1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: %%% VHolm-Staley 17Jan2003
3: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4: \documentclass[11pt]{article}
5: \usepackage{amssymb,latexsym, amsmath}
6: %\documentstyle[aps,prl,multicol]{revtex} % epsf psfig
7: %\documentstyle[twocolumn,psfig,aps]{revtex} % psfig, epsf
8: \usepackage{graphicx}
9: \usepackage{epsfig}
10: \tolerance=10000
11:
12: %\newcommand{\remfigure}[1]{}
13: \newcommand{\remfigure}[1]{#1}
14:
15: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
16: %%% Comments and Todos
17: %
18: \newcommand{\comment}[1]{\vspace{-3 mm}\par
19: \marginpar{\large\underline{}}\noindent
20: \framebox{\begin{minipage}[c]{0.65 \textwidth}
21: \rm #1 \end{minipage}}\vspace{1 mm}\par}
22: %
23: \newcommand{\rem}[1]{}
24: %
25: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
26: % removing non-vanilla commands
27: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
28: \def\mathbf#1{{\bf #1}}
29: \def\mathsf#1{{\sf #1}}
30: \def\mathbb#1{\mathbf{ #1}}
31: \def\mathfrak#1{\mathbf{ #1}}
32: \def\tfrac#1{\frac{ #1}}
33: %\def\mathcal{#1}{\mathbf{ #1}}
34: %
35: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
36:
37: \pagestyle{myheadings}
38: \markright{ D. D. Holm \& M. F. Staley
39: \hfil \underline{Solitons, peakons, ramps/cliffs and leftons}}
40: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
41: %
42: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
43:
44: \begin{document}
45:
46: \title{
47: \vbox to 0pt {\vskip -1cm \rlap{\hbox to \textwidth {\rm{\small
48: BI3692
49: \hfil
50: ACCEPTED FOR {\it PHYS. LETT. A}
51: \hfil
52: %Today's version $\quad$
53: \today}}}}
54: %
55: Nonlinear balance and exchange of stability in dynamics of
56: solitons, peakons, ramps/cliffs and leftons in a
57: 1+1 nonlinear evolutionary pde
58: %\vspace{-2mm}
59: }
60: %
61: \author{
62: Darryl D. Holm$^{1,2}$ and Martin F. Staley$^1$
63: \\$^1$Theoretical Division and Center for Nonlinear Studies
64: \\Los Alamos National
65: Laboratory, MS B284
66: \\ Los Alamos, NM 87545
67: \\ email: dholm@lanl.gov, mstaley@lanl.gov
68: \and
69: $^2$%
70: Mathematics Department\\
71: Imperial College of Science, Technology and Medicine
72: \\ London SW7 2AZ, UK
73: }
74:
75: \date{
76: \comment{
77: \centerline{BI3692}
78: \centerline{ACCEPTED FOR {\it Phys. Lett. A}}
79: \centerline{January 16, 2003}
80: }
81: }
82:
83: \maketitle
84:
85: %\large
86:
87: %\vspace{-5mm}
88:
89: \begin{abstract}
90: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
91: We study exchange of stability in the dynamics
92: of solitary wave solutions
93: under changes in the nonlinear balance in
94: a 1+1 evolutionary partial differential equation
95: related both to shallow water waves and to turbulence.
96: We find that solutions of the equation
97: $
98: m_t
99: +
100: um_x
101: +b\,u_xm
102: =
103: \nu\,m_{xx}
104: $
105: with $m = u - \alpha^2 u_{xx}$ for fluid velocity $u(x,t)$ change their
106: behavior at the special values $b=0,\pm1,\pm2,\pm3$.
107: \\
108: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
109:
110:
111: \noindent
112: PACS numbers: 11.10.Lm, 03.50.-z, 05.45.Yv, 05.45.-a
113:
114: \noindent
115: Keywords: solitary waves, solitons, peakons, nonlinear evolution,
116: turbulence modeling
117:
118:
119: \end{abstract}
120:
121: \newpage
122: \tableofcontents
123:
124: \vskip 4mm
125:
126: %\begin{multicols}{2}
127:
128: \section{Introduction}
129: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
130: \subsection{Linear and nonlinear balances in shallow water waves}
131: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
132: The primary physical mechanism for the propagation of solitary shallow water
133: waves is the balance between nonlinear steepening and linear dispersion.
134: This balance appears {\it uniquely} at linear order in an asymptotic
135: expansion in the Korteweg-de Vries equation (KdV),
136: \begin{equation}\label{kdv}
137: u_t + c_0u_x + \frac{3\epsilon_1}{2}uu_x +
138: \frac{3\epsilon_2}{20}\,u_{xxx}=0
139: \,.
140: \end{equation}
141: Here the expansion parameters satisfy $\epsilon_1 \ge \epsilon_2 >
142: \epsilon_1^2$ and are defined by $\epsilon_1 = a/h$ and $\epsilon_2 =
143: h^2/l^2$, in terms of $a$, $h$ and $l$, which denote wave amplitude, mean
144: water depth, and typical horizontal length scale (e.g., a wavelength),
145: respectively. KdV possesses the famous
146: ${\rm sech}^2$ solitary wave solution $u(x,t)=u_0\,{\rm
147: sech}^2((x-ct)\sqrt{u_0/\gamma}/2\,)$, for
148: $u_0=2(c-c_0)/\epsilon_1$ and $\gamma=3\epsilon_2/(5 \epsilon_1)$, see
149: \cite{AblowitzClarkson[1991]}.
