1: %\documentstyle[eqsecnum,aps,prb,epsf]{revtex4}

2: %\documentclass[twocolumn,prb,floatfix,showpacs]{revtex4}

3: %documentclass[prb,floatfix,showpacs]{revtex4}

4: %\documentclass[prb,floatfix,showpacs]{revtex4}

5: %%% some abbreviations

6: 

7: 

8: \documentclass[preprint,aps]{revtex4}

9: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

10: \usepackage{amsmath}

11: \usepackage{amssymb}

12: \usepackage{graphicx}

13: 

14: %\setcounter{MaxMatrixCols}{10}

15: %TCIDATA{OutputFilter=LATEX.DLL}

16: %TCIDATA{Version=5.00.0.2606}

17: %TCIDATA{<META NAME="SaveForMode" CONTENT="1">}

18: %TCIDATA{BibliographyScheme=Manual}

19: %TCIDATA{LastRevised=Thursday, August 11, 2005 16:40:54}

20: %TCIDATA{<META NAME="GraphicsSave" CONTENT="32">}

21: 

22: \newcommand{\ie}{\emph{i.}$\,$\emph{e.}}

23: \newcommand{\etal}{\emph{et}$\,$\emph{al.}}

24: \newcommand{\romd}{{\operatorname{d}}}

25: \newcommand{\VECe}{{\boldsymbol{e}}}

26: \newcommand{\VECf}{{\boldsymbol{f}}}

27: \newcommand{\VECl}{{\boldsymbol{l}}}

28: \newcommand{\VECn}{{\boldsymbol{n}}}

29: \newcommand{\VECt}{{\boldsymbol{t}}}

30: \newcommand{\VECF}{{\boldsymbol{F}}}

31: \newcommand{\VECX}{{\boldsymbol{X}}}

32: \newcommand{\RR}{\mathbb{R}}

33: %\input{tcilatex}

34: 

35: \begin{document}

36: 

37: \title{Scaling behavior of a one-dimensional correlated\\

38: disordered electronic\\

39: System}

40: \author{Ibrahim Avgin}

41: \affiliation{Department of Electrical and Electronics\\

42: Engineering, Ege University,\\

43: Bornova 35100, Izmir, Turkey}

44: \email{ibrahim@eng.ege.edu.tr}

45: \date{\today}

46: 

47: \begin{abstract}

48: A one-dimensional diagonal tight binding electronic system with correlated

49: disorder is investigated. The correlation of the random potential is

50: exponentially decaying with distance and its correlation length diverges as

51: the concentration of "wrong sign" approaches to $1$ or $0$. The correlated

52: random number sequence can be generated easily with a binary sequence

53: similar to that of a one-dimensional spin glass system. The localization

54: length (LL) and the integrated density of states (IDOS) for long chains are

55: computed. A comparison with numerical results is made with the recently

56: developed scaling technique results. The Coherent Potential Approximation

57: (CPA) is also adopted to obtain scaling functions for both the LL and the

58: IDOS. We confirmed that the scaling functions show a crossover near the band

59: edge and establish their relation to the concentration. For concentrations

60: near to $0$ or $1$ (longer correlation length case), the scaling behavior is

61: followed only for a very limited range of the potential strengths.

62: \end{abstract}

63: 

64: \pacs{71.23.An, 73.20.Fz, 72.15.Rn, 78.30.Ly}

65: \maketitle

66: 

67: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

68: 

69: Theoretical interest in disordered chains remains strong. Recent

70: investigations of the correlated disordered in random or chaotic arrays

71: revealed surprising results such as a possible breakdown of the Anderson criterion

72: for the localization \cite{hy04,lyra02,lyra98,izra99,deyc03}. Various

73: fields have made use of the results obtained from the study of the

74: one-dimensional models. Random microwave transmission in a single-mode wave

75: guide experiment \cite{ul}, transport studies with GaAs-AlGaAs random dimer

76: super lattice systems \cite{loc}, and the photonic band-gap structures \cite%

77: {bay} are a few examples. Recently a new renormalization technique \cite%

78: {rus98,rus02} has been introduced to study the scaling behavior of the well

79: known tight binding chain with long range correlated disorder. The authors

80: found that the Localization Length (LL) shows scaling behavior and a cross

81: over near the band edge. The behavior of disordered magnetic or electronic

82: chains \cite{ishii} (using proper transformations \cite{avgin93,pimentel89})

83: can be described mathematically by a tri-diagonal tight binding model given

84: by

85: 

86: \begin{equation}  \label{tb}

87: (E -\xi_{n}V)\psi_{n}=\psi_{n+1}+\psi_{n-1},

88: \end{equation}

89: 

90: \noindent where $E$ is energy, $V$ is the strength of the random potential

91: with its correlated random sign $\xi_{n}$. $n$ is the site index.

92: 

93: In this paper, we study the scaling properties of a particular form of

94: correlated disorder that is associated with spin glass chains \cite%

95: {pimentel89,avgin93,avgin02,boukahil89}. The random sign has the relation $%

96: \xi _{n}=\xi _{n-1}x_{n}$ with the following distribution $%

97: P(x_{n})=(1-c)\delta (x_{n}-1)+c\delta (x_{n}+1)$ where the $x_{n}$ are

98: uncorrelated between different sites and $c$ is the concentration of "wrong

99: signs". Clearly $\xi _{n}$ is exponentially correlated, \emph{i.}$\,$\emph{e.%

100: }, $\langle \xi _{n}\xi _{m}\rangle =(1-2c)^{|m-n|}$ where one can define a

101: correlation length \cite{vulp89} $l(c)=-1/ln|1-2c|$. For $c>0.5$ the

102: ordering is of "antiferromagnetic type" and $c=0.5$ is uncorrelated case

103: since the correlation length is zero. As seen in Fig.~1, for $c$ approaching 

104: $0$ or $1$ the correlation length diverges.

105: 

106: Previous studies of the disordered chain problem showed that the electronic

107: wave function decays exponentially with a distance \cite{thouless}

108: characterized by the real part of the Lyapunov exponent \cite{derrida84}

109: whose imaginary part is also related to the Integrated Density of States

110: (IDOS) \cite{derrida84}. In our work numerical calculations are carried out

111: for chains of $10^{8}$ sites using the negative eigenvalue counting

112: technique introduced by Dean \cite{dean} providing direct computation of the

113: Lyapunov exponent.

114: 

115: Recently the scaling properties arising from long range correlated disorder 

116: \cite{rus98,rus02} were investigated using the renormalization approach. The

117: authors have argued that their analyses work close to the band edge where

118: the characteristic wavelength diverges since neighboring lattice sites move

119: as blocks. Here we adopt their results for the correlated case discussed

120: above.

121: 

122: The complex Laypunov exponent \cite{derrida84} is defined as $\Gamma(E,V,c)

123: = \lim_{N\to\infty}\frac{1}{N} \ln \frac{\psi_N}{\psi_0}$. The space

124: decimation procedure \cite{rus98} leads to redefining the random potential

125: by its mean value over the block, \emph{i.}$\,$\emph{e.}, $%

126: V_N=\sum_{n=1}^NV\xi_n$.  The square of this scaled potential has finite value

127: proportional to $N$ with increasing $N$ for all concentrations

128: as displayed in Fig.~1.  For $c>0.5$, the summation has alternating values

129: particularly persistent for $c \rightarrow 1$ 

130: but eventually reaches to a limiting value

131: discussed below for larger $N$ (see Fig.~1). 

132: For uncorrelated case exact scaling results 

133: exist for the $LL$ and the $IDOS$ 

134: defined via the Lyapunov exponent. Near the band edges ($E \to 2$ and

135: $V \to 0$) the Lyapunov exponent displays 

136: a scaling law of the form \cite{derrida84}

137: $\Gamma \approx <V^2>^{1/3}h(\frac {2-E}{<V^2>^{2/3}})$ where

138: $h$ explicitly known scaling function. For the correlated

139: case, it was assumed that the Lyapunov exponent \cite{rus98} 

140: has similar power-law relation to the second moment of the redefined

141: random potential  (note that the first moment is zero) 

142: $<V_N^2>$ which can be written for our case as:

143: 

144: \begin{equation}

145: <V_N^2>=\frac{V_N^2}{N}=\frac{1-c}{c}V^2.  \label{vsq}

146: \end{equation}

147: 

148: \noindent Detailed account of this factor is given in Ref.\cite{avgin98,avgin93} 

149: This factor (in fact its $1/3$ power) $\frac{1-c}{c}$ is shown in

150: Fig.~1 and it plays a role in the scaling behavior discussed below. Unlike

151: the correlation length it diverges as $c\rightarrow 0$ while it remains

152: finite as $c\rightarrow 1$.

153: 

154: \begin{figure}[tbp]

155: \centering

156: %\includegraphics[scale=0.8]{91.eps}

157: \includegraphics[angle=270, width=18cm]{f1.eps}

158: \caption{ The lower panel is a plot of the scaling factor 

159: $(\frac {1-c}{c})^{1/3}$ and the correlation length $l(c)$ versus $c$. 

160: $l(c)$ is symmetric about $c=0.5$ and zero at this concentration. The

161: square of the potential $V_{N}^2$ versus the number of cites are

162: presented at the upper panels with respective concentrations.  The lines

163: are $V_{N}^2=\frac {1-c}{c}N$ obtained for a large $N$

164: and the points are calculated by a finite summation of $V_{N}^2$.}

165: \label{fig:scl}

166: \end{figure}

167: 

168: At the band edge $E=2$, according to the analysis of Ref.\cite{rus98} 

169: the space decimation for blocks of b sites 

170: results in the following scaling: the strength of the potential must

171: hold (to preserve the form of the Eq.~(\ref{tb}) 

172: $ V \rightarrow V_{b}=bV$ (see Eq.(7-11) of Ref.\cite{rus98}), hence

173: the second moment transforms 

174: $<V_{N_b}^{2}> = b\frac {1-c}{c} (bV)^2=b^{3}<V_{N}^{2}>$. 

175: This corresponds to their \cite{rus98} case of $\gamma = 1$. 

176: Thus The Lyapunov exponent at the band edge can be written in a form 

177: \cite{derrida84,rus98} 

178: $\Gamma (E=2,V,c)=(<V_{N}^{2}>)^{1/3}=(V^{2}\frac{1-c}{c})^{1/3}$.

179: The power of the Lyapunov exponent is the same as the uncorrelated case 

180: \cite{derrida84} but the coefficient is different. 

181: The factor involving concentration

182: was recovered before using phenomelogical arguments 

183: in the context of the spin glass chain \cite%

184: {avgin93,avgin98}. 

185: The exact results for uncorrelated case $c=0.5$ are well known \cite%

186: {derrida84} given by $IDOS=-{\frac{1}{\pi }}\Im \Gamma =0.159V^{2/3}$ and $%

187: \frac{1}{LL}=\Re \Gamma =0.289V^{2/3}$. In an earlier work we showed that

188: the IDOS and the LL \cite{avgin93,avgin98,avgin02} results for various

189: concentrations scaled with a similar factor where data collapsed well to the

190: exact calculation at $c=0.5$.

191: 

192: For uncorrelated case it is known that there is a cross over limit when $%

193: w\equiv 2-E$ is positive and very small \cite{rus02,derrida84}. In this

194: limit, the Lyapunov exponent can be written in the form

195: 

196: \begin{equation}

197: \Gamma (w,V,c) \sim (V^2 \frac{1-c}{c})^{1/3}F(X)

198: \end{equation}%

199: \ 

200: 

201: \noindent where $X\equiv w(V^{2}\frac{1-c}{c})^{-2/3}$ and $F(X)$ is the

202: scaling function. From the previous studies it was found \cite%

203: {derrida84,rus02} that the scaling function has different behaviors for $X<<1

204: $ and $X>>1$, hence there is a cross over between these two limits. It was

205: argued \cite{rus02} that in this system the cross over is a consequence of

206: the competition between the characteristic length $\lambda $ of the wave

207: function and the LL and the two asymptotic regimes can be obtained by the

208: dominance of the LL or $\lambda $. When $\lambda >>LL$ where $w\rightarrow 0$

209: the wave function is independent of $\lambda $ since it would decay over a

210: distance of $\lambda $ and the LL does not deviate much from its value at

211: the band edge so that $F(x)\rightarrow constant$. In the other limit $%

212: LL>>\lambda $ where the wave function would have considerable oscillations

213: before decaying, the LL has a different functional dependence on the

214: strength of the random potential \cite{rus02,derrida84}, \emph{i.}$\,$\emph{%

215: e.}, $LL\sim <V^{2}>$ since $w$ dominates the dynamics yielding $%

216: F(X)\rightarrow X$.

217: 

218: Insight can be gained if we adopt the Coherent Potential Approximation (CPA)

219: results here. The CPA self energy or the coherent potential has important

220: connections to the system's dynamics which for uncorrelated disorder \cite%

221: {avgin96,avgin2} can be written as $V_{c}\simeq (\frac{<V^{2}>}{2i})^{2/3}$.

222: For correlated random potentials we can substitute here the second moment of

223: the potential given in Eq.~(\ref{vsq}) (so that the correlations between

224: different sites are included) then the coherent potential takes this form $%

225: V_{c}=(\frac{1-c}{c}\frac{V^{2}}{2i})^{2/3}$. The dispersion relation can be

226: obtained from the poles of the configurationally averaged $k$ dependent

227: Green function \cite{avgin96} that $E-\Re V_{c}=2cosk$ where $k$ is a wave

228: vector. For small $k$, it can be reformulated as $k^{2}\simeq \Re V_{c}+w$.

229: Away from the band edge the LL was calculated before \cite%

230: {rus02,derrida84,avgin2} and can be expressed as $LL=\frac{8{\sin }^{2}k}{%

231: <V^{2}>}\simeq \frac{8k^{2}}{<V^{2}>}$. Recalling the fact that in one

232: dimension $k$ is proportional to the IDOS, both the LL and the IDOS can be

233: rearranged yielding the scaling functions respectively $F(x)=LL(\frac{1-c}{c}%

234: V^{2})^{1/3}=8(2^{-5/3}+X)$ and $G(X)=IDOS(\frac{1-c}{c}V^{2})^{-1/3}=

235: {\pi}^{-1}\sqrt{2^{-5/3}+X}$ where $G(X)$ has different asymptotic 

236: behavior when $X>>1$, $G(X)\rightarrow \sqrt{X}$.

237: 

238: We present our numerical and the CPA results below. We have calculated the

239: scaling functions $F(X)$ and $G(X)$ for chains of $10^{8}$ sites using the

240: negative eigenvalue counting technique. As shown in Fig.~1 (semi-logarithmic

241: plot) the correlation length goes to zero as $c\rightarrow 0.5$ and

242: increases with increasing $|c-0.5|$ with a rapid increase developing near $%

243: c\rightarrow 1,0$. The results presented in Fig.~2 are for shorter

244: correlation lengths where  $c=0.1,0.25,0.5,0.75,0.9$ and for $w=10^{-4}$ and 

245: $10^{-3}$. The data clearly indicate that there is a scaling function for

246: both the LL and IDOS. The CPA under estimates the scaling function for the

247: LL but reproduces the IDOS results rather well as shown in figures presented

248: in this section. We have found that for $w>10^{-4}$ the scaling functions

249: some what deviate from the expected scaling behavior near the lower values

250: of $X$. However for $w<10^{-4}$ they follow the predicted results.

251: 

252: \begin{figure}[tbp]

253: \centering

254: \includegraphics[angle=270,width=12cm]{f3.eps}

255: \caption{ plot of $F(X)$ (top) and $G(X)$ (bottom) as function of the

256: scaling variable $X \equiv w (V^2 \frac{1-c}{c})^{-2/3}$ strength $V$. The

257: lines are the CPA results.  The data displayed for the concentrations $%

258: c=0.1,0.25,0.5,0.75,0.9$ and the two energies $w=2-E=10^{-4}$ and $10^{-5}$.}

259: \label{fig:sc1}

260: \end{figure}

261: 

262: \begin{figure}[tbp]

263: \centering

264: \includegraphics[angle=270,width=12cm]{f4.eps}

265: \caption{ plot of $F(X)$ (top) and $G(X)$ (bottom) as function of the

266: scaling variable $X \equiv w (V^2 \frac{1-c}{c})^{-2/3}$ strength $V$. The

267: lines are the CPA results.  The data displayed for the concentrations $%

268: c=0.1,0.05,0.9,0.95$ and the two energies $w=2-E=10^{-6}$ and $10^{-7}$.}

269: \label{fig:sc2}

270: \end{figure}

271: 

272: For concentrations near $0,1$ a region of rapidly increasing correlation

273: length, we produced two sets of data displayed in Fig.~3 and Fig.~4. In

274: Fig.~3 the scaling functions are displayed for concentrations $%

275: c=0.1,0.05,0.9,$and $0.95,$ where the rapid increase in the correlation

276: length starts. The scaling behavior is observed for only $w\geq 10^{-6}$.

277: Note that we obtained similar, even better scaling behavior for the

278: concentrations presented in Fig.~2 with $w\geq 10^{-6}$ values.

279: 

280: The scaling behavior is also clearly revealed in Fig.~4 where the deviations

281: near the lower limit are generally largest for $w=10^{-6}$. The

282: concentrations here are very close to $0,1$, as presented in figure caption,

283: so one has to go to very small values of $w$ (fixed $V$) to see the

284: scaling behavior. The shrinking of the scaling region in $w$ as $c \to 0,1$

285: is consistent with the behavior for $c = 0,1$, where the correlation 

286: length is infinite. The collapse of the data for different values of $c$ 

287: is a feature of the model that reflects the fact that both the 

288: second moment (Eq.~(\ref{vsq}) and the correlation 

289: length are functions of $c$. 

290: The IDOS results also deviate for this range of

291: concentrations. Thus as the correlation length is increased, the scaling

292: behavior can be observed only for smaller and smaller values of $w$;

293: however, we found that $w$ cannot be reduced as much as desired.  Hence 

294: after certain small $w$, deviations starts this time 

295: particularly at higher $X$ sides.  

296: Thus there is a trade-off in building the scaling variable $X$ since

297: it depends on the three parameters $w$, $c$ and $V$ and only

298: delicate combination of these parameters produces scaling and the

299: crossover behavior.  Stronger deviations presented in Fig.~3-4

300: at very small $X$ for the concentrations near $0,1$ show 

301: this sensitive dependence.  Because very small concentrations, the 

302: scaling factor becomes very small or very large, therefore the 

303: Lyapunov exponent must be computed in a region 

304: of a very small or a very large potential strengths at very small 

305: $w$ wherein the computational errors inevitable

306: or the scaling behavior of the Lyapunov exponent holds only very

307: limited potential strengths $V$. 

308: 

309: \begin{figure}[tbp]

310: \centering

311: \includegraphics[angle=270,width=12cm]{f5.eps}

312: \caption{ plot of $F(X)$ (top) and $G(X)$ (bottom) as function of the

313: scaling variable $X \equiv w (V^2 \frac{1-c}{c})^{-2/3}$ strength $V$. The

314: lines are the CPA results. The data displayed for the concentrations $%

315: c=0.01,0.005,0.99,0.995$ and the two energies $w=2-E=10^{-6}$ and $10^{-7}$.}

316: \label{fig:sc3}

317: \end{figure}

318: 

319: We have studied binary correlated disordered chains both analytically and

320: numerically. The LL and the IDOS are computed for various concentrations. We

321: found that they both showed a scaling behavior and a crossover. The behavior

322: of the scaling functions predicted by the renormalization group techniques 

323: \cite{rus02,rus98} is observed. The data calculated numerically for various

324: concentrations with increasing correlation length largely collapsed to

325: scaling functions belonging to the LL and IDOS separately. The scaling

326: factor as a function of concentration plays a key role in the scaling

327: behavior of the LL and IDOS data. The scaling behavior obtained here holds

328: for a more limited range of $w$ values than that obtained by Russ \emph{et}$%

329: \,$\emph{al.} \cite{rus02}. For instance we did not see any scaling behavior

330: when $w>10^{-4}$, whereas they presented the scaling behavior for $%

331: w=10^{-1}-10^{-5}$. For the scaling near the limiting concentrations, some

332: scattered data are obtained for lower values of the scaling variable $X$. We

333: have used scaling arguments and established the scaling functions for both

334: the LL and the IDOS using the CPA results. The CPA results reproduced data

335: for the ILL rather qualitatively but surprisingly for the IDOS it reproduced

336: the data rather well. This study revealed that if the second moment of the 

337: correlated random potential is calculated, then the CPA results can be

338: implemented as the way presented here even though the CPA was developed

339: mainly for noncorrelated random potentials.

340: 

341: We have benefited from discussions with Prof. D. L. Huber. This work is

342: partially sponsored by the Scientific end Technical Research Council of

343: Turkey (TUBITAK).

344: 

345: \begin{thebibliography}{99}

346: 

347: \bibitem{hy04} H. Yamada, Phys. Rev. B \textbf{69}, 14205(2004).

348: 

349: \bibitem{lyra02} R. P. A. Lima and M. L. Lyra, Phys. Rev. B \textbf{65},

350: 104416 (2002).

351: 

352: \bibitem{lyra98} J. V. Kantelhardt, S. Russ, A. Bunde, S. Havlin and I Webman, 

353: Phys. Rev. Lett. \textbf{84}, 198 (2000) and

354: F. A. B. F. de Moura and M. L. Lyra, Phys. Rev.Lett.\textbf{%

355: 81}, 3735 (1998).

356: 

357: \bibitem{izra99} L. Tessieri, J. Phys. A \textbf{35}, 9585 (2002) and

358: F. M. Izrailev and A. A. Krokhin, Phys. Rev. Lett.\textbf{82%

359: }, 4062(1999).

360: 

361: \bibitem{deyc03} L. I. Deyc, M. V. Erementchouk, and A. A. Lyansky, Phys.

362: Rev. B \textbf{67}, 24205(2003).

363: 

364: \bibitem{ul} U. Kuhl, H. J. St{\"o}ckmann Physica \textbf{9 E} 384 (2001).

365: 

366: \bibitem{loc} F. Kuckar, H. Heinrich and G. Bauer, \textit{Localization and

367: confinement of electrons in semiconductors} (Springer-Verlac,Berlin,1997)

368: 

369: \bibitem{bay} M. Bayindir, S. Tariseven and E. Ozbay Appl. Phys. A \textbf{72%

370: } 117 (2001).

371: 

372: \bibitem{rus02} S. Russ, Phys. Rev. B \textbf{66}, 12204(2002).

373: 

374: \bibitem{rus98} S.~Russ, S.~Havlin and I.~Webman, Phil. Mag. B \textbf{77},

375: 1449(1998).

376: 

377: \bibitem{ishii} K. Ishii, Prog. Theor. Phys. Suppl. \textbf{53}, 77 (1973).

378: 

379: \bibitem{avgin93} I. Avgin and D. L. Huber, Phys. Rev. B \textbf{48}, 13625

380: (1993).

381: 

382: \bibitem{pimentel89} I. R. Pimentel and R. B. Stinchcombe, Phys. Rev. B 

383: \textbf{40}, 4947 (1989).

384: 

385: 

386: \bibitem{avgin02} I. Avgin and D. L. Huber, Phys. Rev. B \textbf{66}, 12405

387: (2002).

388: 

389: \bibitem{boukahil89} A. Boukahil and D. L. Huber, Phys. Rev. B \textbf{40},

390: 4638 (1989) and Phys. Rev. B \textbf{50}, 2978 (1994).

391: 

392: \bibitem{vulp89} A. Crisanti, G. Paladin, and G. Vulpiani, Phys. Rev. A 

393: \textbf{39}, 6491 (1989).

394: 

395: \bibitem{thouless} D. J. Thouless, J. Phys. C \textbf{5}, 77 (1972).

396: 

397: \bibitem{derrida84} B. Derrida and E. Gardner, J. Phyique \textbf{45}, 1283

398: (1984).

399: 

400: \bibitem{dean} P. Dean, Proc. Phys. Soc. (London) \textbf{73}, 413 (1959).

401: 

402: \bibitem{avgin98} I. Avgin and D. L. Huber Phys. Rev. B \textbf{57}, 

403: 8269 (1998) and I. Avgin and D. L. Huber Phys. Rev. B. \textbf{56}, 2320

404: (1997).

405: 

406: \bibitem{avgin96} I. Avgin, Phys. Rev. B \textbf{59}, 13554 (1999) and I.

407: Avgin, J.\ Phys.\ Cond.\ Matter \textbf{8}, 8379 (1996).

408: 

409: \bibitem{avgin2} I. Avgin, E. J. Phys. B \textbf{29}, 437 (2002).

410: \end{thebibliography}

411: 

412: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

413: 

414: \end{document}

415: