1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2: \documentclass[eqsecnum,twocolumn,showpacs,amsmath,amssymb]{revtex4}

3: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4: \usepackage{bm}

5: \usepackage{graphicx,subfigure,afterpage}

6: \usepackage{amsfonts}

7: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

8: %% the paper text follows this prescription, the diagram labels and file names follow the original research notes  -- hence the discrepancy

9: %prescription for converting from the parameters of my numerics to more sensible ones

10: %multiply alpha by 10 to get the paper's alpha

11: %multiply beta by 0.8e-4 to get l

12: %mulitly gamm by 4e-4 to get xi

13: 

14: \newcommand{\etal}{{\it et al.}}

15: 

16: \newcommand{\djs}{{\mathfrak{D}}_{\rm M}}

17: \newcommand{\df}{{\mathfrak{D}}_{\rm F}}

18: \newcommand{\di}{{\mathfrak{D}}_{\rm i}}

19: 

20: \newcommand{\tdjs}{{\mathfrak{\tilde{D}}}_{\rm M}}

21: \newcommand{\tdf}{{\mathfrak{\tilde{D}}}_{\rm F}}

22: \newcommand{\tdi}{{\mathfrak{\tilde{D}}}_{\rm i}}

23: 

24: \newcommand{\hdjs}{{\mathfrak{\hat{D}}}_{\rm M}}

25: 

26: \newcommand{\vect}[1]{\underline{#1}}

27: \newcommand{\tens}[1]{\underline{\underline{#1}}}

28: 

29: \newcommand{\vm}{\vect{v}_{\,\rm m}}

30: \newcommand{\vs}{\vect{v}_{\,\rm s}}

31: 

32: \newcommand{\vrel}{\vect{v}_{\, \rm rel}}

33: \newcommand{\vv}{\vect{v}}

34: 

35: \newcommand{\dx}{\Delta x}

36: \newcommand{\ny}{N_y}

37: \newcommand{\dt}{\Delta t}

38: 

39: \newcommand{\etam}{\eta_{\,\rm m}}

40: \newcommand{\etas}{\eta_{\,\rm s}}

41: 

42: \newcommand{\Dm}{\tens{D}_{\,\rm m}}

43: \newcommand{\Ds}{\tens{D}_{\,\rm s}}

44: 

45: \newcommand{\Drel}{\tens{D}_{\,\rm rel}}

46: \newcommand{\Dv}{\tens{D}}

47: 

48: \newcommand{\Omm}{\tens{\Omega}_{\,\rm m}}

49: \newcommand{\Oms}{\tens{\Omega}_{\,\rm s}}

50: 

51: \newcommand{\nablu}{\vect{\nabla}}

52: 

53: \newcommand{\gdotb}{\bar{\dot{\gamma}}}

54: \newcommand{\phib}{\bar{\phi}}

55: 

56: \newcommand{\be}{\begin{equation}}

57: \newcommand{\ee}{\end{equation}}

58: \newcommand{\bea}{\begin{eqnarray}}

59: \newcommand{\eea}{\end{eqnarray}}

60: \newcommand{\lav}{\left\langle}

61: \newcommand{\rav}{\right\rangle}

62: \newcommand{\gdot}{\dot{\gamma}}

63: \newcommand{\bgdot}{\bar{\dot{\gamma}}}

64: \newcommand{\la}{\lambda}

65: \newcommand{\de}{\Delta}

66: \newcommand{\sto}{\rightarrow}

67: \newcommand{\D}{\displaystyle}

68: \newcommand{\taur}{\tau_{\rm r}}

69: \newcommand{\Tf}{T_{\rm f}}

70: \newcommand{\gae}{\stackrel{>}{\scriptstyle\sim}}

71: \newcommand{\lae}{\stackrel{<}{\scriptstyle\sim}}

72: \newcommand{\ie}{{\it i.e.\/}}

73: \newcommand{\eg}{{\it e.g.\/}}

74: \newcommand{\etc}{{\it etc.\/}}

75: \newcommand{\versus}{{\it vs.\/}}

76: \newcommand{\eps}{\epsilon}

77: \newcommand{\Deta}{\Delta{\eta}}

78: \newcommand{\etab}{{\eta}}

79: \newcommand{\etabar}{\bar{\eta}}

80: \newcommand{\bw}{\begin{widetext}}

81: \newcommand{\ew}{\end{widetext}}

82: 

83: \newcommand{\bmini}{\begin{minipage}}

84: \newcommand{\emini}{\end{minipage}}

85: 

86: \newcommand{\eigenvec}{\vect{{\tt{v}}}_{\vect{k},\alpha}}

87: \newcommand{\eigenvecmax}{\vect{{\tt{v}}}_{\vect{k},\Upsilon}}

88: 

89: \newcommand{\phil}{\phi_\ell} 

90: \newcommand{\phic}{\bar{\phi}_{\rm c}}

91: \newcommand{\gdotc}{\bar{\gdot}_{\rm c}}

92: \newcommand{\model}{d-JS-$\phi$}

93: 

94: 

95: \newcommand{\sigmas}{\Sigma_{\rm sel}}

96: \newcommand{\gdoth}{\gdot_{\rm h}}

97: \newcommand{\gdotl}{\gdot_\ell}

98: 

99: %\numberwithin{equation}{section}

100: 

101: \begin{document}

102: 

103: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

104: \title{Flow phase diagrams for concentration-coupled shear banding}

105: \author{S. M. Fielding}

106: \email{physf@irc.leeds.ac.uk} 

107: \author{P. D. Olmsted}

108: \email{p.d.olmsted@leeds.ac.uk}

109: \affiliation{Polymer IRC and

110:   Department of Physics \& Astronomy, University of Leeds, Leeds LS2

111:   9JT, United Kingdom} 

112: \date{\today} 

113: \begin{abstract}

114:   After surveying the experimental evidence for concentration coupling

115:   in the shear banding of wormlike micellar surfactant systems, we

116:   present flow phase diagrams spanned by shear stress (or strain-rate)

117:   and concentration, calculated within the two-fluid, non-local

118:   Johnson-Segalman (\model) model. We also give results for the

119:   macroscopic flow curves $\Sigma(\gdotb,\phib)$ for a range of

120:   (average) concentrations $\phib$. For any concentration that is high

121:   enough to give shear banding, the flow curve shows the usual

122:   non-analytic kink at the onset of banding, followed by a coexistence

123:   ``plateau'' that slopes upwards, $d\Sigma/d\gdotb>0$. As the

124:   concentration is reduced, the width of the coexistence regime

125:   diminishes and eventually terminates at a non-equilibrium critical

126:   point $[\Sigma_{\rm c},\phib_{\rm c},\gdotb_{\rm c}]$.  We outline

127:   the way in which the flow phase diagram can be reconstructed from a

128:   family of such flow curves, $\Sigma(\gdotb,\phib)$, measured for

129:   several different values of $\phib$. This reconstruction could be

130:   used to check new measurements of concentration differences between

131:   the coexisting bands. Our \model\ model contains two different

132:   spatial gradient terms that describe the interface between the shear

133:   bands.  The first is in the viscoelastic constitutive equation, with

134:   a characteristic (mesh) length $l$. The second is in the

135:   (generalised) Cahn-Hilliard equation, with the characteristic length

136:   $\xi$ for equilibrium concentration-fluctuations. We show that the

137:   phase diagrams (and so also the flow curves) depend on the ratio

138:   $r\equiv l/\xi$, with loss of unique state selection at $r=0$.  We

139:   also give results for the full shear-banded profiles, and study the

140:   divergence of the interfacial width (relative to $l$ and $\xi$) at

141:   the critical point.

142: 

143: \end{abstract}

144: \pacs{{47.50.+d}{ Non-Newtonian fluid flows}--

145:      {47.20.-k}{ Hydrodynamic stability}--

146:      {36.20.-r}{ Macromolecules and polymer molecules}

147: }

148: \maketitle

149: 

150: \section{Introduction}

151: \label{sec:intro}

152: 

153: For many complex fluids, the intrinsic constitutive curve of shear

154: stress $\Sigma$ as a function of shear rate $\gdot$ is non-monotonic,

155: admitting multiple values of shear rate at common stress. For

156: semi-dilute wormlike micelles, theory~\cite{cates90,SpenCate94,SCM93}

157: predicts the form ACEG of Fig.~\ref{fig:schem}. In the range

158: $\gdot_{\rm c1}<\gdot<\gdot_{\rm c2}$ where the stress is decreasing,

159: steady homogeneous flow (Fig.~\ref{fig:picture}a) is

160: unstable~\cite{Yerushalmi70}. For an applied shear rate $\bar{\gdot}$

161: in this unstable range, Spenley, Cates and McLeish~\cite{SCM93}

162: proposed that the system separates into high and low shear rate bands

163: ($\gdoth$ and $\gdotl$; Fig.~\ref{fig:picture}b) and that

164: any change in the applied shear rate then merely adjusts the relative

165: fraction of the bands, while the stress $\sigmas$ (which is common to

166: both) remains constant.  The steady state flow curve then has the form

167: ABFG. Several constitutive models augmented with interfacial gradient

168: terms have captured this

169: behaviour~\cite{olmsted99a,lu99,olmstedlu97,spenley96}.

170: 

171: %

172: \begin{figure}[h]

173: \begin{center}

174:  \includegraphics[scale=1.0]{makefigure1.ps}

175: %\input{schem.pstex_t}

176: %\centerline{\psfig{figure=schem.eps,width=8cm}}

177: \caption{Schematic flow curves for  wormlike micelles: the homogeneous

178:   constitutive curve is ACEG; the steady shear-banded flow curve is BF

179:   (without concentration coupling in planar shear) or B'F' (with

180:   concentration coupling, or in a cylindrical Couette device). 

181: \label{fig:schem} } 

182: \end{center}

183: \end{figure}

184: %

185: %

186: \begin{figure}[h]

187:   \includegraphics[scale=0.3]{picture.eps}

188: \caption{(a) Homogeneous shear rate and (b) banded profiles.

189: \label{fig:picture} } 

190: \end{figure}

191: %

192: 

193: Experimentally, this scenario has been widely observed in semidilute

194: wormlike micelles~\cite{berret94b,Call+96,grand97}. The steady state

195: flow curve (which is often attained only after very long

196: transients~\cite{grand97}) has a well defined, reproducible stress

197: plateau $\sigmas$.  Coexistence of high and low viscosity bands has

198: been observed by NMR

199: spectroscopy~\cite{Call+96,MairCall96,MairCall96c,BritCall97}. Further

200: evidence comes from small angle neutron scattering

201: (SANS)~\cite{berret94a,schmitt94,Capp+97,rehage91,berret94b}; and from

202: flow birefringence (FB)~\cite{Decr+95,Makh+95,DCC97,BPD97}, which

203: reveals a (quasi) nematic birefringence band coexisting with an

204: isotropic one (but see~\cite{FisCal01,FisCal00}).

205: 

206: In some systems, the coexistence plateau is not perfectly flat, but

207: slopes upward slightly with increasing shear rate (B'F' in

208: Fig.~\ref{fig:schem}). See, for example, Ref.~\cite{LerDecBer00} for

209: CTAB(0.3M)/${\rm NaNO}_3\rm {(1.79M)}$/H$_2$O at micellar volume

210: fraction $\phi=11\%$. This effect is much more pronounced in other,

211: more concentrated systems that are near an underlying (zero-shear)

212: isotropic-nematic (I-N) transition ($\phi\approx

213: 30\%$)~\cite{schmitt94,BRL98,berret94a}.

214: 

215: In a cylindrical Couette geometry, this upward slope is qualitatively

216: consistent with the inhomogeneous stress arising from the cell

217: curvature: as the high-shear band at the inner cylinder expands

218: outward with increasing applied shear rate, the applied torque must

219: increase to ensure that the interface between the bands stays at the

220: selected stress $\sigmas$~\cite{olmsted99a}.  However a more generic

221: explanation, independent of geometry, is that the shear banding

222: transition is coupled to concentration~\cite{schmitt95,olmstedlu97}.

223: In this case, the properties of each phase must change as the applied

224: shear rate is tracked through the coexistence regime, because material

225: is redistributed between the bands as the high shear band grows to

226: fill the gap.

227: 

228: Generically, one expects flow to be coupled to concentration in

229: viscoelastic solutions where the different constituents (polymer and

230: solvent) have widely separated relaxation

231: timescales~\cite{brochdgen77,HelfFred89,DoiOnuk92,milner93,WPD91,beris94,SBJ97,tanak96}.

232: This was explained by Helfand and Fredrickson (HF)~\cite{HelfFred89}

233: as follows.  In a sheared solution, the parts of an extended polymer

234: molecule (micelle for our purposes) in regions of lower viscosity

235: will, upon relaxing to equilibrium, move more than the parts mired in

236: a region of high viscosity and concentration. A relaxing molecule

237: therefore on average moves towards the higher concentration region.

238: This provides a positive feedback mechanism whereby micelles can move

239: {\em up} their own concentration gradient, and leads to flow-enhanced

240: concentration fluctuations perpendicular to the shear compression

241: axis.

242: This was observed in steadily sheared polymer solutions in the early

243: 1990's~\cite{WPD91}. In a remarkable paper, Schmitt

244: \etal~\cite{schmitt95} discussed the implications of this feedback

245: mechanism for the onset of flow instabilities. Strongly enhanced

246: concentration fluctuations were subsequently observed in the early

247: time kinetics of the shear banding instability in

248: Ref.~\cite{DecLerBer01}.

249: 

250: Recently, therefore, we introduced a model of concentration-coupled

251: shear banding~\cite{FieOlm02,FieOlm02b} by combining the diffusive

252: Johnson Segalman (d-JS) model~\cite{johnson77,olmsted99a} with a

253: two-fluid approach~\cite{brochdgen77,milner91,deGen76,Brochard83} to

254: concentration fluctuations. This ``d-JS-$\phi$" model does not address

255: the microscopics of any particular viscoelastic system, but instead

256: should be regarded as a minimal model that combines (i) a constitutive

257: curve like that of semidilute wormlike micelles (Fig.~\ref{fig:schem})

258: with (ii) the non-local (interfacial) terms required for selection of

259: a unique banded state~\cite{lu99} and (iii) a simple approach to

260: concentration coupling.

261: 

262: In Refs.~\cite{FieOlm02,FieOlm02b}, we examined the linear stability

263: of initially homogeneous shear states in this \model\ model with

264: respect to coupled fluctuations in shear rate $\gdot$, micellar strain

265: $\tens{W}$ and concentration $\phi$. We thereby calculated the

266: ``spinodal'', inside which such homogeneous states are unstable.  We

267: also calculated the selected length scale at which inhomogeneity first

268: emerges during startup flows in the unstable region. In the limit of

269: zero concentration coupling, the unstable region coincides with that

270: of negative slope in the homogeneous constitutive curve, as expected;

271: but no length scale is selected during startup. Concentration coupling

272: enhances this instability at short length scales. It thereby broadens

273: the region of instability, and selects a length scale at which

274: inhomogeneity must emerge.

275: 

276: In the present paper, we compute the corresponding steady-state flow

277: phase diagram (the ``binodals'' and their tie-lines). As far as we are

278: aware, this is the first concrete calculation aimed at qualitatively

279: describing concentration-coupled shear banding for systems such as

280: semi-dilute wormlike micelles.  We start in Sec.~\ref{sec:background}

281: by describing the experimental background in more detail.  We also

282: compare our present calculation with the only other existing one for

283: concentration-coupled shear banded states, in concentrated solutions of

284: rigid rods~\cite{olmsted99c}. In Sec.~\ref{sec:model} we summarize our

285: \model\ model. We then review our results for the spinodal onset of

286: instability in Sec.~\ref{sec:spinodals}. In Sec.~\ref{sec:numerics} we

287: describe our numerical procedure for computing the banded steady

288: states, with brief discussion of our careful study of mesh and finite

289: size effects.  We then (Sec.~\ref{sec:results}) present our results

290: for the flow phase diagrams and shear-banded profiles. We conclude in

291: Sec.~\ref{sec:conclusion}.

292: 

293: \section{Experimental background; theoretical context}

294: \label{sec:background}

295: 

296: In this section, we discuss in more detail the experimental evidence

297: for concentration coupling in the shear banding of wormlike micelles.

298: We survey both (i) concentrated systems near the zero-shear I-N phase

299: transition and (ii) semidilute systems, in which underlying nematic

300: interactions are likely to be less important.  Correspondingly, we

301: compare the present calculation (aimed at the semi-dilute systems)

302: with an earlier calculation of flow phase diagrams in concentrated

303: solutions of rigid rods (near the I-N transition)~\cite{olmsted99c}.

304: 

305: The earliest observations of an upwardly sloping stress plateau in

306: wormlike micelles were made by Schmitt \etal\ \cite{schmitt94} and

307: Berret \etal~\cite{berret94a,BRL98}. Schmitt \etal\ \cite{schmitt94}

308: studied CpClO$_3$/NaClO$_3$(0.05M)/H$_2$O at the high micellar volume

309: fraction $\phi\approx 31\%$, just below the onset of the I-N

310: transition at $\phil=34\%$.  

311: In the steady-state flow curve, the stress increased smoothly up to

312: the critical shear rate $\gdotl$, where it showed a pronounced

313: downward kink before curving upward again for $\gdot>\gdotl$

314: (qualitatively like B'F' in Fig.~\ref{fig:schem}). SANS measurements

315: confirmed a superposition of nematic and isotropic contributions in

316: this regime $\gdot>\gdotl$, with the nematic contribution rising

317: linearly from zero at $\gdot=\gdotl$.

318: 

319: Berret \etal~\cite{berret94a,BRL98} studied CpCl/hexanol/NaCl for

320: several micellar volume fractions, again at a volume fraction just

321: below the onset of the zero-shear I-N transition ($\phil\approx

322: 32\%$). The overall height of the coexistence plateau (which again

323: sloped upwards in $\gdot$) was found to fall with increasing

324: surfactant concentration $\phi\to\phil$, extrapolating to zero at

325: $\phi\gae\phil$, which is already biphasic in zero shear.  They also

326: found an increasing nematic contribution to SANS patterns for

327: increasing shear rates above $\gdotl$. They further used the SANS data

328: to show that the nematic (high shear) band was more concentrated than

329: the low shear band.

330: 

331: As noted above, the majority of existing calculations of shear-banded

332: states have assumed uniform concentration. An important exception is

333: the calculation of Olmsted and Lu~\cite{olmsted99c}. Although this

334: model was aimed at concentrated solutions of rigid rods, it broadly

335: captured some of the experimental phenomenology for the concentrated

336: ($\phi\approx 30\%$) wormlike

337: micelles~\cite{schmitt94,berret94a,BRL98}.  For example, the overall

338: height of the coexistence plateau $\sigmas$ increased from zero as the

339: concentration was reduced below the threshold $\phil$ of the

340: zero-shear I-N biphasic regime.  The coexistence plateau sloped upward

341: markedly in shear rate. In further agreement with experiment, the high

342: shear (nematic) phase had a higher volume fraction of rods.  It should

343: be noted that model of Ref.~\cite{olmsted99c} was explicitly aimed at

344: concentrated systems, which in zero shear are already close to the I-N

345: transition: hence, the dynamics of the relevant order parameter

346: $\tens{Q}$ was driven by a free energy that already contained a phase

347: transition. In contrast, the simple free energy $F_{\rm e}(\tens{W})$

348: we consider below has no underlying phase transition, and flow-induced

349: efects are driven by convective, rather than dissipative

350: (relaxational) dynamics.

351: 

352: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

353: \begin{figure}[h]

354:  \includegraphics[scale=0.35]{Berr94PDiagB.eps}

355: \caption{Height of the coexistence plateau in the system

356:   CpCl/NaSal/brine. The 5 leftmost points are taken from the data of

357:   Ref.~\cite{berret94b}. The righthand point represents the zero-shear

358:   biphasic regime of this system, and is in accordance with the

359:   extrapolation of $G(\phi)/\Sigma_{\rm sel}$ in Ref.~\cite{berret94b}

360:   (see main text for details).

361: \label{fig:coexistence} } 

362: \end{figure}

363: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

364: 

365: There have also been several experimental studies of concentration

366: dependence in the shear banding of more dilute wormlike micellar

367: solutions. For example, Berret \etal~\cite{berret94b} investigated the

368: non-linear rheology of CpCl/NaSal/brine in the concentration range

369: $5\%-20\%$, well below the I-N$_{\rm c}$ transition at $\phi\approx

370: 36\%$. (N$_{\rm c}$ is ``nematic calamitic''.) In contrast to the more

371: concentrated ``prenematic'' systems, the plateau height $\sigmas$

372: decreased with {\em decreasing} concentration: see the left 5 points

373: in Fig.~\ref{fig:coexistence}.  The width of the plateau also

374: decreased so that the difference $\gdot_{\rm h}-\gdot_\ell$ fell to

375: zero at a critical point $\phic,\gdotc,\Sigma_{\rm c}$ (leftmost point

376: in Fig.~\ref{fig:coexistence}). (These trends are the same as those in

377: Fig.~\ref{fig:alpha1.0e-3_gamma0.0:d} below.) In contrast the {\em

378:   scaled} plateau height $\sigmas/G(\phi)$ (where $G$ is the plateau

379: modulus) decreased with {\em increasing} concentration and

380: extrapolated to zero in the zero-shear biphasic (I-N) regime at

381: $\phi\gae\phil\approx 33\%$. According to this extrapolation (which is

382: actually well beyond the final data point at $\phi\approx 22\%$), and

383: in the absence of a divergence in $G(\phi)$, the unscaled plateau

384: height $\sigmas$ must itself fall to zero at $\phi\approx 33\%$

385: (rightmost data point in Fig.~\ref{fig:coexistence}), consistent with

386: the behaviour of the concentrated systems discussed above. To

387: summarise, the plateau height $\sigmas$ appears to be a non-monotonic

388: function of concentration, increasing with $\phi$ through the studied

389: regime $5\%<\phi<20\%$ before (probably) falling to zero in the

390: zero-shear biphasic $I-N$ regime $\phi\approx 33\%$.

391: 

392: Although this experiment showed that shear banding depends on the

393: {\em overall} concentration of the solution, there is relatively little

394: evidence for concentration-{\em coupling} (\ie\ concentration

395: differences between the bands) in such dilute systems, far from the

396: I-N transition.  Indeed, the experiment just described revealed no

397: discernible upward slope in the coexistence plateau. We are not aware

398: of any measurements of concentration differences between the

399: coexisting shear bands in such systems.  Nonetheless, recent

400: experiments on CTAB(0.3M)/${\rm NaNO}_3\rm {(1.79M)}$/H$_2$O at

401: $\phi=11\%$~\cite{LerDecBer00}

402: did reveal a stress plateau with slight upward slope.  Along with the

403: generic expectation that flow should be coupled to concentration in these

404: viscoelastic solutions, this suggests that an explicit calculation of

405: concentration difference in the shear bands of systems far from an I-N

406: transition might be worthwhile.

407: 

408: In this paper, therefore, we present the first such calculation, using

409: our \model\ model~\cite{FieOlm02,FieOlm02b}.  In contrast to the work

410: of Olmsted \etal\ \cite{olmsted99c} for concentrated rigid rods,

411: shear banding in the \model\ model is not due to any underlying nematic

412: feature of the elastic free energy $F^{\rm e}(\tens{W})$. Instead the

413: instability results mainly from the non-linear effects of shear (the

414: intrinsic constitutive curve has a region of negative slope), though

415: it can be strongly enhanced by concentration coupling in systems close

416: to an underlying Cahn-Hilliard (CH) demixing instability (governed by

417: the osmotic free energy $F^{\rm o}(\phi)$).  Indeed, the \model\ model

418: captures a broad crossover between (i) instabilities that are mainly

419: mechanical (governed by the negative slope of the flow curve) and (ii)

420: instabilities that are essentially CH demixing (governed by $F^{\rm

421:   o}(\phi)$), but now triggered by shear.  [Likewise, in practice

422: there should be no sharp distinction between concentrated micelles

423: with an underlying nematic feature in $F^{\rm e}(\tens{W})$ on the one

424: hand and ``non-nematic'' (more dilute) systems on the other: any more

425: refined model should allow a smooth crossover between the two cases.

426: This will be the focus of a future publication~\cite{FieOlm02f}.]

427: 

428: \section{Model}

429: \label{sec:model}

430: 

431: In this section we outline the \model\ model, which couples shear

432: banding instabilities to concentration in a simple way by combining

433: the non-local Johnson-Segalman (d-JS) model~\cite{johnson77} with a

434: 2-fluid framework~\cite{brochdgen77,milner93} for concentration

435: fluctuations. While this description is self-contained, readers

436: are referred to Ref.~\cite{FieOlm02b} for fuller details.

437: 

438: \subsection{Free energy}

439: 

440: In a sheared fluid, one cannot strictly define a free energy.

441: Nonetheless, for realistic shear rates, many internal degrees of

442: freedom of a polymeric solution relax quickly on the timescale of the

443: moving constraints and are therefore essentially equilibrated.

444: Integrating over these fast variables, one obtains a free energy for a

445: given fixed configuration of the slow variables. For our purposes, the

446: relevant slow variables are the fluid momentum and micellar

447: concentration $\phi$ (which are both conserved and therefore truly

448: slow in the hydrodynamic sense), and the micellar strain $\tens{W}$

449: that would have to be reversed in order to relax the micellar stress

450: (which is slow for all practical purposes):

451: %

452: \begin{equation}

453: {W}_{\alpha\beta}=\frac{\partial{R}'_{\alpha}}{\partial

454:   \vect{R}}\cdot\frac{\partial {R}'_{\beta}}{\partial

455:   \vect{R}}-\delta_{\alpha\beta}

456: \end{equation}

457: %

458: where $\delta\vect{R}'$ is the deformed vector corresponding to the

459: undeformed vector $\delta\vect{R}$. 

460: 

461: The resulting free energy is assumed to comprise separate 

462: osmotic and elastic components,

463: %

464: \begin{equation}

465: \label{eqn:free_energy_total}

466: %F=F^{\rm K}(\vect{v})+F^{\rm o}(\phi)+F^{\rm e}(\tens{W},\phi).

467: F=F^{\rm o}(\phi)+F^{\rm e}(\tens{W},\phi).

468: \end{equation}

469: %

470: %The kinetic component is

471: %%

472: %\be

473: %\label{eqn:kinetic_fe}

474: %F^{\rm K}(\vect{v})=\tfrac{1}{2}\int d^3x \rho\vect{v}^2,

475: %\end{equation}

476: %

477: %where $\rho$ is the fluid density and $\vect{v}$ the velocity.  The

478: The osmotic component is

479: %                           

480: \bea 

481: \label{eqn:free_energy}

482: F^{\rm o}(\phi)&=&\int d^3x \left[f(\phi)+\frac{g}{2}(\nablu \phi)^2\right]\nonumber\\

483: &\approx& \tfrac{1}{2} \int d^3q\, (1+\xi^2 q^2)f''|\phi(q)|^2,

484: \end{eqnarray}                             %

485: where $f''$ is the osmotic susceptibility and $\xi$ is the

486: equilibrium correlation length for concentration fluctuations.

487: The elastic component is

488: %

489: \begin{equation}

490: \label{eqn:elastic_fe}

491: F^{\rm e}(\tens{W},\phi)=\tfrac{1}{2}\int d^3x\,G(\phi) {\rm tr} \left[\tens{W}-\log(\tens{\delta}+\tens{W})\right]

492: \end{equation}

493: %

494: in which $G(\phi)$ is the micellar stretching modulus. 

495: 

496: \subsection{Dynamics}

497: 

498: The basic assumption of the two-fluid model is a separate force

499: balance for the micelles (velocity $\vm$; volume fraction $\phi$) and

500: the solvent (velocity $\vs$) within any element of solution.  

501: 

502: 

503: The micellar force balance equation is:

504: %

505: \bw

506: \begin{equation}

507: \label{eqn:micelle}

508: \rho_{\rm m}\,\phi\,\left(\partial_t+\vm.\nablu\right)\vm=\nablu.G(\phi)\,\tens{W}-\phi\,\nablu \frac{\delta F(\phi)}{\delta\phi}+2\,\nablu.\,\phi\, \etam\, \Dm^{0} -\zeta(\phi)\,\vrel-\phi\nablu p.

509: \end{equation}

510: %

511: In this equation, $G(\phi)\tens{W}\equiv

512: 2G(\phi)\tens{W}.\tfrac{\delta F}{\delta\tens{W}}$ is the viscoelastic

513: micellar backbone stress due to deformation of the local molecular

514: strain, while the osmotic stress $\tfrac{\delta F^{\rm o}}{\delta

515:   \phi}$ results from direct monomeric interaction.  The Newtonian

516: stress $2\phi\, \etam\, \Dm^{0}$ describes fast micellar processes

517: (\eg\ Rouse modes) with $\Dm^{0}$ the traceless symmetric micellar

518: strain rate tensor.  The force $\zeta\vrel$, where $\vrel=\vm-\vs$,

519: impedes relative motion; $\zeta$ is the drag coefficient

520: (Eq.~\ref{eqn:relative}).  Incompressibility determines the pressure

521: $p$.

522: 

523: Likewise, the solvent force balance comprises the Newtonian viscous

524: stress, the drag force (equal and opposite to the drag on the

525: micelles), and the hydrostatic pressure:

526: %

527: \begin{equation}

528: \label{eqn:solvent}

529: \rho_{\rm s}(1-\phi)\left(\partial_t+\vs.\nablu\right)\vs=2\nablu.\,(1-\phi)\,

530: \etas\, \Ds^{0}+\zeta(\phi)\vrel-(1-\phi)\nablu p.

531: \end{equation}

532: %

533: %The solvent stress $2(1-\phi)\, \etas\, \Ds^{0}$ is small

534: %for these systems: disregarding it does not change our results.

535: 

536: Equations~\ref{eqn:micelle} and \ref{eqn:solvent} contain the basic

537: assumption of ``dynamical asymmetry'', \ie\ that the viscoelastic

538: stress acts only on the micelles and not on the solvent. Adding them,

539: and assuming equal mass densities $\rho_{\rm m}=\rho_{\rm

540:   s}\equiv\rho$, we obtain the overall force balance equation for the

541: centre of mass velocity, $\vect{v}=\phi\vm+(1-\phi)\vs$:

542: %

543: \begin{equation}

544: \rho \left(\partial_t+\vect{v}.\nablu\right)\vv\equiv D_t\vect{v}

545: =\nablu.G(\phi)\,\tens{W}-\phi\,\nablu\frac{\delta

546:   F(\phi)}{\delta\phi}+2\,\nablu.\,\phi\, \etam\,

547: \Dm^{0}+2\nablu.\,(1-\phi)\, \etas\, \Ds^{0} -\nablu p.    

548: \label{eqn:navier}

549: \end{equation}

550: %

551: Subtracting them (with each predivided by its own volume fraction), we

552: obtain an expression for the relative motion $\vrel=\vm-\vs$, which in

553: turn specifies the concentration fluctuations:

554: %

555: \begin{equation}

556: D_t\phi=-\nablu\cdot\phi(1-\phi)

557: \vrel=-\nablu\cdot\frac{\phi^2(1-\phi)^2}{\zeta(\phi)}\left[\frac{\nablu

558:     \cdot G(\phi)\tens{W}}{\phi}-\nablu\frac{\delta F}{\delta

559:     \phi}+\frac{2\,\nablu\cdot\,\phi\, \etam\,

560:     \Dm^{0}}{\phi}-\frac{2\nablu\cdot\,(1-\phi)\, \etas\,

561:     \Ds^{0}}{1-\phi}\right]

562: \label{eqn:relative}

563: \end{equation} 

564: %

565: which defines the micellar diffusion coefficient $D\equiv

566: f''(\phi)\phi^2(1-\phi)^2/\zeta$.  We have omitted negligible inertial

567: corrections to Eqs.~(\ref{eqn:navier}) and (\ref{eqn:relative})

568: \cite{FieOlm02b}.

569: 

570: %

571: The essence of the 2-fluid model is that the physically distinct

572: elastic and osmotic stresses appear together in the force-balance

573: equation~(\ref{eqn:navier}) and also in the generalised CH

574: equation~(\ref{eqn:relative}). This allows micellar diffusion in

575: response to gradients in concentration {\em and} in the viscoelastic

576: stress. We will see below that this gives rise to a positive HF

577: feedback between concentration and flow~\cite{HelfFred89}, allowing

578: micelles to diffuse up their own concentration gradient.

579: 

580: 

581: 

582: For the dynamics of the viscoelastic micellar backbone strain we use

583: the phenomenological d-JS model~\cite{johnson77,olmsted99a}: 

584: %

585: \begin{equation}

586: \label{eqn:JSd}

587: (\partial_t+\vm.\nablu)\tens{W}=a(\Dm.\tens{W}+\tens{W}.\Dm)+(\tens{W}.\Omm-\Omm.\tens{W})+2\Dm-\frac{\tens{W}}{\tau(\phi)}+\frac{l^2}{\tau(\phi)} \nablu^2 \tens{W},

588: \end{equation}

589: \ew

590: %

591: where $2\Omm=\nablu \vm - (\nablu \vm)^T$ with $(\nablu

592: \vm)_{\alpha\beta}\equiv \partial_{\alpha}(v_{\rm m})_\beta$.

593: $\tau(\phi)$ is the Maxwell time and $l$ is a length scale discussed

594: in Sec.~\ref{sec:interfaces} below. The slip parameter $a$

595: measures the non-affinity of the molecular deformation, \ie\ the

596: fractional stretch of the polymeric material with respect to that of

597: the flow field. For $|a|<1$ (slip) the intrinsic constitutive curve in

598: planar shear is capable of the non-monotonicity of

599: Fig.~\ref{fig:schem}.

600: 

601: We use Eqns.~\ref{eqn:navier},~\ref{eqn:relative} and~\ref{eqn:JSd},

602: together with the incompressibility condition, $\nablu.\vect{v}=0$, as

603: our model for the remainder of the paper.

604: 

605: \subsection{Flow geometry. Boundary conditions}

606: \label{sec:geometry}

607: 

608: We consider idealised planar shear bounded by infinite plates at

609: $y=\{0,L\}$ with $(\vect{v},\nablu v, \nablu \wedge \vect{v})$ in the

610: $(\hat{\vect{x}},\hat{\vect{y}},\hat{\vect{z}})$ directions.  We allow

611: variations only in the flow-gradient direction, and therefore set all

612: other derivatives to zero: $\partial_x\ldots=0$, $\partial_z\ldots=0$.

613: In appendix~\ref{app:fulleqns} we give all the relevant components of

614: the model equations~\ref{eqn:navier},~\ref{eqn:relative}

615: and~\ref{eqn:JSd} in this coordinate system.

616: 

617: The boundary conditions at the plates are as follows.  For the

618: velocity we assume there is no slip.  For the concentration we assume

619: %

620: \begin{equation}

621: \label{eqn:BCconc}

622: \partial_y \phi=\partial^3_y \phi=0,

623: \end{equation}

624: %

625: which ensures (in zero shear at least) zero flux of concentration at

626: the boundaries.  

627: %

628: Following Ref.~\cite{olmsted99a}, for the micellar strain we assume  

629: %

630: \begin{equation}

631: \label{eqn:BCstrain}

632: \partial_y W_{\alpha\beta}=0\;\forall\; \alpha,\beta.

633: \end{equation}

634: %

635: Conditions~\ref{eqn:BCconc} and~\ref{eqn:BCstrain} together ensure

636: zero concentration flux at the boundary even in shear.  For the

637: controlled shear rate conditions assumed throughout,

638: %

639: \begin{equation}

640: \label{eqn:constant_strain_rate}

641: \bar{\gdot}=\int_0^L dy \gdot(y)={\rm constant.}

642: \end{equation}

643: %

644: 

645: \subsection{The interfacial terms}

646: \label{sec:interfaces}

647: 

648: The model contains two different interfacial terms. The first is the

649: gradient term on the RHS of Eqn.~(\ref{eqn:JSd}).  The length $l$ in

650: this term could, for example, be set by the mesh size or by the

651: equilibrium correlation length for concentration fluctuations. Here we

652: assume the former, since the dynamics of the micellar conformation are

653: more likely to depend on gradients in molecular conformation than in

654: concentration.  Physically, one can interpret the gradient term in

655: equation~\ref{eqn:JSd} as resulting dynamically, from the diffusion of

656: stretched molecules across the interface~\cite{elkareh89}, or

657: statically, from nematic interactions between the micelles, or both.

658: There is, at present, no accepted theory for these gradient terms in

659: semi-dilute solutions.  The equilibrium correlation length $\xi$ of

660: course still enters our analysis through our second interfacial term,

661: in the osmotic free energy of Eqn.~\ref{eqn:free_energy}.

662: 

663: Together, $l$ and $\xi$ set the length scale of any interfaces.

664: Throughout this paper, we study the physical limit in which $l$ and

665: $\xi$ are small compared to the system size so that we have a sharp

666: interface connecting two bulk homogeneous phases.  In this case, the

667: solution to Eqns.~\ref{eqn:navier},~\ref{eqn:relative}

668: and~\ref{eqn:JSd} naturally fits the zero-gradient boundary

669: conditions, and is invariant under $y\to y/2$, $l\to l/2$ and $\xi\to

670: \xi/2$. Therefore, a simultaneous reduction in $l$ and $\xi$ by the

671: same factor only changes the overall length of the interface, and not

672: the values of the order parameters in each phase (which determines

673: the phase diagram). However the phase diagram does depend slightly on

674: the ratio $r=l/\xi$: below we will give results for $r=0,r=\infty$ and

675: $r=O(1)$.  This provides a concrete example of the early insight of Lu

676: and co-workers~\cite{lu99}, that the banded state must depend on the

677: nature of the interfacial terms.  This contrasts notably with

678: equilibrium phase coexistence, in which the dynamical equations are

679: integrable and therefore insensitive to interfaces.

680: 

681: \subsection{Model parameters}

682: \label{sec:parameters}

683: 

684: \begin{table}

685:   \begin{center}

686:     \begin{tabular}{|p{2.85cm}|c|c|c|}

687:      \hline

688:      {\small\bf Parameter} & {\small \bf Symbol} $Q$ & {\small \bf Value at $\phi=0.11$} & $\frac{d\log Q}{d\log\phi}$ \\

689:      \hline\hline

690:      {\small Rheometer gap} & $L$ & $0.15\mbox{\,mm}$  & 0 \\

691:      \hline

692:      {\small Maxwell time} & $\tau$ & $0.17\mbox{\,s}$ & 1.1 \\

693:      \hline

694:      {\small Plateau modulus} & $G$ & $232\mbox{\,Pa}$ & 2.2 \\

695:      \hline

696:      {\small Density} & $\rho$ & $10^3\mbox{\,kg\,m}^{-3}$ & 0 \\

697:      \hline

698:      {\small Solvent viscosity} & $\etas$ & $10^{-3}\mbox{\,kg\,m}^{-1}\mbox{s}^{-1}$ & 0 \\

699:      \hline

700:      {\small Rouse viscosity} & $\etam$ & $0.4\mbox{\,kg\,m}^{-1}\mbox{s}^{-1}$ & 0 \\

701:      \hline

702:      {\small Mesh size} & $l$ & 2.6$\times 10^{-8}\mbox{m}$ & -0.73 \\

703:      \hline

704:      {\small Diffusion coefficient} & $D$ & $3.5 \times 10^{-11}\mbox{m}^2\mbox{s}^{-1}$ &0.77 \\

705:      \hline

706:      {\small Drag coefficient} & $\zeta$ & $2.4\times10^{12}\mbox{kg\,m}^{-3}\mbox{\,s}^{-1}$ & 1.54 \\

707:      \hline

708:      {\small Correlation length} & $\xi$ & $6.0\times10^{-7}$m & -0.77 \\

709:      \hline

710:      {\small Slip parameter} & $a$ & $0.92$ & 0 \\

711:      \hline

712:     \end{tabular}

713:   \end{center}

714:  \caption{Experimental values of the model's parameters at volume

715:    fraction $\phi=0.11$ (column 3). Scaling laws for the dependence of

716:    each parameter upon $\phi$ (column 4). In most calculations we use

717:    the reference values of column 3 at $\phi=0.11$, then tune $\phi$

718:    using the scaling laws of column 4. Only where  stated do we allow

719:    the parameters to vary independently.  \label{table:parameters}} 

720: \end{table}

721: 

722: The \model\ model

723: (Eqns.~\ref{eqn:navier},~\ref{eqn:relative},~\ref{eqn:JSd}) has the

724: following parameters: the solvent viscosity $\etas$ and density

725: $\rho$; the plateau modulus $G$; the Maxwell time $\tau$; the Rouse

726: viscosity $\etam$; the mesh size $l$; the osmotic modulus $f''(\phi)$

727: and the equilibrium correlation length $\xi$ (recall

728: Eqn.~\ref{eqn:free_energy}); the drag coefficient $\zeta$ and the slip

729: parameter $a$.  We also need to know the typical rheometer gap, $L$.

730: A reference set of parameter values at $\phi=0.11$ is summarised in

731: table~\ref{table:parameters}.  These values were taken from experiment

732: or calculated using scaling arguments: see Ref.~\cite{FieOlm02b} for

733: details.  Note that explicit data is not available for $f''(\phi)$;

734: however dynamic light scattering gives the diffusion coefficient

735: \begin{equation}

736:   \label{eq:1}

737:   D\equiv \frac{f''(\phi)\phi^2(1-\phi)^2}{\zeta(\phi)}.

738: \end{equation}

739: In this paper we will be guided by these parameter values, but subject

740: to the following considerations.

741: 

742: First, we are only interested in steady states so for convenience can

743: take the limit of zero Reynolds number ($\rho=0$) and rescale the

744: kinetic coefficient $1/\zeta$ so that the diffusion time $L^2/D$ is of

745: order the Maxwell time.  These choices have no effect on the steady

746: state, but make our numerical calculation of it much more efficient

747: (by evolving the {\em dynamical}

748: equations~\ref{eqn:navier},~\ref{eqn:relative} and~\ref{eqn:JSd}).

749: Second, realistic interfaces are much narrower than the typical

750: rheometer gap, with $l$ and $\xi$ both of $O(10^{-4}L)$.  To resolve

751: such interfaces (allowing a minimal 10 numerical mesh points per

752: interface) would therefore require $O(10^5)$ grid points, while in

753: practice we are limited to $O(10^2)$.  We will therefore use

754: artificially large values of $l$ and $\xi$.  However this does not

755: affect the phase diagram, provided the interface is still small

756: compared with the gap size: see Sec.~\ref{sec:numerics} for details,

757: and Fig. \ref{fig:finiteSizeBeta} for a typical banded profile.

758: Finally, we artificially increase the Rouse viscosity $\etam$ by a

759: factor $50$ to ensure, again for numerical convenience, that the shear

760: rate of the high shear phase is not too large. This does

761: quantitatively change the phase diagram, but we checked that the

762: qualitative trends are not affected.

763: 

764: Exploring this large parameter space is a daunting prospect so we

765: shall not, in general, vary the parameters independently of each

766: other. Instead we simply tune the concentration $\phi$, relying on

767: known semi-dilute scaling laws for the $\phi$-dependence of the other

768: parameters (column 4 of table~\ref{table:parameters}).  However we

769: will, in separate $\phi-$sweeps, vary the degree of concentration

770: coupling, which is dictated by ratio of the elastic term

771: $\nablu.G(\phi)\tens{W}$ to the osmotic term $\nablu\frac{\delta

772:   F}{\delta \phi}$ and which we encode in the parameter

773: \begin{equation}

774:   \label{eq:2}

775:   \alpha\equiv \frac{G'(\phi=0.11)}{2f''(\phi=0.11)}

776: \end{equation}

777: (where a prime denotes a derivative).  In other $\phi-$sweeps we will

778: vary the characteristic interface widths $l(\phi=0.11)$ and

779: $\xi(\phi=0.11)$, to investigate any dependence of the phase diagram

780: on the ratio $l/\xi$ in the double limit $l/L\to 0$, $\xi/L\to 0$.  In

781: what follows, we adopt the convenient shorthand of $l$ for

782: $l(\phi=0.11)$ with the understanding that $l$ does actually varies

783: with $\phi$ according to the scaling given in

784: table~\ref{table:parameters}. We do likewise for $\xi$.

785: 

786: Throughout we rescale stress, time and length so that

787: $G(\phi=0.11)=1$, $\tau(\phi=0.11)=1$, and $L=1$.

788: 

789: \section{Intrinsic flow curves; spinodals}

790: \label{sec:spinodals}

791: 

792: %

793: \begin{figure}[tb]

794:   \includegraphics[scale=0.3]{spinodals.eps}

795: \caption{Intrinsic flow curves (dotted lines) for

796:   $\phi=0.11,0.10\ldots0.01$ (downwards). Spinodals for concentration

797:   couplings $\alpha=10^{-2}, 10^{-3}, 10^{-4}$. 

798: \label{fig:spinodals}}

799: \end{figure}

800: %

801: 

802: The homogeneous intrinsic steady state flow curves

803: $\Sigma(\gdotb,\phib)=G(\phib)W_{xy}+\eta(\phib)\gdotb$ that satisfy

804: $\partial_t\vv=\partial_t\phi=\partial_t\tens{W}=0$ are shown as

805: dotted lines in Fig.~\ref{fig:spinodals}. (The average viscosity

806: $\eta(\phi)\equiv\phi\etam+(1-\phi)\etas$.)  The region of negative

807: slope ends at a ``critical'' point $\phib_{\rm c}\approx 0.015$.

808: CPCl/NaSal in brine~\cite{berret94b} shows the same trend. For

809: completeness, in App.~\ref{app:homogen} we give analytical results for

810: the steady state conditions in homogeneous shear flow.

811: 

812: In Ref.~\cite{FieOlm02,FieOlm02b} we linearised in fluctuations about

813: these homogeneous states to find the spinodal region in which the

814: homogeneous states are unstable.  The spinodals are shown in

815: Fig.~\ref{fig:spinodals} for different levels concentration coupling,

816: $\alpha$.  

817: 

818: In the limit of zero concentration coupling $\alpha\to 0$,

819: fluctuations in the ``mechanical variables'', $\tens{W}$ and $\gdot$

820: decouple from those in concentration, and are unstable in the region

821: of negative constitutive slope, as expected. Separately, the

822: concentration could have its own Cahn-Hilliard demixing instability,

823: when the diffusion coefficient $D<0$; however we are interested only

824: in flow-induced instabilities and set $D>0$ throughout.  For finite

825: $\alpha>0$, the region of mechanical instability is broadened by

826: coupling to the concentration fluctuations, as seen in

827: Fig.~\ref{fig:spinodals}. This can be understood as follows.  Consider

828: the first term in the square brackets of Eqn.(\ref{eqn:relative}).

829: This causes micelles to move up gradients in the viscoelastic stress

830: $\tens{W}$, thereby increasing the concentration in stressed regions.

831: If $G'(\phi)>0$ (assumed here), the increased concentration causes the

832: stress to increase further, closing a positive HF~\cite{HelfFred89}

833: feedback loop whereby the micelles can diffuse {\em up} their own

834: concentration gradient.

835: 

836: \section{Numerical details}

837: \label{sec:numerics}

838: 

839: In this section, we outline our numerical procedure for solving the

840: dynamical equations~\ref{eqn:navier},~\ref{eqn:relative}

841: and~\ref{eqn:JSd} and discuss our careful study of time-step, mesh

842: size and finite size effects. Readers who are not interested in these

843: issues can skip this section.

844: 

845: %

846: \begin{figure}[htb]

847:   \includegraphics[scale=0.3]{finiteSizeBeta.eps}

848:   \caption{Main figure: steady banded concentration profile  at

849:     $\alpha=10^{-2}$, $\xi=0$, $\gdotb=7.0$, $\phib=0.15$. Solid lines

850:     $l=0.016$, for $(N_y,\dt)=(100,0.05),(200,0.0125),(400,0.0125)$,

851:     dotted lines $l=0.008$ for $(N_y,\dt)=(200,0.05),(400,0.003125)$,

852:     dashed lines $l=0.004$ for $(N_y,\dt)=(400,0.05),(800,0.05)$.

853:     Upper inset, the same data, enlarged in the left hand phase

854:     (decreasing $\phi$ with increasing $N_y$). Lower inset:

855:     corresponding selected stresses, for the same parameter values and

856:     mesh sizes (decreasing stress with increasing $N_y$).

857: \label{fig:finiteSizeBeta} } 

858: \end{figure}

859: %

860: 

861: We consider variations only in the flow gradient direction, in which we

862: discretise $y\in {0,1}$ on an algebraic grid $y_n=n/N_y$ for

863: $n=0,1\ldots N_y$.  We stored $\phi$ and $\tens{W}$ on these grid points.

864: The velocities $\vm$ and $\vs$ were stored on half grid points

865: $y_{n+1/2}$, and we used linear interpolation between the half and

866: full grid points.  Likewise we discretized time such that $t_n=n\Delta

867: t$.  We evolved the discretized

868: equations~\ref{eqn:navier},~\ref{eqn:relative} and~\ref{eqn:JSd} using

869: the Crank-Nicholson algorithm which is semi-implicit in time, with

870: centred space derivatives

871: 

872: %\bw

873: \begin{figure*}

874:   \centering \subfigure[Tie lines (solid); spinodal (dashed).]{

875:     \includegraphics[scale=0.33]{sigmaVsPhi_alpha1p0e-3_gamma0p0.eps}

876: \label{fig:alpha1.0e-3_gamma0.0:a}

877: }

878: \subfigure[Tie lines (solid); spinodal (dashed); homogeneous

879: constitutive curve (dotted); for $l=0.008$. ]{ 

880: \includegraphics[scale=0.33]{sigmaVsgdot_alpha1p0e-3_gamma0p0.eps}

881: \label{fig:alpha1.0e-3_gamma0.0:b}

882: } \subfigure[Tie lines (solid); spinodal (dashed) for $l=0.008$.]{

883: \includegraphics[scale=0.33]{gdotVsPhi_alpha1p0e-3_gamma0p0.eps}

884: \label{fig:alpha1.0e-3_gamma0.0:c}

885: } \subfigure[Coexistence regime of the macroscopic flow curves (solid);

886: spinodal (dashed); homogeneous constitutive curves (dotted)]{

887:   \includegraphics[scale=0.3]{fc_alpha1p0e-3_gamma0p0_new.eps}

888: \label{fig:alpha1.0e-3_gamma0.0:d}

889: }

890: \caption{Phase diagrams and flow curves for $\alpha=10^{-2}$,

891:   $\xi=0.0$ for small $l/L$. (Recall that $l$ is actually a function

892:   of $\phi$: we are using the convenient shorthand of $l$ for the

893:   value $l(\phi=0.11)$.) (a) Thin (upper) solid lines: tie lines for

894:   $l=0.016,\ny=100,\dt=0.05$. Thick (lower) solid lines: tie lines for

895:   $l=0.008,\ny=200,\dt=0.05$. As described in the main text, we

896:   actually rescaled $l$ in the successive runs of each $\phib-$sweep

897:   (\ie\ as $\phib$ was tracked from $0.15$ down to $\phic$) so that

898:   the interfacial width remained (approximately) constant throughout

899:   the sweep: the value of $l$ in the figure legends refers to the

900:   value used in the first run of the sweep, at $\phib=0.15$. (b,c)

901:   Solid lines: tie lines repeated in the $(\Sigma,\gdot)$,

902:   $(\Sigma,\phi)$ representations for $l=0.008,\ny=200,\dt=0.05$. (d)

903:   Solid lines: macroscopic flow curves for

904:   $\phib=0.11,0.10,\ldots0.04$ (downward). These flow curves were

905:   recontructed from the tie lines of the phase diagrams (using the tie

906:   lines shown in this figure, and some additional ones). Because we

907:   have only calculated tie lines for discrete values of $\Sigma$, in

908:   some cases the reconstructed flow curves stop short of the

909:   single-phase region, and have been continued by eye with a dashed

910:   line. The inset in (d) shows the same data, but on a log-log plot.

911:   The experimentally observed slope $0.3$ is marked as a dot-dashed

912:   line for comparison.  The spinodal is shown in each of Figs a-d as a

913:   dashed line. In (b,d) the thin dotted lines are the intrinsic

914:   (homogeneous) constitutive curves for $\phib=0.11,0.1\ldots0.01$

915:   (downwards).}

916: \label{fig:alpha1.0e-3_gamma0.0}

917: \end{figure*}

918: %\ew

919: 

920: For each run, we seeded an initial profile that was either homogeneous

921: up to a small random contribution, or inhomogeneous according to

922: $\phi=\bar{\phi}[1+\Delta\cos(\pi x)]$ (with $\Delta\approx 0.1$).  We

923: then evolved the discretized

924: equations~\ref{eqn:navier},~\ref{eqn:relative} and~\ref{eqn:JSd} under

925: an imposed wall velocity until a steady banded state was reached. We

926: checked that the homogeneous phases between the interfaces were

927: insensitive to the initial conditions.  However, for the random

928: initial condition several bands could form (and did not coarsen over

929: any accessible timescale). Therefore in most runs we used the

930: co-sinusoidal initial profile, to conveniently obtain just two bands

931: (as in Fig.~\ref{fig:profiles_compared}).

932: 

933: For the dynamics to be independent of time-step, a very small

934: time-step has to be used. However the numerically-attained steady is

935: much less sensitive hence allowing much larger time-steps.  A typical

936: steady state changes by less than $10^{-3}\,\%$ for a factor-two

937: reduction in timestep \ref{fig:finiteSizeBeta}. For the special case

938: of $\xi=0$, timesteps $\Delta t \propto \ny^{-2}$ can be used, since

939: the highest spatial derivative is second order. For $\xi\neq 0$, we

940: have a fourth order derivative in Eqn.~\ref{eqn:relative} and much

941: smaller timesteps $\Delta t\propto \ny^{-4}$ must be used.

942: 

943: In all our calculations, we are interested in the physical limit where

944: the interface width is much smaller than the rheometer gap.  This

945: creates a delicate balance, since narrow interfaces require a very

946: fine grid. Therefore we adopted the following procedure.  For any

947: fixed value of the interfacial lengthscales $l$ and $\xi$, we

948: performed several runs with progressively finer meshes (but always

949: with a small enough time-step) until the shear banded profile and

950: selected stress didn't depend on the mesh. This is quite easy to

951: achieve: a typical steady state presented below changes by less than

952: $0.1\,\%$ upon doubling the number of mesh points. We then reduced $l$

953: and $\xi$ (in fixed ratio) until the order parameters in the

954: homogeneous phases changed by less than $0.5\%$ upon further halving

955: of $l$ and $\xi$ (but always ensuring convergence with respect to the

956: number of grid points).  A sample study of these issues is presented

957: in Fig.~\ref{fig:finiteSizeBeta} for the special case $\xi=0$.

958: 

959: 

960: \section{Results}

961: \label{sec:results}

962: 

963: We now present our results for the steady-state flow phase diagrams,

964: flow curves and shear banded profiles. Because one of our aims is to

965: show that the shear-banded state depends on the nature of the

966: interfacial terms, we consider three separate cases: (A) interfacial

967: terms only in the viscoelastic constitutive equation~\ref{eqn:JSd}

968: ($l\neq 0,\,\xi=0$, $r\equiv l/\xi=\infty$); (B) interfacial terms in

969: both the constitutive and concentration equations,~\ref{eqn:JSd}

970: and~\ref{eqn:relative} ($l\neq 0,\xi \neq 0$, $r=O(1)$); and (C)

971: interfacial terms only in the concentration

972: equation~\ref{eqn:relative} ($l=0,\xi\neq 0,r=0$).

973: 

974: \subsection{Interfacial terms only in the viscoelastic constitutive

975:   equation: $l\neq 0$, $\xi=0$.} 

976: \label{sec:onlyvisco}

977: 

978: In this section, we set the correlation length for concentration

979: fluctuations, $\xi$, to zero and consider small but non-zero values

980: the interfacial lengthscale $l$ in the constitutive

981: equation~\ref{eqn:JSd}.

982: 

983: \subsubsection{Flow phase diagrams}

984: 

985: \begin{figure}[htb]

986: \includegraphics[scale=0.3]{sigmaVsPhi_alpha1p0e-345_gamma0p0.eps}

987: \caption{Phase diagrams for three different degrees of coupling to

988:   concentration for $\xi=0$ and small $l/L$.

989: \label{fig:sigmaVsPhi_alpha1.0e-345_gamma0.0} } 

990: \end{figure}

991: 

992: For any given shear banded profile, the values of the order parameters

993: in each of the two homogeneous phases specify the two ends of one tie line

994: in the phase diagram. Analogously to equilibrium tie lines, the

995: concentrations and strain rates of the coexisting states are related

996: to the mean strain rate $\bar{\dot{\gamma}}$ and mean concentration

997: $\bar{\phi}$ by the lever rule,

998: \begin{align}

999:   \bar{\phi} &= \beta\phi_1 + (1-\beta)\phi_2 \\

1000:   \bar{\dot{\gamma}} &= \beta\dot{\gamma}_1 + (1-\beta)\dot{\gamma}_2,

1001: \end{align}

1002: where $\beta$ is the volume fraction of material in state $(\phi_1,

1003: \dot{\gamma}_1)$.  For each of several values of the concentration

1004: coupling, $\alpha$, we calculated the full phase diagram via a

1005: succession of shear startup runs, all at the critical shear rate

1006: $\gdotc(\alpha)$ (determined from Fig.~\ref{fig:spinodals}), for

1007: average concentrations ranging from $\phib=0.15$ down to the critical

1008: value $\phic(\alpha)$.  For concentrations below the critical point

1009: the response of the system is smooth as a function of stress.  In our

1010: model, this arises because decreasing concentration reduces the

1011: viscosity of the low shear rate branch faster than it reduces the

1012: viscosity of high shear rate branch. Hence the stress maximum

1013: decreases with decreasing concentration, disappearing when the stress

1014: maximum vanishes.  Alternatively, in a more dilute system the plateau

1015: modulus and Maxwell time are both smaller, and one expects a smaller

1016: stress and higher strain rate at the onset of instability.

1017: 

1018: The results for $\alpha=10^{-2}$, which gives rather strong

1019: concentration coupling, are shown in

1020: Fig.~\ref{fig:alpha1.0e-3_gamma0.0}a,b,c.  Because the width,

1021: $\delta$, of the interface in the banded state is set by $l$, but with

1022: a prefactor that diverges at the critical point, in each successive

1023: run we rescaled $l$ so that $\delta$ remained (approximately) equal to

1024: its value ($\ll L$) in the first run at $\phib=0.15$.  We return below

1025: to study the divergence of $\delta/l$ at the critical point.

1026: 

1027: To illustrate the finite size considerations of

1028: Sec.~\ref{sec:numerics} (above), in

1029: Fig.~\ref{fig:alpha1.0e-3_gamma0.0:a} we show the tie lines obtained

1030: for two different (starting) values of $l$. All the results are

1031: converged with respect to mesh fineness and timestep (not explicitly

1032: shown), but the tie lines differ slightly between the two values of

1033: $l$. However all seem to be consistent with one given binodal line: we

1034: do not have any explanation for this apparent consistency.

1035: 

1036: To investigate the effect of reducing the coupling to concentration,

1037: we repeat the phase diagram for $\alpha=10^{-2}$ alongside that for

1038: $\alpha=10^{-3}$ and $\alpha=10^{-4}$ in

1039: Fig.~\ref{fig:sigmaVsPhi_alpha1.0e-345_gamma0.0}.  As expected, the

1040: concentration difference between the bands tends to zero as 

1041: $\alpha\rightarrow0$.

1042: %$\alpha\equiv G''(\phi=0.11)/2f''(\phi=0.11)$. 

1043: 

1044: \subsubsection{Flow curves}

1045: 

1046: So far, we have discussed the flow phase diagrams.  Measurement of

1047: these diagrams still presents an open challenge to experimentalists,

1048: due to the difficulty in measuring the concentration of micelles in

1049: each band (although SANS data has been used to estimate the bands'

1050: concentrations in systems near the I-N transition~\cite{BRL98}). In

1051: this section we discuss the macroscopic flow curves, which are

1052: relativly easily measured using conventional bulk rheology.  However

1053: it is important to realise that a set of flow curves

1054: $\Sigma(\gdotb,\phib)$ measured for {\em several} values of $\phib$

1055: actually contains the same information as the phase diagram:

1056: reconstruction of the latter from the former is described in

1057: Fig.~\ref{fig:reconstruct} A full set of flow curves could therefore

1058: be used to check measurements of concentration differences.

1059: 

1060: \begin{figure}[htb]

1061: \includegraphics[scale=0.3]{calced_fc_alpha_1p0e-3_gamma0p0.eps}

1062: \caption{Macroscopic flow curves for $\alpha=10^{-2}$ at

1063:   $\phib=0.11$. Thick solid line:  reconstructed from the tie lines of

1064:   the phase diagram. Dot dashed and dashed lines:  calculated by

1065:   directly measuring the average stress and strain rate during a

1066:   strain rate sweep for $l=0.016,\ny=100,\dt=0.05$ (dot-dashed) and

1067:   $l=0.008,\ny=200,\dt=0.05$  (dashed). (The slight discrepancy

1068:   between these three curves is discussed in the text.) The thin

1069:   dotted line is the intrinsic (homogeneous) constitutive curve. 

1070: \label{fig:calced_fc_alpha_1.0e-3_gamma0.0} } 

1071: \end{figure}

1072: 

1073: 

1074: \begin{figure}[htb]

1075: \includegraphics[scale=0.3]{fc_alpha1p0e-345_gamma0p0.eps}

1076: \caption{Macroscopic flow curves (from direct measurements of the

1077:   stress and strain rate) for three different degrees of coupling to

1078:   concentration. 

1079: \label{fig:fc_alpha1.0e-345_gamma0.0} } 

1080: \end{figure}

1081: 

1082: %\bw

1083: \begin{figure*}

1084: \subfigure[]{

1085:     \includegraphics[scale=0.19]{reconstruct1.eps}

1086: \label{fig:reconstruct:a}

1087: }

1088: \subfigure[]{

1089: \includegraphics[scale=0.8]{reconstruct2fig.eps}

1090: \label{fig:reconstruct:b}

1091: }

1092: \subfigure[]{

1093: \includegraphics[scale=0.8]{reconstruct3fig.eps}

1094: \label{fig:reconstruct:c}

1095: }

1096: \caption{Reconstruction of the flow phase diagram from a family of

1097:   macroscopic flow curves $\Sigma(\gdotb,\phib)$, measured for several

1098:   different average concentrations $\phib$.  Consider the flow curves

1099:   of Fig.~\ref{fig:reconstruct:a}. The curve that starts at A and ends

1100:   at B is for an average concentration $\phib=0.08$. Points A and B

1101:   are at the edge of the two-phase region.  Reading off the stress

1102:   from Fig.~\ref{fig:reconstruct:a}, A and B give use two points on

1103:   the binodal in Fig.~\ref{fig:reconstruct:b}.  Likewise reading off

1104:   the strain rate, we get points A and B in

1105:   Fig.~\ref{fig:reconstruct:c}. Repeating this for all the circles in

1106:   Fig.~\ref{fig:reconstruct:a}, we can construct many points on the

1107:   binodal in Figs.~\ref{fig:reconstruct:b} and

1108:   \ref{fig:reconstruct:c}, which can then be interpolated over to give

1109:   the full binodal. We now just need to specify the tie lines. In

1110:   Fig.~\ref{fig:reconstruct:b} this is trivial: all tie lines are

1111:   horizontal since the coexistence occurs at common stress (for

1112:   gradient banding). In Fig.~\ref{fig:reconstruct:c}, to get the slope

1113:   of the tie line that starts at $B$ we proceed by recalling that the

1114:   tie line represents constant shear stress. Therefore we find another

1115:   point, D, in Fig.~\ref{fig:reconstruct:a} that is at the same stress

1116:   as point $A$, and read off its average strain-rate. Its average

1117:   concentration is already known. This gives point $C$ in

1118:   Fig.~\ref{fig:reconstruct:c}. Similarly, $D$ is the image of point

1119:   $B$ at constant stress. Repeating this process we can fill in all

1120:   the tie lines of the phase diagram.  }

1121: \label{fig:reconstruct}

1122: \end{figure*}

1123: %\ew

1124: 

1125: In this work, we take the opposite approach for convenience, and

1126: reconstruct the steady-state flow curves from the tie lines of the

1127: phase diagram.  The results are shown in

1128: Fig.~\ref{fig:alpha1.0e-3_gamma0.0:d}.  The inset shows the same data

1129: on a log-log plot, to enable comparison with Ref.\cite{berret94a} in

1130: which the coexistence plateau in a log-log representation was a

1131: reasonably straight line (over the shear-rate range investigated) with

1132: slope $0.3$.  Note that the results shown in

1133: Fig.~\ref{fig:alpha1.0e-3_gamma0.0:d} are in units of $G(\phi=0.11)$

1134: and $\tau(\phi=0.11)$. In Ref.~\cite{BRP94}, Berret replotted the flow

1135: curves in units of $G(\phib)$ and $\tau(\phib)$, finding scaling

1136: collapse of the family $\Sigma(\gdotb,\phib)/G(\phib)$ \versus\ 

1137: $\gdotb\tau(\phib)$ in the low shear regime $\gdotb\to 0$. We do not

1138: find this scaling collapse

1139: (Fig.~\ref{fig:fc_alpha1.0e-3_gamma0.0_scaled}) because we have used

1140: an artificially large high-shear Newtonian contribution $\eta\gdot$

1141: for numerical convenience (recall Sec.~\ref{sec:parameters}): the

1142: overall zero shear viscosity, $G(\phi)\tau(\phi)+\eta(\phi)$ therefore

1143: does not scale as $G(\phi)\tau(\phi)$, even approximately.

1144: 

1145: \begin{figure}[htb]

1146: \includegraphics[scale=0.3]{fc_alpha1p0e-3_gamma0p0_scaled.eps}

1147: \caption{Macroscopic flow curves as shown in

1148:   Fig.~\ref{fig:alpha1.0e-3_gamma0.0:d} above, but now with the stress

1149:   in units of $G(\phi)$ and the strain rate in units of $\tau(\phi)$.

1150: \label{fig:fc_alpha1.0e-3_gamma0.0_scaled} } 

1151: \end{figure}

1152: 

1153: To check the reconstruction of flow curves from the phase diagram, we

1154: also explicitly calculated the flow curve at a single $\phib=0.11$. To

1155: do this, we first performed a shear startup at a given $\gdotb$ in the

1156: unstable region.  We then (without reinitialising the system)

1157: decreased $\gdotb$ in steps to the edge of the coexistence regime,

1158: ensuring that a steady state was reached before measuring the total

1159: stress.  We then reinitialised the system and repeated the entire

1160: procedure, but now with increasing $\gdotb-$jumps. The results are

1161: shown in \ref{fig:calced_fc_alpha_1.0e-3_gamma0.0} for two different

1162: values of $l$.  The slight discrepancy between the directly measured

1163: flow curve ``plateaus'' (i.e. the inhomogeneouse part of the flow

1164: curve) and those reconstructed from the tie lines is due to the finite

1165: size of the interface $\delta$ relative to the cell $L$, and so is

1166: smaller for the smaller value of $\delta/L$. The construction

1167: described in Fig.~\ref{fig:reconstruct} implicitly assumes that

1168: $\delta/L=0$.

1169: 

1170: 

1171: As expected for this value of $\alpha$ (which gives a large

1172: concentration difference between the bands;

1173: Fig.~\ref{fig:alpha1.0e-3_gamma0.0:a}), the steady state flow curve

1174: ``plateau'' slopes strongly upwards in $\gdotb$.  In

1175: Fig.~\ref{fig:fc_alpha1.0e-345_gamma0.0} we compare the (directly

1176: measured) macroscopic flow curve for the three levels of concentration

1177: coupling shown in Fig.~\ref{fig:sigmaVsPhi_alpha1.0e-345_gamma0.0}: as

1178: expected, the slope of the flow curve tends to zero with the degree of

1179: concentration coupling $\alpha$. 

1180: 

1181: The upturn in the measured flow curve at the edge of the coexistence

1182: plateau (apparent at the lower binodal for $\alpha=10^{-2}$ in

1183: Fig.~\ref{fig:fc_alpha1.0e-345_gamma0.0}) results again from the

1184: finite value of $\delta/L$: the interface bumps into the edge of the

1185: rheometer when one of the bands gets very narrow. We expect this

1186: (steady-state) effect to be much less pronounced in experimental

1187: systems, since realistic interfaces are much smaller than those used

1188: in our numerical study. Only near a critical point, where the

1189: interface becomes very broad (for fixed $l$), would we expect to see a

1190: true steady-state bump at the edge of the plateau.  Nonetheless,

1191: pronounced bumps are often apparent in data obtained via upward

1192: strain-rate sweeps.  However in most cases this is likely to be a

1193: metastable effect, so that the bump could be eliminated (or at least

1194: reduced) by reducing the rate of the sweep~\cite{grand97}.

1195: 

1196: As noted in Sec.~\ref{sec:intro}, in a curved Couette geometry the

1197: ``plateau'' (B'F' of Fig.~\ref{fig:schem}) in the flow curve will

1198: slope upwards due to the inhomogeneity of the stress field, even

1199: without concentration coupling. It should be noted that all

1200: calculations in this paper are for a planar shear geometry, and the

1201: slope of our flow curves in the coexistence regime results solely from

1202: concentration coupling. In fact, the slope in

1203: Fig.~\ref{fig:alpha1.0e-3_gamma0.0:d} is far greater than one would

1204: typically expect from curvature effects: for a Couette cell with

1205: radius $R$ and gap $\delta R$, the stress measured at the inner

1206: Couette wall would change by $\delta\Sigma/\Sigma=2\delta R/R$ over

1207: the coexistence regime, and so too would the relative change in torque

1208: through the coexistence ``plateau''.  The slope of

1209: Fig.~\ref{fig:alpha1.0e-3_gamma0.0:d} would therefore require an

1210: atypically large curvature of $\delta R /R\sim0.5$.

1211: 

1212: \subsubsection{Interfacial profiles; divergence of interface width at

1213:   the critical point} 

1214: 

1215: \begin{figure}[htb]

1216: \includegraphics[scale=0.4]{profiles_compared.eps}

1217: \caption{Steady state shear banded profile at $\alpha=10^{-2}$,

1218:   $\gdotb=4.64$, $\phib=0.15$ for two different ratios $r=l/\xi$. The

1219:   thick lines are for $r=\infty$ ($l=0.008$, $\xi=0.0$), $\ny=200$,

1220:   $\dt=0.05$, as considered in this section. The thin lines show the

1221:   corresponding results for   $r=0.4$ ($l=0.008$, $\xi=0.002$),

1222:   $\ny=200$, $\dt=0.00625$ (to be discussed in Sec.~\ref{sec:both}

1223:   below), for comparison.  

1224: \label{fig:profiles_compared} } 

1225: \end{figure}

1226: 

1227: \begin{figure}[htb]

1228: \includegraphics[scale=0.3]{widths_alpha1p0e-3_gamma0p0.eps}

1229: \caption{Scaled interface width $\delta/l$ versus the distance from the critical stress $\Sigma-\Sigma_{\rm c}$. The dotted line is a power $-0.5$. 

1230: \label{fig:widths_alpha1.0e-3_gamma0.0} } 

1231: \end{figure}

1232: 

1233: We now turn to the interfacial profiles and widths.  A full steady

1234: state banded profile for $\alpha=10^{-2}$ (corresponding to the

1235: rightmost/uppermost tie line in

1236: fig.~\ref{fig:alpha1.0e-3_gamma0.0}a,b,c) is shown by the thick lines

1237: in Fig.~\ref{fig:profiles_compared}. As required, the interface is

1238: smooth on the scale of the mesh, but sharp on the scale of the gap

1239: size, \ie\ $L/\ny\ll \delta \ll L\equiv 1$ where $\delta$ is the width

1240: of the interface. Note that the shear rate is negative across the gap

1241: since we have chosen to move the wall at $y=0$; accordingly we have

1242: plotted $-W_{xy}$, since $W_{xy}$ is antisymmetric in shear rate.

1243: $-W_{wy}$ is rather small in the high shear band, as expected from the

1244: underlying constitutive non-monotonicity. Meanwhile $W_{xx}$ is very

1245: large, while $W_{yy}\approx -0.5$ (recall that $\tens{W}$ measures

1246: deformation relative to the unit tensor $\tens{\delta}$): this

1247: corresponds to the micelles being highly stretched along the flow

1248: direction and is consistent with the experimental observation that the

1249: first normal stress difference progressively increases throughout the

1250: banding regime~\cite{rehage91}. The concentration is lower in the high

1251: shear band, where $W_{yy}$ is smaller (more negative): this is a

1252: direct result of the tendency of mielles to move up gradients in

1253: $W_{yy}$, as determined by Eqn.~\ref{eqn:relative} above.

1254: 

1255: In fact the interface width, $\delta$, is slightly different for each

1256: order parameter: we define it to be the distance between the two

1257: points where the change in that order parameter between the two

1258: homogeneous phases is $25\%$ and $75\%$ complete.  For a fixed value

1259: of $l$ (which sets the overall scale of the interface width), $\delta$

1260: diverges at the critical point (for each order parameter).  In

1261: tracking $\phib$ down towards the critical point, therefore, we

1262: continually rescaled $l$ to ensure that the interface width remained

1263: approximately constant. In each case, we measured $\delta/l$, for each

1264: of $W_{xy}$, $W_{xx}$, $W_{yy}$ and $\phi$: see

1265: Fig.~\ref{fig:widths_alpha1.0e-3_gamma0.0}.  According to mean field

1266: theory, the divergence should be of the form $\delta/l \sim

1267: (\Sigma-\Sigma_{\rm c})^{-1/2}$. The power $-1/2$ is accordingly shown

1268: in Fig.~\ref{fig:widths_alpha1.0e-3_gamma0.0} as a guide for the eye.

1269: 

1270: \subsection{Interfacial terms in both the viscoelastic constitutive

1271:   equation, and in the concentration equation: $l\neq 0$, $\xi\neq

1272:   0$.} 

1273: \label{sec:both}

1274: 

1275: %\afterpage{\clearpage}

1276: 

1277: %\bw

1278: \begin{figure}[htb]

1279:   \centering \subfigure[Tie lines]{

1280:     \includegraphics[scale=0.32]{gdotVsPhi_alpha1p0e-3_betaonly_and_both.eps}

1281:     \label{fig:alpha1.0e-3_betaonly_and_both:a}

1282:   } \subfigure[Partially reconstructed macroscopic flow curves

1283:   (solid); homogeneous constitutive curve (dotted).]

1284:   {\includegraphics[scale=0.32]{fc_alpha1p0e-3_betaonly_and_both.eps}

1285:     \label{fig:alpha1.0e-3_betaonly_and_both:b}

1286:     }

1287:   \caption{Phase diagrams and flow curves for $\alpha=10^{-2}$ and

1288:     $r=0.4$ ($l=0.008$, $\xi=0.02$) with the corresponding data for

1289:     $\alpha=10^{-2}$ and $r=\infty$ ($l=0.008$, $\xi=0$) for

1290:     comparison. } 

1291:   \label{fig:alpha1.0e-3_betaonly_and_both}

1292: \end{figure}

1293: %\ew

1294: 

1295: We now study the effect of including interfacial gradient terms in the

1296: concentration equation~\ref{eqn:relative} (so that now $\xi \neq 0$)

1297: as well as in the viscoelastic equation~\ref{eqn:JSd}, $l\neq 0$.

1298: Hence, while in the previous section we considered $r\equiv

1299: l/\xi=\infty$, then, we now consider $r=O(1)$.  In

1300: Fig.~\ref{fig:alpha1.0e-3_betaonly_and_both:a}, we give the phase

1301: diagram for $r=0.4$. Comparing it with our results for $r=\infty$

1302: (also shown in Fig.~\ref{fig:alpha1.0e-3_betaonly_and_both:a}), we see

1303: that the slopes of the tie lines and the overall binodal both depend

1304: quantitatively on $r$. [The difference between the results for

1305: $r=\infty$ and $r=0.4$ is far greater than any ``error'' associated

1306: with the fact that we are not quite in the limit $\dt\to 0$,

1307: $l\ny\to\infty $, $\xi\ny\to\infty $, $l/L\to 0$ and $\xi/L \to 0$.]

1308: This provides a concrete example of the fact that shear-banding

1309: coexistence is determined by, and non-universal with respect to, the

1310: interfacial terms~\cite{lu99}.  As noted above, this contrasts sharply

1311: with the equilibrium case, in which the equations of motion are

1312: integrable and so the phase diagram is independent of the interfacial

1313: terms.  Although conceptually important, this dependence is in

1314: practice rather weak: the overall features of the phase diagram are

1315: unchanged. The critical point is unaffected.

1316: 

1317: 

1318: In Fig.~\ref{fig:alpha1.0e-3_betaonly_and_both:b} we show the

1319: corresponding macroscopic flow curves, reconstructed using the tie

1320: lines of Fig.~\ref{fig:alpha1.0e-3_betaonly_and_both:a}. Because we

1321: only calculated a few tie lines in this case, the recontruction is

1322: rather sparse. Nonetheless, the slight difference between $r=\infty$

1323: and $r=0.4$ is apparent. 

1324: 

1325: In Fig.~\ref{fig:profiles_compared}, we compare a full banded profiles

1326: for $r=\infty$ and $r=0.4$. The slight dependence on $r$ is again

1327: apparent.

1328: 

1329: %\newpage 

1330: 

1331: \subsection{Interfacial gradient terms only in the concentration equation: $l=0$, $\xi\neq 0$.}

1332: \label{sec:onlyconc}

1333: 

1334: %\bw

1335: \begin{figure*}

1336:   \centering \subfigure[Tie lines, only shown for the case $l=r=0$

1337:   when close to the critical point]{

1338:     \includegraphics[scale=0.28]{gdotVsPhi_gammonly_and_both.eps}

1339: \label{fig:gdotVsPhi_gammonly_and_both}

1340: }

1341: \subfigure[Steady-state profiles for initial condition

1342: $\phi(y)=\phib+0.4\cos(\pi y)$: there is no selected smoothly banded

1343: state ]{ \includegraphics[scale=0.33]{profiles_gammonly.eps}

1344: \label{fig:profiles_gammonly}

1345: }\\

1346: \subfigure[Steady-state profiles for initial conditions

1347: $\phi(y)=\phib+0.4\cos(\pi y)$ and $\phi(y)=\phib+0.7\cos(\pi y)$: the

1348: steady state depends on the initial condition.]

1349: {

1350: \includegraphics[scale=0.25]{no_selection.eps}

1351: \label{fig:no_selection}

1352: }

1353: \caption{(a) Phase diagram at $\alpha=10^{-2}$, for $l=0.0$,

1354:   $\xi=0.02$, $r\equiv l/\xi=0.0$, shown with the corresponding data

1355:   for $l=0.008$, $\xi=0.02$, $r=0.4$ for comparison. Tie lines are

1356:   only shown near the critical point because for larger values of

1357:   $\phib$, there is no uniquely selected, smoothly banded state. This

1358:   is shown in Figs. b and c. In Fig.b the steady state profiles from

1359:   left to right at fixed ordinate correspond to tie lines left to

1360:   right in the upper Fig.(a). Fig (c) shows the steady state profile

1361:   in $\gdot$ (upper two curves) and in $\phi$ (lower two curves) for

1362:   $\phib=0.16$ and $\gdotb=4.66$ with initial condition

1363:   $\phi(y)=\phib+0.4\cos(\pi y)$ (solid lines) and with

1364:   $\phi(y)=\phib+0.7\cos(\pi y)$ (dashed lines): the ``selected''

1365:   state depends upon the initial condition -- \ie\ there is no state

1366:   selection for $l=0$ for stresses far enough above the critical

1367:   point.}

1368: \label{fig:alpha1.0e-3_gammonly}

1369: \end{figure*}

1370: %\ew

1371: 

1372: %\afterpage{\clearpage}

1373: 

1374: Finally we set the interfacial length $l$ in the constitutive equation

1375: equal to zero. The constitutive equation is now local, and the only

1376: source of spatial gradients is the equilibrium correlation length for

1377: concentration fluctuations (Eqns.~\ref{eqn:relative}

1378: and~\ref{eqn:free_energy}): $r\equiv l/\xi=0$. In the absence of

1379: concentration coupling, $\alpha=0$, it is known that there is no

1380: uniquely selected, smoothly shear banded state when $l=0$~\cite{lu99}.

1381: Here we investigate whether a smoothly banded state is selected for

1382: $\alpha\neq 0$, by virtue of the interfacial terms in the

1383: concentration equation.

1384: 

1385: Our numerics only gave a smoothly banded profile for stresses near the

1386: critical point, even for the largest accessible values of $\xi$ and

1387: $\ny$. The profiles shown from left to right in

1388: Fig.~\ref{fig:profiles_gammonly} are progressively further above the

1389: critical point. The tie lines correponding to the smooth profiles near

1390: the critical point are shown in

1391: Fig.~\ref{fig:gdotVsPhi_gammonly_and_both}, alongside the correponding

1392: results at $r=0.4$ for comparison.  Consistent with the discussion of

1393: non-universality in the previous section, the phase diagram for

1394: $r=0.0$ is slightly different from that for $r=0.4$ (and is different

1395: again from the case $r=\infty$; not shown).

1396: 

1397: For the spiky profiles, further from the critical point, the binodal

1398: of the associated tie lines is irregular (not shown in

1399: Fig.~\ref{fig:gdotVsPhi_gammonly_and_both}), suggesting that the

1400: steady state is not uniquely selected.  In view of this, a natural

1401: question is whether selection could occur in principle (but is

1402: inaccessible with any realistic mesh due to the pronounced

1403: non-monotonicity in $W_{xy}(y)$), or whether selection cannot occur,

1404: even in principle. In Fig.~\ref{fig:no_selection} we show that the

1405: steady state depends on the initial condition; so state selection

1406: appears to be lost when $l=0$. This numerical observation is backed up

1407: by the following analytical argument.

1408: 

1409: In steady state, the system must obey:

1410: %

1411: \begin{itemize}

1412: \item The force-balance equation,

1413: \begin{equation}

1414: \label{eqn:simplenavier}

1415: S(\gdot,\phi)\equiv G(\phi)W_{xy}[\gdot\tau(\phi)]+\etabar(\phi)\gdot=\Sigma={\rm const.}

1416: \end{equation}

1417: %

1418: \item The (now local) constitutive equation, equation~(\ref{eqn:JSd}), 

1419: \begin{equation}

1420: \label{eqn:solveW}

1421: W_{\alpha\beta}=W_{\alpha\beta}[\gdot \tau(\phi)] \;\;\mbox{for }\;\;\alpha\beta=xx,xy,yy.

1422: \end{equation}

1423: %

1424: \item The steady-state of equation~\ref{eqn:relative}. For the

1425:   purposes of this analytical argument we use a simplified version of

1426:   this equation, which we believe still captures the essential physics:

1427: %

1428: \begin{equation}

1429: \label{eqn:simpleconc}

1430: 0=\partial^2_y\left\{\Lambda-\partial_y^2\phi \right\}

1431: \end{equation}

1432: %

1433: with

1434: %

1435: \begin{equation}

1436: \label{eqn:defineF}

1437: \Lambda=f'(\phi)-\frac{G(\phi)W_{yy}[\gdot\tau(\phi)]}{\phi}.

1438: \end{equation}

1439: %

1440: %$F$ can be written as either a single-valued function of $\gdot$

1441: %(Fig.~\ref{fig:construction}b), or a multi-valued function of $\phi$

1442: %(Fig.~\ref{fig:construction}c).  

1443: Integrating Eqn.~\ref{eqn:simpleconc}

1444: twice, and using the boundary conditions

1445: $\partial_y\phi=0,\partial^3_y\phi=0$ for $y=0,L$, we obtain

1446: %

1447: \begin{equation}

1448: \label{eqn:third}

1449: \mu={\rm const.}=\Lambda-\partial_y^2\phi,

1450: \end{equation}

1451: where $\mu$ is an integration constant.

1452: %

1453: \end{itemize}

1454: %

1455: We now show that a solution satisfying Eqns.~\ref{eqn:simplenavier}

1456: and~\ref{eqn:solveW} cannot in general simultaneously satisfy

1457: Eqn.~\ref{eqn:third}.

1458: 

1459: %

1460: \begin{figure*}[htb]

1461: \includegraphics[scale=0.6]{construction1.eps}

1462: \caption{a) Dotted lines: relation between $\phi$ and $\gdot$ for the

1463:   case of a local constitutive equation, for several values of the

1464:   shear stress, $\Sigma$, calculated using

1465:   Eqns.~(\ref{eqn:simplenavier}) and~(\ref{eqn:solveW}). Solid lines:

1466:   the results of our numerics, showing that the $\phi-$ profile cannot

1467:   properly negotiate the interface, as described in the main text.

1468:   (Note that the stresses used to generate the dotted and the solid

1469:   lines differ slightly, but the overall trend is still clear.) b) The

1470:   function $\Lambda$ of Eqn.~\ref{eqn:defineF} plotted \versus\ 

1471:   $\gdot$ using the relation of Fig.a). c) $\Lambda$ replotted

1472:   \versus\ $\phi$.

1473: \label{fig:construction} } 

1474: \end{figure*}

1475: %

1476: Consider firstly Eqns.~\ref{eqn:simplenavier} and~\ref{eqn:solveW}.

1477: Sustituting $W_{xy}$ from Eqn.~\ref{eqn:solveW} into

1478: Eqn.~\ref{eqn:simplenavier}, we obtain an expression for

1479: $S(\gdot,\phi)$: this is just the family of homogeneous constitutive

1480: curves, as plotted in Fig.~\ref{fig:spinodals} above.  Because the

1481: constitutive equation is local, the solution at all points across the

1482: rheometer cell must lie on one of these intrinsic constitutive curves.

1483: Indeed, as the shear rate changes across the interface, the system

1484: must pass through constitutive curves of differing concentrations to

1485: maintain a uniform stress $\Sigma$. 

1486: %($\partial_y\Sigma=0$ by force balance.) 

1487: In other words, a relation $\phi=\phi(\gdot,\Sigma)$ must be

1488: obeyed. The family of these curves is shown as dotted lines in

1489: Fig.~\ref{fig:construction}a.

1490: For the range of stresses at which $\phi(\gdot,\Sigma)$ is

1491: non-monotonic, $\phi$ must have the form shown in

1492: Fig.~\ref{fig:proof}b in which the derivative $\partial^2_y\phi$

1493: changes sign three times across the interface, as in

1494: Fig.~\ref{fig:proof}c.  [Actually, the forms of

1495: Fig.~\ref{fig:proof}b,c assume that the profile in $\gdot$ increases

1496: monotonically through the interface (Fig.~\ref{fig:proof}a). However

1497: this monotonicity will emerge self consistently from our argument

1498: below.]

1499: 

1500: However we know from Eqn.~\ref{eqn:third} that

1501: $\partial_y^2\phi=\Lambda-\mu$.  $\Lambda$ is plotted in

1502: Fig.~\ref{fig:construction}b,c using Eqn.~\ref{eqn:defineF} together

1503: with the constraint $\phi=\phi(\gdot,\Sigma)$ (imposed from

1504: Eqns.~\ref{eqn:simplenavier} and~\ref{eqn:solveW}, as discussed

1505: above). From this plot we see that, for any $\mu$, a solution that

1506: starts and ends in homogeneous phases (for which

1507: $\partial_y^2\phi=\Lambda-\mu=0$)~\footnote{While this boundary

1508:   condition is intuitively clear (the homogeneous phases are in

1509:   general large compared with the interface), in our numerics we only

1510:   actually imposed the boundary condtion $\partial_y\phi=0$.  We

1511:   therefore did not, a priori, ``overspecify'' the differential

1512:   equation~\ref{eqn:third}. However we shall show below that

1513:   $\partial^2_y\phi$ must {\em automatically} equal zero at each

1514:   interface.}  can only involve at most one sign change of

1515: $\partial_y^2\phi$ between the boundaries.  This inconsistency with

1516: Fig.~\ref{fig:proof}c means that a steady banded solution cannot exist

1517: for these stress values for which $\phi(\gdot,\Sigma)$ is

1518: non-monotonic. To summarize: for stresses far enough above the

1519: critical point that $\phi(\gdot,\Sigma)$ is non-monotonic, a steady

1520: state solution cannot simultaneously satisfy

1521: Eqns.~\ref{eqn:simplenavier} and~\ref{eqn:solveW} (which imply three

1522: sign changes of $\partial_y^2\phi$) at the same time as

1523: Eqn.~\ref{eqn:third} (which only allows one sign change). Therefore

1524: there a steady, smoothly banded profile cannot exist for such

1525: stresses.

1526: 

1527: This argument is consistent with the sharp numerical profiles of

1528: Fig.~\ref{fig:profiles_gammonly}, which are replotted in

1529: Fig.~\ref{fig:construction}a (solid lines): each solution should have

1530: followed a local (dotted) curve $\phi(\gdot)$, but instead has jumped

1531: across the region in which this curve is non-monotonic. (The stresses

1532: used to generate the homogeneous solutions $\phi(\gdot,\Sigma)$ in

1533: Fig.~\ref{fig:construction}a were slightly different from those of the

1534: numerical profiles: however the trend is still clear.) Although

1535: Eqn.~(\ref{eqn:simpleconc}) is highly oversimplified, we believe that

1536: the failure to negotiate the interface due to the conflict described

1537: above is the reason for non-selection in the full, numerically solved

1538: model. 

1539: 

1540: %

1541: \begin{figure}[htb]

1542: \includegraphics[scale=0.6]{proof.eps}

1543: \caption{Assuming that the shear rate varies monotonically across the

1544:   interface (a), then for a relation $\phi(\gdot,\Sigma)$ of

1545:   Fig.~\ref{fig:construction}a that is non-monotonic, the

1546:   concentration $\phi$ must vary as in b), with three sign changes in

1547:   $\phi''\equiv\partial_y^2\phi$ as in c).

1548: \label{fig:proof} } 

1549: \end{figure}

1550: %

1551: 

1552: We return finally to justify our assumption that the shear rate must

1553: increase monotonically through the interface, and to discuss in more

1554: detail the nature of the banded solution when it {\em can} exist (\ie\ 

1555: for stresses near crticial point where $\phi(\gdot)$ is monotonic).

1556: Multiplying Eqn.~\ref{eqn:third} across by $d\phi/dy$, integrating on

1557: $\phi$, and imposing $\partial_y\phi=0$ at each boundary, we find,

1558: {\em for the simplified model of Eqn.~\ref{eqn:simpleconc}},

1559: %

1560: \begin{equation}

1561: \label{eqn:equalareas}

1562: \int_{\phi_l}^{\phi_r}d\phi\,\left[\mu-\Lambda\right]=0,

1563: \end{equation}

1564: %

1565: which is an ``equal areas'' construction. ($\phi_l$ and $\phi_r$

1566: denote the boundary values at $y=0,L$.) 

1567: %

1568: \begin{figure}[htb]

1569: \includegraphics[scale=0.3]{equalareas.eps}

1570: \caption{Of these three proposed constructions specifying the banded

1571:   state (when it is selected, near the critical point), only a) and b)

1572:   are consistent with the boundary conditions $\partial_y\phi=0$. Of

1573:   these, b) is for a finite system for which $\partial_y^2\phi\neq 0$

1574:   at the boundary while a) is for the realistic physical limit in

1575:   which the interface is narrow compared with the gap size,

1576:   connecting two homogeneous phases in which

1577:   $\partial_y^n\phi=0\,\forall\,n$. 

1578: \label{fig:equalareas} } 

1579: \end{figure}

1580: %

1581: If, in addition, we were to impose that $\partial_y^2\phi=0$ at each

1582: boundary, then the construction must automatically be as shown in

1583: Fig.~\ref{fig:equalareas}a. However we did not actually impose this

1584: condition in our numerics, so the construction of

1585: Fig.~\ref{fig:equalareas}b is also possible. This in fact corresponds

1586: to a finite system, where the true homogeneous state

1587: $\partial_y^n\phi=0\,\forall n$ is not quite reached at the

1588: boundaries.  Any other equal areas construction

1589: (Fig.~\ref{fig:equalareas}c) is not possible, for the following

1590: reason.  Consider starting at point C with $\partial_y\phi=0$ (which

1591: we {\em do} impose at the boundary in our numerics).

1592: Eqn.~\ref{eqn:third} then tells us that $\partial_y^2\phi<0$ at this

1593: point, so the function $\phi(y)$ must curve downwards from its

1594: starting point of zero slope.  Therefore $\phi$ locally decreases, and

1595: the system moves to point $C'$.  Repeating this argument, we find that

1596: the system can never cross to the point $D$. By similar reasoning, the

1597: shear rate must rise monotonically through the interface since any

1598: initial fall (from the side of the low shear band) would be similarly

1599: unstable to point $C$ in Fig.~\ref{fig:equalareas}c above.

1600: 

1601: Of course the concentration equation~(\ref{eqn:simpleconc}) is highly

1602: oversimplified. For instance, a more realistic model (such as the one

1603: of Eqn.~\ref{eqn:relative}) would have $\phi$ dependent prefactors to

1604: the $\partial_y^2\phi$ term. {\em The equal areas result of

1605:   Eqn.~(\ref{eqn:equalareas}) is therefore specific to our

1606:   oversimplified Eqn.~(\ref{eqn:simpleconc}), and does not hold in

1607:   general.}  Nonetheless we believe that Eqn.~\ref{eqn:simpleconc}

1608: correctly predicts the absence of a uniquely banded solution for

1609: stresses far above the critical point, via the basic conflict between

1610: the number of sign changes of $\partial_y^2\phi$ across the interface,

1611: described above.

1612: 

1613: \section{Conclusions}

1614: \label{sec:conclusion}

1615: 

1616: In this paper, we have studied the role of concentration coupling in

1617: the shear banding of complex fluids using the two-fluid, non-local

1618: Johnson-Segalman model. We have calculated phase diagrams for

1619: different degrees of coupling between concentration and mechanical

1620: degrees of freedom (molecular strain), and found a phase diagram

1621: qualitatively consistent with experiments on micellar solutions at

1622: dilutions well below the equilibrium isotropic-to-nematic transition

1623: \cite{berret94b}. Specific points to note are as follows.

1624: 

1625: \begin{enumerate}

1626:   

1627: \item The coexistence plateau in the steady-state flow curve slopes

1628:   upward with shear rate, because of the concentration difference

1629:   between the coexisting bands. The overall plateau height and width

1630:   decrease with average concentration, terminating in a

1631:   non-equilibrium critical point.  CPCl/NaSal in

1632:   brine~\cite{berret94b} shows the same trend. 

1633:   

1634: \item 

1635:   

1636:   Of the two coexisting bands, the high shear band has a smaller

1637:   concentration due to the fact that concentration tends to move up

1638:   gradients in the normal micellar strain component $W_{yy}$ (where

1639:   $y$ is the flow-gradient direction).  ($\tens{W}$ describes

1640:   deformation relative to the unit tensor $\tens{\delta}$, and

1641:   $W_{yy}$ is more negative in the high-shear phase than in the lower

1642:   shear phase.) Tie lines of the phase diagram in the $\gdot,\phi$

1643:   plane therefore have negative slope.

1644: 

1645: \item 

1646:   The concentration gap is smaller for smaller values of

1647:   concentration-coupling $\alpha\propto G'(\phi)/f''(\phi)$, and

1648:   tends to zero in the limit $\alpha\to 0$. Accordingly, the

1649:   coexistence region of the steady-state flow curve becomes flat in

1650:   this limit.

1651:   

1652: \item We have described the way in which the flow phase diagram can be

1653:   reconstructed from the family of flow curves $\Sigma(\gdotb,\phib)$,

1654:   measured for several average concentrations $\phib$

1655:   (Fig.~\ref{fig:reconstruct}).

1656:   

1657: \item The phase diagram and flow curves depend slightly on the

1658:   relative size of the interfacial term in the viscoelastic

1659:   constitutive equation to that in the equation that specifies the

1660:   concentration dynamics.  This is a concrete demonstration of how

1661:   stress selection and the coexistence conditions of driven systems

1662:   depend on the nature of the interface, in contrast to equilibrium

1663:   coexistence.

1664:   

1665: \item We find \textit{no} unique state selection when there are no

1666:   gradient terms in the viscoelastic constitutive equation, except for

1667:   stresses that are close to the critical point.  This implies that,

1668:   for a model to reproduce a uniquely selected stress, it is not

1669:   enough to simply have gradient terms only in, for example, the

1670:   concentration dynamics. The dynamical equations of motion for each

1671:   degree of freedom must possess inhomogeneous terms to attain

1672:   selection in all situations. Conversely, in situations where such

1673:   terms are physically absent, one can expect, under certain

1674:   conditions, no selection and hence a range of control parameters

1675:   (shear stress or strain rate) for which the steady states are

1676:   \textit{intrinsically} history-dependent.

1677:   

1678: \item The interface width diverges at the critical point as a power

1679:   law $(\Sigma-\Sigma_{\rm c})^{-n}$ with $n\approx 0.5$, although

1680:   $n$ differs slightly across the different order parameters.

1681: 

1682: \end{enumerate}

1683: 

1684: Although our \model\ model is highly oversimplified, we believe that

1685: it contains the basic ingredients required for a first description of

1686: wormlike micellar surfactant solutions at concentrations well below

1687: the isotropic-nematic (I-N) transition. In particular, it incorporates

1688: the minimal set of realistic degrees of freedom (tensorial order

1689: parameter for the micellar strain together with concentration), and

1690: unifies a non-monotonic flow curve with the Helfand-Fredrickson

1691: coupling between concentration and flow. Similar techniques could be

1692: applied to more involved Cates non-linear theory for wormlike micelles

1693: \cite{cates90,SpenCate94}.

1694: 

1695: We recall a previous calculations by Olmsted \etal\ was aimed at

1696: systems of rigid rods near the I-N transition~\cite{olmsted99c}. In

1697: future work we hope to unify these two approaches into a description

1698: of wormlike micelles that is valid over the entire concentration

1699: range. This should provide a first step towards understanding the

1700: crossover regime in the data of Fig.~\ref{fig:coexistence}, in which the

1701: coexistence plateau stress is a non monotonic function of the micellar

1702: concentration.

1703: 

1704: \begin{acknowledgements}

1705:   We thank J-F Berret and Paul Callaghan for useful discussions, and

1706:   EPSRC GR/N11735 for financial support.

1707: \end{acknowledgements}

1708: 

1709: \appendix

1710: \bw

1711: 

1712: \section{ \model\ Equations in Cartesian Coordinates}

1713: \label{app:fulleqns}

1714: 

1715: In this appendix, we give the components of the \model\ model's

1716: equations for planar shear flow along the $x$ direction, allowing

1717: gradients only in the flow-gradient direction, $y$, as described in

1718: Sec.~\ref{sec:geometry}, above. The $x$ component of force-balance is

1719: (in the zero-Reynolds limit considered in this paper)

1720: %

1721: \be 

1722: 0=\partial_y

1723: \left[ G \left( \phi \right) { W_{xy}} \right] +\,{ \etam}\,\partial_y

1724: \left[\,\phi\,\partial_y { v_{mx}} \right] +\,{ \etas}\,\partial_y

1725: \left[ \, \left( 1-\phi \right) \partial_y { v_{sx}} \right].  \ee 

1726: %

1727: The $y$ component of force-balance is fixed by incompressibility,

1728: $\nablu.\vect{v}=0$, along with the boundary condition $v_y=0$:

1729: %

1730: \be

1731: 0=\phi v_{my}+(1-\phi)v_{sy}.

1732: \ee

1733: %

1734: The relative velocity between the micelles and solvent (again ignoring

1735: inertial terms) is

1736: %

1737: \begin{align}

1738: v_{my}-{ v_{sy}}&=\frac{\phi(1-\phi)}{\zeta} \left\{ \frac{1}{\phi}

1739:   \partial_y \left[ G \left( \phi 

1740:  \right) { W_{yy}} \right] +2\,\frac{1}{\phi}  { \etam}

1741: \,\partial_y \left[  \,\phi\,\partial_y  { v_{my}}   \right] -2\, \frac{1}{1-\phi} { \etas}\,\partial_y \left[ \, \left( 1-\phi

1742:  \right) \partial_y  { v_{sy}}   \right] - \partial_y \mathfrak{F} \right\}\\

1743:  { v_{mx}}-{ v_{sx}}&=\frac{\phi(1-\phi)}{\zeta}\left\{ \frac{1}{\phi}

1744:    \partial_y \left[ G \left( \phi 

1745:  \right) { W_{xy}} \right] +\,\frac{1}{\phi}  { \etam}

1746: \,\partial_y \left[  \,\phi\,\partial_y  { v_{mx}}   \right] - \,\frac{1}{1-\phi} { \etas}\,\partial_y \left[ \, \left( 1-\phi

1747:  \right) \partial_y { v_{sx}}  \right]\right\}\\

1748: \mathfrak{F}&= { f'} \left( \phi \right) - \,g  \partial^2_y \phi  + \tfrac{1}{2}\,{ G'}

1749:  \left( \phi \right)  \left[ { W_{yy}}+{ W_{xx}}-\ln  \left( { W_{yy}}

1750: \,{ W_{xx}}+{ W_{yy}}+{ W_{xx}}+1-{{ W_{xy}}}^{2} \right)  \right].

1751: \end{align}

1752: The evolution of the micellar strain tensor is given by

1753: \begin{align}

1754: \partial_t W_{xy}+{ v_{my}}\,\partial_y  { W_{xy}}&=  \tfrac{1}{2}(a-1) W_{xx} \,\partial_y  { 

1755: v_{mx}} + \tfrac{1}{2}(1+a)\,W_{yy} \partial_y  { v_{mx}}  +a

1756:   \,W_{xy} \partial_y  { v_{my}} + \,\partial_y  { 

1757: v_{mx}}   

1758: -{\frac {{ W_{xy}}}{\tau \left( \phi \right) }}+{\frac {

1759:  l \left( \phi \right) ^{2} \partial^2_y { W_{xy}}

1760:   }{\tau \left( \phi \right) }},\\

1761: \partial_t W_{yy}+{ v_{my}}\,\partial_y  { W_{yy}} &=  (a-1)

1762: \,W_{xy}\partial_y  {  

1763: v_{mx}} +2\,a   \,W_{yy} \partial_y  { v_{my}} + 2\,\partial_y  { v_{my}}   -{\frac {{

1764:  W_{yy}}}{\tau \left( \phi \right) }}+{\frac {  l \left( \phi

1765:  \right)   ^{2} \partial^2_y  { W_{yy}}  }{

1766: \tau \left( \phi \right) }},\\

1767: \partial_t W_{xx}+{ v_{my}}\,\partial_y { W_{xx}} &=  (1+a)\,W_{xy}\partial_y  { 

1768: v_{mx}}  -{

1769: \frac {{ W_{xx}}}{\tau \left( \phi \right) }}+{\frac { l

1770:  \left( \phi \right)   ^{2}\partial^2_y  { W_{xx}} 

1771:  }{\tau \left( \phi \right) }}.

1772: \end{align}

1773: Finally, the concentration dynamics are

1774: %

1775: \bea 

1776: \partial_t\phi&=&-\partial_y \left\{ \frac{\phi^2\, \left( 1-\phi

1777:     \right)^2}{\phi\zeta}

1778:   \left\{ \partial_y \left[ G \left( \phi

1779:       \right) { W_{yy}} \right] +2\,\etam \,\partial_y \left[

1780:       \,\phi\,\partial_y  { v_{my}} 

1781:        \right] -\frac{2\phi}{1-\phi}  {

1782:       \etas}\,\partial_y \left[ \, \left( 1-\phi \right) \partial_y

1783:       { v_{sy}}\right] - \partial_y \mathfrak{F} \right\} \right\}.  \eea

1784: 

1785: \ew

1786: 

1787: \section{Stationary homogeneous  solutions of the \model\ model}

1788: \label{app:homogen}

1789:   

1790: In planar shear, the stationary homogeneous solutions to

1791: Eqns.~(\ref{eqn:navier}-\ref{eqn:JSd}) for given $\gdot$ and $\phi$

1792: are $\vrel\equiv\vm-\vs=0$ and

1793: %

1794: \begin{subequations}

1795: \label{eqn:intrinsic_flow_curve}

1796: \begin{align}

1797: W_{xy}&=\frac{\gdot\tau(\phi)}{1+b\gdot^2\tau^2(\phi)},\\

1798: W_{yy}&= \frac{a-1}{1+a}W_{xx}=

1799: -\frac{1}{(1+a)}\frac{b\gdot^2}{1+b\gdot^2}\\

1800: W_{zz}&=W_{xz}=W_{yz}=0,

1801: \end{align}

1802: 

1803: \end{subequations}

1804: %

1805: where $b=1-a^2$. The steady state shear stress is given by 

1806: \begin{equation}

1807:   \label{eq:3}

1808:   \Sigma_{xy}=G(\phib)W_{xy}+\phi\etam+(1-\phi)\etas\gdot=\textrm{constant}.

1809: \end{equation}

1810: 

1811: %\bibliographystyle{epjsty}

1812: 

1813: %\bibliography{ackerson,actin,articles,banding,barham,berrportdecruppe,books,callaghan,chandcits,crystal,crystal_theory,dnatheory,flowcryst,fredrickson,gelbart,helfrich,hsiao,LCtheory,lifshitz,line_tension,malkus,master,mccoy,membs,mine,noirez,notes,onions,pineweitz,pomeau,poon,pratt,psolutions,rheofolks,ryan,schoot,semenov,sriram,stein,worms2,worms3,worms,zubarev}

1814: 

1815: \begin{thebibliography}{10}

1816: 

1817: \bibitem{cates90}

1818: M.~E. Cates, J.~Phys. Chem. {\bf 94},  371  (1990).

1819: 

1820: \bibitem{SpenCate94}

1821: N.~A. Spenley and M.~E. Cates, Macromolecules {\bf 27},  3850  (1994).

1822: 

1823: \bibitem{SCM93}

1824: N.~A. Spenley, M.~E. Cates, and T.~C.~B. McLeish, Phys. Rev. Lett. {\bf 71},

1825:   939  (1993).

1826: 

1827: \bibitem{Yerushalmi70}

1828: J. Yerushalmi, S. Katz, and R. Shinnar, Chemical Engineering Science {\bf 25},

1829:   1891  (1970).

1830: 

1831: \bibitem{olmsted99a}

1832: P.~D. Olmsted, O. Radulescu, and C.-Y.~D. Lu, J. Rheology {\bf 44},  257

1833:   (2000).

1834: 

1835: \bibitem{lu99}

1836: C.-Y.~D. Lu, P.~D. Olmsted, and R.~C. Ball, Phys. Rev. Lett. {\bf 84},  642

1837:   (2000).

1838: 

1839: \bibitem{olmstedlu97}

1840: P.~D. Olmsted and C.-Y.~D. Lu, Phys.~Rev. {\bf E56},  55  (1997).

1841: 

1842: \bibitem{spenley96}

1843: N.~A. Spenley, X.~F. Yuan, and M.~E. Cates, J.~Phys.~II (France) {\bf 6},  551

1844:   (1996).

1845: 

1846: \bibitem{berret94b}

1847: J.~F. Berret, D.~C. Roux, and G. Porte, J.~Phys.~II (France) {\bf 4},  1261

1848:   (1994).

1849: 

1850: \bibitem{Call+96}

1851: P.~T. Callaghan, M.~E. Cates, C.~J. Rofe, and J.~B. A.~F. Smeulders, J. Phys.

1852:   II (France) {\bf 6},  375  (1996).

1853: 

1854: \bibitem{grand97}

1855: C. Grand, J. Arrault, and M.~E. Cates, J.~Phys.~II (France) {\bf 7},  1071

1856:   (1997).

1857: 

1858: \bibitem{MairCall96}

1859: R.~W. Mair and P.~T. Callaghan, Europhys. Lett. {\bf 36},  719  (1996).

1860: 

1861: \bibitem{MairCall96c}

1862: R.~W. Mair and P.~T. Callaghan, Europhysics Letters {\bf 65},  241  (1996).

1863: 

1864: \bibitem{BritCall97}

1865: M.~M. Britton and P.~T. Callaghan, Phys. Rev. Lett. {\bf 78},  4930  (1997).

1866: 

1867: \bibitem{berret94a}

1868: J.~F. Berret, D.~C. Roux, G. Porte, and P. Lindner, Europhys. Lett. {\bf 25},

1869:   521  (1994).

1870: 

1871: \bibitem{schmitt94}

1872: V. Schmitt, F. Lequeux, A. Pousse, and D. Roux, Langmuir {\bf 10},  955

1873:   (1994).

1874: 

1875: \bibitem{Capp+97}

1876: E. Cappelaere, J.~F. Berret, J.~P. Decruppe, R. Cressely, and P. Lindner, Phys.

1877:   Rev. {\bf E} {\bf 56},  1869  (1997).

1878: 

1879: \bibitem{rehage91}

1880: H. Rehage and H. Hoffmann, Mol. Phys. {\bf 74},  933  (1991).

1881: 

1882: \bibitem{Decr+95}

1883: J.~P. Decruppe, R. Cressely, R. Makhloufi, and E. Cappelaere, Coll. Polym. Sci.

1884:   {\bf 273},  346  (1995).

1885: 

1886: \bibitem{Makh+95}

1887: R. Makhloufi, J.~P. Decruppe, A. Aitali, and R. Cressely, Europhys. Lett. {\bf

1888:   32},  253  (1995).

1889: 

1890: \bibitem{DCC97}

1891: J.~P. Decruppe, E. Cappelaere, and R. Cressely, J. Phys. II (France) {\bf 7},

1892:   257  (1997).

1893: 

1894: \bibitem{BPD97}

1895: J.~F. Berret, G. Porte, and J.~P. Decruppe, Phys. Rev. {\bf E} {\bf 55},  1668

1896:   (1997).

1897: 

1898: \bibitem{FisCal01}

1899: E. Fischer and P.~T. Callaghan, Phys.\ Rev.\ E {\bf 6401},  1501  (2001).

1900: 

1901: \bibitem{FisCal00}

1902: E. Fischer and P.~T. Callaghan, Europhys.\ Lett. {\bf 50},  803  (2000).

1903: 

1904: \bibitem{LerDecBer00}

1905: S. Lerouge, J.~P. Decruppe, and J.~F. Berret, Langmuir {\bf 16},  6464  (2000).

1906: 

1907: \bibitem{BRL98}

1908: J.~F. Berret, D.~C. Roux, and P. Lindner, European Physical Journal B {\bf 5},

1909:   67  (1998).

1910: 

1911: \bibitem{schmitt95}

1912: V. Schmitt, C.~M. Marques, and F. Lequeux, Phys.~Rev. {\bf E52},  4009  (1995).

1913: 

1914: \bibitem{brochdgen77}

1915: F. Brochard and P.-G. de~Gennes, Macromolecules {\bf 10},  1157  (1977).

1916: 

1917: \bibitem{HelfFred89}

1918: E. Helfand and G.~H. Fredrickson, Phys. Rev. Lett. {\bf 62},  2468  (1989).

1919: 

1920: \bibitem{DoiOnuk92}

1921: M. Doi and A. Onuki, J. Phys. II (France) {\bf 2},  1631  (1992).

1922: 

1923: \bibitem{milner93}

1924: S.~T. Milner, Phys.~Rev. {\bf E48},  3674  (1993).

1925: 

1926: \bibitem{WPD91}

1927: X.~L. Wu, D.~J. Pine, and P.~K. Dixon, Phys. Rev. Lett. {\bf 66},  2408

1928:   (1991).

1929: 

1930: \bibitem{beris94}

1931: A.~N. Beris and V.~G. Mavrantzas, J. Rheol. {\bf 38},  1235  (1994).

1932: 

1933: \bibitem{SBJ97}

1934: T. Sun, A.~C. Balazs, and D. Jasnow, Phys. Rev. {\bf E} {\bf 55},  R6344

1935:   (1997).

1936: 

1937: \bibitem{tanak96}

1938: H. Tanaka, Phys. Rev. Lett. {\bf 76},  787  (1996).

1939: 

1940: \bibitem{DecLerBer01}

1941: J.~P. Decruppe, S. Lerouge, and J.~F. Berret, Phys.\ Rev.\ E {\bf 6302},  2501

1942:   (2001).

1943: 

1944: \bibitem{FieOlm02}

1945: S.~M. Fielding and P.~D. Olmsted,   (2002), preprint cond-mat/0207344.

1946: 

1947: \bibitem{FieOlm02b}

1948: S.~M. Fielding and P.~D. Olmsted,   (2002), preprint, cond-mat/0208599.

1949: 

1950: \bibitem{johnson77}

1951: M. Johnson and D. Segalman, J. Non-Newt. Fl. Mech {\bf 2},  255  (1977).

1952: 

1953: \bibitem{milner91}

1954: S.~T. Milner, Phys. Rev. Lett. {\bf 66},  1477  (1991).

1955: 

1956: \bibitem{deGen76}

1957: P.-G. de~Gennes, Macromolecules {\bf 9},  587  (1976).

1958: 

1959: \bibitem{Brochard83}

1960: F. Brochard, J.Phys. (Paris) {\bf 44},  39  (1983).

1961: 

1962: \bibitem{olmsted99c}

1963: P.~D. Olmsted and C.-Y.~D. Lu, Phys. Rev. {\bf E60},  4397  (1999).

1964: 

1965: \bibitem{FieOlm02f}

1966: S.~M. Fielding and P.~D. Olmsted,   (2002), in preparation.

1967: 

1968: \bibitem{elkareh89}

1969: A.~W. El-Kareh and L.~G. Leal, J. Non-Newt. Fl.~Mech. {\bf 33},  257  (1989).

1970: 

1971: \bibitem{BRP94}

1972: J.~F. Berret, D.~C. Roux, and G. Porte, J. Phys. II (France) {\bf 4},  1261

1973:   (1994).

1974: 

1975: \end{thebibliography}

1976: 

1977: 

1978: \end{document}

1979: 

1980: 

1981: