1: \documentstyle[twocolumn,prl,aps,amssymb,epsfig]{revtex}
2: \begin{document}
3: \twocolumn[\hsize\textwidth\columnwidth\hsize\csname@twocolumnfalse\endcsname
4: \draft
5:
6: \title{Dynamical Scaling Properties of Electrons in Quantum Systems
7: with \\Multifractal Eigenstates}
8:
9: \author{Jianxin Zhong,$^{1,2,3}$ Zhenyu Zhang,$^{2,1}$ Michael
10: Schreiber$^{4}$, E. Ward Plummer,$^{1,2}$ and Qian
11: Niu$^{2,5}$}
12:
13: \address{ $^1$Department of Physics, University of Tennessee,
14: Knoxville, Tennessee 37996\\ $^2$Solid State Division, Oak Ridge
15: National Laboratory, Oak Ridge, Tennessee 37831\\
16: $^3$Department of Physics, Xiangtan University, Hunan 411105, P.R. China\\
17: %$^3$Max-Planck-Institut f\"{u}r Str{\"o}mungsforschung, Bunsenstrasse
18: %10, D-37073 G{\"o}ttingen, Germany\\
19: $^4$Institut f\"{u}r Physik,
20: Technische Universit\"{a}t, D-09107 Chemnitz, Germany\\ $^5$Department
21: of Physics, University of Texas at Austin, Austin, Texas 78712}
22:
23: %\date{\today}
24:
25: \maketitle
26:
27: \begin{abstract}%
28:
29: We study the intricate relationships between the dynamical scaling
30: properties of electron wave packets and the multifractality of
31: the eigenstates in quantum systems. Numerical simulations for the Harper
32: model and the Fibonacci chain indicate that the root mean square
33: displacement displays the scaling behavior $r(t)\sim t^\beta$ with
34: $\beta=D_2^\psi$, where $D_2^\psi$ is the correlation dimension of
35: the multifractal eigenstates. The equality can be generalized to
36: $d$-dimensional systems as $\beta=D_2^\psi/d$, as long as the
37: electron motion is ballistic in the effective $D_2^\psi$-dimensional
38: space. This equality should be replaced by $\beta<D_2^\psi/d$ if the
39: motion is non-ballistic, as supported by all known results.
40: \end{abstract}
41:
42: \pacs{PACS numbers:
43: 71.23.Ft, %Electronic structure: Quasicrystals%
44: 71.30.+h, %Metal-insulator transitions and other electronic transitions%
45: 72.15.Rn %Quantum localization%
46: 05.45.+b, %Theory and models of chaotic systems%
47: }
48: ]
49:
50: \narrowtext
51:
52: Spatial extension of eigenstates influences electron dynamics and
53: transport properties of solids. It is well known that extended
54: eigenstates in periodic lattices result in ballistic motion of
55: electrons while localized eigenstates in strongly disordered systems
56: lead to zero mobility. As a consequence, one expects metallic phase
57: and insulating phase, respectively. In between these two extremes
58: stand the critical eigenstates, i.e., the multifractal eigenstates,
59: which are neither extended nor exponentially localized. Multifractal
60: eigenstates exist in various quantum model systems, such as the Harper
61: model of two-dimensional (2D) electrons in the presence of a
62: perpendicular magnetic field\cite{HP}, tight-binding models of
63: quasicrystals \cite{MBC,FT,FIB,ZGRS}, and disordered systems at the
64: metal-insulator transition (MIT) \cite{SE,FG,SG,BHS,OK,TT,HK,SKO}. In
65: recent years, a great deal of effort has been devoted to the analysis
66: of the multifractal nature of the eigenstates. Effort has also been
67: made to establish intrinsic connections between the multifractality of
68: the eigenstates and the transport properties in such systems. As an
69: example, the unusual resistivity increase caused by decreasing
70: temprature or improving structural quality of quasicrystals has been
71: attributed to the multifractal nature of the electronic states
72: \cite{MBC}. Despite all these efforts, a fundamental level of
73: understanding on precisely how the multifractality of the eigenstates
74: affects the electron dynamics is still lacking.
75:
76: The characteristics of electron dynamics is reflected by the time
77: evolution of the corresponding wave packet, whose spreading can be
78: described by the root mean square displacement $r(t)$. For ballistic
79: spreading, $r(t)\sim t$, while in the absence of diffusion, $r(t)\sim
80: t^0$ for large $t$. Numerical simulations \cite{AH} for the Harper
81: model and the Fibonacci quasiperiodic chain \cite{HP,MBC,FT,FIB}
82: showed that $r(t)\sim t^\beta$ with $0<\beta<1$. Such a scaling
83: behavior has also been found in other systems with multifractal
84: eigenstates \cite {OK,HK,UT,PSB}. On the other hand, the energy
85: spectrum of the multifractal eigenstates can be measured by the
86: multifractal dimensions, $D_q^\mu$ \cite{HJKPS}. The second moment,
87: $D_2^\mu$, has been shown to determine the decay of the average
88: probability, $C(t)$, of a wave packet remaining at its initial
89: position: $C(t)\sim t^{-D_2^\mu}$ \cite{KPG,ZMB}. It is naturally
90: tempting to link the spectral multifractal dimensions, $D_q^\mu$, with
91: the dynamical scaling exponent, $\beta$
92: \cite{G,GKGP,EKWA,M,P,KKKG,BB}. Indeed, earlier such efforts have led
93: to a variety of results, including $\beta\leq D_1^\mu$ \cite{G},
94: $\beta =D_{-1}^\mu$ \cite{M,P}, and $\beta\geq D_2^\mu/D_2^\psi$
95: \cite{KKKG}, where $D_2^\psi$ is the correlation dimension of the
96: eigenstates. We note that all these results were derived with the
97: assumption that the energy spectra are singularly continuous. Scaling
98: behavior of $r(t)$ may also exist in higher dimensional systems whose
99: spectra are not singularly continuous \cite{UT,PSB,ZMB}, but it would
100: be even more difficult to derive a unique relation between $\beta$ and
101: $D_q^\mu$.
102:
103: In this Letter, we use a new approach to understand the scaling
104: behavior of electron dynamics. Instead of focusing on the {\sl
105: energy} spectra, we study how the multifractality of the eigenstates
106: in {\sl real} space affects the time spreading of the wave
107: packets. Our study leads to several fundamentally important
108: findings. First, for the 1D Harper model and the Fibonacci chain, our
109: numerical simulations show that $\beta$ is directly given by the
110: correlation dimension of the multifractal eigenstates:
111: $\beta=D_2^\psi$. Second, by interpreting $D_2^\psi$ as the effective
112: dimension of the multifractal space for electron motion, we generalize
113: the above relation to $d$-dimensional systems as
114: $\beta=D_2^\psi/d$. This equality holds as long as the spreading of
115: wave packets in the $D_2^\psi$-space is ballistic in nature, as
116: demonstrated by applying to exactly solvable 2D and 3D quasiperiodic
117: systems. Furthermore, when the spreading in $D_2^\psi$-space is
118: non-ballistic, we have $\beta<D_2^\psi/d$, which is in complete
119: agreement with all the known results for higher dimensional ($d>1$)
120: disordered systems \cite {BHS,OK,TT,HK,SKO}.
121:
122: We start with a 1D lattice of size $L=Na$, where $N$ is the total
123: number of sites and $a$ the lattice constant. The Hamiltonian has the
124: tight-binding form $H=\sum_n\epsilon_n|n><n|+\sum_{m,n}t_{m,n}|m><n|$,
125: where $\epsilon_n$ is the site energy, $t_{m,n}$ the nearest-neighbor
126: hopping elements. Eigen-energies and eigenstates are denoted as $E_m$
127: and $\psi_n(E_m)$, respectively. The multifractal analysis of an
128: eigenstate is based on the standard box-counting procedure: divide the
129: system into a number of boxes of linear size $wL$ $(0<w<1)$, and then
130: calculate the box probability in the $k$th box, $p_k=\sum_{n\in
131: k}|\psi_n(E_m)|^2$. A spatial correlation function $R(w,E)$ is
132: defined as $R(w,E)=\sum_k p_k^2$. In general, $R(w,E)\sim
133: w^{D_2^\psi(E)}$, where $D_2^\psi(E)$ is called the correlation
134: dimension of the eigenstate. Extended eigenstates and localized
135: eigenstates are characterized by $D_2^\psi(E)=1$ and $D_2^\psi(E)=0$,
136: respectively. For multifractal eigenstates, $0<D_2^\psi(E)<1$.
137: Considering the fact that eigenstates contributing to the time
138: spreading of a wave packet may display different multifractal
139: behaviors, we introduce a spectral-averaged spatial correlation
140: function, $R(w)={1\over N}\sum_mR(w,E_m)$, and define a
141: spectral-averaged correlation dimension $D_2^\psi$ if $R(w)\sim
142: w^{D_2^\psi}$ holds. The root mean square displacement for a wave
143: packet initially located at site $n_0$ is given by
144: $r(t,n_0)=\sqrt{\sum_n|n-n_0|^2|\phi_n(t)|^2}$, where the wave
145: function $\phi_n(t)$ can be obtained by
146: $\phi_n(t)=\sum_m\psi^*_{n_0}(E_m)\psi_n(E_m)\exp(iE_mt)$. For
147: inhomogeneous systems, $r(t,n_0)$ varies with $n_0$. In this case,
148: one averages $r(t,n_0)$ over different initial wave packets in order
149: to obtain the dynamics for the whole system. The scaling behavior of
150: the system-averaged root mean square displacement is defined as $r(t)
151: \sim t^\beta$, with $0<\beta<1$.
152:
153: \vspace{1cm}
154: \begin{figure}
155: \centerline{\hspace{1cm}\epsfig{figure=fig1cm.eps,width=0.75\columnwidth}}
156: \vspace{1.5cm}
157: \caption{Multifractal properties of eigenstates for the Harper model
158: of system size $N=3000$. The solid line shows $R(w)\sim
159: w^{D_2^\psi}$ with $D_2^\psi=0.5$. The dotted and the dashed lines
160: show $R(w,E)\sim w^{D_2^\psi(E)}$ for energies $E_e$ and $E_c$ with
161: $D_2^\psi(E_e)=0.38$ and $D_2^\psi(E_c)=0.64$, respectively.}
162: \end{figure}
163:
164: Hamiltonian parameters in the Harper model \cite{HP} are $t_{m,n}=1$
165: and $\epsilon_n=2\cos(2\pi\sigma n)$, where $\sigma =(\sqrt{5}-1)/2$
166: is the golden mean. We find that the correlation function $R(w,E)$ for
167: various eigenstates displays a scaling behavior $R(w,E)\sim
168: w^{D_2^\psi(E)}$, where $0<D_2^\psi(E)<1$ and $D_2^\psi(E)$ is
169: different for different eigenstate. As examples, we show $R(w,E)$ for
170: a system of $N=3000$ in Fig.~1 at energies $E_{\rm e}=2.59751453292$
171: and $E_{\rm c}=0.00000066284$, which respectively correspond to the
172: upper edge and the center of the energy spectrum. The corresponding
173: scaling exponents are $D_2^\psi (E_{\rm e})=0.38$ and $D_2^\psi
174: (E_{\rm c})=0.64$. We also find that the spectral-averaged (over the
175: whole spectrum of $-2.6<E<2.6$) correlation function $R(w)$ displays a
176: scaling behavior $R(w)\sim w^{D_2^\psi}$ with $D_2^\psi=0.5$ as shown
177: in Fig.~1. In Fig.~2, $R(w)$ is compared with $r(t)$ obtained by
178: averaging over $500$ initial wave packets. We find $r(t)\sim t^\beta$
179: with $\beta=0.5$, in agreement with a previous study \cite{P}. We
180: therefore have $\beta=D_2^\psi=0.5$.
181:
182: \vspace{1cm}
183: \begin{figure}
184: \centerline{\hspace{1cm}\epsfig{figure=fig2cm.eps,width=0.7\columnwidth}}
185: \vspace{1.5cm}
186: \caption{Scaling behaviors of $r(t)\sim t^\beta$ (thick lines) and
187: $R(w)\sim w^{D_2^\psi}$ (thin lines) for the whole spectrum (solid
188: lines), the center sub-spectrum (dashed lines), and the edge
189: sub-spectrum (long dashed lines) in the Harper model with
190: $\beta=D_2^\psi=0.5$. System size is $N=3000$.}
191: \end{figure}
192:
193: We can also study how eigenstates in different spectral regions affect
194: the spreading of wave packets by considering
195: $\phi_n(t)=\sum_{E_m\in\Omega}\psi^*_{n_0}(E_m)\psi_n(E_m)\exp(iE_mt)$,
196: where $\Omega$ denotes the spectral region. The corresponding $r(t)$
197: is compared with $R(w)$ obtained by averaging the eigenstates within
198: $\Omega$. Both the spectrum and the eigenstates of the Harper model
199: have self-similar trifurcating structures
200: \cite{HP,MBC,FT,FIB}. Therefore, we expect that the spreading of the
201: wave packets contributed by eigenstates in different sub-spectra
202: follow the same scaling law. In Fig.~2, we show $R(w)$ and $r(t)$ for
203: the center sub-spectrum $-0.2<E<0.2$ and the edge sub-spectrum
204: $1.8<E<2.6$. It is clear that $R(w)$ for different sub-spectra is the
205: same because of the same multifractal behavior of the eigenstates in
206: different sub-spectra. Furthermore, $r(t)$ for different sub-spectra
207: in Fig.~2 displays the same scaling law with $\beta=D_2^\psi=0.5$, as
208: expected. This indicates that the multifractality of the eigenstates
209: indeed plays a crucial role in the spreading of the wave packets. We
210: note that the downward shift of $r(t)$ shown in Fig.~2 for the
211: sub-spectra is due to the decrease in the number of eigenstates
212: contributing to the spreading of the wave packets.
213:
214: We find the same relation $\beta=D_2^\psi$ to hold for the Fibonacci
215: quasiperiodic chain \cite{FIB}. In this model, $t_{i,j}=1$ and
216: $\epsilon_i$ takes $-u$ and $u$ arranged in a Fibonacci sequence. The
217: averaged correlation function has the scaling behavior $R(w)\sim
218: w^{D_2^\psi}$ with $0<D_2^\psi<1$ depending on $u$. The
219: system-averaged root mean square displacement $r(t)$ used to be compared
220: with $R(w)$ was obtained by an average of $500$ wave packets. As
221: examples, we show $R(w)$ and $r(t)$ for the whole spectrum in Fig.~3
222: around $u=1.65$: $\beta=D_2^\psi =0.62$ for $u=1$; $\beta=D_2^\psi
223: =0.5$ for $U=1.65$; and $\beta=D_2^\psi =0.43$ for $u=2$. These three
224: cases cover the super-diffusive, diffusive, and sub-diffusive regimes
225: of motion in real space.
226:
227: \vspace{1cm}
228: \begin{figure}
229: \centering {\hspace{1cm}\epsfig{figure=fig3cm.eps,width=0.7\columnwidth}}
230: \vspace{1.5cm}
231: \caption{Scaling behaviors of $r(t)\sim t^\beta$ (dashed lines) and
232: $R(w)\sim w^{D_2^\psi}$ (solid lines) in the Fibonacci chain with
233: site energies $u=1, 1.65$, and $2$. The corresponding exponents are
234: $\beta=D_2^\psi=0.62, 0.5$, and $0.43$, respectively. System size is
235: $N=2584$.}
236: \end{figure}
237:
238: The relation $\beta=D_2^\psi$ observed in different models can be
239: generalized to higher dimensional systems with multifractal
240: eigenstates. From its definition, $R(w)$ is in fact the average
241: probability for an electron in a box of linear size $wL$. Because
242: $R(w)\sim w^{D_2^\psi}$, the effective number of boxes of the system
243: is $N'=({1\over w})^{D_2^\psi}$, instead of $(1/w)^d$, in order to
244: ensure the normalization $N'\times R(w)\sim 1$. From the standard
245: definition of the volume dimension $D_v$, i.e., {\sl the number of the
246: measured boxes}$\times (wL)^{D_v}=constant$, we have
247: $D_v=D_2^\psi$. Thus $D_2^\psi$ can be viewed as the effective volume
248: dimension for electron motion in the multifractal space (hereafter
249: called the $D_2^\psi$-space). If the motion (i.e., the spreading of
250: the wave packet) is ballistic in $D_2^\psi$-space, the volume, $N'$,
251: visited by the electron is $N'\sim t^{D_2^\psi}$. Because the width of
252: the wave packet $r(t)$ is still measured in the $d$-dimensional real
253: space, we have $r(t)\sim t^{D_2^\psi/d}$, which immediately leads to
254: the result $\beta=D_2^\psi/d$.
255:
256: As tests of the above generalized relation, we consider the exactly
257: solvable models in higher dimensional Fibonacci lattices \cite{UT,ZMB}
258: and the Labyrinth lattices \cite{S}. We find that the equality
259: $\beta=D_2^\psi/d$ holds for all those models. As an example, here we
260: only show the case for 3D Fibonacci lattice, which is constructed in a
261: cubic lattice with constant hopping elements $t_{i,j}=1$ and separable
262: site energy $\epsilon_i=\epsilon_{ix}+\epsilon_{iy}+\epsilon_{iz}$,
263: where $\epsilon_{ix}, \epsilon_{iy}$, and $\epsilon_{iz}$ are the 1D
264: Fibonacci site energies. The energy spectrum is given by
265: $E=E_x+E_y+E_z$ and the corresponding eigenstates are
266: $\psi(E,x,y,z)=\psi_x(E_x)\times \psi_y(E_y)\times \psi_z(E_z)$, where
267: $E_j$ and $\psi_j(E_j)$ $(j=x,y,z)$ are the energy spectra and the
268: eigenstates for the 1D Fibonacci chain, respectively. As a
269: consequence, these eigenstates are multifractal \cite{UT,ZMB}, and
270: their spreading dynamics is non-ballistic in real space: $r(t)\sim
271: t^\beta$ \cite{UT}, where $0<\beta<1$ takes the same value as that for
272: the 1D Fibonacci chain. We note that all the relationships between
273: the energy spectra and the dynamics \cite{G,GKGP,EKWA,M,P,KKKG} cannot
274: give correct prediction for the spreading of wave packets in these
275: higher dimensional quasiperiodic systems. In our approach, for the
276: higher dimensional Fibonacci lattices, it is easy to show that $R(w)$
277: has the scaling behavior with $D_2^\psi=dD_2^{\psi_x}$, where
278: $D_2^{\psi_x}$ is the multifractal dimension of the eigenstates in the
279: 1D Fibonacci chain. Therefore, we have $\beta=D_2^\psi/d$.
280:
281: Electron motion in the effective $D_2^\psi$-space can also be
282: non-ballistic. In such cases, the above equality should be replaced by
283: $\beta<D_2^\psi/d$, because no motion should be faster than ballistic.
284: One special example is the diffusive motion ($\beta=0.5$) in the
285: metallic regime of 3D disordered systems with extended eigenstates
286: ($D_2^\psi=3$), where $\beta<D_2^\psi/3$. As shown below, this
287: inequality is also satisfied by many other known examples involving
288: non-ballistic motion in the $D_2^\psi$-space. At the metal-insulator
289: transition of 3D disordered systems (in the absence of magnetic
290: field), the eigenstates are found to be multifractal with different
291: values of $D_2^\psi$: $D_2^\psi/3=0.57\pm 0.1$\cite {SE},
292: $D_2^\psi/3=0.48-0.6$\cite {SG}, $D_2^\psi/3=0.57\pm 0.07$ \cite
293: {BHS}, and $D_2^\psi/3=0.5\pm 0.07$ \cite {OK}. For such systems in a
294: magnetic field, we have $D_2^\psi/3=0.57\pm 0.07$ \cite {OK},
295: $D_2^\psi/3\approx 0.5$ \cite {TT}, and $D_2^\psi/3=0.43\pm 0.03$
296: \cite{HK}. On the other hand, the corresponding electron dynamics was
297: found to have the same scaling exponent $\beta=0.33$ \cite {OK},
298: showing that $\beta<D_2^\psi/d$ for all these cases (with or without a
299: magnetic field). Existing numerical results for 2D disordered systems
300: satisfy this inequality as well. For the 2D disordered systems in a
301: strong magetic field, eigenstates at the centers of the Landau bands
302: are multifractal with $D_2^\psi/2=0.76\pm 0.03$ and the corresponding
303: spreading has a scaling behavior with $\beta=0.5$ \cite{HK}. At the
304: MIT of the 2D disordered systems with spin-orbit interactions, one
305: finds $D_2^\psi/2=0.84\pm 0.03$ and $\beta=0.5$ \cite{SKO}. It is
306: evident that $\beta< D_2^\psi/d$.
307:
308: In conclusion, we have shown that there exists an intrinsic
309: relationship between the dynamic spreading of electron wave packets
310: and the multifractality of the eigenstates in quantum systems. For the
311: 1D Harper model and the Fibonacci chain, we have $\beta=D_2^\psi$,
312: where $\beta$ is the scaling exponent of the root mean square
313: displacement, and $D_2^\psi$ the correlation dimension of the
314: multifractal eigenstates. We have further provided a conceptually new
315: interpretation of $D_2^\psi$ as the effective dimension for electron
316: motion in the multifractal space. Following this interpretation, the
317: above equality can be generalized to $d$-dimensional systems as
318: $\beta=D_2^\psi/d$, as long as the electron motion is ballistic in the
319: $D_2^\psi$-space. We have demostrated that such an equality applies to
320: exactly solvable 2D and 3D quasiperiodic systems. When the motion in
321: the $D_2^\psi$-space is non-ballistic due to disorder, the equality is
322: replaced by $\beta<D_2^\psi/d$, which is in complete agreement with
323: all the known results for higher dimensional ($d>1$) disordered
324: systems.
325:
326:
327: %\paragraph*{Acknowledgements:}
328: We thank T. Geisel, R. Ketzmerick, J. Bellissard, R.A. R{\"o}mmer,
329: U. Grimm, X.C. Xie and G. Canright for helpful discussions. This
330: research was supported by the US National Science Foundation under
331: Grant No. DMR-9702938, by Oak Ridge National Laboratory, managed by
332: Lockheed Martin Energy Research Corp. for the Department of Energy
333: under contract No. DE-AC05-96OR22464, and in part by the National
334: Natural Science Foundation of China.
335:
336:
337: \begin{references}
338: \frenchspacing
339:
340: \bibitem{HP}
341: P.G. Harper,
342: Proc. Phys. Soc. London A {\bf 68}, 874 (1955);
343: D.R. Hofstadter,
344: Phys. Rev. B {\bf 14}, 2239 (1976);
345: S. Aubry and G. Andre,
346: Ann. Israel Phys. Soc. {\bf 3}, 133.
347:
348: \bibitem{MBC} D. Mayou, in {\sl Lectures on Quasicrystals}, edited by
349: F. Hippert and D. Gratias (Les Editions de Physique, Les Ulis, 1994),
350: pp. 417-462; C. Berger, {\sl ibid}, pp.463-504; C. Sire, {\sl ibid},
351: pp.505-536.
352:
353: \bibitem{FT} T. Fujiwara and H. Tsunetsugu, in {\sl Quasicrystals: The
354: State of the Art}, edited by D.P. Di Vincenzo and P.J. Steinhardt
355: (World Scientific, Singapore, 1991).
356:
357: \bibitem{FIB}
358: M. Kohmoto, L.P. Kadanoff, and C. Tang,
359: Phys. Rev. Lett. {\bf 50}, 1870 (1983);
360: S. Ostlund, R. Pandit, D. Rand, H.J. Schellnhuber, and E.D. Siggia,
361: Phys. Rev. Lett. {\bf 50}, 1873 (1983);
362: M. Kohmoto, B. Sutherland, and C. Tang,
363: Phys. Rev. B {\bf 35}, 1020 (1987);
364: Q. Niu and F. Nori,
365: Phys. Rev. Lett. {\bf 57}, 2057 (1986).
366:
367: \bibitem{ZGRS}
368: J.X. Zhong, U. Grimm, R.A. R{\" o}mer, and M. Schreiber,
369: Phys. Rev. Lett. {\bf 80}, 3996 (1998);
370: J.X. Zhong and T. Geisel,
371: Phys. Rev. E, in press (1999).
372:
373: \bibitem{SE}
374: C.M. Soukoulis and E.N. Economou,
375: Phys. Rev. Lett. {\bf 52}, 565 (1984).
376:
377: \bibitem{FG} F. Wegner, in {\sl Localization and Metal-Insulator
378: Transitions}, edited by H. Fritzsche and D. Adler, (Plenum Press, New
379: York 1985), p. 337.
380:
381: \bibitem{SG}
382: M. Schreiber and H. Grussbach,
383: Phys. Rev. Lett. {\bf 67}, 607 (1991).
384:
385: \bibitem{BHS}
386: T. Brandes, B. Huckestein, and L. Schweitzer,
387: Ann. Physik {\bf 5}, 633 (1996).
388:
389:
390: \bibitem{OK}
391: T. Ohtsuki and T. Kawarabayashi,
392: J. Phys. Soc. Jpn. {\bf 66}, 314 (1996).
393:
394: \bibitem{TT}
395: T. Terao, Phys. Rev. B {\bf 56}, 975 (1997).
396:
397: \bibitem{HK}
398: B. Huckestein and R. Klesse,
399: Cond-mat/9805038, (1998).
400:
401: \bibitem{SKO}
402: L. Schweitzer,
403: J. Phys. Conden. Matter {\bf 7}, L281 (1995);
404: T. Kawarabayashi and T. Ohtsuki,
405: Phys. Rev. B {\bf 53}, 6975 (1996).
406:
407: \bibitem{AH}
408: H. Hiramoto and S. Abe, J. Phys. Soc. Jpn. {\bf 57}, 230
409: (1998); {\sl ibid} 1365 (1988).
410:
411: \bibitem{UT}
412: K. Ueda and H. Tsunetsugu,
413: Phys. Rev. Lett. {\bf 58}, 1272 (1987).
414:
415: \bibitem{PSB}
416: B. Passaro, C. Sire, and V.G. Benza,
417: Phys. Rev. B {\bf 46}, 13751 (1992).
418:
419: \bibitem{HJKPS}
420: T.G. Halsey, M.H. Jensen, L.P. Kadanoff, I. Procaccia and B.I. Shraiman,
421: Phys. Rev. A {\bf 33}, 1141 (1986).
422:
423: \bibitem{KPG}
424: R. Ketzmerick, G. Petschel, and T. Geisel,
425: Phys. Rev. Lett. {\bf 69}, 695 (1992).
426:
427: \bibitem{ZMB}
428: J. X. Zhong and R. Mosseri,
429: J. Phys. Condens. Matter {\bf 7}, 8383 (1995);
430: J.X. Zhong, J. Bellissard, and R. Mosseri,
431: J. Phys. Condens. Matter {\bf 7}, 3507 (1995).
432:
433: \bibitem{G}
434: I. Guarneri,
435: Europhys. Lett. {\bf 10}, 95 (1989);
436: Europhys. Lett. {\bf 21}, (1993).
437:
438: \bibitem{GKGP}
439: T. Geisel, R. Ketzmerick, and G. Petschel,
440: Phys. Rev. Lett. {\bf 66}, 1651 (1991).
441:
442: \bibitem{EKWA}
443: S. N. Evangelou and D. E. Katsanos,
444: J. Phys. A {\bf 26}, L1243 (1993);
445: M. Wilkinson and J. Austin,
446: Phys. Rev. B {\bf 50}, 1420 (1994).
447:
448: \bibitem{M}
449: G. Mantica,
450: Physica (Amsterdam) {\bf 103D}, 576 (1997).
451:
452: \bibitem{P}
453: F. Piechon, Phys. Rev. Lett. {\bf 76}, 4372 (1996).
454:
455: \bibitem{KKKG}
456: R. Ketzmerick, K. Kruse, S. Kraut, and T. Geisel,
457: Phys. Rev. Lett. {\bf 79}, 1959 (1997).
458:
459: \bibitem{BB}
460: H. Schulz-Baldes and J. Bellissard,
461: Rev. Math. Phys. {\bf 10}, 1 (1998).
462:
463: \bibitem{S}
464: C. Sire,
465: Europhys. Lett. {\bf 10}, 483 (1990).
466:
467: \end{references}
468:
469: \end{document}
470:
471:
472:
473:
474: