can coexist bistably
723: together with the conductive state at $r$-values well below $r_s^{TW}(k)$: They 
724: can sustain over a finite lateral length convection with frequencies and bulk
725: wave numbers (big bullets in Fig.~\ref{Fig:3DPsi40om-k-r}) ''ahead'' of the 
726: grey TW surface for reasons that are explained in Ref.~\cite{Jung02}. 
727: This also shows that fronts and LTWs are quite different states. In the 
728: limit $l \to \infty$ the LTW states merge at $r^{LTW}_\infty=r^F_{eq}$ with a 
729: TW whose wave number and frequency is close to the TW saddle-node as shown in 
730: Figs.~\ref{Fig:Psi25v+om+k-r}, \ref{Fig:4Psisom+r-k}, \ref{Fig:3DPsi40om-k-r}.
731: Therefore, $\omega(r^{LTW}_\infty)$ increases when the Soret coupling becomes 
732: more negative but $k(r^{LTW}_\infty)$ decreases. For $\psi \lesssim -0.4$ it 
733: moves towards the tip of the TW nose at $k\simeq \pi$.
734: 
735: %-------------- Sec. II B3 ----------------------------
736:  \subsubsection{Front selection and stability}  
737: %---------------------------------------------------
738: 
739: Simulations of $+$fronts that were done at fixed control parameters 
740: $r, \psi$ with different initial conditions, e.g., different 
741: wave numbers of an initial TW part produced in general a uniquely 
742: selected final front state with the same frequency and the same 
743: asymptotic bulk TW part.
744: During the formation process initial wave structures with the 'wrong' wave
745: patterns propagated out of 
746: the system in the direction of the phase velocity and 
747: were substituted by convection that was selected by the front. That also 
748: explains why our TW boundary condition $f(L)=f(L-\lambda)$ at the 'downstream'
749: end has no measurable influence on the $+$front state even 
750: when $\lambda$ differs from the front-selected value.
751: 
752: The substitution dynamics is documented in Fig.~\ref{Fig:frontrelaxation}. 
753: There a TW bulk part was prepared initially at $x>8$ with a wavelength of 
754: $\lambda=1.85$ and phase velocity $v_p=1.032$. The spatial region to the right
755: of the interface to conduction is then invaded by the front-selected TW pattern
756: that has a smaller bulk wavelength of $\lambda=1.80$ and that propagates with a 
757: faster phase velocity of $v_p=1.258$. The wave number is increased via several 
758: phase annihilating defects.
759: 
760: All our $+$interfaces selected bulk TW wave numbers close to the large-$k$ 
761: branch of the TW saddle-node curve; cf. Figs.~\ref{Fig:4Psisom+r-k}(b) and 
762: \ref{Fig:3DPsi40om-k-r} for $r^{TW}_s(k)$ and Fig.~\ref{Fig:Psi25v+om+k-r}(c)
763: for $k^{TW}_s(r)$. Thus, these wave numbers are too large 
764: to be Eckhaus stable \cite{Baxter92,Kolodner92,Bu99,MeAlBa04}.
765: However, these fully developed TWs were only convectively unstable 
766: \cite{Bu99}: Perturbations could grow but while doing so they
767: were advected sufficiently fast downstream in the direction of the TW phase 
768: propagation so that they could not influence the upstream part of the $+$front 
769: state in a persistent way. In systems with sufficiently long
770: downstream section of the front state the growing fluctuations have sufficient 
771: time --- or are sufficiently fast growing, respectively --- to reach a critical
772:  amplitude at which
773: two neighboring rolls are annihilated \cite{Baxter92,Kolodner92} as, e.g., for
774: the parameters of Fig.~\ref{Fig:irregularfront}.
775: 
776: Because noise cannot be prevented in general one observes then 
777: such phase defects as in Fig.~\ref{Fig:irregularfront} at irregular points in 
778: time and space beyond a certain downstream growth length that is related to 
779: size of the noise and the growth rate. The associated roll-annihilation events 
780: can lead to an effectively reduced mean wave number in the very far downstream
781: region of the convection bulk. Thus, the coherent part of the $+$front 
782: close to the interface to conduction is followed by a second 
783: incoherent, chaotic phase front consisting of the erratically occurring
784: phase defects. This phase front connects the smooth primary Eckhaus unstable 
785: section to a smooth
786: Eckhaus stable TW with smaller wave number that is realized at larger downstream
787: distances.
788: For parameters for which the growth rate of perturbations of the primary
789: front-selected TW is lower than the one of Fig.~\ref{Fig:irregularfront}
790: one does not observe in short systems the erratically occurring phase defects
791:  --- and even less so the Eckhaus stable final downstream TW state. Indeed, 
792: that was the situation for most of our front states. 
793: 
794: We finally mention that we could also generate front states with frequencies
795: larger than those of the laterally extended saddle node TWs [dotted lines in 
796: Fig.~\ref{Fig:Psi25v+om+k-r}(b) and Fig.~\ref{Fig:4Psisom+r-k}(a)], i.e., with  
797: frequencies that lie above the respective dotted line in the
798: respective 3D plot similar to the one of Fig.~\ref{Fig:3DPsi40om-k-r}. However,
799: we suppose that in sufficient long systems and after long enough times
800: these unstable TW realizations develop a $-$front in the downstream bulk 
801: possibly induced by roll annihilating defects \cite{Kolodner92}.
802: 
803: %-------------- Sec. II C ----------------------------
804:  \subsection{$-$Fronts}  
805: %---------------------------------------------------
806: 
807: The right half of Fig.~\ref{Fig:frontpics}(b) shows a typical $-$front. The
808: mean lateral concentration current in the TW bulk part to the left of the 
809: $-$interface to conduction shuffles positive (negative) $\delta C$ in the upper
810: (lower) part of the layer towards the $-$interface. Thus, a large vertical
811: concentration gradient is maintained slightly ahead of it that strongly 
812: stabilizes the conduction regime to the right of the $-$interface:
813: there the mixing number $M$ is even larger than 1. In this way the TW
814: oscillations are damped and the conduction regime is shielded against a rapid
815: invasion of convection. 
816: 
817: The increase of $M(x)$ upon approaching the interface from the convection side 
818: causes --- and is related to --- a similar increase of $v_p(x)$ and $\lambda(x)$.
819: The rolls disperse with growing phase velocity $v_p(x)$ over a short lateral 
820: distance at the interface. The decreasing convection amplitude lowers the mean 
821: concentration current and causes $M(x)$ to grow further. This in turn enhances 
822: $v_p(x)$ and $\lambda(x)$ leading to 
823: smaller convection amplitude and so on. It is therefore the strongly nonlinear 
824: lateral concentration current $<\vec{J}>$ which is responsible for the 
825: rapid self amplified decay process of convection at the $-$interface.
826: 
827: With increasing $r$ the front velocity $v_F^-$ changes sign, becomes positive
828: and continues to grow [Fig.~\ref{Fig:Psi25v+om+k-r}(a) and 
829: Fig.~\ref{Fig:3Psisv+l+w-r}(a)] because the quiescent state becomes
830: less stable when increasing $r$.
831: The slope of $v_F^-(r)$, i.e., the increase of the front velocity is 
832: considerably steeper for negative $v_F^-$ than for positive ones:
833: The strongly stabilizing solutal stratification ahead of the interface hinders 
834: convection to intrude
835: into the quiescent fluid region but favours the latter to replace the TW 
836: part.
837: 
838: The 'upstream' lateral distance over which the $-$interface to conduction 
839: influences the TW to the left of it is definitely smaller than the 'downstream'
840: influence length of the $+$interface on the convective bulk. In the former case
841: one cannot observe 
842: a difference to an extended TW state at an 'upstream' distance of, say, 10-15 
843: rolls while in the latter case the 'downstream' convection properties
844: approach the asymptotic bulk TW behavior only over a significantly longer 
845: distance. So, in particular the phase dilatation at a $-$interface does not  
846: propagate 'upstream' into the TW bulk against the fast phase flow.
847: 
848: This also explains why in the formation process of a $-$front TW 
849: properties that were initially present in a developed form are conserved. In 
850: fact, we could 
851: produce coherent $-$fronts for a fixed $r$ with different wave numbers of the 
852: bulk TW part out of a whole band near the saddle node wave numbers 
853: $k_s^{TW}(r)$. Only for higher $r$ and initial wave numbers away from 
854: $k_s^{TW}$ we observed long-time transient incoherent front behavior.
855: Here this transition to incoherence may correspond to a transition from 
856: a convectively to an absolutely unstable regime concerning the propagation of 
857: phase dilatations in 'upstream' direction.
858: 
859: We should like to stress again that in contrast to $-$fronts which depend on
860: the preparation process the asymptotic 
861: 'downstream' TW part of a $+$front is uniquely selected as discussed in the
862: previous section. Thus, for a particular $r$ we have found only a single 
863: coherent $+$front.
864:  
865: For definiteness and for the sake of comparison with $+$fronts we show in the 
866: Figs. of this paper the properties of $-$front states that have a bulk TW part
867: which itself was selected by a $+$front.
868: This, however, has a slight numerical drawback stemming from the convective
869: Eckhaus instability of this TW part: ever present phase noise (in particular 
870: at the boundary of the 'upstream' TW region) is enhanced on the 'downstream' 
871: way towards the $-$interface. We think that this effect is 
872: responsible for fluctuations in our frequency measurements of $-$fronts. These 
873: data are therefore not shown in Figs.~\ref{Fig:Psi25v+om+k-r}(b) and 
874: Fig.~\ref{Fig:4Psisom+r-k}. However, in our simulations the 'upstream' part of 
875: the $-$fronts were too short to allow for the full development of phase slip 
876: defects.
877: 
878: %-------------- Sec. II D ----------------------------
879:  \subsection{Transient two-front structures}  
880: %---------------------------------------------------
881: 
882: Consider a set-up where a $+$front generates a very long TW part that develops 
883: back into the quiescent fluid via a coherent $-$front. If the convective 
884: part is laterally sufficiently long then this structure appears as a two-front 
885: structure with the TW part being spatially confined between a $+$ and a 
886: $-$interface to conduction. This structure will in general either expand or
887: shrink laterally. Only when the two front velocities $v_F^+$ and $v_F^-$ are 
888: equal, i.e., at the
889: crossing points $r^F_{eq}$ of the curves in Figs.~\ref{Fig:Psi25v+om+k-r}(a) and
890: \ref{Fig:3Psisv+l+w-r}(a) one can have a stationary state, namely, 
891: a LTW with diverging length $l$.
892: 
893: We have simulated such structures at $\psi=-0.25$ and $-0.30$ for Rayleigh 
894: numbers for which $v_F^+<v_F^-$, i.e., to the right of the crossing in 
895: Figs.~\ref{Fig:Psi25v+om+k-r}(a) and \ref{Fig:3Psisv+l+w-r}(a) so that these 
896: two-front structures expand. Their properties are practically identical 
897: to those of the respective single-front states. The two-front structures
898: have a technical advantage over the simulation of  single fronts:
899: we could use a periodic boundary condition that was located in the quiescent
900: region of the former. This avoids the noise that is induced at the upstream TW 
901: boundary of a single $-$front state. With this noise source being absent in our 
902: two-front structures the frequency fluctuations at the $-$
903: interfaces of single $-$front states did not occur. 
904: 
905: When we reduced $r$ then the velocities of the two
906: fronts approached each other along the lines in 
907: Figs.~\ref{Fig:Psi25v+om+k-r}(a) and \ref{Fig:3Psisv+l+w-r}(a) that were
908: obtained from velocity measurements of single-front states. In that way we could
909: reproduce the unique crossing point at $r^F_{eq}$ where $v_F^+=v_F^-$ and
910: where LTWs with diverging $l$ exist with a drift velocity $v_d$ given by the
911: crossing velocity.
912: 
913: In addition to expanding two-front structures we simulated also shrinking ones
914: for a particular parameter combination ($\psi=-0.35, r=1.3586$, for which 
915: $v_F^+=-0.022 > v_F^-=-0.067$) that is located in Fig.~\ref{Fig:3Psisv+l+w-r}(a)
916: at $r<r^F_{eq}$. As an example consider the 
917: large two-front structure as in Fig.~\ref{Fig:LTWapproach}. 
918: The $+$front selects in the bulk of this initial two-front structure a saddle 
919:  node TW with wavelength $\lambda_{plateau}\sim 1.905$. In the course of time 
920:  the velocity  $v_F^-$ of the $-$front 
921: approaches that of the $+$front and a stationary LTW forms with length 
922: $l \simeq 47$ that drifts with the velocity $v_d=v_F^+=-0.022$ and oscillates
923: with the frequency of the $+$front. These values of the $+$front remain
924: practically unchanged in the whole process. 
925: 
926: %\clearpage
927: %-------------- Sec. III ----------------------------
928: \section{Localized traveling wave states} \label{SEC:LTW}
929: %----------------------------------------------------
930: 
931: We produced LTW states with very large length $l$ immediately below the 
932: Rayleigh number $r^{LTW}_\infty = r^F_{eq}$. There, $l$ diverges thus marking 
933: the upper existence boundary of LTW states. And there, the 
934: velocities $v_F^+$ and $v_F^-$ of the $+$ and $-$front states, respectively, 
935: become equal. See Figs.~\ref{Fig:Psi25v+om+k-r} and \ref{Fig:3Psisv+l+w-r} for
936: the corresponding results in the range of $-0.4 \leq \psi \leq -0.25$ that we
937: have investigated in this paper. Upon decreasing $r$ below the threshold 
938: $r^{LTW}_\infty(\psi)$ one finds uniquely selected 
939: LTW states. Depending on parameters they can coexist stably with front states,
940: extended TW states, and the quiescent basic state.
941: 
942: For completeness we include here in Fig.~\ref{Fig:phasediagram} a phase 
943: diagram of the $\psi -r$ plane where
944: all the LTWs that we have numerically obtained in the range 
945: $-0.65 \leq \psi \leq - 0.08$ are shown by vertical bars together
946: with the saddle-node location $r_s^{TW} (k = \pi)$ of extended TWs (full line)
947: and the oscillatory Hopf bifurcation threshold $r_{osc}(k = \pi)$ for TWs 
948: (dashed line). But in this work we focus on the range 
949: $-0.4 \leq \psi \leq -0.25$.
950: 
951: We should like to emphasize again that we found LTWs for 
952: $-0.4 \leq \psi \leq -0.25$ only in the parameter regime below 
953: $r^{LTW}_\infty(\psi)$, i.e.,  
954: to the left of the crossing points in Figs.~\ref{Fig:Psi25v+om+k-r}(a) and 
955: \ref{Fig:3Psisv+l+w-r}(a) where the velocities of independent single fronts 
956: become equal. Thus, LTWs exist at these $\psi$ only for parameters for which 
957: $v_F^+ > v_F^-$, 
958: i.e., for which independent fronts would approach each other: eventually any 
959: convective region between them would shrink to zero and the quiescent conductive
960: state would result if this interface motion would continue without change. 
961: However,
962: the stabilization effects that allow in such a situation a uniquely selected 
963: stable and robust LTW are easily understood with the help of the investigations
964: in the following subsections. On the other hand, for $r>r^F_{eq}$ the front
965: velocities are such that a two-front structure expands.
966: 
967: %-------------- Sec. III A ----------------------------
968: \subsection{Transient dynamics towards the selected LTW}
969: %----------------------------------------------------
970: 
971: A typical transient dynamics towards the uniquely selected LTW is shown in 
972: Fig.~\ref{Fig:frontrelaxation} for $\psi=-0.35$. Here the initial condition was a 
973: very broad two-front structure that was prepared at 
974: $r=1.3586$ where it shrinks with $v_F^+ > v_F^-$ [Fig.~\ref{Fig:3Psisv+l+w-r}(a)]. 
975: In fact, the $-$front moves to the left with a speed that is about three times 
976: higher than that of the $+$front. 
977: 
978: In the following shrinking process 
979: where the $-$front closes up to the $+$front the latter does not change its
980: velocity at all and the former keeps its velocity as long as 
981: the bulk TW part between the two fronts is effectively asymptotic, i.e., without 
982: lateral variation. This behavior reflects the fact that in such broad two-front
983: structures there is practically no interaction between the fronts when
984: their distance is so large that an asymptotic TW part is realized between the 
985: interfaces to conduction. 
986: However, the situation changes when the convective region between the interfaces 
987: becomes less extended since
988: it requires a finite 'downstream' growth length behind a $+$interface over 
989: which the 
990: convection properties still vary with small gradients before the asymptotic TW 
991: is reached. The slow lateral variation is best seen in the mixing
992: number $M(x)$ in Fig.~\ref{Fig:frontpics}(e) reflecting the slow variation of 
993: the concentration distribution and in the related convective contribution 
994: $\langle b \rangle$ [Fig.~\ref{Fig:frontpics}(g)] to the local buoyancy.
995: 
996: When the front separation comes to the order of this length
997: one cannot speak any more of a two-front structure with two independent 
998: fronts: For the parameters of Fig.~\ref{Fig:frontrelaxation} the velocity 
999: of the formerly independent $-$front changes continuously from $v_F^-=-0.067$ 
1000: to $-0.022$, i.e, to the velocity of the preceeding $+$front and a coherent and
1001: robust LTW 
1002: forms which moves with a drift velocity that is determined by the $+$front,
1003: $v_d = v_F^+=-0.022$. During this slowing-down process of the $-$interface 
1004: its structure changes to that of the characteristic decay
1005: interface to conduction of a LTW. Therein the two interfaces, i.e., the former 
1006: $+$ and $-$fronts, respectively, are in a robust
1007: equilibrium with each other at a uniquely selected fixed distance $l$ that 
1008: depends on $r,\psi$ as shown in Fig.~\ref{Fig:3Psisv+l+w-r}(b).
1009: 
1010: So the $-$interfaces of LTWs and front states do not influence significantly the
1011: upstream part of these stuctures. Hence, the local concentration 'barrier'
1012: ahead of the $-$interface does not select the drift velocity of LTWs as
1013: speculated previously \cite{Barten95II}. It is rather the $+$interface that is
1014: the more important one.
1015: 
1016: %-------------- Sec. III B ----------------------------
1017: \subsection{Stabilization via front repulsion}
1018: %----------------------------------------------------
1019: 
1020: Note that it is the 'downstream wake' in the concentration field of the 
1021: preceeding $+$front that
1022: effectively slows down the $-$front: When the latter reaches the region where 
1023: the mixing number $M(x)$ [Fig.~\ref{Fig:frontpics}(e)] starts to decrease towards
1024: the preceeding $+$front, i.e., when the convective contribution 
1025: $\langle b \rangle$ [Fig.~\ref{Fig:frontpics}(g)] to the local buoyancy starts to
1026: increase then the speed of the approaching 
1027: $-$interface has to slow down. This distance over which the $+$front influences 
1028: the $-$front in a two-front structure grows when the Soret 
1029: coupling becomes stronger. For example at $\psi>-0.3$ a separation of about 
1030: 160 rolls between the two interfaces is not sufficient to ensure independence. 
1031: 
1032: The sensitive dependence of the velocity of the $-$interface on
1033: the concentration-induced buoyancy variation in the 'wake' behind the $+$front
1034: is the main reason for the robust localization mechanism of (long) LTWs. The 
1035: invasion of conduction into the convective region via the trailing $-$interface
1036: is stopped at just that well defined distance from the $+$front where the 
1037: concentration-induced convective buoyancy $\langle b \rangle$
1038: has become sufficiently large. The latter increases monotonously towards the 
1039: well mixed region under the $+$interface since this degree of advective mixing 
1040: decreases gradually in the 'wake' behind the $+$interface. See, e.g., 
1041: ref.~\cite{Jung02} for an explanation of the associated interplay of diffusion 
1042: and advection which both reduce concentration gradients and the Soret effect 
1043: which generates them. Of course, the effect of stopping the approaching 
1044: $-$interface at a particular distance from the $+$interface can be interpreted
1045: as an effective repulsive interaction between them.
1046: 
1047: %-------------- Sec. III C ----------------------------
1048: \subsection{Long LTWs}
1049: %----------------------------------------------------
1050: 
1051: The structural similarity 
1052: between long LTWs and fronts is documented in Fig.~\ref{Fig:frontpics}. 
1053: Differences between the full lines (fronts) and the dashed ones (LTW) 
1054: are visible only in the case of the $-$front in
1055: Fig.~\ref{Fig:frontpics}(d,f,h). Here the bulk asymptotic TW that is realized 
1056: to the left of the $-$front differs slightly from the plateau TW in the LTW. 
1057: 
1058: The above mentioned $\delta C$ redistribution via $<\vec{J}>$ enhances the
1059: buoyancy at the $+$interface and leads there to a self-consistent 
1060: stabilization of convection against invasion of conduction at the $+$front.
1061: This mixing effect makes stable LTWs possible even for low heating rates $r$ 
1062: where neither extended TW convection nor fronts exist. 
1063: 
1064: Long LTWs are characterized by a wide TW part with a well developed
1065: plateau with almost no lateral variation in the convection properties like,
1066: e.g., $v_p(x)$ or $M(x)$ \cite{Jung02}. The TW plateau separates the growth and 
1067: the decay part of convection at the $+$ and $-$interface, respectively
1068: and it provides a communication mechanism favoring one direction: The first 
1069: region is shielded from the second one by the fast 'downstream' phase 
1070: propagation. Like in a single $-$front state the $-$interface of the LTW does 
1071: not influence the 'upstream' TW; it only manages the decay transition of the 
1072: TW vortices into the quiesent fluid. Thus, the
1073: $+$front character at the $+$interface is also present in the LTW. And the
1074: properties of long LTWs are dominated by and similar to those of the single 
1075: $+$front at the same $r$ if the latter exists. For example, 
1076: the drift velocities of long LTWs agree with the values
1077: of the corresponding $+$fronts in Figs.~\ref{Fig:Psi25v+om+k-r}, 
1078: \ref{Fig:3Psisv+l+w-r}, \ref{Fig:4Psisom+r-k}, \ref{Fig:3DPsi40om-k-r}, 
1079: \ref{Fig:LTWapproach}, \ref{Fig:4Psisom-r}. Furthermore, they 
1080: continue to show the same linear variation with $r$ as the $+$fronts even where 
1081: the latter cease to exist at smaller $r$, cf., the open
1082: circles in Fig.~\ref{Fig:3Psisv+l+w-r}(a) for the cases of $\psi=-0.35$ and 
1083: $\psi=-0.4$. Similarly, the variation $\omega(r)$ of long LTWs follows the
1084: corresponding one of $+$fronts, cf., open circles and filled triangles in 
1085: Fig.~\ref{Fig:4Psisom-r}. 
1086: 
1087: A comparison of LTW plateau values with extended TWs and front TWs 
1088: is presented in Fig.~\ref{Fig:Psi25v+om+k-r}(b,c) for $\psi=-0.25$, in 
1089: Fig.~\ref{Fig:3DPsi40om-k-r}  for  $\psi=-0.4$, and in 
1090: Fig.~\ref{Fig:4Psisom+r-k} for for all examined $\psi$. At
1091: $r^{LTW}_\infty=r^F_{eq}$ (arrows) 
1092: there is no difference between the fronts, the diverging LTWs, and the extended
1093: saddle TWs. 
1094: 
1095: For decreasing $r$ convection is less stable, the disintegration of the
1096: traveling rolls sets in earlier, and the LTW length $l$ is therefore reduced,
1097: cf., the inset of Fig.~\ref{Fig:Psi25v+om+k-r} and
1098: Fig.~\ref{Fig:3Psisv+l+w-r}(b). The smallest plateau wave numbers of LTWs are 
1099: realized for diverging lengths at $r^{LTW}_\infty(\psi)$. With decreasing $r$ 
1100: one 
1101: finds a slight increase of $k_{plateau}$ while the wave number selected by a 
1102: single $+$front remains close to that of the saddle TW.
1103: 
1104: So this is the LTW bifurcation scenario that we found in the range 
1105: $-0.4 \leq \psi \leq -0.25$ (for a discussion of the scenario at smaller Soret
1106: coupling strength cf. Sec.~\ref{SEC:small-soret}): Approaching 
1107: $r^{LTW}_\infty$ from below LTWs become 
1108: indistinguishable from front states when $l \to \infty$. But further 
1109: below $r^{LTW}_\infty$ LTWs
1110: differ more and more in particular with respect to the bulk 
1111: wave numbers as can be seen in
1112: Figs.~\ref{Fig:4Psisom+r-k} and \ref{Fig:3DPsi40om-k-r}.
1113: However, for long enough LTWs with a well developed spatial bulk plateau 
1114: behavior the frequencies $\omega (r, \psi)$ and drift velocities $v_d(r, \psi)$ 
1115: vary like the corresponding quantities of $+$fronts. This confirms the fact 
1116: \cite{Jung02} that it is the $+$front-like growth 
1117: interface that selects the properties of a long LTW. Shorter LTWs behave
1118: also with respect to the variation of $\omega (r, \psi)$ and $v_d(r, \psi)$ 
1119: somewhat differently.
1120: 
1121: %-------------- Sec. III D ----------------------------
1122: \subsection{Short LTWs}
1123: %----------------------------------------------------
1124: 
1125: Reducing $r$ one eventually arrives for any $\psi$ at the regime of short LTWs
1126: that are marked in Figs~\ref{Fig:Psi25v+om+k-r}, \ref{Fig:3Psisv+l+w-r}, and 
1127: \ref{Fig:4Psisom-r} by shaded circles and that are located close to the dotted 
1128: line in Fig.~\ref{Fig:phasediagram}. Here, the dominant influence of the $+$front
1129: vanishes eventually in the regime of short LTW pulses.
1130: No convection plateau can be identified any more in these structures, cf. the
1131: dotted curves in Fig.~\ref{Fig:frontrelaxation}.
1132: The prototype of a short LTW consists of a growth interface which is followed 
1133: directly by the decay of convection so that the whole pulse has to be seen now 
1134: as one integrated structure that no longer contains front-like independent $+$ 
1135: and $-$interfaces. Hence, short LTWs show a strong lateral
1136: variation of their properties. The shape of their amplitudes superficially 
1137: resemble the pulse solutions of the complex Ginzburg-Landau equation 
1138: \cite{Niemela90,Steinberg91,Kolodner91c,Thual8890,Malomed90,Hakim90,vSaarloos9092}. 
1139: 
1140: Like for long LTWs the stable existence of short pulses below any heating that 
1141: is necessary to sustain 
1142: extended TWs is caused by a lateral $\delta C$ redistribution over the pulse.
1143: Also its frequency $\omega$ is constant in the frame that comoves with the drift
1144: velocity $v_d$ of the pulse like for a long LTW. However, compared to long LTWs
1145: short LTWs provide a qualitatively new convection structure.
1146: They are independent of and cannot even be compared with extended TWs because 
1147: there is no characteristic wavelength or phase velocity. The special 
1148: character of short pulses compared with long LTWs or fronts is 
1149: reflected in the change of the $r$-variation of $v_d, l, w_{max}, \omega$ that 
1150: can be seen in Figs.~\ref{Fig:Psi25v+om+k-r}, \ref{Fig:3Psisv+l+w-r}, and 
1151: \ref{Fig:4Psisom-r} by comparing the dashed circles with the open ones and the
1152: filled triangles.
1153: 
1154: We observed the shortest possible LTWs at the lower end, $r^{LTW}_{min}$, of 
1155: the $r$ band of LTWs. There they seem to end via a saddle 
1156: node bifurcation for pulses \cite{Hakim90}. These minimal pulses always 
1157: contained about 5 convection rolls 
1158: for all Soret couplings $-0.65 \leq \psi \leq -0.08$ that we have 
1159: investigated. This surprising universality of 
1160: $l_{min} \simeq 5$, i.e., its insensitivity to the values of the actual heating
1161: rate $r^{LTW}_{min}$ and the Soret coupling $\psi$ is still unexplained.  
1162:  
1163: Approaching the lower band limit of
1164: LTWs their flow intensity steeply drops [Fig.~\ref{Fig:3Psisv+l+w-r}(c)]
1165: and consequently their frequency increases [Fig.~\ref{Fig:4Psisom-r}] as the
1166: degree of advective mixing of the fluid decreases.
1167: 
1168: %-------------- Sec. III E ----------------------------
1169: \subsection{Comparison with LTW models} \label{SEC:LTWmodels}
1170: %----------------------------------------------------
1171: 
1172: Several attempts have been made to describe LTWs by simple model equations.
1173: Stable pulse solutions of the complex Ginzburg-Landau equation (CGLE) were 
1174: proposed 
1175: as a model for confined binary mixture convection \cite{Thual8890}.
1176: The nonlinear interaction between the local amplitude and frequency seems to be
1177: the essential localization mechanism in this approximation.
1178: Indeed, one could find localized solutions of increasing length up to
1179: the limit of an infinitely long two-front state 
1180: \cite{Malomed90,Hakim90,vSaarloos9092}.
1181: But some basical problems remained: 
1182: Within the CGLE all pulses drift with the same velocity. This is
1183: the critical linear group velocity $v_g$ if the coefficients are derived from an 
1184: asymptotic 
1185: reduction of the full hydrodynamic field equations. But $v_g$ is too 
1186: fast by a
1187: factor of about 20-40 compared with the LTW drift velocity in experiments or 
1188: simulations 
1189: \cite{Cross88Knobloch88,Zimmermann8993,Surko87,Barten91,Luecke92,Barten95II,Jung02,Kolodner91a,Kolodner94}.
1190: Brand and Deissler \cite{Brand8990} introduced asymmetry in the pulse properties
1191: by adding nonlinear gradient terms to the CGLE. 
1192: %???(In fact, already the ordinary CGLE showed asymmetric pulse solutions that 
1193: %do not drift with $v_g$ \cite{Hakim90}). ??? 
1194: A similar extension was given by Bestehorn {\it et al.} \cite{Bestehorn89,Bestehorn888991}
1195: within their framework of order parameter equations. Both could produce a very 
1196: slow drift, even opposite to the phase direction \cite{Bestehorn90}.
1197: But this kind of nonlinear modification of the linear group velocity involves a
1198: balance which seems to be
1199: too fragile to explain the occurrence of small pulse velocities over 
1200: a whole range of $\psi$ and $r$ \cite{RieckeL92}.
1201: 
1202: Another problem was mentioned by van Saarloos and Hohenberg \cite{vSaarloos9092}.
1203: According to their model of a quintic CGLE nonlinear wide pulses 
1204: are expected by counting arguments to exist only in a codimension-2 
1205: submanifold of 
1206: the parameter space. Provided there are no hidden symmetries this seems to be 
1207: incompatible 
1208: with the robust occurrence of LTWs in experiments.
1209: Furthermore, stable pulse solutions seem to exist only in the bistable regime 
1210: whereas LTWs are known to
1211: persist well above the linear onset of extended convection for weakly negative 
1212: $\psi$ 
1213: \cite{Niemela90,Behringer9091,Kolodner9091,Kolodner91a,Kolodner91c,Kolodner91d,Barten91,Barten95II,Luecke98}.
1214: Furthermore, coexisting small stable and wide unstable LTWs were never seen in 
1215: the CGLE but found in experiments by Kolodner \cite{Kolodner94}. Instead, 
1216: stable broad pulse solutions
1217: are found in the model to arise in a saddle node bifurcation together with an 
1218: unstable branch
1219: of smaller 'critical droplets' near the basic state \cite{Hakim90}.
1220: Finally, numerical solutions of the field equations show the 
1221: existence 
1222: of stable LTWs even below the lowest TW saddle node \cite{Jung02}. This makes 
1223: clear that LTWs are
1224: influenced by a localization mechanism that is not contained in the CGLE.
1225: 
1226: Inspired by simulation results of Barten {\it et al.} \cite{Barten91,Luecke92} 
1227: which showed the important role of the concentration field for a LTW Riecke 
1228: \cite{RieckeL92,RieckeD92} 
1229: proposed an extension of the Ginzburg-Landau equations. Within a weakly 
1230: nonlinear expansion he coupled into the standard CGLE as an additional slow 
1231: variable the amplitude ${\mathcal C}$ of an advected mean large-scale 
1232: concentration mode that influences the growth of the critical modes.
1233: A similar idea was advanced already by Glazier {\it et al.} \cite{Kolodner9091}.
1234: The extension can induce an additional amplitude-intability of phase winding 
1235: solutions 
1236: to modulated waves. It may be considered as the origin for pulse formation in 
1237: this ansatz \cite{Riecke01}. 
1238: Riecke showed that the influence of the real ${\mathcal C}$-mode alone on 
1239: the local growth rate (without dispersion) suffices
1240: to generate slowly drifting stable pulse solutions even below a supercritical 
1241: TW-bifurcation \cite{RieckeD92}.
1242: In this way he modelled a new localization mechanism to explain the 
1243: robust occurrence of LTWs in binary mixture convection.
1244: The amplitude ${\mathcal C}$ can be interpreted as a measure for the local 
1245: mixing state or the mean convection-induced deviation of the vertical
1246: concentration gradient from the conductive one. 
1247: In this way his extended complex Ginzburg-Landau equation (ECGLE) contains in a 
1248: sketchy way physical effects like the mixing influence on the growth rate and 
1249: the large-scale concentration redistribution. 
1250: 
1251: Riecke characterized within his model short and long LTWs 
1252: as dispersion-dominated pulses \cite{Riecke96} and states of two fronts 
1253: that are bounded by the ${\mathcal C}$ dynamics \cite{RieckeD95}, respectively.
1254: He proposed an explanation for their coexistence in stable and unstable form, 
1255: respectively,
1256: by the competition between dispersion-dominated and ${\mathcal C}$-dominated 
1257: localization \cite{RieckeD95,RieckeL95}.
1258: 
1259: Note, however, that
1260: in contrast to the model used by Riecke our results show 
1261: % no correlation between drift direction of long LTWs and their stability.
1262: stable long LTWs that drift either in or opposite to
1263: the direction of phase propagation depending on paramters. 
1264: It would be interesting to check whether adding a term of the form 
1265: $v |{\mathcal A}|^2 \partial_x {\mathcal C}$ to the ${\mathcal C}$-equation 
1266: can stabilize forward drifting long pulses within the model since it 
1267: models the concentration 'wake', i.e., the transport of the local mixing state 
1268: in phase direction by the traveling rolls of amplitude ${\mathcal A}$.
1269: 
1270: Numerical and analytical investigations of the ECGLE predict a hysteretic 
1271: transition from
1272: slow to fast drifting pulses or the existence of oscillatory moving ones 
1273: \cite{Riecke96}. But both were never seen in experiments or simulations.
1274: Thus, despite their capability in elucidating some essential mechanisms 
1275: CGLE type models have the drawback so far that they reproduce only 
1276: single aspects of LTWs in a qualitative manner.
1277: Their range of validity and their predictive power is not well known.
1278: And since a satisfactory relation with the 
1279: full field equations has not been established these models remain somewhat 
1280: arbitrary. 
1281: %???Moreover, our simulations show that the strongly nonlinear 
1282: %TW front behavior should be described by a model that can also 
1283: %cover the dynamics of long LTWs.??? 
1284: 
1285: It appears questionable that weakly nonlinear expansions with spatially 
1286: slowly varying mode amplitudes are approriate at all in view of the very 
1287: large P\'eclet numbers, $\cal O$(1000), measuring the strength of the nonlinearity 
1288: in the concentration balance. Thus, so far numerical simulations of the full 
1289: field equations seem 
1290: to be the appropriate tool besides careful experiments to gather insight into 
1291: the specific physical mechanisms for LTW formation in binary mixture convection.
1292: 
1293: %\clearpage
1294: %-------------- Sec. IV ----------------------------
1295: \section{Comparison with experiments and discussion} \label{SEC:Compare-exp}
1296: %----------------------------------------------------
1297: %-------------- Sec. I V A----------------------------
1298: \subsection{Small Soret coupling strength} \label{SEC:small-soret}
1299: %----------------------------------------------------
1300: 
1301: An inspection of Figs.~\ref{Fig:Psi25v+om+k-r}, \ref{Fig:3Psisv+l+w-r}, and
1302: \ref{Fig:phasediagram}
1303: shows that the lower band limit $r^{LTW}_{min}$ for the existence of short 
1304: LTWs and the crossing value, $r^F_{eq}=r^{LTW}_{\infty}$, where the velocities 
1305: of free fronts become equal approach each other when $|\psi|$ decreases. Thus, 
1306: one can foresee an interval of moderately negative $\psi$ where
1307: short and long LTWs can be found close to $r^{LTW}_\infty$. 
1308: In this case the upper band limit for the existence of stable LTWs should be 
1309: defined by a 
1310: backward saddle-node bifurcation at $r_s^{LTW}$ where the branches of stable 
1311: short and unstable 
1312: long LTWs annihilate each other. 
1313: 
1314: Hence, we expect that the upper parts of the
1315: bifurcation diagrams of $l$ versus $r$ in the inset of 
1316: Figs.~\ref{Fig:Psi25v+om+k-r} and in Fig.~\ref{Fig:3Psisv+l+w-r}(b) curve
1317: backwards towards smaller $r$ when $|\psi|$ decreases further below the values
1318: of the two figures. In this way the shape of the curve $l(r)$ would change 
1319: continuously from the form shown in Fig.~\ref{Fig:l-r-scheme}(b) to the one in 
1320: Fig.~\ref{Fig:l-r-scheme}(a). The former shows schematically the bifurcation 
1321: behavior of $l(r)$ that we have determined numerically for $\psi \lesssim -0.25$.
1322: In fact, at $\psi$ slightly larger than -0.25 we expect the appearence of the 
1323: saddle-node in the curves $l(r)$. Fig.~\ref{Fig:l-r-scheme}(a) is a schematic 
1324: representation of experimental results of Kolodner \cite{Kolodner94} for 
1325: $\psi=-0.127$ as presented in his Fig.~5. He stabilized by an adaptive 
1326: heating mechanism long unstable LTWs in coexistence with short stable ones. 
1327: Thus, the unstable LTW solution branch [dashed line in 
1328: Fig.~\ref{Fig:l-r-scheme}(a)] forms for $r^{LTW}_\infty \leq r \leq r_s^{LTW}$ a 
1329: separatrix between the domains of attraction of expanding two-front structures 
1330: to the right of the dashed line in Fig.~\ref{Fig:l-r-scheme}(a) and the domain
1331: to the left of the dashed line leading to stable narrow pulses or the basic 
1332: state. Furthermore, small LTWs that are prepared at $r < r_s^{LTW}$  will 
1333: evolve into expanding two-front structures when $r$ is increased above 
1334: $r_s^{LTW}$.
1335: 
1336: Note that in Kolodner's experiment \cite{Kolodner94} done at $\psi=-0.127$ the 
1337: upper band limit $r_s^{LTW}$ of LTWs lies {\it above} the Hopf bifurcation 
1338: threshold $r_{osc}$ for extended TWs where perturbations of the quiescent fluid
1339: can grow. Therefore, one has to address there questions related to linear 
1340: and nonlinear 
1341: convective versus absolute instability \cite{Chomaz92,Bu99}, to linearly selected
1342: so-called pulled fronts versus nonlinearly selected so-called pushed fronts 
1343: \cite{vSaarloos9092,vSaarloos03}, and to the robustness and stability of 
1344: nonlinear fronts under emission or absorption of TW perturbations that can 
1345: grow in the region occupied by the quiescent fluid.
1346: 
1347: %-------------- Sec. IV B----------------------------
1348: \subsection{Strong Soret coupling strength}
1349: %----------------------------------------------------
1350: 
1351: For stronger negative $\psi$ the measured LTW properties agree qualitatively 
1352: well with our results. For example, the 'arbitrary-width confined states' 
1353: found in experiments \cite{Kolodner93II} for $\psi=-0.253$ 
1354: at a single Rayleigh number are to be identified as two-front structures.
1355: A quantitative comparison is difficult due to the difference in the boundary 
1356: conditions:
1357: % whose implications for the spatiotemporal behavior of LTWs can hardly be assessed:
1358: We simulated two-dimensional convection assuming translational symmetry in 
1359: $y$-direction while the narrow experimental convection channels
1360: impose no-slip conditions at the walls perpendicular to the roll axes.
1361: There are three effects that account for the difference beteen experiments and
1362: simulations. 
1363: 
1364: First, the characteristic Rayleigh numbers
1365: in the experiments are higher. This is already known from the 
1366: suppression of oscillatory or steady convection instabilities in narrow channels
1367: \cite{Platten84,CJT97,AlBa04}.
1368: The no-slip conditions at the side-walls generate a nontrivial $y$-variation 
1369: of the velocity 
1370: field that introduces additional internal friction and that has to be 
1371: compensated by a higher heating rate \cite{Catton72,Ohlsen90,Bensimon90}.
1372: 
1373: Second, the LTW drift velocities in the experiment have the global tendency to
1374: lie below those of the simulations. For example, for $\psi=-0.253$ the
1375: experimental LTWs \cite{Kolodner93II,Kolodner94} move opposite to the phase 
1376: velocity whereas according to our calculations $v_d$ should be around $0.05$.
1377: Again, we attribute this difference to the fact that we neglect gradients
1378: in $y$-direction. They change the concentration redistribution dynamics in
1379: particular at the $+$interface of the LTW which determines the drift velocity.
1380: In this context one has to note that already weak
1381: inhomogeneities in a convection cell can slow down and even pin the LTW 
1382: movement \cite{Kolodner91a,Kolodner91c,Kolodner93II}.
1383: Finally, the influence of the different boundaries on the frequencies, phase 
1384: velocities, and wave numbers of confined states are totally unknown.
1385:  
1386: Also when comparing quantitatively our results for fronts with experiments one 
1387: should take into account the above discussed points.
1388: 
1389: Extrapolating our results for the front velocities to more negative $\psi$
1390: beyond $\psi=-0.4$ we see that already at the lowest TW saddle location 
1391: $r_{min}^{TW}$ two-front stuctures
1392: would expand with $v_F^- > v_F^+$. In other words, the velocity crossing 
1393: point $r^F_{eq}$ is no longer above $r_{min}^{TW}$  but has virtually moved 
1394: below
1395: the lowest TW saddle location where in fact no fronts exist. On the other hand,
1396: LTWs still exist in this $r$-range with length increasing with $r$. However, 
1397: $l(r)$ does not seem to diverge anymore as for $-0.4 \lesssim \psi$ which is
1398: compatible with the absence of fronts moving with the same velocity.
1399: 
1400: %-------------- Sec. IV C----------------------------
1401: \subsection{Wall-attached confined structures}
1402: %----------------------------------------------------
1403: 
1404: Laterally confined convection patches of traveling rolls were found in the 
1405: early experiments 
1406: \cite{Moses87Heinrichs87,Fineberg88Steinberg89,Niemela90,Kolodner91b} 
1407: that were done in narrow rectangular convection channels in the form of 
1408: so-called 
1409: wall-attached confined structures (WACS). They were localized near one of the 
1410: short end walls closing the channel. 
1411: 
1412: These WACS can be understood with our 
1413: knowledge of fronts and free LTWs. For example, for the weakly negative $\psi$ 
1414: used in the early experiments the phase velocity of the WACS was directed
1415: towards the wall to which they were attached. Indeed, for such parameters 
1416: mainly short LTWs occur  
1417: with  drift velocities in phase direction so that they would end as WACS of 
1418: the above described type in finite length channels.
1419: Furthermore, the measured WACS profiles of phase velocity $v_p(x)$, of 
1420: wavelength $\lambda(x)$, and their decrease of frequency
1421: with increasing $r$ \cite{Fineberg88Steinberg89,Steinberg91,Kolodner91b} 
1422: agree qualitatively with the typical behavior of free short LTWs.
1423: 
1424: The connection between WACS and free LTWs was more explicitly demonstrated 
1425: by Kolodner \cite{Kolodner90} for more negative separation ratios 
1426: $\psi=-0.24$ and $-0.408$:
1427: He prepared a free LTW pulse with large phase velocity (a 'fast confined state'
1428: in his terminology) which drifted slowly opposite to the direction of
1429: phase propagation towards an end wall of the convection channel and became 
1430: there a WACS (a 'slow confined state' in his terminology) with lower phase 
1431: velocity being directed away from the wall.
1432: % ((According to our results LTWs at $\psi=-0.24$ should drift in phase direction.
1433: % We adress this discrepancy to the two dimensionality of our simulations. Therefore we neglect
1434: % the strong inhomogenities in the lateral $y$-direction.))
1435: In this WACS the phase generating 'trailing front', i.e., the analogue of
1436: the $+$interface is pinned at the wall and 
1437: therefore without solutal gradients to the quiescent fluid as in a free LTW. 
1438: The absence of these concentration variations at the $+$interface implies and 
1439: allows a lateral concentration redistribution over the whole state at lower 
1440: levels of the mixing number $M$ in WACS as compared to free LTWs. Consequently, 
1441: the phase velocities and frequencies of WACS are smaller than those
1442: of the respective free LTWs. A less dramatic drop of frequency was observed 
1443: also between forward drifting LTWs and short WACS at $\psi=-0.047$ 
1444: \cite{Steinberg91}.
1445: 
1446: Due to their better mixing capability, i.e., smaller $M$ it is 
1447: very probable that short WACS 
1448: % with the growth part attached to a wall
1449: can exist for heating rates below the lower band limit $r^{LTW}_{min}$ of 
1450: stable LTW 
1451: pulses --- at least in the case where the phase velocity is directed away 
1452: from the wall. It would be very interesting to test this conjecture
1453: experimentally, in 
1454: particular for strongly negative $\psi$. There $r^{LTW}_{min}$ itself lies 
1455: already 
1456: well below the range of stable TWs \cite{Jung02} and so the WACS would lie even
1457: lower.
1458: An inportant hint that this conjecture is right is given by Ning {\it et al.} 
1459: \cite{Ning96}. They have performed two-dimensional simulations 
1460: of a finite-length convection channel
1461: with realistic boundary conditions for $\psi=-0.47, \sigma=13.8$, and $L=0.01$. 
1462: Neglecting the slight difference
1463: in $\sigma$ their results should be comparable with our work.
1464: They found short WACS at $r=1.35$ which is according to our results far below 
1465: the TW saddle nodes for this
1466: separation ratio and also below the lower band limit $r^{LTW}_{min}$ of free 
1467: LTWs 
1468: for $\psi=-0.40$. However, these authors claim --- we think, incorrectly ---
1469: that their WACS lie above the saddle-node location $r_s^{TW}$ of extended TWs.
1470: 
1471: %-------------- Sec. IV D----------------------------
1472: \subsection{LTW and front stability}
1473: %----------------------------------------------------
1474: In the previous section we have shown that the $+$interface where the 
1475: convection rolls grow in 'downstream' direction out of the quiescent fluid 
1476: plays the dominant role for the stability of LTWs. While the $-$interface 
1477: where the decaying rolls are advected into the quiescent fluid does 
1478: not play a decisive role. This is clearly confirmed in 
1479: pulse collision experiments \cite{Kolodner9091}.
1480:  
1481: Fast TW pulses --- linear ones with small amplitude as well as nonlinear 
1482: ones with larger amplitude --- were completely absorbed by a LTW when the 
1483: pulses hit the
1484: $-$interface of the LTW, i.e., when the pulse velocity is directed opposite to 
1485: and towards
1486: the phase velocity of the LTW. Then the collision with the pulse affects only 
1487: the $-$interface
1488: itself and perturbations are quickly advected out of the LTW and do not propagate
1489: upstream towards the $+$interface. For the same reason double-LTW states of 
1490: two counter propagating waves can persist over a
1491: long time \cite{Kolodner9091} or even be stable \cite{Kolodner91d}.
1492: Moreover, a pair of LTWs that have their phase propagation directed towards 
1493: each other and that interact with each other via their decay interfaces is seen 
1494: to be stable over a substantially wider $r$-range than two LTW pulses which 
1495: are connected at their growth interfaces \cite{Kolodner91d}. Obviously the 
1496: latter case is more critical for the structural integrity of the involved LTWs.
1497:  
1498: A LTW is most likely destroyed when another wave with the same direction 
1499: of phase propagation infiltrates its phaseflow at the growth region. 
1500: Then, while growing the perturbations can be transmitted into the 
1501: strongly nonlinear bulk part of the LTW and can destroy its coherence.
1502: 
1503: The different selection and stability properties of the $+$ and $-$interfaces
1504: were already observed 
1505: in transient convection behavior in various experiments (see for example 
1506: \cite{Kolodner88,Bensimon90}), however without further investigation.
1507:  
1508: %-------------- Sec. IV E ----------------------------
1509: \subsection{Defected confined states}
1510: %----------------------------------------------------
1511: 
1512: The fact that ($i$) different $-$fronts with different bulk 
1513: TW parts are possible as stable coexisting states
1514: and that ($ii$) a $+$front interface is in general stable against
1515: downflow perturbations opens 
1516: the possibility for another kind of stable long 
1517: confined TWs: Therein rolls with low wavelength grow out of the quiescent 
1518: fluid in a 'normal' growth part. In the bulk an incoherent phase front 
1519: connects this fast wave that is coming from the $+$interface with a slow wave
1520: of higher wavelength and larger amplitude
1521: via spatiotemporal dislocations as, e.g., in Fig.~\ref{Fig:irregularfront}.
1522: The transition could take place via one or more intermediate convection 
1523: states. Eventually this slow TW convection undergoes a decay transition
1524: into the basic state via a coherent $-$interface. Such 'defected 
1525: confined states' are indeed observed in annular containers \cite{Bensimon90,
1526: Behringer9091} and were studied by Kolodner \cite{Kolodner94}. 
1527: He found such structures only for $\psi\le -0.21$.
1528: One may speculate that for these separation ratios the bulk TW that is selected 
1529: by and behind the $+$interface is absolutely unstable against the slow wave 
1530: in the further 'downstream' part.
1531: This could explain the existence of persistent roll pair annihilations without 
1532: the need of fluctuations.
1533: 
1534: We finally mention that the occurrence of phase annihilating dislocations 
1535: between a growth part and the downstream convection was observed for moderately 
1536: negative $\psi$ already in narrow rectangular containers \cite{Kolodner91b}.
1537: Furthermore, end-wall induced Eckhaus instabilities of downstream TW states 
1538: have been seen in simulations \cite{BuLu99}. 
1539: 
1540: %\clearpage
1541: %-------------- Sec. V ----------------------------
1542: \section{Conclusion}
1543: %----------------------------------------------------
1544: 
1545: %1---I---10----I---20----I---30----I---40----I---50----I---60----I---70----I---80
1546:  For parameters where the conductive quiescent fluid is stable and where 
1547:  spatially extended TW solutions bifurcate subcritically out of it we have 
1548:  investigated in quantitative detail relaxed, strongly nonlinear oscillatory 
1549:  convection 
1550:  structures with one or two interfaces to the quiescent fluid, i.e, fronts and
1551:  LTWs, respectively. They are time-periodic global
1552:  nonlinear modes: in the frame that is comoving 
1553:  with the respective front velocity $v_F$ or with the LTW drift velocity $v_d$
1554:  the oscillations have everywhere the same period. 
1555:  
1556:  Fronts come in two varieties.
1557: In a $+$front state the quiescent fluid is located
1558: 'upstream', i.e., phase propagates out of it into convection. 
1559: In a $-$front the quiescent fluid is located
1560: in 'downstream' direction and phase moves out of convection 
1561: into conduction.
1562: 
1563: The lowest Rayleigh number for the existence of fronts is 
1564: the lowest saddle-node location of extended 
1565: TWs, $r^F_{min} = r^{TW}_{min}$: below it there are no TWs to which the 
1566: interface from conduction can 
1567: connect. However, LTWs of {\em finite} length $l$ can coexist bistably
1568: together with the conductive state well below the lowest TW saddle when the
1569: Soret coupling is sufficiently negative, $r^{LTW}_{min} < r^{TW}_{min}$. 
1570: Furthermore, we have arguments 
1571: that WACS at end walls of rectangular channels can exist even at smaller $r$ 
1572: than LTWs. 
1573: 
1574: Central for understanding fronts and LTWs is a large-scale concentration
1575: redistribution that influences the buoyancy at the interfaces to conduction in
1576: different ways than in the bulk TW parts. For example, at the $+$interfaces of
1577: fronts and LTWs alike there is a buoyancy overshoot which is sufficiently 
1578: large to sustain local convection growth there and that can cause even 
1579: invasion of convection into the stable quiescent fluid. At the $-$interface
1580: the lateral buoyancy variation is such as to induce the decay of the approaching
1581: convection rolls into the conductive state. 
1582: 
1583: Front velocities as well as LTW drift velocities are much smaller than the phase
1584: velocities of the carrier waves for reasons that are related to the 
1585: concentration redistribution dynamics. The velocities of $+$fronts decrease 
1586: with growing $r$ while those of $-$fronts increase. At some
1587: $r^F_{eq}$ they become equal so that both fronts move with the same velocity.
1588: At this Rayleigh number the length $l$ of the LTWs diverges and there, and 
1589: strictly speaking only there, the limiting LTW can be seen as a state 
1590: consisting of two fronts. However, $+$fronts and long LTWs have almost 
1591: identical propagation velocities and frequencies. Furthermore, they select a 
1592: similar bulk wave number. The selected frequencies and bulk wave numbers are
1593: close to those of a saddle-node TW. In fact, it is the $+$front-like growth 
1594: interface that selects the properties of long LTWs. 
1595: 
1596: Small amplitude extended TW perturbations of the conductive state oscillate 
1597: with the large Hopf frequency. But the global-mode oscillation 
1598: is restrained by the requirement that its frequency has to allow stable 
1599: developed bulk TW convection. It is interesting to note that the $+$interface 
1600: connecting conduction with
1601: convection selects the largest possible frequency eigenvalue that meets this
1602: requirement, namely the TW saddle-node frequency.
1603: All our $+$fronts select bulk TW wave numbers close to the large-$k$ 
1604: branch of the TW saddle-node curve, i.e., wave numbers that are too large 
1605: to be Eckhaus stable. However, these TWs are only convectively unstable: 
1606: perturbations can grow but while doing so they
1607: are advected sufficiently fast downstream in the direction of the TW phase 
1608: propagation so that they cannot influence the upstream part of the $+$front 
1609: state in a persistent way. 
1610: 
1611: While $+$front states seem to be uniquely selected we could produce for a 
1612: fixed $r$ 
1613: different coherent $-$fronts that were characterized in the bulk part
1614: by different wave numbers and frequencies close to the TW saddles.
1615: The  decay interface adjusts itself to the respective bulk TW part but does not
1616: exert an influence in 'upstream' direction on the bulk convection within a coherent
1617: $-$front. In contrast, the growth under a $+$interface 
1618: induces in downstream direction a long concentration 'wake' that is 
1619: characteristic
1620: for $+$fronts and long LTWs and of special importance for the latter.
1621: 
1622: Here it is interesting to notice that all the interfaces of fronts and LTWs
1623: consist typically only of about 3-4 convection rolls. We furthermore should
1624: like to mention that $+$interfaces of fronts and LTWs always locate a minimal 
1625: wavelength. Its value, $\lambda_{min}\sim 1.4$, is remarkably universal for
1626: {\it all $r$ and $\psi$} that we have simulated. This is unexplained so far. 
1627: 
1628: We have also prepared initial two-front stuctures by connecting a $+$front and 
1629: a $-$front with a common long bulk TW. When $r>r^F_{eq}$ they expand. But at 
1630: $r<r^F_{eq}$ they shrink towards a uniquely selected LTW of fixed length $l$.
1631: Here the 'downstream wake' in the concentration field of the 
1632: preceeding $+$front exerts an effective repulsion on the approaching $-$interface: 
1633: the invasion of conduction via the latter is stopped at a well
1634: defined distance $l$ that is determined by the concentration-induced buoyancy 
1635: levels in the 'wake' of the $+$front. 
1636: 
1637: LTWs shortly below $r^F_{eq}$ (where LTW with diverging $l$ are possible) are 
1638: very long. Their drift velocities, frequencies, and many stuctural properties 
1639: are similar to those of $+$fronts. Decreasing $r$ the LTW length decreases
1640: and one eventually arrives for any $\psi$ at the regime of short LTWs that lies
1641: for strongly negative $\psi$ well below the TW saddle-nodes. These short LTWs
1642: without a convection plateau are qualitatively different structures.
1643: This is also reflected by their drift velocities and frequencies
1644: showing a variation with $r$ that differs from those of long LTWs. 
1645: The shortest possible LTWs are realized at the lower end of 
1646: the $r$ band of LTWs. These minimal pulses always 
1647: contained about 5 convection rolls for all Soret couplings that we have 
1648: investigated. This surprising universality of $l_{min} \simeq 5$ remains to be 
1649: explained.  
1650: 
1651: \clearpage
1652:  %----------------------------------------------
1653:  \begin{thebibliography}{999}
1654:  %----------------------------------------------
1655: 
1656: \bibitem{CH93}
1657:  M.~C.~Cross and P.~C.~Hohenberg,
1658: % {\it Pattern Formation Outside of Equilibrium}
1659:  Rev.~Mod.~Phys. {\bf 65}, 851 (1993).
1660:  %---------------------------
1661:  \bibitem{coupling}
1662:  For 5 weight percent of ethanol mixed into water at $T=20^{\circ} C$ the separation
1663:  ratio measuring the Soret coupling strength \cite{CH93} is 
1664:  $\psi \simeq -0.3$; 
1665:  %\cite{KWM88}.
1666:  %---------------------------
1667:  %\bibitem{KWM88} 
1668:  see P.~Kolodner, H.~L.~Williams, and  C.~Moe,
1669:  %{\it Optical Measurement of the Soret Coefficient of Ethanol-Water Solutions},
1670:  J.~Chem.\ Phys.~{\bf 88}, 6512 (1988).
1671:  %-----------------------------------------
1672:  \bibitem{WKPS85}
1673:  R.~W.~Walden, P.~Kolodner, A.~Passner, and C.~M.~Surko,
1674: % {\it Traveling Waves and Chaos in Convection in Binary Fluid Mixtures}
1675:  Phys.~Rev.~Lett. {\bf 55}, 496 (1985).
1676:  %-----------------------------------------
1677:  \bibitem{Moses87Heinrichs87}
1678:  E.~Moses, J.~Fineberg, and V.~Steinberg,
1679: % {\it Multistability and confined traveling-wave patterns in a convecting binary mixture}
1680:  Phys.~Rev.~A {\bf 35}, R2757 (1987);
1681:  R.~Heinrichs, G.~Ahlers, D.~S.~Cannell,
1682: % {\it Traveling waves and spatial variation in the convection of a binary mixture}
1683:  {\it ibid.} {\bf 35}, R2761 (1987).
1684:  %------------------------------------------
1685:  \bibitem{Behringer9091}
1686:  K.~E.~Anderson and R.~P.~Behringer,
1687: % {\it Long time scales in traveling wave convection patterns}
1688:  Phys.~Lett.~A {\bf 145}, 323 (1990);
1689:  K.~E.~Anderson and R.~P.~Behringer,
1690: % {\it Traveling Wave Convection Patterns in an Annular Cell}
1691:  Physica~D {\bf 51}, 444 (1991).
1692:  %------------------------------------------
1693:  \bibitem{Winkler92}
1694:  B.~I.~Winkler and P.~Kolodner,
1695: % {\it Measurements of the concentration field in nonlinear travelling-wave convection}
1696:  J.~Fluid Mech. {\bf 240}, 31 (1992).
1697:  %------------------------------------------------------------------------
1698:  \bibitem{Kolodner94}
1699:  P.~Kolodner,
1700: % {\it Stable, unstable, and defected confined states of traveling-wave convection}
1701:  Phys.~Rev.~E {\bf 50}, 2731 (1994).
1702:  %------------------------------------------
1703:  \bibitem{Platten96}
1704:  H.~Touiri, J.~K.~Platten, and G.~Chavepeyer,
1705: % {\it Effect of the separation ratio on the transition between travelling waves and steady
1706: % convection in the two-component Rayleigh-B\'enard problem}
1707:  Eur.~J.~Mech.~B {\bf 15}, 241 (1996).
1708:  %------------------------------------------
1709:  \bibitem{Kaplan94}
1710:  E.~Kaplan, E.~Kuznetsov, and V.~Steinberg,
1711: % {\it Burst and collapse in traveling-wave convection of a binary fluid}
1712:  Phys.~Rev.~E {\bf 50}, 3712 (1994).
1713:  %-----------------------------------------
1714:  \bibitem{Surko91}
1715:  C.~M.~Surko, D.~R.~Ohlsen, S.~Y.~Yamamoto, and P.~Kolodner,
1716: % {\it Confined states of traveling-wave convection}
1717:  Phys.~Rev.~A {\bf 43}, R7101 (1991).
1718:  %------------------------------------------
1719:  \bibitem{Aegert01}
1720:  C.~M.~Aegerter and C.~M.~Surko,
1721: % {\it Effects of lateral boundaries on traveling-wave dynamics in binary fluid convection}
1722:  Phys.~Rev.~E {\bf 63}, 46301 (2001).
1723:  %-----------------------------------------*******13
1724:  \bibitem{Ning97b}
1725:  L.~Ning, Y.~Harada, and H.~Yahata,
1726: % {\it Formation Process of the Traveling-Wave State with a Defect in Binary Fluid Convection}
1727:  Prog.~Theor.~Phys. {\bf 98}, 551 (1997).
1728:  %-----------------------------------------
1729:  \bibitem{Batiste01}
1730:  O.~Batiste, E.~Knobloch, I.~Mercader, and M.~Net,
1731: % {\it Simulations of Oscillatory Binary Fluid Convection in Large Aspect Ratio Containers}
1732:  Phys.~Rev.~E {\bf 65}, 016303 (2001).
1733:  %------------------------------------------
1734:  \bibitem{Jung02}
1735:  D.~Jung and M.~L\"ucke,
1736: % {\it Localized Waves without the Extended Waves: Oscillatory Convection
1737: %      of Binary Mixtures with Strong Soret Effect}
1738:  Phys.~Rev.~Lett. {\bf 89}, 054502 (2002).
1739:  %-----------------------------------------
1740:  \bibitem{FL}
1741:  C. F\"utterer and M. L\"ucke,
1742: % {\it Growth of binary fluid convection: the role of the concentration field},
1743:  Phys.~Rev.~E {\bf 65}, 036315 (2002).
1744:  %------------------------------------------------------------------------
1745:  \bibitem{JML04}
1746:  D.~Jung, P.~Matura, and M.~L\"ucke, 
1747: % {\it Oscillatory convection in binary mixtures: Thermodiffusion, solutal
1748: % buoyancy, and advection},
1749:  Eur.~Phys.~J.~E.~{\bf15}, 293 (2004).
1750:  %------------------------------------------------------------------------
1751:  \bibitem{MJL04}
1752:  P.~Matura, D.~Jung, and M. L\"ucke,
1753: % {\it Standing wave oscillations in binary mixture convection: from the onset via 
1754: % symmetry breaking to period doubling into chaos},
1755:  Phys.~Rev.~Lett. {\bf 92}, 254501 (2004).
1756:  %------------------------------------------------------------------------
1757:  \bibitem{Bensimon90}
1758:  D.~Bensimon, P.~Kolodner, C.~M.~Surko, H.~Williams, and V.~Croquette,
1759: % {\it Competing and coexisting dynamical states of travelling-wave convection
1760: %      in an annulus}
1761:  J.~Fluid Mech. {\bf 217}, 441 (1990). 
1762:  %------------------------------------------
1763:  \bibitem{Kolodner92Fronts}
1764:  P.~Kolodner,
1765: % {\it Extended states of nonlinear traveling-wave convection.
1766: %      II. Fronts and spatiotemporal defects}
1767:  Phys.~Rev.~A {\bf 46}, 6452 (1992).
1768:  %------------------------------------------------------------------------
1769:  \bibitem{Bu99}
1770:  P.~B\"uchel,
1771: % {\it Konvektion in zweikomponentigen Fl\"ussigkeiten: Einflu{\ss}  
1772: % von Scherstr\"omung und seitlicher Berandung auf die Strukturbildung},
1773:  Ph.D.\ thesis, Universit\"at des Saarlandes, 1999.
1774:  %------------------------------------------------
1775:  \bibitem{Surko87}
1776:  C.~M.~Surko and P.~Kolodner,
1777: % {\it Oscillatory Traveling-Wave Convection in a Finite Container}
1778:  Phys.~Rev.~Lett. {\bf 58}, 2055 (1987).
1779:  %------------------------------------------
1780:  \bibitem{AlBa04}
1781:  A.~Alonso and O.~Batiste,
1782: % {\it Onset of oscillatory binary fluid convection in three-dimensional cells},
1783:  Theoret.~Comput.~Fluid~Dynamics~{\bf 18}, 239 (2004).
1784:  %-------------------------------------------------------------------
1785:  \bibitem{Kolodner90Glazier91} 
1786:  P.~Kolodner, J.~A.~Glazier, and H.~Williams,
1787: % {\it Dispersive Chaos in One-Dimensional Traveling-Wave Convection}
1788:  Phys.~Rev.~Lett. {\bf 65}, 1579 (1990);
1789:  J.~A.~Glazier, P.~Kolodner, and H.~Williams,
1790: % {\it Dispersive Chaos}
1791:  J.~Stat.~Phys. {\bf 64}, 945 (1991).
1792:  %-----------------------------------------
1793:  \bibitem{Luecke98}
1794:  M.~L\"ucke, W.~Barten, P.~B\"uchel, C.~F\"utterer, St.~Hollinger, and Ch.~Jung,
1795: % {\it Pattern Formation in Binary Fluid Convection and in Systems with Throughflow}
1796:  in {\it Evolution of Structures in Dissipative Continuous Systems},
1797: % Lecture Notes in Physics, {\bf m55},
1798:  edited by F. H. Busse and S. C. M\"uller,    % Lecture Notes in Physics, {\bf m55}
1799:  (Springer, Berlin, 1998), p. 127.
1800:  %-----------------------------------------
1801:  \bibitem{Barten95I}
1802:  W.~Barten, M.~L\"ucke, M.~Kamps, and R.~Schmitz,
1803: % {\it Convection in Binary Fluid Mixtures. I. Extended Traveling-Wave and Stationary States},
1804:  Phys.~Rev.~E {\bf 51}, 5636 (1995).
1805:  %-----------------------------------------****27
1806:  \bibitem{Barten95II}
1807:  W.~Barten, M.~L\"ucke, M.~Kamps, and R.~Schmitz,
1808: % {\it Convection in Binary Fluid Mixtures: II. Localized Traveling Waves},
1809:  Phys.~Rev.~E {\bf 51}, 5662 (1995).
1810:  %-----------------------------------------
1811:  \bibitem{Kolodner88}
1812:  P.~Kolodner, D.~Bensimon, and C.~M.~Surko,
1813: % {\it Traveling-Wave Convection in an Annulus}
1814:  Phys.~Rev.~Lett. {\bf 60}, 1723 (1988).
1815:  %------------------------------------------ 
1816:  \bibitem{Kolodner91a}
1817:  P.~Kolodner,
1818: % {\it Drifting Pulses of Traveling-Wave Convection}
1819:  Phys.~Rev.~Lett. {\bf 66}, 1165 (1991)
1820:  %-----------------------------------------
1821:  \bibitem{Steinberg91}
1822:  V.~Steinberg and E.~Kaplan, 
1823:  in {\it Spontaneous Formation of Space-Time Structures and Criticallit},
1824:  edited by T.~Riste and D.~Sherrington (Kluwer, Boston, 1991), p. 207.
1825:  %-----------------------------------------
1826:  \bibitem{Niemela90}
1827:  J.~J.~Niemela, G.~Ahlers, and D.~S.~Cannell,
1828: % {\it Localized Traveling-Wave States in Binary-Fluid Convection}
1829:  Phys.~Rev.~Lett. {\bf 64}, 1365 (1990).
1830:  %-----------------------------------------
1831:  \bibitem{Kolodner9091}
1832:  P.~Kolodner and J.~A.~Glazier,
1833: % {\it Interaction of localized pulses of traveling-wave convection with propagating disturbance}
1834:  Phys.~Rev.~A {\bf 42}, R7504 (1990);
1835:  J.~A.~Glazier and P.~Kolodner,
1836: % {\it Interaction of nonlinear pulses in convection in binary mixture}
1837:  {\it ibid.} {\bf 43}, 4269 (1991).
1838:  %------------------------------------------
1839:  \bibitem{Kolodner91c}
1840:  P.~Kolodner,
1841: % {\it Drift, shape, and intrinsic destabilization of pulses of traveling-wave convection}
1842:  Phys.~Rev.~A {\bf 44}, 6448 (1991).
1843:  %------------------------------------------
1844:  \bibitem{Kolodner91d}
1845:  P.~Kolodner,
1846: % {\it Collision between pulses of traveling-wave convection}
1847:  Phys.~Rev.~A {\bf 44}, 6466 (1991).
1848:  %------------------------------------------
1849:  \bibitem{Kolodner93II}
1850:  P.~Kolodner,
1851: % {\it Arbitrary-width confined states of traveling-wave convection: Pinning, locking,
1852: %      drift, and stability}
1853:  Phys.~Rev.~E {\bf 48}, R4187 (1993).
1854:  %------------------------------------------
1855:  \bibitem{Barten91}
1856:  W.~Barten, M.~L\"ucke, and M.~Kamps,
1857: % {\it Localized Traveling-Wave Convection in Binary-Fluid Mixtures}
1858:  Phys.~Rev.~Lett. {\bf 66}, 2621 (1991).
1859:  %-----------------------------------------
1860:  \bibitem{Cross868889}
1861:  M.~C.~Cross,
1862: % {\it Traveling and Standing Waves in Binary-Fluid Convection in Finite Geometries}
1863:  Phys.~Rev.~Lett. {\bf 57}, 2935 (1986);
1864: % M.~C.~Cross,
1865: % {\it Structure of nonlinear traveling-wave state in finite geometries}
1866:  Phys.~Rev.~A {\bf 38}, 3593 (1988);
1867: % M.~C.~Cross,
1868: % {\it Non-linear traveling wave states in finite geometries}
1869:  Physica D {\bf 37}, 315 (1989).
1870:  %-----------------------------------------
1871:  \bibitem{Brand86}
1872:  H.~R.~Brand, P.~S.~Lomdahl, and A.~C.~Newell,
1873: % {\it Evolution of the order parameter in situations with broken rotational symmetry}
1874:  Phys.~Lett. A {\bf 118}, 67 (1986);
1875: % {\it Benjamin-Feir turbulence in convective binary fluid mixtures}
1876:  Physica D  {\bf 23}, 345 (1986).
1877:  %-----------------------------------------
1878:  \bibitem{Fineberg90}
1879:  J.~Fineberg, V.~Steinberg, and P.~Kolodner,
1880: % {\it Weakly nonlinear states as propagating fronts in convecting binary mixtures}
1881:  Phys.~Rev.~A {\bf 41}, R5743 (1990). 
1882:  %----------------------------------------- ******40
1883:  \bibitem{Kolodner8889}
1884:  P.~Kolodner and C.~M.~Surko,
1885: % {\it Weakly Nonlinear Traveling-Wave Convection}
1886:  Phys.~Rev.~Lett. {\bf 61}, 842 (1988);
1887:  P.~Kolodner, C.~M.~Surko, and H.~L.~Williams,
1888: % {\it Dynamics of Traveling Waves near the Onset of Convection in Binary Fluid Mixtures}
1889:  Physica D {\bf 37}, 319 (1989).
1890:  %-----------------------------------------
1891:  \bibitem{Fineberg88Steinberg89}
1892:  J.~Fineberg, E.~Moses, and V.~Steinberg,
1893: % {\it Spatially and Temporally Modulated Traveling-Wave Pattern in Convecting Binary Mixtures}
1894:  Phys.~Rev.~Lett. {\bf 61}, 838 (1988);
1895: % J.~Fineberg, E.~Moses, and V.~Steinberg,
1896: % {\it Nonlinear pattern and wave-number selection in convecting binary mixtures}
1897:  Phys.~Rev.~A {\bf 38}, R4939 (1988);
1898:  V.~Steinberg, J.~Fineberg, E.~Moses, and I.~Rehberg,
1899: % {\it Pattern Selection and Transition to Turbulence in Propagating Waves}
1900:  Physica D {\bf 37}, 359 (1989).
1901:  %-----------------------------------------
1902:  \bibitem{Bestehorn89}
1903:  M.~Bestehorn, R.~Friedrich, and H.~Haken,
1904: % {\it Modulated traveling waves in nonequilibrium systems: the 'blinking state'}
1905:  Z.~Phys.~B {\bf 77}, 151 (1989).     
1906:  %-----------------------------------------
1907:  \bibitem{Kolodner93}
1908:  P.~Kolodner,
1909: % {\it Repeated transients of weakly nonlinear traveling-wave convection}
1910:  Phys.~Rev.~E {\bf 47}, 1038 (1993).
1911:  %-----------------------------------------
1912:  \bibitem{Kolodner86}
1913:  P.~Kolodner, A.~Passner, C.~M.~Surko, and R.~W.~Walden,
1914: % {\it Onset of Oszillatory Convection in a Binary Fluid Mixture}
1915:  Phys.~Rev.~Lett. {\bf 56}, 2621 (1986).
1916:  %-----------------------------------------
1917:  \bibitem{Kolodner87}
1918:  P.~Kolodner, C.~M.~Surko, A.~Passner, and H.~L.~Williams,
1919: % {\it Pulses of oscillatory convection}
1920:  Phys.~Rev.~A {\bf 36}, R2499 (1987).
1921:  %-----------------------------------------
1922:  \bibitem{Moses86}
1923:  E.~Moses and V.~Steinberg,
1924: % {\it Flow patterns and nonlinear behavior of traveling waves in a convective binary fluid}
1925:  Phys.~Rev.~A {\bf 34}, R693 (1986).
1926:  %-----------------------------------------
1927:  \bibitem{Kolodner90}
1928:  P.~Kolodner,
1929: % {\it Neutrally stable fronts of slow convective traveling waves}
1930:  Phys.~Rev.~A {\bf 42}, R2475 (1990).
1931:  %-----------------------------------------******48
1932:  \bibitem{Kolodner91b}
1933:  P.~Kolodner,
1934: % {\it Stable and unstable pulses of traveling-wave convection}
1935:  Phys.~Rev.~A {\bf 43}, 2827 (1991).
1936:  %-----------------------------------------
1937:  \bibitem{Yahata91}
1938:  H.~Yahata,
1939: % {\it Travelling convection rolls in a binary fluid mixture}
1940:  Prog.~Theor.~Phys. {\bf 85}, 933 (1991).
1941:  %-----------------------------------------
1942:  \bibitem{Ning96}
1943:  L.~Ning, Y.~Harada, and H.~Yahata,
1944: % {\it Localized Traveling Waves in Binary Fluid Convection}
1945:  Prog.~Theor.~Phys. {\bf 96}, 669 (1996).
1946:  %-----------------------------------------
1947:  \bibitem{Ning97a}
1948:  L.~Ning, Y.~Harada, and H.~Yahata,
1949: % {\it Modulated Traveling Waves in Binary Fluid Convection in an 
1950: %      Intermediate-Aspect-Ratio Rectangular Cell}
1951:  Prog.~Theor.~Phys. {\bf 97}, 831 (1997).
1952:  %-----------------------------------------
1953:  \bibitem{LL66}
1954:  L.~D.~Landau and E.~M.~Lifschitz,
1955:  {\it Course of Theoretical Physics},
1956: % {\it Hydrodynamik}
1957: % Akademie-Verlag, Berlin, 1966.
1958:  (Pergamon Press, Oxford, 1993), Vol.~6.
1959:  %--------------------------- 
1960:  \bibitem{Platten84}
1961:  J.~K.~Platten and J.~C.~Legros,
1962:  {\it Convection in Liquids},
1963:  (Springer, Berlin, 1984).
1964:  %---------------------------
1965:  \bibitem{HLL92}
1966:  W.~Hort, S.~J.~Linz, and M.~L\"ucke,
1967: % {\it Onset of Convection in Binary Gas Mixtures: The Role of the Dufour Effect}
1968:  Phys.~Rev.~A {\bf 45}, 3737 (1992).
1969:  %--------------------------- 
1970:  \bibitem{HL95}
1971:  St.~Hollinger and M.~L\"ucke,
1972: % {\it Influence of the Dufour effect on convection in binary gas mixtures}
1973:  Phys.~Rev.~E {\bf 52}, 642 (1995).
1974:  %--------------------------- 
1975:  \bibitem{LA96}
1976:  J.~L.~Liu and G.~Ahlers,
1977: % {\it  Spiral--Defect Chaos in Rayleigh--B\'enard Convection with Small Prandtl Numbers}
1978:  Phys.~Rev.~Lett. {\bf 77}, 3126 (1996).
1979:  %---------------------------
1980:  \bibitem{LLT83}
1981:  G. W. T. Lee, P. Lucas, and A. Tyler,
1982: % {\it  Onset of Rayleigh-B\'enard Convection in Binary Liquid Mixtures of $^{3}$He in $^{4}$He}
1983:  J.~Fluid~Mech. {\bf 135}, 235 (1983).
1984:  %---------------------------
1985:  \bibitem{MAC-SOLA}
1986:  F.~H.~Harlow and  J.~E.~Welch,
1987: % {\it Numerical Calculation of Time Dependent Viscous Flow or
1988: %     Fluid with Free Surfaces},
1989:  Phys.\ Fluids~{\bf 8}, 2183 (1965);
1990:  J.~E.~Welch, F.~H.~Harlow, J.~P.~Shannon, and  B.~J.~Daly,
1991: % {\it The MAC-Method: A Computing Technique for Solving Viscous,
1992: %      Incompressible, Transient Fluid-Flow Problems Involving
1993: %      Free Surfaces}
1994:  Los Alamos Scientific Laboratory of the University
1995:  of California, Report No. LA-3425, 1966 (unpublished);
1996:  C.~W.~Hirt, B.~D.~Nichols, and  N.~C.~Romero,
1997: % {\it SOLA --- A Numerical Solution Algorithm for
1998: %      Transient Fluid Flow},
1999: % Los Alamos Scientific Laboratory of the University
2000: % of California, 
2001:   {\it ibid.} Report No. LA-5852, 1975 (unpublished).
2002:  %---------------------------
2003:  \bibitem{PT83}  
2004:  R.~Peyret and T.~D.~Taylor,
2005:  {\it Computational Methods in Fluid Flow}
2006:  (Springer, Berlin, 1983).
2007:  %-------------********60
2008:  \bibitem{vSaarloos9092}
2009:  W.~van Saarloos and P.~C.~Hohenberg,
2010: % {\it Pulses and Fronts in the Complex Ginzburg-Landau Equation near a 
2011: %      Subcritical Bifurcation}
2012:  Phys.~Rev.~Lett. {\bf 64}, 749 (1990);
2013: % W.~van Saarloos and P.~C.~Hohenberg,
2014: % {\it Fronts, pulses, sources and sinks in generalized complex Ginzburg-Landau equation}
2015:  Physica D {\bf 56}, 303 (1992).
2016:  %------------------------------------------
2017:  \bibitem{vSaarloos03}
2018:  W.~van Saarloos, 
2019: % {\it Front propagation into unstable states}
2020:  Phys.~Rep. {\bf 386}, 29 (2003); and references therein.
2021:  %------------------------------------------
2022:  \bibitem{vHvSHo93}
2023:  M.~van Hecke, W.~van Saarloos, and P.~C.~Hohenberg,
2024: % {\it Comment on "Absolute and convective instabilities in nonlinear systems"},
2025:  Phys.~Rev.~Lett. {\bf 71}, 2162 (1993).
2026:  %------------------------------------------
2027:  \bibitem{CoWaSMi99}
2028:  P.~Colet, D.~Walgraef, and M.~San~Miguel, 
2029: % {\it Convective and absolute instabilities in the subcritical Ginzburg-Landau 
2030: % equation},
2031:  Eur.~Phys.~J.~B {\bf 11}, 517 (1999). 
2032:  %-------------------------------------------------------------------
2033:  \bibitem{CoCh01}
2034:  A.~Couairon and J.~M.~Chomaz, 
2035: % {\it Pushed global modes in weakly inhomogeneos subcritical flows},
2036:  Physica (Amsterdam) {\bf 158D}, 129 (2001).
2037:  %------------------------------------------
2038:  \bibitem{SzLu03}
2039:  A. Szprynger and M.~L\"ucke, 
2040: % {\it Noise sensitivity of sub- and supercritically bifurcating patterns with 
2041: % group velocities close to the convective-absolute instability},
2042:  Phys.~Rev.~E~{\bf67}, 046301 (2003).
2043:  %-------------------------------------------------------------------
2044:  \bibitem{Rabaud03}
2045:  L. Meignin, P. Gondret, C. Ruyer-Quil, and M. Rabaud,
2046: % {\it Subcritical Kelvin-Helmholtz Instability in a Hele-Shaw Cell},
2047:  Phys. Rev. Lett. {\bf 90}, 234502 (2003).
2048:  %------------------------------------------------*****66
2049:  \bibitem{Buechel00}
2050:  P.~B\"uchel and M.~L\"ucke,
2051: % {\it Localized perturbations in binary fluid convection with and without throughflow}
2052:  Phys.~Rev.~E {\bf 63}, 016307 (2000).
2053:  %------------------------------------------
2054:  \bibitem{HoBuLu97}
2055:  St.~Hollinger, P.~B\"uchel, and M.~L\"ucke, 
2056: % {\it Bistability of slow and fast traveling waves in fluid mixtures},
2057:  Phys.~Rev.~Lett. {\bf78}, 235 (1997).
2058:  %------------------------------------------------
2059:  \bibitem{Linz88}
2060:  S.~J.~Linz, M.~L\"ucke, H.~W.~M\"uller, and J.~Niederl\"ander,
2061: % {\it Convection in Binary Fluid Mixtures: Traveling Waves and Lateral Currents},
2062:  Phys.~Rev. {\bf A38}, 5727 (1988).
2063:  %------------------------------------------
2064:  \bibitem{Barten89}
2065:  W.~Barten, M.~L\"ucke, W.~Hort, and  M.~Kamps,
2066: % {\it Fully Developed Traveling Wave Convection in Binary Fluid Mixtures},
2067:  Phys.~Rev.~Lett. {\bf 63}, 376 (1989).
2068:  %------------------------------------------
2069:  \bibitem{Baxter92}
2070:  G.~W.~Baxter, K.~D.~Eaton, and C.~M.~Surko,
2071: % {\it Eckhaus instability for traveling waves}
2072:  Phys.~Rev.~A {\bf 46}, R1735 (1992).
2073:  %-----------------------------------------
2074:  \bibitem{Kolodner92}
2075:  P.~Kolodner,
2076: % {\it Observations of the Eckhaus instability in one-dimensional traveling-wave convection}
2077:  Phys.~Rev.~A {\bf 46}, R1739 (1992);
2078: % P.~Kolodner,
2079: % {\it Extended states of nonlinear traveling-wave convection. I. The Eckhaus instability}
2080:  {\it ibid.} {\bf 46}, 6431 (1992).
2081:  %-----------------------------------------********73
2082:  \bibitem{MeAlBa04}
2083:  I.~Mercader, A.~Alonso, and O.~Batiste, 
2084: % {\it Numerical analysis of the Eckhaus instability in travelling-wave
2085: % convection in binary mixtures},
2086:  Eur.~Phys.~J.~E.~{\bf15}, 311 (2004).
2087:  %------------------------------------------------
2088:  \bibitem{Thual8890}
2089:  O.~Thual and S.~Fauve,
2090: % {\it Localized structures generated by subcritical instabilities}
2091:  J.~Phys.~France {\bf 49}, 1829 (1988);
2092:  %\bibitem{Fauve90}
2093:  S.~Fauve and O.~Thual,
2094: % {\it Solitary Waves Generated by Subcritical Instabilities in Dissipative Systems}
2095:  Phys.~Rev.~Lett. {\bf 64}, 282 (1990).
2096:  %------------------------------------------
2097:  \bibitem{Hakim90}
2098:  V.~Hakim, P.~Jakobsen, and Y.~Pomeau,
2099: % {\it Fronts vs Solitary Waves in non Equilibrium Systems}
2100:  Europhys.~Lett. {\bf 11}, 19 (1990).
2101:  %------------------------------------------
2102:  \bibitem{Malomed90}
2103:  B.~A.~Malomed and A.~A.~Nepomnyashchy,
2104: % {\it Kinks and solitons in the generalized Ginzburg-Landau equation}
2105:  Phys.~Rev.~A {\bf 42}, 6009 (1990).
2106:  %------------------------------------------
2107:  \bibitem{Luecke92}
2108:  M.~L\"ucke, W.~Barten, and M.~Kamps,
2109: % {\it Convection in binary mixtures: the role of the concentration field}
2110:  Physica D {\bf 61}, 183 (1992).
2111:  %-----------------------------------------
2112:  \bibitem{Cross88Knobloch88}
2113:  M.~C.~Cross and K.~Kim,
2114: % {\it Linear instability and the codimension-2 region in binary fluid convection
2115: %      between rigid impermeable boundaries}
2116:  Phys.~Rev.~A {\bf 37}, 3909 (1988);
2117:  E.~Knobloch and D.~R.~Moore,
2118: % {\it Linear stability of experimental Soret convection}
2119:  {\it ibid.} {\bf 37}, 860 (1988).
2120:  %------------------------------------------
2121:  \bibitem{Zimmermann8993}
2122:  W.~Sch\"opf and W.~Zimmermann,
2123: % {\it Multicritical behaviour in binary fluid convection}
2124:  Europhys.~Lett. {\bf 8}, 41 (1989);
2125: % W.~Sch\"opf and W.~Zimmermann,
2126: % {\it Convection in binary fluids: Amplitude equations, codimension-2 bifurcation, and thermal fluctuations}
2127:  Phys.~Rev.~E {\bf 47}, 1739 (1993).
2128:  %------------------------------------------
2129:  \bibitem{Brand8990}
2130:  H.~R.~Brand and R.~J.~Deissler,
2131: % {\it Interaction of Localized Solutions for Subcritical Bifurcations}
2132:  Phys.~Rev.~Lett. {\bf 63}, 2801 (1989);
2133:  R.~J.~Deissler and H.~R.~Brand, 
2134: % {\it The Effect of Nonlinear Gradient Terms on Localized States near a Weakly
2135: %      Inverted Bifurcation}
2136:  Phys.~Lett.~A {\bf 146}, 252 (1990).
2137:  %------------------------------------------
2138:  \bibitem{Bestehorn888991}
2139:  M.~Bestehorn, R.~Friedrich, and H.~Haken,
2140: % {\it The oscillatory instability of a spatially homogeneous state in large aspect ratio 
2141: %      systems of fluid dynamics}
2142:  Z.~Phys.~B {\bf 72}, 265 (1988);
2143: % {\it Two-dimensional traveling wave patterns in nonequilibrium systems}
2144: % Z.~Phys.~B 
2145:  {\bf 75}, 265 (1989);
2146: % {\it Traveling waves in nonequilibrium systems}
2147:  Physica D {\bf 37}, 295 (1989);
2148:  M.~Bestehorn,
2149: % {\it Stationary and Travelling Pulses of the 2D Complex Ginzburg-Landau Equation}
2150:  Europhys.~Lett. {\bf 15}, 473 (1991).
2151:  %------------------------------------------
2152:  \bibitem{Bestehorn90}
2153:  M.~Bestehorn and H.~Haken,
2154: % {\it Traveling waves and pulses in a two-dimensional large-aspect-ratio system}
2155:  Phys.~Rev.~A {\bf 42}, 7195 (1990).
2156:  %------------------------------------------
2157:  \bibitem{RieckeL92}
2158:  H.~Riecke,
2159: % {\it Self-Trapping of Traveling-Wave Pulses in Binary Mixture Convection}
2160:  Phys.~Rev.~Lett. {\bf 68}, 301 (1992).
2161:  %------------------------------------------
2162:  \bibitem{RieckeD92}
2163:  H.~Riecke,
2164: % {\it Ginzburg-Landau equation coupled to a concentration field in binary-mixture convection}
2165:  Physica D {\bf 61}, 253 (1992).
2166:  %------------------------------------------
2167:  \bibitem{Riecke01}
2168:  A.~Roxin and H.~Riecke,
2169: % {\it Destabilization and localization of traveling waves by an advected field}
2170:  Physica D {\bf 156}, 19 (2001).
2171:  %------------------------------------------
2172:  \bibitem{Riecke96}
2173:  H.~Riecke,
2174: % {\it Solitary waves under the influence of a long-wave mode}
2175:  Physica D {\bf 92}, 69 (1996).
2176:  %------------------------------------------
2177:  \bibitem{RieckeD95}
2178:  H.~Herrero and H.~Riecke,
2179: % {\it Bound pairs of fronts in a real Ginzburg-Landau equation coupled to a mean field}
2180:  Physica D {\bf 85}, 79 (1995).
2181:  %------------------------------------------
2182:  \bibitem{RieckeL95}
2183:  H.~Riecke and W.~J.~Rappel,
2184: % {\it Coexisting Pulses in a Model for Binary-Mixture Convection}
2185:  Phys.~Rev.~Lett. {\bf 75}, 4035 (1995).
2186:  %------------------------------------------
2187:  \bibitem{Chomaz92} 
2188:  J.~M.~Chomaz, 
2189: % {\it Absolute and Convective Instabilities in Nonlinear Systems},
2190:  Phys.~Rev.~Lett.~{\bf 69}, 1931 (1992).
2191:  %------------------------------------------------*****90
2192:  \bibitem{CJT97}
2193:  Ch.~Jung,
2194:  {\it Numerische Simulationen dreidimensionaler Konvektionsstrukturen in 
2195:  bin\"aren Fluiden mit positiver Soretkopplung},
2196:  Ph.D.~thesis, Universit\"at des Saarlandes, 1997.
2197:  %-------------------------------------------------------------------
2198:  \bibitem{Catton72}
2199:  I.~Catton,
2200: % {\it The effect of insulating vertical walls on the onset of motion in a fluid heated from below}
2201:  J.~Heat Mass Transfer {\bf 15}, 665 (1972).
2202:  %------------------------------------------
2203:  \bibitem{Ohlsen90}
2204:  D.~R.~Ohlsen, S.~Y.~Yamamoto, C.~M.~Surko, and P.~Kolodner,
2205: % {\it Transition from Traveling-Wave to Stationary Convection in Fluid Mixtures}
2206:  Phys.~Rev.~Lett. {\bf 65}, 1431 (1990).
2207:  %-------------------------------------------------------------------
2208:  \bibitem{BuLu99}
2209:  P.~B\"uchel and M.~L\"ucke, 
2210: % {\it Sidewall induced Eckhaus instability of propagating wave convection 
2211: % in binary fluid mixtures},
2212:  Entropie~{\bf 218}, 22 (1999).
2213:  %-------------------------------------------------------------------
2214: \end{thebibliography}
2215: 
2216:  \clearpage
2217: 
2218:  %-----------------FIG 1 --------------------------------------------------
2219:  \begin{figure}
2220:  \includegraphics[clip=true,angle=0,width=14cm]{./FIG1.eps}
2221:  \caption{Typical structures of a coherent $+$front (left) and of a $-$front 
2222:  (right) with nearly the same asymptotic wavelength $\lambda= 1.90$.
2223:  Only the vicinity of the respective interfaces between
2224:  convection and conduction is shown. 
2225:  Both fronts propagate to the left ($v^+_F=-0.022, v_F^-=-0.067$), i.e., 
2226:  opposite to the direction of the phase velocity $v_p$.
2227:  (a, b) Color coded snapshot of concentration deviation $\delta C$ from its 
2228:  global 
2229:  mean in a vertical cross section of the layer. The color code is shown at the
2230:  right end of (b).
2231:  (c, d) Instantaneous lateral wave profile at midheight, $z=0$, of $\delta C$ 
2232:  (green),  vertical velocity $w$ (blue), and its envelope.
2233:  Arrows mark the positions where $w$ has grown up to $v_p$. 
2234:  (e, f) Mixing number $M$ (green), Eq.~(\ref{Eq:M(x)}), and phase velocity 
2235:  $v_p$ (black) of the nodes
2236:  of $w$ in the comoving frames. The variation of $\lambda(x)=2\pi v_p(x)/\omega$
2237:   is the
2238:  same since the frequency $\omega$ is a {\it global} constant.
2239:  (g, h) Time averaged deviations from the conductive state at $z=-0.25$ for 
2240:  concentration
2241:  (green), temperature (red), and their sum ($<b>$) measuring the convective 
2242:  contribution to the buoyancy.
2243:  (i, j) Streamlines of the averaged concentration current $<\vec{J}>$ (green) and
2244:  velocity field $<\vec{u}>$ (blue). The latter results from $<b>$ and documents 
2245:  roll shaped
2246:  contributions of $<\vec{u}><\delta C>$ to $<\vec{J}>$ at the interfaces.
2247:  Thick blue and green arrows indicate $<\vec{u}>$ and transport of positive
2248:  $\delta C$ (alcohol surplus), respectively.
2249:  Temporal averaging is always performed in the frame comoving with the 
2250:  respective front velocity. Dashed lines show the decay
2251:  part of the long LTW that coexists at the same parameters 
2252:  ($r=1.3586, \psi=-0.35, L=0.01$) with the fronts. 
2253:  Differences between the interfaces of the $+$front and the corresponding LTW 
2254:  interface are not visible on the scale of the above plots.
2255:  \label{Fig:frontpics}}
2256:  \end{figure}
2257:  %\clearpage
2258: %-----------------FIG 2 --------------------------------------------------
2259:  \begin{figure}
2260:  %
2261:  \includegraphics[clip=true,width=8.0cm,angle=0]{./FIG2.eps}
2262:  \caption{ Front and LTW bifurcation properties versus reduced Rayleigh number 
2263:  $r=R/R^0_c$ for $\psi=-0.25$.
2264:  Left and right pointing triangles with triangles denote $+$ and $-$fronts, 
2265:  respectively. Open and shaded circles refer to long and  short LTWs, 
2266:  respectively.
2267:  (a) Front velocities of relaxed single-front states (thick dashed lines with 
2268:  filled triangles), of
2269:  expanding two-front states (thin dashed lines with open triangles), and drift 
2270:  velocities of LTWs (circles).
2271:  The inset shows the drastic increase of LTW length $l$ at
2272:  $r^{LTW}_\infty=r^F_{eq}$ where the 
2273:  front velocities of the $+$ and $-$ single-front states become equal.  
2274:  (b) Frequencies of front states and of long LTWs in comparison with the rest 
2275:  frame frequencies of laterally periodic TWs. The saddle-node
2276:  vicinities of the latter are shown by full lines for several wave numbers 
2277:  $k=2\pi/\lambda$. TW states with frequencies above the dotted line of 
2278:  saddle-node TWs are unstable.   
2279:  (c) Wave numbers selected by front states in the bulk part far away from the
2280:  interface (triangles) and in the central part of long LTWs (open circles). 
2281:  Horizontal lines indicate the laterally periodic TWs that are shown in (b) by
2282:  full lines. The continuum of these TW states is bounded in the $r-k$ plane by 
2283:  the dotted line of TW saddle nodes, 
2284:  $k_s^{TW}(r)$. Here we show only the large-$k$ branch of it (cf. the dotted
2285:  line marked $r_s^{TW}$ in Fig.~\ref{Fig:3DPsi40om-k-r} for another 
2286:  perspective).
2287:  %\label{Fig:Psi25fronts}
2288:  \label{Fig:Psi25v+om+k-r}}
2289:  \end{figure}
2290:  %\clearpage
2291:  %-----------------FIG 3-----------------------------------------------------------
2292:  \begin{figure}
2293:  \includegraphics[clip=true, angle=0,width=8.5cm]{./FIG3.eps}
2294:  \caption{ Front states and LTWs for different $\psi$.
2295:  (a) Front velocities, $v_F$, of $+$fronts (left pointing triangles), of  $-$ 
2296:  fronts (right pointing triangles), and drift velocities, $v_d$, of LTWs
2297:  (circles) versus $r$. Note that short LTWs (shaded circles) and long LTWs 
2298:  (open circles) show different $v_d(r)$-behavior.
2299:  The latter varying linearly with $r$ is nearly indistinguishable from $v^+_F(r)$. 
2300:  Arrows mark the low-$r$ existence boundary $r_{min}^{TW}$ of laterally 
2301:  periodic TWs
2302:  and with it of front states. LTWs exist below this threshold \cite{Jung02} with
2303:  drift velocities that show the above mentioned linear variation
2304:  with $r$ as long as the LTWs are long enough. To identify an LTW as a long one
2305:  we required a clearly visible plateau in the spatial properties. 
2306:  (b) Length $l$ of the LTWs of (a) measured as the distance between 
2307:  the half maximum values of the envelope of the vertical velocity field $w$ 
2308:  [cf., blue line in Fig.~\ref{Fig:frontpics}(c,d)].
2309:  (c) Maximal vertical flow velocities $w_{max}$ of LTWs.
2310:  \label{Fig:3Psisv+l+w-r}}
2311:  \end{figure} 
2312:  %\clearpage
2313:  %-------------Fig 4 ---------------------------------------------------
2314:  \begin{figure}
2315:  \includegraphics[clip=true, angle=0,width=8.5cm]{./FIG4.eps}
2316:  \caption{ Wave properties of $+$fronts (dashed lines with triangles), 
2317:  long LTWs (circles), and laterally extended saddle-node TWs (dotted 
2318:  lines) for different $\psi$. The wave numbers of the two former refer to
2319:  plateau values in the bulk part away from the interface. 
2320:  (a) Frequency $\omega$ (for TWs in the rest frame and for LTWs and fronts in
2321:  the comoving frame) versus wave number $k$.
2322:  (b) The same convection structures in the $k-r$ plane.
2323:  The wave numbers and frequencies of LTWs with
2324:  $l \to \infty$ (arrows) coincide at $r^{LTW}_\infty=r^F_{eq}$ with those of 
2325:  the respective $+$  fronts.
2326:  \label{Fig:4Psisom+r-k} }
2327:  \end{figure}
2328:  %\clearpage
2329:  %------------Fig 5------------ ----------------------------------------------------
2330:  \begin{figure}
2331:  \includegraphics[clip=true, angle=-90,width=8.5cm]{./FIG5.eps}
2332:  \caption{ Laterally periodic TWs, $+$fronts, and LTWs
2333:  in the three dimensional $k-r-\omega$ parameter space.
2334:  Grey, nose-shaped surface \cite{Jung02} denotes TWs. They are unstable when
2335:  $\omega$ is above the dotted line of saddle nodes. Its projection onto the
2336:  $k-r$ is marked by $r^{TW}_s(k)$. A particular bifurcation branch for a given 
2337:  $k$ (e.g., the long-dashed line for $k\sim \pi$) starts backwards with a large
2338:  Hopf frequency (not shown) and becomes stable by a saddle-node bifurcation at 
2339:  the dotted line.
2340:  The big plus signs on the TW surface mark asymptotic $+$fronts. The small
2341:  plusses at large $k$ denote the 
2342:  highest wave numbers occurring at the $+$front interface. Long LTWs are 
2343:  represented with their plateau values by big bullets. 
2344:  They coexist with fronts (big plusses) in a very narrow $r$ interval at 
2345:  $r^{LTW}_\infty$ close to the tip of the TW nose. The small bullets at large $k$ 
2346:  denote the largest local wave numbers occurring at the interface of convection
2347:  growth. Each horizontal line indicates at fixed $r$ and $\omega$ the spatial 
2348:  variation 
2349:  of the local wave number $k$ within a $+$front or a long LTW from the growth 
2350:  interface to the asymptotic plateau value. The $k$ variation of LTWs from 
2351:  plateau to the decay interface into conduction is not shown.
2352:  \label{Fig:3DPsi40om-k-r}}
2353:  \end{figure}
2354:  %\clearpage
2355:  %----------Fig 6---------- ----------------------------------------------------
2356:  \begin{figure}
2357:  \includegraphics[clip=true, width=15cm]{./FIG6.eps}
2358:  \caption{ Typical spatio-temporal evolution of a $+$front. Shown are the 
2359:  extrema positions of the vertical velocity field $w$. 
2360:  The initial condition at time $t=-5$ (not visible) consisted of an extended TW 
2361:  for $x>8$ with wavelength $\lambda=1.85$ and phase velocity $v_p=1.032$ 
2362:  and quiescent fluid for $x<8$.
2363:  Boundary conditions are conduction at $x=0$ and $f(x=160)= f(x=160-1.85)$ 
2364:  that imposes at $x=160$ a wavelength of $\lambda=1.85$.
2365:  First, a pulse that causes a quite regular sequence of roll-pair annihilation 
2366:  events (lower line of defects) propagates to the right with velocity greater 
2367:  than 
2368:  $v_p$. The intermediate wave pattern resulting from this primary sequence of
2369:  defects is then transformed via further, somewhat erratically occurring
2370:  phase defects into the fast asymptotic TW with $\lambda=1.80,v_p=1.258$ that is
2371:  favoured by the $+$front. The boundary condition 
2372:  at $x=160$ that imposes a 'wrong' wavelength there does not influence the 
2373:  bulk behavior which is selected by the front.
2374:  Parameters are $r=1.237, \psi=-0.25$. 
2375:  \label{Fig:frontrelaxation}}
2376:  \end{figure}
2377:  %\clearpage
2378:  %--------Fig 7------------ ----------------------------------------------------
2379:  \begin{figure}
2380:  \includegraphics[clip=true,width=15cm]{./FIG7.eps}
2381:  \caption{ Spatio-temporal dynamics of a front-selected TW pattern that is 
2382:  convectively Eckhaus unstable. Shown are the  extrema positions of the 
2383:  vertical velocity field $w$. The front selects a bulk TW 
2384:  with wavelength $\lambda \sim 1.72$ that is strongly Eckhaus unstable:
2385:  While beeing advected 'downstream' phase deformations (that are caused, e.g.,
2386:  by computer 'noise') grow sufficiently fast to reach a critical amplitude 
2387:  within the system length. Then two neighboring rolls are annihilated. That 
2388:  increases the wavelength and decreases the phase velocity towards Eckhaus 
2389:  stable values. Parameters are $r=1.26, \psi=-0.25$.
2390:  \label{Fig:irregularfront}}
2391:  \end{figure}   
2392: %\clearpage
2393: %-----------Fig 8--- --------------------------------------------------
2394:  \begin{figure}
2395:  \includegraphics[clip=true, angle=0, width=8.5cm]{./FIG8.eps}
2396:  \caption{ Evolution of the lateral profiles of the wavelength (top) and of 
2397:  the vertical flow amplitude $w_{max}/10$ (bottom) after 
2398:  starting with a very long two-front structure at $r=1.3586, \psi=-0.35$. 
2399:  The $+$front selects in the bulk of the initial two-front structure a 
2400:  saddle-node TW with wavelength $\lambda_{plateau}\sim 1.905$. With  
2401:  $v^-_F<v^+_F<0$
2402:  the $-$front approaches the $+$front and doing so the velocity of the former
2403:  goes monotonously towards $v_F^+$. This transient process ends in a long LTW 
2404:  (dashed line) of constant length $l \simeq 47$ with a plateau wavelength of $1.873$.
2405:  Its drift and frequency is effectively the same as the respective values of 
2406:  the $+$front which remain unchanged all the time.
2407:  For comparison the profiles of a short LTW at the same $\psi$ but smaller  
2408:  $r=1.3220$ are shown with dotted lines.
2409:               \label{Fig:LTWapproach}}
2410:  \end{figure}
2411: %\clearpage
2412: %----------Fig 9 ------------------------------------------
2413:  \begin{figure}
2414:  \includegraphics[clip=true, angle=0, width=8.5cm]{./FIG9.eps}
2415:  \caption{Phase diagram in the $\psi -r$ plane. The vertical bars indicate 
2416: the range of stable existence of those LTWs that we
2417: have numerically simulated. Full and dashed lines refer to the saddle node
2418: location $r_s^{TW}$ of extended TWs and to their 
2419: oscillatory Hopf bifurcation threshold $r_{osc}$, respectively; both for a 
2420: wave number $k = \pi$ . For $\psi \leq -0.25$ the upper existence boundary of
2421: LTWs was determined by the requirement that $l$ remained below about 120 in 
2422: our numerical set-up. The dotted line guides the eye along the lower 
2423: band limit $r^{LTW}_{min}$ of LTWs. Parameters are $L = 0.01, \sigma = 10$.
2424:  \label{Fig:phasediagram}}
2425:  \end{figure}
2426:  %\clearpage
2427:  %----------Fig 10------ ----------------------------------------------------
2428:  \begin{figure}
2429:  \includegraphics[clip=true, angle=0,width=8.5cm]{./FIG10.eps}
2430:  \caption{ Frequency $\omega$ of $+$fronts (dashed lines with triangles) and 
2431:  of LTWs (full lines with circles) in the respective comoving frame versus 
2432:  $r$ for different $\psi$. Open and shaded circles refer to long and short LTWs,
2433:  respectively. The frequencies of the former are the same as those of the 
2434:  fronts while short LTWs differ. Arrows indicate the lower limit of existence 
2435:  of the fronts at $r^F_{min}=r^{TW}_s(k \simeq \pi)$. 
2436:  %\label{Fig:fandLvaluesB}
2437:  \label{Fig:4Psisom-r}}
2438:  \end{figure}
2439:  %\clearpage
2440:  %----------Fig 11----------------------------------------------------------
2441:  \begin{figure}
2442:  \includegraphics[clip=true, angle=0,width=8.5cm]{./FIG11.eps}
2443:  \caption{ Schematic bifurcation diagrams of LTW length $l$ versus $r$. (a)
2444:   experimental results \cite{Kolodner94} obtained for $\psi=-0.127$ with dashed
2445:   line denoting unstable states (cf., text for further explanation); (b) 
2446:   numerically obtained bifurcation behavior for $-0.4 \leq \psi \leq -0.25$. 
2447:  \label{Fig:l-r-scheme}}
2448:  \end{figure}
2449:  %------------------------------------------------------------------------
2450: 
2451:  
2452:  
2453:  
2454:  \end{document}
2455:  
2456: