cond-mat0211037/m.tex
1: %RECOMMENDED%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: %
3: \documentstyle[aps,prb,epsfig,psfig,twocolumn]{revtex}
4: 
5: 
6: %%upright Greek letters (example below: upright "mu")
7: \newcommand{\greeksym}[1]{{\usefont{U}{psy}{m}{n}#1}}
8: \newcommand{\umu}{\mbox{\greeksym{m}}}
9: \newcommand{\udelta}{\mbox{\greeksym{d}}}
10: \newcommand{\uDelta}{\mbox{\greeksym{D}}}
11: \newcommand{\uPi}{\mbox{\greeksym{P}}}
12: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
13: 
14: 
15: %OPTIONAL%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
16: %
17: %\usepackage{amstex}   % useful for coding complex math
18: %\mathindent\parindent % needed in case "Amstex" is used
19: %
20: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
21: 
22: %AUTHOR_STYLES_AND_DEFINITIONS%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
23: %
24: \def\xref#1#2#3#4#5{#1:\ #2\ {\bf #3},~#4~(#5)}
25: \def\xbe{\begin{equation}}
26: \def\xee{\end{equation}}
27: %Please reduce your own definitions and macros to an absolute
28: %minimum since otherwise the editor will find it rather
29: %strenuous to compile all individual contributions to a
30: %single book file
31: %
32: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
33: 
34: \begin{document}
35: %
36: \title{Conductance statistics near the Anderson transition}
37: %
38: %
39: %
40: \author{Peter Marko\v{s}
41: }
42: \address{Institute of Physics, Slovak Academy of Sciences, D\'ubravsk\'a cesta 9, 842 28 bratislava, Slovakia\\
43: e-mail address: markos@savba.sk}
44: %
45: \maketitle         
46: 
47: \section{Introduction}
48: \label{markos-one}
49: %
50: 
51: Since the pioneering work of the Anderson \cite{Anderson-58} 
52: we know that disorder localizes electrons.
53: At zero temperature  any small amount of the disorder
54: localizes all electrons in the one-dimensional systems \cite{Mott-61}. In higher dimension ($d>2$ for
55: systems with orthogonal symmetry)
56: weak disorder does not destroy the metallic regime. Only when the strength of the disorder increases over
57: a certain specific value (critical disorder), the electron becomes localized. This phenomenon
58: - transition from the metallic to the insulating regime due to an increase of disorder -
59: is called Anderson transition
60: \cite{MD}.
61: 
62: 
63: Scaling theory of the Anderson transition uses  the conductance $g$ \cite{LT} as the order parameter. It is
64: supposed \cite{Wegner-76,AALR,GLK} 
65: that the system size dependence of the conductance is determined only by the value 
66: of the conductance itself:
67: \xbe
68: \frac{\partial\ln g}{\partial\ln L}=\beta(g)
69: \label{scaling}
70: \xee
71: where $\beta(g)$ is an analytical function of $g$. $\beta$ is positive (negative) in the metallic (localized)
72: regimes, respectively. For dimension $d>2$ the 
73: function $\beta(g)$ changes its sign, being positive for $g\gg 1$ and negative in the limit $g\to-\infty$. 
74: There is an unstable 
75: fixed point $g_c$ defined as the solution of $\beta(g=g_c)=0$. 
76: System-size independent critical conductance  $g_c$ 
77: represents the critical point of the Anderson transition.
78: 
79: Relation (\ref{scaling}) contains no
80: information about the {\it microscopic} structure of the model.  
81: This means that the Anderson transition is universal.  The form of the $\beta$ function
82: is determined 
83: only by the physical symmetry and dimension of the system 
84: \cite{McKK-93}.
85: 
86: Soon after the formulation of the scaling theory of localization it became clear that
87: the conductance $g$ is not a  self-averaged quantity.  
88: Reproducible fluctuations of the conductance were found both in the metallic and
89: in the insulating regimes \cite{WW,Fowler}. 
90: The knowledge of the mean value $\langle g\rangle$ is therefore
91: not sufficient for complete description of the
92: transport properties.  One has to deal with the conductance distribution $P(g)$
93: \cite{ATAF,BS}
94: or, equivalently, with all cummulants of the conductance.  This is easier in
95: the metallic regime, where $P(g)$ is Gaussian and the conductance fluctuations are universal 
96: \cite{LSF,Imry,DMPK,Beenakker,Pnato,BMcK,G,PZIS,APM}
97: and independent on the value of the mean conductance and/or the system size.
98: The width of the distribution depends only on the dimension, physical symmetry
99: of the system \cite{LSF} and on the boundary conditions \cite{BHMMcK,RMS-01c}.
100: In the insulator, conductance wildly fluctuates within the ensemble of
101: macroscopically equivalent ensembles. It is the logarithm of the conductance 
102: which is distributed
103: normally in the the limit of large system size \cite{ATAF,BS,Pnato,MK-AP}. 
104: 
105: At present we have no complete analytical theory  able 
106: to describe   the conductance statistics at  the critical point.  
107: Analytical results are known only for systems of dimension $d=2+\epsilon$
108: ($\epsilon\ll 1$) \cite{BS,AKL,CS}.
109: Expression for conductance cummulants \cite{AKL}
110: enabled to estimate the shape of the critical conductance distribution $P_c(g)$
111: \cite{CS}. $P_c(g)$ is system size independent and Gaussian in the neighbor
112: of the mean value $\langle g\rangle\sim\epsilon^{-1}$. The distribution possesses
113: power law tail $P_c(g)\sim g^{-1-2/\epsilon}$ for $g\to\infty$, and the delta-function
114: peak $\delta(g)$. 
115: These results can not be applied to three dimensional (3D) system
116: ($\epsilon=1$) where $\langle g\rangle\sim 1$ \cite{MK-PM}.
117: Numerical simulations are  therefore crucial 
118: in 3D system. The first systematic numerical 
119: analysis of the conductance statistics in 3D was done in
120: \cite{MK-PM} and was followed by a series of papers  
121: \cite{M-EL,SO-97,M-PRL,SOK-00,SO-01,RMS-01a,SMO-02}.  In 2D systems,
122: critical conductance distribution was numerically 
123: studied in the regime of quantum Hall effect
124: \cite{2D} and in systems with the spin-orbit interaction
125: \cite{M-EL,M-JPF,RMS-01b}.
126: 
127: \smallskip
128: 
129: This paper  reviews our recent numerical data for   the conductance distribution. 
130: We address the question of the shape, universality and the
131: scaling of the critical conductance distribution.
132: 
133: \smallskip
134: 
135: In numerical calculation of the conductance
136: we suppose that  two opposite sites of the sample are connected to semi-infinite perfect leads
137: and use the multichannel Landauer formula 
138: \cite{Landauer}
139: which relates the conductance $g$ (in units $2e^2/\hbar$) to the transmission matrix  $t$:
140: \xbe
141: g={\rm Tr}~t^\dag t =\sum_{i=1}^{N_{\rm open}}\cosh^{-2}\frac{z_i}{2}
142: \label{Landauer}
143: \xee
144: In (\ref{Landauer})   we introduced  the variables  $z_i$, $i=1,2\dots N_{\rm open}$, 
145: ($z_1<z_2<\dots$) which 
146: parametrize the eigenvalues of the matrix $t^\dag t$. 
147: In the limit $L_z>>L$, $z_i$ converges to $2L_z/\lambda_i$ where $\lambda_i$ is the $i$th
148: localization length of the quasi-one dimensional (quasi-1D) system 
149: \cite{Pnato,MK-PM}.
150: $N_{\rm open}$ is the number of open channels.
151: 
152: Owing to relation (\ref{Landauer}), the analysis  of the conductance can be reduced 
153: to the calculation of the eigenvalues of the matrix
154: $t^\dag t$. The general formula for $t$ was derived in \cite{Ando-91,PMcKR}. 
155: 
156: 
157: According to the scaling theory,
158: critical exponents of the Anderson transition  as well as 
159: conductance statistics do not
160: depend on the microscopic details of the model.  
161: In numerical simulations, we consider the Anderson Hamiltonian
162: \xbe
163: {\cal H}=W\sum_n \varepsilon_n c^\dag_nc_n+\sum_{[nn']}\tau_{[nn']}c^\dag_nc_{n'}.
164: \label{hamiltonian}
165: \xee
166: 
167: \begin{figure}[t!]
168: \begin{center}
169: %\includegraphics[width=.44\textwidth]{w_detail_2a.eps}~~
170: %\includegraphics[width=.47\textwidth]{g_var_g.eps}
171: \includegraphics[width=.22\textwidth]{markos_fig1a.eps}
172: \includegraphics[width=.22\textwidth]{markos_fig1b.eps}
173: \end{center}
174: \caption[]{{\it Left\,} The $L$ - dependence of the mean conductance $\langle g\rangle$ for 
175: different values of the disorder in the critical region. Data show that 
176: $16.45<W_c<16.55$. More exact estimation of the critical disorder was done in
177: \cite{SMO-01}. {\it Right\,}  The
178: unambiguous dependence var $g$ vs $\langle g\rangle$ in the critical region of the metal-insulator transition.
179: This agrees with the single parameter scaling theory.}
180: \label{m1}
181: \end{figure}
182: 
183: 
184: 
185: \noindent In (\ref{hamiltonian}) $n$ counts the  sites of the $d$-dimensional lattice and $[nn']$ 
186: means nearest neighbor sites. 
187: The hopping term $\tau$ equals to unity for orthogonal systems, unless the  anizotropy 
188: is considered.  When spin orbit scattering
189: is considered, $\tau$ is a $2\times 2$ matrix \cite{EZ,Ando-89}.  
190: Parameter $W$ measures the strength of the disorder. For box distribution of 
191: random energies $\varepsilon_n$  
192: ($P(\varepsilon)=1$ for $|\varepsilon|\le 1/2$ and $P(\varepsilon)=0$ otherwise), 
193: the 3D Anderson model (\ref{hamiltonian})
194: exhibits Anderson transition at $W_c=16.5$.
195: 
196: 
197: \section{Finite size scaling}
198: 
199: As it was discussed in the Section \ref{markos-one}, the conductance $g$ is not 
200: the self-averaged quantity. To avoid  
201: statistical fluctuations, quasi-1D systems were introduced
202: \cite{PS} and  
203: the localization length $\gamma$  is calculated
204: instead of the conductance.
205: In the limit  $L_z/L\to\infty$  ($L_z$ and $L$ are the
206: length and the  width  of the system, respectively)
207: $\gamma$ is a self-averaged quantity which converges to its mean value.
208: Finite size scaling is then used \cite{McKK} for the analysis of the disorder 
209: and the system width dependence of $\gamma$. It is assumed that the 
210: variable $\Lambda=\gamma/L$ is a function of only one parameter: 
211: $\Lambda(L,W)=\Lambda(L/\xi(W))$.
212: Here $\xi=\xi(W)$ is the correlation length which   diverges in the vicinity of the critical point
213: $\xi(W) =|W-W_c|^{-\nu}$.
214: Critical exponents  $\nu$  and $s=(d-2)\nu$ characterize the critical behavior of 
215: the localization length and of the conductance, respectively
216: \cite{Wegner-76}. 
217: 
218: Finite size scaling analysis of the quasi-1D data enabled to test the universality of the Anderson transition
219: \cite{ostatni} and
220: provided us with the more accurate estimation of the critical exponent
221: $\nu\approx 1.57$ \cite{SO-99}.  
222: 
223: 
224: \subsection{Scaling of the mean conductance}
225: 
226: Verification of the scaling theory of the localization requires
227: the proof of the universal scaling of the mean conductance and 
228: of the entire conductance distribution in the critical regime. 
229: Single parameter scaling of various mean values,
230: $\langle g\rangle$, $\exp\langle \ln g\rangle$, and $1/\langle g^{-1}\rangle$,
231: was proved numerically for the 3D Anderson model
232: \cite{SMO-01}. 
233: Statistical ensembles of more than $N_{\rm stat}\ge 10^6$ cubes
234: of the size from $4^3$ to $18^3$  were 
235: collected for various values of the disorder $W$.
236: Fig. \ref{m1} shows typical data for the
237: system size dependence of the mean conductance. In agreement with (1), 
238: $\langle g\rangle$
239: increases (decreases) with the system size in the metallic (localized) regime.
240: By the use of the general fitting procedure \cite{SO-99}, 
241: the critical disorder and the  critical exponent
242: $\nu\approx 1.57$ were obtained. 
243: 
244: 
245: Data in fig. \ref{m1}  confirm that the variance, 
246: var $g=\langle g^2\rangle-\langle g\rangle^2$ is an unambiguous function of the
247: mean $\langle g\rangle$ in the critical regime.  This supports, but still
248: does not prove the single parameter
249: scaling theory.  General proof  of the theory requires 
250: verification of 
251: the single parameter scaling  of all conductance cummulants. This
252: is numerically
253: impossible since  higher cummulants  are fully determined by rare
254:  events with very large values of the conductance. 
255: 
256: \subsection{Scaling of the conductance distribution}
257: 
258: As higher cummulants are not treatable numerically, we test 
259: the scaling of the conductance
260: distribution by the analysis of 
261: the scaling of  percentiles $g_\alpha$
262: \cite{SMO-02}.
263: Percentile $g_\alpha$ is defined as
264: \xbe
265: \alpha=\int_0^{g_\alpha} P(g) {\rm d}g.
266: \label{percentile}
267: \xee
268: Owing to (\ref{percentile}), the probability to find $g<g_\alpha$ equals to $\alpha$.
269: Of course, the 
270: percentile $g_\alpha$ is a function of disorder 
271: and system size: $g_\alpha=g_\alpha(L,W)$. 
272: Single parameter scaling of percentiles has been proved
273: for several values of $\alpha$ \cite{SMO-02}. 
274: 
275: \begin{figure}[t!]
276: \begin{center}
277: %\includegraphics[width=.47\textwidth]{g3g.eps}
278: %\includegraphics[width=.50\textwidth]{4D_hw_1a.eps}
279: \includegraphics[width=.22\textwidth]{markos_fig2a.eps}
280: \includegraphics[width=.24\textwidth]{markos_fig2b.eps}
281: \end{center}
282: \caption[]{{\it Left} Critical conductance distribution $P_c(g)$ of the 3D Anderson model, obtained for
283: statistical ensembles of $N_{\rm stat}=10^6$ samples of the size $10^3 - 18^3$.
284: $P_c(g)$ is system size independent and decreases faster than exponentially when $g>1$.
285: The main properties of the critical distribution are
286: shown in insets:
287: {\it Left inset} shows in details the small - $g$ behavior and proves that
288: the distribution decreases to zero as $g\to 0^+$. 
289: {\it Right inset} shows the discontinuity of the  derivation ${\rm d} P(g)/{\rm d} g$ 
290: at $g=1$.
291: {\it Right} figure presents the distribution
292: $P_c(\ln g)$ at the critical point for the three dimensional  ($W_c=16.5$, {\it open symbols\,}) 
293: and four dimensional (4D) ($W_c=34.3$, {\it full symbols\,}) systems of various system size.
294: }
295: \label{m2}
296: \end{figure}
297: 
298: \begin{figure}[t!]
299: \begin{center}
300: %\includegraphics[width=.35\textwidth]{g2b.eps}~~~\includegraphics[width=.35\textwidth]{g2a.eps}
301: \includegraphics[width=.2\textwidth]{markos_fig3a.eps}
302: \includegraphics[width=.2\textwidth]{markos_fig3b.eps}
303: \end{center}
304: \caption{Conductance distribution for ensembles of cubes $6^3$ ({\it dashed lines}) and 
305: $12^3$ ({\it solid line}):  (\textbf{a}) 
306: The  metallic regime  $W=15.4$  (the critical disorder is $W_c=16.5$).
307: Distribution $P(g)$ moves toward higher values of the 
308: conductance as the system size increases and $P(g)$ becomes Gaussian when $L\to\infty$.
309: (\textbf{b}) 
310: Insulator $W=17.6$
311: Mean conductance decreases as the system size increases and the distribution 
312: of the $\ln g$ becomes  Gaussian 
313: in the limit $L\to\infty$.
314: }
315: \label{m3}
316: \end{figure}
317: 
318: Suppose that  $g_\alpha$ and $g_\beta$  ($\alpha<\beta$) obey the single parameter scaling. Then
319: $g_\gamma$ ($\alpha<\gamma<\beta$) scales, too.
320: Therefore, in contrast to the analysis of the conductance 
321: cummulants, it is enough to analyze
322: only a few percentiles. Next, if $g_\alpha$ and $g_\beta$ scale, 
323: then the difference  $g_\beta-g_\alpha$
324: scales. Scaling of percentiles assures thus the scaling of the entire conductance 
325: distribution. Of course, this analysis is not applicable to the limit  $\alpha\to 1$, because available  statistical
326: ensembles are never  big enough to provide us with sufficient information about the tail of the distribution.
327: 
328: 
329: 
330: \section{Critical conductance distribution}
331: 
332: At present, we have no analytical description of the critical conductance distribution in 3D systems.
333: Analytical results were obtained only for the conductance cummulants 
334: in the dimension $d=2+\epsilon$ close to the lower critical dimension  ($\epsilon\ll 1$)
335: \cite{AKL}. In spite of the non-universality of higher order conductance cummulants
336: \xbe
337: \langle \delta g^n\rangle =\left\{\begin{array}{l} \epsilon^{n-2} ~~~ n<n_0=\epsilon^{-1}\\
338: 
339:                                                    \sim L^{\epsilon n^2-n} ~~~ n>\epsilon^{-1}
340: 					\end{array}\right.
341: \label{kumulanty}
342: \xee
343: the critical distribution $P_c(g)$ was shown  to be universal and $L$-independent in the  limit $L\to\infty$
344: \cite{CS}. However, theoretical analysis of the form of the critical distribution,
345: is applicable only in the limit of very small $\epsilon$
346: \cite{MK-PM}.  
347: 
348: \subsection{The form of the critical conductance distribution}
349: 
350: All what we know about the $P_c(g)$ in 3D is based on the 
351: numerical data.  In Fig.  \ref{m2} we present $P_c(g)$ for the 3D Anderson model.
352: Data confirm that the critical conductance distribution is system size independent, as required
353: \cite{BS}. The shape of  $P_c(g)$ differs considerably from  
354: the conductance distribution in  the metallic and in the insulating
355: regimes. To explain the typical properties of the critical conductance distribution, 
356: we use  our knowledge about statistical properties of parameters  $z$ (\ref{Landauer}).
357: 
358: \begin{figure}[t!]
359: \begin{center}
360: %\includegraphics[width=.5\textwidth]{g_2D.eps}
361: \includegraphics[width=.26\textwidth]{markos_fig4.eps}
362: \end{center}
363: \caption[]{Comparison of the critical conductance distribution of two different 2D models with
364: spin-orbit coupling: {\it full symbols} Ando model \cite{Ando-89} {\it open symbols} 
365: Evangelou-Ziman model \cite{EZ}.  
366: Squares $40\times 40$ and $80\times 80$ were simulated to prove the system size
367: independence of the critical distribution. Note also the non-analytic behavior for $g=1$.
368: }
369: \label{m4}
370: \end{figure}
371: 
372: 
373: Decrease of $P_c(g)$ to zero when $g\to 0$ is
374: visible when very large statistical ensembles are studied
375: (left inset of Fig. \ref{m2}).
376: Due to Eq. (\ref{Landauer}),
377: small conductance means that $z_1$ is large.  From numerical data we know that
378: the distribution  $P(z_1)$ is similar  to Wigner surmises 
379: and decreases  as $\exp -z_1^2$ for large $z_1$. Consequently,
380: $\ln P_c(\ln g)$ decreases as $-\ln^2g$ (right figure in Fig. \ref{m2}), and 
381: guarantees that $\lim_{g\to 0}P_c(g)=0$ \cite{MK-PM,M-PRL}. 
382: 
383: 
384: Large $g$ behavior of $P_c(g)$ is determined by the chance that many parameters  $z_i$ are  small.   
385: Statistical analysis of parameters $z_i$ showed that for $i>1$ the
386: distribution $P(z_i)$ is Gaussian with 
387: mean value $\langle z_i\rangle\propto \sqrt{i}$ and
388: var $z_i\propto \langle z_i\rangle^{-1}$
389: \cite{M-PRL}. The probability to find a sample with small value of  
390: higher $z_i$  is therefore very small: $P_c(g\approx i)$ decreases as $\exp-i^{3/2}$
391: \cite{M-PRL}.
392: Fig. \ref{m2} indeed shows  very fast decrease of the probability $P_c(g)$ for  
393: $g>1$. The chance to have $g>1$ is only $3\%$.
394: Probability to find  large values of  $g$ drastically decreases: 
395: we found that in the ensemble of $10^7$ samples ($L=10$) 
396: only 470 samples have $g>2$  and only one sample has  $g>3$ \cite{RMS-01a}. 
397: The analysis of  the contribution of the first two channels is  therefore 
398: sufficient for the understanding of the qualitative properties  of the critical distribution.
399: 
400: Numerical data also show that the critical distribution is non-analytical
401: at  $g=1$.  Right inset of Fig. \ref{m2} shows the discontinuity of  the first
402: derivative ${\rm d} P_c(g)/{\rm d} g$. 
403: The same   non-analyticity was found  in  the 4D systems (right figure
404: \ref{m2}), in the unitary \cite{SO-97} and symplectic (Fig. \ref{m5}) systems, and also in the
405: weakly disordered quasi-1D systems \cite{MuttW-99,M-PRB02a}.
406: %This can be also qualitatively explained.  The distribution of the smallest parameter $z_1$ is {\it linear} for small $z_1$, independently on the physical symmetry \cite{Pnato,M-JPF}: $P(z_1)\propto c\times z_1$ for $z_1\to 0$.  Probability that the first channels is completely open, is therefore non-zero: $P(g_1=1) = c$ \cite{M-PRL}.  However, small $z_1$ is usually accompanied by very large value of $z_2$. The probability to have $g$ slightly larger than unity decreases therefore very quickly: $P(g=1+\epsilon)\ll P(g=1)$.  
407: 
408: Present description of the critical distribution is based on the analysis of 
409: statistical properties of parameters $z$. It is applicable to any system, for which
410: the mean values of parameters $z$ are of the order of unity. Then only a few (two, or three) channels
411: contribute to the conductance. 
412: This  analysis is, however, not applicable to systems close to the lower 
413: critical dimension
414: $d=2+\epsilon$. Here, the mean conductance $\langle g\rangle\sim\epsilon^{-1}$, 
415: which means that number of channels which contribute to the conductance, is large, $\sim\epsilon^{-1}$. 
416: It is therefore no surprise that the
417: critical conductance distribution found in \cite{CS} differs from that shown in
418: fig. \ref{m2}.
419: 
420: 
421: \begin{figure}[t!]
422: \begin{center}
423: %\includegraphics[width=.42\textwidth]{frac_g.eps}
424: %\includegraphics[width=.55\textwidth]{Fig1b.ps}
425: \includegraphics[width=.2\textwidth]{markos_fig5a.eps}
426: \includegraphics[width=.2\textwidth]{markos_fig5b.ps}
427: \end{center}
428: \caption[]{Critical conductance distribution for  bifractal lattices. 
429: Legend gives spectral dimension.
430: 5th generation of fractals was used to calculate the distribution. Bifractal is 
431: linear along the $z$ axis and fractal in a plane perpendicular to  $z$.
432: {\it Right} figure shows the structure of the fractal (the 3rd generation) 
433: cross section.  All fractals have the same  fractal dimension $d_{\rm f}=\ln 3/\ln2$.
434: Spectral dimension $d_{\rm s}=\ln 9/\ln 5$ for fractals A and B, and $=\ln 9/\ln 6$ for fractal C.
435: Note that although  
436: bifractals A and B have the same both fractal and spectral dimensions, 
437: they posses different critical conductance distribution. $P_c(g)$ depends on the lattice
438: topology \cite{TM-PRB}}
439: \label{m5}
440: \end{figure}
441: 
442: 
443: \subsection{Conductance distribution in the critical regime}
444: 
445: Fig. \ref{m3} shows that  the  typical properties of the
446: critical conductance distribution
447: hold also for $P(g)$  in the neighbor of the
448: critical point. This is because the statistical properties of parameters $z$ depend 
449: continuously
450: on the disorder in the critical regime. Critical properties of 
451: $P(g)$ survive until the system size $L$ exceeds the correlation length $\xi$.
452: Only when  $\xi\ll L$
453: the distribution  typical for the  metallic or the  insulating regime can be observed.
454: 
455: 
456: 
457: \subsection{Universality}
458: 
459: Single parameter scaling theory of localization supposes that the critical conductance
460: distribution is universal. Its form does not depend on the microscopic
461: details of the model.
462: As an example, we present in Fig. \ref{m4} 
463: the critical conductance distribution for two 2D models with spin-orbit scattering: 
464: Evangelou-Ziman model \cite{EZ}  and Ando model \cite{Ando-89}.
465: In spite of the different 
466: microscopic definition of both models, $P_c(g)$ is universal
467: \cite{M-EL}.  Universality of $P_c(g)$ with respect to various distributions of random energies $\varepsilon_n$ was 
468: confirmed in \cite{M-EL} and \cite{SO-01}. 
469: 
470: As was shown already in  fig. \ref{m2},
471: the shape of $P_c(g)$ depends on the dimension of the model. Ref. \cite{SO-97}
472: confirmed that also the physical
473: symmetry influences the  form of $P_c(g)$.
474: Less expectable was the observation \cite{BHMMcK,SOK-00}
475: that  $P_c(g)$ 
476: as well as the spectral statistics \cite{BMP}
477: depend  on the {\it boundary conditions} in the transversal direction.
478: Nevertheless, this is consistent with the original definition of the conductance
479: as a measure of the sensitivity of the energy spectrum of the system
480: to the change of the boundary conditions \cite{LT}.
481: 
482: 
483: For completeness, let us note that the critical conductance distribution depends also on other parameters of
484: the model: lattice topology \cite{TM-PRB}, anizotropy \cite{RMS-01b} and, of course, on the length of the 
485: system. We believe that these non-universalities could be compensated by the change of another model  parameter
486: (see, for instance \cite{RMS-01b}, where the anizotropy is compensated by the length of the system).
487: 
488: \subsection {Dimension dependence}
489: 
490: Right figure in Fig.  \ref{m2} compares the critical conductance
491:  distribution for 3D and 4D cubes.
492: As supposed, the maximum of  $P_c(g)$ for 4D is shifted toward smaller conductances, because
493: the critical disorder increases as the spatial dimension increases and
494: higher disorder means lower mean conductance \cite{Wegner-76,CS}.
495: Qualitatively, however, both distributions are  very similar:
496: $P_c(\ln g)$ decreases as $\exp[-\ln^2 g]$ for $\ln g\to -\infty$  and possesses
497: the non-analyticity 
498: at $\ln g=0$. This similarity is not surprising, 
499: because the form of the distribution
500: is determined mostly by the statistics of  $z_1$ and $z_2$,
501: which are qualitatively similar in 3D and 4D
502: \cite{M-PRL}. 
503: 
504: Surprisingly, the relation (\ref{kumulanty})  seems to hold also
505: for $\epsilon=1$ and 2, at least for the first two cummulants.
506: We obtained numerically that $\langle g\rangle=0.285$ 
507: for 3D and 0.135
508: for 4D, so that 
509: $\langle g\rangle_{\rm 3D} \approx 2\langle g\rangle_{\rm 4D}$.
510: For the second cummulants we found
511: ${\rm var}~g_{\rm 3D}\approx {\rm var} g_{\rm 4D}\approx 0.17$ \cite{TM-PRB}.
512: 
513: \smallskip
514: 
515: More interesting is the investigation of the $P_c(g)$ in the systems
516: of dimension $2+\epsilon$
517: \cite{TM-PRB}. As we are not able to create the $d$-dimensional hyper-cubes 
518: with non-integer $d$ in computers, we simulated the transport on bifractal latices 
519: \cite{SG,TM-PRB}. Bifractals
520: are linear along the propagation direction and possess the fractal lattice in the cross section
521: (fig. \ref{m5}).
522: We proved that the critical  exponent $\nu$ depends 
523: only on the {\it spectral} dimension of the lattice. Mean 
524: conductance, var $g$   and the critical distribution $P_c(g)$  depend, however, on the lattice topology.
525: For instance, fig. \ref{m5} shows that
526: bifractals A and B have the same spectral dimension, but different critical distribution.
527: This is the reason why  obtained data can not be used for the verification of relations 
528: (\ref{kumulanty}). 
529: 
530: In Fig. \ref{m5} we present $P_c(g)$ for  three different 
531: bifractals. As expected, $\langle g\rangle$ increases
532: and the distribution converges to Gaussian when $\epsilon\to 0$. However, we found neither the
533: $\delta$ - function   peak at $g=0$
534: nor  the power-law tail of the distribution
535: for $g\gg \langle g\rangle$, predicted by the theory \cite{CS}. 
536: 
537: 
538: 
539: 
540: 
541: \section{Conductance distribution in non-critical regime}
542: 
543: Although we have no analytical theory of the conductance statistics in the critical regime, we can learn some
544: typical properties of the conductance distribution from the analysis of the 
545: quasi-1D weakly disordered systems \cite{MuttW-99}. 
546: Starting from the Dorokhov - Mello-Pereyra-Kumar equation \cite{DMPK} for the
547: probability distribution of parameters $z$,
548: the conductance distribution 
549: $P(g)$ can be calculated.
550: $P(g)$ depends on the length of the system. 
551: 
552: 
553: \begin{figure}[t!]
554: \begin{center}
555: %\includegraphics[width=.32\textwidth]{g106a.eps}~~~
556: %\includegraphics[width=.315\textwidth]{g165a.eps}~~~
557: %\includegraphics[width=.31\textwidth]{g49_la.eps}%\hspace*{.47\textwidth}
558: \includegraphics[width=.2\textwidth]{markos_fig6a.eps}
559: \includegraphics[width=.2\textwidth]{markos_fig6b.eps}
560: \includegraphics[width=.2\textwidth]{markos_fig6c.eps}
561: \end{center}
562: \caption[]{Comparison of the conductance distribution of the 3D ({\sl dashed lines})
563: and  quasi-1D 
564: ({\it solid line}).
565: {\bf (a)} Metallic ($\langle g\rangle>1$)  {\bf (b)} critical 
566: ($\langle g\rangle\approx 1$) and {\bf (c)} localized ($\langle g\rangle\ll 1$) 
567: regimes are shown. 
568: In the quasi-1D systems, the strength of the disorder is fixed to $W=4$ and the length of the system is 
569: tuned to obtain the same
570: mean conductance as in the 3D system.}
571: \label{m6}
572: \end{figure}
573: 
574: \begin{figure}[t!]
575: \begin{center}
576: %\includegraphics[width=.47\textwidth]{g_38a.eps}~~~~~
577: %\includegraphics[width=.47\textwidth]{g_60.eps}
578: \includegraphics[width=.22\textwidth]{markos_fig7a.eps}
579: \includegraphics[width=.22\textwidth]{markos_fig7b.eps}
580: \end{center}
581: \caption[]{Distribution of  
582: $-\ln g$ for strongly localized regime.
583: Note that the distribution is not symmetric. It possesses the long tail
584: for small values of the argument, and decreases much faster on the opposite side.
585: {\it Solid} lines are a  Gaussian distribution  with the mean and variance 
586: found from numerical data. Size of the cubes is
587: $L=10$. {\it Left}: $W=38$. Note that the sharp decrease of the distribution at $\ln g=0$, discussed in Sect.
588: 3.1, is still present \cite{MuttW-99}.  {\it Right}:  $W=60$. 
589: }
590: \label{m7}
591: \end{figure}
592: 
593: \noindent For short samples, $P(g)$ is Gaussian
594: \cite{Pnato} (see also fig. \ref{m6}a).
595: When  the length of the system increases, the conductance decreases and
596: the system passes from the metallic regime into  the localized one.
597: For the intermediate system length, where $\langle g\rangle\approx 1$, $P(g)$ 
598: is expected to be {\it qualitatively} 
599: similar to the critical conductance distribution
600: \cite{MuttW-99}. This is shown in
601: Fig \ref{m6}, where we compare the conductance distribution of quasi-1D
602: and 3D systems of the same mean conductance.
603: 
604: Fig. \ref{m6} also shows that the analogy between 3D and quasi-1D systems is not exact.
605: To understand {\it quantitative} differences between the quasi-1D and 3D systems, 
606: we analyzed the spectrum of parameters $z$. 
607: In the metallic regime, the difference between 3D and quasi-1D systems
608: is only in the  value of var $g$ \cite{LSF}.  $P(g)$ is Gaussian and
609: the spectrum of parameters $z$ is linear $\langle z_i\rangle\propto i$ \cite{Imry,Pnato}
610: independently on the dimension of the system \cite{M-JPCM-95}.
611: In contrast to the metallic regime,
612: the spectrum of $z$ becomes dimension dependent
613: in the critical region.
614: Both quasi-1D \cite{MH-JPCM,M-JPCM-95} and 3D-dimensional 
615: \cite{M-PRB02a} numerical studies confirmed that at the critical point
616: \xbe
617: \langle z_i\rangle^{d-1}\propto i
618: \label{dd}
619: \xee
620: ($d>2$). 
621: Owing to (\ref{dd}), the difference $\Delta=\langle z_2-z_1\rangle$ is smaller that
622: $\langle z_1\rangle$ in 3D, while it equals to $\langle z_1\rangle$ in the quasi-1D system.
623: The contribution of the second channel is therefore more important in 3D than in the quasi-1D.
624: This explains longer tail of the distribution in the 3D system (fig. \ref{m6}b).
625: 
626: 
627: It is commonly believed that the distribution $P(\ln g)$ is
628: Gaussian in the insulating phase, independently on the dimension of the system.
629: This is,however, not true. The
630: spectrum of $z$ depends namely on the dimension of the system also in the localized regime. 
631: For the 3D systems
632: it  was proved numerically \cite{M-PRB02a,M-JPCM-95}
633: that the difference $\Delta=\langle z_2-z_1\rangle$ is {\it constant},
634: independent on the disorder and on  the  system size. Therefore
635: the second channel influences always the statistics of the conductance.
636: More than  its contribution to the  value of the conductance it is important
637: that constant  value of $\Delta$ prevents
638: the distribution $P(z_1)$ to develop into the Gaussian form. 
639: While the values 
640: $z_1\ll \langle z_1\rangle$ are still possible, the probability to find systems
641: with much higher values
642: $z_1\gg \langle z_1\rangle$ is  strongly suppressed.
643: The distribution $P(z_1)$ is therefore not symmetric. 
644: The same is true for the distribution $P(-\ln g)$ (Fig. \ref{m7})
645: which  possesses a long tail for
646: small values of $|\ln g|$   and decreases much faster than Gaussian for
647: $|\ln g|\to\infty$. 
648: 
649: Note that in weakly disordered quasi-1D systems $\Delta\sim \langle z_1\rangle$.
650: The distance  $\langle z_2-z_1\rangle$ is much larger than the width of the distribution $P(z_1)$.
651: Higher channels therefore do not influence the distribution of $z_1$ and $P\ln g$  is Gaussian.
652: 
653: 
654: \section{Conclusion}
655: 
656: We reviewed  recent  progress in numerical studies of the statistics of the conductance
657: in the critical regime. 
658: Numerical analysis confirms  that
659: the conductance distribution in 3D Anderson model obeys 
660: single parameter scaling. 
661: Analysis of the statistics of the eigenvalues of the transmission matrix
662: enables us to understand the main features of the
663: conductance distribution in the critical regime.
664: Although we  still have no 
665: analytical  description of  the conductance statistics in the critical regime,
666: we hope that results of numerical experiments will inspire  theoreticians to formulate
667: the general analytical theory of the Anderson transition.
668: 
669: \vspace*{-0.5cm}
670: 
671: \begin{thebibliography}{8.}
672: \addcontentsline{toc}{section}{References}
673: 
674: \bibitem{Anderson-58} \xref{P.~W.~Anderson}{Phys.~Rev.}{109}{1492}{1958}
675: \bibitem{Mott-61} \xref{N.~F.~Mott, W.~D.~Twose}{Adv.~Phys.}{10}{107}{1961}
676: \bibitem{MD} N.~F.~Mott, E.~A.~Davis: Electron processes in non-crystalline materials, Clarendon Press, Oxford 1979
677: \bibitem{LT} \xref{J.~T.~Edwards, D.~J.~Thouless}{J.~Phys.~C}{5}{807}{1972};
678: \xref{D.~C.~Licciardello, D.~J.~Thouless}{J.~Phys.~C}{8}{4157}{1975}
679: 
680: \bibitem{Wegner-76} \xref{F.~Wegner}{Z.~Phys.~B}{25}{327}{1976}
681: 
682: \bibitem{AALR} \xref{E.~Abrahams, P.~W.~Anderson, D.~C.~Licciardello, T.~V.~Ramakrishnan}{Phys.~Rev.~Lett.}{42}{673}{1979}
683: \bibitem{GLK} \xref{L. P. Gorkov, A. I. Larkin, D. E. Khmelnitskii}{JETP Lett.}{30}{228}{1979}
684: \bibitem{McKK-93} \xref{B. Kramer, A. MacKinnon}{Rep. Prog. Phys}{56}{1469}{1993}
685: 
686: \bibitem{WW} \xref{S. Washburn, R. A. Webb}{Adv. Phys.}{35}{375}{1986}
687: \bibitem{Fowler} \xref{A. B. Fowler, J. J. Wainer, R. A. Webb}{IBM J. Res. Dev.}{32}{372}{1988}
688: \bibitem{ATAF}\xref{P. W. Anderson, D. J. Thouless, E. Abrahams, D. S. Fisher}{Phys. Rev. B}{22}{3519}{1980}
689: \bibitem{BS}\xref{B. Shapiro}{Philos. Mag. B}{56}{1032}{1987}; Phys. Rev. Lett. {\bf 65} 1510 (1990)
690: 
691: \bibitem{LSF} \xref{P.~A.~Lee, A.~D.~Stone}{Phys. Rev. Lett.}{55}{1622}{1985}; \xref{P.~A.~Lee, A.~D.~Stone, H.~Fukuyama}{Phys. Rev. B}{35}{1039}{1987}
692: 
693: \bibitem{Imry} \xref{I. Imry}{Europhys. Lett.}{1}{249}{1986}
694: \bibitem{DMPK} \xref{O.~N.~Dorokhov}{JETP Lett.}{36}{318}{1982}; \xref{P.~A.~Mello, P.~Pereyra, N.~Kumar}{Ann. Phys. (NY)}{181}{290}{1988}
695: \bibitem{Beenakker} \xref{C. W. J. Beenakker}{Rev. Mod. Phys.}{69}{731}{1997}
696: \bibitem{Pnato} J.-L. Pichard: 'Random transfer matrix theory and conductance fluctuations'. In: {\it Quantum Coherence in Mesoscopic Systems} NATO ASI Ser. B: Physics vol. {\bf 254} ed. by B. Kramer (Plenum Press, 1991)
697: \bibitem{BMcK} \xref{P. M. Bell, A. MacKinnon}{J. Phys.: Condens. Matt.}{6}{5423}{1994}
698: \bibitem{G}   \xref{N. Giordano}{Phys. Rev. B}{38}{4746}{1988} 
699: \bibitem{PZIS}\xref{J.-L. Pichard, N. Zannon, I. Imry, A. D. Stone}{J. Phys. France}{51}{587}{1990}
700: \bibitem{APM}\xref{Y. Avishai, J.-L. Pichard, K. A. Muttalib}{J. Phys. France}{3}{1387}{1991}
701: \bibitem{BHMMcK}\xref{D. Braun, E. Hofstetter, G. Montambaux, A. MacKinnon}{Phys. Rev. B}{64}{155107}{2001}
702: \bibitem{RMS-01c}\xref{M. R\"uhl\"ander, P. Marko\v{s}, C. M. Soukoulis}{Phys. Rev. B}{64}{172202}{2001}  % boundary
703: 
704: 
705: 
706: \bibitem{MK-AP}\xref{P. Marko\v{s}, B. Kramer}{Ann. Physik (Leipzig)}{2}{339}{1993}
707: 
708: \bibitem{AKL}\xref{B. L. Altshuler, V. E. Kravtsov, I. V. Lerner}{Sov. Phys. JETP}{64}{1352}{1986}; JETP Lett. {\bf 43}, 441 (1986)
709: 
710: \bibitem{CS}\xref{A. Cohen, B. Shapiro}{Int. J. Mod. Phys. B}{6}{1243}{1992}
711: \bibitem{MK-PM}\xref{P. Marko\v{s}, B. Kramer}{Philos. Mag. B}{68}{357}{1993}
712: \bibitem{M-EL}\xref{P. Marko\v{s}}{Europhys. Lett.}{26}{431}{1994}   % universality of P(g)
713: \bibitem{SO-97}\xref{K. Slevin, T. Ohtsuki}{Phys. Rev. Lett.}{78}{4083}{1997};
714: \bibitem{M-PRL}\xref{P. Marko\v{s}}{Phys. Rev. Lett.}{83}{588}{1999}
715: \bibitem{SOK-00}\xref{K. Slevin, T. Ohtsuki, T. Kawarabayashi}{Phys. Rev. Lett.}{84}{3915}{2000}
716: \bibitem{SO-01}\xref{K. Slevin, T. Ohtsuki}{Phys. Rev. B}{63}{45108}{2001}
717: \bibitem{RMS-01a}\xref{M. R\"uhl\"ander, P. Marko\v{s}, C. M. Soukoulis}{Phys. Rev. B}{64}{212202}{2001}  % symmetry
718: \bibitem{SMO-02} K. Slevin, P. Marko\v{s}, T. Ohtsuki: cond-mat/0208157
719: 
720: \bibitem{2D}\xref{Z.~Wang, B.~Jovanovic, D.~H.~Lee}{Phys. Rev. Lett.}{77}{4426}{1996}; \xref{V.~Plerou, Z.~Wang}{Phys. Rev. B}{58}{1967}{1998}; \xref{X.~Wang, Q.~Li, C.~M.~Soukoulis}{Phys. Rev. B}{58}{R3576}{1998}
721: 
722: \bibitem{M-JPF}\xref{P. Marko\v{s}}{J. Phys. I France}{4}{551}{1994}
723: \bibitem{RMS-01b}\xref{M. R\"uhl\"ander, P. Marko\v{s}, C. M. Soukoulis}{Phys. Rev. B}{64}{193103}{2001}  % anizotropy
724: 
725: \bibitem{Landauer}\xref{R.~Landauer}{IBM~J.~Res.~Dev.}{1}{223}{1957}; \xref{R.~Landauer}{Phil.~Mag.}{21}{683}{1970}; \xref{E.~N.~Economou, C.~M. Soukoulis}{Phys.~Rev.~Lett.}{46}{618}{1981}
726: \bibitem{Ando-91}\xref{T. Ando}{Phys. Rev. B}{44}{8017}{1991}
727: \bibitem{PMcKR}\xref{J. B. Pendry, A. MacKinnon, P. J. Roberts}{Proc. P. Soc. London A}{437}{67}{1992}
728: \bibitem{EZ}\xref{S. N. Evangelou, T. Ziman}{J. Phys. C}{20}{L235}{1987}
729: \bibitem{Ando-89}\xref{T. Ando}{Phys. Rev. B}{40}{5325}{1989}
730: 
731: \bibitem{PS}\xref{J.-L. Pichard,  G. Sarma}{J. Phys. C}{14}{L127}{1981}; {\it ibid}, L167
732: \bibitem{McKK}\xref{A. MacKinnon, B. Kramer}{Phys. Rev. Lett.}{47}{1546}{1981}
733: \bibitem{ostatni}\xref{B.~Bulka, B.~Kramer, A.~MacKinnon}{Z. Phys. B: Condens. Matt.}{60}{13}{1985}; \xref{B.~Kramer, K.~Broderix, A.~MacKinnon, M.~Schreiber}{Physica A}{167}{163}{1990}; \xref{A.~MacKinnon}{J. Phys.: Condens. Matt}{6}{2511}{1994}
734: %; \xref{P. Cain, R. A. R\"omer, M. Schreiber}{Ann. Phys. (Leipzig)}{8}{SI-33}{1999}
735: 
736: \bibitem{SO-99}\xref{K. Slevin, T. Ohtsuki}{Phys. Rev. Lett.}{82}{382}{1999}
737: \bibitem{SMO-01}\xref{K. Slevin, P. Marko\v{s}, T. Ohtsuki}{Phys. Rev. Lett.}{86}{3594}{2001}
738: 
739: %\bibitem{CRS}\xref{A. Cohen, Y. Roth, B. Shapiro}{Phys. Rev. B}{38}{12125}{1988}
740: %\bibitem{Roberts}\xref{P. J. Roberts}{J. Phys.: Condens. Matt.}{4}{7795}{1992}
741: %\bibitem{DLA}\xref{L. I. Deych, A. A. Lisyansky, B. L. Altshuler}{Phys. Rev. Lett.}{84}{2678}{2000}
742: %\bibitem{VMMS}{P. Vagner, P. Marko\v{s}, M. Mo\v{s}ko, T. Sch\"apers}: cond-mat/0207636
743: %\bibitem{SP}\xref{K. Slevin, J. B. Pendry}{J. Phys.: Condens. Matt.}{2}{2821}{1990}
744: 
745: 
746: \bibitem{MuttW-99}\xref{K. A. Muttalib, P. W\"olfle}{Phys. Rev. Lett.}{83}{3013}{1999};
747: V.~A.~Gopar, K.~A.~Muttalib, P.~W\"olfle: cond-mat/0205643;
748: K.~A.~Muttalib, P.~W\"olfle, A.~Garc\'{\i}a-Mart\'{\i}n, V.~A.~Gopar: cond-mat/0210293
749: \bibitem{M-PRB02a}\xref{P. Marko\v{s}}{Phys. Rev. B}{65}{104207}{2002} % non-metalic regimes
750: 
751: 
752: 
753: 
754: %\bibitem{AC}\xref{A. Cohen}{Phys. Rev. B}{51}{10406}{1995}
755: 
756: 
757: 
758: %\bibitem{EXMK-01}\xref{S. Evangelou, S. J. Xiong, P. Marko\v{s}, D. Katsanos}{Phys. Rev. B}{63}{144526}{2001}
759: 
760: \bibitem{BMP}\xref{D. Braun, G. Montambaux, M. Pascaud}{Phys. Rev. Let.}{81}{1062}{1998}
761: 
762: \bibitem{TM-PRB}\xref{I. Trav\v enec, P. Marko\v{s}}{Phys. Rev. B}{65}{113109}{2002}
763: \bibitem{SG}\xref{M. Schreiber, H. Gru\ss bach}{Phys. Rev. Lett.}{76}{1687}{1996}
764: %\bibitem{M-JPA-97}\xref{P. Marko\v{s}}{J. Phys. A: Mat. Gen.}{30}{1471}{1997}   % universal scaling ...
765: %\bibitem{M-JPA-00}\xref{P. Marko\v{s}}{J. Phys. A: Mat. Gen.}{33}{L393}{2000}   % scaling of LE
766: 
767: %\bibitem{Mnato} P. Marko\v{s} 'Universality of the metal-insulator transition'. In: {\it Quantum Dynamics of Submicron Structures} NATO ASI Ser. E: Applied Sciences, vol {\bf 291}, ed. by H. A. Cerdeira, B. Kramer, G. Sch\"on (Kluwer, 1995) pp. 99-102
768: 
769: \bibitem{MH-JPCM}\xref{P. Marko\v{s}, M. Henneke}{J. Phys.: Condens. Matt.}{6}{L765}{1994}
770: \bibitem{M-JPCM-95}\xref{P. Marko\v{s}}{J. Phys.: Condens. Matt.}{7}{8361}{1995} % phenomenological ...
771: 
772: %\bibitem{PA}\xref{J.-L. Pichard, G. Andre}{Europhys. Lett.}{2}{477}{1986}
773: 
774: 
775: 
776: 
777: \end{thebibliography}
778: 
779: 
780: \end{document}
781: