2 1 0 2 TRANSPORT VIA CLASSICAL PERCOLATION AT QUANTUM n HALL PLATEAU TRANSITIONS a J 3 1 MARTINAFLO¨SER Institut N´eel, CNRS and Universit´e Joseph Fourier, B.P. 166, 25 Avenue des Martyrs, 38042 ] l Grenoble Cedex 9, France l a h - SERGEFLORENS s e Institut N´eel, CNRS and Universit´e Joseph Fourier, B.P. 166, 25 Avenue des Martyrs, 38042 m Grenoble Cedex 9, France . t a THIERRYCHAMPEL m Universit´e Joseph Fourier Grenoble I / CNRS UMR 5493, Laboratoire de Physique et - d Mod´elisation des Milieux Condens´es, B.P. 166, 38042 Grenoble, France n o c Weconsidertransportpropertiesofdisorderedtwo-dimensionalelectrongasesunderhigh [ perpendicularmagneticfield,focusinginparticularonthepeaklongitudinalconductivity 1 σxpxeak atthequantumHallplateautransition.Weusealocalconductivitymodel,valid at temperatures high enough such that quantum tunneling is suppressed, taking into v account the random drift motion of the electrons in the disordered potential landscape 7 andinelasticprocessesprovidedbyelectron-phononscattering.Adiagrammaticsolution 7 of this problem is proposed, which leads to a rich interplay of conduction mechanisms, 27 whereclassicalpercolationeffectsplayaprominentrole.Thescalingfunctionforσxpxeak isderivedinthehightemperaturelimit,whichcanbeusedtoextractuniversalcritical . 1 exponentsofclassicalpercolationfromexperimentaldata. 0 2 PACSnumbers:73.43.Qt,64.60.ah,71.23.An 1 : v 1. Introduction i X ThequantumHalleffect1,2 intwo-dimensionalelectrongases(2DEG)followsfroma r a disorder-inducedlocalizationprocesspeculiartothesituationoflargeperpendicular magnetic fields B. While the formation of discrete Landau levels (LL) at energies E =(cid:126)ω (cid:0)n+ 1(cid:1) can account for the existence of robust quantum numbers (with n c 2 ω = eB/m(cid:63) thecyclotronfrequency,e= e theelectroncharge,m(cid:63) theeffective c | | −| | mass,(cid:126)Planck’sconstantdividedby2π,andnapositiveinteger),theexistenceofa macroscopicnumberoflocalizedstatesinthebulkof2DEGisanessentialaspectof quantumHallphysics3.Manyquestionsareyetstillopenthirtyyearsaftertheinitial discovery: i) for metrological purposes4, which physical processes are limiting the plateauquantizationoftheHallconductivityneartheuniversalvalueσ =ne2/h?; xy ii) what is the nature of the localization/delocalization transition from one plateau 1 2 to the next5,6,7, whereupon highly dissipative transport sets in? Theoretically, the problem in its full complexity requires to understand the quantum dynamics of electronssubjecttotheLorentzforceandrandomlocalelectricfields,possiblywith the inclusion of dissipative processes such as electron-electron and electron-phonon interaction, that are sensitive issues when one considers transport properties. Sofar,alotofattentionwasturnedtowardstheunderstandingofthedelocaliza- tion process in terms of a zero-temperature quantum percolation phase transition, which still remains a challenge for the theory5,6,7, despite intriguing experimen- tal evidence from transport8,9 and local scanning tunneling spectroscopy10. In this framework, the quantum tunneling and interference of the guiding center trajecto- rieswithinacomplexpercolationclusterallowdissipationtodevelopinanon-trivial way at the quantum Hall transition. Obviously, increasing temperature from abso- lute zero will generate inelastic processes limiting the coherence between saddle points of the disorder lanscape, so that the quantum character of the transition be- comesprogressivelyirrelevant.Inthatcase,asimplerquasiclassicaltransporttheory becomesvalid11,12,13,14,whichincorporatesthe fastcyclotronmotion withthe slow guiding center drifting, and takes into account inelastic contributions to transport. The transport problem does not become however totally trivial, because classical percolation in the related advection-diffusion regime is still not fully understood15. The aim of the present paper is two-fold. First, we will show in Sec. 2 that high mobilitysamplesdisplayaveryrichtemperaturebehaviorforthepeaklongitudinal conductivity σpeak (at the plateau transition), leading to a complex succession of xx transportcrossoverswithuniversalpowerlaws,seeFig.1.Second,wewillpresentin 1 T1−κ /h] Tβ 1 Tα 1 Vxx 2[e − T2κ 1 ) 0.5 − T ( k a peσxx T1−κ I I Vxy (lB/ξ)2v (lB/ξ)v v ¯hωc/4 T [a.u.] Fig.1. Left:measurementoflongitudinalVxx andHallVxy voltageswithappliedcurrentI ina two-dimensionalsamplewithpercolatingrandomchargeinhomogeneities.Right:sketch(onlog-log scale) of the temperature dependence of the peak longitudinal conductivity σxpxeak at the plateau transition.Theexistenceofahierarchyofenergyscales(thatareindicatedbydashedlines)results inseveralcrossoversbetweenuniversalpower-laws,asdecribedlaterinthetext. Sec. 3 a general diagrammatic formalism14 allowing to compute dissipative trans- port dominated by classical percolation effects in quantum Hall samples such as 3 depicted in Fig. 1. In particular, we will be able to give strong support to a pre- viously conjectured11,12,13,15 critical exponent κ = 10/13 for the peak longitudinal conductivityσpeak inthehightemperatureregimeoftheplateautransition.Anuni- xx versal scaling function describing the crossover for temperatures near the cyclotron energy (see the V-shaped part of the curve in Fig. 1) will be computed, giving a way to extract κ from experiments. 2. Classically percolating transport at the plateau transition 2.1. Local conductivity model The starting point of transport calculations in the high temperature regime of the quantum Hall effect is a purely classical model11,14, where the continuity equation ∇ j = 0 (i.e. the continuum version of Kirchoff’s law) is solved from the micro- · scopicknowledgeofOhm’slawj(r)=σˆ(r)E(r),definingherethelocalconductivity tensor16,17 that relates the local electric field to the local current density. Due to the existence of several energy scales, the disorder-induced spatial variations of σˆ(r) have a strong temperature dependence, which in turn affects the macroscopic transport properties. Drastic simplifications occur in the regime of high magnetic fields17,18, wherethe combination of Lorentzforce and localelectrostatic potentials inducesaslowdriftmotioninthedirectionorthogonaltothecrossedmagneticand localelectricfields.Thisvindicatesthefirstsimplificationoftheconductivitytensor (cid:18) (cid:19) σ σ (r) σˆ(r)= 0 − H , (1) σ (r) σ H 0 where σ encodes dissipative processes such as electron-phonon scattering, which 0 willbeassumedtobeuniforminthebulkofthesample,andσ (r)isthelocalHall H component, whose spatial dependence originates from charge density fluctuations due to disorder in the sample. We will discuss below the various regimes that can be expected for σ (r) depending on the range of temperature T. For this purpose, H we need to introduce the energy scales associated with the local disorder potential (cid:113) (cid:10) (cid:11) V(r), and we define its typical amplitude v = [V(r)]2 and correlation length ξ. Wewillassumethroughoutthatthedisorderissmoothatthescaleofthemagnetic length l = (cid:112)(cid:126)/eB (l = 8nm at B = 10T), so that l /ξ is a small parameter. B B B In what follows, a centered Gaussian distribution will be considered for the local electrostatic potential V(r). The local conductivity model introduced here is valid at temperatures high enough so that phase-breaking processes, such as electron- phononscattering,occuronlengthscalesthatareshorterthanthetypicalvariations of disorder. However, quantum mechanics may still be important to determine the microscopics of the conductivity tensor, as we argue below. 2.2. Percolation effects in quantum Hall transport: phenomenology TheoccurenceofpercolationeffectsinthequantumHallregimecanbeunderstood alreadyfromaquasiclassicalperspective.Inthehighmagneticfieldlimit,cyclotron 4 andguidingcentermotionsfullydecouple,givingrisetoLandauquantizationonone hand,andtomainlyclosedtrajectoriesoftheguidingcenterontheotherhand,that followtheequipotentialsofthedisorderlandscape.Intuitively,theelectricalcurrent contributing to macroscopic transport will thus follow a percolation backbone. The crucial role of inelastic processes, controlled by the longitudinal component σ in 0 Eq. (1), can be understood by the fact that such current-carrying extended states mustpassthroughmanysaddle-pointsonthedisorderlandscape.However,thedrift velocityassociatedtotheguidingcenteridenticallyvanishesatthesepoints,sothat havingafiniteσ isessentialtoconnectthedifferentvalleysofthepotentialprofile. 0 The technical difficulty lies in evaluating the macroscopic conductivity in the limit where σ is much smaller than the amplitude variations of the Hall component 0 σ (r), but yet does not fully vanish. This regime cannot simply be accessed from H the σ 0 limit, because the transport equation becomes singular. The strategy 0 → developped in Ref. 14 and Sec. 3 will be to extrapolate from high orders of the perturbatively controlled σ expansion to the case of small dissipation. 0 →∞ Assuming that a critical state is established in the small σ limit due to the 0 scale invariant nature of the percolation backbone, one can infer from dimensional analysis that the macroscopic longitudinal conductivity scales as11,14: σxx ∝σ01−κ[(cid:10)σH2 (cid:11)−(cid:10)σH(cid:11)2]κ/2, (2) where κ is a non-trivial exponent previously conjectured11,12,13,15 to be κ=10/13, see Sec. 3 for a diagrammatic approach to this result. Based on simple microscopic argumentsforthelocalHallconductivityσ (r)thatweintroducenow,itispossible H tounderstandfromEq.(2)varioustransportregimesthatarerelevantforquantum Hall systems. In all what follows, we will assume that electron-phonon processes dominate in the longitudinal component19, leading to the temperature dependence σ (T) T. 0 ∝ 2.3. A hierarchy of transport crossovers 2.3.1. Fully classical regime: (cid:126)ω T c (cid:28) Attemperatureshigherthanthecyclotronenergy,bothcyclotronanddriftmotions are classical, so that the classical Hall’s law prevails: σ (r)=(e/B)n(r), with n(r) H the local electronic density, which undergoes smooth spatial fluctuations in case of high mobility samples11. In relatively clean samples, the amplitude v of disorder fluctuations remain small compared to the classical cyclotron energy, so that the Hall conductivity follows at first order the spatial variations of the local potential: en σ (r)= +AV(r). (3) H B withnthetotalelectrondensityandAaconstanttobedeterminedbelow.Thus,for a Gaussian distributed disorder, the local Hall conductivity displays Gaussian fluc- tuations and is weakly dependent of temperature. Using the percolation Ansatz (2) 5 for the macroscopic longitudinal conductivity, we find: σ vκT1 κ T3/13, (4) xx − ∝ ∝ which shows already a first non-trivial behavior in temperature12,14 connected to classical percolation, where the conductivity mildly decreases as temperature is lowered, see also Fig. 1. 2.3.2. Formation of Landau levels: v T (cid:126)ω c (cid:28) (cid:28) Astemperaturecrossesthecyclotronenergy,Landaulevelsstarttoemerge,andthe e2 (cid:88)∞ local density is given by Pauli’s principle: σ (r)= n [E V(r) µ] with H F m h − − m=0 µ the chemical potential and n (E) = 1/(eE/T +1) the Fermi-Dirac distribution F (we set Boltzmann’s constant k =1 in what follows). We will neglect spin effects B for simplicity in what follows (Landau levels are assumed spin non-degenerate). In the considered temperature range v T, the Fermi distribution can be linearized, (cid:28) which leads to Eq. (3) with A = (e2/h)((cid:126)ω ) 1 in the case (cid:126)ω T considered c − c (cid:28) previously, and more generally to: en e2 (cid:88)∞ σ (r)= + n (E µ)V(r). (5) H B h (cid:48)F m− m=0 The local conductivity remains Gaussian, but acquires now an extra temperature dependence from the Fermi function, which can be illustrated in the case of the plateau ν ν+1 transition, which leads for T (cid:126)ω to c → (cid:28) en e2 1 σpeak(r)= + V(r). (6) H B h 4T Using the percolation Ansatz (2), we find in the considered temperature range: (cid:16)v(cid:17)κ 1 1 σpeak T1 κ , (7) xx ∝ T − ∝ T2κ 1 ∝ T7/13 − so that the peak longitudinal conductivity strongly increases below T (cid:46)ω /4 (this c crossoverscale,aswellasthecompletescalingfunctionwillbedeterminedinSec.3), see also Fig. 1. 2.3.3. Two-fluids regime: (l /ξ)v T v B (cid:28) (cid:28) The peak longitudinal conductivity cannot diverge at vanishing temperature, and thelaw(7)mustbecut-offbyadditionalphysicalprocesses.Indeed,byfurtherlow- ering the temperature, the Fermi distributions becomes sharp at the scale T v, (cid:28) and the local Hall conductivity σ (r) now assumes rapid spatial variations be- H tween quantized values νe2/h and (ν +1)e2/h, with ν the filling factor. The local conductivity model now reads e2 e2 σ (r)= ν+ Θ[V(r)+µ E ], (8) H ν h h − 6 introducing the step function Θ. The transport properties of this two-fluids model were considered extensively in previous works20. It was found using duality argu- ments that the local conductivity Eq. (8) leads in the (unphysical) limit of zero temperature to an exact value for the longitudinal peak conductivity in the σ 0 0 → limit: σpeak = e2/h. This result would seem at first sight at odds with the scaling xx Ansatz(2),whichpredictsapowerlawvanishingofσ atsmallσ .Onmathemati- xx 0 calgrounds,themodelEq.(8)isquitepeculiarinthesensethatthefluctuationsof theHallconductivity[(cid:10)σ2 (cid:11) (cid:10)σ (cid:11)2]areactuallydivergingatthepeakvalue,inval- H − H idating the Ansatz. Yet, the existence of a finite and universal value σpeak = e2/h xx seemsstillphysicallysurprisingfromtheargumentationgiveninSec.2.2,wherewe argued that the fully opened current lines at σ have a vanishing drift velocity at 0 the saddle points of disorder. However, for such bimodal distribution of the local Hall conductivity Eq. (8) and in contrast to any continuous conductivity distribu- tion, the drift velocity does not vanish anymore at the saddle points, which allows to establish a macroscopic current even in the absence of dissipation mechanisms. This simple argument allows to understand why the percolation scaling Ansatz (2) does not apply to the two-fluids model of Dykhne and Ruzin20. However, we will see below that other processes invalidate the model Eq. (8) in the limit of zero temperature. Mreover, we can also infer how the “exact” value e2/h is approached from above. Indeed, the sharp Fermi function in Eq. (8) is always smeared on the scale T, recovering a continuous (but strongly non-Gaussian) distribution, leading likely to power-law deviations from the exact zero-temperature result e2/h: e2 σpeak = BTα (9) xx h − with a new critical exponent α > 0 that is to our knowledge still unknown, and B some constant. The fact that the peak longitudinal conductivity levels off at low temperatures towards values close (but not strictly equal) to e2/h has been noted from experimental data20, see also Fig. 1. 2.3.4. Wavefunction corrections: (l /ξ)2v T (l /ξ)v B B (cid:28) (cid:28) The low-temperature two-fluids conductivity model Eq. (8) relies on the high mag- netic field limit, and is strictly speaking only correct in the limit l 0. However, B → quantum corrections will occur for finite l /ξ due to the fact that the electronic B wavefunctions are not infinitely sharp transverse to the guiding center motion, but rather spread on the scale of the magnetic length l . For this reason, the local Hall B conductivity σ (r) will not undergo infinitely sharp steps from a quantized value xx to the next as in Eq. (8), but rather rapid but smooth rises on the scale l , see B Refs.17, 18. Because the wavefunctions extend transversely in a Gaussian manner, the resulting form of the Hall conductivity is easily understood (here for the lowest Landau level): e2 e2 (cid:90) d2R σH(r)= h + h πl2 Θ[V(R)−µ−Eν]e−(r−R)2/lB2. (10) B 7 Clearly,thetwo-fluidmodelEq.(8)isrecoveredinthelimitl /ξ 0,butforamore B → realistic smooth disorder, the correlation length ξ does not exceed a few hundreds of nanometers. In that case, the sharp step in Eq. (8) is smoothened whenever the new energy scale (l /ξ)v sets in. Interestingly, we recover now a continuous B conductivity distribution where the percolation Ansatz (2) should apply. Because the spatial fluctuations of the local Hall conductivity are no more controlled by temperature, we can infer without detailed calculation the following powerlaw for the peak longitudinal conductivity: σpeak T1 κ T3/13. (11) xx ∝ − ∝ Thusthepeakconductivityshoulddecreaseagainbycoolingthesampletoverylow temperatures, as evidenced experimentally8, see also Fig. 1. 2.3.5. Onset of quantum tunneling: T (l /ξ)2v B (cid:28) By further cooling towards the limit of zero temperature, a new energy scale (l /ξ)2v emerges,associatedtoquantumtunnelingatthesaddlepoints18.Forhigh B mobility samples, one can assume that transport remains incoherent between the widely separated saddle points, so that quantum interference effects can be ne- glected,andthelocalconductivitymodelEq.(1)stillapplies(ifnot,non-localeffects in the spirit of Ref. 21 must be accounted for). Here, the precise form of the local conductivitytensorisnotyetfullyunderstood,althoughaquasilocalapproachthat incorporates quantum tunneling can be developped18. For this reason, the precise scaling form of the peak longitudinal conductivity is still unknown in this regime, althoughaslowerdecreasethanEq.(11),leadingtoakinkatT =(l /ξ)2v,canbe B expected, due to the onset of the quantum processes allowing to transfer electrons above the saddle points: σpeak Tβ 0<β <3/13. (12) xx ∝ Such behavior was also observed experimentally in low temperature studies of the peak longitudinal conductivity8, see also Fig. 1. 3. Diagrammatic approach to classical percolating transport 3.1. Systematic weak coupling expansion and extrapolation to the percolation regime Our goal in this section is to discuss how classical percolation features in quantum Halltransportcanbecapturedanalyticallybyadiagrammaticapproach14,allowing torecoverthepercolationAnsatz(2)andaccurateestimatesofthecriticalexponent κ discussed in Sec. 2.3. Building on earlier works22,23 for the case of the local conductivity tensor Eq. (1), one can show by standard techniques that the disorder averaged longitudinal macroscopic conductivity reads (cid:18) (cid:19) (cid:18) (cid:10) (cid:11)(cid:19) σxx −σxy = (cid:10)σ0(cid:11)− σH +(cid:10)χˆ(r)(cid:11), (13) σ σ σ σ xy xx H 0 8 χˆ(r) = + + +... h i r r1 r r1 r2 r3 r r1 r2 r3 Fig. 2. Diagrammatic expansion in the case of Gaussian fluctuations of the local conductivity. Wigglylinesareassociatedtodisorderaverages,andsolidlinestotheGreen’sfunctionEq.(15) where χˆ(r) obeys the equation of motion: (cid:90) χˆ(r)=δσ(r)(cid:15)ˆ+δσ(r) d2r (cid:15)ˆˆ (r r)χˆ(r). (14) (cid:48) 0 (cid:48) (cid:48) G − (cid:10) (cid:11) WehaveintroducedabovetheHallconductivityfluctuationsδσ(r) σ (r) σ , H H ≡ − the antisymmetric 2 2 tensor (cid:15)ˆ, and the Green’s function: × ∂ ∂ (cid:90) d2p eipr [ˆ ] (r)= · . (15) G0 ij ∂r ∂r (2π)2σ p2+0+ i j 0 | | In previous analyses of Eq. (14), several methods were proposed, such as a (cid:10) (cid:11) mean-fieldtreatment23,lowestorderperturbationtheory24 inpowersof [δσ]2 /σ2, 0 or self-consistent Born approximation22. Clearly these approaches are insufficient to capture the critical percolation behavior in the strong coupling limit σ 0. 0 → However, the small dissipation Ansatz (2) ressembles the critical behavior typical of phase transitions, and leads hope that Pad´e extrapolation techniques of a suf- ficiently high order perturbative calculation could bridge the gap from weak (i.e. (cid:10) (cid:11) (cid:10) (cid:11) [δσ]2 σ2) to strong coupling (i.e. [δσ]2 σ2). The calculation actually sim- (cid:28) 0 (cid:29) 0 plifies for the case of Gaussian fluctuations of the local Hall conductivity δσ(r), whichappliestothehighesttemperatureregimesconsideredinEq.(3)andEq.(5). By symmetry considerations, one finds that the Hall component is not affected in the high temperature regime, namely classical Hall’s law σ = en/B holds. By xy dimensional analysis, the longitudinal conductivity reads: (cid:88)∞ δσ2 n σ =σ + a (cid:104) (cid:105) (16) xx 0 n=1 n σ02n−1 with dimensionless coefficients a collecting all diagrams of order n in perturba- n tiontheoryin δσ2 /σ2.Thelongitudinalconductivityσ thusreceivesnon-trivial (cid:104) (cid:105) 0 xx corrections that will lead to percolation effects in the limit σ 0. 0 → The methodology to compute the large σ expansion relies in iterating Eq. (14) 0 tothedesiredorder,averagingoverdisorderowingtotherelation(5),andevaluating theresultingmultidimensionalintegral,eitheranalyticallyornumerically,seeFig.2. In order to simplify the calculations, we considered spatial correlations of disorder of the form δσ(r)δσ(r) = δσ2 e r r(cid:48)2/ξ2, with correlation length ξ, allowing (cid:48) −| − | (cid:104) (cid:105) (cid:104) (cid:105) us to compute the series Eq. (16) up to sixth loop order14, see Table 1. Standard extrapolation techniques allow us to extract14 the estimate κ = 0.767 0.002 for ± thecriticalexponentappearinginEq.(2),quiteclosetothepreviouslyconjectured value11,12,13,15 κ=10/13 0.769. (cid:39) 9 Table1. Coefficientsan oftheperturba- tiveseries(16)uptosixthlooporder. Order Method Coefficientan 1 Analytical 1 2 2 Analytical 1 − 1log(2) 8 2 3 Analytical 0.2034560502 4 Numerical −0.265±0.001 5 Numerical 0.405±0.001 6 Numerical −0.694±0.001 3.2. High temperature crossover function for σpeak xx We finally provide a simple scaling function describing the crossover from the high temperature regime above the cyclotron energy T (cid:126)ω to the intermediate sit- c (cid:29) uation v T (cid:126)ω , where Gaussian fluctuation of the local Hall conductivity c (cid:28) (cid:28) stillarise,seeEq.(5).Fromthisexpression,wecanconnectthetypicalfluctuations (cid:113) (cid:10) (cid:11) of the Hall conductivity to the width v = [V(r)]2 of the disorder distribu- (cid:12) (cid:12) tion: (cid:113)(cid:10)[δσ(r)]2(cid:11) = e2v(cid:12)(cid:12)(cid:88)∞ n (E µ)(cid:12)(cid:12), so that the high temperature crossover h (cid:12)(cid:12) (cid:48)F m− (cid:12)(cid:12) m=0 function reads from the scaling Ansatz (2): (cid:12) (cid:12)κ (cid:12)e2 (cid:88)∞ (cid:12) σxx =σ01−κ(cid:12)(cid:12)(cid:12)hv n(cid:48)F(Em−µ)(cid:12)(cid:12)(cid:12) . (17) m=0 NotethatfortheGaussianmodelstudiedhere,adimensionlessprefactorinEq.(17) happens14 to be quite close to 1, and has not been written. At temperatures such thatv T (cid:126)ω andattheν ν+1plateautransition,i.e.forµ=(cid:126)ω (ν+1/2), c c (cid:28) (cid:28) → we thus find: σxpxeak =σ01−κ(cid:12)(cid:12)(cid:12)(cid:12)eh24vT(cid:12)(cid:12)(cid:12)(cid:12)κ, (18) recoveringexpression(7)forthelongitudinalconductivityinthelimitv T (cid:126)ω . c (cid:28) (cid:28) Wecanalternativelyre-expressthesumoverLandaulevelsinEq.(17)byusing Poisson summation formula25 in the limit T <µ, giving: σxx =σ01−κeh2(cid:126)vωc(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)1+(cid:88)+∞(−1)lcos(cid:18)2(cid:126)πωlcµ(cid:19)sinh4(cid:16)π2(cid:126)2lωπkc2BlkTBT(cid:17)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)κ (19) l=1 (cid:126)ωc vindicating expression (4) for the peak longitudinal conductivity in the T (cid:126)ω c (cid:29) limit. Either Eq. (17) or Eq. (19) can be used to extract the critical exponent κ from experimental data in the range of temperatures near the cyclotron energy. 4. Perspectives As a conclusion, we list several issues that could be addressed in further develop- ments of the present work. 10 What is the magnetic field behavior of σ at high temperature? xx • Can one extract reliably the classical exponent κ from experiments? • Are the exponents α and β of the low temperature regime related to κ? • How do the finite probe currents affect the Hall plateau quantization? • What are the fundamental differences between 2DEGs and graphene26? • Is a more realistic description of electron-phonon conductivity19 needed? • CanonedescribethecrossovertoDrudebehavioratlowmagneticfields27? • Can one implement transport calculations using diagrammatic QMC 28? • Acknowledgments We thank S. Bera, A. Freyn, B. Piot, W. Poirier, M. E. Raikh, V. Renard and F. Schoepferforstimulatingdiscussions,andANR“Metrograph”forfinancialsupport. References 1. K. Von Klitzing, G. Dorda, and M. Pepper, Phys. Rev. Lett. 45, 494 (1980). 2. R.E.PrangeandS.M.Girvin,TheQuantumHallEffect,(Springer,NewYork,1987). 3. M. Janssen, O. Viehweger, U. Fastenrath, and J. Hadju, Introduction to the Theory of the Integer Quantum Hall Effect (VCH, Germany, 1994). 4. J. Matthews and M. E. Cage, J. Res. Natl. Inst. Stand. Technol. 110, 497 (2005). 5. B. Huckestein, Rev. Mod. Phys. 67, 357 (1995). 6. B. Kramer, T. Ohtsuki and S. Kettemann, Phys. Rep. 417, 211 (2005). 7. F. Evers and A. D. Mirlin, Rev. Mod. Phys. 80, 1355 (2008). 8. H. P. Wei et al., Phys. Rev. B 45, 3926 (1992). 9. W. Li et al., Phys. Rev. B 81, 033305 (2010). 10. K. Hashimoto et al., Phys. Rev. Lett. 101, 256802 (2008). 11. S. H. Simon and B. I. Halperin, Phys. Rev. Lett. 73, 3278 (1994). 12. D. G. Polyakov and B. I. Shklovskii, Phys. Rev. Lett. 74, 150 (1995). 13. M. M. Fogler and B. I. Shklovskii, Sol. State Comm. 94, 503 (1995). 14. M. Flo¨ser, S. Florens, and T. Champel, Phys. Rev. Lett. 107, 176806 (2011). 15. M. B. Isichenko, Rev. Mod. Phys. 64, 961 (1992). 16. M. R. Geller and G. Vignale, Phys. Rev. B 50, 11714 (1994). 17. T. Champel, S. Florens and L. Canet, Phys. Rev. B 38, 125302 (2008). 18. T. Champel and S. Florens, Phys. Rev. B 80, 125322 (2009). 19. H. L. Zhao and S. Feng, Phys. Rev. Lett. 70, 4134 (1993). 20. A. M. Dykhne and I. M. Ruzin, Phys. Rev. B 50, 2369 (1994). 21. H. U. Baranger and A. D. Stone, Phys. Rev. B 40, 8169 (1989). 22. Y. A. Dreizin and A. M. Dykhne, Sov. Phys. JETP 36, 127 (1972). 23. D. Stroud, Phys. Rev. B 12, 3368 (1975). 24. C. Timm, M. E. Raikh and F. von Oppen, Phys. Rev. Lett. 94, 036602 (2005). 25. T. Champel and V. P. Mineev, Philos Mag. B 81, 55 (2001). 26. T. Champel and S. Florens, Physical Review B 82, 045421 (2010). 27. D.G.Polyakov,F.Evers,A.D.MirlinandP.Wo¨lfle,Phys.Rev.B64,205306(2001). 28. E. Gull et al., Rev. Mod. Phys. 83, 349 (2011).