1: \documentclass[fleqn]{elsart}
2: \usepackage{amsmath}
3: \usepackage{bm}
4: \usepackage{url}
5: \usepackage{graphicx}
6:
7: %\renewcommand{\label}[1]{\quad(\textit{#1})}
8: %\renewcommand{\ref}[1]{\textit{#1}}
9: %\def\dotsc{\dots}
10: %\def\mod{\mathop{\rm mod}\nolimits}
11: %\newcommand{\text}[1]{{\rm #1}}
12:
13: \renewcommand{\vec}[1]{\hm{#1}}
14: \def\B{\text{B}}
15:
16:
17: \newcommand{\lvec}[1]{\line(-1,0){#1}\put(-#1,0)%
18: {\put(-0.8,0){\line(4,1){4}}\put(-0.8,0){\line(4,-1){4}}}}
19: \newcommand{\rvec}[1]{\line(1,0){#1}\put(0,0)%
20: {\put(0.8,0){\line(-4,1){4}}\put(0.8,0){\line(-4,-1){4}}}}
21:
22: \def\drawgrid{%
23: \put(2,1){$0$}%
24: \put(0,5){\vector(1,0){130}}%
25: \put(124,1){$\pi$}%
26: \put(128,7){$k$}%
27: \put(5,0){\vector(0,1){45}}%
28: \put(1,42){$A$}%
29: \put(5,17){\line(-1,0){1}}%
30: \put(2,13){$1$}%
31: \put(5,29){\line(-1,0){1}}%
32: \put(2,25){$2$}%
33: \put(5,41){\line(-1,0){1}}%
34: \put(2,37){$3$}%
35: \put( 15,5){\line(0,-1){1}}%
36: \put( 25,5){\line(0,-1){1}}%
37: \put( 35,5){\line(0,-1){1}}%
38: \put( 45,5){\line(0,-1){1}}%
39: \put( 55,5){\line(0,-1){1}}%
40: \put( 65,5){\line(0,-1){1}}%
41: \put( 75,5){\line(0,-1){1}}%
42: \put( 85,5){\line(0,-1){1}}%
43: \put( 95,5){\line(0,-1){1}}%
44: \put(105,5){\line(0,-1){1}}%
45: \put(115,5){\line(0,-1){1}}%
46: \put(125,5){\line(0,-1){1}}%
47: \put( 15,5){\dashbox{1}(0,36){}}% 1
48: \put( 25,5){\dashbox{1}(0,36){}}% 2
49: \put( 35,5){\dashbox{1}(0,36){}}% 3
50: \put( 45,5){\dashbox{1}(0,36){}}% 4
51: \put( 55,5){\dashbox{1}(0,36){}}% 5
52: \put( 65,5){\dashbox{1}(0,36){}}% 6
53: \put( 75,5){\dashbox{1}(0,36){}}% 7
54: \put( 85,5){\dashbox{1}(0,36){}}% 8
55: \put( 95,5){\dashbox{1}(0,36){}}% 9
56: \put(105,5){\dashbox{1}(0,36){}}% 10
57: \put(115,5){\dashbox{1}(0,36){}}% 11
58: \put(125,5){\dashbox{1}(0,36){}}% 12
59: }
60:
61: \def\drawgridA{%
62: \put(2,1){$0$}%
63: \put(0,5){\vector(1,0){130}}%
64: \put(124,1){$\pi$}%
65: \put(128,7){$k$}%
66: \put(5,0){\vector(0,1){45}}%
67: \put(1,42){$A$}%
68: \put(5,41){\line(-1,0){1}}%
69: \put(2,37){$1$}%
70: \put( 15,5){\line(0,-1){1}}%
71: \put( 25,5){\line(0,-1){1}}%
72: \put( 35,5){\line(0,-1){1}}%
73: \put( 45,5){\line(0,-1){1}}%
74: \put( 55,5){\line(0,-1){1}}%
75: \put( 65,5){\line(0,-1){1}}%
76: \put( 75,5){\line(0,-1){1}}%
77: \put( 85,5){\line(0,-1){1}}%
78: \put( 95,5){\line(0,-1){1}}%
79: \put(105,5){\line(0,-1){1}}%
80: \put(115,5){\line(0,-1){1}}%
81: \put(125,5){\line(0,-1){1}}%
82: \put( 15,5){\dashbox{1}(0,36){}}% 1
83: \put( 25,5){\dashbox{1}(0,36){}}% 2
84: \put( 35,5){\dashbox{1}(0,36){}}% 3
85: \put( 45,5){\dashbox{1}(0,36){}}% 4
86: \put( 55,5){\dashbox{1}(0,36){}}% 5
87: \put( 65,5){\dashbox{1}(0,36){}}% 6
88: \put( 75,5){\dashbox{1}(0,36){}}% 7
89: \put( 85,5){\dashbox{1}(0,36){}}% 8
90: \put( 95,5){\dashbox{1}(0,36){}}% 9
91: \put(105,5){\dashbox{1}(0,36){}}% 10
92: \put(115,5){\dashbox{1}(0,36){}}% 11
93: \put(125,5){\dashbox{1}(0,36){}}% 12
94: }
95:
96: \def\drawgridB{%
97: \put(2,1){$0$}%
98: \put(0,5){\vector(1,0){130}}%
99: \put(124,1){$\pi$}%
100: \put(128,7){$k$}%
101: \put(5,0){\vector(0,1){45}}%
102: \put(1,42){$A$}%
103: \put(5,23){\line(-1,0){1}}%
104: \put(2,19){$1$}%
105: \put(5,41){\line(-1,0){1}}%
106: \put(2,37){$2$}%
107: \put( 15,5){\line(0,-1){1}}%
108: \put( 25,5){\line(0,-1){1}}%
109: \put( 35,5){\line(0,-1){1}}%
110: \put( 45,5){\line(0,-1){1}}%
111: \put( 55,5){\line(0,-1){1}}%
112: \put( 65,5){\line(0,-1){1}}%
113: \put( 75,5){\line(0,-1){1}}%
114: \put( 85,5){\line(0,-1){1}}%
115: \put( 95,5){\line(0,-1){1}}%
116: \put(105,5){\line(0,-1){1}}%
117: \put(115,5){\line(0,-1){1}}%
118: \put(125,5){\line(0,-1){1}}%
119: \put( 15,5){\dashbox{1}(0,36){}}% 1
120: \put( 25,5){\dashbox{1}(0,36){}}% 2
121: \put( 35,5){\dashbox{1}(0,36){}}% 3
122: \put( 45,5){\dashbox{1}(0,36){}}% 4
123: \put( 55,5){\dashbox{1}(0,36){}}% 5
124: \put( 65,5){\dashbox{1}(0,36){}}% 6
125: \put( 75,5){\dashbox{1}(0,36){}}% 7
126: \put( 85,5){\dashbox{1}(0,36){}}% 8
127: \put( 95,5){\dashbox{1}(0,36){}}% 9
128: \put(105,5){\dashbox{1}(0,36){}}% 10
129: \put(115,5){\dashbox{1}(0,36){}}% 11
130: \put(125,5){\dashbox{1}(0,36){}}% 12
131: }
132:
133:
134:
135: \begin{document}
136: \begin{frontmatter}
137: \journal{Physica D}
138: \title{Stability of low-dimensional bushes of vibrational modes in the
139: Fermi-Pasta-Ulam chains}
140: \author{G.M.~Chechin},
141: \ead{chechin@phys.rsu.ru}
142: \author{D.S.~Ryabov},
143: \author{K.G.~Zhukov}
144: \address{Department of Physics, Rostov State University, Zorge 5,
145: Rostov-on-Don, 344090 Russia}
146: \begin{abstract}
147: Bushes of normal modes represent the exact excitations in the nonlinear
148: physical systems with discrete symmetries~[Physica~D \textbf{117} (1998)
149: 43]. The present paper is the continuation of our previous paper~[Physica~D
150: \textbf{166} (2002) 208], where these dynamical objects of new type were
151: discussed for the monoatomic nonlinear chains. Here, we develop a simple
152: crystallographic method for finding bushes in nonlinear chains and
153: investigate stability of one-dimensional and two-dimensional vibrational
154: bushes for both FPU-$\alpha$ and FPU-$\beta$ models, in particular, of
155: those revealed recently in~[Physica~D \textbf{175} (2003) 31].
156: \end{abstract}
157: \begin{keyword}
158: Nonlinear dynamics; Discrete symmetry; Anharmonic lattices; Normal mode
159: interactions
160: \PACS 05.45.-a; 45.90.+t; 63.20.Ry; 63.20.Dj
161: \end{keyword}
162: \end{frontmatter}
163:
164: \maketitle
165:
166:
167:
168: \section{Introduction}
169:
170: This paper expands upon our previous work~\cite{FPU1} where an outline of
171: the general theory of the bushes of modes~\cite{DAN1,DAN2,PhysD} was
172: presented in connection with the FPU-chain dynamics. Nevertheless, it seems
173: reasonable to recapitulate here some basic notions and ideas.
174:
175:
176:
177: \subsection{The concept of bushes of modes}
178:
179: Bushes of modes can be considered as a new type of \textit{exact}
180: excitations in \textit{nonlinear} systems with discrete symmetries, such as
181: molecules and crystals. The simplest way for introducing this concept can
182: be described as follows.
183:
184: Let us have an $N$-particle Hamiltonian system characterized by a symmetry
185: group $G_0$ in its equilibrium state. Let us also assume that this system
186: permits the harmonic approximation and, therefore, we can introduce a
187: complete set of normal modes. Note that every normal mode possesses its own
188: symmetry group $G_j$ which is a subgroup of the group $G_0$. Let us now
189: excite only one normal mode at the initial instant by imposing the
190: appropriate initial conditions. We call this mode by the term ``root
191: mode''. Normal modes are independent of each other in the harmonic
192: approximation. However, if we take into account some anharmonic terms of
193: the Hamiltonian, the excitation will transfer from the root mode to a
194: number of other normal modes with zero amplitudes at the initial instant,
195: so-called ``secondary modes''. Because of certain \textit{selection rules}
196: for the excitation transfer between modes of different symmetry, the number
197: of the secondary modes can often be rather \textit{small}.
198:
199: \textit{Definition}. The complete collection of the root mode and all
200: secondary modes, corresponding to it, forms a \textit{bush of normal
201: modes}. The number of modes in this collection is the \textit{dimension} of
202: the given bush.
203:
204: To avoid any misunderstanding at this point, let us note that one can
205: consider normal modes, determined for a linear system, as a basis for a
206: decomposition of different dynamical regimes in the nonlinear system (for
207: more details, see below).
208:
209: The energy of the initial excitation turns out to be trapped in the bush
210: simply because of the above definition. The number of modes in the bush
211: does not change in time while the amplitudes of these modes do change.
212:
213: For many systems, we can find one-dimensional, two-dimensional,
214: three-dimensional and so on bushes of modes. Note that one-dimensional
215: bushes can be treated as the \textit{similar nonlinear normal modes}
216: introduced by Rosenberg forty years ago~\cite{Rosenberg}.
217:
218: Every bush possesses its own symmetry group $G$ which is a
219: \textit{subgroup} of the symmetry group of the Hamiltonian of the
220: considered system. In the above discussed case, when the bush is excited by
221: perturbing the root mode, the symmetry group $G$ of the bush coincides with
222: that of this root mode. The symmetries of all other modes of the bush are
223: \textit{greater} than or \textit{equal} to its full symmetry $G$. The
224: development of the above ideas leads to the following important statement
225: of the general theory~\cite{FPU1,DAN1,PhysD}:
226:
227: \textit{Proposition}. Different nonlinear dynamical regimes of a physical
228: system can be classified by subgroups of its symmetry group $G_0$
229: (so-called ``parent'' group)\footnote{We prefer to use this term because
230: the considered theory is valid not only for the Hamiltonian systems, but
231: also for the dissipative systems.}.
232:
233: It is important to emphasize that the considered bushes of modes are
234: \textit{symmetry determined} objects: the sets of their modes \textit{do
235: not depend} on the interparticle interactions in the physical system.
236: Taking into account the specific character of these interactions can only
237: \textit{reduce} the dimension of the bush.
238:
239:
240:
241: \subsection{Some properties of bushes of modes}
242:
243: In general case, we must speak about bushes of \textit{symmetry-adapted}
244: modes rather than normal modes. Indeed, we consider the symmetry determined
245: bushes whose finding does not need any information about the interactions
246: in the system. The symmetry-adapted coordinates are the \textit{basis
247: vectors} of the \textit{irreducible representations} (irreps) of the parent
248: symmetry group $G_0$ and they can be found by group-theoretical methods
249: only, without any information about the interactions. In contrast, the
250: normal coordinates, used for construction of normal modes, must be
251: obtained, generally, by diagonalization of the Hamiltonian matrix and,
252: therefore, the interparticle interactions must be taken into account.
253:
254: In the geometrical sense, a bush of modes is an \textit{invariant manifold}
255: singled out by its symmetry and decomposed into the basis vectors of the
256: irreducible representations of the appropriate parent symmetry group.
257:
258: This decomposition plays a very important role in our approach. Indeed,
259: different irreps describe the transformation properties of the dynamical
260: variables corresponding to \textit{different physical characteristics} of
261: the considered system. For example, some irreps are active in the infrared
262: or Raman experiments, while others are nonactive in optics, but play the
263: essential role in neutronographics, etc.
264:
265: In this paper, we discuss bushes of \textit{vibrational} modes describing
266: the time-dependent displacements of the particles of nonlinear chains from
267: their equilibrium positions. Speaking about the symmetry group of a single
268: mode or of the bush as integral object, we mean the specific symmetry of
269: the \textit{patterns} of the instantaneous displacements of all particles
270: which correspond to the mode or to the bush. Considering the symmetry group
271: of the displacement pattern and the collection of the normal modes
272: contained in a given bush, we deal with the \textit{geometrical} aspect of
273: the bush. This aspect is independent of the interparticle interactions in
274: the chain.
275:
276: On the other hand, we deal with the \textit{dynamical} aspect of the bush
277: when consider the differential equations describing the time-dependence of
278: the amplitudes of the bush modes. In this sense, the given bush represents
279: a \textit{reduced} dynamical system whose dimension is, in general, less
280: (often, considerably less!) than the dimension of the original physical
281: system. Obviously, the dynamical aspect of the bush, in contrast to the
282: geometrical aspect, depends essentially on the interparticle interactions.
283:
284: The modes of the bush are coupled by the so-called ``force interactions'',
285: while their coupling to all other modes is brought about by ``parametric
286: interactions''~\cite{PhysD}. The latter interactions can lead to a loss of
287: stability of the bush, if the amplitudes of its modes become sufficiently
288: large~\cite{PhysD,FPU1,Okta}, because of the phenomenon similar to the
289: well-known parametric resonance. In this case, the spontaneous breaking of
290: the symmetry of the system's vibrational state takes place and, as a
291: consequence, the given bush transforms into a bush of larger dimension.
292: Precisely this mechanism of stability loss by the bushes of the FPU chains
293: is discussed in this paper. The other mechanisms of the loss of bush
294: stability will be considered elsewhere.
295:
296:
297:
298: \subsection{Some history}
299:
300: The general concept of bushes of modes for physical systems with discrete
301: symmetry was introduced in~\cite{DAN1}. Actually, this concept appeared as
302: a generalization of the notion of the complete condensate of primary and
303: secondary order parameters, whose theory was developed in our papers
304: devoted to phase transitions in crystals~\cite{PhysStatSol,Acta}. The
305: discussion of properties of bushes of modes, the group-theoretical methods
306: for their construction, some theorems about the bush structure can be found
307: in~\cite{DAN1,DAN2,PhysD}. The bushes of modes for different classes of
308: mechanical systems with point and space symmetry were investigated
309: in~\cite{DAN2,PhysD,Okta,NNM,ENOC3,C60}. In particular, let us mention the
310: investigation of
311:
312: -- all possible ``irreducible bushes'' and symmetry determined similar
313: nonlinear normal modes for the mechanical systems with any of the 230 space
314: groups~\cite{NNM};
315:
316: -- low-dimensional bushes for all point groups of crystallographic
317: symmetry~\cite{ENOC3};
318:
319: -- bushes of vibrational modes for the fullerene $C_{60}$ (``buckyball''
320: structure)~\cite{C60};
321:
322: -- bushes of modes and their stability for the octahedral structure with
323: Lennard-Jones potential~\cite{Okta}.
324:
325: A review of the group-theoretical methods and the appropriate computer
326: algorithms for treating the condensates of order parameters and bushes of
327: modes can be found in~\cite{Comput}. A remarkable computer program
328: \texttt{ISOTROPY} by Stokes and Hatch, capable to deal with many
329: group-theoretical methods of crystal physics and, in particular, with
330: bushes of modes, is now available on the Internet as free
331: software~\cite{Isotropy}.
332:
333: In~\cite{PoggiRuffo}, Poggi and Ruffo revealed some exact solutions for the
334: FPU-$\beta$ chain dynamics, by analyzing the appropriate dynamical
335: equations, and investigated their stability. These solutions correspond to
336: certain subsets of normal modes (so-called ``subsets of I type'') ``where
337: energy remains trapped for suitable initial conditions''.
338:
339: In~\cite{FPU1}, we showed that these subsets are the simplest cases of the
340: bushes of modes for the monoatomic nonlinear chains and that they can be
341: found without taking into account the information about the specific
342: interactions in the FPU-$\beta$ model. We discussed there the
343: symmetry-determined bushes of modes with respect to the translational group
344: $T$, in detail, and the bushes with respect to the dihedral group $D$,
345: partially. Also, in the paper~\cite{FPU1}, the stability of the former
346: bushes for the FPU-$\alpha$ chain was studied.
347:
348: In a recent paper~\cite{BRink}, Bob Rink developed a method for finding all
349: possible symmetric invariant manifolds for the monoatomic chains. In spite
350: of the fact that this paper was carried out independently of our approach,
351: based on the concept of bushes of modes, and that it is written in a more
352: rigorous mathematical style, the main idea and the group-theoretical method
353: are very similar to those in our previous papers. Actually, Bob Rink found
354: all symmetry-determined bushes of modes for the monoatomic chains with
355: respect to the dihedral symmetry group. Moreover, he revealed some bushes
356: for the FPU-$\beta$ chain which are brought about by the additional
357: symmetry of this mechanical system connected with the evenness of its
358: potential (hereafter we call them ``additional bushes''). However, let us
359: note that the general idea of the classification of the symmetric manifolds
360: (bushes of modes) by the \textit{subgroups} of the symmetry group of the
361: Hamiltonian was not emphasized explicitly in this paper.
362:
363: Finally, we would like to comment on two papers by
364: Shinohara~\cite{Shinohara}, devoted to finding ``type I subsets of modes''
365: (bushes, in our terminology) in nonlinear chains with fixed endpoints. The
366: author does not use any symmetry-related method and prefers to analyze the
367: specific structure of the dynamical equations (practically, in the manner
368: similar to that of Poggi and Ruffo in~\cite{PoggiRuffo}). Moreover, he
369: writes that the ``approach, which relies on the direct analysis of the
370: mode-coupling coefficients, is much simpler'' than the one based on the
371: group-theoretical methods. In contrast to this opinion, we think that the
372: situation is quite the opposite! In particular, we have tried to
373: demonstrate in the present paper that the symmetry-related methods are much
374: simpler and much more transparent than those based on investigating the
375: structure of the dynamical equations. This simplicity becomes especially
376: apparent when one deals with physical objects more complex than monoatomic
377: chains, such as molecules and crystals characterized by nontrivial symmetry
378: groups.
379:
380: Note that comparing our results and those from~\cite{Shinohara}, one must
381: take into account that we use periodic boundary conditions, while Shinohara
382: discusses the nonlinear chains with fixed endpoints (the former boundary
383: conditions are more general, since the latter conditions can be considered
384: as their special case). We would also like to note that the stability
385: problem for the bushes of modes (invariant manifolds or ``modes subsets of
386: I type'') is not considered in~\cite{Shinohara}. On the other hand, this
387: problem seems to be important for treating the induction phenomenon
388: discussed in those papers.
389:
390:
391: \subsection{The structure of the present paper}
392:
393: In Section~2, we consider a simple crystallographic method for obtaining
394: bushes of modes in the space of atomic displacements which differ from that
395: in our previous paper~\cite{FPU1} and in the paper~\cite{BRink} by Bob
396: Rink. A major advantage of this method is its remarkable clarity. Actually,
397: we obtain the invariant manifolds corresponding to the subgroups of the
398: dihedral group and then decompose them into the normal modes of the
399: monoatomic chain. Here we restrict ourselves to the discussion of one- and
400: two-dimensional bushes only (about other bushes see~\cite{FPU2a}).
401:
402: The dynamical description of the bushes of normal modes is presented
403: briefly in Section~3.
404:
405: The general approach to analyzing the parametric stability of the bushes of
406: normal modes is discussed in Section~4.
407:
408: Sec.~5 is devoted to investigation of the stability of the one-dimensional
409: (1D) bushes in the FPU-$\alpha$ and FPU-$\beta$ chains. Only one 1D bush,
410: B$[\hat{a}^2,\hat{i}]$, can be obtained for the monoatomic chain with an
411: even number of particles\footnote{It was denoted by the symbol B$[2a]$ in
412: the paper~\cite{FPU1}.}, if the translational group $T$ is considered as
413: the appropriate parent group of this mechanical system. The stability of
414: the bush B$[\hat{a}^2,\hat{i}]$ was studied for the FPU-$\alpha$ and
415: FPU-$\beta$ chains in the papers~\cite{FPU1} and~\cite{PoggiRuffo},
416: respectively.\footnote{The stability of this bush, consisting of only one
417: so-called $\pi$-mode, was also discussed in~\cite{Page}.} However, we
418: obtain some additional one-dimensional bushes considering the dihedral
419: group $D$ as the parent symmetry group~\cite{FPU1,FPU2a}. Moreover, as was
420: already mentioned, there exist still more 1D bushes for the FPU-$\beta$
421: model because of the evenness of its potential~\cite{BRink}. Therefore, it
422: is interesting to compare the regions of stability of all these bushes in
423: both FPU-$\alpha$ and FPU-$\beta$ models. We perform such a comparison in
424: this section. An exceptional case is represented by the bush
425: B$[\hat{a}^4,\hat{a}\hat{i}]$ for the FPU-$\alpha$ chain for which the
426: threshold of stability loss is exactly equal to zero (this result can be
427: proved analytically).
428:
429: We discuss here the stability of 1D bushes not only for the FPU chains with
430: the finite number of particles~($N$), but for the continuum limit
431: ($N\rightarrow\infty$), as well.
432:
433: In Section~6 we discuss the shape of the stability regions of the
434: two-dimensional (2D) bushes for the FPU-$\alpha$ and FPU-$\beta$ chains.
435: Indeed, thresholds of the loss of stability of the 1D bushes depend only on
436: their energy, while the similar thresholds for multi-dimensional bushes
437: depend on the complete set of the initial conditions for the dynamical
438: equations of the bush. We demonstrate, with the aid of the appropriate
439: plots, how nontrivial the stability regions of the 2D bushes for the FPU
440: chains can be. The dependence of these regions on the total number of
441: particles in the chains is partially discussed.
442:
443: In Conclusion (Sec.~7), the obtained results are summarized and some new
444: directions for the bush theory are outlined.
445:
446:
447:
448:
449: \section{Simple crystallographic method for finding bushes of vibrational
450: modes in the monoatomic chains}
451:
452: We consider the longitudinal vibrations of $N$-particle monoatomic chains
453: with periodic boundary conditions. Let the $N$-dimensional vector
454: \[\vec{X}(t)=\{x_1(t),x_2(t),\dotsc,x_N(t)\}\]
455: describe all the displacements, $x_i(t)$, of the individual particles
456: (atoms) from their equilibrium states at the instant $t$. In accordance
457: with the above-mentioned periodic boundary conditions, we demand for
458: arbitrary time $t$:
459: \begin{equation}
460: x_{N+1}(t)\equiv x_1(t).
461: \label{eq1}
462: \end{equation}
463:
464: In the equilibrium state, a given chain is invariant under the action of
465: the operator $\hat{a}$ which shifts the chain by the lattice spacing $a$.
466: This operator generates the translational group
467: \begin{equation}
468: T=\{\hat{E},\hat{a},\hat{a}^2,\dotsc,\hat{a}^{N-1}\},\quad\hat{a}^N=\hat{E},
469: \label{eq2}
470: \end{equation}
471: where $\hat{E}$ is the identity element and $N$ is the order of the cyclic
472: group~$T$. The operator $\hat{a}$ induces the cyclic permutation of all
473: particles of the chain and, therefore, it acts on the ``configuration
474: vector'' $\vec{X}(t)$ as follows:
475: \[\hat{a}\vec{X}(t)\equiv
476: \hat{a}\{x_1(t),x_2(t),\dotsc,x_{N-1}(t),x_N(t)\}=
477: \{x_N(t),x_1(t),x_2(t),\dotsc,x_{N-1}(t)\}.\]
478:
479: The full symmetry group of the monoatomic chain contains also the inversion
480: $\hat{i}$, with respect to the center of the chain, which acts on the
481: vector $\vec{X}(t)$ in the following manner:
482: \[\hat{i}\vec{X}(t)\equiv
483: \hat{i}\{x_1(t),x_2(t),\dotsc,x_{N-1}(t),x_N(t)\}=
484: \{-x_N(t),-x_{N-1}(t),\dotsc,-x_2(t),-x_1(t)\}.\]
485:
486: The complete set of all products $\hat{a}^k\hat{i}$ of the pure
487: translations $\hat{a}^k$ ($k=1,2,\dotsc,N-1$) with the inversion $\hat{i}$
488: forms the so-called dihedral group $D$ which can be written as a direct sum
489: of two cosets $T$ and $T\cdot\hat{i}$:
490: \begin{equation}
491: D=T\oplus T\cdot\hat{i}.
492: \label{eq500}
493: \end{equation}
494: The dihedral group is non-Abelian group induced by two generators
495: ($\hat{a}$ and $\hat{i}$) with the following generating relations
496: \begin{equation}
497: \hat{a}^N=\hat{E},\quad\hat{i}^2=\hat{E},\quad\hat{i}\hat{a}=\hat{a}^{-1}\hat{i}.
498: \label{eq170a}
499: \end{equation}
500:
501: Let us consider the geometrical interpretation of the symmetry elements of
502: the dihedral group $D$. For simplification, we will use the reflection
503: $\hat{\sigma}$ in the mirror plane orthogonal to the chain instead of the
504: inversion $\hat{i}$. Indeed, the actions of $\hat{\sigma}$ and $\hat{i}$ on
505: the vectors of the one-dimensional space are equivalent to each other
506: ($\hat{\sigma}x=\hat{i}x=-x$). Then we can use the well-known theorem of
507: crystallography: ``Any mirror plane and the translation $\vec{A}$
508: orthogonal to it generate a new mirror plane, parallel to the former plane
509: and displaced away from it by $\vec{A}/2$''.
510:
511: Let the number of particles in the chain ($N$) be an \textit{even} integer.
512: Then it is easy to check that the products
513: $\hat{a}^k\hat{\sigma}\equiv\hat{a}^k\hat{i}$ ($k=\pm 1,\pm 2,\pm
514: 3,\dotsc$) generate the family of parallel planes depicted by vertical
515: lines\footnote{For simplicity, the hats above the operators are dropped in
516: all our pictures.} in Fig.~\ref{fig1}. For odd $k$, these planes pass
517: \textit{through} the atoms, while for the even $k$ they pass
518: \textit{between} atoms, exactly at the middle of the neighboring particles.
519: Note that, for odd $N$, everything will be vice versa: for even $k$, the
520: planes pass through the atoms and for the add $k$ -- between them. In
521: Fig.~\ref{fig1}, we show a part of the family of the above mentioned planes
522: near the center of the chain. It can be verified that the symmetry elements
523: $\hat{a}^k\hat{\sigma}\equiv\hat{a}^k\hat{i}$ ($k=1,2,3,\dotsc$) are
524: located to the right of the center of the chain, while the elements
525: $\hat{\sigma}\hat{a}^k\equiv\hat{i}\hat{a}^k$ are located to the left from
526: it\footnote{Note that
527: $\hat{a}^k\hat{i}=\hat{i}\hat{a}^{-k}=\hat{i}\hat{a}^{N-k}$.}.
528: \begin{figure}
529: \centering
530: \setlength{\unitlength}{1mm}
531: \begin{picture}(100,40)(0,0)
532: \put(0,20){\line(1,0){100}}
533: \put(10,5){\line(0,1){35}}
534: \put(10,0){$ia^4$}
535: \put(20,0){\line(0,1){35}}
536: \put(20,20){\circle*{2}}
537: \put(20,37){$ia^3$}
538: \put(30,5){\line(0,1){35}}
539: \put(30,0){$ia^2$}
540: \put(40,0){\line(0,1){35}}
541: \put(40,20){\circle*{2}}
542: \put(40,37){$ia$}
543: \put(50,5){\line(0,1){35}}
544: \put(50,0){$i$}
545: \put(60,0){\line(0,1){35}}
546: \put(60,20){\circle*{2}}
547: \put(60,37){$ai$}
548: \put(70,5){\line(0,1){35}}
549: \put(70,0){$a^2i$}
550: \put(80,0){\line(0,1){35}}
551: \put(80,20){\circle*{2}}
552: \put(80,37){$a^3i$}
553: \put(90,5){\line(0,1){35}}
554: \put(90,0){$a^4i$}
555: \end{picture}
556: \caption{\label{fig1} Symmetry elements of the dihedral group $D$ near the
557: center of the monoatomic chain.}
558: \end{figure}
559:
560: Let us now consider the subgroups $G_j$ of the dihedral group $D$, keeping
561: in mind that every subgroup $G_j\subset D$ induces the certain bush of
562: modes.
563:
564: Each group $G_j$ contains its own translational subgroup $T_j\subset T$,
565: where $T$ is the above discussed full translational group~(\ref{eq2}). If
566: $N$ is divisible by 4 (for example, we consider below the case $N=12$)
567: there exists the subgroup $T_4=[\hat{a}^4]$ of the group $T=[\hat{a}]$.
568: Note that in square brackets we write down only generators of the
569: considered group, while the complete set of group elements is written in
570: curly brackets (see, for example, Eqs.~(\ref{eq2})).
571:
572: If a vibrational state of the chain possesses the symmetry group
573: $T_4=[\hat{a}^4]\equiv\{\hat{E},\hat{a}^4,\hat{a}^8,\dotsc,\hat{a}^{N-4}\}$,
574: displacements of the atoms, which are apart by the distance $4a$ from each
575: other in the equilibrium state, turn out to be equal. Therefore, for the
576: case $N=12$, we can write the following atomic displacement pattern:
577: \begin{equation}
578: \vec{X}(t)=\{x_1(t),x_2(t),x_3(t),x_4(t)~|~x_1(t),x_2(t),x_3(t),x_4(t)~|~x_1(t),%
579: x_2(t),x_3(t),x_4(t)\},
580: \label{eq11}
581: \end{equation}
582: where $x_i(t)$ ($i=1,2,3,4$) are arbitrary functions of time. The
583: pattern~(\ref{eq11}) determines the bush corresponding to the symmetry
584: group $T_4=[\hat{a}^4]$. Thus, the complete set of the atomic displacements
585: can be divided into $N/4$ (in our case, $N/4=3$) identical subsets, and
586: each of them determines the so-called ``Extended Primitive Cell''
587: (EPC)\footnote{This term is used in the theory of phase transitions in
588: crystals.}. For the bush~(\ref{eq11}), the EPC contains four atoms, and the
589: vibrational state of the whole chain is described by three such EPC. In
590: other words, the EPC for the vibrational state with the symmetry group
591: $T_4=[\hat{a}^4]$ (its size is equal to $4a$) is four times larger than the
592: primitive cell ($a$) of the chain in the equilibrium state.
593:
594: It is essential that some symmetry elements of the dihedral group $D$
595: disappear as a result of the symmetry reduction
596: $D=[\hat{a},\hat{i}]\rightarrow T_4=[\hat{a}^4]$. In our case, there are no
597: nontrivial symmetry elements of the group $D$ inside the EPC and, as a
598: consequence, there are no restrictions on the displacements belonging to
599: one and the same extended cell (this fact is obvious from
600: Eq.~(\ref{eq11})).
601:
602: There are four other subgroups of the dihedral group $D$ to which the same
603: translational subgroup $T_4=[\hat{a}^4]$ correspond:
604: \begin{equation}
605: [\hat{a}^4,\hat{i}],\quad [\hat{a}^4,\hat{a}\hat{i}],\quad
606: [\hat{a}^4,\hat{a}^2\hat{i}],\quad [\hat{a}^4,\hat{a}^3\hat{i}].
607: \label{eq50}
608: \end{equation}
609: These subgroups possess \textit{two generators}, namely, the common
610: translational generator $\hat{a}^4$ and different inversion elements
611: $\hat{a}^k\hat{i}$ ($k=0,1,2,3$). These inversion elements differ from each
612: other by the \textit{position} of the center of inversion (see
613: Fig.~\ref{fig1} where they are depicted by the vertical lines representing
614: the mirror planes).
615:
616: Note that subgroups $[\hat{a}^4,\hat{a}^k\hat{i}]$ with $k>3$ are
617: equivalent to those from the list~(\ref{eq50}), because the second
618: generator $\hat{a}^k\hat{i}$ can be multiplied by $\hat{a}^{-4}$,
619: representing the inverse element with respect to the first generator
620: $\hat{a}^4$. Thus, there exist only five subgroups of the dihedral group
621: (with $N\mod 4=0$) constructed on the basis of the translational group
622: $T_4=[\hat{a}^4]$, namely, this group and the four groups from the
623: list~(\ref{eq50}).
624:
625: Now, let us consider the bushes corresponding to the
626: subgroups~(\ref{eq50}).
627:
628: 1. The subgroup $[\hat{a}^4,\hat{i}]$ consists of the following six
629: elements:
630: \begin{equation}
631: \hat{E},\hat{a}^4,\hat{i},\hat{a}^4\hat{i},
632: \hat{a}^8\hat{i}\equiv\hat{i}\hat{a}^4.
633: \label{eq51a}
634: \end{equation}
635:
636: The diagram of this subgroup, similar to that of the full dihedral group
637: $D$ in Fig.~\ref{fig1}, is depicted in Fig.~\ref{fig2}.
638: \begin{figure}
639: \centering
640: \setlength{\unitlength}{1mm}
641: \begin{picture}(100,40)(0,0)
642: \put(0,20){\line(1,0){100}}
643: \put(10,5){\line(0,1){35}}
644: \put(10,0){$ia^4$}
645: \put(20,20){\circle*{2}}
646: \put(20,22){$x_1$}
647: \put(40,20){\circle*{2}}
648: \put(40,22){$x_2$}
649: \put(50,5){\line(0,1){30}}
650: \put(50,0){$i$}
651: \put(60,20){\circle*{2}}
652: \put(60,22){$x_3$}
653: \put(80,20){\circle*{2}}
654: \put(80,22){$x_4$}
655: \put(90,5){\line(0,1){35}}
656: \put(90,0){$a^4i$}
657: \put(40,37){\vector(-1,0){30}}
658: \put(60,37){\vector(1,0){30}}
659: \put(43,36){EPC ($4a$)}
660: \linethickness{0.5mm}
661: %\put(20,20){\vector(1,0){10}}
662: \put(20,20){\rvec{10}}
663: %\put(40,20){\vector(1,0){7}}
664: \put(40,20){\rvec{7}}
665: %\put(60,20){\vector(-1,0){7}}
666: \put(60,20){\lvec{7}}
667: %\put(80,20){\vector(-1,0){10}}
668: \put(80,20){\lvec{10}}
669: \end{picture}
670: \caption{\label{fig2} Vibrational bush corresponding to the subgroup
671: $G=[\hat{a}^4,\hat{i}]$.}
672: \end{figure}
673:
674: Two symmetry elements ($\hat{i}\hat{a}^4$ and $\hat{a}^4\hat{i}$), situated
675: at the boundaries of the EPC, do not produce any additional restrictions on
676: the atomic displacements \textit{inside} this EPC. Indeed, they connect
677: with each other the atomic displacements in the adjacent extended primitive
678: cells, while the equivalence of the displacements of the atoms with the
679: same positions in the adjacent EPC is guaranteed by the translation
680: $\hat{a}^4$ which is the first generator of the considered subgroup
681: $[\hat{a}^4,\hat{i}]$.
682:
683: On the other hand, the inversion $\hat{i}$ at the center of the EPC
684: (Fig.~\ref{fig2}), surviving in the subgroup $[\hat{a}^4,\hat{i}]$, demands
685: the displacements of the atoms, symmetrically situated with respect to this
686: center, to be equal in value, but opposite in sign:
687: \begin{equation}
688: x_1(t)=-x_4(t),\quad x_2(t)=-x_3(t).
689: \label{eq51}
690: \end{equation}
691: These relations between the atomic displacements are shown in
692: Fig.~\ref{fig2} by the appropriate arrows. As a consequence, the atomic
693: displacements, corresponding to the bush B$[\hat{a}^4,\hat{i}]$ with the
694: symmetry group $[\hat{a}^4,\hat{i}]$, can be written (for $N=12$) in the
695: following form:
696: \begin{equation}
697: \vec{X}(t)=\{x_1(t),x_2(t),-x_2(t),-x_1(t)~|~x_1(t),x_2(t),-x_2(t),
698: -x_1(t)~|~x_1(t),x_2(t),-x_2(t),-x_1(t)\}
699: \label{eq149}
700: \end{equation}
701:
702: Since Eqs.~(\ref{eq51}) hold for an arbitrary moment $t$, we will drop the
703: argument $t$ and write down the \textit{vibrational bush in the $X$-space}
704: (configuration space) as follows:
705: \begin{equation}
706: \B[\hat{a}^4,\hat{i}]=|x_1,x_2,-x_2,-x_1|.
707: \label{eq52}
708: \end{equation}
709: Thus, we point out the atomic displacements in only one EPC.
710:
711: 2. For the subgroup
712: \begin{equation}
713: [\hat{a}^4,\hat{a}\hat{i}]\equiv\{\hat{E},\hat{a}^4,\hat{a}^8,\hat{a}\hat{i},
714: \hat{a}^5\hat{i}\equiv\hat{i}\hat{a}^7,\hat{a}^9\hat{i}\equiv\hat{i}\hat{a}^3\},
715: \label{eq53}
716: \end{equation}
717: we obtain the diagram depicted in Fig.~\ref{fig3}. In this diagram, one can
718: see two symmetry elements, $\hat{i}\hat{a}^3$ and $\hat{a}\hat{i}$, which
719: pass, respectively, through the first and third atoms of the represented
720: EPC. Therefore, the displacements of these atoms must satisfy the relations
721: $x_1(t)=-x_1(t)$, $x_3(t)=-x_3(t)$. In turn, this means that $x_1(t)\equiv
722: 0$, $x_3(t)\equiv 0$, i.e.\ the atoms located on the inversion elements
723: (which are equivalent, in the one-dimensional space, to the mirror planes!)
724: must be \textit{immovable}.
725: \begin{figure}
726: \centering
727: \setlength{\unitlength}{1mm}
728: \begin{picture}(100,40)(0,0)
729: \put(0,20){\line(1,0){100}}
730: \put(10,0){\line(0,1){40}}
731: \put(20,5){\line(0,1){30}}
732: \put(20,0){$ia^3$}
733: \put(20,20){\circle*{2}}
734: \put(21,22){$x_1=0$}
735: \put(40,20){\circle*{2}}
736: \put(40,22){$x_2$}
737: \put(60,5){\line(0,1){30}}
738: \put(60,0){$ai$}
739: \put(60,20){\circle*{2}}
740: \put(61,22){$x_3=0$}
741: \put(80,20){\circle*{2}}
742: \put(80,22){$x_4$}
743: \put(90,0){\line(0,1){40}}
744: \put(40,37){\vector(-1,0){30}}
745: \put(60,37){\vector(1,0){30}}
746: \put(43,36){EPC ($4a$)}
747: \linethickness{0.5mm}
748: %\put(40,20){\vector(1,0){10}}
749: \put(40,20){\rvec{10}}
750: %\put(80,20){\vector(-1,0){10}}
751: \put(80,20){\lvec{10}}
752: \end{picture}
753: \caption{\label{fig3} Vibrational bush corresponding to the subgroup
754: $G=[\hat{a}^4,\hat{a}\hat{i}]$.}
755: \end{figure}
756:
757: On the other hand, the symmetry element $\hat{a}\hat{i}$ is situated in the
758: middle between the second and fourth atoms of the EPC and, therefore, their
759: possible displacements must satisfy the equation $x_2(t)=-x_4(t)$. Thus, we
760: have the following displacement pattern for the bush
761: B$[\hat{a}^4,\hat{a}\hat{i}]$
762: \begin{equation}
763: \vec{X}(t)=\{0,x(t),0,-x(t)~|~0,x(t),0,-x(t)~|~0,x(t),0,-x(t)\}
764: \label{eq150a}
765: \end{equation}
766: with \textit{only one} independent dynamical variable $x(t)$. We can write
767: this \textit{one-dimensional} vibrational bush in the abbreviated form:
768: \[\B[\hat{a}^4,\hat{a}\hat{i}]=|0,x,0,-x|.\]
769:
770: 3. The diagram of the symmetry elements of the subgroup
771: \[[\hat{a}^4,\hat{a}^2\hat{i}]\equiv\{\hat{E},\hat{a}^4,\hat{a}^8,
772: \hat{a}^2\hat{i},\hat{a}^6\hat{i}\equiv\hat{i}\hat{a}^6,
773: \hat{a}^{10}\hat{i}\equiv\hat{i}\hat{a}^2\}\]
774: is depicted in Fig.~\ref{fig4}.
775: \begin{figure}
776: \centering
777: \setlength{\unitlength}{1mm}
778: \begin{picture}(100,40)(0,0)
779: \put(0,20){\line(1,0){100}}
780: \put(10,0){\line(0,1){40}}
781: \put(20,20){\circle*{2}}
782: \put(20,22){$x_1$}
783: \put(30,5){\line(0,1){30}}
784: \put(30,0){$ia^2$}
785: \put(40,20){\circle*{2}}
786: \put(40,22){$x_2$}
787: \put(60,20){\circle*{2}}
788: \put(60,22){$x_3$}
789: \put(70,5){\line(0,1){30}}
790: \put(70,0){$a^2i$}
791: \put(80,20){\circle*{2}}
792: \put(80,22){$x_4$}
793: \put(90,0){\line(0,1){40}}
794: \put(40,37){\vector(-1,0){30}}
795: \put(60,37){\vector(1,0){30}}
796: \put(43,36){EPC ($4a$)}
797: \linethickness{0.5mm}
798: %\put(20,20){\vector(1,0){7}}
799: \put(20,20){\rvec{7}}
800: %\put(40,20){\vector(-1,0){7}}
801: \put(40,20){\lvec{7}}
802: %\put(60,20){\vector(1,0){9}}
803: \put(60,20){\rvec{9}}
804: %\put(80,20){\vector(-1,0){9}}
805: \put(80,20){\lvec{9}}
806: \end{picture}
807: \caption{\label{fig4} Vibrational bush corresponding to the subgroup
808: $G=[\hat{a}^4,\hat{a}^2\hat{i}]$.}
809: \end{figure}
810: There are two elements, $\hat{i}\hat{a}^2$ and $\hat{a}^2\hat{i}$ surviving
811: in the transition from the full dihedral group $D=[\hat{a},\hat{i}]$ to the
812: considered subgroup $[\hat{a}^4,\hat{a}^2\hat{i}]$. These elements are
813: located between two first and two last atoms of the depicted EPC.
814: Therefore, $x_1(t)=-x_2(t)$, $x_3(t)=-x_4(t)$ and the corresponding
815: displacement pattern is
816: \begin{equation}
817: \vec{X}(t)=\{x_1(t),-x_1(t),x_2(t),-x_2(t)~|~x_1(t),-x_1(t),x_2(t),
818: -x_2(t)~|~x_1(t),-x_1(t),x_2(t),-x_2(t)\}.
819: \label{eq57}
820: \end{equation}
821:
822: There are two degrees of freedom in~(\ref{eq57}), $x_1(t)$ and $x_2(t)$.
823: Thus, the \textit{two-dimensional} bush with the symmetry group
824: $[\hat{a}^4,\hat{a}^2\hat{i}]$ can be written as
825: \[\B[\hat{a}^4,\hat{a}^2\hat{i}]=|x_1,-x_1,x_2,-x_2|.\]
826:
827: 4. The diagram of the symmetry elements for the last subgroup of the
828: list~(\ref{eq50})
829: \[[\hat{a}^4,\hat{a}^3\hat{i}]\equiv\{\hat{E},\hat{a}^4,\hat{a}^8,
830: \hat{a}^3\hat{i},\hat{a}^7\hat{i}\equiv\hat{i}\hat{a}^5,
831: \hat{a}^{11}\hat{i}\equiv\hat{i}\hat{a}\}\]
832: is depicted in Fig.~\ref{fig5}.
833: \begin{figure}
834: \centering
835: \setlength{\unitlength}{1mm}
836: \begin{picture}(100,40)(0,0)
837: \put(0,20){\line(1,0){100}}
838: \put(10,0){\line(0,1){40}}
839: \put(20,20){\circle*{2}}
840: \put(20,22){$x_1$}
841: \put(40,5){\line(0,1){30}}
842: \put(40,0){$ia$}
843: \put(40,20){\circle*{2}}
844: \put(41,22){$x_2=0$}
845: \put(60,20){\circle*{2}}
846: \put(60,22){$x_3$}
847: \put(80,5){\line(0,1){30}}
848: \put(80,0){$a^3i$}
849: \put(80,20){\circle*{2}}
850: \put(71,22){$x_4=0$}
851: \put(90,0){\line(0,1){40}}
852: \put(40,37){\vector(-1,0){30}}
853: \put(60,37){\vector(1,0){30}}
854: \put(43,36){EPC ($4a$)}
855: \linethickness{0.5mm}
856: %\put(20,20){\vector(1,0){10}}
857: \put(20,20){\rvec{10}}
858: %\put(60,20){\vector(-1,0){10}}
859: \put(60,20){\lvec{10}}
860: \end{picture}
861: \caption{\label{fig5} Vibrational bush corresponding to the subgroup
862: $G=[\hat{a}^4,\hat{a}^3\hat{i}]$.}
863: \end{figure}
864: We see from this diagram that $x_2(t)\equiv 0$, $x_4(t)\equiv 0$,
865: $x_1(t)=-x_3(t)$, and, therefore,
866: \begin{equation}
867: \vec{X}(t)=\{x(t),0,-x(t),0~|~x(t),0,-x(t),0~|~x(t),0,-x(t),0\}.
868: \label{eq58}
869: \end{equation}
870: The corresponding \textit{one-dimensional} bush can be written as follows
871: \[\B[\hat{a}^4,\hat{a}^2\hat{i}]=|x,0,-x,0|.\]
872:
873: We have just considered all the subgroups of the dihedral group $D$, which
874: contain the common translational group $T_4=[\hat{a}^4]$, and the
875: corresponding vibrational bushes. In the same way, we can find the
876: subgroups of the group $D$ corresponding to the arbitrary translational
877: group $T_m=[\hat{a}^m]$\footnote{Such subgroup exists in the case $N\mod
878: m=0.$} and then obtain the bushes induced by these subgroups~\cite{FPU2a}.
879: We will not discuss this method any further in the present paper, because
880: all the bushes of modes (symmetric invariant manifolds) for the dihedral
881: group were recently obtained by Bob Rink in~\cite{BRink}. Nevertheless, let
882: us note that the above described crystallographic method for finding bushes
883: of vibrational modes seems to be more straightforward than that used
884: in~\cite{BRink}. Actually, the former method represents a variant of our
885: general method of the splitting of the space group orbits
886: (SO)~\cite{Comput,Acta}, adapted for the sufficiently simple case of the
887: dihedral group $D$. The SO method was developed for finding the so-called
888: complete condensate of primary and secondary order parameters in the
889: framework of the theory of phase transitions in crystals. It was applied to
890: such nontrivial space groups as $O_h^1$~\cite{Acta},
891: $O_h^7$~\cite{Crystal}, etc.
892:
893: The notion of the complete condensate of order parameters is equivalent to
894: the notion of the bush of modes in the \textit{geometrical} (only
895: geometrical!) sense. Three different methods for finding the condensate of
896: order parameters were considered in~\cite{PhysStatSol,Comput,Acta}: the
897: ``direct method'', which is similar to that by Bob Rink~\cite{BRink}, the
898: above mentioned SO method and the method based on the concept of the
899: irreducible representations of the appropriate symmetry groups. The last
900: method was used, in particular, for treating bushes of vibrational modes in
901: nonlinear chains in the paper~\cite{FPU1}.
902:
903: Now let us consider the notion of the bush equivalence. There are two pairs
904: of equivalent bushes of modes induced by the subgroups from the
905: list~(\ref{eq50}). Indeed, shifting the displacement pattern~(\ref{eq150a})
906: of the bush B$[\hat{a}^4,\hat{a}\hat{i}]$ by the lattice spacing $a$, we
907: obtain the displacement pattern~(\ref{eq58}) of the bush
908: B$[\hat{a}^4,\hat{a}^3\hat{i}]$\footnote{We can imagine that our chain
909: represents a ring, because of the periodic boundary conditions.}. The
910: pattern~(\ref{eq149}) of the bush B$[\hat{a}^4,\hat{i}]$ can be transformed
911: into the pattern~(\ref{eq57}) of the bush B$[\hat{a}^4,\hat{a}^2\hat{i}]$
912: by the same shifting.
913:
914: The importance of the notion of bush equivalence is brought about by the
915: fact that equivalent bushes represent \textit{identical dynamical systems}:
916: they are described by equivalent equations of motions (see below). It can
917: be proved, in the general case, that the \textit{conjugate subgroups} of
918: the parent symmetry group induce equivalent bushes\footnote{Sometimes, we
919: call equivalent bushes by the term ``dynamical domains''. The term
920: ``domain'' is borrowed from the theory of phase transition in crystals.}.
921: Remember that two subgroups $G_1$ and $G_2$ of the same group $G$
922: ($G_1\subset G$, $G_2\subset G$) are called conjugate to each other
923: ($G_1\sim G_2$), if there exists at least one element $g_0$ of the group
924: $G$ which transfers $G_1$ into $G_2$ by the transformation
925: \begin{equation}
926: G_2=g_0^{-1}G_1g_0\quad(g_0\in G).
927: \label{eq60}
928: \end{equation}
929: It can be easy checked the following conjugations:
930: \begin{equation}
931: \left[\hat{a}^4,\hat{i}\right]\sim\left[\hat{a}^4,\hat{a}^2\hat{i}\right],\quad
932: \left[\hat{a}^4,\hat{a}\hat{i}\right]\sim\left[\hat{a}^4,\hat{a}^3\hat{i}\right].
933: \label{eq71a}
934: \end{equation}
935: For example, using Eqs.~(\ref{eq170a}), we obtain the relation
936: \[[\hat{a}^4,\hat{a}\hat{i}]=\hat{a}^{-1}\cdot[\hat{a}^4,\hat{a}^3\hat{i}]\cdot a,\]
937: which proves the equivalence of the subgroups $[\hat{a}^4,\hat{a}\hat{i}]$
938: and $[\hat{a}^4,\hat{a}^3\hat{i}]$, if we take into account the
939: definition~(\ref{eq60}) with $g_0=\hat{a}$. Thus, the both above considered
940: one-dimensional bushes B$[\hat{a}^4,\hat{a}\hat{i}]$ and
941: B$[\hat{a}^4,\hat{a}^3\hat{i}]$ turn out to be equivalent to each other, as
942: well as the both two-dimensional bushes, B$[\hat{a}^4,\hat{i}]$ and
943: B$[\hat{a}^4,\hat{a}^2\hat{i}]$. Moreover, it can be proved that for any
944: even $m$ and $k=0,1,2,\dotsc,m-1$, all bushes of the form
945: B$[\hat{a}^m,\hat{a}^k\hat{i}]$ with $k$ of the \textit{same evenness} are
946: equivalent. For odd $m$, all the bushes of this form are equivalent.
947:
948: Let us summarize the above said about the construction of the vibrational
949: bushes for nonlinear monoatomic chains.
950:
951: -- Each subgroup of the dihedral group singles out a certain bush of
952: vibrational modes.
953:
954: -- Every subgroup is determined by one or two generators and contains the
955: translational subgroup $T_k=[\hat{a}^k]$ (it is trivial, $T_0$, for the
956: subgroups $[\hat{a}^m\hat{i}]$, $m=0,1,2,\dotsc,N-1$).
957:
958: -- The translational subgroup $T_k=[\hat{a}^k]$ determines the extended
959: primitive cell (EPC) for the vibrational state of the chain.
960:
961: -- The symmetry elements $\hat{a}^m\hat{i}$, belonging to the second coset
962: of the dihedral group (see Eq.~(\ref{eq500})) and surviving at the symmetry
963: lowering $D\equiv[\hat{a},\hat{i}]\rightarrow[\hat{a}^k,\hat{a}^m\hat{i}]$,
964: produce certain restrictions on the displacements inside the EPC.
965:
966: -- The conjugate subgroups induce the equivalent bushes of modes. The
967: number of the equivalent bushes of a given type is equal to the
968: \textit{index}\footnote{The index of the subgroup $G$ of the group $G_0$ is
969: equal to the ratio $\|G_0\|/\|G\|$, where $\|G_0\|$ and $\|G\|$ are the
970: orders of these two groups, $G_0$ and $G$.} of the corresponding subgroup
971: in the dihedral group.
972:
973: In addition, note that there exist subgroups to which no vibrational bushes
974: correspond. In our case, the subgroup $[\hat{a}^2,\hat{a}\hat{i}]$ does not
975: induce the vibrational bush, because the EPC, corresponding to this
976: subgroup, contains only two atoms through which the inversion elements,
977: $\hat{i}\hat{a}$ and $\hat{a}\hat{i}$, pass and, therefore, these atoms
978: cannot move.
979:
980: The bushes obtained with the aid of the described method represent
981: invariant manifolds determined in the \textit{configuration} space. Let us
982: now consider the bushes in the \textit{modal} space.
983:
984: We consider the normal modes for the monoatomic chain in the form used
985: in~\cite{PoggiRuffo}:
986: \begin{equation}
987: \vec{\psi}_k=\left\{\frac{1}{\sqrt{N}}\left.\left[\sin\left(\frac{2\pi
988: k}{N}n\right) +\cos\left(\frac{2\pi
989: k}{N}n\right)\right]\right|n=1,2,\dotsc,N\right\}.
990: \label{eq70}
991: \end{equation}
992: Here the subscript $k$ refers to the mode, while the subscript $n$ refers
993: to the atom. The modes $\vec{\psi}_k$ ($k=0,1,2,\dotsc,N-1$) form an
994: orthonormal basis in the modal space and we can decompose the set of the
995: atomic displacements $\vec{X}(t)$, corresponding to a given bush, in this
996: basis
997: \begin{equation}
998: \vec{X}(t)=\sum_{k=0}^{N-1}\nu_k(t)\vec{\psi}_k
999: \label{eq71}
1000: \end{equation}
1001: For example, we obtain the following decompositions for the bushes
1002: B$[\hat{a}^4,\hat{i}]$ and B$[\hat{a}^4,\hat{a}^2\hat{i}]$:
1003: \begin{eqnarray}
1004: \vec{X}(\B[\hat{a}^4,\hat{i}])&\equiv&\{x_1(t),x_2(t),-x_1(t),-x_2(t)|%
1005: x_1(t),x_2(t),-x_1(t),-x_2(t)|x_1(t),x_2(t),-x_1(t),-x_2(t)\}\nonumber\\
1006: &=&\mu(t)\vec{\psi}_{N/2}+\nu(t)\vec{\psi}_{3N/4},
1007: \label{eq72}\\
1008: \vec{X}(\B[\hat{a}^4,\hat{a}^2\hat{i}])&=&\tilde{\mu}(t)\vec{\psi}_{N/2}+\tilde{\nu}(t)\vec{\psi}_{N/4}.
1009: \label{eq73}
1010: \end{eqnarray}
1011: Only vectors $\vec{\psi}_{N/2}$, $\vec{\psi}_{N/4}$ and $\vec{\psi}_{3N/4}$
1012: from the complete basis~(\ref{eq70}) contribute to these two-dimensional
1013: bushes:
1014: \begin{eqnarray}
1015: \vec{\psi}_{N/2}&=&\frac{1}{\sqrt{N}}(-1,1,-1,1,-1,1,-1,1,-1,1,-1,1),
1016: \label{eq74}\\
1017: \vec{\psi}_{N/4}&=&\frac{1}{\sqrt{N}}(1,-1,-1,1,1,-1,-1,1,1,-1,-1,1),
1018: \label{eq74a}\\
1019: \vec{\psi}_{3N/4}&=&\frac{1}{\sqrt{N}}(-1,-1,1,1,-1,-1,1,1,-1,-1,1,1).
1020: \label{eq74b}
1021: \end{eqnarray}
1022: Remember that these bushes are equivalent to each other, i.e.\ they
1023: represent the so-called ``dynamical domains''. For the bush
1024: B$[\hat{a}^4,\hat{i}]$ we can find the following relations between the old
1025: dynamical variables $x_1(t)$, $x_2(t)$ (relating to the configuration
1026: space) and the new dynamical variables $\mu(t)$, $\nu(t)$ (relating to the
1027: modal space):
1028: \begin{equation}
1029: \begin{array}{l}
1030: \mu(t)=-\frac{\sqrt{N}}{2}[x_1(t)-x_2(t)],\\
1031: \nu(t)=-\frac{\sqrt{N}}{2}[x_1(t)+x_2(t)].
1032: \end{array}
1033: \label{eq75}
1034: \end{equation}
1035: Thus, each of the above bushes consists of two modes. One of these modes is
1036: the \textit{root} mode ($\vec{\psi}_{3N/4}$ for the bush
1037: B$[\hat{a}^4,\hat{i}]$ and $\vec{\psi}_{N/4}$ for the bush
1038: B$[\hat{a}^4,\hat{a}^2\hat{i}]$), while the other mode $\vec{\psi}_{N/2}$
1039: is the secondary mode. Indeed, we have already discussed (see Introduction)
1040: that the symmetry of the secondary modes must be higher or equal to the
1041: symmetry of the root mode. In our case, as it can be seen from
1042: Eqs.~(\ref{eq74a}), the translational symmetry of the mode
1043: $\vec{\psi}_{N/4}$ is $\hat{a}^4$ (acting by this element on~(\ref{eq74a})
1044: we obtain the same displacement pattern), while the translational symmetry
1045: of the mode $\vec{\psi}_{N/2}$ is $\hat{a}^2$, which is two times higher
1046: than that of $\vec{\psi}_{N/4}$
1047: (see~(\ref{eq74})-(\ref{eq74b}))\footnote{Note that the full symmetry of
1048: the modes $\vec{\psi}_{N/4}$ and $\vec{\psi}_{N/2}$ are
1049: $[\hat{a}^4,\hat{a}^2\hat{i}]$ and $[\hat{a}^2,\hat{i}]$, respectively.}.
1050:
1051: One-dimensional (1D) and two-dimensional (2D) vibrational bushes for the
1052: FPU chains in configuration and modal spaces can be seen in
1053: Tables~\ref{table1},\ref{table2}. Note that in these tables, as well as in
1054: Tables~\ref{table25},\ref{table3},\ref{table4} we list one-dimensional
1055: bushes completely, but two-dimensional bushes partially: only those 2D
1056: bushes are given whose stability properties are studied in the present
1057: paper (see Sec.~6). The complete list of 2D bushes in the FPU chains can be
1058: found in Appendix.
1059:
1060: \begin{table}
1061: \centering
1062: \caption{\label{table1} Representation of 1D and 2D bushes of vibrational
1063: modes for the FPU chains in the configuration space (classification by the
1064: dihedral group~$D$)}
1065: \begin{tabular}{|l|c|c|}
1066: \hline
1067: Bush&Displacement pattern&Dimension\\
1068: \hline
1069: B$[a^2,i]$&$|x,-x|$&1\\
1070: \hline
1071: B$[a^3]$&$|x_1,x_2,x_3|^*$&2\\
1072: \hline
1073: B$[a^3,i]$&$|x,0,-x|$&1\\
1074: B$[a^3,ai]$&$|0,x,-x|$&1\\
1075: B$[a^3,a^2i]$&$|x,-x,0|$&1\\
1076: \hline
1077: B$[a^4,i]$&$|x_1,x_2,-x_2,-x_1|$&2\\
1078: B$[a^4,a^2i]$&$|x_1,-x_1,x_2,-x_2|$&2\\
1079: \hline
1080: B$[a^4,ai]$&$|0,x,0,-x|$&1\\
1081: B$[a^4,a^3i]$&$|x,0,-x,0|$&1\\
1082: \hline
1083: B$[a^6,ai]$&$|0,x_1,x_2,0,-x_2,-x_1|$&2\\
1084: B$[a^6,a^3i]$&$|x_1,0,-x_1,x_2,0,-x_2|$&2\\
1085: B$[a^6,a^5i]$&$|x_1,x_2,0,-x_2,-x_1,0|$&2\\
1086: \hline
1087: \end{tabular}
1088:
1089: \raggedright
1090: \small{$^*$ Here $x_3$ is not an independent variable: its value must be
1091: determined by the condition of \textit{immobility} of the mass center of
1092: the chain, $x_1+x_2+x_3=0$.}
1093: \end{table}
1094:
1095: \begin{table}
1096: \centering
1097: \caption{\label{table2} Representation of 1D and 2D bushes of vibrational
1098: modes for the FPU chains in the modal space (classification by the dihedral
1099: group~$D$). Here
1100: $\varepsilon_1=\sin\left(\frac{\pi}{12}\right)=\frac{\sqrt{2}(\sqrt{3}-1)}{4}$,
1101: $\varepsilon_2=\cos\left(\frac{\pi}{12}\right)=\frac{\sqrt{2}(\sqrt{3}+1)}{4}$,
1102: $\varepsilon_3=\frac{\sqrt{2}}{2}$}
1103: \begin{tabular}{|l|c|}
1104: \hline
1105: Bush&Representation in modal space\\
1106: \hline
1107: B$[a^2,i]$&$\nu\vec{\psi}_{N/2}$\\
1108: \hline
1109: B$[a^3,i]$&$\nu(\varepsilon_1\vec{\psi}_{N/3}+\varepsilon_2\vec{\psi}_{2N/3})$\\
1110: B$[a^3,ai]$&$\nu(\varepsilon_2\vec{\psi}_{N/3}+\varepsilon_1\vec{\psi}_{2N/3})$\\
1111: B$[a^3,a^2i]$&$\nu\varepsilon_3(\vec{\psi}_{N/3}-\vec{\psi}_{2N/3})$\\
1112: \hline
1113: B$[a^4,i]$&$\nu_1\vec{\psi}_{3N/4}-\nu_2\vec{\psi}_{N/2}$\\
1114: B$[a^4,a^2i]$&$\nu_1\vec{\psi}_{N/4}+\nu_2\vec{\psi}_{N/2}$\\
1115: \hline
1116: B$[a^4,ai]$&$\nu\varepsilon_3(\vec{\psi}_{N/4}+\vec{\psi}_{3N/4})$\\
1117: B$[a^4,a^3i]$&$\nu\varepsilon_3(\vec{\psi}_{N/4}-\vec{\psi}_{3N/4})$\\
1118: \hline
1119: B$[a^6,ai]$&$\nu_1(\varepsilon_1\vec{\psi}_{N/6}+\varepsilon_2\vec{\psi}_{5N/6})+\nu_2(-\varepsilon_2\vec{\psi}_{N/3}-\varepsilon_1\vec{\psi}_{2N/3})$\\
1120: B$[a^6,a^3i]$&$\nu_1(\varepsilon_2\vec{\psi}_{N/6}+\varepsilon_1\vec{\psi}_{5N/6})+\nu_2(\varepsilon_1\vec{\psi}_{N/3}+\varepsilon_2\vec{\psi}_{2N/3})$\\
1121: B$[a^6,a^5i]$&$\nu_1\varepsilon_3(\vec{\psi}_{N/6}-\vec{\psi}_{5N/6})+\nu_2\varepsilon_3(\vec{\psi}_{N/3}-\vec{\psi}_{2N/3})$\\
1122: \hline
1123: \end{tabular}
1124: \end{table}
1125:
1126: \begin{table}
1127: \centering
1128: \caption{\label{table25} Additional 1D and 2D bushes, for the FPU-$\beta$
1129: chain (these bushes were obtained in~\cite{BRink})}
1130: \begin{tabular}{|l|c|c|c|}
1131: \hline
1132: ~~Bush&Dim.&Displacement pattern&Representation in the modal space\\
1133: \hline
1134: \begin{tabular}{l}B$[a^3,iu]$\\B$[a^3,aiu]$\\B$[a^3,a^2iu]$\end{tabular} &
1135: 1 &
1136: \begin{tabular}{l}$|x,-2x,x|$\\$|-2x,x,x|$\\$|x,x,-2x|$\end{tabular} &
1137: \begin{tabular}{l}$\nu(\varepsilon_2\vec{\psi}_{N/3}-\varepsilon_1\vec{\psi}_{2N/3})$\\
1138: $\nu(\varepsilon_1\vec{\psi}_{N/3}-\varepsilon_2\vec{\psi}_{2N/3})$\\
1139: $\nu\varepsilon_3(\vec{\psi}_{N/3}+\vec{\psi}_{2N/3})$\end{tabular}\\
1140: \hline
1141: \begin{tabular}{l}B$[a^4,iu]$\\B$[a^4,a^2iu]$\end{tabular} &
1142: 1 &
1143: \begin{tabular}{l}$|x,-x,-x,x|$\\$|x,x,-x,-x|$\end{tabular} &
1144: \begin{tabular}{l}$\nu\vec{\psi}_{N/4}$\\$\nu\vec{\psi}_{3N/4}$\end{tabular}\\
1145: \hline
1146: \begin{tabular}{l}B$[a^6,ai,a^3u]$\\B$[a^6,a^3i,a^3u]$\\B$[a^6,a^5i,a^3u]$\end{tabular} &
1147: 1 &
1148: \begin{tabular}{l}$|0,x,x,0,-x,-x|$\\$|x,0,-x,-x,0,x|$\\$|x,x,0,-x,-x,0|$\end{tabular} &
1149: \begin{tabular}{l}$\nu(\varepsilon_1\vec{\psi}_{N/6}+\varepsilon_2\vec{\psi}_{5N/6})$\\
1150: $\nu(\varepsilon_2\vec{\psi}_{N/6}+\varepsilon_1\vec{\psi}_{5N/6})$\\
1151: $\nu\varepsilon_3(\vec{\psi}_{N/6}-\vec{\psi}_{5N/6})$\end{tabular}\\
1152: \hline
1153: \begin{tabular}{l}B$[a^4,aiu]$\\B$[a^4,a^3iu]$\end{tabular} &
1154: 2 &
1155: \begin{tabular}{l}$|x_1,x_2,x_3,x_2|^{**}$\\$|x_2,x_1,x_2,x_3|^{**}$\end{tabular} &
1156: \begin{tabular}{l}$\nu_1\varepsilon_3(\vec{\psi}_{N/4}-\vec{\psi}_{3N/4})+\nu_2\vec{\psi}_{N/2}$\\
1157: $\nu_1\varepsilon_3(\vec{\psi}_{N/4}+\vec{\psi}_{3N/4})+\nu_2\vec{\psi}_{N/2}$\end{tabular}\\
1158: \hline
1159: \begin{tabular}{l}B$[a^4,a^2u]$\end{tabular} &
1160: 2 &
1161: \begin{tabular}{l}$|x_1,x_2,-x_1,-x_2|$\end{tabular} &
1162: \begin{tabular}{l}$\nu_1\vec{\psi}_{N/4}+\nu_2\vec{\psi}_{3N/4}$\end{tabular}\\
1163: \hline
1164: \begin{tabular}{l}B$[a^6,iu]$\\B$[a^6,a^2iu]$\\B$[a^6,a^4iu]$\end{tabular} &
1165: 2 &
1166: \begin{tabular}{l}$|x_1,x_2,x_3,x_3,x_2,x_1|^{**}$\\$|x_1,x_1,x_2,x_3,x_3,x_2|^{**}$\\$|x_1,x_2,x_2,x_1,x_3,x_3|^{**}$\end{tabular} &
1167: \begin{tabular}{l}$\nu_1(\varepsilon_2\vec{\psi}_{N/6}+\varepsilon_1\vec{\psi}_{5N/6})+\nu_2(\varepsilon_2\vec{\psi}_{N/3}-\varepsilon_1\vec{\psi}_{2N/3})$\\
1168: $\nu_1\varepsilon_3(\vec{\psi}_{N/6}-\vec{\psi}_{5N/6})+\nu_2\varepsilon_3(\vec{\psi}_{N/3}+\vec{\psi}_{2N/3})$\\
1169: $\nu_1(\varepsilon_1\vec{\psi}_{N/6}+\varepsilon_2\vec{\psi}_{5N/6})+\nu_2(\varepsilon_1\vec{\psi}_{N/3}-\varepsilon_2\vec{\psi}_{2N/3})$\end{tabular}\\
1170: \hline
1171: \begin{tabular}{l}B$[a^6,i,a^3u]$\\B$[a^6,a^2i,a^3u]$\\B$[a^6,a^4i,a^3u]$\end{tabular} &
1172: 2 &
1173: \begin{tabular}{l}$|x_1,x_2,x_1,-x_1,-x_2,-x_1|$\\$|x_1,-x_1,x_2,-x_1,x_1,-x_2|$\\$|x_1,x_2,-x_2,-x_1,-x_2,x_2|$\end{tabular} &
1174: \begin{tabular}{l}$\nu_1(-\varepsilon_1\vec{\psi}_{N/6}+\varepsilon_2\vec{\psi}_{5N/6})+\nu_2\vec{\psi}_{N/2}$\\
1175: $\nu_1\varepsilon_3(\vec{\psi}_{N/6}+\vec{\psi}_{5N/6})-\nu_2\vec{\psi}_{N/2}$\\
1176: $\nu_1(\varepsilon_2\vec{\psi}_{N/6}-\varepsilon_1\vec{\psi}_{5N/6})+\nu_2\vec{\psi}_{N/2}$\end{tabular}\\
1177: \hline
1178: \end{tabular}
1179:
1180: \raggedright
1181: \small{$^{**}$ See footnote to Table~\ref{table1}.}
1182: \end{table}
1183:
1184:
1185:
1186: \section{Dynamical description of the bushes of modes}
1187:
1188: Up to this point, we've been considering bushes as geometrical objects. In
1189: the present section, we address their dynamical aspect. For this purpose,
1190: one would need to obtain the equations of motion for the appropriate
1191: dynamical variables (in the configuration space or in the modal space). It
1192: can be done with the aid of the Lagrange method. This method is very useful
1193: for the mechanical systems with some constraints on the natural dynamical
1194: variables (for example, on the displacements of the individual particles).
1195: Note that we deal now with precisely this case, and the additional
1196: constraints on the dynamical variables brought about by the
1197: \textit{symmetry} causes. Indeed, the displacement pattern $\vec{X}(t)$,
1198: corresponding to a given bush B$[G]$, is determined by the condition of its
1199: invariance under the action of the bush symmetry group $G$:
1200: $\hat{G}\vec{X}(t)=\vec{X}(t)$. Namely this condition permits us to
1201: introduce the generalized coordinates which represent dynamical variables
1202: of the reduced Hamiltonian system associated with the considered bush of
1203: modes.
1204:
1205: We have already used the Lagrange method for the derivation of the bush
1206: dynamical equations in the paper~\cite{Okta} devoted to nonlinear
1207: vibrations of the octahedral mechanical systems with Lennard-Jones
1208: potential. Now we apply this method to the FPU chains.
1209:
1210: The Hamiltonian of these nonlinear chains can be written as follows:
1211: \begin{equation}
1212: H=T+V=\frac{1}{2}\sum_{n=1}^N \dot{x}_n^2+\frac{1}{2}\sum_{n=1}^N
1213: (x_{n+1}-x_n)^2 +\frac{\gamma}{p}\sum_{n=1}^N (x_{n+1}-x_n)^p.
1214: \label{eq80}
1215: \end{equation}
1216: Here $p=3$, $\gamma=\alpha$ for the FPU-$\alpha$ model, and $p=4$,
1217: $\gamma=\beta$ for the FPU-$\beta$ model\footnote{The constants $\alpha$
1218: and $\beta$ can be removed from the Hamiltonian~(\ref{eq80}) by a trivial
1219: scaling, but we keep them for the clarity of consideration.}. $T$ and $V$
1220: are the kinetic and potential energy, respectively. We also assume the
1221: periodic boundary conditions~(\ref{eq1}) to be valid. Let us consider the
1222: set of atomic displacements corresponding to the two-dimensional bush
1223: B$[\hat{a}^4,\hat{i}]$
1224: \begin{equation}
1225: \vec{X}(t)=\{x,y,-y,-x|x,y,-y,-x|x,y,-y,-x|\cdots\},
1226: \label{eq81}
1227: \end{equation}
1228: where we rename $x_1(t)$ and $x_2(t)$ from Eq.~(\ref{eq149}) as $x(t)$ and
1229: $y(t)$, respectively. Substituting the atomic displacements from
1230: Eq.~(\ref{eq81}) into the Hamiltonian~(\ref{eq80}) corresponding to the
1231: FPU-$\alpha$ chain we obtain the following expressions for the kinetic
1232: energy $T$ and the potential energy $V$:
1233: \begin{equation}
1234: T=\frac{N}{4}(\dot{x}^2+\dot{y}^2),
1235: \label{eq82}
1236: \end{equation}
1237: \begin{equation}
1238: V=\frac{N}{4}(3x^2-2xy+3y^2)+\frac{N\alpha}{2}(x^3+x^2y-xy^2-y^3).
1239: \label{eq83}
1240: \end{equation}
1241: These equations are valid for the arbitrary FPU-$\alpha$ chain for which
1242: $N\mod 4=0$. The size of the extended primitive cell for the vibrational
1243: state (\ref{eq81}) is equal to $4a$ and, therefore, we can restrict
1244: ourselves calculating the energies $T$ and $V$, by summing over only one
1245: EPC.
1246:
1247: The Lagrange equations can be written as
1248: \begin{equation}
1249: \frac{d}{dt}\left(\frac{\partial
1250: L}{\partial\dot{x}_j}\right)-\frac{\partial L}{\partial x_j}=0,
1251: \label{eq84}
1252: \end{equation}
1253: where $L=T-V$. The subscript $j$ is used here for numbering the dynamical
1254: variables (in our case, $x_1(t)=x(t)$, $x_2(t)=y(t)$). Using
1255: Eqs.~(\ref{eq82}) and~(\ref{eq83}), we obtain from~(\ref{eq84}) the
1256: following dynamical equations for the considered bush
1257: B$[\hat{a}^4,\hat{i}]$:
1258: \begin{equation}
1259: \left\{\begin{array}{lll}
1260: \ddot{x}+(3x-y)+\alpha(3x^2+2xy-y^2)&=&0,\\
1261: \ddot{y}+(3y-x)+\alpha(x^2-2xy-3y^2)&=&0.
1262: \end{array}\right.
1263: \label{eq85}
1264: \end{equation}
1265: Let us emphasize that these equations do not depend on the number $N$ of
1266: the particles in the chain (but the relation $N\mod 4=0$ must hold).
1267:
1268: Eqs.~(\ref{eq85}) are written in terms of the atomic displacements $x(t)$
1269: and $y(t)$. From these equations, it is easy to obtain the dynamical
1270: equations for the bush in terms of the normal modes $\mu(t)$ and $\nu(t)$.
1271: Using the relations~(\ref{eq75}) between the old and new variables, we find
1272: the following equations for the bush B$[\hat{a}^4,\hat{i}]$ in the modal
1273: space
1274: \begin{eqnarray}
1275: \ddot{\mu}+4\mu-\frac{4\alpha}{\sqrt{N}}\nu^2=0,
1276: \label{eq86}\\
1277: \ddot{\nu}+2\nu-\frac{8\alpha}{\sqrt{N}}\mu\nu=0.
1278: \label{eq86b}
1279: \end{eqnarray}
1280: The Hamiltonian for the bush B$[\hat{a}^4,\hat{i}]$, considered as a
1281: two-dimensional dynamical system, can be written in the modal space as
1282: follows:
1283: \begin{equation}
1284: H[\hat{a}^4,\hat{i}]=\frac{1}{2}(\dot{\mu}^2+\dot{\nu}^2)+(2\mu^2+\nu^2)-%
1285: \frac{4\alpha}{\sqrt{N}}\mu\nu^2.
1286: \label{eq87}
1287: \end{equation}
1288:
1289: We already emphasized that identical dynamical equations correspond to
1290: equivalent bushes of modes. Moreover, the dynamical equations of the given
1291: \textit{type} are often associated with many bushes of very
1292: \textit{different} symmetry groups (these equations differ from each other
1293: by numerical values of their pertinent parameters only). Because of this
1294: reason, we can say that such bushes belong to the same class of ``dynamical
1295: universality''~\cite{DAN1,DAN2}. As an example, let us point out that all
1296: one-dimensional bushes for the FPU-$\beta$ models belong to one and the
1297: same class of dynamical universality. Indeed, the dynamics of all these
1298: bushes is described by the Duffing equation $\ddot{\mu}+A\mu+B\mu^3=0$, but
1299: with different values of the constants $A$ and $B$.
1300:
1301: Dynamical equations for all one-dimensional and two-dimensional bushes of
1302: vibrational modes for the FPU-chains are presented in
1303: Tables~\ref{table3},\ref{table4}. In the first column, we present the
1304: complete sets of equivalent bushes, to which the same dynamical equations
1305: (see column 2 and 3) correspond.
1306:
1307: Table~\ref{table4} describes the ``additional'' bushes\footnote{They are
1308: additional with respect to the bushes of normal modes associated with the
1309: dihedral group.}, which were found in~\cite{BRink}. They are brought about
1310: by the evenness of the potential of the FPU-$\beta$ model and the
1311: additional symmetry operator $\hat{u}$ is used in the description of these
1312: bushes. This operator changes the signs of all atomic displacements:
1313: $\hat{u}\vec{X}(t)=-\vec{X}(t)$. Note that the operator $\hat{u}$ is
1314: contained in the description of the bushes only in combinations with some
1315: other symmetry operators, $\hat{i}$, $\hat{a}^k$ or their products.
1316:
1317:
1318: \begin{table}
1319: \centering
1320: \caption{\label{table3} Dynamical equations for 1D and 2D bushes in the FPU
1321: chains, induced by the subgroups of the dihedral group}
1322: \begin{tabular}{|l|c|c|}
1323: \hline
1324: ~~Bush&Equations for the FPU-$\alpha$ chain&Equations for the FPU-$\beta$ chain\\
1325: \hline
1326: \begin{tabular}{l}B$[a^2,i]$\end{tabular} &
1327: $\ddot{\nu}+4\nu=0$ &
1328: $\ddot{\nu}+4\nu=-16\frac{\beta}{N}\nu^3$\\
1329: \hline
1330: \begin{tabular}{l}B$[a^3,i]$\\B$[a^3,ai]$\\B$[a^3,a^2i]$\end{tabular} &
1331: $\ddot{\nu}+3\nu=\frac{3\sqrt{6}}{2}\frac{\alpha}{\sqrt{N}}\nu^2$&
1332: $\ddot{\nu}+3\nu=-\frac{27}{2}\frac{\beta}{N}\nu^3$\\
1333: \hline
1334: \begin{tabular}{l}B$[a^4,ai]$\\B$[a^4,a^3i]$\end{tabular} &
1335: $\ddot{\nu}+2\nu=0$ &
1336: $\ddot{\nu}+2\nu=-4\frac{\beta}{N}\nu^3$\\
1337: \hline
1338: \begin{tabular}{l}B$[a^3]$\end{tabular} &
1339: $\begin{array}{l}
1340: \ddot{\nu}_1+3\nu_1=\frac{3\sqrt{3}}{2}\frac{\alpha}{\sqrt{N}}(-\nu_1^2-2\nu_1\nu_2+\nu_2^2)\\
1341: \ddot{\nu}_2+3\nu_2=\frac{3\sqrt{3}}{2}\frac{\alpha}{\sqrt{N}}(-\nu_1^2+2\nu_1\nu_2+\nu_2^2)
1342: \end{array}$ &
1343: $\begin{array}{l}
1344: \ddot{\nu}_1+3\nu_1=-\frac{27}{2}\frac{\beta}{N}\nu_1(\nu_1^2+\nu_2^2)\\
1345: \ddot{\nu}_2+3\nu_2=-\frac{27}{2}\frac{\beta}{N}\nu_2(\nu_1^2+\nu_2^2)
1346: \end{array}$\\
1347: \hline
1348: \begin{tabular}{l}B$[a^4,i]$\\B$[a^4,a^2i]$\end{tabular} &
1349: $\begin{array}{l}
1350: \ddot{\nu}_1+2\nu_1=-8\frac{\alpha}{\sqrt{N}}\nu_1\nu_2\\
1351: \ddot{\nu}_2+4\nu_2=-4\frac{\alpha}{\sqrt{N}}\nu_1^2
1352: \end{array}$ &
1353: $\begin{array}{l}
1354: \ddot{\nu}_1+2\nu_1=-8\frac{\beta}{N}\nu_1(\nu_1^2+3\nu_2^2)\\
1355: \ddot{\nu}_2+4\nu_2=-8\frac{\beta}{N}\nu_2(3\nu_1^2+2\nu_2^2)
1356: \end{array}$\\
1357: \hline
1358: \begin{tabular}{l}B$[a^6,ai]$\\B$[a^6,a^3i]$\\B$[a^6,a^5i]$\end{tabular} &
1359: $\begin{array}{l}
1360: \ddot{\nu}_1+\nu_1=-\sqrt{6}\frac{\alpha}{\sqrt{N}}\nu_1\nu_2\\
1361: \ddot{\nu}_2+3\nu_2=\frac{\sqrt{6}}{2}\frac{\alpha}{\sqrt{N}}(-\nu_1^2+3\nu_2^2)
1362: \end{array}$ &
1363: $\begin{array}{l}
1364: \ddot{\nu}_1+\nu_1=-\frac{3}{2}\frac{\beta}{N}\nu_1(\nu_1^2+3\nu_2^2)\\
1365: \ddot{\nu}_2+3\nu_2=-\frac{9}{2}\frac{\beta}{N}\nu_2(\nu_1^2+3\nu_2^2)
1366: \end{array}$\\
1367: \hline
1368: \end{tabular}
1369: \end{table}
1370:
1371: \begin{table}
1372: \centering
1373: \caption{\label{table4} Dynamical equations for the additional 1D and 2D
1374: bushes, for the FPU-$\beta$ chain}
1375: \begin{tabular}{|l|c|}
1376: \hline
1377: ~~Bush&Equations\\
1378: \hline
1379: \begin{tabular}{l}B$[a^3,iu]$\\B$[a^3,aiu]$\\B$[a^3,a^2iu]$\end{tabular} &
1380: $\ddot{\nu}+3\nu=-\frac{27}{2}\frac{\beta}{N}\nu^3$\\
1381: \hline
1382: \begin{tabular}{l}B$[a^4,iu]$\\B$[a^4,a^2iu]$\end{tabular} &
1383: $\ddot{\nu}+2\nu=-8\frac{\beta}{N}\nu^3$\\
1384: \hline
1385: \begin{tabular}{l}B$[a^6,ai,a^3u]$\\B$[a^6,a^3i,a^3u]$\\B$[a^6,a^5i,a^3u]$\end{tabular} &
1386: $\ddot{\nu}+\nu=-\frac{3}{2}\frac{\beta}{N}\nu^3$\\
1387: \hline
1388: \begin{tabular}{l}B$[a^4,aiu]$\\B$[a^4,a^3iu]$\end{tabular} &
1389: $\begin{array}{l}
1390: \ddot{\nu}_1+2\nu_1=-4\frac{\beta}{N}\nu_1(\nu_1^2+6\nu_2^2)\\
1391: \ddot{\nu}_2+4\nu_2=-8\frac{\beta}{N}\nu_2(3\nu_1^2+2\nu_2^2)
1392: \end{array}$\\
1393: \hline
1394: \begin{tabular}{l}B$[a^4,a^2u]$\end{tabular} &
1395: $\begin{array}{l}
1396: \ddot{\nu}_1+2\nu_1=-8\frac{\beta}{N}\nu_1^3\\
1397: \ddot{\nu}_2+4\nu_2=-8\frac{\beta}{N}\nu_2^3
1398: \end{array}$\\
1399: \hline
1400: \begin{tabular}{l}B$[a^6,iu]$\\B$[a^6,a^2iu]$\\B$[a^6,a^4iu]$\end{tabular} &
1401: $\begin{array}{l}
1402: \ddot{\nu}_1+\nu_1=-\frac{3}{2}\frac{\beta}{N}\nu_1(\nu_1^2+9\nu_2^2)\\
1403: \ddot{\nu}_2+3\nu_2=-\frac{27}{2}\frac{\beta}{N}\nu_2(\nu_1^2+\nu_2^2)
1404: \end{array}$\\
1405: \hline
1406: \begin{tabular}{l}B$[a^6,i,a^3u]$\\B$[a^6,a^2i,a^3u]$\\B$[a^6,a^4i,a^3u]$\end{tabular} &
1407: $\begin{array}{l}
1408: \ddot{\nu}_1+\nu_1=-\frac{\beta}{N}\nu_1(\frac{3}{2}\nu_1^2+3\sqrt{2}\nu_1\nu_2+12\nu_2^2)\\
1409: \ddot{\nu}_2+4\nu_2=-\frac{\beta}{N}(\sqrt{2}\nu_1^3+12\nu_1^2\nu_2+16\nu_2^3)
1410: \end{array}$\\
1411: \hline
1412: \end{tabular}
1413: \end{table}
1414:
1415:
1416:
1417: \section{Stability of bushes of normal modes}
1418:
1419: The term ``stability'' is often used in various senses. Let us explain what
1420: we mean by this term in the present paper.
1421:
1422: In general case, stability of the bushes of modes was discussed
1423: in~\cite{DAN1,PhysD}, while their stability in the FPU-$\alpha$ chain is
1424: considered in~\cite{FPU1}. As well as in these papers, we will discuss here
1425: the stability of a given bush of normal modes with respect of its
1426: interactions\footnote{There is an essential difference between the
1427: interactions of the modes which belong and which do not belong to a given
1428: bush: we speak about ``force interaction'' in the former case and about
1429: `parametric interaction'' in the last case~\cite{PhysD}.} with the modes
1430: which \textit{do not belong} to this bush. Let us illustrate this idea with
1431: the following example.
1432:
1433: For the FPU-$\alpha$ chain, the two-dimensional bush B$[\hat{a}^4,\hat{i}]$
1434: is described by Eqs.~(\ref{eq86}),(\ref{eq86b}). These equations allow a
1435: solution of the form
1436: \begin{equation}
1437: \mu(t)\neq 0,\quad \nu(t)\equiv 0.
1438: \label{eq90}
1439: \end{equation}
1440: (The vibrational regime of this type can be excited by imposing the
1441: appropriate initial conditions: $\mu(t_0)=\mu_0\neq 0$, $\dot{\mu}(t_0)=0$,
1442: $\nu(t_0)=0$, $\dot{\nu}(t_0)=0$). Substitution of the
1443: solution~(\ref{eq90}) into~(\ref{eq86}) produces the dynamical equation of
1444: the one-dimensional bush B$[\hat{a}^2,\hat{i}]$ (see Table~\ref{table4})
1445: consisting of only one mode $\mu(t)$
1446: \begin{equation}
1447: \ddot{\mu}+4\mu=0
1448: \label{eq92}
1449: \end{equation}
1450: with the trivial solution\footnote{We can consider the initial phase in the
1451: solution~(\ref{eq93}) to be equal to zero~\cite{FPU1}.}
1452: \begin{equation}
1453: \mu(t)=\mu_0\cos(2t).
1454: \label{eq93}
1455: \end{equation}
1456: Substituting~(\ref{eq93}) into Eq.~(\ref{eq86b}), we obtain
1457: \begin{equation}
1458: \ddot{\nu}+\left[2-\frac{8\alpha\mu_0}{\sqrt{N}}\cos(2t)\right]\nu=0.
1459: \label{eq94}
1460: \end{equation}
1461: This equation can be easily transformed into the standard form of the
1462: Mathieu equation
1463: \begin{equation}
1464: \ddot{\nu}+[a-2q\cos(2t)]\nu=0.
1465: \label{eq95}
1466: \end{equation}
1467:
1468: On the other hand, there exist domains of stable and unstable motion of the
1469: Mathieu equation~(\ref{eq95}) in the $a-q$ plane of its intrinsic
1470: parameters, $a$ and $q$. The one-dimensional bush B$[\hat{a}^2,\hat{i}]$ is
1471: stable for the sufficiently small amplitudes $\mu_0$ of the mode $\mu(t)$,
1472: but it becomes unstable as the result of the increasing of this amplitude.
1473: This phenomenon, similar to the well-known parametric resonance, takes
1474: place for those $\mu_0$ which get into the domains of unstable motion of
1475: the Mathieu-type equation~(\ref{eq94}). The loss of stability of the
1476: dynamical regime~(\ref{eq90}) (the bush B$[\hat{a}^2,\hat{i}]$) manifests
1477: itself in the appearance of the mode $\nu(t)$ which was identically equal
1478: to zero for the vibrational state~(\ref{eq90}). As a result, the dimension
1479: of the original one-dimensional bush B$[\hat{a}^2,\hat{i}]$ increases and
1480: this bush transforms into the two-dimensional bush B$[\hat{a}^4,\hat{i}]$.
1481: This transformation is accompanied by the breaking of the symmetry of the
1482: vibrational state (the symmetry of the bush B$[\hat{a}^2,\hat{i}]$ is twice
1483: larger than that of the bush B$[\hat{a}^4,\hat{i}]$). In general case, we
1484: consider a given bush as a stable dynamical object, if the complete
1485: collection of its modes (and, therefore, its dimension) does not change in
1486: time. All other modes in the system, according to the definition of the
1487: bush as the full collection of active modes, possess zero amplitudes, they
1488: are ``sleeping'' modes. If we increase intensity of bush vibrations, some
1489: sleeping modes, because of the parametric interactions with the active
1490: modes, can lose their stability and become excited. In this situation, we
1491: speak about the \textit{loss of stability} of the \textit{original bush},
1492: since the dimension of the vibrational state (the number of active modes)
1493: becomes larger, while its symmetry becomes lower. As a consequence of the
1494: stability loss, the original bush transforms into another bush of higher
1495: dimension.
1496:
1497: As it was described above, the loss of stability of the bush
1498: B$[\hat{a}^2,\hat{i}]$ with respect to its transformation into the bush
1499: B$[\hat{a}^4,\hat{i}]$ can be investigated with the aid of the Mathieu
1500: equation. But if our nonlinear chain contain $N$ particles, we must analyze
1501: the stability of the bush B$[\hat{a}^2,\hat{i}]$ not only relatively to the
1502: mode $\nu(t)$ of the bush B$[\hat{a}^4,\hat{i}]$, but with respect to
1503: \textit{all the other} modes, too. This can be done by the Floquet method.
1504:
1505: The stability of the bush B$[\hat{a}^2,\hat{i}]$\footnote{In the
1506: paper~\cite{FPU1}, this bush was denoted by the symbol B$[2a]$.} in the
1507: FPU-$\alpha$ chain for arbitrary even $N$ was investigated by us
1508: in~\cite{FPU1}. Let us recall some points of this investigation which are
1509: needed for analyzing the stability of the other one-dimensional bushes.
1510:
1511: In the modal space, the dynamical equations of the FPU-$\alpha$ chain can
1512: be written as follows (see Eqs.~(27) from the paper~\cite{FPU1}):
1513: \begin{eqnarray}
1514: \ddot{\nu}_j+\omega_j^2\nu_j=-\frac{\alpha}{\sqrt{N}}\sum_{k=0}^{N-1}\nu_k&&\left[
1515: (\nu_{j+k}+\nu_{-j-k})\left(2\sin\frac{2\pi}{N}k+\sin\frac{2\pi}{N}j\right)\right.\nonumber\\
1516: &&\;\left.+(\nu_{j-k}-\nu_{-j+k})\left(2\sin\frac{2\pi}{N}k-\sin\frac{2\pi}{N}j\right)\right],\label{eq100}
1517: \end{eqnarray}
1518: \[\omega_j^2=4\sin^2\left(\frac{\pi j}{N}\right),\quad j=0,1,2,\cdots,N-1.\]
1519: Here $\nu_j(t)$ are the time-dependent coefficients in front of the basis
1520: vectors $\vec{\psi}_j$~(\ref{eq70}) in the decomposition~(\ref{eq71}). For
1521: brevity, we call $\nu_j(t)$ by the term ``mode'', in spite of the fact that
1522: this term, rigorously speaking, corresponds to the product
1523: $\nu_j(t)\vec{\psi}_j$. The numbers of all modes $\nu_j$ in~(\ref{eq100})
1524: are assumed to be reduced to the interval $1\le i\le N-1$ by adding $\pm
1525: N$, since these numbers are defined modulo $N$. Hereafter, the mode $\nu_0$
1526: is excluded from the consideration of the \textit{vibrational} bushes
1527: because it corresponds to the motion of the chain as a whole.
1528:
1529: The bush B$[\hat{a}^2,\hat{i}]$ consists of only one mode $\nu_{N/2}(t)$
1530: and looks as
1531: \begin{equation}
1532: \B[\hat{a}^2,\hat{i}]=\nu_{N/2}(t)\vec{\psi}_{N/2},
1533: \label{eq102}
1534: \end{equation}
1535: where $\vec{\psi}_{N/2}$ is determined by Eq.~(\ref{eq74}). The dynamical
1536: equation for the mode $\nu_{N/2}(t)$ can be obtained from~(\ref{eq100})
1537: supposing that all the other modes are equal to zero: $\nu_j(t)\equiv 0$,
1538: $j\neq \frac{N}{2}$. This equation reads
1539: \begin{equation}
1540: \ddot{\nu}_{N/2}+\omega_{N/2}^2\nu_{N/2}=0.
1541: \label{eq104}
1542: \end{equation}
1543: Its solution can be written in the form
1544: \begin{equation}
1545: \nu_{N/2}(t)=A\cos(2\tau)
1546: \label{eq105}
1547: \end{equation}
1548: (the initial phase in this solution, as well as in Eq.~(\ref{eq93}), may be
1549: chosen equal to zero~\cite{FPU1}). \textit{Linearizing} the
1550: system~(\ref{eq100}) near the exact solution~(\ref{eq102}) with respect of
1551: all the modes $\nu_j$ ($j\neq \frac{N}{2}$), we obtain the following
1552: approximate equations
1553: \begin{equation}
1554: \ddot{\nu}_j+\omega_j^2\nu_j=-\frac{8\alpha}{\sqrt{N}}\sin\left(\frac{2\pi j}{N}\right)\nu_{N/2}\nu_{N/2-j},
1555: \label{eq106}
1556: \end{equation}
1557: \[\omega_j^2=4\sin^2\left(\frac{\pi j}{N}\right),\quad j=1,2,\cdots,\frac{N}{2}-1,\frac{N}{2}+1,\cdots,N-1.\]
1558: The system~(\ref{eq106}) splits into a number of subsystems containing one
1559: or two equations only. Indeed, the mode $\nu_j$ in~(\ref{eq106}) is
1560: connected with the mode $\nu_{\tilde{j}}$, where $\tilde{j}=\frac{N}{2}-j$,
1561: and vice versa, the mode $\nu_{\tilde{j}}$ is connected with
1562: $\nu_{N/2-\tilde{j}}\equiv \nu_j$. Therefore, we have for
1563: $j=1,2,\dotsc,\frac{N}{2}-1$ (see Eqs.~(\ref{eq106}) the following pairs of
1564: linearized equations which are \textit{independent} from all the other
1565: equations:
1566: \begin{equation}
1567: \begin{array}{lll}
1568: \ddot{\nu}_j&+4\sin^2\left(\frac{\pi j}{N}\right)\nu_j&=-\gamma\sin\left(\frac{2\pi j}{N}\right)\nu_{N/2-j}\cos(2\tau),\\
1569: \ddot{\nu}_{N/2-j}&+4\cos^2\left(\frac{\pi j}{N}\right)\nu_{N/2-j}&=-\gamma\sin\left(\frac{2\pi j}{N}\right)\nu_{j}\cos(2\tau),
1570: \end{array}\label{eq107}
1571: \end{equation}
1572: \[\gamma=\frac{8\alpha A}{\sqrt{N}},\quad j=1,2,\cdots,\frac{N}{4}-1.\]
1573: Here we replaced $\nu_{N/2}(t)$ with $A\cos(2\tau)$ from Eq.~(\ref{eq105}).
1574: For $j=\frac{N}{2}+1,\dotsc,N-1$, we obtain equations equivalent to
1575: Eqs.~(\ref{eq107})~\cite{FPU1}.
1576:
1577: If $j=\frac{N}{2}-j$ and, therefore, $j=\frac{N}{4}$, Eqs.~(\ref{eq107})
1578: reduce to a pair of identical equations of the form
1579: \begin{equation}
1580: \ddot{\nu}_{N/4}+2\nu_{N/4}=-\gamma\nu_{N/4}\cos(2\tau),
1581: \label{eq108}
1582: \end{equation}
1583: which is the Mathieu equation. Actually, this is the above considered case,
1584: describing the transformation of the bush B$[\hat{a}^2,\hat{i}]$ with the
1585: mode $\mu(t)\equiv\nu_{N/2}(t)$ into the two-dimensional bush
1586: B$[\hat{a}^4,\hat{i}]$ containing the modes $\mu(t)\equiv\nu_{N/2}(t)$ and
1587: $\nu(t)\equiv\nu_{3N/4}(t)$.
1588:
1589: It is convenient to rewrite Eqs.~(\ref{eq107}) as follows:
1590: \begin{equation}
1591: \begin{array}{lcl}
1592: \ddot{x}+4\sin^2(k)x&=&\gamma\sin(2k)y\cos(2\tau)\\
1593: \ddot{y}+4\cos^2(k)y&=&\gamma\sin(2k)x\cos(2\tau),
1594: \end{array}
1595: \label{eq109}
1596: \end{equation}
1597: where $k=\frac{\pi j}{N}$, $x(t)=\nu_j(t)$ and $y(t)=\nu_{N/2-j}(t)$. Thus,
1598: studying the loss of stability of the bush B$[\hat{a}^2,\hat{i}]$, brought
1599: about by its interactions with the modes $\nu_{N/4}$ (and $\nu_{3N/4}$), is
1600: reduced to analyzing the Mathieu equation~(\ref{eq108}), while the loss of
1601: its stability with respect to all other modes reduces to analyzing the
1602: equations~(\ref{eq109}).
1603:
1604: Eqs.~(\ref{eq109}) are linear differential equations with periodic
1605: coefficients and, therefore, they can be studied with the aid of the
1606: Floquet theory. The system~(\ref{eq109}) is very interesting, and we would
1607: like to cite the following fragment of the text of our paper~\cite{FPU1},
1608: devoted to its properties:
1609:
1610: ``Our computation of the multiplicators for the system~(\ref{eq109}) as
1611: eigenvalues of the monodromic matrix reveals a surprising fact! Indeed, it
1612: seems that the critical value $\gamma_c$ of the constant $\gamma$
1613: from~(\ref{eq107}), corresponding to the loss of stability of the bush
1614: B$[\hat{a}^2,\hat{i}]$, must depend on the mode number $j$, since the
1615: coefficients of these equations depend explicitly on $\frac{j}{N}$.
1616: However, this is not true. We found that $\gamma_c$ does not depend on
1617: $\frac{j}{N}$, at least up to $10^{-5}$, and coincides with that of the
1618: Mathieu equation~(\ref{eq108}):
1619: \begin{equation}
1620: \gamma_c=2.42332.
1621: \label{eq110}
1622: \end{equation}
1623: Then from Eq.~(\ref{eq107}) we obtain
1624: \begin{equation}
1625: \alpha A_c=0.30292\sqrt{N}.
1626: \label{eq111}
1627: \end{equation}
1628: This nontrivial fact means that the original bush B$[\hat{a}^2,\hat{i}]$
1629: loses its stability with respect to all modes $\nu_j$ ($j\neq \frac{N}{2}$)
1630: \textit{simultaneously}, i.e.\ for the same value $A_c$ of the amplitude of
1631: the mode $\nu_{N/2}(t)$. In other words, all modes of the $N$-particle
1632: FPU-$\alpha$ chain are excited parametrically because of interactions with
1633: B$[\hat{a}^2,\hat{i}]$ when $\alpha A$ reaches its critical value $\alpha
1634: A_c=0.30292\sqrt{N}$, and, therefore, the bush B$[\hat{a}^2,\hat{i}]$
1635: transforms at once into the bush B$[\hat{a}^N]$ of trivial symmetry''.
1636:
1637: One may think that the above-mentioned fact means that there exists such
1638: transformation of Eqs.~(\ref{eq107}) which removes the dependence of these
1639: equations on the parameter $\frac{j}{N}$ (or the dependence on $k$ of
1640: Eqs.~(\ref{eq109})). Nevertheless, this doesn't turn out to be correct.
1641: Indeed, in Fig.~\ref{fig6}, we show (in white color) not only the first
1642: stability domain for the different modes $\nu_k$ for $0\leq k\leq\pi$
1643: ($k=\frac{\pi j}{N}$), but also the second stability domain. The former
1644: domain looks as a white horizontal band near the $k$-axis, while the latter
1645: domain looks as two white triangles above this band (the vertical axis
1646: corresponds to the value $A$ of the amplitude of the $\nu_{N/2}(t)$ mode of
1647: the bush B$[\hat{a}^2,\hat{i}]$). This means that there is no dependence of
1648: the excitation conditions for the modes $\nu_j$ ($j\neq\frac{N}{2}$) on the
1649: number $j$ for the first stability domain, but such a dependence is very
1650: obvious for the second stability domain.
1651:
1652: Since we use similar stability diagrams for all one-dimensional bushes, let
1653: us describe their structure in more detail. In Fig.~\ref{fig6}, each point
1654: $(A,k)$ determines a certain value of the root mode amplitude $A$ of the
1655: bush B$[\hat{a}^2,\hat{i}]$ and a certain value $k=\frac{\pi j}{N}$, which
1656: is connected with the index $j$ of a fixed mode of the FPU-$\alpha$ chain.
1657: The black points $(A,k)$ correspond to the case where the mode
1658: $j=k\frac{N}{\pi}$ becomes excited because of its parametric interaction
1659: with the mode of the bush B$[\hat{a}^2,\hat{i}]$. The white color denotes
1660: the opposite case: the corresponding mode $j$, being zero at the initial
1661: instant, continues to be zero in spite of its interaction with the
1662: considered bush. Dashed vertical lines correspond to the modes
1663: ($j=1,2,\dotsc,12$) for the FPU chain with $N=12$ particles. Considering
1664: stability diagrams, we must make one remark concerning normalization of the
1665: normal modes. In Eq.~(\ref{eq70}), there is the normalizing factor
1666: $\frac{1}{\sqrt{N}}$ which leads to the relation
1667: $\left|\vec{\psi}_k\right|^2=1$. Because of this factor, the absolute
1668: values of the individual atomic displacements corresponding to a given mode
1669: $\vec{\psi}_k$ tend to zero for $N\rightarrow\infty$. Such normalization of
1670: the normal modes is not convenient for studying the bush stability. Indeed,
1671: the critical value of atomic displacements, associated with the loss of
1672: stability, say, for the bush B$[\hat{a}^2,\hat{i}]$ is equal to
1673: $x_c=0.30292/\alpha$ (this result can be easy obtained from
1674: Eq.~(\ref{eq111}) (see~\cite{FPU1}). This value $x_c$ does not depend on
1675: the number $N$ of particles in the FPU chain. Because of this reason,
1676: depicting stability diagrams of one-dimensional and two-dimensional bushes
1677: in Figs.~\ref{fig26}-\ref{fig31}, we use different normalization of the
1678: normal modes than that in Eq.~(\ref{eq70}), namely, we simply reject the
1679: factor $\frac{1}{\sqrt{N}}$ which presents in Eq.~(\ref{eq70}). Then, the
1680: maximal value of the atomic displacements corresponding to every normal
1681: mode $\vec{\psi}_k$ turns out to be equal to $\sqrt{2}$.
1682:
1683: We have just discussed the stability of the bush B$[\hat{a}^2,\hat{i}]$.
1684: Now we intend to consider the stability of all the other one-dimensional
1685: bushes in the FPU-$\alpha$ and FPU-$\beta$ chains. There are three such
1686: bushes for \textit{every} monoatomic nonlinear chain with dihedral symmetry
1687: group, but for the FPU-$\beta$ model, there exist three additional
1688: one-dimensional bushes because of the evenness of its
1689: potential~\cite{BRink}. For investigating stability of these bushes (three
1690: for the FPU-$\alpha$ chain and six for the FPU-$\beta$ chain), we shall
1691: also use the Floquet method.
1692:
1693: There are two difficulties in implementing of this method for
1694: one-dimensional bushes in the general case, as compared to that for the
1695: above considered bush B$[\hat{a}^2,\hat{i}]$. Indeed, the dynamical
1696: equations for the bushes B$[\hat{a}^2,\hat{i}]$ and
1697: B$[\hat{a}^4,\hat{a}\hat{i}]$ in the FPU-$\alpha$ model are reduced to the
1698: harmonic oscillator equation (with different frequencies), while the
1699: dynamical equations for all the other bushes turn out to be
1700: \textit{nonlinear}. As a consequence, the period $T$ of the oscillations,
1701: described by the bush of such a type, depends on the amplitude $\mu_0$ of
1702: these oscillations ($T=T(\mu_0)$). Therefore, we must construct the
1703: monodromy matrix by integration of the linearized dynamical equations, for
1704: the considered chain, over the period $T(\mu_0)$. This period must be found
1705: for every value $\mu_0$ before obtaining the monodromy matrix.
1706:
1707: The second difficulty is that the linearized dynamical system (near the
1708: appropriate exact solution to the given bush) can be split into the
1709: subsystems whose dimension (3 or 4) can be higher than that for the case of
1710: the bush B$[\hat{a}^2,\hat{i}]$. As a result, we must deal, respectively,
1711: with $6\times 6$ or $8\times 8$ monodromy matrices.
1712:
1713:
1714:
1715: \section{Stability of one-dimensional bushes for the FPU chains} We will
1716: consider the stability of the bushes for the FPU-$\alpha$ and FPU-$\beta$
1717: models separately.
1718:
1719: \subsection{FPU-$\alpha$ chain} The exact dynamical equations for the
1720: FPU-$\alpha$ chain in the modal space are given by Eqs.~(\ref{eq100}).
1721:
1722: We want to linearize these equations in the vicinity of the solution
1723: described by a given one-dimensional bush. There are three one-dimensional
1724: bushes associated with the subgroups of the dihedral group of the
1725: monoatomic chains with the appropriate divisibility properties of the
1726: number $N$ of the particles (see Table~\ref{table2}):
1727: B$[\hat{a}^2,\hat{i}]$ for $N\mod 2=0$, B$[\hat{a}^3,\hat{i}]$ for $N\mod
1728: 3=0$, B$[\hat{a}^4,\hat{a}\hat{i}]$ for $N\mod 4=0$. The following root
1729: modes correspond, respectively, to them: $\nu(t)\vec{\psi}_{N/2}$,
1730: $\nu(t)[\varepsilon_1\vec{\psi}_{N/3}+\varepsilon_2\vec{\psi}_{2N/3}]$,
1731: $\nu(t)[\vec\psi_{N/4}+\vec\psi_{3N/4}]$. Note that the root modes for the
1732: bushes B$[\hat{a}^3,\hat{i}]$ and B$[\hat{a}^4,\hat{a}\hat{i}]$ are the
1733: certain superpositions of the \textit{conjugate modes}\footnote{According
1734: to~\cite{FPU1}, we call two \textit{real} modes $\vec{\psi}_j$ and
1735: $\vec{\psi}_{N-j}$ by the term ``conjugate'' since normal modes
1736: $\vec{\phi}_j$ and $\vec{\phi}_{N-j}$, corresponding to them in the
1737: \textit{complex} form, are complex conjugate. The frequencies of the
1738: conjugate normal modes are equal.} with time-independent coefficients
1739: $\left[\varepsilon_1=\sin\left(\frac{\pi}{12}\right),
1740: \varepsilon_2=\cos\left(\frac{\pi}{12}\right)\right]$.
1741:
1742: Let us consider the bush B$[\hat{a}^3,\hat{a}^2\hat{i}]$\footnote{This
1743: bush, being equivalent to the bush B$[\hat{a}^3,\hat{i}]$, turns out to be
1744: more suitable for the next calculations.}. Linearizing the FPU-$\alpha$
1745: dynamical equations near the solutions
1746: $\vec{X}_0(t)=\nu(t)\varepsilon_3[\vec{\psi}_{N/3}-\vec{\psi}_{2N/3}]$ of
1747: this bush $\left[\varepsilon_3=\frac{\sqrt{2}}{2}\right]$, we must consider
1748: the amplitudes $\nu_{N/3}(t)=\varepsilon_3\nu(t)$ and
1749: $\nu_{2N/3}(t)=-\varepsilon_3\nu(t)$ in Eqs.~(\ref{eq100}) to be much
1750: greater than those of all other modes or, more precisely, $\nu_j(t)$ with
1751: $j\neq\frac{N}{3},\frac{2N}{3}$ must be treated as the infinitesimal
1752: values. Then we retain in Eqs.~(\ref{eq100}) only linear terms with respect
1753: to the modes $\nu_j(t)$ ($j\neq\frac{N}{3},\frac{2N}{3}$) which do not
1754: contribute to the given bush. The result of this computational procedure
1755: can be written as follows
1756: \begin{eqnarray}
1757: \ddot\nu_j+\omega_j^2\nu_j=-\frac{\alpha\sqrt 6}{\sqrt N}\nu(t)
1758: &&\left[\left(1+\sqrt 3\sin\frac{2\pi j}{N}-\cos\frac{2\pi j}{N}\right)\nu_{\frac{N}{3}+j}+\right.\nonumber\\
1759: &&\;\left.\left(1-\sqrt 3\sin\frac{2\pi j}{N}-\cos\frac{2\pi j}{N}\right)\nu _{\frac{2N}{3}+j}\right],\label{eq5182}
1760: \end{eqnarray}
1761: \[j=1,2,\cdots,N-1.\]
1762: It can be easy seen from this form of linearized dynamical system that, for
1763: every fixed $j$, the modes $\nu_j$, $\nu_{N/3+j}$ and $\nu_{2N/3+j}$, being
1764: coupled by Eq.~(\ref{eq5182}), are decoupled from all other modes.
1765: Therefore, we obtain for $j=1,2,\cdots,\frac{N}{3}-1$ the following
1766: $3\times 3$ systems of ordinary differential equations with
1767: \textit{periodic coefficients} proportional to the function $\nu(t)$:
1768: \begin{equation}
1769: \begin{array}{l}
1770: \ddot\nu_j+\omega_j^2\nu_j=-\frac{\alpha\sqrt 6}{\sqrt N}\nu(t)\left[ {\left( {1 - 2\cos \left( {\frac{{2\pi j}}{N} + \frac{\pi }{3}} \right)} \right)\nu _{\frac{N}{3} + j} + \left( {1 - 2\sin \left( {\frac{{2\pi j}}{N} + \frac{\pi }{6}} \right)} \right)\nu _{\frac{{2N}}{3} + j} } \right], \\
1771: \ddot\nu_{\frac{N}{3}+j}+\omega_{\frac{N}{3}+j}^2\nu_{\frac{N}{3}+j}=-\frac{\alpha\sqrt 6}{\sqrt N}\nu(t)\left[ {\left( {1 - 2\cos \left( {\frac{{2\pi j}}{N} + \frac{\pi }{3}} \right)} \right)\nu _j + \left( {1 + 2\cos \frac{{2\pi j}}{N}} \right)\nu _{\frac{{2N}}{3} + j} } \right], \\
1772: \ddot\nu_{\frac{2N}{3}+j}+\omega_{\frac{2N}{3}+j}^2\nu_{\frac{2N}{3}+j}=-\frac{\alpha\sqrt 6}{\sqrt N}\nu(t)\left[ {\left( {1 - 2\sin \left( {\frac{{2\pi j}}{N} + \frac{\pi }{6}} \right)} \right)\nu _j + \left( {1 + 2\cos \frac{{2\pi j}}{N}} \right)\nu _{\frac{N}{3} + j} } \right]. \\
1773: \end{array}
1774: \label{eq5183}
1775: \end{equation}
1776: In turn, the function $\nu(t)$ is the solution to the dynamical equation
1777: \[\ddot{\nu}+3\nu=\frac{3\sqrt{6}}{2}\frac{\alpha}{\sqrt{N}}\nu^2\]
1778: of the considered bush B$[\hat{a}^3,\hat{a}^2\hat{i}]$. Thus, the stability
1779: of the bush B$[\hat{a}^3,\hat{a}^2\hat{i}]$ can be investigated by the
1780: Floquet method applied to the systems~(\ref{eq5183}) for
1781: $j=1,2,\cdots,\frac{N}{3}-1$.
1782:
1783: Similarly to the above said, we can obtain the $4\times 4$ systems of
1784: differential equations for investigation of stability of the bush
1785: B$[\hat{a}^4,\hat{a}\hat{i}]$. All systems of such a type, which must
1786: subjected to the further Floquet analysis, are given for the FPU-$\alpha$
1787: chain in Table~\ref{table5}.
1788:
1789:
1790:
1791: \subsection{FPU-$\beta$ chain}
1792: We have obtained the linearized dynamical equations for one-dimensional
1793: bushes in the FPU-$\alpha$ chain by linearizing the \textit{exact}
1794: nonlinear equations~(\ref{eq100}) defined in the modal space. This method,
1795: namely, obtaining the exact equations in the modal space with their
1796: subsequent linearization, is not the simplest method for our purpose.
1797: Indeed, we can, firstly, linearize the dynamical equations in the
1798: \textit{configuration} space and only then transform the obtained equations
1799: to the modal space. In this manner, we can obtain the linearized equations
1800: in the modal space for different nonlinear chains, in particular, for the
1801: FPU-$\beta$ chain. Moreover, we want to find the dynamical equations
1802: linearized near the solution
1803: \begin{equation}
1804: \vec{X}_0(t)=\nu(t)\vec{c},
1805: \label{eq400}
1806: \end{equation}
1807: which describes the exact nonlinear regime corresponding to the considered
1808: one-dimensional bush. Here $\vec{c}$ is a constant vector which determines
1809: the directions of displacements of all particles in the chain, while
1810: $\nu(t)$ is a time-dependent function satisfied the dynamical equation of a
1811: given 1D bush
1812: \begin{equation}
1813: \ddot\nu+\omega^2\nu=F(\nu).
1814: \label{eq401}
1815: \end{equation}
1816: Here $F(\nu)$ is a certain nonlinear function.
1817:
1818: Let us then suppose that
1819: \begin{equation}
1820: \vec{X}(t)=\vec{X}_0(t)+\vec{\delta}(t)
1821: \label{eq410}
1822: \end{equation}
1823: with $\vec{\delta}(t)=\left\{\delta_1(t),\cdots,\delta_N(t)\right\}$ being
1824: an infinitesimal vector, and substitute $\vec{X}(t)$ in this form into the
1825: dynamical equations of the nonlinear chain in the \textit{configuration
1826: space}. For the chain with dynamical equations\footnote{For the
1827: FPU-$\alpha$ and FPU-$\beta$ chains, $m$ is equal to $2$ and $3$,
1828: respectively.}
1829: \begin{equation}
1830: \ddot{x}_k=(x_{k+1}+x_{k-1}-2x_k)+\gamma\left[(x_{k+1}-x_k)^m-(x_k-x_{k-1})^m\right],\quad k=1,2,\cdots,N,
1831: \label{eq415}
1832: \end{equation}
1833: and with periodic boundary conditions $x_{N+1}(t)=x_1(t)$, $x_0(t)=x_N(t)$,
1834: after linearization with respect to $\vec{\delta}(t)$, we obtain
1835: \begin{equation}
1836: \ddot{\vec{\delta}}=B(t)\vec{\delta}.
1837: \label{eq420}
1838: \end{equation}
1839: Here time-periodic matrix $B(t)$ is almost three-diagonal (see
1840: Fig.~\ref{fig50}) and can be determined by the equations
1841: \[\begin{array}{l}
1842: b_{j,j}=-2-\tilde\gamma(t)\left[(c_{j+1}-c_j)^{m-1}+(c_j-c_{j-1})^{m-1}\right],\\
1843: b_{j,j-1}=1+\tilde\gamma(t)(c_j-c_{j-1})^{m-1},\\
1844: b_{j,j+1}=1+\tilde\gamma(t)(c_{j+1}-c_j)^{m-1},
1845: \end{array}\]
1846: where $\tilde\gamma(t)=\gamma m\nu^{m-1}(t)$.
1847:
1848: \begin{figure}
1849: \centering
1850: \begin{tabular}{|c|c|c|c|c|c|c|c|}
1851: \hline
1852: *&*&~~~~& & & & &*\\
1853: \hline
1854: *&*&*&~~~~& & & & \\
1855: \hline
1856: &*&*&*&~~~~& & & \\
1857: \hline
1858: & &*&*&*&~~~~& & \\
1859: \hline
1860: & & &*&*&*&~~~~& \\
1861: \hline
1862: & & & &*&*&*&~~~~\\
1863: \hline
1864: ~~~~& & & & &*&*&*\\
1865: \hline
1866: *&~~~~& & & & &*&*\\
1867: \hline
1868: \end{tabular}
1869: \caption{\label{fig50} The structure of the matrix $B$ from Eq.~\ref{eq420}
1870: (for $N=8$).}
1871: \end{figure}
1872:
1873: The vector Eq.~(\ref{eq420}) represents the linearized dynamical equations
1874: of the considered chain~(\ref{eq415}) in the configuration space. Now, we
1875: must transform these equations to the \textit{modal} space. Let
1876: \begin{equation}
1877: \vec{\delta}(t)=\sum_{j=1}^{N-1}\nu_j(t)\vec{\psi}_j,
1878: \label{eq430}
1879: \end{equation}
1880: where the basis vectors $\vec{\psi}_j$ are determined by Eq.~(\ref{eq70}).
1881: Substituting $\vec{\delta}(t)$ from Eq.~(\ref{eq430}) into
1882: Eq.~(\ref{eq420}) and taking into account that these basis vectors are
1883: orthonormal, one can obtain the final dynamical equations of the chain,
1884: linearized near the solution~(\ref{eq400}) to a given one-dimensional bush.
1885: It must be noted, that this transformation, being simple in principle, is
1886: practically very cumbersome. We performed it with the aid of the
1887: mathematical packet \texttt{MAPLE}. The results of this procedure are
1888: listed for all one-dimensional bushes in the FPU-$\alpha$ and FPU-$\beta$
1889: chains in Table~\ref{table5} and Table~\ref{table6}, respectively.
1890:
1891: As can be seen from Table~\ref{table6}, the full systems of linearized
1892: differential equations corresponding to different one-dimensional bushes in
1893: the FPU-$\beta$ chain split into certain subsystems whose dimensionalities
1894: are equal to $1$, $2$, $3$.\footnote{Let us note that splitting of the
1895: general dynamical system, linearized near a given bush, into independent
1896: subsystems of small dimensionalities, is brought about by
1897: \textit{symmetry-related} causes. Some general results concerning such
1898: splitting will be considered elsewhere.}
1899:
1900: These systems contain time-periodic coefficients proportional to
1901: $\nu^2(t)$, where $\nu(t)$ is the solution to the Duffing equation
1902: \[\ddot{\nu}+\omega^2\nu=\gamma\nu^3.\]
1903: The coefficients $\omega^2$ and $\gamma$ are, naturally, different for
1904: different bushes.
1905:
1906: Using Tables~\ref{table5} and~\ref{table6}, we can investigate the
1907: stability of all one-dimensional bushes, with the aid of the Floquet
1908: method, for the FPU-$\alpha$ and FPU-$\beta$ models, respectively.
1909:
1910: In our bush stability diagrams (see Figs.~\ref{fig6}-\ref{fig14}), we
1911: depict regions of stability (white color) and instability (black color) for
1912: individual modes $j=1,2,3,\dotsc,N-1$ of the FPU chains. The loss of
1913: stability by any mode, which does not belong to a given bush, leads to the
1914: loss of stability by the entire bush. The fact that we can study stability
1915: of individual modes allows us to investigate the bush stability not only
1916: for the chains with finite number of particles, but also for the case of
1917: thermodynamical (continuum) limit $N\rightarrow\infty$ (see a discussion in
1918: the next section).
1919:
1920: Practically, we study the eigenvalues of the monodromy matrices for a
1921: certain subsets of sleeping modes. We choose as many points on the
1922: horizontal $k$-axis ($k=\frac{\pi j}{N}$), as it is necessary for obtaining
1923: sufficiently accurate picture. In some cases, the density of the analyzed
1924: $k$-points corresponds to $N\sim 10^5$. Usually, we recognize instability
1925: of a given set of sleeping modes, if there are eigenvalues of the
1926: corresponding monodromy matrix with modules exceeding $1$ by $10^{-5}$.
1927:
1928: In Table~\ref{table7}, for the case $N=12$, we present the thresholds of
1929: the loss of stability for all one-dimensional bushes in the FPU-$\alpha$
1930: and FPU-$\beta$ chains. In the stability diagrams
1931: (Figs.~\ref{fig6}-\ref{fig14}), all possible modes for this case
1932: ($j=1,2,3,\dotsc,12$) are depicted by vertical dashed lines. Using these
1933: diagrams, one can understand, in some sense, the \textit{cause} of the
1934: given threshold values of bush stability for the FPU chains with $N=12$
1935: (see below).
1936:
1937:
1938:
1939: \subsection{Diagrams of stability for one-dimensional bushes in the
1940: FPU-$\alpha$ chain}
1941: First of all, we would like to emphasize, that studying the stability of
1942: bushes of normal modes, we can restrict ourselves to only one specimen from
1943: the set of ``dynamical domains'' which represent equivalent bushes (see the
1944: previous section).
1945:
1946: Now we consider the stability diagrams for the bushes in the FPU-$\alpha$
1947: chain.
1948:
1949: \subsubsection{Bush B$[\hat{a}^2,\hat{i}]$}
1950: The stability diagram for this bush (see Fig.~\ref{fig6}) has already been
1951: discussed. Let us recall that in all our diagrams, for one-dimensional
1952: bushes, the stability regions are white, while unstable regions are black.
1953:
1954: \subsubsection{Bush B$[\hat{a}^3,\hat{i}]$}
1955: In the configuration and modal spaces this bush looks, respectively, as
1956: follows:
1957: \begin{equation}
1958: \vec{X}(t)=\left\{x(t),0,-x(t)|x(t),0,-x(t)|\cdots\right\}=
1959: \nu(t)\left\{\varepsilon_1\vec{\psi}_{N/3}+\varepsilon_2\vec{\psi}_{2N/3}\right\}.
1960: \label{eq200a}
1961: \end{equation}
1962:
1963: The stability diagram for the bush B$[\hat{a}^3,\hat{i}]$ is presented in
1964: Fig.~\ref{fig7}. The domain depicted in grey color corresponds to the
1965: disruption of the FPU-$\alpha$ chain\footnote{Obviously, this phenomenon is
1966: impossible for the FPU-$\beta$ chain.} (atoms escape from the potential
1967: wells in which the vibrational motion can occur). Thus, we must distinguish
1968: between the stability of the chain itself and the stability of a given
1969: \textit{vibrational} regime, i.e., of a bush of modes, in this chain.
1970:
1971: Let us imagine a horizontal line in Fig.~\ref{fig7} corresponding to a
1972: given value $A$ of the bush root mode. This line can partially pass through
1973: the white and partially through the black regions of the considered
1974: stability diagram. Actually, the parts of this line belonging to the black
1975: (unstable) region represent the set of modes which are excited because of
1976: the interactions with the bush root mode at the fixed level of its
1977: amplitude ($A$). If $A$ is very small, our horizontal line $A=const$
1978: intersects the \textit{narrow} unstable regions near the mode numbers $j=0,
1979: \frac{N}{3}, \frac{2N}{3}$ and $N$.
1980:
1981: Let us suppose that $N=12$ and $A<<1$. Then it is easy to see from
1982: Fig.~\ref{fig7} that only two modes, corresponding to $j=\frac{N}{3}$ and $j=\frac{2N}{3}$,
1983: are excited in our mechanical system\footnote{Note that the mode $\nu_0$
1984: (and the equivalent to it mode $\nu_{12}$) must be excluded because of the
1985: momentum conservation law~\cite{FPU1}.}. This fact is consistent with the
1986: stability of the bush B$[\hat{a}^3,\hat{i}]$ for small values of its root
1987: mode amplitude. Indeed, according to Eq.~(\ref{eq200a}), this bush contains
1988: only two (conjugate) modes $\nu_{N/3}(t)$ and $\nu_{2N/3}(t)$ with the
1989: certain relation between their amplitudes:
1990: \begin{equation}
1991: \nu_{N/3}(t)=\varepsilon_1\nu(t),\quad \nu_{2N/3}(t)=\varepsilon_2\nu(t)
1992: \label{eq201a}
1993: \end{equation}
1994: Moreover, we can find the \textit{threshold of the stability loss} of the
1995: bush B$[\hat{a}^3,\hat{i}]$ directly from the diagram in Fig.~\ref{fig7},
1996: taking into account that all the possible modes for the case $N=12$ are
1997: marked by vertical dashed lines. This threshold is the minimal vertical
1998: distance from the horizontal coordinate axis to the black (unstable) region
1999: for $j=1,2,3,5,6,7,9,10,11$. The stability thresholds for all
2000: one-dimensional bushes in the FPU-$\alpha$ and FPU-$\beta$ chains for
2001: $N=12$ are given in Table~\ref{table7}. From this table, we find that the
2002: stability threshold for the bush B$[\hat{a}^3,\hat{i}]$ (the critical value
2003: $x_c$ of the atomic displacements) is equal to $x_c=0.203$. The
2004: corresponding critical values of the root mode amplitude ($A_c$) and the
2005: energy ($E_c$) are equal to $A_c=0.166$, $E_c=0.047$.
2006:
2007: Let us now increase the value $A$ of the root mode amplitude of the bush
2008: B$[\hat{a}^3,\hat{i}]$. Such an increase leads to the broadening of the
2009: unstable regions near the $j=0, \frac{N}{3}, \frac{2N}{3}, N$ which
2010: intersects the line $A=const$ in Fig.~\ref{fig7}. This fact does not affect
2011: the stability of the bush B$[\hat{a}^3,\hat{i}]$ for $N=12$, as long as
2012: there are no modes $\nu_j$ with $j=1,2,\cdots,11$ (except the modes
2013: $\nu_{N/3}$ and $\nu_{2N/3}$) which fall into the above-mentioned
2014: broadening unstable intervals.
2015:
2016: On the other hand, if $N\rightarrow\infty$ (continuum limit), the mode
2017: density, i.e.\ the density of vertical dashed lines in Fig.~\ref{fig7},
2018: increases and there appear narrow groups of modes near the values
2019: $j=0,\frac{N}{3},\frac{2N}{3},N$ ($k=0,\frac{\pi}{3},\frac{2\pi}{3},\pi$).
2020: This case corresponds to the onset of the modulation instability of the
2021: considered bush of modes. Note that we then have a bush of \textit{narrow
2022: wave packets} which gradually collapses on a long time scale\footnote{We
2023: will discuss this phenomenon elsewhere.}. Thus, we conclude that the bush
2024: B$[\hat{a}^3,\hat{i}]$, being stable for small number of particles in the
2025: chain ($N$), for a fixed value of its root mode amplitude, becomes unstable
2026: in the continuum limit $N\rightarrow\infty$.
2027:
2028: \subsubsection{Bush B$[\hat{a}^4,\hat{a}\hat{i}]$} The one-dimensional bush
2029: B$[\hat{a}^4,\hat{a}\hat{i}]$ can be written as follows
2030: \begin{equation}
2031: \vec{X}(t)=\left\{0,x(t),0,-x(t)|0,x(t),0,-x(t)|\cdots\right\}=\frac{\sqrt{2}}{2}\nu(t)\left[\vec{\psi}_{N/4}+\vec{\psi}_{3N/4}\right].
2032: \label{eq150b}
2033: \end{equation}
2034: Thus, it represents a superposition of two conjugate modes
2035: $\vec{\psi}_{N/4}$ and $\vec{\psi}_{3N/4}$ with equal amplitudes
2036: \begin{equation}
2037: \nu_{N/4}(t)=\nu_{3N/4}(t)=\frac{\sqrt{2}}{2}\nu(t).
2038: \label{eq151}
2039: \end{equation}
2040:
2041: According to the Table~\ref{table7}, the threshold\footnote{This is the
2042: upper boundary of the first stability domain of the considered bush.} of
2043: the stability loss for the bush B$[\hat{a}^4,\hat{a}\hat{i}]$ in the
2044: FPU-$\alpha$ chain, in contrast to the similar thresholds for all other
2045: bushes, is equal to \textit{zero}. In particular, the loss stability
2046: threshold (LST) for the same bush in the FPU-$\beta$ chain is equal to
2047: $1.161$ (for the case $N=12$). Let us discuss this unusual case in more
2048: detail.
2049:
2050: In Fig.~\ref{fig8}, we see that besides the modes $\nu_{N/4}(t)$ and
2051: $\nu_{3N/4}(t)$ contributing to B$[\hat{a}^4,\hat{a}\hat{i}]$, the mode
2052: $\nu_{N/2}(t)$ is also excited for \textit{arbitrarily small} values of the
2053: root mode amplitude of the considered bush. On the other hand, all the
2054: results presented in Table~\ref{table7} and in Figs.~\ref{fig6}-\ref{fig14}
2055: are obtained \textit{numerically} by the Floquet method and, partially, by
2056: a straightforward integration of the dynamical equations of the FPU-chains.
2057: Since numerical results are only approximate, it is desirable to prove
2058: \textit{analytically} that the LST for the bush
2059: B$[\hat{a}^4,\hat{a}\hat{i}]$ in the FPU-$\alpha$ chain is indeed equal to
2060: zero exactly.
2061:
2062: In the general case, the presence of three modes $\nu_{N/4}(t)$,
2063: $\nu_{N/2}(t)$, $\nu_{3N/4}(t)$, in the dynamical regime, means that we
2064: have the \textit{three-dimensional} bush B$[\hat{a}^4]$ (see~\cite{FPU2a})
2065: with
2066: \begin{equation}
2067: \vec{X}(t)=\nu_{N/4}(t)\vec{\psi}_{N/4}+\nu_{N/2}(t)\vec{\psi}_{N/2}+\nu_{3N/4}(t)\vec{\psi}_{3N/4}.
2068: \label{eq144}
2069: \end{equation}
2070: Therefore, one can suppose that the initially excited one-dimensional bush
2071: B$[\hat{a}^4,\hat{a}\hat{i}]$ transforms spontaneously into the
2072: three-dimensional bush B$[\hat{a}^4]$:
2073: \begin{equation}
2074: \B[\hat{a}^4,\hat{a}\hat{i}]\rightarrow\B[\hat{a}^4].
2075: \label{eq160a}
2076: \end{equation}
2077: This transformation is accompanied by the lowering of the symmetry, because
2078: of the loss of the invertion element $\hat{a}\hat{i}$. In the dynamical
2079: picture of the above transformation, the new mode $\nu_{N/2}(t)$ appears
2080: and the specific relation between the modes $\nu_{N/4}(t)$ and
2081: $\nu_{3N/4}(t)$, presented in the bush B$[\hat{a}^4,\hat{a}\hat{i}]$, is
2082: destroyed:
2083: \[\nu_{N/4}(t)\neq\nu_{3N/4}(t).\]
2084:
2085: Thus, we must now prove that the bush B$[\hat{a}^4,\hat{a}\hat{i}]$ loses
2086: its stability with respect to the interaction exactly with the mode
2087: $\nu_{N/2}$. So, let us exclude all the other modes of the FPU-$\alpha$
2088: chain from our consideration and focus only on the
2089: transformation~(\ref{eq160a}). Obviously, if we prove that the stability
2090: threshold of the bush B$[\hat{a}^4,\hat{a}\hat{i}]$, for the
2091: transformation~(\ref{eq160a}), is zero, then it is also zero with respect
2092: to the interactions with \textit{all modes} of the chain.
2093:
2094: Because of the above said, let us consider the dynamical equations of the
2095: three-dimensional bush B$[\hat{a}^4]$ which can be written as follows
2096: \begin{equation}
2097: \begin{array}{l}
2098: \ddot{\mu}_1+2\mu_1=-\frac{8\alpha}{N}\mu_1\mu_2,\\
2099: \ddot{\mu}_2+4\mu_2=-\frac{4\alpha}{N}\left(\mu_1^2-\mu_3^2\right),\\
2100: \ddot{\mu}_3+2\mu_3=\frac{8\alpha}{N}\mu_2\mu_3.
2101: \end{array}
2102: \label{eq155}
2103: \end{equation}
2104: Here we renamed the modes $\nu_{N/4}(t)$, $\nu_{N/2}(t)$ and
2105: $\nu_{3N/4}(t)$ as $\mu_1(t)$, $\mu_2(t)$ and $\mu_3(t)$, respectively.
2106:
2107: In the dynamical regime corresponding to the bush
2108: B$[\hat{a}^4,\hat{a}\hat{i}]$ only one degree of freedom is excited (see
2109: Eq.~(\ref{eq150b})). In our new notation it can be written as
2110: \begin{equation}
2111: \vec{X}(t)=\mu_1(t)\vec{\psi}_{N/4}+\mu_3(t)\vec{\psi}_{3N/4},
2112: \label{eq170b}
2113: \end{equation}
2114: where $\mu_1(t)=\mu_3(t)$. Substituting $\mu_1(t)=\mu_3(t)$, $\mu_2(t)=0$
2115: into Eqs.~(\ref{eq155}), we find that the first and third equations are
2116: both reduced to the equation of the harmonic oscillator
2117: \begin{equation}
2118: \ddot{\mu}_1+2\mu_1=0,
2119: \label{eq156}
2120: \end{equation}
2121: while both sides of the second equation turn out to be zero. Thus,
2122: \begin{equation}
2123: \mu_1^{(0)}(t)=A\cos(\sqrt{2}t+\phi_0)
2124: \label{eq157}
2125: \end{equation}
2126: and we can linearize Eqs.~(\ref{eq155}) near the dynamical
2127: regime~(\ref{eq157}). Note that one can assume $\phi_0=0$ in
2128: Eq.~(\ref{eq157}), without loss of generality, since it is possible to
2129: shift the time variable $t$ in the dynamical equations of the bush
2130: B$[\hat{a}^4]$ by the appropriate value.
2131:
2132: According to the general idea of linearization, let us suppose that
2133: \begin{equation}
2134: \begin{array}{lll}
2135: \mu_1(t)&=&\mu_1^{(0)}(t)+\delta_1(t),\\
2136: \mu_2(t)&=&\delta_2(t),\\
2137: \mu_3(t)&=&\mu_1^{(0)}(t)+\delta_3(t).
2138: \end{array}
2139: \label{eq160b}
2140: \end{equation}
2141: Here $\delta_i$ ($i=1,2,3$) are small corrections to the original dynamical
2142: regime~(\ref{eq170b}) $\left\{\mu_1^{(0)}(t),0,\mu_1^{(0)}(t)\right\}$
2143: which is determined by the one-dimensional bush
2144: B$[\hat{a}^4,\hat{a}\hat{i}]$. Let us now substitute~(\ref{eq160b}) into
2145: Eqs.~(\ref{eq155}) and neglect all nonlinear in $\delta_i$ terms. Then we
2146: have:
2147: \begin{equation}
2148: \begin{array}{lll}
2149: \ddot{\delta}_1+2\delta_1&=&-8\alpha A\cos(\sqrt{2}t)\delta_2,\\
2150: \ddot{\delta}_2+4\delta_2&=&-8\alpha A\cos(\sqrt{2}t)(\delta_1-\delta_3),\\
2151: \ddot{\delta}_3+2\delta_3&=&8\alpha A\cos(\sqrt{2}t)\delta_2.
2152: \end{array}
2153: \label{eq162}
2154: \end{equation}
2155:
2156: Subtracting the third equation from the first equation and introducing new
2157: variables $\gamma_1(t)=\delta_1(t)-\delta_3(t)$, $\gamma_2(t)=\delta_2(t)$,
2158: we obtain finally
2159: \begin{equation}
2160: \begin{array}{lcl}
2161: \ddot{\gamma}_1+2\gamma_1&=&-16\alpha A\cos(\sqrt{2}t)\gamma_2,\\
2162: \ddot{\gamma}_2+4\gamma_2&=&-8\alpha A\cos(\sqrt{2}t)\gamma_1.
2163: \end{array}
2164: \label{eq163}
2165: \end{equation}
2166: Namely for these equations, we must find analytically the nontrivial
2167: solution ($\gamma_1(t)\neq 0$, $\gamma_2(t)\neq 0$)
2168:
2169: Note that the same system can be obtained from the linearized equations for
2170: investigation of the stability of the bush B$[\hat{a}^4,\hat{a}^3\hat{i}]$
2171: which is presented in Table~\ref{table5} for the FPU-$\alpha$ chain.
2172: Indeed, let us assume $j=0$ in the last system from this table, keeping in
2173: mind the relation $k=\frac{\pi j}{N}$. Then we have
2174: \begin{equation}
2175: \begin{array}{lllll}
2176: \ddot{\nu}_0&&&=&0,\\
2177: \ddot{\nu}_{N/4}&+&\omega_{N/4}^2\nu_{N/4}&=&-\frac{4\sqrt{2}\alpha}{\sqrt{N}}\mu(t)\nu_{N/2}(t),\\
2178: \ddot{\nu}_{N/2}&+&\omega_{N/2}^2\nu_{N/2}&=&-\frac{4\sqrt{2}\alpha}{\sqrt{N}}\mu(t)\left[\nu_{N/4}(t)+\nu_{3N/4}(t)\right],\\
2179: \ddot{\nu}_{3N/4}&+&\omega_{3N/4}^2\nu_{3N/4}&=&-\frac{4\sqrt{2}\alpha}{\sqrt{N}}\mu(t)\nu_{N/2}(t),
2180: \end{array}
2181: \label{eq164}
2182: \end{equation}
2183: where $\omega_j^2=4\sin^2(\frac{\pi j}{N})$, while
2184: $\mu(t)=A\cos(\sqrt{2}t)$ is the solution to the equation
2185: $\ddot{\mu}+2\mu=0$. Summing the second and forth equations of the
2186: system~(\ref{eq164}) leads us again to Eqs.~(\ref{eq163}), where
2187: $\gamma_1(t)=\nu_{N/4}(t)+\nu_{3N/4}(t)$, $\gamma_2(t)=\nu_{N/2}(t)$.
2188:
2189: Thus, the analysis of the stability loss by the bush
2190: B$[\hat{a}^4,\hat{a}\hat{i}]$ is reduced to the problem of parametric
2191: excitation of the nontrivial solution to Eqs.~(\ref{eq163}). On the other
2192: hand, it is well known that the onset of unstable motion for the Mathieu
2193: equation can be investigated with the aid of different asymptotic methods
2194: of nonlinear mechanics (averaging method, multiscale method, etc.).
2195: In~\cite{FPU2b}, we use one of such a type methods, namely, the
2196: normalization procedure based on the Poincare-Dulac theorem, for studying
2197: the parametric excitation in the system~(\ref{eq163}). Using this
2198: procedure, we remove, step by step, the so-called \textit{nonresonance}
2199: terms of a given order to higher orders in the small parameter (in our
2200: case, this is the root mode amplitude $A$) with the aid of the appropriate
2201: nonlinear transformations of the dynamical variables $\gamma_i(t)$. In such
2202: a way, the normalized system of differential equations with respect to new
2203: variables $\tilde{\gamma}_i(t)$ is obtained and, as a rule, this system
2204: turns out to be simpler than the original system.
2205:
2206: In~\cite{FPU2b}, it was shown that the normalized equations up to $O(A^3)$,
2207: obtained from Eqs.~(\ref{eq163}), can be solved exactly. The corresponding
2208: solution for $\tilde{\gamma}_1(t)$ contains the growing exponent
2209: $\exp(\sqrt{2} \alpha^2 A^2 t)$ which increases from zero for the
2210: \textit{arbitrary small} value of the root mode amplitude $A$. For the old
2211: variable $\gamma_1(t)$ this exponent, being decomposed into the power
2212: series with respect to $A$, produces the unremovable secular term of the
2213: form\footnote{The full approximate analytical solution to
2214: Eqs.~(\ref{eq163}) represents the power series for $\gamma_1(t)$ and
2215: $\gamma_2(t)$ with respect to the small parameter $A$. Since this solution
2216: is rather cumbersome, we do not present it here.}
2217: \[\alpha^2A^2\sqrt{2}t\sin(\sqrt{2}t+\delta).\]
2218:
2219: From the above said, it is clear that the parametric excitation in the
2220: system~(\ref{eq163}) takes place for the arbitrary small value of the root
2221: mode amplitude $A$ and, therefore, the bush B$[\hat{a}^4,\hat{a}\hat{i}]$
2222: possesses the stability domain of zero size, as it has already been
2223: concluded directly from Fig.~\ref{fig8}.
2224:
2225:
2226: \subsection{Diagrams of stability for one-dimensional bushes in the
2227: FPU-$\beta$ chain}
2228: \subsubsection{Bush B$[\hat{a}^2,\hat{i}]$}
2229:
2230: The stability diagram corresponding to the bush B$[\hat{a}^2,\hat{i}]$ in
2231: the FPU-$\beta$ chain is presented in Fig.~\ref{fig9}. From this diagram,
2232: one can see that for any \textit{finite} $N$ and for sufficiently small
2233: root mode amplitude $A$, there exists, in our mechanical systems, only one
2234: mode $\nu_{N/2}(t)$, which is the root (and single!) mode of the bush
2235: B$[\hat{a}^2,\hat{i}]$. On the other hand, if we \textit{fix} arbitrary
2236: small value of $A$ and then begin to increase the number of particles $N$,
2237: such $N$ can be always found for which some other modes, close to the mode
2238: $\nu_{N/2}(t)$, turn out to be excited (see the center of the
2239: diagram~\ref{fig9}). In other words, the threshold of the loss of stability
2240: of the considered bush decreases with increasing $N$. Thus, the bush
2241: B$[\hat{a}^2,\hat{i}]$ becomes unstable in the continuum limit
2242: $N\rightarrow\infty$.
2243:
2244: Another interesting fact can be revealed with the aid of the
2245: diagram~\ref{fig9}. Namely, there are certain intervals of mode numbers on
2246: r.h.s.\ and l.h.s.\ of this diagram which cannot be excited because of the
2247: interactions with the mode $\nu_{N/2}(t)$. Note that the both above
2248: mentioned facts were found\footnote{The authors of this paper used the
2249: direct integration of the dynamical equations of the FPU-$\beta$ chain in
2250: the modal space, while we use the Floquet method.} and discussed, in rather
2251: different terms, in the paper~\cite{PoggiRuffo}.
2252:
2253: \subsubsection{Bush B$[\hat{a}^3,\hat{i}]$}
2254:
2255: The form of the bush B$[\hat{a}^3,\hat{i}]$ in the configuration and modal
2256: spaces was given in Eq.~(\ref{eq200a}) The corresponding stability diagram
2257: is shown in Fig.~\ref{fig10}. From this diagram, we find that, for small
2258: amplitudes, the bush B$[\hat{a}^3,\hat{i}]$ is not stable in the continuum
2259: limit $N\rightarrow\infty$, but can be stable for finite $N$. For example,
2260: for $N=12$ the stability takes place up to the value $A_1$ of the root
2261: mode\footnote{This mode is a certain superposition,
2262: $\varepsilon_1\vec{\psi}_{N/3}+\varepsilon_2\vec{\psi}_{2N/3}$, of the
2263: conjugate modes $\vec{\psi}_{N/3}$ and $\vec{\psi}_{2N/3}$.} amplitude~$A$.
2264: Note that three horizontal dashed lines are depicted in Fig.~\ref{fig10}:
2265: the lower line is described by the equation $A=A_1$, while the middle and
2266: upper lines correspond to the equations $A=A_2$ and $A=A_3$, respectively.
2267:
2268: For $A_1<A<A_2$ all the modes, with the only exception of the modes with
2269: $j=\frac{2N}{12}=\frac{N}{6}$, $j=\frac{6N}{12}=\frac{N}{2}$ and
2270: $j=\frac{10N}{12}=\frac{5N}{6}$, are excited.\footnote{Note that only the
2271: results of a \textit{linear} stability analysis are presented in our
2272: stability diagrams: we find those modes, which are excited because of the
2273: direct parametric excitation with the root mode of the considered bush. On
2274: the other hand, if some new modes appear in the mechanical system, we must
2275: also take into account their interactions with other modes (it is the
2276: problem of nonlinear stability analysis).} Therefore, for $N=12$, the bush
2277: with trivial symmetry is excited. Indeed, the excitation of the mode with
2278: $j=\frac{N}{12}$ can lead, in principle, to excitation of all the modes
2279: with $j=\frac{kN}{12}$ ($k=1,2,\cdots,11$).
2280:
2281: It is very interesting that there exists also the \textit{upper} value
2282: ($A_3$) for the loss of stability of the bush B$[\hat{a}^3,\hat{i}]$.
2283: Indeed, for $A>A_3$ (see Fig.~\ref{fig10}) this bush turns out to be stable
2284: for \textit{all} values of the number ($N$) of particles in the chain.
2285: Thus, the bush B$[\hat{a}^3,\hat{i}]$ becomes stable even in the continuum
2286: limit, for such values of the root mode amplitude $A$.
2287:
2288: \subsubsection{Bush B$[\hat{a}^4,\hat{a}\hat{i}]$}
2289: The description of the bush B$[\hat{a}^4,\hat{a}\hat{i}]$ in the
2290: configuration and modal spaces has already been given by Eq.~(\ref{eq150b}):
2291: \[\vec{X}(t)=\left\{0,x,0,-x|0,x,0,-x|\dots\right\}=\varepsilon_3\nu(t)\left[\vec{\psi}_{N/4}+\vec{\psi}_{3N/4}\right].\]
2292: The stability diagram for this bush for the FPU-$\beta$ chain is
2293: represented in Fig.~\ref{fig12}. From this diagram, one can see that, for
2294: any finite $N$, the bush B$[\hat{a}^4,\hat{a}\hat{i}]$ is stable for
2295: sufficiently small root mode amplitudes $A$. For example, for $N=12$, this
2296: bush loses its stability only for $A>A_1$ (see Fig.~\ref{fig12}). For very
2297: large $N$, the bush B$[\hat{a}^4,\hat{a}\hat{i}]$ practically also remains
2298: stable for sufficiently small $A$ because the intervals of the numbers of
2299: the excited modes turn out to be very narrow (these intervals are situated
2300: near $j=\frac{N}{4}$ and $j=\frac{3N}{4}$). In the continuum limit
2301: $N\rightarrow\infty$, rigorously speaking, the bush
2302: B$[\hat{a}^4,\hat{a}\hat{i}]$ appears to be unstable, but its instability
2303: is weak.
2304:
2305: We have just discussed the stability of all one-dimensional bushes for the
2306: FPU-$\beta$ chain which are induced by the dihedral symmetry group. Now let
2307: us consider the stability of three ``additional'' one-dimensional bushes
2308: whose existence is brought about by the evenness of the potential of the
2309: FPU-$\beta$ chain.
2310:
2311: \subsubsection{Bush B$[\hat{a}^3,\hat{i}\hat{u}]$}
2312: The description of this bush in configuration and modal spaces looks as
2313: follows:
2314: \begin{equation}
2315: \vec{X}(t)=\left\{x,-2x,x|x,-2x,x|\dots\right\}=\nu(t)\left[\varepsilon_2\vec{\psi}_{N/3}-\varepsilon_1\vec{\psi}_{2N/3}\right].
2316: \label{eq200b}
2317: \end{equation}
2318: The stability diagram of the bush B$[\hat{a}^3,\hat{i}\hat{u}]$ is given in
2319: Fig.~\ref{fig11}. It seems similar to that of the above discussed bush
2320: B$[\hat{a}^3,\hat{i}]$ (see Fig.~\ref{fig10}). Therefore, the discussion of
2321: the stability properties of the bush B$[\hat{a}^3,\hat{i}\hat{u}]$ can be
2322: practically reduced to that of the bush B$[\hat{a}^3,\hat{i}]$.
2323:
2324: \subsubsection{Bush B$[\hat{a}^4,\hat{i}\hat{u}]$}
2325: This bush can be described by the equation
2326: \begin{equation}
2327: \vec{X}(t)=\left\{x,-x,-x,x|x,-x,-x,x|\dots\right\}=\nu(t)\vec{\psi}_{N/4}.
2328: \label{eq201b}
2329: \end{equation}
2330: The corresponding stability diagram is represented in Fig.~\ref{fig13} and
2331: looks as ``rabbit ears''. For root mode amplitudes $A>A_1$ (see
2332: Fig.~\ref{fig13}) this bush is stable for \textit{arbitrary} $N$. It is
2333: interesting that for the case $N=12$ there are no excited modes, certainly,
2334: besides the mode $\vec{\psi}_{N/4}$ (or the mode $\vec{\psi}_{3N/4}$ for
2335: the equivalent bush B$[\hat{a}^4,\hat{a}^2\hat{i}\hat{u}]$ which looks in
2336: the modal space as $\vec{X}(t)=\nu(t)\vec{\psi}_{3N/4}$). This is in
2337: agreement with Table~\ref{table7} from which we find that this bush is
2338: stable, \textit{at least}, up to the root mode amplitude $A<20$.
2339:
2340: \subsubsection{Bush B$[\hat{a}^6,\hat{a}\hat{i},\hat{a}^3\hat{u}]$}
2341: This bush can be described as follows:
2342: \begin{equation}
2343: \vec{X}(t)=\left\{0,x,x,0,-x,-x|0,x,x,0,-x,-x|\dots\right\}=\nu(t)\left[\varepsilon_1\vec{\psi}_{N/6}+\varepsilon_2\vec{\psi}_{5N/6}\right].
2344: \label{eq202}
2345: \end{equation}
2346:
2347: One of the distinguishing features of the stability diagram for this bush
2348: (see Fig.~\ref{fig14}) is the presence of three white regions in the
2349: ``black sea'' of instability. They correspond to the second stability zones
2350: for some modes whose numbers fall into these white regions. Another
2351: interesting feature of the stability diagram for the considered bush is the
2352: very narrow unstable intervals in the ``white sea'' of stability (see the
2353: black lines which resemble parabolas). We did not examine these unstable
2354: parabolas in detail, because they seem to be not very essential for
2355: consideration of the stability of the bush
2356: B$[\hat{a}^6,\hat{a}\hat{i},\hat{a}^3\hat{u}]$. Indeed, they represent too
2357: narrow regions of unstable movements.
2358:
2359: \subsection{About numerical values of root mode amplitudes for the FPU
2360: chains}
2361: The FPU models of $\alpha$ and $\beta$ type, considered as exact mechanical
2362: systems, do not contain the interparticle distance $a$. As a consequence,
2363: one cannot estimate the reasonable values of the root mode amplitudes for
2364: the above discussed bushes. This is an important problem because atomic
2365: displacements in real crystals cannot, as a rule, exceed $10-15\%$ of the
2366: equilibrium interparticle distance.
2367:
2368: On the other hand, we can consider FPU system as a certain approximation to
2369: more realistic models of one-dimensional crystals. Let us consider a
2370: nonlinear monoatomic chain with the Lennard-Jones potential $U(r)$
2371: describing the interaction between the nearest particles:
2372: \begin{equation}
2373: U(r)=\varepsilon\left[\left(\frac{\sigma}{r}\right)^{12}-\left(\frac{\sigma}{r}\right)^6\right].
2374: \label{eq300}
2375: \end{equation}
2376: Here, $r$ is the interparticle distance, while $\varepsilon$ and $\sigma$
2377: are certain constants.
2378:
2379: Let us expand the force $f(r)$ acting on one particle, in Taylor series
2380: near the equilibrium interparticle distance $a=\sigma\sqrt[6]{2}$ up to the
2381: terms of the second order:
2382: \begin{equation}
2383: f(r)=-\frac{9\sqrt[3]{4}}{\sigma^2}(r-a)+\frac{189\sqrt{2}}{2\sigma^3}(r-a)^2.
2384: \label{eq301}
2385: \end{equation}
2386: Then we can write the Newton equations of motion in the form
2387: \begin{equation}
2388: m\ddot{x}_i=\frac{9\sqrt[3]{4}\varepsilon}{\sigma^2}\left(-2x_i+x_{i+1}+x_{i-1}\right)+
2389: \frac{189\sqrt{2}\varepsilon}{2\sigma^3}\left[\left(x_{i+1}-x_i\right)^2-\left(x_i-x_{i-1}\right)^2\right],
2390: \label{eq302}
2391: \end{equation}
2392: where $m$ is the mass of one particle.
2393:
2394: Introducing new dimensionless time and space variables
2395: \begin{equation}
2396: t'=\sqrt{\frac{9\sqrt[3]{4}\varepsilon}{m\sigma^2}}t,\quad
2397: x'_n=\frac{21\sqrt[3]{2}}{a}x_n,
2398: \label{eq303}
2399: \end{equation}
2400: we obtain the following equations:
2401: \begin{equation}
2402: \ddot{x}'_i=(-2x'_i+x'_{i+1}+x'_{i-1})+[(x'_{i+1}-x'_i)^2-(x'_i-x'_{i-1})^2].
2403: \label{eq304}
2404: \end{equation}
2405: These equations precisely correspond to the FPU-$\alpha$ model considered
2406: in the present paper.
2407:
2408: On the other hand, the dimensionless variables $x'_i$, according to
2409: Eq.~(\ref{eq303}), are proportional to the ratio of the real atomic
2410: displacements $x_i$ and the equilibrium interparticle distance $a$. As it
2411: has already been noted, for real crystal, $x_i/a$ must not exceed
2412: $10-15\%$. The dimensionless displacements $x'_i$ from the FPU-$\alpha$
2413: model~(\ref{eq304}) can be considerably larger than this value, because of
2414: the coefficient $21\sqrt[3]{2}\approx 29.7$ before $x_i/a$. For example, if
2415: $x_i\sim 0.1a$, then $x'_i\sim 3$.
2416:
2417: Let us now recall that, studying the bush stability, we use such
2418: normalization of the normal modes, that their amplitudes are of the same
2419: order as the atomic displacements corresponding to them. Thus, the root
2420: mode amplitudes $A$ at about $3$ seem to be rather reasonable. It is
2421: precisely because of this fact that we depict the bush stability diagram
2422: up to $A$ of such an order of magnitude.
2423:
2424: \section{Stability of two-dimensional bushes for the FPU chains}
2425: In the previous section, we considered the stability of one-dimensional
2426: bushes with respect to increasing their root mode amplitudes. Such an
2427: increasing leads to strengthening of the parametric interactions of the
2428: root mode of a given bush with all the other modes which do not belong to
2429: this bush and, finally, to a loss of its stability. Actually, the threshold
2430: of the loss of stability of any one-dimensional bush depends only on its
2431: \textit{total energy} (the energy of the initial excitation), but not on
2432: the distribution of this energy between the kinetic and potential
2433: counterparts~\cite{FPU2b}. In two last columns of Table~\ref{table7} we
2434: give the energy thresholds of the loss of stability for all one-dimensional
2435: bushes in the FPU chains. In contrast to one-dimensional bushes, the
2436: stability of the \textit{multi-dimensional} bushes depends not only on the
2437: energy of the initial excitation, but also on the complete set of the
2438: initial conditions. Let us consider this problem in more detail.
2439:
2440: The four-dimensional phase space corresponds to the two-dimensional bushes
2441: considered below in the present paper. In Fig.~\ref{fig15}-\ref{fig24}, we
2442: depict only certain planar sections of the four-dimensional stability
2443: domains, aiming to show how nontrivial the boundary of these domains can
2444: be.
2445:
2446: The complete set of initial conditions for the dynamical equations of a
2447: given two-dimensional bush determines the concrete values of
2448: $\nu_1(0),\dot{\nu}_1(0),\nu_2(0),\dot{\nu}_2(0)$, where $\nu_i(t)$ and
2449: $\dot{\nu}_i(t)$ ($i=1,2$) are the two modes of the considered bush and
2450: their velocities, respectively. We depict the bush stability diagrams (see
2451: Fig.~\ref{fig15}-\ref{fig31}) in certain planes which are chosen by fixing
2452: two of the four above-mentioned initial values. Most frequently, we specify
2453: $\dot{\nu}_1(0)=0,\dot{\nu}_2(0)=0$ and change $\nu_1(0),\nu_2(0)$ in some
2454: intervals near their zero values. Therefore, the initial kinetic energy
2455: assumed to be equal to zero and the dependence of bush stability on the
2456: initial mode amplitudes $\nu_1(0)$ and $\nu_2(0)$ is studied.
2457:
2458: The diagrams discussed in the previous section allow us to analyze
2459: stability of one-dimensional bushes for the FPU chains with arbitrary
2460: number ($N$) of particles. Unfortunately, we cannot obtain similar diagrams
2461: for two-dimensional bushes. Stability diagrams considered in this section
2462: depend on the number of particles and, therefore, they must be constructed
2463: individually for each specific $N$. Because of this reason, we consider
2464: stability of 2D bushes only for fixed $N$ (for the most diagrams $N=12$).
2465: As a consequence, we do not discuss the stability properties of the 2D
2466: bushes with translation symmetry $[\hat{a}^5]$, $[\hat{a}^8]$ and
2467: $[\hat{a}^{10}]$ (the complete list of 2D bushes for the FPU chains can be
2468: found in Appendix).
2469:
2470: Firstly, let us consider the stability of the two-dimensional bush
2471: B$[\hat{a}^4,\hat{i}]$ in the FPU-$\alpha$ chain with $N=12$ particles. The
2472: stability domain for this bush, in the plane of the initial values of its
2473: both modes $[\nu_1(0),\nu_2(0)]$, is presented in \textit{black}
2474: color\footnote{In stability diagrams for one-dimensional bushes, we use
2475: white color for stability regions and black color for instability regions.
2476: In contrast, for two-dimensional bushes, white color corresponds to
2477: instability domains, while black color is used for stability domains. It
2478: seems to us that this manner of depicting stability diagrams turns out to
2479: be more expressive.} in the center of Fig.~\ref{fig77}. This domain,
2480: resembling a beetle, is depicted on the background of the energy level
2481: lines. The bush B$[\hat{a}^4,\hat{i}]$ loses its stability when we cross
2482: the boundary of the black region in any direction. Moreover, it can be
2483: shown that the considered bush transforms into \textit{different} bushes of
2484: larger dimensions at different points of the boundary. Which bushes appear
2485: near the various parts of the boundary of the stability domain of the
2486: bush~B$[\hat{a}^4,\hat{i}]$ will be discussed elsewhere. Now we merely want
2487: to emphasize that the loss of stability of this bush is not determined only
2488: by the energy of the initial excitation (the boundary of stability domain
2489: does not coincide with any energy level line!), but depends on the relation
2490: between the values of its both modes $[\nu_1(0),\nu_2(0)]$, as well. Thus,
2491: the stability domain of the bush B$[\hat{a}^4,\hat{i}]$ is strongly
2492: anisotropic: the linear size of this domain in different directions, from
2493: zero point $\nu_1(0)=0,\nu_2(0)=0$, turns out to be different.
2494:
2495: The stability diagram of the bush B$[\hat{a}^4,\hat{i}]$ in the
2496: FPU-$\alpha$ chain, depicted in Fig.~\ref{fig77}, was obtained by the
2497: direct numerical experiment for the case $\dot\nu_1(0)=0,\dot\nu_2(0)=0$.
2498: Therefore, initial kinetic energy was equal to zero and the total
2499: excitation energy was purely potential.
2500:
2501: It is interesting to consider the case where the excitation energy is
2502: distributed at the initial instant, in some a manner, between its kinetic
2503: and potential counterparts. The appropriate stability diagram is depicted
2504: in Fig.~\ref{fig78}. Actually, this is simply another section of the same
2505: four-dimensional stability domain of the considered bush which is
2506: determined by the equations $\dot\nu_1(0)=0.2,\dot\nu_2(0)=0.1$.
2507:
2508: Comparing Fig.~\ref{fig77} and Fig.~\ref{fig78}, one can see the effect of
2509: the ``wind blowing from the right to the left''. This effect leads to a
2510: loss of the symmetry of the diagram in Fig.~\ref{fig77} with respect to the
2511: vertical axis passing through the origin.
2512:
2513: Thus, the threshold of the loss of stability of the bush
2514: B$[\hat{a}^4,\hat{i}]$ in the FPU-$\alpha$ chain depend essentially not
2515: only on the distribution of the initial energy between its two modes, but
2516: also between the kinetic and potential counterparts. Let us repeat, once
2517: more, that instead of dealing with the four-dimensional stability diagram,
2518: we investigate different \textit{sections} of this domain.
2519:
2520: The stability diagrams for two other two-dimensional bushes, B$[\hat{a}^3]$
2521: and B$[\hat{a}^6,\hat{a}\hat{i}]$, in the FPU-$\alpha$ chain are depicted
2522: in Fig.~\ref{fig15} and Fig.~\ref{fig17}, respectively. These diagrams, as
2523: well as the diagrams in Fig.~\ref{fig18}-\ref{fig24} for the FPU-$\beta$
2524: chain, correspond to the initial conditions $\dot\nu_1(0)=0$,
2525: $\dot\nu_2(0)=0$, i.e.\ the kinetic energy of the initial excitation is
2526: supposed to be equal to zero.
2527:
2528: Let us emphasize that the stability diagrams for the \textit{same} bushes
2529: in the FPU-$\alpha$ and FPU-$\beta$ chains look very different. As an
2530: example, one can compare Fig.~\ref{fig16} and Fig.~\ref{fig19} for the bush
2531: B$[\hat{a}^4,\hat{i}]$ in FPU-$\alpha$ and FPU-$\beta$ chains,
2532: respectively. It is interesting to note that the last diagram
2533: (Fig.~\ref{fig19}) demonstrates the presence of several regions of
2534: instability (white color) in the stable sea (black color).
2535:
2536: In Figs.~\ref{fig20}-\ref{fig24}, we represent the stability diagrams
2537: corresponding to four additional two-dimensional bushes for the FPU-$\beta$
2538: chain (as it has already been said, they are induced by the evenness of the
2539: potential of this model).
2540:
2541: The most nontrivial picture corresponds to the stability diagram of the
2542: bush B$[\hat{a}^4,\hat{a}^2\hat{u}]$. It resembles a chivalrous cross with
2543: two crossed swords (along the diagonals of the diagram). Note that these
2544: swords describe, actually, the stability of two equivalent one-dimensional
2545: bushes $\nu(t)[\vec{\psi}_3+\vec{\psi}_9]$ and
2546: $\nu(t)[\vec{\psi}_3-\vec{\psi}_9]$, while the two-dimensional bush
2547: B$[\hat{a}^4,\hat{a}^2\hat{u}]$ is
2548: $\nu_1(t)\vec{\psi}_3+\nu_2(t)\vec{\psi}_9$.
2549:
2550: For many bushes, the stability domains depend essentially on the number $N$
2551: of particles in the chain. The area of the corresponding regions of
2552: stability, in our diagrams, tends often to decrease with the increasing of
2553: $N$. We demonstrate such a behavior in the cases $N=8,12,16,20,24,48$ for
2554: the bush B$[\hat{a}^4,\hat{i}]$ in the FPU-$\alpha$ chain (see
2555: Figs.~\ref{fig26}-\ref{fig31})\footnote{We do not discuss the continuum
2556: limit $N\rightarrow\infty$ for two-dimensional bushes in the present
2557: paper.}. Note that the stability domain of this bush shrinks only along the
2558: horizontal axis $\nu_1(0)$ when $N\rightarrow\infty$. This is the
2559: consequence of the fact that when $\nu_1(t)\rightarrow 0$ the bush
2560: B$[\hat{a}^4,\hat{i}]=\nu_1(t)\vec{\psi}_{3N/4}-\nu_2(t)\vec{\psi}_{N/2}$
2561: (or its equivalent form
2562: B$[\hat{a}^4,\hat{a}^2\hat{i}]=\nu_1(t)\vec{\psi}_{N/4}+\nu_2(t)\vec{\psi}_{N/2}$)
2563: tends to the one-dimensional bush
2564: B$[\hat{a}^2,\hat{i}]=\nu_2(t)\vec{\psi}_{N/2}$. On the other hand, the
2565: latter bush possesses, for $N\rightarrow\infty$, a finite one-dimensional
2566: interval of stability with respect to increasing of the initial value
2567: $\nu_2(0)$ of its single mode $\nu_2(t)$.
2568:
2569:
2570:
2571: \section{Conclusion}
2572:
2573: The bushes of normal modes as \textit{exact} nonlinear excitations in the
2574: physical systems with discrete (point or space) symmetry groups were
2575: introduced and discussed in our previous
2576: papers~\cite{DAN1,DAN2,PhysD,Okta,NNM,ENOC3,C60}. In general case, bushes
2577: can be rather complex, for example, the bushes in the $C_{60}$ fullerene
2578: structure~\cite{C60}.
2579:
2580: In~\cite{FPU1}, we considered bushes in monoatomic nonlinear chains which
2581: represent the simplest systems with translational symmetry. Mainly, we
2582: studied there the bushes induced by the group $T$ of pure translations and
2583: discussed only in part those induced by the dihedral symmetry group $D$.
2584: The stability of a few simplest bushes was also studied in~\cite{FPU1}.
2585: Nevertheless, there are some more one-dimensional and two-dimensional
2586: bushes in the FPU-chains, in particular, those obtained by Bob Rink
2587: in~\cite{BRink}, whose stability still has not been investigated. On the
2588: other hand, the problem of the bush stability is one of the most important
2589: problems in the general theory of the bushes of normal modes. Because of
2590: this, in the present paper, we discuss in detail the stability of all
2591: one-dimensional and some two-dimensional bushes in both FPU-$\alpha$ and
2592: FPU-$\beta$ chains.
2593:
2594: Let us summarize the results obtained in the previous sections of this
2595: paper.
2596:
2597: 1. A simple crystallographic method for finding bushes in nonlinear chains
2598: is developed.
2599:
2600: 2. The stability of all one-dimensional bushes in the FPU-$\alpha$ (three
2601: bushes) and in the FPU-$\beta$ (six bushes, including those obtained
2602: in~\cite{BRink}) are investigated. The stability diagrams which can be used
2603: for studying stability of one-dimensional bushes for the FPU chains with
2604: any finite and infinite number of particles (continuum limit
2605: $N\rightarrow\infty$) are presented. They were obtained by Floquet method.
2606:
2607: 3. The stability diagrams of different type, which represent certain
2608: sections of four-dimensional stability domains, corresponding to some
2609: two-dimensional bushes in the FPU-$\alpha$ (three bushes) and in the
2610: FPU-$\beta$ (seven bushes), are found. These diagrams demonstrate
2611: explicitly that the bush stability domains depend not only on the energy of
2612: the initial excitation, but on the complete set of the initial conditions
2613: in the bush phase space.
2614:
2615: The above mentioned diagrams for one-dimensional and two-dimensional bushes
2616: show how nontrivial the shape of the stability domains for the bushes of
2617: normal modes can be.
2618:
2619: In this paper, we consider only the main mechanism for the loss of bush
2620: stability, wherein it is brought about by the parametric excitations of the
2621: bush modes with those modes which do not belong to the given bush.
2622: Nevertheless, one can consider some different reasons for a bush to lose
2623: its stability, such as a finite temperature of the thermal bath, the
2624: presence of impurities in the chain structure, etc. These problems, as well
2625: as the methods for excitation of the bushes of normal modes in the
2626: FPU-chains will be considered in a separate paper.
2627:
2628:
2629: \section*{Acknowledgments}
2630: We are very grateful to Prof.\ V.P.~Sakhnenko for his friendly support and
2631: to O.E.~Evnin for his valuable help with the language corrections in the
2632: text of this paper.
2633:
2634:
2635:
2636: \clearpage
2637:
2638: \section*{Appendix. Two-dimensional bushes of vibrational modes for the FPU
2639: chains}
2640: \begin{table}[hb]
2641: \centering
2642: \caption{\label{tableA1} Representation of 2D bushes of vibrational modes
2643: for the FPU chains in the configuration space (classification by the
2644: dihedral group~$D$)}
2645: \begin{tabular}{|l|c|}
2646: \hline
2647: Bush&Displacement pattern\\
2648: \hline
2649: B$[a^3]$&$|x_1,x_2,x_3|$\\
2650: \hline
2651: B$[a^4,i]$&$|x_1,x_2,-x_2,-x_1|$\\
2652: B$[a^4,a^2i]$&$|x_1,-x_1,x_2,-x_2|$\\
2653: \hline
2654: B$[a^5,i]$&$|x_1,x_2,0,-x_2,-x_1|$\\
2655: B$[a^5,ai]$&$|0,x_1,x_2,-x_2,-x_1|$\\
2656: B$[a^5,a^2i]$&$|x_1,-x_1,x_2,0,-x_2|$\\
2657: B$[a^5,a^3i]$&$|x_1,0,-x_1,x_2,-x_2|$\\
2658: B$[a^5,a^4i]$&$|x_1,x_2,-x_2,-x_1,0|$\\
2659: \hline
2660: B$[a^6,ai]$&$|0,x_1,x_2,0,-x_2,-x_1|$\\
2661: B$[a^6,a^3i]$&$|x_1,0,-x_1,x_2,0,-x_2|$\\
2662: B$[a^6,a^5i]$&$|x_1,x_2,0,-x_2,-x_1,0|$\\
2663: \hline
2664: \end{tabular}
2665: \end{table}
2666:
2667: \begin{table}[h]
2668: \centering
2669: \caption{\label{tableA2} Additional 2D bushes for the FPU-$\beta$ chain}
2670: \begin{tabular}{|l|c|}
2671: \hline
2672: Bush&Displacement pattern\\
2673: \hline
2674: B$[a^4,aiu]$&$|x_1,x_2,x_3,x_2|$\\
2675: B$[a^4,a^3iu]$&$|x_2,x_1,x_2,x_3|$\\
2676: \hline
2677: B$[a^4,a^2u]$&$|x_1,x_2,-x_1,-x_2|$\\
2678: \hline
2679: B$[a^5,iu]$&$|x_1,x_2,x_3,x_2,x_1|$\\
2680: B$[a^5,aiu]$&$|x_1,x_2,x_3,x_3,x_2|$\\
2681: B$[a^5,a^2iu]$&$|x_1,x_1,x_2,x_3,x_2|$\\
2682: B$[a^5,a^3iu]$&$|x_1,x_2,x_1,x_3,x_3|$\\
2683: B$[a^5,a^4iu]$&$|x_1,x_2,x_2,x_1,x_3|$\\
2684: \hline
2685: B$[a^6,iu]$&$|x_1,x_2,x_3,x_3,x_2,x_1|$\\
2686: B$[a^6,a^2iu]$&$|x_1,x_1,x_2,x_3,x_3,x_2|$\\
2687: B$[a^6,a^4iu]$&$|x_1,x_2,x_2,x_1,x_3,x_3|$\\
2688: \hline
2689: B$[a^6,i,a^3u]$&$|x_1,x_2,x_1,-x_1,-x_2,-x_1|$\\
2690: B$[a^6,a^2i,a^3u]$&$|x_1,-x_1,x_2,-x_1,x_1,-x_2|$\\
2691: B$[a^6,a^4i,a^3u]$&$|x_1,x_2,-x_2,-x_1,-x_2,x_2|$\\
2692: \hline
2693: B$[a^8,i,a^4u]$&$|x_1,x_2,x_2,x_1,-x_1,-x_2,-x_2,-x_1|$\\
2694: B$[a^8,a^2i,a^4u]$&$|x_1,-x_1,x_2,x_2,-x_1,x_1,-x_2,-x_2|$\\
2695: B$[a^8,a^4i,a^4u]$&$|x_1,x_2,-x_2,-x_1,-x_1,-x_2,x_2,x_1|$\\
2696: B$[a^8,a^6i,a^4u]$&$|x_1,x_1,x_2,-x_2,-x_1,-x_1,-x_2,x_2|$\\
2697: \hline
2698: B$[a^8,ai,a^4u]$&$|0,x_1,x_2,x_1,0,-x_1,-x_2,-x_1|$\\
2699: B$[a^8,a^3i,a^4u]$&$|x_1,0,-x_1,x_2,-x_1,0,x_1,-x_2|$\\
2700: B$[a^8,a^5i,a^4u]$&$|x_1,x_2,0,-x_2,-x_1,-x_2,0,x_2|$\\
2701: B$[a^8,a^7i,a^4u]$&$|x_1,x_2,x_1,0,-x_1,-x_2,-x_1,0|$\\
2702: \hline
2703: B$[a^{10},ai,a^5u]$&$|0,x_1,x_2,x_2,x_1,0,-x_1,-x_2,-x_2,-x_1|$\\
2704: B$[a^{10},a^3i,a^5u]$&$|x_1,0,-x_1,x_2,x_2,-x_1,0,x_1,-x_2,-x_2|$\\
2705: B$[a^{10},a^5i,a^5u]$&$|x_1,x_2,0,-x_2,-x_1,-x_1,-x_2,0,x_2,x_1|$\\
2706: B$[a^{10},a^7i,a^5u]$&$|x_1,x_1,x_2,0,-x_2,-x_1,-x_1,-x_2,0,x_2|$\\
2707: B$[a^{10},a^9i,a^5u]$&$|x_1,x_2,x_2,x_1,0,-x_1,-x_2,-x_2,-x_1,0|$\\
2708: \hline
2709: \end{tabular}
2710: \end{table}
2711:
2712:
2713: \clearpage
2714:
2715: \begin{thebibliography}{}
2716: \bibitem{FPU1}
2717: G.M.~Chechin, N.V.~Novikova, A.A.~Abramenko,
2718: Physica~D \textbf{166} (2002) 208.
2719: \bibitem{DAN1}
2720: V.P.~Sakhnenko, G.M.~Chechin,
2721: Dokl.\ Akad.\ Nauk \textbf{330} (1993) 308 [Phys.\ Dokl.\ \textbf{38}
2722: (1993) 219];
2723: \bibitem{DAN2}
2724: V.P.~Sakhnenko, G.M.~Chechin,
2725: Dokl.\ Akad.\ Nauk \textbf{338} (1994) 42 [Phys.\ Dokl.\ \textbf{39}
2726: (1994) 625].
2727: \bibitem{PhysD}
2728: G.M.~Chechin, V.P.~Sakhnenko,
2729: Physica~D \textbf{117} (1998) 43.
2730: \bibitem{Rosenberg}
2731: R.M.~Rosenberg,
2732: J.\ Appl.\ Mech.\ \textbf{29} (1962) 7.
2733: \bibitem{Okta}
2734: G.M.~Chechin, A.V.~Gnezdilov, M.Yu.~Zekhtser,
2735: Int.\ J.\ Non-Linear Mech.\ \textbf{38} (2003) 1451.
2736: \bibitem{PhysStatSol}
2737: G.M.~Chechin, T.I.~Ivanova, V.P.~Sakhnenko,
2738: Phys.\ Status Solidi (b) \textbf{152} (1989) 431.
2739: \bibitem{Acta}
2740: G.M.~Chechin, E.A.~Ipatova, V.P.~Sakhnenko,
2741: Acta Cryst.\ \textbf{A49} (1993) 824.
2742: \bibitem{NNM}
2743: G.M.~Chechin, V.P.~Sakhnenko, H.T.~Stokes, A.D.~Smith, D.M.~Hatch,
2744: Int.\ J.\ Non-Linear Mech.\ \textbf{35} (2000) 497.
2745: \bibitem{ENOC3}
2746: G.M.~Chechin, V.P.~Sakhnenko, M.Yu.~Zekhtser, H.T.~Stokes, S.~Carter,
2747: D.M.~Hatch,
2748: in \textit{World Wide Web Proceedings of the Third ENOC Conference},
2749: \url{http://www.midit.dtu.dk}.
2750: \bibitem{C60}
2751: G.M.~Chechin, O.A.~Lavrova, V.P.~Sakhnenko, H.T.~Stokes, D.M.~Hatch,
2752: Fizika tverdogo tela \textbf{44} (2002) 554.
2753: \bibitem{Comput}
2754: G.M.~Chechin,
2755: Comput.\ Math.\ Applic.\ \textbf{17} (1989) 255.
2756: \bibitem{Isotropy}
2757: H.T.~Stokes, D.M.~Hatch,
2758: The software package \texttt{ISOTROPY} (this package is available on the
2759: Internet at \url{http://www.physics.byu.edu/~stokesh/isotropy.html}).
2760: \bibitem{PoggiRuffo}
2761: P.~Poggi, S.~Ruffo,
2762: Physica~D \textbf{103} (1997) 251.
2763: \bibitem{BRink}
2764: B.~Rink,
2765: Physica~D \textbf{175} (2003) 31.
2766: \bibitem{Shinohara}
2767: S.~Shinohara,
2768: J.\ Phys.\ Soc.\ Japan \textbf{71} (2002) 1802;
2769: S.~Shinohara,
2770: Progr.\ Theor.\ Phys.\ Suppl.\ \textbf{150} (2003) 423.
2771: \bibitem{FPU2a}
2772: G.M.~Chechin, D.S.~Ryabov, K.G.~Zhukov,
2773: Construction of bushes of modes for the nonlinear monoatomic chains,
2774: Electronic journal ``Investigated in Russia'', \textbf{137} (2003)
2775: 1616-1644, \url{http://zhurnal.ape.relarn.ru/articles/2003/137.pdf}.
2776: \bibitem{Page}
2777: K.V.~Sandusky, J.B.~Page,
2778: Phys.\ Rev.\ B \textbf{50} (1994) 866.
2779: \bibitem{Crystal}
2780: V.V.~Krivtsova, G.M.~Chechin,
2781: Crystallography \textbf{35} (1990) 1319.
2782: \bibitem{FPU2b}
2783: G.M.~Chechin, D.S.~Ryabov, K.G.~Zhukov,
2784: Investigation the stability of one-dimensional and two-dimensional
2785: bushes of vibrational modes for the Fermi-Pasta-Ulam chains, Electronic
2786: journal ``Investigated in Russia'', \textbf{161} (2003) 1945-1964,
2787: \url{http://zhurnal.ape.relarn.ru/articles/2003/161.pdf}.
2788: \end{thebibliography}
2789:
2790:
2791:
2792: \clearpage
2793:
2794: \begin{table}
2795: \centering
2796: \caption{\label{table5} Linearized dynamical equations for studying
2797: stability of 1D bushes for the FPU-$\alpha$ chain. Here $\nu(t)$ is the
2798: vibrational mode representing a given bush, while $k=\frac{\pi j}{N}$
2799: determines the number $j$ of ``sleeping'' modes}
2800: \begin{tabular}{|l|}
2801: \hline
2802: B$[a^2,i]$: $\ddot\nu+4\nu=0$\\
2803: $\left\{\begin{array}{l}
2804: \ddot\nu_k+4\nu_k\sin^2k=-\frac{8\alpha}{N}\nu\nu_{\frac{\pi}{2}-k}\sin 2k\\
2805: \ddot\nu_{\frac{\pi}{2}-k}+4\nu_{\frac{\pi}{2}-k}\cos^2k=-\frac{8\alpha}{N}\nu\nu_k\sin 2k
2806: \end{array}\right.$\\
2807: \hline
2808: B$[a^3,a^2i]$: $\ddot\nu+3\nu=\frac{3\alpha\sqrt{6}}{2\sqrt{N}}\nu^2$\\
2809: $\left\{\begin{array}{l}
2810: \ddot\nu_k+4\nu_k\sin^2k=-\frac{\alpha\sqrt{6}}{\sqrt{N}}\nu
2811: \left[\left(1-2\cos\left(2k+\frac{\pi}{3}\right)\right)\nu_{\frac{\pi}{3}+k}+
2812: \left(1-2\sin\left(2k+\frac{\pi}{6}\right)\right)\nu_{\frac{2\pi}{3}+k}\right]\\
2813: \ddot\nu_{\frac{\pi}{3}+k}+4\nu_{\frac{\pi}{3}+k}\sin^2\left(k+\frac{\pi}{3}\right)=-\frac{\alpha\sqrt{6}}{\sqrt{N}}\nu
2814: \left[\left(1-2\cos\left(2k+\frac{\pi}{3}\right)\right)\nu_k+
2815: \left(1+2\cos 2k\right)\nu_{\frac{2\pi}{3}+k}\right]\\
2816: \ddot\nu_{\frac{2\pi}{3}+k}+4\nu_{\frac{2\pi}{3}+k}\cos^2\left(k+\frac{\pi}{6}\right)=-\frac{\alpha\sqrt{6}}{\sqrt{N}}\nu
2817: \left[\left(1-2\sin\left(2k+\frac{\pi}{6}\right)\right)\nu_k+
2818: \left(1+2\cos 2k\right)\nu_{\frac{\pi}{3}+k}\right]\\
2819: \end{array}\right.$\\
2820: \hline
2821: B$[a^4,a^3i]$: $\ddot\nu+2\nu=0$\\
2822: $\left\{\begin{array}{l}
2823: \ddot\nu_k+4\nu_k\sin^2k=-\frac{2\alpha\sqrt{2}}{\sqrt{N}}\nu
2824: \left[\left(1-\sqrt{2}\cos\left(2k+\frac{\pi}{4}\right)\right)\nu_{\frac{\pi}{4}+k}+
2825: \left(1-\sqrt{2}\sin\left(2k+\frac{\pi}{4}\right)\right)\nu_{\frac{3\pi}{4}+k}\right]\\
2826: \ddot\nu_{\frac{\pi}{4}+k}+4\nu_{\frac{\pi}{4}+k}\sin^2\left(k+\frac{\pi}{4}\right)=-\frac{2\alpha\sqrt{2}}{\sqrt{N}}\nu
2827: \left[\left(1-\sqrt{2}\cos\left(2k+\frac{\pi}{4}\right)\right)\nu_k+
2828: \left(1+\sqrt{2}\sin\left(2k+\frac{\pi}{4}\right)\right)\nu_{\frac{\pi}{2}+k}\right]\\
2829: \ddot\nu_{\frac{\pi}{2}+k}+4\nu_{\frac{\pi}{2}+k}\cos^2k=-\frac{2\alpha\sqrt{2}}{\sqrt{N}}\nu
2830: \left[\left(1+\sqrt{2}\sin\left(2k+\frac{\pi}{4}\right)\right)\nu_{\frac{\pi}{4}+k}+
2831: \left(1+\sqrt{2}\cos\left(2k+\frac{\pi}{4}\right)\right)\nu_{\frac{3\pi}{4}+k}\right]\\
2832: \ddot\nu_{\frac{3\pi}{4}+k}+4\nu_{\frac{3\pi}{4}+k}\cos^2\left(k+\frac{\pi}{4}\right)=-\frac{2\alpha\sqrt{2}}{\sqrt{N}}\nu
2833: \left[\left(1-\sqrt{2}\sin\left(2k+\frac{\pi}{4}\right)\right)\nu_k+
2834: \left(1+\sqrt{2}\cos\left(2k+\frac{\pi}{4}\right)\right)\nu_{\frac{\pi}{2}+k}\right]\\
2835: \end{array}\right.$\\
2836: \hline
2837: \end{tabular}
2838: \end{table}
2839:
2840: \begin{table}
2841: \centering
2842: \caption{\label{table6} Linearized dynamical equations for studying
2843: stability of 1D bushes for the FPU-$\beta$ chain}
2844: \begin{tabular}{|l|}
2845: \hline
2846: B$[a^2,i]$: $\ddot\nu+4\nu=-\frac{16\beta}{N}\nu^3$\\
2847: $\left\{\begin{array}{l}
2848: \ddot\nu_k+4\left(1+\frac{12\beta}{N}\nu^2\right)\nu_k\sin^2k=0
2849: \end{array}\right.$\\
2850: \hline
2851: B$[a^3,a^2i]$: $\ddot\nu+3\nu=-\frac{27\beta}{2N}\nu^3$\\
2852: $\left\{\begin{array}{l}
2853: \ddot\nu_k+4\nu_k\sin^2k=-\frac{18\beta}{N}\nu^2\sin k
2854: \left[2\nu_k\sin k-\nu_{\frac{\pi}{3}+k}\sin\left(k+\frac{\pi}{3}\right)+\nu_{\frac{2\pi}{3}+k}\sin\left(k+\frac{2\pi}{3}\right)\right]\\
2855: \ddot\nu_{\frac{\pi}{3}+k}+4\nu_{\frac{\pi}{3}+k}\sin^2\left(k+\frac{\pi}{3}\right)=-\frac{18\beta}{N}\nu^2\sin\left(k+\frac{\pi}{3}\right)
2856: \left[-\nu_k\sin k+2\nu_{\frac{\pi}{3}+k}\sin\left(k+\frac{\pi}{3}\right)-\nu_{\frac{2\pi}{3}+k}\sin\left(k+\frac{2\pi}{3}\right)\right]\\
2857: \ddot\nu_{\frac{2\pi}{3}+k}+4\nu_{\frac{2\pi}{3}+k}\sin^2\left(k+\frac{2\pi}{3}\right)=-\frac{18\beta}{N}\nu^2\sin\left(k+\frac{2\pi}{3}\right)
2858: \left[\nu_k\sin k-\nu_{\frac{\pi}{3}+k}\sin\left(k+\frac{\pi}{3}\right)+2\nu_{\frac{2\pi}{3}+k}\sin\left(k+\frac{2\pi}{3}\right)\right]\\
2859: \end{array}\right.$\\
2860: \hline
2861: B$[a^4,a^3i]$: $\ddot\nu+2\nu=-\frac{4\beta}{N}\nu^3$\\
2862: $\left\{\begin{array}{l}
2863: \ddot\nu_k+4\left(1+\frac{6\beta}{N}\nu^2\right)\nu_k\sin^2k=0
2864: \end{array}\right.$\\
2865: \hline
2866: B$[a^3,a^2iu]$: $\ddot\nu+3\nu=-\frac{27\beta}{2N}\nu^3$\\
2867: $\left\{\begin{array}{l}
2868: \ddot\nu_k+4\nu_k\sin^2k=-\frac{18\beta}{N}\nu^2\sin k
2869: \left[2\nu_k\sin k+\nu_{\frac{\pi}{3}+k}\sin\left(k+\frac{\pi}{3}\right)-\nu_{\frac{2\pi}{3}+k}\sin\left(k+\frac{2\pi}{3}\right)\right]\\
2870: \ddot\nu_{\frac{\pi}{3}+k}+4\nu_{\frac{\pi}{3}+k}\sin^2\left(k+\frac{\pi}{3}\right)=-\frac{18\beta}{N}\nu^2\sin\left(k+\frac{\pi}{3}\right)
2871: \left[\nu_k\sin k+2\nu_{\frac{\pi}{3}+k}\sin\left(k+\frac{\pi}{3}\right)+\nu_{\frac{2\pi}{3}+k}\sin\left(k+\frac{2\pi}{3}\right)\right]\\
2872: \ddot\nu_{\frac{2\pi}{3}+k}+4\nu_{\frac{2\pi}{3}+k}\sin^2\left(k+\frac{2\pi}{3}\right)=-\frac{18\beta}{N}\nu^2\sin\left(k+\frac{2\pi}{3}\right)
2873: \left[-\nu_k\sin k+\nu_{\frac{\pi}{3}+k}\sin\left(k+\frac{\pi}{3}\right)+2\nu_{\frac{2\pi}{3}+k}\sin\left(k+\frac{2\pi}{3}\right)\right]\\
2874: \end{array}\right.$\\
2875: \hline
2876: B$[a^4,iu]$: $\ddot\nu+2\nu=-\frac{8\beta}{N}\nu^3$\\
2877: $\left\{\begin{array}{l}
2878: \ddot\nu_k+4\left(1+\frac{6\beta}{N}\nu^2\right)\nu_k\sin^2k=-\frac{12\beta}{N}\nu^2\nu_{\frac{\pi}{2}-k}\sin 2k\\
2879: \ddot\nu_{\frac{\pi}{2}-k}+4\left(1+\frac{6\beta}{N}\nu^2\right)\nu_{\frac{\pi}{2}-k}\cos^2k=-\frac{12\beta}{N}\nu^2\nu_k\sin 2k\\
2880: \end{array}\right.$\\
2881: \hline
2882: B$[a^6,a^5i,a^3u]$: $\ddot\nu+\nu=-\frac{3\beta}{2N}\nu^3$\\
2883: $\left\{\begin{array}{l}
2884: \ddot\nu_k+4\nu_k\sin^2k=-\frac{6\beta}{N}\nu^2\sin k
2885: \left[2\nu_k\sin k+\nu_{\frac{\pi}{3}+k}\sin\left(k+\frac{\pi}{3}\right)-\nu_{\frac{2\pi}{3}+k}\sin\left(k+\frac{2\pi}{3}\right)\right]\\
2886: \ddot\nu_{\frac{\pi}{3}+k}+4\nu_{\frac{\pi}{3}+k}\sin^2\left(k+\frac{\pi}{3}\right)=-\frac{6\beta}{N}\nu^2\sin\left(k+\frac{\pi}{3}\right)
2887: \left[\nu_k\sin k+2\nu_{\frac{\pi}{3}+k}\sin\left(k+\frac{\pi}{3}\right)+\nu_{\frac{2\pi}{3}+k}\sin\left(k+\frac{2\pi}{3}\right)\right]\\
2888: \ddot\nu_{\frac{2\pi}{3}+k}+4\nu_{\frac{2\pi}{3}+k}\sin^2\left(k+\frac{2\pi}{3}\right)=-\frac{6\beta}{N}\nu^2\sin\left(k+\frac{2\pi}{3}\right)
2889: \left[-\nu_k\sin k+\nu_{\frac{\pi}{3}+k}\sin\left(k+\frac{\pi}{3}\right)+2\nu_{\frac{2\pi}{3}+k}\sin\left(k+\frac{2\pi}{3}\right)\right]\\
2890: \end{array}\right.$\\
2891: \hline
2892: \end{tabular}
2893: \end{table}
2894:
2895: \begin{table}
2896: \centering
2897: \caption{\label{table7} Thresholds of bush stability for the FPU-$\alpha$
2898: and FPU-$\beta$ chains with $N=12$. Here we give the maximal value of the
2899: displacement $x$, as well as the maximal energy for which the corresponding
2900: bushes lose their stability}
2901: \begin{tabular}{|c|c|c|c|c|c|}
2902: \hline
2903: Bush&Displacement pattern&\multicolumn{4}{c}{Threshold of the bush stability}\vline\\
2904: \cline{3-6}
2905: &&\multicolumn{2}{c}{in displacement}\vline&\multicolumn{2}{c}{in energy (per atom)}\vline\\
2906: \cline{3-6}
2907: &&FPU-$\alpha$&FPU-$\beta$&FPU-$\alpha$&FPU-$\beta$\\
2908: \hline
2909: B$[a^2,i]$&$|x,-x|$&$0.302$&$0.112$&$0.183$&$0.025$\\
2910: \hline
2911: B$[a^3,i]$&$|x,0,-x|$&$0.203$&$0.268$&$0.047$&$0.085$\\
2912: \hline
2913: B$[a^4,ai]$&$|0,x,0,-x|$&$0$&$1.161$&$0$&$0.675$\\
2914: \hline
2915: B$[a^4,i,a^2u]$&$|x,x,-x,-x|$&&$>20$&&$>10000$\\
2916: \hline
2917: B$[a^6,ai,a^3u]$&$|x,0,-x,-x,0,x|$&&$0.488$&&$0.079$\\
2918: \hline
2919: B$[a^3,iu]$&$|x,-2x,x|$&&$0.157$&&$0.074$\\
2920: \hline
2921: \end{tabular}
2922: \end{table}
2923:
2924:
2925: \clearpage
2926:
2927: \begin{figure}
2928: \centering
2929: \setlength{\unitlength}{1mm}
2930: \begin{picture}(130,45)(0,0)
2931: \put(5,5){\includegraphics[width=120mm,height=36mm]{aB2ai.eps}}
2932: \drawgridB
2933: \put(65,5){\line(0,1){36}}% 6
2934: \end{picture}
2935: \caption{\label{fig6} Regions of stability (white color) of different modes
2936: of the FPU-$\alpha$ chain, interacting parametrically with the
2937: one-dimensional bush B$[a^2,i]$.}
2938: \end{figure}
2939:
2940: \begin{figure}
2941: \centering
2942: \setlength{\unitlength}{1mm}
2943: \begin{picture}(130,45)(0,0)
2944: \put(5,5){\includegraphics[width=120mm,height=36mm]{aB3ai.eps}}
2945: \drawgridA
2946: \put(45,5){\line(0,1){36}}% 4
2947: \put(85,5){\line(0,1){36}}% 8
2948: \end{picture}
2949: \caption{\label{fig7} Regions of stability (white color) of different modes
2950: of the FPU-$\alpha$ chain, interacting parametrically with the
2951: one-dimensional bush B$[a^3,i]$.}
2952: \end{figure}
2953:
2954: \begin{figure}
2955: \centering
2956: \setlength{\unitlength}{1mm}
2957: \begin{picture}(130,45)(0,0)
2958: \put(5,5){\includegraphics[width=120mm,height=36mm]{aB4aai.eps}}
2959: \drawgridA
2960: \put(35,5){\line(0,1){36}}% 3
2961: \put(95,5){\line(0,1){36}}% 9
2962: \end{picture}
2963: \caption{\label{fig8} Regions of stability (white color) of different modes
2964: of the FPU-$\alpha$ chain, interacting parametrically with the
2965: one-dimensional bush B$[a^4,ai]$.}
2966: \end{figure}
2967:
2968:
2969: \begin{figure}
2970: \centering
2971: \setlength{\unitlength}{1mm}
2972: \begin{picture}(130,45)(0,0)
2973: \put(5,5){\includegraphics[width=120mm,height=36mm]{bB2ai.eps}}
2974: \drawgrid
2975: \put(65,5){\line(0,1){36}}% 6
2976: \end{picture}
2977: \caption{\label{fig9} Regions of stability (white color) of different modes
2978: of the FPU-$\beta$ chain, interacting parametrically with the
2979: one-dimensional bush B$[a^2,i]$.}
2980: \end{figure}
2981:
2982: \begin{figure}
2983: \centering
2984: \setlength{\unitlength}{1mm}
2985: \begin{picture}(130,45)(0,0)
2986: \put(5,5){\includegraphics[width=120mm,height=36mm]{bB3ai.eps}}
2987: \drawgridB
2988: \put(45,5){\line(0,1){36}}% 4
2989: \put(85,5){\line(0,1){36}}% 8
2990: \put( 5, 8.9){\dashbox{1}(120,0){}}% A1
2991: \put(20, 9.1){$A_1$}
2992: \put( 5,16.7){\dashbox{1}(120,0){}}% A2
2993: \put(20,16.9){$A_2$}
2994: \put( 5,21.5){\dashbox{1}(120,0){}}% A3
2995: \put(20,21.7){$A_3$}
2996: \end{picture}
2997: \caption{\label{fig10} Regions of stability (white color) of different
2998: modes of the FPU-$\beta$ chain, interacting parametrically with the
2999: one-dimensional bush B$[a^3,i]$.}
3000: \end{figure}
3001:
3002: \begin{figure}
3003: \centering
3004: \setlength{\unitlength}{1mm}
3005: \begin{picture}(130,45)(0,0)
3006: \put(5,5){\includegraphics[width=120mm,height=36mm]{bB4aai.eps}}
3007: \drawgrid
3008: \put(35,5){\line(0,1){36}}% 3
3009: \put(95,5){\line(0,1){36}}% 9
3010: \put( 5,15.0){\dashbox{1}(120,0){}}% A1
3011: \put(20,15.2){$A_1$}
3012: \end{picture}
3013: \caption{\label{fig12} Regions of stability (white color) of different
3014: modes of the FPU-$\beta$ chain, interacting parametrically with the
3015: one-dimensional bush B$[a^4,ai]$.}
3016: \end{figure}
3017:
3018: \begin{figure}
3019: \centering
3020: \setlength{\unitlength}{1mm}
3021: \begin{picture}(130,45)(0,0)
3022: \put(5,5){\includegraphics[width=120mm,height=36mm]{bB3aiu.eps}}
3023: \drawgrid
3024: \put(45,5){\line(0,1){36}}% 4
3025: \put(85,5){\line(0,1){36}}% 8
3026: \put( 5, 7.8){\dashbox{1}(120,0){}}% A1
3027: \put(20, 8.0){$A_1$}
3028: \put( 5,17.0){\dashbox{1}(120,0){}}% A2
3029: \put(20,17.2){$A_2$}
3030: \put( 5,24.2){\dashbox{1}(120,0){}}% A3
3031: \put(20,24.4){$A_3$}
3032: \end{picture}
3033: \caption{\label{fig11} Regions of stability (white color) of different
3034: modes of the FPU-$\beta$ chain, interacting parametrically with the
3035: one-dimensional bush B$[a^3,iu]$.}
3036: \end{figure}
3037:
3038: \begin{figure}
3039: \centering
3040: \setlength{\unitlength}{1mm}
3041: \begin{picture}(130,45)(0,0)
3042: \put(5,5){\includegraphics[width=120mm,height=36mm]{bB4aiu.eps}}
3043: \drawgridB
3044: \put(35,5){\line(0,1){36}}% 3
3045: \put(95,5){\line(0,1){36}}% 9
3046: \put( 5,26.0){\dashbox{1}(120,0){}}% A1
3047: \put(20,26.2){$A_1$}
3048: \end{picture}
3049: \caption{\label{fig13} Regions of stability (white color) of different
3050: modes of the FPU-$\beta$ chain, interacting parametrically with the
3051: one-dimensional bush B$[a^4,iu]$.}
3052: \end{figure}
3053:
3054: \begin{figure}
3055: \centering
3056: \setlength{\unitlength}{1mm}
3057: \begin{picture}(130,45)(0,0)
3058: \put(5,5){\includegraphics[width=120mm,height=36mm]{bB6aaia3u.eps}}
3059: \drawgridB
3060: \put( 25,5){\line(0,1){36}}% 2
3061: \put(105,5){\line(0,1){36}}% 10
3062: \end{picture}
3063: \caption{\label{fig14} Regions of stability (white color) of different
3064: modes of the FPU-$\beta$ chain, interacting parametrically with the
3065: one-dimensional bush B$[a^6,ai,a^3u]$.}
3066: \end{figure}
3067:
3068:
3069: \clearpage
3070:
3071: \begin{figure}
3072: \centering
3073: \setlength{\unitlength}{1mm}
3074: \includegraphics[width=90mm]{2DaB4ai_energy.eps}
3075: \caption{\label{fig77} Stability diagram of the bush B$[a^4,i]$ in the
3076: FPU-$\alpha$ chain with $N=12$ for the case $\dot\nu_1(0)=0$,
3077: $\dot\nu_2(0)=0$.}
3078: \end{figure}
3079:
3080: \begin{figure}
3081: \centering
3082: \setlength{\unitlength}{1mm}
3083: \includegraphics[width=90mm]{2DaB4ai_wind.eps}
3084: \caption{\label{fig78} Stability diagram of the bush B$[a^4,i]$ in the
3085: FPU-$\alpha$ chain with $N=12$ for the case $\dot\nu_1(0)=0.2$,
3086: $\dot\nu_2(0)=0.1$.}
3087: \end{figure}
3088:
3089:
3090: \twocolumn
3091:
3092: \begin{figure}
3093: \centering
3094: \setlength{\unitlength}{1mm}
3095: \includegraphics[width=60mm]{2DaB3a.eps}
3096: \caption{\label{fig15} Stability diagram for the bush B$[a^3]$ in the
3097: FPU-$\alpha$ chain.}
3098: \end{figure}
3099:
3100: \begin{figure}
3101: \centering
3102: \setlength{\unitlength}{1mm}
3103: \includegraphics[width=60mm]{2DaB4ai.eps}
3104: \caption{\label{fig16} Stability diagram for the bush B$[a^4,i]$ in the
3105: FPU-$\alpha$ chain.}
3106: \end{figure}
3107:
3108: \begin{figure}
3109: \centering
3110: \setlength{\unitlength}{1mm}
3111: \includegraphics[width=60mm]{2DaB6aai.eps}
3112: \caption{\label{fig17} Stability diagram for the bush B$[a^6,ai]$ in the
3113: FPU-$\alpha$ chain.}
3114: \end{figure}
3115:
3116:
3117:
3118: \begin{figure}
3119: \centering
3120: \setlength{\unitlength}{1mm}
3121: \includegraphics[width=60mm]{2DbB3a.eps}
3122: \caption{\label{fig18} Stability diagram for the bush B$[a^3]$ in the
3123: FPU-$\beta$ chain.}
3124: \end{figure}
3125:
3126: \begin{figure}
3127: \centering
3128: \setlength{\unitlength}{1mm}
3129: \includegraphics[width=60mm]{2DbB4ai.eps}
3130: \caption{\label{fig19} Stability diagram for the bush B$[a^4,i]$ in the
3131: FPU-$\beta$ chain.}
3132: \end{figure}
3133:
3134: \begin{figure}
3135: \centering
3136: \setlength{\unitlength}{1mm}
3137: \includegraphics[width=60mm]{2DbB6aai.eps}
3138: \caption{\label{fig22} Stability diagram for the bush B$[a^6,ai]$ in the
3139: FPU-$\beta$ chain.}
3140: \end{figure}
3141:
3142:
3143: \clearpage
3144: \begin{figure}
3145: \centering
3146: \setlength{\unitlength}{1mm}
3147: \includegraphics[width=60mm]{2DbB4aa2u.eps}
3148: \caption{\label{fig20} Stability diagram for the bush B$[a^4,a^2u]$ in the
3149: FPU-$\beta$ chain.}
3150: \end{figure}
3151:
3152: \begin{figure}
3153: \centering
3154: \setlength{\unitlength}{1mm}
3155: \includegraphics[width=60mm]{2DbB4aaiu.eps}
3156: \caption{\label{fig21} Stability diagram for the bush B$[a^4,aiu]$ in the
3157: FPU-$\beta$ chain.}
3158: \end{figure}
3159:
3160: ~
3161:
3162: ~
3163:
3164: ~
3165:
3166: ~
3167:
3168: ~
3169:
3170: ~
3171:
3172: ~
3173:
3174: \begin{figure}\centering
3175: \setlength{\unitlength}{1mm}
3176: \includegraphics[width=60mm]{2DbB6aiu.eps}
3177: \caption{\label{fig23} Stability diagram for the bush B$[a^6,iu]$ in the
3178: FPU-$\beta$ chain.}
3179: \end{figure}
3180:
3181: \begin{figure}
3182: \centering
3183: \setlength{\unitlength}{1mm}
3184: \includegraphics[width=60mm]{2DbB6aia3u.eps}
3185: \caption{\label{fig24} Stability diagram for the bush B$[a^6,i,a^3u]$ in
3186: the FPU-$\beta$ chain.}
3187: \end{figure}
3188:
3189: ~
3190:
3191: ~
3192:
3193: ~
3194:
3195: ~
3196:
3197: ~
3198:
3199: ~
3200:
3201: ~
3202:
3203: \clearpage
3204:
3205: \begin{figure}
3206: \centering
3207: \setlength{\unitlength}{1mm}
3208: \includegraphics[width=60mm]{2DaB4ai_8.eps}
3209: \caption{\label{fig26} Bush B$[a^4,i]$ in FPU-$\alpha$ with $N=8$.}
3210: \end{figure}
3211:
3212: \begin{figure}
3213: \centering
3214: \setlength{\unitlength}{1mm}
3215: \includegraphics[width=60mm]{2DaB4ai_12.eps}
3216: \caption{\label{fig27} Bush B$[a^4,i]$ in FPU-$\alpha$ with $N=12$.}
3217: \end{figure}
3218:
3219: \begin{figure}
3220: \centering
3221: \setlength{\unitlength}{1mm}
3222: \includegraphics[width=60mm]{2DaB4ai_16.eps}
3223: \caption{\label{fig28} Bush B$[a^4,i]$ in FPU-$\alpha$ with $N=16$.}
3224: \end{figure}
3225:
3226: \begin{figure}
3227: \centering
3228: \setlength{\unitlength}{1mm}
3229: \includegraphics[width=60mm]{2DaB4ai_20.eps}
3230: \caption{\label{fig29} Bush B$[a^4,i]$ in FPU-$\alpha$ with $N=20$.}
3231: \end{figure}
3232:
3233: \begin{figure}
3234: \centering
3235: \setlength{\unitlength}{1mm}
3236: \includegraphics[width=60mm]{2DaB4ai_24.eps}
3237: \caption{\label{fig30} Bush B$[a^4,i]$ in FPU-$\alpha$ with $N=24$.}
3238: \end{figure}
3239:
3240: \begin{figure}
3241: \centering
3242: \setlength{\unitlength}{1mm}
3243: \includegraphics[width=60mm]{2DaB4ai_48.eps}
3244: \caption{\label{fig31} Bush B$[a^4,i]$ in FPU-$\alpha$ with $N=48$.}
3245: \end{figure}
3246:
3247: \end{document}
3248: