physics0304092/q.tex
1: %----------------- Settings ----------------------------
2:  \documentclass[preprint, aps, pre, eqsecnum, amsmath,amssymb,showpacs]{revtex4}
3:  %\documentclass[preprint, aps, pre, preprintnumbers,amsmath,amssymb,showpacs]{revtex4}
4:  %\documentclass[preprint, 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:  %\usepackage{psfig}
10: 
11:  \renewcommand{\arraystretch}{1.5}  %faktor 1.5 fuer tabellenzeilenabstand                                             
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{Spiral and Taylor vortex fronts and pulses in axial through-flow}                                
31: 
32:  \author{A.~Pinter, M.~L\"ucke, and Ch.~Hoffmann}                                   
33:  \affiliation{Institut f\"ur Theoretische Physik, Universit\"at des Saarlandes,      
34:  Postfach 151150, \\ D-66041 Saarbr\"ucken, Germany}                              
35: 
36:  \date{\today} 
37:                                                                    
38: %-------------- Abstract --------------------------------
39:  \begin{abstract}
40: %-----------------------------------------------------
41: The influence of an axial through-flow on the spatiotemporal
42: growth behavior of different vortex structures in  
43: the Taylor-Couette system with radius ratio $\eta=0.5$ is determined.  
44: The Navier Stokes equations (NSE) linearized around the basic 
45: Couette-Poiseuille flow are solved numerically with a shooting method in a wide
46: range of through-flow strengths $Re$ and different rates of co- and 
47: counterrotating cylinders for
48: toroidally closed vortices with azimuthal wave number $m=0$ and for spiral vortex
49: flow with $m= \pm 1$. For each of these three different vortex varieties we have
50: investigated ({\it i}) axially extended vortex structures, ({\it ii}) axially 
51: localized vortex pulses, and ({\it iii}) vortex fronts.
52: The complex dispersion relations of the linearized NSE for vortex modes with 
53: the three different $m$ are evaluated for real axial wave numbers for ({\it i}) 
54: and over the plane of complex axial wave numbers for ({\it ii}, {\it iii}).
55: We have also determined the Ginzburg-Landau amplitude equation (GLE) approximation 
56: in order to analyze its predictions for the vortex stuctures 
57: ({\it ii}, {\it iii}). Critical bifurcation thresholds for 
58: extended vortex structures are evaluated. The boundaries between 
59: absolute and convective instability of the basic state for vortex pulses
60: are determined with a saddle-point analysis of the dispersion relations.
61: Fit parameters for power-law expansions of the boundaries up to $Re^4$ are 
62: listed in two tables. Finally, the linearly selected front behavior of growing 
63: vortex structures is investigated using saddle-point analyses of the dispersion 
64: relations of NSE and GLE. For the two front intensity 
65: profiles (increasing in positive or negative axial direction) we have determined
66: front velocities, axial growth rates, and the wave numbers and frequencies of
67: the unfolding vortex patterns with azimuthal wave numbers $m=0, \pm 1$, 
68: respectively.  
69:  \end{abstract}
70:  
71: \pacs{PACS number(s): 47.20.-k, 47.54.+r, 47.32.-y, 47.10.+g}
72: %****************************************************************
73: % 47.20.-k : Hydrodynamic stability
74: % 47.54.+r   Pattern selection; pattern formation
75: % 47.32.-y   Rotational flow and vorticity
76: % 47.10.+g : General theory (of fluid dynamics)  
77: % 47.60.+i   Flows in ducts, channels, nozzles, and conduits
78: %****************************************************************
79:  \maketitle 
80:                                                                      
81:  %\newpage
82:  % skip 2/6inch
83:  %\vskip2pc
84: 
85:  %im gegenstz zu \widetext (z.b. bei formeln)
86:  %halbe zeitenbreite f"ur spaltentext
87:  %\narrowtext                                                                     
88: %\tableofcontents
89:  %\draft
90:  
91: 
92: %-------------- Sec. I ---------------------------------
93:  \section{INTRODUCTION} \label{SEC.INTRO}
94: The Taylor-Couette system \cite{review} of fluid flow in the annulus between
95: concentric cylinders with the inner 
96: and the outer one rotating with different velocities is one of the simplest 
97: examples of a driven nonlinear dissipative system that shows spontaneous pattern
98: formation out of an unstructured basic state that is stable at small driving
99: \cite{CRO-HOH}. 
100: This basic flow state is stationary and axially and azimuthally homogeneous and
101: shows only a radial variation across the annular gap. It consists of a
102: superposition of circular Couette flow (CCF) in azimuthal direction 
103: and of an annular Poiseuille flow (APF) in axial direction if as in our case 
104: an axial through-flow is imposed. Axially periodic vortex flow solutions 
105: bifurcate \cite{CHO-IOO,GOL-STE-SCH-LAN} out of this homogeneous 
106: basic flow when the rotation rate of the inner cylinder is sufficiently 
107: high. These primary bifurcation thresholds to periodic vortex stuctures have
108: been the aim of many linear stability analyses of the basic flow state  
109: \cite{DIP-PRI,TAK-JAN,NG-TUR,LAN-TAG-KOS-SWI-GOL,GEB-GRO,MES-MAR}.
110: 
111: For the radius ratio $\eta =0.5$ and the parameter ranges of rotation
112: rates and through-flow investigated here in this work three spatiotemporally
113: differing primary vortex structures are relevant: Rotationally 
114: symmetric, toroidally closed vortices with
115: azimuthal wave number $m=0$ that move in downstream direction with the APF --- 
116: for shortness we call this flow state Taylor vortex flow (TVF) although the 
117: presence of an axial through-flow modifies the genuine stationary TVF
118: stucture. And, furthermore, spiral vortex flow (SPI) consisting of either left 
119: spiral vortices (L-SPI) with $m=1$ or right spiral vortices (R-SPI) with 
120: $m=-1$.
121: 
122: L-SPI and R-SPI are axial mirror images of each other in the absence of axial 
123: through-flow with the latter breaking the mirror symmetry of the former. 
124: While rotating azimuthally into the same direction as the inner cylinder 
125: L-SPI propagate axially opposite to R-SPI. This spiral dynamics is largely 
126: induced by the advective properties of the basic flow state.
127: Furthermore, without through-flow the symmetry degenerate bifurcation treshold 
128: for these two symmetry degenerate SPI
129: solutions is simultaneously also the bifurcation threshold for a vortex flow 
130: solution called ribbons \cite{CHO-IOO}. This solution consists right at threshold
131: of a linear superposition
132: of L-SPI and R-SPI with equal amplitude and it becomes further away from 
133: threshold a genuine nonlinear vortex flow solution. However, here 
134: we are dealing only with linear vortex flow fields that may be superimposed 
135: with arbitrary amplitudes as well as wave numbers and that are evolving 
136: separately from each other according to the linear field equations. Thus we do 
137: not need to discuss
138: ribbons separately from our general investigation of linear vortex modes with
139: general axial and azimuthal wave numbers. 
140: 
141: In this work we quantitatively determine the influence of an axial through-flow
142: on the spatiotemporal growth properties of linear perturbations of the basic
143: flow state with azimuthal wave numbers $m=0$ and $m=\pm 1$, i.e., of toroidally
144: closed vortices and of spiral vortices, repectively. In each case we investigate
145: ({\it i}) axially extended structures, ({\it ii}) pulses of axially localized 
146: wave packets of vortices, and ({\it iii}) vortex fronts. 
147: 
148: In Sec.~\ref{SEC.SYSTEM} we describe the system, we
149: briefly review the linearized Navier-Stokes equations (NSE) for the eigenvalue 
150: problem describing vortex perturbations, and
151: we give details of our numerical procedure to solve the eigenvalue problem.
152: In Sec.~\ref{SEC.EXTENDED_VORTICES}
153: we discuss the spatiotemporal structure, symmetry properties, and
154: bifurcation thresholds for onset of axially extended
155: vortex perturbations of the form $e^{i(kz + m\varphi)}$ with real axial wave 
156: number $k$ and different azimuthal wave numbers $m$ in the absence and presence 
157: of an axial through-flow. In Sec.~\ref{SEC.LOCAL-PERTUB}
158: we consider axially localized wave packets 
159: consisting of superpositions of vortex eigenmodes of the linear NSE. Here we
160: determine among others the boundary between convective and absolute instability
161: of the basic flow against growth of vortices with a particular $m$ by a 
162: saddle point analysis of the linear complex dispersion relation of the NSE
163: over the plane of complex axial wave numbers. In addition we also determine 
164: the Ginzburg-Landau amplitude 
165: equation (GLE) approximation for the dispersion relation for the sake of
166: comparison. In Sec.~\ref{FRONTS} we evaluate the spatiotemporal properties of
167: linearly selected vortex fronts using a saddlepoint analysis of the dispersion 
168: relation. Also here we compare with GLE results. The final section contains a
169: summary.
170: 
171: %\clearpage
172: %-------------- Sec. II --------------------------------
173:  \section{System} \label{SEC.SYSTEM} 
174: %--------------------------------------------------------- 
175: Here we describe the system and we provide definitions and equations. Then we
176: briefly review the linearized equations for the eigenvalue problem describing 
177: vortex perturbations of the basic flow state. Finally
178: we give details of our numerical procedure to solve the eigenvalue problem.
179: %-------------- Sec. II A--------------------------------
180:  \subsection{Setup} \label{SEC.SETUP} 
181: %--------------------------------------------------------- 
182: We consider the flow of an incompressible fluid in the annulus between two
183: concentric cylinders of inner radius $r_1$ and outer radius $r_2$ with a gap width
184: $d=r_2-r_1$. The boundary conditions at $r_1$ and $r_2$ are no-slip. The angular
185: velocity of the inner and outer cylinder is $\Omega_{1}$ and $\Omega_{2}$, 
186: respectively. The associated Reynolds numbers are 
187: \begin{equation}
188: R_{1}=\frac{d}{\nu}r_1\Omega_{1}, \quad
189: R_{2}=\frac{d}{\nu}r_2\Omega_{2},
190: \end{equation}
191: where $\nu$ is the kinematic viscosity. An externally imposed axial through-flow
192: is measured by the axial Reynolds number
193: \begin{equation}
194: Re=\frac{d}{\nu} \langle w \rangle
195: \end{equation}
196: where the mean axial velocity $\langle w \rangle$ averaged over the annular 
197: cross section describes the total through-flow. We use also the relative control 
198: parameters 
199: \begin{equation}
200: \mu = \frac{R_{1}}{R_{1c}(Re)}-1,\quad \epsilon = \frac{R_{1}}{R_{1c}(Re=0)}-1,
201: \end{equation}
202: measuring the relative distance of the inner Reynolds number $R_{1}$ from
203: the critical onset $R_{1c}$ of axially extended spiral vortices or Taylor 
204: vortices in the presence and in the absence $(Re=0)$ of through-flow, 
205: respectively \cite{epsT}. In this notation 
206: \begin{equation} \label{epsilon_c}
207: \mu_c = 0 \quad \mbox{and} \quad 
208: \epsilon_c (Re) =  \frac{R_{1c}(Re)}{R_{1c}(Re=0)} -1
209: \end{equation}
210: is the critical threshold for onset of the vortex flow in question. The relation 
211: between $\mu$ and $\epsilon$ is 
212: \begin{equation}
213: \mu= \frac{\epsilon-\epsilon_{c}(Re)}{1+\epsilon_c (Re)}.
214: \end{equation}
215: With infinitely long cylinders the only relevant parameter
216: characterizing the geometry is the radius ratio $\eta=r_1/r_2$.
217: 
218: The velocity field $\bf u$ of the fluid is described by the Navier-Stokes 
219: equations (NSE) for incompressible fluid flow 
220: \begin{equation} \label{NAVIER2}
221: \partial_t {\mathbf u}= \Nabla^2 {\bf u} -
222:  R_{1} ({\bf u}\cdot \Nabla){\bf u} - \Nabla p,\quad  \Nabla \cdot {\bf u} = 0.
223: \end{equation}
224: Here and in the following we scale positions by the gap width $d$, the velocity
225: ${\bf u}$ by the velocity $r_1\Omega_{1}$ of the inner cylinder, time $t$ by
226: the momentum diffusion time $d^{2}/\nu$ across the gap, and the pressure $p$ by 
227: $\rho r_1 \Omega_{1}\nu/d$ with $\rho$ denoting the constant mass density of the
228: fluid. Furthermore we decompose the velocity field 
229: \begin{equation}
230: {\bf u} = u{\bf e}_r + v{\bf e}_{\varphi} + w{\bf e}_z
231: \end{equation}
232: into radial ($u$), azimuthal ($v$), and axial ($w$) components
233: using cylindrical coordinates $r,\varphi,z$.
234: %-------------- Sec. II B--------------------------------
235:  \subsection{Basic flow state} \label{SEC.BASIC_FLOW} 
236: %---------------------------------------------------------  
237: The basic flow state ${\bf u}_0$ that is realized
238: in the absolutely stable regime of inner Reynolds numbers $R_1$ below the 
239: thresholds for onset of Taylor and spiral vortex flow is rotationally symmetric,
240: axially homogeneous, and constant in time. It consists of a linear superposition 
241: of circular Couette flow (CCF) in azimuthal direction, ${\bf e}_{\varphi}$, and 
242: of annular Poiseuille flow (APF) in axial direction, ${\bf e}_z$,
243: \begin{equation}\label{u_0}
244: {\bf u}_0 = v_{CCF}(r){\bf e}_{\varphi}+w_{APF}(r){\bf e}_{z}
245: \end{equation} 
246: without any radial component. Here
247: \begin{equation} 
248: v_{CCF}(r)=Ar+B/r,
249: \end{equation}
250: and 
251: \begin{equation}
252: w_{APF}(r)=Re\frac{r^{2}+C ln(r)+D}{E},
253: \end{equation}
254: with 
255: \begin{eqnarray}
256: A&=&-\frac{\eta^{2}-\Omega_{2}/\Omega_{1}}{\eta(1+\eta)},\\
257: B&=&\frac{\eta(1-\Omega_{2}/\Omega_{1})}{(1-\eta)(1-\eta^{2})},\\
258: C&=&\frac{1+\eta}{(1-\eta)ln(\eta)},\\
259: D&=&\frac{(1+\eta)\ln(1-\eta)}{(1-\eta)\ln(\eta)}-\frac{1}{(1-\eta)^{2}},\\
260: E&=&-\frac{1}{2}R_{1}\frac{1-\eta^{2}+(1+\eta^{2})\ln(\eta)}{(1-\eta)^{2}\ln(\eta)}.
261: \end{eqnarray} 
262: %-------------- Sec. II C ------------------------------- 
263:  \subsection{Linear eigenvalue problem of vortex perturbations} 
264:  \label{SEC.LIN-STAB-ANALYSIS}
265: %---------------------------------------------------------------------- 
266: Let $\psi =(u,v,w,p)$ abbreviate the deviation fields from the basic flow state
267: (\ref{u_0}). Then the general solution of the NSE linearized in the deviation 
268: fields can be written as a superposition of modes of the form
269:  \begin{equation}\label{Ansatz}
270: \psi(r,\varphi,z,t) = \phi(r)e^{i(kz + m\varphi)}e^{\sigma t}
271: \end{equation} 
272: with axial wave number $k=2\pi/\lambda$ and integer azimuthal wave number $m$. 
273: The complex amplitude functions
274: \begin{equation}
275: \phi(r) = \left[ U(r),V(r), W(r),P(r) \right]
276: \end{equation}
277: depend on the mode indices $k,m$ and the radial coordinate $r$. The characteristic
278: exponent $\sigma(k,m)$ is in general complex. It is decomposed here as follows
279: \begin{equation}
280: \sigma = \Re \sigma + i\, \Im \sigma = \gamma - i \omega 
281: \end{equation}
282: into the growth rate $\gamma$ and the characteristic frequency $\omega$ of the 
283: $k-m$ mode. Substituting the above solution ansatz into the linearized NSE yields
284: \begin{eqnarray} 
285: \label{Linear1}
286: & & \sigma U = \left( \partial_r^2  + \frac{1}{r} \partial_r 
287: - \frac{1+m^2}{r^2} - k^2 \right) U + 2 F V \nonumber\\
288: & & \qquad - \partial_r P -im \left( \frac{2}{r^2} V + F U \right)- ik H U \\
289: \label{Linear2} 
290: & & \sigma V = \left( \partial_r^2 + \frac{1}{r}\partial_r 
291: - \frac{1+m^2}{r^2} - k^2 \right) V + 2GU \nonumber\\
292: & & \qquad -\frac{im}{r} P + im \left( \frac{2}{r^2} U - F V \right)-ikH V \\
293: \label{Linear3}
294: & & \sigma W = \left( \partial_r^2  + \frac{1}{r} \partial_r 
295: - \frac{m^2}{r^2} - k^2 \right) W -ikP \nonumber\\
296: & & \qquad -im F W - I U-ikH W \\
297: \label{Linear4}
298: & & \quad 0 = \partial_r U + \frac{1}{r} U + \frac{im}{r} V + ikW . 
299:   \end{eqnarray}
300: The solution of this eigenvalue problem yields the characteristic exponent 
301: $\sigma$ and the associated eigenfunctions $\phi(r)$ as functions of $k,m$. 
302: Here 
303:  \begin{eqnarray} \label{FGHIKonstanten}
304: F(r) & = &\frac{R_1}{r} v_{CCF}(r) \quad , \quad G = -R_1 A,\\
305: H(r) & = & R_1 w_{APF}(r) \quad , \quad I(r) = \partial_r H(r)
306: \end{eqnarray}
307: are quantities defining the basic flow state (\ref{u_0}). The latter enters via
308: the linearized advective term of the NSE.
309: 
310: In order to rewrite (\ref{Linear1} - \ref{Linear4}) into a system of first-order
311: differential equations -- which is advantageous for numerical reasons -- we
312: introduce three additional complex amplitude functions 
313: \cite{LAN-TAG-KOS-SWI-GOL}
314: \begin{eqnarray}
315: \label{DefX}
316: X &=& \partial_r U + \frac{1}{r} U - P \, , \\
317: Y &=& \partial_r V + \frac{1}{r} V \, , \\
318: Z &=& \partial_r W \, .  
319: \end{eqnarray}
320: Using (\ref{DefX}) and the continuity equation (\ref{Linear4}) one can then
321: eliminate the pressure in Eq.(\ref{Linear2}) by $P=-X-ikW-\frac{im}{r}V.$
322: All in all one obtains in this way a system of 6 coupled, first-order 
323: differential equations
324: \begin{equation} \label{resultEW}
325: \partial_r {\bf X} = {\cal L}\ {\bf X},
326: \end{equation}
327: for the six variables
328: \begin{equation} \label{start.vector}
329: {\mathbf X}=\left( U,V,W,X,Y,Z \right)^{T}
330: \end{equation}
331: with 
332: \begin{equation} \label{Result.Operator}
333: {\cal L} = \left( 
334: \begin{array}{cccccc} 
335: -\frac{1}{r} & - i\frac{m}{r} & -ik & 0 & 0 & 0\\  
336: 0 & -\frac{1}{r} & 0 & 0 & 1 & 0 \\ 
337: 0 & 0 & 0 & 0 & 0 & 1\\ 
338: L  & 2(i\frac{m}{r^2} - F) & 0 & 0 & 0 & 0 \\ 
339: - 2(i\frac{m}{r^2} + G) & L + \frac{m^2}{r^2}  & \frac{mk}{r} & -i\frac{m}{r} & 0 & 0 \\ 
340:  I & \frac{mk}{r} &  L + k^2 & -ik & 0 & -\frac{1}{r} \\
341: \end{array} \right) 
342: \end{equation}
343: where 
344: \begin{equation} \label{defineL}
345: L = \sigma + \frac{m^2}{r^2} + k^2 + imF + ikH.
346: \end{equation}
347: %-------------- Sec. II D -----------------------------
348:  \subsection{Numerical procedures}  \label{SEC.NUM-PROC1}
349: %------------------------------------------------------------------ 
350:  We have solved the eigenvalue equations (\ref{resultEW}) numerically with a
351:  standard shooting method subject to the six boundary conditions, 
352:  \begin{equation} \label{result.boundary}
353: U=V=W=0 \quad \textrm{at} \quad r_1=\eta/(1-\eta) \,\, 
354: \textrm{and} \,\, r_2 = 1/(1-\eta) \, ,
355: \end{equation}
356: which make the eigenvalue spectrum discrete.
357: To integrate from $r_1$ to $r_2$ we used a fourth-order Runge-Kutta formula 
358: \cite{STO-BUR} with two step widths ( $\Delta r = 1/200$ and  1/400,
359: respectively) for a Richardson
360: extrapolation. A Newton-Raphson method \cite{PRE-TEU-VET-FLA} was then used to 
361: find the roots of the complex 
362:  determinant of the $3 \times 3$ matrix which ensures the vanishing of 
363:  $U,V,W$ at the outer cylinder. We therefore vary in the Newton-Raphson 
364: procedure only two of the parameters ($\sigma,\eta, R_{1}, R_{2}, Re, m, k$)
365: \cite{DIP-PRI} that enter into (\ref{resultEW}) while keeping the others 
366: fixed. In this way we
367: determine on the one hand the marginal threshold values of $R_1$ and $\omega$ 
368: with $\gamma=0$ for which the basic state is marginally stable against the
369: growth of an extended perturbation with given $m$ and real axial wave number 
370: $k$ at specified parameters $\eta, R_2, Re$. On the other hand we calculate for
371: given $m, \eta, R_1, R_2, Re$ the complex eigenvalue $\sigma$ over the 
372: {\em complex} wave number plane -- including as special case also the real 
373: $k$-axis. In each case we are interested only in the vortex modes with the
374: largest growth rates for which the associated amplitude functions $\phi(r)$
375: display the least radial variation with the fewest number of nodes.
376: 
377: We present here results for the radius ratio $\eta=0.5$ in a range of outer
378: Reynolds numbers $-150 \leq R_2 \leq 50$. In this parameter regime the vortex 
379: perturbations with the
380: largest growth rates have in the absence of through-flow azimuthal wave numbers
381: of either $m=0$ or $m=\pm 1$ \cite{{LAN-TAG-KOS-SWI-GOL}}. We investigate here 
382: linear properties of such vortices with $m=0$ and $m=\pm 1$ in a through-flow of
383: Reynolds numbers $0\leq Re \leq 20$.
384: %\clearpage
385: %-------------- Sec. III  -----------------------------
386:  \section{Axially extended vortex structures} 
387:  \label{SEC.EXTENDED_VORTICES}
388: %------------------------------------------------------ 
389: Here we discuss the spatiotemporal structure, symmetry properties, and
390: bifurcation thresholds for onset of axially extended
391: vortex perturbations with real axial wave number $k$ and different azimuthal 
392: wave numbers $m$ in the absence and presence of an axial through-flow.
393: %-------------- Sec. III A -----------------------------
394:  \subsection{Spatiotemporal structure} \label{SEC.STRUCTURE}
395: %------------------------------------------------------ 
396: Structure and dynamics of the vortex modes (\ref{Ansatz}) are dominated by the 
397: fact that their phases are constant on any cylindrical surface,
398: $r=const$, along lines given by the equation
399:  \begin{equation} \label{z_phase}
400: z_0 = - \frac{m}{k}\varphi + \frac{\omega(k,m)}{k}t \, .
401: \end{equation}
402: Here the constant phase coming from the amplitude $\phi(r)$ is suppressed.
403: Thus, on the $\varphi-z$ plane of such an 'unrolled' cylindrical surface these
404: lines of constant phase are straight.
405: %------------------------------------------------------ 
406: \subsubsection{Taylor vortex like patterns --- $m$=0}
407:  %------------------------------------------------------ 
408:  For rotationally symmetric Taylor vortex like perturbations the line pattern of
409:  constant phases, $kz_0 = n 2 \pi$, is parallel to ${\bf e}_\varphi$. The $m$=0
410:  pattern is stationary for $Re=0$. Only for finite through-flow it
411:  propagates axially with phase velocity
412:  \begin{equation} \label{w_phase_TVF}
413: w_{phase} = \frac{\omega}{k}
414: \end{equation}
415: that is proportional to $Re$. The main reason for this is that the azimuthal 
416: flow of the basic
417: CCF state is precisely parallel to the vortex lines of constant phase
418: while the APF flow being perpendicular to them can advect them. The latter
419: happens in our axially periodic system that does not exert any phase pinning 
420: at the axial boundaries as soon as $Re > 0$. 
421:  %------------------------------------------------------ 
422: \subsubsection{Spiral patterns --- $m \neq 0$}
423:  %------------------------------------------------------ 
424: The vortex modes (\ref{Ansatz}) with axial wave number $m \neq 0$ have spiral
425: structure. When $m/k$ is positive (negative) the lines of constant phase
426: $z_0(\varphi,t)$ (\ref{z_phase}) wind in a left spiral L-SPI (right spiral 
427: R-SPI) around the cylindrical surface $r=const$ with negative (positive) slope
428: $\partial_{\varphi} z_0 = -m/k$. The lines of constant phase and with it the
429: whole spiral stucture rotates in $\varphi$ rigidly with angular velocity
430:   \begin{equation} \label{phidot_phase}
431: \dot{\varphi}_{SPI} = \frac{\omega}{m} \, .
432: \end{equation}
433: In the absence of an externally imposed through-flow, $Re$=0, this rotation
434: proceeds for L-SPI and R-SPI alike into the same direction as the rotation of 
435: the inner cylinder. The reason is that
436: the spiral perturbations are advected by the inner part of the azimuthal CCF
437: which is relevant for the centrifugal instability leading to vortex
438: generation. A model explaining this effect is presented in \cite{HOF-LUE}. 
439: 
440: There
441: are two immediate consequences of this advective origin of the spiral dynamics
442: induced by the inner cylinder's rotation: ({\it i}) With the latter being
443: by definition positive -- in this work the inner cylinder is taken to rotate in  
444: positive $\varphi$-direction -- also $\omega (k,m) /m $ is positive for $Re$=0.
445: Hence, say, an $m=1$ ($m=-1$) spiral has positive (negative) frequency for 
446: $Re$=0. ({\it ii}) Then a L-SPI (R-SPI) being defined by  $m/k > 0$ ($m/k < 0$)
447: propagates for $Re$=0 upwards (downwards) with positive (negative) axial phase 
448: velocity
449: \begin{equation} 
450: w_{phase} = \frac{\omega}{k} = \frac{m}{k} \dot{\varphi}_{SPI}
451: \end{equation}
452: that is directly related to its {\em positive} angular
453: velocity $\dot{\varphi}_{SPI}$.
454: 
455: An externally applied axial through-flow changes the axial phase velocities and
456: frequencies of the $m=0$ and $m \neq 0$ vortex modes roughly proportional to 
457: $Re$, i.e., 
458: \begin{equation} 
459: w_{phase}(k,m,Re) - w_{phase}(k,m,Re=0) \propto Re \, .
460: \end{equation}
461: Simultaneously the rotation rates, $\dot{\varphi}_{SPI} = w_{phase} k /m$, of
462: the spirals are changed accordingly. Thus, in each case the vortex frequencies 
463: are largely determined by the basic state's advective properties, i.e., by the 
464: combination of azimuthal advection by $v_{CCF}$ and axial advection by 
465: $w_{APF}$.
466: %-------------- Sec. III B -----------------------------
467:  \subsection{Symmetries} \label{SEC.SYMM-EIGENVALUE}
468: %------------------------------------------------------ 
469: Here we consider symmetry properties of axially extended vortex perturbations of
470: the form (\ref{Ansatz}) with real wave number $k$. Symmetry relations between 
471: different vortex fronts with complex wave number $Q$ are discussed later on in 
472: Sec.~\ref{SEC.FRONT-SYMM}.
473: 
474: Table \ref{TAB.MOD-SYMM} shows the symmetry transformations that leave the
475: eigenvalue problem unchanged. They reflect ({\it i}) that ${\cal L}$ transforms
476: under complex conjugation (indicated by an overline) as 
477: \begin{equation} \label{Lcomplexconj}
478: \overline {{\cal L}(k,m,\sigma)} = {\cal L} (-k,-m, \overline {\sigma}) 
479: \end{equation}
480: and ({\it ii}) that the NSE (\ref{NAVIER2}) are invariant under an axial 
481: reflection $(z,Re,w) \longrightarrow (-z,-Re,-w)$.
482: 
483: Thus one infers, for example, that the growth rate (frequency) of the 
484: characteristic exponent $ \sigma = \gamma - i \omega$ for $m = 0$ vortices is 
485: an even (odd) function of $k$ and $Re$. For perturbations with $m\ne 0$ one 
486: finds that
487: \begin{eqnarray}
488: \gamma(k,m,Re) =  \gamma(-k,-m,Re) &=& \gamma(-k,m,-Re) =  \gamma(k,-m,-Re)  \\
489: \omega(k,m,Re) = -\omega(-k,-m,Re) &=& \omega(-k,m,-Re) = -\omega(k,-m,-Re)
490: \end{eqnarray}  
491: and that the spatiotemporal structure including the amplitude functions of a 
492: L-spiral perturbation ($k/m > 0$) at $Re > 0$ is the same as that of a R-spiral
493: ($k/m < 0$) at $Re < 0$. Note, however, that any finite through-flow breaks 
494: the axial mirror symmetry between L- and R-spirals at $Re=0$ so that, among
495: others,
496: \begin{equation}
497: \gamma_R (- Re) = \gamma_L (Re) \ne \gamma_R (Re) = \gamma_L (- Re)
498: \end{equation} 
499: when $Re \ne 0$. But the symmetry relations are such that it
500: suffices to investigate, say, positive $k$ combined together with 
501: either ({\em i}) $m>0$ only for positive {\em and} negative $Re$ or, equivalently, 
502: ({\em ii}) $m$ positive {\em and} negative for $Re>0$ only in order to get the
503: complete linear information on both spiral vortex types. 
504: %-------------- Sec. III c -----------------------------
505: \subsection{Bifurcation thresholds} \label{SEC.RESULTS1}
506: %----------------------------------------------------
507: Figure~\ref{FIG.R1c_R2} shows the {\em critical} bifurcation thresholds 
508: $R_{1c}(R_2)$ for $m=0$ and $m=\pm 1$ vortex patterns with the respective
509: {\em critical} wave numbers, $k_c(R_2)$, as functions of $R_2$ in the absence of
510: through-flow. 
511: 
512: The vertical lines in Fig.~\ref{FIG.R1c_R2} mark the two outer
513: Reynolds numbers $R_2 = 0$ and $R_2 = -125$ for which we show in 
514: Fig.~\ref{FIG.R1c_Re} as representative examples how the critical thresholds
515: evolve with through-flow Reynolds number $Re$. The above discussed
516: symmetry relation $\gamma_R (- Re) = \gamma_L (Re)$ between the growth rates 
517: of R- and L-spirals implies the corresponding relation between the respective 
518: bifurcation thresholds (full and dashed lines in Fig.~\ref{FIG.R1c_Re}).
519: In the remainder of this paper we therefore restrict ourselves without loss of
520: information to positive $Re$.
521: 
522: For small $Re>0$ the axial flow stabilizes the basic state against growth of
523: TVF ($m=0$) and R-SPI ($m=-1$) perturbations. On the other hand, the 
524: bifurcation threshold for 
525: L-SPI ($m=1$) vortex patterns that propagate into the same direction as the
526: through-flow decreases at small $Re$ and increases only at larger $Re$. 
527: The upwards shift of the $m=0$ threshold with increasing $Re$ is 
528: stronger than the one for $m=1$. Thus, eventually the latter comes to lie below
529: the former and consequently the growth of L-spirals propagating into the same
530: direction as the through-flow is favored for sufficiently large $Re$ even when
531: $R_2=0$.
532: 
533: Dotted lines in Fig.~\ref{FIG.EPSILON} show the reduced critical threshold
534: curves $\epsilon_c(Re)$ (\ref{epsilon_c}) as functions of $Re$ for the two
535: representative outer Reynolds numbers $R_2=0$ and $R_2=-125$. The other lines
536: in Fig.~\ref{FIG.EPSILON} are discussed in Sec. \ref{SEC.LOCAL-PERTUB}. Our
537: numerical results for $\epsilon_c(Re)$ that were obtained in steps of 
538: $\delta Re = 1$ were fitted in the range $Re=-20\cdots20$ to the following 
539: expression
540: \begin{equation} \label{fitfunction}
541: f=a_{1}Re+a_{2}Re^{2}+a_{3}Re^{3}+a_{4}Re^{4} \, .
542: \end{equation}
543: The fit parameters $a_n$ are listed for different $R_2$ in Tables 
544: \ref{TAB.FITPARA} and \ref{TAB.FITPARA2} for TVF ($m=0$) and L-SPI $(m=1)$,
545: respectively. The threshold curves for R-SPI $(m=-1)$ are obtained according to
546: Sec.~\ref{SEC.SYMM-EIGENVALUE} from those for L-SPI by $Re \rightarrow -Re$, 
547: i.e., by changing the sign of the odd coefficients in Table~\ref{TAB.FITPARA2}.
548: %\clearpage
549: %-------------- Sec. IV -----------------------------
550:  \section{Localized vortex perturbations} \label{SEC.LOCAL-PERTUB}
551: %-------------------------------------------------------------
552: So far we have considered axially extended vortex perturbations of the form
553: $\psi(r,\varphi,z,t) = \phi(r)e^{i(kz + m\varphi)}$
554: which are single eigenmodes of the operator ${\cal L}$~(\ref{Result.Operator}).
555: For supercritical control parameters, $R_1 > R_{1c}(m)$, a finite band of axial
556: wave numbers can grow and with it also axially localized wave packets 
557: consisting of superpositions of vortex modes.
558: %****************** IV A**********************************************
559: \subsection{Vortex packets} \label{SEC.PACKETS}
560: %****************************************************************
561: Let us consider first an infinitesimal initial perturbation with azimuthal 
562: wave number $m$ that is axially localized, i.e., a wave packet that
563: consists of a superposition of vortex modes of different $k$ but common $m$ --
564: an initial perturbation containing different $m$-modes would be just a sum of
565: the above described vortex packets that would evolve independently of each other
566: as long as the linear description is valid.
567:  
568: After fast transients have decayed a pulse like perturbation survives
569: with axial wave numbers within the unstable band centered around the wave 
570: number of maximal growth $k_{max}(m) \simeq k_c(m)$. Since the above described
571: wave packet contains vortex modes that can grow the pulse will grow as well. 
572: Simultaneously the pulse center travels axially for small $\epsilon$ with 
573: the critical goup velocity
574: \begin{equation}
575: v_{g}=\left. \frac{\partial \omega(k)}{\partial k} \right|_{k_{c}}.
576: \end{equation}
577: Hence when at a bifurcation threshold the frequency is nonzero with a finite 
578: group velocity then the supercritical spatiotemporal growth behavior of an 
579: axially {\it localized} vortex perturbation differs significantly from an axially 
580: {\it extended} vortex mode. The growth of the latter is axially uniform which is
581: not the case for the former. 
582: 
583: Furthermore, 
584: one has to distinguish between two different supercritical regimes for the 
585: former: ({\em i}) In the so-called convectively unstable parameter regime 
586: of the basic state the vortex packet moves with the velocity $v_{g}$ faster 
587: away than it grows -- while growing in the frame comoving with $v_{g}$ the 
588: pulse moves out of the system so that the basic state is restored 
589: \cite{b&b,HUE}. In other words, the two fronts that join 
590: the pulse intensity envelope to the structureless state propagate 
591: both in the direction in which the packet center moves.
592: {\em(ii)} In the so called absolutely unstable parameter regime 
593: the growth rate of the packet is so large that one front propagates in the 
594: laboratory frame opposite to the center motion. Thus, the packet expands not 
595: only into the direction of the pulse motion but also opposite to it 
596: \cite{b&b} so that eventually the initial perturbation can fill the entire 
597: system. But the linear growth analysis of the vortex fields does not 
598: determine in what nonlinear final stable state the system will end nor what 
599: possible intermediate nonlinear transient behavior might occur.
600: 
601: However, this analysis has an important implication for experiments with 
602: through-flow: 
603: Developed vortex structures can be seen in finite systems with vortex 
604: suppressing inlet conditions only in the absolutely unstable regime which is 
605: typically realized at larger $R_1 > R_{1c}$ (cf. Fig.~\ref{FIG.EPSILON}) if one 
606: leaves aside noise-sustained patterns  \cite{Deissler85,BAC-Pre94} in the 
607: convectively unstable regime. In this latter regime the vortex front that 
608: connects to the zero amplitude inlet condition moves downstream (we assume that
609: the fronts of our forwards bifurcating vortex structures are linearly selected 
610: thus excluding the buildup of nonlinear fronts that might revert their 
611: propagation direction). In the absolutely unstable regime, on the other hand, 
612: an upstream  motion of this front is stopped by the  inlet condition at a
613: characteristic downstream growth length from the inlet. This growth length
614: of the downstream evolving vortex structure diverges \cite{MLK89,BLRS96} when 
615: approaching the boundary between convective and absolute instability from the 
616: latter regime. 
617:  
618: It is remarkable that based on the experimental observations of Takeuchi and 
619: Jankowski \cite{TAK-JAN} such a behavior was discussed already in 1979 [cf. 
620: figures 4 and 5(a) and their discussion in Sec. 6 of ref.~\cite{TAK-JAN}], albeit
621: without invoking the concept of absolute and convective instability \cite{b&b} 
622: which was introduced to a broader fluid dynamics community only a few years 
623: later \cite{HUE}.
624: 
625: %-------------- Sec. IV B ---------------------------
626:  \subsection{Boundary between convective and absolute instability: 
627:  saddle point analysis} \label{SEC.SADDLE-POINT}
628: %---------------------------------------------------------------
629: The boundary between convective and absolute instability of the basic flow 
630:  against growth of vortices with a particular azimuthal wave number $m$ 
631:  is marked by those parameter combinations for which one of the fronts of the 
632:  linear packet of $m$-vortices reverts its propagation direction in the 
633:  laboratory  frame: In the convectively unstable parameter regime this front 
634:  propagates in the same direction as the
635:  center of the packet, in the absolutely unstable regime it moves opposite to
636:  it, and right on the boundary between the two regimes the front is stationary
637:  in the laboratory frame. 
638:  
639:  This parameter combination can be determined 
640: by a saddle point analysis of the linear complex dispersion relation
641: $\sigma(Q)$ over the plane of complex wave numbers \cite{HUE,CRO-HOH}
642: \begin{equation}
643: Q = \Re Q + i\Im Q = k - i K \, . 
644: \end{equation}
645: Here we do not display the dependence of 
646: \begin{equation}
647: \sigma(Q) = \Re \sigma(Q) + i\Im \sigma(Q) = \gamma(Q) - i \omega(Q) 
648: \end{equation}
649: on the parameters $R_1, R_2, Re, \eta$ and we also do not indicate that the
650: dispersion relations for vortex perturbations with different azimuthal wave
651: numbers $m$ are different. The boundary condition of no growth for a front that 
652: is stationary in the laboratory system is  
653: \begin{equation} \label{Wachsbed.}
654: \Re\sigma = 0
655: \end{equation}
656: at the appropriate saddle position, $Q^\ast$, of $\sigma(Q)$ in the complex
657: wave number plane \cite{HUE}. Here $Q^\ast$ follows from 
658: \begin{equation} \label{saddle}
659: \frac{d\sigma(Q)}{dQ}\Big|_{Q^\ast} = 0.  
660: \end{equation}
661: For fixed $m, R_2, Re,\eta$ eqs.(\ref{Wachsbed.},\ref{saddle}) yield 
662: $R_{1c-a}$. Here and in the following the subscript $c-a$ identifies
663: boundaries between convective and absolute instability. Thus, e.g., the basic 
664: flow state is convectively [absolutely] unstable against vortex perturbations 
665: with azimuthal wave number $m$ for $R_{1c}(m) < R_1 < R_{1c-a}(m)$ 
666: [$R_1 > R_{1c-a}(m)$] and absolutely stable when $R_1 < R_{1c}(m)$. 
667: 
668: We are interested here in the $Re$ dependence of these thresholds and we will 
669: discuss to that end the reduced boundary quantities 
670: \begin{equation}
671: \mu_{c-a}(Re)=\frac{R_{1c-a}(Re)}{R_{1c}(Re)}-1
672: \end{equation}
673: and
674: \begin{equation}
675: \epsilon_{x}(Re)=\frac{R_{1x}(Re)}{R_{1c}(Re=0)}-1,
676: \ \textrm{where}\ x=c\ \textrm{or}\ x=c-a \, .
677: \end{equation}
678: Because of the symmetries of $\sigma(Q)$, of its saddle-point $Q^\ast$, and of
679: the resulting boundary $R_{1c-a}(m, Re)$ one has
680: \begin{eqnarray}
681: k^{\ast}(-m,-Re)&=&k^{\ast}(m,Re)\\
682: K^{\ast}(-m,-Re)&=&-K^{\ast}(m,Re)\\
683: \mu_{c-a}(-m,-Re)&=&\mu_{c-a}(m,Re)\\
684: \epsilon_{c-a}(-m,-Re)&=&\epsilon_{c-a}(m,Re)
685: \end{eqnarray}
686: so that it suffices again to investigate only positive through-flow.
687: 
688: %-------------- Sec. IV C ---------------------------
689:  \subsection{Numerical procedures} \label{SEC.NUM-PROC2}
690: %----------------------------------------------------------------
691: In order to determine the boundary between convective and 
692: absolute instability via the solution of eqs.~(\ref{Wachsbed.}, \ref{saddle})
693: one has to evaluate the dispersion relation $\sigma(Q)$ for complex $Q$ 
694: \cite{TAG-EDW-SWI,BAC-Pre94}. 
695: To that end we solved the eigenvalue problem (\ref{resultEW}) of
696: the full field equations for {\em complex} $Q$ (cf., Sec.~\ref{SEC.NSE1}). 
697: In addition
698: and for comparison we used for $\sigma(Q)$ the Ginzburg-Landau amplitude 
699: equation approximation which only requires knowledge of 
700: $\sigma(k)$ along the real $k$-axis.  
701: %-------------- Sec. IV C 1 -------------------------
702:  \subsubsection{Ginzburg-Landau amplitude equation approximation} 
703:  \label{SEC.GLA1} 
704: %---------------------------------------------------------------------------------
705: Within this approximation the dispersion relation $\sigma(Q)$ for vortex 
706: modes is expanded in $Q$ and $R_1$ around the critical point $k_c, R_{1c}$
707: \begin{eqnarray}
708: \sigma(Q,R_1)&=&\sigma_c+\left(\frac{\partial \sigma}{\partial Q}
709: \right)_c(Q-k_c)+\frac{1}{2}\left(\frac{\partial^2 \sigma}{\partial
710: Q^2} \right)_c (Q-k_c)^2\nonumber\\
711: & &+\left(\frac{\partial \sigma}{\partial R_1}
712: \right)_c(R_1-R_{1c}) + h.o.t. \\
713: &=&-i\omega_c-iv_g(Q-k_c)  
714: -\frac{\xi_0^2}{\tau_0}(1+ic_1)(Q-k_c)^2 \nonumber\\
715: & &+ \frac{(1+ic_0)}{\tau_0}\mu + h.o.t. \, . \label{Ginz-Landau-Naeherung}
716: \end{eqnarray}
717: The expansion coefficients in the above expressions appear also in the linear 
718: parts of the Ginzburg-Landau amplitude equation \cite{CRO-HOH}. They are 
719: obtained from the numerical solution of the eigenvalue problem (\ref{resultEW}) 
720: of the full field equations for real $k$ close to the 
721: critical point \cite{REC-LUE-MUE,PIN}.  
722: 
723: Within the GLE approximation one gets from eqs.~(\ref{Wachsbed.}, \ref{saddle})
724: \begin{equation} \label{QstarGLE}
725:  Q^\ast=\underbrace{k_c-\frac{c_1 \tau_0 v_g}{2(1+c_1^2)\xi_0^2}}_{k^\ast}
726:  -i \underbrace{\frac{\tau_0 v_g}{2(1+c_1^2)\xi_0^2}}_{K^\ast},
727: \end{equation}
728: \begin{equation}
729: \mu_{c-a}=\frac{v_{g}^{2}\tau_{0}^{2}}{4(1+c_{1}^{2})\xi_{0}^{2}}, 
730: \end{equation}
731: and
732: \begin{equation}
733: \epsilon_{c-a}(Re)= \epsilon_{c}(Re) + 
734: \frac{v_g^2\tau_0^2}{4(1+c_1^2)\xi_0^2} \left[1 + \epsilon_c(Re)\right].
735: \end{equation}
736: Note that the GLE coefficients $v_g, \tau_0, \xi_0, c_0, c_1$ depend on 
737: $m, R_2, Re, \eta$. For $m=0$ and $R_2=0$ also the nonlinear coefficients of the
738: GLE have been obtained for several $\eta$ as functions of $Re$ \cite{REC-LUE-MUE}.
739: 
740: %-------------- Sec. IV C 2 -------------------------
741:  \subsubsection{Dispersion relation $\sigma(Q)$ of the NSE for complex $Q$} 
742:  \label{SEC.NSE1}
743: %-----------------------------------------------------------------------------
744: To assess the quality of the GLE results for the boundary between convective and
745: absolute instability we determined the dispersion relation of the NSE not only
746: for real $k$ close to the critical point but for complex axial wave numbers $Q$
747: that lie in the vicinity of the relevant saddle locations $Q^\ast$ of 
748: $\sigma(Q)$. Having determined $\sigma(Q)$ as described in 
749: Sec.~\ref{SEC.NUM-PROC1} we solved the equations (\ref{Wachsbed.},\ref{saddle})
750: for $R_{1c-a}$ in the form
751: \begin{equation}
752: \gamma (Q^\ast)=0 ; \quad
753: \frac{\partial \gamma(Q)}{\partial K} \Big|_{Q^\ast}=0 ; \quad
754: \frac{\partial \gamma(Q)}{\partial k} \Big|_{Q^\ast}=0
755: \end{equation}
756: that follows from using the Cauchy-Riemann relations for $\sigma(Q)$.
757: 
758: 
759: %-------------- Sec. IV D ----------------------------
760:  \subsection{Results} \label{SEC.RESULTS2}
761: %------------------------------------------------------------------------
762: 
763: In Fig.~\ref{FIG.SATTEL} we show the $Re$-variation of the real and imaginary
764: parts of the relevant saddle-point $Q^\ast = k^\ast - i K^\ast$ at the boundary  
765: $R_{1c-a}$ for two characteristic cases: (a) L-SPI ($m=1$) at $R_2=0$ and (b) 
766: R-SPI ($m=-1$) at $R_2=-125$. In each case full (dashed) lines were obtained
767: from the correct NSE (approximate GLE) dispersion relation $\sigma(Q)$. 
768: 
769: Case (a) is representative for a situation where the GLE approximation
770: reasonably well reproduces the correct result -- at least for small $Re$ -- and
771: starts to deviate significantly only for larger $Re$. On the other hand, in case
772: (b) the GLE approximation to $k^\ast$ displays a smooth variation with $Re$
773: that reflects the smooth variation of the saddle $Q^\ast_{GLE}$ (\ref{QstarGLE})
774: of $\sigma_{GLE}(Q)$ (\ref{Ginz-Landau-Naeherung})
775: while the real part $k^\ast$ of the saddle location of the correct dispersion
776: relation undergoes a dramatic change around $Re \simeq 0.8$. The reason is that
777: the surfaces of $\gamma(Q) = \Re \sigma(Q)$ over the $Q$-plane for the two 
778: largest eigenvalues intersect and change their order along the $\gamma$ axis. 
779: Thus, the saddle that is relevant for the $c-a$ transition and that has the 
780: largest $\gamma$ value switches from one eigenvalue surface to another. Similarly
781: the eigenvalue surface might have another saddle and their ordering
782: in $\gamma$ changes.
783: In contrast to that the GLE approximation produces only a
784: single eigenvalue tracing out the smooth surface $\sigma_{GLE}(Q)$ 
785: (\ref{Ginz-Landau-Naeherung}).
786: 
787: To give an impression of such an intersection of the correct dispersion 
788: surfaces we show in Fig.~\ref{FIG.2EW} their real parts over the complex 
789: $Q-$plane at $Re=5,\mu=\mu_{c-a}$. Full lines in Fig.~\ref{FIG.SATTEL-SCHNITTE} 
790: show different constant-$k$ sections through them in the intersection range. 
791: At $Re=5,\mu=\mu_{c-a}$ the saddle has moved already away from the
792: intersection region. The saddle coordinates 
793: ($Q^\ast \simeq 4.73 -i 1.76, \gamma^\ast=0$) are indicated in
794: Figs.~\ref{FIG.SATTEL}, \ref{FIG.2EW} by full dots.
795: 
796: While the saddle locations $Q^\ast_{GLE}$ of the GLE approximation 
797: $\sigma_{GLE}(Q)$ (\ref{Ginz-Landau-Naeherung}) can differ substantially from
798: the saddle $Q^\ast$ of the correct dispersion relation the difference in the
799: boundaries between convective and absolute instability is typically much less
800: pronounced -- cf, e.g., the dashed and full curves for $\epsilon_{c-a}$ in  
801: Fig.~\ref{FIG.EPSILON}. There the GLE results (dashed lines) agree in each case
802: quite well with the correct boundaries (full lines) up to, say $Re =5$.
803: However, as a consequence of the typical increase of the reduced boundary
804: $\mu_{c-a}$ with increasing $Re$ the quality of the GLE results for 
805: $\mu_{c-a}$ generally
806: deteriorates. After all the GLE is strictly valid only for $\mu \to 0$.
807: Fig.~\ref{FIG.EPSILON-RE=20} shows an example ($m=1, R_2=-125$) where the GLE
808: prediction for the boundary shows even a qualitative different variation with 
809: $Re$ for, say, $Re>15$. We have observed similar behavior -- partly at larger
810: $Re$ -- also for other combinations of $m, R_2$. 
811: 
812: Tables~\ref{TAB.FITPARA} and \ref{TAB.FITPARA2} contain the coefficients for
813: fitting our results for $\epsilon_{c-a}(m=0, R_2, Re)$ and 
814: $\epsilon_{c-a}(m=1, R_2, Re)$ in the same way as described in 
815: Sec.~\ref{SEC.RESULTS1}. Also here the  
816: boundary curves for R-SPI $(m=-1)$ are obtained according to
817: Sec.~\ref{SEC.SYMM-EIGENVALUE} from those for L-SPI $(m=1)$ by 
818: $Re \rightarrow -Re$, i.e., by changing the sign of the odd coefficients in 
819: Table~\ref{TAB.FITPARA2}.
820: %\clearpage
821: %-------------- Sec. V ------------------------------
822:  \section{Fronts and pulses} \label{FRONTS}
823: %---------------------------------------------------------------
824: The vortex fronts that we are investigating here and that appear also as 
825: constituents of vortex pulses are 
826: perturbations of the basic state where the fields (locally) have the form
827: \begin{equation} \label{Einhullende}
828: \psi  \sim e^{i (k^\ast z - \omega^\ast t)}
829: e^{K^\ast (z - v^\ast t)}e^{im\varphi}
830: \end{equation}
831: in the laboratory frame. This form describes long-time spatiotemporal
832: properties of uniquely selected linear fronts \cite{CRO-HOH}.
833: Axial front velocity $v^\ast$, axial growth rate $K^\ast$ of the front envelope,
834: wave number $k^\ast$ of the vortex pattern under the front, and its frequency 
835: $\omega^\ast$ are determined 
836: by the saddle behavior of the linear complex dispersion relation 
837: $\sigma(Q,m) = \gamma (Q,m) - i \omega (Q,m)$ of the field equations over the 
838: plane of complex axial wave numbers $Q = k - i K$. Since superpositions of vortex
839: perturbations with different $m$ evolve independently from each other within the
840: linear description as mentioned already in Sec.~\ref{SEC.PACKETS} we consider 
841: here only fronts of vortex perturbations that have a common azimuthal wave 
842: number $m$ and we do not always display the latter explicitly. 
843: 
844: The saddle condition is \cite{CRO-HOH} 
845: \begin{equation} \label{saddlecondition}
846: \frac{d}{dQ} \left[\sigma \left(Q\right)  \: \: + \: \: i \: v \: Q \right] 
847: \Big|_{Q=Q^\ast}  =  0 .
848: \end{equation} 
849: And the stationarity requirement that the temporal growth rate of the front 
850: vanishes in the frame comoving with the front velocity $v^\ast$ demands that 
851: \begin{equation}
852: \label{gammasaddle}
853: 0 = \Re \left[ \sigma (Q) + i v Q \right] \Big|_{Q = Q^\ast}
854: = \gamma (Q^\ast) + v^\ast K^\ast .
855: \end{equation}
856: We combine (\ref{saddlecondition}) and (\ref{gammasaddle}) into the three 
857: equations 
858: \begin{equation} \label{Bedingungen2}
859: v^\ast = - \frac{\gamma (Q^\ast)}{K^\ast}
860: = - \frac{\partial \gamma (Q)}{\partial K} \Big|_{Q^\ast} \quad ; \quad
861: \frac{\partial \gamma (Q)}{\partial k} \Big|_{Q^\ast} = 0 
862: \end{equation}
863: that we have solved for $v^\ast, k^\ast, K^\ast$.
864: 
865: %-------------- Sec. V A ----------------------------
866:  \subsection{Notation} \label{SEC.FRONT-NOTATION}
867: %---------------------------------------------------
868: 
869: The front envelope of (\ref{Einhullende}) varies axially with $e^{K^\ast z}$. 
870: If $K^\ast>0$, then the perturbation 
871: grows at $z=-\infty$ out of the basic state. We call such a front to be of type 
872: $+$ and identify the associated front
873: properties by a subscript $+$. On the other hand, for $K^\ast<0$ we have a front 
874: of type $-$ with an intensity envelope that joins at $z=\infty$ with the basic 
875: state. So the two subscripts $\pm$ identify
876: the axial variations of the front envelopes. A pulse-like perturbation of the 
877: basic state would consist suffiently away
878: from its center of a $+$ front for $z \to -\infty$ and of a $-$ front 
879: for $z \to \infty$. Correspondingly the eqs.~(\ref{Bedingungen2}) have two
880: different solutions: one of them describes a 
881: $-$ front and the other one a $+$ front. A schematic plot of the
882: different envelope types can be seen in Fig. \ref{FIG.FRONTDIRECT}.
883: 
884: We identify the vortex pattern 
885: that is unfolded under a front either by its azimuthal wave number $m$ or by
886: the superscripts T, L, R. Here T refers to a TVF-like pattern of vortices with
887: $m=0$, L denotes L-spiral vortices with $m=1$, and R identifies R-spiral vortices
888: with $m=-1$. Hence $m$ or equivalently the 
889: superscripts identify the spatiotemporal structure of the vortex pattern growing
890: under the front. In this work we restrict ourselves to these three vortex 
891: varieties. Since they appear under the above
892: described two front envelopes there are six different fronts that we have
893: investigated here.
894: 
895: %-------------- Sec. V B ----------------------------
896:  \subsection{Results} \label{SEC.RESULTS3}
897: %--------------------------------------------------------
898: 
899: In Fig.~\ref{FIG.FRONTEN} the results of our investigations are shown for 
900: fronts with azimuthal wave numbers $m$ = 0, 1, and -1 for axial
901: through-flow Reynolds numbers $Re$ =0, 10, and 20. In each case the outer 
902: cylinder is at rest, $R_2=0$. The front properties are presented as functions of
903: $\mu$. Within each 2$\times$2 block of figures (a)-(i) the left column shows 
904: real part $k^\ast$ and negative imaginary part $K^\ast$ of the saddle point and 
905: the right column shows the front velocity $v^\ast$ and the frequency 
906: $\omega^\ast$, respectively. They all start at $\mu = 0$ since vortex growth is 
907: possible only above the critical threshold, i.e., for $R_1 > R_{1c}(Re)$.
908: The two {\em critical} fronts at this threshold are degenerate ($k^\ast_- = 
909: k^\ast_+ = k_c, \, \omega^\ast_- = \omega^\ast_+ = \omega_c, \, 
910: v^\ast_- = v^\ast_+ = v_g$) with vanishing axial growth rates 
911: $K^\ast_- = K^\ast_+ = 0$. For $\mu>0$, however, $+$ and $-$ fronts differ from
912: each other thus reflecting differences in the respective saddle points.
913: 
914: The variation of $k^\ast, K^\ast, v^\ast$, and $\omega^\ast$ with $\mu$ and 
915: $Re$ is best understood by
916: comparison with the corresponding GLE approximation (cf. Sec.~\ref{SEC.GLA2}) and
917: by invoking symmetry relations (cf. Sec.~\ref{SEC.FRONT-SYMM}). We therefore
918: continue the discussion of our results in the following two sections 
919: \ref{SEC.GLA2} and \ref{SEC.FRONT-SYMM}. 
920: 
921: %-------------- Sec. V B 1 --------------------------
922:  \subsubsection{Ginzburg-Landau amplitude equation approximation} \label{SEC.GLA2}
923: %----------------------------------------------------------------------------------
924: 
925: The saddle point analysis (\ref{saddlecondition} - \ref{Bedingungen2}) of the
926: GLE approximation (\ref{Ginz-Landau-Naeherung}) yield the following
927: front properties
928: %\begin{eqnarray} \label{GL-Front1}
929: \begin{eqnarray} 
930: K^{\ast}_{\pm}&=&\pm \sqrt{\frac{\mu}{(1+c_{1}^{2})\xi_{0}^{2}}},\label{K-GLE} \\
931: k^{\ast}_{\pm}&=&k_{c}-c_{1}K^{\ast}_{\pm},\label{k-GLE}\\
932: v^{\ast}_{\pm}&=&v_{g}-2(1+c_{1}^{2})\frac{\xi_{0}^{2}}{\tau_{0}}K^{\ast}_{\pm},
933: \label{v-GLE}\\
934: \gamma^{\ast}_{\pm}&=&-v^{\ast}_{\pm}K^{\ast}_{\pm},\label{gamma-GLE}\\
935: \omega^{\ast}_{\pm}&=& \omega_c +v_g (k^\ast_\pm - k_c)
936: +(c_1-c_0)\frac{\mu}{\tau_0} \label{omega-GLE}\, .
937: %\label{GL-Front2}
938: \end{eqnarray}
939: for the two fronts with $K_+^{\ast} >0 $ and $K_-^{\ast} < 0 $, respectively.
940: {\em All} quantities appearing in (\ref{K-GLE} - \ref{omega-GLE}) depend on 
941: whether they refer to T ($m$ = 0), L ($m$ = 1), or R ($m$ = -1) vortex fronts.
942: These GLE results are shown in Fig.~\ref{FIG.FRONTEN} by dashed lines.
943: 
944: They reasonably
945: well describe the small-$\mu$ behavior of the correct front properties 
946: (full lines in Fig.~\ref{FIG.FRONTEN}) which
947: were obtained from the correct dispersion relation of the NSE: As predicted 
948: by the small-$\mu$ GLE approximation (\ref{K-GLE} - \ref{omega-GLE}) one finds
949: that for small $\mu$ the axial growth rates 
950: $K^\ast$ vary $\propto \sqrt \mu$, that consequently also $k^\ast - k_c$
951: and $v^\ast - v_g$ vary $\propto \sqrt \mu$, and that
952: $\omega^\ast - \omega_c$ can have in addition also a contribution 
953: $\propto \mu$ when $c_1 - c_0 \neq 0$. The latter is the case for $m=\pm 1$
954: irrespective of $Re$ and for $m=0$ if $Re \neq 0$.
955: 
956: Note that for TVF fronts with $m=0, Re = 0$ the GLE predicts $\omega^\ast = 0$
957: whereas the correct dispersion in Fig.~\ref{FIG.FRONTEN}(a) seems to show for 
958: small $\mu$ a variation of 
959: $\omega^\ast \propto \mu^2$ that is beyond the range of applicability of the 
960: GLE. Thus, under the linear part of a moving 
961: TVF front there should be a non zero phase propagation in the laboratory 
962: frame with phase velocity $\omega^\ast / k^\ast$. The analogous behavior was
963: found also for convection rolls in the Rayleigh-B\'enard system \cite{BUE-LUE}.
964: 
965: At the boundary between convective and absolute
966: instability, $\mu =\mu_{c-a}$, the velocity 
967: $v_+^{\ast}$ and the temporal growth rate $\gamma^{\ast}_+$ of the $+$ front 
968: vanish in the laboratory frame while $v_-^{\ast}$ is positive there
969: (and given by 2 $v_g$ within the GLE approximation). Note that
970: according to Fig.~\ref{FIG.EPSILON} and 
971: tables~\ref{TAB.FITPARA} and \ref{TAB.FITPARA2} $\mu_{c-a}$ is zero for $m=0,
972: Re=0$ and very small for $m= \pm 1, Re=0$. Only for large enough
973: through-flow $\mu_{c-a}$ becomes sizeable. 
974: In the convectively unstable regime $0 < \mu < \mu_{c-a}$ both, the $+$ front 
975: as well as the $-$ front of a vortex pulse move into the same downstream 
976: direction as the through-flow, $0 < v_+^{\ast} < v_-^{\ast}$. In the absolutely
977: unstable regime $\mu > \mu_{c-a}$, however, the $+$ front moves upstream and 
978: the $-$front moves downstream, $v_+^{\ast} < 0 < v_-^{\ast}$. 
979: 
980: In ref.~\cite{NG-TUR} it was remarked that the phase velocity of spiral patterns 
981: in axial flow through a system of radius ratio $\eta=0.95$ \cite{SNY}
982: deviates from the critical one, $\omega_c/k_c$, of axially extended vortex
983: perturbations. While we have done only 
984: calculations for $\eta=0.5$ our results of Fig.~\ref{FIG.FRONTEN} and of the
985: small-$\mu$ GLE approximation (\ref{K-GLE} - \ref{omega-GLE}) shed some light on
986: the existence of such deviations. Since the experimental vortex structures grow
987: in downstream direction under intensity fronts one strictly speaking would have 
988: to compare with the phase velocity under such fronts that connect in the 
989: absolutely unstable regime to the fully developed downstream vortex pattern. 
990: Such a nonlinear analysis has been done for $m=0$ patterns \cite{BLRS96} but not
991: for spirals. However, already the phase velocities $\omega_+^\ast /k_+^\ast$
992: under our linear $+$ fronts that grow in downstream direction differ from the 
993: corresponding critical phase velocities $\omega_c /k_c$ of axially extended 
994: patterns --- cf. Fig.~\ref{FIG.FRONTEN} and eqs.~(\ref{K-GLE} - \ref{omega-GLE}).
995: 
996: %-------------- Sec. V B 2 --------------------------
997:  \subsubsection{Symmetries} \label{SEC.FRONT-SYMM}
998: %-------------------------------------------------------
999: The front properties shown in Fig.~\ref{FIG.FRONTEN} are largely influenced by
1000: the symmetry properties of the system {\em without} through-flow although 
1001: a finite $Re$ changes them.
1002: 
1003: Invariance of the field equations under $z \to -z$ for $Re=0$ implies that 
1004: stationary perturbations with
1005: $\omega_{c}=0$ (TVF) under a $+$ front are mirror images of those under a $-$ 
1006: front. This implies for $Re=0$ the symmetry relations
1007: \begin{equation} \label{symmet1}
1008: (K,\omega,v)_{+}^{\ast T}=-(K,\omega,v)_{-}^{\ast T};\quad 
1009: k_{+}^{\ast T}=k_{-}^{\ast T}.
1010: \end{equation}
1011: These symmetry properties of the two front types of T-perturbations can be seen
1012: in Fig. \ref{FIG.FRONTEN}(a).
1013: 
1014: Now consider L and R perturbations. Here the invariance of the field equations 
1015: under  $z \to -z$ for $Re=0$ implies
1016: first of all that a spatially extended L-SPI with uniform amplitude is the 
1017: mirror image of a spatially extended R-SPI.
1018: Furthermore, a L-SPI under a $+$ front with positive $K$ is symmetry degenerate 
1019: with a R-SPI under a $-$ front with
1020: negative $K$. Similarly a R-SPI under a $+$ type front is the mirror image of a
1021:  L-SPI under a $-$ front. This implies for $Re=0$ the symmetry relations   
1022: \begin{equation} \label{symmet2}
1023: (K,\omega,v)_{+}^{*L}=-(K,\omega,v)_{-}^{*R};\quad k_{+}^{*L}=k_{-}^{*R},
1024: \end{equation}
1025: \begin{equation} \label{symmet3}
1026: (K,\omega,v)_{+}^{\ast R}=-(K,\omega,v)_{-}^{\ast L};\quad 
1027: k_{+}^{\ast R}=k_{-}^{\ast L}.
1028: \end{equation}
1029: They can be seen to be realized in Fig.~\ref{FIG.FRONTEN}(b) and 
1030: Fig.~\ref{FIG.FRONTEN}(c).
1031:  
1032: The GLE approximation (\ref{K-GLE}) - (\ref{omega-GLE}) shows beyond the relations 
1033: (\ref{symmet1})-(\ref{symmet3}) the following additional relations
1034: \begin{equation}
1035: k_{+}^{\ast T}=k_{-}^{\ast T}=k_c^T \quad ; 
1036: \quad \omega_{+}^{\ast T}=\omega_{-}^{\ast T}=0
1037: \end{equation}
1038: \begin{equation} \label{GLE-Symmetrie2}
1039: K_{+}^{\ast L}=-K_{-}^{\ast L}=K_{+}^{\ast R}=-K_{-}^{\ast R},
1040: \end{equation}  
1041: \begin{equation}
1042: \omega_{+}^{\ast L}=\omega_{-}^{\ast L}=-\omega_{+}^{\ast R}=-\omega_{-}^{\ast R}
1043: \end{equation}
1044: that follow from the fact, that $\omega_c^T = v_g^T = c_0^T = c_1^T = 0$,
1045: $k_c^R=k_c^L$, $v_g^R=-v_g^L$, 
1046: $\tau_0^R=\tau_0^L$, $c_0^R=-c_0^L$, $c_1^R=-c_1^L$, and
1047: $\xi_0^{R}=\xi_0^{L}$ for $Re=0$ \cite{PIN}.
1048: 
1049: %-------------- Sec. VI ----------------------------
1050:  \section{Summary}
1051: %---------------------------------------------------- 
1052:  We have determined the influence of an axial through-flow on the spatiotemporal
1053: growth behavior of structurally different vortex perturbations of the basic 
1054: Couette-Poiseuille flow in
1055: the Taylor-Couette system with radius ratio $\eta=0.5$. To that end we
1056: have solved the linearized NSE numerically with a shooting method for vortex 
1057: perturbations with azmuthal wave numbers $m=0$ (TVF), $m=1$ (L-SPI), and $m=-1$
1058: (R-SPI) in a wide range of the parameters $Re, R_1$, and $R_2$. Here symmetry
1059: properties allowed us to restrict ourselves to positive through-flow Reynolds
1060: numbers $Re$. For each of the three different vortex varieties we have
1061: investigated ({\it i}) axially extended vortex structures with homogeneous 
1062: amplitudes, ({\it ii}) axially localized vortex pulses consisting of a linear
1063: superposition of axially extended vortex modes with different real axial wave
1064: numbers $k$, and ({\it iii}) vortex fronts. 
1065: 
1066: Central to our analysis is the determination of the complex dispersion 
1067: relations $\sigma(Q)$ of the linearized NSE for vortex modes with the three 
1068: different $m$. We have evaluated $\sigma$ over the plane of complex wave 
1069: numbers $Q =  k - i K$ for patterns ({\it ii}, {\it iii}) and along 
1070: the real $k$-axis for pattern ({\it i}). We have also determined 
1071: the Ginzburg-Landau amplitude equation approximation $\sigma_{GLE}(Q)$ in order
1072: to analyze its predictions for the vortex stuctures ({\it ii}, {\it iii}) in 
1073: comparison with the correct NSE dispersion relation $\sigma(Q)$. In each case 
1074: symmetry relations are elucidated.
1075: 
1076: First we have
1077: evaluated the critical bifurcation thresholds $R_{1c}(Re, R_2, m)$ for axially 
1078: extended vortex structures. Then using a saddle-point analysis of $\sigma(Q)$
1079: we have determined the boundaries $R_{1c-a}(Re, R_2, m)$ between absolute and
1080: convective instability of the basic state at which one of the fronts of the
1081: expanding vortex pulses reverts its propagation direction in the laboratory 
1082: frame. Here we have elucidated also in some detail how the different saddle
1083: topologies of $\sigma(Q)$ and of $\sigma_{GLE}(Q)$ explain some of the
1084: shortcomings of the latter. Fit parameters for power-law expansions of
1085: the reduced boundaries $\epsilon_c, \epsilon_{c-a}$, and $\epsilon_{c-a}^{GLE}$
1086: up to $Re^4$ are listed in two tables. 
1087: 
1088: Finally we have determined the linearly selected front 
1089: behavior of growing vortex patterns with $m=0, \pm 1$ for $R_2=0$ under two 
1090: different types of front intensity envelopes: type $+$ shows growth in positive 
1091: z-direction while type $-$ locates growth in negative z-direction. The 
1092: combination of the
1093: three different dynamics of the constituent vortex modes ($m=0, \pm 1$) and of
1094: the two different spatial intensity profiles ($+$, $-$) leads to six different 
1095: fronts $\sim e^{i (k^\ast z - \omega^\ast t)}e^{K^\ast (z - v^\ast t)}$
1096: in the laboratory frame. Their velocity $v^\ast$, spatial growth rate $K^\ast$,
1097: wave number $k^\ast$, and frequency $\omega^\ast$ as
1098: determined via a saddle point analysis of the respective dispersion
1099: relations differ in general from each other in the presence of a through-flow.
1100: 
1101: \acknowledgments
1102: This work was supported by the Deutsche Forschungsgemeinschaft.
1103: 
1104: \clearpage
1105: %----------------- Bibliography ---------------------
1106:  \begin{thebibliography}{999}
1107: %---------------------------------------------------- 
1108: \bibitem{review} 
1109: For a review see refs.~\cite{DIPS-Springer-81,DON,TAG,CRO-HOH}.
1110: %-----------------------------------------------
1111: \bibitem{DIPS-Springer-81} 
1112: R.~C.~DiPrima and H.~L.~Swinney, in
1113: {\it Hydrodynamic Instabilities and Transition to Turbulence}, edited 
1114: by H.~L.~Swinney and J.~P.~Gollub (Springer, Berlin, 1981), p.~139.
1115: %---------------------------------------------------------
1116: \bibitem{DON}
1117: R.~J.~Donnelly, 
1118: % {\it Taylor-Couette Flow: The early days}, 
1119: Physics Today {\bf 44}, 332 (1991). 
1120: %---------------------------------------------------- 
1121:  \bibitem{TAG}
1122:  R.~Tagg, 
1123:  %{\it The Couette-Taylor Problem},
1124:  Nonlinear Science Today {\bf 4}, 1 (1994).
1125: %---------------------------------------------------- 
1126: \bibitem{CRO-HOH}
1127:  M.~C.~Cross and P.~C.~Hohenberg, 
1128:  %{\it Pattern formation outside of equilibrium}, 
1129:  Rev. Mod. Phys. {\bf 65}, 851 (1993).
1130: %---------------------------------------------------- 
1131:  \bibitem{CHO-IOO} 
1132: P.~Chossat and G.~Ioos, 
1133: {\it The Couette-Taylor Problem} (Springer, Berlin, 1994).
1134: %----------------------------------------------------
1135: \bibitem{GOL-STE-SCH-LAN}
1136: M.~Golubitsky, I.~Stewart, D.~G.~Schaeffer, and W.~F.~Langford,
1137: {\it Case Study 6: The Taylor-Couette system}, in {\it Singularities and Groups in
1138: Bifurcation Theory}, Vol. 2 (Springer, New York, 1988), p. 485.
1139: %---------------------------
1140: \bibitem{DIP-PRI} 
1141: R.~C.~DiPrima and A.~Pridor,
1142: %{\it Yhe stability of viscous flow between rotating concentric cylinders with an
1143: % axial flow}
1144: Proc.~R.~Soc.~London Ser. A {\bf 366}, 555 (1979). 
1145: %----------------------------------------------------
1146: \bibitem{TAK-JAN}
1147:  D.~I.~Takeuchi and D.~F.~Jankowski, 
1148:  %{\it A numerical and experimental investigation of the stability of spiral Poiseuille flow}, 
1149:  J. Fluid. Mech. {\bf 102}, 101 (1981).
1150: %---------------------------------------------------- 
1151:  \bibitem{NG-TUR}
1152:  B.~S.~Ng and E.~R.~Turner, 
1153:  %{\it On the linear stability of spiral flow between rotating cylinders}, 
1154:  Proc. R. Soc. Lond. A {\bf 382}, 83 (1982).
1155: %---------------------------------------------------- 
1156: \bibitem{LAN-TAG-KOS-SWI-GOL}
1157:  W.~F.~Langford, R.~Tagg, E.~Kostelich, H.~Swinney, and M.~Golubitsky,
1158:  %{\it Primary instabilities and bicriticality in flow between counter-rotating cylinders}, 
1159:  Phys. Fluids {\bf 31}, 776 (1988).
1160: %---------------------------------------------------- 
1161: \bibitem{GEB-GRO}
1162:  Th.~Gebhardt and S.~Grossmann,
1163:  %{\it The Taylor-Couette eigenvalue problem with independently rotating cylinders}, 
1164:  Z. Phys. {\bf B 90}, 475 (1993).
1165: %---------------------------------------------------- 
1166: \bibitem{MES-MAR}
1167: A.~Meseguer and F.~Marques,
1168: %{\it On the competition between centrifugal and shear instability in spiral Poiseuille flow}, 
1169: J. Fluid Mech. {\bf 455}, 129 (2002).
1170: %---------------------------------------------------------
1171: \bibitem{epsT}
1172: In the literature one uses also $\epsilon_T = \frac{T_{1}}{T_{1c}(Re=0)}-1$ as
1173: control parameter where the inner cylinder's Taylor number $T_1$ is proportional
1174: to $\Omega_1^2$. Thus, $\epsilon_T = 2 \epsilon + \epsilon^2$.
1175: %----------------------------------------------------
1176: \bibitem{STO-BUR}   
1177:  J.~Stoer and R.~Bulirsch,
1178:  {\it Einf\"uhrung in die Numerische Mathematik II}
1179:  (Springer Verlag, Berlin, 1972).
1180: %---------------------------------------------------- 
1181:  \bibitem{PRE-TEU-VET-FLA}
1182:  W.~Press, S.~Teukolsky, W.~Vetterling, and B.~Flannery, {\it Numerical Recipes}
1183:  (Cambridge University Press, Cambridge, 1992).
1184: %---------------------------------------------------- 
1185: \bibitem{HOF-LUE}
1186:  C.~Hoffmann and M.~L\"ucke, 
1187:  %{\it ??? Spirals and TVF ???},
1188: manuscript in preparation.
1189: %----------------------------------------------------
1190: \bibitem{b&b}
1191: R.~J.~Briggs, 
1192: {\it Electron-Stream Interaction with Plasmas},
1193: Research Monograph No. 29 (M.I.T. Press, Cambridge Mass, 1964);
1194: A.~Bers, 
1195: {\it Linear Waves and Instabilities} 
1196: (Physique des Plasmas, New York, 1975).
1197: %----------------------------------------------- 
1198: \bibitem{HUE}
1199:  P.~Huerre, in {\it Instabilities and Nonequilibrium Structures}, edited by
1200:  E.~Tirapegui and D.~Villaroel (Reidel, Dordrecht, 1987), p. 141; P.~Huerre and
1201:  P.~A.~Monkewitz, Annu. Rev. Fluid Mech. {\bf 22}, 473 (1990).
1202: %------------------------------------------------------------------------------
1203: \bibitem{Deissler85}
1204: R.~J.~Deissler,
1205: %{\it Noise-Sustained Structure, Intermittency, and the Ginzburg-Landau Equation},
1206: J.~Stat.~Phys.~{\bf 40}, 371 (1985).
1207: %------------------------------------------------------------------------------
1208: \bibitem{BAC-Pre94} 
1209: K.~L.~Babcock, G.~Ahlers, and D.~S.~Cannell, 
1210: %{\it Noise amplification in open Taylor-Couette flow},
1211: Phys.~Rev.~E {\bf 50}, 3670 (1994).
1212: %----------------------------------------------------
1213: \bibitem{MLK89}
1214:  H.~W.~M\"uller, M.~L\"ucke, M.~Kamps,
1215: % {\it Convective patterns in horizontal flow},
1216: Europhys.~Lett.~ {\bf 10}, 451 (1989);
1217: %------------------------------------------------------------------------------ 
1218: \bibitem{BLRS96}
1219: P.~B\"uchel, M.~L\"ucke, D.~Roth, R.~Schmitz,
1220: %{\it Pattern selection in the absolutely unstable regime as a nonlinear eigenvalue problem: 
1221: % Taylor vortices in axial flow },
1222: Phys.~Rev.~E {\bf 53}, 4764 (1996).
1223: %------------------------------------------------------------------------------
1224: \bibitem{TAG-EDW-SWI}
1225: R.~Tagg, W.~S.~Edwards, and H.~L.~Swinney,
1226: %{\it Convective versus absolute instability in flow between counterrotating cylinders}
1227: Phys.~Rev.~A {\bf 42}, 831 (1990). 
1228: %---------------------------------------------------------
1229:  \bibitem{REC-LUE-MUE}
1230:  A.~Recktenwald, M.~L\"ucke, and H.~W.~M\"uller, 
1231:  %{\it Taylor vortex formation in axial through-flow: Linear and weakly nonlinear analysis}, 
1232:  Phys. Rev. E {\bf 48}, 4444 (1993).
1233: %---------------------------------------------------- 
1234: \bibitem{PIN}
1235:  A.~Pinter,
1236:  %{\it Spiralwirbel im Taylor-Couette System},
1237:  Diploma thesis, Universit\"at des Saarlandes, Saarbr\"ucken, 2001.
1238: %---------------------------------------------------- 
1239: \bibitem{BUE-LUE}
1240:  P.~B\"uchel and M.~L\"ucke, 
1241:  %{\it Localized perturbations in binary fluid convection with and without throughflow}, 
1242:  Phys. Rev. E {\bf 63}, 016307 (2000).
1243: %---------------------------------------------------- 
1244:  \bibitem{SNY}
1245:  H.~A.~Snyder, 
1246:  %{\it Experiments on the stability of spiral flow at low axial Reynolds numbers}, 
1247:  Proc. R. Soc. Lond. A {\bf 285}, 198 (1962).
1248: %----------------------------------------------------
1249: \end{thebibliography}
1250: \clearpage
1251: 
1252: %----------------- Figures-------------------------------------------------
1253: %----------------- FIG 1 --------------------------------------------------
1254:  \begin{figure} 
1255:  \includegraphics[clip=true,width=10cm,angle=0]{./FIG/fig1.ps}
1256:  \caption{ Critical bifurcation thresholds $R_{1c}(R_2)$ for $m=0$ and 
1257:  $m=\pm 1$ vortex patterns with the respective
1258: critical wave numbers, $k_c(m, R_2)$, as functions of $R_2$ in the absence of
1259: axial through-flow. The vertical lines mark the two 
1260: representative outer Reynolds numbers $R_2 = 0$ and $R_2 = -125$ that are 
1261: investigated in more detail in this work. The radius ratio is $\eta=0.5$.
1262: \label{FIG.R1c_R2}}
1263:  \end{figure}
1264: %----------------- FIG 2 --------------------------------------------------
1265:  \begin{figure}
1266:  \includegraphics[clip=true,width=10cm, angle=0]{./FIG/fig2.ps}
1267:  \caption{ Evolution of the critical bifurcation thresholds $R_{1c}(Re)$  
1268:  for $m=0$ and $m=\pm 1$ vortex patterns 
1269: with through-flow Reynolds number $Re$. The two outer Reynolds 
1270: numbers $R_2 = 0$ and $R_2 = -125$ are marked in Fig.~\ref{FIG.R1c_R2}. 
1271: The radius ratio is $\eta=0.5$. 
1272: \label{FIG.R1c_Re}}
1273:  \end{figure}
1274: %----------------- FIG 3 --------------------------------------------------
1275:  \begin{figure} 
1276: \includegraphics[clip=true,width=9cm, angle=0]{./FIG/fig3.ps}
1277:  \caption{Stability boundaries of the basic flow state as functions of the
1278:  through-flow Reynolds number $Re$. Dotted lines show the reduced critical 
1279:  bifurcation thresholds 
1280:  $\epsilon_c$ (\ref{epsilon_c}) for axially extended vortex patterns.  
1281:  The full (dashed) boundary lines, $\epsilon_{c-a}$, between the 
1282:  convectively and absolutely unstable parameter regions for vortex growth 
1283:  were obtained from the eigenvalues of the full NSE (the GLE approximation) --
1284:  cf. Sec.~\ref{SEC.LOCAL-PERTUB}. The radius ratio is $\eta=0.5$.
1285:  \label{FIG.EPSILON}}
1286:  \end{figure}
1287: %----------------- FIG 4 --------------------------------------------------
1288:  \begin{figure}
1289:  \includegraphics[clip=true,width=10cm, angle=0]{./FIG/fig4.ps}
1290:  \caption{Evolution of the coordinates $k^\ast$ and $K^\ast$ of the saddle point
1291:   that determines the boundary, $R_{1c-a}$, between convective and absolute 
1292:   instability with through-flow Reynolds number $Re$. Full (dotted) lines are
1293:   evaluated with the correct NSE (approximate GLE) dispersion relation. Filled
1294:   circles in (b) at $Re=5$ 
1295:   mark the coordinates of the saddle in 
1296:   Fig.~\ref{FIG.2EW}. The radius ratio is $\eta=0.5$. 
1297:  \label{FIG.SATTEL}}  
1298:  \end{figure} 
1299: %----------------- FIG 5 --------------------------------------------------
1300:  \begin{figure}
1301:  \includegraphics[clip=true,width=10cm, angle=0]{./FIG/fig5.ps}
1302:  \caption{Real parts $\gamma (Q)$ of the two biggest eigenvalues of the NSE over
1303:  the complex $Q-$plane for $m=-1$, $R_2=-125$, $Re=5$, 
1304:  $\mu=\mu_{c-a}$, $\eta=0.5$. The filled circle  
1305:  marks the saddle at 
1306:  $\gamma (Q^\ast) = 0$ that determines the boundary beteen convective and
1307:  absolute instability. 
1308:  \label{FIG.2EW}}  
1309:  \end{figure}  
1310: %----------------- FIG 6 --------------------------------------------------
1311:  \begin{figure}
1312:  \includegraphics[clip=true,width=10cm, angle=0]{./FIG/fig6.ps}
1313:  \caption{Full lines are sections through the surfaces of $\gamma (Q)$ 
1314:  (Fig.~\ref{FIG.2EW}) for the two biggest eigenvalues of the NSE at constant 
1315:  values of $k$ in the vicinity of the intersection of the surfaces. Dashed
1316:  lines show $\gamma (K)$ obtained from the GLE approximation to the dispersion
1317:  relation. Parameters are $m=-1$, $R_2=-125$, $Re=5$, $\mu=\mu_{c-a}$, 
1318:  $\eta=0.5$. 
1319:  \label{FIG.SATTEL-SCHNITTE}}  
1320:  \end{figure} 
1321: %----------------- FIG 7 --------------------------------------------------
1322:  \begin{figure}
1323:  \includegraphics[clip=true,width=10cm, angle=0]{./FIG/fig7.ps}
1324:  \caption{Stability boundaries of the basic flow state as functions of the
1325:  through-flow Reynolds number $Re$ for $m=1$, $R_2=-125$. The full (dashed) 
1326:  boundary lines, $\epsilon_{c-a}$, between the 
1327:  convectively and absolutely unstable parameter regions for vortex growth 
1328:  were obtained from the eigenvalues of the full NSE (the GLE approximation) --
1329:  cf. Sec.~\ref{SEC.LOCAL-PERTUB}. Dotted lines show the reduced critical 
1330:  bifurcation thresholds 
1331:  $\epsilon_c$ (\ref{epsilon_c}) for axially extended vortex patterns.  
1332:  The radius ratio is $\eta=0.5$.
1333:  \label{FIG.EPSILON-RE=20}}  
1334:  \end{figure} 
1335: %----------------- FIG 8 --------------------------------------------------
1336:  \begin{figure}
1337:  \includegraphics[clip=true,width=10cm, angle=0]{./FIG/fig8.ps}
1338:  \caption{Schematic plot of different vortex fronts. Thick full lines in (a) and 
1339:  thin full lines in (b) show intensity envelopes  of $+$ type and of $-$ type 
1340:  fronts connecting to the basic state  at $z \to -\infty$ and at 
1341:  $z \to \infty$, respectively. This line convention --- thick ones for $+$ type 
1342:  fronts and thin ones for $-$ type fronts --- is used also in 
1343:  Fig.~\ref{FIG.FRONTEN}. Dashed lines indicate the vortex field 
1344:  growing under the front.
1345:  \label{FIG.FRONTDIRECT}}  
1346:  \end{figure} 
1347: %----------------- FIG 9 --------------------------------------------------
1348:  \begin{figure}
1349:  \includegraphics[clip=true, width=13cm, angle=0]{./FIG/fig9.ps}
1350:  \caption{Front properties of vortices with azimuthal wave numbers $m=0, 1, -1$ 
1351:  in systems with and without through-flow $Re$ for $R_2$ = 0, $\eta=0.5$ as 
1352:  functions of  $\mu = R_1/R_{1c}(Re) -1$. Within each
1353:  2$\times$2 block of figures (a)-(i) the left column shows the axial wave number
1354:  $k^\ast$ and the axial growth rate $K^\ast$. The right column shows the front 
1355:  velocity $v^\ast$ and the frequency $\omega^\ast$ in the laboratory frame. 
1356:  Thick lines correspond to $+$ fronts, thin lines to $-$ fronts, respectively.
1357:   Full (dashed) lines result from the saddle point analysis of the  
1358:   dispersion relation of the NSE (GLE approximation). 
1359:  \label{FIG.FRONTEN}}  
1360:  \end{figure}
1361:  \clearpage 
1362:  
1363: %----------------- Tables--------------------------------------------------
1364: %----------------- TAB 1 --------------------------------------------------
1365:  \begin{table}[h] 
1366:  \begin{center}
1367:  \begin{tabular}{c|cccc} 
1368:  %\hline
1369:  &\multicolumn{4}{c}{Operation}\\
1370:  &\ \ A&\ \ B&\ \ C&\ \ D \\
1371:  \hline
1372:  \hline
1373:  $k \rightarrow$&$\quad k$&$\ -k$&$\ -k$&$\quad k$\\
1374:  %\hline
1375:  $m \rightarrow$&$\quad m$&$\ -m$&$\quad m$&$\ -m$\\
1376:  %\hline
1377:  $Re \rightarrow$&$\quad Re$&$\quad Re$&$\ -Re$&$\ -Re$\\
1378:  %\hline
1379:  %&\multicolumn{4}{c}{Transformation}\\
1380:  %\hline
1381:  %\hline
1382:  $\sigma \rightarrow$&$\quad \sigma$&$\quad {\overline \sigma}$&$\quad \sigma$&$\quad {\overline \sigma}$\\
1383:  %\hline
1384:  $U \rightarrow$&$\quad U$&$\quad {\overline U}$&$\quad U$&$\quad {\overline U}$\\
1385:  %\hline
1386:  $V \rightarrow$&$\quad V$&$\quad {\overline V}$&$\quad V$&$\quad {\overline V}$\\
1387:  %\hline
1388:  $W \rightarrow$&$\quad W$&$\quad {\overline W}$&$\ -W$&$\ -{\overline W}$\\
1389:  %\hline
1390:  $X \rightarrow$&$\quad X$&$\quad {\overline X}$&$\quad X$&$\quad {\overline X}$\\
1391:  %\hline
1392:  $Y \rightarrow$&$\quad Y$&$\quad {\overline Y}$&$\quad Y$&$\quad {\overline Y}$\\
1393:  %\hline
1394:  $Z \rightarrow$&$\quad Z$&$\quad {\overline Z}$&$\ -Z$&$\ -{\overline Z}$\\
1395:  %\hline
1396:  \end{tabular}
1397:  \end{center}
1398:  \caption{Transformation behavior of the eigenvalues and eigenfunctions of the
1399:  eigenvalue problem (\ref{resultEW} - \ref{defineL}) under symmetry operations.
1400:  Here A denotes the identity, B complex conjugation, C axial reflection, and 
1401:  D complex conjugation (indicated by an overbar) combined with axial reflection.
1402:  \label{TAB.MOD-SYMM}}
1403:  \end{table}
1404: %----------------- TAB 2 --------------------------------------------------
1405:  \begin{table}[h] 
1406:  \begin{tabular}{|c||c|c|c|c|c|c|c|c|c|}
1407:  \hline
1408:  $R_{2}$&-150&-125&-100&-75&-50&-25&0&25&50\\
1409:  \hline
1410:  %\multicolumn{10}{c}{} \\
1411:  \hline
1412:  \multicolumn{10}{|c|}{$\epsilon_{c}=a_{2}Re^{2}+a_{4}Re^{4}$} \\
1413:  \hline
1414:  $a_{2}*10^{4}$&0.590&0.795&1.089&1.502&2.083&3.115&3.679&2.447&1.307\\
1415:  \hline
1416:  $a_{4}*10^{9}$&5.181&3.966&1.151&-2.804&-3.385&-15.62&-34.97&-14.55&-0.958\\
1417:  \hline
1418:  %\multicolumn{10}{c}{} \\
1419:  \hline  
1420:  \multicolumn{10}{|c|}{$\epsilon_{c-a}=a_{2}Re^{2}+a_{4}Re^{4}$} \\
1421:  \hline
1422:  $a_{2}*10^{3}$&1.181&1.451&1.854&2.516&3.906&5.588&6.238&4.338&2.483\\
1423:  \hline
1424:  $a_{4}*10^{6}$&0.116&0.139&0.182&0.293&-2.583&-4.642&-5.716&-3.682&-1.740\\
1425:  \hline 
1426:  \multicolumn{10}{|c|}{$\epsilon_{c-a}^{GLE}=a_{2}Re^{2}+a_{4}Re^{4}$} \\
1427:  \hline
1428:  $a_{2}*10^{3}$&1.161&1.416&1.800&2.517&4.068&6.547&7.732&5.152&2.762\\
1429:  \hline
1430:  $a_{4}*10^{6}$&-0.164&-0.268&-0.481&-1.011&-2.106&-3.807&-4.900&-3.022&-1.425\\
1431:  \hline 
1432:  \end{tabular}
1433:  \caption{Fitparameters for the $Re$-dependence of the stability boundaries of
1434:  the basic state against growth of TVF $(m=0)$ perturbations for different
1435:  $R_2$: reduced critical bifurcation threshold $\epsilon_c$ of axially extended 
1436:  vortex patterns and boundary $\epsilon_{c-a}$ between convectively and 
1437:  absolutely unstable parameter regime. Here $\epsilon_{c-a}^{GLE}$ is obtained
1438:  from the GLE approximation (cf. Sec.~\ref{SEC.GLA1}). 
1439:  \label{TAB.FITPARA}}
1440:  \end{table}
1441: %----------------- TAB 3 --------------------------------------------------
1442:  
1443:  \begin{table}
1444:  \begin{tabular}{|c||c|c|c|c|c|c|c|c|c|}
1445:  \hline
1446:  $R_{2}$&-150&-125&-100&-75&-50&-25&0&25&50\\
1447:  \hline
1448:  %\multicolumn{10}{c}{} \\
1449:  \hline
1450:  \multicolumn{10}{|c|}{$\epsilon_{c}=a_{1}Re+a_{2}Re^{2}+a_{3}Re^{3}+a_{4}Re^{4}$} \\
1451:  \hline
1452:  $a_{1}*10^{3}$&-1.582&-2.365&-3.513&-4.887&-6.143&-5.833&-3.372&-1.402&-0.514\\
1453:  \hline
1454:  $a_{2}*10^{4}$&0.957&1.199&1.496&1.925&2.520&3.174&3.197&1.806&0.581\\
1455:  \hline
1456:  $a_{3}*10^{7}$&-9.181&-9.001&-8.485&-7.678&-6.741&-4.172&-3.832&-3.572&-3.685\\
1457:  \hline
1458:  $a_{4}*10^{9}$&1.018&-0.378&-0.180&-0.563&-3.673&-21.77&-36.15&-5.707&19.92\\
1459:  \hline
1460:  %\multicolumn{10}{c}{} \\
1461:  \hline
1462:  \multicolumn{10}{|c|}{$\epsilon_{c-a}=a_{0}+a_{1}Re+a_{2}Re^{2}+a_{3}Re^{3}+a_{4}Re^{4}$} \\
1463:  \hline
1464:  $a_{0}*10^{3}$&2.358&5.497&3.775&-1.237&4.095&12.25&18.96&10.09&2.610 \\
1465:  \hline
1466:  $a_{1}*10^{3}$&-3.268&-3.578&-3.048&-0.449&0.926&1.159&1.012&1.217&1.455 \\
1467:  \hline
1468:  $a_{2}*10^{3}$&1.532&1.964&2.618&3.539&4.351&5.535&6.036&4.357&2.603 \\
1469:  \hline
1470:  $a_{3}*10^{6}$&8.408&11.87&14.97&9.992&6.487&4.919&3.680&1.367&-0.143 \\
1471:  \hline
1472:  $a_{4}*10^{6}$&-0.169&-0.368&-0.815&-1.977&-2.809&-4.068&-4.827&-3.203&-1.549 \\
1473:  \hline
1474:  \multicolumn{10}{|c|}{$\epsilon_{c-a}^{GLE}=a_0+a_1Re+a_2Re^2+a_3Re^3+a_4Re^4$} \\
1475:  \hline
1476:  $a_{0}*10^{3}$&1.101&-2.541&1.405&2.529&3.458&3.998&3.641&1.320&0.062\\
1477:  \hline
1478:  $a_{1}*10^{3}$&-4.117&0.755&-1.044&-0.318&0.831&2.063&1.495&1.377&1.625\\
1479:  \hline
1480:  $a_{2}*10^{3}$&2.030&2.792&2.820&3.435&4.437&6.166&7.265&4.981&2.723 \\
1481:  \hline
1482:  $a_{3}*10^{6}$&-8.845&-30.34&-7.121&4.780&10.58&8.229&-8.191&-12.58&-10.03 \\
1483:  \hline
1484:  $a_{4}*10^{6}$&-0.822&-2.713&-1.136&-1.004&-1.674&-2.923&-3.799&-2.234&-0.854 \\
1485:  \hline
1486:  \end{tabular}
1487:  \caption{Fitparameters for the $Re$-dependence of the stability boundaries of
1488:  the basic state against growth of L-SPI $(m=1)$ perturbations for different
1489:  $R_2$: reduced critical bifurcation threshold $\epsilon_c$ of axially extended 
1490:  vortex patterns and boundary $\epsilon_{c-a}$ between convectively and 
1491:  absolutely unstable parameter regime. Here $\epsilon_{c-a}^{GLE}$ is obtained
1492:  from the GLE approximation (cf. Sec.~\ref{SEC.GLA1}). The results for 
1493:  R-SPI $(m=-1)$ perturbations are obtained according to
1494: Sec.~\ref{SEC.SYMM-EIGENVALUE} from those for L-SPI $(m=1)$ by 
1495: $Re \rightarrow -Re$, i.e., by changing the sign of the odd coefficients in 
1496: the table.
1497: \label{TAB.FITPARA2}}
1498:  \end{table}
1499: %--------------------------------------------------------------------------- 
1500: \end{document}   
1501: