Chapter 1

The integer quantum Hall e↵ect I

H.L. Stormer: Nobel Lecture

Learning goals
• We know the basic phenomenology of the quantum Hall e↵ect (QHE)
• We know the structure of the lowest Landau level (LLL)
• We understand the role of disorder for the QHE.

• K. von Klitzing, G. Dorda, and M. Pepper, Phys. Rev. Lett. 45, 494 (1980)

In large parts of this chapter we follow reference [2].

1.1 Preliminaries
The Lorentz force acting on charged particles moving in a two-dimensional plane leads to a
build-up of charges perpendicular to the direction of motion. This is the classical Hall e↵ect
first discussed by Edwin Hall in 1879 [3]. To understand this, let us consider a two-dimensional
system which is translationally invariant. We move to a frame moving with v where we
therefore see a current
J = nev, (1.1)
where n is the particle density and e the electron charge. In the laboratory frame we have E = 0
and B = Bẑ. Hence, in the moving frame we obtain

E=v^B and B = Bẑ. (1.2)

We can express the electric field as

B
E= J ^ ẑ. (1.3)
ne

5
Figure 1.1: Measured Hall resistivity as a function of an applied back-gate which leads to a
change in the particle density n. The pronounced plateau is the hallmark of the quantum Hall
e↵ect. Figure taken from Ref [1].

The resistivity ⇢ is defined as the relation between the current and the electric field E µ = ⇢µ⌫ J ⌫ .
We thus find ✓ ◆ ✓ ◆
B 0 1 ne 0 1
⇢= ) = . (1.4)
ne 1 0 B 1 0
We see that owing to the non-zero xy the longitudinal resisitivity ⇢xx = xx = 0 is equal to the
longitudinal conductivity. Moreover, the Hall resistivity is proportional to the magnetic field
B
⇢xy = . (1.5)
ne
This is in striking contrast to the seminal discovery of von Klitzing and his co-workers in 1980
[1], see Fig. 1.1. The only ingredient in our theoretical model so far, however, was translational
symmetry. In the following, we first take steps towards a quantum mechanical understanding of
electrons in a magnetic field before we come back to the issue of translational symmetry breaking
via disorder.

1.2 Classical Lagrangian

To motivate how the magnetic field enters our quantum mechanical description, we recall that the
classical equations of motions are reproduced by the following Lagrangian L = m µ µ
2 ẋ ẋ eẋµ Aµ .

@ @L @L
µ
+ µ =0 ) mẍ = eB ẏ and mÿ = eB ẋ. (1.6)
@t ẋ @x
The canonical momentum is given by pµ = @L ẋµ = mẋ
µ eAµ and therefore the Hamiltonian
reads
1
H(xµ , pµ ) = ẋµ pµ L(xµ , ẋµ ) = (pµ + eAµ ) (pµ + eAµ ) . (1.7)
2m
With this small detour into classical mechanics we are now in the position to tackle the quantum
mechanical problem of a particle in a magnetic field.

6
1.3 Landau levels
We have to solve for the eigenstates of the following Hamiltonian
1
H= (p + eA)2 . (1.8)
2m
As only the vector potential A enters the Hamiltonian we have to chose an appropriate gauge.
For now we choose the Landau gauge where A = xB ŷ. We check that r ^ A ⌘ B = (@x Ay
@y Ax )ẑ = Bẑ. Inserted into the above Hamiltonian we obtain

1 h 2 i
H= px + (py + exB)2 . (1.9)
2m
We immediately observe that this Hamiltonian has a translational symmetry in y direction. We
therefore choose the following ansatz for the wave function (x, y) = eiky fk (x). With this ansatz
we obtain a family of one-dimensional problems (one per momentum k in y-direction)
r
~2 @x2 1 2 eB ~
hk = + m!c2 x + kl2 with !c = and l= . (1.10)
2m 2 m eB
We see that we are dealing with a (displaced) one-dimensional harmonic oscillator. The charac-
teristic frequency is known as the the cyclotron frequency !c . The displacement is proportional
to the y-momentum and measured in the natural length scale, the magnetic length l. Solving
the harmonic oscillator we find that
1
1. ✏k = ~!c s + 2 with s 2 N.

2. For s = 0, i.e., the LLL the wave function is a Gaussian centered at Xk = kl2
1 1
(x+kl2 )2 1 1
(x X k )2
(x, y) = q eiky e 2l2 =q eiky e 2l2 , (1.11)
⇡ 1/2 Ly l ⇡ 1/2 Ly l

where Ly is the extent in y-direction as shown in Fig. 1.2.

3. We have a vastly degenerate system. The number of degenerate states in each LL is given
by
Z 2
Ly Lx /l Lx Ly Lx Ly B
N= dk = 2
= , (1.12)
2⇡ 0 2⇡l 0

where 0 = h/e is the magnetic flux quantum. In other words, per magnetic flux quantum
that penetrates the sample we have one state per Landau level.

Before we continue we should remind ourselves that in the case of huge degeneracies any pertur-
bation might have dramatic e↵ect. Moreover, the choice of basis can facilitate the description
of these e↵ects. For the case of a magnetic field, the choice of gauge determined the shape of
the basis wave-functions. We will come back to this point later.

Xk = kl2 x

Figure 1.2: Eigenfunctions of the LLL in the Landau gauge.

7
1.4 Currents
We set out to understand the Hall conductivity. To make further progress, we need to calculate
e
currents. We evaluate the current operator in y direction, Jy = m (py + eAy ), in the LLL
eigenfunctions
Z (x Xk )2 (x Xk )2
e
h |Jy | i = dx e 2l2 (~k + eBx)e 2l2 (1.13)
m⇡ 1/2 l
Z (x Xk )2
Z
e!c 2 e!c ↵2
= dx e l 2 (x + kl ) = d↵ e l2 ↵ = 0. (1.14)
⇡ 1/2 l ⇡ 1/2 l
The last equality holds as the integrand is odd under ↵ ! ↵. In other words, no net current
is flowing as shown in Fig. 1.3.

Jy (x)

+
x

Figure 1.3: Current distribution in the lowest Landau level.

For a current to flow, we need to add an electric field in x-direction V (x) = eEx. We still
are translationally invariant in y-direction and the one-dimensional problem is changed to

~2 @x2 1 2
hk = + m!c2 x + kl2 + eEx (1.15)
2m 2
✓ ◆
~2 @x2 1 eE 2 1
= 2
+ m!c x + kl +2
eEXk0 + mv̄, (1.16)
2m 2 m!c2 2

where the center of the Gaussians is shifted

Xk0 = kl2 eE/m!c2 (1.17)

and an additional energy 12 mv̄ 2 with v̄ = E/B is arises from the drift of the electrons. The
immediate conclusion is that the new energy depends on k, i.e., the huge degeneracy is lifted
1 1
✏k = ~!c + eEXk0 + mv̄ 2 . (1.18)
2 2
With an energy that depends on k we can also calculate a non-zero group velocity

1 @✏k eE @Xk0 eE 2 E
vgroup = = = l = = v̄. (1.19)
~ @k ~ @k ~ B
We this we reach the classical result
ne
hJy i = ev̄ ) xy = . (1.20)
B
We this result we reach the same conclusion as with the classical manipulations based entirely
on translational symmetry in the beginning of this chapter. In order to make further progress
we should take a closer look at the finite extent of a realistic sample as well as on disorder e↵ects
to understand the quantization of xy .

8
1.5 Edge states

We try to build an understand-

ing of the influence of the edges s=2
of a two dimensional sample by
considering a strip which is finite s=1
in x-direction and infinite (or pe-
riodic) in y-direction. The ba- LLL (s = 0)
sis wave functions of the LLL, or Xk
equivalently the gauge choice for
A, which we used above is opti-
mally tailored to this geometry. V (x): sample edges
Remember that the wave func- Figure 1.4: Edge states from curved Landau levels.
tions are localized in x-direction
with a typical extent l. If we now consider a potential V (x) that confines the electrons to a finite
region which is smooth over the length-scale l, we can expect the wave function to remain ap-
proximately Gaussian. However, the wave functions centered in the vicinity of the edges will be
lifted in energy. As the position of the wave function is linked to the momentum k in y-direction
we obtain dispersive edge channels.
In order to determine how the current is distributed we again
calculate the group velocity.
(
1 @✏k 1 @✏k @Xk l2 @✏k < 0 right edge
vgroup = = = = .
~ @k ~ @Xk @k ~ @Xk > 0 left edge
(1.21)
Figure 1.5: Classical These simple manipulations reveal that the two opposite edges carry
skipping orbits. opposite current. This can also be understood from the classical
“skipping orbits” picture as shown on the left.
In order to calculate xy we now apply a voltage VH between the
two edges (in x-direction) and calculate the resulting current along the sample (in y-direction).
Moreover, we assume that the Fermi energy EF lies in between two Landau levels.
The obtain the total current Iy we sum over the contri-
bution evk of all occupied states
Z 1 EF
dk 1 @✏k
Iy = e nk , (1.22)
1 2⇡ ~ @k

where nk is the occupation probability of the k’th mode. VH

Under the assumption that we only fill the LLL and that Figure 1.6: Voltage bias.
we are at zero temperature the occupation numbers only take the values nk = {1, 0}. Under
these assumptions we arrive at
Z
e µR e
Iy = d✏ = (µR µL ), (1.23)
h µL h
where µR/L are the respective chemical potentials on the two sides. As we can write eVH =
µR µL we arrive at
e2 e2
Iy = VH ) xy = . (1.24)
h h
Let us move now the Fermi energy in between any two LL and we immediately conclude that
e2
xy = ⌫ , (1.25)
h
where the integer ⌫ counts the number of filled LL’s.

9
 Figure 1.7: Chemical potential stuck to Landau levels.

1.5.1 The e↵ect of disorder

The above result is strongly suggestive that one dimensional edge channels are responsible for
the transport in the quantum Hall e↵ect. Generically the current carried by a one-dimensional
channel is given by
e2
I = |T |2 , (1.26)
h
where |T |2 denotes the probability for an electron to be transmitted through a disordered region.
However, our edge channels are chiral where the electrons have no way to be back-scattered and
therefore |T |2 = 1. These arguments explain why even in the case of a disordered sample xy
can be quantized. However, we did not yet reconcile a quantized xy with the general result
xy = ne/B for a clean system.
We assumed the Fermi energy to lie between two Landau levels. Let us see under which
conditions this can be the case. We assume the sample to be Lx wide and the edge region which
is curved up to extend over the length W ⌧ Lx . From the finite size (or periodic) quantization
in y-direction we know that the momentum can take the values ki = L2⇡y i with i 2 Z. Hence, we
find for the centers of the Localized wave functions Xi = L2⇡y l2 i. We now count how many wave
functions fit into the bulk and how many into the edge:
W Lx
edge : / Ly , bulk : / Lx Ly . (1.27)
Xi Xi 1 Xi Xi 1

We see that there are extensively many bulk states but only a sub-extensive number of edge
states as shown in Fig. 1.7.
Translated to a fixed density but varying magnetic field B we find that for almost all values
of B the Fermi energy will lie in the bulk, not the edge! Meaning, our assumption that the
we have a completely filled LL and the relevant physics is happening only on the edge was not
justified. Hence we need to get a better understanding of disorder e↵ects.

References
1. V. Klitzing, K., Dorda, G. & Pepper, M. “New Method for High-Accuracy Determination
of the Fine-Structure Constant Based on Quantized Hall Resistance”. Phys. Rev. Lett. 45,
494 (1980).
2. Topological aspects of low dimensional systems (eds Comtet, A., Jolicoeur, T., Ouvry, S. &
David, F.) (Springer-Verlag, Berlin, 1999).

10
3. Hall, E. “On a New Action of the Magnet on Electric Currents”. Amer. J. Math. 2, 287
(1879).

11