150:
151: The Benjamin-Bona-Mahoney equation (BBM),
152: \begin{equation}\label{bbm}
153: u_t + c_0u_x + \frac{3\epsilon_1}{2}uu_x
154: - \frac{3\epsilon_2}{20} c_0^{-1}\,u_{xxt}
155: =
156: 0
157: \,,
158: \end{equation}
159: has the same sech$^2$ shape for its solitary wave, but with $\gamma$
160: replaced by $\gamma\,'=c\gamma/c_0$. BBM is asymptotically equivalent to KdV
161: at order $O(\epsilon_1,\epsilon_2)$. However, the linear dispersion
162: relation for BBM is a better match than that for KdV in the physical
163: application to shallow water waves.
164:
165: Beyond KdV at linear order, the asymptotic expansion at quadratic order
166: in the small parameters $\epsilon_1$ and $\epsilon_2$ does not
167: produce a unique wave equation for unidirectional shallow water
168: waves \cite{DGH[2003]}. Instead, the asymptotic expansion
169: produces an entire {\it family} of shallow water wave equations that are
170: asymptotically equivalent to each other at quadratic order in the shallow
171: water expansion parameters. The equations in this family are related
172: amongst themselves by a continuous, three-parameter group
173: of nonlinear, nonlocal transformations of variables given by,
174: \begin{equation}
175: \label{KodamaKdV}
176: u =
177: v + \epsilon_1(a_1v^2+a_2v_x\partial^{-1}_x{v})
178: + \epsilon_2a_3 v_{xx}
179: \,,
180: \end{equation}
181: in which $(a_1,a_2,a_3)$ are real parameters. This transformation group
182: was first introduced for determining normal forms of shallow water
183: equations by Kodama in
184: \cite{Kodama[1985&1987]}.
185:
186: Among the family of asymptotically equivalent shallow water wave
187: equations at quadratic order accuracy in the small parameters $\epsilon_1 =
188: a/h$ and $\epsilon_2 = h^2/l^2$ are several equations that are completely
189: integrable. As for KdV at linear order, these integrable shallow water
190: equations at quadratic order possess soliton solutions that interact via
191: elastic collisions. In particular, the equation in the KdV hierarchy with
192: fifth-order derivatives is present among these integrable equations, as
193: shown in \cite{LiZhiSibgatullin[1997]}.
194:
195: The family of shallow water equations at quadratic order accuracy that are
196: asymptotically equivalent under Kodama transformations (\ref{KodamaKdV})
197: contains the following sub-family, derived in \cite{DGH[2003]},
198: \begin{equation}
199: \label{b-eqn}
200: \underbrace{\
201: m_t \
202: }_{\hbox{Evolution}}
203: \!\!\!+
204: \underbrace{\
205: c_0u_x
206: + \frac{3\epsilon_2}{20}u_{xxx} \
207: }_{\hbox{Dispersion}}
208: +
209: \underbrace{\
210: \epsilon_1(um_x + b\,mu_x) \
211: }_{\hbox{Nonlinearity}}
212: = 0\,,
213: \end{equation}
214: where $m = u - (19\epsilon_2/60) u_{xx}$. The set of 1+1 evolutionary
215: equations (\ref{b-eqn}) is nonlocal, dispersive and nonlinear. The
216: nonlinearity in this set of asymptotically equivalent shallow water
217: equations is parameterized by the real constant $b$, which depends on the
218: group parameters $(a_1,a_2,a_3)$ in the Kodama transformation
219: (\ref{KodamaKdV}). An asymptotically equivalent shallow water equation for
220: any $b \not = -1$ may be achieved by a Kodama transformation. However, the
221: case $b=-1$ violates the asymptotic ordering and the corresponding Kodama
222: transformation is singular in this case \cite{DGH[2003]}.
223:
224: The cases $b=2$ and $b=3$ are special values for the $b-$equation
225: (\ref{b-eqn}). The case $b=2$ restricts (\ref{b-eqn}) to
226: the integrable Camassa-Holm equation (CH) \cite{CH[1993]}.
227: The case $b=3$ in (\ref{b-eqn}) is the Degasperis-Procesi equation (DP)
228: \cite{DP[1999]}, which was shown to be integrable in \cite{DHH2002}.
229: These two cases exhaust the integrable candidates for (\ref{b-eqn}), as
230: was shown using Painlev\'e analysis in \cite{DHH2002}. The b-family of
231: equations (\ref{b-eqn}) was also shown in \cite{MN[2002]} to admit the
232: symmetry conditions necessary for integrability only in the cases $b=2$ for
233: CH and
234: $b=3$ for DP.
235:
236: KdV (\ref{kdv}) and the cases $b=2$ (CH) and $b=3$ (DP) of equation
237: (\ref{b-eqn}) are three completely integrable Hamiltonian equations
238: that possess solitons as traveling waves. In KdV, the balance that
239: confines the traveling wave soliton occurs between nonlinear
240: steepening and linear dispersion. This is the leading order asymptotic
241: balance for shallow water waves. However, even in the absence of linear
242: dispersion, the parameter $b$ in equation (\ref{b-eqn}) introduces
243: additional possibilities for balance, including the nonlinear/nonlocal
244: balance in the following (rescaled) dispersionless case of CH that was
245: studied previously for
246: $b=2$ in \cite{CH[1993]},
247: \begin{equation}
248: \label{b-family-inviscid}
249: m_t + um_x + b\,mu_x = 0\,,
250: \hbox{ with }
251: m = u - \alpha^2 u_{xx}
252: \,.
253: \end{equation}
254: The nonlinear/nonlocal balance in this equation, even in the absence of
255: linear dispersion, can still produce a confined solitary traveling wave
256: pulse $u(x,t)=ce^{-|x-ct|/\alpha}$, called the {\it peakon}
257: \cite{CH[1993]}. The peakon moves with speed equal to its amplitude and has
258: a jump in derivative at its peak. Peakons for either $b=2$ or $b=3$ are
259: true solitons that interact via elastic collisions under CH dynamics, or DP
260: dynamics, respectively. In addition, the CH and DP initial value problems
261: are both completely integrable as Hamiltonian systems by using the inverse
262: spectral transform (IST) method for an isospectral linear eigenvalue
263: problem whose purely discrete spectrum gives the asymptotic speeds of the
264: peakons
265: \cite{CH[1993]}, \cite{DHH2002}. Figure \ref{Gaussian-to-Peakons-CH} shows
266: the evolution under dispersionless CH for the case $b=2$ in
267: equation (\ref{b-family-inviscid} ) of a Gaussian initial
268: velocity distribution of unit area and width $5\alpha$. Peakon
269: solutions exist for equation (\ref{b-family-inviscid}) with any value of
270: $b$. However, we shall find that the stability of these peakon solutions
271: requires
272: $b>1$.
273:
274: The properties of dispersionless CH for the case $b=2$ in
275: equation (\ref{b-family-inviscid}) and also the related class of
276: dispersionless equations consisting of (\ref{b-family-inviscid}) with
277: $u=g*m=\int_{-\infty}^\infty g(x-y)m(y)dy$ for the case $b=2$ and an even
278: kernel $g(x)=g(-x)$ including $g(x)=e^{-|x|/\alpha}$ are studied further in
279: Fringer and Holm \cite{FH[2001]}. The properties of dispersionless DP for
280: the case $b=3$ in equation (\ref{b-family-inviscid}) are studied further in
281: \cite{DHH2002}, \cite{HS2003}.
282:
283: %\vspace{6.5cm}
284:
285: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
286: \rem{
287: \begin{figure}
288: \begin{center}
289: \leavevmode{\hbox{\epsfig{
290: figure=fig1a.ps, scale=0.7
291: }}
292: }
293: \caption{\label{Gaussian-to-Peakons-CH}
294: Evolution under equation (\ref{b-family-inviscid}) of a Gaussian initial
295: velocity
296: distribution of unit area and width $5\alpha$.
297: }
298: \end{center}
299: \end{figure}
300: }
301:
302: %\vskip-6.5cm
303: \remfigure{
304: \begin{figure}[htbp]
305: \centering
306: \psfig{file=fig1a.ps,height=6.5cm}
307: \caption[]{
308: Evolution under equation (\ref{b-family-inviscid}) with $b=2$ of a
309: Gaussian initial velocity distribution of unit area and width $5\alpha$.
310: }
311: \label{Gaussian-to-Peakons-CH}
312: \end{figure}
313: }
314: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
315:
316: The dispersionless limit of KdV in (\ref{kdv}), upon rescaling velocity
317: $u$, yields the Burgers equation
318: \begin{equation}
319: u_t + uu_x - \nu\,u_{xx}=0
320: \,,
321: \end{equation}
322: in which we now also add constant viscosity $\nu$. The characteristic
323: Burgers solution is the classic ramp and cliff shown arising from the
324: Gaussian initial condition in Figure
325: \ref{Gaussian-to-ramp/cliff-Burgers}. In the ramp/cliff solution, nonlinear
326: steepening is balanced by linear viscosity to produce the ``cliff'' whose
327: width is controlled by the magnitude of viscosity $\nu$. The ``ramp'' is
328: the self-similar $u\approx{x}/t$ part of the solution for which the viscous
329: term vanishes.
330:
331: %\vspace{6.5cm}
332:
333: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
334: \rem{
335: \begin{figure}
336: \begin{center}
337: \leavevmode{\hbox{\epsfig{
338: figure=fig2a.ps, scale=0.7
339: }}
340: }
341: \caption{\label{Gaussian-to-ramp/cliff-Burgers}
342: The Burgers solution is the classic ramp/cliff shown arising from the
343: Gaussian initial condition.
344: }
345: \end{center}
346: \end{figure}
347: }
348:
349: %\vskip-6.5cm
350: \remfigure{
351: \begin{figure}[htbp]
352: \centering
353: \psfig{file=fig2a.ps,height=6.5cm}
354: \caption[]{
355: The Burgers solution is the classic ramp/cliff shown arising from
356: the Gaussian initial condition.
357: }
358: \label{Gaussian-to-ramp/cliff-Burgers}
359: \end{figure}
360: }
361: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
362:
363: The equations KdV, BBM, CH, DP, the other $b-$equations and Burgers all
364: preserve the area $M=\int_{-\infty}^\infty{u\,dx}$ (total momentum) for a
365: solution that vanishes at spatial infinity.
