1: % This is file JFM2esam.tex
2: % first release v1.0, 20th October 1996
3: % release v1.01, 29th October 1996
4: % release v1.1, 25th June 1997
5: % (based on JFMsampl.tex v1.3 for LaTeX2.09)
6: % Copyright (C) 1996, 1997 Cambridge University Press
7:
8: \NeedsTeXFormat{LaTeX2e}
9:
10: \documentclass{jfm}
11: \usepackage{epsfig}
12:
13: % See if the author has AMS Euler fonts installed: If they have, attempt
14: % to use the 'upmath' package to provide upright math.
15:
16: \ifCUPmtlplainloaded \else
17: \checkfont{eurm10}
18: \iffontfound
19: \IfFileExists{upmath.sty}
20: {\typeout{^^JFound AMS Euler Roman fonts on the system,
21: using the 'upmath' package.^^J}%
22: \usepackage{upmath}}
23: {\typeout{^^JFound AMS Euler Roman fonts on the system, but you
24: dont seem to have the}%
25: \typeout{'upmath' package installed. JFM.cls can take advantage
26: of these fonts,^^Jif you use 'upmath' package.^^J}%
27: \providecommand\upi{\pi}%
28: }
29: \else
30: \providecommand\upi{\pi}%
31: \fi
32: \fi
33:
34: % See if the author has AMS symbol fonts installed: If they have, attempt
35: % to use the 'amssymb' package to provide the AMS symbol characters.
36:
37: \ifCUPmtlplainloaded \else
38: \checkfont{msam10}
39: \iffontfound
40: \IfFileExists{amssymb.sty}
41: {\typeout{^^JFound AMS Symbol fonts on the system, using the
42: 'amssymb' package.^^J}%
43: \usepackage{amssymb}%
44: \let\le=\leqslant \let\leq=\leqslant
45: \let\ge=\geqslant \let\geq=\geqslant
46: }{}
47: \fi
48: \fi
49:
50: % See if the author has the AMS 'amsbsy' package installed: If they have,
51: % use it to provide better bold math support (with \boldsymbol).
52:
53: \ifCUPmtlplainloaded \else
54: \IfFileExists{amsbsy.sty}
55: {\typeout{^^JFound the 'amsbsy' package on the system, using it.^^J}%
56: \usepackage{amsbsy}}
57: {\providecommand\boldsymbol[1]{\mbox{\boldmath $##1$}}}
58: \fi
59:
60:
61: %%% Example macros (some are not used in this sample file) %%%
62:
63: % For units of measure
64: \newcommand\dynpercm{\nobreak\mbox{$\;$dynes\,cm$^{-1}$}}
65: \newcommand\cmpermin{\nobreak\mbox{$\;$cm\,min$^{-1}$}}
66:
67: % Various bold symbols
68: \providecommand\bnabla{\boldsymbol{\nabla}}
69: \providecommand\bcdot{\boldsymbol{\cdot}}
70: \newcommand\biS{\boldsymbol{S}}
71: \newcommand\etb{\boldsymbol{\eta}}
72: \newcommand\xib{\boldsymbol{\xi}}
73: \newcommand\psib{\boldsymbol{\psi}}
74: \newcommand\Gb{\boldsymbol{\Gamma}}
75: \newcommand\Lamb{\boldsymbol{\Lambda}}
76: \newcommand\bchi{\boldsymbol{\chi}}
77: \newcommand\bPhi{\boldsymbol{\Phi}}
78: % For multiletter symbols
79: \newcommand\Real{\mbox{Re}} % cf plain TeX's \Re and Reynolds number
80: \newcommand\Imag{\mbox{Im}} % cf plain TeX's \Im
81: \newcommand\Rey{\mbox{\textit{Re}}} % Reynolds number
82: \newcommand\Pran{\mbox{\textit{Pr}}} % Prandtl number, cf TeX's \Pr product
83: \newcommand\Pen{\mbox{\textit{Pe}}} % Peclet number
84: \newcommand\Ai{\mbox{Ai}} % Airy function
85: \newcommand\Bi{\mbox{Bi}} % Airy function
86:
87: % For sans serif characters:
88: % The following macros are setup in JFM.cls for sans-serif fonts in text
89: % and math. If you use these macros in your article, the required fonts
90: % will be substitued when you article is typeset by the typesetter.
91: %
92: % \textsfi, \mathsfi : sans-serif slanted
93: % \textsfb, \mathsfb : sans-serif bold
94: % \textsfbi, \mathsfbi : sans-serif bold slanted (doesnt exist in CM fonts)
95: %
96: % For san-serif roman use \textsf and \mathsf as normal.
97: %
98: \newcommand\ssC{\mathsf{C}} % for sans serif C
99: \newcommand\sfsP{\mathsfi{P}} % for sans serif sloping P
100: \newcommand\slsQ{\mathsfbi{Q}} % for sans serif bold-sloping Q
101:
102: % Hat position
103: \newcommand\hatp{\skew3\hat{p}} % p with hat
104: \newcommand\hatR{\skew3\hat{R}} % R with hat
105: \newcommand\hatRR{\skew3\hat{\hatR}} % R with 2 hats
106: \newcommand\doubletildesigma{\skew2\tilde{\skew2\tilde{\Sigma}}}
107: % italic Sigma with double tilde
108:
109: % array strut to make delimiters come out right size both ends
110: \newsavebox{\astrutbox}
111: \sbox{\astrutbox}{\rule[-5pt]{0pt}{20pt}}
112: \newcommand{\astrut}{\usebox{\astrutbox}}
113:
114: \newcommand\GaPQ{\ensuremath{G_a(P,Q)}}
115: \newcommand\GsPQ{\ensuremath{G_s(P,Q)}}
116: \newcommand\p{\ensuremath{\partial}}
117: \newcommand\tti{\ensuremath{\rightarrow\infty}}
118: \newcommand\kgd{\ensuremath{k\gamma d}}
119: \newcommand\shalf{\ensuremath{{\scriptstyle\frac{1}{2}}}}
120: \newcommand\sh{\ensuremath{^{\shalf}}}
121: \newcommand\smh{\ensuremath{^{-\shalf}}}
122: \newcommand\squart{\ensuremath{{\textstyle\frac{1}{4}}}}
123: \newcommand\thalf{\ensuremath{{\textstyle\frac{1}{2}}}}
124: \newcommand\Gat{\ensuremath{\widetilde{G_a}}}
125: \newcommand\ttz{\ensuremath{\rightarrow 0}}
126: \newcommand\ndq{\ensuremath{\frac{\mbox{$\partial$}}{\mbox{$\partial$} n_q}}}
127: \newcommand\sumjm{\ensuremath{\sum_{j=1}^{M}}}
128: \newcommand\pvi{\ensuremath{\int_0^{\infty}%
129: \mskip \ifCUPmtlplainloaded -30mu\else -33mu\fi -\quad}}
130:
131: \newcommand\etal{\mbox{\textit{et al.}}}
132: \newcommand\etc{etc.\ }
133: \newcommand\eg{e.g.\ }
134: \newcommand\GammaA{\Gamma_{A}}
135: \newcommand\GammaB{\Gamma_{B}}
136: \newcommand{\dint}{\displaystyle\int}
137: \newcommand{\dsum}{\displaystyle\sum}
138: \newcommand{\dfrac}{\displaystyle\frac}
139: \newtheorem{lemma}{Lemma}
140: \newtheorem{corollary}{Corollary}
141:
142:
143: \def\a{\alpha}
144: \def\b{{\bf b}}
145: \def\g{\gamma}
146: \def\c{{\bf c}}
147: \def\cd{\!\cdot\!}
148: \def\d{\delta}
149: \def\dd{\mbox{d}}
150: \def\ve{\varepsilon}
151: \def\eps{\epsilon}
152: \def\o{\omega}
153: \def\r{{\bf r}}
154: \def\k{{\bf k}}
155: \def\q{{\bf q}}
156: \def\k{{\bf k}}
157: \def\u{{\bf u}}
158: \def\v{{\bf v}}
159: \def\x{{\bf x}}
160: \def\y{{\bf y}}
161: \def\l{\ell}
162: \def\o{\omega}
163: \def\p{{\bf p}}
164: \def\s{\sigma}
165: \def\vp{\varphi}
166: \def\D{\Delta}
167: \def\G{{\bf G}}
168: \def\H{{\bf H}}
169: \def\K{{\bf K}}
170: \def\L{{\bf L}}
171: \def\Q{{\bf Q}}
172: \def\R{{\bf R}}
173: \def\U{{\bf U}}
174: \def\V{{\bf V}}
175: \def\W{{\bf W}}
176: \newcommand{\gb}[1]{{\bf#1}}
177: \newcommand{\gbout}[1]{}
178:
179: \begin{document}
180:
181: \title[KH Instability with Periodic Plates]
182: %{Stop Bands in the Kelvin-Helmholtz Instability}
183: {Kelvin-Helmholtz instabilities across periodic plates}
184:
185: %{Bragg Scattering of the Kelvin-Helmholtz Instability}
186:
187:
188: \author[Tom Chou]
189: {Tom Chou}
190:
191: \affiliation{Dept. of Biomathematics and
192: Dept. of Mathematics, UCLA, Los Angeles, CA 90095-1766 USA}
193:
194: \maketitle
195:
196: \begin{abstract}
197: We consider the linear stability of two inviscid fluids, in the
198: presence of gravity, sheared past each other and separated by an
199: flexible plate. Conditions for exponential growth of velocity
200: perturbations are found as functions of the flexural rigidity of the
201: plate and the shear rate. This Kelvin-Helmholtz instability is then
202: analysed in the presence of plates with spatially periodic (with
203: period $a$) flexural rigidity arising from, for example, a periodic
204: material variation. The eigenvalues of this periodic system are
205: computed using Bloch's Theorem (Floquet Theory) that imposes specific
206: Fourier decompositions of the velocity potential and plate
207: deformations. We derive the nonhermitian matrix whose eigenvalues
208: determine the dispersion relation. Our dispersion relation shows
209: that plate periodicity generally destabilises the flow, compared to a
210: uniform plate with the same mean flexural rigidity. However, enhanced
211: destabilisation and stabilization can occur for disturbances with
212: wavelengths near an even multiple of the plate periodicity. The
213: sensitivity of flows with such wavelengths arises from the
214: nonpropagating, ``Bragg reflected'' modes coupled to the plate
215: periodicity through the boundary condition at the plate.
216: \end{abstract}
217:
218:
219: \section{Introduction}\label{sec:INTRO}
220:
221: When two immiscible fluids are sheared relative to each other, the
222: flow, and the normally flat interface between them, may become
223: unstable. This classical Kelvin-Helmholtz instability
224: (\cite{CHAND,DRAZIN,WHITHAM}) arises from perturbations of the
225: fluid velocity that grows into a vortex sheet due to local Bernoulli
226: variations in pressure. The presence of gravity also tunes this
227: instability, depending on the mass density difference and shear rate
228: between the two fluids. If the denser fluid is in lower
229: gravitational potential, there is a critical shear rate beyond which
230: small perturbations grow exponentially. This instability first arises
231: in modes of the fluid velocity perturbation that have wavevectors
232: greater than a critical wavevector. All velocity perturbations with
233: sufficiently small wavelength become unstable as long as the shear
234: rate exceeds a critical value. However, very small wavelength
235: perturbations can be stabilised if additional restoring forces from,
236: say, an interfacial surface tension (\cite{WHITHAM}), or a separating
237: elastic plate are included. The Kelvin-Helmholtz instability, and its
238: generalisations, have been thoroughly studied, for example, in the
239: context of knotting in astrophysical jets (\cite{REES}), cloud top
240: entrainment instabilities (\cite{REISS}), magneto-hydrodynamic
241: instabilities (\cite{CHAND} and \cite{FRANK}), and the flutter of panels
242: (\cite{MILES}).
243:
244: In the context of potential applications involving fluid-structure
245: interactions, we consider the Kelvin-Helmholtz instability in the
246: presence of an infinitesimally thin bendable plate separating the two
247: inviscid flows, as shown in Figure \ref{SLAB0}. We assume the plate
248: is thin and inextensible, such that only bending modes can be excited.
249: Since elastic restoring forces stabilize short wavelength
250: perturbations, instabilities arise only for perturbations that have
251: wavevectors within a certain window of values. Furthermore, we
252: consider an elastic membrane that has spatially periodic properties,
253: such as mass density or bending rigidity, which might arise, for
254: example, in an engineered composite structure. By itself (in the
255: absence of fluid), a periodic plate supports stable bending waves that
256: propagate {\it except} at band gaps, {\it i.e.} frequencies
257: corresponding to waves that have half-integer wavelengths in one
258: period of the structural variation (\cite{AM} and \cite{LIGHT}). Waves
259: near these wavelengths have vanishing group velocities and do not
260: propagate in an infinite periodic medium due to their multiple
261: ``Bragg'' reflections and deconstructive interference. Properties of
262: wave propagation through periodic media have been computed in numerous
263: physical settings, including acoustic waves moving through materials
264: with periodic sound speeds (\cite{SIGALAS}), electromagnetic
265: propagation through periodic dielectrics (\cite{LIGHT}), and water
266: wave propagation over periodic structures
267: (\cite{CHEN,CHOU,CHAN,HU,MEI1,PETER,PORTER}).
268:
269: \begin{figure}
270: \begin{center}
271: \includegraphics[height=2.0in]{slab0.eps}
272: \end{center}
273: \caption{Discontinuous shear flow separated by a periodically flexible
274: plate.}
275: \label{SLAB0}
276: \end{figure}
277:
278: In this paper, we combine the periodic stabilizing membrane with an
279: otherwise unstable shear flow and study the complex dispersion
280: relation. A velocity perturbation with any wavevector, convected by
281: the base laminar flow forms a travelling wave. This travelling wave,
282: when coupled to the intrinsic periodicity of the elastic plate may be
283: Bragg reflected and not able to propagate within the region of
284: periodicity, altering its stability characteristics.
285: %Within this context, we explore the effects of a periodic restoring force on the
286: %Kelvin-Helmholtz instability.
287: In order to explore this heuristic argument further, we formulate in
288: the next section the linear problem in terms of a vorticity-free
289: velocity perturbation and a small deformation of the separating
290: elastic membrane. We review, in turn, a uniform plate subject to
291: destabilizing flow, and a free periodic plate that supports stable
292: waves. The fully coupled problem, a periodic plate in destabilising
293: flow, is then presented in the Results section. We show that the
294: stability properties are affected most when the wavelengths of the
295: most unstable modes are close to the periodicity of the plate
296: flexibility. Implications and extensions to related flow systems are
297: discussed the Discussion and Conclusions.
298:
299: \section{Linear Model}\label{sec:LINEAR}
300:
301: Consider the fluid-structure interaction depicted in Figure
302: \ref{SLAB0}. Two incompressible fluids flowing uniformly in two
303: different directions, ${\bf U}_{+}$ and ${\bf U}_{-}$, in the $x-y$
304: plane are separated by a negligibly thin elastic plate. Gravitational
305: acceleration is in the $-{\bf z}$ direction. Since viscosity will be
306: neglected, we will consider only potential flows. The total velocity
307: in each region ($\pm$) is defined by the uniform flow $\U_{\pm}$ plus
308: a small irrotational perturbation $\v_{\pm} \equiv \nabla \vp_{\pm}$.
309: Upon imposing incompressibility, the upper and lower velocity
310: potentials, $\vp_{\pm}$, satisfy Laplace's equation:
311:
312: \begin{equation}
313: (\nabla_{\perp}^{2} + \partial_{z}^{2})\vp_{\pm}(\r,z,t) = 0,\quad
314: \nabla_{\perp}^{2} = \partial_{x}^{2}+\partial_{y}^{2}.
315: \label{LAPLACE}
316: \end{equation}
317:
318: Since $\partial_{z}\varphi\vert_{z=0} \approx v_{z}$ is the vertical
319: velocity of the surface, the kinematic conditions at the
320: plate-separated interface are
321:
322: \begin{equation}
323: \partial_{t}h(\r,t) + \U_{\pm}\cdot\nabla_{\perp}h(\r,t) -
324: \lim_{\ve\rightarrow 0}\partial_{z}\vp_{\pm}(\r,\pm\ve,t) = 0,
325: \label{KINEMATIC}
326: \end{equation}
327:
328: \noindent where $h(\r,t)$ is the normal displacement of the plate.
329: The linearized normal stress balance at the interface,
330: including the normal forces arising from the flexible plate, is
331:
332: \begin{equation}
333: \left[\rho\dot{\vp}+\rho\U\cd\nabla_{\perp}\vp+\rho gh(\r,t)\right]^{+}_{-}
334: =m(\r)\ddot{h}(\r,t) + \nabla_{\perp}^{2}(D(\r)\nabla_{\perp}^{2})h(\r,t),
335: \label{NSTRESS}
336: \end{equation}
337:
338: \noindent where $\left[X\right]^{+}_{-} \equiv \lim_{\ve \rightarrow
339: 0} X(z=\ve)-X(z=-\ve)$. In Eq. \ref{NSTRESS}, $m(\r)$ and $D(\r)$ are the
340: mass per unit area and the flexural rigidity of the plate,
341: respectively. Since $h(\r,t)$ and $\vp(\r,z,t)$ are defined
342: only up to a constant, we need only consider the nonzero wavevector
343: decompositions of the plate displacement and velocity potential,
344:
345: \begin{equation}
346: \displaystyle h(\r,t) = \sum_{\q\neq 0}\!\int\!h_{\q}(\omega)
347: e^{i\q\cdot\r-i\omega t}{\dd \o \over 2\pi} \quad \mbox{and} \quad
348: \vp_{\pm}(\r,z,t) = \sum_{\q\neq 0}\!\displaystyle
349: \int\!\varphi^{\pm}_{\q}(\omega) e^{i\q\cdot\r-i\omega t
350: \mp\vert\G-\q\vert z}{\dd \o \over 2\pi}.
351: \label{EXPANSIONS}
352: \end{equation}
353:
354:
355: %\begin{equation}
356: %\begin{array}{rl}
357: %\displaystyle h(\r,t) & \displaystyle = \sum_{\q}'\displaystyle \int{\dd \o \over 2\pi}
358: %e^{i\q\cdot\r-i\omega t}h_{\q}(\omega) \qquad \mbox{and} \\[13pt]
359: %\displaystyle \vp_{\pm}(\r,z,t) & \displaystyle = \sum_{\q}'\displaystyle \int{\dd
360: %\o \over 2\pi} \varphi^{\pm}_{\q}(\omega)
361: %e^{i\q\cdot\r-i\omega t}e^{\mp\vert\G-\q\vert z},
362: %\end{array}
363: %\label{EXPANSIONS}
364: %\end{equation}
365:
366: %\noindent where $\sum_{\q}'$ signifies the sum $\sum_{\q\neq 0}$
367: %that excludes the zero-wavevector component.
368:
369: \noindent In Eqs. \ref{EXPANSIONS}, we have also assumed a continuous
370: time-harmonic decomposition. The representation of $\vp_{\pm}(\r, z,
371: t)$ automatically satisfies Eq. \ref{LAPLACE} and the boundary
372: conditions at $z\rightarrow \pm\infty$.
373:
374: The periodic plate mass density and rigidity functions that obey
375: $m(\r)= m(\r+\R)$ and $D(\r)=D(\r+\R)$, where $\R=n_{1}a_{1}{\bf
376: e}_{1} + n_{2}a_{2}{\bf e}_{2}\,$ ($n_{1},n_{2}$ integers) is any
377: basis vector describing the periodicity. Here, $a_{1}, a_{2}$ are the
378: periodicities of $m(\r)$ and $D(\r)$ in the ${\bf e}_{1}, {\bf e}_{2}$
379: directions. The periodic functions $m(\r)$ and $D(\r)$ can thus be
380: decomposed into a Fourier sum over reciprocal lattice vectors $\G$,
381:
382:
383: \begin{equation}
384: m(\r)=\sum_{G}m(\G)e^{i\G\cdot\r} \,\quad \mbox{and}\quad\,
385: D(\r) = \sum_{G}D(\G)e^{i\G\cdot\r}.
386: \label{DECOMP}
387: \end{equation}
388:
389:
390: \noindent The reciprocal lattice vectors $\G = n_{1}\G_{1} + n_{2}\G_{2}$
391: are defined by the reciprocal lattice basis vectors
392:
393: \begin{equation}
394: \G_{1} = {2\pi\over a_{1}} {{\bf e}_{2}\times{\bf z}\over
395: {\bf e}_{1}\cdot({\bf e}_{2}\times {\bf z})} \,\quad\mbox{and}\quad\,
396: \G_{2} = {2\pi\over a_{2}} {{\bf z}\times{\bf e}_{1}\over
397: {\bf e}_{1}\cdot({\bf e}_{2}\times {\bf z})}.
398: \end{equation}
399:
400:
401: Finally, we assume a large plate with dimensions $L_{1}\times L_{2}$
402: that fit an integer multiple of the period ($L_{1} = N_{1}a_{1}$ and
403: $L_{2} = N_{2}a_{2}$) and impose periodic boundary
404: conditions on the entire system: $m(\r + N_{i}a_{i}{\bf e}_{i}) =
405: m(\r)$ and $D(\r + N_{i}a_{i}{\bf e}_{i}) = D(\r)$ for $i=1,2$. Under
406: this construction, the wavevector sum in the decompositions
407: \ref{EXPANSIONS} are taken over $\q = \sum_{n_{i}\neq
408: 0}(n_{i}/N_{i})\G_{i}$. In the $L_{1},L_{2}\rightarrow \infty$ limit,
409: sums over $\q$ become the Cauchy principle value of the corresponding
410: Riemann integral.
411:
412:
413: After substitution of Eqs. \ref{EXPANSIONS} into the
414: kinematic conditions \ref{KINEMATIC}, we find
415: the relationship
416:
417: \begin{equation}
418: \varphi^{\pm}_{\q}(\o) = \pm {i(\o - \U_{\pm}\cd\q)\over \vert
419: \q\vert}h_{\q}(\o), \qquad \q\neq 0.
420: \label{PHIH}
421: \end{equation}
422:
423: \noindent Upon substitution of Eqs. \ref{EXPANSIONS}, \ref{DECOMP},
424: and \ref{PHIH} into Eq. \ref{NSTRESS}, and exploiting the
425: orthogonality of the $e^{i\q\cdot\r}$ basis functions,
426:
427: \begin{equation}
428: \begin{array}{l}
429: \left[(\rho_{+}+\rho_{-})\o^{2} -
430: 2\o (\rho_{+}\U_{+} + \rho_{-}\U_{-})\cd\q +
431: \rho_{+}(\q\cd\U_{+})^{2}+\rho_{-}(\q\cd\U_{-})^{2} +
432: (\rho_{+}-\rho_{-})g\vert\q\vert\right]h_{\q} \\[13pt]
433: \:\hspace{2cm} =
434: -\o^{2}\vert\q\vert\sum_{\G'}m(\G')h_{\q-\G'} +
435: \sum_{\G'}\vert\q\vert^{3}\vert\q-\G'\vert^{2}D(\G')h_{\q-\G'}.
436: \end{array}
437: \label{INTEGRAL}
438: \end{equation}
439:
440: \noindent Equation \ref{INTEGRAL} shows that each coefficient $h_{\q}$
441: is linked with $N_{1}\times N_{2}$ other coefficients
442: $h_{\q-\G'}$. This problem can be solved numerically by truncating
443: $h_{\q-\G} \approx 0$ for large $\vert\G\vert$ where $m(\G)$ and
444: $D(\G)$ are small. Alternatively, Eq. \ref{INTEGRAL} can be organized
445: according to the ``reduced zone scheme,'' by defining $\k = \q-\G$,
446: where $\G$ is such that $\k$ is restricted in the first Brillouin zone
447: defined by $\vert\k\cdot{\bf e}_{i}\vert < \vert\G_{i}\cdot{\bf
448: e}_{i}\vert/2$ (\cite{AM}). In this representation, we need only
449: consider the $N_{1}\times N_{2}$ eigenvalues, each with index $\k$
450: inside the first Brillouin zone. The problem is thus cast into a
451: quadratic eigenvalue problem defined by
452:
453: \begin{equation}
454: \sum_{\G'}\left[A_{\k}(\G,\G')\omega^{2} + B_{\k}(\G,\G')\omega +
455: C_{\k}(\G,\G')\right]h_{\k-\G'}(\o) = 0,
456: \label{MATRIX2}
457: \end{equation}
458:
459:
460: \noindent where
461:
462: \begin{equation}
463: \begin{array}{l}
464: \displaystyle A_{\k}(\G,\G') = {\rho_{+}+\rho_{-} \over
465: \vert\k-\G\vert}\d_{\G,\G'} + m(\G'-\G) \\[13pt] \displaystyle
466: B_{\k}(\G,\G') = - 2\left[\rho_{+}{\U_{+}\cd(\k-\G)\over \vert\k-\G\vert} +
467: \rho_{-}{\U_{-}\cd(\k-\G)\over
468: \vert\k-\G\vert}\right]\d_{\G,\G'}\\[13pt] \displaystyle C_{\k}(\G,\G') =
469: \left[\rho_{+}{(\U_{+}\cd(\k-\G))^{2}\over \vert\k-\G\vert}+
470: \rho_{-}{(\U_{-}\cd(\k-\G))^{2}\over
471: \vert\k-\G\vert}+(\rho_{+}-\rho_{-})g\right]\d_{\G,\G'} \\[13pt]
472: \:\displaystyle
473: \hspace{6cm} -\vert\k-\G\vert^{2}\vert\k-\G'\vert^{2}D(\G'-\G).
474: \end{array}
475: \end{equation}
476:
477: \noindent Equation \ref{MATRIX2} is a matrix equation for
478: $h_{\k-\G}(\o)$ that has solutions for only certain
479: $\o(\k)$.
480: %This first result determines the stability of the
481: %coupled, periodic system. Analogous equations can be developed for
482: %related flow stability problems such as flow in a periodically
483: %compliant pipe (a periodic Orr-Sommerfeld problem). In cases where a
484: %nonuniform flow field is present, we expect qualitatively similar
485: %quadratic eigenvalue equations.
486: The full dispersion relation can be recovered by assigning the
487: appropriate translation $\G$ in the eigenvalues $\o(\k-\G)$ to
488: $\omega(\q)$ for wavevectors $\q$ outside the first Brillouin zone.
489: The system becomes unstable when $\omega(\q)$ acquires an imaginary
490: component.
491:
492: In anticipation of a length scale $a$ associated with the periodicity
493: in the plate properties, we henceforth nondimensionalise all
494: parameters and eliminate $\rho_{-}, g$ and $a$ according to
495:
496:
497: \begin{equation}
498: \begin{array}{l}
499: \displaystyle \r \rightarrow \r/a, \, \, \G \rightarrow a\G,\,\,
500: \k\rightarrow a\k,\,\, \o \rightarrow \sqrt{{a\over g}}\o, \,\, \U_{\pm} \rightarrow {\U_{\pm}\over \sqrt{ag}} \\[13pt]
501: \displaystyle D\rightarrow {D \over a^{4}\rho_{-}g},\,\, m \rightarrow {m\over
502: a\rho_{-}},\,\,\,\mbox{and}\,\,\, h_{\k}(\G) \rightarrow h_{\k}(\G)/a.
503: \end{array}
504: \end{equation}
505:
506: %\begin{eqnarray}
507: %\left[\rho_{+}{(\U_{+}\cd(\G-\q)+\o)^{2}\over \vert\G-\q\vert}
508: %+\rho_{-}{(\U_{-}\cd(\G-\q)+\o)^{2}\over \vert\G-\q\vert}
509: %+(\rho_{+}-\rho_{-})g\right]h_{\q}(\G)\hspace{2cm}\:\nonumber \\
510: %\:\hspace{2cm}=
511: %\sum_{\G'}\left[\vert\G-\q\vert^{2}\vert\G'-\q\vert^{2}
512: %D(\G-\G')-m(\G-\G')\o^{2}\right]h_{\q}(\G').
513: %\hspace{1cm}\:
514: %\label{MATRIX}
515: %\end{eqnarray}
516:
517:
518: %\noindent which is a matrix equation for $h_{\q}(\G,\o)$ that only has
519: %solutions for certain dispersion relations $\o(q)$. Upon
520: %nondimensionalizing all distances by a fixed length $a$ (which will be
521: %idenstified with the plate periodicity),
522:
523:
524:
525: \subsection{A Uniform Plate in the Presence of Flow}
526:
527: First consider the standard Kelvin-Helmholtz instability in the
528: presence of a uniform elastic plate, where $m=m_{0}$ and $D(\r) =
529: D_{0}$ are constants. \cite{MILES01} treated a similar problem of a
530: thin boundary layer over an elastic plate. Since $D(\G) =
531: D_{0}\delta_{\G,0}$ and $m(\G) = m_{0}\delta_{\G,0}$,
532: Eq. \ref{MATRIX2} is diagonal and is satisfied when
533:
534: \begin{equation}
535: \Gamma(q)\omega_{\pm}(\q) = (\g\U_{+}+\U_{-})\cd\q \pm
536: \sqrt{\Delta(\q)},
537: \label{OMEGA}
538: \end{equation}
539:
540: \noindent where $\Gamma(q)\equiv \g+1+m_{0}q$, and
541: the discriminant
542:
543: \begin{equation}
544: \Delta(\q) \equiv
545: \Gamma(q)(D_{0}q^{5}+(1-\g)q)-\g((\U_{+}-\U_{-})\cd\q)^{2}-m_{0}q
546: (\g(\U_{+}\cd\q)^{2}+(\U_{-}\cd\q)^{2}).
547: \label{DELTA}
548: \end{equation}
549:
550: \noindent The eigenvalues $\o_{\pm}$ become complex when $\Delta(\q)$
551: becomes negative, leading to exponentially growing perturbations. In
552: the limit where the influence of the plate is negligible ($m_{0},
553: D_{0}\approx 0$) the dispersion relation reduces to that of the
554: standard Kelvin-Helmholtz instability. The inclusion of membrane
555: stiffness ($D_{0}> 0$) tends to stabilise small wavelength modes. For
556: any given $m_{0}\geq 0$, the pairs of critical $D_{0}$ and $U$
557: (denoted $D_{*}$ and $U_{*}$) that give the onset of instability at a
558: chosen wavevector $q_{*}$ can be expressed explicitly in two flow
559: configurations where $\U_{\pm}$ and the $\q$ corresponding to the most
560: unstable velocity perturbations are all collinear. When $\U_{\pm}$ are
561: collinear, the onset of instability arises when $\Delta(\q) = 0$ is
562: solved by two critical repeated, real roots, $q_{*}$, in addition to
563: the root at $q=0$ and two complex conjugate roots. For a static
564: configuration of the lower fluid ($\U_{-}=0$ in the reference frame of
565: the plate), we find
566:
567: \begin{equation}
568: \begin{array}{rcl}
569: \displaystyle D_{*} &=& \displaystyle {(1+m_{0}q_{*})^2 -
570: \g^{2}(1+2m_{0}q_{*})-\g m_{0}^{2}q_{*}^{2}\over
571: q_{*}^{4}(3(1+\g+m_{0}^{2}q_{*}^{2})+2(\g+3)m_{0}q_{*})}\\[13pt]
572: \displaystyle U_{*}^{2} &=& \displaystyle
573: {4(1-\g)(1+\g+m_{0}q_{*})^{2} \over
574: \g q_{*}(3(1+\g+m_{0}^{2}q_{*}^{2})+2(\g+3)m_{0}q_{*})}.
575: \end{array}
576: \label{DUSTAR}
577: \end{equation}
578:
579:
580: %\begin{equation}
581: %D_{*}= {27\over 256}{\g^{4}(\Delta\U\cd\hat{\q})^{8}\over
582: %(1-\g^{2})^{3}(\g+1)}
583: %\end{equation}
584:
585: \begin{figure}[h!]
586: \begin{center}
587: \includegraphics[height=2.2in]{disp.eps}
588: \end{center}
589: \caption{(a) The curve relating the critical value of the shear
590: $U_{*}$ to the constant flexural rigidity $D_{0}$, for various values
591: of the uniform plate mass density $m_{0}$. For velocities
592: greater than $U_{*}$, the flow becomes unstable. For different plate
593: mass densities, the sets of $D_{*}, U_{*}$ that give rise to an
594: initial instability at wavevectors $q_{*}=\pi$ and $q_{*}=2\pi$ are
595: shown by the black arcs spanning the curves. (b) The dispersion
596: relation for a uniform plate just after the onset of instability.}
597: \label{DU}
598: \end{figure}
599:
600: \noindent Alternatively, for a chosen value of $D_{0}$, the critical
601: $U_{*}$ and the corresponding initially unstable wavevector can be
602: found implicitly. The critical velocities $U_{*}$, as a function of
603: $D_{0}$ and $m_{0}$ are shown in Fig. \ref{DU}(a). Note that the
604: critical velocity $U_{*}$ {\it decreases} as the plate mass density
605: $m_{0}$ increases. The limit $m_{0}\rightarrow \infty$ is different
606: from the $\rho_{\pm}\rightarrow 0$ limit treated in the next section.
607: Here, the plate flexibility remains constant as its mass density
608: increases. Although the magnitude of $\o_{\pm}(q)$ decreases, the
609: discriminant (Eq. \ref{DELTA}) becomes negative at smaller $U_{*}$ due
610: to the increased inertial forces of he plate back on the fluid.
611:
612: The two branches of the full dispersion relation $\omega_{\pm}(q)$ are
613: plotted in Fig. \ref{DU}(b). We have chosen parameters ($D_{*}(\pi)$
614: and $U_{*}^{2}$ defined by Eqs. \ref{DUSTAR}) such that the onset of
615: instability arises at $q_{*}\approx \pi$. For $U = 1/\sqrt{2} <
616: U_{*}(q_{*}=\pi) = 0.756320$, $\o_{\pm}(q)$ are real, as shown by the dashed
617: curves. We also plot (solid curves) the dispersion relation for $U =
618: 0.761577 > U_{*}(q_{*}=\pi) = 0.75632$, whereupon a window of instability
619: opens near $q_{*}=\pi$. The magnitude of the complex parts of
620: $\o_{\pm}(q)$ are indicated by the height of the shaded region.
621:
622: In the second flow configuration, we assume a vanishing net momentum in the
623: reference frame of the plate: $\U_{-}=-\gamma \U_{+}$. In this case, we find
624: independent of $m_{0}$,
625:
626: \begin{equation}
627: D_{*} = {1-\g \over 3q_{*}^{4}}\quad \mbox{and} \quad U_{*}^{2}
628: ={4(1-\g) \over 3\g(\g+1)q_{*}},
629: \label{DUSTAR2}
630: \end{equation}
631:
632: \noindent and the symmetry $\omega_{+}(q) = -\omega_{-}(q)$. For
633: $D_{0} \lesssim D_{*}$ and/or $U\gtrsim U_{*}$, as defined by
634: Eq. \ref{DUSTAR} or Eq. \ref{DUSTAR2}, $\omega(q\approx q_{*})$
635: becomes complex and instability arises for a small window of
636: wavevectors near $q_{*}$.
637:
638: %\begin{figure}
639: % \begin{center}
640: % \includegraphics[height=5in]{disp.eps}
641: % \end{center}
642: %\caption{(a) The dispersion relation $\o_{\pm}(q)$ (black and red
643: %curves) for $\g=1/2$, $m=0.2$. Both the stable ($U < U_{*}(\pi) =
644: %0.75632$, dashed curves) and unstable ($U > \U_{*}(\pi) = 0.75632$,
645: %solid curves) are shown. In the unstable case the magnitude of the
646: %complex part of $\o_{\pm}(q)$ is indicated by the vertical hatched
647: %lines. (b) Dispersion relation of an isolated, but periodic plate.}
648: %\label{DISP}
649: %\end{figure}
650:
651:
652: \subsection{Dispersion Relation for a Free Periodic Plate}\label{subsec:FP}
653:
654: Now consider the limit of an isolated (but periodically structured)
655: plate where the inertia of the bounding fluid is negligible. If the
656: plate were uniform, the dispersion relation in the limit
657: $\rho_{\pm}\rightarrow 0$ is simply $m_{0}\o^{2} = D_{0}q^{4}$. For a
658: periodically structured plate ($m(\G\neq 0)\neq 0, D(\G\neq 0) \neq
659: 0$), the dispersion relation can be found from Eq. \ref{MATRIX2} with
660: the simplified matrices $A(\G-\G') = m(\G-\G'), B(\G-\G')=0$, and
661: $C(\G-\G') = -\vert\k+\G\vert^{2}\vert\k+\G'\vert^{2}D(\G-\G')$.
662: Henceforth, we assume the periodicities in the plate to be sinusoidal
663: in the ${\bf x}$-direction and take the forms
664:
665: \begin{equation}
666: m(\r) = m_{0} + 2m_{1}\cos 2\pi x; \quad D(\r) = D_{0} + 2D_{1}\cos 2\pi x,
667: \end{equation}
668:
669: Although the mean mass density and bending rigidity are unchanged from
670: $m_{0}$ and $D_{0}$, respectively, the variations $m_{1}$ and/or
671: $D_{1}$ will change the dispersion relation. Specifically, periodic
672: variations will break the dispersion relation at wavevectors
673: corresponding to waves that cannot propagate through an infinite,
674: periodic medium. At each of these wavevectors (the ``Bragg planes''),
675: the waves (half-integers of which fit into a period) become
676: standing. Corresponding to these nonpropagating waves are gaps in the
677: frequency (band gaps, or stop bands) within which an infinitely
678: periodic material cannot be excited.
679:
680: \begin{figure}[h!]
681: \begin{center}
682: \includegraphics[height=2.2in]{dispbg.eps}
683: \end{center}
684: \caption{(a) The normalized dispersion relation
685: $\sqrt{m_{0}/D_{0}}\,\o_{\pm}(q)$ for $m_{1}=0$, $D_{1}/D_{0} = 0.1$
686: (dashed curve) and $D_{1}/D_{0}=0.25$ (solid curve). (b) A blow-up of
687: the dispersion relation for $D_{1}=0, m_{1}/m_{0}=0.1$ (dashed curve)
688: and $m_{1}/m_{0}=0.25$ (solid curve).}
689: \label{DISPBG}
690: \end{figure}
691:
692: Figures \ref{DISPBG}a,b show the dispersion relation of an isolated
693: plate with periodicities in the bending rigidity (a), and mass density
694: (b). The frequency gaps increase as the contrast $m_{1}$ or $D_{1}$
695: is increased. From consideration of the eigenvectors, the standing
696: modes near the band gaps can be shown to have either nodes or
697: antinodes over the high flexural rigidity (or high mass density)
698: regions, depending on which side of the band gap the wavevector $q$
699: lies. These behaviour generally arise in wave propagation through
700: periodic media. When the plate is coupled to flow, it will be the
701: near equal-wavelength modes immediately straddling the band gaps, with
702: different frequencies and $\sim 90^{\circ}$ out-of-phase, that will
703: qualitatively affect the stability near the band gaps.
704:
705: \section{Results for the Coupled Problem}\label{sec:FCP}
706:
707: We now combine the known results of the previous sections and consider
708: the fully coupled problem where fluid instabilities and Bragg
709: reflection interact. For simplicity, we consider the plate periodicity
710: wavevectors ($\G = 2\pi n/a {\bf x}$) and the uniform flows ${\bf
711: U}_{\pm}=U_{\pm}{\bf x}$ to be aligned. In the full problem, the
712: dispersion relation $\o(q)$ may be both complex (signalling regions of
713: instability) and discontinuous at band gaps where
714: $(\partial\omega/\partial q) = 0$. The full dispersion relation is
715: found from solving the quadratic eigenvalue problem given by
716: Eq. \ref{MATRIX2}. This problem can be expressed in linear eigenvalue
717: form,
718:
719: \begin{equation}
720: ({\bf M}-\omega {\bf I})\cdot\etb = 0,
721: \label{LINEAREIGEN}
722: \end{equation}
723:
724: \noindent with $\etb = (h_{\k}, \o h_{\k})^{T}$, and
725:
726: \begin{equation}
727: {\bf M} = \left(\begin{array}{cc}
728: {\bf 0} & {\bf I} \\[13pt]
729: -{\bf A}^{-1}{\bf C} & -{\bf A}^{-1}{\bf B} \end{array}\right).
730: \label{ACBMATRIX}
731: \end{equation}
732:
733: The dispersion relation $\o(\k)$ is found from the eigenvalues of the
734: nonhermitian matrix ${\bf M}$. For the parameters explored, we find
735: numerical convergence of the lowest handful of eigenvalues with 40 or
736: fewer modes $\G, \G'$. Therefore, we truncate the system at
737: approximately 60 modes (where ${\bf M}$ is a $120\times 120$
738: matrix). This is more than sufficient to obtain numerical accuracy for
739: the lower eigenvalues at all relevant wavevectors. The eigenvalues of
740: ${\bf M}$ are sorted from the numerical solutions and replotted in the
741: extended zone scheme where the wavevector $0\leq \vert\q\vert <
742: \infty$. When unfolding the eigenvalues $\omega(\k)$, we make use of
743: the fact the functions $\omega(0\leq \vert\q\vert < \infty)$ in the
744: normal extended zone scheme come in pairs $\omega_{\pm}(\q)$ and have
745: the symmetry property $\omega_{\pm}(-\vert\q\vert) =
746: -\o_{\mp}(\vert\q\vert)$. This property, along with the consideration
747: of the corresponding eigenvectors at each wavevector $k$ allows us to
748: reconstruct $\omega_{\pm}(q)$ from $\omega_{\pm}(k)$.
749:
750: \begin{figure}
751: \begin{center}
752: \includegraphics[height=4.0in]{9b.eps}
753: \end{center}
754: \caption{(a-c) The complex dispersion relation for a system just above
755: the instability threshold. The instability is first triggered at
756: $q_{*} = \pi/2$ (which arises for parameters $\gamma =1/2,
757: m_{0}=0.2, D_{0} \leq D_{*} = 0.0298349$, and $U^{2} \geq U_{*}^{2} =
758: 1.19809$). As $D_{1}$ increases, the instability bubble
759: (Im$\{\o_{\pm}\}$) increases in size as more wavevectors become
760: unstable. (d-f) Dispersion relation for instabilities that first arise
761: at $q_{*}=3.14\lesssim \pi$. As $D_{1}$ is increased, the modes with $q < \pi$ are
762: destabilised, while those with $q>\pi$ are stabilised. (g-i)
763: Increasing $D_{1}$ when the incipient instability starts at
764: $q_{*}=3.25 >\pi$ destabilises modes with $q> \pi$ and stabilises
765: those with $q< \pi$.}
766: \label{FULL}
767: \end{figure}
768:
769: In Figures \ref{FULL}(a-c) we show the effects of increasing the
770: strength of the periodic rigidity $D_{1}/D_{0}$ for the case
771: $\U_{-}=0$. In Fig. \ref{FULL}(a) we show a system with an incipient
772: instability at $q \approx q_{*} = \pi/2$. As $D_{1}$ is increased
773: (Figs. \ref{FULL}(b-c)) the bubble of instability grows modestly,
774: while gaps in the dispersion relation also appear. The destabilising
775: effect of rigidities with periodicity $a \gg 2\pi/q_{*}$ can be
776: understood in terms of a slight effective softening of the plate to
777: bending modes with wavelengths greater than $a$\footnote{The effective
778: long wavelength bending rigidity of a plate made from alternating
779: equal strips of two materials with individual rigidities
780: $D_{0}\pm\Delta D$ is $D_{eff} = 2/(1/(D_{0}+\Delta D)+1/(D_{0}-\Delta
781: D)) \approx D_{0} - (\Delta D)^2/D_{0}$}.
782:
783: Note that the gaps here do not appear at the Brillouin zone edge
784: $q=\pi$ defined by a system's static periodic structure as in the free
785: plate problem (Figs. \ref{DISPBG}) and in most other settings.
786: Rather, they appear symmetrically around the band edge due to a
787: doppler shift arising from the nonvanishing net momentum of the
788: system. Gaps occurring away from the zone boundaries have been shown
789: by \cite{BERMEL} to arise in electromagnetic propagation through
790: deformed cholesteric elastomers.
791:
792: Now consider an incipient instability that arises at higher
793: wavevectors, near the first band gap at $q=\pi$. Assume for
794: simplicity, the zero net momentum ($\U_{-}=-\gamma \U_{+}$) case where
795: ${\bf B}_{\k}(\G,\G') = 0$ in Eq. \ref{MATRIX2} and
796: $\omega_{\pm}(q)=-\o_{\mp}(q)$. Figures \ref{FULL}(d-f) show the
797: effects of increasing $D_{1}$ when the initial instability starts at
798: $q_{*}=3.14 \lesssim \pi$. As the plate heterogeneity is increased,
799: modes with wavevector $q < \pi$ are destabilised, while those with
800: $q>\pi$ are stabilised. Conversely, if the incipient instability
801: arises at $q_{*} \gtrsim \pi$, as in Figs. \ref{FULL}(g-i) (where
802: $q_{*}=3.25$), increasing $D_{1}$ will destabilise modes with $q>\pi$,
803: while stabilising those with $q<\pi$. The discontinuous change in the
804: eigenvalues across the Bragg plane $q=\pi$ occurs first in the
805: imaginary component and the growth rate $\mbox{Im}\{\o(q)\}$ is
806: discontinuous. As the contrast $D_{1}/D_{0}$ is further increased,
807: $\mbox{Re}\{\o(q)\}$ also becomes discontinuous. This behavior can be
808: explicitly seen with perturbation analysis of Eq. \ref{ACBMATRIX} by
809: truncating the matrix and finding the four eigenvalues as $q
810: \rightarrow \pi$.
811:
812:
813:
814:
815:
816:
817:
818: \section{Discussion and Conclusions}
819:
820: We have analysed the properties of the instabilities of two fluids
821: uniformly flowing past each other, separated by an elastic plate with
822: periodic bending rigidity and/or mass density. Our generalization of
823: the Kelvin-Helmholtz instability can be quantified by finding the
824: eigenvalues and eigenvectors of a nonhermitian matrix
825: (Eq. \ref{MATRIX2}). The imaginary components of the eigenvalues
826: $\o(q)$ determine the rate of unstable growth for eigenmodes of
827: wavevector $q$. We find that an imposed plate periodicity generally
828: destabilises the flow but that the destabilisation mainly occurs in
829: the band of wavevectors on the side of the initially most unstable
830: mode. Modes with wavevectors on the other side of the band gap from
831: the most unstable wavevector $q_{*}$ are stabilised as $D_{1}$ is
832: increased.
833:
834: The destabilisation/stabilisation that occurs on either side of the
835: band gap can be understood in terms of $\sim 90^{\circ}$ out-of-phase
836: standing modes with wavevectors $q=(2n-1)\pi \pm \varepsilon$
837: immediately straddling the $n^{th}$ band gap. For example, if $q_{*}
838: \gtrsim (2n-1)\pi$, increasing $D_{1}$ destabilises the waves with $q
839: \approx q_{*}$ since their antinodes occur predominantly over regions
840: of lower rigidity. Modes with $q < (2n-1)\pi$, on the other side of
841: the $n^{th}$ band gap, can be significantly stabilised since they are
842: approximately $90^{\circ}$ out-of-phase, and their antinodes sample a
843: stiffer plate. The same argument applies for $q_{*} \lesssim
844: (2n-1)\pi$.
845:
846: Our results suggest possible stability control strategies. By adding
847: restoring forces (such as bending rigidity) with periodicity near the
848: wavelength of the most unstable mode, stability can be enhanced or
849: diminished at wavelengths either less than or greater than the band
850: gap wavelength. Stabilisation of specific modes against
851: Kelvin-Helmholtz instabilities can be achieved, but at the expense of
852: destabilisation of other nearby modes.
853:
854: Extensions of our approach to stability analysis in other
855: fluid-structure systems should be possible. For example, the
856: Orr-Sommerfeld problem (\cite{ORSZAG71}) in nonuniform Poiseuille flow
857: through periodically elastic pipes can be similarly solved. A
858: physical realization of flow through elastically periodic pipes may
859: have arisen in the experiments of \cite{KRAMER} in which damping
860: elements were periodically embedded along an elastic pipe. The
861: Kelvin-Helmholtz type instability has been only briefly mentioned in
862: this context by \cite{BENJAMIN} and \cite{LANDAHL}, and only in
863: uniform pipes. Moreover, the effects of periodicity on other classes
864: of instabilities ({\it e.g.} Tollmien-Schlichting) have not been
865: explored. It would also be interesting to consider the effects of
866: dissipation, in both the fluid flow and the viscoelastic plate, on the
867: stability properties of a periodic system.
868:
869:
870: \acknowledgements
871:
872: The author was supported by the US National Science Foundation through
873: grant DMS-0349195, and the US National Institutes of Health via grant
874: K25AI058672.
875:
876:
877: \begin{thebibliography}{}
878:
879: \bibitem[Ashcroft \& Mermin (1976)]{AM} \textsc{Ashcroft, N. W. \&
880: Mermin, N. D.} 1976 {\it Solid State Physics}. Thomson Learning, Inc.
881:
882: %\bibitem[Bender \& Orszag (1978)]{BENDER} \textsc{Bender, C. M. \&
883: %Orszag, S. A.} 1978 {\it Advanced Mathematical Methods for Scientists and
884: %Engineers}. McGraw-Hill, New York.
885:
886: \bibitem[Benjamin (1960)]{BENJAMIN} \textsc{Benjamin, T. Brooke} 1960
887: {Effects of a flexible boundary on hydrodynamic stability}
888: \textit{J. Fluid Mech.} {\bf 9}, 513-533.
889:
890: \bibitem[Bermel \& Warner (2002)]{BERMEL} \textsc{Bermel, P. A. \&
891: Warner, M.} 2002 {Photonic band structure of cholesteric elastomers}
892: \textit{Phys. Rev.} E {\bf 65}, 056614.
893:
894: %\bibitem[Binding (2001)]{BINDING} \textsc{Binding, P.} 2001 {Elliptic
895: %Eigenvalue problems with eigenparameter dependent boundary
896: %conditions}. \textit{J. Differential Equations} \textbf{174} No. 1
897: %239-247
898:
899: \bibitem[Blanford \& Rees (1974)]{REES} \textsc{Blandford, R. D. \&
900: Rees M. J.} 1974 {A `twin-exhaust' model for double radio sources.}
901: \textit{MNRAS}, \textbf{169}, 395-415.
902:
903: \bibitem[Chandrasekar (1961)]{CHAND} \textsc{S. Chandrasekar} 1961
904: {Hydrodynamic and Hydromagnetic Stability} (Clarendon Press, Oxford)
905:
906: \bibitem[Chen {\it et al.} (2004)]{CHEN} \textsc{Chen, L. S., Kuo,
907: C. H., Ye, Z. \& Sun, X.} 2004 {Band gaps in the propagation and
908: scattering of surface water waves over cylindrical steps.}
909: \textit{Phys. Rev. E} \textbf{69}, 066308.
910:
911: \bibitem[Chou (1998)]{CHOU} \textsc{Chou, T.} 1998 {Band structure of
912: surface flexural-gravity waves along periodic interfaces}.
913: \textit{J. Fluid Mech.} \textbf{369}, 333-350.
914:
915: \bibitem[Drazin \& Reid (2004)]{DRAZIN} \textsc{Drazin, P. G. \&
916: Reid, W. H.} 2004 {Hydrodynamic Stability} (Cambridge University Press,
917: Cambridge, UK).
918:
919: \bibitem[Frank {\it et al.} (1996)]{FRANK} \textsc{Frank, A., Jones,
920: T. W., Ryu, D. \& Gaalaas, J. B.} 1996 {The Magnetohydrodynamic
921: Kelvin-Helmholtz Instability: A two-dimensional
922: study}. \textit{Astrophysical J.} {\bf 460}, 777-793.
923:
924: \bibitem[Hu \& Chan (2005)]{CHAN} \textsc{Hu, X. \& Chan, C. T.} 2005
925: {Refraction of Water Waves by Periodic Cylinder Arrays}.
926: \textit{Phys. Rev. Lett.} {\bf 95}, 154501.
927:
928: \bibitem[Hu {\it et al.} (2003)]{HU} \textsc{Hu, X., Shen, Y., Liu,
929: X., Fu, R. and Zi, J.} 2003 {Complete band gaps for liquid surface
930: waves propagating over a periodically drilled bottom.}
931: \textit{Phys. Rev. E} \textbf{68}, 066308.
932:
933: %\bibitem[Ibanescu {\it et al.} (2004)]{IBAN} \textsc{Ibanescu, M.,
934: %Johnson, S. G., Roundy, D., Luo, C., Fink, Y., and Joannopolous,
935: %J. D.} 2004 {Anomalous Dispersion Relations by Symmetry Breaking in
936: %Axially Uniform Waveguides}. \textit{Phys. Rev. Lett.} {\bf 92},
937: %063903.
938:
939: \bibitem[Joannopolous, Meade, \& Winn (1995)]{LIGHT}
940: \textsc{Joannopolous, J., Meade, R. D., \& Winn, J. N.} 1995 {\it Photonic
941: Crystals: Molding the Flow of Light}. Princeton University Press,
942: Princeton, NJ.
943:
944: \bibitem[Kr\"{a}mer (1960)]{KRAMER} \textsc{Kramer, M. O.} 1960
945: {Boundary layer stabilization by distributed damping}.
946: {\it J. Amer. Soc. Nav. Engrs.} {\bf 72}, 25-33.
947:
948: \bibitem[Landahl (1961)]{LANDAHL} \textsc{Landahl, M. T.} 1961
949: {On the stability of a laminar incompressible boundary layer over a flexible surface}
950: {\it J. Fluid Mech.} {\bf 13}, 609-632.
951:
952: \bibitem[Linton \& Evans (1990)]{LINTON} \textsc{Linton, C. M. \&
953: Evans, D. V.} 1990 {The interaction of waves with arrays of vertical
954: circular cylinders}. {\it J. Fluid Mech.} {\bf 215}, 549-569.
955:
956: \bibitem[Miles (1956)]{MILES} \textsc{Miles, J. W.} 1956
957: {On the aerodynamic stability of thin panels}
958: {\it J. Aero. Sci.} {\bf 23}, 77.
959:
960: \bibitem[Miles (2001)]{MILES01} \textsc{Miles, J.} 2001
961: {Stability of inviscid shear flow over a flexible boundary}
962: {\it J. Fluid Mech.} {\bf 434}, 371-378.
963:
964: \bibitem[Naciri \& Mei (1988)]{MEI1} \textsc{Naciri, M. \& Mei, C. C.}
965: 1988 {Bragg scattering of water waves by a doubly periodic seabed.}
966: \textit{J. Fluid Mech.} \textbf{192}, 51-74.
967:
968: \bibitem[Orszag (1971)]{ORSZAG71} \textsc{Orszag, S.A.} 1971 {Accurate
969: Solution of the Orr Sommerfeld stability equation}. \textit{J. Fluid
970: Mech.} \textbf{50}, 689-703.
971:
972: \bibitem[Peter, Meylan, \& Linton (2006)]{PETER} \textsc{Peter, M. A.,
973: Meylan, M. H. \& Linton, C. M.} 2006 {Water-wave scattering by a
974: periodic array of arbitrary bodies}. \textit{J. Fluid Mech.} {\bf 548},
975: 237-256.
976:
977: \bibitem[Porter \& Evans (1996)]{PORTER} \textsc{Porter, R. \& Evans,
978: D. V.} 1996 {Wave scattering by periodic arrays of
979: breakwaters}. \textit{Wave Motion} \textbf{23}, 95-120.
980:
981: \bibitem[Reiss \& Corona (1977)]{REISS} \textsc{Reiss, N. M. \& Corona,
982: T. J.} 1977 {An Investigation of a Kelvin-Helmholtz Billow Cloud}.
983: \textit{Bulletin of the American Meteorological Society} {\bf 58}, 159-162.
984:
985:
986: %\bibitem[Sen \& Arora (1988)]{SEN} \textsc{Sen, P. K. \& Arora, D. S.}
987: %1988 {On the stability of laminar boundary-layer flow over a flat
988: %plate with a compliant surface} \textit{J. Fluid Mech.} \textbf{197}
989: %201-240.
990:
991: \bibitem[Sigalas \& Economou (1992)]{SIGALAS} \textsc{Sigalas, M. M. \&
992: Economou, E. N.} 1992 {Elastic and acoustic wave band structure}.
993: \textit{J. Sound and Vibration} {\bf 158}, 377-382.
994:
995:
996: \bibitem[Whitham (1999)]{WHITHAM} \textsc{Whitham, G. B.} 1999 {Linear
997: and Nonlinear Waves}. (Wiley Inter-Science).
998:
999:
1000: \end{thebibliography}
1001:
1002: \end{document}
1003:
1004:
1005:
1006: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1007: %%%%%%%%%%%%%%% END %%%%%%%%%%%%%%%%%%
1008: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1009:
1010:
1011: %in the small and large plate density limits.
1012: %When $m_{0}q \ll 1$,
1013:
1014: %\begin{equation}
1015: %\Delta(q) \approx D_{0}q^{5}+(1-\g)q - {\g \over \g +1}(\U\cd\q)^{2} -
1016: %m_{0} q{\g^{2}\over (\g+1)^{2}} (\U\cd\q)^{2} +
1017: %O(m^{2}q^{2}).
1018: %\end{equation}
1019:
1020: %\noindent In this limit, the relations
1021: %between $q_{*}, U_{*}$ and $D_{0}$ are
1022:
1023: %\begin{equation}
1024: %q_{*} = q_{0} + m_{0} {\g \over 3(\g+1)} q_{0} + O(m_{0}^{2}),
1025: %\quad q_{0} = \sqrt{1-\g \over 3D_{0}},
1026: %\label{Q0}
1027: %\end{equation}
1028:
1029: %\noindent
1030:
1031: %\begin{equation}
1032: %U_{*}^{2} \approx \left[{256(1-\g^{2})^{3}(1+\g) D_{0} \over
1033: %27\g^{4}}\right]^{1/4} -{4\over 3}(1-\g)m_{0} + O(m_{0}^{2}),
1034: %\label{U0}
1035: %\end{equation}
1036:
1037: %\noindent respectively.
1038:
1039: %In the case of extremely massive plates ($m_{0} \rightarrow \infty$),
1040: %we find
1041:
1042: %\begin{equation}
1043: %q_{*} \approx \left({1-\g \over 3D_{0}}\right)^{1/4} + {\g \over 3 m_{0}}
1044: %+ O(m_{0}^{-2}),
1045: %\label{QINF}
1046: %\end{equation}
1047:
1048:
1049: %\noindent and
1050:
1051: %\begin{equation}
1052: %U_{*}^{2} \approx {4D_{0} \over \g}\left({1-\g\over
1053: %3D_{0}}\right)^{3/4} + {4D_{0}\over m_{0}}\sqrt{{1-\g\over 3D_{0}}} +
1054: %O(m_{0}^{-2}).
1055: %\label{UINF}
1056: %\end{equation}
1057: