nlin0412002/PJL.tex
1: %% -*- coding: euc-jp -*- 
2: %$Id: PJL.tex,v 1.18 2005/06/16 12:16:34 axel Exp $
3: %% ****** Start of file PJL.tex ****** %
4: %%
5: %%   This is the paper for the phenomena of Liquid Crystal.
6: %%   For We found Phase Jump Line of MBBA Cell.
7: %%   And I reported.
8: %%
9: %%   Copyright (c) 2004 TATSUMI Soichi
10: %%   It is the first paper of my life.
11: %%   I've great thanked with Axel Rossberg, and Masaki Sano.
12: %%   I want to turn this paper as a first step of my life for reserchers.
13: %%
14: %%   Then it'll be started as it follows.
15: %%
16: 
17: %\listfiles
18: \documentclass[%
19: %prl%
20: ,preprint%
21: %,twocolumn%
22: ,secnumarabic%
23: %,tightenlines%
24: ,amssymb, amsmath%
25: %,nobibnotes%
26: %, aps%
27: , pre%
28: %, superscriptaddress
29: ]{revtex4}
30: %\usepackage{acrofont}%NOTE: Comment out this line for the release version!
31: %\usepackage{docs}%
32: \usepackage{times}
33: %\usepackage{CJK}
34: %\usepackage{srctex} %for jumping directly to the current position in
35:   %the DVI file.
36: \usepackage{bm}%
37: \usepackage[%
38: %draft,%
39: dvips]{graphicx}
40: 
41: %\usepackage[colorlinks=true,linkcolor=blue]{hyperref}%
42: %\nofiles
43: 
44: \expandafter\ifx\csname package@font\endcsname\relax\else
45:  \expandafter\expandafter
46:  \expandafter\usepackage
47:  \expandafter\expandafter
48:  \expandafter{\csname package@font\endcsname}%
49: \fi
50: 
51: \newcommand{\DegC}{\mbox{\rm\char'27\kern-.3em C}}
52: \newcommand{\dx}{\partial_x}
53: \newcommand{\dy}{\partial_y}
54: \newcommand{\dz}{\partial_z}
55: \newcommand{\dt}{\partial_t}
56: \newcommand{\sech}{\mathop{\mathrm{sech}}}
57: \newcommand{\cm}[1]{({\small \sf #1})} % use for comments
58: \renewcommand\cm[1]{\ignorespaces} % this turns comments off
59: 
60: \begin{document}
61: %\begin{CJK*}{JIS}{song}
62: 
63: \title{Observation of stable phase jump lines in
64:   convection of a twisted nematic}%
65: \author{Soichi Tatsumi}
66: \email{tatsumi@daisy.phys.s.u-tokyo.ac.jp}
67: \affiliation{Department of Physics, The University of Tokyo, 7-3-1
68:   Hongo, Bunkyo-ku Tokyo,113-0033}
69: \author{A. G. Rossberg}
70: \affiliation{Graduate School of Environment and Information Sciences, 
71: Yokohama National University, 79-7 Tokiwadai, Hodogaya-Ku
72: Yokohama 240-8501 }
73: \author{Masaki Sano}\affiliation{Department of Physics, The University
74:   of Tokyo, 7-3-1 Hongo, Bunkyo-ku Tokyo,113-0033}
75: 
76: %\date{\today}
77: \begin{abstract}
78:   We report observations of stable, localized, line-like structures
79:   in the spatially periodic pattern formed by nematic
80:   electroconvection, along which the phase of the pattern jumps by
81:   $\pi$.  With increasing electric voltage, these lines form a grid-like 
82:   structure that goes over into a structure indistinguishable from the
83:   well known grid pattern.  We present theoretical arguments that
84:   suggest that the twisted cell geometry we are using is indirectly
85:   stabilizing the phase jump lines, and that the PJL lattice is caused
86:   by an interaction of phase jump lines and a zig-zag instability of
87:   the surrounding pattern.
88: \end{abstract}
89: \maketitle
90: 
91: \section{Introduction}
92: Usually, topological defects in the phase of periodic spatial
93: modulations have co-dimension two.  They are point-like in two
94: dimensions, line-like in three dimensions, and in one dimension they
95: reduce, for dynamical systems, to short events, localized in space and
96: time\cite{Bray994}.  Here we report the observation of stable, self-organized phase
97: defects that extend to line-like objects in 2d: \textit{Phase jump
98:   lines} (PJL) stretched parallel to the wavevector of the modulation,
99: characterized by a change of the phase of the spatial modulation by
100: $\pi$ on a short distance (Fig.\ref{SetUp}).
101: 
102: The phenomenon was observed in the periodic pattern of convection
103: rolls forming in thin layers of the nematic liquid crystal MBBA
104: (4-methoxybenzylidene-4'-butylaniline) under the influence of an
105: electric ac field ($\parallel \!\!\hat z$) that is oriented normal to
106: the ($x$-$y$) layer and the alignment of the nematic director.
107: Electroconvection (EC) in nematic liquid crystals has been studied
108: extensively.  After early studies by E. Dubois-Violette and others
109: \cite{dubdusol} aimed at understanding the basic mechanism of
110: convection, EC later became popular as a model for pattern formation
111: in anisotropic systems.  Depending on the experimental conditions, the
112: convection rolls are aligned either preferentially normal to the
113: symmetry axis ($\hat x$) of the system---defined by the direction at
114: which the nematic director is anchored at the boundaries---or oblique
115: at some preferred angel.  In the oblique-roll case, two convective
116: modes related to each other by a reflection at the $x$ axis coexist
117: \cite{zk,bzk}.  Near the onset of convection, descriptions in terms of
118: one (normal rolls) or two (oblique rolls) time-dependent
119: Ginzburg-Landau equations could be established \cite{annrev}.  These
120: equations are of potential type and predict a relaxation to a simple
121: equilibrium state.
122: 
123: The more puzzling was the rich dynamics of the system observed for
124: stronger electric fields, with scenarios such as those named
125: ``fluctuating Williams domains'', ``chevron patterns'', ``abnormal
126: rolls'', ``grid patterns'', or ``defect lattices'' being reported
127: \cite{kck,asr,huhiroka,RokrChev,Nasuno989-a,Sano992,Nasuno995,Dennin998,Ribotta986,Scherer002,Rudroff999,Sasa990,Oikawa004}.
128: Only in recent years it became clear that most of these observations can
129: be understood by the non-relaxational interaction of the convective
130: mode(s) with only one additional, weakly damped mode that describes an
131: in-plane rotation of the nematic director. A Ginzburg-Landau-type
132: model for the non-relaxational coupling of these two (or three) modes,
133: called \textit{standard model} below, explains most of the experimentaly
134: observed scenarios at least qualitatively, and predicts a few more
135: \cite{roth,Komineas003}.
136: 
137: There are several indications in the literature that hint at the
138: existence of stable PJL in nematic electroconvection.  For example,
139: for the defect-chaotic regime of normal rolls called ``fluctuating
140: Williams domains'' it is well known that the ``shape'' of topological
141: defects, particularly near the events of defect creations and
142: annihilations, can be highly anisotropic.  Their extension in $x$
143: direction is considerably larger than in $y$ direction.  As such this
144: deformation could be explained by the anisotropy of the linear
145: dispersion relation of the convective mode, which can be removed by a
146: simple rescaling of the $x$ axis.  But in some cases the defects
147: stretch considerably and are then more naturally characterized as
148: ``phase jump lines'' \cite{Sasa992,Buka001,Cladis_book001}.  These
149: structures seem to be related to weakly unstable saddles corresponding
150: to the stable PJL reported here.  
151: Stable PJL have also been seen in numerical simulations of the
152: standard model \cite{zhao00:_secon_instab_compl_patter_anisot_convec}.
153: Yet, a general theory is missing, and it remains unclear if or under
154: which conditions stable PJL would be observable in electroconvection.
155: 
156: In our experiments, the conventional EC setup was modified by twisting
157: the nematic within the layer: Instead of anchoring the nematic
158: director at the surfaces of the layer parallel to the $x$ axis, the
159: anchoring direction is rotated by a fixed angle in opposite directions
160: at the two surfaces \cite{Hertrich994,bosta,Rossberg000,Delev000}.
161: This geometry seems to be favorable for PJL.
162: 
163: In the following we first describe the details of the twisted EC
164: setup, and describe the observed PJL and structures they form.  We
165: then discuss some theoretical ideas that understanding these
166: observations.
167: 
168: \section{Experimental Setup}
169: \begin{figure}[bth]
170:  \includegraphics[width=.9\linewidth]{SetUp_f.eps}
171:  \caption{Top: rubbing direction and the anchoring direction of
172:    nematic. Bottom: a typical phase jump line at 400Hz, 13.10 Volts.
173:    The white bar corresponds to 100$\mu m$.}
174:  \label{SetUp}
175:  \vspace{-.5cm}
176: \end{figure}
177: The nematic liquid crystal MBBA doped with 0.1wt\%
178: tetra-n-butylammonium bromide to stabilize the conductivity was
179: sandwiched between two glass plates covered with transparent indium
180: tin oxide thin-film electrodes. The lateral extensions of the liquid
181: crystal layer were $2\mathrm{cm} \times 2\mathrm{cm}$ and its
182: thickness $d=50 \mu \mathrm{m}$.  The surfaces of the plates were
183: coated with polyvinyl alcohol and rubbed in order to attain a planar
184: anchoring of the nematic molecules.  Temperature of the cell was
185: controlled at $25\pm0.01 \DegC $. A standard wave-generator was
186: used to apply ac voltages $V=\mathcal{O}(100\mathrm{V})$ of
187: frequencies $f$ up to several $\mathrm{kHz}$ to the electrodes.  The
188: structure of the cell is shown schematically in Fig.~\ref{SetUp}.  The
189: anchoring direction for each plate is represented by an arrow.  By
190: rotating the anchoring direction by an angle $\alpha=\pi/4$ out of the
191: $x$ axis at the lower plate, and in opposite direction by the same
192: amount at the upper plate, the nematic director is twisted along the
193: $z$-axis in the ground state \footnote{This geometry allows two
194:   equivalent ground states: left twist and right twist.  For the
195:   experiments a region of the sample without domain walls (uniform
196:   twist) was selected.}  (Fig.~\ref{SetUp}).
197: 
198: \section{Experimental Results}
199: \begin{figure}[bth]
200:  \includegraphics[width=\linewidth]{PhaseDiagram.eps}
201:  \caption{Phase Diagram of our experiments. Solid circles represent the
202:    onset voltage of convection rolls (Williams Domain); open square
203:    the onset voltage of PJL; open circle represents the onset of the
204:    PJL grid pattern (GP); solid square the onset voltage of the
205:    dynamic scattering mode (DSM1).  The inset shows the part of the
206:    phase diagram that we discuss in this paper.  The points a, b and c
207:    in the inset correspond to Figs.~\ref{Main}a-c.
208:    \label{phase}}
209: % \vspace{-.5cm}
210: \end{figure}
211: The control parameters of the system are the externally applied
212: voltage and its frequency.  A phase diagram of the observed patterns
213: is shown in the Fig.~\ref{phase}.  The cutoff frequency $f_c$, where
214: the conductive mode of electroconvection is superseded by the
215: dielectric mode, was above $1\,\mathrm{kHz}$.  Measurements were
216: performed for $f<f_c$, {\it i.e.}, in the conductive regime.  As the
217: voltage was increased, normal rolls with a wavevector parallel to the
218: $x$-axis (at an angle of $\alpha$ to each anchoring direction) were
219: observed.  Besides the well-known states including normal rolls
220: (parallel rolls), grid pattern (rectangular cells), and dynamic
221: scattering modes (3D turbulent states), we found stably aligned phase
222: jump lines (PJL) in a wide frequency range of the conductive regime.
223: 
224: 
225: 
226: \begin{figure*}[bth]
227:  \includegraphics[width=\linewidth]{Change.eps}
228:  \caption{\label{Main} The upper row shows images of convection
229:    patterns with PJL, the lower row the modulus of the corresponding
230:    2D Fourier transforms.  These three pair of figures represent the
231:    transition from PJL to GP.  Fig.~a corresponds to 13.93 V
232:    ($\epsilon=0.47$), Fig.~b to 15.91 V ($\epsilon=0.91$), and Fig.~c to
233:    19.23 V ($\epsilon=1.79$), all at 400 Hz. The length of the scale
234:    bars is 200$\mu m$. }
235: \end{figure*}
236: Figures~\ref{Main}a-c show the shadow-graph image of PJL observed
237: under a polarization microscope (polars $\parallel x$).  PJL can have
238: finite length.  At the endpoints of PJL the phase difference between
239: the adjacent domains gradually decreases to zero.  As the voltage is
240: increased, the typical distance $L$ between PJL in $y$ direction
241: decreases.  
242: 
243: In order to determine $L$, which is below expressed in terms of the
244: wave number $K:=2\pi/L$, quantitatively, two methods were employed.
245: %
246: The first makes direct use of the Fourier transforms of recorded
247: images (Figs.~\ref{Main}b',c'). The other method is to count the
248: number $n$ \footnote{The number $n$ averaged over a full image was
249:   determined by first summing the $x$-extension of all PJL in the
250:   image and dividing this number by the $x$-width of the image.}
251: of PJL along a line of length $E$ parallel to $\hat y$, which gives
252: $K=2\pi n/E$.  For high voltage, the Fourier analysis was preferred.
253: At lower voltages we used the counting method, since the small number
254: of PJL would not lead to sharp peaks in the Fourier transform.  We
255: verified that in the intermediate range the results of the two methods
256: coincide.
257: 
258: In Fig.~\ref{length} the measured wave number $K$ is plotted against
259: the normalized distance $\epsilon = (V^2 -V_c^2)/V_c^2$ from the
260: threshold of convection $V_c$.  The data indicate that $K$ goes to
261: zero following a square-root law 
262: \begin{align}
263:   \label{Klaw}
264:  K= c\, (\epsilon - \epsilon_1)^{1/2} 
265: \end{align}
266: with $\epsilon_1>0$.  For $f=400\mathrm{Hz}$ we obtain
267: $c=50.5\mathrm{mm^{-1}}$ and $\epsilon_1=0.22$ and for $f=600Hz$
268: data is well described by $c=95.3\mathrm{mm^{-1}}$ and
269: $\epsilon_1=0.31$.
270: 
271: Some observations indicate that there are at least two different
272: regimes of PJL.  At low $\epsilon$, the spacing between PJL is very
273: irregular (Fig.~\ref{Main}a,b) and the PJL continue to move, although
274: very slowly.  At high $\epsilon$, where the spacing between PJL is
275: smaller, it becomes also more regular (as can be seen from the sharp
276: peaks in the Fourier transform of Fig.~\ref{Main}c shown in
277: Fig.~\ref{Main}c'), and a steady state is reached.  We refer to the
278: resulting structure as a \textit{PJL grid}. The onset voltage of PJL
279: grids is well above the onset of PJL (Fig.~\ref{phase}).  At this
280: point the $K$ \textit{vs} $\epsilon$ relation also starts to deviate
281: significantly from the square-root law valid for smaller $\epsilon$,
282: as shown in Fig.~\ref{length}.  The deviation is towards larger $K$,
283: i.e.\ towards denser PJL, contrary to what one would expect from a
284: repulsive interaction between PJL. The PJL grids are visually
285: indistinguishable from the well-known grid pattern\cite{Kai976}
286: consisting of two overlapping sets of oblique
287: rolls.  If the two structures are indeed identical this means that we
288: found a new route from normal rolls to oblique rolls, different from
289: the conventional transition \textit{via} a zig-zag instability.  But
290: further investigations are required to confirm this hypothesis.  For
291: the low-$\epsilon$ range, on the other hand, some first steps towards
292: a better understanding have been made.
293: 
294: 
295: 
296: \begin{figure}[bth]
297:  \includegraphics[width=\linewidth]{PhaseShift.eps}
298:  \caption{The relation between $\epsilon$ and the inverse distance $K$
299:    between PJL.  The upper panel corresponds to $f=400$~Hz, the lower
300:    one to $f=600$~Hz.  The dashed lines are fits to the square-route
301:    law~(\ref{Klaw}).  The insets show $\epsilon $ - $\epsilon _1$
302:    \textit{vs} $K$ on a log-log scale.  The onsets of convection is at
303:    11.50 Volts for 400Hz and at 21.41 Volts for 600Hz.\label{length}
304:    }
305: % \vspace{.5cm}
306: \end{figure}
307: 
308: 
309: 
310: \section{Theory}
311: A clear theoretical picture of PJL, including the conditions for their
312: existence as well a description of their interaction with the
313: surrounding roll pattern and with each other, has not yet emerged.
314: Even though other mechanisms leading to stable PJL appear to exist
315: \cite{softmodesX}, the simulations of Zhao
316: \cite{zhao00:_secon_instab_compl_patter_anisot_convec} strongly
317: suggest that the observed PJL can be described by the standard model
318: for normal-roll electroconvection.  We shall here investigate three
319: aspects related to this problem that help  understanding why PJL
320: stabilize in twisted cells and how they are forming PJL grids.
321: 
322: The standard model \cite{rohekrpe,roth,rzkr,zth,Komineas003} is given by
323: \begin{subequations}
324:   \label{Aphi-dim}
325:   \begin{align}
326:     \begin{split}
327:     \label{Aphi-A-dim}
328:     \tau \partial_t A =& i \beta A \dy \phi + \Big(\epsilon +
329:     \xi_x^2 \partial_x^2 + \xi_y^2 \partial_y^2 \\ & + 2iq_c\,\xi_y^2\,C_1
330:     \phi\dy -C_2q_c^2\xi_y^2\phi^2 -g |A|^2\Big) A,
331:   \end{split}
332: \\
333:     \label{Aphi-phi-dim}
334:     \begin{split}
335:       \gamma_1 \partial_t \phi =& -h \phi + K_3 \partial_x^2 \phi +
336:       K_1 \partial_y^2 \phi \\&+2 \Gamma \mathop{\mathrm{Im}}\{A^*
337:       (\partial_y - i q_c \phi)A\}.
338:   \end{split}
339: \end{align}
340: \end{subequations}
341: It describes the interaction of two spatio-temporally slow variables:
342: The complex amplitude $A$ of the convection pattern $\sim
343: (Ae^{iq_cx}+\text{c.c.})$, and the amount $\phi$ of in-plane rotation
344: of the nematic director.  By the requirements of structural stability
345: and according to theoretical and experimental results for the
346: untwisted geometry, the time-constants $\tau$ and $\gamma_1$, the
347: coherence lengths $\xi_x$, $\xi_y$, the critical wave number $q_c$,
348: the geometry factors $C_1$ and $C_2$, the Landau coefficient $g$, the
349: damping constant $h$, and the two elastic constants $K_1$ and $K_3$
350: are all positive.  The sign of $\beta$ depends on the applied
351: frequency and the experimental geometry and is difficult to determine
352: \textit{a priori}.  The value of $\Gamma$ is always found to be
353: negative and we shall assume this hereafter.  It is easily seen that
354: with negative $\Gamma$ convection rolls ($A \ne 0$) have a
355: destabilizing effect on $\phi$.  This effect is crucial for most of
356: the complex patterns observed in electroconvection, and PJL are no
357: exception.
358: 
359: \subsection{Stabilization of PJL}
360: \label{sec:stability}
361: 
362: After scaling out the natural $\epsilon$ dependence of $A$, $\phi$ and
363: length and time scales ($\dx\sim\dy\sim A\sim\phi\sim\epsilon^{1/2}$,
364: $\dt\sim\epsilon$), the standard model~(\ref{Aphi-dim}) can be
365: reformulated by the following, equivalent system of PDE:
366: \begin{subequations}
367:   \label{Aphi}
368:   \begin{align}
369:     \label{Aphi-A}
370:     \tau \partial_t A =& (1 + \partial_x^2 + \partial_y^2 + 2i\,C_1
371:     \phi\dy-C_2\phi^2 -|A|^2) A + i \beta A \dy \phi, \\
372:     \label{Aphi-phi}
373:     \partial_t \phi =& -h\, \phi + \partial_y^2 \phi +
374:     K_3 \partial_x^2 \phi -2 |\Gamma| \mathop{\mathrm{Im}}\{A^* (\partial_y - i
375:     \phi)A\}.
376: \end{align}
377: \end{subequations}
378: The coefficients and variables in system~(\ref{Aphi}) are not
379: identical to those denoted with the same symbols in
380: system~(\ref{Aphi-dim}), but are related to these by simple scaling
381: transformations.  For all coefficients but $h$ this relation is
382: independent of $\epsilon$.  The coefficient $h$ in
383: Eq.~(\ref{Aphi-phi}) scales as $\epsilon^{-1}$, i.e., $h$ plays the
384: role of the main control parameter in the rescaled
385: equations~(\ref{Aphi}).  As $h$ decreases, the first instability of
386: the homogeneous solution $A=1$, $\phi=0$ is to abnormal rolls at
387: $h=2|\Gamma|$ if $\beta<0$ and to zig-zag modulations at
388: $h=2|\Gamma|(1+\beta)$ if $\beta>0$.
389: 
390: The system also has a steady-state solution
391: \begin{align}
392:   \label{PJL}
393:   \phi=0, \quad A=\tanh
394:   \left(
395:      y/\sqrt{2}
396:   \right)
397: \end{align}
398: that describes PJL.  For large $h$, i.e.\ close to the threshold of
399: convection, PJL are unstable.  This can be seen by considering the
400: limit $h\to\infty$ where $\phi=0$ is fixed.  Then solution~(\ref{PJL})
401: is unstable to the linear mode $\delta A=i\,\sech( y/\sqrt{2})$ with
402: growth rate $\tau^{-1}$ \cite{langer967}.  However, by lowering $h$
403: the additional degree of freedom $\phi$ can stabilize the PJL by an
404: interaction through the last term in Eq.~(\ref{Aphi-phi}).  For a
405: discussion of this process, we linearize the system~(\ref{Aphi}) with
406: respect to perturbations of solution~(\ref{PJL}) of the form $\delta
407: A= [a_r(y)+i a_i(y)]\exp(\sigma t)$, $\delta \phi=f(y) \exp(\sigma
408: t)$, which leads to
409: \begin{align}
410:   \label{ar}
411:   0 =& \left( -\sigma\,\tau +1
412:     -3\,{{\tanh}^2}{\frac{y}{{\sqrt{2}}}}+\partial_y^2 \right) a_r
413:   ,
414: \end{align}
415: and
416: \begin{subequations}
417: \label{linearized}
418: \begin{align}
419:   \label{ai}
420:   \begin{split}
421:     0 =& \left(- {\sqrt{2}}\,{ C_1}\,
422:       {{\sech}^2}{\frac{y}{{\sqrt{2}}}} +
423:       \beta\,\tanh{\frac{y}{{\sqrt{2}}}}\,\partial_y \right) f
424:     \\
425:     & + \left( -\sigma\,\tau + {{\sech}^2}{\frac{y}{{\sqrt{2}}}}
426:       +\partial_y^2 \right) a_i,
427: \end{split}
428:   \\
429:   \label{delta}
430:   \begin{split}
431:     0 =& \left(-\sigma-h -
432:       2|\Gamma|{{\tanh}^2}{\frac{y}{{\sqrt{2}}}}+\partial_y^2\right) 
433:     f \\
434:     & - \left({\sqrt{2}}\,|\Gamma|
435:       {{\sech}^2}{\frac{y}{{\sqrt{2}}}}-
436:       2\,|\Gamma|\,\tanh{\frac{y}{{\sqrt{2}}}}\partial_y \right)
437:     a_i .
438: \end{split}
439: \end{align}
440: \end{subequations}
441: We are looking for eigenfunctions of these equations which are bounded
442: for $y\to\pm\infty$.  Equation~(\ref{ar}) is decoupled from
443: Eqs.~(\ref{ai},\ref{delta}).  Apart from the translational mode
444: $a_r=\dy \tanh(y/\sqrt{2})$, Eq.~(\ref{ar}) has bounded solutions only
445: with $\tau \sigma\le-3/2$ and does not contribute to the linear
446: stability problem.  The $y$ axis can be divided into an inner region
447: $|y|\lesssim 5$ near the PJL, and two outer regions where the
448: system~(\ref{linearized}) simplifies.  For $y>0$, for example, the
449: substitutions $\sech\to 0$, $\tanh\to 1$ lead to
450: \begin{subequations}
451:   \label{outer}
452:   \begin{align}
453:     \label{ai-out}
454:     0=&(-\sigma\tau+\dy^2) a_i+\beta \dy f,\\
455:     \label{delta-out}
456:     0=&(-\sigma+2|\Gamma|\mu+\dy^2)f-2 |\Gamma| \dy a_i,
457:   \end{align}
458: \end{subequations}
459: where we introduced the abbreviation $\mu=1-h/2|\Gamma|$.
460: In general, the eigenmodes of Eq.~(\ref{linearized}) have to be
461: calculated numerically.  For example this can be done by ``shooting''
462: solutions starting at $y=0$ with $a_i=1$, $f=f_0$, $a_i'=f'=0$ into
463: the $y>0$ outer region such as to match
464: $a_i(y_1),a_i'(y_1),f(y_1),f'(y_1)$  for
465: sufficiently large $y_1$ with a bounded outer solution.  
466: We shall here only be interested in the
467: cases where the PJL stabilize prior to any destabilization of the
468: homogeneous state, i.e., for the cases that $\mu<-\beta$ and $\mu<0$.  The
469: bounded outer solutions are then generally given by linear
470: combinations of the two solutions of~(\ref{outer}) of the form
471: $a_i,f\sim\exp(\lambda y)$ with $\lambda\le 0$ and
472: \begin{align}
473:   \label{flat}
474:   \lambda^2=&\frac{\mu\sigma\tau}{\beta+\mu}+\mathcal{O}(\sigma^2)
475:   \intertext{or}
476:   \label{steep}
477:   \lambda^2=&-2|\Gamma|\,(\beta+\mu)+\mathcal{O}(\sigma),
478: \end{align}
479: respectively.  The value of $\mu$ where the PJL stabilize can be found
480: by considering the limit $\sigma\to 0^+$ of the
481: systems~(\ref{linearized}) and~(\ref{outer}).  Two particularities of
482: this limit will be used here: (i)~The outer solution corresponding
483: to~(\ref{flat}) goes over into a simple phase shift ($f=0$,
484: $a_i=\text{const}.$).  Thus, $f$ contains only contributions from the
485: component corresponding to~(\ref{steep}) and must satisfy
486: \begin{align}
487:   \label{phi-steep}
488:   \dy f=-\sqrt{-2|\Gamma|(\beta+\mu)}\,f.
489: \end{align}
490: (ii)~By Eq.~(\ref{ai-out}) the quantity
491: \begin{align}
492:   \label{conserved}
493:   u=\dy a_i+\beta f\quad(=0)
494: \end{align}
495: is conserved along $y$ in the outer region
496: \cite{zhao00:_secon_instab_compl_patter_anisot_convec}.  When
497: $\mu<-\beta$ and~(\ref{phi-steep}) holds, this implies $u=0$ for
498: bounded solutions.  Equations~(\ref{phi-steep},\ref{conserved}) are
499: necessary and sufficient conditions for the inner solutions to be
500: bounded when continued to $y\to\infty$.  When the shooting method is
501: used, the two conditions can be satisfied by adjusting the initial
502: value $f_0$ of $f$ at $y=0$ and the value of the control parameter
503: $\mu$.
504: 
505: With $C_1>1$ we could find neutrally stable modes of PJL with
506: $\mu<-\beta$ for all parameter sets tested, in particular for positive
507: as well as negative $\beta$.  For example,
508: Fig.~(\ref{fig:stable-mode}) shows the neutrally stable mode of the
509: PJL for parameters $\Gamma=-0.6$, $\beta=0.3$, $C_1=1.1$, and
510: $\tau=0.5$.  These parameters are suggested by numerical calculations
511: for similar systems~\cite{rohekrpe}.  For this case we verified that
512: there is no unstable mode, i.e., Fig.~(\ref{fig:stable-mode}) shows
513: the critical mode of a PJL.  (For smaller $|\Gamma|$, e.g. with
514: $\Gamma=-0.06$, there is another mode with positive growth rate.  Then
515: the feed-back loop through $\phi$ is too weak to stabilized the PJL.)
516: \begin{figure}[tbp]
517:   \centering
518:   \includegraphics[width=\linewidth,keepaspectratio,clip]{neutral_mode.eps}
519:   \caption{Typical neutral mode of a phase jump line at the critical
520:     control parameter. Solid: $\dy a_i$, dashed: $-\beta f$.}
521:   \label{fig:stable-mode}
522: \end{figure}
523: 
524: As $C_1$ is lowered, the critical $\mu$ approaches $-\beta$ and the
525: decay of the $f$ component in the far field becomes flatter, as
526: described by Eq.~(\ref{phi-steep}). Thus $a_i$ increases/decreases
527: approximately linearly over a wide range according
528: to~(\ref{conserved}).  For $\mu\to-\beta$, one expects the critical
529: mode to converge to an unbounded secular solution with constant $\dy
530: a_i$ and $f$ in the outer range.
531: 
532: It turns out that for $C_1=1$ the linearized
533: equations~(\ref{linearized}) have the analytic solution
534: \begin{align}
535:   \label{analytic-mode}
536:   f=1,\quad a_i=\sqrt{2}\,(1-\mu)+\mu\, y\, \tanh\frac{y}{\sqrt{2}}.
537: \end{align}
538: For $\mu=-\beta$ this solution satisfies conditions~(\ref{phi-steep})
539: and (\ref{conserved}) and is of the expected form.  Combined with our
540: numerical calculations, we conclude that the critical $\mu$ reaches
541: $-\beta$ exactly when $C_1=1$.  Thus, for $\beta_y>0$, the PJL can
542: stabilize prior to the zig-zag instability at $\mu=-\beta$ only if
543: $C_1>1$.  
544: %
545: %Unfortunately, the singularity at $\mu=-\beta$ itself it
546: %quite difficult to unfold analytically.
547: 
548: For $\beta<0$ the first instability of the homogeneous state and the
549: far field in system~(\ref{linearized}) is towards abnormal rolls at $\mu=0$.  
550: %% Even though for
551: %% $\beta$ slightly smaller than $0$ ... ? \cm{what's going on there?}
552: In contrast to the zig-zag instability at $\mu=-\beta$, the outer
553: equations~(\ref{outer}) have for $\mu=0$ no secular solutions with
554: $\sigma=0$.  Instead, the solution space corresponding to $\lambda=0$
555: [Eq.~(\ref{flat})] is spanned by two linearly independent constant
556: solutions, the pure phase shift $(a_i,f)=(1,0)$ and the pure director
557: rotation $(a_i,f)=(0,1)$.  Since both are bounded, bounded solutions
558: of~(\ref{linearized}) for $\mu=0$ are obtained whenever the
559: contribution to the outer solutions corresponding to Eq.~(\ref{steep})
560: with $\lambda>0$ vanishes.  When the shooting method is used, this can
561: be achieved by adjusting a single parameter, e.g. $f_0$.  Thus, at the
562: abnormal-roll instability $\mu=0$, a neutrally stable linear mode of
563: the PJL can always be found.
564: 
565: By the interaction of $f$ and $a_i$ in the core region, a certain
566: linear combination of the constant outer modes is singled out,
567: characterized by a specific ratio $r=a_i/f$ (for $y\to+\infty$).  When
568: $C_1=1$, for example, the analytic solution~(\ref{analytic-mode})
569: gives $r=\sqrt{2}$.  A perturbative analysis of the outer
570: equations~(\ref{linearized}) for small $\sigma\sim\dy\sim\mu$ under
571: the constraint $a_i/f=r+\mathcal{O}(\mu)$ yields a solution with
572: $a_i,f\sim\exp(\lambda_s y)$ (as $y\to+\infty$) where
573: \begin{subequations}
574: \label{muexpand}
575: \begin{align}
576:   \label{lambdas}
577:   \lambda_s=&\frac{2 r \tau |\Gamma| \mu}{2 r^2 |\Gamma|
578:     \tau-\beta}+O(\mu^2) 
579:   \intertext{with the growth rate}
580:   \label{sigmas}
581:   \sigma=&\frac{2 \beta |\Gamma| \mu}{2 r^2 |\Gamma|
582:     \tau-\beta}+O(\mu^2).
583: \end{align}
584: \end{subequations}
585: When $\lambda_s<0$ this solution is bounded and an eigenmode of the
586: PJL.  Since $\sigma/\lambda_s=\beta/r\tau+\mathcal{O}(\mu^2)$, this
587: mode can be stable only for $r<0$, which is the case only for values
588: of $C_1$ much smaller or much larger than one.  Usually one would
589: assume $r>0$.  With $-2|\Gamma|\, r^2\,\tau<\beta<0$, for example,
590: (\ref{muexpand}) then describe an unstable linear mode that becomes
591: neutrally stable as $\mu$ approaches zero from below.
592: 
593: In conclusion, these considerations show that within the framework of
594: the model~(\ref{Aphi}) a stabilization of PJL of the form~(\ref{PJL})
595: prior to any destabilization of the homogeneous state---as we seem to
596: observe it in the experiments---is possible, but likely only for
597: $C_1>1$ and $\beta\gtrsim 0$. 
598: 
599: \subsection{The effect of the twisted anchoring on $C_1$}
600: \label{sec:c1}
601: In order to test if the twisted anchoring used in our experiment can
602: cause a higher value of $C_1$, we analyzed a model for the $z$
603: dependence of director twist $\hat \phi(z)$ and convection mode $\hat
604: A(z)$ and their interaction \cite{softmodesX,twisted}.  Keeping only
605: terms that are relevant for determining $C_1$, the model reads in
606: appropriate dimensionless units
607: \begin{subequations}
608:   \label{zequations}
609:   \begin{align}
610:     \label{zA}
611:     \tau \dt \hat A=&\left[u-(P-\hat \phi)^2+\dz^2 \right] \hat A\\
612:     \label{zphi}
613:     \dt \hat \phi=&\dz^2 \hat \phi + 2\Gamma |\hat A|^2 (P-\hat \phi).
614:   \end{align}
615: \end{subequations}
616: The coefficient $u$ in Eq.~(\ref{zA}) is a control parameter that
617: depends on the applied voltage. $P$ corresponds to the angle of the
618: wavevector of a plane-wave convection mode to the $x$ axis and the
619: term $-(P-\hat \phi)^2$ in Eq.~(\ref{zA}) describes the preference of
620: the convective modes to align normal to the director.  The last term
621: ($\dz^2$) describes the diffusion of hydrodynamic variables along $z$.
622: In Eq.~(\ref{zphi}) the first term describes the stiffness of the
623: in-plane director to twist and the second term the torque generated by
624: misaligned rolls on the director.  We assume layers of thickness $2$
625: without loss of generality and impose the boundary conditions $\hat
626: A(\pm 1)=0$ and, corresponding the twisted anchoring, $\hat
627: \phi(+1)=b$, $\hat \phi(-1)=-b$.  The isotropy of the system is broken
628: only by the boundary condition.  In the bulk, isotropy is preserved,
629: which is reflected by the equivariance under $P\to P+\delta$, $\hat
630: \phi\to\hat \phi+\delta$.  The model can be derived analytically in
631: the limit that the wavelength of the roll pattern is much smaller than
632: the layer thickness \cite{softmodesX}.  However, since this condition
633: is not satisfied for our experimental system, we expect the model to
634: describe the experiment only qualitatively.
635: 
636: The coefficient $C_1$ entering the 2d description~(\ref{Aphi}) can be
637: obtained by the following method.  Noting that in the basic state
638: ($\hat A=0$) the director relaxes to $\hat \phi=b z$, and that the
639: fundamental mode of linear excitations of $\hat \phi$ is $\hat
640: \phi_1=\cos(\pi z/2)$, we assume a weakly excited $\hat \phi=b z+K
641: \hat \phi_1$ and search numerically for the threshold value
642: $u_\text{crit}$ of the control parameter $u$ at which the lowest
643: eigenmode $\hat A(z)=\hat A_\text{crit}(z)$ of Eq.~(\ref{zA}) becomes
644: critical ($\dt=0$).  The orientation $P$ of the roll wavevector is
645: chosen such as to minimize $u_\text{crit}$.  This corresponds to
646: letting $A=\exp(iPy)$ in Eq.~(\ref{Aphi-A}) with $P=C_1 \phi$.
647: Following general prescriptions (e.g.\ \cite{manbook,bkbook}), the
648: value of $C_1$ is then given by the ratio of projections of two
649: components of the r.h.s.\ of Eq.~(\ref{zphi}) onto its adjoint
650: eigenvector ($\equiv \hat \phi_1$)
651: \begin{align}
652:   \label{c1}
653:   C_1=\frac{\displaystyle \int_{-1}^1 \hat \phi_1 \, 2\Gamma |\hat A|^2 P\, dz}
654:   {\displaystyle \int_{-1}^1 \hat \phi_1 \, 2\Gamma |\hat A|^2 \hat \phi\, dz}
655: \end{align}
656: in the limit $K \to 0$.  
657: 
658: Without twist ($b=0$) we reproduce the analytic result \footnote{This
659:   result does not depend on the absence or presence of
660:   flexoelectric effects, as a note in Ref.~\cite{Komineas003} might
661:   suggest.}  $C_1=256/(27\pi^2)=0.96$ \cite{linddpl,softmodesX}.  With
662: increasing twist $b$, the value of $C_1$ increases, until it reaches a
663: maximum $C_1=1.14$ at $b=8.1$ (Fig.~\ref{fig:c1}).  The result is
664: independent of all other system parameters.  Due to the purely
665: geometric nature of this effect, it is plausible to assume that the
666: twist leads to an increased value of $C_1$ also in our experimental
667: system.
668: \begin{figure}[bth]
669:   \centering
670:   \includegraphics[width=\linewidth,keepaspectratio,clip]{c1.eps}
671:   \caption{The geometry factor $C_1$ in dependence of the twist $b$
672:   for the simple model~(\ref{zequations}).  The dash-dotted line at
673:   $C_1= 1$ is a guide to the eye.}
674:   \label{fig:c1}
675: \end{figure}
676: 
677: \subsection{Coupling to zig-zag modulations}
678: \label{sec:zig-zag}
679: Simulations~\cite{zhao00:_secon_instab_compl_patter_anisot_convec}
680: show that PJL can also coexist with zig-zag modulations of the roll
681: pattern.  The PJL are then located at preferred phase angles of the
682: zig-zag modulation.  Thus, the distance $L$ between neighbouring PJL
683: will be controlled by the wavelength $\Lambda$ of the zig-zag
684: modulation.  Even though they are not clearly visible in the
685: experiments, small zig-zag modulations are probably also controlling
686: the distance between PJL in PJL grids.
687: 
688: To test this hypothesis, the wavelength of zig-zags is estimated.  For
689: an easy comparison with experiments, dimensional units are used in
690: this calculation.  In the absences of strong variations of the pattern
691: amplitude $|A|$, the dynamics of the roll pattern can be described in
692: terms of the phase $\theta=\arg A\,(\mathop\mathrm{mod}2\pi)$ of the
693: pattern alone.  Together with the coupling to $\phi$, this is
694: description becomes \cite{rzkr,Zhao000}
695: \begin{subequations}
696:   \label{thetaphi}
697:   \begin{align}
698:     \label{thetaphi-theta}
699:     \tau\dt \theta=&
700:     \xi_x^2\dx^2\theta+\xi_y^2\dy^2\theta+\beta \dy \phi+r\phi^2\dy\phi, \\
701:     \label{thetaphi-phi}
702:     \gamma_1\dt\phi=&
703:     2|\Gamma|(\epsilon-\epsilon_\text{AR})\phi-g\phi^3+K_3\dx^2\phi+
704:     K_1\dy^2\phi-2|\Gamma|\epsilon\dy\theta
705:   \end{align}
706: \end{subequations}
707: with positive $r,g$, the threshold of abnormal rolls
708: $\epsilon_\text{AR}=h/(2|\Gamma|)$ and all other parameters identical
709: to those entering Eq.~(\ref{Aphi-dim}).  While the derivation of the
710: nonlinear terms in this system hinges on the assumption of that
711: spatial and temporal variations of $\theta$ and $\phi$ are slow on the
712: scales $\xi_{x,y}\epsilon^{-1}$ and $\tau\epsilon^{-1}$, which is
713: strictly satisfied only for small $\epsilon-\epsilon_\text{AR}$ and
714: $\beta$, the linear part is valid uniformly in $\epsilon\le0$ and
715: $\beta$, as long as the Ginzburg-Landau description (\ref{Aphi-dim})
716: is valid.
717: 
718: For stationary solutions that depend only on $y$ ($\dt,\dx=0$)
719: Eq.~(\ref{thetaphi-theta}) can be integrated to
720: \begin{align}
721:   \label{theta-integrated}
722:   \xi_y^2\dy \theta+\beta \phi+r\phi^3/3=J=\text{const.}
723: \end{align}
724: In the simple case that rolls are not tilted on the average ($\left<
725:   \dy\theta \right>_y=0$) the integration constant $J$ vanishes.
726: Using Eq.~(\ref{theta-integrated}) to eliminate $\theta$ from
727: Eq.~(\ref{thetaphi-phi}) then yields an equation for $\phi$ of the
728: form~\cite{Zhao000}
729: \begin{align}
730:   \label{phi-alone}
731:   K_1\dy^2\phi=-2 |\Gamma|((1+\beta/\xi_y^2)
732:   \epsilon-\epsilon_\text{AR})\phi+g'(\epsilon)\phi^3.
733: \end{align}
734: While the wavelength of zig-zag modulations can become arbitrary large
735: when the nonlinearity in Eq.~(\ref{phi-alone}) becomes effective, the
736: wavelength of small-amplitude modulations is sharply determined as
737: \begin{align}
738:   \label{wavelength1}
739:   \begin{split}
740:     \Lambda=&2\pi \sqrt{\frac{K_1}{2|\Gamma|
741:         ((1+\beta/\xi_y^2)\epsilon-\epsilon_\text{AR})}}\\
742: =&2\pi \sqrt{\frac{K_1
743:         \epsilon_\text{ZZ}}{2|\Gamma|\epsilon_\text{AR}
744:         (\epsilon-\epsilon_\text{ZZ})}}.
745: \end{split}
746: \end{align}
747: where $\epsilon_\text{ZZ}=\epsilon_\text{AR}/(1+\beta/\xi_y^2)$
748: denotes the threshold of the zig-zag instability.  In order to
749: estimate the value of $2|\Gamma|\epsilon_\text{AR}/K_1$, recall that
750: the linear part of Eq.~(\ref{thetaphi-phi}) holds also for
751: $\epsilon=0$ (no convection).  In this case the relaxation of small
752: in-plane rotations of the director is easily described by the
753: linearized Leslie-Erickson \cm{spell?} equations.  For the untwisted
754: case, for example, they reduce to
755: \begin{align}
756:   \label{LE}
757:   \gamma_1\dt n_y=(k_{33}\dx^2+k_{11}\dy^2+k_{22}\dz^2) n_y.
758: \end{align}
759: Considering only the slowest mode of the system for fixed boundary
760: conditions at $z=\pm d/2$, $n_y=\phi \cos(\pi z/d)$, and projecting
761: onto the corresponding adjoint $\cos(\pi z/d)$, yields
762: \begin{align}
763:   \label{rederived}
764:   \gamma_1\dt \phi=(k_{33}\dx^2+k_{11}\dy^2-k_{22}\frac{\pi^2}{d^2}) \phi.
765: \end{align}
766: Comparing of Eqs.~(\ref{thetaphi-phi}) and (\ref{rederived}) shows
767: that $2|\Gamma|\epsilon_\text{AR}/K_1=\pi^2 k_{22}/d^2 k_{11}$.  For
768: MBBA at 25$^\circ$C, for example, the ratio $k_{11}/k_{22}$ is
769: $\approx 1.6$.  For the twisted geometry one would expect some mixture
770: of $k_{11}$, $k_{22}$, and $k_{33}$ to enter $K_1$ instead of just
771: $k_{11}$, but since in the projection onto the adjoined mode the
772: central part $z\approx 0$ dominates, and there the director is still
773: aligned parallel to $\hat x$, these corrections are presumably small.
774: Assuming a relation $L=m\,\Lambda$ between zig-zag wavelength
775: $\Lambda$ and PJL distance $L$, one obtains the result
776: \begin{align}
777:   \label{L}
778:  \begin{split}
779:   L=&2 d m\sqrt{\frac{k_{11}
780:   \epsilon_{ZZ}}{k_{22}(\epsilon-\epsilon_{ZZ})}}. \\
781:   \Leftrightarrow
782:   K=&\frac{\pi}{dm}\sqrt{\frac{k_{22}}{k_{11}}\frac{\epsilon-\epsilon_{ZZ}}{\epsilon_{ZZ}}}. 
783:  \end{split}
784: \end{align}
785: Even when $\epsilon_\text{ZZ}$ is fitted from the measured data for
786: $K$, Eq.~(\ref{L}) makes a nontrivial prediction because
787: $\epsilon_\text{ZZ}$ appears at two different positions in the
788: formula.
789: 
790: Using the fitted experimental curves, we find that Eq.~(\ref{L}) holds
791: with $m=2.1$ for the $400\,\mathrm{Hz}$ data and with $m=0.94$ for the
792: $600\,\mathrm{Hz}$ data.  That is, in both cases we find a value of
793: $m$ of order one.  In view of the coarseness of the theoretical
794: estimates, and allowing for some experimental error, we conclude that
795: the observations fully supports the hypothesis that the lattice
796: constant of the PJL-grid is controlled by small zig-zag modulations.
797: The mechanism by which the amplitude of zig-zag modulations or,
798: equivalently, the zig-zag wavelength is kept small remains unclear,
799: however.
800: 
801: \section{Conclusion}
802: In this study, we reported the observation of stable PJL and described
803: a process by which these PJL from a lattice that finally evolves to a
804: structure indistinguishable from the well know grid pattern as the
805: applied voltage increases.  We argued that the stabilization of PJL is
806: presumably an indirect effect of the twisted geometry of the nematic
807: in the basic state: The twist leads to values of the coefficient
808: $C_1>1$, which in turn enables the stabilization of PJL prior to a
809: destabilization of the roll pattern.
810: 
811: \bibliographystyle{apsrev}
812: %perhaps we can have multiple bib files:
813: \bibliography{bibview1_3,bibtatsumi} %bib file of A.R.
814: %\bibliography{bibtatsumi} %bib file of TATSUMI Soichi 
815: %\bibliography{bibfilename} %your bib file  
816: 
817: %\end{CJK*}
818: \end{document}
819: 
820: %%% Local Variables: 
821: %%% mode: latex
822: %%% mode: flyspell
823: %%% TeX-master: t
824: %%% End: 
825: 
826: % LocalWords:  euc jp axel JIS Soichi Tatsumi Hongo Bunkyo ku Tokiwadai Masaki
827: % LocalWords:  Hodogaya Sano spatio methoxybenzylidene butylaniline MBBA PJL ar
828: % LocalWords:  butylammonium PJLs Ginzburg eigenmode Eq ZZ relaxational GLE PDE
829: % LocalWords:  annihilations Im Eqs const crit Lindner eigenmodes
830: 
831: