Coulomb blockade in Quantum Dots

Additional reading complementing the lecture notes.1

Version: 23/04/2007 - Andreas Fuhrer/Carina Fasth

In the resonant tunnelling diode we have treated electrons as non interacting par-
ticles. We have discussed their wave nature, which led to quantized energy states
in small and coherent cavities. Qualitatively this is due to the requirement that an
integer number of Fermi wavelengths has to fit between the barriers.
Here we begin in the opposite limit by neglecting space quantization effects in
terms of self-interference and discussing single electron charging of small metallic
islands. Due to the small Fermi-wavelength in metals the energy spectrum of such
a system is quasi-continuous and the system can be treated classically, except that
due to the quantization of charge an integer number of particles needs to reside on
the island.

Vg

gate

Cg

source dot CΣ drain

Cs Cd
Vsd

Figure 1: Schematic drawing of a typical electrode arrangement for a single

electron transistor. The island has index 0, the source has index 1, the drain
index 2 and the gate index 3. The elements of the capacitance matrix Cij are
numbered accordingly.
1
Text adapted from Ref.[1]

1
Single electron charging

The device which we want to consider is a so-called single electron transistor [

Fig. 1(a)] where a small island with the self-capacitance CΣ is weakly coupled to
source and drain contacts via tunnel barriers. At low enough temperatures and
small bias voltage, the energy cost to add an extra electron onto the island may
exceed the thermal energy (kB T  e2 /CΣ ) and the current through the island is
blocked. This is the Coulomb blockade effect.
It was first suggested in the early 50’s by Gorter [2] as an explanation for the
observation [3, 4] of an anomalous increase of the resistance of thin granular metallic
films with a reduction in temperature. More than 30 years later Fulton and Dolan [5]
observed Coulomb blockade effects in a microfabricated metallic sample and initiated
a huge number of experimental and theoretical studies. Today there are many text
books and reviews [6, 7, 8, 9] on single electron systems both in metals and in
semiconductor systems.

(a) ai
n
dr

tunnel barriers

ce
d

ur
an

de

so
isl

tro
ec
el
e
at
G

500 nm

http://www-drecam.cea.fr/drecam/spec/Pres/Quantro/Qsite/gallery/SET.htm

(b) 100
source
drain 80 1/α(ΔE+e2/C)
I(pA)

60

40
e/Cg
20
Back gate 0
2.0 2.4 2.8 3.2 3.6 4.0 4.4 4.8
Vg(V)

Figure 2: (a) Single electron transistor fabricated using shadow evaporation

techniques. The metallic island is connected to source and drain contacts
through two oxide tunnel barriers. (yellow). On the peak the number of
electrons can fluctuate between N and N+1, in between transport is blocked
due to the Coulomb blockade effect. Coulomb ocillations through a metallic
device are periodic and show conductance peaks with identical heights (b)
Nanowire quantum dot connected to source and drain. Coulomb peaks show
different heights and are irregularly spaced.

2
Coulomb blockade in metallic islands
We start with the situation shown in Fig. 1 where a small metallic particle(dot) is
surrounded by a number of electrodes connected to voltage sources. As discussed
in the lecture the charge state of such an arrangement can be described using a
capacitance matrix which relates the induced charge on each of the electrodes to the
potential change Vj .
Xn
Q̃i = Cij Vj (1)
j=0

Here, Cij are the elements of the capacitance matrix describing the arrangement of
conductors and Q̃i denotes the charge that is induced on the island by the electro-
static potentials Vj on the gates and in the source and drain contacts. We will denote
the dot with index 0, the source and drain contacts with index 1 and 2 respectively
and the gate with index 3. Generally there will be more than one gate electrode
(index 4 ... ) which will make the capacitance matrix of the system larger.
The electrostatic potential of the isolated dot in the presence of the gates is then
given by  
n
1  X
V0 (Q0 ) = Q0 − Qbg − C0j Vj  (2)

CΣ | {z } j=1
Q̃0

In typical Coulomb blockade experiments the bias voltage (VSD ) is much smaller
than the gate voltages and if we choose the electrostatic zero-point to be equal to
the electro-chemical potential at source (j = 1) and drain (j = 2) we can identify the
voltage applied to the gates with Vj in the above equation. Qbg is the charge that
remains on the dot (index 0) if all potentials are put to zero. This can vary from
device to device e.g. because of the doping in Pthe region of the dot. Due to overall
n
charge neutrality the Cij obey the relation i=0 CP ij = 0 and the self capacitance
of the island is therefore given by CΣ := C00 = − nj=1 C0j . If the electron charge
Q0 = −eN on the island is quantized, the total electrostatic energy needed to charge
the dot with N electrons at fixed Vj is
Z −eN n
!
e2 N 2 Qbg X C0j
U (N ) = V0 (Q0 )dQ0 = + eN + Vj (3)
0 2CΣ CΣ j=1
CΣ

C
Here, the external gates induce the displacement charge nj=1 C0jΣ Vj , which can be
P
varied continuously by changing the gate voltages. (In electrostatics this is often
refered to as the charging energy, which makes things confusing when talking about
quantum dots!) From this we can now calculate the energy needed to charge the
island with an additional electron assuming that there are already N electrons on
the island
n
!
e2
 
1 Qbg X C0j
µ̃N +1 = U (N + 1) − U (N ) = N+ +e + Vj (4)
CΣ 2 CΣ j=1
C Σ

3
This difference between the total energies for N+1 and N electrons is often referred
to as the addition energy and will lead to the chemical potential as discussed below
(sometimes, however, this is also called the charging energy). Here we put a ˜ in
order to denote that this is the difference between the electrostatic energies only,
not including the single particle energies. µ̃N +1 increases linearly with the number
of electrons on the dot if the gate voltages are kept constant. By tuning the gates it
is possible to tune µ̃N +1 to lie between the electro chemical potentials in source and
drain, allowing electrons to tunnel on and off the dot one at a time. A continuous
change of the gate voltage then leads to the typical Coulomb blockade oscillations
in the conductance through the dot as shown in Fig. 2(a). On a conductance peak
the number of electrons can fluctuate between N and N+1 while in the blockade
region between two peaks the charge on the island is fixed. Being able to tune a
current of sequential single electron tunneling by a gate voltage gave this device the
name single electron transistor. In these considerations we have used only classical
arguments to describe the properties of such a device. This is called the classical
Coulomb blockade regime and it is a good description for metallic systems with a
continuous density of states.

Coulomb blockade in semiconductor quantum dots

We have mentioned before that the Fermi wavelength in semiconductor heterostruc-
tures is much larger than in metallic systems, due to the relatively small electron
density in semiconductors. This means that in semiconductor nanostructures size
quantization as well as Coulomb blockade effects will be important. The simplest
model which combines both the Coulomb blockade effect and the energy spectrum
of a quantum dot is the constant-interaction model [10].

Constant interaction model

In the constant interaction model it is assumed that the total energy of the island
is given by the sum of its single-particle energies plus the electrostatic energy U(N):
N N n
!
2 2
X X eN Qbg X C0j
E(N ) = i + U (N ) = i + + eN + Vj (5)
i=1 i=1
2CΣ CΣ j=1
CΣ

Since the electrochemical potential is defined as the energy required to add the Nth
electron to a conductor, we can write µN for a dot as
n
!
e2
 
1 Qbg X C0j
µN = E(N ) − E(N − 1) = N + N− +e + Vj . (6)
CΣ 2 CΣ j=1
C Σ

The lever arm αi of gate i on the dot is defined by the ratio

C0i
αi := − (7)
CΣ

4
 a) b)

µN+1
µS µS
e2/C+∆N+1 µD µN µD
µN

µN-1

UN(VG)
UN+1(VG)

Figure 3: Energy diagram for a quantum dot. The two tunnel barriers connect
the dot to the source and drain contacts. (a) Here, transport is blocked and
the dot contains a fixed number of N electrons. (b) The gate voltage was
tuned in order to align the chemical potential in the dot with that of source
and drain. In this situation the number of electrons on the dot can fluctuate
between N-1 and N giving rise to a peak in the conductance.

The value of αi is always positive and in an experiment it typically varies with

large changes in gate voltage. For metallic gates this nonlinearity is small and the
leverarm is well defined. However in quantum dots where regions of a 2DEG are
used as so-called in-plane gates the lever arm might change slowly as a function of
the voltage applied to the gate making the relation between the chemical potential
on the dot and gate voltage non-linear.
Transport through a quantum dot system can now be viewed in terms of the
energy diagram of a double tunnel barrier arrangement like the one schematically
depicted in Fig.3. Here, we first consider the situation at very small bias and low
temperatures (eVbias , kB T << e2 /CΣ ). In Fig. 3(a) transport through the dot is
blocked due to the Coulomb blockade effect, with N electrons on the dot. By
decreasing the gate voltage, the chemical potential µN inside the dot [equation (6)]
is raised until it aligns with that of the drain contact (µD = µN ) and an electron
can leave the dot. If at the same time µS & µD a current can flow and the number
of electrons on the dot will fluctuate between N and N-1. When the gate voltage
further decreases and µS < µN the dot is left with one electron less (N-1) and the
current is again blocked.
For a small bias we therefore have a peak in the conductance whenever µN ≈
µS ≈ µD . If this is used to solve equation (6) for the gate voltage we find that the
position of the peak maximum is given by
" n
#
2
 
(N ) 1 e 1 X eQ bg
VG = N + N− −e αj Vj + − µS . (8)
eαG CΣ 2 j=4
CΣ

5
Addition spectroscopy
In the constant interaction model the evolution of the peak position as a function of
an external parameter B such as the magnetic field has a very simple interpretation.
Under the assumption that the capacitance coefficients Cij and the background
(N )
charge stay constant, the peak maximum position VG [equation (8)] is a measure
of the change of the corresponding single-particle level N (B) as a function of an
external parameter such as the magnetic field. The separation between two peaks
is given by
 
2
(N ) 1 1  e
e∆VG = (µN − µN −1 ) = (N − N −1 ) +  (9)
αG αG | {z } CΣ
∆N

where ∆N is the single-particle level spacing. The influence of the space quantization
can be nicely seen when we look at a measurement of a semiconducting nanowire
quantum dot in Fig. 2(b). In contrast to Fig. 2(a) the Coulomb blockade peaks
are unequally spaced and have various amplitudes. By subtracting 1/αG (e2 /CΣ )
from each peak spacing it is therefore possible to reconstruct the single-particle
spectrum of the dot. This method was first used in an experiment by McEuen [11]
to reconstruct the energy spectrum of a dot in the quantum Hall effect regime.
In order to analyze a spectrum in a quantitative way it is necessary to know the
value of the lever arm of the gate which is tuned. The easiest way to determine the
lever arm is by measuring the Coulomb blockade diamonds, namely current-voltage
characteristics through the dot as a function of the gate voltage.

Charge stability diagrams and Coulomb diamonds

To this end we consider the same situation as above but for finite bias voltage
Vbias . We assume that the bias is applied symmetrically to the source and drain
contacts, which means µS = µ0 + eVbias /2 and µD = µ0 − eVbias /2 where µ0 is the
electrochemical potential in both contacts without an additional bias voltage. This
now leads to a set of requirements for the situation where a configuration with N
electrons on the dot is stable. For Vbias > 0 one finds:

µN < µ0 − eVbias /2
µN +1 > µ0 + eVbias /2

for Vbias < 0:

µN < µ0 + eVbias /2
µN +1 > µ0 − eVbias /2

These inequalities can be translated into what we will call borderline-equations

describing the line where the Coulomb blockade is lifted at the edge of the diamond

6
 +

Figure 4: Coulomb blockade diamonds. (a) The current is blocked in the diamond shaped
areas shaded in grey and dark blue. In these areas the number of electrons in the dot,
N , is constant. Conductance peaks occur on the VG -axis at points where neighboring
diamonds touch (black dots). (b) Excited states move as lines parallel to the borderlines
of the diamonds in the regions where the current is not blocked.

shaped region. Using the relation (6) we find for Vbias > 0
"   n
#
1 1 X eQbg
VG = N + Ec N − − µ0 + e(1 − αS + αD )Vbias /2 − e αj Vj +
eαG 2 j=4
CΣ

"   n
#
1 1 X eQbg
VG = N +1 + Ec N + − µ0 − e(1 + αS − αD )Vbias /2 − e αj Vj +
eαG 2 j=4
CΣ

Here, αS , αD are the lever arms of the source and drain contacts. If the dot
is symmetric, e.g the tunnel barriers have the same geometry, then αS = αD and
the borderlines have exactly the opposite slope ±1/(2αG ). The two lines cross at
eVbias = ∆N +1 +e2 /CΣ . Therefore, comparing the extent of the diamonds for positive
bias with the separation of the corresponding Coulomb peaks gives a measure of
the lever arm. Furthermore it turns out that the difference of the slopes of the
two border lines is 1/αG irrespective of the lever arms of source and drain. The
situation is shown in detail in Fig. 4(a). In the central diamond (dark blue region)
the probability of finding N electrons on the quantum dot is unity and the dot is
in a stable N-electron configuration.The light blue diamonds extending from the
Coulomb peaks (black dots) denote the regions where the probability for finding N

7
electrons on the dot is between 0 and 1 and the electron number can fluctuate by one.
Further away from the gate axis (green areas), the large bias (eVbias > e2 /CΣ ) allows
for two electrons to tunnel at the same time. In a measurement of the differential
conductance (∂I/∂V ) as a function of DC-bias Vbias and gate voltage the border
lines will show up as peaks since this is where the current through the dot changes
and a new transport channel opens/closes (cf. Fig. 5).

30
20
Bias Voltage (mV)

10
0 0 1 2 3 4
-10
-20
-30
-0.95 -0.9 -0.85 -0.8
Gate Voltage (V)

Figure 5: Measurement of Coulomb blockade diamonds in the differential conductance

through a quantum dot. Blue(red) indicate regions of low(high) differential conductance
∂I/∂Vbias respectively. The second diamond is a bit larger than it’s neighboring ones
indicating that the next orbital level is being occupied. The arrows at the top of the figure
show how the levels are being filled with spins.

Excited states
In the above discussion we have assumed that only a single level N contributes to
the transport through the dot. This is not true if eVbias > ∆N +1 . Here ∆N +1 is
the energy difference between the N + 1 electron ground state and it’s first excited
state. Within the constant interaction picture this is equivalent to the N + 1st
single particle energy spacing ∆N +1 = N +1 − N . Therefore, additional single-
particle levels become accessible within the bounds given by the bias voltage and
lead to an increase of the current through the dot and to additional boundary
lines in the differential conductance. This is shown in Fig. 4(b) where for each
Coulomb peak additional lines occur outside the blockade diamonds. For all the

8
elastic processes (energy conserving regarding the dot energy) the border lines for
each Coulomb blockade peak are simply shifted by the single-particle level spacings
1/αG ∆. (indicated by the bold dots on the gate axis. In Vbias direction the line of the
ground state and that of the excited state are separated by 2 ∗ ∆ in analogy to the
separation of the two tips of a Coulomb diamonds.) In real dots with many electrons
there are also more complex collective excitations [12] that have to be considered.
In addition, not all processes have the same amplitude, which means that some
of the lines in the diamonds are suppressed. This can be seen in Fig. 5 where
the measurement of such diamonds in a quantum dot is shown. The differential
conductance is plotted on a logarithmic colorscale in order to bring out the excited
states more clearly.

Amplitude and line shape of the conductance peaks

Up to this point we have treated the dot as an isolated system at zero temperature.
In a real system the dot is connected to source and drain. If we continue in the
language of the single-particle picture then each level i will have a slightly different
coupling ΓSi , ΓD
i due to the different overlap of its wavefunction with source and
drain contacts. It follows that the corresponding level has a finite width given by
~Γi = ~ΓSi + ~ΓD i . This average inverse lifetime is used to discriminate two limiting
regimes in theory, namely the weak coupling limit ~Γ  kB T and the strong coupling
regime kB T . ~Γ.

Weak-coupling regime
A model for this regime at finite temperatures, was first discussed in 1991 by
Beenakker [10]. Here, the energy levels in the dot are treated as delta functions in the
density of states inside the dot (~Γ  kB T ). Here, we only give the limiting form
which is important for most measurements taken in the weak coupling limit, namely
the conductance in the quantum-Coulomb blockade regime (~Γ  kB T  ∆, e2 /CΣ ).
A single level then contributes to the current on a Coulomb peak and the conductance
is given by
  (i) 
e2

1 1
 −1 αG VG − VG
−2 
G(i) = + cosh  (10)
4kB T ΓSi ΓD
i 2k B T

(i)
Here, VG is the position of the i’th Coulomb peak on the gate voltage axis. With
increasing temperature the amplitude of the peak decays with 1/T . Apart from the
temperature dependence, the peak amplitude also reflects the coupling to source and
drain given by the series resistance of the two barriers for level i . A clear signature
of this regime is therefore the uncorrelated variation of the Coulomb peak height as
the gate voltage is tuned. Note, that we have neglected spin in this consideration.
In the simplest case if we assume degenerate spin levels the peak amplitude would

9
lead to peak pairs with similar amplitudes since the same orbital level is charged
with spin up and spin down.

for the interested reader - When the temperature smearing of the Fermi-
function in source and drain becomes comparable to a neighboring single-
particle level spacing, the additional channel can lead to an anomalous in-
crease of the conductance maximum with temperature. Such a situation is
shown in Fig. 6. Three levels are considered where the middle level (2) and
the uppermost level (3) are quite close (∆ = 20meV ). The lever arm of the
gate is assumed to be unity and the coupling of the three levels is chosen as
indicated in Fig. 6(a) where the middle level has the weakest coupling. Here,
the amplitude of the central peak first increases with temperature when level
3 starts to contribute to the current, then the current drops as expected. A
similar behavior is expected for an increase in the bias voltage. The central
peak will increase in amplitude once the excited state with strong coupling is
energetically accessible. This is shown in Fig. 6(c). The conductance is small
at Vbias = 0 and suddenly increases for eVbias > ∆3 . The opposite is true for
peak 3. Once the first excited state gets a finite occupation probability the cur-
rent suddenly drops. This leads to strips of negative differential conductance
(NDC) where the current drops as the bias is increased (red regions marked
by arrows). Qualitatively the long lifetime of the excited state blocks the se-
quential tunneling current through the resonant channel due to the Coulomb
blockade which permits only one additional electron on the dot. Figure 6(d)
shows the case where the coupling of the middle level is made asymmetric with
respect to source and drain. This leads to asymmetries in the diamonds and in
some cases may lead to the complete suppression of one of the borderlines. In
real dots it is quite difficult to make the barriers exactly the same. Apart from
these orbital effects states with a large groundstate spin will lead to similar
features in a phenomenon called spin blockade [13].

In all of these cases the peak shape stays roughly the same, except for a slight
increase of the width. This means that the FWHM (full width half maximum) of a
peak can be used as an upper bound for the electron temperature:
eαG ∆VGF W HM
T ≤ √ ≈ 3.29K/mV αG ∆VGF W HM (11)
4acosh( 2)kB
Here, αG ∆VGF W HM is the width of the Coulomb peak in units of the dot energy.
In the case where many levels contribute to each Coulomb peak, namely where
~Γ, ∆  kB T  e2 /CΣ the peaks all have the same height given by the series
addition of the resistance of the two barriers
 −1
1 1
+
ΓS ΓD

10
 a) b)
αG=1
Τ=250mK
Conductance (e2/h) ΕC=250meV Τ=150mK
Γ3=0.1 Τ=50mK
∆ =20meV
Γ2=0.01 3 0.05
∆2=80meV
1 2 3
Γ1=0.08

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
c) NDC d) Gate Voltage (V)

suppressed 0.05
0.2 0.2
Bias Voltage (V)

Bias Voltage (V)

dI/dV (arb. units)

0.1 0.1
2 3 2 3
0 0 0
∆3
-0.1 -0.1

-0.2 -0.2
-0.05
0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Gate Voltage (V) Gate Voltage (V)

Figure 6: Calculations of peaks and diamonds in the model by Beenakker.

(a)Three single-particle levels are considered with different coupling and sepa-
ration ∆. (b) Peak-shapes at different temperatures. The central peak shows
an anomalous temperature behavior. (c) Differential conductance through the
dot as a function of both gate and bias voltage. When the excited states are
accessible due to a finite bias they can strongly modify the current and even
lead to regions of negative differential conductance (NDC). (d) Asymmetric
coupling ΓD S
2 > Γ2 . Some of the borderlines of the diamonds are completely
suppressed.

11
This is called the classical Coulomb blockade regime since size quantization is no
longer relevant (e.g. in metallic SET’s). The peak shape is very similar to the
quantum limit except that the width is larger by a factor of 5/4.

Strong-coupling regime
The strong coupling limit (kB T . ~Γ) cannot be treated in the formalism discussed
above. The case of noninteracting electrons (e2 /CΣ  kB T, ~Γ, ∆) is closely related
to the problem of resonant transmission through a double barrier [14, 15] as we
have discussed it in the lecture. The line-shape of a single resonant level is simply a
Lorenzian and its conductance is given by the well known Breit-Wigner formula [16]
−1
2e2 h2 Γ

1 1
G= + D (12)
h Γ S Γ 2
αG (VGmax − VG )2 + (hΓ/ 2)2

Here, Γ = ΓS + ΓD is assumed to be independent on the level index i and the factor

2 is for spin degeneracy. Since charge effects are neglected, the tunneling no longer
needs to be sequential (=one electron at a time) and the two channels for each spin
direction add up. In this limit higher order processes are important [9] and in the
case of hΓ > ∆ inelastic transitions in the dot reduce the conductance and broaden
the resonant level [14, 15].

Transmission phase of a quantum dot

Recently several experiments have investigated the transmission phase of a quantum
dot embedded in one of the arms of an Aharonov Bohm (AB)-ring [17, 18, 19, 20].
The basic idea was to extract the phase of the transmission through the quantum
dot from the phase shift of the AB-oscillations in the conductance through the ring
as a function of magnetic field. In order to explain a phase shift of π, the complex
transmission amplitude of a Breit-Wigner resonance was used [18](see problem 5.12
in the book by Davies)

ΓN /2
tQD = iCN . (13)
EF − EN + iΓN /2
Here, CN is a complex amplitude, EF the energy of the electrons transmitted
through the device and EN and ΓN the energy and width of the resonance in the
quantum dot. The resulting behavior of the phase and amplitude of tQD (EF − EN )
are shown in Fig. 7 for CN = 1 and ΓN = 20.

In-plane magnetic field

In a magnetic field parallel to the plane of the quantum dot (for a semiconducting
nanowire perpendicular to the wire, for a dot in a 2DEG parallel to the surface) the

12
 a) b)
Amplitude Phase
1 1.5

0.8 1

0.5
0.6
EF-EN
0.4 -100 -50 50 100
-0.5
0.2
-1

-100 -50 50 100 EF-EN -1.5

Figure 7: Amplitude |tQD |(a) and phase arg tQD (b) of the transmission
tQD (EF − EN ) through a quantum dot as calculated using a Breit-Wigner
Ansatz for the conductance resonance line shape.

chemical potential of the dot has two additional terms

1 e2 −1)

∗
µN = N + (N + ) − γN B||2 + s(N
z
)
− s(N
z g µB B|| (14)
2 CΣ
One is called the diamagnetic shift which is parabolic in B|| and describes the shift
of the energy levels due to a squeezing of the wavefunctions in strong magnetic
fields. The other term is the Zeeman splitting which is linear in B. Here, we restrict
ourselves to a short discussion of the Zeeman term since other magnetic field effects,
namely on the spectrum of a harmonic oscillator potential, are treated in the book.

Zeeman splitting
So far we have neglected spin-effects altogether in our considerations. In the zero
magnetic field case this means that we have simply assumed degenerate spin states
or in other words the filling of each orbital level first with a spin up and then a
spin down electron. Within the constant interaction picture this leads to a sequence
of spin pairs namely Coulomb peaks which are separated only by the interaction
energy 1/αG e2 /CΣ . In this case the groundstate spin of the dot switches between
sz = 0 for an even number of electrons and sz = ±1/2 for odd N. Since the addition
energy results from the difference µN = E(N ) − E(N − 1) the Coulomb peaks will
be shifted proportional to

−1)

∗ 1
∆Ezeeman = s(N
z
)
− s(N
z g µB B|| = ± g ∗ µB B|| (15)
2
here sNz is the ground state spin for N electrons on the dot and g∗ is the effective
g-factor for the electrons in the dot.
In these considerations we have assumed that the lever arm αG is independent of
the magnetic field. In addition, the diamagnetic shift of both the contacts as well as
the dot will obscure this picture of a linear Zeeman splitting. A possible solution to

13
this is found by the assumption that both αG and γN are slowly varying functions
of the gate voltage which means that more reliable information can be obtained
from differences of neighboring peak positions. Here, we have only discussed the
simplest case of including spin in the constant interaction picture. In a system
where exchange effects are important (e.g. orbital filling according to Hunds rules.)
this picture breaks down.

14
Bibliography

[1] A. Fuhrer, Ph.D. thesis, Swiss Federal Institute of Technology (2003).

[2] C. J. Gorter, Physica 17, 777 (1951).

[3] R. Lambeir, A. Vanitterbeek, and G. J. Vandenberg, Physica 16, 907 (1950).

[4] N. Mostovetch and B. Vodar, Compt,Rend 230, 934 (1950).

[5] T. A. Fulton and G. J. Dolan, Phys. Rev:lett. 59, 109 (1987).

[6] P. Hawrylak, L. Jacak, and A. Wojs, Quantum dots (Springer Verlag, Berlin,
1998).

[7] H. Grabert and M. H. Devoret, Single Charge Tunneling (Plenum Press, New
York, 1992).

[8] T. Ando, Y. Arakawa, K. Furuya, S. Komiyama, and H. Nakashima, eds.,

Mesoscopic Physics and Electronics (Springer, 1998).

[9] L. P. Kouwenhoven, C. M. Marcus, P. L. McEuen, S. Tarucha, R. M. Wester-

velt, and N. S. Wingreen, in Nato ASI conference proceedings, edited by L. P.
Kouwenhoven, G. Schön, and L. L. Sohn (Kluwer, Dordrecht, 1997), pp. 105–
214.

[10] C. W. J. Beenakker, Phys. Rev. B 44, 1646 (1991).

[11] P. L. McEuen, E. B. Foxman, U. Meirav, M. A. Kastner, Y. Meir, N. S.

Wingreen, and S. J. Wind, Phys. Rev. Lett. 66, 1926 (1991).

[12] D. Pfannkuche and S. Ulloa, Phys. Rev. Lett 74, 1194 (1995).

[13] D. Weinmann, W. Häusler, and B. Kramer, Phys. Rev. Lett. 74, 984 (1995).

[14] M. Büttiker, Phys. Rev. B. 33, 3020 (1986).

[15] A. D. Stone and P. A. Lee, Phys. Rev. Lett. 54, 1196 (1985).

[16] G. Breit and E. Wigner, Phys. Rev. 49, 519 (1936).

15
Bibliography

[17] A. Yacoby, M. Heiblum, D. Mahalu, and H. Shtrikman, Phys. Rev. Lett. 74,
4047 (1995).

[18] R. Schuster, E. Buks, M. Heiblum, D. Mahalu, and V. U. H. Shtrikman, Nature

385, 417 (1997).

[19] Y. Ji, M. Heiblum, and H. Shtrikman, Phys. Rev. Lett. 88, 076601 (2002).

[20] K. Kobayashi, H. Aikawa, S. Katsumoto, and Y. Iye, Phys. Rev. Lett 88, 256806
(2002).

[21] Y. M. Blanter, A. D. Mirlin, and B. A. Muzykantskii, Phys. Rev. Lett. 78, 2449
(1997).

[22] H. U. Baranger, D. Ullmo, and L. I. Glazman, Phys. Rev. B 61, R2425 (2000).

[23] P. W. Brouwer, Y. Oreg, and B. I. Halperin, Phys. Rev. B 60, R13977 (1999).

[24] H. Jiang, H. U. Baranger, and W. Yang, Phys. Rev. Lett. 90, 026806 (2003).

[25] W. J. D. Haas, J. D. Boer, and G. J. van der Berg, Physica 1, 1115 (1933).

[26] J. Kondo, Progr. Theor. Phys. 37, 37 (1964).

[27] L. I. Glazman and M. E. Rai, JETP Lett. 47, 452 (1988).

[28] T. K. Ng and P. A. Lee, Phys. Rev. Lett 61, 1768 (1988).

[29] D. Goldhaber-Gordon, H. Shtrikman, D. Mahalu, D. Abusch-Magder,

U. Meirav, and M. A. Kastner, Nature 391, 156 (1998).

[30] F. D. M. Haldane, Phys. Rev. Lett 40, 416 (1978).

[31] S. M. Cronenwett, T. H. Oosterkamp, and L. P. Kouwenhoven, Science 281,

540 (1998).

[32] J. Schmid, J. Weis, K. Eberll, and K. v. Klitzing, Physica B 256, 182 (1998).

[33] H. Jeong, A. M. Chang, and M. R. Melloch, Science 293, 2221 (2001).

[34] J. Nygard, D. H. Cobden, and P. E. Lindelof, Nature 408, 342 (2000).

[35] W. Liang, M. P. Shores, M. Bockrath, J. R. Long, and H. Park, Nature 417,

725 (2002).

[36] U. F. Keyser, C. Fuhner, S. Borck, R. J. Haug, M. Bichler, G. Abstreiter, and

W. Wegscheider (2002), cond-mat/0206262.

16