mesoscopic physics in carbon based devices · mesoscopic physics in carbon based devices i/...
TRANSCRIPT
Mesoscopic Physics in Carbon Based Devices
I/ Mesoscopic Physics and Quantum Transport in Graphene
The fundamental properties of graphene (transport)
Discovered in 2004 by Geim and Novoselov who received the Nobel Prize in 2010
A perfect 2D metal: single sheet of atoms!
Band Structure: a semi‐metal with relativistic particles.
The valence band and the conduction band touch each other at two non‐equivalent points of
the Brillouin’s zone (K & K’). Those points are called “Dirac points”. Very close to K and K’, the
dispersion relation is linear: ∆ ( stands for electrons or holes)
This dispersion relation is typical of massless relativistic particles (for non‐relativistic
particles, ∝ ). Indeed, the energy in relativistic physics is given by
where is the momentum of the particle . Two different limits can be considered:
a. The case of a massive particle ( ≫ )
In this case, the energy can be approximate by:
112
2
As is constant, we recover the usual quadratic dispersion relation ∝ of a massive
particle.
b. The case of a relativistic massless particle ( ≪ )
In this case, the energy can be approximate by:
112
At the smallest order in / , the dispersion relation is linear. This case is similar to the case
of photons. Fundamental differences are nevertheless important to note: in the case of
graphene, the particles are charged, they are fermions and the “speed of light” is about 106
m/s instead of 3.108 m/s for photons.
Very high mobility.
mobilityμ: drift velocity of the electron μ and the conductivity is directly related to the
mobility : μ ⇔ μ ⁄
Metals Semi‐conductors (2DEG) Semi‐conductors (bulk)
Typically, the mobility of graphene samples can be up to 200 000 cm²/Vs at liquid helium
temperature.
Band structure: ambipolar behavior
changing the position of thanks to a gate voltage allows us to measure either electrons or
holes.
II/ Magnetotransport in 2D‐metals: from the Hall effect to the
quantum Hall effect
We will have interest in this part in the magnetotransport properties of massive charged particles in
conventional metals or semiconductors. We start with the classical Hall effect and introduce the
quantum mechanics to explain the Shubnikov‐de Haas oscillations. At higher magnetic fields, we show
that a quantum phase emerges due to the quantum Hall effect (QHE). This regime exhibits purely
quantum transport properties. As we will see later on, massless particles (graphene or topological
insulators) behaves similarly at low magnetic field (in the classical regime of the Hall effect) but show a
different quantum behavior (SdH oscillations as well as QHE) so that such magnetotransport
experiments are a very convenient way to prove the massless nature of charge carriers.
1. Classical Hall effect
Dirty 2D‐metal embedded in a low magnetic field. Deviation of the charge carriers due to the Lorenz
force: ∧ ∧ .
→ Charge non‐equilibrium that induces a transverse electric field which counteracts the effect of
the Lorenz Force once the steady state is reached: ⇒ ∧ .
We have then:
∙ ∧ ∙
∙
We can define the Hall coefficient
Remarks about (or ):
=107 S/m =3.10‐2 S/□ =10‐6 ‐104 S/m
n=1029 m‐3 n=1‐2.1011 cm‐2 n=1015‐1020 cm‐3
μ=6.10‐4 m2/Vs μ up to 3,5.103 m2/Vs @4K μ=10‐2 m2/Vs
‐ It does not depend on the geometry of the sample
‐ It is a direct way to measure the charge carrier density
‐ It gives the sign of the charge carrier (test of the ambipolar behavior of graphene or Tis)
2. Hall effect in Graphene
Graphene is a semi‐metal that shows an ambipolar behavior when a sufficient gate voltage is applied.
Indeed, the gate effect make possible the tuning of the Fermi energy that can move from the conduction
band to the valence band or the opposite. It is then possible to study either the transport of electrons
or holes and to switch on a single sample from the one to the other by choosing adequately the gate
voltage.
Fig. 1 Gate effect in a graphene sheet: applying a gate effect allow us to tune the position of the Fermi energy in the conduction band or the valence band
Using standard fabrication technics, it is possible to pattern a Hall bar and to measure the classical Hall
effect.
Experimental method for producing a sample.
For a given magnetic field, the Hall voltage is given by and depends on the gate voltage
through the charge carrier density . It is judicious to plot the charge density with
respect to the gate voltage. A change in the sign of (or of ) is a clear indication that changed
its sign, namely that the type of charge carrier changed.
Fig 2 Hall bar patterned in a graphene sheet and measurement of Hall effect that shows a ambipolar behavior (from Novoselov et al.)
3. Case of an infinite 2D‐metal with no disorder (massive fermions)
→ Classical deriva on
Cyclotron orbitals of the charge (no disorder).
∧ with
⇒ 0 where is the cyclotron frequency
The charge motion describes orbitals so that in the case of a non‐disordered material, the switching of
the magnetic field leads to a transition to an insulating state (I=0)!
The radius of the orbitals is given by the cyclotron radius:
cossin
⁄ sin ⁄ cos 1
We distinguish a clean metal from a disordered one by the ability for charge carriers under magnetic
field to form orbitals without to be scattered. A clean metal corresponds to the limit . A non‐
intuitive consequence of this model is that the introduction of disorder restores the conductance!
Two ways to reach this regime : to improve the material properties (longer ) or to use high magnetic
field (up to 80T!)
4. Landau Levels and Shubnikov‐de Haas oscillations
To solve the problem of a clean metal under magnetic field, we need to introduce quantum mechanics.
First thing to do is to solve the Hamiltonian and to find the Eigen energies.
i) Naïve derivation of the Landau Levels in a semi‐classical approach In quantum mechanics, the charge carrier is represented by a complex wave function that has a
phase . Considering the latter case of charge carriers of a 2D metals with no disorder embedded in
a magnetic field, a semi‐classical approach would consist in keeping the classical trajectories and adding
a quantum phase that accounts for the circulation along the cyclotron orbital :
∮ ∙
In the presence of a magnetic field, the total phase acquired by a quasi‐particle will also be given by the
circulation of the potential vector and we end up with
/ ∙
According to Onsager and to the Bohr‐Sommerfeld quantification rules, the phase acquired on a closed
cyclotron orbital should satisfy: 2 with ∈ and ∈ . As the scalar products ∙
and ∙ are constant all along the orbital, we end up for an electron ( ) with:
/2 2 2
With ⁄ ⁄ ⁄
Finally, we have
/2 ⁄
2⁄
2⁄
It is not possible to determine the value of in a semi‐classical approach but the quantum mechanics
tell us that 1/2.
In the semi‐classical approach, we consider that the magnetic field only change the trajectories and not
the general expression of the Hamiltonian so that we have to consider the case of a Fermi sea with
massive free electrons. As a result, the energy of a massive charge coupled to a magnetic field will be
still given (only in our semi‐classical approach) by 2⁄ . We finally find the Landau solution
of the Hamiltonian for a free electron embedded in a magnetic field:
12
The consequence of switching on the magnetic field will be the quantification of the energy.
This rough approach reveals the underlying physics and gives the key ingredients to understand the
physics of the magnetotransport in a non‐disordered material: due to interferences along orbitals, we
open gaps in the density of states. In a similar way than for the classical case, a pure material is a material
for which the disorder is low enough (or the magnetic field high enough) to allow the formation of
cyclotron orbitals .
ii) Quantum mechanics treatment
In a general way, we have: | | which can be simplified if 0. We have then
| | .
For a Fermi sea of massive Fermions without magnetic field, . If the magnetic field ∧
is turned on, one have to do the Peierl’s substitution: ↔ with . The
Hamiltonian that has to be solved in the case of electrons is:
2|
2| |
This is a classical problem of the quantum mechanics. This have been solved by Lev Landau in 1930 (he
was 21 years old!). As suggested by the semi‐classical model, gaps open in the energy spectrum and
only some discrete values of the energy will be allowed:
The major achievement of the quantum mechanics is to show the existence of a zero point energy
that is a direct consequence of the Heisenberg inequality∆ ∆ ~ . It is a fundamental property
and we will see that this zero energy points is not present for massless Dirac Fermions like graphene
quasi‐particles!
Fig. 1 Density of states at 2D for a zero magnetic field, for a non‐disordered 2D metal and for a disordered 2D‐metal
When the magnetic field is high enough to open some gaps into the DOS, the transport properties are
determined by the position of the Fermi energy (for an infinite sample, we have the relation
/ ):
‐ If lies between to Landau Levels (LL), the material will be insulating as predicted by the
classical model
‐ If is pined on a LL, the density of states is ≠ 0 at so that transport should occurs.
Nevertheless, even an infinitesimal disorder will localize those states such that 0. The
sample will also be in an insulating state!
In the case of an infinite 2D metal (with no edges) with an infinitesimal disorder, the switching of a
magnetic field induces a transition to an insulating state as predicted by the classical approach!
This simple picture does not take into account the effect of the edges (in the QHE regime, very
important!) as well as the effect of the disorder that induces broadening of the LL. In the regime where
the broadening induced by the disorder is larger than the energy spacing between two LLs ∆ ,
magnetic field is not strong enough to open gaps into the energy spectrum and the DoS does not vanish
for any energies : as in the classical case, the disorder restores the conduction into the bulk!
Nevertheless, the magnetic field induces a fluctuation of the DOS which exhibits maximum at energies
corresponding to a LL as soon as the magnetic field is strong enough to close the cyclotron orbitals
≲ . Such a fluctuating DOS has some strong implication for the transport properties of a the
metals: for a given sample, the position of the Fermi energy is fixed by the doping level in the sample.
Sweeping the magnetic field shifts the position of the LL with respect to since ∝ and the
fluctuates as well as the conductance of the sample. A minimum of the conductance (at a
magnetic field ) corresponds to the crossing of a LL.
Such oscillations, called Shubnikov‐de Haas (SdH) oscillations, are very useful to determine the charge
carrier density of a 2D metal (thanks to the value of the different ), the effective mass (the thermal
dependence of ∆ )and the mean free path (the dependence of with respect to the
magnetic field or the onset of the oscillations that corresponds to ≲ ⇔ ≲ ⇔ ≲ ).
iii) Temperature dependence of SdHO Not only the mobility is required to be good for the observation of SdH oscillations ( ≳ 1). The
temperature needs also to be low enough: the inequality: ≳ has to be satified. For usual
materials such as GaAs, this means that ≲ 10 . For graphene, the effective mass is much smaller and
→ 0 so that the cyclotron energy is much larger and the temperature requirement for the observation
of SdH oscillations becomes ≲ 1000 : SdHO can be observed at room temperature in graphene!
The temperature dependence of the SdHO gives a direct access to the cyclotron mass that is equivalent
to the effective mass in the case of massive fermions.
Fig. 2 Example of SdH oscillations measured in GaAs based heterostructures at low temperature (from Malcoff et al.)
iv) SdHO for massless fermions In a very general case, it can be shown that the parameter that determines the temperature evolution
of the SdH oscillations will be the cyclotron mass that is expressed, in the case of masseless fermions
as:
For massive particles, the energy is dominated by the term which leads to an that is
independent on the charge carrier density.
For relativistic massless particles in graphene, we have:
√√
The cyclotron mass depends on the charge carrier density in the case of massless fermions and is
independent on then charge carrier density for massive charge carriers. The measurement of the
temperature dependence of the SdH oscillations and the study of the density dependence of the
cyclotron mass gives an unambiguous proof of the nature (classical or relativistic) of the charge carriers
that are involved in the transport properties of a sample.
Fig. 3 Example of the measurement of the cyclotron mass in a graphene sheet. The mass is found to depend on the charge carrier density n like n1/2 which an unambiguous proof that charge carriers in graphene are massless (from Novoselov et al.)
5. The quantum Hall effect in non‐relativistic fermions (electrons)
We have seen that the classical model as well as the quantum mechanics expect both an insulating
transition when the magnetic field is turned on for an infinite 2D metal with no disorder. As we will see,
the introduction of edges drastically changes the conclusion of both models.
i) Classical model of a finite (with edges) 2D‐metal with no disorder
Fig. 1 Formation of chiral classical edge channels
In the classical model, the “bulk” remains insulating but we have the formation of conducting stripes
which width is about . Those channels should not be back‐scattered and should perfectly conduct the
current. Very importantly, the transport is chiral (see the above figure 4).
ii) Quantum mechanics
The introduction of edges in the sample can be modeled in the Hamiltonian by the introduction of an
infinite repulsive potential outside of the sample in order to retain the charge carriers into the sample.
We assume here that this potential is adiabatic: its effect is simply to shift the position of the Eigen
energies depending on the position considered in the sample. This is a good assumption if the potential
changes slowly at the scale of .
2⇒
12
Into the bulk of the sample, we should have no influence of the edges: we have the formation of the
previous LLs. Close to the edges, those LLs are shifted up such that even if the Fermi energy lies in a
gap into the bulk of the sample, it crosses a LL somewhere close to the edge. It can be shown that close
to the edges, the LLs are not localized states anymore such that the introduction of edges induces a
chiral conduction as predicted by the classical model. Those states, called “edge states” or “edge
channels” have fascinating properties:
‐ They are purely ballistic (ideal conductors).
‐ They are chiral
‐ They are purely 1D channels
The number of edge channels is given by the number of LLs filled into the bulk of the sample (filling
factor ν). When the magnetic field increases, the number of edge channels decreases. When the Fermi
energy lies between two LL into the bulk of the sample, the current will be carried by the edge channels
only.
6. Semi‐classical derivation of the conductance of a 1D channel
The determination of the conductance (in a two point configuration) of such an edge channel is
equivalent to the determination of a 1D conductor which transmission 1 (the probability for a charge carrier to from the left to the right or from the right to the left without to be backscattered).
‐ Each contact absorbs all incoming charge carrier (black body model)
‐ Energy relaxations occurs into the contact only (length of the channel )
Fig. 2 Sketch of the principle of the conductance measurement of a 1D channel in a 2 probes configuration
The current is given (at zero temperature) by:
The factor ½ comes from the fact that half of the charge carriers have a positive velocity, the rest having
a negative velocity.
22
With the spin degeneracy. Taking into account the fact that we obtain finally:
2
Which results in
This relation is very general and does not depend on the dispersion relation.
Important remark:
‐ the conductance is not infinite even if for a perfectly transmitted channel
‐ the conductance is related to the fundamental constants e (the charge of an electron) and h
(the Plank constant). It does not depend on the geometry of the sample nor on the charge density!
7. Hall bar in the Quantum Hall effect regime
In this section, we determine the typical voltages (Hall voltage and longitudinal voltage) that are
measured on a Hall bar in the different field regimes (hall regime, Shubnikov‐de Haas regime and QHE
regime).
i) The classical Hall regime
This regime corresponds to ⁄ ≲ 1
ii) The Shubnikov‐de Haas regime
This regime corresponds to ⁄ ≳ 1 and ≪ where is the magnetic
length (see below).
iii) The quantum Hall regime
At 1, exhibits plateaus which value is given by fundamental physics constant (e and h) and
does not depend on the geometry nor on any other parameters of the metal (density, mean free path,
effective mass,…). The corresponding Hall resistance of the plateau gives the Klitzing constant
25812,…Ωwhich defines the quantum of conductance/resistance. Because of its
independence to the various parameters, the quantum Hall effect is used in metrology to determine
the value of .
At , it is possible to show that the Hall voltage is given by . To do so, we need to take into
account the fact that the emission rate of charge carrier of a contact is given by per channel.
Another important point is that in a four probe configuration, the longitudinal resistance of a ballistic
conductor is founded to vanish. This results is in agreement with the common sense, on the contrary to
the case of a two probe measurement. This indicates that the resistance of the 1D perfectly transmitted
channel in a two probe configuration is an access resistance due to the strong reduction of the number
of conducting channels between the macroscopic contact and the 1D conductor.
iv) Experimental results
8. Landau Levels and the quantum Hall effect in Graphene
Considering again the Hall bar of graphene and locking now at the longitudinal resistance at higher
magnetic field ( ≳ 1). In conventional metal, i.e. with massive charge carriers, the temperature
evolution of the SdH oscillations is directly related to the mass of the charge carriers. This is a very
convenient and very conventional way to determine this mass. In the case of graphene, the situation is
completely different since the charge carriers are massless fermions. The quantum Hall effect in
graphene
Increasing further the magnetic field until approaches few , we could expect to enter the
quantum Hall effect regime. Nevertheless this regime is expected to be very different than the one of
massive fermions because of the vanishing of the effective mass of the charge carriers. Indeed, in the
case of massive fermions, the cyclotron frequency is given by → ∞ when → 0. Hence, the
energy of the LLs should diverge:
12 →
∞
There is a breakdown of this model in the case of massless Dirac fermions! It can be shown than in the
relativistic case, the cyclotron frequency can be given by √2 and the cyclotron radius has to be
written in a more general way as that does not directly depend on the mass any more. Using
similar argument than previously, one can try to infer what would be the energy spectrum of massless
Dirac fermions under a magnetic field. Taking into account the fact that the ‐vector is quantized along
cyclotron orbital (⇔ / with ∈ ) and that the dispersion relation at zero magnetic field
reads: | |, we have:
(Wrong formulae!)
Quantum Hall
plateaus
Spin‐
degenerescence
lifted
Figure 3. First observation of the Quantum Hall effect by K. von Klitzing (Physical Review Letters 45, 494 (1980) )
As in the case of massive fermions, this formulae is wrong but gives the key ingredients of the quantum
Hall effect with massless fermions and particularly the fact that ∝ √ (for massive fermions, ∝). Solving the Hamiltonian of massless Dirac fermions under a magnetic field gives the following
eigenvalues :
√
Very important differences with the case of massive particles have to be mentioned:
‐ The LLs are not equally placed in energy
‐ There is a zero energy mode
‐ There is positives (electrons) as well as negatives (holes) solutions
‐ Even if the cyclotron energy in graphene does not→ ∞, it remains much larger than the one of
usual massive particles. Typically, ~1000 at 10 . This make possible the observation
of the quantum Hall effect at room temperature!
Figure 4 Graph LL massive vs Massless
Example of the first experimental observation of such quantum Hall effect for massless Dirac Fermions
(from Novoselo et al.):
As we can see, on contrary to the case of the QHE in GaAs based heterostructures for which the LLs
have just a spin degeneracy (the reason why we have plateaus at multiple of 2 ⁄ instead of ⁄ ),
in the case of graphene, we have a spin degeneracy and a valley degeneracy (two valleys and ′ in the Brillouin zone) which leads to steps of 4 ⁄ .
The LL of zero energy is equally composed of electrons and holes. In a similar picture than the one we
used for electrons where the LLs are shifted at high energy when approching the edges of the sample,
the LLs of holes are shifted at low energy (the holes need to be retained into the sample) and the zero
energy LL will split into two LLs: one for electrons and the other one for holes. Each of those LLs has a
two times lower degeneracy which leads to plateaus of2 ⁄ instead of4 ⁄ . The symmetry
electron‐hole is respected.
Fig. 5 Shift of the LLs of graphene close to the edge of the ample and splitting of the zero energy mode
Not only fundamental research:
Ref :
‐ Mesoscopic Physics : Datta
‐ Graphene : Review of Das Sarma. The nature paper of Novoselov
‐ Quantum Mechanics : Feynman / Landau / Cohen‐Tannoudjii