Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Bra–ket notation wikipedia , lookup
Quantum state wikipedia , lookup
Magnetic monopole wikipedia , lookup
Density matrix wikipedia , lookup
Ferromagnetism wikipedia , lookup
Aharonov–Bohm effect wikipedia , lookup
Symmetry in quantum mechanics wikipedia , lookup
12 The Integer Quantum Hall E↵ect The Hall e↵ect has played already an interesting role in classical electrodynamics, since it provides a unique tool for determining the sign of the current carriers. This task is challenging since the electric current produced by particles of charge q moving at velocity v, is the same as that produced by particles of the opposite charge ( q) moving at the opposite direction (at the same speed and density of the moving charges). The two cases can however be distinguished by the sign of the Hall resistance which is measured in the presence of the combination of crossed electric (E) and magnetic (B) fields. The Lorentz force on moving charges F = qv ⇥ B deflects the current in a transverse direction, e.g. towards a lateral edge of the conductor in which the current flows. The charge accumulation there results in a transversal voltage di↵erence between the lateral edges of the conductor, which is named VHall after the discoverer of this e↵ect. In the steady state the corresponding electric field counters the transversal drift, that is: q v ⇥ B + q EHall = 0 . (12.1) Since from macroscopic-scale measurements one can determine the direction of qv as well as of EHall , the above relation allows to deduce the sign of the current carriers’ charge q. The Hall, i.e., transversal resistance is: ⇢H = VHall Ichannel (12.2) According to the classical explanation of electrical conduction, at given density of the current carriers the Hall resistance would be proportional to the applied magnetic field (as is implied by (12.2) and the Drude picture of electrical conductance). 141 142 The Integer Quantum Hall E↵ect In their 1975 paper on a quantum calculation of the Hall conductance, T. Ando, Y. Matsumoto and Y. Uemura [21] noted (italics added here): • “[The Green function G N or XN is determined by the self-consistency equa(1) tion (2.8) of II.] From the above equation, one can conclude that XY van2 ishes and XY becomes nec/H = e (N + 1)/2⇡~, when the Fermi level lies in energy gaps between adjacent N-th and N + 1-th Landau levels at zero temperature. ...” Figure 12.1 A sketch of the Hall resistance (⇢ x,y ), plotted along with the direct resistance (⇢ x,x ), for two dimensional samples at low temperature. The Hall conductance exhibits plateaux of integer multiples of e2 /h as a function of the perpendicular magnetic field. The direct conductivity ( xx = ⇢ x,x /(⇢2x,x +⇢2x,y )), which is shown here in unrelated units, vanishes over the plateaux. The significance of this point was realized in the 1980 experimental work K. von Klitzing who, working with samples provided by G. Dorda and M. Pepper, turned the quantization of the Hall conductance (within its plateaux) into a tool for a very precise determination of e2 /h. This constant’s value was reportedly determined by such means up to the relative precision of 10 9 . Within five years the work was awarded the Nobel prize. More surprises were in the store, in particular the discovery of the Fractional Quantum Hall E↵ect [242], i.e. observation of plateaux with Hall conductance at fractional multiples of e2 /h. We shall however not get to this subject here. The successful observation of IQHE required to work at low temperatures 12.1 Laughlin’s charge pump 143 and high magnetic fields but not necessarily with ‘clean’ samples.1 In fact, in was noted that the e↵ect is improved by disorder (in the form of impurities in the sample). Soon after the experiments have focused attention of the IQHE phenomenon theorists started to grasp the beautiful mathematical structure behind it. In fact, the quantization has topological aspect. Its robust explanation combines functional analysis, topology (in the space of operators), and probability. In particular, it was understood that in the absence of direct conductance the quantum Hall conductance of two dimensional systems can be expressed both as a Chern number and as a Fredholm index. The precision of integers is protected by their topological nature! The e↵ect provides also an example of the interesting phenomenon that adiabatic transport [21] tends to be quantized in situations which are enabled by the vanishing of the direct conductivity. 12.1 Laughlin’s charge pump There have by now been a number of di↵erent derivations of the IQHE. Let us present here a slightly formal one, which fits neatly with the general setup of random operators discussed in these lecture notes. The argument starts from Laughlin’s observation that Hall conductance can be probed in a charge pump Gedankenexperiment [172] which is outlined in Figure 12.1. Induced EMF ⊗ Hall current B Increasing magnetic flux Φ (up to sign) wire leads Figure 12.2 Laughlin’s charge pump, based on the Hall e↵ect (drawing based on [27]): A conducting sheet is wrapped into a cylinder bounded by a pair of rings connected to lead wires. In the presence of non vanishing transversal magnetic field B, increase in the enclosed magnetic flux results in charge transport between the rings which form the cylinder boundary. Deforming this picture, the cylinder can be stretched and flattened to form a ’Corbino disk’ whose boundaries consist of a pair of concentric rings. Thus, 1 The temperature range has since been brought up, in particular working with graphene - the subject of another Nobel prize. Random Operators c M. Aizenman & S. Warzel DRAFT 144 The Integer Quantum Hall E↵ect one considers a system in which the charges are confined to a plane and the magnetic field is changed through an adiabatic process which results in an increase of the flux through the smaller ring. As described by the Faraday law, responding to changes in the magnetic flux is a time-dependent electric field E whose integral along a loop C is: I d C E · dl = . (12.3) dt C where C is the change in the flux enclosed by C. By Lenz’ law, the current which would result from this field would tend to diminish the rate of change of the enclosed flux. However, we are now interested in the transversal current. The latter may be present if in addition to the above electromotive force there is also a transversal magnetic field throughout the lattice (B , 0). The induced current density is ! D H j= E. (12.4) H D where D and H (the direct and the Hall conductance) are elements of the bulk homogenized conductivity tensor within the plane. Correspondingly, the rate of the induced charge transport across the contour C is I dQ = j · n d` dt C I I = D E · n d` + H E · dl (12.5) The last integral on the right-hand side of (12.5) is the induced electromotive force and tied by Lenz’s law (12.3) to the flux change rate. The first term vanishes in situations in which the direct conductivity D is zero. To keep this condition, the process will be carried out adiabatically. Under this condition, integrating (12.5) and (12.3) over time one obtains an expression for the Hall conductance as the ratio of the transported charge to the flux change: H = Q B . (12.6) 12.2 The magnetic Hamiltonian Let us recall from (3.7) that a natural way to incorporate the magnetic field into the Hamiltonian is to replace the regular Laplacian by its magnetic version, i.e. 12.3 Charge transport as an index 145 consider (A) H = with ( (A) )(x) = X ⇣ + V(!) e iqA(x,y) (y) |x y|=1 (12.7) ⌘ (x) (12.8) where A(x, y) = A(y, x) |x y|,1 corresponds to the vector potential integrated over the edge from x to y. The magnetic flux through the lattice plaquettes (P, which carry orientation) is given by the discrete version of the relation B = Curl A, i.e. X BP = A(b) (12.9) b2@P where b is summed over the oriented edges which form the boundary of P. Of relevance to the discussion which follows is the observation that the insertion of a unit flux (measured in units h/q) at a point a 2 R2 \Z2 in the plane corresponds to a gauge transformation which is given by the unitary mapping: (Ua ) (x) := e i✓a (x) (x) , where ✓a (x) := Arg(x a) 2 ( ⇡, ⇡] is the argument of the vector x the latter being identified with C. (12.10) a 2 R2 , 12.3 Charge transport as an index ΔΦ ΔQ Figure 12.3 The Avron-Seiler-Simon Hilbert hotel based on the Hall e↵ect Y. Avron, R. Seiler and B. Simon [26] stretched the Corbino disk picture even further, and considered the setup in which the flux is inserted through one of the plaquettes of a regular lattice and the outer ring is taken out to Random Operators c M. Aizenman & S. Warzel DRAFT 146 The Integer Quantum Hall E↵ect infinity (at least on the lattice scale). Furthermore, they in e↵ect propose that the lattice itself can serve as a charge reservoir and instead of looking at the charge flow through the system one could try to make sense of the change in the total charge within the infinite lattice due to a finite increase in the magnetic flux. We now turn to a more explicit review of this approach to the IQHE, assuming now that the system’s valance electrons (which have charge q) can be treated as a Fermi gas at zero temperature. Under the assumption that the Fermi energy falls within a spectral gap, or, as emphasized by J. Bellisard [32], within a ’mobility gap’, i.e. a regime of Anderson localization, one may expect that under an adiabatic flux insertion, the system’s state would evolve from the Fermi projection PEF = 1[H E F ] (12.11) to the corresponding projection for the modified Hamiltonian. The final state is particularly easy to write explicitly if the flux increment is h/e (or an integer multiple of it) since in that case the final state is given by Ua PE F Ua† with Ua the gauge transformation (12.10). Thus, it is tempting to argue that ⇣ ⌘ Q = q tr Ua PE F Ua† PE F , (12.12) to the extent that the di↵erence makes sense. The latter question is not quite trivial: i) if the di↵erence is calculated by evaluating the traces separately it becomes (1 1) which is ill defined, and ii) had the di↵erence of the two projection operators been trace class the difference trace would vanish since locally the densities are equal: , Ua PE F Ua† x i . (12.13) ⌘ As it turns out, the di↵erence Ua PE F Ua† PE F is not trace class, so the above trivial answer does not apply. However, if E F is in the localization regime the di↵erence is a compact operator, and thus instead of (12.12) one may take the index which is defined as follows. h x , PE F xi ⇣ = h x Definition 12.1 The index of two orthogonal projections P, Q on a Hilbert space H whose di↵erence P Q is a compact operator is Index(P, Q) := dim ker(P Q 1) dim ker(Q P 1) . (12.14) It may be added that the dimensions are well defined since the assumed compactness of P Q ensures that dim ker(P Q ± 1) < 1. In the special case where P and Q are trace class the index is given by (12.12), that is, it equals the di↵erence of the dimensions of PH and QH. A more generally useful formula is (12.17), which is explained within the proof of Proposition 12.2. 12.4 A calculable expression for the index 147 In this terminology, the tentative proposal (12.12) is revised to: ⇣ ⌘ Q = q Index Ua PE F Ua† , PE F , (12.15) When defined, this index takes integer values, and through (12.6) it yields H 2 (q2 /h) Z. Furthermore, the value which it yields for the Hall conductance H coincides with that provided by the Kubo formula, as was shown in [26, 32], and is explained below following [8]. 12.4 A calculable expression for the index It may be worth stressing that the notion of index comes with certain subtleties which are actually of relevance in the present context: 1. When defined, the index takes only integer values, and it is robust in its independence of the flux insertion point, and by implication also independent of the disorder (in the ’almost sure’ sense). 2. However, the relative index of two projections Index(P, Q) is defined only under some strict In ⇣ conditions. ⌘ the case of interest here, a necessary condi† tion for Index Ua PE F Ua , PE F to make sense is that the Fermi projection P has rapidly decaying matrix elements. For that, localization plays an essential enabling role. This caveat well reflects also the observed physics, with the Hall conductance changing over regimes where it is not quantized. 3. Formal manipulations, in which questions of convergence are ignored, could be misleading. Splitting traces or invoking cyclicity of the trace without paying attention to trace-class restrictions (as is allowed in finite dimensions) may easily read to contradictions, or to the wrong impression that the index is always zero. From the index definition one may conclude the following elementary properties: Index(P, Q) = Index(Q, P) = Index(P? , Q? ) = Index UPU 1 , UQU 1 , (12.16) where P? := 1 P and Q? := 1 Q stand for the projections onto the orthogonal complement and U : H ! H is any invertible linear map. In computing the index we shall rely on: Proposition 12.2 Let P, Q be a orthogonal projects on a Hilbert space H. Random Operators c M. Aizenman & S. Warzel DRAFT 148 The Integer Quantum Hall E↵ect Then for any n with which (P Q)2n+1 is trace class one has Index(P, Q) = tr (P Q)2n+1 . (12.17) Proof The strange independence from n of the the value in (12.17) is explained by the observation that the spectrum of P Q is antisymmetric, counting multiplicity, except possibly for the values 1 and ( 1). This is established in [26] by considering the pair of anticommuting operators C := P Q and S := P Q? , which are easily seen to satisfy: S 2 + C2 = 1 and S C + CS = 0 . (12.18) For any eigenvalue 2 (C)\{0} and a corresponding (normalized) eigenfunction ' 2 H one has: CS ' = S C' = S', (12.19) i.e., unless S ' = 0, the functions S ' is an eigenfunction corresponding to the reflected eigenvalue. In the exceptional case S ' = 0 implies 2 ' = C 2 ' = ' S 2 ' = ' and hence 2 {±1}. Since with the exception of the eigenvalues ±1 the spectrum of C (counted with its multiplicity) is invariant under reflections, at any odd power at which the eigenvalues are summable one gets cancellations, and the contribution left standing refers to just the term which are counted in the definition of the index. Thus: X 2n+1 tr C 2n+1 = dim ker(C ) = dim ker(C 1) dim ker(C + 1) 2 (C) = Index(P, Q) . (12.20) ⇤ Guided by the criterion provided⇣ by Proposition ⌘12.2, one would want to know under what conditions is T = Ua PE F Ua† PE F a compact operator, and more precisely under what conditions and for what powers p > 0 is kT k p := (tr |T | p )1/p < 1 ? (12.21) In this case T is said to be in the Schatten-p class. The following lemma spells a sufficient condition for that. Lemma 12.3 Let T : `2 (Zd ) ! `2 (Zd ) be a bounded linear operator with kernel T (x, y) = h x , T y i. Then for any p 2 [1, 1): X⇣X ⌘ 1p kT k p (12.22) |T (x, y)| p . y2Zd x2Zd 12.4 A calculable expression for the index 149 P Proof We rewrite T = ↵2Zd T ↵ in terms of the operators defined by: T ↵ (x, y) := T (x, y) x ↵,y . The triangle inequality yields X X kT k p kT ↵ k p = kT ↵† T ↵ k1/2 p/2 ↵2Zd = X⇣X ↵2Zd x2Zd ↵2Zd |T (x + ↵, x)| p ⌘ 1p (12.23) . ⇤ To answer the question posed in (12.21) let us note that ⇣ h x , Ua PE F Ua† y i PE F (x, y) = h x , PE F y i ei[✓a (y) ✓a (x)] ⌘ 1 . (12.24) Assuming the o↵-diagonal matrix elements h x , PE F y i decay rapidly, but terms with bounded distance |x y| do not, the factor involving |ei[✓a (y) ✓a (x)] 1| = 2| sin( 12 \(x, a, y))| yields decay at the rate of O(1/|x|). Thus, one is led to expect that in the regime of exponential localization the condition (12.21) would be met in the two dimensional system considered here for all k > 2. In ⇣ ⌘3 particular, the index should be computable through tr Ua PE F Ua† PE F . In addition to the above estimates we shall find useful also the following property of the unitary gauge transformations Ua . Lemma 12.4 The unitary transformations Ua : `2 (Z2 ) ! `2 (Z2 ), defined by the gauge transformation (Ua )(x) = e i✓a (x) (x), with a 2 R2 \Z2 , satisfy: 1. The operator di↵erences Ub Ua are in the Schatten-3 class, and furthermore for all a, b 2 R2 \Z2 with |a b| 1: kUb Ua k3 := (tr |Ub Ua |3 )1/3 C |a b|1/3 , (12.25) with a uniform constant C < 1. 2. The unitary operators Ua transform covariantly, U x(A) † Ua U x(A) = Ua x , under the magnetic translations U x(A) , x 2 Z2 that are given by U x(A) (y) = e iBq(y1 x2 x1 y2 )/2 (y x). Proof For a proof of the first assertion we note that Ub Ua is diagonal in the basis of position eigenfunctions x , and thus X 3 kUb Ua k33 = e i✓a (x) e i✓b (x) . (12.26) x2Z2 By elementary geometric considerations: ( ) |b a| |ei✓a (x) ei✓b (x) | min 2, p , |x a||x b| Random Operators c M. Aizenman & S. Warzel (12.27) DRAFT 150 The Integer Quantum Hall E↵ect for all a, b 2 R2 and x 2 Z2 . Therefore, for min{|x a|, |x b|} |b a|/2 the summand in (12.26) is of the order O(1/|x|3 ), which is summable in d = 2 dimensions. The claimed bound (12.25) follows by summation over unit increments in |b a|. The second assertion is a direct consequence of the way the magnetic translations are set up. ⇤ 12.5 Evaluating the charge transport index under the mobility gap assumption The above preparatory statements enable the following general result which allows to conclude quantization of the quantum Hall conductance at T = 0 and Fermi energy E F within the regime of Anderson localization. Theorem 12.5 (Existence of charge transport index) Let P 2 Kmc be a covariant family of orthogonal projections on `2 (Z2 ) and assume that its kernel P(x, y) = h x , P y i satisfies: X x2Z2 ⇣ h i⌘1/3 |x| E |P(0, x)|3 < 1. (12.28) Then there is a set ⌦0 of full measure P(⌦0 ) = 1 such that for all ! 2 ⌦0 : 1. P(!) Ua P(!)Ua† is Schatten-3 class for all a 2 R2 \Z2 , so that the charge transport index Ca (P)(!) = Index(Ua P(!)Ua† , P(!)) (12.29) is well defined and takes integer values, 2. the value of Ca (P)(!) does not change with a 2 R2 \Z2 , and 3. it is almost surely equal to its mean value. Proof 1. For n 2 Z2 let Bn = {a 2 R2 : ka nk1 < 1}. Lemma 12.3 allows to 12.5 Evaluating the charge transport index under the mobility gap assumption 151 deduce the uniform estimate: 2 3 66 77 E 664 sup kP Ua PUa† k3 775 a2B0 \Z2 = X h ⇣X ⇣ E sup P(x + y, x) 1 y2Z2 a2B0 \Z2 X⇣X y2Z2 x2Z2 ei(✓a (x) ✓a (x+y)) x2Z2 h i E |P(x, x + y)|3 sup e i✓a (x+y) e a2B0 \Z2 X h i1 ⇣ X E |P(0, y)|3 3 sup e y2Z2 x2Z2 i✓a (x+y) a2B0 \Z2 ⌘ 3 ⌘ 13 i i✓a (x)) 3 i✓a (x)) 3 e ⌘ 13 ⌘ 13 (12.30) < 1 where the second step is by Jensen’s inequality. By an estimate similar to (12.27) the last term in the right side is shown to be bounded by C(1 + |y|). This proves the finiteness of the right side. As a consequence, there is a full measure set of ! for which P(!) Ua P(!)Ua† is Schatten-3 class for all a 2 B0 \Z2 . Let ⌦0 be the countable intersection of lattice shifts of this set. Then ⌦0 is still of full measure and for all ! 2 ⌦0 , the operator di↵erence P(!) Ua P(!)Ua† is Schatten-3 class for all a 2 R2 \Z2 . On this set, Index(Ua P(!)Ua† , P(!)) is therefore well defined and integer valued for all a 2 R2 \Z2 . 2. The independence of Ca (P)(!) from a is implied by its quantization and continuity in a over the connected domain R2 \Z2 . The continuity relies on the continuity of the Schatten-3 norm k·k3 = (tr |·|3 )1/3 of the gauge transformations which yields Ub P(!)Ub† d(a, b; !) := Ua P(!)Ua† (Ua 2 kUa Ub )P(!)Ua† 3 3 + Ub P(!)(Ua† Ub k3 C |a Ub† ) b|1/3 . 3 (12.31) Here the last inequality is by Lemma 12.4 and holds for |a b| 1 and almost all !. The proof of the continuity of the index is completed by the estimate tr(P(!) + 3 kP Ua P(!)Ua† )3 Ua P(!)Ua† k3 tr(P(!) 2 Ub P(!)Ub† )3 d(a, b; !)3 d(a, b; !) + 3 kP(!) Ua P(!)Ua† k23 (12.32) d(a, b; !) , which is a consequence of an elementary algebraic identity for the di↵erence of (P Ua PUa† )3 (P Ub PUb† )3 and Hölder’s inequality for Schatten norms. 3. The invariance of Ca (P) under lattice shifts of a, and the covariance of the Schrödinger operators under magnetic shifts, implies that as a function of ! Ca (P) is invariant under the ergodic action of this group. It is therefore almost surely constant, and hence also equal to its mean value. ⇤ Random Operators c M. Aizenman & S. Warzel DRAFT 152 The Integer Quantum Hall E↵ect 12.6 Correspondence with and quantization of the Streda-Kubo Hall conductance The next result will show that under the localization assumption (12.28) on the Fermi projection PE F , the explicit formula for the charge-transport index, C(PE F ) = Index(Ua PE F Ua† , PE F ) (12.33) with a 2 (Z2 )⇤ an arbitrary point on the dual lattice, agrees also with the Streda-Kubo formula (11.21) for Hall conductance (at zero temperature). That is somewhat reassuring since the justification of the above approach did include a certain creative ansatz. The relation can also be used to explain the Hall conductance plateaux. The following statement, in a slightly di↵erent form, dates back to [32]. The present version is based on [8]. Use is make here also of the continuity of the integrated density of states, which is known for all ergodic random operators of the subsequent form [70]. Theorem 12.6 (Quantization and plateaux of Hall conductance) Let H(!) = (A) + V(!) be a random operator on `2 (Z2 ) with a constant perpendicular magnetic field B , 0 and a bounded ergodic random potential V. Then over any value of the Fermi energy E F at which ⇠(E F ) := X x2Z2 |x| E h 0 , PE F (H) xi 3 1/3 < 1 (12.34) the zero-temperature Streda-Kubo Hall conductance 1,2 (E F ) defined by (11.21) coincides with the charge transport index described by Theorem 12.5 times q2 /2⇡ ⌘ q2 /h, i.e. almost surely: 1,2 (E F ) = q2 C(PE F ) . h (12.35) Furthermore, the value of 1,2 (E F ), which in this case is an integer multiple of e2 /h, stays constant as E F ranges over any interval I in which (12.34) holds with supE2I ⇠(E) < 1. Proof 1. In the setting of Theorem 12.5 with P = PE F one gets for the charge 12.6 Correspondence with and quantization of the Streda-Kubo Hall conductance 153 transport index: ⇣ ⌘3 C(P) := E tr Ua PUa† P X = E [P(x, u) P(u, v) P(v, x)] x,u,v2Z2 ⇣ ⌘⇣ ⇥ ei(✓a (u) ✓a (x)) 1 ei(✓a (v) X = 2i E [P(x, u) P(u, v) P(v, x)] x,u,v2Z2 = 2i X ✓a (u)) ⌘⇣ 1 ei(✓a (x) ✓a (v)) ⌘ 1 ⇥ sin \(u, a, x) + sin \(v, a, u) + sin \(x, a, v) E [P(0, u) P(u, v) P(v, 0)] u,v2Z2 a2(Z2 )⇤ ⇥ sin \(u, a, 0) + sin \(v, a, u) + sin \(0, a, v) . (12.36) Here \(u, a, x) = ✓a (x) ✓a (u) is the angle of the segment enclosed by the halflines anchoring in a and passing through u and x respectively, cf. Figure ??. In the last line, we have used the homogeneity which allows us to shift x to the origin and sum over a 2 (Z2 )⇤ instead. In the last expression the sum over a 2 Z2 can be carried out with the help of a striking formula of A. Connes [60] which is presented in Proposition 12.7 below. Applying it (with g(↵) = sin ↵) one has: X a2(Z2 )⇤ sin \(u, a, 0) + sin \(v, a, u) + sin \(0, a, v) = ⇡ u ^ v = ⇡(u2 v1 u1 v2 ) . (12.37) Therefore the charge transport index is given by C(P) = 2⇡i X u,v2Z2 = = E [P(0, u) P(u, v) P(v, 0)] u ^ v (12.38) 2⇡i E [h 0 , PX2 PX1 P 0 i h 0 , PX1 PX2 P 0 i] 2⇡ 2⇡i E [h 0 , P [[X2 , P] , [X1 , P]] P 0 i] = 2 1,2 , q where the last expression relates directly to the transversal conductance as given the Streda-Kubo formula (11.21). 2. The constancy of the Hall conductance over I under the uniformity assumption, follows from its quantization and continuity. To prove the latter, let E E 0 in I. Then applying the representation (12.38), an elementary telescopic decomposition, translation invariance and Hölder’s inequality with Random Operators c M. Aizenman & S. Warzel DRAFT 154 2q The Integer Quantum Hall E↵ect 1 +r 1 = 1, one gets: 1,2 (E) 6 1,2 (E X u,v2Z2 0 ) h E P# (0, u) q i1/q h q i1/q E P# (u, v) E [| P(v, 0)|]1/r |u| |u 0 12 BBB X h i1/q CCC q CCC E [| P(0, 0)|]1/r 6 BBB@ |u| E P# (0, u) A v| (12.39) u2Z 2 where P# is either P( 1,E] or P( 1,E 0 ] and P := P( 1,E] P( 1,E 0 ] = P(E 0 ,E] . For the last inequality also use was made of the Cauchy-Schwarz inequality in Hilbert space, | P(v, 0)|2 | P(0, 0)| | P(v, v)|, followed by the CauchySchwarz inequality for the expected value and translation invariance. By the continuity of the integrated density of states [70] the expected value E [| P(0, 0)|] tends to zero as E # E 0 . Thus, under the assumption that ⇠(E) is uniformly bounded over I the quantized Hall conductance is also continuous there. ⇤ 12.7 Appendix: Connes’ area formula In relating the charge-transport index to the Hall conductivity as given by the Kubo formula, a vital role is played by Connes’ area formula. Following is its formulation and derivation, using the streamlined argument of Colin de Verdière, cf. [25]. Proposition 12.7 For a fixed triplet x1 , x2 , x3 2 Z2 , let ↵ j (a) 2 ( ⇡, ⇡) be the angle of view from a 2 (Z2 )⇤ of x j+1 relative to x j in the positive angular orientation, with the index interpreted cyclicly and ↵ j (a) = 0 if a lies between them. Then the following equality holds 3 X X g(↵ j (a)) = 2⇡ Area(x1 , x2 , x3 ) (12.40) a2(Z2 )⇤ j=1 for any antisymmetric bounded function g : ( ⇡, ⇡) ! R which for small ↵ = 0 behaves as: g(↵) = ↵ + O(↵3 ) . with Area(·) the triangle’s oriented area. (12.41) 12.7 Appendix: Connes’ area formula 155 x 2 x 3 a x1 Figure 12.4 The three angles in the sum (12.40) Proof We may assume the triangle to be positively oriented. The statement (12.40) is true for g(↵) = ↵. Indeed, for each ↵ 2 (Z2 )⇤ : 8 9 > 1 > > > > > > < 1 > = ↵ j (a) = 2⇡ > > > > 2 > > > : 0 > ; j=1 3 X 8 > inside > > > < for a > on the boundary > > > : outside 9 > > > > = the triangle. > > > > ; Thus, for g(↵) = ↵ the left side of (12.40) is 2⇡ times the number of dual lattice sites inside within the triangle (counting a boundary site with weight 12 ). This number is the same for triangles obtained by the lattice translation and reflection symmetry operations. Since this set of triangles tiles the plane, the number of enclosed dual sites must equal the triangle’s area. The above observation reduces (12.40) to the statement that for f (↵) = g(↵) ↵ 3 X X f (↵ j (a)) = 0 . (12.42) a2(Z2 )⇤ j=1 A significant di↵erence between f and g is that the individual terms f (↵ j (a)) are summable in a 2 (Z2 )⇤ , since by assumption f (↵ j (a)) = O(|a| 3 ) for |a| ! P 1. This allows to split the sum into three terms: a2(Z2 )⇤ f (↵ j (a)), j = 1, 2, 3. Each term is antisymmetric with respect to a reflection which is a symmetry of the lattice Z2 . Explicitly: the reflection with respect to the midpoint of the corresponding edge, (x j+1 + x j+2 )/2 2 (Z/2)2 . Thus the individual sums (at given j) vanish, and hence (12.42) holds. ⇤ Random Operators c M. Aizenman & S. Warzel DRAFT 156 The Integer Quantum Hall E↵ect Exercises 12.1 Let P be an orthogonal projection and U be a unitary operator on some Hilbert space H and set Q := UPU † . Assume that P Q is Schatten2n + 1 class for some n 2 N0 . 1. Show that (P PQP)n+1 = (P Q)2n+2 P and similarly (Q QPQ)n+1 = (P Q)2n+2 Q and prove that Index(P, Q) = tr(P PQP)n+1 tr(Q QPQ)n+1 . 2. Consider F := PUP on the Hilbert space PH. Conclude that 1 F † F and 1 FF † are Schatten-n + 1 class on PH and that Index(Q, P) = tr(1 F † F)n+1 tr(1 FF † )n+1 , where the trace extends over PH. 3. Use the isospectrality of 1 F † F on (ker F)? with 1 FF † on (ker F † )? to prove Index(Q, P) = dim ker F dim ker F † := IndexFN (F) . [The right side defines the Fredholm-Noether index of F.] 12.2 To be expanded Notes We used the notations H = 1,2 = xy for the Hall conductance and ⇢H = ⇢1,2 = ⇢ x,y for the Hall resistance. The relation x,y = ⇢ x,y1 is valid only over the regime where the direct conductance vanishes D = xx = yy = 0, which coincides with the Hall plateaux. Transport quantization is present also in other situations of quantum charge pumps, which are driven adiabatically by a cyclic increase in some of a quantum system’s parameters, cf. [240]. In this more general context it is the analog of the Hall resistance, rather than conductance, which is expected to be quantized. Exercise 12.1 concerns the relation, which was introduced in [26], of the charge-transport index with the Fredholm-Noether index of the operator PUa P on the Hilbert space P`2 (Z2 ): Ca (P) = IndexFN (PUa P) . The independence of Ca (P) from a can be viewed as homotopy invariance within a norm-continuous family of operators of finite index, cf. [174]. This can also be deduced from the invariance of the Fredholm-Noether index under compact perturbations (since by Lemma 12.4 Ua Ub are compact operators).