366:
367: \subsection{Problem statement for the present work}
368: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
369: In this paper, we shall treat the constant $b$ in the dispersionless
370: case of equation (\ref{b-eqn}) as a {\it bifurcation parameter}. We
371: shall study numerically the exchange of stability among three types of
372: solitary wave solutions, as $b$ is varied. This is done by using numerical
373: simulations to investigate its initial value problem on a periodic domain
374: of length $L\gg\alpha$ for various values of the parameter $b$. When
375: necessary, we shall also add viscosity in the form%
376: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
377: \footnote{{\bf Relation of the $b-$equation (\ref{b-eqn}) to 3D turbulence
378: closure models.} Higher-dimensional representatives of the viscous
379: $b-$family of equations (\ref{b-eqn}) have appeared recently in 3D
380: turbulence modeling. The $b=0$ case of equation (\ref{visc-b-eqn}) is the
381: 1D version of Leray's regularization of the Navier-Stokes equations in 3D
382: \cite{Leray[1934]}. Leray regularization has recently been revived as an
383: approach to deriving turbulence closure models for large eddy simulations
384: of turbulent mixing layers in 3D \cite{GH[2002]}. For 3D studies of the
385: turbulence model corresponding to the $b-$equation (\ref{b-eqn}) with
386: $b=2$, see \cite{Chen-etal[1998]},
387: \cite{Chen-etal[1999c]},
388: \cite{FHT[2001]}.}
389: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
390: \begin{equation}\label{visc-b-eqn}
391: m_t\
392: +\
393: \underbrace{\ \ um_x\ \
394: }
395: _{\hspace{-2mm}\hbox{convection}\hspace{-2mm}}\
396: +\
397: \underbrace{\ \ b\,u_xm\ \
398: }
399: _{\hspace{-2mm}\hbox{stretching}\hspace{-2mm}}\
400: =\
401: \underbrace{\ \
402: \nu\,m_{xx}\
403: }
404: _{\hspace{-2mm}\hbox{viscosity}
405: }
406: \,,
407: \end{equation}
408: with $m = u - \alpha^2 u_{xx}$. We shall treat the following cases, which
409: we shall show analytically are special values:
410: $b=0,\,\pm1,\,\pm2,\,\pm3$.
411:
412:
413: \rem{
414: Recently, the following
415: equation of this type was singled out for further study based on formal
416: asymptotics in
417: \cite{DP[1999]},
418: \begin{equation}
419: \label{DP-nonlin}
420: m_t + um_x + 3mu_x = 0\,,
421: \hbox{ with }
422: m = u - \alpha^2 u_{xx}
423: \,.
424: \end{equation}
425: This equation (DP) was shown to be completely integrable by IST and also to
426: support peakons as soliton solutions in \cite{DHH[2002]}. Subsequently, the
427: b-family of equations introduced in \cite{DHH[2002]}, \cite{footnote}
428: \begin{equation}
429: \label{b-family}
430: m_t + um_x + bmu_x = 0\,,
431: \hbox{ with }
432: m = u - \alpha^2 u_{xx}
433: \,,
434: \end{equation}
435: with constant parameter $b$ was shown \cite{MN[2002]} to admit the
436: symmetry conditions necessary for integrability only in the cases $b=2$
437: for CH in (\ref{b-family-inviscid}) and $b=3$ for DP in (\ref{DP-nonlin}).
438: }
439:
440:
441: \section{Exchange of stability}
442: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
443:
444: %\vspace{6.5cm}
445:
446: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
447: \rem{
448: \begin{figure}
449: \begin{center}
450: \leavevmode{\hbox{\epsfig{
451: figure=fig3a.ps, scale=0.7
452: }}
453: }
454: \caption{\label{peakon-to-ramp/cliff-to-peakons}
455: Exchange of stability between the classic ramp/cliff for $b=0$ and the
456: peakon for $b=2$.
457: }
458: \end{center}
459: \end{figure}
460: }
461:
462: %\vskip-6.5cm
463: \remfigure{
464: \begin{figure}[htbp]
465: \centering
466: \psfig{file=fig3a.ps,height=6.5cm}
467: \caption[]{
468: Exchange of stability between the classic ramp/cliff for $b=0$ and the
469: peakons for $b=2$.
470: }
471: \label{peakon-to-ramp/cliff-to-peakons}
472: \end{figure}
473: }
474: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
475:
476:
477: \subsection{Numerics for exchange of stability from $b=0$ to
478: $b=\pm2$}
479: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
480: We begin by considering the initial value problem for case $b=0$ in
481: equation (\ref{visc-b-eqn}). The first part of Figure
482: \ref{peakon-to-ramp/cliff-to-peakons} shows for
483: $b=0$ that a peakon of width $w=5\alpha$ is unstable to
484: forming a ramp/cliff. Namely, the initial peakon develops into a
485: Burgers-type ramp/cliff solution for $b=0$, just as it would have done for
486: the Burgers equation. Note, however, that the case $b=0$ in equation
487: (\ref{visc-b-eqn}) is {\it not} the Burgers equation. At time $T=2500$, we
488: stopped this part of the numerical calculation and used the result as the
489: initial condition for continuing the $b-$equation evolution with $b$ changed
490: to $b=2$, a case that supports peakons. The remainder of Figure
491: \ref{peakon-to-ramp/cliff-to-peakons} shows the exchange of stability after
492: $b=0$ is changed to $b=2$. Thus, the Burgers-type ramp/cliff solution
493: is stable for $b=0$ in equation (\ref{visc-b-eqn}). However, when treated
494: as an initial condition, the ramp/cliff solution is unstable and breaks
495: into a train of peakons for $b=2$. Figure \ref{final-states/b=0/b=2} shows
496: the spatial profiles of these solutions at the initial time, at the end of
497: the
498: $b=0$ evolution, and at the end of the $b=2$ evolution. A similar exchange
499: of stability occurs between the ramp/cliff for $b=0$ and the train of
500: peakons for $b=3$.
501:
502: %\vspace{6.5cm}
503:
504: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
505: \rem{
506: \begin{figure}
507: \begin{center}
508: \leavevmode{\hbox{\epsfig{
509: figure=fig4a.ps, scale=0.7
510: }}
511: }
512: \caption{\label{final-states/b=0/b=2}
513: Spatial profiles of the solutions at the initial time and at the end times
514: of the $b=0$ evolution and the $b=2$ evolution.
515: }
516: \end{center}
517: \end{figure}
518: }
519:
520: %\vskip-6.5cm
521: \remfigure{
522: \begin{figure}[htbp]
523: \centering
524: \psfig{file=fig4a.ps,height=6.5cm}
525: \caption[]{
526: Spatial profiles of the solutions at the initial time and at the
527: end times of the $b=0$ evolution and the $b=2$ evolution.
528: }
529: \label{final-states/b=0/b=2}
530: \end{figure}
531: }
532: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
533:
534: If we switch from $b=0$ to $b=-2$ instead of $b=2$, then Figure
535: \ref{peakon-to-ramp/cliff-to-leftons} shows that the ramp/cliff at
536: time $T=2500$ evolves into a train of {\it leftward moving} solitary waves
537: (leftons). Thus, a different exchange of stability occurs between
538: ramp/cliff solutions for $b=0$ and leftons for $b=-2$ (and for
539: $b=-3$ as well). Figure \ref{final-states/b=0/b=-2} plots the comparison
540: between
541: the stationary solution $u=u_0\,{\rm sech}^2\,(x/(2\alpha))$ for $b=-2$
542: and the final states for an initial Gaussian of width $w=10$ evolved to
543: leftons.
544: Similar agreement occurs
545: for the case $b=-3$, whose stationary solution is $u=u_0\,{\rm
546: sech}\,(x/\alpha)$. Thus, confined initial states move leftward and evolve
547: into stationary solutions for $b=-2$ and $b=-3$.
548:
549: %\vspace{6.5cm}
550:
551: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
552: \rem{
553: \begin{figure}
554: \begin{center}
555: \leavevmode{\hbox{\epsfig{
556: figure=fig5a.ps, scale=0.7
557: }}
558: }
559: \caption{\label{peakon-to-ramp/cliff-to-leftons}
560: Exchange of stability between the classic ramp/cliff for $b=0$ and the
561: leftons for $b=-2$.
562: }
563: \end{center}
564: \end{figure}
565: }
566:
567: %\vskip-6.5cm
568: \remfigure{
569: \begin{figure}[htbp]
570: \centering
571: \psfig{file=fig5a.ps,height=6.5cm}
572: \caption[]{
573: Exchange of stability between the classic ramp/cliff for $b=0$ and the
574: leftons for $b=-2$.
575: }
576: \label{peakon-to-ramp/cliff-to-leftons}
577: \end{figure}
578: }
579: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
580:
581: %\vspace{6.5cm}
582:
583: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
584: \rem{
585: \begin{figure}
586: \begin{center}
587: \leavevmode{\hbox{\epsfig{
588: figure=fig6a.ps, scale=0.7
589: }}
590: }
591: \caption{\label{final-states/b=0/b=-2}
592: Comparison between leftons and the stationary solution
593: $u=u_0\,{\rm sech}^2\,(x/(2\alpha))$ for $b=-2$.
594: }
595: \end{center}
596: \end{figure}
597: }
598:
599: %\vskip-6.5cm
600: \remfigure{
601: \begin{figure}[htbp]
602: \centering
603: \psfig{file=fig6a.ps,height=6.5cm}
604: \caption[]{
605: Comparison between leftons and the stationary solution
606: $u=u_0\,{\rm sech}^2\,(x/(2\alpha))$ for $b=-2$.
607: }
608: \label{final-states/b=0/b=-2}
609: \end{figure}
610: }
611: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
612:
613: \subsection{Analytical discussion of the exchange of stability}
614: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
615: The exchange of stability as the parameter
616: $b$ varies can be understood analytically as arising from the following
617: properties of the $b-$equation (\ref{visc-b-eqn}).
618:
619: \noindent$\bullet$
620: For $0\le{b}\le1$, the $L^{1/b}$ norm of the momentum
621: $\|m\|_{L^{1/b}}=\Big(\int_{-\infty}^\infty|m|^{1/b}\Big)^b$
622: is preserved by (\ref{visc-b-eqn}), in the inviscid case for which
623: $\nu=0$. Thus, the ramp/cliff solution for $b=0$ forms in a regime of
624: classical analytical control in which $|m|$ possesses a maximum principle.
625: Numerically, we observe: (1) the cliff's width is controlled by the
626: magnitude of $\alpha$; and (2) the viscosity $\nu$ can be made negligible
627: without measurably affecting the computed solution.
628:
629: \noindent$\bullet$
630: For $1<{b}\le3$ in the inviscid case, an inflection point with negative
631: slope may be shown to develop into a vertical slope in finite time.
632: However, numerically, we observe the nonlinear evolution conspires to
633: eliminate such singular inflection points, e.g., by forming peakons, which
634: are weak solutions.
635:
636: \noindent$\bullet$
637: For $b=-1$ the lowest order nonlinearity in equation (\ref{visc-b-eqn}) (the
638: steepening term $uu_x$) has coefficient zero. The evolution then develops
639: from the higher order nonlinear terms. This can be seen by rewriting
640: the $b-$equation (\ref{visc-b-eqn}) equivalently as
641: \begin{equation}\label{visc-b-eqn-uform}
642: u_t + (b+1)uu_x - \nu\,u_{xx}
643: =
644: \alpha^2\big(u_t + uu_x - \nu\,u_{xx}+ \frac{b-3}{2}u_x^2\big)_{xx}
645: \,.
646: \end{equation}
647: Thus, solutions with small curvature are nearly stationary for $b=-1$. The
648: leading order nonlinearity is absent, so the dynamics is governed by the
649: balance between viscous diffusion and higher-order terms. The case
650: $\alpha\to\infty$ in (\ref{visc-b-eqn-uform}) is interesting, as well.
651: However, this case will be discussed elsewhere.
652:
653: \noindent$\bullet$
654: Equation (\ref{visc-b-eqn}) has stationary $(c=0)$
655: solutions $u=u_0\,{\rm sech}^2(x/(2\alpha))$ for $b=-2$ and
656: $u=u_0\,{\rm sech}(x/\alpha)$ for $b=-3$. Figure
657: \ref{final-states/b=0/b=-2} shows that the evolution of equation
658: (\ref{visc-b-eqn}) rapidly approaches the stationary solution for $b=-2$.
659: (A similar result holds for $b=-3$.) The linearized evolution of
660: (\ref{visc-b-eqn}) around these stationary solutions is stable for all
661: wavelengths. Therefore, for any nonzero viscosity, these stationary
662: solutions have a nontrivial basin of attraction. Note: equation
663: (\ref{visc-b-eqn}) with either $b=-2$, or $b=-3$ is not Hamiltonian. The
664: mathematical properties of the $b-$equation in these cases is discussed
665: in \cite{HS2003}.
666:
667: \subsection{Confirmations of our numerical methods}
668: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
669: In our numerical runs we advanced the $b-$equation (\ref{visc-b-eqn})
670: with an explicit, variable timestep fourth/fifth order Runge-Kutta-Fehlberg
671: (RKF45) predictor/corrector. We selected the timestep for numerical stability
672: by trial and error, while our code selected the timestep for numerical accuracy
673: (not to exceed the timestep for numerical stability). We used a very
674: strict relative error tolerance per timestep,
675: $\epsilon=10^{-8}$.
676:
677: We computed spatial derivatives using 4th order finite differences, generally
678: at resolutions of $2^{13}=8192$ or $2^{14}=16384$ zones.
679: To invert the Helmholtz operator in determining $u(x,t)$ from
680: $m(x,t)$, we used the Fourier transform. When the numerical approximation
681: of the nonlinear terms had aliasing errors in the high wavenumbers, we
682: removed these errors by applying a high pass filtered artificial viscosity.
683:
684: The quality of the numerical convergence may be checked analytically in the
685: collision of two peakons of asymptotic speeds $c_1$ and $c_2$, for which the
686: minimum peakon separation with $b>1$ is
687: %
688: \begin{equation}\label{peakon-qmin}
689: e^{-|q_{min}|/\alpha} = 1-
690: \Big(\,\frac{4c_1c_2}{(c_1+c_2)^2}\Big)^{{1}/(b-1)}.
691: \end{equation}
692: %
693: For $b=2$, $c_1=1$, $c_2=1/2$ and $\alpha=5$, formula (\ref{peakon-qmin})
694: implies $q_{min}=10\,{\rm ln}\,3=10.9861$. Our numerical results with the
695: resolution of $2^{14}$ zones yield $q_{min}=11.0049$. The small
696: discrepancy, less than $0.2\%$, occurs largely because our numerical
697: measurement of $q_{min}$ is obtained by examining the peakon positions at
698: each internal timestep in the code without interpolation, although one is
699: unlikely to land exactly on the time at which the minimum separation
700: occurs. The code's true accuracy is better than the above measure
701: indicates, because the intermediate steps involved in advancing the
702: solution from one discrete time to the next with an RKF45 method cancel the
703: higher-order discretization errors.
704:
705: Likewise, for peakons with $b=3$, formula
706: (\ref{peakon-qmin}) gives the minimum separation when $c_1=1$, $c_2=1/2$,
707: and $\alpha=5$, as
708: $q_{min}=5\,{\rm ln}\,(3/(3-\sqrt{8}))=14.3068$. In this case, our numerical
709: results yield $q_{min}=14.2924$, a discrepancy of only $0.1\%$.
710:
711: %\vspace{6.5cm}
712:
713: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
714: \rem{
715: \begin{figure}
716: \begin{center}
717: \leavevmode{\hbox{\epsfig{
718: figure=fig7a.ps, scale=0.7
719: }}
720: }
721: \caption{\label{scores-of-peakon-collisions/b=3/Gaussian/w=10}
722: Scores of peakon collisions arise for (\ref{b-family-inviscid})
723: with $b=3$ starting from an initial
724: Gaussian velocity distribution of width $w=10$.
725: }
726: \end{center}
727: \end{figure}
728: }
729:
730: %\vskip-6.5cm
731: \remfigure{
732: \begin{figure}[htbp]
733: \centering
734: \psfig{file=fig7a.ps,height=6.5cm}
735: \caption[]{
736: Scores of peakon collisions arise for $b=3$ (DP) starting from an initial
737: Gaussian velocity distribution of width $w=10$.
738: }
739: \label{scores-of-peakon-collisions/b=3/Gaussian/w=10}
740: \end{figure}
741: }
742: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
743:
744: Of course, the two-peakon collision is rather simple compared to the
745: plethora of other multi-wave dynamics that occurs in this problem, as in
746: Figures \ref{scores-of-peakon-collisions/b=3/Gaussian/w=10} and
747: \ref{DP-peakons-2peakon-ic/w=5/alpha=1}. For this reason, we also checked
748: the convergence of our numerical algorithms by verifying that the relative
749: phases of the peakons in the various figures remained invariant under grid
750: refinement. Moreover, the integrity of the waveforms in our figures
751: attests to the convergence of the numerical algorithm -- after scores of
752: collisions, the peakon waveforms are still extremely well preserved. The
753: preservation of these peakon/soliton waveforms after so many collisions
754: would not have occurred unless the numerics had converged well.
755:
756: %\vspace{6.5cm}
757:
758: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
759: \rem{
760: \begin{figure}
761: \begin{center}
762: \leavevmode{\hbox{\epsfig{
763: figure=fig8a.ps, scale=0.7
764: }}
765: }
766: \caption{\label{DP-peakons-2peakon-ic/w=5/alpha=1}
767: Under DP in (\ref{DP-nonlin}) many peakons of unit width for $\alpha=1$
768: emerge from an initial pair of peakons of width $w=5$.
769: }
770: \end{center}
771: \end{figure}
772: }
773:
774: %\vskip-6.5cm
775: \remfigure{
776: \begin{figure}[htbp]
777: \centering
778: \psfig{file=fig8a.ps,height=6.5cm}
779: \caption[]{
780: Under DP in (\ref{b-family-inviscid}) with $b=3$, many peakons of unit
781: width for $\alpha=1$ emerge from an initial pair of peakons of width
782: $w=5$. }
783: \label{DP-peakons-2peakon-ic/w=5/alpha=1}
784: \end{figure}
785: }
786: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
787:
788: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
789: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
790: \section{Conclusions and further comments}\label{Conclusions}
791: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
792: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
793:
794: Equation (\ref{b-family-inviscid}) comprises a family of
795: reversible, parity invariant, evolutionary 1+1 PDEs that arise in the
796: zero-dispersion case of shallow water waves at quadratic order in the
797: standard asymptotic expansion,
798: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
799: \begin{equation}
800: \label{b-family-inviscid-sum}
801: m_t + um_x + b\,mu_x = 0\,,
802: \hbox{ with }
803: m = u - \alpha^2 u_{xx}
804: \,.
805: \end{equation}
806: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
807: The paper identified the bifurcations and exchanges of stability of its
808: traveling wave solutions as a function of the nonlinear balance parameter
809: $b$.
810:
811: Numerical computations of the initial value problem for
812: (\ref{b-family-inviscid-sum}) when $b>1$ showed the emergence of its
813: stable particle-like peakon solutions and their interactions. These may be
814: obtained analytically by superposing $N$ peakon traveling wave solutions
815: for the inviscid case,
816: $u(x,t)=ce^{-|x-ct|}$ as
817: \begin{equation}\label{peakon-soln-1}
818: u(x,t)=\sum_{i=1}^Np_i(t)e^{-|x-q_i(t)|}
819: \quad\hbox{and}\quad
820: m(x,t)=\sum_{i=1}^Np_i(t)\delta(x-q_i(t))\,,
821: \end{equation}
822: for any real constant $b$. For any $b>1$, the peakons are stable and
823: undergo particle-like dynamics in terms of the moduli variables $p_i(t)$
824: and $q_i(t)$, with $i=1,\dots,N$. The peakon dynamics studied for $b>1$
825: in this framework displayed all of the classical soliton interaction
826: behavior for peakons found in \cite{CH[1993]}, \cite{FH[2001]} for the case
827: $b=2$. This behavior included pairwise elastic scattering of peakons,
828: dominance of the initial value problem by confined pulses and asymptotic
829: sorting according to height -- all without requiring complete
830: integrability. Thus, the ``emergent pattern'' for $b>1$ in the nonlinear
831: evolution governed by the $b-$equation (\ref{b-family-inviscid-sum}) was
832: the rightward moving peakon train, ordered by height.
833:
834: A second type of emergent pattern of the initial value problem for
835: (\ref{b-family-inviscid-sum}) occurs in the parameter region
836: $0\le{b}<1$. This is the classic Burgers-type ramp/cliff structure as in
837: Figure \ref{final-states/b=0/b=2}. In contrast, for the parameter region
838: ${b}<-1$ a third type of behavior arises, consisting of leftward
839: moving structures as in Figure \ref{peakon-to-ramp/cliff-to-leftons}.
840: That the solution behavior should depend on the value of
841: $b$ is clear from the velocity form of equation
842: (\ref{b-family-inviscid-sum}) written in (\ref{visc-b-eqn-uform}).
843:
844: \paragraph{Three regions of $b$.} We found that the solution behavior for
845: the $b-$equation (\ref{b-family-inviscid-sum}) changes its character near
846: the two boundaries, $b=\pm1$, of the following three regions in the balance
847: parameter $b$.
848: \begin{description}
849: \item (B1)
850: In the stable peakon region $b>1$, the Steepening Lemma for the
851: $b-$equation (\ref{b-family-inviscid-sum}) proven in \cite{CH[1993]} for
852: $b=2$ and in \cite{HS2003} for $1<b\le3$ allows inflection points with
853: negative slopes to escape verticality by producing a jump in spatial
854: derivative at the peak of a traveling wave that eliminates the inflection
855: points altogether. Peakon behavior dominates this region. When $b\le1$ we
856: found the solution behavior of the $b-$equation
857: (\ref{b-family-inviscid-sum}) changed its character
858: and excluded the peakons entirely.
859:
860: \item (B2)
861: In the Burgers region $0\le{b}\le1$, the $L^{1/b}$ norm of the
862: variable $m$ is controlled%
863: \footnote{For $b=0$, this is a maximum principle for $|m|$.}
864: and the solution behavior is dominated by ramps and cliffs, as for
865: the usual Burgers equation.
866: Similar Burgers-type ramp/cliff solution properties hold for the region
867: $-1\le{b}\le0$, for which the $L^{1/b}$ norm of the variable $1/|m|$ is
868: controlled. At the boundary of the latter region, for $b=-1$, the active
869: transport equation (\ref{b-family-inviscid-sum}) admits stationary plane
870: waves as exact nonlinear solutions. However, shallow water asymptotic
871: ordering is broken at $b=-1$.
872:
873: \item (B3)
874: In the steady pulse region $b<-1$, pulse trains form that move
875: leftward from a positive velocity
876: initial condition (instead of moving rightward, as for $b>-1$). These
877: leftward-moving pulse trains are found to approach a steady state.
878:
879: \end{description}
880:
881: At the two boundaries, $b=\pm1$, of these three regions in $b$, the
882: emergent solutions of the $b-$family of equations
883: (\ref{b-family-inviscid-sum}) are observed numerically to exchange
884: stability. Although we discussed some analytical indications, a full
885: explanation of these properties remains open for future research.
886:
887: \subsection*{Acknowledgments}
888: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
889: We are grateful to A. Degasperis, A.~N.~W.~Hone,
890: J.~M.~Hyman, S.~Kurien, C.~D.~Levermore, R.~Lowrie and E.~S.~Titi for their
891: thoughtful insights and remarks. This work was supported by the US DOE under
892: contracts W-7405-ENG-36 and the Applied Mathematical Sciences Program
893: KC-07-01-01.
894:
895: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
896: \begin{thebibliography}{99}
897:
898: \bibitem{AblowitzClarkson[1991]}
899: M.J.~Ablowitz and P.A.~Clarkson,
900: {\it Solitons, Nonlinear Evolution Equations and
901: Inverse Scattering}, Cambridge University Press (1991).
902:
903: \bibitem{DGH[2003]}
904: H.R.~Dullin, D.D.~Holm and G. Gottwald,
905: \rem{An integrable shallow water equation with linear and nonlinear
906: dispersion}
907: {\it Phys. Rev. Lett.}, {\bf 87}, 194501 (2001).
908:
909: \bibitem{Kodama[1985&1987]}
910: Y. Kodama,
911: \rem{On integrable systems with higher order corrections}
912: {\it Phys. Lett. A} \textbf{107}, 245--249 (1985);
913: \rem{Normal forms for weakly dispersive wave equations}
914: Ibid \textbf{112}, 193-196 (1985);
915: \rem{On solitary-wave interaction}
916: Ibid \textbf{123}, 276--282 (1987).
917:
918: \bibitem{LiZhiSibgatullin[1997]}
919: Z. Li and N.R. Sibgatullin,
920: \rem{An Improved Theory of Long Waves on the Water Surface}
921: \textit{J. Appl. Maths. Mechs.} \textbf{61}, 177-182 (1997).
922:
923: \bibitem{CH[1993]} R.~Camassa and D.D.~Holm,
924: \rem{An integrable shallow water equation with peaked solitons}
925: {\it Phys. Rev. Lett.} {\bf71} 1661 (1993).
926:
927: \bibitem{DP[1999]}
928: A. Degasperis and M. Procesi, Asymptotic integrability, in
929: {\it Symmetry and Perturbation Theory}, edited by A. Degasperis and G.
930: Gaeta, World Scientific (1999) pp.23-37.
931:
932: \bibitem{DHH2002}
933: A. Degasperis, D. D. Holm and A. N. W. Hone,
934: {A new integrable equation with peakon solutions.}
935: {\it Theoret. and Math. Phys.} {\bf133}, 1463-1474 (2002).
936: http://xxx.lanl.gov/abs/nlin.SI/0205023.
937:
938: \bibitem{MN[2002]} A.S.~Mikhailov and V. Novikov, (private communication).
939:
940: %\bibitem{footnote1}
941:
942: \bibitem{FH[2001]} O.~Fringer and D.D.~Holm,
943: {\it Physica D} {\bf150} 237 (2001).
944:
945: \bibitem{HS2003} D.D.~Holm and M.F. Staley,
946: Wave structures and nonlinear balances in a family of 1+1
947: evolutionary PDEs.
948: To appear in {\it SIAM J. Dyn. Syst.} (2003).
949:
950: \bibitem{Leray[1934]}
951: J. Leray,
952: Sur le mouvement d'un liquide visqueux emplissant l'espace,
953: {\it Acta Math.} {\bf63}, 193-248 (1934).
954: Reviewed, e.g., in P. Constantin, C. Foias, B.
955: Nicolaenko \& R. Temam, {\it Integral manifolds and inertial
956: manifolds for dissipative partial differential equations}.
957: Applied Mathematical Sciences, {\bf70}, (Springer-Verlag, New
958: York-Berlin, 1989).
959:
960:
961: \bibitem{GH[2002]} B.G.~Geurts and D.D.~Holm,
962: Alpha-modeling strategy for LES of turbulent mixing. In {\it Turbulent Flow
963: Computation}, edited by D. Drikakis and B. G. Geurts,
964: Kluwer: London, pp. 237-278.
965:
966: \bibitem{Chen-etal[1998]}
967: S.Y.~Chen, C.~Foias, D.D.~Holm, E.J.~Olson, E.S.~Titi {and} S.~Wynne,
968: {\it Phys. Rev. Lett.} {\bf 81} 5338 (1998). Ibid, {\it Physica D} {\bf133} 49
969: (1999). Ibid, {\it Phys. Fluids} {\bf11} 2343 (1999).
970:
971: \bibitem{Chen-etal[1999c]}
972: S.Y.~Chen, D.D.~Holm, L.G.~Margolin {and} R.~Zhang,
973: {\it Physica D} {\bf133} 66 (1999).
974:
975: \bibitem{FHT[2001]}
976: C.~Foias, D.D.~Holm {and} E.S.~Titi,
977: {\it Physica D} {\bf 152} 505 (2001).
978: Ibid, {\it Dyn. and Diff. Eqns.} {\bf 14} 1 (2002).
979:
980: \end{thebibliography}
981:
982:
983: %\end{multicols}
984:
985:
986:
987: \end{document}
988:
989: