1:
2: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
3: \ProvidesFile{v3.tex}
4: [1999/12/01 v1.4c Il Nuovo Cimento]
5: \documentclass[varenna]{cimento}
6: \usepackage{graphicx}
7: \usepackage{multicol}
8:
9: \title{Metallic conduction, apparent metal-insulator\\
10: transition and related phenomena in two-dimensional electron
11: liquid}
12: \shorttitle{Metal-insulator transition in 2D}
13: \author{V.~M.~Pudalov
14: %\thanks{}
15: } \institute{P.~N.~Lebedev Physics Institute - 53 Leninskii
16: prospekt, Moscow, 119991 Russia}
17:
18: \PACSes{\PACSit{71.30.+h, 73.40.Qv, 71.27.+a}
19: {%\ldots
20: } }
21: \begin{document}
22:
23: \maketitle
24:
25: \begin{abstract}
26: %This paper results from the lecture given in the summer of
27: %2003, at Enrico Fermi school in Varenna.
28: The paper introduces a reader to
29: a relatively young
30: field of
31: the physics of strongly interacting and disordered
32: 2D electron system, in particular, to the phenomena of the metallic conduction
33: and the apparent metal-insulator transition in 2D. The paper briefly overviews the
34: experimental data on the electron transport, magnetotransport, and spin-magnetization in 2D
35: and on the electron-electron interaction effects.
36: \end{abstract}
37:
38: \section{Introduction}
39: Resistivity of materials around us ranges by more than 30 orders
40: of magnitude, from $10^{20}$ to $10^{-12}$ Ohm\,m. The variety of
41: materials are classified as Metals (M) and Insulators (I), which
42: are two distinct classes of matter only in the $T\rightarrow 0$
43: limit. The principle difference between the two classes is in the
44: character of the electronic wavefunctions, which are spatially
45: localized in insulators and extended in metals.
46:
47: It is well-known,
48: that subjected to even insignificant changes of pressure,
49: dopant density, etc.
50: many materials exhibit transformations between the two states \cite{mott}.
51: The M-I transitions (MIT), unrelated to changes in the lattice structure and symmetry,
52: are of a special interest because they are considered to be
53: continuous Quantum phase transitions, occurring
54: at $T=0$.
55: The M-I transitions usually take place as density
56: of electrons decreases;
57: the latter necessarily
58: leads to the increase of the effective
59: strength of $e-e$ interaction.
60: A dimensionless ratio $r_s$ of the potential (Coulomb) energy
61: to kinetic (Fermi) energy is often used to quantify interaction
62: in 2D, $r_s \propto n^{-1/2}$ with $n$ being the electron density \cite{ando_review}.
63: On the other hand, the decrease in the density is accompanied with
64: the increase in the effective disorder, particularly, due to the weakening of the
65: screening of potential fluctuations. Thus, both, interactions and disorder
66: are important in the vicinity of the MIT; the separation and interpretation
67: of their effects represents a hard task \cite{Nagaoka_review}.
68:
69: In the limit of strong interactions $r_s \rightarrow \infty$ and zero disorder,
70: the groundstate of the 2D system
71: is believed to be the Wigner crystal (WC) of electrons.
72: Correspondingly, at zero temperature and zero disorder, for 2D system, there is at least one
73: critical point on the interaction axes,
74: quantum melting of the Wigner crystal \cite{ceperley,ando_review}.
75: A weak disorder
76: is expected to stabilize the WC \cite{chui&tanatar,giamarchi}.
77: As disorder increases further, WC looses
78: long-range order and crosses over to a localized
79: phase.
80: Thus, in the limit of
81: strong disorder and weak interactions, the ground state of the system
82: is built of the single-particle
83: localized states \cite{efros&shklovskyi}.
84: The experiments always take place at non-zero disorder
85: and temperature.
86: The subject of this paper
87: is related to phenomena which occur due to the combined action of
88: strong inter-electron interactions and disorder.
89:
90: The outline of this paper is as follows:
91: Sections 1 and 2 introduce a reader to
92: the field and briefly review the main results in
93: the semiclassical and quantum frameworks, correspondingly.
94: Section 3 describes major results on transport in 2D systems, which are
95: discussed in terms of the apparent metal-insulator transition in 2D.
96: Section 4 describes quantitative studies of the effects of electron-electron interaction;
97: the experimental results are compared with available theories in various regimes.
98:
99: \vspace{0.1in}
100: \underline{1.1. Mott semiclassical picture of the MIT}.\\
101: At finite temperature, or for finite system
102: size $L$, the metal and insulator states are not well-defined; one can be mislead
103: by attempting to distinguish these two states according to
104: their conduction. In fact, the two states are inherently related with two
105: distinct types of electron wavefunctions, localized and extended ones.
106: The latter are classified according to the localization length $\xi$: if $\xi >
107: (r_2-r_1)$ the waves are extended (metal) and $\Psi \propto
108: \cos(kr)$; if $\xi < (r_2-r_1)$ the waves are localized
109: (insulator) and $\Psi \propto \exp(-r/\xi)\cos(kr)$. Note, the
110: electron spacing, ($r_2-r_1) \sim \lambda_F = 2\pi/k_F = 1120
111: \left(10^{11} g_v/2n \right)^{1/2}$\AA\ (the valley
112: degeneracy $g_v=2$ corresponds to (100)Si-crystal plane).
113:
114: For the extended states, when $\lambda_F \ll l <L$ (with $l$ being the
115: mean free path) the electron motion results in a
116: classical Drude conduction $\sigma = ne^2\tau_p/m=ne^2 l/v_Fm$.
117: As disorder increases, $l$ decreases. However, $l$ can not be
118: less than $\lambda_F$ (the so called Ioffe-Regel criterium
119: (1960)):
120: $l_{\rm min}=\lambda_F = 2\pi/k_F$.
121: Thus, we obtain the minimum metallic conductivity \cite{mott}
122: $\sigma_{\rm min}(l_{\rm min})=2\pi e^2/h \quad {\rm in\ 2D}$.
123: We used here
124: $k_F^{2D}=\sqrt{2\pi n}$. Note, that $h/e^2=25812.7$\,Ohm.
125: In this semiclassical picture \cite{mott},
126: the M-I transition is caused by the disappearance
127: of the localized (or, vice versa, extended) states.
128:
129: \vspace{0.1in}
130: \underline{1.2. Basics of the 2D semiconducting devices}\\
131: The 2D electron and hole systems discussed in this lecture consist of the bulk
132: carriers confined in a two-dimensional potential well; the latter is
133: formed at the interface between two semiconductors
134: or semiconductor/insulator interface. Figure 2 schematically shows
135: crossview of the silicon metal-oxide-semiconductor (MOS) stucture \cite{ando_review}.
136: The 2D layer of electrons is confined at the interface between Si and SiO$_2$, grown at the
137: top of the Si-substrate. On the top of the insulating layer,
138: a thin metallic film (gate) is deposited.
139: Throughout this lecture, the electrons and holes are the same quasiparticles
140: and for simplicity will be often called ``electrons''.
141:
142: When the positive voltage $V_g$ is applied to the gate (relative to
143: the source, or drain contact, or relative to the Si-bulk),
144: the conduction $E_c$ and valence $E_v$ band edges
145: bend down as Figure \ref{Energy band_Si-MOS} shows. For a sufficiently high gate voltage,
146: the bottom of the conduction band
147: decreases below the bulk Fermi level and the resulting triangular potential well
148: starts populating with electrons \cite{ando_review}. At room temperatures, when the bulk Si
149: is conducting, redistribution of the electrons between the bulk and the 2D
150: layer occurs in accord with the Poisson
151: equation \cite{ando_review}. When the gate voltage is applied or changed at low temperatures,
152: the bulk conduction is frozen-out, the equilibrium with bulk is not achieved,
153: and the required electrons come into (or out) the 2D layer
154: from the potential contacts. The contacts are lithographically defined dopant diffusion areas
155: with high concentration of electrons supplied by the dopants.
156: \vspace{0.1in}
157: \begin{figure}[htb]
158: \begin{minipage}[t]{.47\textwidth}
159: \includegraphics[width=2.4in]{Fig1.eps}
160: \caption{Schematic structure of a silicon field effect transistor.
161: The 2D electron gas is
162: confined at the Si side of the Si/SiO$_2$ interface.
163: Drain and source are diffusion areas
164: used as ohmic contacts to the 2D layer.
165: Metallic gate is deposited on the top of the SiO$_2$ layer
166: \protect\cite{ando_review}.}
167: \label{Si-MOSFET schematics}
168: \end{minipage}
169: \hfil
170: \begin{minipage}[t]{.47\textwidth}
171: \begin{center}
172: \includegraphics[width=1.9in]{Fig2c.eps}
173: \caption{Schematic energy band diagram of the Si-MOS structure. $E_C$ ($E_V$) is the bottom
174: (top) of the conduction (valence) band, $E_A$-bulk
175: acceptors energy level, $E_F$ - Fermi level,
176: $Z$-axes is directed to the bulk, perpendicular to the surface.}
177: \label{Energy band_Si-MOS}
178: \end{center}
179: \end{minipage}
180: %\vspace{-0.15in}
181: \end{figure}
182: For our purpose, it is essential
183: only that the layer of electrons confined at the interface
184: is two-dimensional,
185: and that its charge $Q$ is controlled by voltage $V_g$
186: applied to the gate:
187: $Q\equiv ne = C(V_g-V_t)$, with $V_t$, the so called threshold voltage.
188: Thus, the Si-MOS system may be viewed as
189: a plane capacitor,
190: whose one plate is formed by the 2D layer of electrons whereas the gate
191: serves as another plate.
192:
193: \section{Quantum transport at zero field}
194: \underline{2.1. Various types of transport:
195: Delocalized states: diffusive and ballistic transport.}\\
196: There are two regimes of the metallic-like transport:
197: diffusive for $l < l_\varphi< L$, and ballistic regime for $l> L$.
198: The phase breaking length $l_\varphi = \sqrt{D\tau_\varphi}$ (with $\tau_\varphi$,
199: the phase breaking time, $D$ - diffusion coefficient)
200: is related with
201: large changes of electron energy and is similar (not equivalent though)
202: to inelastic length.
203: The relevant scattering processes for $\tau_\varphi$
204: at high temperatures are (i) electron-phonon scattering,
205: and for low temperatures - (ii) electron-impurity and
206: (iii) electron-electron scattering.
207: In general, $\tau_\varphi \propto T^{-p}$ with $p=1 - 2$, depending
208: on the dominating scattering process.
209: Since electron-electron collisions conserve the total momentum,
210: the transport time $l_p$
211: (momentum relaxation)
212: is determined by the two former processes and is insensitive to
213: electron-electron scattering.
214:
215: \vspace{0.1in}
216: \underline{2.2. Localized states}.\\
217: If an energy barrier $\Delta$ separates the energy of the
218: electron states
219: from conduction (or valence) band, and the
220: wavefunctions of localized states
221: don't overlap, the conduction occurs
222: via temperature activated transitions whose probability is
223: $\sigma \propto \exp(-\Delta/T)$.
224: When the probability of the temperature activation
225: is too low (e.g., for high $\Delta$, or for low temperatures), transport occurs via
226: elastic tunneling between the localized states. This is the so-called
227: hopping conduction regime, where
228: the characteristic temperature dependence of the conductivity is
229: $\sigma \propto \exp((-T_0/T)^p)$,
230: where $p=1/2$ or 1/3, depending on a specific model \cite{efros&shklovskyi}.
231: Usually, hopping conduction %always
232: dominates at low temperatures.
233:
234: \vspace{0.1in}
235: \underline{2.3. Electron's Phase
236: Coherence and Transport.}\\
237: In the quantum-mechanical picture, the electron waves propagate and interfere.
238: The interference gives
239: rise to quantum corrections to the
240: conductivity \cite{altshuler&aronov,lee_review}.
241: In 2D system of non-interacting electrons (Fermi-gas), as $T\rightarrow 0$:
242: \begin{equation}
243: \sigma = \sigma_D
244: -\frac{e^2}{\pi h}\ln(\tau_{\varphi}/\tau) \sim
245: \sigma_D + \frac{e^2}{\pi h}\ln(T\tau),
246: \end{equation}
247: where $\sigma_D$ is the semiclassical Drude conductivity.
248: The single-particle interference thus may be viewed as a quantum ``backscattering''.
249: Weak logarithmic dependence is a typical attribute of the 2D system at high electron
250: density (low $r_s$), and may be observed at high conductivity
251: $\sigma \gg e^2/h$ in the low-temperature
252: diffusive regime $T\tau \ll 1, \tau \ll \tau_\varphi$ (see fig.~\ref{Gmax}).
253:
254: \begin{figure}
255: \begin{center}
256: \includegraphics[width=3.2in]{Fig3c.eps}
257: \caption{$\sigma(T)$ dependences
258: for high densities \protect\cite{Gmax} (from top to bottom,
259: in units of $10^{11}$\,cm$^{-2}$):
260: 32.1, 42.94,
261: 48.4, 53.8, 64.7,
262: 75.6,
263: 86.5.}
264: \label{Gmax}
265: \end{center}
266: %\vspace{-0.25in}
267: %\vspace{-0.1in}
268: \end{figure}
269:
270: \vspace{0.1in}
271: \underline{2.4. Suppression of the weak localization in $H_\perp$ fields}\\
272: The quantum interference corrections originate from
273: small loops of self-intersections on the trajectories, which quantum
274: particles propagate in both directions, due to the time reversal symmetry
275: \cite{altshuler&aronov,lee_review}.
276: The amplitudes of the wavefunction for a particle passing the loop clockwise and
277: counterclockwise acquires additional factors:
278: $
279: A_1 \rightarrow A_1\exp\left(\pm i\pi\Phi/\Phi_0\right)
280: $
281: with $\Phi_0=h/e$ being the flux quantum and $\Phi$ the magnetic flux through the loop.
282: The magnetic field applied perpendicular to the 2D plane of motion,
283: destroys the interference by reducing the probability for a particle to return.
284: The interference breaks down when the phase difference $\Delta\varphi$ becomes $\sim 1$:
285: \begin{equation}
286: \Delta \varphi \sim \frac{H D \tau_\phi}{\Phi_0} \sim 1 \hspace{0.5in} H\sim \frac{h}{eD\tau_\varphi}
287: \end{equation}
288: The ``negative magnetoresistance'' effect is
289: a tool to determine $\tau_\varphi$ \cite{lee_review,fukuyama82}.
290:
291: \vspace{0.1in}
292: \underline{2.5. Single-particle scaling theory of localization (E.\ Abrahams et al., 1979)
293: \cite{gang4}}\\
294: The one-parameter scaling theory considers
295: %addresses a question on
296: the dependence of the conductivity
297: on the system size.
298: When the size
299: $L$ is changed, the effective disorder is changed; it is
300: assumed that the only measure of this is the conductance $G$.
301: The above assumption is equivalent to
302: introducing a function $\beta$ such that
303: \begin{equation}
304: \beta = \frac{L}{G} \frac{dG}{dL}=\frac{d\ln G}{d\ln L},
305: \end{equation}
306: where $\beta$ is a universal function of $G$ solely.
307:
308: One can consider how the scaling hypothesis fits various obvious limits:\\
309: (1) In the Ohm's law region ($G \gg e^2/h$), $\beta= d-2$ and
310: the conductance
311: does not depend on $L$
312: for the two-dimensional case, $d=2$.\\
313: (2) If the states are localized,
314: $ G(L) \propto \exp(-L/\xi)$.
315: Therefore,
316: $\beta(G)=\ln(G/G_0)$. \\
317: The two distinct limits enable to plot the asymptotes of $\beta$ in fig.~\ref{beta}\,a.
318: One may expect that
319: between these two limits, $\beta $ should vary smoothly,
320: as shown schematically in fig.~\ref{beta}\,a.
321:
322: \begin{figure}
323: \begin{center}
324: \includegraphics[width=2.4in]{Fig4c.eps}
325: \caption{a) Schematic behavior of the scaling function for
326: non-interacting electrons in different dimensions $d$. Bold dot depicts
327: the critical point of the M-I transition for $d=3$.
328: b)Spin-orbit scattering case for $d=2$.} \label{beta}
329: \end{center}
330: \vspace{-0.2in}
331: \end{figure}
332:
333:
334:
335: \underline{Note:}
336: It is not necessary to vary $L$ in order to move along the scaling trajectories.
337: For finite temperatures, $l_\varphi$ causes inelastic cut-off, so that the temperature
338: tunes the effective sample length.
339:
340: In $d=3$, as fig.~\ref{beta}\,a shows, the scaling function $\beta <0$
341: for very small $G$ (strong disorder). As conductance increases, $\beta$
342: passes through zero at a certain critical value $G_c$.
343: This is a repulsive point or an ``unstable critical point'' in the renorm-group
344: terminology.
345: For low initial $G<G_c$, increasing the size of the system will cause $G$ to decrease
346: and the behavior of the system approaches that for the localized states.
347: In contrast, when the initial conductance is high,
348: $G>G_c$, increasing the size of the system will leads to the Ohmic regime
349: (metallic conduction). Such behavior and the corresponding metal-insulator transition
350: are consistent with the Mott' semiclassical picture
351: described above.
352:
353: In lower dimensions, $d<3$,
354: $\beta $ is always negative and all states are localized. As system size $L$ increases,
355: the system moves to the
356: strongly localized regime. The scaling function describes a crossover from
357: weak localization (localization length
358: $\xi >L$) to strong localization ($\xi <L$) regime.
359:
360: It follows from the
361: one-parameter scaling theory that
362: ``there is no true metallic conduction in $d<3$'' \cite{gang4}.
363: For $d=1$, this conclusion is obvious since the negative scaling function has
364: substantially large amplitude $|\beta| \geq 1$. In accordance with eq.~(4),
365: this causes the conductance to decrease quickly as $L$ increases,
366: approaching
367: that for the strongly localized case.
368: However, the $d=2$ case is more tricky: for high $G>>1$, the scaling function is so
369: close to zero, that minor correction may potentially change
370: the sign of $\beta$, therefore the situation requires a special attention.
371:
372: \vspace{0.1in}
373: \underline{a) Spin-orbit scattering case} (Hikami, Larkin, Nagaoka (1980)).
374: The electron interference is destroyed by spin-orbit random scattering
375: and the weak localization ``back-scattering'' correction vanishes
376: \cite{hikami80,altshuler&aronov,lee_review}.
377: More over, the spin-orbit scattering induces
378: a positive correction (``forward scattering'') so that
379: the $\beta$-function becomes $\propto + 1/2G$ for high $G$,
380: as schematically shown in fig.~\ref{beta}\,b. Note, that this results
381: is valid only for a hypothetical 2D system of non-interacting particles.
382:
383: \vspace{0.1in}
384: \underline{b) Interaction quantum corrections in the diffusive regime}.
385: Altshuler-Aronov corrections take into account
386: $e-e$ interactions in the $T\tau \ll 1$ limit
387: \cite{altshuler&aronov,finkelstein}:
388: \begin{equation}
389: \delta\sigma =-\frac{e^2}{2\pi^2 \hbar}\left[1 +
390: 3\left(1-\frac{\ln\left(1+F_0^a\right)}{F_0^a}\right)\right]\ln(\hbar/k_BT\tau),
391: \end{equation}
392: where $F_0^a$ is the Fermi-liquid interaction parameter, dependent on $r_s$
393: (see further for more details).
394: In the $F_0^a\rightarrow 0$ limit,
395: the total quantum correction $\delta\sigma_{\rm loc}
396: +\delta\sigma_{so}+\delta\sigma_{ee}$
397: is usually negative (backscattering). However, when
398: $F_0^a<0$ and $|F_0^a|$ is large, the total quantum correction may
399: lead to delocalization (``forward scattering''). This case will be discussed
400: in Sec.~\ref{sec:e-e}
401: in more details.
402:
403: \vspace{0.1in}
404: \underline{c) Interaction quantum corrections to the transport in the ballistic regime}
405: $T\tau \gg 1$
406: have been calculated
407: in Refs. \cite{ZNA,gornyi}; they are discussed in
408: Sec.~\ref{sec:e-e}.
409:
410: \section{An apparent MIT in 2D}
411: A great body of experimental data accumulated since 1979
412: support the conclusions of the scaling theory \cite{bishop,uren}.
413: An example is shown in fig.~\ref{Si46_Si39}\,a.
414: \begin{figure}
415: \begin{center}
416: %\includegraphics[angle=-90,width=4.8in]{Fig5c.eps}
417: \includegraphics[width=6.0in]{Fig5c1.eps}
418: \caption{Resistivity vs temperature for
419: disordered Si-MOS samples: (a) with peak mobility $\mu=1500$cm$^2$/Vs,
420: (b) $\mu=5100$cm$^2$/Vs.
421: The densities (in units of $10^{11}$\,cm$^{-2}$) on the left panel
422: span from 3.85 to 37.0 (from top to bottom),
423: on the right panel -- from 1.95 to 5.44.}
424: \label{Si46_Si39}
425: \end{center}
426: \vspace{-0.3in}
427: \end{figure}
428: The uppermost curves in figs.~\ref{Si46_Si39} show that
429: the resistivity for a very disordered (low-mobility) Si-MOS sample at low densities
430: exhibits an exponentially strong increase with cooling.
431: This is consistent with motion to the left hand side along the
432: logarithmic part of the $\beta$-function in fig.~\ref{beta}\,a.
433: For higher densities (lower curves in fig.~\ref{Si46_Si39}),
434: the exponential increase of the resistivity is replaced by
435: a weak temperature dependence. The latter one
436: at very high carrier densities in Si-MOS or in GaAs samples
437: is logarithmic (see, e.g. fig.~\ref{Gmax}) and consistent
438: with the theory ($d\rho/dT >0$).
439:
440: As sample mobility increases, and, correspondingly,
441: the density of the crossover (from the exponential to logarithmic $T$-dependence)
442: decreases \cite{highmu}, the temperature derivative $d\rho/dT$ obtains
443: ``wrong'' positive sign
444: (compare figs.~\ref{Si46_Si39}\,a, b, and fig.~\ref{high_mu}). This effect was noticed
445: more than two decades ago \cite{cham80,smith86,dorozh88}
446: and discussed in terms of the temperature-dependent
447: screening \cite{stern80,das86,gd86}.
448: As will be shown below, the sign of $d\rho/dT$
449: is determined by the
450: interaction parameter $F_0^a$, which in its turn, depends primarily on the electron density.
451:
452:
453: The strong indication that the prediction of insulating states in
454: 2D may not be universally valid was obtained in 90s in studies of
455: the quantum Hall liquid to insulator transitions
456: \cite{reentrant_dio,dioPRB92,hall93}.
457: Extended states, which in high magnetic field $B$ are located at
458: centers of Landau bands, were found experimentally to merge and
459: remain in a finite energy range as $B \rightarrow 0$. This
460: behavior is not expected within the scaling theory, which predicts
461: that the extended states ``float up'' in energy in that limit
462: \cite{halperin,khmelnitskii,laughlin}. A serious challenge to the
463: scenario of the insulating ground state in 2D arose in 1994 and
464: the years following, when metallic behavior of electrical
465: transport in zero magnetic field was observed in high mobility
466: Si MOSFET structures \cite{PRB94,PRB95} (see fig.~\ref{high_mu}).
467:
468: Subsequently, qualitatively similar metallic states were found in
469: other high mobility 2D systems (for an extensive bibliography,
470: see Ref.~\cite{sk_RMP}). These include $p-$ and $n-$Si/SiGe--
471: \cite{coleridge,lam97,Brunthaler}, $p$-GaAs--
472: \cite{pepper,shahar,papadakis,mills,noh03}, $n$-GaAs--
473: \cite{ngaas,ensslin}, $n-$AlAs--heterostructures \cite{alas},
474: inverted Si-on-insulator structures \cite{SOI} and back-gate
475: biased Si-MOS structures \cite{fowler97}. The most important
476: findings of these studies can be summarized as follows:\\
477:
478: \underline{Features in zero magnetic field}:
479: \begin{itemize}
480: \item If the density of 2D carriers $n$ is larger than a sample-dependent
481: value $n_c$, the $\rho(T)$ dependence is metallic-like, i.e. $d\rho/dT>0$
482: (see fig.~\ref{high_mu}).
483: The
484: changes in $\rho$ can be almost an order of magnitude
485: (in high mobility Si-MOSFETs \cite{PRB94}).
486:
487:
488: \item
489: The strong metallic-like $\rho(T)$-dependence is characteristic
490: of a wide range of densities
491: rather than of the critical regime; it sets in
492: for relatively high temperatures $T \leq T_F$
493: in the ballistic rather than diffusive regime \cite{weakloc} (see fig.~\ref{Domains_Si43}).
494:
495: \begin{figure}[htb]
496: \begin{minipage}[t]{.48\textwidth}
497: \includegraphics[width=2.5in]{Fig6c.eps}
498: \caption{Resistivity vs temperature for HIGH mobility Si-MOS
499: sample \protect\cite{JETPL98scaling} in the critical regime
500: $n\approx n_c$. Electron density decreases from 1.326 in steps of
501: 0.0218 (from bottom to top) and in unites of $10^{11}$cm$^{-2}$.}
502: \label{high_mu}
503: \end{minipage}
504: \hfil
505: \begin{minipage}[t]{.47\textwidth}
506: \includegraphics[width=2.4in]{Fig7c.eps}
507: \caption{Resistivity vs $T/E_F$ in a wide range of densities.
508: The diffusive/ballistic border for $T\tau$ \protect\cite{ZNA}
509: is calculated based on the data \protect\cite{weakloc} with
510: renormalized $m^*$.}
511: \label{Domains_Si43}
512: \end{minipage}
513: %\vspace{-0.1in}
514: \end{figure}
515:
516: \item When $n<n_c$, both the sign ($d\rho/dT<0$) and the functional (exponential)
517: form of the $\rho(T)$ dependence are characteristic of the insulating behavior
518: (see figs.~\ref{Si46_Si39},\ref{high_mu}).
519: \item At much higher densities ($n>30\times n_c$), the derivative $d\rho(T)/dT$ is
520: metallic (positive)
521: at higher temperatures $T\sim T_F$, but $\rho$ exhibits an slow (logarithmic)
522: ``insulating'' upturn at lower $T$, consistent with
523: %weak localization,
524: eq.~(2)
525: \cite{Gmax,hamilton99,simmons99} - see fig.~\ref{Gmax}.
526: Qualitatively, conduction
527: at high density is consistent with the weak localization picture.
528:
529: \item
530: The $\log(R(T))$ data show a reflection symmetry with respect to
531: $n_c$ \cite{PRB94,PRB95} (see fig.~\ref{high_mu}), with exclusion of the low-temperature data,
532: which in the $T\rightarrow 0$ limit diverge for $n<n_c$, but do not fall to
533: zero for $n>n_c$.
534: This signifies that at intermediate temperatures, the $\rho(T)$-dependence
535: is empirically represented by a logarithmic function $\rho(T) \propto
536: \exp\left[\pm(T_0/T)^p\right]$ with $p \sim 1 $ and the sign equal to that of $(n_c -n)$
537: \cite{JETPL98scaling};
538: such dependence obviously fits the one-parameter scaling function in the vicinity
539: of a critical point shown in fig.~\ref{beta}\,b.
540: \item
541: For high-mobility samples, the resistances of the metallic and
542: insulating phases, measured as a functions of $T$ and $n$,
543: exhibit scaling properties with respect to $T/T_0$
544: \cite{PRB94,PRB95,JETPL98scaling} demonstrated in
545: fig.~\ref{scaling}. The energy scale $T_0$ goes to zero on both
546: sides of the transition; such behaviors are reminiscent of a
547: critical point of a continuous quantum phase transition
548: \cite{vlad_97}. This interpretation, however,
549: seems to work only for intermediate temperatures
550: and fails in the $T\rightarrow 0$ limit. For a more detailed review
551: and phenomenological discussion of the experimental data in the critical regime,
552: see ref.~\cite{akk}.
553: \begin{figure}[htb]
554: \begin{minipage}[t]{.48\textwidth}
555: \includegraphics[width=2.4in]{Fig8c.eps}
556: \caption{Resistivity for high-$\mu$ sample
557: scaled vs $T_0/T$ \protect\cite{JETPL98scaling}}
558: \label{scaling}
559: \end{minipage}
560: \hfil
561: \begin{minipage}[t]{.48\textwidth}
562: \includegraphics[width=2.4in]{Fig9c.eps}
563: \caption{$R(H_\parallel)$
564: dependences for different densities \protect\cite{anyso}.}
565: \label{R(H)}
566: \end{minipage}
567: %\vspace{-0.2in}
568: \end{figure}
569:
570: \vspace{0.1in}
571: \underline{Features in the in-plane field}
572: \item In Si-MOSFET and p-GaAs structures, $\rho$ increases strongly
573: with $H_\parallel$ \cite{pudalov97,breakdown98,suppression97,okamoto99,yoon99,anyso,eng02}
574: and shows saturation at a field, which is approximately equal to that for
575: complete spin polarization \cite{vitkalov_doubling}; this is
576: most sharply pronounced in Si-MOS structures
577: \cite{suppression97,breakdown98,pudalov97,anyso}.
578: \item
579: As $B_\parallel$ couples to spins rather than to orbital motion,
580: it follows that the spin-related mechanism of interactions
581: (i.e. exchange) is of major importance. More specifically,
582: when $H_\parallel$ field switches off the spin-degrees of freedom,
583: the $R(T)$ dependence flattens and MIT disappears \cite{vitkalov_01,disorder}.
584: \end{itemize}
585:
586: \vspace{0.1in} As $n_c$ is quite small ($\sim 10^{10}-10^{11}$
587: cm$^{-2}$), the corresponding $r_s$ values are large, an order of
588: 10; therefore, it is reasonable to assume the $e-e$ interaction
589: to be one of the major driving forces in the above listed phenomena. In
590: the following section, the $e-e$ interaction effects will be
591: discussed in more detail. We shall show, in particular, how the
592: listed features find a natural explanation within the frameworks
593: of the quantum interaction corrections at intermediate
594: temperatures.
595:
596:
597: \section{Quantitative studies of the electron-electron interactions}
598: \label{sec:e-e}
599: \underline{4.1. Fermi-liquid renormalization of electron parameters in 2D systems}\vspace{0.05in}\\
600: The Landau Fermi-liquid (FL) theory \cite{landau,pines}
601: introduces a number of interaction parameters
602: to describe the interacting system in terms of the
603: renormalized $g^*$-factor Land\'{e},
604: effective mass $m^*$ etc.:
605: \begin{equation}
606: \label{eq:FL-renormalization}
607: F_0^a = \frac{2}{g^*}-1, \qquad F_1^s =2\left(\frac{m^*}{m_b} - 1\right).
608: \end{equation}
609:
610: Recently, these parameters have been determined
611: experimentally in a wide range of densities, in measurements of the
612: quantum oscillations of conductivity (Shubnikov-de Haas (SdH) effect)
613: in tilted or crossed weak magnetic fields
614: \cite{okamoto99,gm,crossed,zhu}. In these experiments, the renormalized spin susceptibility
615: $\chi^*$ is obtained with no assumptions:
616: \begin{equation}
617: \label{eq:chi*}
618: \chi^*= g^*g_b m^* \mu_B \frac{e}{h}, \qquad \frac{\chi^*}{\chi_b} = \frac{g^*m^*}{2m_b}=
619: \frac{\left(F_1^s+2\right)}{\left(F_0^a +1\right)},
620: \end{equation}
621: where $g_b$, $m_b$, and $\chi_b$ are the corresponding band values.
622: Figure~\ref{chi_g_m} shows the $r_s$-dependence of the
623: measured renormalized $\chi^*$ and $m^*$-values
624: \cite{okamoto99,gm}; the renormalized $g^*$-values are obtained
625: from these two sets of data as $\chi^*/m^*$.
626: Other experimental approaches consist in scaling the
627: $\sigma(H/H_0)$-data for different densities
628: \cite{vitkalov_scalingMR_01,vitkalov_scalingMR_02,shashkin_R(B)scaling},
629: or in fitting the $\sigma(T)$ and $\sigma(H_\parallel)$
630: dependences to the calculated quantum corrections \cite{ZNA}, using
631: $F_0^a$ (and in some cases, also $m^*$) as fitting parameter
632: \cite{shashkin_fitting_s(T)02,aleiner,prosk_PRL02,prosk_JPhysA03}.
633:
634: The resulting $F_0^a$ values obtained for 2D electrons in GaAs
635: \cite{zhu} and in Si-MOS samples
636: \cite{shashkin_R(B)scaling,vitkalov_scalingMR_02,gm,aleiner,prosk_JPhysA03,kvon_111Si_2002,shashkin_fitting_s(T)02}
637: qualitatively agree with each other \cite{granada03} but differ
638: substantially from the data for 2D holes in GaAs
639: \cite{prosk_PRL02}, and, especially, from the data taken for low
640: carrier density \cite{noh03}; the reason for this discrepancy is
641: not clear yet \cite{granada03}.
642:
643: \begin{figure}
644: \begin{center}
645: \includegraphics[width=1.0\textwidth]{Fig10.eps}
646: %\vspace{-0.4in}
647: \caption{Renormalized spin susceptibility (a), effective
648: mass (b), and $g^*$-factor (c) measured with Si-MOS
649: samples\protect\cite{gm,granada03}.} \label{chi_g_m}
650: \end{center}
651: %\vspace{-0.4in}
652: \end{figure}
653:
654: Figure~\ref{chi_g_m} shows a strong increase in $\chi^*$, $m^*$,
655: and $g^*$ with $r_s$, which should significantly affect transport
656: at low densities. The growth in $\chi^*$ might reflect a tendency
657: to either ferromagnetic or antiferromagnetic transition; besides,
658: within the FL-description eq.~(\ref{eq:FL-renormalization}), the
659: $g$-factor should diverge as $F_0^a \rightarrow -1$. Therefore, a
660: separate, interesting question is whether or not $\chi^*$, $m^*$
661: and $g^*$ diverge as $r_s$ increases, and whether the divergency
662: (if any) occurs at the MIT. Numerical calculations can not give an
663: unambiguous answer \cite{senatore,paola02,ceperley01} for the
664: single-valley system, whereas the 2D valley system is expected to
665: be more stable and to remain unpolarized up to the WC phase.
666:
667: As a result of the renormalization, the energy scale $E_F$ diminishes.
668: However, in the vicinity of the transition (i.e. at the sample-dependent density $n=n_c$),
669: $m^*$ does not diverge \cite{polariz,granada03}; this may be easily seen,
670: e.g., from non-vanishing amplitude of the SdH oscillations measured at $T=0.03$\,K
671: for $r_s$ values up to 9.5
672: %densities down to $7.7\times 10^{10}$cm$^{-2}$
673: \cite{polariz}. Thus, the MI-transition is not caused by simply
674: vanishing the $E_F$ value and the quasiparticles. In
675: Refs.~\cite{shashkin_R(B)scaling,vitkalov_scalingMR_01} the
676: in-plane magnetoresistance $\rho(H_\parallel)$ data have been
677: discussed in terms of a developing ferromagnetic instability at or
678: very close to $n_c$. However, direct measurements of the (a)
679: magnetization of 2D electrons \cite{reznikov}, (b) period, and
680: (c) sign of the SdH oscillations \cite{polariz,granada03} do not
681: confirm such possibility: the spin susceptibility remains finite
682: at $n \geq 7.7\times 10^{10}$cm$^{-2}$, the density which is
683: lower than the critical $n_c$-value for many samples.
684:
685: The strong enhancement of the weak field spin susceptibility was predicted also
686: by the renorm-group (RG) theory \cite{finkelstein,castellani_chi_84,castellani_chi_86};
687: however the RG-theory is based on consideration of the diffusive modes of electron-electron
688: interaction in the regime $T\tau \ll 1$,
689: whereas the renormalized parameters (Fig.~10) are measured in
690: the ballistic interaction regime $T\tau >1$ (for a more detailed discussion, see Section 4.3).
691:
692: \vspace{0.1in}
693: \underline{4.2. Implementation of the measured FL parameters to the metallic-like transport.}
694: \vspace{0.05in}\\
695: A considerable progress has been also achieved in the theory of quantum transport:
696: in Refs.~\cite{ZNA,gornyi}
697: the interaction corrections
698: to the conductivity have been calculated beyond the diffusive regime, in terms of
699: the FL interaction parameters.
700: In this section, we make
701: a comparison of the data with the theory
702: and conclude that the ``metallic'' drop of conductivity with cooling in the regime
703: $\sigma \gg e^2/h$ can be accounted for by the interaction effects in electron ``liquid''
704: at temperatures which correspond to the ballistic regime $T\tau >1$.
705: This observation suggests that the anomalous
706: ``metallic'' conduction in 2D, at least for densities
707: not too close to the critical density, is
708: the finite-temperature phenomenon rather than the
709: signature of a new quantum ground state.
710:
711:
712: The theory \cite{ZNA}
713: considers backscattering of electrons at the scattering centers
714: and at the Friedel oscillations of the density of surrounding electrons.
715: The interference
716: between these scattering processes gives rise
717: to the quantum corrections, which are calculated to higher orders
718: in the interactions and leading order to the temperature.
719: The interference
720: gives rise to quantum corrections to the Drude conductivity
721: (in units of $e^2/\pi \hbar$):
722: \begin{eqnarray}
723: \label{correction}
724: & \sigma(T)-\sigma_D = \delta \sigma_C(T) + 15\delta\sigma_T(T) %+ \delta\sigma_{\rm loc}(T)
725: . &
726: \end{eqnarray}
727: \begin{eqnarray}
728: {\rm Here}\qquad \qquad \qquad \delta\sigma_C = x\left[1-\frac{3}{8}f(x)\right] -
729: \frac{1}{2\pi}\ln\left(\frac{E_F}{T}\right) \qquad{\rm and} \nonumber\\
730: \delta\sigma_T = A(F_0^a)x\left[1-\frac{3}{8}t\left(x,F_0^a\right)\right]-
731: B(F_0^a)\frac{1}{2\pi}\ln\left(\frac{E_F}{T}\right)
732: \end{eqnarray}
733: are the
734: interaction contributions in the ,,charge'' (exchange term and singlet part of the correlation terms)
735: and triplet channels,
736: respectively; $x=T\tau k_B/\hbar$,\hspace{0.2in}
737: %$\delta\sigma_{\rm loc}(T)= 1/(2\pi) \ln
738: %(\tau/\tau_\varphi(T))$ is the weak localization contribution,
739: $A(F_0^a)=F_0^a/(1+F_0^a)$, and $B(F_0^a) = 1-\ln(1+F_0^a)/F_0^a$.
740: The prefactor 15 to $\delta\sigma_T$ reflects
741: enhanced number of triplet components:
742: two valleys and two spins produce 4 pseudospin components; interaction of two particles
743: involves $4\times 4 =16$ channels, one of them being singlet.
744:
745: \vspace{0.1in}
746: \underline{Diffusive regime, $T\tau \ll 1$}.
747: For %sufficiently
748: large $|F_0^a|$ and $F_0^a<0$, $\delta\sigma(T)$ becomes positive.
749: E.g., for $F_0^a = -0.3$ (i.e. $g^* \approx 3$)
750: \begin{equation}
751: \delta\sigma = \frac{1}{2\pi}\left(1-15\times 0.2 +1 \right)\ln T = -\frac{1}{2\pi}\ln T.
752: \end{equation}
753: Thus, due to valley degeneracy, the quantum correction is constructive
754: and produces a ``forward scattering''.
755:
756: \vspace{0.1in}
757: \underline{Ballistic regime, $T\tau >1$}.
758: The quantum correction $\delta\sigma(T)$ is quasilinear in $T$:
759: \begin{equation}
760: \delta\sigma(T) \approx \left[1+15 \frac{F_0^a}{1+F_0^a}\right](T\tau)
761: \end{equation}
762:
763: \vspace{0.1in}
764: \underline{Comparison of the theory with experiment at high $G\gg e^2/h$ (see Fig.~\ref{Domains_Si43})}.
765: \begin{figure}
766: \begin{center}
767: \includegraphics[width=3.2in]{Fig11c.eps}
768: \end{center}
769: \vspace{-0.2in}
770: \caption{Conductivity for sample Si22 vs
771: temperature \protect\cite{loc02}. The electron densities, from top to bottom are: $n=$
772: 5.7, 6.3, 6.9, 7.5, 8.1, 8.7, 9.3, 10.5, 11.7, 12.9, 14.1, 16.5,
773: 18.9, 21.3, 23.7, 28.5, 35.7 (in units of $10^{11}$cm$^{-2}$).
774: Dots represent the data, solid lines - the
775: theoretical curves with $F_0^a$ and $m^*$ from
776: Ref.~\protect\cite{gm}.}
777: \label{s(T)_comparison}
778: \vspace{-0.25in}
779: \end{figure}
780:
781: The terms in eq.~(\ref{correction}) are functions of $T\tau/\hbar$ and
782: $F_0^a$.
783: The momentum relaxation time $\tau$
784: may be found from the Drude resistivity $\rho_D\equiv
785: \sigma_D^{-1}$ using the renormalized effective mass $m^*$
786: determined in Ref. \cite{gm}.
787: Thus, the comparison shown in fig.~\ref{s(T)_comparison}
788: has no fitting parameters, beyond $\sigma_D$.
789: The determination of the latter value is transparent
790: \cite{aleiner}, it is obtained
791: by a quasilinear extrapolation of the experimental
792: $\sigma(T)$ data to $T=0$, by setting the logarithmic (diffusive) terms to zero.
793: Almost entire temperature range in fig.~\ref{s(T)_comparison} belongs to the ballistic regime.
794: In view of the quantitative agreement of the experimental data with theory,
795: we conclude that the initial slope of the $\sigma(T)$
796: dependence is well described by the quantum interaction corrections
797: eqs.~(\ref{correction})
798: with $F_0^a(r_s)$ and $m^*(r_s)$ values
799: determined independently (see fig.~\ref{chi_g_m}).
800: Several experimental teams came
801: to a similar conclusion
802: for $p$-GaAs/AlGaAs \cite{prosk_JPhysA03,noh03} and
803: Si MOSFETs \cite{shashkin_fitting_s(T)02,kvon_111Si_2002,vitkalov_R(B)03,aleiner}.
804:
805: \vspace{0.1in}
806: We conclude this section by listing the achieved results:
807: \begin{itemize}
808: \item
809: Not too close to the critical regime, $\sigma \gg e^2/h$, the ``metallic''
810: strong $T$-dependence results from
811: the quantum interactions at intermediate range of $T$'s.
812: \item
813: As $T$ decreases further,
814: in the diffusive regime, $\sigma(T)$ is expected to vary $\propto
815: \ln T$ with $d\sigma/dT <0$ for sufficiently large
816: $g^*$ (large $|F_0^a|$) and large number of valleys.
817: Then, for $T <T_0$
818: (determined by the intervalley $h/\tau_{vv}$ and spin-spin scattering rate $h/\tau_{ss}$),
819: the enhancement factor (15) should vanish and the localizing terms are expected to overwhelm.
820: \item
821: If the true metallic phase in 2D exists, it must
822: of a non-FL type \cite{vlad_97,chakravarty_98,chakra_wigner_glass}.
823: Several non-FL theoretical models
824: have been recently suggested, such as
825: phase separation model (Wigner solid inclusions in the FL)
826: \cite{spivak}, an electron-paired state with broken U(1) symmetry \cite{pp},
827: magnetic phase transition etc. However, so far no clear signatures of the
828: non-FL behavior have been detected experimentally.
829: \end{itemize}
830: \vspace{0.1in}
831: %\newpage
832: \underline{4.3. Transport in the critical regime, $\sigma \sim e^2/h$ and $n\approx n_c$}\\
833: In view of the encouraging comparison with the theory of
834: interaction corrections in the high density regime, it seems
835: attractive to extend the comparison to the critical regime, where
836: $\sigma \sim e^2/h$ (see fig.~\ref{Domains_Si43}). Quantum
837: corrections theory is inapplicable for large conductivity changes
838: $\delta\sigma/\sigma \sim 1$ and in the vicinity of the critical
839: point, where $\sigma \sim e^2/h$. The method now commonly in use
840: is a generalization of the nonlinear $\sigma$-model theory, which
841: has been developed by Finkel'stein \cite{finkelstein} and by
842: Castellani, Di Castro et al. \cite{castellani_chi_84,castellani_MIT_84}. The
843: renormalization-group method is an extension of the same
844: diagrams of the interaction theory to the case of the strong
845: disorder and interaction \cite{belitz}. The RG equations describe
846: renormalization of both, disorder (i.e. $\rho$) and interaction
847: parameters
848: as the length scale
849: varies with $T$ \cite{finkelstein,castellani_MIT_84,dicastro}:
850:
851:
852: \begin{eqnarray}
853: \label{RG} \frac{d\rho}{dy} & = &
854: \frac{\rho^2}{\pi}\left[1+1-3\left(\frac{1+\gamma_2}{\gamma_2}
855: \ln(1+\gamma_2)-1\right)\right]\nonumber\\
856: \frac{d\gamma_2}{d y} & = & \frac{\rho(1+\gamma_2)^2}{2\pi},
857: \end{eqnarray}
858: where $y=-\ln(T\tau)$, $\gamma_2$ is the
859: Fermi-liquid amplitude for large angle scattering
860: ($\gamma_2 \rightarrow -F_0^a/2$ in the weak-coupling limit),
861: and $\rho$ is in units of $h/e^2$.
862: The 1st term in the square brackets is the weak localization contribution,
863: the 2nd term is the $e-e$ interaction correction in the singlet channel
864: \cite{altshuler&aronov,lee_review},
865: and the last term is the contribution from 3 triplet modes.
866: Due to the difference in symmetry
867: of the singlet and triplet wavefunctions, the singlet and triplet terms have different sign,
868: and favors localization and delocalization, correspondingly. For two-valley system, factor 3
869: should be again replaced by 15,
870: and the weak-localization term also becomes twice larger \cite{punnoose}.
871: \begin{figure}
872: \begin{center}
873: \includegraphics[width=5.0in]{Fig12.eps}
874: \caption{a) Comparison of the $\rho(T)$ data normalized to its
875: maximum value ($1.35 h/e^2$) with solution of the RG equations (solid line)
876: eq.~(\protect\ref{RG}); b)
877: Renormalized interaction parameter $\gamma_2$
878: \protect\cite{punnoose}, calculated from the 2nd Eq.(11) with
879: $\rho(T)$ data shown on the left panel. The ,,0''-point at the $\ln(T\tau)$ scale
880: corresponds to $T=1.8$K. Sample Si15, electron density $n=0.88\times 10^{11}$cm$^{-2}$.
881: %Solid line is a guide to the eye.
882: }
883: \label{Finkelstein_Si15}
884: \end{center}
885: \vspace{-0.3in}
886: \end{figure}
887:
888: The above equations have a universal solution, which has been %recently
889: compared with experimental
890: $\rho(T)$ data in Ref.~\cite{punnoose}.
891: We present in fig.~\ref{Finkelstein_Si15}
892: a similar comparison,
893: which %indeed
894: confirms a certain similarity between the $\rho(T)$-data and the theory.
895: There is a limitation though, the
896: eqs.~(\ref{RG}) are perturbative, i.e. are derived in the lowest order in $\rho$,
897: therefore, the comparison with experimental data for $\rho(e^2/h) \sim 1$ can be only qualitative.
898:
899: The application of the RG perturbative equations to the experimental data
900: is based on the following assumptions:\\
901: (i) the critical regime $n\approx n_c$ belongs to the diffusive
902: domain of {\it e-e}-interactions $T\tau \ll 1$;\\
903: (ii) the $\rho(T)$ maximum
904: signifies
905: a turning point from localization
906: to delocalization behavior due to the renormalization
907: of the interaction parameter (primarily, $\gamma_2$) with $T$;\\
908: (iii) the decrease in $\rho$ with cooling takes place mainly for the account
909: of the strong growth in the interaction parameter $\gamma_2$ (and hence, $F_0^a$).
910:
911: \vspace{0.1in}
912: The first assumption does not agree well with the empirical diagram
913: in fig.~\ref{Domains_Si43}; therefore, its application would
914: require to re-interpret the definition of the diffusive/ballistic border,
915: i.e. to find a different interpretation
916: for the $\tau(n)$ or $F_0^a(n)$ data in the critical regime.
917: In Ref.~\cite{punnoose},
918: the assumption (ii) was presumed to
919: be fulfilled in the most clean samples; however, some
920: high mobility Si-MOS samples do not exhibit a pronounced $\rho(T)$ maximum
921: in the critical regime \cite{kk,JETPL98scaling} (see, e.g., fig.~\ref{high_mu}).
922: Furthermore, the $\rho(T)$
923: data in the critical regime, as a rule, show a strong
924: sample-dependent and non-universal behavior \cite{JETPL98scaling,cooldowns}.
925: It is possible though that the relevant disorder is related with
926: short-range fluctuating potential at the interface
927: which may favor inter-valley scattering and, thus reduce a
928: number of the triplet terms in eq.~(\ref{RG})
929: but is not significant for the mobility at low $T$.
930:
931: Regarding assumption (iii), figure~\ref{Finkelstein_Si15}\,b shows that the interaction parameter
932: $\gamma_2$, determined from the experimental data, starts growing
933: at rather high temperatures,
934: $\sim 1.8$\,K for $n\approx n_c$; the latter behavior of $\gamma_2$
935: is necessary in the RG-theory to turn down the resistivity flow. Thus, the renormalization
936: of $\gamma_2$ should occur in a quite accessible range of temperatures
937: (0.3 - 1.8)K. However,
938: the anticipated strong $T$-dependence
939: of the renormalized spin susceptibility, eq.~(\ref{eq:chi*}),
940: \cite{castellani_chi_84}
941: is not observed experimentally at these temperatures in SdH-effect measurements:
942: $\chi^*/\chi_b \propto g^*m^*/2m_b$ changes only within 2\%
943: in the above temperature range \cite{granada03}.
944:
945: An alternative theoretical approach \cite{vlad_97} suggests that the critical
946: behavior in the vicinity of $n_c$ is dominated by the physics of
947: the insulating state, namely by the melting of the disordered
948: Wigner crystal \cite{chakra_wigner_glass}, electron glass
949: \cite{thakur99}, or an inhomogeneous state consisting of
950: inclusions of WC into the 2D liquid \cite{spivak}. Corresponding
951: theories has not reached yet a fully predictive stage. We note that
952: signatures of the glassy behavior were reported in Ref.~
953: \cite{snezana02}.
954:
955: \vspace{0.05in}
956: \underline{4.4. Transport in the presence of the in-plane field}\\
957: Figure \ref{R(H)} shows a typical behavior of $\rho(B_\parallel)$.
958: At low densities $n\sim n_c$, as field increases, transport
959: becomes temperature activated (hopping regime)
960: \cite{disorder}.
961: When $n$ approaches $n_c$, by definition,
962: $n_c(H_\parallel)$ tends to
963: $n_c(H_\parallel =0) \equiv n_c$, therefore, the range of magnetic fields
964: where the magnetoresistance (MR) may be studied in the regime of ``metallic'' non-activated conduction,
965: shrinks to zero.
966:
967: \vspace{0.1in}
968: \underline{Regime of high densities $n\gg n_c$, $G\gg e^2/h$.}
969: According to the theory of interaction corrections \cite{ZNA},
970: \begin{equation}
971: \label{eq:s(B)}
972: \delta\sigma(H_\parallel)\equiv \sigma(0,T)-\sigma(E_Z,T)
973: \approx \frac{e^2}{\pi\hbar}f(F_0^a) \frac{T\tau}{\hbar} \left[ K_b\left(\frac{E_z}{2T},F_0^a\right) \right]
974: \end{equation}
975: where $K_b \approx (E_Z/2T)^2 f(F_0^a)$ in the low field limit $E_Z/2T \ll 1$, and
976: $K_b \approx (E_Z/T)f_2(F_0^a)$ in the limit $E_Z/2T \gg 1$ limit.
977: Correspondingly, as field increases, $\delta\sigma(H)$
978: should increase initially $\propto H^2/T$, and than $\propto H$.
979:
980: Experimental data, in general, show similar behavior \cite{vitkalov_R(B)03,gm,aleiner}
981: with a transition from parabolic to linear dependence;
982: however, in contrast to
983: the temperature dependence of $\rho$,
984: the agreement with theory \cite{ZNA} is only qualitative \cite{aleiner}.
985: In high fields, $g^*\mu H_\parallel \gg T$,
986: the magnetoresistance
987: deviates substantially from the theory,
988: and the deviations are sample dependent \cite{aleiner}.
989: As density decreases and field increases,
990: the deviations of the measured MR from the theory grow
991: and reach a factor of two at $n=1.5n_c$ \cite{aleiner,disorder}.
992: The
993: measured $\sigma(H,T)$
994: scales with temperature somewhat different from eq.~(\ref{eq:s(B)}),
995: which predicts $\delta\sigma \propto -(H^2/T)$ in low fields.
996:
997: \vspace{0.05in}
998: \underline{Critical regime of densities $n\approx n_c$, and $\sigma \approx e^2/h$.}
999: As mentioned above,
1000: the magnetoresistance studies in the critical regime
1001: are restricted to the fields vanishing to zero
1002: as $n\rightarrow n_c$. Nevertheless, down to densities $n\approx 1.1n_c$.
1003: the magnetotransport can be safely studied in the regime $g^*\mu H_\parallel \ll T$.
1004: The experimental data in low fields scales
1005: as $\delta\sigma \propto -(H^2/T^p)$,
1006: where $p$ increases from 1.1 to 1.6 as
1007: density decreases from $5n_c$ to $1.2 n_c$.
1008: Comparing this empirical scaling law
1009: with that predicted by the RG theory
1010: \cite{finkelstein,castellani_98},
1011: \begin{equation}
1012: \sigma(H,T)-\sigma(0,T) = -0.084\frac{e^2}{\pi h}\gamma_2\left(\gamma_2 +1\right)
1013: \left(\frac{g\mu H}{k T}\right)^2,
1014: \end{equation}
1015: we conclude that the interaction
1016: parameter $\gamma_2$ (roughly, $\propto |F_0^a|$) {\it decreases} as temperature decreases.
1017: This result is not consistent with
1018: %contradicts
1019: the main idea of the two-parameter scaling, where $\gamma_2$ is
1020: expected to increase and $\chi^*$ diverges $\propto T^{-4/3}$
1021: \cite{castellani_chi_84} as temperature decreases (i.e., the
1022: length scale increases). It is important also that the direct SdH
1023: measurements in low $H_\parallel$ fields do not confirm strong
1024: $T$-dependence of $F_0^a$ \cite{granada03}. It might be that the
1025: critical regime where the RG-description is applicable is much
1026: narrower, than the range of densities ($\approx 20\% n_c$) where
1027: the $\rho(T)$ data exhibit a pronounced maximum. If this is the
1028: case, the $\rho(H_\parallel,T)$ analysis should be restricted
1029: also to much lower magnetic fields. We note that right at the
1030: transition and for lower densities $n\leq n_c$, the $\rho(T)$
1031: displays an approximately $T$-activated exponential dependence
1032: $\rho(H,T) \propto \exp(\Delta /T)$ with $\Delta \propto
1033: (n-n_c(H)) \propto H$ \cite{disorder}; the resulting dependence
1034: $\exp(H/T)$ is not easy to distinguish from the
1035: $(H/T)^2$-dependence.
1036:
1037: In refs.~\cite{disorder,anyso,cooldowns,aleiner},
1038: it was suggested that transport
1039: in the critical regime is driven not only by ``universal'' effects of interactions among
1040: the itinerant electrons, but
1041: also by interactions of itinerant electrons with localized
1042: ones, the latter issue (short-range aspect of interactions)
1043: is missing in most of the existing theories.
1044:
1045: \vspace{0.05in}
1046: \underline{Homogeneity.}
1047: There is an important
1048: issue, whether or not a spatial inhomogeneity
1049: develops in the 2D electron system as the carrier density decreases; such
1050: inhomogeneity was suggested to
1051: cause the percolation-type MI-transition \cite{shi02,meir99}.
1052: Though this issue was considered in a number of theoretical papers, there is no clear
1053: experimental evidence for such mechanism to be the major driving force in the experimentally observed
1054: phenomena in high mobility samples.
1055:
1056: \section{Summary}
1057:
1058: I will present below a brief pedagogical summary of the issues which has been learnt
1059: in the considered field, and those which still remain unanswered.
1060: \begin{itemize}
1061: \item
1062: For high densities and high conductivities,
1063: in the ballistic regime $T\tau \gg 1$,
1064: it is now widely accepted that
1065: the metallic $T$-dependence of the conductivity
1066: is a finite temperature effect caused by $e-e$ interactions. However,
1067: (i) in the diffusive regime $T\tau \ll 1$, a thorough comparison with theory
1068: is missing; some data reveal a substantial disagreement \cite{aleiner} with theory,
1069: which may be attributed, e.g.,
1070: to the intervalley scattering;
1071: (ii)
1072: a successful comparison with interaction theory has not been
1073: demonstrated yet for the same set of $\rho(H, T)$-data for both zero and non-zero in-plane
1074: field.
1075: \item
1076: For low densities, in the critical regime $n \approx n_c$,
1077: the observed one-parameter critical behavior
1078: of conduction $\sigma(T)$ seems to be characteristic of
1079: intermediate temperatures only. For much lower temperatures,
1080: in the diffusive regime,
1081: the potential critical behavior remains unexplored.
1082: \item
1083: So far, all data find a reasonable explanation within the FL theory;
1084: no clear signature of the non-FL behavior was detected.
1085: \item
1086: Both, the growth of the spin susceptibility as density decreases and the
1087: non-monotonic $T$-dependence of resistivity
1088: in the critical regime $n \approx n_c$, are in a qualitative agreement
1089: with the RG theory. However, some other experimental results are inconsistent
1090: with the theory; the discrepancy requires more detailed
1091: analysis of the experimental data and the RG scenario of the MIT in 2D.
1092: \item
1093: The behavior of the renormalized spin susceptibility and effective mass for 2D holes at
1094: high $r_s$ values remains puzzling; the reason of the discrepancy
1095: between the 2D hole and electron systems is not clear.
1096: \item
1097: By now,
1098: there is no complete theory which would include
1099: short-range and high-energy aspects
1100: of $e-e$ interactions, inter-valley scattering, etc.
1101: \item
1102: The interesting ideas of the potential spontaneous polarization transition
1103: in the spin or valley system require investigations at substantially
1104: higher $r_s$ values.
1105: \item
1106: The detailed experimental verification of the possible formation of local moments,
1107: melting of the Wigner glass and other non-FL scenarios is currently lacking.
1108:
1109:
1110: \end{itemize}
1111:
1112:
1113:
1114:
1115: \acknowledgments
1116: %This paper is based on the lecture given at the 2003 summer school in Varenna.
1117: I am grateful to the Italian Physical Society for the hospitality,
1118: and to G.~Giulliani and G.~Vignale for stimulating fruitful atmosphere at
1119: the 2003 Enrico Fermi school.
1120: I would like to thank
1121: C.~Di Castro, D.~Ceperley, G.~Senatore, and B.~Tanatar for interesting discussions.
1122: The reviewed data were taken with M.~Gershenson, H.~Kojima,
1123: E.~Dizhur, G.~Brunthaler, A.~Prinz, and G.~Bauer, whom I would
1124: like to specially thank. The high mobility Si-MOS samples were
1125: manufactured in collaboration with M.~Vernikov, L.~Pazinich, and
1126: S.~G.~Semenchinskii. The research done by our experimental group
1127: was supported by INTAS, RFBR, and Russian grants from the
1128: Ministry for science and education, and the Presidential program
1129: of the support of leading scientific schools.
1130:
1131:
1132: \begin{thebibliography}{0}
1133:
1134: \bibitem{mott}\BY{Mott~N.~F.} \TITLE{Metal-Insulator Transitions}
1135: (London, Tailor \& Hobson) 1990.
1136:
1137: \bibitem{ando_review}\BY{Ando~T., Fowler~A.~B. \atque Stern~F.}
1138: \IN{Rev. Mod. Phys.}{54}{1982} 432.
1139:
1140: \bibitem{Nagaoka_review}for a review see, \TITLE{Anderson Localization} edited by
1141: \NAME{Nagaoka~Y. \atque Fukuyama~H.} (Springer, New York) 1982.
1142:
1143: \bibitem{ceperley}\BY{Tanatar~B. \atque Ceperley~D.~M.} \IN{Phys. Rev. B}{39}{1989}{5005}.
1144: \bibitem{chui&tanatar}\BY{Chui~S.~T. \atque Tanatar~B.}
1145: \IN{Phys. Rev. Lett.}{74}{1995}{458}.
1146:
1147: \bibitem{giamarchi}\BY{R. Chitra, T. Giamarchi, P. Le Doussal}
1148: \IN{Phys. Rev. B}{65}{2001}{035312}.
1149:
1150: \bibitem{efros&shklovskyi}\BY{Efros A.~L. \atque Shklovsii~B.~I.}
1151: \TITLE{Electronic Properties of Doped Semiconductors}
1152: in \TITLE{Springer Series in Solid State Sciences}
1153: {45} (Springer Verlag, New York, 1984).
1154:
1155: \bibitem{altshuler&aronov}\BY{Al'tshuler~B.~L.\atque Aronov~A.~G.}
1156: in \TITLE{Electron-electron interactions in disordered systems} ed. by
1157: \NAME{Efros A.~L.
1158: \atque Pollack M.} (North-Holland, Amsterdam, 1985), pp.~1-154.
1159:
1160: \bibitem{lee_review}\BY{Lee~P.~A. \atque Ramakrishnan~T.~V.}
1161: \IN{Rev. Mod. Phys.}{57}{1985}{287}.
1162:
1163: \bibitem{Gmax} \BY{Pudalov~V.~M., Brunthaler~G., Prinz~A. \atque Bauer~G.}
1164: \IN{Phys. Rev. B}{60}{1999}{2154}.
1165:
1166: \bibitem{fukuyama82} \BY{Fukuyama~H.} \IN{Surf. Sci.} {113}{1982}{489}.
1167:
1168: \bibitem{gang4}\BY{Abrahams~E.~A., Anderson~P.~W., Licciardello~D.~C. \atque
1169: Ramakrishnan~T.~V.}\IN{Phys. Rev. Lett.}{42}{1979}{673}.
1170:
1171: \bibitem{hikami80}\BY{Hikami~S.~A., Larkin~A.~I. \atque Nogaoka~Y.}
1172: \IN{Progr. Theor. Phys. Lett.}{63}{1980}{707}.
1173:
1174: \bibitem{finkelstein}\BY{Finkelstein~A.~M.} \IN{Sov.Phys. JETP}{57}{1983}{97};
1175: \IN{Z. Phys. B}{56}{1984}{189}; \IN{Sov. Phys. Reviews, Sec. A}{14}{1990}{1}.
1176:
1177:
1178: \bibitem{ZNA}\BY{Zala G., Narozhny B.~N. \atque Aleiner I.~L.}
1179: \IN{Phys. Rev. B}{64}{2001}{214204}.
1180: Phys. Rev. B {\bf 65}, 020201 (2001).
1181:
1182: \bibitem{gornyi}\BY{Gornyi~I.V., Mirlin~A.~D.}
1183: \IN{Phys. Rev. Lett.}{90}{2003}{076801}.\IN{Phys. Rev. B}{69}
1184: {2004}{045313}.
1185:
1186: \bibitem{bishop}\BY{Bishop~D.~J., Tsui~D.~C. \atque Dines~R.C.}
1187: \IN{Phys. Rev. Lett.}{44}{1980}{1153}.
1188:
1189: \bibitem{uren}\BY{Uren~M.~J., Davis~R.A., Kaveh~M. \atque Pepper~M.}
1190: \IN{J. Phys. C}{14}{1981}{5737}.
1191:
1192: \bibitem{highmu}As sample mobility increases,
1193: the disorder and resistivity decrease
1194: for the given carrier density. Correspondingly,
1195: the onset of the exponential insulating behavior
1196: is shifted to lower $n_c$ values,
1197: thus resulting in
1198: larger values of $r_s$ and of the interaction parameter $|F_0^a|$.
1199:
1200: \bibitem{cham80}\BY{Cham~K.~M. \atque Wheeler~R.~G.}
1201: \IN{Phys. Rev. Lett.}{44}{1980}{1472}.
1202:
1203: \bibitem{smith86}\BY{Smith~ R.~P. \atque Stiles~P.~J.}
1204: \IN{Solid State Commun.}{58}{1986}{511}.
1205:
1206: \bibitem{dorozh88}\BY{Vyrodov~E.A., Dolgopolov~V.T., Dorozhkin~S.~I. \atque Zhitenev~N.~B.}
1207: \IN{Sov. Phys. JETP}{67}{1988}{998}.
1208:
1209: \bibitem{stern80}\BY{Stern~F.} \IN{Phys. Rev. Lett.}{44}{1980}{1469}.
1210:
1211:
1212: \bibitem{das86}\BY{Das Sarma~S.} \IN{Phys. Rev. B}{33}{1986}{5401}.
1213:
1214: \bibitem{gd86}\BY{Gold~A. \atque Dolgopolov~V.~T.} \IN{Phys. Rev. B}{33}{1986}{1076}.
1215:
1216:
1217: \bibitem{reentrant_dio}\BY{D'Iorio~M., Pudalov~V.~M. \atque Semenchinskii~S.~G.}
1218: \IN{Phys. Lett A}{150}{1990}{422}.
1219:
1220: \bibitem{dioPRB92}\BY{D'Iorio~M., Pudalov~V.~M. \atque Semenchinskii~S.~G.}
1221: \IN{Phys. Rev. B}{46}{1992}{15992}.
1222:
1223: \bibitem{hall93}\BY{Pudalov~V.~M., Campbell~J. \atque D'Iorio~M.}
1224: \IN{JETP Lett.}{57}{1993}{608}.
1225:
1226:
1227: \bibitem{halperin} \BY{Halperin B.~I.} \IN{Phys. Rev. B}{25}{1982}{2185}.
1228:
1229: \bibitem{khmelnitskii}\BY{Khmelnitskii D.~E.} \IN{Phys. Lett. A}{106}{1984}{182}.
1230:
1231:
1232: \bibitem{laughlin}\BY{Laughlin R.~B.} \IN{Phys. Rev. Lett.}{52}{1984}{2304}.
1233:
1234: \bibitem{PRB94}\BY{Kravchenko S.~V., Kravchenko G.~V.,
1235: Furneaux~J.~E., Pudalov~V.~M. \atque D'Iorio M.} \IN{Phys. Rev. B}{50}{1994}{8039}.
1236:
1237: \bibitem{PRB95}\BY{Kravchenko S.~V., Mason~W.~E., Bowker~G.~E.,
1238: Furneaux~J.~E., Pudalov~V.~M. \atque D'Iorio~M.}\IN{Phys. Rev. B}{51}{1995}{7038}.
1239:
1240: \bibitem{sk_RMP} \BY{Abrahams~E., Kravchenko~S.~V. \atque
1241: Sarachik~S.~V.} \IN{Rev. Mod. Phys.}{73}{2001}{251}.
1242:
1243:
1244: \bibitem{coleridge}\BY{Coleridge P.~T., Williams R.~L., Feng~Y. \atque Zawadzki~P.}
1245: \IN{Phys. Rev. B}{56}{1997}{12764}.
1246:
1247: \bibitem{lam97}\BY{Lam~J., D'Iorio~M., Brown~D. \atque Lafontaine~H.}
1248: \IN{Phys. Rev. B}{56}{1997}{12741}.
1249:
1250: \bibitem{Brunthaler}\BY{Brunthaler~G., Prinz~A., Pillwein~G., Bauer~G.,
1251: Brunner~K., Abstreiter~G.,
1252: Dietl~T. \atque Pudalov~V.~M.} in \TITLE{Proc. ICPS-25} (Berlin,
1253: Springer-Verlag, 2001) p.785.
1254:
1255: \bibitem{pepper}\BY{Simmons~M.~Y., Hamilton~A.~R., Pepper~M., Linfield~E.~H.,
1256: Rose~P.~D, Ritchie~D.~A., Savchenko~A.~K. \atque Griffiths~T.~G.}
1257: \IN{Phys. Rev. Lett.}{80}{1998}{1292}.
1258:
1259: \bibitem{shahar}\BY{Hanein~Y., Meirav~U., Shahar~D.,
1260: Li~C.~C., Tsui~D.~C. \atque Shtrikman Hadas}
1261: \IN{Phys. Rev. Lett.}{80}{1998}{1288}.
1262:
1263: \bibitem{papadakis}\BY{Papadakis~S.~J., De Poortere~E.~P., Manoharan~H.~C.,
1264: Shayegan~M. \atque Winkler R.}
1265: \IN{Science}{283}{1999}{2056}.
1266:
1267: \bibitem{mills}\BY{Mills~A.~P.~Jr.,
1268: Ramirez~A.~P., Pfeiffer~L.~N. \atque West~K.~W.}
1269: \IN{Phys. Rev. Lett}{83}{1999}{2805}.
1270:
1271: \bibitem{noh03}\BY{Noh~H., Lilly~M.~P., Tsui~D.~C., Simmons~J.~A.,
1272: Pfeiffer~L.~N. \atque West~K.~W.} \IN{Phys. Rev. B}{68}{2003}{165308}.
1273:
1274: \bibitem{ngaas}\BY{Hanein~Y., Shahar~D.,
1275: Yoon~J., Li~C.~C. \atque Tsui~D.~C. \atque
1276: Shtrikman Hadas} \IN{Phys. Rev. B}{58}{1998}{13338}.
1277:
1278: \bibitem{ensslin}\BY{Ribeiro~E., Jaggi~R.~D., Heinzel~T., Ensslin~K.,
1279: Medeiros-Ribeiro~G. \atque Petroff~P.~M.}
1280: \IN{Phys. Rev. Lett.}{82}{1999}{996}.
1281:
1282: \bibitem{alas}\BY{Papadakis~S.~J. \atque Shayegan~M.}
1283: \IN{Phys. Rev. B}{57}{1998}{15068}.
1284:
1285: \bibitem{SOI}\BY{Brunthaler~G., Prinz~A., Pillwein~G., Bauer~G.,
1286: Pudalov~V.~M., Lindelof~P.~E. \atque Ahopelto~J.}
1287: \IN{Physica E}{13}{2002}{691}.
1288:
1289: \bibitem{fowler97}\BY{Popovic~D., Fowler~A.~B. \atque Washburn~S.}
1290: \IN{Phys. Rev. Lett.}{79}{1997}{1543}.
1291:
1292: \bibitem{weakloc}\BY{Brunthaler G., Prinz A., Bauer G. \atque
1293: Pudalov V.~M.} \IN{Phys.~Rev.~Lett.}{87}{2001}{096802}.
1294:
1295: \bibitem{hamilton99}\BY{Hamilton~A.~R., Simmons~M.~Y.,
1296: Pepper~M., Linfield~E.~H., Rose~P.~D., \atque Ritchie~D.~A.}
1297: \IN{Phys. Rev. Lett.}{82}{1999}{1542}.
1298:
1299: \bibitem{simmons99}\BY{Simmons~M.~Y., Hamilton~A.~R., Pepper~M., Linfield~E.~H.,
1300: Rose~P.~D. \atque Ritchie~D.~A.}\IN{Phys. Rev. Lett.}{84}{2000}{2489}.
1301: cond-mat/9910368.
1302:
1303:
1304: \bibitem{JETPL98scaling}\BY{Pudalov~V.~M., Brunthaler~G., Prinz~A. \atque Bauer~G.}
1305: \IN{JETP Lett.}{68}{1998}{442}.
1306:
1307:
1308: \bibitem{vlad_97}\BY{Dobroslavljevi\'c~V.,
1309: Abrahams~E., Miranda~E. \atque Chakravarty~Sudip} \IN{Phys. Rev.
1310: Lett.}{79}{1997}{455}.
1311:
1312: \bibitem{akk}\BY{Altshuler~B.~L., Maslov~D.~L. \atque Pudalov~V.M.}
1313: \IN{Physica E}{9}{209}{201}.
1314:
1315: \bibitem{pudalov97}\BY{Pudalov~V.~M., Brunthaler~G., Prinz~A. \atque
1316: Bauer~G.} \IN{JETP Lett.}{65}{1997}{932}.
1317:
1318: \bibitem{breakdown98}\BY{Pudalov~V.~M., Brunthaler~G., Prinz~A. \atque
1319: Bauer~G.} \IN{Physica B}{249-251}{1998}{697}.
1320:
1321: \bibitem{suppression97}\BY{Simonian~D.,
1322: Kravchenko~S.~V., Sarachik~M.~P. \atque Pudalov~V.~M.}
1323: \IN{Phys. Rev. Lett.}{79}{1997}{2304}.
1324:
1325: \bibitem{okamoto99}\BY{Okamoto~T., Hosoya~K., Kawaji~S. \atque Yagi~A.}
1326: \IN{Phys. Rev. Lett.}{82}{1999}{3875}.
1327:
1328: \bibitem{yoon99}\BY{Yoon~Jongsoo, Li~C.~C., Shahar~D., Tsui~D.~C. \atque Shayegan~M.}
1329: \IN{Phys. Rev. Lett.}{84}{2000}{4421}.
1330:
1331: \bibitem{anyso}\BY{Pudalov~V.~M., Brunthaler~G., Prinz~A. \atque
1332: Bauer~G.} \IN{Phys. Rev. Lett.}{88}{2002}{076401}.
1333:
1334: \bibitem{eng02}\BY{Eng~K., Feng~X.~G., Popovi\'{c}~D. \atque Washburn~S.}
1335: \IN{Phys. Rev. Lett.}{88}{2002}{136402}.
1336:
1337:
1338: \bibitem{vitkalov_doubling}\BY{Vitkalov~S.~A., Zheng~H., Mertes~K.~M., Sarachik~M.~P.
1339: \atque Klapwijk~T.M.}\IN{Phys. Rev. Lett.}{85}{2000}{2164}.
1340:
1341: \bibitem{vitkalov_01}\BY{Mertes~K.~M., Zheng~H., Vitkalov~S.~A., Sarachik~M.~P.
1342: \atque Klapwijk~T.M.}
1343: \IN{Phys. Rev. B}{63}{2001}{041101}.
1344:
1345: \bibitem{disorder}\BY{Pudalov~V.~M., Brunthaler~G., Prinz~A. \atque Bauer~G.}
1346: cond-mat/0103087.
1347:
1348: \bibitem{landau}\BY{Lifshitz~E.~M. \atque Pitaevskii~L.~P.}
1349: \TITLE{Statistical physics} (Pergamon Press, Oxford, 1980).
1350:
1351: \bibitem{pines}\BY{Pines David \atque Nozi\'{e}res Philippe}
1352: \TITLE{The theory of quantum liquids} (W.~A.~Benjamin, Inc., New York, 1966).
1353:
1354: \bibitem{gm}\BY{Pudalov~V.~M., Gershenson~M.~E., Kojima~H., Butch~N.,
1355: Dizhur~E.~M., Brunthaler~G., Prinz~A. \atque Bauer G.}
1356: \IN{Phys. Rev. Lett.}{88}{2002}{196404}.
1357:
1358: \bibitem{crossed}\BY{Gershenson~M.~E., Pudalov~V.~M., Kojima~H., Butch~N.,
1359: Dizhur~E.~M., Brunthaler~G., Prinz~A. \atque Bauer G.}
1360: \IN{Physica E}{12}{2002}{585}.
1361:
1362: \bibitem{zhu}\BY{Zhu~J., Stormer~H.~L., Pfeiffer~L.N., Baldwin~K.~W.
1363: \atque West~K.~W.} \IN{Phys. Rev. Lett.} {90}{2003}{056805}.
1364:
1365: \bibitem{vitkalov_scalingMR_01}\BY{Vitkalov~S.~A., Zheng~H., Mertes~K.~M.,
1366: Sarachik~M.~P. \atque Klapwijk~T.~M.}
1367: \IN{Phys. Rev. Lett.}{87}{2001}{086401}.
1368:
1369: \bibitem{vitkalov_scalingMR_02}\BY{Vitkalov~S.~A.,
1370: Sarachik~M.~P. \atque Klapwijk~T.M.}
1371: \IN{Phys. Rev. B}{65}{2002}{201106}.
1372:
1373:
1374: \bibitem{shashkin_R(B)scaling}\BY{Shashkin~A.~A., Kravchenko~S.~V.,
1375: Dolgopolov~V.~T. \atque Klapwijk~T.~M.}
1376: \IN{Phys. Rev. Lett.}
1377: {87}{2001}{086801}.
1378:
1379:
1380: \bibitem{shashkin_fitting_s(T)02}\BY{Shashkin~A.~A., Kravchenko~S.~V.,
1381: Dolgopolov~V.~T. \atque Klapwijk~T.~M.}
1382: \IN{Phys. Rev. B}{66}{2002}{73303}.
1383:
1384:
1385: \bibitem{aleiner}\BY{Pudalov~V.~M., Gershenson~M.~E., Kojima~H.,
1386: Brunthaler~G., Prinz~A., \atque Bauer G.}\IN{Phys. Rev. Lett.}{91}{2003}{126403}.
1387:
1388: \bibitem{prosk_PRL02}\BY{Proskuryakov~Y.~Y., Savchenko~A.~K.,
1389: Safonov~S.~S., Pepper~M., Simmons~M.~Y. \atque Ritchie~D.~A.}
1390: \IN{Phys. Rev. Lett.}{89}{2002}{076406}.
1391:
1392: \bibitem{prosk_JPhysA03}\BY{Proskuryakov~Y.~Y., Savchenko~A.~K.,
1393: Safonov~S.~S., Pepper~M., Simmons~M.~Y. Ritchie~D.~A. Linfield~E.H. \atque
1394: Kvon~Z.D.}\IN{J. Phys. A}{36}{2003}{9249}.
1395:
1396:
1397: \bibitem{kvon_111Si_2002}\BY{Kvon~Z.~D., Estibals~O., Gusev~G.~M. \atque Portal~J.~C.}
1398: \IN{Phys. Rev. B}{65}{2002}{161304}.
1399:
1400: \bibitem{granada03}\BY{Pudalov~V.~M., Gersheson~M.~E., \atque
1401: Kojima~H.} Chapter 19 in: {\it Proceedings of the
1402: EURESCO conference ``Fundamental Problems of Mesoscopic Physics''
1403: (Granada, Sept., 2003), Kluwer} (2004) 309; cond-mat/0401396.
1404:
1405: \bibitem{senatore}\BY{Senatore~G., Moroni~S. \atque Varzano~D.}
1406: \IN{Sol. St. Commun.}{119}{2001}{333}.
1407:
1408: \bibitem{paola02}\BY{Attaccalite~C., Moroni~S., Gori-Giorgi~P.
1409: \atque Batchelet~G.~B.} \IN{Phys.~Rev.~Lett.}{88}{2002}{256601}.
1410:
1411: \bibitem{ceperley01} \BY{Bernu~B., Candido~L. \atque Ceperley~D.}
1412: \IN{Phys. Rev. Lett.}{86}{2001}{870}.
1413:
1414: \bibitem{polariz}\BY{Pudalov~V.~M., Gershenson~M. \atque Kojima~H.}
1415: cond-mat/0110160.
1416:
1417:
1418: \bibitem{reznikov}\BY{Prus~O., Yaish~Y., Reznikov~M., Sivan~U. \atque Pudalov~V.M.}
1419: \IN{Phys. Rev. B}{67}{2003}{205407}.
1420:
1421:
1422: \bibitem{castellani_chi_84}\BY{Castellani~C., Di Castro~C., Lee~P.A.,
1423: Ma~M., Sorella~S. \atque Tabet E.} \IN{Phys.~Rev. B}{30}{1984}{1596}.
1424:
1425: \bibitem{castellani_chi_86}\BY{Castellani~C., Di Castro~C. \atque Sorella~S.}
1426: \IN{Phys.~Rev. B}{34}{1986}{1349}.
1427:
1428: \bibitem{loc02}\BY{Kojima~H., Gershenson~M.~E., Pudalov~V.~M., Brunthaler~G.,
1429: Prinz~A. \atque Bauer~G.} \IN{J. Phys. Soc. Jpn. Suppl. A}{72}{2003}{57}.
1430:
1431: \bibitem{vitkalov_R(B)03}\BY{Vitkalov~S.~A., James~K., Narozhny~B.~N.,
1432: Sarachik~M.~P. \atque Klapwijk~T.M.} \IN{Phys. Rev. B}{67}{2003}{113310}.
1433:
1434: \bibitem{chakravarty_98}\BY{Chakravarty~S., Yin~L. \atque Abrahams~E.}
1435: \IN{Phys. Rev. B}{58}{1998}{559}.
1436:
1437: \bibitem{chakra_wigner_glass}\BY{Chakravarty~S., Kivelson~S., Nayak~C. \atque Voelker~K.}
1438: \IN{Phil. Mag. B}{79}{1999}{859}.
1439:
1440:
1441: \bibitem{spivak}\BY{Spivak~B.} \IN{Phys Rev. B}{64}{2001}{085317}.
1442:
1443: \bibitem{pp}\BY{Phillips~P., Wan~Y., Martin~I., Knysh~S. \atque Dalidovich~D.}
1444: \IN{Nature(London)}{395}{1998}{253}.
1445:
1446: \bibitem{castellani_MIT_84}\BY {Castellani~C., Di Castro~C., Lee~P.~A. \atque Ma~M.}
1447: \IN{Phys. Rev. B}{30}{527}{1984}.
1448:
1449: \bibitem{belitz}for a review, see: \BY{Belitz~D. \atque Kirkpatrick~T.~R.}
1450: \IN{Rev. Mod. Phys.}{66}{1994}{261}.
1451:
1452: \bibitem{dicastro}\BY{Castellani~C., Kotliar~G. \atque Lee~P.~A.}
1453: \IN{Phys. Rev. Lett.}{59}{1987}{323}.
1454:
1455: \bibitem{punnoose}\BY{Punnoose A. \atque Finkelstein~A.~M.} \IN{Phys. Rev.
1456: Lett.}{88}{2002}{016802}.
1457:
1458: \bibitem{kk}\BY{Kravchenko~S.~V. \atque Klapwijk~T.~M.} \IN{Phys. Rev.
1459: Lett.}{84}{2000}{2909}.
1460:
1461: \bibitem{cooldowns}\BY{Pudalov~V.~M., Gershenson~M.~E. \atque Kojima~H.}
1462: \IN{J. Phys. Soc. Jpn.}{72}{2003} Suppl. A, p.57.
1463:
1464:
1465: \bibitem{thakur99}\BY{Thakur~J.~S. \atque Neilson~D.} \IN{Phys. Rev.
1466: B}{59}{1999}{5280}. \BY{Pastor~A.~A. \atque Dobrosavljevic~V.} \IN{Phys. Rev.
1467: Lett.}{83}{1999}{4642}.
1468:
1469:
1470: \bibitem{snezana02}\BY{Bogdanovich~S. \atque Popovi\'{c}}
1471: \IN{Phys. Rev. Lett.}{88}{2002}{236401}.
1472:
1473: \bibitem{castellani_98}\BY{Castellani~C., Di Castro~C. \atque Lee~P.~A.}
1474: \IN{Phys. Rev. B}{57}{1998}{9381}.
1475:
1476: \bibitem{shi02}\BY{Shi~J. \atque Xie~X.C}
1477: \IN{Phys. Rev. Lett.}{88}{2002}{086401}.
1478:
1479: \bibitem{meir99}\BY{Meir~Y.} \IN{Phys. Rev. Lett.}{83}{1999}{3506}.
1480:
1481: \end{thebibliography}
1482: \end{document}
1483: \endinput
1484: