Download as pdf or txt
Download as pdf or txt
You are on page 1of 32

University of Amsterdam

Science Park 904, 1098 XH Amsterdam

A Historical Perspective on the Quantum Hall Effect


Wout Neutkens

Amsterdam, 21 january 2013


Institute for Theoretical Physics (ITFA)
Master Theoretical Physics, Track CE
Supervisor: AMM Pruisken
Examinator: B. Nienhuis
2nd Examinator: A. de Visser
Student: W. Neutkens
Student number: 0306347
Abstract
Around 1980, von Klitzing et al. 1 discovered the quantized behav-
ior of a cold, 2D electron gas in a randomly disordered potential that
is placed in a high magnetic eld perpendicular to the surface. At that
time, the eld theoretical description of this electron setting was incom-
plete and the quantum Hall eect could only be explained by semi-classical
arguments. These ad-hock arguments became very popular in the liter-
ature because they provided a good temporal framework for research on
the quantum Hall eect. When, four years later, the fundamental prob-
lems of the eld theoretical description were solved by Levine, Libby and
Pruisken 24 the physics community coudn't accomodate a new, eld the-
oretical framework. `A historical perspective on the quantum Hall eect'
gives a brief overview of the semi-classical arguments and then emphasizes
on the historical formation of a (modern) eld theoretical perspective and
its relation to experiment.

2
Contents
1 The Early Quantum Hall Eect 4
1.1 The Discovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Extended versus Localized States . . . . . . . . . . . . . . . . . 5
1.3 Quantized Conductance . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 The Percolation Picture . . . . . . . . . . . . . . . . . . . . . . . 8
1.5 The True Quantum Hall Regime . . . . . . . . . . . . . . . . . . 9

2 The Modern Quantum Hall Eect 10


2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Anderson Localization . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3 No QHE from Anderson Localization . . . . . . . . . . . . . . . . 13
2.4 No alternative for Anderson Localization . . . . . . . . . . . . . . 14
2.5 The Non-Linear Sigma Model and Disorder . . . . . . . . . . . . 15
2.6 The Instanton Vacuum . . . . . . . . . . . . . . . . . . . . . . . . 16
2.7 Importance of the Edge . . . . . . . . . . . . . . . . . . . . . . . 17
2.8 General Features are Super-Universal . . . . . . . . . . . . . . . . 18

3 Experiments on Critical Behavior 20


3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2 Quantum Hall Laboratory . . . . . . . . . . . . . . . . . . . . . 21
3.3 The Critical Exponent . . . . . . . . . . . . . . . . . . . . . . . 22
3.4 Fermi-liquid principles . . . . . . . . . . . . . . . . . . . . . . . . 25

4 Other issues 26
4.1 Coulomb Interactions . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2 Super-Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

5 Summary 28
References 29

3
1 The Early Quantum Hall Eect
1.1 The Discovery

The quantum Hall eect (QHE) occurs when a 2D electron gas is cooled down
to low temperatures and placed in a high magnetic eld. When we x a current
in the longitudinal direction of a rectangular sample, the electrons start to move
and a Lorentz force deects them in the perpendicular direction. This causes
an electric force in the perpendicular direction that eventually compensates the
Lorentz force. We measure the voltages in both directions and obtain the resis-
tance tensor. The perpendicular voltage dierence divided by the longitudinal
current is then called the `Hall resistance'. If the magnetic eld and the tem-
perature are tuned to the quantum regime, the Hall resistance shows a series of
plateaus as a function of the magnetic eld (gure 1.1). On these plateaus the
longitudinal resistance vanishes, but there are peaks at the transitions between
plateaus. As rst observed by von Klitzing et al. 1 in 1980, the values for the
Hall resistance at which the plateaus occur are extremely accurate. These values
were somehow not inuenced by minor variations in dierent samples, but they
were precisely determined by just integer multiples of the fundamental units of
electrical charge and Plank's constant. Therefore, the QHE could be used to
provide an independent measurement of the fundamental ne structure coupling
constant appearing in electromagnetic interactions and even as a standard of
resistance. But truly interesting for theoretical physics was the universality, the
sample-independence of this result.

Figure 1.1: The quantum Hall eect in a GaAs-GaAlAs heterojunction, recorded


at 30mK. The Hall resistance xy forms a `staircase' of plateaus. The longitu-
dinal resistance XX forms peaks at the transitions between plateaus. (Image
Courtesy of D.R. Leadley, Warwick University 1997)

4
How can measurements on dierent samples return exactly the same, univer-
sal values? These samples all have dierent boundaries, electron densities and
other features that should inuence the measured resistance. In condensed mat-
ter, we often start with an innite, homogeneous system without boundaries.
Any universal result that is obtained from this theory would not be immediately
the result that one measures. Experimentalists use tricks to extract the universal
result from their measurements. In the quantum Hall case, it is the universal
result that is directly obtained from the dierent samples. Are we somehow
very lucky here? For example, the longitudinal resistance is proportional to
the length of the sample. Dierent samples will therefore yield dierent results
that are all proportional to the presumed universal result. The length of every
sample has to be measured very carefully to extract this result, which would
be completely impossible in the Quantum Hall case. But, the Hall resistivity is
the perpendicular voltage dierence divided by the longitudinal current. If we
rescale the length, the longitudinal current and perpendicular voltage remain
the same. If on the other hand we rescale the width, the longitudinal current
and the perpendicular voltage dierence both change, but they change equally.
The longitudinal resistance will change upon rescaling. It is only because of
two-dimensions and it is only the Hall resistance that doesn't depend on sample
dimensions. The precision of measuring a universal result for the longitudinal
resistance will always be limited by the small variations in sample dimensions
that are present if we compare dierent samples. It is therefore at least pos-
sible that our two-dimensional setting provides us a result that is independent
of small variations in the dimensions of the sample. The fact that the result is
also independent of the entire geometry of the 2D samples as a whole, and more
importantly, that it is even independent of the many kinds of disorder in the
samples is not at all explained!

1.2 Extended versus Localized States

As it is well known, a magnetic eld makes a moving electron to circle around


some (possibly unxed) guiding center. In the two-dimensional electron gas, this
causes the energy levels of a free electron to become quantized. This is visible
in the quantum regime, where the thermal energy is smaller than the level
separation. In the staircase picture (gure 1.1), the horizontal plateaus do not
occur because the energies are quantized. In any disordered sample, the energy
levels (called Landau Levels) are broadened and the disorder broadening of the
energy levels is so large that their tails are overlapping. It is not a band gap,
but the immobility of the carrier system that prohibits a change in resistance.
The horizontal plateaus indicate that the carrier system has to be extremely
immobile. But surprisingly, the sudden changes between those plateaus and the
peaks in the longitudinal resistance indicate that the carrier system can also be
very mobile. Here, the Fermi energy approaches the center of a Landau band.
In other words, near the band center the localization lengths of the electrons
are macroscopically large, while in the tails of the Landau bands, they are
microscopically small (see gure 1.2). Obviously, we are dealing with some sort

5
of phase transition at (or near) every band center. The entire staircase picture,
consisting of very at plateaus and phase transitions in between, is visible as
a function of the magnetic eld. Increasing the magnetic eld introduces more
possibilities to form electrons per Landau level. Since the total number of
electrons remains constant, increasing the magnetic eld sweeps us through the
staircase picture.

Figure 1.2: A Landau band. The states in the centre (indicated by a black
line) are completely extended, while the rest of the states (shaded area) are
completely localized.

The Hall resistance remains nite after the Fermi energy is again in a region
of localized states, because even when all states are immobile in the bulk of the
system, current is still able to ow through electron states that extend along the
edge of the sample. In our two-dimensional system the conductance is related
to the resistance by matrix inversion. Since at the plateaus the longitudinal
resistance is zero, the Hall conductance is just the inverse of the Hall resistance.
Every time we shift the Fermi energy over the center of a Landau band, the
Hall conductance increases exactly with e2 /h. We will rst discuss the exact
dierence between the conductance plateaus (subsection 1.3) and then we will
discuss why there can be very localized states as well as very extended states in
the system (subsection 1.4).

1.3 Quantized Conductance

An argument that became famous in the literature on the QHE, Laughlin's


gauge argument, assumes the existence of extended states (which is a very non-
trivial assumption). Laughlin 5 considers the 2D electron gas in the geometry
of a loop, where the magnetic eld points perpendicular to the surface. He
then inserts a magnetic ux through the loop adiabatically. This changes the
phase of the quantum mechanical wave functions. The extended states, that
go all the way around the loop, can be gauge transformed back to the original

6
situation if their phase has been changed by any multiples of 2 . This phase
change alters the potential energy in the direction along the loop and if a gauge
transformation was possible, the total electronic energy must have remained the
same. But, it can be compensated by a change in energy over a perpendicular
potential. Thus, the wave phase is allowed to change by a multiple of 2 and
then the Hall conductance changes with an integer multiple of e2 /h.
It is important that the Fermi energy is in a region of well-localized states,
otherwise the wave phases are allowed to change by non-integer multiples of
2 . What is then the mechanism by which electric charge is transported across
the ribbon? It is possible that there is no such mechanism and the `integer'
equals zero. But just after Laughlin's publication, Halperin 6 claries the charge
transfer across the ribbon. At an edge, the conning potential shifts up to
innity and the Landau levels will cross the Fermi energy (see gure 1.3). An

Figure 1.3: The Landau levels that are under the Fermi energy, will cross this
energy at the edges

electron in an extended state can get excited into a localized state just above
the Fermi energy on one edge of the sample. This leaves a hole behind from
a state that extends to the other edge of the sample. A localized electron just
below the Fermi energy on the other edge can thus ll this hole and eectively,
charge has been transferred across the ribbon.
The geometry of a loop can be related to the rectangular Hall bar by con-
sidering the formation of extended edge states. These edge states should be
somehow not inuenced by all the microscopic bumps and so on in the sample.
Butticker 7 explained this due to the absence of back scattering at the edge.
Instead of going around in a cyclotron motion, a particle reverses its motion
every time it scatters from the edge resulting in series of semicircles following
the edge equipotential line. This motion was called a `skipping orbit' and it
was shown that disorder on the edges, even large disorder from the contacts
used to measure the voltages, altered only the shape of those orbits, but not the
extended edge states themselves.

7
1.4 The Percolation Picture

We have seen that there had to be extended states near the center of the Lan-
dau bands as well as localized states elsewhere. To understand this a percola-
tion concept was introduced 8;9 . Percolation can have dierent manifestations,
ranging from current percolation to the percolation of rigidity. In physics it is
always a critical phenomenon. In the context of the quantum Hall eect, one
rst assumes that the spatial correlation of the disorder potential is large com-
pared to the magnetic length. Thus, we have a very smooth potential landscape
and since an electron cannot change its energy, its guiding center moves along
equipotential lines. The disorder potential can be compared to a landscape with
mountains and valleys. At the tail of a Landau band, we are either near the top
of a mountain, or near the bottom of a valley. In this situation, the equipoten-
tial lines are closed shapes with a mean radius. But, if we approach the center
of a Landau band, the radius of the shapes becomes bigger and when valleys
turn into mountains, the radius of a valley circle exceeds the system size and
electrons can only move along mountain circles or along the edges. Thus, in the
tails of a Landau band (the valleys or mountains), the electrons are localized
and in the center of the band and at the edges, the states are extended (see
gure 1.4). Since the equipotential circles are very big near the center of a Lan-
dau band, there are many points where they are close to each other. Quantum
tunneling then causes diusion of electrons over multiple valleys or mountains.
This implies that the critical aspects of the valley/mountain phase transition
are not determined by classical percolation, but by quantum percolation.

Figure 1.4: A random contour map. The black dots indicate areas where an
electron cyclotron motion takes place. Such an area cannot be too large for
the percolation picture to apply. The black line indicates a percolating shape
with a radius that is larger than the sample size. For larger samples, the energy
window where percolating states occur becomes smaller.

8
1.5 The True Quantum Hall Regime

Conclusively, we obtain a picture where for every disorder broadened Landau


band under the Fermi energy, there are extended states near the band center
that give rise to a quantization condition. At the edges, the Landau level shifts
up to the Fermi energy allowing an edge current to ow and resulting in a
quantized Hall conductance.
To explain the Quantum Hall Eect at a fundamental level, we will need
our modern techniques of Quantum Field Theory. The argumentations from
above only clarify the rst measurements on the quantum Hall eect. Their
major shortcoming is that they are formulated entirely from a semi-classical
perspective. A statistical eld theory describing the quantum Hall eect re-
quires the formulation of certain abstract operators acting on elds that live
in a particular space. These operators are then weighed by a set of coupling
constants that is evaluated from a perturbative (renormalization) scheme. This
gives rise to an entirely dierent language and perspective. Take for example
localization, quantization and edge currents. In the percolation picture, local-
ization of the electron states was due to closed equipotential shapes while in
the eld theoretical language, localization is described by the general theory of
Anderson localization - that also applies to non-smooth randomness. Also, we
have seen that the quantization of the Hall current is explained with Laughlin's
gauge argument, while in eld theory, the quantization arises due to topological
invariants in the mapping structure of the elds. Finally, the edge currents are
described by one-dimensional `massless chiral edge excitations', these are very
dierent from the edge currents described by Butticker.
Anderson localization in the presence of a magnetic eld, topological invari-
ants and massless chiral edge excitations have all been subject of Pruisken's
work. It is this work that is the starting point of a mathematical description
of the QHE. For example, by understanding more about his Quantum Field
Theory, we may infer what happens to the percolation picture when the disor-
der correlation length is of the order of the magnetic length, instead of much
larger. Although such an understanding may lead to a clear mental picture of
the percolation mechanism (which may be important for education, nummerics,
etc..) the path of theoretical physics narrows down to the most fundamental as-
pect of the Quantum Hall measurement. Since the renormalization behavior of
the conductance parameter describes this measurement from a Quantum Field
Theory perspective, we will have to take a look at the the transitions that occur
in between the staircase plateaus (In 1988, it would be experimentally veried 10
that these were true quantum phase transitions).
Unfortunately, `looking at the phase transitions' was often done from the
perspective of the percolation picture, which also happens to be a critical phe-
nomenon. In smoothly disordered samples, there is indeed a low temperature
regime where power-law scaling reminiscent of this picture can be observed,
But in the limit of zero temperature, or innite system size, the critical aspects
are dierent. Although this research can be meaningful for the same practical
applications, it is obviously not the same study of critical aspects that we are

9
interested in. Still, these dierent kinds of research are often intertwined in the
literature. We have to remember that the percolation picture is irrelevant in
the sense of a eld theoretical description of the Quantum Hall Eect.
This statement about percolation is actually very similar to an earlier state-
ment by Mott and Thouless even before the QHE was discovered. In that
context, the same two-dimensional, disordered electron system as the QHE was
studied, but then without a magnetic eld (we will look at this situation in more
detail later on). To investigate the critical aspects of the QHE from a theoretical
perspective we will have to learn more about Pruisken's work on the subject.
Since the appropriate eld theory was already found by him three decades ago
and he and his collaborators have been working on it ever since, we can only
lightly delve into his work in the present thesis.
Also, there are real measurements on the phase transitions possible in the
Quantum Hall laboratory. These subtle measurements are currently the deep-
est connection of the theoretical framework with its laboratory. But, the mea-
surements on the QHE phase transitions are technically very challenging. In
smoothly disordered samples we have already stated that the percolation criti-
cal phenomenon interferes with the true quantum phase transition. Therefore,
these smoothly disordered samples have to be of subkelvin temperatures in or-
der to measure the critical exponents of interest. But, from the perspective of
eld theory, there is more to quantum phase transitions than just their critical
exponents. For example, since both the Hall conductance and the longitudinal
conductance undergo the same phase transition, their behavior has been studied
in a renormalization group ow diagram (see gure 2.3 on page 19). Even less
cold, or less ideal samples are then suitable to study critical behavior, albeit
only in a more general sense.
Today it is found 11 that for long-ranged disorder correlations there are two
regimes with power law scaling behavior as the temperature is lowered. In-
deed, the regime of lowest temperature has a critical exponent that is univer-
sal and independent of the disorder correlation length (see gure 1.5). In the
zero-temperature limit the smoothly disordered system always ows to the true
quantum regime. In this limit, percolation breaks down because length scales
much larger than the separation between tunneling points will be revealed. The
tunneling ensemble can then again be imagined as a disordered potential and
we end up with an Anderson transition, just like in the non-smoothly disor-
dered system. We can therefore say that, in the disordered electron gas, the
percolation picture is `unstable' with respect to scaling.

2 The Modern Quantum Hall Eect


2.1 Introduction

The modern language of the Quantum Hall Eect has its roots in a eld theo-
retical description of localization in electron systems. Just before the quantum
Hall eect was discovered, it was strongly believed that all electronic states in

10
Figure 1.5: Two regimes of criticality for a smoothly disordered system (recent
experimental results 11 ). One with critical exponent = 0.58 and one with
= 0.42. Above: For the lowest temperatures, the exponent saturated because
of nite sample size. Below: The squared dots indicate a system with a shorter
disorder correlation length. Its critical exponent remains = 0.42 for higher
temperatures.

two dimensions had to be localized by the mechanism of Anderson localization.


This highly nontrivial and fundamental result was based on a novel formulation
of Anderson localization, many years after Anderson's original discovery 12 . The
success of Anderson localization was based on the possibility to apply renormal-
ization group ideas on the disordered electron system 13 . Because the problem
of localization in electron systems was described in the modern language of eld
theory, the results of Anderson localization were believed to be `proven facts'
instead of theoretical pictures (like the percolation picture) or other models that
could be very close to the reality of the experiment.
The most practical and interesting result, was that in two dimensions, all
electronic states had to be localized. This means that even a very good two
dimensional conductor, with macroscopically large localization lengths in the
quantum regime, will always be a perfect insulator on an innite sample. The
little disorder that is present in the `conductor' will localize the electronic states
for some length that is large enough. Such a result was considered a victory

11
for the renormalization group theory, but the Quantum Hall Eect was still to
be discovered. It doesn't take much imagination to understand that the QHE
cannot be in accordance with `all states are localized in 2D'. It did take a lot of
imagination to disprove the `proven facts' of Anderson localization.
If Anderson localization is wrong, then either quantum eld theory or renor-
malization group theory must be wrong and that is quite an alarming situa-
tion in theoretical physics. But this situation was almost three decades ago,
when topological issues were only beginning to become important in statistical
physics. Pruisken and his colleagues were at the frontier of understanding that
those issues are as important as we know today. As we will see, their unexpected
solution to the `extended states' problem is in fact topological in nature and it
is able to reconcile Anderson localization with the results of the Quantum Hall
Eect.

2.2 Anderson Localization

The theory of Anderson localization is a statistical quantum model of electrons


in the presence of disorder. Since every sample in reality is somewhat disor-
dered, we can imagine that Anderson localization is in fact a vast subject in
statistical physics. Localization of electrons in general can be due to the re-
pulsive Coulomb interactions (`Mott localization' ) as well as due to disorder
(`Anderson localization'). There is weak and strong Anderson localization. In
the path integral formulation we can imagine an electron to follow many paths
of which some return to the original location. For every such self-returning path,
there is a path in the opposite direction which is otherwise exactly the same
and the electron traversing this path has the same phase shift as the electron
traversing the original path. Thus, after scattering, the electron constructively
self-interferes at its origin, causing the weak localization. For more heavy dis-
order, wave interferences everywhere around the electron stop the wave from
propagating through the sample, and there is strong localization.
With the renormalization group, the Anderson localization problem could be
formulated as a scaling theory. This was motivated by a relation on a quantum
mechanical level found by Edwards and Thouless 14 between the conductance
and the response to perturbations of boundary conditions. As for scaling, this
means that the change in the conductance with the system size only depends
on the conductance itself. This gives rise to a one-parameter scaling theory and
a beta function can be calculated perturbatively. In the Anderson theory, it
was found that for an initially large conductance (little disorder), the scaling
corrections are very small but negative in two dimensions. Since the corrections
are small, scaling the system only slightly alters the conductance. In this regime
the electrons are weakly localized. But, the beta function is always negative in
two dimensions and scaling to larger sizes will decrease the conductivity. For
little conductance (initially strong disorder) the beta function is very negative
and scaling the system induces large, negative corrections to the conductance.
This is the regime of strong localization and the two-dimensional system will
always ow to a perfect insulator (see gure 2.1). Therefore the conclusion could

12
be reached that for any given amount of disorder, no states with innitely large
localization lengths exist in two dimensions.

Figure 2.1: Beta functions of some d dimensional disordered systems. Arrows


indicate the renormalization ow upon scaling to larger system sizes. In two
dimensions, corrections to the conductance can be small, but they always remain
negative. (Image courtesy E. Abrahams 15 )

2.3 No QHE from Anderson Localization

By renormalizing the general action (formula 2.1 on page 16) for the disordered
electron gas, the beta function as in gure 2.1 can be obtained and hence the
famous result of Anderson localization theory that `all states are localized in
two dimensions'. Around 1980, when the quantum Hall eect was discovered,
the eect of the magnetic eld was included in Anderson theory by considering
the symmetry requirement for the eld matrices not to be orthogonal, but only
of unitary nature. Then, weak localization becomes even weaker because self-
interference requires time-reversal symmetry, which is broken in the unitary
ensemble. In this case, the localization lengths of the electrons can become
macroscopically large and the nite Hall conductance was naively explained.
But in fact, the Hall conductance is so robustly quantized that it doesn't change
within measurable precision for larger system sizes. Then the mean localization
length, which is of Gaussian curvature, has to be so extremely large that this is
irreconcilable with the results of unitary Anderson models. In other words, the
quantum Hall eect says that `there are extended states in two dimensions'.

13
The fundamental issue here is that the statements `all states are localized'
and `there are extended states' (one from eld theory and one from experiment),
are contradicting. From our modern perspective on this situation we know that
topological considerations will solve this kind of issue. Around 1980, many
kinds of `pictures', `arguments' or `mechanisms' were invented to understand
or describe the Quantum Hall Eect, but these approaches obviously couldn't
(and didn't) solve our fundamental issue.

2.4 No alternative for Anderson Localization

In our introduction to the Quantum Hall Eect, a number of such explanations


like Lauglin's gauge argument or Landau level quantization were reviewed and
the problem with the percolation picture was already indicated. In order to
clearly separate those early inventions (that are still appearing in the modern
literature on the QHE) from the modern perspective we will again indicate the
dierence between the percolation picture and Anderson localization before we
continue.
In the context of the QHE, percolation ideas were introduced in attempts 8;16;17
to understand the delocalization mechanism of the electrons in their disordered
landscape (see subsection 1.4). However, even before the discovery of the quan-
tum Hall eect, practically the same percolation ideas were introduced in the
disordered electron gas without a magnetic eld and these ideas were already
found to be wrong! From the percolation perspective, the conclusion that ex-
tended states could exist in two dimensions was reached by Cohen and Jortner 18
and that there was a mobility edge near half-lling of the disordered potential.
Then, after Cohen and Jortner, Thouless and Mott independently published the-
oretical papers 19;20 that demonstrated the invalidity of percolation and other
mere intuitive ideas about the mobility edge.
Without a magnetic eld, percolation predicts that the electron states are
localized in valleys (lakes) for low lling fractions and that for high lling frac-
tions all states are extended (a big, `extended' sea with only a few islands, see
gure 2.2). This perspective radically changes when we let eld theory de-
scribe the system and thereby include all quantum eects like tunneling and
self-interference. As we have seen in subsection 2.2, renormalization predicts no
extended states in two dimensions and no phase transition. Even for high lling
fractions (the big sea with little islands) the states are essentially localized, but
their localization lengths are so large that we have to imagine the system on
much larger length scales to get the picture. The results of eld theory are there-
fore the very non-intuitive ideas of Anderson localization, where all states are
essentially localized in two dimensions and the former `mobility edge' becomes
a nite-size (or nite-temperature) eect.
So before the quantum Hall eect was invented, the conclusion was already
reached that the localization problem could not at all be described by any
percolation mechanisms. Thouless and other theoreticians that reached those
conclusions about the invalidity of percolation 21 , still continued to investigate
the same percolation ideas after the quantum Hall eects discovery. Because

14
the Quantum Hall Eect could not at all be explained from the perspective of
Anderson eld theory, they tried to describe the electron gas again with the
percolation picture. The disagreement between percolation and eld theory
was apparently less impressive than the problem of `No QHE from Anderson
localization'.

Figure 2.2: with only few islands as a metaphor for a disordered system with
high lling fraction. The fact that the electrons in the sea are able to perco-
late through the system is pretty obvious from this picture. Still, Anderson
localization predicts that there are no extended states in any 2D disordered
system. This shows that the percolation picture is not predictive at all in the
true (innite) quantum regime.

2.5 The Non-Linear Sigma Model and Disorder

To describe the disordered electron gas as an eective eld theory, a non-linear


sigma model can be used 22 . The non-linear sigma model is actually a very
general model, where the target manifold has any nontrivial curvature (often a
sphere). It can be used when a global symmetry group is spontaneously broken
by the ground states, which possess only the symmetry of a subgroup. In the
modern context of the quantum Hall eect, the special unitary SU (N + M ) Lie
group is being investigated and the vacuum possesses only a S(U (N ) U (M ))
symmetry. The Grassmanian eld matrices Q have N + M components and
satisfy the non-linear constraint of unitarity. The action that describes the

15
dynamics of the isolated system without a magnetic eld is given by


S= drtr(Q)2 (2.1)
8
Where is the unrenormalized conductance. Besides the disordered electron
gas, where the phenomenon of Anderson localization occurs, the non-linear
sigma model contains other systems like anti-ferromagnetic spin chains in the
SO(3) model. This system is obtained by setting N = M = 1 (SU (2) is the
double covering group of SO(3)).
The two dimensional electron gas without disorder has a vector eld with
only two components. The N +M component eld matrices are studied because
the disorder must be taken into account. This can be done by averaging the
logarithm of the partition function (the entropy) over the disorder potential.
This is possible because any observables are expected to be self-averaging with
respect to the disorder. The entropy, or the logarithm of the partition function,
can be transformed into a polynomial function by performing the `replica trick'.
The problem of averaging the logarithm of the partition function can be reduced
to the problem of averaging a sum of powers of the partition function. By the
denition of the partition function, taking powers of the partition function with
itself is like taking `replicas' of the same system. It is often sucient to just
consider this set of replicas as a grand system, of which the dimension equals
the amounth of replicas.
The two-dimensional electron system is therefore studied as an N + M di-
mensional system. The N + M components from the elds in the SU (N + M )
Lie group as mentioned above are the replicas of the two-dimensional elds. In
the end, analytic continuation to zero replicas (N, M 0) has to be performed
to obtain the real, two-dimensional electron gas. Sometimes this analytic con-
tinuation mechanism introduces extra parameters in the model and this is called
`replica symmetry breaking'. If there is replica symmetry breaking, the phases
in the replicated system, do not guarantee the same phases real system. For
the Quantum Hall Eect, this doesn't matter, because its most prominent fea-
tures are already apparent in the replicated model (see subsection 2.8). By
considering the replica trick, it is possible to include a general disorder distri-
bution function. Not an unknown, specic realization of the disorder potential
has to be provided, but a general distribution function suces, with certain
characteristics such as the disorder correlation length.

2.6 The Instanton Vacuum

Levine, Libby and Pruisken 24 managed to solve the fundamental shortcoming
of localization theory by introducing the theta angle in the Anderson model.
The theta angle came from an unexpected `angle' (Quantum Chromodynam-
ics, QCD) because such a bizarre topological issue wasn't considered by then
in any of the statistical eld theories of nite samples. Apart from topology,
the theories of QCD and Anderson theory share quite fundamental similarities.

16
For example, we have seen that in Anderson theory there is a crossover be-
tween weak localization at small scales to strong localization at large scales. In
Quantum Chromodynamics, there is a crossover between weak connement at
small scales to strong connement at large scales. This phenomenon is called
`asymptotic freedom' and strong connement is a necessary ingredient for any
realistic quantum eld theory about quarks.
The similarity that those theories didn't share at that time was the theta
angle. This angle is equal to the weight of an operator that measures a purely
topological property of the eld. It can therefore be used as a label for the
dierent topological sectors in the nite system. But, in QCD, its physical
interpretation remained completely mysterious. Under the normal conditions
where QCD applies, the theta angle is very small and has no eect on the quark
connement, but it was argued by 't Hooft that above a critical value =
the connement breaks down and a new phase appears. The meaning of this
phase was unknown but it was clear the critical value could never be observed
in the universe. It was therefore considered a very uninteresting mathematical
artifact of some particular QCD model. But, when we apply this uninteresting
`artifact' to the eld theory of Anderson localization 23 , it does exactly what this
theory couldn't describe to be happening for the quantum Hall eect! Here, in
the very `realistic' situation of the QHE, the theta angle obtains a clear physical
meaning: it is the unrenormalized Hall conductance xy itself. The action of
the modern Anderson model including its topological properties is then written
as 24
xx xy
S= 2
drtr(Q) + drtr[Q(Q Q)] (2.2)
8 8
The rst part including the longitudinal conductance xx is the same as in
equation 2.1(no magnetic eld). The integration corresponding to the theta
angle is an integration over the eld variables of a modern version of 't Hooft's
instanton vacuum, called the theta vacuum. Since the theta vacuum is topolog-
ical in origin, any continuous deformations around the instanton eld will leave
it invariant. The unconventional renormalization procedure of the theta angle
that encompasses the non-perturbative sectors of the original Anderson theory
is originally invented by 't Hooft, due to the concept of instantons.
The interesting properties of the laboratory system at criticality are then de-
scribed by a topological picture of the instanton vacuum. According to Pruisken,
the topological theta vacuum picture can be used to describe the phase transi-
tion that appears in the 2D quantum electron system.

2.7 Importance of the Edge

Without the need to understand the QHE with Anderson's localization theory,
the important features of the instanton vacuum didn't get uncovered by the
physics community. Early attempts 2527 to understand the scaling behavior
of the instanton vacuum failed because the topology of the system was not
appropriately included. This led to the idea that 't Hooft's instanton approach
wasn't interesting in the eld theory of QCD.

17
The reasoning was globaly that by considering the large N expansion of the
CP N 1 model, which is a also a SU (N ) non-linear sigma model, the theory be-
comes exactly solvable and a certain electromagnetic eld appears (in analogy
to the Hall current). But the electromagnetic eld that was obtained is a free
eld with a charge that can take any value and an energy that is non-periodic
in theta. In the Hall context this would mean that the Hall conductivity is
unquantized and with no phase transitions. In the QCD context this would
mean that no phase transitions can occur and that the theta angle is just an
extra parameter which doesn't change the fundamental properties of the quark
system. In this sense, the entire QHE didn't appear in the sigma model. To
explain this, it was argued that the QHE would have to be a very specialized
feature of the theta angle concept, which only appears after analytic continua-
tion, by replica symmetry breaking. The unknown replica symmetry breaking
mechanism was supposed to introduce extra parameters in the theory allowing
the systems criticallity. Because the whole situation could now be explained,
the theta angle wasn't supposed to provide any fundamental information about
the instanton vacuum or QCD. But (and that is a very big `but'), by introduc-
ing the boundary of the system, Pruisken was able to show explicitly 24 that
for any replica limit (also the large N expansion of the CP N 1 model) the 2D
sigma model displays the same scaling behavior as observed in the QHE. This
should have fundamentally altered the physics communities vision on the basic
properties of the (in)famous non-linear sigma model.

2.8 General Features are Super-Universal

For the general action given by equation 2.2, the topological term can be rewrit-
ten as a one-dimensional integral over the boundary of the system. It turns out 28
that the instanton vacuum generically displays massless chiral edge excitations.
The eld congurations at the edge classify topologically dierent sectors. The
bulk variables of the system can be integrated out for each topological sector.
Hereby, an eective action for the edge can be formulated, allowing the one-
dimensional edge to be renormalized separately. This edge is the edge of the
bulk of the system. In a 2D simply connected system such an edge is topologi-
cally equivalent to a circle. The edge itself as an object is obviously a non-simply
connected system. It isn't an extra `shell' around the bulk, or any other sepa-
rate channel, but it is the natural edge of the eective system (quantum eld).
Renormalizaton by instantons of the critical action for the bulk-edge provides
us with scaling functions that can be combined in a general ow diagram. This
ow diagram consists of xed points that are determined by the topology of the
edge. For the quantum Hall eect, this general ow diagram actually explains
the staircase picture (gure 1.1)! On larger scales, conductances renormalize
to the attractive xed points (that are in the strong coupling regime) and the
Hall plateaus are formed (see gure 2.3). The robustness of these plateaus is
thus an emergent property of the large-scale limit. The plateaus don't `blur' be-
cause of the disorder as would be the case with energy levels. It is the opposite
that is true. If the disorder correlation length becomes smaller, we are driven

18
closer to the attractive xed points, enhancing the precision of the plateaus1 .
Conclusively, because of the nontrivial topology of the systems edge, the scale
invariance property of the system becomes detectable in the macroscopic Hall
conductance. The scaling behavior thereby captures all the basic features of the
quantum Hall eect.

Figure 2.3: Flow diagram in the plane of conductances, with H in units of


e/h. As the system ows to the strong coupling regime, the Hall conductance
aquires a quantized value.

It is established 29 that the structure of xed points doesn't apply speci-


cally to the QHE, but that it is a general future of the instanton vacuum. As
mentioned, the non-linear sigma model with arbitrary number of eld compo-
nents is a very general theory about spontaneous symmetry breaking. Since it
is found that the number of eld components don't alter the scaling functions,
it is even more exciting that the xed point structure can be directly measured
in a quantum Hall experiment. Particular realizations of the instanton vacuum
with a certain number of eld components correspond to systems in various
universality classes, with dierent critical exponents. The general features of
the theta vacuum apply to all these classes and are thus called `super-universal'.
The most notable super-universal features are: massless chiral edge excitations,
the robust quantization of the Hall current (or more generally, the topologi-
cal charge) and the scaling behavior governed by a periodic structure of xed
points.
In short, Pruisken found that, by super-universality, the non-linear sigma
1 There are many kinds of disorder. Here we refer to disorder with a small correlation

length. This is like comparing a polished surface with a rough one.

19
model with theta angle describes and explains the QHE. Because it now doesn't
matter which M, N are chosen in the SU (M +N ) model, we can take M = N 1
and N (the large N expansion) to obtain the same general features as
in the replica limit M, N 0. This is a great advantage, because the CP N 1
model is exactly solvable and thereby we have a very unexpected model that
is able to describe the physics of the theta vacuum from the strong coupling
side. At criticality, the electron gas delocalizes and the edge excitations become
relevant. Therefore, the symmetry of the theta vacuum is broken and multiple
phases appear. In the strong coupling regime (the true quantum regime) the
phases and their transitions become visible!
In gure 2.3, we see how the conductances renormalize from weak to strong
coupling. The periodicity in theta (in units of = , = 2 , etc..) is an exact
result of the CP N 1 model. The staircase picture (gure 1.1) can be understood
as follows. When we start with any non-critical value for the unrenormalized
Hall conductance, the renormalized or measured Hall conductance takes exact
values, independent of sample-geometry. The electrons here are strongly local-
ized except for massless chiral edge excitations that form the Hall conductance.
At criticality, the unrenormalized conductances experience the inuence of a
saddle-point ( = 12 n) and we have a true quantum phase transition. Such a
transition is characterized by a unique critical exponent that can be measured in
the laboratory. The value of this critical exponent is not a super-universal fea-
ture of the instanton vacuum. In Pruisken's modern work, the critical aspects
of the delocalization mechanism are being investigated. Because of the large
sensitivity of the critical system, even the innitely long ranged Coulomb in-
teractions become relevant. The Fermi-liquid approximation is then inaccurate
making the theory extremely complicated.

3 Experiments on Critical Behavior


3.1 Introduction

In collaboration with Pruisken, H.P. Wei 30;31 investigated the scaling behavior
of the QHE. By lowering the temperature, the system scales to larger eective
system sizes and a ow diagram in accordance with gure 2.3 was observed. Af-
ter more precise measurements 10 Wei observed power law scaling and universal
exponent values for the rst three plateau transitions. This was in agreement
with the scaling behavior that was derived by Pruisken 32 on the basis of the
renormalization group theory in the presence of a theta angle. Eventually, the
measurements demonstrated that it is possible to observe true quantum criti-
cal behavior in the QHE and hence to research the instanton vacuum and its
super-universal aspects experimentally.

20
3.2 Quantum Hall Laboratory

It is not an easy task to perform measurements on criticality. The narrow


regime approaching self-similarity comes with an exponential sensitivity for the
observables. Any critical exponent has to be tted on a log-log plot requir-
ing a window that spans decades of the involved physical quantities (instead
of multiples) if it is to be unambiguously determined. The quantum Hall ef-
fect is actually only a very rare occasion where the theory of continuous phase
transitions can even be tested. For example, measuring the critical exponent of
a normal liquid-gas transition involves a trip to outer space, just because the
earth gravitational eld introduces a small density gradient enough to make the
critical point vary too signicantly. A similar situation happens in the QHE,
when there are macroscopic inhomogeneities that cause variations in the elec-
tron density over the sample (in the lowest phase transition of the QHE there
is a clever way to avoid this problem 33 ). If the width of the uncertainty in the
critical point's value isn't much smaller than the width of the phase transition
itself, criticality cannot be observed.
The accessibility of the scaling regime then depends sensitively on the sam-
ple choice. In the past decades, a lot of technological progress has been made
in the fabrication of all kinds of samples. There are very clean materials and
samples with a high electron mobility, but also samples with controlled impu-
rity additions and samples with a more homogeneous electron density. As we
have seen, the correlation length of the disorder potential always plays a crucial
role in observing quantum scaling phenomena 34 . If the disorder is too smooth,
one observes traces of the percolation mechanism (see gure 1.5) inuencing the
critical exponent. This is the case for a GaAs-AlGaAs heterostructure where the
ionized impurities are in the opposite layer, away from the two-dimensional elec-
tron gas and therefore inducing a smooth disorder potential landscape. Indeed
the measurements on scaling that were being performed on the GaAs-AlGaAs
heterostructure, yielded non-universal critical exponents 35 or even the absence
of scaling 36 . On the opposite side, when the disorder correlation length is too
small, the states are not suciently localized to observe them de localize. In
both cases, very large eective system sizes, at inaccessibly low temperatures,
are needed to obtain true quantum criticallity.
A modern material, graphene, has such a high electron mobility that the
quantum Hall eect can even be measured at room temperature. It would be
a great opportunity to measure the critical behavior from room temperature
down to millikelvins, which is a truly huge temperature window. But again this
is going to be dicult, if not impossible. At high temperatures, the Landau
levels in graphene are valley- as well as spin-degenerate, which means that we
would be measuring transitions between two or four levels simultaneously. The
degeneracy alters the scaling behavior in an uncontrollable fashion. Only at low
temperatures thermal excitations between the levels are rare enough to study
phase transitions independently.
In the Quantum Hall laboratory, strong coupling problems can be researched
from a unique perspective. It might seem strange that such a large, `dirty' sys-

21
tem provides such `pure' and fundamental information. But also in the classical
regime we see a surprisingly universal result arising from the Hall experiment.
Over a relatively large range of magnetic eld strengths, the longitudinal re-
sistance is independent of the magnetic eld up to a precision of a few parts
per-mil 33 . This means that in the xx , xy plane, the experimental data exactly
follows a semi-circle. For the lowest electron density, we see in the inset of gure
3.1 that the system is departing the semi-circle for lower temperatures and is
owing to the strong localization regime, where eventually the quantum Hall
eect appears (which is not at all visible on this scale).

Figure 3.1: In the classical regime, the conductivities follow a semi-circle with
great precision. Inset: the colder system departs from the semi-circle. (Image
courtesy, Pruisken et al. 33 )

3.3 The Critical Exponent

In 1988, After testing many samples, H.P. Wei's choice to measure scaling in
the quantum regime, fell on an InGaAs-InP sample where electron scattering
occurs due to the dierent atoms in the alloy. It was therefore expected that the
disorder correlation length was of the order of the lattice spacing. To extract
the critical exponent, temperature scaling of the eective sample size was used.
The following steps that were derived by Pruisken 32 , illustrate how the critical
exponent can be identied. Here, we can see how renormalization group theory
points out the critical exponents in the phase transitions of the QHE.
The eld theory of Anderson localization implies that the scaling dependence
of the conductance is a universal function F of only the conductance itself.
The Anderson theory with theta angle implies this for the combination of the
longitudinal conductance 0 and the Hall conductance H . Near a xed point,

22
the conductances scale algebraically with length. In general, the environment of
the xed point is nonlinear and we must work within curvilinear coordinates 37 .
Suppose we have a starting point for scaling L0 and we denote the corresponding
conductances as:
H0
H (L0 , B) (3.1)

00 0 (L0 , B)
(Where H 0
can be identied as the lling fraction of the Landau band) Then
near the xed point ( 12 , 0 ) we can express the curvilinear coordinates (H , 0 )
as Taylor series around the xed point:
 
1 1 2
0 0 0
H = (H ) 1 + (0 0 ) + (H ) + ... (3.2)
2 2
 
1 2
0 = (00 0 ) 0 0
1 + (0 0 ) + (H ) + ...
2
(Where the expansion is dierentiable). As in the linear case, these nonlinear
coordinates also scale algebraically with length in the large scale limit 37 . Thus,
if we change the length scale L0 L we will have:
 yH
L
H H (3.3)
L0
 y0
L
0 0
L0
Which allows us to conveniently dene the scaling elds, that are the renormal-
ized coordinates (elds):  yH
L
X H (3.4)
L0
 y0
L
Y 0
L0
We can then express the macroscopic or measured conductances in terms of
universal scaling functions FH,0 :

H (B, L) = FH (X, Y ) (3.5)

0 (B, L) = F0 (X, Y )
The experimentally measured resistances RH,0 also depend on these scaling
variables since they are directly related to the conductances by matrix inversion.
We can thus dierentiate them with respect to the magnetic eld:
RH,0 RH,0 X RH,0 Y
= + (3.6)
B X B Y B

23
Because Y scales as an irrelevant parameter, we take Y 0 as L while
X is xed. Here, B enters the equations via the starting points for scaling, the
bare conductances 00 , H
0
. Also derivatives renormalize and we can again apply
the renormalization trick (including an arbitrary function f ):
 yH  yH  yH
X L L L
(B, L) = f( X) = f (0) (3.7)
B L0 L0 L0

Where in the last equality we used L . If we now analyze at the maximum


slope of RH (which is not the critical point):
 max  yH  max
RH L RH
= (X)f (0) (3.8)
B L0 X

The expression between square brackets on the RHS does not depend on B or
L anymore, because X is at a xed value that ensures the maximum (and B
and L enter the equations only via X ).
For R0 , the maximum slope is related to the half width of the visible peak
(see gure 1.1), which is approximately Gaussian 32 . This gives a width, 4B ,
and we can write:
 yH  1
B L X
4B 4H = (3.9)
H L0 B
And, as in equation 3.7, the expression between square brackets doesn't depend
on B or L.
The eective sample size scales with temperature provided Lef f  L and
we can substitute temperatures for lengths to obtain:
 max  
RH T
(3.10)
B T0
 
T
4B
T0
We therefore have two independent measurements that give the value of the
critical exponent . For a list of dierent temperatures we can sweep B over the
phase transition, nd the maximum and measure the critical exponent without
knowing the location of the critical point. This provides access to the scaling
regime and allows the observation and proof of criticality.
H.P. Wei was able to measure a critical exponent = 0.42 10 that was the
same in the transitions at the centers of the rst three Landau levels. The
general way in which scaling theory introduces itself in a model implies in this
case that the critical exponent is the same for the longitudinal as well as the
Hall resistance. Luckily the mixing of data that appears because the resistances
are taken from dierent parts of the sample doesn't spoil the opportunity to
measure a critical exponent.

24
Over a decade later, de Visser et al. 38 preformed measurements on the same
InGaAs-InP heterostructure but found = 0.57 for the critical exponent. A
calculation on mesoscopic level 33 could explain the dierence with the value =
0.42 due to electron density inhomogeneities. Those density inhomogeneities
introduce a kind of disorder with a macroscopic localization lenght and therefore
they have to be treated in an entirely dierent fashion. The critical exponent
value that H.P. Wei was measuring had a lower value because he was measuring
an eective exponent.
But still, the true critical exponent wasn't found. In very recent years,
the Princeton group 11 measured the exponents on state-of-the-art Alx Ga1-x As-
AlGaAs samples with a controllable alloy scattering and very little sample in-
homogeneities. Adjusting the alloy, they found critical exponents between the
values 0.42 and 0.58. Crucially, the data of the dierent alloys all converged to
= 0.42 (the rst `H.P. Wei value') for the zero temperature limit. With this
experiment we can clearly observe that if the temperature is lowered to make
the eective sample size large enough, the true quantum critical behavior of the
system becomes visible and the critical exponent value becomes unambiguous
(see gure 1.5).
Conclusively, De Visser et al. measured a larger value = 0.57 in the
InGaAs-InP sample. This had to be accounted for by an unforeseen, long ranged
disorder correlation (again a disorder of a dierent kind). The fact that H.P.
Wei found a value already in 1988 that was equal to the value of the latest
measurements on scaling in the QHE today might seem mysterious, but in
physics we must consider it a complete coincidence...

3.4 Fermi-liquid principles

It is standard in scaling experiments on the QHE, to interpret the correlation


length exponent using Fermi liquid ideas (or free electron theory). As we have
seen in subsection 3.3, the canonical correlation length exponent can be ex-
tracted by assuming nitely ranged Coulomb interactions. We assume that
the only eect of the Coulomb interactions is to dephase the electrons after
they collide. The radius of the electrons quantum coherent eld is then lim-
ited by its inelastic scattering behavior. The inelastic scattering length has to
be much larger than the localization length for Anderson theory to apply. If
this is the case, the length scale with which the measured conductance scales is
well-determined by an eective system size. This eective system size induced
by Coulomb scattering scales with temperature. The critical scaling exponent
that can be associated is called the inelastic scattering exponent. Unfortunately
this exponent cannot be calculated, but has to be taken as a phenomenological
quantity.
The assumption of an eective system size is generally believed to be valid in
the regime of strong localization. We can now identify the critical exponent as
the ratio of the inelastic scattering length exponent p (how length relates to tem-
perature) and the localization (correlation) length exponent = 1/yH and write
= p/2 . In a modern context, the dephasing mechanism is called dynamical

25
screening and a dynamical critical exponent z is used to describe how length
relates to time instead of temperature. Because the time between electron-
electron collisions is inversely proportional to the temperature, the measured
critical exponent is identied as = 1/z .
The canonical correlation length exponent was calculated numerically 39;40
as well as analytically 41 and values were found between = 2.3 2.4. This was
considered to be in agreement (or at least not in dis agreement) with experiments
on nite size scaling 42 and current scaling 43 , where can be measured directly.
Because we have = 1/z , agreement of the calculated value = 2.3 2.4 with
the well-known experimental value = 0.42 required z 1. But if we assume
that Coulomb interactions only dephase electrons, z = 2 is the appropriate `free
electron' value to work with. Then, a number of calculations on the dynamical
dimension of interacting electrons 44;45 suggested that it is possible to put z 1,
while maintaining the Fermi-liquid properties of free electrons. Therefore =
2.3 2.4 was accepted for the quantum Hall regime and Fermi-liquid principles
(as well as the percolation picture) remained to be pursued in the context of
critical behavior in the QHE.
More recently, in 2009, Li et al. 46 published new results (see gure 1.5) on
scaling in the quantum Hall regime using state-of-the-art GaAs-AlGaAs sam-
ples with a controllable alloy scattering and very little sample inhomogeneities.
Measuring = 0.42 and assuming z 1, they obtained = 2.4 conrming the
original, accepted value for correlation length exponent in the Quantum Hall
regime. However, in the same year, the correlation length exponent was calcu-
lated again in a Chalker-Coddington network model 47 and it was demonstrated
that in fact = 2.6. We must have z > 1 48 , so when using = 1/z , the
value = 2.6 is in disagreement with the value = 0.42. This indicates that
the theory of free electrons is not sucient when the system becomes quantum
critical.

4 Other issues
4.1 Coulomb Interactions

The quantum Hall eect cannot be in the Fermi-liquid universality class and
innitely ranged Coulomb interactions have to be taken into account. At criti-
cality, the localization length becomes innite and thereby at least comparable
to the inelastic scattering length even for very low temperatures. This is not
in accordance with the assumption of a localization length that is much smaller
than the inelastic scattering length. The reason why the concept of dynamical
screening is going to help us out here remains unclear. If we want to work in a
complete, eld theoretical language we must forget about inelastic scattering or
Coulomb screening in the critical regime and work with a theory including in-
nitely ranged Coulomb interactions. The numerical network simulations that
are based on percolation ideas are calculating a canonical correlation length
exponent that can only be related to the quantum Hall laboratory due to the

26
concept of dynamical screening.
The very dicult task of including interactions was undertaken by Pruisken
and his collaborators. Their starting point was an obscure method used by
Finkel'stein 49 involving eld matrices of innite size. Basic solutions like topo-
logical excitations, nding global symmetries and identifying physical observ-
ables that were obtained in the context of Fermi-liquid theory become mean-
ingless when innite interactions are included. Pruiskens ground work 28;5052 is
the starting point for studying Coulomb interactions in the quantum Hall eect,
and hence for studying its critical regime. In Pruiskens work, the new (quantum
Hall) universality class is called the F invariant universality class.

4.2 Super-Symmetry

Besides constructing a fundamental theory for the quantum Hall eect, it can be
important to invent mathematical toy models that are useful during the dier-
ent phases in the research on the quantum Hall eect. For example, to calculate
the exact value of the critical exponent Zirnbauer 53 tried to construct a discrete
vertex model that falls in the same universality class as the quantum Hall tran-
sition. The super-symmetric versions of the non-linear sigma model with theta
angle and the Chalker-Coddington network model are shown by Zirnbauer to
possess the same global symmetries at criticality. Global symmetries dene a
universality class, so both models must be in the same class. Since the super-
symmetric Chalker-Coddington network model does not t within the usual
framework of integrable models, he needs to construct another vertex model
that is in the same class and integrable. With this vertex model the critical
exponent could be calculated. But, since Zirnbauer used the non-linear sigma
model without Coulomb interactions, this critical exponent is unfortunately still
the Fermi-liquid universality class. Constructing the Zirnbauer vertex model is
therefore unimportant in the context of the quantum Hall eect.
Interestingly, because of the super-symmetric relation between the non-linear
sigma model with theta angle and the Chalker-Coddington vertex model, the
sigma model is a continuum version of the discrete, Chalker-Coddington vertex
model. Since this vertex model is a random hopping model including a phase
dependence, it represents the percolation picture including quantum tunneling.
This explains why semi-classical lattice percolation was so close to explaining
the quantum Hall eects critical behavior even though it didn't make any sense
in the general explanation of the staircase picture. At the Fermi-liquid crit-
ical point, the non-linear sigma model without Coulomb interactions and the
percolation mechanism share the same universality class. By considering only
the critical exponent itself, the semi-classical percolation mechanism appeared
to be a valid substitute for the Anderson localization mechanism, but since
the universality class of the quantum Hall transition is the formerly unknown
F invariant universality class, there is no such valid substitute for studying
Anderson theory in the QHE.

27
5 Summary
The quantum Hall eect has an important and fundamental role in modern
condensed matter physics. Its history of over tree decades involved thorough
theoretical and experimental research. The QHE involves some of the most
interesting features in todays laboratory, like a quantum eld of 2D electrons,
quantum magnetization, phase transitions in a random potential and edge cur-
rents around an insulating bulk system. Semi-classical arguments like perco-
lation of electronic states, skipping orbits resulting in conducting edge states
and a Fermi-liquid approximation for the critical exponent can never provide
a complete, theoretically justied framework. This is a general problem in the
area of topological insulators.
For the QHE, the modern picture is based on the notion of instantons in
the theory of Anderson localization. The non-linear sigma model with M + N
components describes the 2D disordered electron gas. With the renomaliza-
tion group theory, Anderson localization can be proven, including the fact that,
without a magnetic eld, `All 2D states are localized'. The de -localization mech-
anism is constructed by including the theta angle, a topological property of 2D
electrons in a magnetic eld. Renormalization by instantons of the theta angle
provides an eective theory for the edge of the system. This edge-theory has
super-universal features (massless chiral edge excitations, robust quantization
of the Hall conductance, etc..) that explain the quantum Hall eect.
The calculation of the exact value of the critical exponent remains very di-
cult. In the critical regime, the Fermi-liquid approximation becomes invalid and
hence electrons do not behave like free particles that scatter o each other after
traveling their inelastic scattering length. In the future, a lot of interesting the-
oretical work can be done on including innitely ranged Coulomb interactions
in the non-linear sigma model. Experimentally, there are many opportunities
to continue on the QHE, like the QHE in graphene and studying other topo-
logical insulators. The critical behavior of such mesoscopic quantum devices is
in my opinion the most interesting phenomenon in all of our condensed matter
laboratories.

28
References
[1] K. von Klitzing, G. Dorda, and M. Pepper. New method for high-accuracy
determination of the ne-structure constant based on quantized Hall resis-
tance. Physical Review Letters, 45(6):494497, 1980.

[2] H. Levine, S.B. Libby, and A.M.M. Pruisken. Theory of the quantized Hall
eect (iii). Nuclear Physics B, 240:7190, 1984.

[3] H. Levine and S.B. Libby. Theory of the quantized Hall eect (I). Nuclear
Physics B, 142(1):147154, 1984.
[4] S.B. Libby and A.M.M. Pruisken. Theory of the quantized Hall eect (ii).
Nuclear Physics B, 240:4970, 1984.
[5] R.B. Laughlin. Quantized Hall conductivity in two dimensions. Physical
Review B, 23(10):5632, 1981.
[6] B. Halperin. Quantized Hall conductance, current-carrying edge states, and
the existence of extended states in a two-dimensional disordered potential.
Physical Review B, 25(4):21852190, 1982.
[7] M. Bttiker. Absence of backscattering in the quantum Hall eect in mul-
tiprobe conductors. Physical Review B, 38(14):9375, 1988.

[8] R.F. Kazarinov and S. Luryi. Quantum percolation and quantization of Hall
resistance in two-dimensional electron gas. Physical Review B, 25(12):7626,
1982.

[9] S. Luryi and R.F. Kazarinov. Theory of quantized Hall eect at low tem-
peratures. Physical Review B, 27(2):1386, 1983.

[10] H.P. Wei, D.C. Tsui, M.A. Paalanen, and A.M.M. Pruisken. Experiments
on delocalization and university in the integral quantum Hall eect. Phys-
ical review letters, 61(11):12941296, 1988.
[11] W. Li, W. Pan, D.C. Tsui, L.N. Pfeier, and K.W. West. Crossover from
the nonuniversal scaling regime to the universal scaling regime in quantum
Hall plateau transitions. Physical Review B, 81(3):033305, 2010.

[12] P.W. Anderson. Absence of diusion in certain random lattices. Physical


Review, 109(5):14921505, 1958.
[13] E. Abrahams, P.W. Anderson, D.C. Licciardello, and T.V. Ramakrishnan.
Scaling theory of localization: Absence of quantum diusion in two dimen-
sions. Physical Review Letters, 42(10):673676, 1979.

[14] J.T. Edwards and D.J. Thouless. Numerical studies of localization in dis-
ordered systems. Journal of Physics C: Solid State Physics, 5(8):807820,
1972.

29
[15] E. Abrahams. Scaling at the metal-insulator transition in two dimensions.
Physica E: Low-dimensional Systems and Nanostructures, 3(1-3):6978,
1998.

[16] R.E. Prange and R. Joynt. Conduction in a strong eld in two dimensions:
The quantum Hall eect. Physical Review B, 25(4):2943, 1982.

[17] S.M. Apenko. The quantised Hall eect in strong magnetic elds. Journal
of Physics C: Solid State Physics, 18:11971203, 1985.
[18] M.H. Cohen and J. Jortner. Inhomogeneous transport regime in disordered
materials. Physical Review Letters, 30(15):699702, 1973.

[19] D.J. Thouless. Electrons in disordered systems and the theory of localiza-
tion. Physics Reports, 13(3):93142, 1974.

[20] N. Mott. Transport in Disordered Materials. Physical Review Letters,


31(7):466467, 1973.

[21] D.J. Thouless. Theory of the quantized Hall eect. Surface Science,
142:147154, 1984.

[22] L. Schfer and F.J. Wegner. Disordered system with n orbitals per site: La-
grange formulation, hyperbolic symmetry, and goldstone modes. Zeitschrift
fr Physik B: Condensed Matter, 38(2):113126, 1980.
[23] H. Levine and S.B. Libby. Electron delocalization by a magnetic eld in
two dimensions. Physical review letters, 51(20):19151918, 1983.

[24] A.M.M. Pruisken. Super universality of the quantum Hall eect and the
"large N picture" of the theta angle. International Journal of Theoretical
Physics, 48(6):17361765, 2009.
[25] I. Aeck. Testing the instanton method. Physics Letters B, 92(1-2):149
152, 1980.

[26] I. Aeck. The role of instantons in scale-invariant gauge theories. Nuclear


Physics B, 162:461477, 1980.
[27] E. Witten. Instatons, the quark model, and the 1/N expansion. Nuclear
Physics B, 149(2):285320, 1979.
[28] A.M.M. Pruisken and I.S. Burmistrov. Theta renormalization, electron-
electron interactions and super universality in the quantum Hall regime.
Annals of Physics, 322(6):12651334, 2007.
[29] A.M.M. Pruisken, R. Shankar, and N. Surendran. General topological fea-
tures and instanton vacuum in quantum Hall and spin liquids. Physical
Review B, 72(3):035329, 2005.

30
[30] H.P. Wei and A.M.M. Pruisken. An experimental test of two-parameter
scaling in the integral quantum Hall eect. Surface Science, 170:238242,
1986.
[31] H.P. Wei, D.C. Tsui, and A.M.M. Pruisken. Localization and scaling in the
quantum Hall regime. Physical Review B, 33(2):14881491, 1985.
[32] A.M.M. Pruisken. Universal singularities in the integral quantum Hall
eect. Physical review letters, 61(11):12971300, 1988.
[33] B. Karmakar, M.R. Gokhale, A.P. Shah, B.M. Arora, D.T.N. de Lang,
A. de Visser, L.A. Ponomarenko, and A.M.M. Pruisken. The eects of
macroscopic inhomogeneities on the magnetotransport properties of the
electron gas in two dimensions. Physica E: Low-dimensional Systems and
Nanostructures, 24(3-4):187210, 2004.
[34] H.P. Wei, S.Y. Lin, D.C. Tsui, and A.M.M. Pruisken. Eect of long-range
potential uctuations on scaling in the integer quantum Hall eect. Physical
Review B, 45(7):3926, 1992.
[35] K. von Klitzing. Experiments on scaling in heterostructures under quantum
hall conditions. Physical Review B, 43(8):68286831, 1991.
[36] N.Q. Balaban, U. Meirav, and I. Bar-Joseph. Absence of scaling in the
integer quantum Hall eect. Physical Review Letters, 81(22):49674970,
1998.
[37] F. Wegner. Corrections to Scaling Laws. Physical Review B, 5(11):4529
4536, 1972.
[38] R.T. van Schaijk, A. de Visser, S.M. Olsthoorn, H.P. Wei, and A.M.M
Pruisken. Probing the plateau-insulator quantum phase transition in the
quantum hall regime. Physical review letters, 84(7):156770, 2000.
[39] J.T. Chalker and P.D. Coddington. Percolation, quantum tunnelling and
the integer Hall eect. Journal of Physics C: Solid State Physics, 21:2665,
1988.
[40] B. Huckestein and B. Kramer. One-parameter scaling in the lowest Landau
band: Precise determination of the critical behavior of the localization
length. Physical Review Letters, 64(12):14371440, 1990.
[41] I.M. Sokolov and G.V. Mil'nikov. Semiclassical localization in a magnetic
eld. JETP Lett, 48(9):536540, 1988.
[42] S. Koch, R.J. Haug, K. von Klitzing, and K. Ploog. Size-dependent anal-
ysis of the metal-insulator transition in the integral quantum Hall eect.
Physical review letters, 67(7):883886, 1991.
[43] H.P. Wei and L.W. Engel. Current scaling in the integer quantum Hall
eect. Physical Review B, 50(19):609612, 1994.

31
[44] D.H. Lee. Transitions between Hall plateaux and the dimerization transi-
tion of a Hubbard chain. Philosophical Magazine Letters, 73(3):145152,
1996.

[45] S.R. Yang, A. MacDonald, and B. Huckestein. Interactions, localization,


and the integer quantum Hall eect. Physical Review Letters, 74(16):3229
3232, 1995.

[46] W. Li, C.L. Vicente, J.S. Xia, W. Pan, D.C. Tsui, L.N. Pfeier, and K.W.
West. Scaling in plateau-to-plateau transitions: a direct connection of
quantum Hall systems with the Anderson localization model. Physical
Review Letters, 102(21):216801, 2009.
[47] K. Slevin and T. Ohtsuki. Critical exponent for the quantum Hall transi-
tion. Physical Review B, 80(4):041304, 2009.

[48] M.A. Baranov and A.M.M. Pruisken. (Mis-) handling gauge invariance in
the theory of the quantum Hall eect. II. Perturbative results. Physical
Review B, 60(24):1682116837, 1999.
[49] A.M. Finkel'stein. Weak localization and coulomb interaction in disordered
systems. Zeitschrift fur Physik B: Condensed Matter, 56(3):189196, 1984.

[50] A.M.M. Pruisken. Cracking Coulomb interactions in the quantum Hall


regime. Europhysics Letters, 31(9):543548, 1995.

[51] A.M.M. Pruisken, M.A. Baranov, and I.S. Burmistrov. Localization,


Coulomb interaction, topological principles and the quantum Hall eect.
ArXiv: cond-mat/0104387 (unpublished).

[52] A.M.M. Pruisken, M.A. Baranov, and I.S. Burmistrov. The theta vacuum
reveals itself as the fundamental theory of the quantum Hall eect II. The
Coulomb interaction. ArXiv: cond-mat/0206012 (unpublished).

[53] M.R. Zirnbauer. Toward a theory of the integer quantum Hall transition:
Continuum limit of the Chalker-Coddington model. Journal of Mathemat-
ical Physics, 38(4):2007, 1997.

32

You might also like