cond-mat0511233/p5.tex
1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: \documentclass[twocolumn,showpacs,amsmath,amssymb,bibnotes]{revtex4}
3: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4: \usepackage{bm}
5: \usepackage{graphicx,psfig,epsfig,subfigure,afterpage}
6: \usepackage{amsfonts}
7: %\usepackage{mcite}
8: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
9: 
10: \newcommand{\be}{\begin{equation}}
11: \newcommand{\ee}{\end{equation}}
12: \newcommand{\bea}{\begin{eqnarray}}
13: \newcommand{\eea}{\end{eqnarray}}
14: \newcommand{\gdot}{\dot{\gamma}}
15: \newcommand{\gdotbar}{\overline{\dot{\gamma}}}
16: \newcommand{\ie}{{\it i.e.\/}}
17: \newcommand{\eg}{{\it e.g.\/}}
18: \newcommand{\etc}{{\it etc.\/}}
19: \newcommand{\versus}{{\it vs.\/}}
20: \newcommand{\etal}{{\it et al.\/}}
21: \newcommand{\bw}{\begin{widetext}}
22: \newcommand{\ew}{\end{widetext}}
23: \newcommand{\lae}{\stackrel{<}{\scriptstyle\sim}}
24: 
25: \newcommand{\tmax}{t_{\rm max}}
26: \newcommand{\Tband}{T_{\rm b}}
27: \newcommand{\Nbase}{N_{\rm base}}
28: 
29: \newcommand{\ommax}{\omega_{\rm max}}
30: 
31: \newcommand{\vecv}[1]{\mathbf{{#1}}}
32: \newcommand{\tens}[1]{\mathbf{{#1}}}
33: \newcommand{\nablu}{{\bf \nabla}}
34: 
35: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
36: \begin{document}
37: 
38: \title{Nonlinear dynamics of a shear banding interface}
39: \author{S. M. Fielding}
40: \email{suzanne.fielding@manchester.ac.uk}
41: \affiliation{School of Mathematics, University of Manchester, Booth
42:   Street East, Manchester M13 9EP, United Kingdom } 
43:  \author{P. D. Olmsted}
44: %\email{p.d.olmsted@leeds.ac.uk} 
45: \affiliation{Polymer IRC and School of Physics \& Astronomy,
46:   University of Leeds, Leeds LS2 9JT, United Kingdom}  
47: \date{\today}
48: \begin{abstract}
49: %  We study numerically the dynamics of a shear banding interface in
50: %  two dimensional planar shear flow, within the non-local Johnson
51: %  Segalman model. Consistent with a recent linear stability analysis,
52: %  we show that an initially flat interface is unstable with respect to
53: %  small undulations. We then demonstrate that the instability
54: %  saturates in finite amplitude fluctuations, and study the resulting
55: %  nonlinear interfacial dynamics. We report a non equilibrium transition
56: %  between simple travelling interfacial waves and periodically
57: %  breaking waves.  When multiple shear bands are present we find
58: %  erratic interfacial dynamics and a stress response suggesting low
59: %  dimensional chaos.
60:   
61:   We study numerically the nonlinear dynamics of a shear banding
62:   interface in two dimensional planar shear flow, within the non-local
63:   Johnson Segalman model. Consistent with a recent linear stability
64:   analysis, we find that an initially flat interface is unstable with
65:   respect to small undulations for sufficiently small ratio of the
66:   interfacial width $\ell$ to cell length $L_x$. The instability
67:   saturates in finite amplitude interfacial fluctuations. For
68:   decreasing $\ell/L_x$ these undergo a non equilibrium transition
69:   from simple travelling interfacial waves with constant average wall
70:   stress, to periodically rippling waves with a periodic stress
71:   response. When multiple shear bands are present we find erratic
72:   interfacial dynamics and a stress response suggesting low
73:   dimensional chaos.
74: \end{abstract}
75: \pacs{{47.50.+d}, %{ Non-Newtonian fluid flows}--
76:      {47.20.-k}, %{ Hydrodynamic stability}--
77:      {36.20.-r}.%{ Macromolecules and polymer molecules}
78:      } 
79: \maketitle
80: 
81: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
82: 
83: Complex fluids such as polymers, liquid crystals and surfactant
84: solutions have mesoscopic structure that is readily perturbed by flow
85: \cite{LarsonComplex}.  For example, wormlike surfactant micelles with
86: lengths of the order of microns can be induced to stretch, disentangle and
87: entangle, and break or increase in length.  Their mechanical response
88: is therefore highly non-Newtonian, with shear flows inducing normal
89: stresses (\textit{e.g.}  $\sigma_{xx}-\sigma_{yy}$), and with the
90: shear stress $\sigma_{xy}$ being a nonlinear function of applied shear
91: rate $\gdot$.  Recent work on such fluids has led to a fairly
92: consistent picture of ``shear banding'': coexistence in shear flow of
93: viscously thicker (nascent) and thinner (flow-induced) bands of
94: material flowing at different local shear rates, for an overall
95: average imposed shear rate.
96: %For a given system, the shear stress at which
97: %coexistence occurs is unique and highly reproducible (``selected'').
98: 
99: This phenomenon can be described by constitutive models for which the
100: shear stress is a non-monotonic function of shear rate for homogeneous
101: flow. This leads to a separation into bands of differing shear rate
102: that coexist at a common shear stress \cite{SCM93b,olmsted99d}.
103: %To predict a uniquely selected stress at
104: %coexistence, one must invoke spatially non local (``diffusive'') terms
105: %to account for gradients across the banding
106: %interface~\cite{johnson77,olmsted99a}.  
107: Although most studies have assumed a flat interface between the bands
108: and (with a few exceptions \cite{fielding04}) predicted a
109: time-independent banded state, an accumulating body of data
110: %with resolution on the order of microns and milliseconds, 
111: has demonstrated that the average shear stress, and the banding
112: interface, can fluctuate
113: \cite{Becu.Manneville.ea04,Lopez-Gonzalez.Holmes.ea04,WunColLenArnRou01,Holmes.Lopez-Gonzalez.ea03,HBP98,WFF98,Manneville.Salmon.ea04,chaos2000}.
114: An important question is then whether these fluctuations resemble
115: small amplitude capillary waves stabilised by surface tension, or
116: whether they arise from an underlying instability, stabilised at large
117: amplitude by nonlinearities.  In this Letter we give strong evidence
118: supporting the latter scenario, via the first theoretical study of the
119: nonlinear dynamics of a shear banding interface.
120: 
121: 
122: \textit{The model -- } The generalised Navier Stokes equation
123: for a viscoelastic material in a Newtonian solvent of viscosity $\eta$
124: and density $\rho$ is:
125: \begin{equation}
126: \label{eqn:NS}
127: \rho(\partial_t + \vecv{v}.\nablu)\vecv{v} = \nablu .(\tens{\Sigma} +
128: 2\eta\vecv{D} -P\tens{I}), 
129: \end{equation}
130: where $\vecv{v}(\vecv{r})$ is the velocity field. The pressure $P$ is
131: determined by incompressibility, $\vecv{\nabla}\cdot\vecv{v}=0$.  The
132: viscoelastic stress $\vecv{\Sigma}(\vecv{r})$ evolves according to the
133: non-local (``diffusive'') Johnson Segalman (DJS)
134: model~\cite{johnson77,olmsted99a}
135: %
136: \begin{gather}\label{eqn:DJS}
137: (\partial_t
138: +\vecv{v}\cdot\nablu )\,\tens{\Sigma} 
139: - a(\tens{D}\cdot\tens{\Sigma}+\tens{\Sigma}\cdot\tens{D}) \\- 
140: (\tens{\Sigma}\cdot\tens{\Omega} - \tens{\Omega}\cdot\tens{\Sigma})  
141:  = 2 G\tens{D}-\frac{\tens{\Sigma}}{\tau}+ \frac{\ell^2}{\tau }\nablu^2 
142:  \tens{\Sigma}, \nonumber
143: \end{gather}
144: %
145: with plateau modulus $G$ and relaxation time $\tau$. $\tens{D}$ and
146: $\tens{\Omega}$ are the symmetric and antisymmetric parts of the
147: velocity gradient tensor, $(\nablu \vecv{v})_{\alpha\beta}\equiv
148: \partial_\alpha v_\beta$.  For $a=1$ and $\ell=0$ this model reduces
149: to the Oldroyd B model, which is motivated by considering an ensemble
150: of beads paired by springs (simplified polymer chains) into dumbbells.
151: Stress is generated as the flow deforms the dumbbells, and is relaxed
152: on the timescale $\tau$ for the springs to regain equilibrium length.
153: To capture shear thinning the DJS model invokes a ``slip parameter''
154: $a$ with $|a|<1$ to give non-affine dumbbell
155: deformation~\cite{johnson77}. The constitutive curve
156: $T_{xy}=\Sigma_{xy}(\gdot, a)+\eta\gdot$ for homogeneous planar shear
157: $\vecv{v}=y\gdot\vecv{\hat{x}}$ is then capable of non-monotonicity,
158: allowing shear banding. The non-local diffusive term in
159: Eqn.~\ref{eqn:DJS} accounts for spatial gradients across the banding
160: interface on a length scale $\ell$ \cite{olmsted99a}. It arises
161: naturally in models of liquid crystals, and diffusion of polymer
162: molecules~\cite{elkareh89}. In the context of one dimensional (1D)
163: calculations, it has been shown to give a unique (selected) stress
164: $T_{xy}^*$ at which banding occurs~\cite{olmsted99a}, as seen
165: experimentally.
166: 
167: We study 2D boundary-driven flow between parallel plates at $y=0,L$
168: with plate conditions
169: $\partial_y\Sigma_{\alpha\beta}=0\;\forall\;\alpha,\beta$ for the
170: viscoelastic stress, and no slip or permeation for the velocity.  In
171: the flow direction we consider a domain $x=\{0,L_x\}$, with periodic
172: boundary conditions.  We choose $a=0.3, \eta=0.05$ throughout, and
173: units in which $G=1,\tau=1$ and $L=1$.
174: %$G=\tau=L=1$.
175: 
176: For an imposed average shear rate $\gdotbar\equiv[v_y(L)-v_y(0)]/L$ in
177: the region of decreasing stress, $d T_{xy}/d\gdot<0$, homogeneous flow
178: is unstable~\cite{Yerushalmi70}. A 1D ($y$) calculation then predicts
179: separation into bands of shear rates $\gdot_1=0.66, \gdot_2=7.09$, at
180: a selected shear stress $T^{\ast}_{xy}=0.506$.  Recent
181: analysis~\cite{SMF2005} showed this stationary 1D banded state to be
182: linearly unstable to 2D ($x, y$) perturbations corresponding to
183: undulations of the interface with wavevector
184: $\vecv{q}=q_x\vecv{\hat{x}}$. The most unstable mode has $q_xL\approx
185: 2\pi$, and the instability involves feedback of the normal stress with
186: velocity fluctuations across the interface. In this Letter, we study
187: the fate of the interface in the \textit{nonlinear} regime, and
188: demonstrate it to be restabilised at the level of finite amplitude
189: undulations.
190: 
191: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
192: \begin{figure}[htb]
193:   \centering
194: %\includegraphics[scale=0.3]{./figures/dispersion1.eps}
195: \includegraphics[scale=0.3]{./dispersion1.eps}
196:  \caption{(a) Growth rate $\omega(q_x)$ of perturbations about an
197:    initially 1D banded state. ($\circ$, \textbf{---}): predictions of
198:    linear stability \cite{SMF2005} for $\ell=0.005,\, 0.01,\, 0.015$.
199:    ($\Box$): early time dynamics of the full 2D numerics for
200:    $\ell=0.005,\,L_x=4,\,6,\,8$; $\ell=0.01,\,L_x=2,\,4,\,6,\,8$;
201:    $\ell=0.015,\,L_x=2,4,8$. (b) Power spectrum $P(q_x)$ in the
202:    travelling wave regime for $\ell=0.005,\,L_x=1$ ($\circ$);
203:    $\ell=0.01,\,L_x=2$ ($\Box$); $\ell=0.015,\,L_x=2,\,4,\,6$
204:    ($\triangle$).}
205:  \label{fig:dispersion}
206: % nonlinear startup results; default Nx Ny Dt as per 2DDJS3 notes pages 3. ie for l=(0.015,0.01,0.005) take Ny=(400,400,800) and Lx=2,4,6 take Nx=(100,200,300) and for all take Dt=0.000316.
207: \end{figure}
208: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
209: \textit{Numerical scheme---} We enforce incompressibility by
210: eliminating the velocity in favour of a stream-function $\psi$. At
211: each time step we update the local parts of Eqn.~\ref{eqn:DJS} using
212: an explicit Euler algorithm within a finite difference scheme on a
213: rectangular grid of $(N_x, N_y)$ nodes, using third order upwinding
214: for the convective term $(\vecv{v}.\nablu)\tens{\Sigma}$. We then take
215: a Fourier transform $x\to q_x$ in the flow direction and update the
216: nonlocal part of Eqn.~\ref{eqn:DJS} using the semi-implicit
217: Crank-Nicolson algorithm. We finally update Eqn.~\ref{eqn:NS} at zero
218: Reynolds number, $\rho=0$. We attain convergence to one percent with
219: respect to increasing spatial and temporal resolution. This ensures
220: that the results shown are converged to the eye, apart from the stress
221: signal $\bar{T}_{xy}(t)$ in Fig.~\ref{fig:break}: this shows slight
222: quantitative, but not qualitative, changes. Runs at finer resolution
223: are prohibitively time consuming. We study two initial conditions:
224: (IC1) the stationary banded state, predicted by a 1D calculation, of
225: two bands separated by flat interface; and (IC2) a homogeneous
226: unstressed fluid, corresponding to shear startup from rest.
227: 
228: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
229: \begin{figure}[htb]
230:   \centering
231: %\includegraphics[scale=0.3]{./figures/phase.eps}
232: \includegraphics[scale=0.3]{./phase.eps}
233:  \caption{``Phase diagram''  for 
234: %shear rate
235:    $\bar{\gdot}=2.0$, showing the different non-linear regimes for IC1
236:    (accurate to  spacing between numbers) and the numbers of
237:    linearly unstable modes (Ref.~\cite{SMF2005}). }
238:  \label{fig:phasediag}
239: % \caption{``Phase diagram'' showing the number of linearly unstable
240: % modes (Ref.~\cite{SMF2005}) for shear rate
241: % $\bar{\gdot}=2.0$; and different non-linear regimes (for IC1),
242: % separated by solid lines (accurate to the spacing between the numbers
243: % shown).}  \label{fig:phasediag}
244: \end{figure}
245: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
246: \textit{Linear regime---} First we consider the early time evolution
247: of an initially flat interface, IC1. For a wide enough interface
248: $\ell$ and small enough system size $L_x$ this 1D solution is stable.
249: For smaller $\ell/L_x$, however, interfacial undulations are predicted
250: by the linear analysis of Refs.~\cite{SMF2005} to become unstable. Our
251: numerics successfully reproduce this instability during the initial
252: evolution away from IC1: the eigenvectors and growth rates
253: $\omega(q_x)$ match the analytical results of Refs.~\cite{SMF2005},
254: Fig~\ref{fig:dispersion}.
255: %(verifying our numerical procedure). 
256:  Beyond this early-time regime,
257: the interface is restabilised by nonlinear effects, and the
258: instability saturates in undulations of finite large amplitude.
259: Depending on the distance from the onset of linear instability,
260: Fig.~\ref{fig:phasediag}, the ultimate attractor corresponds either to
261: a steady travelling interfacial wave, or to periodically rippling
262: waves.
263: 
264: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
265: \begin{figure}[b]
266:   \centering
267: %\includegraphics[scale=0.475]{./figures/travelSmaller.eps}
268: %\includegraphics[scale=0.475]{./figures/travelling.eps}
269: \includegraphics[scale=0.95]{./travelling.eps}
270:  \caption{Greyscale of order parameters for travelling wave
271:    in the $(x,y)$ plane for $\ell=0.015$, $L_x=6$, and upper wall
272:    velocity $V\equiv \gdotbar L=2$ to the right.
273: \label{fig:travel}}
274: \end{figure}
275: %% results file results/gdot2.0_l0.015_Lx6.0_Nx400_Ny800_Dt0.0001_tmax200.0_tref1.0_tswi0.0_temp1.0e-10_convection3_rungeFlag0_printState
276: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
277: 
278: 
279: \textit{Travelling wave---} For values of $\ell/L_x$ marginally inside
280: the unstable regime (Fig.~\ref{fig:phasediag}) the ultimate attractor
281: comprises a travelling wave $\vecv{A}=\vecv{A}(y,x-ct)$ for all order
282: parameters $\vecv{A}=\{\Sigma_{xx},\Sigma_{xy},\Sigma_{yy},\psi\}$.
283: Since this satisfies periodic boundary conditions, the wall-averaged
284: shear stress is constant in time, with a value
285: $\overline{T}_{xy,ss}\simeq0.51-0.54$ that depends on $\ell$ and $L_x$ and
286: is slightly higher than the selected stress $T_{xy}^{\ast}=0.506$ of
287: the 1D calculation. In Fig.~\ref{fig:travel} the wave-speed $c=1.15$,
288: compared with a horizontal velocity $v_x$ that varies between $1.439$
289: and $1.484$ along the interface.
290: %as $x$ \textbf{[is this as a function of y rather than
291: %  x?]} is swept across the cell.
292: To analyse the structure of this
293: state we define the following power spectrum,
294: %
295: \begin{equation}
296:   \label{eq:1}
297:    P(q_x)=\sum_{i=1}^4\int_0^L\,dy| A_i(q_x,y)|^2.
298: \end{equation}
299: %
300: %
301: \begin{figure}[t]
302:   \centering
303: %\includegraphics[scale=0.45]{./figures/breakDrawInterface.eps}
304: %\includegraphics[scale=0.45]{./figures/breakDrawInterfaceSmaller.eps}
305: %\includegraphics[scale=0.45]{./figures/Figbreak.eps}
306: %\includegraphics[scale=0.3]{./figures/stress.eps} 
307: \includegraphics[scale=0.9]{./Figbreak.eps}
308: \includegraphics[scale=0.3]{./stress.eps} 
309: \caption{Rippling wave  at $\ell=0.005, L_x=4, \gdotbar=2$.  Top:
310:   Greyscale of $\Sigma_{xx}(x,y)$.  Upper wall moves to the
311:   right. White line in d): interface height defined in text. Bottom:
312:   average wall stress. Inset: $(N_x,N_y,dt)=(200,800,0.0003)$, solid
313:   line; $(400,800,0.0001)$, dotted; $(400,800,0.0003)$, dashed;
314:   $(500,800,0.0002)$, dot-dashed.}
315: %  average wall stress. Inset: $(N_x,N_y,dt)=(200,800,0.0003)$ ---; $(400,800,0.0001)$ $\cdot\cdot\cdot$; $(400,800,0.0003)$ - - - ;
316: %  $(500,800,0.0002)$ $-\cdot-$.}
317: \label{fig:break}
318: \end{figure}
319: %\end{figure}
320: %% results file results/gdot2.0_l0.005_Lx4.0_Nx400_Ny800_Dt0.0001_tmax200.0_tref1.0_tswi0.0_temp1.0e-10_convection3_rungeFlag0_printState
321: %
322: %\begin{figure}[t]
323: %  \centering
324: %\includegraphics[scale=0.3]{./figures/stress.eps}
325: % \caption{Stress {\it vs.} time for
326: % parameters of Fig.~\ref{fig:break}. Inset: zoom  around
327: % time interval of Fig.~\ref{fig:break} (arrow).} \label{fig:stress}
328: %% results file results/gdot2.0_l0.005_Lx4.0_Nx400_Ny800_Dt0.0001_tmax200.0_tref1.0_tswi0.0_temp1.0e-10_convection3_rungeFlag0_printState
329: %
330: %
331: \begin{figure}[t]
332:   \centering
333: %\includegraphics[scale=0.15]{./figures/interface.eps}
334: \includegraphics[scale=0.45]{./interface.eps}
335:  \caption{Rippling wave regime. Greyscale of interface height (defined
336:    in text) in  $(x,t)$ plane. Parameters as in
337:    Fig.~\ref{fig:break}. Time $t=150\ldots160$ upwards. Upper: raw
338:    data. Lower:  transformed as $x\to x-ct$, $c$ extracted by
339:    eye.}\label{fig:interface} 
340: \end{figure}
341: %% results file results/gdot2.0_l0.005_Lx4.0_Nx400_Ny800_Dt0.0001_tmax200.0_tref1.0_tswi0.0_temp1.0e-10_convection3_rungeFlag0_printState
342: %
343: As seen in Figs.~\ref{fig:dispersion} and~\ref{fig:phasediag}, while
344: the linear stability analysis predicted only a few unstable modes, the
345: ultimate nonlinear state has active modes at higher $q_x$,
346: Fig.~\ref{fig:dispersion}b.  Interestingly, the dominant mode is still
347: close to that of the linear analysis, despite being in a nonlinear
348: regime with a finite interfacial displacement $\delta h$.
349: %(In the linear calculation, $\delta h\to 0$.) 
350: Indeed, for all cases in Fig. \ref{fig:dispersion}b, we find
351: empirically that the dominant mode has the longest wavelength that is
352: both consistent with periodic boundary conditions and linearly
353: unstable.
354: 
355: 
356: %\textbf{[Find wavespeed; simple motion at more or less local velocity,
357: %  or other, such as following from the wavelength and $\tau$ or other
358: %  time?]}  
359: 
360: 
361: \textit{Rippling wave---} For smaller $\ell/L_x$, deeper inside the
362: unstable regime (Fig.~\ref{fig:phasediag}), we see a new regime in
363: which the travelling wave now periodically ``ripples''.  The
364: wall-averaged stress $\overline{T}_{xy}$ is periodic in time
365: (Fig.~\ref{fig:break}, again for IC1), with variations of the order of
366: one percent, and an average value larger than the 1D selected stress
367: $T_{xy}^{\ast}$. The interface height $h(x)$ is shown as a white line
368: in Fig.~\ref{fig:break}d and as a greyscale over a time window in
369: Fig.~\ref{fig:interface}.  Because the structure is quite complicated
370: near a rippling wave (Figs.~\ref{fig:break}cd), the location of $h(x)$
371: depends on its precise definition. We take $h(x)$ as that value of $y$
372: at which $\Sigma_{xx}$ lies half way between its values at the two
373: walls. (An alternative might be $\int dy\,y \,| \partial_y
374: \Sigma_{xx}(y)|$.)  This provides a fairly reliable measure, subject
375: to a small kink in the rippling region (Fig.~\ref{fig:break}d).
376: 
377: \textit{Robustness to initial conditions---} So far, we have
378: considered only IC1, two bands separated by a single flat interface.
379: To test robustness to ICs, we now study startup from rest (IC2).  Here
380: the system develops either (i) two bands that show the same dynamics
381: as with IC1, or (ii) multiple bands that have erratic dynamics
382: suggestive of low dimensional chaos, Fig.~\ref{fig:multiple1}.  This
383: is the counterpart of startup from rest in 1D planar shear flow, which
384: typically yields random configurations of multiple bands, because the
385: uniform shear stress $T_{xy}(y)=T^*_{xy}$ allows interfaces to reside
386: at any $y$ value. (To achieve IC1, the system was prebiased to form
387: two bands.)  In contrast, Couette flow between concentric cylinders
388: has a stress $T_{xy}\sim1/r^2$ ($r$=radius), which allows only a
389: single stationary interface at $r^*\sim1/\sqrt{T^*_{xy}}$, as
390: demonstrated in 1D in Ref.~\cite{GrecBall97,radulescu99a}. By analogy,
391: we anticipate that curvature should eliminate the multiple band case
392: (ii) in 2D.  Although a true Couette calculation has not been
393: performed, we implement a ``poor man's'' version by adding a biasing
394: stress-gradient in the $y$ direction while the bands are forming. For
395: strong enough bias, we then indeed find case (i), just two bands. An
396: open question is whether, even with IC1, a transition exists in the
397: numerically inaccessible regime of small $\ell/L_x$ (far bottom right
398: in Fig.~\ref{fig:phasediag}) to chaotic dynamics of a single
399: interface.
400: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
401: \begin{figure}[t]
402:   \centering
403: \includegraphics[scale=0.57]{./multiplebands.eps}
404: \includegraphics[scale=0.29]{./multiple2.eps}
405: %\includegraphics[scale=0.57]{./figures/multiplebands.eps}
406: %\includegraphics[scale=0.29]{./figures/multiple2.eps}
407:  \caption{Multiple bands after startup from rest for
408:    $\ell=0.005$, $L_x=2$, $\gdotbar=2$. Top: greyscale of
409:    $\Sigma_{xx}(x,y)$ at time $t=120$, reconstructed from the first
410:    $15$ Fourier modes. Bottom: Shear stress $T_{xy}(t)$.
411:    \label{fig:multiple1}}
412: \end{figure} 
413: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
414: 
415: To summarise, we have studied numerically the nonlinear dynamics of a
416: shear banding interface in 2D planar shear flow, within the non-local
417: DJS model, for a system of length $L_x$ with periodic boundary
418: conditions in $x$. Upon decreasing the ratio of the interfacial width
419: $\ell$ to the system length $L_x$ the stable two band state separated
420: by a single interface undergoes successive transitions to travelling
421: waves and rippling travelling waves.  Multiple shear bands can also be
422: found, depending on initial conditions, which show irregular
423: interfacial dynamics and a corresponding stress signal suggestive of
424: low dimensional chaos.  Depending on the regime, this dynamics is
425: qualitatively similar to several recent experiments in wormlike
426: micellar solutions that studied the interfacial dynamics and the
427: associated stress response
428: \cite{Becu.Manneville.ea04,Lopez-Gonzalez.Holmes.ea04,chaos2000,WFF98}.
429: %The rippling instability resembles the Kelvin-Helmholtz instability;
430: %however, that instability is inertial in origin and the waves
431: %typically break. In the current zero Reynolds number breaking waves
432: %may be prevented due to the finite interfacial width $\ell$, which
433: %prohibits sharp variations.
434: 
435: In earlier work, Yuan \etal~\cite{Jupp.Yuan04} used a 2D finite
436: element algorithm to evolve a related JS model in planar shear.
437: Contrary to our results, they reported a stable interface. However,
438: they considered rather small values of $L_x$, perhaps in the stable
439: regime, and the finite element technique may have introduced
440: stabilising numerical diffusion.  Related recent work by
441: Onuki~\cite{Furukawa.Onuki05} on a two-fluid version of the DJS model
442: \cite{fielding03a} at imposed stress also showed a variety of
443: interesting time-dependent behaviour due to the interplay between
444: stress and concentration degrees of freedom; it would be interesting
445: to study the phenomena found here in that model. Indeed, an open
446: question is the extent to which the instability seen in this DJS model
447: is ubiquitous among other models of shear banding, such as those for
448: liquid crystals. Future work should also extend to larger systems and
449: smaller interfaces, likely to lead to more complex dynamical states.
450: 
451: SMF thanks Prof. Ajdari for his hospitality at the ESPCI in Paris
452: where this work was partly carried out, and UK EPSRC GR/S29560/01 for
453: financial support. We thank O. Harlen and H. Wilson for helpful
454: discussions.
455: 
456: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
457: %\bibliographystyle{prsty}
458: %\bibliography{articles,banding,berrportdecruppe,books,callaghan,LCtheory,malkus,master,rheofolks,sriram,worms,shear05,bord04,rheochaosPRL,worms3,latest}
459: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
460: \begin{thebibliography}{10}
461: 
462: \bibitem{LarsonComplex}
463: R.~G. Larson, {\em The Structure and Rheology of Complex Fluids} (Oxford
464:   University Press, New York, 1999).
465: 
466: \bibitem{SCM93b}
467: N.~A. Spenley, M.~E. Cates, and T.~C.~B. McLeish, Phys. Rev. Lett. {\bf 71},
468:   939  (1993).
469: 
470: \bibitem{olmsted99d}
471: P.~D. Olmsted, Curr. Op. Coll. Int. Sci. {\bf 4},  95  (1999).
472: 
473: 
474: \bibitem{fielding04}
475: S.~M. Fielding and P.~D. Olmsted, Phys. Rev. Lett. {\bf 92},  084502  (2004);
476: A. Aradian and M.~E. Cates, Europhys. Lett. {\bf 70},  397  (2005).
477: 
478: \bibitem{Becu.Manneville.ea04}
479: L. Becu, S. Manneville, and A. Colin, Phys. Rev. Lett. {\bf 93},  018301
480:   (2004).
481: 
482: \bibitem{Lopez-Gonzalez.Holmes.ea04}
483: M.~R. Lopez-Gonzalez, W.~M. Holmes, P.~T. Callaghan, and P.~J. Photinos, Phys.
484:   Rev. Lett. {\bf 93},  268302 (2004).
485: 
486: \bibitem{WunColLenArnRou01}
487: A.~S. Wunenburger, A. Colin, J. Leng, A. Arneodo, and D. Roux, Phys.\ Rev.\
488:   Lett. {\bf 86},  1374  (2001).
489: 
490: 
491: \bibitem{Holmes.Lopez-Gonzalez.ea03}
492: W.~M. Holmes, M.~R. Lopez-Gonzalez, and P.~T. Callaghan, Europhys. Lett. {\bf
493:   64},  274  (2003).
494: 
495: \bibitem{HBP98}
496: Y.~T. Hu, P. Boltenhagen, and D.~J. Pine, J. Rheol. {\bf 42},  1185  (1998).
497: 
498: \bibitem{WFF98}
499: E.~K. Wheeler, P. Fischer, and G.~G. Fuller, J. Non-Newt. Fl. Mech. {\bf 75},
500:   193  (1998).
501:   
502: \bibitem{Manneville.Salmon.ea04} S. Manneville, J.~B. Salmon, L. Becu,
503:   A. Colin, and F. Molino, Rheol. Acta {\bf 43}, 408 (2004); J.~B.
504:   Salmon, L. Becu, S. Manneville, and A. Colin, Eur. Phys. J. E {\bf
505:     10}, 209 (2003).
506: 
507: %\bibitem{CatHeaAjd02}
508: %M.~E. Cates, D.~A. Head, and A. Ajdari, Phys.\ Rev.\ E {\bf 66},  025202
509: %  (2002).
510: 
511: \bibitem{johnson77}
512: M. Johnson and D. Segalman, J. Non-Newt. Fl. Mech {\bf 2},  255  (1977).
513: 
514: \bibitem{olmsted99a}
515: P.~D. Olmsted, O. Radulescu, and C.-Y.~D. Lu, J. Rheology {\bf 44},  257
516:   (2000).
517: 
518: \bibitem{elkareh89}
519: A.~W. El-Kareh and L.~G. Leal, J. Non-Newt. Fl.~Mech. {\bf 33},  257  (1989).
520: 
521: \bibitem{Yerushalmi70}
522: J. Yerushalmi, S. Katz, and R. Shinnar, Chemical Engineering Science {\bf 25},
523:   1891  (1970).
524:   
525: \bibitem{SMF2005} S.~M. Fielding, Phys. Rev. Lett. {\bf 95}, 134501
526:   (2005); H.~J. Wilson and S.~M. Fielding, submitted to JNNFM.
527: 
528: %\bibitem{wilson2005}
529: %H.~J. Wilson and S.~M. Fielding, in preparation.
530: 
531: \bibitem{GrecBall97}
532: F. Greco and R.~C. Ball, J. Non-Newt. Fl. Mech. {\bf 69},  195  (1997).
533: 
534: \bibitem{radulescu99a}
535: O. Radulescu and P.~D. Olmsted, J. Non-Newt. Fl. Mech {\bf 91},  141  (2000).
536: 
537: \bibitem{chaos2000}
538: R. Bandyopadhyay, G. Basappa, and A.~K. Sood, Phys. Rev. Lett. {\bf 84},  2022
539:   (2000).
540: 
541: \bibitem{Jupp.Yuan04}
542: L. Jupp and X.~F. Yuan, J. Non-Newton. Fluid Mech. {\bf 124},  93  (2004);
543: X.~F. Yuan, Europhys. Lett. {\bf 46},  542  (1999).
544: 
545: \bibitem{Furukawa.Onuki05}
546: A. Furukawa and A. Onuki, Physica D {\bf 205},  195  (2005).
547: 
548: \bibitem{fielding03a}
549: S.~M. Fielding and P.~D. Olmsted, Phys. Rev. Lett. {\bf 90},  224501  (2003).
550: 
551: \end{thebibliography}
552: 
553: 
554: \end{document}
555: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
556: 
557: