1: %
2: \documentclass[12pt]{iopart}
3: %\newcommand{\gguide}{{\it Preparing graphics for IOP journals}}
4: %Uncomment next line if AMS fonts required
5: \usepackage{iopams}
6: \usepackage{graphicx}
7: \begin{document}
8:
9: \title
10: [Critical conductance and wavefunction of two dimensional Ando model]
11: {Critical regime of two dimensional Ando model: relation between critical conductance and
12: fractal dimension of electronic eigenstates}
13:
14: \author{P. Marko\v{s}$^1$, L. Schweitzer$^2$}
15:
16: \address{$^1$Institute of Physics, Slovak Academy of Sciences, D\'ubravsk\'a
17: cesta 9, 854 11 Bratislava, Slovakia}
18: \address{$^2$Physikalisch-Technische Bundesanstalt, Bundesallee 100, 38116
19: Braunschweig, Germany}
20: \eads{\mailto{peter.markos@savba.sk}, ~\mailto{ludwig.schweitzer@ptb.de}}
21: \begin{abstract}
22: The critical two-terminal conductance $g_c$ and the spatial fluctuations of
23: critical eigenstates
24: are investigated for a disordered two dimensional model of non-interacting
25: electrons subject to spin-orbit scattering (Ando model). For square samples,
26: we verify numerically the relation $\sigma_c=1/[2\pi(2-D(1))]\,e^2/h$ between
27: critical conductivity $\sigma_c=g_c=(1.42\pm 0.005)\,e^2/h$ and the fractal
28: information dimension of the electron wave function, $D(1)=1.889\pm 0.001$.
29: Through a detailed numerical scaling analysis of the two-terminal conductance
30: we also estimate the critical exponent $\nu=2.80\pm 0.04$ that governs the
31: quantum phase transition.
32: \end{abstract}
33:
34: %Uncomment for PACS numbers title message
35: %\pacs{00.00, 20.00, 42.10}
36: % Keywords required only for MST, PB, PMB, PM, JOA, JOB?
37: %\vspace{2pc}
38: %\noindent{\it Keywords}: Article preparation, IOP journals
39: % Uncomment for Submitted to journal title message
40: %\submitto{\JPA}
41: % Comment out if separate title page not required
42: \maketitle
43:
44:
45:
46: \section{Introduction}
47: Quantum phase transitions in two dimensional (2d) disordered systems have been the
48: subject of continuing research for many decades. In particular, systems of
49: non-interacting particles that are invariant under time reversal, but lack the
50: symmetry of spin rotation are of special importance due to their property of a
51: complete Anderson transition \cite{HLN80,Mac84,Mac85,EZ87,Weg89,And89}.
52: Thus, 2d systems with spin-orbit interaction,
53: which belong to the symplectic symmetry class \cite{ZP88},
54: exhibit an energy spectrum with regions of either localized or extended states
55: separated by quantum critical points at which the correlation length diverges
56: according to $\xi(E)\sim |E-E_c|^{-\nu}$ with a universal exponent $\nu$.
57: The energetical position of these points $E_c$ depend on disorder strength and
58: spin-orbit interaction. The corresponding critical eigenstates were found to be
59: multi-fractal objects showing strong amplitude fluctuations as well as
60: long-range spatial and energetical correlations \cite{CDEN93,Sch95,SP99}.
61:
62: Our knowledge about the critical properties like critical energy and disorder,
63: or about the exponents that govern their scaling, originate mainly
64: from numerical investigations of quantities like localization length
65: \cite{Mac85,EZ87,And89,Fea91,Eva95,ASO02} or spectral correlations of
66: eigenvalues \cite{SZ97} which both are not directly accessible to experimental
67: detection. The aim of the present paper is to present scaling results of the
68: two-terminal conductance \cite{ES81}, a physical quantity that can easily be
69: measured in experiment, and to examine a recently proposed relation between
70: the average critical conductance and the generalized fractal dimensions of
71: critical eigenstates. Therefore, we first present our numerical data for the
72: two-terminal conductance and obtain the critical parameters of the model by a
73: finite-size scaling analysis.
74: In the metallic regime, the mean conductance is expected to increase
75: with increasing system size while a decrease is usually found in the localized
76: regime. At the critical point, the mean conductance $g_c$ should be a universal,
77: system size independent quantity and equal to the critical conductivity
78: $\sigma$. Hence, the size scaling of our numerical data enables us to
79: estimate the critical disorder $W_c$, and to calculate the critical
80: conductance $g_c$.
81: We also find the critical exponent $\nu$ which governs the energy and disorder
82: dependence of the conductance near the critical point.
83:
84: Then we discuss the fractal properties of critical wave functions and
85: calculate fractal dimensions $D(q)$.
86: Non-uniform spatial amplitude distributions of electron wave functions
87: that exhibit multi-fractal properties, are a characteristic feature of the critical
88: regime at metal-insulator transitions. It is well known that the corresponding
89: generalized multi-fractal dimensions influence the time dependent transport
90: properties \cite{BHS96}. Here, we show that they also determine the
91: conductance at a quantum critical point.
92: Our main result is a confirmation of the relation $D(q)=2-q/[\beta 4\pi^2
93: \varrho D \hbar]$ between the fractal dimensions and the conductivity (per spin
94: direction), $\sigma=e^2 \varrho D$, proposed by Fal'ko and Efetov
95: \cite{FE95a}. We also examine the validity
96: of Janssen's formula \cite{Jan94,Jan98} which relates the
97: Lipschitz-H\"older exponent $\alpha_0$ of the maximum of the multi-fractal
98: distribution $f(\alpha)$ to the scaling parameter
99: $\Lambda_c=1/[\pi(\alpha_0-d)]$
100: for quasi-1d systems.
101: We find, however, our result for the symplectic Ando model ($d=2$) to be at
102: variance with this proposal.
103:
104: \section{The model and methods}
105:
106: \subsection{Hamiltonian}
107:
108: We study the two dimensional (2d) Ando model \cite{And89} defined on a square
109: lattice by the Hamiltonian
110: \begin{equation}\label{hamiltonian}
111: {\cal H}=W\sum_n \varepsilon_nc^{\dag}_{n}c_{n}^{}+\sum_{[nn']}V_{nn'}c^{\dag}_{n}c_{n'}^{}.
112: \end{equation}
113: The $\varepsilon_n$ are uncorrelated random on-site energies chosen
114: according to a box probability distribution in the range $|\varepsilon_n|\le 1/2$.
115: The parameter $W$ measures the strength of the disorder.
116: The spin dependent hopping
117: terms between nearest neighbour sites in the direction parallel and
118: perpendicular to the current flow are
119: \begin{equation}
120: V_\parallel=V\left(
121: \begin{array}{rr}
122: c & s\\
123: -s & c
124: \end{array}
125: \right),
126: ~~\textrm{and}~~~V_\perp=V\left(
127: \begin{array}{rr}
128: c & -is\\
129: -is & c
130: \end{array}
131: \right),
132: \end{equation}
133: with $c^2+s^2=1$.
134: The parameter $s$ determines hopping accompanied with a change of the
135: spin of the electron. As usual, it is set to $s=0.5$. The special case $s=0$
136: corresponds to a system with orthogonal symmetry which exhibits no
137: metal-insulator transition in 2d.
138: Energies are measured in units $V=1$ and lengths
139: in units of the lattice constant $a=1$.
140:
141: It is well known that for Fermi energy $E_F\approx 0$, the Ando model exhibits
142: a metal-insulator transition as a function of disorder
143: at $W=W_c\approx 5.8$
144: \cite{And89,Fea91}.
145: The critical exponent for various $2d$ systems with symplectic symmetry
146: has been estimated in numerous works using different numerical methods
147: \cite{Fea91,SZ97,MJH98,Min98,YO98,ASO02,ASO04},
148: leading to rather inconsistent results. The critical conductance
149: was studied in \cite{OSK04}. We will discuss some of these results later together
150: with our own results.
151:
152:
153: \subsection{Conductance}
154:
155: The two-terminal conductance of a particular sample is calculated from
156: \begin{equation}\label{Landauer}
157: g=\Tr t^\dag t,
158: \end{equation}
159: where $t$ is the transmission matrix. Its elements $t_{\alpha\beta}$ determine the
160: transmission amplitude from channel $\alpha$ into channel $\beta$.
161: Two semi-infinite ideal leads are attached to the disordered sample and
162: periodic boundary conditions are applied in the transverse direction.
163: The Fermi energy is fixed at $E_F=0.01$ and spin-orbit scattering is assumed
164: to be absent in the leads.
165:
166: We use the algorithm of Pendry \etal \cite{PMR92} for our numerical calculations.
167: The size of the sample varied from $L=20$ to $L=200$.
168: We concentrate on the analysis of the $L$-dependence of the mean conductance
169: in the neighbourhood of the critical point at $W_c\sim 5.80$. Because of the absence of
170: self-averaging of the conductance, we need to analyze data for a large number
171: of macroscopically identical samples which differ only by the respective microscopic
172: realization of the disorder.
173: For each disorder strength, $5.71\le W\le 5.96$, and each system size, we
174: collected a statistical ensemble of $N_{\rm stat}\ge 10^5$
175: samples and calculate the mean conductance $\langle g\rangle$ and variance
176: $\textrm{var} g=\langle g^2\rangle -\langle g\rangle^2$. Here,
177: $\langle\dots\rangle$ means averaging over a statistical ensemble.
178: Mean values $\langle g\rangle$
179: with the uncertainty $\delta g=(\textrm{var} g/N_{\rm stat})^{1/2}$ are
180: used in the scaling analysis.
181: As a typical mean value $\langle g\rangle\approx 1.4$ and $\textrm{var}
182: g \approx 0.36$, the relative uncertainty
183: $\delta g/\langle g\rangle$ of our data is of order of 0.0015. Typical
184: numerical data are shown in \fref{fig1} and \ref{fig2}.
185:
186: \begin{figure}[t!]
187: \begin{center}
188: \includegraphics[clip,width=0.65\textwidth]{ms-fig1.eps}
189: \end{center}
190: \caption{The mean conductance $\langle g(L)\rangle$ as a function of the
191: system size $L$ for
192: various values of disorder strength $W$. Within the interval of system sizes
193: $30\le L\le 200$, $\langle g(L)\rangle$ is almost constant for disorder
194: $W=5.84$. This indicates that
195: finite size effects are absent and irrelevant scaling terms negligible.
196: Solid lines are fits to \eref{scaling} with critical parameters
197: $g_c=1.42$, $W_c=5.838 $, $\nu=2.80$ and $A=-0.23$.
198: }
199: \label{fig1}
200: \end{figure}
201:
202:
203: \subsection{Critical wave function}
204: Electron wave functions at quantum critical points of disordered systems exhibit
205: a peculiar spatial structure \cite{Weg80a,Aok83}. For our symplectic model,
206: the critical wave functions for systems sizes of up to $L=260$ were obtained
207: by direct diagonalization using a Lanczos-algorithm. The fractality of the
208: spatial fluctuations of the modulus of the normalized eigenstates $\psi_E(r)$
209: was determined from the power-law scaling of the $q$-th moment defined
210: via a 'box-probability'
211: \begin{equation}
212: P(q,\lambda)=\sum_i^{N(l)}(\sum_{r\in\Omega_i(l)}|\psi_E(r)|^2)^q
213: \sim \lambda^{\tau(q)}.
214: \label{boxprob}
215: \end{equation}
216: The generalized fractal dimensions $D(q)=\tau(q)/(q-1)$ or
217: the so called $f(\alpha(q))$-distri\-bution
218: was thus derived \cite{HP83,Hea86,CJ89},
219: where $\Omega_i(l)$ is the the i-th box of size $l=\lambda L$.
220: The $\tau(q)$ and $f(\alpha(q))$ are related by a Legendre transform.
221:
222:
223: \subsection{Scaling analysis}
224: \label{fss}
225: At the critical point $W\equiv W_c$, the conductance does not depend on
226: the system size. In accordance with the scaling theory of localization
227: \cite{MK81}, we assume that in the
228: neighbourhood of the critical point the $L$ dependence of $g$ is governed by
229: the critical exponent $\nu$,
230: \begin{equation}\label{scaling}
231: g_{\rm sc}(W,L)=g_c+A(W-W_c)L^{1/\nu}.
232: \end{equation}
233: We collected numerical data $g(W,L)$ for more than $N=200$ values of $W$ and $L$
234: and obtain the critical parameters, $g_c$, $W_c$ and $\nu$ from a fit of the
235: numerical data to the scaling ansatz (\ref{scaling}).
236: We made sure that more sophisticated fits \cite{SO99} which include higher
237: order terms in disorder are not needed because
238: our data exhibit already a perfect linear $W$-dependence (see \fref{fig2}),
239: Also, subject to the uncertainty of our raw data, we found that possible
240: irrelevant scaling fields \cite{SO99,Mac94} are weak and do only marginally
241: influence our results.
242:
243: \begin{figure}[t]
244: \begin{center}
245: \includegraphics[clip,width=0.5\textwidth]{ms-fig2.eps}
246: \end{center}
247: \caption{Disorder dependence of the mean conductance $\langle g(W)\rangle$ for
248: $L=70$, 100, 140, and 200. Solid lines are linear fits \eref{linear}.
249: }
250: \label{fig2}
251: \end{figure}
252:
253: The most simple analysis of critical parameters can be done as follows: first,
254: we calculate parameters of the linear $W$-dependence of the conductance
255: \begin{equation}\label{linear}
256: g(W,L)=g_0(L)+Wg_1(L).
257: \end{equation}
258: Comparing \eref{linear} with \eref{scaling}, we see that the slope $g_1$ depends
259: on the system size as
260: \begin{equation}\label{slope}
261: g_1(L)\propto L^{1/\nu}.
262: \end{equation}
263: Then, a power-law fit $g_1(L)$ \textit{vs.\@} $L$ gives us the critical exponent.
264:
265: We also minimize numerically the function
266: \begin{equation}\label{F}
267: F=\frac{1}{N}\sum_{LW} \frac{1}{(\delta g)^2}
268: \left[\langle g(W,L)\rangle-g_{\rm sc}(W,L)\right]^2
269: \end{equation}
270: with respect to unknown parameters $W_c$, $\nu$ and $g_c$ and $A$. To estimate
271: the accuracy of the final result, we repeat $N_m$ times the minimization of the
272: function $F$ with input data $\langle g(W,L)\rangle$
273: randomly fluctuating within their respective error bars.
274: To understand the role of the finite size effects, we use in the scaling analysis
275: only data for systems with sizes $L_{\rm min}\le L\le L_{\rm max}$ and study
276: how the obtained critical parameters depend on the choice of $L_{\rm min}$ and
277: $L_{\rm max}$.
278:
279: \begin{figure}[t!]
280: \begin{center}
281: \includegraphics[clip,width=0.5\textwidth]{ms-q1d.eps}
282: \end{center}
283: \caption{%
284: The scaling parameter $\Lambda$ calculated for quasi-1d geometry versus
285: disorder strength for varying system width $M$. The data confirm a
286: critical disorder $W_c\approx 5.835$ and $\Lambda_c\simeq 1.87\pm 0.02$.
287: }
288: \label{q1d}
289: \end{figure}
290:
291:
292: \section{Results and discussions}
293:
294: \subsection{Critical parameters}
295: Our analysis showed that neither the critical disorder nor the critical
296: conductance depend on the choice of the minimal and maximal system size,
297: $L_{\rm min}$ and $L_{\rm max}$. From all the data we concluded that
298: \begin{equation}
299: W_c=5.838\pm 0.007
300: \end{equation}
301: The estimate of the critical disorder differs considerably from the result
302: of Fastenrath \etal \cite{Fea91} ($W_c=5.74$),
303: obtained from scaling analysis of the localization length of quasi-1d systems.
304: Since previous results were derived from small system's widths,
305: the discrepancy could be explained as an effect of finite size corrections.
306: To support this assumption, we performed numerical simulations
307: for quasi-1d systems of size $M\times L$. Figure \ref{q1d} shows the
308: calculated data for the ratio $\Lambda=\lambda_M/M$,
309: the localization length $\lambda_M$ divided by the system width $M$
310: \cite{MK81,PS81}. For our purposes, we run $\Lambda$
311: with an accuracy of only 1\%. This was sufficient to confirm
312: that our estimation of the critical disorder derived from the conductance
313: calculations indeed coincides with the one obtained from the localization length.
314: Also, we get an estimate of the critical value $\Lambda_c$
315: of the scaling parameter $\Lambda$,
316: \begin{equation}\label{lambdac}
317: \Lambda_c\simeq 1.87\pm 0.02
318: \end{equation}
319: which agrees well with a recent result of Asada \etal \cite{ASO04}
320: obtained for a SU(2) model.
321:
322: From our conductance data, the critical exponent $\nu$ was obtained for the first
323: time using finite size scaling of the calculated electrical two-terminal conductance,
324: which is, unlike the localization length, an easy way to measure physical quantity.
325: With the finite size scaling analysis described in section~(\ref{fss}), our
326: estimation of the critical exponent gives
327: \begin{equation}\label{nu}
328: \nu=2.8\pm 0.04,
329: \end{equation}
330: which is close to a numerical result $\nu=2.746\pm 0.009$ for a SU(2) model
331: published recently \cite{ASO04}, but considerably larger than $\nu=2.05\pm 0.08$
332: \cite{And89} and smaller than $\nu=2.88\pm 0.15$ as obtained in \cite{Min98}.
333: In contrast to the critical conductance and critical disorder, the critical
334: exponent is sensitive to the system size as can be seen in \fref{fig3}.
335: The estimate of $\nu$ is therefore highly non-trivial and error bars obtained from
336: various fit procedures differ.
337:
338: \begin{figure}[t!]
339: \begin{center}
340: \includegraphics[clip,width=0.5\textwidth]{ms-fig3.eps}
341: \end{center}
342: \caption{%
343: The critical exponent $\nu$ obtained from numerical data
344: with $L_{\rm min}\le L\le L_{\rm max}$. Full symbols
345: shows how $\nu$ depends on the choice of $L_{\rm max}$ for
346: fixed minimal system size $L_{\rm min}$. Open symbols show
347: how $\nu$ depends on $L_{\rm min}$ for fixed $L_{\rm max}$,
348: $L_{\rm max}=140$ ({\large$\circ$}),
349: and $L_{\rm max}=200$ ($\bigtriangleup$). The shaded area
350: highlights our estimate of the critical exponent $\nu=2.80\pm 0.04$.
351: }
352: \label{fig3}
353: \end{figure}
354:
355: \subsection{Critical conductance}
356: The critical two-terminal conductance obtained by us for square samples with
357: periodic boundary conditions in the transverse direction,
358: \begin{equation}\label{gc}
359: g_c=1.42\pm 0.005,
360: \end{equation}
361: can be compared with the result for a SU(2) model studied recently in \cite{OSK04}.
362: From the corresponding numerical data for size $L\le 97$ (T. Ohtsuki, private
363: communication) we calculated $\lim_{L\to\infty}g(L)=1.411$, which is in
364: very good agreement with our present result.
365:
366: Assuming for square systems $\sigma \equiv g/L^{(2-d)}$, we have at the
367: critical point $g_c=\sigma_c$. As mentioned above, for $L>20$ our critical
368: conductance data do not exhibit any finite size effect. We do not expect that
369: the numerical analysis causes any inaccuracy of the estimation of $g_c$.
370: Still, when comparing $g_c$ with $\sigma_c$, we have to
371: keep in mind that small corrections due to the properties of leads (we assume
372: perfect leads without spin-orbit scattering) might be responsible for a small
373: difference between $\sigma_c$ and $g_c$. However, we do not expect this
374: difference to be larger than $\sim L^{-1}$ and neglect it for further purposes.
375:
376:
377: \Fref{fig4} shows the critical conductance distribution $p(g)$. We present
378: data for disorder $W=5.84$ and system size $82\le L \le 200$.
379: We see that indeed $p(g)$ does not depend on the system size at the
380: critical point.
381: Besides the
382: shape of the distribution, which is known also from previous studies,
383: we want to point out the non-analytical behaviour
384: of the distribution at the point $g=g_{\rm non-an}$ which
385: is slightly larger than 2. This agrees with a qualitative
386: estimation of Muttalib \etal \cite{MWGG03},
387: although their work concerns quasi one dimensional
388: systems with orthogonal symmetry.
389:
390: \begin{figure}[t!]
391: \begin{center}
392: \includegraphics[clip,width=0.5\textwidth]{ms-fig4.eps}
393: \end{center}
394: \caption{Critical conductance distribution $p(g)$.
395: Inset: the derivative $\partial p(g)/\partial g$ which shows $g_{\rm
396: non-an}$ to be slightly larger than 2.
397: }
398: \label{fig4}
399: \end{figure}
400:
401:
402: \subsection{Fractal dimensions and Fal'ko-Efetov relation}
403:
404: In figure~\ref{D1scal} the $f(\alpha(q))$-distributions of 10 critical
405: eigenfunctions of systems with $L=260$ are shown together with a parabolic fit
406: $f(\alpha(q))=d-(\alpha(q)-\alpha_0)^2/[4(\alpha_0-d)]$
407: with $\alpha_0=\alpha(q=0)=2.107\pm 0.005$. This value is lesser than
408: the one obtained previously for smaller sample size $L=150$ \cite{Sch95}.
409: The inset exhibits the logarithm of $P(1,\lambda)$ plotted versus $\ln
410: \lambda$ for $2\le l \le 25$. The data belonging to critical eigenstates
411: obtained from 10 disorder realizations clearly obey power-law scaling.
412:
413: From the multi-fractal analysis of critical eigenfunctions of $L=260$ samples,
414: we obtained an information dimension $D(1)=1.889\pm 0.001$
415: using 10 different realizations with disorder strength $W=5.83$.
416: With this data, we checked the validity of a formula suggested by Fal'ko and
417: Efetov \cite{FE95a} which should hold in the limit $1\le \varrho D h$,
418: \begin{equation}\label{sigma-dq}
419: D(q) = 2-\frac{q}{\beta 4\pi^2\varrho D \hbar},
420: \label{FalEfe}
421: \end{equation}
422: where $\beta=1/2,~1$ and 2 for or orthogonal, unitary and symplectic symmetry, respectively.
423: With $\sigma=e^2\varrho D$ (diffusion constant $D$ and density
424: of states $\varrho$) we get
425: \begin{equation}\label{efetov}
426: \sigma=\displaystyle{\frac{q}{2\pi\beta[2-D(q)]}\frac{e^2}{h}}.
427: \end{equation}
428:
429: \begin{figure}[t]
430: \begin{center}
431: \includegraphics[clip,width=0.55\textwidth]{falfa_D1.eps}
432: \end{center}
433: \caption{%
434: The $f(\alpha(q))$-distributions of 10 critical eigenstates at $E\approx 0$
435: and $W_c=5.83$ for $q=0.0,\pm 0.2, \pm 0.5, \pm 0.8, \pm 1.0, \pm 1.5, \pm
436: 2.0$, and $\pm 3.0$. The parabolic fit is determined by
437: $\alpha_0=2.107\pm 5\cdot 10^{-3}$.
438: The inset shows the scaling of $\ln P(1,\lambda)$ vs.\@
439: logarithm of the box size $\lambda$. Data are obtained from ten $L=260$
440: samples with different disorder realization. The dashed line has
441: steepness $\tau(1)=1.889$.
442: }
443: \label{D1scal}
444: \end{figure}
445: Taking $q=1$, we found our results for $\sigma_c$ and $D(1)$ to be in very
446: good agreement with (\ref{efetov}).
447: Therefore, this relation is fulfilled also for the symplectic symmetry in a similar
448: manner as found previously for the quantum Hall case \cite{SM05}. In comparing
449: these results one has to keep in mind that in the present paper two spin
450: channels were considered, i.e., the conductance has to be divided by 2, whereas
451: in the QHE case only one spin direction was taken into account.
452: Note also that the value of the disorder used in the multi-fractal
453: analysis, $W=5.83$, differs slightly form the critical disorder
454: $W_c=5.838$. Therefore, the conductance $g(W=5.83)\approx 1.43$ mentioned
455: in ref.~\cite{SM05} is a little larger than $g_c$.
456: The validity of the proposed linear relationship (\ref{FalEfe}) holds at least
457: for $q \le 1.5$ which can be seen in figure~\ref{Dq} where $D(q)$ calculated
458: according to (\ref{boxprob}) is plotted versus $q$.
459:
460:
461: \subsection{Comparison with Janssen's formula}
462: A formula that connects fractal properties of critical eigenstates in
463: square samples with the finite size scaling variable $\Lambda=\lambda_M/M$ of
464: quasi-1d systems was conjectured by Janssen \cite{Jan94,Jan98}. At the
465: critical point, $\Lambda_c$ is scale independent and depends only on the
466: Lipschitz-H\"older exponent $\alpha_0$ and the spatial dimension of the
467: system $d=2$
468: \begin{equation}\label{Lambda}
469: \Lambda_c=\displaystyle{\frac{1}{\pi(\alpha_0-d)}}.
470: \end{equation}
471: \begin{figure}[t]
472: \begin{center}
473: \includegraphics[clip,width=0.5\textwidth]{D_q_L2U25_M260W5p83.eps}
474: \end{center}
475: \caption{%
476: The generalized fractal dimension $D(q)$ of a critical eigenfunction as a
477: function of $q$ calculated via the box-probability (\ref{boxprob}). The
478: linearity $D(q)=2-q/k$ holds for $q\lesssim 1.5$, with $k=0.112$ leading to
479: $\sigma_c=0.71\,e^2/h$.
480: }
481: \label{Dq}
482: \end{figure}
483: Equation (\ref{Lambda}) was reported to hold for 2d disordered systems in the quantum
484: Hall regime \cite{Jan94,Huc94}. In the present symplectic system, however, it is not
485: met by our results.
486: Also, assuming the parabolic approximation to be valid, we can use the relation
487: between $D(1)$ and $\alpha_0$,
488: \begin{equation}
489: D(1)=\alpha(1)=4-\alpha_0,
490: \end{equation}
491: and connect (\ref{efetov}) and (\ref{Lambda}) to obtain a
492: relation between $\Lambda_c$ and the critical conductivity
493: \begin{equation}\label{univ}
494: \sigma_c=\frac{\Lambda_c}{2\beta}\frac{e^2}{h}.
495: \end{equation}
496: Again, contrary to what is observed in the QHE regime, the relation
497: (\ref{univ})
498: is not satisfied for a symplectic system described by the Ando model.
499: A possible reason for this failure might be that in the present 2d model, the
500: scaling parameter $\Lambda_c$ may not represent the typical localization
501: length of the system as has been assumed in \cite{Jan94}.
502:
503:
504: \section{Conclusion}
505:
506: We studied the electrical two-terminal conductance and the spatial
507: fluctuations of electron eigenfunctions near the metal-insulator transition
508: of the two dimensional Ando model. Using finite-size scaling, we obtained for
509: this symplectic model a critical exponent $\nu=2.8\pm 0.04$, which governs
510: the size dependence of the conductance in the critical regime,
511: and a critical disorder $W_c=5.838\pm 0.007$.
512: Our results for the critical conductance and for the fractal dimension of
513: critical eigenstates confirm the validity of Fal'ko and Efetov's prediction
514: (\ref{sigma-dq}). However, comparison of the fractal dimensions with the
515: critical value of the finite size scaling parameter $\Lambda_c$, calculated
516: for quasi one-dimensional systems, indicates that Janssen's formula
517: (\ref{Lambda}) is not fulfilled by our data in the present model.
518:
519: Unlike the localization
520: length, the two-terminal conductance should be easily accessible in experiments.
521: Our numerical confirmation of the relation between fractal dimensions
522: of critical eigenfunctions and the critical conductance provides an
523: additional argument for universality at the metal-insulator transition
524: in 2d symplectic models.
525:
526:
527: \ack
528: We thank T. Ohtsuki for sending us numerical data of the
529: critical conductance from the SU(2) model.
530: PM thanks APVT Grant No. 51-021602 for partial financial support, and LS
531: for the hospitality of the Slovak Academy of Sciences.
532:
533: \section*{References}
534: %\bibliographystyle{iopart-num}
535: %\bibliography{paPers_DB,ludwig}
536: \providecommand{\newblock}{}
537: \begin{thebibliography}{10}
538: \expandafter\ifx\csname url\endcsname\relax
539: \def\url#1{{\tt #1}}\fi
540: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi
541: \providecommand{\eprint}[2][]{\url{#2}}
542: % Bibliography created with iopart-num.bst, v1.0
543:
544: \bibitem{HLN80}
545: Hikami S, Larkin A~I and Nagaoka Y 1980 {\em Prog. Theor. Phys.\/} {\bf 63}(2)
546: 707--710
547:
548: \bibitem{Mac84}
549: MacKinnon A 1984 in B~Kramer, G~Bergmann and Y~Bruynseraede, eds, {\em
550: Localization, Interaction, and Transport Phenomena\/} (Berlin Heidelberg:
551: Springer-Verlag) vol~61 of {\em Springer Series in Solid-State Sciences\/} pp
552: 90--98
553:
554: \bibitem{Mac85}
555: MacKinnon A 1985 {\em Z. Phys. B\/} {\bf 59} 385--390
556:
557: \bibitem{EZ87}
558: Evangelou S~N and Ziman T 1987 {\em J. Phys. C: Solid State Phys.\/} {\bf 20}
559: L235--L240
560:
561: \bibitem{Weg89}
562: Wegner F 1989 {\em Nuclear Physics\/} {\bf B316}(663-678)
563:
564: \bibitem{And89}
565: Ando T 1989 {\em Phys. Rev. B\/} {\bf 40}(8) 5325--5339
566:
567: \bibitem{ZP88}
568: Zanon N and Pichard J-L 1988 {\em J. Phys. France\/} {\bf 49} (6) 907-920
569:
570: \bibitem{CDEN93}
571: Chalker J~T, Daniell G~J, Evangelou S~N and Nahm I~H 1993 {\em J. Phys.:
572: Condens. Matter\/} {\bf 5} 485--490
573:
574: \bibitem{Sch95}
575: Schweitzer L 1995 {\em J. Phys.: Condens. Matter\/} {\bf 7} L281--L285
576:
577: \bibitem{SP99}
578: Schweitzer L and Potempa H 1999 {\em Physica A\/} {\bf 266} 486--491
579:
580: \bibitem{BHS96}
581: Brandes T, Huckestein B and Schweitzer L 1996 {\em Annalen der Physik\/} {\bf 5} 633-651
582:
583: \bibitem{Fea91}
584: Fastenrath U, Adams G, Bundschuh R, Hermes T, Raab B, Schlosser I, Wehner T and
585: Wichmann T 1991 {\em Physica A\/} {\bf 172} 302--308
586:
587: \bibitem{Eva95}
588: Evangelou S~N 1995 {\em Phys. Rev. Lett.\/} {\bf 75}(13) 2550--2553
589:
590: \bibitem{ASO02}
591: Asada Y, Slevin K and Ohtsuki T 2002 {\em Phys. Rev. Lett.\/} {\bf 89}(25)
592: 256601
593:
594: \bibitem{SZ97}
595: Schweitzer L and {Kh Zharekeshev} I 1997 {\em J. Phys.:\ Condens.\ Matter\/}
596: {\bf 9} L441--L445
597:
598: \bibitem{ES81}
599: Economou E~N and Soukoulis C~M 1981 {\em Phys. Rev. Lett.\/} {\bf 46}(9) 618
600:
601: \bibitem{FE95a}
602: Fal'ko V~I and Efetov K~B 1995 {\em Phys. Rev. B\/} {\bf 52}(24) 17413--17429
603:
604: \bibitem{Jan94}
605: Janssen M 1994 {\em Int. J. Mod. Phys. B\/} {\bf 8}(8) 943--984
606:
607: \bibitem{Jan98}
608: Janssen M 1998 {\em Physic Reports\/} {\bf 295}(1-2) 1--91
609:
610: \bibitem{MJH98}
611: Merkt R, Janssen M and Huckestein B 1998 {\em Phys. Rev. B\/} {\bf 58}(8)
612: 4394--4405
613:
614: \bibitem{Min98}
615: Minakuchi K 1998 {\em Phys. Rev. B\/} {\bf 58}(15) 9627--9630
616:
617: \bibitem{YO98}
618: Yakubo K and Ono M 1998 {\em Phys. Rev. B\/} {\bf 58}(15) 9767--9772
619:
620: \bibitem{ASO04}
621: Asada Y, Slevin K and Ohtsuki T 2004 {\em Phys. Rev. B\/} {\bf 70} 035115
622:
623: \bibitem{OSK04}
624: Ohtsuki T, Slevin K and Kramer B 2004 {\em Physica E\/} {\bf 22} 248--251
625:
626: \bibitem{PMR92}
627: Pendry J~B, MacKinnon A and Roberts P~J 1992 {\em Proc. R. Soc. Lond. A\/} {\bf
628: 437} 67--83
629:
630: \bibitem{Weg80a}
631: Wegner F 1980 {\em Z. Physik B\/} {\bf 36} 209--214
632:
633: \bibitem{Aok83}
634: Aoki H 1983 {\em J. Phys. C: Solid State Phys.\/} {\bf 16} L205--L208
635:
636: \bibitem{HP83}
637: Hentschel H~G~E and Procaccia I 1983 {\em Physica D\/} {\bf 8} 435--444
638:
639: \bibitem{Hea86}
640: Halsey T~C, Jensen M~H, Kadanoff L~P, Procaccia I and Shraiman B~I 1986 {\em
641: Phys. Rev. A\/} {\bf 33}(2) 1141--1151
642:
643: \bibitem{CJ89}
644: Chhabra A and Jensen R~V 1989 {\em Phys. Rev. Lett.\/} {\bf 62}(12) 1327--1330
645:
646: \bibitem{MK81}
647: MacKinnon A and Kramer B 1981 {\em Phys. Rev. Lett.\/} {\bf 47}(21) 1546--1549
648:
649: \bibitem{SO99}
650: Slevin K and Ohtsuki T 1999 {\em Phys. Rev. Lett.\/} {\bf 82}(2) 382--385
651:
652: \bibitem{Mac94}
653: MacKinnon A 1994 {\em J. Phys.: Condens. Matter\/} {\bf 6} 2511--2518
654:
655: \bibitem{PS81}
656: Pichard J~L and Sarma G 1981 {\em J. Phys. C: Solid State Phys.\/} {\bf 14}
657: L617--L625
658:
659: \bibitem{MWGG03}
660: Muttalib K~A, W\"olfle P, Garc\'{i}a-Mart\'{i}n A and Gopar V~A 2003 {\em
661: Europhys. Lett.\/} {\bf 61}(1) 95--101
662:
663: \bibitem{SM05}
664: Schweitzer L and Marko\v{s} P 2005 {\em Phys. Rev. Lett.\/} {\bf 95}, 256805-1
665:
666: \bibitem{Huc94}
667: Huckestein B 1994 {\em Phys. Rev. Lett.\/} {\bf 72}(7) 1080--1083
668:
669: \end{thebibliography}
670:
671:
672: \end{document}
673: