24 Quantum Hall Effect

The quantum Hall effect is an astonishing macroscopic phenomenon that oc-

curs when electrons are confined to an approximately planar 2D sample, and
subject to a perpendicular mangetic field of large uniform strength b and low
temperature. When the classically-measured Hall conductance (actually, Hall
resistance, i.e., transverse voltage response to an applied current) is measured,
one discovers an astonishing quantization,

e2
σHall = νint · , when ν ≈ integer, (24.1)
h
where ν ∼ 1b is the R-valued “filling factor”, and νint is the nearest integer
to ν. The factors e and h are the elementary electron charge and Planck’s
constant, respectively. Thus σHall looks like a staircase function of 1b . The
flat steps (called “plateau”) are possibly the flattest curves ever measured
experimentally, and the integer nature of the steps’ values has been verified to
∼ 10−10 uncertainty since its original discovery by Klaus von Klitzing in 1980.
The 1985 Nobel Prize in Physics was awarded for this unexpected experimental
discovery.
Classical models of charged particles in magnetic fields fail to predict such
quantized conductances, so the above effect is called the Integer Quantum Hall
Effect (QHE). The use of quantum mechanical models to provide a mechanism
to perfectly “round off” σHall to the nearest integer, involves deep mathematics
and physics. Still, there remain many open problems, and one must be careful
not to use tautologous arguments (e.g. of the form integer=integer).

2
Quantum kilogram. The obvious interpretation of Eq. (24.1) is that eh is
an a priori known quantity, then we measure σHall and verify the hypothesis
2
that it is an integer multiple of eh . Planck’s constant h is presumed to be some
universal quantum of action,

h = 6.626 . . . × 10−34 · kg m2 s−1 , (24.2)

for some numerical factor to be experimentally determined with better and

better accuracy over time.
However, in the first place, “kg” was defined to the mass of some physical
prototype stored in Paris, and this mass was not completely stable over time.

262
So the “true” number 6.626 . . . does not make sense once the required preci-
sion exceeds the uncertainty in the old definition of “kg”. Consequently, the
statement (24.1) had an inherent fuzziness.
The QHE has breached this fuzziness threshold. Today, it is instead pre-
sumed that the quantization of σHall is exact (more precisely, approaches exact-
ness as the sample size is taken to infinity). So the QHE grants direct access to
e2
h
, via the fundamental “step” of Hall conductance. Combined with another
macroscopic quantum phenomenon — a quantum tunneling effect between two
superconducting materials, called the Josephson effect — which grants access
h
to 2e , one therefore has access to h (and also e).
In this way, h is actually much more precisely known than the old “kg”,
and it makes sense to redefine a “new kg” as follows. Agree on some fixed
numerical ratio in (24.2), based on the best currently known ratios (which
have some small uncertainty). In 2019, scientists committed to this idea, and
officially redefined the new “kg” via the equation

h = 6.62607015 × 10−34 · kg m2 s−1 .

24.1 Landau Hamiltonian on a surface

The QHE is quite an involved subject. We will only have time to investigate
a closely related spectral phenomenon in the quantum mechanics of charged
particles in magnetic fields, called Landau quantization.
On an oriented Riemannian 3-manifold, there is a Hodge duality between
1-forms and 2-forms. What is customarily called a magnetic vector field B is,
equivalently, a 1-form B♭ , Hodge dual to a 2-form B = ⋆B♭ . The 2-form B
is more fundamental, in the sense that it does not need an “external” choice
of orientation on the 3-manifold. The divergence-free condition for magnetic
vector fields corresponds to dB = 0.
Example 24.1. In Euclidean R3 , with the usual right-handed orientation con-
vention for the coordinate vector fields ∂x , ∂y , ∂z , a magnetic vector field B =
b ∂z of uniform strength b ∈ R in the z-direction corresponds to the 2-form
B = b dx ∧ dy.
Let X (the “sample”) be a connected, simply-connected, embedded 2-
submanifold in R3 (the “laboratory”). There is an induced Riemannian metric
g on X. We abuse notation and continue to write B for the restriction of the
the magnetic 2-form to X.

263
 To describe a charge-q particle on X under the influence of B ∈ Ω2 (X),
we need a Hermitian line bundle L → X with unitary connection ∇, whose
curvature is
iq
− B = F ∇. (24.3)
ℏ
For simplicity, we treat the case of spinless particles35 with positive unit charge
q = e, and suppress the physical unit ℏe . (For electrons, with q = −e, we change
the sign of B in the calculations.) Then the free Hamiltonian is the connection
Laplacian for ∇, called the Landau Hamiltonian, denoted
HB := ∆L .
The reason why only the curvature −iB of ∇ occurs in the notation, rather
than ∇ itself, is as follows.
1. The line bundle L → X is trivializable. This is because it is restricted
from a line bundle on the ambient contractible R3 . So for a unitary
connection ∇, there exists a global gauge potential A on X with
dA = F ∇ = −iB.
(A is a 1-form on X valued in u(1) = iR.)
2. Suppose à is the global gauge potential for another unitary connection
˜ on L, with F ∇˜ = F ∇ . Then
∇
˜
d(Ã − A) = F ∇ − F ∇ = 0.
Thus η = −i(Ã − A) is a closed R-valued 1-form. As X is simply-
connected36 , this implies that η = dΛ for some 0-form Λ. Then
à = A + iη = A + e−iΛ d(eiΛ ),
showing that there is a globally-defined gauge transformation, eiΛ which
˜
converts ∇ into ∇.
35
This simplifies the discussion, but the spinful case can also be treated with the techniques
that follow. In real samples, the free electron model is only an approximation, and it is the
ℏ2
effective mass of the “effectively free” electron which enters the 2m factor in the Landau
Hamiltonian. The spin contribution to the magnetic moment, as encoded in the second
term of (23.4), generally involves the free electron mass.
36
Recall from the Aharonov–Bohm effect discussion (Section 9.3), that when X is
multiply-connected, there are gauge-inequivalent unitary connections with the same cur-
vature.

264
 For calculations, it will be useful to pick an orientation on X, thus a Rie-
mannian volume form vol ≡ volX,g . Then B can be written as
B = B · vol, B ∈ C ∞ (X; R).
The scalar function B represents the magnetic field strength felt by an electron
confined to X. Accordingly, we write HB for the Landau Hamiltonian on the
line bundle L = LB .

24.2 Euclidean plane Landau Hamiltonian

The simplest example has X being a Euclidean plane R2 , and B being a
uniform magnetic 2-form. We pick an orientation, and coordinates (x, y), such
that the volume form is vol = dx ∧ dy, and the uniform magnetic field strength
B ≡ B(x, y) = b > 0.
We may pick a global gauge such that
i
A = − b(x dy − y dx) ⇒ F ∇ = dA = −ib dx ∧ dy = −ib · vol.
2
In this gauge, the covariant derivatives ∇∂x , ∇∂y act as
ib ib
∇∂x ∼ ∂x + y, ∇∂y ∼ ∂y − x, (24.4)
2 2
and the Landau Hamiltonian is represented as the operator
Hb = −∇2∂x ,∂x − ∇2∂y ,∂y = −∇∂x ∇∂x − ∇∂y ∇∂y
iby 2 ibx 2
= −(∂x + 2
) − (∂y − 2
). (24.5)
There is a SUSY way to solve for the spectrum of Hb . Considering the Dirac
operator twisted by LB , (the operator (21.1) with ∂x , ∂y replaced by the co-
variant derivatives (24.4)),
−i∂x − ∂y + ib2 (x + iy) ∂ − 4b z̄
   
0 0
Db = = −2i ¯ b .
−i∂x + ∂y − ib2 (x − iy) 0. ∂ + 4z 0
Above, we introduced complex coordinates z = x + iy for R2 = C, and the
holomorphic/antiholomorphic derivatives ∂ = 21 ∂x − i∂y and ∂¯ = 12 ∂x + i∂y for
later use. A short calculation shows that
 
2 Hb − b 0
Db = .
0 Hb + b

265
 The SUSY arguments from Section 7.3.2 can now be used37 to deduce

σ(Hb ) = (2N + 1)b. (24.6)

These eigenvalues are called Landau levels. Furthermore, the lowest Landau
level’s eigenspace is
b
ker(Hb − b) = ker(Db2 ) = ker(Db ) = ker(∂¯ + z)
4
n − 4b |z|2
= spanC {z e : n ∈ N}. (24.7)

This Hilbert subspace of L2 (C) is well-known in complex analysis, and is called

the Bargmann–Fock–Segal space.
For b = 0, the Laplace spectrum is well-known to be σ(∆) = [0, ∞). The
dramatic “discretization” of the spectrum once b > 0, Eq. (24.6), is called
Landau quantization, first discovered in 1930.
Remark. The discretized spectrum (24.6) is reminiscent of the simple harmonic
oscillator spectrum. For the latter, we learned from the SUSY construction of
the ground state that it is identifiable with the Fredholm index of a Dirac-type
operator. In the Landau Hamiltonian case, the ground state space, (24.7), is
infinite dimensional, so the “index” has to be counted in a “renormalized”
manner. Technically, this involves the K-theory of a certain operator algebra
(omitted, ongoing research).

24.2.1 Landau quantization for general surfaces (sketched)

An actual sample X is only approximately planar, and an actual magnetic
field is only approximately uniform. In experiments, one does not need to
precisely realize the geometric idealizations of the Euclidean plane model in
Section 24.2. Let us explain why the qualitative results of Section 24.2 still
hold in the geometric context of Section 24.1, with general (X, g) and general
B = B · vol.
First, let us compute the curvature of the spinor line bundles S ± → X.
Spin
Recall Eq. (22.17) for the spin curvature 2-form R(·, ·) = F ∇ (·, ·) in terms
37
The operator Hb can be shown to be essentially self-adjoint on compactly-supported
smooth functions (e.g. Theorem X.34 of Reed–Simon), and we close it to a self-adjoint
operator. Likewise essential self-adjointness of Db follows from a similar argument to Lemma
12.4. Both will generally fail if the underlying manifold is not a complete Riemannian
manifold.

266
of the Riemann tensor. In two dimensions, we only need to understand the
endomorphism R(e1 , e2 ) for {e1 , e2 } a local oriented orthonormal frame,
2
1X
R(e1 , e2 ) = g(R(e1 , e2 )ei , ej )c(ei )c(ej )
4 i,j=1
1

= g(R(e1 , e2 )e1 , e2 )c(e1 )c(e2 ) + g(R(e1 , e2 )e2 , e1 )c(e2 )c(e1 )
4
1

= − g(R(e1 , e2 )e2 , e1 )c(e1 )c(e2 ) − g(R(e2 , e1 )e1 , e2 )c(e1 )c(e2 )
4
2
1X
=− g(R(ej , ei )ei , ej )c(e1 )c(e2 )
4 i,j=1
Sc
=− c(e1 )c(e2 ).
4
Since c(e1 )c(e2 ) = ∓i on S ± , we deduce that R(e1 , e2 ) = ± i 4Sc on S ± . So
S±
F∇ = ± i 4Sc · vol, and

S ± ⊗LB Sc
F∇ = −i(∓ + B)vol.
4
Thus, up to unitary gauge-equivalence, we may identify
± ⊗L
∆S B
= HB∓ Sc .
4

Put these calculations into the Lichnerowicz identity, Eq. (23.6),

(X,g) Sc
/ LB )2 = ∆S⊗LB +
(D − iBc(e1 )c(e2 )
4 !
HB− Sc + Sc
4
−B 0
= 4
Sc
0 HB+ Sc + 4
+B
4

Equivalently, by replacing B with B + Sc

4
, we get
 
(X,g) 2 HB − B 0
(D/ L Sc ) = Sc . (24.8)
B+ 4 0 H B+ Sc + 2
+B
2

Now, let us suppose the true magnetic field strength B is a constant b > 0
plus some perturbation Bpert with ||B||∞ = supx∈X |Bpert (x)| ≪ b. If the

267
scalar curvature of X is not too negative,
 
Sc Sc
inf + b − 3||Bpert ||∞ > 0 ⇒ inf + B > 2||Bpert ||∞ , (24.9)
2 2
then HB+ Sc + Sc2
+ B remains a strictly positive operator with spectral gap at
2
least (0, c), where c > 2||Bpert ||∞ . Spectral SUSY says that HB − B also has
spectral gap at least (0, c), thus HB = (HB − B) + b + Bpert has spectral gap
at least
(b + ||Bpert ||∞ , b + c − ||Bpert ||∞ ).
Informally, the Euclidean plane model’s Landau levels b, 3b, . . . become
spread out into “Landau bands”. The bottom two bands do not overlap,
provided (24.9) holds.
Remark (Optional). The Landau Hamiltonian with constant B = b on a hyper-
bolic plane (constant negative scalar curvature) had been studied previously
by Comtet–Houston. Once the scalar curvature is overcome by a sufficiently
large field b (see Eq. (24.9)), isolated Landau levels can form in the spectrum.

24.3 More about QHE (Optional)

We have explained the geometric origin of the formation of Landau levels/bands
in the spectrum of magnetic Laplacians. As we outlined, an isolated Landau
level (or Landau band) realizes a Dirac “index”.
Now, we are not actually studying a single electron, but a whole density
of independent electrons occupying X. Although a Landau band’s spectral
subspace38 is infinite dimensional, it has a finite density of states, which can
be calculated to be ebh
. The electrons occupy these states one by one, starting
from the lowest energy one. The density of states is varied as b is varied, but
the density of electrons is some fixed material constant µ. Thus the filling
factor ν = µh
eb
is controlled experimentally by varying 1b .
Each completely filled Landau level/band is understood by physicists to
contribute one unit of Hall conductance,
e2
ν filled Landau levels/bands ↔ ν × Hall conductance, ν ∈ N.
h
(24.10)
38
Recall that the spectral projection for an isolated energy band [c, d] is constructed by
applying a (smoothened) characteristic function χ[c,d] to Hb in its functional calculus.

268
Assuming a so-called Kubo formula for the Hall conductance, the heuristic
(24.10) can be made rigorous in the planar geometry example (e.g. Avron–
Seiler–Simon’ 1994 paper, Charge Deficiency, Charge Transport and Com-
parison of Dimensions.). A more interesting statement is that this quan-
tization heuristic is still true for general geometries. See arXiv:2308.02819,
arXiv:2401.07449, for a recent approach using the idea of “macroscopic index
theory”.39 .
Regardless, the heuristic Eq. (24.10) is only applicable to the very special
situation of exactly integer filling, in which case, the claim “integer = integer”
is not so surprising. The true experimental surprise is that σHall remains an
integer even when ν is significantly varied near integer values,

(ν filled Landau levels) ↔ νint units of Hall conductance, ν ≈ integer.

Jumps occur when ν is close to a half-integer.

The key to understanding the rounding off of σHall is that the true experi-
mental setting is “dirty”. The electron actually experiences some electrostatic
potential V coming from all sorts of impurities. So we really need to study
the spectrum of the magnetic Schrödinger operator HB + V , where again, V
is assumed to be small but nonzero, and may even be random (i.e. basically
indeterminate to the experimentalist). The effect of this “dirt” is to modify
the spectrum of HB slightly, by introducing lots of new localized eigenstates
with energies near the Landau levels of the clean model. In this regard, the
Bpert term in B = b + Bpert could be viewed as “geometrical dirt”.
Now we can understand what happens when the filling factor ν (of a dirty
sample’s Landau bands) is not exactly integral — the “remainder” involves
(un)filling of the localized eigenstates of the dirty model, which does not con-
tribute to the Hall conductance. In other words, as long as ν is approximately
an integer, only νint matters, and we may apply the claim (24.10).

39
Recently, it has become popular to characterize such quantizations as being “topologi-
cal”, which may be confusing to topologists. What is really meant is a certain robustness
against deformations of a geometric-analytic object (e.g. the Hamiltonian). In certain ide-
alized situations, the origin of the analytic robustness is identifiable with a topological
invariant, and this is basically an index theorem. The “topological invariant” has almost
nothing to do with the topology of the sample X itself.

269