1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: \documentclass[twocolumn,english,aps,pre,showpacs]{revtex4}
3: \usepackage[T1]{fontenc}
4: \usepackage[latin1]{inputenc}
5: \usepackage{float}
6: \usepackage{amsmath}
7: \usepackage{graphicx}
8: \usepackage{amssymb}
9:
10: \makeatletter
11:
12:
13: \usepackage{babel}
14: \makeatother
15: \begin{document}
16:
17: \preprint{preprint}
18:
19:
20: \title{Phonon modes in the Frenkel-Kontorova chain:\\
21: exponential localization and the number theory properties of frequency bands}
22:
23:
24: \author{O.V. Zhirov}
25: \email{zhirov@inp.nsk.su}
26: \affiliation{Budker Institute of Nuclear Physics, 630090 Novosibirsk, Russia}
27:
28: \author{G. Casati}
29: \email{giulio.casati@uninsubria.it}
30: \affiliation{International Center for the Study of Dynamical
31: Systems, Universit\`a degli Studi dell'Insubria and\\
32: Istituto Nazionale per la Fisica della Materia,
33: Unit\`a di Como, Via Valleggio 11, 22100 Como, Italy and\\
34: Istituto Nazionale di Fisica Nucleare,
35: Sezione di Milano, Via Celoria 16, 20133 Milano, Italy}
36:
37: \author{D.L. Shepelyansky}
38: \email{dima@irsamc.ups-tlse.fr}
39: \affiliation{Laboratoire de Physique Quantique, UMR 5626 du CNRS, Universit\'{e}
40: Paul Sabatier, 31062 Toulouse, France}
41:
42:
43: \date{\today{}}
44:
45: \begin{abstract}
46: We study numerically phonon modes of the classical one-dimensional
47: Frenkel-Kontorova chain, in the regime of pinned phase characterized
48: by the phonon gap and devil's staircase, as well as by a large number
49: of states (configurational excitations), which energy splitting from the ground
50: state is exponentially small. We demonstrate, these states behave like
51: disorder media: their phonon modes are {\it exponentially} localized,
52: in contrast to the phonon modes in the ground state, where phonons are
53: {\it prelocalized} only.
54:
55: We demonstrate also, the phonon frequency spectrum of the ground state has an
56: hierarchical structure, a direct manifestation of hierarchical spatial structure,
57: found for the ground state of the FK chain in our recent work.
58: \end{abstract}
59: %\pacs{05.45.-a, 63.70.+h, 61.44.Fw }
60: \pacs{PACS numbers: 05.45.Mt}
61: \maketitle
62:
63:
64:
65: \section{Introduction}
66:
67: The most trivial disorder originates in media due to random static
68: impurities (see, e.g. \cite{Pastur}). However, another but very interesting
69: possibilities are glasses, which have a huge number of (meta)stable
70: degenerated states. Originally glassy system has
71: a homogeneous Hamiltonian with no intrinsic random parameter, and disorder
72: occurs in it dynamically. Recently \cite{Zh01} we have demonstrated that a popular
73: Frenkel-Kontorova model \cite{FK38} presents an example of glassy system,
74: which has a lot of static states, known as {\it configurational excitations}
75: of the classical ground state, with energy splitting extremely (exponentially)
76: small. As it was shown in \cite{Zh03}, this model has a nontrivial quantum
77: dynamics, the quantum phase transition: if quantum parameter exceeds some
78: critical value, the "pinned" glassy phase turns into "sliding" phonon gas.
79:
80:
81: The Frenkel-Kontorova model (FK) \cite{FK38} is widely used
82: \cite{Naba67,Pokr84,Ying71,Piet81,Au78,Au83a,Flor96,Weis96,Brau97,Cons00}
83: in the solid state physics to get insight on generic properties of
84: noncommesurate systems. Its ground state, which is rather quasiperiodical
85: \cite{Au78,Au83b,Zh01} than periodical, attracts also an attention
86: \cite{Au78,Au83a,Au90,Burk96,Hu00} as some interplay \cite{Burk96}
87: between order and disorder \cite{Sin77}.
88:
89: This model describes a chain of atoms/particles interacting with elastic
90: forces, placed in periodic potential, which period differs from a
91: mean interparticle distance. The ground state (GS) of this model is
92: defined as a static, equilibrium configuration of the chain, that
93: corresponds to the {\it absolute} minimum of the chain potential
94: energy. The ground state is unique and has some special order of particles
95: in the chain, that was discovered by Aubry \cite{Au78,Au83b} more
96: than twenty years ago. The positions of atoms in the chain are described
97: by an area preserving map, which is well known in the field of dynamical
98: chaos as the Chirikov standard map \cite{Chir79}. The ratio of the
99: mean interparticle distance to a period of the external potential
100: in the FK model determines the rotation number of the invariant curves
101: of the map, while the amplitude of the periodic potential gives the
102: value of the dimensionless parameter $K$. For $ K<K_{c} $
103: the KAM curves are smooth and the spectrum of longwave phonon excitations
104: in the chain is characterized by a linear dispersion law starting
105: from zero frequency. In this regime the chain can freely slide along
106: the external field (the ``sliding'' phase). On the contrary, for
107: $ K>K_{c} $ the KAM curves are destroyed and replaced by an invariant
108: Cantor set, which is called cantorus. In this regime the phonon spectrum
109: has a gap, and the chain is pinned (``pinned'' phase). Later,
110: on the example of Ising spin model to which the FK model can be approximately
111: reduced \cite{Be80} it has been shown \cite{Vall86} that the GS
112: has some well defined hierarchical structure, which particular detailes
113: are determined by number properties of the ratio of the mean interparticle
114: distance to the period of the external field. Recently our numerical
115: study \cite{Zh01} of the original FK model in the pinned phase has shown,
116: that the GS has indeed an hierarchical structure, but in some important
117: detailes different from predicted in \cite{Vall86}.
118:
119: In short, we put our attention to the striking fact, that
120: in the pinned phase of the FK chain there are some particles, which positions
121: exponentially close to bottoms of wells of the external potential:
122: corresponding external force acting to such particle is extremely close to zero.
123: Obviously, in static equilibrium each of these particles can be considered as some
124: dummy ``glue'' that only couples two adjacent parts of the chain,
125: which ends has almost identical (exponentially close) tension forces.
126:
127: Another important observation is that small deviations of glue particles from
128: their well bottoms are {\it groupped}
129: into well defined hierarchically ordered scales. Now, if one cut the
130: chain into fragments via glue particles, which belong to level with
131: least deviation from well bottoms, one gets several fragments of {\it two}
132: sizes, or two species of some ``bricks''\footnote{%
133: The number of brick species depends on the number properties of the
134: ratio of a mean interparticle distance to a period of the external
135: field, the chain ``rotation number'' $\nu $. The most simple picture
136: quoted here assumes that this rotational number approximates the golden
137: mean value.
138: }. Then one may repeat the procedure, cutting ``bricks'' via ``glue''
139: particles that belong to next scale of deviations and getting two
140: new species of smaller bricks, and so far. At the last step one gets
141: two species of smallest possible bricks with {\it no} glue particle
142: inside: smaller bricks $ A^{(0)} $ which consist of 2 particles
143: inside a single well, and larger brick $ B^{(0)} $ which consist
144: of 4 particles (two pairs in two adjacent wells). In this way one
145: gets an hierarchically ordered set of brick species
146: $ \left\{ A^{(i)},B^{(i)}\right\} $
147: with very simple composition rules \cite{Zh01}:
148:
149: \begin{eqnarray}
150: A^{(i+1)} & = & B^{(i)}gA^{(i)}gB^{(i)},\label{cr1} \\
151: B^{(i+1)} & = & B^{(i)}gA^{(i)}gB^{(i)}gA^{(i)}gB^{(i)},\label{cr2}
152: \end{eqnarray}
153:
154: where symbol $ g $ denotes of an insertion of glue particle, which
155: ``glues'' two adjacent bricks. The difference of tension forces
156: at boundaries of bricks $ A^{(i)} $ and $ B^{(i)} $ is exponentially
157: small and decrease rapidly with a number of the hierarcy level. In
158: principle, these rules are sufficient to construct a GS for a FK chain
159: of any length, if the rotation number parameter of the chain approximates
160: the mean golden value $ \nu =(\sqrt{5}-1)/2 $.
161:
162: Besides the GS there exists ``configurational excitation states''
163: (CES), presented by static equilibrium configurations corresponding
164: to {\it local} (rather than absolute) minima of the chain potential,
165: with energy very close to GS. Within the picture just outlined above
166: CES correspond to different permutations of bricks \cite{Zh01},
167: therefore the number of them can be combinatorically huge.
168: At any accessible small temperature their contributions can dominate
169: over the contribution of GS.
170:
171: In this paper we address to phonon excitations of the chain, small
172: vibrations around static GS and CES configurations of the chain. These
173: excitations are relevant as for heat transport properties \cite{Tong99,Hu00}
174: of the chain, as for some quantum effects \cite{Hu01}, especially
175: in the quasiclassical limit. As in the previous paper, we concentarte
176: on the case of pinned phase of the chain, which corresponds to a nonzero
177: phonon gap. We start with analysis of the structure of the phonon
178: frequency spectrum in the GS. It is well known, that this spectrum
179: is splitted into bands \cite{Au83a,Pe83} but, to our knowledge, up
180: to now there is no clear explanation for origin of its splitting into
181: particular bands and subbands. We show, that this splitting is a direct
182: consequence of particular spatial structure of the chain in its GS.
183: We have found, that this structure is also hierarchically ordered,
184: with definite resemblance and distinctions with respect to a spatial
185: structure of underlying GS.
186:
187: Localisation properties of phonons in incommensurate one-dimentional
188: chains are intesively studied in recent works \cite{Burk96,Keto97,Tong99,Hu00},
189: with strong indications \cite{Keto97,Tong99}, that phonon modes in
190: the GS of FK chain are not localized, and even at edges of frequency
191: bands they are rather {\it prelocalized}, than localized. We study
192: also phonon properties as in GS, as in CES of the chain. Our results
193: confirm, that in the GS phonon modes are only prelocalized. However,
194: the situation appears to be quite different for CES. Even for CES,
195: which energy splitting (in natural problem scale) $ \Delta U\leq 10^{-12} $
196: there are phonon modes, which are localized exponentially. Moreover,
197: for CES with higher splitting \( \Delta U \) we see, that there are
198: entire bands of exponentially localized phonon modes.
199:
200:
201: \section{The model.}
202:
203:
204: The Hamiltonian of the FK model is
205:
206: \begin{equation}
207: \label{HFK}
208: H=\sum _{i=1}^{s}\frac{P_{i}^{2}}{2}+\frac{(x_{i}-x_{i-1})^{2}}{2}-K\cos x_{i}.
209: \end{equation}
210:
211: The first term in the Hamiltonian is a kinetic energy, where we put
212: masses of particles $ m=1 $, the second term describes interparticle
213: interaction with elasticity coefficient put to unity, while the third
214: term corresponds to particle interaction with external periodical
215: field with coupling constant $ K $. All $ s $ particles are
216: distributed over $ r $ period/wells of the external potential,
217: which period, without any loss of generality is taken equal to $ 2\pi $.
218: The ratio $ \nu =r/s $ gives \cite{Au78} the rotational number
219: of corresponding standard map \cite{Chir79}.
220:
221: We assume periodical boundary conditions: $ P_{0}\equiv P_{s} $,
222: $ x_{0}\equiv x_{s}-L $, where $ L=2\pi r $ is the length of
223: the chain. In our subsequent analysis we take (as some {\it typical}
224: example of FK chain) the chain with $ r/s=377/610 $ as an approximation
225: of the golden mean value $ \bar{\nu }=(\sqrt{5}-1)/2 $, and parameter
226: $ K=2 $ well above the critical value
227: $ K_{c}(\nu =\bar{\nu })=0.971\ldots $\cite{Gree79}.
228: The technique to obtain GS and CES is described in our previous work
229: \cite{Zh01}. Let us stress, that validity of the GS can be checked
230: either by direct analysis \cite{Zh01} of its spatial structure, or
231: proved by monotonous behavior \cite{Au83b} of its hall function.
232: Note, that the GS is almost degenerate with enormous number of CES: there are
233: hundreds of states with $ \Delta U=U_{ECS}-U_{GS}\leq 10^{-80} $,
234: while the number of ECS with $ \Delta U\leq 10^{-12} $ exceeds
235: $ 10^{9} $!
236:
237: Effective Hamiltonian for phonon modes can be obtained by expansion
238: up to second order terms of the original FK chain Hamiltonian (\ref{HFK})
239: around a chosen static configuration of the chain:
240: \begin{equation}
241: \label{Hph}
242: H^{ph}=\sum ^{s}_{i=1}\frac{\Pi _{i}^{2}}{2}
243: +\frac{1}{2}\sum ^{s}_{i,k=1}\frac{\partial U}{\partial x_{i}\partial x_{k}}\psi _{i}\psi _{k},
244: \end{equation}
245: where $ \psi _{i}=x_{i}-\bar{x}_{i} $ are small deviations of particles
246: from their equilibrium static positions $ \bar{x}_{i} $, and $ \Pi _{i} $
247: are corresponding particle momenta. The elasticity matrix
248: \begin{equation}
249: \label{Rik}
250: R_{ik}\equiv \frac{\partial ^{2}U}{\partial x_{i}\partial x_{k}}=
251: K\cos (\bar{x}_{i})\delta _{ik}-\delta _{i,k+1}-\delta _{i+1,k}.
252: \end{equation}
253: Solving the eigenvalue problem
254: \begin{equation}
255: \label{EigVal}
256: \left( R-\omega ^{2}I\right) \psi =0,
257: \end{equation}
258: numerically, we get both the spectrum of phonon frequencies and corresponding
259: vectors of phonon modes.
260:
261:
262: \section{Phonon frequency spectrum}
263:
264: \label{sPhFr}
265:
266: It is well known for a long time \cite{Au83a,Pe83}, that frequency
267: spectrum of phonons in the Frenkel-Kontorova chain is splitted into
268: several bands. Main features of the spectrum for a chain in the GS
269: look as very universal: (i) the number of main bands are independent
270: of the parameter $ K $ as well as of a length of the chain;
271: (ii) it was also noticed in \cite{Burk96} (without any explanation, either)
272: that boundaries of the bands correspond to Fibonacci numbers;
273: (iii) replacement in interparticle interactions elastic forces by the Lenard-Jones
274: potential does not change qualitatively a pattern of splitting phonon
275: spectrum into band \cite{Burk96}.
276:
277: Now we can explain these and more features as direct consequence of
278: particular spatial structure \cite{Zh01} of the GS, which main relevant
279: detailes are summarized in Introduction. The key point of our explanation
280: is that the GS consists of large number of almost identical elements.
281: For clarity sake, we consider a particular case of the FK chain with
282: $ r/s=377/610 $, but all our arguments can be easily generalizied
283: for a chain of different size.
284:
285: Applying the picture of CS spatial structure outlined in Introduction
286: to our particular case, we can see, that our CS can be presented as
287: a composition of two bricks of 3d level of hierarcy:
288: \begin{equation}
289: \label{xL3}
290: 610=(g232g376),
291: \end{equation}
292: with subsequent expansion, in accordance with composition rules (\ref{cr1}),(\ref{cr2}),
293: into bricks of 2nd level:
294: \begin{equation}
295: \label{xL2}
296: 232=(88g54g88),\quad 376=(88g54g88g54g88),
297: \end{equation}
298: which, in turn, are expanded into bricks of 1st level:
299: \begin{equation}
300: \label{xL1}
301: 54=(20g12g20),\quad 88=(20g12g20g12g20),
302: \end{equation}
303: that, at last, are expanded into basic bricks of zero level:
304: \begin{equation}
305: \label{xL0}
306: 12=(4g2g4),\quad 20=(4g2g4g2g4).
307: \end{equation}
308: In this way it is easy to calculate, that the chain can be cut into
309: pieces either as
310: \begin{equation}
311: \label{xS0}
312: 144\times g+55\times (2)+89\times (4),
313: \end{equation}
314: or
315: \begin{equation}
316: \label{xS1}
317: 34\times g+13\times (12)+21\times (20),
318: \end{equation}
319: or
320: \begin{equation}
321: \label{xS2}
322: 8\times g+3\times 54+5\times 88,
323: \end{equation}
324: or \( 2\times g+(232)+(376) \), see (\ref{xL3}). Mutual static disturbance
325: of bricks is small and decreases exponentially with a level of hierarcy,
326: therefore we have at any level of hierarcy a sequence in some order
327: of two species of almost identical structures (bricks).
328:
329: Now let us make some important remark. In our study in \cite{Zh01}
330: of {\it static} structure of the chain, the glue particles play a
331: some passive role only, which results in their specification as some
332: dummy ``glue''. However, in {\it dynamical}
333: problem of motion particles in the chain, that we now address to,
334: the role of glue particles becomes more important, since they have
335: masses. Actually, in this case one should consider
336: as repeating structures {\it elementary cells}, which we get if attach
337: to each brick (e. g. at the left side) one glue particle\footnote{%
338: Note, that at any level of hierarcy a number of glue particles is
339: equal to the total number of bricks.
340: }.
341:
342: At the basic level of hierarcy one have 55 elementary cells of 3 particles
343: and 89 elementary cells of 5 particles, let them be called as cells
344: \( \alpha ^{(0)}=(g2) \) and \( \beta ^{(0)}=(g4) \), respectively.
345: Each elementary cell, being isolated (and, e.g. periodically closed)
346: has its own eigenfrequencies: three for a cell \( \alpha ^{(0)} \)
347: and five for a cell \( \beta ^{(0)} \). Then, if these cells in the
348: chain be really isolated, then, see Fig.\ref{FrGS} we get three bands
349: each of 55 degenerated states (solid lines) which belongs to cells
350: \( \alpha ^{(0)} \), and five bands each of 89 degenerated states
351: (dotted lines).
352:
353:
354: \begin{figure}
355: {\centering \resizebox*{0.9\columnwidth}{!}{\rotatebox{90}{\includegraphics{figp1.ps}}}
356: \par}
357: \caption{\label{FrGS}Frequency spectrum for FK model in the ground state.
358: In the left box:the total spectrum of chain, in the middle and right
359: boxes the central band and its central subband are shown with greater
360: resolution. Solid and dashed horizontal lines show eigenfrequencies
361: for isolated cells
362: \protect\( \alpha ^{(0)}\protect \) and \protect\( \beta ^{(0)}\protect \)
363: respectively. }
364: \end{figure}
365:
366:
367: Due to cells interactions all degenerated frequencies are really splitted
368: into finite width bands, see Fig.\ref{FrGS}. The closer inspection
369: of frequency spectrum shows, that seven largest breaks cut the exact
370: spectrum in bands $ n=(1,89) $, $ (90,144) $, $ (145,233) $,
371: $ (234,322) $, $ (323,377) $, $ (378,467) $, $ (468,523) $,
372: $ (524,610) $ so the band widths have the same order 89, 55, 89,
373: 89, 55, 89, 55, 89, exactly as the sequence of solid and dotted horizontal
374: lines in Fig.\ref{FrGS}!
375:
376: Now, let us remind, that the sequence of cells $ \alpha ^{(0)} $
377: and $ \beta ^{(0)} $ in our GS is not random but they belong to
378: 34 cells of the next hierarcy level, see (\ref{xS1}): 13 cells $ \alpha ^{(1)}=(g12)=(g4g2g4) $
379: and 21 cells $ \beta ^{(1)}=(g20)=(g4g2g4g2g4) $. Again, if this
380: cells would be decoupled, we could see bands, that consist of 13 and
381: 21 degenerated states. Indeed, in the middle part of the Fig.\ref{FrGS},
382: where the central band is shown with greater resolution, we see a
383: sequence of bands, that contain 21,13, 21, (21+13), 21, 13, 21 number
384: of states, which is similar to that we have at the main level of hierarcy!\footnote{%
385: Here two central subbands of widths 21 and 13 are merged due to their
386: overlapping.
387: } However, at higher resolution (the right box in Fig.\ref{FrGS})
388: the spectrum in central band becomes structureless.
389:
390: This is not surprising, since the ``cells'' we have introduced
391: are not weakly coupled objects with respect to phonon modes. In fact,
392: what cells provides, is that some chain fragments have fixed periods;
393: these fragments form locally their band structure. If bands of different
394: fragments overlap, their levels are collectivised into one common
395: band. The central band is the case, however in this case band levels
396: are less sensitive to small variations introduced by extra regularities
397: of next levels of hierarcy of spatial chain structure.
398:
399: However, a quite new interesting phenomenon occurs, if one consider
400: band, which belongs to a cell $ \beta ^{(0)} $ but does {\it not}
401: overlap with any band of the cell $ \alpha ^{(0)} $. Phonons with
402: frequencies inside this band will be damped along the cells of the
403: kind $ \alpha ^{(0)} $, these cells will play a role of some
404: ``potential barriers'', which decouple cells of the kind $ \beta ^{(0)} $
405: each from other. Our chain contains 89 cells \( \beta ^{(0)} \) and
406: 55 cells $ \alpha ^{(0)} $, which is like to 89 particles separated
407: by 55 barriers, or distributed among 55 wells. Now we have got a new
408: effective Frenkel-Kontorova chain, where cells $ \beta ^{(0)} $
409: play role some particles with potential barriers $ \alpha ^{(0)} $
410: among them. For a clarity sake, let us take more graphical notations
411: for these effective ``barrier'' $ \wedge \, =\alpha ^{(0)} $,
412: and ``particle'' $ \bullet \, =\beta ^{(0)} $. Then, seconary
413: bricks, which occur in this effective FK chain, are
414: $ \widetilde{A}=\wedge \bullet \bullet \wedge $
415: and $ \widetilde{B}=\wedge \bullet \bullet \wedge \bullet \bullet \wedge $,
416: and corresponding elementary cells can be obtained adding to bricks
417: one ``glue'' particle at the left:
418: $ \widetilde{\alpha }=\bullet \wedge \bullet \bullet \wedge $,
419: $ \widetilde{\beta }=\bullet \wedge \bullet \bullet \wedge \bullet \bullet \wedge $,
420: having again three and five eigenfrequencies, respectively. Our effective
421: chain has 8 cells $ \widetilde{\alpha } $ and 13 cells $ \widetilde{\beta } $.
422: Now, if in the frequency spectrum of new {\it effective} FK chain
423: we take a band, which belongs to the cell $ \widetilde{\beta } $
424: but not to the cell $ \widetilde{\alpha } $, it must contain 13
425: states.
426:
427: We can repeat the whole procedure ones more; then we come to next
428: level effective Frenkel-Kontorova chain with 13 effective particles
429: $ \widetilde{\beta } $ distributed among 8 potential barriers $ \widetilde{\alpha } $,
430: and to next generation cells, one of the kind $ \widetilde{\widetilde{\alpha }} $
431: and two of the kind $ \widetilde{\widetilde{\beta }} $. Numerical
432: data presented in Fig.\ref{FrGS4} confirm our picture in all the
433: detailes. Here we consider the fourth band of spectra, which is well
434: resolved from other bands, see, the left box of Fig.\ref{FrGS4}.
435: It contains 89 states, the number of cells $ \beta ^{(0)} $. In
436: the middle box we take again the fourth band which in turn contains
437: 13 states, the number of cells $ \widetilde{\beta } $. At last,
438: in the right box we show by open circles 3 states of the cell
439: $ \widetilde{\widetilde{\alpha }} $
440: and by closed ones $ 2\times 5 $ states of two cells $ \widetilde{\widetilde{\beta }} $.
441: \begin{figure}
442: {\centering \resizebox*{0.9\columnwidth}{!}{\rotatebox{90}{\includegraphics{figp2.ps}}} \par}
443:
444:
445: \caption{\label{FrGS4}The same, as in Fig.\ref{FrGS}, but here a {\it fourth}
446: band and its {\it fourth} subband are shown with greater resolution.
447: In right figure box open and closed circles correspond to phonon modes
448: of the cells \protect\( \widetilde{\widetilde{\alpha }}\protect \)
449: and \protect\( \widetilde{\widetilde{\beta }}\protect \) , respectively.}
450: \end{figure}
451:
452:
453: Note, that this new kind of hierarcy is quite different from that
454: we found in \cite{Zh01} with respect to spatial structure, since
455: the transformation rules between levels of hierarcy are more complicated.
456:
457: In conclusion of this section it should be stressed, that all the universal features
458: of the global band structure mentioned at the beginning are goverened
459: by nearest order in the chain. In contrast, the fine structure depends
460: crucially on the far order.. The latter is destroyed in the CES, therefore
461: in CES the fine structure is washed out, and frequency spectra become
462: smoother, see Fig.\ref{FrGS_EX}. In particular, we see also in Fig.\ref{FrGS_EX}
463: how two lowest bands are merged into common one.
464: \begin{figure}
465: {\centering \resizebox*{0.9\columnwidth}{!}{\rotatebox{90}{\includegraphics{figp3.ps}}} \par}
466:
467:
468: \caption{\label{FrGS_EX}Comparison of the ground state (left box) and excited
469: state (rigth box) frequency spectra.}
470: \end{figure}
471:
472:
473:
474: \section{Phonon localization}
475:
476: \label{sPhLoc}
477:
478: The ordering of bricks, which persists in the GS, is gradually destroyed
479: with configurational excitation of the chain. The lowest excitations
480: are equivalent destruction of largest bricks due to permutations of
481: bricks of preceeding level of hierarcy: the smaller bricks are destroyed,
482: the higher is the excitation energy \cite{Zh01}. The number of almost
483: degenerate CES, originating from different bricks permutation is combunatorically
484: large, therefore some arbitrary chosen CES has rather {\it random}
485: sequence of bricks, and is similar to disordered media, e.g. spin
486: glass system.
487:
488: From the very beginning it is clear, that perfect randomization of
489: bricks order may result in Anderson localization of phonon modes.
490: A new interesting point is, that FK model does {\it not} require
491: any external disorder: randomization of bricks order in FK chain occurs
492: dynamically. Another interesting feature, which characterizes a degree
493: of chaotization of bricks in CES, is that all the examples of exponential
494: localization are obtained from {\it single} arbitrary taken CES,
495: without averaging over any encemble of nearest CES.
496:
497: A typical quantity used traditionally in studies of localization phenomena
498: (see, in particular \cite{Burk96,Tong99}) is the participation ratio
499: (PR), defined as
500: \begin{equation}
501: \label{iprDef}
502: R=\frac{1}{s}\left( \sum _{i=1}^{s}\psi ^{4}_{i}\right) ^{-1},
503: \end{equation}
504: where $ \psi $ is a normalized ($ \sum _{i}\psi _{i}^{2}=1 $)
505: vector of the phonon eigenstate. Its value correspond to a chain fraction
506: occupied by the localized state, but whether the state is localized
507: exponentially can be unclear. Typical feature of exponentially localized
508: state is that components outside the center of localization are exponentially
509: small. Meanwhile, this components do not contribute in PR at all.
510:
511: To our opinion a better characteristics to indicate an {\it exponential}
512: localization can be the generalized mean geometrical value (MGV),
513: defined as\begin{equation}
514: \label{mgv}
515: W=s\cdot (\prod _{i=1}^{s}\psi ^{2})^{1/s}=s\exp (\frac{1}{s}\sum ^{s}_{i=1}\ln (\psi ^{2})),
516: \end{equation}
517: where, a normalization factor $ s $ provides that for extended
518: states $ W $ be order of unity. The reason in favor to $ W $
519: is that it essentially better probes the exponentially small components
520: (tails) of the phonon eigenstate. In particular, for a typical exponentially
521: localized state $ |\psi _{i}|\sim \ell ^{-1/2}\exp (-|i-i_{0}|/\ell ) $,
522: $ \ell $ is a localization length, and value of the MGV: $ W\sim \ell ^{-1}\exp (-s/2\ell ) $,
523: i.e. becomes \emph{exponentially} small. Moreover, one can get from
524: the value of MGV an estimate of the localization length as
525: \begin{equation}
526: \label{Lloc}
527: \ell =-s/2\ln \left( \frac{W\ell }{s}\right) \simeq -s/2\ln \left( -W/2\ln W\right) .
528: \end{equation}
529: Note an important difference between the estimate of the localization
530: length as inverse participation ratio and our estimate (\ref{Lloc}).
531: The former estimates a size of the domain where the eigenstate is
532: localized, and is sensitive to particular short range dynamics for
533: formation of the given state. On the contrary, the latter is related
534: to the rate of the exponential falloff at the tails of the eigenstates,
535: and characterizes properties of the disordered media.
536:
537: Now let us turn to localization properties of phonons in the FK chain.
538: As in the previous section, we concentrate on the numerical study
539: of the chain with the rotation number $ \nu =r/s=377/610 $, and
540: parameter $ K=2 $. In Fig.\ref{fGS_PR-MGV} we present our comparison
541: of PR and MGV for GS. One can see, that both quantities $ R $ and
542: $ W $ look very similar. Note, that $ W $ for GS is well distinct
543: from zero (dotted line). This means, that all phonon states are \emph{not}
544: exponentially localized; the smallness of $ R $ and $ W $ for
545: some phonon modes means only that this modes are prelocalized only,
546: which agree with earlier studies \cite{Keto97,Tong99}.
547:
548:
549: \begin{figure}
550: {\centering \resizebox*{0.9\columnwidth}{!}{\rotatebox{90}{\includegraphics{figp4.ps}}} \par}
551:
552:
553: \caption{\label{fGS_PR-MGV}Comparison of PR (upper plot) and MGV (bottom
554: plot) versus phonon mode number \protect\( n\protect \) for the ground
555: state of the chain.}
556: \end{figure}
557:
558:
559: In order to see, how the exponential localization manifest itself
560: in PR and MGV, let us address to CES. We start with a typical CES,
561: which is still in energy very close to GS: $ \Delta U=U_{CES}-U_{GS}=10^{-12} $.
562: Despite to, that energy splitting is small, this CES belong to 5th
563: band in the energy structure of CES \cite{Zh01}, and it is one of
564: $ 10^{9} $ CESs with $ \Delta U\lesssim 10^{-12} $. In Fig.\ref{eLoc}
565: we plot $ R $ and $ W $ for the same CES. It is seen that the
566: behaviour of PR remains in main details the same, while in the MGV
567: plot there are points, where the curve touches a \emph{zero} line.
568: This means, that $ W $ at these points is exponentially small,
569: i.e. there is an \emph{exponential} localization of these modes. Note,
570: that the PR plot at the same points has no clear indications that
571: these modes are localized exponentially.
572:
573:
574: \begin{figure}
575: {\centering \resizebox*{0.9\columnwidth}{!}{\rotatebox{90}{\includegraphics{figp5.ps}}} \par}
576:
577:
578: \caption{\label{eLoc}Comparison of PR (upper plot) and MGV (bottom plot)
579: versus phonon mode number \protect\( n\protect \) for a typical CES
580: with \protect$ \Delta U=10^{-12}\protect $.}
581: \end{figure}
582:
583:
584: Some typical examples of nonlocalized mode \( (n=305) \) and localized
585: one \( (n=378) \) for this CES is presented in Fig.\ref{exMods}.
586: We see, that the latter is perfectly localized exponentially, with
587: a localization length \( \ell \sim 12 \), which characterizes a correlation
588: length of disorder. For comparison we show also the same mode in the
589: GS in a ``prelocalized'' state.
590:
591:
592: \begin{figure}
593: {\centering \resizebox*{0.9\columnwidth}{!}{\rotatebox{90}{\includegraphics{figp6.ps}}} \par}
594:
595:
596: \caption{\label{exMods}Phonon eigenvector component distribution versus the
597: component number \protect\( i\protect \). Examples of non-localized
598: and localized modes.}
599: \end{figure}
600:
601:
602: In fact, CES with \( \Delta U=10^{-12} \) corresponds an early localization
603: of phonon modes: only small fraction of the is exponentially localized,
604: as seen from Fig.\ref{eLoc}. To get insight, what modes are localized
605: at the first turn, in Fig.\ref{wFr} we plot MGV as a function of
606: phonon frequency. We see, that localized modes are located at edges
607: of the frequency spectrum bands.
608:
609:
610: \begin{figure}
611: {\centering \resizebox*{0.9\columnwidth}{!}{\rotatebox{90}{\includegraphics{figp7.ps}}} \par}
612:
613:
614: \caption{\label{wFr}MGV versus the frequency \protect\( \omega \protect \)
615: of the phonon. It is seen, that phonon are exponentially localized
616: ot the edges of ferquency bands.}
617: \end{figure}
618:
619:
620: Now, how this picture localization looks for CES with higher splitting
621: from GS? We expect \cite{Zh01}, that at the splitting \( \Delta U=10^{-12} \)
622: the largest robust elementary cells\footnote{%
623: Let us remind (see previous section), that an elementary cell is a
624: brick with one glue particle added.
625: }
626: are \( 13 \) and \( 21 \), that take part in mutual permutations
627: only, while larger cells are destroyed. In the range of splitting
628: \( \Delta U=10^{-9}\div 10^{-8} \) the cell \( 21 \) can dissociate
629: \cite{Zh01},
630: that increase a number of smaller cells \( 13 \) and decrease a corelation
631: length of the disorder. Next, at splitting \( \Delta U=10^{-5}\div 10^{-4} \)
632: the cell \( 13 \) can dissociate too, and permutations of cells \( 5 \)
633: come into play, that decrease a correlation length of the disorder
634: even more. In apparent agreement with our expectations decrease of
635: the disorder scale results in a total exponential localization of
636: substantial fraction of phonon modes, especially in high frequency
637: regoin, as seen from Fig.\ref{exW}.
638:
639: \begin{figure}
640: {\centering \resizebox*{0.9\columnwidth}{!}{\rotatebox{90}{\includegraphics{figp8.ps}}} \par}
641:
642: \caption{\label{exW}The MGV versus phonon mode number \protect\( n\protect \)
643: in two CES.}
644: \end{figure}
645:
646:
647: In Fig.\ref{loc2EX} we plot fo the same two CES our estimate for
648: localization length (\ref{Lloc}). Two dotted lines show levels \( \ell =5 \)
649: and \( 13 \), which correspond to expected sizes of largest robust
650: elementary cells, which survive at energy splitting \( \Delta U\sim 10^{-4} \)
651: and \( 10^{-8} \), respectively. We see, that minimal localization
652: length follows the size of maximal robust structure in the chain.
653:
654:
655: \begin{figure}
656: {\centering \resizebox*{0.9\columnwidth}{!}{\rotatebox{90}{\includegraphics{figp9.ps}}} \par}
657:
658:
659: \caption{\label{loc2EX}Localization length \protect\( \ell \protect \) versus
660: phonon mode number \protect\( n\protect \). Horizontal dotted lines
661: correspond to levels \protect\( \ell =5\protect \) and \protect\( 13\protect \).}
662: \end{figure}
663:
664:
665: Since the chain is not homogeneous, localization properties are not
666: homogeneous too. As seen from Figs. \ref{exW},\ref{loc2EX}, the
667: localization is maximal at the high frequency part of phonon spectrum,
668: while low frequency part seems nonlocalized. In fact, at least longwave
669: modes of phonons show a clear tendency to be localized too, see Fig.\ref{lwStates}.
670: \begin{figure}
671: {\centering \resizebox*{0.9\columnwidth}{!}{\rotatebox{90}{\includegraphics{figp10.ps}}} \par}
672:
673:
674: \caption{\label{lwStates}Profiles of phonon eigenstates: eigenvector component
675: distribution versus component number. The upper corresponds to CS,
676: while next two present CES with splitting from GS \protect\( \Delta U=6\cdot 10^{-8}\protect \)
677: and \protect\( 2.4\cdot 10^{-4}\protect \) respectively. }
678: \end{figure}
679:
680:
681:
682: \section{Discussion and conclusions.}
683:
684: In this paper we have studied properties of phonon modes in the Frenkel-Kontorova
685: chain, taken in the regime of pinned phase. Spatial "brick" structure of the ground
686: state of the chain found in \cite{Zh01}, appears to be very useful for
687: understanding the fine structure of the chain phonon spectrum. Actually, similar
688: analysis can also be performed for electronic spectrum, studied in recent work
689: \cite{Tong02}.
690:
691: However, it is obvious, that in a real physical scale the ground state is highly
692: degenerated with a huge number of the static configurational excitations states (CES)
693: of the chain. Actually this means that the true ground state of the chain is
694: practically inaccessible. CES has properties quite different from that of the ground state:
695: their spatial structure is rather chaotic \cite{Zh01}. As a result, they can cause
696: the Anderson-like (exponential) localization of phonon modes similar to that seen in
697: disordered media. This means, that results of previous
698: studies of phonon\cite{Burk96,Keto97,Tong99,Hu00} and electron properties\cite{Tong02}
699: should be revized or extended to more realistic states of the chain.
700:
701: Configurational excitation states (CES) can provide a possibility
702: to study a gradual transition from the order to disorder. It is important,
703: that the number of CES grows with energy splitting from the GS very
704: fast. Even at very small splitting the number of ECS is huge. It is curious
705: that each particular CES has intrinsic chaos, which reminds the situation in
706: classical spin glass \cite{Mez97}, but in contrast to the latter,
707: the chaos in ECS arises dynamically, without any external noise.
708:
709: The quantization of FK model in small $ \hbar$ limit is in essence
710: the quantization of its phonon modes \cite{Hu01}. Therefore localization of phonon
711: modes means localization of quantum states. Note also, that phonon
712: quantization problem is very close to electron quantization problem \cite{Tong02},
713: where transition from localization to delocalized state is interesting
714: as a insulator-metal transition .
715:
716: This work was supported in part by EC RTN network contract HPRN-CT-2000-0156.
717: One of us (O.V.Z.) thanks Cariplo fundation, INFN and RFBR grant No.04-02-16570
718: for financial support. Support from the PA INFN ``Quantum transport
719: and classical chaos'' is gratefully acknowledged.
720:
721: \begin{thebibliography}{99}
722: \bibitem{Pastur} J.M. Ziman, {\it Models of disorder} (Cambridge University Press, Cambridge,1979).
723: \bibitem{Zh01} O.V.\ Zhirov, G.\ Casati and D.L.Shepelyansky, Phys.\ Rev.\ E {\bf 65}, 026220 (2002).
724: \bibitem{FK38} Y.I.\ Frenkel and T.K.\ Kotorova, Zh.Eksp.Teor.Fiz.{\bf 8}, 1340 (1938).
725: \bibitem{Zh03} O.V.\ Zhirov, G.\ Casati and D.L.\ Shepelyansky, Phys.\ Rev.\ E {\bf 67}, 056209 (2003).
726: \bibitem{Naba67} F. Nabarro, {\it Theory of Crystall Dislocations} (Clarendon, Oxford,
727: 1967).
728: \bibitem{Pokr84} V.L.\ Pokrovsky and A.L.\ Talapov, {\it Theory of Incommesurate Crystals},
729: Soviet Scientifical Reviews Supplement Series Physics, vol.1 (Harwood,
730: London, 1984).
731: \bibitem{Ying71} S.C.\ Ying, Phys. Rev.\ B{\bf 3}, 4160 (1971).
732: \bibitem{Piet81} L.\ Pietronio, W.R.\ Schneider and S.\ Str\"{a}ssler, Phys. Rev.\ B {\bf 24},
733: 2187(1981).
734: \bibitem{Au78} S.\ Aubry, in {\it Solitons and Condensed Matter Physics}, ed. by
735: Bishop\ A.R. and Schneider\ T., (Springer, N.Y., 1978).
736: \bibitem{Au83a} S.\ Aubry, J. Phys. (France), {\bf 44}, 147 (1983).
737: \bibitem{Flor96} L.M.\ Floria and J.J.\ Mazo, Adv. Phys., {\bf 45}, 505 (1996).
738: \bibitem{Weis96} M.\ Weiser and F.J.\ Elmer, Phys. Rev.\ B{\bf 53}, 7539 (1996);
739: Z.\ Phys.\ B\textbf{104}, 55 (1997).
740: \bibitem{Brau97} O.M.\ Braun, T.\ Dauxois, M.V.\ Paliy, and M.\ Peyrard, Phys. Rev. Lett.
741: {\bf 78}, 1295 (1997); Phys. Rev.\ E, {\bf 55}, 3598 (1997).
742: \bibitem{Cons00} L.\ Consoli, H.J.F.\ Knops, and A.\ Fasolino, Phys. Rev. Lett. {\bf 85},
743: 302 (2000).
744: \bibitem{Au83b} S.\ Aubry, Physica\ D, {\bf 7}, 240 (1983); S.\ Aubry and P.Y.\ Le\ Daeron, Physica\ D
745: {\bf 8}, 381 (1983).
746: \bibitem{Au90} S.\ Aubry and G.\ Abramovici, Physica\ D {\bf 43}, 199 (1990).
747: \bibitem{Burk96} E.\ Burkov, B.E.C.\ Koltenbach, and L.W.\ Bruch, Phys. Rev.\ B {\bf 53},
748: 14179 (1996).
749: \bibitem{Hu00} B.\ Hu, B.\ Li, and H.\ Zhao, Phys. Rev.\ E {\bf 61}, 3828 (2000).
750: \bibitem{Sin77} Ya.\ G.\ Sinai,{\it Introduction to Ergodic Theory}, (Princeton University
751: Press, 1977).
752: \bibitem{Chir79} B.V.\ Chirikov, Phys. Rep. {\bf 52}, 263 (1979).
753: \bibitem{Be80} H.U.\ Beyeler, L.\ Pietronero and S.\ Str\"{a}ssler, Phys. Rev.\ B {\bf 22},
754: 2988(1980).
755: \bibitem{Vall86} F.\ Vallet, R.\ Schilling and S.\ Aubry, Europhys. Lett., {\bf 2},
756: 815 (1986); J. Phys.\ C {\bf 21}, 67 (1988).
757: \bibitem{Tong99} P.\ Tong, B.\ Li, and B.\ Hu, Phys. Rev.\ B {\bf 59}, 8639 (1999).
758: \bibitem{Hu01} B.\ Hu, B.\ Li, and H.\ Zhao, Europhys. Lett. {\bf 53} 342 (2001).
759: \bibitem{Keto97} J.A.\ Ketoja and I.I.\ Satija, Physica\ D {\bf 104}, 239 (1997).
760: \bibitem{Pe83} M.\ Peyrard and S.\ Aubry, J.\ Phys.\ C {\bf 16}, 1593 (1983).
761: \bibitem{Gree79} J.M.\ Greene, J. Math. Phys., {\bf 20}, 6 (1979);\\
762: R.S.\ MacKay, Physica\ D {\bf 7}, 283 (1983).
763: \bibitem{Mez97} M.\ M\'{e}zard, G.\ Parizi, and M.A.\ Virasoro,{\it Spin Glass Theory and
764: Beyond}, (Wold Sci., Singapore, 1997).
765: \bibitem{Tong02} P.\ Tong, B.\ Li and B.\ Hu, Phys. Rev. Lett. {\bf 88}, 046804 (2002).
766: \end{thebibliography}
767: \end{document}
768: