1: \documentclass[preprint, aps, pre, eqsecnum,amsmath,amssymb,showpacs]{revtex4}
2:
3: \usepackage[final]{graphicx}
4:
5: \renewcommand{\arraystretch}{1.5}
6: \newcommand{\Nabla}{\mbox{\bf\boldmath $\nabla$}}
7: \newcommand{\ve}[1]{\stackrel{\rightharpoonup}{#1}}
8: \renewcommand{\vec}[1]{{\bf #1}}
9: \newcommand{\Vxi}{\mbox{\bf\boldmath $\xi$}}
10: \newcommand{\Vphi}{\mbox{\boldmath $\varphi$}}
11: \newcommand{\Vzeta}{\mbox{\boldmath $\zeta$}}
12: \newcommand{\VPsi}{\mbox{\boldmath $\Psi$}}
13: \newcommand{\Overrightarrow}[1]{\stackrel{\textstyle\rightarrow}{#1}}
14: \newcommand{\ra}{\right\rangle}
15: \newcommand{\la}{\left\langle}
16: \newcommand{\bean}{\begin{eqnarray}}
17: \newcommand{\eean}{\end{eqnarray}}
18: \newcommand{\bea}{\begin{eqnarray*}}
19: \newcommand{\eea}{\end{eqnarray*}}
20: \newcommand{\beq}{\begin{equation}}
21: \newcommand{\eeq}{\end{equation}}
22: \newcommand{\D}{\displaystyle}
23: \newcommand{\quer}{\overline}
24: \newcommand{\ul}{\underline}
25: \newcommand{\ob}{\overbrace}
26: \newcommand{\ub}{\underbrace}
27: \newcommand{\greeksym}[1]{{\usefont{U}{psy}{m}{n}#1}}
28: \newcommand{\umu}{\mbox{\greeksym{m}}}
29: \newcommand{\udelta}{\mbox{\greeksym{d}}}
30: \newcommand{\uDelta}{\mbox{\greeksym{D}}}
31: \newcommand{\uPi}{\mbox{\greeksym{P}}}
32: \def\lesssim{\mathrel{\mathpalette\vereq<}}
33: \def\vereq#1#2{\lower3pt\vbox{\baselineskip1.5pt \lineskip1.5pt
34: \ialign{$\hfill##\hfil$\crcr#2\crcr\sim\crcr}}}
35: \def\gtrsim{\mathrel{\mathpalette\vereq>}}
36:
37: \pagestyle{myheadings}
38: \markboth{\today}{\today}
39: \bibliographystyle{prsty}
40:
41:
42: \begin{document}
43:
44: \title{\bf Spiral Vortices and Taylor Vortices
45: in the Annulus between Rotating Cylinders and the Effect of an Axial Flow}
46:
47: \author{Ch.~Hoffmann, M.~L\"{u}cke, and A.~Pinter}
48: \affiliation{Institut f\"{u}r Theoretische Physik, \\ Universit\"{a}t
49: des Saarlandes, D-66041~Saarbr\"{u}cken, Germany\\}
50:
51: \date{\today}
52:
53: \begin{abstract}
54: We present numerical simulations of vortices that appear via primary
55: bifurcations out of the unstructured circular Couette flow in the Taylor-Couette
56: system with counter-rotating as well as with co-rotating cylinders. The full,
57: time dependent
58: Navier-Stokes equations are solved with a combination of a finite difference
59: and a Galerkin method for a fixed axial periodicity length of the vortex
60: patterns and for a finite system of aspect ratio 12 with rigid nonrotating ends
61: in a setup with radius ratio $\eta=0.5$. Differences in structure,
62: dynamics, symmetry properties, bifurcation and stability behavior between
63: spiral vortices with azimuthal wave numbers $M=\pm1$ and $M=0$ Taylor vortices
64: are elucidated and compared in quantitative detail. Simulations in axially
65: periodic systems and in finite systems with stationary rigid ends are compared
66: with experimental spiral data.
67: In a second part of the paper we determine how the above listed
68: properties of the $M=-1,0,1$
69: vortex structures are changed by an externally imposed axial through-flow
70: with Reynolds numbers in the range $-40 \le Re \le 40$.
71: Among others we investigate when left handed or right handed spirals or
72: toroidally closed vortices are preferred.
73: \end{abstract}
74:
75:
76: \pacs{PACS number(s): 47.20.-k, 47.32.-y, 47.54.+r, 47.10.+g}
77:
78: \maketitle
79: \clearpage
80:
81:
82: \section{Introduction}
83: Spiral vortices appearing in the annular gap between the concentric rotating
84: cylinders of the Taylor-Couette system \cite{T94} are a rather interesting
85: example for the
86: spontaneous formation of a helicoidal structure out of a homogeneous state of a
87: nonlinear forced
88: system when the forcing exceeds a critical threshold. Like the competing
89: toroidally closed Taylor vortices the spiral vortex structures
90: bifurcate out of the unstructured basic state of circular Couette flow (CCF)
91: that is
92: stable at small rotation rates of the inner cylinder. The spiral pattern breaks
93: the rotational symmetry of the annular gap. It oscillates in time by rotating
94: azimuthally as a whole thereby propagating axially. The Taylor vortex flow
95: (TVF), on the other hand is rotationally symmetric and stationary.
96:
97: The spiral pattern is effectively one dimensional (like TVF) and stationary
98: when seen from a co-moving frame \cite{CI94}: the spiral fields do not depend on
99: time $t$,
100: axial coordinate $z$, and azimuthal angle $\varphi$ separately but only via the
101: combined phase variable $\phi= kz + M\varphi -\omega(k, M) t$. Here $k$ and $M$
102: are the axial and azimuthal wave numbers, respectively, and $\omega$ the
103: frequency. In the $\varphi-z$ plane of an 'unrolled' cylindrical surface the
104: lines of constant phase, $\phi = \phi_0$, are straight with slope $-M/k$ as
105: shown in Fig.~\ref{FIG:phases}.
106: An azimuthal wave
107: number $M>0$ implies a left handed spiral (L-SPI) while $M<0$ refer to
108: right handed spirals (R-SPI) with our convention of taking $k$ to be positive.
109: L-SPI and R-SPI being mirror images of each
110: other under the operation $z \to -z$ are symmetry degenerate flow states. Which
111: of them is realized in a particular experimental or numerical setup depends on
112: the initial conditions.
113:
114: With the lines of constant phase in the $\varphi-z$ plane being oriented for
115: both spiral types obliquely to the azimuthal 'wind' of the basic CCF both
116: spirals are advectively rotated by the latter like rigid objects. Their common
117: angular velocity is $\dot{\varphi}_{SPI} = \omega(k,M)/M$.
118: This advection enforced rigid-body rotation of the spiral vortices is also
119: reflected by the fact that the axial phase velocities
120: $w_{ph} = \omega /k = \dot{\varphi}_{SPI} M /k$ of an L-SPI
121: ($M>0$) and of an R-SPI ($M<0$) are opposite to each other, see
122: Fig.~\ref{FIG:phases}.
123: By the same token the rotationally symmetric ($M=0$) structure of toroidally
124: closed Taylor vortices is stationary ($\omega=0$):
125: the lines of constant phases being parallel to the azimuthal CCF the latter
126: cannot advect these vortices. However, an externally imposed axial
127: through-flow can advect Taylor vortices as well as spiral vortices.
128:
129: The external through-flow breaks the mirror symmetry between L-SPI and R-SPI.
130: It changes their rotation and propagation dynamics as well as their structural
131: properties and their bifurcation behavior in different ways. This is the topic
132: of our investigation.
133:
134: In his review \cite{T94} Tagg remarks that systematic investigation of
135: non-axisymmetric vortex states that appear via primary bifurcations out of the CCF
136: state started remarkably late in the history of the Taylor-Couette problem.
137: Krueger et al. \cite{KGD66} predicted in 1966 primary transitions to
138: non-axisymmetric
139: rotating-wave flow which then were observed in experiments by Snyder
140: \cite{S68} who had presented experimental evidence for different types of
141: stable helical flow (referred to as 'spirals') a few years earlier. In 1985,
142: an experimental survey was published by Andereck et al. \cite{ALS85}
143: which classified a large variety of different flow states, including some
144: spiral types like linear, modulated, interpenetrating, and wavy spirals etc.
145: An extensive numerical linear stability analysis was then performed for a
146: wide range of radius ratios by Langford et al. \cite{LTKSG88}. At this
147: time, Tagg et al. \cite{TESM88} experimentally observed a transition from
148: CCF to axially standing and azimuthally
149: traveling waves (ribbons) and found numerically calculated wave speeds to be
150: in agreement with experimental results. Edwards \cite{E90} studied the
151: transition from CCF to traveling waves. More recent experiments were done
152: with a system of radius ratio $\eta=0.5$ \cite{SP99}. Spiral solutions in a
153: relatively narrow gap
154: with axially periodic boundary conditions were obtained numerically with a
155: pseudo-spectral method using co-rotating helicoidal coordinates which were
156: adapted to the expected spiral \cite{AMS98}.
157:
158: Various effects of an externally imposed axial through-flow in the Taylor-Couette
159: system have been explored since the early 1930 so that the list of publications
160: cannot be discussed here -- see, e.g, Ref.~\cite{BLRS96} for a
161: partial and far from complete compilation. We mention here in addition
162: a few, more recent papers on selected topics beyond those listed in
163: Ref.~\cite{BLRS96}:
164: linear analysis of the competition between shear and centrifugal instability
165: mechanisms \cite{GG93,MM02}; linear SPI and TVF fronts and
166: pulses \cite{PLH03}; weakly nonlinear bifurcation analysis of axially extended
167: spiral, ribbon, and mixed vortex states with homogeneous amplitudes
168: \cite{RL93,CI94}; theoretical/numerical investigation of the nonlinear pattern
169: selection in the absolutely unstable regime under downstream evolving intensity
170: envelopes \cite{BLRS96}; theoretical/numerical analysis of noise-sustained
171: patterns in the convectively unstable regime \cite{noise-effects} (related
172: experiments are listed in \cite{BLRS96}); analysis of the changes
173: in the noise sensitivity across the convective-absolute stability boundary
174: \cite{noise-sensitivity}; measurements of velocity fields by particle image
175: velocimetry \cite{WL99}.
176:
177: In this work we explore in a detailed quantitative investigation the
178: spatio-temporal structures as well as the bifurcation properties of spirals and
179: TVF in a setup with co- and
180: counter-rotating cylinders of fixed radius ratio $\eta=0.5$ with and without an
181: externally imposed axial through-flow. Most calculations were done for
182: axially periodic boundary conditions that impose the wave length of the vortex
183: pattern. However also a few simulations of finite systems with rigid,
184: non-rotating lids were done to compare with experiments and to study the effect
185: of phase propagation suppressing boundaries.
186: The calculations were done with a time dependent finite differences
187: method in the $r-z$ plane combined with a spectral decomposition in $\varphi$
188: which yields by construction only the stable flows. However, by selectively
189: suppressing destabilizing modes we also were able to trace out the unstable
190: TVF and SPI solution branches. We do not include in this work results on
191: ribbons \cite{TESM88}, i.e., nonlinear combinations of L and R spirals
192: \cite{CI94} since they were unstable for the parameters investigated here.
193:
194: In Sec.~\ref{SEC:System} we present the notation for describing the Taylor
195: Couette system and we describe our numerical method. In Sec.~\ref{SEC:SPI-TVF}
196: we review the spatio-temporal properties of TVF and SPI and we present results
197: on their bifurcation behavior and flow structure in the absence of
198: through-flow. In particular we provide detailed comparisons of the bifurcation
199: and structural properties of these primary vortex states. Also comparisons with
200: experiments are presented and discussed.
201: In Sec.~\ref{SEC:ETF} we elucidate the effect of an external through-flow on
202: structure, dynamics, and bifurcation properties of TVF and SPI for counter-rotating
203: cylinders and stationary outer cylinder. The last section contains a summary of
204: the main results.
205:
206: \section{System} \label{SEC:System}
207: We report results obtained numerically for a Taylor-Couette system with co-
208: and counter-rotating cylinders. The ratio $\eta=r_1/r_2$ of the radii $r_1$
209: and $r_2$ of the inner and outer cylinders, respectively, was fixed at the
210: value $\eta=0.5$ for which also experiments have been made recently
211: \cite{SP99}.
212:
213: \subsection{Theoretical description}
214: We consider the fluid in the annulus between the cylinders to be isothermal
215: and incompressible with kinematic viscosity $\nu$. The gap width $d=r_2-r_1$
216: is used as the unit of length and the momentum diffusion time $d^2/ \nu$
217: radially across the gap as the time unit so that velocities are reduced by
218: $\nu /d$. To characterize the driving of the system, we use the Reynolds
219: numbers
220: \begin{eqnarray}
221: R_1=r_1\Omega_1 d/\nu \,\, ; R_2=r_2\Omega_2 d/\nu \, .
222: \end{eqnarray}
223: They are just the reduced azimuthal velocities of the fluid at the inner and
224: outer cylinder, respectively, where $\Omega_1$ and $\Omega_2$ are the
225: respective angular velocities of the cylinders. The inner one is always
226: rotating counterclockwise so that $\Omega_1$ and $R_1$ are positive.
227: We explore positive as well as negative values of $R_2$
228: corresponding to co- as well as counter rotation of the cylinders,
229: respectively. And we elucidate the effect of an externally imposed axial
230: through-flow.
231:
232: Within the above described scaling, the NSE take the form
233: \begin{eqnarray}
234: \partial_t {\bf u} = \Nabla^2 {\bf u} -
235: ({\bf u}\cdot \Nabla){\bf u} - \Nabla p \,.
236: \end{eqnarray}
237: Here $p$ denotes the pressure reduced by $\rho \nu^2/d^2$ and $\rho$ is the mass
238: density of the fluid. Using cylindrical coordinates, the velocity field
239: \begin{eqnarray}
240: {\bf u}=u\,{\bf e}_r + v\,{\bf e}_\varphi + w\,{\bf e}_z
241: \end{eqnarray}
242: is decomposed into a radial component $u$, an azimuthal one $v$, and an
243: axial one $w$.
244:
245: We have solved the resulting equations subject to no slip conditions at the
246: cylinders. In Sec.~\ref{SEC:CER} we
247: present simulations of systems with axial size $\Gamma=12$ and
248: rigid stationary ends bounding the annulus axially in order to compare
249: with experiments \cite{SP99}. For the main part (c.f. Secs.~\ref{SEC:SPI-TVF}
250: and \ref{SEC:ETF}) of this work we imposed, however, axially periodic boundary
251: conditions at $z=0$ and $z=\Gamma=1.6$. So the axial wavelength of the TVF
252: and the SPI patterns investigated there is $\lambda=1.6$ and the wave number
253: is $k=2\pi/\lambda=3.927$.
254:
255: \subsection{Numerical method}\label{SEC:Num-method}
256: The calculations were done with a finite differences
257: method in the $r-z$ plane combined with a spectral decomposition in $\varphi$.
258: Since we have been studying also finite length cylinders, say, with lids
259: bounding
260: the annulus vertically, we do not use here a spectral decomposition in axial
261: direction which for axially periodic systems is a generic alternative.
262: The discretization (a FTCS - Forward Time, Centered
263: Space algorithm) has been done on staggered grids in the $r-z$ plane
264: following the procedure of Ref. \cite{HNR75}. It yields simple
265: expressions for the derivatives, it does not require boundary conditions for the
266: pressure, and it avoids difficulties with boundary conditions for more than one
267: velocity field component at the same position.
268: We used homogeneous grids with discretization lengths $\Delta r=\Delta
269: z=0.05$ which have showed to be more accurate than non-homogeneous grids.
270: Time steps were $\Delta t<1/3600$.
271:
272: Azimuthally all fields $f=u,v,w,p$ were expanded as
273: \begin{eqnarray}\label{EQ-expansion}
274: f(r,\varphi,z,t)=
275: \sum_{m=-m_{max}}^{m_{max}} f_m(r,z,t)\,e^{im\varphi} .
276: \end{eqnarray}
277: For the flows investigated here a truncation of the above Fourier expansion
278: at $m_{max}=8$ was sufficient to properly resolve the anharmonicities in the
279: fields. The system of coupled equations for
280: the amplitudes $f_m(r,z,t)$ of the azimuthal normal modes
281: $-m_{max} \leq m \leq m_{max}$ is solved with the FTCS algorithm.
282: Pressure and velocity fields are iteratively adjusted to each other with the
283: method of 'artificial compressibility' \cite{PT83}
284: \begin{eqnarray}
285: \label{dp-n}
286: dp^{(n)} &=& - \beta \, \Nabla \cdot {\bf u}^{(n)} \qquad (0<\beta<1) \\
287: p^{(n+1)} &=& p^{(n)} + dp^{(n)} \\
288: \label{u-nplus1}
289: {\bf u}^{(n+1)} &=& {\bf u}^{(n)} - \Delta t \, \Nabla (dp^{(n)}) \,.
290: \end{eqnarray}
291: The pressure correction $dp^{(n)}$ in the $n$-th
292: iteration step being proportional to the divergence of ${\bf u}^{(n)}$
293: is used to adapt the velocity field ${\bf u}^{(n+1)}$.
294: The iteration loop (\ref{dp-n}-\ref{u-nplus1}) is executed for each azimuthal
295: Fourier mode separately. It is iterated until $\Nabla \cdot {\bf u}$ has
296: become sufficiently small for each $m$ mode considered --
297: the magnitude of the total divergence never exceeded 0.02 and typically
298: it was much smaller. After that the next FTCS time step was executed.
299:
300: For code validation we compared SPI solutions with experiments \cite{SP99} and
301: TVF solutions with previous numerical simulations \cite{BLRS96} and close to
302: onset also with Ginzburg-Landau results \cite{REC-LUE-MUE}. Furthermore, we
303: compared bifurcation thresholds of the nonlinear SPI and TVF solutions with
304: the respective stability boundaries of the linearized NSE \cite{LTKSG88,PLH03}
305: obtained by a shooting method that is described in detail in
306: \cite{PLH03}. As expected from our experience with primary vortex
307: structures in the Taylor-Couette and Rayleigh-Benard problem lie the MAC FTCS
308: bifurcation thresholds for our discretization typically 1 - 2 \% below the
309: respective linear stability thresholds. This deviation significantly reduces
310: for finer discretizations. We also investigated how the nonlinear solutions
311: change when varying $m_{max}$ and/or the grid spacing. From these analyses we
312: conservatively conclude that typical SPI frequencies have an error of less than
313: about 0.2\% and that typical velocity field amplitudes can be off by about
314: 3 - 4\%. Time steps were always well below the von Neumann stability
315: criterion and by more than a factor of three below the Courant-Friederichs-Lewy
316: criterion.
317:
318: In order to trace out the unstable parts of bifurcation branches of TVF and SPI
319: solutions we applied different stabilization methods that are described in
320: Sec.~\ref{RFA}.
321:
322: \section{Spiral vortices and Taylor vortices}
323: \label{SEC:SPI-TVF}
324: In this section we first briefly review spatio-temporal properties of
325: spiral vortices ($M\neq0$) and Taylor vortices ($M=0$) in the absence of any
326: externally enforced axial through-flow. Here $M$ is the azimuthal wave number
327: of the respective vortex structure.
328: Then we present our results on the bifurcation behavior of $M=0$ and $M=\pm1$
329: vortex solutions and on their flow structure.
330:
331: They both grow out of the basic CCF state,
332: ${\bf u}_{CCF} = v_{CCF}(r){\bf e}_{\varphi}$, that is rotationally symmetric,
333: axially homogeneous, and time translationally invariant. Here in our system
334: with $\eta=1/2$ the radial profile of its azimuthal velocity reads
335: \begin{equation}\label{v_CCF}
336: v_{CCF}(r) =\frac{2R_2-R_1}{3}r + \frac{4R_1-2R_2}{3}\frac{1}{r} \, .
337: \end{equation}
338:
339: \subsection{Spatio-temporal structure}
340:
341: The spiral vortex structure is periodic in $\varphi,z$, and $t$. It
342: rotates uniformly as a whole like a rigid object in
343: azimuthal direction thereby translating with constant phase velocity in axial
344: direction --- the spiral fields $f(r,\varphi,z,t)$ do not depend on
345: $\varphi,z,t$ separately but only the phase combination
346: \begin{equation}\label{phase}
347: \phi= kz + M\varphi -\omega(k, M) t \, .
348: \end{equation}
349: Here $k$ is the axial wave number that we always take to be positive and
350: $\omega(k, M)$ is the frequency. Thus, with
351: $f(r,\varphi,z,t) = F(r, \phi)$, the spiral pattern is
352: one dimensional. Comparing the Fourier decompositions
353: \begin{subequations}
354: \begin{eqnarray}
355: f(r,\varphi,z,t) = \sum_{m,n} f_{m,n}(r,t)\,e^{i(m\varphi + nkz)}
356: = \sum_{\nu} F_{\nu}(r)\,e^{i\nu \phi} = F(r,\phi)
357: \end{eqnarray}
358: one finds that
359: \begin{eqnarray}
360: f_{m,n}(r,t) = \delta_{m,nM} \, e^{-i n\omega t} F_n(r)\,.
361: \end{eqnarray}
362: \end{subequations}
363: Thus only the mode combinations $m=nM$ appear in a SPI with azimuthal wave
364: number $M$.
365:
366: The SPI phase is constant, $\phi_0$, on a cylindrical surface,
367: $r=const$, along lines given by the equation
368: \begin{equation} \label{z_phase}
369: z_0 = - \frac{M}{k}\varphi + \frac{\omega(k,M)}{k}t + \frac{1}{k}\phi_0 \, .
370: \end{equation}
371: Thus, on the $\varphi-z$ plane of such an 'unrolled' cylindrical surface these
372: lines of constant phase are straight with slope $-M/k$. And an azimuthal wave
373: number $M>0$ implies a left handed spiral (L-SPI) while $M<0$ refer to
374: right handed spirals (R-SPI) with our convention of taking $k$ to be positive.
375: L-SPI and R-SPI being mirror images of each
376: other under the operation $z \to -z$ are symmetry degenerate flow states. Which
377: of them is realized in a particular experimental or numerical setup depends on
378: the initial conditions.
379:
380: The lines of constant phase and with it the whole spiral structure rotates
381: in $\varphi$ with angular velocity
382: \begin{equation} \label{phidot_phase}
383: \dot{\varphi}_{SPI} = \frac{\omega}{M} \, .
384: \end{equation}
385: Its direction strongly depends on the inner cylinder's rotation due to the
386: influence of the CCF. The latter decisively determines the shape of the
387: linear spiral eigenmodes that can grow beyond the stability boundary of the CCF state
388: against perturbations with azimuthal wave number $M\neq0$. In the
389: parameter range explored here the spirals rotate into the same direction as the
390: inner cylinder, i.e., into positive $\varphi$-direction so that
391: $\omega(k, M)/M$ is always positive, i.e., $\omega = sign (M) |\omega|$.
392: From this rigid rotation one immediately infers from Eq.(\ref{phase}) that
393: the axial phase velocity
394: \begin{equation}
395: w_{ph} = \frac{\omega}{k} = \frac{M}{k} \dot{\varphi}_{SPI}
396: \end{equation}
397: of an L-SPI ($M>0$) is positive and of an R-SPI ($M<0$) it is negative.
398:
399: For the rotationally symmetric ($M=0$) structure of toroidally closed Taylor
400: vortices the lines of constant phases are parallel to ${\bf e}_\varphi$.
401: This $M=0$ pattern is stationary ($\omega=0$). The main reason is that
402: the azimuthal flow of the basic CCF state being precisely parallel to the
403: vortex lines of constant phase cannot advect them. However, an axial
404: mean flow, being perpendicular to them can advect them: an externally
405: enforced axial through-flow of strength $Re$ causes a non-zero
406: axial phase velocity of the Taylor vortex pattern that grows linearly with $Re$,
407: at least when phase pinning
408: effects are absent as for axially periodic boundary conditions.
409:
410: \subsection{Bifurcation behavior}
411: In the parameter regime considered here the bifurcation thresholds for
412: nonlinear SPI and TVF solutions, i.e., the linear stability boundaries of the
413: CCF state against
414: $M=\pm 1$ and $M=0$ vortex perturbations \cite{LTKSG88} differ only slightly
415: from each other. For our fixed wave number of $k=3.927$ they intersect at
416: $(R_1^s=95.25, R_2^s=-73.69)$ where these two different vortex modes are
417: bi-"critical" in the sense that their growth rates are simultaneously zero.
418: The stability boundaries were obtained with a shooting method from the
419: linearized NSE. The nonlinear SPI and TVF solutions that were determined
420: with the numerical method described in Sec.~\ref{SEC:Num-method} give
421: bifurcation thresholds that differ as a result of the FTCS discretization
422: errors by at most 2\% from the linear stability analysis. However, this
423: difference can grow with externally applied through-flow up to, say, 5\%
424: at $Re\simeq40$ (c.f. Sec.~\ref{SEC:ETF}) when the discretization is not refined.
425:
426: \subsubsection{Radial flow amplitudes of TVF and SPI}
427: \label{RFA}
428:
429: The bifurcation of both, TVF and SPI solutions is forward as
430: shown by the bifurcation surface over the $R_1 - R_2$ plane of Fig.~\ref{FIG:BIF-u}.
431: There the respective vortex solution is characterized by the primary
432: Fourier amplitude, $|u_{m,n}|$, of the radial flow intensity at mid gap,
433: $r=r_1 + 0.5$, taken
434: as order parameter with $m$ denoting the azimuthal mode index and $n$ referring
435: to the axial one, respectively. Thus, Fig.~\ref{FIG:BIF-u} shows $|u_{0,1}|$ for
436: the TVF solution by thin lines and $|u_{1,1}|=|u_{-1,1}|$ for the two symmetry
437: degenerate $M=\pm 1$ SPI solutions by thick lines, respectively. In each
438: case stable (unstable) solutions are represented by full (dashed) lines.
439: The different stability regions labelled A - E are explained in the caption of
440: Fig.~\ref{FIG:BIF-u}.
441:
442: The stability of the vortex states refers to our system with fixed axial
443: periodicity length. Thus, e.g., Eckhaus or Benjamin-Feir instabilities
444: \cite{CRO-HOH} that can destabilize periodic patterns in infinite and large
445: systems do not occur here. Furthermore, our periodic
446: boundary conditions allowing free phase propagation enhance the
447: existence range as well as the stability range of SPI solutions in comparison
448: with, say, Ekman vortex generating stationary lids that axially close the
449: annulus in an experimental setup. The latter suppress phase propagation
450: in their vicinity so that phase generating and phase destroying defects near
451: opposite boundaries are necessary for the realization of spirals in the bulk
452: of such systems.
453:
454: In our setup TVF is for $R_2 > R_2^{s}$ stable close to onset. And it remains
455: so at least up to the
456: largest value of $R_1=130$ shown in Fig.~\ref{FIG:BIF-u} -- for larger $R_1$ TVF
457: eventually undergoes an oscillatory instability. For more negative
458: $R_2 < R_2^{s}$
459: TVF is unstable at onset (region C in Fig.~\ref{FIG:BIF-u}) but becomes stable at
460: larger $R_1$ in region E. The unstable TVF solution branch in region C was
461: obtained by suppressing any $m\neq0$ modes in the field
462: representation (\ref{EQ-expansion}), i.e., by allowing only rotationally
463: symmetric solutions. Lifting this mode restriction infinitesimal $m\neq0$
464: perturbations drive the system in the parameter region C of Fig.~\ref{FIG:BIF-u}
465: away from the unstable TVF solution into a stable SPI state.
466:
467: Spirals, on the other hand, are for $R_2 < R_2^{s}$ stable close to onset and
468: remain so at least up to the largest value of $R_1=130$ shown in
469: Fig.~\ref{FIG:BIF-u} while for $R_2 > R_2^{s}$ they are unstable at onset
470: (region D in Fig.~\ref{FIG:BIF-u}). But then they become stable at larger
471: $R_1$ in region E. The unstable SPI solution branch in region D was
472: obtained by suppressing $m=0$ contributions to the radial velocity field $u$
473: at mid gap location. This stabilized the SPI solution against the growth of
474: TVF. Lifting this restriction of the available mode space the unstable SPI
475: solutions in region D decay into stable Taylor vortices.
476:
477: In the relatively large region E both, SPI as well as TVF solutions coexist
478: bistably and the final vortex structure to be found here depends on the initial
479: conditions and the driving history of $R_1, R_2$. Note in particular that for
480: our periodic boundary conditions the region E with stable spirals extends to
481: positive $R_2$, i.e., to a situation with co-rotating cylinders.
482:
483: \subsubsection{SPI frequencies}
484:
485: In Fig.~\ref{FIG:BIF-om} the spiral frequencies $\omega$ are plotted over the same
486: control
487: parameter range as the radial flow amplitudes in Fig.~\ref{FIG:BIF-u}. Also
488: here we include -- for the sake of comparison with Fig.~\ref{FIG:BIF-u} -- the
489: identification of the different stability regions of TVF and SPI solution by
490: the symbols A-E explained in the caption of Fig.~\ref{FIG:BIF-u}. At onset $\omega$
491: agrees within the numerical accuracy of our nonlinear code with the eigenvalue
492: resulting from the linear stability analysis of the CCF state.
493:
494: The nonlinear SPI frequencies further away from onset
495: vary smoothly: the bifurcation surface of $\omega$ in Fig.~\ref{FIG:BIF-om}
496: has the shape of a cloth that hangs down from a frame given by the linear
497: onset spiral frequencies $\omega(R_{1,stab})$ at the stability threshold
498: $R_{1,stab}(R_2)$ of CCF. The location of minimal $\omega$ on the
499: bifurcation surface is shown by a thick line in Fig.~\ref{FIG:BIF-om}.
500: Thus, the nonlinear SPI frequencies are typically smaller than the linear
501: ones but do not deviate substantially from them.
502:
503: Since the linear onset frequencies show a characteristic variation along the
504: bifurcation threshold, $R_{1,stab}(R_2)$, that dictates the form of the
505: whole $\omega$ bifurcation surface we discuss them in some detail. They,
506: furthermore allow for a simple, yet semiquantitive
507: explanation of the phenomenon of rigid body rotation of
508: spirals in terms of a passive advection dynamics of $M = \pm 1$ vortex
509: perturbations, $e^{i\phi}$, with lines of constant phase, $\phi=kz +M\varphi
510: -\omega t$, that are oriented obliquely to the "wind" of the basic azimuthal
511: CCF. To that end we compare in Fig.~\ref{FIG:om-model} the onset spiral
512: frequency $\omega(R_{1,stab})$ at the stability threshold $R_{1,stab}(R_2)$
513: of CCF with the "model" frequency
514: $\omega^{\mbox{model}}(R_{1,stab})$ which is
515: evaluated also at the stability threshold $R_{1,stab}(R_2)$. Here
516: \begin{eqnarray} \label{EQ:om-model}
517: \omega^{\mbox{model}}= \langle \omega_{CCF}(r) \rangle =
518: \frac{2}{r_0^2-r_1^2} \int\limits_{r_1}^{r_0}\omega_{CCF}(r)\,r\,dr
519: \end{eqnarray}
520: is the mean of the rotation rate of the CCF, $\omega_{CCF}=v_{CCF}/r$. For
521: $R_2 < 0$ the averaging is done over
522: the radial domain between inner cylinder, $r_1$, and the first zero, $r_0$,
523: of $v_{CCF}(r)$ (\ref{v_CCF}). Thus, at the stability threshold
524: $R_{1,stab}(R_2)$ one has
525: \begin{eqnarray}
526: r_0^2= \frac{2R_2 - 4R_{1,stab}}{2R_2 - R_{1,stab}}
527: \end{eqnarray}
528: when $R_2 < 0$. However, when $R_2 \geq 0$, i.e., when $v_{CCF}$ remains
529: positive throughout the gap $r_0$ is replaced by $r_2$.
530: The restriction of the radial average to the range between $r_1$ and $r_0$ is
531: motivated by an argument of largely hand-waving nature:
532: the linear eigenfunctions for marginally stable SPI modes are
533: somewhat centered to this range where the growth of vortex perturbations is
534: supported.
535:
536: Fig.~\ref{FIG:om-model} shows that the onset spiral
537: frequency $\omega(R_{1,stab})$ agrees perfectly well with
538: the mean CCF rotation frequency (\ref{EQ:om-model}) when $R_2 > 0$. For
539: $R_2 < 0$ the model ansatz (\ref{EQ:om-model}) for the global spiral rotation
540: rate overestimates slightly the spiral
541: frequency since Eq.(\ref{EQ:om-model}) does not contain contributions from
542: negative CCF rotation rates between $r_0$ and $r_2$. In fact, if one extends
543: in an ad hoc way the averaging domain slightly beyond $r_0$ then the agreement
544: improves significantly. Thus, the onset spiral frequency $\omega(R_{1,stab})$,
545: i.e., the frequency eigenvalue
546: can be seen as the mean rotation rate of the CCF -- albeit weighted
547: appropriately by the critical eigenfunctions.
548:
549: \subsection{Flow structure of TVF and SPI}
550:
551: In this section we elucidate the flow structure of spiral vortices in comparison
552: with Taylor vortices. To that end we consider the radial velocity field.
553: In Fig.~\ref{FIG:u-profiles} we show the axial profiles of
554: $u(z)$ at mid gap position for $R_1 = 130$ being fixed and various $R_2$ that cover
555: the whole interval between the bifurcation thresholds, c.f. Fig.~\ref{FIG:BIF-u}
556: and the inset of Fig.~\ref{FIG:u-profiles}. Full (dashed) lines refer to negative
557: (positive) $R_2$. In each case the axial position of maximal radial outflow is
558: chosen to lie at $z=0.5 \lambda$. For the sake of better visibility two axial
559: periods of the vortex profiles are shown.
560:
561:
562: \subsubsection{Anharmonicity: TVF versus SPI}
563:
564: Typically SPI are less anharmonic than TVF. Also the profiles of both are
565: less anharmonic for positive $R_2$ than for negative $R_2$ and the degree of
566: anharmonicity increases when $R_2$ becomes more negative. For the mirror
567: symmetric TVF this anharmonicity growth comes from a widening (narrowing) of
568: the axial range $\Delta_{in}$ ($\Delta_{out}$) of radial inflow over which
569: $u<0$ ($u>0$) and the corresponding decrease (increase) of the inflow
570: (outflow) velocity. For the L-SPI that propagate in
571: Fig.~\ref{FIG:u-profiles} into positive $z$-direction the anharmonicity
572: grows mainly by flattening (steepening) of the wave profiles ahead of
573: (behind) the crests. However, $\Delta_{in} /
574: \Delta_{out}$ increases also for SPI albeit less than for TVF.
575:
576: The variation of the anharmonicity of the vortex profiles can be read off more
577: quantitatively from the results of an axial Fourier analysis. To that end we show in
578: Fig.~\ref{FIG:anharm} the ratios
579: $|u_n/u_1|$ of the n-th and first axial Fourier modes of the profiles of
580: Fig.~\ref{FIG:u-profiles} as a function of $R_2$ for fixed $R_1$. With growing
581: distances from the bifurcation thresholds at positive and negative $R_2$ the
582: anharmonicity grows for TVF as well as for SPI. It does so most precipitously
583: near the thresholds at negative $R_2$ of about -150 in Fig.~\ref{FIG:anharm}.
584:
585: At negative $R_2$ the anharmonicity of TVF can be for rapidly counter rotating
586: cylinders already close to threshold so large that $|u_2/u_1| > 1$. This
587: property reflects the fact that for sufficiently negative $R_2$ Taylor vortices
588: are effectively smaller in size than the gap width. There
589: are two main reasons for this size reduction which are both connected to the
590: tendency of vortices to have circular shape: ({\it i}) the axial periodicity length
591: $\lambda=1.6$ reduces the {\it axial} vortex size relative to the gap and, more
592: importantly,
593: ({\it ii}) the TVF intensity is {\it radially} restricted not to extend significantly
594: beyond the zero of CCF at $r_0$ since according to the Rayleigh criterion
595: $m=0$ radial momentum transport is suppressed by opposite pressure gradients for
596: $r > r_0$ where the CCF stratification of the squared angular momentum density
597: is stable. With $R_2$ becoming more negative $r_0$ moves inwards and the radial
598: size of Taylor vortices reduce.
599:
600: However, the $m=0$ Rayleigh criterion does not apply to SPI. Their $m \neq0$ radial
601: momentum transport extends further beyond $r_0$. Therefore SPI vortices fill
602: out the whole gap more than Taylor vortices, c.f. Fig.~\ref{FIG:arrows_TVF_SPI}, and
603: consequently they are less anharmonic.
604:
605: \subsubsection{Mirror symmetry breaking of SPI}
606:
607: TVF shows axial mirror symmetry around the position of maximal
608: radial outflow, $z=0.5 \lambda$, in Fig.~\ref{FIG:u-profiles}. In order
609: to measure the degree to which this symmetry is broken in SPI we have used the
610: asymmetry parameter
611: \begin{eqnarray} \label{EQ:P}
612: P=\frac{\int |u(z')-u(-z')|\,dz'}{\int |u(z')+u(-z')|\,dz'}
613: \end{eqnarray}
614: evaluated at mid gap with $z'=0$ locating the largest radial SPI
615: outflow at this $r$-value. In this way we found, e.g., for the spirals of
616: Fig.~\ref{FIG:u-profiles} that the smallest $P\simeq 0.2$ occurs
617: for spirals with the smallest frequency $\omega_{min} \simeq 23.4$ at
618: $R_2 \simeq -74$. Increasing $R_2$ from this value all the way toward
619: the upper SPI bifurcation threshold at $R_2 \simeq 48$ the frequency increases
620: but
621: $P$ remains roughly unchanged at about 0.2. On the other hand, when decreasing
622: $R_2$ from -74 the asymmetry parameter increases with increasing $\omega$
623: up to $P\simeq 1$ close to the lower SPI bifurcation threshold $R_2 \simeq -158$.
624: Thus, fast propagating spirals at large negative $R_2$ show the largest
625: mirror symmetry breaking.
626:
627: \subsection{Comparison with experimental results}
628: \label{SEC:CER}
629:
630: In order to check our numerical results we made a few comparisons with
631: experiments. For example, in Fig.~\ref{FIG:exp_num_u} we show the axial profile of the
632: radial flow $u(z)$
633: of an L-SPI at $r_1 + 0.4$. Symbols denote Laser-Doppler velocimetry
634: measurements \cite{SP99} and the full line a numerical simulation, both done in a
635: setup of height $\Gamma=12$ with rigid, non-rotating lids at both
636: ends of the annulus. In each case the spirals were monitored at mid-height of
637: the cylinders where they had the common wavelength $\lambda \simeq 1.76$.
638: Since absolute experimental velocities were not available
639: we have scaled the experimental maximum in Fig.~\ref{FIG:exp_num_u} to that of our
640: simulation (full line). Without knowledge of the experimental error-bars we
641: consider the agreement between symbols and full line to be satisfactory.
642:
643: The dashed line shows a numerical profile obtained for axially periodic
644: boundary conditions
645: imposing the wavelength $\lambda = 1.6$. It differs slightly from the SPI
646: profile (full line) in the bulk part of the $\Gamma=12$ system with rigid ends.
647: The difference is presumably related to the fact that the axial flow, and in
648: particular the mean-flow $w_0$ (\ref{Defw_0}), is different in these two cases
649: as discussed in Sec.~\ref{SEC:axial-veloc}.
650:
651: In Fig.~\ref{FIG:exp_num_om} we compare the frequency variation of experimental and
652: numerical L-SPI with $R_1$. Symbols and the full line come from Laser-Doppler
653: velocimetry measurements \cite{SP99} and numerical simulations, respectively,
654: of the aforementioned Taylor-Couette setup ($\eta$=0.5) of height $\Gamma=12$
655: with rigid, non-rotating lids at both ends. Note that not only the frequency values
656: of these experimental and numerical SPI states agree reasonably well with each
657: other but also
658: their existence range in $R_1$. Its lower end marks the oscillatory onset.
659: At the upper end in $R_1$ these SPI lose their stability to TVF --- in
660: experiments as well as in the simulations.
661:
662: However, under axially periodic boundary conditions the existence range of
663: stable SPI extends to significantly larger values of $R_1$ lying outside of the
664: plot range of Fig.~\ref{FIG:exp_num_om}.
665: The dashed line in Fig.~\ref{FIG:exp_num_om} refers to simulations done with axially
666: periodic conditions ($\lambda = 1.6$) that allow for a free propagation of
667: phase. And, in addition, they allow the Reynolds-stress-sustained mean
668: axial flow $w_0$ (\ref{Defw_0}) to have a finite {\em net} part
669: $<w>$ (\ref{Def_w_net}) that is negative for our parameters -- c.f.
670: Sec.~\ref{SEC:axial-veloc}. In order to compare with the SPI frequencies for
671: rigid end conditions we subtract from the oscillation frequencies under
672: periodic boundary conditions (dashed line) the pure Galilean contribution
673: $\langle w \rangle k$ and obtain the dash-dotted line. Note how close the latter
674: lies to the SPI frequencies in the system with rigid end conditions. Thus, we
675: find that the SPI frequency differences \cite{RL93} for the two different end
676: boundary conditions are mostly due to whether the Galilean contribution
677: $\langle w \rangle k$ is suppressed or not.
678:
679: \section{External Through-flow}
680: \label{SEC:ETF}
681: Here we discuss the influence of an externally imposed axial through-flow on
682: spiral and on Taylor vortices. Since the effect of an
683: axial through-flow on TVF has been investigated for $R_2=0$ in several works,
684: we focus our investigation on SPI vortices.
685:
686: The through-flow is enforced by adding in the NSE for the axial velocity
687: component a constant
688: pressure gradient of size $\partial_z p_{APF}$ throughout the annulus. In the
689: absence of any vortex flow, i.e., for sub-critical control parameters
690: this pressure gradient, $\partial_z p_{APF}$, drives an annular
691: Poiseuille flow (APF) with a radial profile of the axial through-flow velocity
692: given by
693: \begin{equation}\label{wAPF}
694: w_{APF}(r)= \frac{\partial_z p_{APF}}{4}
695: \left[r^2 +
696: \frac{1+\eta}{(1-\eta)\ln\eta}\ln r +
697: \frac{(1+\eta)\ln(1-\eta)}{(1-\eta)\ln\eta} -
698: \frac{1}{(1-\eta)^2} \right]
699: \end{equation}
700: We checked that our numerical code reproduces this analytical solution
701: (\ref{wAPF}) of the NSE. We use its mean to define the through-flow Reynolds number
702: by
703: \begin{equation} \label{DefRe}
704: \left<w_{APF}(r)\right> = Re =
705: - \frac{\partial_z p_{APF}}{8}
706: \frac{1 - \eta^2 + (1+\eta^2)\ln\eta}{(1-\eta)^2\ln\eta}.
707: \end{equation}
708: Hence positive (negative) $Re$ implies an axial flow, $w_{APF}(r)$, in positive
709: (negative) $z$-direction. The last equality in Eq.~(\ref{DefRe}) establishes the
710: relation between the externally applied additional axial pressure gradient and the
711: through-flow Reynolds number $Re$.
712:
713: \subsection{Counter-rotating cylinders}
714: Fig.~\ref{FIG:SPIvsRe} shows how the through-flow influences L-SPI, R-SPI, and TVF
715: at the fixed characteristic driving combination $R_1=120, R_2=-100$ that is
716: located in Figs.~\ref{FIG:BIF-u} and ~\ref{FIG:BIF-om} in the region C close to the
717: border to region E. For this parameter combination TVF is unstable when $Re=0$
718: and it remains unstable in the $Re$-range shown in Fig.~\ref{FIG:SPIvsRe}. This is
719: of relevance for the through-flow induced transitions between L-SPI and R-SPI (c.f.
720: further below).
721:
722: \subsubsection{Bifurcation behavior}
723: \label{SEC:bif_R_2=-100}
724:
725: We present in Fig.~\ref{FIG:SPIvsRe}(a) the
726: primary Fourier amplitudes, $|u_{m,n}|$, of the radial flow intensity at mid gap
727: versus $Re$. These are $|u_{1,1}|$ for the $M=1$ L-SPI, $|u_{-1,1}|$ for the
728: $M=-1$ R-SPI, and $|u_{0,1}|$ for TVF. Fig.~\ref{FIG:SPIvsRe}(b) shows their axial
729: phase velocity, $w_{ph}=\omega /k$, and Fig.~\ref{FIG:SPIvsRe}(c) shows the
730: {\em net} mean axial flow
731: \begin{equation} \label{Def_w_net}
732: \langle w \rangle
733: = \frac{1}{\pi (r_2^2-r_1^2)}\int\limits_0^{2\pi}\int\limits_{r_1}^{r_2}
734: w(r,\varphi,z,t)rdr\,d\varphi \, .
735: \end{equation}
736:
737: For $Re=0$ the two spirals are mirror images of each other: their radial
738: velocities are the same and all respective axial velocities have the same
739: magnitude but opposite direction. Note that the SPI Reynolds stresses drive
740: an axial flow to be discussed further below. Its net mean, $\langle w
741: \rangle$ (\ref{Def_w_net}), is directed opposite to the phase velocity,
742: $w_{ph}$, of the respective spiral when $Re=0$. Note, however, the
743: difference in size between $|w_{ph}| \simeq 7.1$ and
744: $|\langle w \rangle| \simeq 1.1$ \cite{error}.
745:
746: A finite through-flow breaks the mirror symmetry between the $M=1$ L-SPI and
747: the $M= -1$ R-SPI. Their radial flow amplitudes evolve with through-flow
748: as shown in Fig.~\ref{FIG:SPIvsRe}(a). We performed also a
749: linear stability analysis of the combined CCF-APF state. It shows that for our
750: control parameters $R_1=120, R_2=-100$ the amplitudes of the $M=\pm 1$ SPI
751: solutions go to zero at the bifurcation threshold values of $Re=\mp 19.07$
752: and $Re=\pm 50.95$. The numerical solutions of the full nonlinear NSE showed
753: in addition that the L-SPI (R-SPI) is unstable near the first threshold,
754: $Re\simeq-19$ ($Re\simeq 19$), and that it is
755: stable near the second one, $Re \simeq 50$ ($Re \simeq -50$).
756:
757: For small through-flow -- say, for $-6 \lesssim Re \lesssim 6$ in
758: Fig.~\ref{FIG:SPIvsRe} -- the two spiral solutions coexist bistably; their particular
759: realization depending on initial conditions. However, with increasing $|Re|$ that
760: spiral suffers a through-flow enforced loss of stability for which
761: the phase velocity changes sign. This happens roughly
762: when the through-flow has become sufficiently strong to revert an originally
763: adverse axial phase propagation. For example, the $M=-1$ R-SPI of
764: Fig.~\ref{FIG:SPIvsRe} propagate for small $Re \lesssim 6.6$ axially
765: downwards (i.e. opposite to the externally imposed through-flow) as for $Re=0$,
766: then become stationary, and finally propagate upwards in through-flow direction for
767: $Re \gtrsim 6.6$.
768: Similarly, by symmetry, the $M=1$ L-SPI propagates in a small negative through-flow
769: upwards against the through-flow for $Re \gtrsim -6.6$ and downwards, i.e.,
770: in through-flow direction for $Re \lesssim -6.6$.
771:
772: The direction of the imposed through-flow is the preferred one for stable phase
773: propagation: A spiral that has started at small $|Re|$ to move against the wind
774: dies out -- or, more precisely, becomes unstable -- when the wind becomes
775: sufficiently strong to turn it back. Only that SPI is stable at large
776: $|Re| \gtrsim 7.2$ in Fig.~\ref{FIG:SPIvsRe} that keeps propagating
777: into the preferred direction of the through-flow. The other one is unstable at
778: large $|Re|$.
779:
780: The through-flow enforced loss of stability of one SPI state
781: and the transition to the remaining stable one is indicated schematically
782: in Fig.~\ref{FIG:SPIvsRe}(a) by vertical arrows. However, we should like to
783: stress that
784: the transition is somewhat complex extending over the through-flow interval
785: $6\lesssim |Re|\lesssim 7.2$ the center of which locates the zero of $w_{ph}$
786: at $|Re| \simeq 6.6$.
787: In this interval there are stable, mixed states with
788: finite L- and R-SPI modes. Their amplitudes seem to vary largely
789: continuously with $Re$ (with possibly some saddle-node discontinuity)
790: between the pure SPI solutions: the amplitude of the spiral that loses the
791: stability competition decreases with growing $|Re|$ towards zero while the amplitude
792: of the winning one increases from zero to the pure monostable final SPI state.
793:
794: Note that since TVF is unstable for the parameters of Fig.~\ref{FIG:SPIvsRe}
795: it does not offer an alternative transition to a final $M=0$ state as for
796: the parameters of Sec.~\ref{SEC:inner_cylinder_at_rest}. There, for $R_2=0$,
797: the through-flow induces a transition to stable TVF rather than to the
798: stably coexisting SPI with preferred propagation direction. Only when TVF
799: is eliminated there does the transition occur to the then monostable spiral ---
800: for details see Sec.~\ref{SEC:inner_cylinder_at_rest}.
801:
802: We also made a few calculations in a regime
803: where TVF stably coexists with SPI for counter-rotating cylinders. Also then
804: the through-flow induces preferably
805: a transition to stable TVF rather than to the stable SPI state. Thus, when the
806: through-flow destabilizes, e. g., the $M=-1$ R-SPI then typically the $M=0$ TVF
807: modes grow rather than the $M=1$ L-SPI modes.
808:
809: \subsubsection{Axial velocities $w_{ph}, w_0$, and $\langle w \rangle$}
810: \label{SEC:axial-veloc}
811:
812: In the through-flow range shown in Fig.~\ref{FIG:SPIvsRe} the phase velocity
813: $w_{ph}$ and the net mean flow $\langle w \rangle$ vary roughly linearly with
814: $Re$. The slopes $\partial w_{ph}/\partial Re$ and
815: $\partial \langle w \rangle /\partial Re$ for SPI as well as for TVF are
816: roughly 1.
817:
818: While the phase of the $M=\pm 1$ SPI reverts its propagation direction at $Re
819: \simeq \mp 6.6$ the net mean flow changes sign already at
820: $Re \simeq \pm 1.2$. The reversal of the latter does not seem to have any
821: consequence. But the through-flow enforced reversal of the phase velocity
822: seems to be responsible for the
823: destabilization of the SPI that propagate at small $|Re|$ against the wind, i.e.,
824: into the "wrong" direction.
825:
826: In Fig.~\ref{FIG:w_0} we show how the radial profiles of the mean axial flow
827: \begin{equation} \label{Defw_0}
828: w_0(r) = \frac{1}{2\pi}\int\limits_0^{2\pi}w(r,\varphi,z,t)\,d\varphi
829: \end{equation}
830: of spirals shown in Fig.~\ref{FIG:SPIvsRe} evolve with the through-flow in the
831: range $-4 \leq Re \leq 14$. We checked that $w_0$ is independent of $z$ and $t$
832: and that our spirals propagating in the externally imposed axial
833: pressure gradient still have the SPI symmetry, i.e., the flow fields depend on
834: $z, \varphi, t$ only via the phase combination $\phi$ (\ref{phase}) with an
835: oscillation frequency $\omega$ that is modified by the through-flow. Then one
836: finds from the NSE for the $m=0$ azimuthal mode of the axial velocity field,
837: \begin{equation} \label{reystr}
838: \left(\partial_r + \frac{1}{r}\right) \partial_r w_0
839: =\left(\partial_r + \frac{1}{r}\right) \left( u w \right)_0
840: +\partial_z p_0 \, ,
841: \end{equation}
842: that the SPI mean flow can be driven by Reynolds stresses and/or by mean axial
843: pressure
844: gradients. For $Re=0$ the pressure is enforced to be axially periodic, hence
845: $\partial_z p_0(Re=0) = 0$. So in that case the mean axial flow is driven
846: solely by the nonlinear Reynolds stresses. They are rather large. For example
847: for the R-SPI propagating at
848: $Re=0$ in negative $z$-direction with phase velocity $w_{ph} \simeq -7.1$ the maximum
849: of $w_0(r)$ is about 3, i.e, directed opposite to the phase propagation and
850: almost half as large in magnitude as $w_{ph}$. The {\em net} mean flow
851: $\langle w \rangle$ (\ref{Def_w_net}) is for this case still about 1.1
852: and also opposite to $w_{ph}$.
853:
854: As an aside we mention that rigid axial end
855: conditions enforce $\langle w \rangle=0$ throughout the annulus. They generate
856: an adverse axial pressure gradient that compensates the Reynolds stresses
857: \cite{ETDS91} so that $w_0$ is practically zero in the bulk part where SPI are
858: realized. Only in the Ekman region $w_0$ becomes finite showing TVF behavior
859: there.
860:
861: For the R-SPI of Fig.~\ref{FIG:w_0} propagating at $Re>0$ opposite to the
862: external through-flow the maximal mean flow is located roughly at mid-gap.
863: However, for the SPI propagating into the direction of the external
864: through-flow, i.e., the R-SPI for $Re<0$ and the L-SPI for $Re>0$ the extremum
865: of $w_0(r)$ is shifted towards the inner cylinder.
866: The mean flow profiles of the spirals of Fig.~\ref{FIG:w_0} are given within
867: about 5\% by the superposition
868: \begin{equation}
869: \label{EQ:composition_w_0}
870: w_0(r;Re) \simeq w_0 (r; Re=0) + w_{APF} (r; Re)
871: \end{equation}
872: of the pure, Reynolds stress generated flow $w_0(Re=0)$ of the respective SPI
873: plus the pure, pressure gradient enforced APF flow $w_{APF}(Re)$
874: (\ref{wAPF}). This holds for L-SPI as well as for R-SPI, irrespective of whether
875: they propagate into the direction of the through-flow or against it.
876:
877: \subsubsection{Spiral profiles}
878: The through-flow changes the structure of the SPI. This is documented in
879: Figs.~\ref{FIG:arrows_SPI_Re} and \ref{FIG:u_L-SPI_Re}. The arrows in
880: Fig.~\ref{FIG:arrows_SPI_Re} representing the $u,w$ vector field of L-SPI in
881: the $r-z$ plane show the effect of imposing an axial through-flow that
882: increases from $Re=-5$(a) to $Re=10$(d) in steps of 5. Note, however, that the
883: externally imposed axial pressure gradient does not just add $w_{APF}(r)$ to
884: the axial velocity field $w$. It also modifies all vector field components
885: of the SPI. The axial profile of the radial flow $u(z)$ for example is changed
886: by the through-flow as shown in Fig.~\ref{FIG:u_L-SPI_Re} for increasing $Re$.
887: Here the axial asymmetry of the upwards propagating L-SPI is reduced by
888: steepening up the leading part of $u(z)$ ahead of the wave crests. This
889: reduction of the mirror-asymmetry of the radial flow of L-SPI grows
890: somewhat linearly with increasing $Re$. As an aside we mention that on the
891: other hand the TVF profiles of
892: $u(z)$ become with increasing $Re$ more and more asymmetric --- the mirror
893: asymmetry parameter $P$ (\ref{EQ:P}) increases for TVF linearly with $Re$.
894:
895: \subsection{Non-rotating outer cylinder}
896: \label{SEC:inner_cylinder_at_rest}
897: We have investigated the influence of an externally imposed axial through-flow on
898: TVF and SPI also for stationary outer cylinder, $R_2=0$.
899:
900: \subsubsection{Bifurcation behavior}
901:
902: In Fig.~\ref{FIG:bif_SPI_TVF_R2=0} we show the bifurcation behavior of TVF
903: and SPI as a function of through-flow Reynolds number $Re$ for $R_2=0$,
904: $R_1=100$. This parameter combination lies well within the region E of
905: Fig.~\ref{FIG:BIF-u} in which TVF, L-SPI, and R-SPI are all stable at
906: $Re=0$.
907:
908: Switching on the through-flow one sees in Fig.~\ref{FIG:bif_SPI_TVF_R2=0}(a)
909: how the dominant modes of these vortex structures vary with $Re$. That SPI
910: loses its stability for which the through-flow enforces a reversal of the
911: phase propagation as in the case of counter rotating cylinders
912: (Fig.~\ref{FIG:SPIvsRe}). Thus, also here the direction of the imposed
913: through-flow is the preferred one for stable SPI at large $|Re|$. A spiral
914: that has started at small $|Re|$ to move against the through-flow becomes
915: unstable when the latter becomes sufficiently strong to turn it back. On the
916: other hand that SPI remains stable at large $|Re|$ that keeps propagating
917: into the preferred direction of the through-flow.
918:
919: As in Fig.~\ref{FIG:SPIvsRe} the loss of stability takes place in the vicinity
920: of the Reynolds number where the
921: axial phase velocity $w_{ph}$ [Fig.~\ref{FIG:bif_SPI_TVF_R2=0}(b)] of the
922: respective SPI goes through zero. This happens in
923: Fig.~\ref{FIG:bif_SPI_TVF_R2=0} for the $M=\pm 1$ SPI at $Re \simeq \mp
924: 6.4$. However here we found the
925: transition from the then unstable SPI to occur to the stable TVF solution
926: [c.f. arrows in Fig.~\ref{FIG:bif_SPI_TVF_R2=0}(a)] rather than to the other
927: stable SPI.
928:
929: We have also investigated briefly the situation where the TVF solution was
930: numerically eliminated (here, suppressing $m=0$ modes of the
931: $u$-field at mid-gap position turned out to be an efficient way to globally
932: reduce TVF towards zero).
933: Also then the SPI that is unfavored by the through-flow loses its
934: stability. However, with TVF being unavailable as final state the
935: transition occurs in this case to the favored SPI in a way that seems to be
936: similar to the one described in Sec.~\ref{SEC:bif_R_2=-100}.
937:
938: Without the above described numerically imposed mode restriction TVF is stable
939: for moderate through-flow rates while at sufficiently large $|Re|$ SPI
940: are stable \cite{BP90,LDM92,TS94b}. For our parameters TVF decays at
941: $Re \simeq \pm34$ into a $M=\pm 1$ SPI as indicated by arrows in
942: Fig.~\ref{FIG:bif_SPI_TVF_R2=0}(a).
943:
944: For small through-flow the phase velocity
945: $w_{ph}$ and the net mean flow $\langle w \rangle$ vary roughly linearly with
946: $Re$. The initial slopes $\partial w_{ph}/\partial Re$ and
947: $\partial \langle w \rangle /\partial Re$ are for SPI as well as for TVF
948: roughly 1. However, at larger $Re$ one sees in
949: Fig.~\ref{FIG:bif_SPI_TVF_R2=0}(c) that in particular $\langle w \rangle$ shows
950: nonlinear corrections.
951:
952: \subsubsection{Phase diagram}
953:
954: Fig.~\ref{FIG:phasediagram} shows the phase diagram of TVF, R-SPI, and L-SPI
955: for stationary outer cylinder in
956: the control parameter plane spanned by $Re$ and $R_1$. The {\em existence} range
957: of the vortex states is bounded from below by the
958: bifurcation threshold (full line in Fig.~\ref{FIG:phasediagram}) of the
959: respective vortex solution out of the combined CCF-APF basic state. These
960: bifurcation thresholds result from a linear stability analysis of the CCF-APF
961: state \cite{PLH03}. The one for TVF increases quadratically for small $Re$.
962: Also the SPI threshold curves in Fig.~\ref{FIG:phasediagram} have a somewhat
963: parabolic shape, however with minima shifted to finite $Re$. Thus, the threshold
964: for L-SPI first decreases for small positive $Re$ but eventually increases at
965: larger $Re$. By symmetry the R-SPI threshold curve in
966: Fig.~\ref{FIG:phasediagram} is the mirror image under $Re \to -Re$ of the L-SPI
967: threshold curve. Hence small through-flow destabilizes (stabilizes) the CCF-APF
968: state against spirals that propagate into (against) the through-flow direction.
969:
970: Note that for small $Re$ in Fig.~\ref{FIG:phasediagram} TVF bifurcates first
971: when increasing $R_1$. But for sufficiently large $Re$ the bifurcation sequence
972: of TVF and SPI is reversed since the bifurcation threshold for TVF curves up
973: faster with increasing $Re$ than the one for L-SPI. After their intersection
974: stable SPI bifurcate first out of the CCF-APF state. Hence, for example in
975: region E of Fig.~\ref{FIG:phasediagram} only stable L-SPI exist, in region D
976: TVF exists but only as unstable solution, and in region B they exist bistably.
977:
978: The dashed lines in Fig.~\ref{FIG:phasediagram} are stability boundaries of the
979: vortex solutions. Different regions of Fig.~\ref{FIG:phasediagram} between
980: various stability boundaries and bifurcation thresholds are identified with
981: the respective stability properties of the vortex states in the caption of
982: Fig.~\ref{FIG:phasediagram}.
983:
984: \section{Summary}
985: We have numerically simulated vortex flow structures of different azimuthal wave
986: numbers $M$ in the Taylor-Couette system with
987: counter-rotating as well as with co-rotating cylinders. In particular we have
988: investigated the effect of an externally imposed axial through-flow
989: on the spatio-temporal properties and on the bifurcation behavior of
990: $M=1$ L-spirals, $M=-1$ R-spirals, and $M=0$ Taylor vortices.
991:
992: To that end we first have determined for zero
993: through-flow, $Re=0$, the bifurcation surfaces of the appropriate order
994: parameters characterizing SPI and TVF solutions over the $R_1 - R_2$ control
995: parameter plane of the inner and outer cylinder's Reynolds numbers.
996: For the parameter combinations explored in this work these
997: bifurcations out of the basic CCF state are forward and their order of
998: appearance determines the stability of the respective bifurcating vortex state:
999: the vortex solution that bifurcates second
1000: is unstable. But it eventually becomes stable with
1001: increasing distance from the bifurcation threshold so that, e.g., for larger
1002: $R_1$ there is a large region in the $R_1 - R_2$ plane with bistability of TVF
1003: and SPI. In particular the existence region of stable SPI extends for axially
1004: periodic boundary conditions even to positive $R_2$ with co-rotating cylinders.
1005: Unstable solution branches were obtained by selectively suppressing
1006: destabilizing modes. Stable ribbons, i.e., nonlinear combinations of $M=\pm 1$
1007: spirals were not found.
1008:
1009: Simulations of axially finite systems with rigid,
1010: non-rotating lids showed in good agreement with experiments how the
1011: stable existence range of SPI is reduced by stationary Ekman vortices
1012: which suppress phase propagation at the two ends. Also the frequencies and the
1013: wave profiles of the spiral vortices in the
1014: bulk of the numerical and experimental systems agreed well with each other.
1015: Spiral profiles obtained for periodic and rigid end conditions do not differ
1016: much. On the other hand, the respective frequencies differ basically by the
1017: Galilean contribution $\langle w \rangle k$. Here $\langle w \rangle $ is the net
1018: axial mean flow that the nonlinear Reynolds stresses of a spiral with axial wave
1019: number $k$ sustains with
1020: axially periodic end conditions but not with impermeable ends.
1021:
1022: Furthermore, we showed how the phenomenon of rigid body rotation of spirals can
1023: be understood quantitatively in terms of the passive advection dynamics of
1024: $M = \pm 1$ vortex perturbations whose lines of constant phase are oriented
1025: obliquely to the azimuthal CCF. The onset spiral frequency is
1026: the mean rotation rate of the CCF, albeit weighted appropriately by the
1027: critical eigenfunctions with the consequence that L-SPI as well as R-SPI rotate
1028: into the same direction as the inner cylinder. The nonlinear SPI frequencies are
1029: typically smaller than the linear ones but do not deviate substantially from them.
1030:
1031: A finite through-flow breaks the mirror symmetry between the L-SPI and the R-SPI
1032: and it changes the structure of the SPI. The externally imposed axial
1033: pressure gradient does not just add the annular Poiseuille flow $w_{APF}(r)$ to
1034: the axial velocity field. It modifies the SPI structure, e.g., the profiles of
1035: the radial flow in a characteristic way.
1036:
1037: For $Re=0$ L-SPI propagate axially upwards and R-SPI downwards.
1038: When they are initially stable they continue to coexist bistably
1039: for small through-flow. However, they are no longer mirror images of each other
1040: and their phase velocities differ by an amount $\propto Re$. Then, with
1041: increasing $|Re|$ that spiral loses its stability for which
1042: the through-flow enforces the phase velocity to change direction. Only that SPI
1043: is stable at large $|Re|$ that keeps propagating
1044: into the preferred direction of the through-flow. The other one is unstable at
1045: large $|Re|$.
1046:
1047: The SPI that loses stability upon reverting its propagation direction --- i.e.
1048: the R-SPI (L-SPI) for positive (negative) $Re$ --- preferentially undergoes a
1049: transition to propagating TVF provided the latter is available as {\em stable}
1050: vortex state. Otherwise the transition is to the then monostable L-SPI (R-SPI).
1051: Such a situation was explored in detail for negative $R_2$ where TVF was
1052: unstable and for other parameter combinations where the TVF solution was
1053: eliminated numerically.
1054:
1055: Also the situation where initially at $Re=0$ all three vortex solutions are
1056: stable was elucidated for different $R_1 -R_2$ parameter combinations and in
1057: more detail for stationary outer cylinder, $R_2=0$. Here, a complete phase
1058: diagram was determined in the control parameter plane spanned by $Re$ and $R_1$.
1059: We found that small through-flow destabilizes (stabilizes) the basic CCF-APF
1060: state against spirals that propagate into (against) the through-flow direction.
1061: For sufficiently large $Re$ the bifurcation sequence
1062: of TVF and SPI is reversed since the bifurcation threshold for TVF curves up
1063: faster with increasing $Re$ than the one for L-SPI. After their intersection
1064: stable SPI bifurcate first out of the CCF-APF state. Then there opens up a
1065: region at sufficiently large positive $Re$ in which only stable L-SPI but no
1066: Taylor vortices exist for stationary outer cylinder.
1067:
1068:
1069: \section*{Acknowledgments}
1070: We thank A. Schulz for communicating the
1071: experimental data referred to in this paper.
1072:
1073: \begin{thebibliography}{99}
1074: \addcontentsline{toc}{section}{References}
1075: \bibitem{T94}
1076: {\it The Couette-Taylor Problem},
1077: Nonlinear Science Today {\bf 4}, 1 (1994).
1078: \bibitem{CI94}
1079: P.~Chossat and G.~Iooss,
1080: {\it The Couette-Taylor Problem},
1081: (Springer, Berlin, 1994).
1082: \bibitem{KGD66}
1083: E.~R.~Krueger, A.~Gross, and R.~C.~DiPrima,
1084: J.~Fluid~Mech. {\bf 24}, 521 (1966).
1085: \bibitem{S68}
1086: H.~A.~Snyder,
1087: Phys.~Fluids {\bf 11}, 728 (1968).
1088: \bibitem{ALS85}
1089: C.~D.~Andereck, S.~S.~Liu, and H.~L.~Swinney,
1090: J.~Fluid~Mech. {\bf 164}, 155 (1986).
1091: \bibitem{LTKSG88}
1092: W.~F.~Langford, R.~Tagg, E.~Kostelich, H.~L.~Swinney, and M.~Golubitsky,
1093: Phys.~Fluids {\bf 31}, 776 (1988).
1094: \bibitem{TESM88}
1095: R.~Tagg, W.~S.~Edwards, H.~L.~Swinney, and P.~S.~Marcus,
1096: Phys.~Rev.~A {\bf 39}, 3734 (1989).
1097: \bibitem{E90}
1098: W.~S.~Edwards,
1099: in {\it Instability and Transition, Vol. II},
1100: edited by M.~Y.~Hussaini, (Springer, Berlin, 1990), p 408.
1101: \bibitem{SP99}
1102: A.~Schulz and G.~Pfister,
1103: in {\it Physics of Rotating Fluids},
1104: edited by C. Egbers and G. Pfister, Lecture Notes in Physics 549,
1105: (Springer, Berlin, 2000), p. 37; and unpublished.
1106: \bibitem{AMS98}
1107: J.~Antonijoan, F.~Marqu\`{e}s, and J.~S\'{a}nchez,
1108: Phys.~Fluids {\bf 10}, 829 (1998).
1109: \bibitem{BLRS96}
1110: P. B\"uchel, M. L\"ucke, D. Roth, and R. Schmitz,
1111: Phys.~Rev.~E {\bf 53}, 4764 (1996).
1112: \bibitem{GG93}
1113: Th.~Gebhardt and S.~Grossmann,
1114: Z. Phys. {\bf B 90}, 475 (1993).
1115: \bibitem{MM02}
1116: A.~Meseguer and F.~Marques,
1117: J.~Fluid~Mech. {\bf 455}, 129 (2002).
1118: \bibitem{PLH03}
1119: A.~Pinter, M.~L\"ucke, and Ch.~Hoffmann,
1120: Phys.~Rev.~E {\bf 67}, 026318 (2003).
1121: \bibitem{RL93}
1122: R.~Raffai and P.~Laure,
1123: Eur.~J.~Mech.~B/Fluids {\bf 12}, 277 (1993).
1124: \bibitem{noise-effects}
1125: J.~B.~Swift, K.~L.~ Babcock, and P.~C.~Hohenberg,
1126: Physica A {\bf 204}, 625 (1994);
1127: R.~J.~ Deissler,
1128: Phys.~ Rev.~E {\bf 49}, R31 (1994);
1129: M.~L\"ucke and A.~Szprynger,
1130: Phys.~Rev.~E {\bf 55}, 5509 (1997).
1131: \bibitem{noise-sensitivity}
1132: K.~L.~Babcock, G.~Ahlers, and D.~S.~Cannell,
1133: Phys.~Rev.~E {\bf 50}, 3670 (1994);
1134: A.~Szprynger and M.~L\"ucke,
1135: Phys.~Rev.~E {\bf 67}, 046301 (2003).
1136: \bibitem{WL99}
1137: S.T. Wereley and R. M. Lueptow,
1138: Phys.~Fluids {\bf 11}, 3637 (1999).
1139: \bibitem{HNR75}
1140: C.~W.~Hirt, B.~D.~Nichols, and N.~C.~Romero,
1141: {\it SOLA --- A Numerical Solution Algorithm for Transient Fluid Flow},
1142: (Los Alamos Scientific Laboratory of the University of California, LA-5852, 1975).
1143: \bibitem{PT83}
1144: R.~Peyret and T.~D.~Taylor,
1145: {\it Computational Methods in Fluid Flow},
1146: (Springer, Berlin, 1983).
1147: \bibitem{REC-LUE-MUE}
1148: A.~Recktenwald, M.~L\"ucke, and H.~W.~M\"uller,
1149: Phys.~Rev.~E {\bf 48}, 4444 (1993).
1150: \bibitem{CRO-HOH}
1151: M.~C.~Cross and P.~C.~Hohenberg,
1152: Rev.~Mod.~Phys. {\bf 65}, 851 (1993).
1153: \bibitem{error}
1154: In Fig. 5(a) of Ref. \cite{HL} the quantity labelled "Axial mean flow
1155: $-\langle w \rangle$" is too small by a factor of $2\pi$.
1156: \bibitem{HL}
1157: Ch.~Hoffmann and M.~L\"ucke,
1158: in {\it Physics of Rotating Fluids}, Lecture Notes in Physics 459,
1159: edited by C.~Egbers and G.~Pfister (Springer, Berlin, 2000), p. 55.
1160: \bibitem{ETDS91}
1161: W.~S.~Edwards, R.~P.~Tagg, B.~C.~Dornblaser, and H.~L.~Swinney,
1162: Eur.~J.~Mech.~B/Fluids {\bf 10}, 205 (1991).
1163: \bibitem{BP90}
1164: K.~B\"{u}hler and F.~Polifke,
1165: in {\it Nonlinear Evolution of Spatio-temporal Structures in Dissipative
1166: Continuous Systems},
1167: edited by F.~Busse and L.~Kramer,
1168: (Plenum Press, New York, 1990), p.21.
1169: \bibitem{LDM92}
1170: R.~M.~Lueptow, A.~Docter, and K.~Min,
1171: Phys.~Fluids A {\bf 4}, 2446 (1992).
1172: \bibitem{TS94b}
1173: A. Tsameret and V. Steinberg,
1174: Phys.~Rev.~E {\bf 49}, 4077 (1994).
1175:
1176: \end{thebibliography}
1177:
1178: \clearpage
1179:
1180:
1181: \begin{figure}[ht]
1182: \begin{center}
1183: \includegraphics[width=8.5cm]{fig01.eps}
1184: \end{center}
1185: \caption{Lines of constant phases, $\phi=const$, for spirals in the
1186: $\varphi - z$ plane. Arrows indicate their velocities.}
1187: \label{FIG:phases}
1188: \end{figure}
1189: \begin{figure}[ht]
1190: \begin{center}
1191: \includegraphics[width=14.5cm]{fig02.eps}
1192: \end{center}
1193: \caption{
1194: Order parameter bifurcation surfaces of TVF (thin lines) and of
1195: SPI (thick lines) over the $R_1 - R_2$ plane. Shown are primary
1196: Fourier amplitudes, $|u_{m,n}|$, of the radial flow intensity at mid gap,
1197: $r=r_1 + 0.5$, with axial mode index $n=\pm 1$. The azimuthal one is $m=0$ for
1198: TVF and $m =\pm 1$ for SPI, respectively. In each case full (dashed) lines
1199: denote stable (unstable) solutions.}
1200: \begin{tabular}{|c|c|c|c|c|c|}
1201: \hline
1202: region & A & B & C & D & E \\\hline\hline
1203: TVF & - & stable & unstable & stable & stable \\\hline
1204: SPI & stable & - & stable & unstable & stable \\\hline
1205: \end{tabular}
1206: \label{FIG:BIF-u}
1207: \end{figure}
1208: \begin{figure}[ht]
1209: \begin{center}
1210: \includegraphics[width=14.5cm]{fig03.eps}
1211: \end{center}
1212: \begin{quote}
1213: \caption{Bifurcation diagram of $M= \pm 1$ spiral frequencies $\omega$
1214: over the $R_1 - R_2$ plane. The thick line locates the minima.
1215: The different stability regions A -E of TVF and
1216: of SPI solutions (c.f. caption of Fig.~\ref{FIG:BIF-u}) in the $R_1 - R_2$ plane
1217: are included for better comparison with Fig.~\ref{FIG:BIF-u}. }
1218: \label{FIG:BIF-om}
1219: \end{quote}
1220: \end{figure}
1221: \begin{figure}[ht]
1222: \begin{center}
1223: \includegraphics[width=14.5cm]{fig04.eps}
1224: \end{center}
1225: \begin{quote}
1226: \caption{Linear frequency $\omega(R_{1,stab})$ of $M= 1$ spiral at onset,
1227: $R_{1,stab}(R_2)$, in comparison with frequency
1228: $\omega^{model}(R_{1,stab})$ (\ref{EQ:om-model}) resulting from rigid-body
1229: rotation model.}
1230: \label{FIG:om-model}
1231: \end{quote}
1232: \end{figure}
1233: \begin{figure}[ht]
1234: \begin{center}
1235: \includegraphics[width=12.5cm]{fig05.eps}
1236: \end{center}
1237: \begin{quote}
1238: \caption{
1239: Axial profiles of the radial velocity $u(z)$ at mid gap position for $R_1 = 130$ and
1240: various $R_2$ (along the thick horizontal line in the inset) covering the whole
1241: interval between the bifurcation thresholds marked TVF and SPI, respectively, in the
1242: inset; see also Fig.~\ref{FIG:BIF-u}. Full (dashed) lines refer to negative
1243: (positive) $R_2$. In each case the maximal radial outflow is
1244: chosen to lie at $z=0.5 \lambda$. For better visibility two axial periods of the
1245: vortex profiles are shown. The $M=1$ L-SPI are propagating in positive $z$-direction.
1246: Parameters are $\eta=0.5$, $k=3.927$.}
1247: \label{FIG:u-profiles}
1248: \end{quote}
1249: \end{figure}
1250: \begin{figure}[ht]
1251: \begin{center}
1252: \includegraphics[width=13.5cm]{fig06.eps}
1253: \end{center}
1254: \begin{quote}
1255: \caption{Anharmonicity of TVF and SPI. The ratios $|u_n/u_1|$ of the axial Fourier
1256: modes of the profiles of $u(z)$ shown in
1257: Fig.~\ref{FIG:u-profiles} are displayed here as functions of $R_2$ for fixed
1258: $R_1=130$. The bifurcation thresholds are located at the zeroes. Parameters are
1259: $\eta=0.5$, $k=3.927$.}
1260: \label{FIG:anharm}
1261: \end{quote}
1262: \end{figure}
1263: \begin{figure}[ht]
1264: \begin{center}
1265: \includegraphics[width=14.5cm]{fig07.eps}
1266: \end{center}
1267: \caption{Velocity field ($u,w$) of TVF (left) and L-SPI (right) in an
1268: $r-z$ plane.
1269: Vertical lines locate the zero of the azimuthal CCF flow $v_{CCF}(r)$.
1270: Parameters are $\eta=0.5$, $k=3.927$, $R_1 = 120$, $R_2 = -100$.}
1271: \label{FIG:arrows_TVF_SPI}
1272: \end{figure}
1273: \begin{figure}
1274: \begin{center}
1275: \includegraphics[width=14.5cm]{fig08.eps}
1276: \end{center}
1277: \begin{quote}
1278: \caption{
1279: Comparison of experimental and numerical axial profiles of the radial velocity
1280: $u(r_1 + 0.4, z)$ of an L-SPI. For better visibility more than one period is shown.
1281: Symbols and the full line denote Laser-Doppler
1282: velocimetry measurements \cite{SP99} and numerical simulations, respectively, of a
1283: Taylor-Couette setup of height $\Gamma=12$ with rigid, non-rotating lids at both
1284: ends. Both refer to the bulk region at mid height with a common local wavelength of
1285: $\lambda \simeq 1.76$. There the experimental maximum of $u$ is scaled to our
1286: simulation
1287: result. Dashed line refers to a simulation done with axially periodic conditions
1288: imposing a wavelength of $\lambda = 1.6$. Common parameters are
1289: $\eta=0.5$, $R_1=111$ with $R_2=-95$ for the experiments and $R_2=-96$ for the
1290: simulations.}
1291: \label{FIG:exp_num_u}
1292: \end{quote}
1293: \end{figure}
1294: \begin{figure}
1295: \begin{center}
1296: \includegraphics[width=14.5cm]{fig09.eps}
1297: \end{center}
1298: \begin{quote}
1299: \caption{Comparison of the frequency variation of experimental and numerical
1300: L-SPI with $R_1$. Symbols and the full line come from Laser-Doppler
1301: velocimetry measurements \cite{SP99} and numerical simulations, respectively, of a
1302: Taylor-Couette setup ($\eta$=0.5) of height $\Gamma=12$ with rigid, non-rotating lids
1303: at both
1304: ends that enforce the net mean axial flow $\langle w \rangle$ (\ref{Def_w_net})
1305: to vanish. Dashed line refers to a simulation done with axially periodic conditions
1306: ($\lambda = 1.6$). They allow for a finite Reynolds-stress-sustained
1307: $\langle w \rangle$ that is negative for our parameters. Upon subtracting this
1308: Galilean contribution
1309: $\langle w \rangle k$ from the oscillation frequency under periodic
1310: boundary conditions (dashed line) one obtains the dash-dotted line that lies
1311: close to the SPI
1312: frequencies with rigid end conditions. Common parameters are $R_2=-96$, however,
1313: $R_2=-100$ for the full line.}
1314: \label{FIG:exp_num_om}
1315: \end{quote}
1316: \end{figure}
1317: \begin{figure}[ht]
1318: \begin{center}
1319: \includegraphics[width=14.5cm]{fig10.eps}
1320: \end{center}
1321: \begin{quote}
1322: \caption{Influence of an external through-flow on vortex structures. (a) Primary
1323: Fourier amplitudes of the radial flow field at mid gap for the
1324: $M=1$ L-SPI ($u_{1,1}$), the $M=-1$ R-SPI ($u_{-1,1}$), and for TVF ($u_{0,1}$).
1325: (b) Axial phase velocity
1326: $w_{ph}=\omega /k$. (c) Net mean axial flow $\langle w \rangle$ (\ref{Def_w_net}).
1327: Full (dashed) lines with filled (open) symbols refer to stable (unstable) states.
1328: Arrows indicate transitions after loss of stability, see text for details. TVF is
1329: unstable in the
1330: $Re$-range shown here for our parameters $R_1=120, R_2=-100, \eta=0.5, k=3.927$.}
1331: \label{FIG:SPIvsRe}
1332: \end{quote}
1333: \end{figure}
1334: \begin{figure}[ht]
1335: \begin{center}
1336: \includegraphics[width=14.5cm]{fig11.eps}
1337: \end{center}
1338: \begin{quote}
1339: \caption{
1340: Radial profiles of the axial mean flow $w_0(r)$ (\ref{Defw_0})
1341: of spirals shown in Fig.~\ref{FIG:SPIvsRe} for axial
1342: Reynolds numbers $-4\leq Re \leq 14$ increasing in steps of 2. Thick line refers
1343: to $Re=0$. The transition from R- to L-SPI occurs around
1344: $Re\simeq7$, c.f. text. Parameters are $R_1=120, R_2=-100, \eta=0.5, k=3.927$.}
1345: \label{FIG:w_0}
1346: \end{quote}
1347: \end{figure}
1348: \begin{figure}[ht]
1349: \begin{center}
1350: \includegraphics[width=14.5cm]{fig12.eps}
1351: \end{center}
1352: \begin{quote}
1353: \caption{Velocity field ($u,w$) of L-SPI in an $r-z$ plane for $Re$=-5
1354: (a), 0(b), 5(c), 10(d) .
1355: Parameters are $\eta=0.5$, $k=3.927$, $R_1 = 120$, $R_2 = -100$.}
1356: \label{FIG:arrows_SPI_Re}
1357: \end{quote}
1358: \end{figure}
1359: \begin{figure}[ht]
1360: \begin{center}
1361: \includegraphics[width=14.5cm]{fig13.eps}
1362: \end{center}
1363: \begin{quote}
1364: \caption{The effect of an external through-flow on the axial profiles of the
1365: radial velocity of L-SPI. Lines show $u(z)$ at mid gap position for
1366: $Re=-5$ to $Re=20$ in steps of 5. Thick one refers to $Re=0$. In each case the
1367: maximal radial outflow is chosen to lie at $z=0.5 \lambda$. Parameters are
1368: $R_1 = 130$, $R_2=-100$, $\eta=0.5$, $\lambda=1.6$.}
1369: \label{FIG:u_L-SPI_Re}
1370: \end{quote}
1371: \end{figure}
1372: \begin{figure}[ht]
1373: \begin{center}
1374: \includegraphics[width=14.5cm]{fig14.eps}
1375: \end{center}
1376: \begin{quote}
1377: \caption{Influence of an external through-flow on vortex structures. (a) Primary
1378: Fourier amplitudes of the radial flow field at mid gap for the
1379: $M=1$ L-SPI ($u_{1,1}$), the $M=-1$ R-SPI ($u_{-1,1}$), and for TVF ($u_{0,1}$).
1380: (b) Axial phase velocity
1381: $w_{ph}=\omega /k$. (c) Net mean axial flow $\langle w \rangle - Re$. Full
1382: (dashed) lines with filled (open) symbols refer to stable (unstable) states.
1383: Arrows indicate transitions after loss of stability, see text for details.
1384: Parameters are $R_1=100, R_2=0, \eta=0.5, k=3.927$.}
1385: \label{FIG:bif_SPI_TVF_R2=0}
1386: \end{quote}
1387: \end{figure}
1388: \begin{figure}[ht]
1389: \begin{center}
1390: \includegraphics[width=14.5cm]{fig15.eps}
1391: \end{center}
1392: \begin{quote}
1393: \caption{$R_1 - Re$ phase diagram of TVF, R-SPI, and L-SPI for stationary
1394: outer cylinder. Solid lines represent linear stability thresholds of the basic
1395: state, i.e., bifurcation thresholds of the respective vortex solutions out of
1396: the combined CCF-APF. Dashed lines are stability boundaries of the vortex
1397: states. The phase diagram is symmetric under $Re \to -Re$. Parameters are
1398: $R_2=0, \eta=0.5, k=3.927$.}
1399: \begin{tabular}{|c|c|c|c|c|c|c|c|c|}
1400: \hline
1401: region & A & B & C & D & E & F & G & H \\\hline\hline
1402: TVF & s & s & s & u & - & s & s & s \\\hline
1403: R-SPI & s & u & - & - & - & u & - & - \\\hline
1404: L-SPI & s & s & s & s & s & u & u & - \\\hline
1405: \end{tabular}
1406: s: stable; u: unstable; -: nonexistent.
1407: \label{FIG:phasediagram}
1408: \end{quote}
1409: \end{figure}
1410:
1411: \end{document}
1412:
1413: