INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT

LEO P. KOUWENHOVEN
Dept. of Applied Physics and DIMES
Delft University of Technology, 2600 GA Delft, The Netherlands
GERD SCHO N
Institut fur Theoretische Festkorperphysik
Universtitat Karlsruhe, 76128 Karlsruhe, Germany
AND
LYDIA L. SOHN
Dept. of Physics, Princeton University
Princeton, NJ 08544 - 0708, USA

In this introductory chapter several basic concepts, relevant for meso-

scopic electron transport, will be described. The aim is to provide a basis
for several of the following Chapters of this volume. We, therefore, describe
in the rst Section various aspects of electron quantum transport in two-
dimensional electron gases. This includes an estimate of typical material
parameters as well as comments on the fabrication. We then describe the
quantization of the conductance in point contacts and the edge state pic-
ture of the Quantum Hall e ect. In the second Section we describe the
theory of single{electron tunneling in systems with strong charging and
Coulomb-blockade e ects. Here we restrict ourselves to the simplest case
where low order perturbation theory is sucient. We rst discuss metallic
low-capacitance junction systems and then indicate the relevant extensions
when dealing with transport through quantum dots with discrete levels.
Many further, equally important aspects of mesoscopic transport can not
be covered here. Examples are interference and weak localization e ects,
level statistics or the many body description of solids. Fortunately, some of
those will be covered in the more specialized Chapter of this volume.
2 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN

Figure 1. (a) Schematic cross section of a GaAs/AlGaAs heterostructure. The 2DEG is

located at the interface between the GaAs substrate and the AlGaAs top layer. (b) Hall
bar with six Ohmic contacts (shaded squares).

1. Electron Transport
1.1. 2DEG SYSTEMS AND THE QUANTUM HALL EFFECT
In this Section we introduce basic properties of a two-dimensional electron
gas (2DEG) in a GaAs/AlGaAs heterostructure (see Fig. 1a). On the GaAs
substrate a layer of typically 100 nm AlGaAs is grown. Somewhere halfway
in the AlGaAs layer there is a thin layer where the Ga atoms are replaced by
Si donor atoms. With a proper amount of Si one nds that at low temper-
ature the only mobile electrons are located at the GaAs/AlGaAs interface.
These free electrons are attracted by the GaAs since they can lower their
energy in this smaller band gap material. They are also held as close as
possible to their ionized Si+ donors and thus they form a thin conducting
layer near the GaAs/AlGaAs interface (for a review on growth of GaAs
heterostructures see Ref. [1]). Since GaAs and AlGaAs can form a nearly
perfect interface on the atomic scale and since the Si donors are spatially
separated, the electrons experience very little scattering. Typical mean free
paths are 10 m and the record is close to 100 m. Before we discuss bal-
listic mesoscopic devices we rst review a few important electron transport
properties of 2DEG's. In Table 1.1 we have summarized a number of useful
relations and some typical values for GaAs/AlGaAs heterostructures.
For transport experiments one rst de nes a so-called Hall-bar of typi-
cally 0.1mm1mm. At the edge of the Hall-bar one then locally evaporates
a number of metal squares of roughly 50m50m in size. By heating the
whole sample this metal di uses into the semiconductor where at some
point it makes electrical contact to the 2DEG. Good contacts are char-
acterized by linear current-voltage traces and are therefore called Ohmic
contacts. Fig. 1b shows a Hall-bar with 6 Ohmic contacts in a con guration
that allows measurements of the Hall resistance Rxy and the longitudinal
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 3

Figure 2. Measurement of the Hall resistance Rxy and the longitudinal resistance Rxx
at 0.1 K. The mobility is  = 1:5
106 cm2 /Vs, the temperature is T = 80mK. (Figure
provided by R. L. Willett.)

resistance Rxx. The sample is current biased and resulting voltages can be
measured as a function of magnetic eld.
Fig. 2 shows a remarkable measurement of Rxy and Rxx as a function
of magnetic eld measured at a temperature of 80mK. The Hall resistance
shows plateaus where the longitudinal resistance has minima of virtually
zero resistance. The striking aspect of this data is that the plateaus are
precisely quantized at integer and fractional multiples of h=e2 . The integer
plateaus are known as the integer quantum Hall e ect (QHE), and corre-
spondingly, the fractional plateaus are referred to as the fractional QHE.
The precision of the quantization of the former is so accurate that it now
forms the international standard of resistance. An important quantity in
the quantum Hall regime is the lling factor  = hns =eB . It is equal to
the number of electrons divided by the number of ux quanta 0 = h=e.
For lling factor  = 1 the system is in the center of the rst plateau at
h=e2, for  = 2 in the center of the second plateau at h=2e2, and so forth.
The lling factor is, for instance, convenient for determining the electron
density. The oscillations in the longitudinal resistance, which have minima
4 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN
at the same magnetic elds where the Hall resistance shows plateaus, are
called Shubnikov-de Haas oscillations. The temperature dependence of the
oscillation minima is an accurate determination of the mobility and mean
free path of the 2DEG.
Table 1.1. Useful relations and system parameters (including spin de-
generacy). We have chosen typical values for e , and ns , from which the
other values are deduced.
{ electron mobility e = 106 cm2 /Vs (typical value)
{ scattering time  = me /e = 38 ps
{ e ective mass m = 0.067 m0
{ electron density ns = 2:8 1015 m,2 (typical value)
{ Fermi energy EF = h2 ns=m = 10 meV
{ Fermi velocity vF = (2EF=m)1=2 = 2:3 105 m/s
{ Fermi wavelength F = (2=ns)1=2 = 47 nm
{ elastic mean free path l = vF  = 8:7 m
{ cyclotron radius at EF rc = m vF =eB = 88 nm at B = 1 T
{ angular cyclotron frequency !c = eB=m = 2:6 1012 rad/s
{ magnetic length lB = (h=eB )1=2 = 8:1 nm at B = 10 T
1.2. E-BEAM FABRICATION OF A SUBMICRON SEMICONDUCTOR
DEVICE
We brie y outline a standard procedure for fabricating small devices in a
2DEG. We start from a 2DEG con ned in a GaAs/AlGaAs heterostruc-
ture. To laterally con ne the electrons one must de ne a pattern on top of
the heterostructure. As an example, we describe e-beam lithography, which
is a technique also used in the fabrication of chips. Depending on the de-
sired pattern, one can choose between many variations in the lithography
process steps. One procedure is shown schematically in Fig. 3. An organic
resist lm (100nm thick) is spun onto the substrate. Exposing the re-
sist with an electron beam results in a molecular-mass di erence between
the exposed and unexposed parts (see Fig. 3a). An appropriate developer
removes only the exposed resist, resulting in the mask pattern shown in
Fig. 3b. Evaporated material now sticks only at the substrate where the
resist has been removed (see Fig. 3c). The mask itself can be removed by
dissolving the remaining resist (lift-o ), leaving a small pattern on top of
the substrate (see Fig. 3d). The minimum resolution of such a pattern with
present day electron-beam lithography facilities is about 20 nm.
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 5

Figure 3. Outline of the electron-beam lithography procedure for fabricating submicron

structures.

Figure 4. (a) Cross-section of a GaAs/AlGaAs heterostructure with typical layer thick-

nesses. A negative voltage VG applied to the metal split-gate con nes the electrons lat-
erally in the 2DEG.
(b) Top-view of a QPC. The dotted line indicates the depletion region in the 2DEG,
which is tuned by VG . The two wide 2DEG regions act as reservoirs, emitting electrons
through the QPC with energies up to their electrochemical potentials 1 and 2 . A
voltage di erence V = (1 , 2 )=e results in a net current I through the QPC.

From this point there are basically two ways to transfer the pattern to
the 2DEG. One way is to use the pattern as an etch mask. Etching removes
the portion of the 2DEG not protected by the pattern. The boundaries of
the etched pattern cause a depletion region, such that the conducting width
in the 2DEG is unknown and often much smaller than the de ned width [2].
6 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN
Another, more exible way to transfer the pattern to the 2DEG is to use it
as a gate [3, 4]. Applying a negative voltage to the gate depletes the electron
gas beneath it, thereby con ning electron motion to the ungated region. For
a split-gate geometry, shown in Fig. 4, this results in a narrow conducting
channel. The advantage of the split-gate technique is that the conducting
width of the point contact in the 2DEG can be tuned from the de ned
lithographic width of the pattern to zero, by making the gate voltage more
negative. Transport between the two wide 2DEG regions in Fig. 4b occurs
only through the point contact and can be studied as a function of the
width by changing the gate voltage. The width W of the constriction can
be made comparable to the Fermi wavelength, so this device is called a
quantum point contact (QPC). The actual induced potential in the 2DEG
is unknown, but self-consistent calculations [5] indicate that it has a saddle-
shape (see Fig. 5a). In the constriction, electrons are con ned in the lateral
x-direction and slowed down by the presence of a potential barrier in the
y -direction. Making the gate voltage more negative, simultaneously reduces
the width and increases the barrier height. For zero-width or a barrier which
is higher than the Fermi energy EF of the 2DEG, the QPC is pinched-o and
electron transport between the wide 2DEG regions is impossible. Fig. 5b
shows a scanning electron micrograph of a double point contact device.

Figure 5. (a) Saddle-shaped potential induced in the 2DEG upon application of a

negative gate voltage, resulting in lateral con nement in the x-direction and a potential
barrier in the longitudinal y-direction (from Beenakker, and van Houten in Ref. [6]).
(b) Scanning electron micrograph of a double-QPC device. The white areas are the Au
gates, and the marker is 1 m long. The QPCs are 250 nm wide and are separated by
1.5 m.

1.3. QUANTIZED CONDUCTANCE OF A POINT CONTACT

The resistance of a point contact in the classical ballistic regime is known as
the Sharvin resistance [7]. The Sharvin resistance is entirely due to elastic
backscattering at the geometrical narrowing of the ballistic point contact.
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 7
Dissipative processes, which bring the electron system into thermodynamic
equilibrium, take place far away from the point contact (i.e. several times
the inelastic mean free path). Therefore, the cause of the resistance in this
system is spatially separated from its corresponding Joule heating. The
classical conductance of a Sharvin point contact in a 2DEG is [8]
2 2D
Gs = e dNdE vF W : (1)
The quantum mechanical 2D density of states per unit area, including
a factor 2 for spin degeneracy, is dN 2D=dE = m = h , while vF = h kF =m
is the Fermi velocity. We can rewrite the classical equation (1) to a semi-
classical version which includes the conductance quantum 2e2 =h
2 2
G = 2e kF W = 2e 2W :
s h  h F (2)
The Fermi wave-vector kF or wavelength F are related to the 2D elec-
tron density ns by kF = 2=F = (2ns)1=2. The semi-classical form of the
Sharvin conductance is continuous and linear in the width W . However,
Eq. 2 suggests that we can expect deviations due to the wave nature of
electrons whenever F is of order W . We give a derivation of the quantum
version of the Sharvin conductance below, but rst discuss the experimental
results.
The conductance of a point contact is measured by passing a current
I through the sample and measuring the voltage V between the current
source and drain (see Fig. 4b). Fig. 6 shows the conductance G in units
of 2e2 =h = (12906 ),1 versus gate voltage VG measured at B = 0. As-
suming that the width varies linear with gate voltage, we indeed see that
on average the conductance decreases linearly when the constriction is nar-
rowed. However, around this classical dependence we see that G changes
in quantized steps of 2e2 =h. At VG = ,2:2V, the conductance becomes
zero, corresponding to a pinched-o point contact. Fig. 6 further shows
that on increasing the temperature the conductance quantization gradu-
ally disappears [9]. The conductance G(VG) is roughly linear at 4.2 K, in
accordance with the classical dependence of Eq. 1. Although the classical
result cannot explain the quantization, we note that the plateau values are
obtained in Eq. 2 whenever the width W is an integer multiple of F=2.
Eq. 2 predicts that an increase in W of F =2 (which is 21 nm in this sam-
ple) increases G by 2e2 =h. In total 16 steps were observed between pinch-o
and VG = ,0:3V were the constriction is just formed in the 2DEG. The 16
steps give an estimate of the width W of about 340 nm, somewhat larger
than the lithographic width of 250 nm, but consistent with the schematic
depletion pro le shown in Fig. 4(a). These considerations are reminiscent
8 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN

Figure 6. Conductance versus gate voltage at B = 0 and di erent temperatures. In-

creasing the temperature thermally averages the higher plateaus rst (from Ref. [9]).

of the states of a particle-in-a-1D-box, which, as we show below, is the basic

idea behind the conductance quantization.
We note that the conductance quantization is not as exact as the quan-
tum Hall e ect. First, a series resistance ( 100 ) originating from the
wide 2DEG regions has been subtracted to line up the plateaus at their
quantized values [8]. Furthermore, the plateaus are not completely at.
This may be due to scattering at impurities in the vicinity of the QPC or,
as we discuss below, the abruptness of the constriction.
We now discuss that the conductance quantization for transport through
1D subbands. If the potential which describes the transition from the
wide 2DEG regions to the narrowest point in the QPC varies suciently
smoothly (i.e. adiabatically), the potential variation in the x- and y -directions
may be decoupled [10] (see Fig. 5a). The narrowest point forms the bottle-
neck of the QPC in the sense that it completely determines the transport
properties. In this case we can calculate transport through a QPC from the
Hamiltonian 2 p2
H = 2pmx + eV (x) + 2my  ; (3)
where V (x) describe the transport at the bottleneck. For the con nement in
the lateral x-direction, we follow Berggren et al. [11] and choose a parabolic
con ning potential V (x) = 1=2mw02x2 . Self-consistent calculations of Laux
et al. [5] have shown that small split-gate samples have a con nement close
to such a parabola. The advantage of using a parabolic potential is that the
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 9
resulting Schrodinger equation can be written in the form of a harmonic
oscillator having energy eigenvalues
h2 k2
En = (n , 12 )h!0 + 2my (n = 1; 2; :::) (4)
which contains a free-electron kinetic energy dispersion in the longitudinal
y-direction. In the lateral x-direction the energy states, indexed by n =
1; 2; :::, are quantized and separated in energy by h !0 .
Because the electron motion is free in one direction only, Eq. 4 describes
1D subbands. Fig. 7 shows the 1D subband dispersion versus longitudinal
wave-vector ky . The right-going electrons, with a velocity h vn = dEn =dky ,
originate from the left 2DEG reservoir, which at zero temperature populates
all the states up to its electrochemical potential 1 . Similarly, the left-going
electron states are occupied up to 2 , the electrochemical potential of the
right 2DEG reservoir (see Fig. 4b).
A voltage di erence V = (1 , 2 )=e between the two reservoirs results
in a net current I , which is carried by the (uncompensated) electron states
in the energy interval between 1 and 2 . Note that we de ne the Fermi
energy as EF = 1 = 2 when V = 0. The net current I at zero temperature
is
N Z 1
dE 12 ddNEn vn (E )Tn(E ) ;
X
I=e (5)
n=1 2
which includes the transmission probability of the n-th subband Tn (E ) to
describe possible scattering events. Here N denotes the number of occupied
subbands, i.e. the largest number for which EN (ky = 0) < EF . The 1D spin-
degenerate density of states is dNn =dE = 2= (dEn=dky ),1 . The important
aspect of 1D transport is the cancelation of the energy dependence in the
product of velocity and density of states (dNn =dE )vn = 4=h. For small
voltages (eV  EF ), one can take Tn (E ) = Tn (EF ). Substituting this
in Eq. 5, one nds that the conductance G = I=V = eI=(1 , 2 ) is
independent of energy
2 e N
2X
G= h Tn(EF ) : (6)
n=1
Eq. 6 is known as theP2-terminal Landauer formula [12]. If no backscattering
takes place, so that Nn=1 Tn (EF) = N , Eq. 6 reduces to
2
G = 2he N ; (7)
10 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN

Figure 7. Energy En versus longitudinal wave-vector ky from Eq. 4 at the bottleneck of

a QPC assuming a parabolic con nement potential. The 1D subbands are separated by
h !0 . A net current results from the uncompensated occupied electron states in the interval
between 1 and 2 , the electrochemical potentials of the two wide 2DEG reservoirs.

demonstrating that each occupied subband contributes 2e2=h to the con-

ductance. The subbands are called 1D current channels to emphasize that
each channel carries the same amount of current.
In the experiment, a decreasing VG increases the barrier in the QPC,
and simultaneously increases the lateral con nement and consequently the
energy splitting. Both e ects increase the subband energies. As long as EF
is between two subband bottoms, N is constant and G is quantized. If a
subband bottom moves through EF, N changes by 1 and G by 2e2 =h.
Several numerical calculations [13] have shown that Eq. 7 gives an ac-
curate description of a QPC with the assumptions that impurity scattering
is absent and that the potential variations are smooth. Sharp potential
variations, possibly present at the entrance and exit of the QPC [14] or
originating from impurities [15, 16] can give rise to backscattering and
therefore destroy the quantization. The assumption of an adiabatic con-
striction in Eq. 3 is therefore no longer valid. At T > 0 the reservoirs inject
electronsR with a Fermi-Dirac distribution, which averages the conductance
G(T ) = dE (,df=dE )G(E ). Comparing this with the temperature depen-
dence of the curves in Fig. 6, it is found that the subband separation grad-
ually increases from about 1 meV at VG = ,1V to 3 meV at VG = ,2:1V.
This illustrates that thermal averaging has a stronger e ect on the higher
plateaus, as observed in Fig. 6.
1.4. DEPOPULATION OF 1D MAGNETO-ELECTRIC SUBBANDS
We now turn to the case of a QPC with an applied magnetic eld in the
z-direction. In the Hamiltonian of Eq. 3, the magnetic eld B is included
by substituting (p , eA) for the momentum p. In the Landau gauge for
the vector potential A = Ay = Bx, the Schrodinger equation is once again
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 11

Figure 8. 1D subband dispersion for three values of the magnetic eld, illustrating
magnetic depopulation. The energy splitting is electric (h!0 ) in (a), and hybrid (h!,
with !2 = !02 + !c2 ) in (b). For large magnetic elds (c) the 1D subbands are Landau
levels with an energy splitting of h !c.

that of a harmonic oscillator, but now with energy eigenvalues [11]

1 h 2 ky2
En = (n , 2 )h! + 2m : (8)
B
These eigenvalues describe hybrid magneto-electric subbands. With ! 2 =
!02 + !c2, where the cyclotron frequency is !c = eB=m , the energy separa-
tion is now a combination of the electrical con nement and the magnetic
eld. mB = m! 2 =!02 is a magnetic eld dependent e ective mass yielding
a smaller dispersion for larger magnetic elds. The in uence of the mag-
netic eld on the subband dispersion is shown schematically in Fig. 8. For
B = 0 the subbands are determined by the electrical con nement only. A
small magnetic eld increases the subband splitting and reduces the dis-
persion. For large magnetic elds, the subbands have the magnetic energy
separation and a vanishing dispersion. In this case, the subbands are the
well-known Landau levels. It can be seen from Fig. 8 that on increasing the
magnetic eld, the number of occupied subbands decreases. This process is
known as the depopulation of magneto-electric subbands.
One can show that the velocity and density of states also cancel in a
magnetic eld and that Eqs. ( 6) and ( 7) are still valid [17, 18]. From
the above analysis it follows that a gradual transition exists between the
quantized conductance G = N 2e2=h at zero magnetic eld (with N the
number of occupied electric subbands in Fig. 8a) to the quantum Hall
conductance GH = NL 2e2=h at a high magnetic eld (with NL the number
of occupied Landau levels in Fig. 8c). We note that our particular choice
of a parabolic con nement does not a ect the general conclusions, such as
the cancelation of velocity with the density of states, and the conductance
quantization at zero and non-zero magnetic eld.
12 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN

Figure 9. QPC conductance versus gate voltage at 0.6 K for several values of magnetic
eld. The increasing width of the plateaus demonstrate the increasing energy splitting in
a magnetic eld. The curves have been o set for clarity (from Ref. [18]).

Fig. 9 shows the conductance of a QPC versus gate voltage for several
values of the magnetic eld [18]. As can be seen, the quantization is pre-
served in a magnetic eld. Above B = 1 T, spin-resolved plateaus develop
at odd multiples of e2 =h. The depopulation can be seen from the fact that
at a xed gate voltage, the number of plateaus (or, equivalently, the num-
ber of occupied subbands), decreases with increasing magnetic eld. From
the measurements of Fig. 9, one can deduce subband splittings of about
1 meV at VG = ,1V and 3 meV at VG = ,2V, in agreement with the
values obtained from the temperature dependence [9]. A third independent
way to determine the subband splittings is by measuring the non-linear
current-voltage characteristics of a QPC [19].
1.5. ELECTRON TRAJECTORIES IN A LOW MAGNETIC FIELD
In the previous Section, we discussed the in uence of a magnetic eld on
the subband dispersion in momentum space. We now consider the electron
motion in real space, which yields a simple physical picture of the QHE
and associated e ects. To elucidate the quantized electron motion in a
high magnetic eld, we rst discuss classical electron trajectories in a low
magnetic eld.
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 13

Figure 10. (a) Schematic classical electron trajectories in a magnetic eld.

(b) Corresponding quantum picture of the energy states of Eq. (10) along a cross-section
of the 2DEG, illustrating the formation of edge channels at the boundary of the 2DEG
(formed by the electrostatic potential energy eV (x; y)) where the Landau levels intersect
the Fermi energy EF .

In the absence of an electric eld E , the balance of the Lorentz force

FL = evB and the centripetal force F = m v 2=r leads to a cyclotron motion
of the electrons, with an angular frequency !c = eB=m and at the Fermi
energy a cyclotron radius rc = mvF =eB (see Fig. 10a). When the electric
eld E = ,rV (x; y ) =6 0, the electrons have a net drift velocity vD = E=B.
At the boundary of the sample, where E is large, the collisions at the edge
result in skipping orbits. The electrons skip with the drift velocity along
the edge of the sample. The direction of the velocity is opposite for the two
opposite edges (see Fig. 10a).
The skipping orbit motion of electrons along a 2DEG boundary in a
small magnetic eld has been observed in an electron focusing experiment
by van Houten et al. [20]. The geometry shown in Fig. 5b with two ad-
jacent QPCs with a separation of L = 3m was used, where one QPC
injects electrons into the 2DEG and the second QPC collects them. The
injected electrons are focused by the magnetic eld on the boundary be-
tween injector and collector at distances 2p rc(p = 1; 2; :::). Focusing into
the collector occurs when 2p rc = L. In the experiment the collector voltage
is measured as a function of magnetic eld B . From the condition 2p rc = L
it follows that the largest number of electrons reaches the collector when
Bfoc = 2pmvF =eL which leads to periodic oscillations in the collector sig-
nal. This is shown in Fig. 11 together with a calculation. On average the
voltage increases linearly with magnetic eld, which is expected for the
classical Hall resistance VHall=I = B=ens . For curve (a), however, we see
14 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN

Figure 11. Bottom: Electron focusing (T = 50 mK, L = 3:0m) in the con guration
depicted in the inset. The two traces a and b are measured with interchanged current
and voltage leads, and demonstrate the injector-collector reciprocity as well as the repro-
ducibility of the ne structure. Top: Calculated classical focusing spectrum corresponding
to the experimental trace a (50 nm wide point contacts were assumed). The dashed line
is the extrapolation of the classical Hall resistance seen in reverse elds. (From Ref. [20].)

additional large oscillations with ne structure around the average Hall re-
sistance for positive magnetic eld. These oscillations occur at the expected
focusing elds. For a negative magnetic eld no focusing signal is observed
since now the injected electrons are de ected away from the collector.
Curve (b) is taken with the current and voltage probes interchanged.
The relation between (a) and (b) demonstrates Buttiker's reciprocity rela-
tion [21],
Rij;kl(B ) = Rkl;ij (,B ) : (9)
It implies that upon interchange of the current probes i and j with the
voltage probes k and l, one obtains the same resistance at the opposite
magnetic eld. This fundamental symmetry relation is clearly con rmed
in Fig. 11, including the symmetry in the ne-structure. The latter has
been explained by extending the classical focusing calculation of the top
section to include quantum interference e ects between trajectories being
injected at di erent angles [20]. This focusing experiment demonstrates
that the collisions at the boundary between injector and collector are highly
specular, since di usive boundary scattering would average the oscillations.
In a subsequent experiment by Spector et al. [22] focusing signals were
observed up to a distance between injector and detector as large as 64 m.
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 15
1.6. THE QUANTUM HALL EFFECT ON MESOSCOPIC LENGTH SCALES
1.6.1. Edge Channels
In a high magnetic eld the electron motion is quantized. The ux  en-
closed by an electron in a cyclotron orbit equals an integer times the ux
quantum 0 = h=e, and the quantized electron energies are:
En = (n , 21 )h!c + eV (x; y) (10)
where n = 1; 2; ::: is the spin-degenerate Landau level index, and we have
ignored the Zeeman energy splitting gB B=2. We assume that the elec-
trostatic potential V (x; y ) is at in the interior of the sample and rises at
the boundary. Electrostatic variations due to impurities are ignored, be-
cause we are dealing with ballistic samples. Fig. 10b shows schematically
the Landau energy levels of Eq. 10. The electron states at the left boundary
are occupied up to 1 , the electrochemical potential of the current source,
and at the right boundary up to 2 , the electrochemical potential of the
current sink (see Fig. 4b). At the two sample boundaries, the electron states
have opposite velocity directions, similar to the classical case of Fig. 10a.
The relevant electron states for linear transport are only those at the
Fermi energy. As can be seen, these are located at the sample boundaries,
where the Landau levels intersect the Fermi energy (En = EF ), and they
extend in the direction perpendicular to the cross-section of Fig. 10b. The
intersections are the current-carrying states, which are known as edge chan-
nels [17, 23, 24]. The net current I only results from the uncompensated
states in the interval between 1 and 2 . The total current carried by
the states below 2 is zero. The transport through edge channels is 1-
dimensional [17]. Edge channels can therefore also be viewed as 1D current
channels, each carrying a current In = 2e=h(1 , 2 ). With the Hall volt-
age VH = (1 , 2 )=e measured between the two sample boundaries, this
directly gives the quantized Hall conductance GH = NL In =VH = NL 2e2=h.
Up to now we have ignored all scattering processes. Buttiker [25] has
pointed out that due to the spatial separation of the electron states with op-
posite velocity, backscattering requires scattering from one sample bound-
ary to the other. Backscattering is therefore suppressed when the edge
states between 1 and 2 are not connected by extended electron states.
This is the case in Fig. 10b, where the Fermi energy is between two bulk
Landau levels.
1.6.2. Selective Probing of Edge Channels
The above description of transport in the quantum Hall e ect (QHE)
regime, known as the Landauer-Buttiker formalism, is reviewed in Ref. [26].
This edge channel description gives an appealing physical picture of the
16 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN
QHE. The question now arises about how to prove the existence of edge
channels and whether they can really be viewed as independent current
channels. We now discuss an experiment involving two adjacent QPCs
which directly probes the transport through a particular edge channel. The
theory for this experiment provides a simple illustration of Buttiker's multi-
probe formalism which has been used very successfully to describe di erent
kind of mesoscopic transport phenomena.
In Section 1.5 we discussed the electron focusing from one QPC to an
adjacent second QPC by small magnetic elds (B  1T). These elds are
too small to quantize the electron motion, and the focusing can be explained
in terms of classical cyclotron motion. We consider now the same geometry
in the high eld regime (B  1T). Fig. 12a shows schematically two adja-
cent QPCs A and B de ned in a 2DEG with an applied magnetic eld such
that two edge channels are occupied. The 2-terminal conductances GA and
GB of the individual QPCs measure the number of transmitted channels
and are quantized in multiples of 2e2=h (see also Fig. 9). The Hall conduc-
tance GH is normally thought to be independent of the characteristics of
the current and voltage probes and to correspond directly to the number of
occupied (spin-degenerate) Landau levels NL in the 2DEG, GH = NL2e2 =h.
This is not true for the situation in Fig. 12a and we discuss here that for
describing transport through mesoscopic conductors one has to include the
properties of the measurement contacts.

Figure 12. Left: Geometry used to observe an anomalous quantization of the Hall
conductance, for which QPC A is used as current probe and QPC B as voltage probe.
Right: Comparison between the Hall conductance GH and the 2-terminal conductance
of the current probe GA . The voltage probe conductance is kept xed at GB = 2e2 =h.
Although the number of occupied Landau levels in the 2DEG is unchanged, GH follows
the largest probe conductance (from Ref. [27])

We rst assume that the edge channels shown in Fig. 12a are inde-
pendent, i.e. no scattering events occur between di erent edge channels
or, equivalently, electrons travel with conservation of quantum-subband-
number. In this case the transport is adiabatic. The edge channels are oc-
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 17
cupied up to the electrochemical potential of the last Ohmic contact they
have left. (We assume ideal Ohmic contacts, meaning that all incoming
electrons are absorbed and all outgoing states are occupied up to an av-
erage electrochemical potential.) Consequently, in the region between the
two QPCs, the two edge channels have an unequal population. Channel-2 is
occupied up to 1 , while current probe 5 populates channel-1 up to 5 . This
process is called selective population of edge channels. If the voltage probe
B detects all edge channels like an ideal Ohmic contact, the regular quan-
tum Hall conductance is measured. However, this is not the case in Fig. 12a
where the voltage probe selectively detects only the rst channel. The sec-
ond channel is neither populated by the current probe nor detected by the
voltage probe and, therefore, is not measured. The Hall conductance G54;61
equals 2e2=h instead of 4e2=h which a regular Hall measurement would give.
The current and voltage contacts do not measure all the 2DEG properties,
but only those properties they "see", or couple to. In general one can show
that in the absence of inter-edge channel scattering this Hall conductance
is given by [27]
GH = maxfGA ; GBg (11)
implying that GH is completely determined by the characteristics of the
probes and is independent of the number of occupied Landau levels in the
2DEG.
Fig. 12b compares the measured probe conductances GA and GB with
the Hall conductance GH . The magnetic eld is kept xed at 3.8 T, corre-
sponding to NL = 2 in the bulk 2DEG. The voltage on QPC B , de ning the
voltage probe, is also xed such that only the rst edge channel is transmit-
ted, and therefore GB = 2e2 =h. The voltage VgA on QPC A, de ning the
current probe, is varied, resulting in a decreasing GA . Again we note that a
normal Hall measurement would give a constant conductance GH = 4e2 =h.
However, Fig. 12b shows a Hall conductance which virtually follows the
largest probe conductance in agreement with Eq. 11 (for VgA > ,1:5V,
GH = GA > 2e2=h; for ,2:1V < VgA < ,1:5V, GH = GA = GB = 2e2=h
corresponding to the edge channel ow of Fig. 12a; and for VgA < ,2:1V,
GH = GB = 2e2=h). The experiment demonstrates that on short distances
of order m the transport through edge channels is adiabatic, implying
that they can be viewed as independent 1D current channels. Subsequent
experiments have shown that a non-equilibrium population can persist up
to larger distances (several tens of m's) [28]. In particular, it is found that
the top most channel is virtually decoupled from the lower edge channels
even over macroscopic distances of several times 100 m [28, 29, 30, 31, 32].
These experiments have clari ed the important role of measurement probes
in the QHE regime.
18 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN

Figure 13. Two-terminal conductance of a quantum point contact at B = 7T. At zero

gate voltage the conductance is G = e2 =h implying that the 2DEG is spin-polarized at
lling factor 1. On decreasing the gate voltage fractional plateaus are observed (from
Ref. [37].

1.6.3. The Importance of Electron-Electron Interactions for the QHE on

Mesoscopic Length Scales
In our discussion we neglected Coulomb interactions between the electrons.
For many experimental observations it is sucient to give a description
in terms of a non-interacting electron model. There are several exceptions,
however. First of all, in very high mobility samples not only are plateaus
observed at integer lling factors, but also at fractional lling factors (see
Fig. 2). The theory for this fractional QHE needs to incorporate electron-
electron interactions [33]. Recently, a new appealing formulation for the
fractional QHE has been given in terms of composite particles [34]. Another
recent development is the prediction [35] and observation [36] of so-called
skyrmions.
Also, on mesoscopic length scales it has recently become appreciated
that electron-electron interactions can be important under certain circum-
stances. First of all, the FQHE can also be observed at short ballistic length
scales [37]. Fig. 13 shows the conductance of a QPC at 7 T. At zero gate
voltage the conductance is equal to e2 =h showing that the lling factor of
the bulk 2DEG is equal to one (i.e. spin resolved). On making the gate
voltage more negative the conductance decreases. However, at fractional
values 2/3 and 1/3 times e2 =h it shows quantized plateaus. The QPC used
in this experiment shows quantized conductance at B = 0 implying that
fractional plateaus can exist in a ballistic sub-micron device [37, 38]. Also,
selective population and detection experiments of fractional edge channels
have been performed [39, 40] in a similar way as in Section 1.6.2 for the
case of integer edge channels. These experiments on the fractional QHE in
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 19
the ballistic regime can not be explained with the non-interacting electron
models of Sections 1.6.1 and 1.6.2. Instead interacting models have been
proposed by Beenakker [41], MacDonald [42] and Chang [43]. These models
are the basis of recent models on self-consistent calculations of edge chan-
nels [44]. From these electrostatic calculations it follows that edge channels
have a nite width. Regions where the Fermi energy is pinned in a Landau
level are called a compressible liquid whereas regions where the Fermi en-
ergy is between Landau levels are called an incompressible liquid. Although,
Coulomb interactions give edge channels a nite width the selective popu-
lation experiments suggest that they remain 1D channels. In 1D Coulomb
interactions are known to give rise to a new type of electron liquid. This
so-called Luttinger liquid is discussed by Fisher and Glazman in this vol-
ume including predictions for correlated electron transport in fractional
edge channels. Altogether, we can conclude this Section with the statement
that the QHE is still a very interesting system for studying new electronic
properties.

2. Single{electron tunneling
2.1. REVIEW
The electron transport in two-dimensional electron gases displays the quan-
tum coherence of the electronic states, as described in the previous Section.
Another important aspect of mesoscopic systems is the role of interactions.
We mentioned already their importance in connection with the fractional
QHE. Interaction e ects also in uence strongly electron tunneling through
mesoscopic systems, and they lead to strong correlation e ects. In this
Section we will discuss situations where it is sucient to approximate the
Coulomb interaction by an e ective capacitance model. This applies for
metal junction systems where after a relaxation of the space distribution
of the electrons the remaining Coulomb interaction is well described by
the geometric capacitance of the junctions. The capacitance model also
works remarkably well for the transport through small quantum dots. (See
the Chapter of Kouwenhoven et al. for many facts on quantum dots.) The
\charging e ects" allow us to control single electron charges, which leads
to a variety of single{electron e ects, e. g. to the suppression of tunneling,
a phenomenon known as the Coulomb blockade.
Single{electron e ects have been studied for more than a decade, and
a large number of papers, incl. several reviews have been devoted to the
topic. Kulik and Shekhter [45] and Averin and Likharev [46] developed
the perturbation theory of single{electron tunneling in metal junctions and
discussed several consequences. Initial scepticism against the new concepts
20 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN
was overcome when experiments proved to be successful. After early experi-
ments by Fulton and Dolan [47], important breakthroughs were achieved in
Delft by Mooij and Geerligs and in Saclay by Devoret, Esteve, Urbina and
further members of these groups [48, 49]. Their work is well summarized in
the proceedings of an earlier NATO ASI Single Charge Tunneling [50].
In this Section we introduce the concepts and basic description of single{
electron tunneling in systems with strong charging e ects. For de niteness
we rst consider metallic electrodes with a large density of electron states.
We study how the charging energy depends on the number of electrons
and the transport and gate voltages applied to various parts of the system.
The simplest model systems demonstrating these features are the \single{
electron box" and the \single{electron transistor". We then derive within
perturbation theory the single{electron tunneling rates. In low capacitance
systems it is crucial to account for the change in the charging energy associ-
ated with the tunneling process. A master-equation description accounts for
the large-scale features of the current-voltage characteristic of the single{
electron transistor. In the Coulomb-blockade regime, where single{electron
tunneling is suppressed, higher-order processes such as coherent \cotunnel-
ing" of electrons through several junctions become observable. Finally we
discuss single{electron tunneling through quantum dots, where the level
quantization becomes observable as well..
Several extensions have been described in the literature. We mention
only a few:
{ The mesoscopic junction systems studied here are small such that charg-
ing e ects and higher-order quantum processes play a role. On the other
hand, they are large enough such that macroscopic current and voltage
probes and sources can be coupled to the system. This makes the meso-
scopic system susceptible to the in uence of the electric circuit. A complete
description has to include this circuit. The in uence of the electrodynamic
environment on single{electron tunneling has been reviewed in the article
by Ingold and Nazarov in Ref. [50].
{ In this introduction we study tunnel junctions with two normal-conducting
electrodes (NN). If the system or part of it is superconducting the combi-
nation of single{electron tunneling, Cooper pair tunneling and Andreev
re ection leads to further highly interesting e ects [51]. For a review see
the Chapter of Fazio and Schon in this volume.
{ Many of the single{electron e ects can be described within simple pertur-
bation theory. A necessary requirement is that the resistance of the tunnel
barriers is high compared to the quantum resistance RK = h=e2  25:8 k .
For tunnel junctions with lower resistance a more general formulation is
needed [52, 53, 54]. A systematic description of tunneling in systems with
strong charging e ects is presented later in the Chapter of Schoeller.
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 21
(a) (b)

Figure 14. a) An overlap junction with an oxide layer (hatched region), b) schematic
diagram for a tunnel junction.

2.2. CHARGING ENERGY AND SINGLE{ELECTRON DEVICES

2.2.1. The Energy Scale
Modern lithographic techniques allow the fabrication of narrow metallic
lines with widths down to approximately 20 nm, as well as tunnel junctions
in overlap regions of such lines, as illustrated in Fig. 14. The structures are
grown by evaporation of the metal, e. g. Aluminum, through masks onto
the substrate. Tunnel junctions can be produced by shadow evaporation
techniques, which involves two stages of evaporation at di erent angles.
Between the two stages the rst layer is allowed to oxidize. The junction is
formed in the overlap region. They can be produced reliably with areas of
S = (100 nm)2 and below. The oxide layer is roughly d = 10  A thick, and
the dielectric constant of the oxide is   10. Using the classical expression
for the capacitance we arrive at C = S=(4d)  10,15F.
The capacitance introduces an energy scale, the charging energy corre-
sponding to a single{electron charge (,e),
2
EC  2eC ; (12)
which characterizes all charging e ects. It is of the order of EC  10,4eV
if the capacitance is C = 10,15 F , which corresponds to a temperature
EC=kB  1K. In a tunneling process the electrostatic energy changes by an
amount of the order of magnitude of EC . Hence we expect in the sub-Kelvin
regime electron transport to be a ected by charging e ects. Similar prop-
erties have been observed in semiconductor nanostructures, for instance
in quantum dots in 2-dimensional electron gases. The Coulomb energy in
these systems again can be characterized by a capacitance which depends
on the size of the dot and also may lie in the range of 10,15F or less.
Charging e ects play a role in granular materials and ultimately even in
molecular systems. Here the capacitance may be as low as 10,18 F, making
single{electron tunneling observable even at room temperatures.
2.2.2. Single{Electron Box
We now analyze in more detail the charging energy of simple systems of
tunnel junctions. It depends on the electron number in various parts of
the particular system and the applied voltages. The rst example is the
22 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN
n

CJ CG

VG

Figure 15. The single{electron box.

single{electron box, shown schematically in Fig. 15. It consists of a small

metallic island, coupled via a tunnel junction with capacitance CJ to an
electrode and via a capacitor CG to a gate voltage source VG. We choose
the reference such that for VG = 0 the lowest energy state of the system is
charge-neutral, i. e. the electrons on the island compensate the charge of the
ions; consequently there are n = 0 excess electrons on the island. If a gate
voltage is applied the number of excess electrons on the island can change
due to tunneling across the junction in discrete steps to n = 1; 2; :::.
While the total number of electrons on an island is integer, the charge
is spatially distributed and in general shifted relative to the positive back-
ground. If a voltage is applied the surface charges on the capacitor plates,
which are of equal magnitude but opposite sign on the two sides of each
junction, are in general non-quantized. They are determined by the integer
n and the non-quantized applied voltage. We obtain the charging energy
from the following elementary arguments: the total excess charge of the
box splits into two parts on the left and right capacitor plate ,ne =
QL + QR. The corresponding voltage drops add to the applied voltage
VG = QL=CJ , QR =CG, and the charging energy is Q2L =2CJ + Q2R =2CG.
The relevant free energy is a Legendre transform of this energy, which also
contains the work done by the voltage source ,VGQR . Elimination of QL
and QR in favor of n and VG yields, up to a contribution which does not
depend on the variable n, the result

Ech(n; QG) = (ne ,2CQG) :

2
(13)
Here C = CJ + CG is the total capacitance of the island. The e ect of the
voltage source is contained in the \gate charge" de ned as QG = CGVG .
The charging energy Ech is plotted in Fig. 16 as function of the gate
charge for di erent excess electron numbers n. With increasing gate voltage,
the electron number corresponding to the lowest energy state increases. It
does so in discrete steps from n to n + 1 at the degeneracy points QG =e =
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 23

Ech(n,QG)/EC
1

n=0 n=1 n=2

0
-1 0 1 2
QG/e

Figure 16. The charging energy of a single{electron box as a function of the gate voltage
for di erent numbers n of electron charges on the island.

1.5

0.5
<n>

-0.5

-1

-1.5
-1.5 -1 -0.5 0 0.5 1 1.5

QG/e

Figure 17. The average number of electron charges hni on the island of a single{electron
box as a function of the gate charge (voltage) for di erent temperatures T=EC = 0
(dashed steps), 0.02, 0.05, 0.1, 0.2, 0.4, and 1 (nearly linear).

n + 1=2. Under the same conditions, the voltage of the island, Visland =
@Ech=@QG, displays a sawtooth dependence on the applied voltage.
At nite temperatures the steps and sawtooth dependence are washed-
out, as follows from the classical statistical average
1
hn(QG)i = Z1
X
n e,Ech (n;QG )=kB T ; (14)
ch n=,1
where Zch is an obvious normalization. The result is displayed in Fig. 17 for
di erent temperatures. The stepwise increase has been observed experimen-
tally, e.g. by the Saclay group (see results in Ref. [50]). Their measurement
24 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN
n

CL CR
CG

VG
VL VR

Figure 18. The SET transistor.

procedure will be discussed below. The experiments agree well with theo-
retical expectations. However, it is crucial to control heating and the noise
from the measurement setup, which usually is at a temperature higher than
that of the cryostat.
2.2.3. Single{Electron Transistor
Another fundamental example is provided by the single{electron transistor
shown in Fig. 18. Here an island is coupled via two tunnel junctions to
a transport voltage source V = VL , VR such that a current can ow.
The island is, furthermore, coupled capacitively to a gate voltage VG. The
charging energy of the system depends again on the integer number of
electrons n on the island and the continuous voltages. Some algebra along
the lines outlined for the electron box produces again Ech(n; QG) = (ne ,
QG )2=2C . For the transistor C = CL + CR + CG is the total capacitance
of the island, i.e. the sum of the two junction capacitances and the gate
capacitance, and all three voltage sources de ne the gate charge
QG = CGVG + CLVL + CRVR : (15)
In a tunneling process, increasing the island charge from n to n + 1, the
charging energy changes by
  2
Ech(n + 1; QG) , Ech(n; QG) = n + 2 , e eC :
1 Q G (16)
The second energy di erences are equally spaced and can be tuned by the
gate voltage. The situation is illustrated in the energy scheme shown in
Fig. 19. The di erences in charging energy are plotted in the center. We
further display the Fermi levels of the two leads which are shifted by the
applied potentials VL=R.
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 25

Ech(2,Q G) - Ech(1,QG)

eV L
Ech(1,Q G) - Ech(0,Q G)

eVR
Ech(0,Q G) - Ech(-1,Q G)

Ech(-1,Q G) - E ch(-2,Q G)

eVG

Figure 19. In the island the energy di erences corresponding to the addition or removal
of an electron charge are shown. They can be shifted by the gate voltage VG . The elec-
trochemical potentials of the leads are shifted relative to each other by the transport
voltage V = VL , VR .

For de niteness we assume that the energy of the electrons in the left
lead is higher than that in the right lead. Then at low temperature tunneling
from the left lead to the island (transition from n to n +1) is possible if the
electrochemical potential in the left lead eVL is high enough to compensate
for the increase in charging energy of the island
eVL > Ech(n + 1; QG) , Ech(n; QG) : (17)
Similarly tunneling from the island (transition from n +1 to n) to the right
lead is possible at low temperature only if
Ech(n + 1; QG) , Ech(n; QG) > eVR : (18)
Both conditions have to be satis ed simultaneously in order for a current
to ow through the transistor. It is obvious from the gure that at low
transport voltages, depending on the gate voltage VG we may be either in
a Coulomb blockade regime or have a nite current. By varying the gate
voltage we produce the Coulomb oscillations, i.e. the e-periodic dependence
of the conductance on QG .
Additional devices can be examined (for a review see Esteve's article
in Ref. [50]): (i) The electron trap is similar to the electron box except
that it contains at least two junctions in series. In contrast to the box
the trap has metastable charge states. (ii) Two traps can be combined to
build the electron turnstile, which can serve as a current source [48]. A
suitable ac-gate voltage with frequency f allows the controlled transfer of
26 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN
a single{electron per cycle. Hence the current is I = ef . (iii) Finally, in
single{electron pumps a current is driven by two phase-shifted ac-voltages
applied to di erent islands [49]. In this case a current I = ef is transported
even at vanishing transport voltages. The turnstile and pump can serve
as a current standards, if one manages to minimize the e ect of missed
cycles, of thermal uctuations, and of quantum uctuations. This requires
low frequencies, low temperatures, and a design (many junctions) which
minimizes higher order quantum tunneling processes.
Many properties of single{electron systems can be understood by con-
sidering only the energy of di erent charge con gurations. However, a de-
tailed understanding of the I -V characteristics requires knowledge of the
tunneling rates of the electrons, which will be next topic.
2.3. TUNNELING RATES AND I -V CHARACTERISTICS
In this Subsection we introduce the Hamiltonian of the SET transistor.
Using simple golden-rule arguments we derive the rate for the transfer of a
single electron charge across the tunnel barriers. It depends crucially on the
change in the charging energy. The transition rates enter a master equation,
from which we obtain the current-voltage characteristic. If a tunneling pro-
cess would increase the charging energy it is suppressed at low temperature.
This phenomenon is called \Coulomb blockade". This \orthodox theory"
was developed in Refs. [45, 46].
2.3.1. The Single{electron Tunneling Rate
For de niteness we consider a SET transistor, shown in Fig. 18, which
consists of a metallic island coupled via tunneling barriers to two leads and
capacitively to an external gate voltage. Its Hamiltonian is
H = HL + HI + HR + Hch + Ht : (19)
Here, HL = k; k cyk; ck; describes the noninteracting electrons with wave
P

vector k in the left lead, with similar expressions for the island HI (with
states denoted by q ) and the right lead HR . We allow that the leads have dif-
ferent electrochemical potentials. The Coulomb interaction Hch is assumed
to depend only on the total charge on the island, as discussed above,

Hch = (^ne ,2CQG ) :

2
(20)
The
P
number operator of excess electrons on the island is given by n^ =
y
q; cq; cq; , N+ , where the number of positively charged ions of the
island has been subtracted. Charge transfer processes are described by the
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 27
standard tunneling Hamiltonian, for instance tunneling in the left junction
between the states k and q by
X
Ht;L = Tk;q cyk; cq; + h:c: : (21)
k;q;
We determine the transition rates by golden-rule arguments. The tun-
neling rate of an electron in the left junction is
Z 1 Z 1
+ (n) = 1 d d f ( )[1 , f ( )] (E +  ,  ) : (22)
L e2 Rt;L ,1 k q L k I q ch q k
,1
It describes the tunneling from one of the many states k in the left lead
to one of the available states q in the island, In this process the electron
number is increased from n to n + 1. The crucial point is that the energy,
which is conserved as expressed by the  -function, contains the energies of
the electron states k=q , but also the charging energy. The latter depends on
the electron number and the applied voltages VG and VL=R. In the process
considered it changes by
Ech = Ech(n + 1; QG) , Ech(n; QG) , eVL : (23)
We further introduced the tunnel conductance of the left junction
1 = 4e2 N (0) N (0) jT j2 : (24)
Rt;L h I I L L
It depends on the tunnel matrix elements Tk;q , which at this stage can
be considered as constants, as well as the densities of states at the Fermi
level, NI=L(0), and the volumes, I=L, of the island and lead. Equivalent
expressions apply for the reverse process L, (n + 1), decreasing the island
charge from n +1 to n, and for the tunneling processes in the other barriers.
In equilibrium the distribution functions fI=L are Fermi functions, and
the integrals over the electron states in Eq. (22) can be performed explicitly.
The resulting \single{electron tunneling" (SET) rate is [46]
+ 1 Ech
L (n) = e2Rt;L exp[Ech=kBT ] , 1 : (25)
At low temperatures, kB T  jEchj, if the charging energy would increase
in a tunneling process, the tunneling is suppressed, ! 0. This phe-
nomenon is called \Coulomb blockade" of electron tunneling. If charging
energy is gained the rate is
+(n) = 1 jE j for E  0 ; T ! 0 : (26)
L e2Rt;L ch ch
28 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN
At nite temperatures all processes are allowed. The forward and backward
rates satisfy detailed balance, L+(n)= L,(n + 1) = exp[,Ech=kB T ].
A familiar limit of what is described above is a single voltage-biased tun-
nel junction, where Ech is replaced by ,eV , independent of n. In this case
(25) yields a linear current- voltage relation, It = e[ + , , ] = V=Rt. We
can also reverse the argument. The two requirements | (i) a linear char-
acteristic in the voltage-biased case and (ii) detailed balance | uniquely
determine the expression for the rate to be of the form (25).
2.3.2. Master Equation for Sequential Tunneling
Given the electron tunneling rates we can set up a master equation for
the probability P (n; t) to nd the island in a state with n electrons. The
probability changes by tunneling in the left and right junctions. Hence
d + , + ,
dt P (n; t) = , [ L (n) + L (n) + R (n) + R (n)] P (n; t)
+ [ L+(n , 1) + R+ (n , 1)] P (n , 1; t)
+ [ L,(n + 1) + R, (n + 1)] P (n + 1; t) : (27)
The rates and probabilities also determine the current. In the left junction
the current is
X
IL(t) = ,e [ L+(n) , L,(n)] P (n; t) : (28)
n
In most cases we apply dc-voltages and are interested in the dc-current. In
this case we need only the stationary solution of the master equation, and
the currents in the left and right junctions are equal I = IL = IR .
As an example we consider a junction with symmetric bias VL = ,VR =
V=2. At low temperatures and low transport voltages (except at symmetry
points) only two di erent charge states { and those transitions which con-
nect both { have an appreciable probability. For instance, if ne < QG <
(n +1)e we need to consider only P (n) and P (n +1) and the two transitions
+ +
L (n) and R (n) increasing the island, charge from n,to n + 1 electrons, as
well as the two reverse transitions L (n + 1) and R (n + 1). The energy
changes determining the rates L+(n) and L, (n + 1) are
"  2 #
Ech =  n + 2 , e eC , eV2 ;
1 Q G (29)
respectively, while for the transitions in the right junction eV is replaced
by ,eV . In the 2-state limit the stationary probability and current become
, ,
P (n) = L (n + 1) + R (n + 1) ; P (n + 1) = 1 , P (n)

 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 29

2IRtC/e
1

0.5

2
1.5
0.5 1 VC/e
1 0.5
QG/e 1.5
2
Figure 20. The current of a symmetric transistor is shown as a function of gate and
transport voltage. At low temperatures and low transport voltages V C=e < 1 only two
charge states play a role, and the Coulomb oscillations are clearly demonstrated. At larger
transport voltages more charge states are involved.
+ , + ,
I = ,e L (n) R (n + 1) , R (n) L (n + 1) ; (30)

where  = L+ (n) + +(n) + ,(n + 1) + ,(n + 1).
R L R
This expression is readily analyzed by inspection of (29). At low tem-
peratures the tunneling process in the left junction from n to n + 1, with
rate L+ (n), is allowed when QG , (n +1=2)e  ,V C=2. On the other hand,
the transition which carries on the charge to the right electrode with rate
,
R (n + 1) is allowed when QG , (n + 1=2)e  V C=2. Both coexist in a
window of width CV around QG = (n +1=2)e. The other two processes are
not allowed simultaneously, in fact they are suppressed in the window just
mentioned. Therefore, at low temperature the current is
"  2#
1 4 e Q
I = 4R V , C 2 V e , n , 2 G 1 for , V2C  QG , n , 1  V C ;
t e e 2 2e
(31)
while it vanishes outside the window. For simplicity we have assumed in (31)
that the two junctions have the same tunneling resistance Rt = Rt;L = Rt;R.
The current through a symmetric SET transistor is plotted as a function
of transport and gate voltages in Fig. 20. For gate voltages such that QG=e
is close to an integer, the current vanishes below a threshold bias voltage
Vth(QG = ne) = e=C . This is a manifestation of the Coulomb blockade.
At non-integer values of QG=e the threshold voltage is lower Vth(QG ) =
30 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN
3
2.5
I(Rt,L+Rt,R)C/e

2
1.5
1
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
VC/e
Figure 21. Coulomb staircase: The current of an asymmetric transistor with di erent
tunneling resistances in the two junctions Rt;R = 10Rt;L is shown as a function of the
transport voltage for QG = 0 (pronounced Coulomb blockade, QG =e = 0:25 (intermedi-
ate), and QG =e = 0:5 (linear at low voltage).
minn f2jQG , (n + 1=2)ej=C g. One nds a series of evenly spaced peaks
centered around half-integer values of QG =e = n + 1=2, each of parabolic
shape as given by Eq. (31). These are called \Coulomb oscillations".
The strong dependence of I (QG; V ) on the gate voltage makes the SET
transistor a highly sensitive \electrometer". Small changes of polarization
charges by fractions of an electron charge in uence a macroscopic measure-
ment current. It has been used, for instance, to measure the charge in the
electron box hn(QG)i in Fig. 15.
For larger transport voltages more charge states play a role even at
low temperatures. For illustration we consider a junction with symmetric
bias VL = ,VR = V=2 and QG = 0, where the lowest energy state has
n = 0 electrons in the island. At transport voltages exceeding a threshold
Vth;0 = e=C tunneling sets in to the state with n = 1. Above this voltage
the electrochemical potential in the left lead is sucient to compensate
the increase in charging energy of the island. Since this state with n = 1
is unstable against a tunneling process in the right junction, a current is
transported through the system. At the same voltage tunneling processes
involving the state with n = ,1 are possible. At still higher voltages further
charge states jnj  1 play a role. This leads to a series of voltages Vth;n =
(2n +1)e=C , each marking the threshold above which another pair of charge
states becomes populated and a new channel for the conductance opens.
The increase in conductance is limited, however, due to the normalization
condition for the P (n). Still for suitable parameters (di ering conductances
of the two junctions or di erent capacitances) the current increases in the
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 31
shape of a staircase. The phenomenon got accordingly the name \Coulomb
staircase" [55]. The behavior is demonstrated in the plot of Fig. 21.
2.4. HIGHER ORDER TUNNELING PROCESSES
If sequential single{electron tunneling is suppressed by the Coulomb block-
ade, higher-order processes such as coherent \cotunneling" through several
junctions becomes crucial (Averin and Nazarov in Ref. [50]). As a spe-
ci c example we consider a SET transistor, biased such that the current
in lowest-order perturbation theory vanishes (see Fig. 19). At low tem-
peratures sequential tunneling is exponentially suppressed in this regime
since the energy of a state with an excess charge on the island lies above
the Fermi levels of the leads. On the other hand, if a transport voltage is
applied, a higher-order tunneling process transferring an electron charge
coherently through the total system is energetically allowed. In this case
the state with an excess electron charge in the island exists only virtually.
Standard second-order (or fourth, depending on the counting) perturbation
theory yields the rate
2
2 X hijHtj ih jHtjf i (E , E ) :
i!f = h E , EI i f (32)

The energy of the intermediate virtual state lies above the initial one,
E , Ei > 0, but it enters only into the denominator rather than into
the exponent of the sequential tunneling rate. Hence the higher-order rate
is nonzero even at very low temperatures.
When analyzing the process we have to pay attention to the following:
(i) There are actually two channels which add coherently. Either an electron
tunnels rst from the left lead onto the island, and then an electron tunnels
from the island to the other lead. In this case the increase in charging
energy of the intermediate state compared with the initial one is EL =
Ech(n + 1; QG) , Ech(n; QG) , eVL. Or an electron tunnels rst out of the
island to the right lead, and another electron from the left lead replaces
the charge. In this case the increase in energy of the intermediate state is
ER = Ech(n , 1; QG) + eVR , Ech(n; QG). Both amplitudes have to be
added before the matrix element is squared.
(ii) The leads contain a macroscopic number of electrons. Therefore, with
overwhelming probability the outgoing electron will come from a di erent
state than the one which the incoming electron occupies. Hence, after the
process an electron-hole excitation is left in the island, which explains why
it is called \inelastic" cotunneling. This scenario is visualized in Fig. 22.
32 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN

Figure 22. In an inelastic cotunneling process two electrons tunnel coherently, i.e. in a
single qutnum process, in the left and right junction. The result is the transfer of charge
through the system even in the Coulomb blockade regime. A particle-hole excitation is
left in the island.

Transitions involving di erent excitations are added incoherently. The

resulting rate for inelastic cotunneling is
h
cot = 2e4R R
t;L t;R
Z Z Z Z
 dk dq dq0 dk0 f (k )[1 , f (q )]f (q0 )[1 , f (k0 )]
k2L q2I q 0 2I k0 2R
" #2
1 1
  + E ,  +  0 + E ,  0 (eV + k , q + q0 , k0 ) : (33)
q L k k R q
At T = 0 the integrals can be performed analytically. The result for in the
Coulomb blockade regime (eV  EL; ER) is
 2
h
 1 1 V3 :
cot = 12eR R + (34)
t;L t;R EL ER
At nite temperatures forward and backward processes occur. They obey
a detailed balance relation cot(,V ) = exp(,eV=kBT ) cot(V ). The current
then is
 2 h
I (V )  12e2Rh R E1 + E1
i
(eV )2 + (2kBT )2 V : (35)
t;L t;R L R
In the Coulomb blockade regime of a SET transistor the V 3 depen-
dence of the cotunneling current has been observed. In systems with N
junctions a corresponding N-th order process (or 2N-th order, depending
on the counting) leads to a current I / V 2N ,1 . As an example we con-
sider N=4 junctions with C = 10,15F and tunneling resistance Rt. In this
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 33
,

case (see D. Esteve in Ref. [50]) cot = 2:5  10,3 =sec (V=V)7 (k =Rt)4 .
These cotunneling processes limit the accuracy of the single{electron turn-
stile even under the most favorable situations, i.e. low T and low frequency,
where thermally activated multi-electron transfer processes and missed cy-
cles play little role.
The expression for the cotunneling rate presented above displays sev-
eral important properties: (i) The expansion parameter is the dimensionless
tunneling conductance RK =Rt, where the quantum resistance RK serves as
reference. (ii) The approximate expression given diverges when the interme-
diate and initial or nal states are degenerate. This divergence is removed by
life-time broadening e ects. The complete cotunneling theory, recently an-
alyzed in Ref. [56], describes well the logarithmic temperature dependence
observed in th experiments of the Saclay group on junctions with tunneling
resistances comparable to the quantum resistance [57]. (iii) Higher order
processes and eventually resonant tunneling processes are most essential
near the points of degeneracy of the charging energy, QG =e = n + 1=2. The
Chapter of Schoeller in this volume presents the theoretical framework to
describe tunneling beyond perturbation theory.
There exists also the process of \elastic cotunneling" where one electron
tunnels through the total system, leaving no excitations in the island. It is
the dominant process in the Coulomb blockade regime of tunneling through
a single-level quantum dot. However, in the metallic junction it is usually
not important, since its rate is smaller by a prefactor / 1= INI (0) (i.e.
inversely proportional to the number of states of the island) compared to the
inelastic cotunneling rate. Thepexception is the range of very low voltages
and temperatures kB T; eV  EC = N (0), since elastic cotunneling yields
a current which is linear in the applied voltage.
Single{electron tunneling is also in uenced by the response and the uc-
tuations of the electromagnetic circuit where it is embedded. This is partic-
ularly important in single junctions. In this case Coulomb blockade e ects
can only be observed if the junction is in series with a nearby large resistor
of the order of the quantum resistance or larger. In systems of junctions the
tunneling resistance of one junction usually provides the required decou-
pling of the remaining junctions and single{electron e ects are observable.
Here we do not have the space to present this theory. It is reviewed in the
article by Ingold and Nazarov in Ref. [50].
2.5. SINGLE{ELECTRON TUNNELING THROUGH QUANTUM DOTS
In this Subsection we describe the in uence of Coulomb blockade phenom-
ena on single{electron tunneling through ultrasmall quantum dots. The
important di erence compared to the metallic case is the quantization of
34 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN
the energy levels inside the quantum dot with typical separation  . It can be
resolved in transport experiments, when the level spacing exceeds the tem-
perature  > kB T . In this case one encounters the phenomenon of resonant
tunneling through discrete levels.
Many of the experiments showing single{electron e ects in quantum
dots can be explained by lowest order perturbation theory in tunneling. As
in the metallic case, the theory is based on a classical master equation with
golden-rule rates which describe incoherent transport through the device.
This means that the electrons tunnel sequentially, i.e. they loose their phase
memory before the next tunneling process occurs. The \orthodox theory",
initially developed for metallic islands [45, 46], was later generalized to
quantum dots with discrete spectra [58, 59, 60, 61]. In this Section we de-
scribe this golden-rule theory for the simplest systems. Quantum dots with
exact many-body wave functions in the few electron limit have been stud-
ied in Ref. [62, 63, 64], while coupled quantum dots have been considered
in Ref. [65, 66]. The e ect of time-dependent elds have been described in
Refs [67, 68] for metallic junctions and in Ref. [69] for the Coulomb block-
ade model, and in Refs. [70, 71, 72] for the metallic case in the presence of
a heat bath. Various extensions, incl. experimental results and further ref-
erences, will be reviewed in the Chapter of Kouwenhoven et al. later in this
volume. A powerful theory which allows a consistent treatment of higher
order tunneling processes and further generalizations will be presented in
the Chapter by Schoeller.
The electron tunneling through a quantum dot is described by the
Hamiltonian H = Hres + HD + Ht, where
2 3
X X
Hres = 4 kr ayk;r ak;r + eVrn^r 5 ; (36)
r=L;R k;
X
HD = Esjs >< sj ; (37)
s
X X
Ht = Tklr ayk;r al;D + h:c: ; (38)
r=L;R k;l;

describe the reservoirs, the dot, and the tunneling, respectively.

The reservoirs are assumed to have noninteracting single-particle states
labeled by the reservoir index r, wave vector k and spin  . The voltage of
reservoir r is denoted by Vr , and n^ r denotes the total particle number.
The eigenstates of the isolated dot are denoted by js > with energy Es .
For the Coulomb blockade model, the states js > of the dot are speci ed
by the set of all occupation numbers for the single particle states: js >=
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 35
jfnl;Dg >. In this case, the dot energy is given by
X
Es = lDn^l;D + EC (n , nG )2 ; (39)
l;
where n^ is the particle number operator of the dot.PThe charging energy
EC = e2 =2C and the gate charge QG = ,enG = i=L;R;G Ci Vi coincide
with the expressions (13) and (15) introduced in the metallic case. The
general notation in terms of the many-body wave functions js > is intro-
duced here to include cases where the states of the dot cannot be described
by single particle states, see e.g. Refs. [73, 62, 63, 74].
The tunneling part describes charge transfer processes between the
reservoirs and the dot. The tunneling matrix elements are conveniently
combined in the spectral function
,r;ll0 (E ) = 2h Tklr Tklr 0  (E , kr ) ;
X
(40)
k
which depends on the energy and the single-particle states l; l0 involved.
A typical value of the spectral function de nes an energy scale , which
characterizes the intrinsic broadening of the single-particle states of the dot
by quantum uctuations from tunneling. If ,  kBT , thermal uctuations
dominate over quantum uctuations, and we can use golden-rule theory.
The reservoirs are treated as large systems in equilibrium described by
the grand canonical density matrix eq res. The tunneling rate for a transition
0
of the dot from state js > to js > when p = 1 charges have been added
from reservoir r to the dot is given by the golden-rule expression
p 2 X
eq 0 0 0 2
r;ss0 = h res( )j < sjHt j s > j  (Es,Es0 +E ,E0 ,pr ) :
0
nr (0 )=nr ()+p
(41)
Here  denote the states of the reservoirs with particle numbers nr ().
The energy conserving  -function includes the change r = eVr of the elec-
trostatic energy, which is regarded here as the e ective electrochemical
potential of reservoir r. The change of the electrostatic energy of the dot is
included in Es , Es0 .
Inserting the tunneling Hamiltonian Ht from Eq. (38) into the golden-
rule rate (41) yields
X
p
r;ss0 = ,r;ll0 (Es ,Es0 ,r ) < sjaylD js0 >< s0jal0 D js > frp (Es ,Es0 ) ; (42)
ll0
where fr+ (! ) = f (! , r ) is the Fermi distribution of reservoir r, while
fr, = 1 , fr+
36 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN
The rates can be used as an input for a master equation. Consequently,
the stationary dc-probability distribution Ps for the dot and the stationary
dc-tunneling current in reservoir r can be calculated from
X
0 = [ ss0 P (s0 ) , s0 s P (s)] ; (43)
sX
0

Ir = e [ r+;ss0 P (s0 ) , r,;s0 s P (s)] ; (44)

ss 0

P p P
with ss0 = rp r;ss0 . Current conservation r Ir = 0 follows from the
property X
[ r+;ss0 , r,;ss0 ] = ss0 (ns , ns0 ) ; (45)
r
where ns denotes the number of particles in the dot in state s.
The rates satisfy the detailed balance relation r,;s0 s = r+;ss0 = exp[ (Es ,
Es0 , r )]. As a consequence, the equilibrium solution of the master equation
(43) is the grand canonical distribution
P eq(s) = 1 e, (Es,ns) ; Z (46)
which applies when all electrochemical potentials are the same r = . In
this case the dc-current (44) is zero.
Using detailed balance we can write the tunneling rates as
+ = fr+ (Es , Es0 ) r;ss0 ; , , 
r;ss0 r;s0s = fr (Es , Es0 ) r;ss0 ; (47)
where
= r+;ss0 + r,;s0s

r;ss0 (48)
is the sum of tunneling `in' and tunneling `out' rates. As a consequence the
current (44) can be expressed as
X
  + 0 , 
Ir = e r;ss0 fr (Es , Es0 )P (s ) , fr (Es , Es0 )P (s) : (49)
ss0
A current can ow through the structure if both the tunneling `in' and
tunneling `out' rate are non-zero. At low temperatures we consider the
transition between two dot states sn $ sn+1 , with n and n + 1 particles
in the dot, respectively. For tunneling `in' from reservoir r we need
E =
Esn+1 , Esn < r , and a similar relation for tunneling `out' to reservoir r0 .
Both conditions are satis ed simultaneously if the excitation energy
E lies
in the window of the e ective potentials of the reservoirs, r0 <
E < r .
The energy- and state-dependence of the spectral function ,r;ll0 (! ) ac-
counts for the energy-dependence of the density of states in the reservoirs,
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 37
and mesoscopic uctuations of the conductance peak heights. Here we con-
centrate on the simplest case and set
,r;ll0 (! )  ll0 ,r ; (50)
In this case we nd from (42) that the sum of the tunneling `in' and `out'
rates de ned in (48), di ers from this scale
X

r;ss0 = ,r j < sjaylDjs0 > j2 (51)
l
only by a constant factor, since the Fermi functions have canceled
P
out.
Inserting this relation in (49), and using current conservation r Ir = 0 to
eliminate the term which is independent of the Fermi functions, we nd
Ir = e ,r,,r0 j < sjaylDjs0 > j2[P (s) + P (s0 )]
X X

r0 ss0 l
[fr(Es , Es0 ) , fr0 (Es , Es0 )] ; (52)
P
with , = r ,r . This formula is frequently used in the literature. In linear
response, it reduces to a form rst derived by Beenakker [59]. In this case
we set r =  + eVr, and nd
X
Ir = Grr0 (Vr , Vr0 ) ; (53)
r0
where the conductance is
Grr0 = ,e2 ,r,,r0 j < sjaylD js0 > j2 [P (s)eq + P (s0 )eq]f 0 (Es , Es0 , ) :
X X

ss0 l
(54)
In the general case, many excitation energies Es , Es0 can lie between
r and r0 and are relevant for transport. However, only those transitions
s0 ! s will occur for which the initial probability P (s0 ) is not too small.
For temperatures and bias voltages smaller than the level spacing  and the
charging energy Ech, only the ground states s0n of the dot will have nonzero
occupation probability. This means that only the energies

n = Es0n+1 , Es0n (55)
are relevant. The transition from s0n to an excited state sn+1 does not occur
since, upon increasing the gate voltage, the transition s0n ! s0n+1 occurs
sooner and afterwards the dot is already in the n + 1-particle ground state.
Thus, we obtain the same physical picture as shown in Fig. 19 with the
38 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN

Figure 23. Linear conductance versus  for two doubly degenerate levels with 1 = 0,
2 = 25,, T = 5,, EC = 75,, and ,L = ,R = ,=2. The distance between the second and
third main resonance is larger due to the nite level spacing. All resonances involving
excited states are hidden.

only di erence that the distance
between adjacent excitation energies is

no longer a constant.
This behavior is re ected in the formula (54) for the conductance matrix
in linear response. Due to the derivative of the Fermi function, the conduc-
tance will be maximal when  coincides with one of the excitation energies
within a range set by the temperature. The energy di erence Es , Es0 is
varied experimentally by the gate voltage. Thus, the conductance shows
a series of resonances with varying distance between the peaks and a line
shape which is approximately given by the derivative of the Fermi distri-
bution function. Between the resonances, transport is not possible and the
system is in the Coulomb blockade regime. As an example, we show the
Coulomb oscillations in Fig. 23 for the Coulomb blockade model with two
spin-degenerate levels with energies 1 < 2 . According to (39) the excita-
tion energies, describing the energy changes of the dot when a particle is
added in level l to a state with n particles, are given by

nl = lD + 2nEC : (56)
Here lD = lD + EC (1 , nG ) describes the e ective level position, which is
tuned by the gate voltage. As a consequence, we observe four resonances
corresponding to the excitation energies
01 = 1 ,
11 = 1 + 2EC ,
22 =
2 + 4EC, and
32 = 2 + 6EC . As explained above, all other excitation
energies
02 = 2 ,
12 = 2 + 2EC,
21 = 1 + 4EC , and
31 = 1 + 6EC
are hidden because they involve excited states.
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 39

Figure 24. The dc current in nonlinear response versus eV = e(VL , VR ), with

VL = ,VR = V=2 and CG VG =e = 1 xed, for two doubly degenerate levels with 1 = 0,
2 = 50,. The other parameters are T = 5,, EC = 75,, and ,L = ,R = ,=2. All
one-particle excitations of the dot are visible.

At nite bias voltage all excitations are in principle visible since the
excited states acquire a nite occupation probability. This holds at least
in the absence of certain selection rules arising from the matrix element
< sjaylD js0 > in (52). For a constant density of states of the leads the I-V-
characteristic shows steps each time a new excitation becomes relevant. This
result is shown in Fig. 24 for the same example as before. Equivalently, the
di erential conductance dI=dV shows peaks as function of the bias voltage.
As can be seen, all eight excitation energies mentioned before are visible.
The e ects of strong correlations on the dot are not only re ected in
the increase of the distance between adjacent resonances but also in the
line shape of an individual peak as a function of the applied gate voltage.
To show this explicitly, we consider the Coulomb blockade model for a
single spin-degenerate state with energy  on the dot. For EC  T we can
restrict ourselves to the transition between an empty and a singly occupied
dot, n = 0; 1. According to relation (56), the excitation energy for this
transition is given by
= . From the master equation and (52) we nd
Ir = 2e ,r,,r0 P 1,r00
X
[fr (
) , fr0 (
)] ; (57)
r0 1 + r00 , fr00 (
)
Grr0 () = ,2e2 ,r,,r0 1 + f (
1 0
, ) f (
, ) : (58)
P
The current contains an asymmetry factor 1=(1 + r ,,r fr (
)) which is
absent either for a nondegenerate level or for the noninteracting case EC =
40 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN
0. This factor arises from correlations since double occupancy of the dot
is forbidden. This gives rise to particle-hole asymmetry and, consequently,
to an asymmetric line shape of the di erential conductance as a function
of
as shown in Fig. 25 for nite bias voltage. The maximal value of the
conductance in linear response is given by

Gmax = 4 e2 ,r,r0 2 : (59)

rr0
h , 3T
For a nondegenerate level or for the noninteracting case with one degen-
erate level, the factor 2=3 has to be replaced by 1=2 or 1, respectively.
This can be easily understood. At the maximum of the conductance sev-
eral states of the dot have the same probability. For large charging energy
the doubly occupied state can be excluded, and we have two degenerate
excitations which can be used for transport, and three possible states of
the dot (the empty dot and two degenerate states with one electron). Each
excitation contributes equally to the current but has to be multiplied with
the probability 1=3 of the initial state. This explains the factor 2=3. For
a nondegenerate level we have only one excitation and two states, result-
ing in a factor 1=2. For a noninteracting model with one degenerate level
we have four excitations (two for each transition n = 0 ! n = 1 and
n = 1 ! n = 2) and four possible states, giving a factor 1. The reduc-
tion of the current by Coulomb repulsion is obvious, since certain processes
are blocked. In contrast to the noninteracting case, we have seen that the
presence of degenerate states does not give rise to a pure multiplicative
factor of the degeneracy. The reason is that Coulomb interaction induces a
correlation between the levels. When one level is occupied, the other is not
allowed to be occupied due to the strong on-site Coulomb repulsion.
Selection rules occur due to the matrix element j < sjaylD js0 > j2 in (52).
Spin conservation allows only transitions where the total spin of the states
s and s0 di ers by 1=2. For the discussion of spin blockade e ects and
related negative di erential conductances we refer to Refs. [62, 63, 74].
Conclusions
Many further fundamental concepts of mesoscopic electron transport
could not be included in this Introduction, but will be covered in follow-
ing Chapters. Interference and localization e ects and extensions will be
discussed in the Chapters of Stern and of Imry. The Chapter by Kouwen-
hoven, Markus, McEuen, Tarucha, Westervelt, and Wingreen will cover the
wide eld of electron transport through quantum dots. Eaves includes in
his Chapter concepts related to chaos. De Jong and Beenakker review noise
properties of electron transport. Also Buttiker and Christen's Chapter deals
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 41

Figure 25. The di erential conductance as a function of
=  for a two-fold degen-

erate level with large charging energy EC so that double occupancy can be neglected.
T = 0:25,, L = ,R = 15,, and ,L = ,R = ,=2.

with extensions to time-dependent phenomena. A systematic discussion of

tunneling beyond perturbation theory is presented in Schoeller's Chapter,
while Fisher and Glazman discuss transport in 1D interacting systems.
Superconductivity adds further degrees of freedom to mesoscopic elec-
tron transport. The properties of normal metal { superconductor heterostruc-
tures are described in the Chapter of the Saclay group, while Fazio and
Schon describe single-charge tunneling in superconducting junction systems
and further review the theory of quantum transport in NS heterostructures.
The very existence of superconductivity in ultrasmall particles is investi-
gated in the article of Ralph et al., while van Wees and Takayanagi address
transport through semiconductor { superconductor systems.
The eld of scanning probe microscopy has advanced substantially in
recent years. It is reviewed by Sohn et al. The transport through quantum
point contacts still reveals new results as described in two Chapters by van
Ruitenbeek and by Garcia et al.
We included in this Book also two peripheral Chapters. Yamamoto de-
scribes concepts of quantum optics to solid state physicists, while DiVin-
cenzo reviews the novel eld of quantum computing.
References
1. Molecular Beam Epitaxy: Fundamentals and Current Status, M. A. Herman and
H. Sitter, Springer-Verlag, NY (1985); The Technology and Physics of Molecular
Beam Epitaxy, ed. E. H. C. Parker, Plenum Press, NY (1985)
2. H. van Houten, B. J. van Wees, M. G. J. Heijman, and J. P. Andre, Appl. Phys. Lett.
49, 1781 (1986).
42 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN
3. T. J. Thornton, M. Pepper, H. Ahmed, D. Andrews, and G. J. Davies, Phys. Rev.
Lett. 56, 1198 (1986).
4. H. Z. Zheng, H. P. Wei, D. C. Tsui, and G. Weimann, Phys. Rev. B 34, 5635 (1986).
5. S. E. Laux, D. J. Frank, and F. Stern, Surf. Sci. 196, 101 (1988).
6. H. Ehrenreich and D. Turnbull, eds. , Solid State Physics (Academic Press, New
York, 1991).
7. Yu. V. Sharvin, Zh. Eksp. Teor. Fiz. 48, 984 (1965) [Sov. Phys. JETP 21, 655
(1965)].
8. B. J. van Wees, H. van Houten, C. W. J. Beenakker, J. G. Williamson, L. P. Kouwen-
hoven, D. van der Marel, and C. T. Foxon, Phys. Rev. Lett. 60, 848 (1988).
9. B. J. van Wees, L. P. Kouwenhoven, E. M. M. Willems, C. J. P. M. Harmans,
J. E. Mooij, H. van Houten, J. G. Williamson, and C. T. Foxon, Phys. Rev. B 43,
12431 (1991).
10. L. I. Glazman, G. B. Lesovik, D. E. Khmel'nitskii, and R. I. Shekter, JETP Lett
48, 238 (1988); L. I. Glazman, and A. V. Khaetskii, J. Phys. : Condens. Matter 1,
5005 (1989).
11. K. F. Berggren, T. J. Thornton, D. J. Newson, and M. Pepper, Phys. Rev. Lett.
57, 1769 (1986).
12. R. Landauer, IBM J. Res. Dev. 1, 223 (1957); Phys. Lett. 85A, 91 (1981);
J. Phys. Condens. Matter 1, 8099 (1989).
13. A. Szafer and A. D. Stone, Phys. Rev. Lett. 62, 300 (1989); E. G. Haanappel and
D. van der Marel, Phys. Rev. B 39, 5484 (1989); G. Kirczenow, Phys. Rev. B 39,
10452 (1989).
14. Several authors have obtained transmission resonances in the calculated conduc-
tance of narrow constrictions, see Ref. [13].
15. J. Nixon and J. Davies, Phys. Rev. B 41, 7929 (1990).
16. J. G. Williamson, C. E. Timmering, C. J. P. M. Harmans, J. J. Harris, and
C. T. Foxon, Phys. Rev. B 42, 7675 (1990).
17. B. I. Halperin, Phys. Rev. B 25, 2185 (1982).
18. B. J. van Wees, L. P. Kouwenhoven, H. van Houten, C. W. J. Beenakker, J. E. Mooij,
C. T. Foxon, and J. J. Harris, Phys. Rev. B 38, 3625 (1988).
19. L. P. Kouwenhoven, B. J. van Wees, C. J. P. M. Harmans, J. G. Williamson, H. van
Houten, C. W. J. Beenakker, C. T. Foxon, and J. J. Harris, Phys. Rev. B 39, 8040
(1989).
20. H. van Houten, B. J. van Wees, J. E. Mooij, C. W. J. Beenakker, J. G. Williamson,
and C. T. Foxon, Europhys. Lett. 5, 721 (1988); C. W. J. Beenakker, H. van Houten,
and B. J. van Wees, Europhys. Lett. 7, 359 (1988); H. van Houten et al., Phys. Rev. B
39, 8556 (1989).
21. M. Buttiker, Phys. Rev. Lett. 57, 1761, (1986).
22. J. Spector, H. L. Stormer, K. W. Baldwin, L. N. Pfei er, and K. W. West,
Surf. Sci. 228, 283 (1990).
23. P. Streda, J. Kucera, and A. H. MacDonald, Phys. Rev. Lett. 59, 1973 (1987).
24. J. K. Jain and S. A. Kivelson, Phys. Rev. Lett. 60, 1542 (1988).
25. M. Buttiker, Phys. Rev. B 38, 9375 (1988).
26. M. Reed, ed. , Semiconductors and Semimetals (Academic Press, New York, 1990).
27. B. J. van Wees, E. M. M. Willems, C. J. P. M. Harmans, C. W. J. Beenakker, H. van
Houten, J. G. Williamson, C. T. Foxon, and J. J. Harris, Phys. Rev. Lett. 62, 1181
(1989).
28. G. Muller, D. Weiss, S. Koch, K. von Klitzing, H. Nickel, W. Schlapp, and R. Losch,
Phys. Rev. B 42, 7633 (1990).
29. S. Komiyama, H. Hirai, S. Sasa, and S. Hiyamizu, Phys. Rev. B 40, 12566 (1989).
30. B. J. van Wees, E. M. M. Willems, L. P. Kouwenhoven, C. J. P. M. Harmans,
J. G. Williamson, C. T. Foxon, and J. J. Harris, Phys. Rev. B 39, 8066 (1989).
31. B. W. Alphenaar, P. L. McEuen, R. G. Wheeler, and R. N. Sacks, Phys. Rev. Lett.
64, 677 (1990).
 INTRODUCTION TO MESOSCOPIC ELECTRON TRANSPORT 43
32. P. L. McEuen, A. Szafer, C. A. Richter, B. W. Alphenaar, J. K. Jain, A. D. Stone,
R. G. Wheeler, and R. N. Sacks, Phys. Rev. Lett. 64, 2062 (1990).
33. For a review see T. Chakraborty and P. Pietilainen, The fractional quantum Hall
e ect, Springer series in Solid-State Sciences 85 (Springer-Verlag 1988).
34. For a review see S. C. Zhang, Int. J. of Mod. Phys. B, 6, 25 (1992).
35. S. L. Sondhi, A. Karlhede, S. A. Kivelson, and E. H. Rezayi, Phys. Rev. B 47, 16419
(1993); and H. A. Fertig, L. Brey, R. Cote, and A. H. MacDonald, Phys. Rev. B 50,
11018 (1994).
36. S. E. Barrett, G. Dabbagh, L. N. Pfei er, K. W. West, and R. Tycko,
Phys. Rev. Lett. 74, 5112 (1995).
37. L. P. Kouwenhoven, unpublished.
38. G. Timp, R. Behringer, J. E. Cunningham, R. E. Howard, Phys. Rev. Lett. 63,
2268 (1989).
39. L. P. Kouwenhoven, B. J. van Wees, N. C. van der Vaart, C. J. P. M. Harmans
C. E. Timmering, and C. T. Foxon, Phys. Rev. Lett 64, 685 (1990).
40. A. M. Chang and J. E. Cunningham, Phys. Rev. Lett. 69, 2114 (1992).
41. C. W. J. Beenakker, Phys. Rev. Lett. 64, 216 (1990).
42. A. H. MacDonald Phys. Rev. Lett. 64, 220 (1990).
43. A. M. Chang, Solid State Commun. 74, 871 (1990).
44. P. L. McEuen, E. B. Foxman, J. Kinaret, U. Meirav, M. A. Kastner, N. S. Wingreen,
and S. J. Wind, Phys. Rev B 45, 11419 (1992); D. B. Chklovskii, B. I. Shklovskii,
and L. I. Glazman, Phys. Rev. B 46, 4026 (1992).
45. I. O. Kulik and R. I. Shekhter, Zh. Eksp. Teor. Fiz. 68, 623 (1975)
[Sov. Phys. JETP 41, 308 (1975)].
46. D. V. Averin and K. K. Likharev, J. Low Temp. Phys. 62, 345 (1986); and in
Mesoscopic Phenomena in Solids, B. L. Altshuler, P. A. Lee and R. A. Webb, eds. ,
p. 173 (Elsevier, Amsterdam, 1991).
47. T. A. Fulton and G. J. Dolan, Phys. Rev. Lett. 59, 109 (1987).
48. L. J. Geerligs et al. , Phys. Rev. Lett. 64, 2691 (1990).
49. H. Pothier et al. , Physica B 169, 573 (1991), Europhys. Lett. 17, 249 (1992)
50. Single Charge Tunneling, NATO ASI Series, Vol. B 294, eds. H. Grabert and
M. H. Devoret, (New York, Plenum Press 1992).
51. Mesoscopic Superconductivity, Proceedings of the NATO ARW, edited by
F. W. J. Hekking, G. Schon, and D. V. Averin, Physica B 203 (1994).
52. G. Schon and A. D. Zaikin, Phys. Rep. 198, 237 (1990).
53. L. I. Glazman and K. A. Matveev, Sov. Phys. JETP. 71, 1031 (1990); K. A. Matveev,
Sov. Phys. JETP 72, 892 (1991).
54. J. Konig, H. Schoeller, G. Schon, and R. Fazio, in Quantum Dynamics of Submicron
Structures, NATO ASI Series B 291, edited by H. Cerdeira et al. (Kluwer 1995),
p. 221; J. Konig, H. Schoeller, and G. Schon, Europhys. Letters 31, 31 (1995).
55. K. Mullen, E. Ben-Jacob, R. C. Jaklevic, and Z. Schuss, Phys. Rev. B 37, 98 (1988).
56. J. Konig, H. Schoeller, and G. Schon, Phys. Rev. Lett. 78, 4482 (1997).
57. P. Joyez, V. Bouchiat, D. Esteve, C. Urbina, and M. H. Devoret, preprint.
58. D. V. Averin and A. N. Korotkov, Zh. Eksp. Teor. Fiz. 97, 1661 (1990) [Sov.
Phys. JETP 70, 937 (1990)]; D. V. Averin, A. N. Korotkov, and K. K. Likharev,
Phys. Rev. B 44, 6199 (1991).
59. C. W. J. Beenakker, Phys. Rev. B 44, 1646 (1991).
60. Y. Meir, N. S. Wingreen, and P. A. Lee, Phys. Rev. Lett. 66, 3048 (1991).
61. H. van Houten, C. W. J. Beenakker, and A. A. M. Staring, in Ref. [50].
62. D. Weinmann, W. Hausler, W. Pfa , B. Kramer, and U. Weiss, Europhys. Lett. 26,
467 (1994); D. Weinmann et al Phys. Rev. Lett. 74, 984 (1995).
63. K. Jauregui, W. Hausler, D. Weinmann, and B. Kramer, Phys. Rev. B 53 R1713
(1996).
64. D. Pfannkuche and S. E. Ulloa, Phys. Rev. Lett. 74, 1194 (1995).
65. R. H. Blick, R. J. Haug, J. Weis, D. Pfannkuche, K. von Klitzing, and K. Eberl,
44 L.P. KOUWENHOVEN, G. SCHO N, AND L.L. SOHN
Phys. Rev. B 53, 7899 (1996).
66. G. Klimeck, G. Chen, and S. Datta, Phys. Rev. B 50, 2316 (1994); Phys. Rev. B
50, 8035 (1994).
67. L.P. Kouwenhoven, S. Jauhar, K. McCormick, D. Dixon, P.L. McEuen,
Yu.V. Nazarov, N.C. van der Vaart, and C.T. Foxon, Phys. Rev. B 50, 2019 (1994).
68. I. A. Devyatov and K. K. Likharev, Physica B 194-196, 1341 (1994).
69. C. Bruder and H. Schoeller, Phys. Rev. Lett. 72, 1076 (1994).
70. G. Ingold and Yu. V. Nazarov, in Ref. [50].
71. A. A. Odintsov, G. Falci, and G. Schon, Phys. Rev. B 44, 13089 (1991).
72. K. Flensberg, S. M. Girvin, M. Jonson, D. R. Penn, and M. D. Stiles, Phys. Scripta
T 42, 189 (1992).
73. D. Pfannkuche, V. Gudmundsson, and P. A. Maksym, Phys. Rev. B 47, 2244 (1993).
74. W. Hausler, Ann. Physik 5, 401 (1996).