Вы находитесь на странице: 1из 91

Coherent Electron Transport in Triple Quantum Dots

Adam Schneider
Centre for the Physics of Materials
Department of Physics
McGill University
Montréal, Québec
Canada

A Thesis submitted to the


Faculty of Graduate Studies and Research
in partial fulfillment of the requirements for the degree of
Master of Science

c Adam Schneider, 2008



Contents

Acknowledgments ix
1 Introduction 1
1.1 Experimental Background . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.1 Single Quantum Dots: The Coulomb Blockade . . . . . . . . . 2
1.1.2 Double Quantum Dots: Stability Diagrams . . . . . . . . . . . 5
1.1.3 Triple Quantum Dots: Chains and Rings . . . . . . . . . . . . 9
1.2 Theoretical Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.2.1 Transport Basics . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.2.2 Beyond Sequential Tunneling . . . . . . . . . . . . . . . . . . 18
1.2.3 Triple Dot Rings and Aharonov-Bohm Oscillations . . . . . . 20
1.2.4 What’s Next? . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2 Model and Results 25
2.1 Model Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.2 Zero-Bias Conductance . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.2.1 No Bath Coupling Limit . . . . . . . . . . . . . . . . . . . . . 33
2.2.2 Strong Bath Coupling Limit . . . . . . . . . . . . . . . . . . . 38
2.2.3 Aharanov-Bohm Oscillations . . . . . . . . . . . . . . . . . . . 42
2.3 Negative Differential Resistance and The “Dark State” . . . . . . . . 43
3 Details of Calculation 50
3.1 Quantum Master Equations . . . . . . . . . . . . . . . . . . . . . . . 50
3.1.1 Coherent System Evolution . . . . . . . . . . . . . . . . . . . 51
3.1.2 Lead-Dot Coupling: A Markovian Master Equation . . . . . . 52
3.1.3 Lead-Dot Coupling: FGR Transition Rates . . . . . . . . . . . 57
3.1.4 Bath-Dot Coupling: Dephasing and Eigenstate Transitions . . 61
3.1.5 Solving A Quantum Master Equation . . . . . . . . . . . . . . 63
3.2 Analytic Diagonalization . . . . . . . . . . . . . . . . . . . . . . . . 66
3.2.1 Equivalent 2-path Double Dot . . . . . . . . . . . . . . . . . . 68
3.3 Other Master Equations . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.3.1 The Analogous 2-Path Double Dot Master Equation . . . . . . 70
3.3.2 Tunneling into Charge States . . . . . . . . . . . . . . . . . . 72
4 Conclusions 74
References 77

iii
List of Figures

1.1 Network of capacitors and voltage nodes for a single quantum dot with
one gate voltage and two lead contacts having additional tunnel cou-
plings to pass a current. . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2 Network of capacitors and voltage nodes for a double quantum dot
with one gate voltage and one lead contact for each dot. The dots
are tunnel coupled to each other as well as to one lead each so that a
current can be passed through the dots. . . . . . . . . . . . . . . . . . 5

1.3 (a) Stability diagram for a double dot system with no inter-dot ca-
pacitive coupling. (b) Stability diagram for double dots with non-zero
inter-dot coupling. These figures were taken from the the review Elec-
tron Transport Through Double Quantum Dots by van der Wiel et al.[10] 7

1.4 Stability diagram measured by Gaudreau et al. for the first ever triple
dot systems(image shown bottom left). This image was taken from The
Stability Diagram of a Few Electron Artificial Triatom, by Gaudreau
et al.[26] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.5 Schematic of a triple dot chain system, in which only dots a and c cou-
ple to external lead contacts, while the middle dot a is tunnel coupled
to both dots a and c. . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

1.6 Both and image (left), and schematic (right) of the triple dot star
system created by Rogge et al. In this case three dots are in a triangular
shape, each dot is tunnel coupled to a lead. The middle dot connected
to the source lead, and the other two connected to drains allowing two
non-interacting paths. This image was taken from Two Path Transport
Measurements on a Triple Quantum Dot[28]. . . . . . . . . . . . . . . 12

1.7 Double quantum dot schematic in which the tunnel rate from the left
lead to dot a is ΓL , and from dot b to the right lead is ΓR , while the
two dots have tunnel coupling tab . . . . . . . . . . . . . . . . . . . . . 16

1.8 Schematic of the two dot energies, and two lead chemical potentials,
in the high bias voltage regime. . . . . . . . . . . . . . . . . . . . . . 18

2.1 Schematic of triple dot setup with inter-dot tunneling t and t0 , lead tun-
neling rates Γ, magnetic flux Φ, and lead chemical potentials µLa ,µLb
and µR as shown. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

iv
List of Figures v

2.2 (Left) Conductance versus ¯ = (b + a )/2 (where a = b ) and c with


Φ = 0, at a temperature of at 50mK. Interdot tunnel couplings are
set to |tab | = 67.8µeV, |tac | = |tcb | = 9.7µeV , dot charging energies
are ECa = 3.1meV, ECb = 2.8meV, ECc = 2.28meV , and the capacitive
couplings of the dots are Eab = .86meV, Eac = .28meV, Ebc = .49meV .
The lead-dot tunnel rate Γ = .1µeV and the temperature is 50mK
(4.3µeV ).(Right) Stability diagram from simulated QPC, where light
blue corresponds to the electron equally in dots a and b, dark blue the
dots are empty, and red the electron is in dot c. . . . . . . . . . . . . 31
2.3 a = b = t0 , with t0 = 67.8µeV  t = 9.7µeV , and φ = 0. Top Left:
Eigenstates energies, with Ground state (red), 1st excited state (green),
2nd excited state (blue), and empty state (black). Chemical potentials
(dashed black) for Vbias = 1µeV . Bottom Left: Eigenstate probabili-
ties, with color code as above. Top Right: Charge state Energies, with
dot a (red), dot b (green) dot c (blue), and empty state (black). Bot-
tom Right: Charge state probabilities. All other parameters are kept
fixed as previously stated. . . . . . . . . . . . . . . . . . . . . . . . . 32
2.4 (Top Left): Eigenstates energies, (Bot Left): Eigenstate probabilities,
(Bot Right): Charge state probabilities. Now with φ = π, and all
other parameters are kept fixed as previously stated. Color coding as
previously used. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.5 Differential conductance versus dot energy c and enclosed flux Φ, for
the case of zero relaxation and a = b = t0 and t  t0 , at a temperature
of 50 mK. All other parameters are fixed as before. . . . . . . . . . . 35
2.6 g · (kB T /Γ) versus electrostatic energy c for various Φ: (Top Left)
φ = 0, (Top Right) φ = 1.5π/2, (Bot Left) φ = 1.9π/2, (Bot Right)
φ = π. For the case of zero relaxation (ie Λ = 0) and a = b = t0
and t  t0 , at a temperature of 50 mK. All other parameters are fixed
as before. The red curve corresponds to current passing through the
ground state, while the green curve in the first excited state. . . . . . 36
2.7 (Left) Eigenstate energies versus dot energy c for Φ = 0, in the case
of zero relaxation and a = b = t0 and t = t0 , at a temperature of 50
mK. All other parameters are as before. Color coding of eigenstates is
as before. (Right) g · (kB T /Γ) versus dot energy c and enclosed flux
Φ, for the case of zero relaxation and a = b = t0 and t = t0 , at a
temperature of 50 mK. All other parameters are as before. . . . . . . 37
2.8 g · (kB T /Γ) versus dot energy c and enclosed flux Φ, for t0  t in the
case of strong bath coupling. All other parameters are as before. . . . 38
2.9 g · (kB T /Γ) versus dot energy c for various enclosed flux Φ, in the
strong bath coupling limit with a = b and all other parameters as
before, at a temperature of 50 mK. (Top Left) φ = 0, (Top Right)
φ = 1.5π/2, (Bot Left) φ = 1.9π/2, (Bot Right) φ = π. . . . . . . . . 40
2.10 g · (kB T /Γ) versus dot energy c , for various bath coupling strengths,
with enclosed flux Φ/Φ0 = 0.5 and all other parameters as before, at
50 mK. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
vi List of Figures

2.11 (Top Left) g · (kB T /Γ) versus Φ/Φ0 for various values of c , with no
bath coupling (Λ = 0), and all other parameters left as before. Black
is centered at the resonance peak, red to the left of the resonance
peak, and green to the right of the resonance peak. (Top Right) Fast
Fourier transforms of the AB oscillations shown in to the left. (Bot
Left)g · (kB T /Γ) versus Φ/Φ0 for various values of c , with strong bath
coupling, and all other parameters left as before. Black is centered at
the resonance peak, red to the left of the resonance peak, and green to
the right of the resonance peak. (Bot Right) Fast Fourier transforms
of the AB oscillations shown in to the left. . . . . . . . . . . . . . . . 43
2.12 Current versus bias voltage, for no relaxation and strong relaxation,
with all parameters as set earlier. . . . . . . . . . . . . . . . . . . . . 45
2.13 Current versus dot energy c and magnetic flux Φ/Φ0 for Vbias =
500µeV (Left) and Vbias = −500µeV (Right). . . . . . . . . . . . . . . 46
2.14 Current versus dot energy c for Vbias = 500µeV for various values of
Φ at 50 mK. (Top Left)φ = 0,(Top Right)φ = 0.1π/2,(Mid Left)φ =
π/2,(Mid Right)φ = 1.9π/2,(Bot)φ = π. . . . . . . . . . . . . . . . . . 48
2.15 Current versus dot energy c for Vbias = −500µeV for various values of
Φ at 50 mK. (Top Left)φ = 0,(Top Right)φ = 0.1π/2,(Mid Left)φ =
π/2,(Mid Right)φ = 1.9π/2,(Bot)φ = π. . . . . . . . . . . . . . . . . . 49
3.1 Schematic of a double dot system connected via two tunnel couplings
representing two spatially separated paths enclosing a magnetic flux Φ 69
Abstract

We use a quantum master equation approach to study the transport properties of a


triple quantum dot ring. Unlike double quantum dots and triple quantum dot chains,
this geometry gives two transport paths with a relative phase sensitive to magnetic
flux via the Aharonov-Bohm effect. This gives rise to a coherent population trapping
effect and what is known as a “dark state”. Unlike other master equation techniques
valid only in the high bias voltage limit, our treatment reproduces such results as
well as giving an analytic zero-bias conductance formula. As well as providing a more
robust signature of this “dark state” physics, our model further predicts a negative
differential resistance in connection with high bias rectification already predicted.

vii
Résumé

Nous utilisons une approche déquation quantique maı̂tresse pour étudier les propriétés
de transport des points quantiques triples en forme d’anneau. Contrairement aux
points quantiques doubles et triples en forme de chaı̂nes, cette géométrie offre deux
chemins pour le transport avec une phase quantique relative qui est sensible au flux
magnétique en raison de l’effet Aharonov-Bohm. Ceci méne à un effet de piégeage de
population cohérent et cela est connu sous le nom d’un “état sombre”. Contrairement
à d’autres techniques d’équation maı̂tresse qui sont seulement valides dans la limite
d’un potentiel électrique élevé, notre méthode reproduit les résultats de ces derniers
en plus de donner une expression analytique pour la conductance différentielle de
zéro potentiel électrique. En plus de donner une optique plus robuste de la physique
“détats sombres”, notre modèle prédit une résistance différentielle négative qui est
reliée au phénomène déjà prédit de rectification à potentiel élevé.

viii
Acknowledgments

Adam would like to acknowledge his supervisor Aashish Clerk for the countless hours
of guidance and insight, as well as Louis Gaudreau and Andy Sachradja for the useful
conversations and data. Adam would also like to thank all of his friends and especially
his family for all of the support and encouragement needed to carry on.

ix
x Acknowledgments

.
1
Introduction

As semiconductor fabrication has reached the nanoscale, quantum effects have become
increasingly significant. When a device confines electrons to a small enough region, a
so called quantum dot is formed, having a discrete atom-like energy spectrum. This
makes semiconductor quantum dots not only a promising medium for next generation
quantum computers, but also a novel means to study analogs of fundamental atomic
and molecular properties. This earned them the name “artificial atoms” as coined by
the title of a paper by Kaster[1]. Easily tunable to contain only a few electrons at a
time, quantum dots provide an ideal grounds to study coherent transport, as covered
extensively in the review by Hackenbroich et al. [2]. A series of experiments by
Yacoby, Schuster et al.[3, 4, 5] carefully study the phase coherent electron transport
through a single quantum dot. With double dot fabrication underway for several
years already[6, 7, 8, 9], many of their electronic transport properties are also now
very well understood, as outlined in the review by van der Wiel et al. [10]. Most
recently, triple dot systems have begun to be realized experimentally, showing novel
coherent transport properties.

After a brief overview of the physics and experimental characterization of single


and double dot systems, we will briefly review three recent triple dot experiments.
From these experiments, we will see how a triple dot can have a geometry funda-
mentally distinct from a double dot system, as well as the novel effects that result.
With this recent experimental progress in controlled triple dot fabrication, there have
been several recent theoretical efforts towards understanding coherent transport in
triple dot systems, all of which leave some questions unanswered. Before outlining

1
2 1 Introduction

a few interesting new triple dot models already proposed, we will first give a brief
overview of some basic transport properties of single and double dot systems. After
highlighting the success and shortcomings of a few recent triple dot theory papers,
we will then see the regimes to which they are limited, motivating the presentation
of our novel triple dot model.

1.1 Experimental Background

1.1.1 Single Quantum Dots: The Coulomb Blockade

A quantum dot (QD) is formed when a potential well confines electrons to a small
enough region, that a discrete atom-like energy spectrum is formed. The fabrication
of such a device generally begins with with the formation of a heterostructure of two
different semiconductor materials (GaAs/AlGaAs). The free electrons of the system
are thus confined to the interface of the GaAs and the AlGaAs, restricting their
motion to a two-dimensional plane creating a two-dimensional electron gas (2DEG).
The free electrons in the 2DEG can be further confined to a point by either etching
techniques, or more commonly via metal gate electrodes patterned on the surface of
the 2DEG stack. With clever patterning of these gate electrodes, these dot structures
can be coupled to electrical contacts via tunnel barriers allowing transport properties
of such devices to be studied[10]. The voltages of the gate electrodes can then be
tuned experimentally, controlling the electrostatic energies of the electrons trapped
in the dot’s potential well.

As show in figure 1.1, a single quantum dot coupled to two lead contacts can be
treated as three small chunks of metal, modeled by a capacitor and voltage node
network. The dot in the center is both capacitively coupled and tunnel coupled
to the leads such that a voltage difference between the two leads can pass a current,
while the dot’s gate voltage is only capacitively coupled to control the dot’s potential.
The systems electrostatic energy is derived in terms of the voltages and capacitive
couplings between each element, as shown in the review by van der Wiel et al.[10] to
1.1 Experimental Background 3

Vg
=
Cg

L d R
CL ,R L CR ,R R
Figure 1.1: Network of capacitors and voltage nodes for a single quantum dot with one gate voltage
and two lead contacts having additional tunnel couplings to pass a current.

give
[−(n − n0 )|e| + CL VL + Cg Vg + CR VR ]2
U (n) = , (1.1)
2C1
where C1 = CL + Cg + CR is a sum of all capacitances connected to the dot, n is the
number of electrons in the dot and n0 is the number of electrons on the dot with no
applied voltage. The above equation can be reduced to the more compact form,

E(n) = EC (n − N )2 , (1.2)

where N = Cg Vg /|e| is the now dimensionless gate voltage. EC = e2 /C1 is the


charging energy of the dot, or energy required to add another electron to the dot.
This equation tells us how many electrons will be in the dot for a specific value of
gate voltage, this being the n for which E(n) is minimized. Although classically we
do not think of the amount of charge on a capacitor a quantized, in the weak coupling
regime such that the rate at which electrons tunnel on and off the dot is much less
that the charging energy, we will have a well defined number of electrons on the dot,
quantizing the amount of charge. Up until now our description of a quantum dot has
actually been purely classical, only considering the electrostatic potential energy of
an electron.
In general even a free electron can have any additional amount of kinetic energy,
what makes a quantum dot quantum is that this confinement of a single electron
results in a discrete particle-in-a-box-like energy spectrum of possible excited states.
The electrostatic energy equation above gives the ground state energy for different
4 1 Introduction

electron numbers in the dot, not this full spectrum of each state. This intrinsic level
spacing of the dot’s excited states is inversely proportional to the dot’s diameter which
when reaching the nanometer scale makes single states easy to isolate. In the central
model of this thesis we will assume that this level spacing is actually large enough to
be completely ignored. In this case, although a quantum dot will only have a ground
state, by tuning its gate voltage its potential well is altered possibly resulting in a
change in the ground state configuration.

The amount that the gate voltage must change in order for the ground state to
change in number by one electron, is proportional to the charging energy of the dot.
This change in ground state number occurs whenever E(n) = E(n + 1), a degeneracy
which corresponds to gate voltages of N = (n+1)/2 in dimensionless units (as defined
after equation 1.2). By slightly offsetting the chemical potentials of the left and right
lead contacts a bias voltage is created and a current will attempt to pass through
the dot from one lead to the other. For a bias voltage and lead temperature both
much less than the charging energy of the dot, current will pass most easily at the
degeneracy point just mentioned as an electron can be added and then removed with
no cost in energy. Away from this degeneracy point, the leads will not be able to
provide the required charging energy to add an electron to the dot. This is known
as the Coulomb blockade regime as discussed by Averin and Grabert[11, 12]. In
this regime, measuring the current through the dot as a function of its gate voltage,
a series of sharp peaks known as Coulomb Blockade peaks are found[3, 4]. These
peaks are spaced apart by a voltage proportional to the charging energy of the dot,
occurring at the gate voltage values specified earlier. These peaks do of course have
some width, which is a function of the temperature of the the leads as well as the
strength of the lead-dot coupling. Although two charge states need not be completely
degenerate, far from the degeneracy points only higher order non-energy conserving
processes are possible (but vanishingly rare). We will see that in double and triple
dot systems current resonances like these can be much sharper than temperature.

Although the current through a single quantum dot only is only a function of the
1.1 Experimental Background 5

magnitude of the transmission through it, the transmitted wave function is in fact
a complex quantity with a possibly well defined phase relative to the incident wave.
From clever experiments by Yacoby et al.[3, 4, 5], this phase factor was in fact probed
by placing a single quantum dot in one arm of a two-path interferometer for electrons.
This is what as known as an Aharanov-Bohm (AB) ring, as the relative phase of
two transport paths is proportional to any magnetic flux enclosed within them[13].
Therefore, the current produced by the two interfering paths oscillates with respect
to the magnetic field strength. Tuning the dots gate voltage to a Coulomb peak to get
a current, one can then vary the magnetic field to see oscillations. What was found
was that the oscillations on either side of a Coulomb peak were always out of phase
by π. This abrupt phase flip is now understood in terms of an Onsagger symmetry
present in two terminal devices, as discussed extensively in the review on coherent
transport in mesoscopic devices by Hackenbroich et al.[2]. Later we will see how these
oscillations also become possible in triple dot ring systems, as well as consider the
effects of decoherence on these oscillations.

1.1.2 Double Quantum Dots: Stability Diagrams


Vg1 Vg2 =

C g1 C g2

L a b R

CL ,R L Cm ,R m CR ,R R

Figure 1.2: Network of capacitors and voltage nodes for a double quantum dot with one gate voltage
and one lead contact for each dot. The dots are tunnel coupled to each other as well as to one lead
each so that a current can be passed through the dots.

A double quantum dot system is created when some extra gate electrodes are
added to create a double potential well, with each dot coupled to one lead. This
results in two dots both capacitively and tunnel coupled to each other and one lead
each, while only coupled capacitively to two gate electrodes Vg1 and Vg2 as shown in
figure 1.2. Again, the electrostatic potential for the system modeled as a capacitor
6 1 Introduction

network was done by van der Wiel et al.[10], resulting in the now slightly more
cumbersome equation

1 1
U (na , nb ) = n2a ECa + n2b ECb + na nb ECm + f (Vg1 , Vg2 )
2 2
1
f (Vg1 , Vg2 ) = − [Cg1 Vg1 (na ECa + nb ECm ) + Cg2 Vg2 (na ECm + nb ECb )] (1.3)
|e|
1 1 1
+ 2 [ Cg21 Vg21 ECa + Cg22 Vg22 ECb + Cg2 Vg2 Cg2 Vg2 ECm ],
e 2 2
2
e2
where, ECi = Ci
(1 − CCamCb )−1 is the charging energy for the ith dot, and Ci is the sum
e2 Ca Cb
of all capacitances coupling to the ith dot. ECm = ( 2
Cm Cm
− 1)−1 is the energy of the
capacitive coupling of the two dots. Fortunately this can again be simplified greatly
to
X
E(na , nb ) = ECi (ni − Ni )2 + ECm (na − Na )(nb − Nb ). (1.4)
i=a,b

The dimensionless gate voltages of each dot are defined as before in equation 1.2, and
na and nb now represent the number of electrons in each dot, a and b. We can now
have states with different electron number or configuration, (na , nb ), the pair which
minimizes E(na , nb ) being the ground state configuration. We will now see how a
map of the ground state configuration for different values of the two gate voltages is
both measured and understood.
A charge stability diagram is a map specifying the ground state charge config-
uration of the system with respect to its two gate voltages as shown in figure 1.3
from the review by van der Wiel et al.[10]. Regions of parameter space with different
ground state configurations are separated by lines. With Cm = 0 we see that each
independent gate voltage can be tuned to add an electron to its respective dot, gen-
erating a set of horizontal and a set of vertical lines, one for each dot. The spacing of
these parallel lines is proportional to the charging energy of the particular dot they
represent and are thus called charging lines. Although the point of intersection of
two perpendicular charging lines is a quadruple point (QP) where four charge states
are degenerate, any capacitive coupling between the two dots lifts this degeneracy
splitting the the QP into two triple points (see dashed box in right of figure 1.3). In
1.1 Experimental Background 7

general, only triple points are available in double dot systems, and one must look to
systems of more dots for degeneracies involving more states.

When Cm 6= 0, changing one dots energy effects the other dot, resulting in nega-
tively sloped set of parallel charging lines for each dot (as seen in the right of figure
1.3). With strong inter-dot coupling it is therefor possible to add an electron to one
dot, by varying the gate voltage of the other dot. With non-zero inter-dot coupling,
one also notices positively sloped lines connecting the two triple points created by
the split QP (see dashed box in right of figure 1.3). These positively sloped lines
generally separate two regions of different charge configuration, but equal number.
At the triple point found at the left end of these lines for example, the ground state
is threefold degenerate between (na , nb ), (na + 1, nb ), (na , nb + 1). It is at these point
that sequential tunneling across the dots is maximized. Sequential tunneling is the
simplest form of essentially classical transport through a double dot system and will
be discussed in detail in section 1.2.1. This occurs when the classical electrostatic
energies of the three states just mentioned are equal, allowing the system to easily
switch from one to the other shuffling electrons across the system one by one. After
measuring the stability diagram, the system can then be tuned to any degeneracy
point of interest.

Figure 1.3: (a) Stability diagram for a double dot system with no inter-dot capacitive coupling. (b)
Stability diagram for double dots with non-zero inter-dot coupling. These figures were taken from
the the review Electron Transport Through Double Quantum Dots by van der Wiel et al.[10]
8 1 Introduction

Measuring the stability diagram is achieved by placing a quantum point contact


(QPC) near the double dot system[14, 15, 16, 17]. A QPC is a separate current
channel that is highly sensitive to the charge configuration of the nearby dot system.
When the number or configuration of charges in the quantum dot system changes, the
potential in the nearby QPC will change giving a sudden change in its conductance,
or a spike in the derivative of its conductance versus the gate voltage being changed.
It is ∂/∂Vgi of gQP C , that is actually plotted in a stability diagram, which gives lines
separating regions with different ground state charge configurations. Although the
classical electrostatic energy of equation 1.4 will tell us where these lines should be, it
cannot explain the variable thickness found in experimentally measured lines[26]. This
thickness is due to quantum tunneling between the two states being separated. This
tunneling allows superpositions of the two states to exist when their energy difference
is small enough. The charge transfer lines are thus much broader and generally less
well defined than the charging lines separating states with a different electron number.
Furthermore, it is found experimentally that at the point where two charging lines
of different dots cross, the lines develop a curvature, and an anti-crossing is formed
by the splitting of the quadruple point into two triple points. This curvature as
well as the width of these charge addition lines are quantum mechanical effects not
captured by the electrostatic Hamiltonian. A fully quantum mechanical Hubbard
Hamiltonian (see line one of equation 2.2 for example) combining the inter-dot tunnel
couplings with the electrostatic energies must then be used to fully reproduce an
experimentally measured stability diagram. In this way the dot charging energies,
gate voltage coupling constants, and inter-dot tunneling strengths can all be extracted
from experiment.

In this full quantum Hamiltonian for a double dot system, quantum tunneling
allows the system to exist in a superposition of electronic charge states. There have
in turn been many proposals for the use of semiconductor quantum dots as qubits[18,
19, 20, 21, 22, 23, 24] in next generation quantum computers. Although many exploit
the intrinsic quantum spin of the electrons (which we neglect in this entire thesis),
1.1 Experimental Background 9

charge qubits have also been proposed. This is a system where the two possible charge
states correspond to the 0 and the 1 of the bit, were now superposition states with long
lifetimes are desired. In our triple dot model, we will not only consider novel coherent
transport properties potentially useful for future quantum computation algorithms.

1.1.3 Triple Quantum Dots: Chains and Rings


A triple quantum dot can be formed in much the same way as a double dot, as
for instance when one dot in a double dot split in two to make the first ever triple
dot ring of three potential wells arranged in a triangle[25, 26](see device image and
schematic in figure 1.4). In this case two dots are coupled to one lead on the left,
while the third dot couples to the lead on the right. Another star-like triple dot ring
was recently intentionally created (see figure 1.6), where each dot has its own lead
so that current could be measured through two independent paths simultaneously.
A third sort of triple dot was also created using electrons already confined to one
dimension in a carbon nanotube. In this case a triple dot chain is formed in which
only one transport path is possible as depicted in figure 1.5. Although the triple dot
chain system is not fundamentally distinct from a double dot, in a triple dot ring
two possible transport paths enclose a magnetic flux Φ when a perpendicular field is
applied, giving rise to Aharonov-Bohm oscillations not possible in double dot systems.
The first ever triple dot system was recently created and characterized by Gaudreau
et al.[25, 26], when one dot in a double dot system was split in two forming a triple
dot ring. They were able to identify the system as a triple dot from the stability
diagram (see figure 1.4), even though it had been designed as a double dot. For a
double dot system where each dot is coupled to a gate voltage, the stability diagram
should have a set of parallel charging lines for each dot. Since one dot is mainly
coupled to one gate voltage, its charging lines should be nearly perpendicular to that
axis, while the other dots charging lines are perpendicular to the other gate voltage
axis (as previously discussed in section 1.1.2). What was found was a single set of
charging lines for the right dot c, perpendicular to the x-axis gate voltage (see vertical
lines in figure 1.4 labeled with a green c). Instead of a single set of horizontal charging
10 1 Introduction

Figure 1.4: Stability diagram measured by Gaudreau et al. for the first ever triple dot systems(image
shown bottom left). This image was taken from The Stability Diagram of a Few Electron Artificial
Triatom, by Gaudreau et al.[26]

lines for the left dot however, two sets of nearly parallel lines were found, as can been
seen labeled by alternating red and blue dots in figure 1.4. The two new lines are
hard to distinguish as they are nearly parallel, but in fact one set of lines is brighter
than the other as it couples to the QPC more than the other. These two lines do
however form an anti-crossing where they should intersect as is marked by the large
dashed oval in figure 1.4. The fact that the anti-crossing formed by dot a and dot b
lines is much more pronounced than those between dots a and c or b and c (marked
in smaller dashed circles in figure 1.4) is because the tunnel coupling between dots a
and b is much greater than between either dot and dot c. This limit will in fact turn
out to be very convenient in for our model, as will be explained in section 2.2.
Although this triple dot only has two gate voltages, in general with three dots
one should have three gate voltages, and the stability diagram would become a three
dimensional object. Viewing a planar projection of the stability diagram we do still
recover the same features as before with a set of parallel charge addition lines for each
dot, their anti-crossings measuring the respective inter-dot tunnel couplings. Even
though the triple dot in figure 1.4 only has two main gate voltages labeled as 1B and
5B, there are still additional gate voltages necessary to define the full dot potential
wells, labeled 2B, 3B, 4B, 3T . In fact, while the main two gate voltages are used
1.1 Experimental Background 11

to make the stability diagram, these additional gates can be tuned to modify it. In
this way, Gaudreau et al.[25] were able to bring two triple points (one from the large
dashed oval and one from the small dashed oval) together, forming quadruple point
between four states, such as (na , nb , nc ) ∈ {(0, 0, 0), (1, 0, 0), (0, 1, 0), (0, 0, 1)}.
As discussed is section 1.1.2, the QP just mentioned is the point at which sequential
tunneling across the dot is maximized. Sweeping one of the gate voltages to move
the system through this QP, resulted in generally ordinary resonances with a width
set by temperature. However, since the three dots could be in a ring enclosing a
magnetic flux, Gaudreau et al. then measured the current versus magnetic field for
fixed gate voltages on either side of this resonance. In order to find Aharonov-Bohm
oscillations[26], it was necessary to be close enough to the resonance to get enough
current flowing, but too close to the resonance peak and the oscillations disappeared.
Measuring these oscillations in current, at fixed gate voltages on either side of this
resonance, it was found that the oscillations had flipped by π; most of the time.
Although these oscillations are of a similar nature to those for the Coulomb peaks
in a single dot (as discussed earlier), we now have more than one dimension through
phase space with which to cross a resonance, not all of which lead to this flip in
the AB phase. We aim to shed some light on the geometric mechanism behind this
flipping in the presentation of our model in section 2.2, and our analytic calculations
in section 3.2.

ΓL t ab t bc ΓR
L a b c R

Figure 1.5: Schematic of a triple dot chain system, in which only dots a and c couple to external
lead contacts, while the middle dot a is tunnel coupled to both dots a and c.

More and more triple dot systems are now being built and characterized, such
as the carbon nanotube (CNT) triple dot chain of Grove-Rasmussen et al.[27] and
the triple dot ring created by Rogge et al.[28, 29]. As CNTs can have remarkably
high coherence lengths[27], they are a perfect medium with which to create coherently
coupled quantum dot systems. With electrons already confined to one dimension, top
12 1 Introduction

gates can be added which pinch the tube off into a chain of dots (as depicted in figure
1.5). In this system, they were able to clearly resolve that dot b was strongly coupled
to dot a on the left, but weakly coupled dot c on the right. Although the CNT prevents
the dots from forming a ring to enclose any magnetic flux, the high coherence time
makes them ideal for observing higher order effects due to the two strongly coupled
dots. In these experiments each dot had its own gate voltage, but those of the left
and right dots both cross-couple to the middle dot. This was discovered in stability
diagrams of both Vg1 vs Vg2 as well as Vg2 vs Vg3 . Here it was found that two of the
dots were more strongly coupled, resulting in a much bigger avoided crossing of the
two dots charging lines, and in turn a much bigger shift in the triple point formed
there.

Figure 1.6: Both and image (left), and schematic (right) of the triple dot star system created by
Rogge et al. In this case three dots are in a triangular shape, each dot is tunnel coupled to a lead.
The middle dot connected to the source lead, and the other two connected to drains allowing two
non-interacting paths. This image was taken from Two Path Transport Measurements on a Triple
Quantum Dot[28].

In Rogge’s triple dot star, each dot has its own lead, but the three dots are still not
all tunnel coupled to complete a closed ring[28]. Opposite to the work of Gaudreau
et al., rather than run the current from two dots to one dot, the current runs in two
independent paths from one dot to the other two (see figure 1.6). In this case the
two paths do not interfere, and no AB oscillations can be observed as the dots are
not tunnel coupled into a complete ring. However, Rogge et al. were able to map
the current for each path while blocking the second path, and then overlay the two
with the charge stability map. Although the two paths are independent, one was
1.2 Theoretical Overview 13

found to have two weakly coupled dots displaying atom-like electronic states, while
the other contained two strongly coupled dots displaying molecule-like states[29].
They were then able to measure, signatures of these states in higher order transport
measurements, rather than just stability diagrams. In this thesis, we will further
investigate the effects of strong dot coupling and molecular states in a geometry
creating two interfering transport paths. First of all, we shall present some basic
transport theory applied to single and double quantum dots, before reviewing a few
newer models of triple dots and their shortcomings.

1.2 Theoretical Overview

We begin this section with an overview of the basic electronic transport properties of
single and double dots, as have been well summarized in the review papers by Van
der Wiel et al.[10] and Hackenbroich et al.[2]. These well established foundations
are forcing theorists to investigate more detailed models in search of more exotic
phenomena[30, 31, 32] devices continue to become more sophisticated. For instance,
Hettler et al. predict a novel negative differential conductance (NDC) effect[38], due
to strong inter-dot tunneling in a double dot system. With the recent experimental
realization of triple dot systems, theory papers are beginning to analyze the new ring
geometry attained. A simple non-interacting tight binding approach by Delgado et
al.[33] gives some results on AB oscillation periodicity. More sophisticated master
equation approaches do treat interactions and make novel predictions of rectifying
“dark states”[40, 41], but are only valid in the high bias voltage regime. After re-
viewing the success’ and shortcomings of these three papers, we will present a novel
master equation technique, which treats electron-electron interactions while being
valid for both high and low bias voltages. Not only will this model fill in the nec-
essary gaps left by these other theories, but it will bring new understanding to the
geometric mechanism behind the “dark state” rectification previously overlooked.
14 1 Introduction

1.2.1 Transport Basics


Reviewing elementary transport, we can begin with the simplest one dimensional
problem of solving the Schrodinger equation for a square potential barrier. In this
problem, solved in every undergraduate quantum mechanics course one finds an in-
coming wave, a reflected wave, and a transmitted wave. Of course it is required that
T + R = 1, where T is the probability of the wave being transmitted through the
barrier and R is the probability of it being reflected back. We can now consider a
single quantum dot (or any system) coupled to two leads, as a coherent scattering
region. Once the transmission probability T is known for the system, the correspond-
ing current due to an applied voltage can easily be calculated with the Landauer
formula. This treatment rests on the assumptions that the current is composed of
a flow of independent non-interacting electrons, which scatter elastically through/off
some form of potential barrier. The resulting formula is
Z
e
I= dE[f (E − µL ) − f (E − µR )]T (E), (1.5)
h
where E is the energy of the incoming electron, while µL = eVbias and µR = 0 specify
the voltage applied across the system. The zero-bias conductance of the system is
defined as g = ∂I/∂Vbias |Vbias =0 , which is then given by
Z  
∂f (E)
g = dE − T (E) (1.6)
∂E
Of course the transmitted wave function will have a phase, but the current through
a system is a function of T = |t|2 , where the phase of t cannot be directly measured.
The problem now is to calculate the transmission probability of the system. This can
be done via scattering theory as done by Delgado et al.[33], or with Green’s functions.
For details of the green’s function approach see the book by Bruus and Flensberg[34].
For a single quantum dot with a single level coupled to a left and right lead via ΓL
and ΓR , the transmission probability is found to be
ΓL ΓR /4
T (E) = , (1.7)
(E − d )2 + (Γ/4)2
where d is the energy of the single dot level, and Γ = ΓL + ΓR is the sum of the
tunneling rates into and out of the dot via the left and right leads. Substituting
1.2 Theoretical Overview 15

this transmission probability back in to the Landauer equation for conductance, and
letting E = x · kB T , results in
Z
1 1 αΓ/kB T Γ̃/kB T
g= dx 2 2 , (1.8)
4 cosh (x/2) (1 + α)

x2 + Γ̃/kB T

(1+α)
where I have let ΓL = Γ, ΓR = αΓ, and Γ̃ = 4
Γ. This makes the last term a
lorentzian of width Γ̃, so in the limit that Γ/kB T → 0, we get a delta function which
kills the integral, giving
π 2α Γ
g= . (1.9)
8 (1 + α) kB T
We have thus seen that the conductance through a single resonant level produces a
Lorentzian with respect to the energy difference of the dot level and the lead electrons,
which in the limit of weak lead-dot coupling (as will be used in the central works of
this thesis) has a peak that scales as Γ/kB T . Thus if we choose α = 1 for equal lead
couplings, we find that Γ/kB T · g = π/8 ' 0.39.
Another way to calculate the current through a single quantum dot level is via the
use of a master equation. In this approach, we consider two probabilities: P0 is the
probability that the dot will be empty, and Pd is the probability that the dot will be
full. If we then assume a low bias voltage regime in which an electron can tunnel on
and off the dot via both the left and right leads, we can write the classical master
equation

Ṗ0 = Γoff Pd − Γon P0


(1.10)
Ṗd = Γon P0 − Γoff Pd ,

where Γon = ΓL0 + ΓR0 and Γoff = Γ0L + Γ0R . Transitions on and off the dot can
occur via the left or right lead, resulting in the rates Γi0 = Γi f (d − µi /e) and
Γ0i = Γi [1 − f (d − µi /e)], where i ∈ {L, R} and d is the electrostatic energy of
the electron on the dot (often rescaled to be zero). The Fermi functions are present
to either require and electron in the lead with the energy of the dot level, or a hole
in the lead with that energy. For a derivation of these rates via Fermi’s golden rule,
see section 2.2. Solving these equations for the steady state probability distribution,
16 1 Introduction

with the additional normalization constraint that P0 + Pd = 1 results in the solutions

1 Γon
P0 = , Pd = P0 . (1.11)
1 + ΓΓoff
on
Γoff

Letting ΓL = ΓR = Γ, we can then solve for the current between the left lead and the
dot (which must equal that trough the right), giving

I = e(ΓL0 P0 − Γ0L Pd )
eΓ fL + fR
= (fL − (1 − fL )( )) (1.12)
D 2 − fL − fR

= (fL − fR ),
D
where we have abbreviated fi ≡ f (d − µi /e), and the denominator D ≡ 1 + (fL +
fR )/(2 − fL − fR ). If we now set the bias voltage such that µL = Vbias and µR = 0,
then the conductance can be calculated via the formula g = ∂I/∂Vbias |Vbias =0 , giving
Γ −∂f (E)
g(E) = ( )
D ∂E (1.13)
Γ f (E)[1 − f (E)]
= .
kB T D
Again, we see that the conductance scales as Γ/kB T , but now we see that the reso-
nance width is proportional to kB T due to the derivative of the Fermi function. For
simplicity, we have used (−∂f (E)/∂E) = f (E)[1 − f (E)]/kB T , where the two fermi
functions now tell us that we need an electron available in the lead from f (E), as
well as an empty spot in the a lead from 1 − f (E). In this classical master equation
approach, all of the quantum mechanics is taken into account by the Γ’s which can
be calculated via Fermi’s golden rule, as is done in section 3.1.3.

ΓL tab ΓR
L a b R

Figure 1.7: Double quantum dot schematic in which the tunnel rate from the left lead to dot a is
ΓL , and from dot b to the right lead is ΓR , while the two dots have tunnel coupling tab .

In order to generalize this approach to a double dot system (depicted in figure


1.7), we must use a quantum master equation for the density matrix of the now three
1.2 Theoretical Overview 17

possible states, (na , nb ) ∈ {(0, 0), (1, 0), (0, 1)}. The density matrix and quantum
master equation will be described in section 3.1.2, or can be seen in detail in the
quantum optics text by Scully[35]. In this case, we will only three diagonal entries
giving the probability of these three possible states of the system. Although the full
master equation involves the off diagonal elements as well which will not be shown
here (see section 3.1.2), this can be reduced to an equivalent classical master equation
of probabilities and classical transition rates:

Ṗ0 = −(ΓL0 + ΓR0 )P0 + Γ0L Pa + Γ0R Pb

Ṗa = ΓL0 P0 − Γ0L Pa − Γ̃(Pa − Pb )


(1.14)
Ṗb = ΓR0 P0 − Γ0R Pb + Γ̃(Pa − Pb )
|tab |2 Γoff
Γ̃ = Γoff 2
,
~ ( 2 ) + ( ∆E
2
~
)2

where Γ̃ is the effective classical rate at which electrons tunnel back and forth between
dots a and b. For an explicit derivation of this classical master equation for a double
dot see section 3.3. In the limit that Γ0L = ΓR0 = 0 and lead tunneling can only occur
from left to right, these equations can be solved in the steady state limit to give the
high bias voltage sequential tunneling current formula as first derived by Nazarov et
al.[36] and Stoof[37],

ΓR |tab |2 /~2
I(∆E) = e Γ2R
. (1.15)
|tab |2 ΓR
(∆E/~)2 + 4
+ ~2
(2 + ΓL
)

where in this equation, ΓL is the rate at which electrons tunnel from the left lead onto
dot a, ΓR is the tunnel rate from dot b to the right lead, tab is the tunnel coupling
between the two dots (as shown in figure 1.7), and ∆E = E(1, 0) − E(0, 1). This
is sequential tunneling, the lowest order most classical form of transport in which
an electron hops across the dots, one dot at a time. In this equation the resonance
width is set by ΓR not by the temperature of the electrons in the leads. Via the
uncertainty principle, these decay rates into and out of the dot via the leads (or the
states lifetimes), are proportional to an energy uncertainty or intrinsic level linewidth
also denoted Γ, as shown in figure 1.8. This decay rate or tunnel rate is proportional
18 1 Introduction

to the quantum tunneling strength between the lead and dot. Thus in the weak lead-
dot coupling regime with high bias voltages, as two dot levels are swept past each
other, the current resonance will be proportional to the Γ’s of the two dot levels. For
Γ  kB T , this results in a resonance much sharper than temperature. This sharp
resonance would not be possible in low bias transport, as the temperature of the leads
would always smear things out. This will not be the case however, in triple dot rings.

Energy

kB T

ΓL
µL ΓR

kB T
µR
Figure 1.8: Schematic of the two dot energies, and two lead chemical potentials, in the high bias
voltage regime.

1.2.2 Beyond Sequential Tunneling


As we have just seen, sequential tunneling is the lowest order form of tunneling and
can be described by an effectively classical master equation (with some work). If
one thinks of the tunneling hops as Fermi’s golden rule transitions, higher order
transition rates exist which involve two or more hops through virtual states. The
difference between a regular state and a virtual state is that the electron is there for
such a short time that the uncertainty principle allows transitions to a virtual state
that do not conserve energy. These are known as co-tunneling events, as they involve
multiple simultaneous transitions which can even involve more than one electron.
In fact, a single electron appears more delocalized over the entire system during
these processes, rather than being highly localized on one dot after another. These
highly delocalized states are known as molecular states, caused by the strong tunnel
1.2 Theoretical Overview 19

coupling between the dots. These molecular states are in fact energy eigenstates of
the Hamiltonian including both the electrostatic charge state energies as well as the
inter-dot tunneling.

In a theory paper, Hettler et al.?? use a novel technique in which lead electrons
tunnel directly into the molecular states of a strongly coupled double dot system.
Rather than a sequential tunneling current resonance when the two dots electrostatic
energies are degenerate, high current is found when the molecular state energies are
on resonance with the leads[38]. The molecular states of the system correspond to
those in which the matrix representation of the system Hamiltonian is diagonal (as
derived for a double dot by van der Wiel et al.[10]). These two molecular eigenstate
energies can of course not be degenerate, and have a difference proportional to the
inter-dot tunneling. However, only one eigenstate is needed for transport as it can be
composed of a superposition of both dots. Including Coulomb interactions, Hettler’s
model allows for a system with two electrons, where both molecular eigenstates are
filled. With this model, Hettler et al. were able to investigate a novel mechanism for
an interesting effect known as negative differential conductance (NDC), in which the
current through the system actually decreases as a function of bias voltage.

The first and most trivial case of NDC found is due to the added capacitive coupling
of the lead contacts bias voltage to the charge state electrostatic energies. Thus, as
the bias voltage is increased, the eigenstate energy can actually be shifted out of the
bias voltage window causing the current to suddenly shut off. This effect is of course
also possible in the sequential tunneling regime, as the lead voltages actually couple
directly to the charge state electrostatic potentials, only modifying the molecular
states indirectly. The second and more interesting mechanism of NDC found is a direct
consequence of this co-tunneling via the molecular eigenstates. Since the molecular
eigenstates have an energy difference, it is not until a sufficiently high bias voltage
that the higher energy state will become available. In the case of strong inter-dot

tunneling tab > ΓL ΓR /2 it was found that at high biases, the second eigenstate
filling doubled the current like an ordinary extra transport channel. However, in
20 1 Introduction

the case of tab < ΓL ΓR /2 it was found that the current actually dropped as the
resonance was broadened[38]. Actually, Hettler didn’t provide any explanation for
this effect, as it was first discovered by Djuric using a more conventional master
equation technique[39]. We will present another NDC mechanism only possible in
triple dot rings, by using a similar technique of tunneling the lead electrons directly
into the systems molecular eigenstates. However, before moving on to our model of
a triple dot ring, let first consider some other recent models and their shortcomings.

1.2.3 Triple Dot Rings and Aharonov-Bohm Oscillations


Due to the recent success in triple dot fabrication, theory papers have begun to deal
with the AB oscillations found in triple dot rings. One such paper comes from Delgado
et al.[33], in which a non-interacting tight-binding approach is used. Discretizing the
leads as a one dimensional chain, a transfer matrix T can be written to transform
the wavefunction amplitude at one lead to that at the other lead. The transmission
through the system is then a function of the matrix elements of T , giving the simplest
expression if first diagonalized. Diagonalizing this matrix requires the molecular
eigenstates and energies of the triple dot system, which Delgado et al. have found
for the specific case in which each dot has equal energy, and all three have equal
coupling to one another. Although sequential tunneling would be maximal here, for
strongly coupled dots we expect the current peak to be shifted off this naive triple
point. Furthermore, we find solving for the eigenstates in the limit that tab  tac to
be simple and illustrative. Delgado et al. did find Aharanov-Bohm oscillations with
a periodicity of Φ0 , as well as anomalous sharp dips in the conductance at Φ0 /2, but
the cause of which remained a mystery. We will not only explain these dips, but also
see how they are smoothed out by decoherence effects.
Another effort to analyze the triple dot ring was made my Michaelis et al.[40]
who found an all electronic analogue of coherent population trapping in quantum
optics known as a “dark state”[42, 43]. Here it was found that this “dark state”
provides a novel mechanism for rectification in the high bias voltage regime. In this
work Michaelis et al. used a quantum master equation for the density matrix of the
1.2 Theoretical Overview 21

system, coupling the left lead to two dots, which can tunnel to a third dot which
couples to a right lead (as in figure 2.1, but with t0 = 0). Although this does not
allow magnetic flux to be enclosed by the dots giving AB oscillations, the effect found
is related. Looking at the steady state current limit, it was found that an anti-
bonding superposition state of the first two a,b dots is an eigenstate of the system
which cannot tunnel into the third dot due to deconstructive interference of the two
possible paths. The rectification behavior found was thus explained by the fact that
this anti-bonding eigenstate can still decay back into the left lead if the bias voltage is
reversed[40]. Although we too find this high bias rectification behavior, we will show
that the mechanism behind it is slightly more subtle than this, involving a second
molecular state to complete the rectification. Michaelis et al. also introduce dephasing
terms to the master equation representing charge noise in the system. Decoherence
of the “dark state” in turn prevents it from fully blocking the current. In this case it
is found that in the limit of no dephasing the current is blocked by the “dark state”,
but in the limit of strong dephasing the current is again blocked due the quantum
Zeno effect of the environment repeatedly measuring the occupancy of each dot. For
intermediate dephasing rates, the current that leaks through provides a measure of
the decoherence rate or coherence time of the system.

Although the “dark state” responsible for the current blocking discussed above
is found for a system in which all three dots have the same energy, and there is no
tunneling between the left two dots, in further works continued by Emary et al.[41]
the robustness of this “dark state” is explored for less arbitrary choices of parameters.
It is found that while the right dot c can have any energy, if the left two dots a and
b have energies a = ∆ and b = −∆ then a “dark state” can always be found by
offsetting the tunneling from a to c and b to c such that ∆ = ∆0 = tab (t2ac −t2bc )/2tac tbc .
The current is then compared to this energy detuning ∆ − ∆0 where it is found that
the current blocking is lifted as the “dark state” is lost, as expected. AB oscillations
were then studied where it was found that the “dark state” caused the current to
drop to zero with a periodicity of Φ0 when tab  tac , but with a periodicity of Φ0 /2
22 1 Introduction

when tab = tac . Although the mechanism behind this difference in periodicity was
not well understood, many references to systems with Φ0 periodicity[3, 4, 44] as well
as systems with Φ0 /2[45, 46] are given. Emary et al. use the same decoherence
treatment as Michaelis, in which there is no energy cost for dephasing. In our work
we will present a treatment of dephasing in the systems eigenstate basis, highlighting
the connection between charge dephasing and energy relaxation, as well as the effects
on “dark states” in strongly coupled dots.

1.2.4 What’s Next?


In this thesis we will present a theoretical model for coherent transport through
a triple dot ring, resulting in a novel negative differential resistance effect which
clarifies the mechanism behind the “dark state” rectification predicted by Michaelis
et al.[40]. We use a quantum master equation like Michaelis and Emary, but rather
than having the leads tunneling to the dots themselves, we couple the leads directly to
the system’s molecular eigenstates, as done by Hettler[38]. The lead tunnel rates then
follow from Fermi’s golden rule (FGR). This approach is justified in the limit that
inter-dot tunneling is much stronger than coupling to the leads, and should not be
neglected in the lead coupling as done by Emary (See discussion in section 3.1.2). This
treatment is valid for both zero bias conductance, as well as the high bias voltages in
which it reproduces the results of Michaelis and Emary. Furthermore, we couple our
system to a bosonic bath as done by Brandes et al.[32]. However, unlike the induced
dephasing terms derived by Michaelis and Emary by neglecting the strong inter-dot
tunneling, we find a novel way of including this inter-dot tunneling. The effect of this
bath coupling is thus found to induce transitions between molecular eigenstates of the
system. These transitions are clearly inelastic (unless molecular states are degenerate)
either taking energy from the bath or giving energy to the bath. This results in the
molecular eigenstates being non-stationary states, allowing the system to relax to the
ground state, or be thermally excited to higher energy molecular states. In the strong
bath coupling limit we find the system’s molecular states thermally distributed, in
equilibrium with the bath temperature.
1.2 Theoretical Overview 23

Furthermore, in the zero bias regime with inter-dot tunneling much stronger than
temperature, we find that transport involves only the ground and first excited molec-
ular eigenstates. This allows the molecular eigenstates to be solved for analytically,
leading to a rather simple analytic expression for the conductance in both the strong
and no bath coupling limits. Cases of intermediate relaxation and high bias voltage
can then be evaluated numerically without loss of any of the qualitative understand-
ing already gained. From these results we find that the ground state can be tuned
to a “dark state” via magnetic field, which results in the conductance being reduced
to zero. This is understood by an effective tunnel coupling of dots a and b to dot c,
which when reduced to zero passes current in neither direction. In fact this effective
coupling sets the width of the conductance resonance, which allows for features much
sharper than temperature, even at low bias voltages. The width of this resonance we
will see actually provides a novel measure of inter-dot tunneling unaffected by thermal
broadening. If the second excited state is tuned to be a “dark state”, we then find
that the ground state is no longer a “dark state”. In this case, it is not until a forward
bias is increased until the chemical potential is greater than the excited “dark state’s”
energy, that the current is suddenly blocked. Although the ground state could still
pass a current, in the Coulomb blockade regime once an electron gets stuck in the
“dark state” it also prevents a second electron from entering the system. Now if the
bias voltage is reversed, an electron first tunnels from the right lead to dot c and then
cannot tunnel into the excited a,b “dark state” but will still always be able to take the
available ground state. Thus the mechanism for rectification requires two molecular
eigenstates, one dark and one bright. Dephasing will of course lift the blockade of the
“dark state”, but investigating the robustness of the rectification and NDR effects we
find that they also provide a novel measure of the intrinsic electron-phonon coupling
strength.

The remaining chapters of this thesis will proceed as follows: In chapter two we
will give a brief overview of our model and techniques used to find solutions. We
will then present the results of this model, starting with zero-bias conductance in the
24 1 Introduction

limits of strong and no bath, where illustrative analytic solutions can be found. We
then consider the effects of large bias voltages, predicting a unique NDR effect, as
well as clarifying the rectification mechanism proposed by Michaelis et al. In chapter
three we will go through all the details and derivations behind our model and results.
Finally, chapter four will give the conclusions of this thesis.
2
Model and Results

µLa Γ a t
Γ µR
t! Φ c
µLb
Γ
b t

Figure 2.1: Schematic of triple dot setup with inter-dot tunneling t and t0 , lead tunneling rates Γ,
magnetic flux Φ, and lead chemical potentials µLa ,µLb and µR as shown.

2.1 Model Hamiltonian


We model a triangular triple dot geometry as shown in figure 2.1, where electrostatic
gate voltages are used to define the confining potentials of the three dots arranged
in a ring. The electrostatic energy of this triple dot system is easily generalized from
that of the double dot system given in equation 1.4, resulting in
X X
E(na , nb , nc ) = ECi (ni − Ni )2 + ECij (ni − Ni )(nj − Nj ). (2.1)
i=a,b,c i6=j=a,b,c

As we saw earlier, ECi is the charging energy for the ith dot, ECij is the energy of
the capacitive coupling of the ith and j th dots, where now i, j ∈ {a, b, c}. Inspired by
the experiments of Gaudreau et al.[25, 26], we use two experimentally tunable gate
voltages Vg1 and Vg2 cross-coupled to the three dimensionless gate voltages N1 , N2 , N3 .
These two experimentally tunable gate voltages, along with the experimental coupling

25
26 2 Model and Results

parameters, could then be used to reproduce the stability diagrams of Gaudreau et


al., allowing us to tune our system to a quadruple point as studied experimentally. In
fact, we use charging energies, capacitive couplings, and tunnel coupling parameters
from this experiment, as specified in figure 2.2.
Working in the Coulomb blockade regime, as the charging energy of the the dots
is the largest energy scale in the system, allows us to truncate our system’s Hilbert
space to states with only zero electrons or one electron in either dot a,b, or c. Thus we
consider only the four possible charge states (0, 0, 0), (1, 0, 0), (0, 1, 0), (0, 0, 1), denoted
by the kets |0i, |ai, |bi, |ci respectively. Due to the nanometer size of the dots, their
intrinsic level spacing is large enough that each dot is assumed to have only a single
energy level. With the empty state’s energy rescaled to be zero, the addition energy of
dot a for instance is defined as a ≡ (1, 0, 0) − (0, 0, 0). Although the dimensionless
gate voltages Ni cross couple to the two gate voltages of the experiment, they can
easily be remapped to control individual state energies. Henceforth, we shall thus
refer to sweeping electrostatic energies rather than gate voltages.
To add quantum tunneling to our model we write a Hubbard-like system Hamil-
tonian Hsys , as well as a component describing the two leads Hres , and their coupling
the dot states HL and HR . In second quantized notation this takes the form:

H = Hsys + HL + HR + Hres ,
X X
Hsys = i fˆi† fˆi + tij fˆi† fˆj ,
i={a,b,c} i6=j
X
HL = [(wk,a fˆa† + wk,b fˆb† )ĉk + H.c.],
(2.2)
k
X
HR = [wk,a fˆc† ĉk + H.c.],
k
X X
Hres = Lk ĉ†k ĉk + R †
k ĉk ĉk ,
k k

where the charge state creation operators, defined as fˆi ≡ |0ihi| for i ∈ {a, b, c},
are not canonical fermionic operators, as only one dot can be occupied at a time.
Therefore, despite its appearance, this Hamiltonian does in fact treat electron-electron
interactions in the strongly interacting limit that only one electron is allowed in the
2.1 Model Hamiltonian 27

system at a time. The ĉk operators are however, canonical fermionic operators for
the fermi sea of lead reservoir electron states. For simplicity, we take each dots lead
couplings to be equal, wk,a = wk,b = wk,c ≡ w, and define the first order lead-dot
tunnel rate Γ ≡ 2πg0 |w|2 (which is derived in section 3.1.3), where g0 is the average
density of states in the leads near the chemical potential. The inter-dot tunneling
parameters are defined as tab = |t0 | and tac = tcb = |t|eiφ/2 , where φ = 2πΦ/Φ0 is the
relative phase difference of the two paths, due to the magnetic flux enclosed. Here
we take |tac | = |tbc | only for convenience, and not requirement of our model. Hsys
describes the coherent inter-dot tunneling and Hres the independent lead reservoir
energies, while HL,R couple the appropriate dots to the left and right lead reservoirs
respectively. As the addition energies and tunnel couplings come from experiment,
the remaining parameter Γ can be chosen such that the peak current has an amplitude
matching experiment.
As discussed earlier, we will use a density matrix and quantum master equation
as has been done by several others[40, 41, 47, 48]. Since the Hamiltonian in equation
2.2 describes both the four system states as well as a continuum of lead states, the
resulting density matrix can become very large. Since we only care about the state
of the system, in section 3.1.2 we derive a quantum master equation for the reducesd
density operator of the system states via a partial trace of the general time evolution
equation i~ρ̇S = TrR [H, ρ], as has been done in similar systems by many others[47, 48].
As we find that the lead coupling is also much less than the temperature of the leads,
we can ignore any intrinsic level broadening of the system states. The master equation
can then be derived by perturbatively coupling the leads to the system states with
an offset chemical potential in the leads to give a bias voltage. This is standard in
the treatment of open quantum systems[40, 41] and derived in section 3.1.2 of this
thesis. Now, considering the regime in which inter-dot tunneling is much stronger
than lead-dot tunneling, it is natural to work in the molecular eigenstate basis of the
system. We thus diagonalize Hsys giving
X
H̃sys = Ei f˜i† f˜i , (2.3)
i=1,2,3
28 2 Model and Results

where the f˜i operators are for the system’s molecular eigenstates, defined as |Ei i =
ai |ai + bi |bi + ci |ci, with E1 < E2 < E3 . This diagonalization can be done analytically
in the illustrative limit that t0  t as is derived in detail in section 3.2, but nu-
merical diagonalization can also easily be used for any other parameter values. This
is where our approach differs from others. By perturbatively coupling the leads to
the molecular eigenstates via Fermi’s golden rule (FGR), rather than to the charge
states. In fact this perturbative coupling to the molecular eigenstates turns out be
be higher order than simply coupling to the charge states. This approach is distinct
from the standard approach of coupling the leads to the localized charge states, in
that transport now occurs when the molecular eigenstate energies E1,2,3 , rather than
the localized charge state energies a,b,c , are aligned within kB Tleads of the chemical
potential of the leads. This is a well justified approach for behavior beyond sequential
tunneling as has already been confirmed by the stability diagrams of strongly coupled
double-dot systems[38].
To make our model more realistic we treat inelastic effects due to electron-phonon
interactions and electromagnetic noise resulting in fluctuating dot potentials. This is
modeled by coupling each dot to a bosonic bath of oscillators, introducing two new
terms to the Hamiltonian, similar to those used by Brandes et al.[32]:
X
He−p = (λj,α n̂α )(âj,α + â†j,α ),
j,δ
X (2.4)
Hp = ωj,α â†j,α âj,α ,
j,α

where âj,α are canonical bosonic operators, j sums the bosonic modes, and α ∈
{a, b, c} gives each dot its own independent bath. The bath coupling factors λj,α = λ
are taken to be equal for simplicity, and the bath mode energies ωj,α are taken to have a
linear density of states, for simple Ohmic dissipation. Interpreting the sum of bosonic
operators as coordinates of the bath, j (âj + â†j ) = X̂, we see that the bath couples
P

to the charge as would a fluctuating electromagnetic potential. Treating the bath-


dot coupling similarly to the lead-dot coupling results in additional master equation
terms as derived in section 3.1.2. Emary et al.[41] give a standard treatment in the
2.1 Model Hamiltonian 29

charge basis ignoring inter-dot tunneling, which results in ordinary charge dephasing
terms. But with tunneling much greater than bath coupling, we work in the molecular
eigenstate basis, where these fluctuations are found to induce incoherent transitions
between molecular states as is derived in section 3.1.4. Taking the bosonic bath to be
in thermal equilibrium with a temperature Tbath = Tleads , these incoherent transitions
add energy to or remove energy from the system, bringing it into thermal equilibrium
with the bath. In the strong coupling limit that these transition rates are much
greater than the lead-dot tunnel rates (but still less than inter-dot tunneling), the
system’s molecular states are found to be in a thermal distribution with a temperature
Tbath . The dephasing treatment by Emary et al.[41] in the charge state basis includes
no energy exchange between the system and bath. Inducing molecular eigenstate
transitions with no energy cost in the high bias voltage regime where all eigenstates
are relevant is therefore only valid in the high bath temperature limit where energy
of any amount is always available.

After adding additional master equation terms to model the incoherent eigenstate
transitions (see section 3.1.2 for complete master equation), we now have 16 poten-
tially coupled differential equations to solve: one for each element of the density
matrix. Taking the steady state limit and projecting this master equation onto the
eigenstate basis results in only four coupled algebraic equations necessary to charac-
terize the conductance. In the zero bias regime these can be solved analytically in
the limits of both strong and no bath coupling. Intermediate bath coupling strengths
can then be investigated numerically, illustrating the robustness of the “dark state”
against decoherence. We will also be able to investigate the role of the “dark state”
in the Aharonov-Bohm oscillations. Finally, we consider the effects of finite bias volt-
age and present the resulting negative differential resistance and rectification effects,
explained in terms of this “dark state” physics.
30 2 Model and Results

2.2 Zero-Bias Conductance

Considering only the four possible charge states |0i, |ai, |bi, |ci (since Γ  tij 
EC ), we study transport near a quadruple point (QP) at which all four states have
equal electrostatic energies. Naively, we expect the conductance peak here from
sequential tunneling, as all energy levels are aligned for the electron to hop along
from one dot to the other. However, with leads tunneling directly into the molecular
eigenstates due to strong inter-dot tunneling, it is now a molecular eigenstate energies
which must equal zero for transport to occur. This corresponds to the conductance
resonance (left of figure 2.2) being shifted off the electrostatic charge state QP, which
has coordinates (0,0) marked with an x in the right side of figure 2.2. This plot of
IQP C = 0 · P0 + 0.15Pa + 0.25Pb + 0.6Pc , simulates the stability diagram, where dark
blue corresponds to the ground state being |0i, light blue corresponds to |ai or |bi, and
red to |ci. Here we see that the current peak when a = b = 67.8µeV = t0 , showing
how strong inter-dot tunneling shifts the resonance away from the electrostatic QP.
In addition to being energetically aligned with the lead chemical potentials, transport
also requires a molecular eigenstates to be delocalized over dots a and b as well as
dot c to connect both leads and pass a current. In the light blue region the charge
is localized evenly over dots a and b, and in the red only dot c. In the broad yellow
region however, these states merge and the electron is delocalized over the entire
system.

From figure 2.2 we see that the resonance peak occurs around the point roughly
where a = b = t0 . In figure 2.3 we compare the energies and steady state probabilities
of both the four molecular eigenstates and four electrostatic charge states of the
system at a horizontal gate voltage slice through the resonance shown in figure 2.2.
From the charge state energies, we see that both dots a and b have equal and constant
energies, while the energy of dot c is swept linearly though zero. Then looking at the
eigenstate probabilities, we see that each eigenstate is most often composed of either
dot c when its line is straight but sloped, and composed equally of dots a and b when
horizontal. The two horizontal eigenstate energy lines correspond to the two possible
2.2 Zero-Bias Conductance 31

Conductance · (kB T /Γ)

IQP C (AU )
!¯ (meV )
!¯ (meV )

(e2 /h)
!c (meV ) !c (meV )

Figure 2.2: (Left) Conductance versus ¯ = (b + a )/2 (where a = b ) and c with Φ = 0, at a tem-
perature of at 50mK. Interdot tunnel couplings are set to |tab | = 67.8µeV, |tac | = |tcb | = 9.7µeV , dot
charging energies are ECa = 3.1meV, ECb = 2.8meV, ECc = 2.28meV , and the capacitive couplings
of the dots are Eab = .86meV, Eac = .28meV, Ebc = .49meV . The lead-dot tunnel rate Γ = .1µeV
and the temperature is 50mK (4.3µeV ).(Right) Stability diagram from simulated QPC, where light
blue corresponds to the electron equally in dots a and b, dark blue the dots are empty, and red the
electron is in dot c.

eigenstates formed by the strongly coupled a and b dots, which will be denoted ψ+
and ψ− and derived subsequently. We can also see here that an eigenstate made of
dot c will form an anti-crossing when its energy becomes degenerate with either ψ+
or ψ− .

To better understand the delocalization of these molecular eigenstates, we can


decompose the system Hamiltonian (line 2 in equation 2.2) into left and right parts,
by writing Hsys = Hab + Hc , where Hab = a n̂a + b n̂b + tab fˆa† fˆb + tba fˆb† fˆa and Hc =
Hsys − Hab . Taking tunneling between dots a and b much greater than between a
and c, or b and c, as found experimentally by Gaudreau et al.[26], these molecular
eigenstates can easily be found by first diagonalizing Hab (as described in detail in
chapter 3.2). This results in bonding and anti-bonding molecular states of dots a and
32 2 Model and Results

0.2

0.15 0.15

0.1 0.1
Energy (meV)

Energy (meV)
0.05 0.05

0 0

−0.05 −0.05

−0.1 −0.1

−0.15 −0.15

−0.2 −0.2
−0.1 −0.05 0 0.05 0.1 0.15 −0.1 −0.05 0 0.05 0.1 0.15
! c (meV) ! c (meV)

1 1

0.8 0.8
Probability

Probability
0.6 0.6

0.4 0.4

0.2 0.2

0 0

−0.1 −0.05 0 0.05 0.1 0.15 −0.1 −0.05 0 0.05 0.1 0.15
! c (meV) ! c (meV)

Figure 2.3: a = b = t0 , with t0 = 67.8µeV  t = 9.7µeV , and φ = 0. Top Left: Eigenstates


energies, with Ground state (red), 1st excited state (green), 2nd excited state (blue), and empty
state (black). Chemical potentials (dashed black) for Vbias = 1µeV . Bottom Left: Eigenstate
probabilities, with color code as above. Top Right: Charge state Energies, with dot a (red), dot b
(green) dot c (blue), and empty state (black). Bottom Right: Charge state probabilities. All other
parameters are kept fixed as previously stated.

b denoted |ψ+ i and |ψ− i respectively:

θ θ
|ψ+ i = cos |ai + sin |bi
2 2
θ θ
|ψ− i = − sin |ai + cos |bi
2 2 (2.5)
1 1p
± = (a + b ) ∓ (a − b )2 + 4|t0 |2
2 2
2|t0 |
tan θ ≡ .
b − a
In the zero bias regime of differential conductance, with − − + = 2|t0 |  |t|, kB T ,
we find the conductance peak around + = c = 0, and can neglect the higher energy
|ψ− i state. Although we will not be able to neglect the higher energy anti-bonding
2.2 Zero-Bias Conductance 33

state in the high bias regime, these approximate analytic expressions allow for simple
illustrative analytic equations for the systems conductance. The coupling between
|ψ+ i and |ci is then given by the following matrix element,

|X|2 = |hψ+ |Hc |ci|2


(2.6)
= |t|2 (1 + sin θ cos 2πΦ/Φ0 ),

where Φ is the enclosed magnetic flux, and Φ0 is the fundamental magnetic flux
quantum. We can then diagonalize the 2 × 2 Hamiltonian resulting for these two
states (see equation 3.38), to find analytic expressions for the ground and first excited
states and their resulting energy anti-crossing. The size of the anti-crossing between
the ground and first excited state in figure 2.3 is proportional to |X|, across which
the two eigenstates delocalize and switch composition. The sensitivity to magnetic
flux is due to the dots’ ring-like geometry which creates two interfering paths that
due to destructive interference can cause the effective coupling |X| to go to zero.

As shown in figure 2.4, for |X| = 0 at Φ/Φ0 = 1/2, an anti-crossing of diameter


zero is formed between the two states. An effective coupling of zero results in a “dark
state”, in which the electron is trapped in the |ψ+ i state for forward bias voltages, or
in |ci for reverse biases. This is coherent population trapping, as |ψ+ i is a coherent
superposition state. Figure 2.4 also shows the charge state and molecular eigenstate
probabilities, from which we see that when |X| = 0 then region over c in which the
the probability is delocalized over all of the dots is significantly reduced. In the next
sections we will present analytic solutions for the conductance through such molecular
eigenstates as a function of magnetic flux for both the cases of no bath coupling and
strong bath coupling.

2.2.1 No Bath Coupling Limit

Solving the master equation in this limit of no bath coupling (as shown explicitly in
section 3.1.5), we find the analytic formula for the conductance, g ≡ ∂I/∂Vbias |Vbias =
34 2 Model and Results

0.2

0.15

0.1

Energy (meV)
0.05

−0.05

−0.1

−0.15

−0.2
−0.1 −0.05 0 0.05 0.1 0.15
! c (meV)

1 1

0.8 0.8
Probability

Probability
0.6 0.6

0.4 0.4

0.2 0.2

0 0

−0.1 −0.05 0 0.05 0.1 0.15 −0.1 −0.05 0 0.05 0.1 0.15
! c (meV) ! c (meV)

Figure 2.4: (Top Left): Eigenstates energies, (Bot Left): Eigenstate probabilities, (Bot Right):
Charge state probabilities. Now with φ = π, and all other parameters are kept fixed as previously
stated. Color coding as previously used.

0 with I ≡ − Γ0i ρii ), to reduce to:


P
i=1,2 (Γi0 ρ00

−∂f (E) 1 − f (E2 )


  
T1L T1R
g = 2πΓ |E1
T1L + T1R ∂E 1 − f (E1 )f (E2 )
(2.7)
−∂f (E) 1 − f (E1 )
  
T2L T2R
+ 2πΓ |E2 ,
T2L + T2R ∂E 1 − f (E1 )f (E2 )
where f (E) is a fermi function with temperature Tleads , with E1 and E2 being the
ground and first excited state energies (see equation 3.41). The Fermi function deriva-
tive terms are the usual terms for resonant tunneling through single level, ensuring
the energetic availability of the state, giving a resonance width proportional to kB T
as shown in section 1.2.1. The second Fermi function terms are new blocking fac-
tors which say that transport through one eigenstate prevents transport through the
other eigenstate. This makes sense as both eigenstates are composed of the same
2.2 Zero-Bias Conductance 35

Conductance · (kB T /Γ)


Φ/Φ0

(e2 /h)
!c (meV )
Figure 2.5: Differential conductance versus dot energy c and enclosed flux Φ, for the case of zero
relaxation and a = b = t0 and t  t0 , at a temperature of 50 mK. All other parameters are fixed
as before.

physical dots, and only one electron is allowed in the system at a time. TiL and TiR
are matrix elements which specify how well the ith eigenstate couples to the left and
right leads respectively (see equation 3.42 for expressions and derivation).Since the
composition of these molecular eigenstates is a function of the dot energies and the
magnetic flux, the matrix elements are not just constant pre-factors anymore, but
satisfy the following expression (derived in section 3.2):
|X|
TiL TiR = |X| , (2.8)
4|X|2 + (∆E)2
for i ∈ {1, 2}, where ∆E = (+ − c ). We now clearly see that the resonance is mod-
ulated by a Lorentzian of width |X|, the effective coupling of ψ+ to dot c. This term
is to ensure that a molecular state is sufficiently delocalized as well as energetically
available. From the form of |X| in equation 2.6, we see that when θ = π/2 (see
equation 2.5) and when φ = π, we get perfect destructive interference between the
two paths, resulting in |X| = 0. Thanks to the Coulomb blockade, this makes |ψ+ i
a “dark state”, trapping the electron in dots a and b. Although an electron in |ψ+ i
can still tunnel back into the left lead if the bias voltage was reversed, |X| would
still be zero and the electron would then get stuck in |ci. In the zero-bias limit with
36 2 Model and Results

t0  kB T , “Dark state” current blocking is symmetric with respect to the sign of the
bias voltage.

0.14
0.25
0.12
(e 2 /h)

(e 2 /h)
0.1 0.2

0.08
g · (k B T /Γ)

g · (k B T /Γ)
0.15

0.06
0.1
0.04

0.05
0.02

0 0
−0.1 −0.05 0 0.05 0.1 −0.1 −0.05 0 0.05 0.1
! c (meV) ! c (meV)
−3
x 10
10

0.25
8
(e 2 /h)

(e 2 /h)
0.2
6
g · (k B T /Γ)

g · (k B T /Γ)

0.15

4
0.1

2
0.05

0 0
−0.1 −0.05 0 0.05 0.1 −0.1 −0.05 0 0.05 0.1
! c (meV) ! c (meV)

Figure 2.6: g · (kB T /Γ) versus electrostatic energy c for various Φ: (Top Left) φ = 0, (Top Right)
φ = 1.5π/2, (Bot Left) φ = 1.9π/2, (Bot Right) φ = π. For the case of zero relaxation (ie Λ = 0)
and a = b = t0 and t  t0 , at a temperature of 50 mK. All other parameters are fixed as before.
The red curve corresponds to current passing through the ground state, while the green curve in the
first excited state.

In figure 2.5, the zero-bias conductance is plotted versus the enclosed flux Φ/Φ0
and the charge state energy c , showing that the resonance width goes to zero when
Φ/Φ0 = .5 and |X| = 0. The resonance has maximal width at Φ/Φ0 = 0, 1, where
|X| is maximal. However, at these points we can also notice that the conductance
resonance shifts to the right. This is because as the effective coupling |X| grows,
the energy of the ground state in the vicinity of the anti-crossing is lowered. This
causes the majority of the current to pass through the ground state for large values of
|X|. On the other hand, this also causes the conductance maximum to be for smaller
|X|, as the ground and first excited state’s energies are closer together and can both
2.2 Zero-Bias Conductance 37

contribute. Here the resonance is narrower and more centered around c = 0. These
effects will become clearer by looking at horizontal line cuts of figure 2.5
In figure 2.6 we see four line cuts for various magnetic flux, with the contributions
of each molecular eigenstate in a different color. In the top two plots we see the
first excited state begin to contribute as |X| decreases. In the bottom two plots we
notice the two states contribute more equally and sharply near X = 0. Now that at
this point exactly, the current peak is actually three orders of magnitude smaller and
essentially zero. We also notice that when the current is passed mainly through the
ground state, the resonance curve (marked in red) is asymmetric. This is due to the
ground state composition changing, the energy of which is constant when the state is
composed of dots a and b, while changing linearly when composed of dot c. A single
eigenstate will therefor have an asymmetric resonance due to its reconfiguration as it
delocalizes across the resonance. It is required that kB T > |X| the excited state to
contribute, but at this point the asymmetric resonance is lost due to deconstructive
interference responsible for reducing |X|.

0.2

Conductance · (kB T /Γ)


0.15

0.1
Energy (meV)

0.05
Φ/Φ0

−0.05

(e2 /h)
−0.1

−0.15

−0.2
−0.1 −0.05 0 0.05 0.1 0.15
! c (meV) !c (meV )

Figure 2.7: (Left) Eigenstate energies versus dot energy c for Φ = 0, in the case of zero relaxation
and a = b = t0 and t = t0 , at a temperature of 50 mK. All other parameters are as before. Color
coding of eigenstates is as before. (Right) g · (kB T /Γ) versus dot energy c and enclosed flux Φ,
for the case of zero relaxation and a = b = t0 and t = t0 , at a temperature of 50 mK. All other
parameters are as before.

In the case that t = t0 , we cannot necessarily ignore the third molecular eigenstate,
unless t  kB T . The energy difference of the a, b bonding and anti-bonding states,
38 2 Model and Results

+ −− , will now be small enough that both |ψ+ i and |ψ− i can couple to |ci simultane-
ously. This can be seen in the right of figure 2.7, where there are no longer clearly two
distinct anti-crossings. Even if all three of the systems molecular eigenstates don’t
contribute to the conductance, analytic expressions will now not be easily found. The
eigenstates and their energies can however be calculated numerically, as was done in
the left of figure 2.7. The conductance in this case can be calculated numerically,
and is as is done for the conductance in figure 2.7. There does still remain some
asymmetry towards the right for the same reason as before, but to a lesser extent as
t0 is now smaller. A sharp peak and pinch in the resonance at Φ/Φ0 = 0.5 is still
found. However, the current is no longer blocked completely, but without analytic
expressions for the molecular eigenstates this cannot be well understood.

2.2.2 Strong Bath Coupling Limit

Conductance · (kB T /Γ)


Φ/Φ0

(e2 /h)

!c (meV )
Figure 2.8: g · (kB T /Γ) versus dot energy c and enclosed flux Φ, for t0  t in the case of strong
bath coupling. All other parameters are as before.

In the limit that the incoherent eigenstate transition rates (equation 3.27) induced
by the bath are much greater than the lead-dot tunneling rates the master equation
is also analytically solvable (see section 3.1.5). In this case we find our molecular
eigenstate probabilities to have a thermal distribution in equilibrium with the bath
2.2 Zero-Bias Conductance 39

temperature (which we take equal to the lead temperature for simplicity). In this limit
we are able to find an analytic conductance formula independent of the particular
relaxation rates that reduces nicely to:
2πβΓ h
g= T1L T1R · f (E1 )(1 − f (E1 ))
D
+ T2L T2R · f (E2 )(1 − f (E2 ))eβ(E1 −E2 )
(2.9)
+ T2L T1R · f (E2 )(1 − f (E1 ))
i
+ T1L T2R · f (E1 )(1 − f (E2 ))eβ(E1 −E2 ) ,

where the normalization factor D ≡ T1 (1+f (E2 )eβ(E1 −E2 ) )+T2 (f (E2 )+eβ(E1 −E2 ) ),with
β = 1/(kB Tbath ) and Ti ≡ TiL + TiR . Here we see the same two processes as before
in the first two lines of equation 2.7, where and electron enters one state and leaves
via the same state. In the last two lines however, we find an additional two processes
in which the molecular eigenstates are mixed, entering via one state and leaving via
the other. The matrix elements in the third term tell us that an electron enters
into the excited state, and then leaves via the ground state by emitting energy to
into the bath. In the fourth term we see that the electron is thermally raised from
the ground to excited state before leaving the system, and thus requires and extra
Boltzmann factor to represent this energy cost. We also see that rather than blocking
factors on the first two processes, an extra Boltzmann factor is present instead, so
that transport through the excited state is proportionally less likely the greater its
energy. Simplifying equation 2.9 we find
|X|2

(1 + sin θ) h
g = πβ (1 − f (E1 ))f (E1 )
2D (4|X|2 + (∆E)2 )
i
+ (1 − f (E2 ))f (E2 )eβ(E1 −E2 ) (2.10)
(∆E)2
  
βE1
+ 1+ 2f (E1 )f (E2 )e .
4|X|2 + (∆E)2
The first two lines in equation 2.9 can still contribute an arbitrarily sharp resonance
of width proportional to |X| as with no bath coupling. But, now these can be over-
whelmed by the two new processes represented in the last line of equation 2.10. As
seen in figure 2.8, the resonance width no longer goes to zero at Φ/Φ0 = 0.5. Al-
40 2 Model and Results

though X still goes to zero and the first term in equation 2.10 contributes no current,
the second term due to the charge relaxation induced by the bath, gives a resonance
of width proportional to kB T due equally to both molecular states.

0.14

0.2
0.12
(e 2 /h)

(e 2 /h)
0.1
0.15
0.08
g · (k B T /Γ)

g · (k B T /Γ)
0.06 0.1

0.04
0.05
0.02

−0.1 −0.05 0 0.05 0.1 −0.1 −0.05 0 0.05 0.1


! c (meV) ! c (meV)

0.4
0.5
0.35 0.45
(e 2 /h)

(e 2 /h)
0.3 0.4
0.35
0.25
0.3
g · (k B T /Γ)

g · (k B T /Γ)

0.2 0.25

0.15 0.2
0.15
0.1
0.1
0.05
0.05

−0.1 −0.05 0 0.05 0.1 −0.1 −0.05 0 0.05 0.1


! c (meV) ! c (meV)

Figure 2.9: g · (kB T /Γ) versus dot energy c for various enclosed flux Φ, in the strong bath coupling
limit with a = b and all other parameters as before, at a temperature of 50 mK. (Top Left) φ = 0,
(Top Right) φ = 1.5π/2, (Bot Left) φ = 1.9π/2, (Bot Right) φ = π.

Plotting this conductance, we see that the resonance in figure 2.8 is still somewhat
curved towards the right, for values of Φ 6= 0. By looking at a series of horizontal
line cuts for various values of flux (as shown in figure 2.9) we see that when |X| is
maximal at Φ = 0, the current still passes almost entirely through the ground state.
This is because the energy difference between the two states is still to large to be
overcome by the charge relaxation. As the effective coupling |X| is decreased, the
energy difference between the first and ground state is decreased, and the eigenstate
transitions induced by the bath become strong enough to generate some current.
Curves in blue and yellow represent the processes in lines three and four of equation
2.2 Zero-Bias Conductance 41

2.9, where an electron enters the ground state and is raised to the excited state before
leaving and visa versa. Finally, when |X| = 0 we see that the resonance is actually
at a maximum, due entirely from processes that scatter between molecular states.
In this case we again see the equal contribution from both molecular eigenstates,
creating a symmetric resonance unlike that of the ground state alone. From figure 2.8
we also notice that the conductance is also maximal when Φ/Φ0 = 0.5 exactly, and
the effective coupling |X| = 0. This is because the eigenstates formed by this effective
coupling have an energy difference proportional to |X|. Thus when |X| = 0 scattering
between the two states has no energy cost. As |X| is increased, their energies are
shifted apart causing the resonance to bend again as the ground state passes the
majority of the current.

0.014
Λ = 10−2
0.012
Λ = 10−1
Conductance (e2/h)

0.01 Λ=1
Λ = 101
0.008
Λ = 102
0.006

0.004

0.002

0
−2 0 2
c ( meV ) x 10
−2

Figure 2.10: g · (kB T /Γ) versus dot energy c , for various bath coupling strengths, with enclosed
flux Φ/Φ0 = 0.5 and all other parameters as before, at 50 mK.

Although an analytic solution for arbitrary bath coupling strengths is not easily
obtained, the system can still be solved numerically. From figure 2.10, with bath
coupling |λ|2 = 1, where λ = αj = βj = δj , we find a resonance much sharper than
temperature still remains at Φ/Φ0 = 0.5. With the relevant relaxation rates Γ12 and
Γ21 (see equation 3.27), we define a dephasing rate Γφ ≡ (Γ12 + Γ21 )/2. Here we find
that when |X|, Γφ < kB T , we get a resonance width set my max(|X|, Γφ ). We will
thus find a resonance sharper than temperature (a signature of “dark state” physics)
42 2 Model and Results

as long as Γφ < |X| < kB T , but once |X| < kB T < Γφ we will have reached the
strong relaxation limit. With the relaxation rates given by equation 3.27, we find
the strong relaxation limit (at 50mK) is reached with a bath coupling |λ|2 = 100,
while the critical value at which “dark state” physics can still be resolved is of order
unity. We can now investigate the effect of this dephasing on the Aharanov-Bohm
oscillations caused by this “dark state” physics.

2.2.3 Aharanov-Bohm Oscillations


Examining vertical linecuts of figure 2.5 and figure 2.8, we see how the Aharonov-
Bohm oscillations are affected by strong bath induced dephasing. With no dephasing
(top left of figure 2.11), we see that if c is aligned perfectly at the resonance peak (line
shown in black), the AB oscillation has a periodicity of Φ0 . Moving off resonance to
either side, we find a sharp dip into the central peak (this can be seen slightly in the
black curve due to the resolution in gate voltage used), which results in oscillations
with periodicity closer to Φ0 /2. Analyzing the spectrum of these oscillations via the
fast Fourier transform (shown to the right in figure 2.11), we find that the black curve
is mainly only the first harmonic, which the red and green curves have significant
higher harmonic components. As circling around the triple dot ring twice before
leaving will enclose twice the flux, it is trajectories like these that give rise to the
higher harmonics. In general it should be more likely to circle the ring only once, and
less likely for additional loops. However, it has been reported from the NRC group in
Ottawa that in many cases the third harmonic is dominant. In some of these curves
it appears that the higher harmonics are more dominant that the first harmonic, but
further investigations in this area are required. Looking now at the bottom of figure
2.11, we see the effects of strong dephasing on the AB osciallations. Here we see that
none of the curves go to zero when |X| = 0 as expected. However, on the right side
of the resonance the period of the oscillations still tries to double. This is do to the
way the resonance curves to the right for non-zero values of X as was noted in figure
2.8. There is however, no sign of the phase os these oscillations flipping from one
side of the resonance to the other. This is actually to be expected here, as only ψ+
2.3 Negative Differential Resistance and The “Dark State” 43

is involved in transport in these low bias cases. As ψ+ and ψ− couple to dot c with
strengths that oscillate out of phase, if the resonance were composed of both of these
states (one on either side), then the phase would flip as the current changed from
flowing through one to the other. This has not been further explored. At this point
we will investigate the novel effects found in connecting the low bias and high bias
voltage regimes.

0.5 0.7

0.6
( eh )

0.4
2

Fourier Amplitude
0.5
Conductance

0.3
0.4

0.2 0.3

0.2
0.1
0.1

0 0
0.2 0.4 0.6 0.8 1 0 1 2 3 4 5
Φ/Φ 0 1/B [AU]

0.5 0.7

0.6
( eh )

0.4
2

Fourier Amplitude

0.5
Conductance

0.3
0.4

0.2 0.3

0.2
0.1
0.1

0 0
0.2 0.4 0.6 0.8 1 0 1 2 3 4 5
Φ/Φ 0 1/B [AU]

Figure 2.11: (Top Left) g · (kB T /Γ) versus Φ/Φ0 for various values of c , with no bath coupling
(Λ = 0), and all other parameters left as before. Black is centered at the resonance peak, red to the
left of the resonance peak, and green to the right of the resonance peak. (Top Right) Fast Fourier
transforms of the AB oscillations shown in to the left. (Bot Left)g · (kB T /Γ) versus Φ/Φ0 for various
values of c , with strong bath coupling, and all other parameters left as before. Black is centered at
the resonance peak, red to the left of the resonance peak, and green to the right of the resonance
peak. (Bot Right) Fast Fourier transforms of the AB oscillations shown in to the left.

2.3 Negative Differential Resistance and The “Dark State”


Rectification in a triple dot ring was first reported in the works of Michaelis et al.[40],
where it was found that |ψ− i (with θ = π/2) is an eigenstate of the system which was
44 2 Model and Results

unable to tunnel into dot c (ie a “dark state”) for forward bias voltages, but could
still decay into the left lead with reversed bias. From our treatment at zero-bias when
t0  t, we found a conductance peak near + = c = 0 , depending on the strength
of the effective coupling |X|. Using the magnetic field, the effective coupling of these
two states can be tuned to zero, in which case no current can pass with either forward
or reverse bias voltages. In the forward bias case the electron is found to be trapped
in the |ψ+ i state, whereas for a reversed bias the electron gets stuck in |ci. Although
we do not find the “dark state” physics to provide rectification for zero-bias (and
t0  t), in agreement with Michaelis et al.[40] we do find rectification in the limit of
high bias voltage. Furthermore, by looking at current versus bias voltage we find a
negative differential resistance effect which will complete the explanation of this high
bias rectification.

In the high bias regime that |E+ − E− |  |µL − µR |, we can no longer throw
away the higher energy anti-bonding state |ψ− i of the a, b dots. However, because
tunneling between dots a and b is much greater than from either to dot c, if we keep
the energies of |ci and |ψ+ i aligned their coupling strength will still be given by |X|,
and |ψ− i will remain decoupled. Similarly, if |ψ− i is aligned with |ci they would have
a coupling Y ≡ hψ− |Hsys |ci, and |ψ+ i would be decoupled. The form of this coupling
to the anti-bonding state is |Y |2 = |t|2 (1 − sin θ cos 2πΦ/Φ0 ), making it π out of phase
with |X| with respect to magnetic flux. Thus at Φ = 0, |ψ− i is a “dark state” while
|X| is maximized, and when Φ = Φ0 /2 |ψ+ i is a “dark state”, and |Y | is maximized.

Starting off in the low bias regime, we tune the magnetic flux to zero, keeping
|ψ+ i aligned with |ci where |X| is maximal. Now the anti-bonding state |ψ− i has
an energy difference of − − + = 2t0  kB T , so it is not until µL > 2t0 that the
anti-bonding “dark state” can begin to play a role. Using a symmetric bias voltage
this occurs when eVbias > 4t0 , in which case the current shuts off as shown in figure
2.12. Although the bonding state can still pass a current, in the steady state limit, an
electron will eventually enter into the anti-bonding state and become stuck provided
the bias voltage is high enough that it can’t tunnel back out into the left lead and try
2.3 Negative Differential Resistance and The “Dark State” 45

−7
x 10
3
No Rel
2 Strong Rel

Current (Amps)
1

−1

−2

−3
−5 0 5
Vbias (eV) −4
x 10

Figure 2.12: Current versus bias voltage, for no relaxation and strong relaxation, with all parameters
as set earlier.

again. Thanks to the Coulomb blockade, one trapped electron is enough to block all
current. If the bias voltage is now reversed, an electron fist tunnels from the right lead
into dot c, where it can then still has the option of tunneling to either the bonding or
anti-bonding state. In this case, |X| and |Y | cannot both go to zero simultaneously,
and one state will always be available to pass current. Whereas in the case of forward
bias, as soon as either |X| or |Y | become zero current is blocked as can be seen in
figure 2.13.
With a qualitative understanding of the effects of relaxation, we can easily investi-
gate its effect in the high bias regime numerically. Although in the forward bias case
an electron will eventually scatter into the anti-bonding state, it will now be able to
tunnel into the bonding state instead of getting stuck. Thus only a small amount of
relaxation is necessary to cause a “dark state” to leak some current through, but in
the strong relaxation limit, having both ψ+ and ψ− to contribute actually increases
the total current as seen in figure 2.12. Therefore, the ratio of high bias current to low
bias current could give a measure of level of relaxation and in turn the bath coupling.
We can also see that in the forward bias case having the two levels in the bias window
increases the total current, whereas in the negative bias case, the current remains
unchanged. There is therefore still some asymmetry between the forward and reverse
high bias cases, even in the presence of strong relaxation.
46 2 Model and Results

Current (Ampères)
Current (Ampères)
Φ/Φ0

Φ/Φ0
!c (meV ) !c (meV )

Figure 2.13: Current versus dot energy c and magnetic flux Φ/Φ0 for Vbias = 500µeV (Left) and
Vbias = −500µeV (Right).

Full color maps of the current in the high bias voltage regime for both forward and
reverse bias are plotted in figure 2.13. In the forward bias case, we see that the current
shuts off when either X or Y = 0, whereas in the reverse bias case one state always
passes a current. Although this was expected due to our explanation of the high bias
rectification, there is still a striking difference in the two colormaps. For forward bias,
a zig-zag pattern is formed, while in the reverse bias case we find vertical stripes of
current. The separation of these two stripes is proportional to the inter-dot tunneling
t0 , which if brought sufficiently low would cause the two resonances to merge. In this
case it might be possible to get a phase flip in the AB oscillations, but this has not
been investigated.

To better understand the features that were found in the colormaps, it is noted
that in both cases, the lead-dot tunneling and inter-dot tunnelings do not change.
All that has changed in the two cases is the probability distribution of the molecular
eigenstates (and in turn the charge states). Thus in figures 2.14 and 2.15 we look at
these probability distributions for various magnetic fields in both forwards and reverse
bias cases. By looking at eigenstate energies shown in figure 2.3, we can then tell
which charge states each eigenstate is composed of. In the forward bias case, we see
that for φ = 0 and φ = π the probability is almost entirely in the a, b state which has
gone dark. Tuning φ slightly away from these two points, a region of delocalization
2.3 Negative Differential Resistance and The “Dark State” 47

opens up at the point of c where that particular “dark state” occurs. As one involves
the ψ+ state while the other the ψ− state, the probability must shift from one to
the other, which we see happen at φ = π/2. This is the only symmetric way for
the probability to shift from one to the other, and is a consequence of the geometry
with which the leads couple to the dots. In the reverse bias case we can see that the
electron is most often found in dot c, with some probability spilling into the ψ states
when one or the other has non-zero effective couplings X and Y . In this case the
probabilities is delocalized at two values of c for φ = π/2. This is how asymmetric
way in which the two leads couple to the three dots becomes manifest. This would
not be the case however, if there was a symmetric geometry in which the two leads
couple be connected.
We will now explicitly go through all the details and derivations behind this model
and approach. Then we will summarize the successes of the model, and what outlooks
remain.
48 2 Model and Results

1 1

0.8 0.8
Probability

Probability
0.6 0.6

0.4 0.4

0.2 0.2

0 0

−0.2 −0.1 0 0.1 0.2 −0.2 −0.1 0 0.1 0.2


! c (meV) ! c (meV)

1 1

0.8 0.8
Probability

Probability

0.6 0.6

0.4 0.4

0.2 0.2

0 0

−0.2 −0.1 0 0.1 0.2 −0.2 −0.1 0 0.1 0.2


! c (meV) ! c (meV)

0.8
Probability

0.6

0.4

0.2

−0.2 −0.1 0 0.1 0.2


! c (meV)

Figure 2.14: Current versus dot energy c for Vbias = 500µeV for various values of Φ at 50 mK. (Top
Left)φ = 0,(Top Right)φ = 0.1π/2,(Mid Left)φ = π/2,(Mid Right)φ = 1.9π/2,(Bot)φ = π.
2.3 Negative Differential Resistance and The “Dark State” 49

1 1

0.8 0.8
Probability

Probability
0.6 0.6

0.4 0.4

0.2 0.2

0 0

−0.2 −0.1 0 0.1 0.2 −0.2 −0.1 0 0.1 0.2


! c (meV) ! c (meV)

1 1

0.8 0.8
Probability

Probability
0.6 0.6

0.4 0.4

0.2 0.2

0 0

−0.2 −0.1 0 0.1 0.2 −0.2 −0.1 0 0.1 0.2


! c (meV) ! c (meV)

0.8
Probability

0.6

0.4

0.2

−0.2 −0.1 0 0.1 0.2


! c (meV)

Figure 2.15: Current versus dot energy c for Vbias = −500µeV for various values of Φ at 50 mK.
(Top Left)φ = 0,(Top Right)φ = 0.1π/2,(Mid Left)φ = π/2,(Mid Right)φ = 1.9π/2,(Bot)φ = π.
3
Details of Calculation

3.1 Quantum Master Equations


In order to write an equation for the time evolution of a system, it is convenient to use
a density matrix which contains all relevant information about the state of the system.
For some arbitrary system and complete basis of states {ψα }, the diagonal elements
of the density matrix will equal the state probabilities hψα |ρ|ψα i = |ψα |2 , but the off
diagonal elements are hψα |ρ|ψβ i = ψα∗ ψβ . If the system is only occupying a single
state of a basis set, there will be a single 1 along the diagonal of the matrix, with all
other entries zero. For a system to have probability distributed over multiple states,
we maintain that the electron be found somewhere and that the matrix have a trace of
unity. The off-diagonal elements however, correspond to the expectation value of the
coherence between two of the states. In our triple dot model Hamiltonian (equation
2.2), we now have four charge states coupled to a continuum of lead states resulting
in a potentially infinite dimensional Hilbert space. Rather than solve differential
equations equations tracking the time evolution of every lead state as well as the
system states, we trace out the lead degrees of freedom and derive a master equation
for the reduced density matrix of the four system states. This results in a perturbative
treatment of leads coupling to the system, as will be derived explicitly in subsequent
sections.

In this section we will first review the coherent time evolution of a density operator,
followed by the derivation of the master equation for the time evolution of the reduced
density matrix for a single quantum dot perturbatively coupled to a lead in this way.

50
3.1 Quantum Master Equations 51

We will then see how to extend this treatment to that of our full triple dot system
where strong inter-dot tunneling cannot be neglected. In this case we will show how
this perturbative lead coupling is more adequately treated in the basis of the system’s
molecular eigenstates. Repeating this treatment for the additional Hamiltonian terms
in equation 2.4 for the bosonic bath coupling, we then derive the additional master
equation terms as used in the dephasing treatment of Michaelis and Emary[40, 41].
Again we will see how to generalize the resulting master equation terms to treat a
triple dot system without neglecting the strong inter-dot tuenneling. Finally we put
all these components together to present the full quantum master equation used in
this thesis, and explain how analytic and numerical solutions can be found.

3.1.1 Coherent System Evolution


In order to see how the time evolution of a density operator is written, let us first
imagine a simple system of states for which it is diagonal. Denoting these states |ψα i,
where each is occupied with a probability pα , the density operator for the system is

X
ρ̂ = pα |ψα ihψα |. (3.1)
α

We can then write the time evolution of the density operator from the time evolution
of states,
X
i~ρ̂˙ = pα (i~|ψ̇α ihψα | + i~|ψα ihψ̇α |)
X
= pα (Hsys |ψα ihψα | − |ψα ihψα |Hsys ) (3.2)

= [Hsys , ρ].

In general, we our density matrix will not be diagonal, but the final line of this
equation can be used for any Hamiltonian and a complete basis set.
Using our triple dot system Hamiltonian Hsys as written in equation 2.2 results in
the following master equation:

−i X i X
ρ̇ = i (n̂i ρ − ρn̂i ) − tij (fi† fj ρ − ρfi† fj ), (3.3)
~ i=a,b,c ~ i6=j=a,b,c
52 3 Details of Calculation

which results in 16 coupled differential equations for each element in the 4 × 4 density
matrix of the four system states {|0i, |ai |bi, |ci}. We can however use a basis set in
which the system Hamiltonian is diagonal (such as equation 2.3), in which case the
master equation takes the form

i X ˆ˜† ˆ˜
ρ̇ = − i (fi fi ρ − ρfˆ˜i† fˆ˜i ). (3.4)
~ i

Although both of these mater equations are completely equivalent, it is now much
simpler to project this master equation into the eigenstate basis in which Hsys is
already diagonal. In this case the 4 differential equations for the diagonal elements
of the density operator decouple from the rest, making the task of finding analytic
solutions much simpler(provided you can diagonalize the Hamiltonian in the first
place).
We will now see how to write a master equation for the reduced density matrix
of a single quantum dot coupled to a lead reservoir, and then two possible ways to
generalize the treatment to a triple quantum dot system coupled to two leads. As
this treatment is perturbative, we will see that the use of the eigenstate basis will in
fact differ non-trivially from that of the charge state basis.

3.1.2 Lead-Dot Coupling: A Markovian Master Equation


As an illustrative derivation of the master equation terms that will come from the
Hamiltonian in equation 2.2, let us first consider a single dot with a single level
coupled to a single continuum of lead states described by the following slightly simpler
Hamiltonian:

H = HS + HR + HSR

HS = d d† d
(3.5)
X †
HR = k ck ck
k
X
HSR = t(c†k d + d† ck ),
k
3.1 Quantum Master Equations 53

where we consider HS + HR ≡ H0 while HSR ≡ Hint . We now choose to work in the


interaction picture, where the relevant operators become

ρ̃ = eiH0 t/~ ρSch e−iH0 t/~


(3.6)
iH0 t/~ −iH0 t/~
H̃SR = e HSR e ,

where ρSch is actually an operator composed of wavefunctions in the Schrödinger


picture, while HSR is a usual time independent Schrödinger picture Hamiltonian.
This transformation accounts for the free evolution of the system and reservoir due
to HS and HR respectively, where the only remaining dynamics come from their
interaction resulting in
1
ρ̃˙ = [H̃SR , ρ̃]. (3.7)
i~
H̃SR is now time dependent and requires the use of the time dependent operators
d(t) = de−id t/~ and ck (t) = ck e−ik t/~ . Substituting these into the last line in equation
3.5 and regrouping all of the time dependence into the reservoir operators results in

H̃SR = S † FR (t) + FR† (t)S, (3.8)

−i(k −d )t/~


P
where S = d and FR (t) = t k ck e .
We are now in a position to analyze equation 3.7, but first must consider two
necessary approximations that will be made: 1) The reservoir is large, and 2) The
reservoir has a broad bandwidth. The first assumption implies that the reservoir is not
significantly effected by the interaction, and can be approximated as stationary state
in thermal equilibrium at sum temperature T . The result of the second assumption is
that the reservoir has a correlation time τc which is much shorter than time scale τS
at which the system evolves. Together these two assumptions result in what is called
the “Born-Markov Approximation”, allowing us to average our system over a time
scale ∆t which lies between the reservoirs correlation time and the system’s time scale
of evolution, τc  ∆t  τS . This approach is equivalent to a perturbation theory in
τc /τS , providing equations of motion which are valid for times greater than ∆t.
With these approximations at hand, we formally integrate equation 3.7 over the
54 3 Details of Calculation

time scale ∆t, resulting in

Z ∆t
1
ρ̃(t + ∆t) − ρ̃(t) = dt0 [H̃SR (t0 ), ρ̃(t0 )]. (3.9)
i~ t

Similarly expanding ρ̃(t0 ) and substituting above (which is valid for t0 − t < ∆t),
results in the second order expansion

Z t+∆t Z t+∆t Z t0
1 0 0 0 1 0
∆ρ̃ = dt [H̃SR (t ), ρ̃(t )] + dt dt00 [H̃SR (t0 ), [H̃SR (t00 ), ρ̃(t00 )]],
i~ t (i~)2 t t
(3.10)
where ∆ρ̃ = ρ̃(t + ∆t) − ρ̃(t). Now rather than solving for the density matrix for the
system and all the reservoir degrees of freedom, since we are only interested in what
the system states are doing we solve for the reduced density matrix of the system by
tracing out the reservoir states, where ρ˜S = TrR ρ̃.

Z t+∆t Z t+∆t Z t0
1 0 0 0 1 0
∆ρ̃S = dt TrR [H̃SR (t ), ρ̃(t )]+ dt dt00 TrR [H̃SR (t0 ), [H̃SR (t00 ), ρ̃(t00 )]].
i~ t (i~)2 t t
(3.11)

To proceed in solving this equation, we must make the further assumption that
the system and reservoir are initially factorizable, allowing the density operator to be
written as

ρ̃(t) = ρ̃S (t) ⊗ ρ̃R (t). (3.12)

For short times thereafter, the density operator remains approximately factorizable,
with a correction term that can be neglected since ∆t  τS . This approximation is
valid for a reservoir with no memory, for which τc = 0. This allows us to substitute
equation 3.12 into equation 3.11 and proceed to evaluate the partial traces over the
reservoir states.

Defining ∆ρ̃S = ∆ρ̃(1) + ∆ρ̃(2) , we will first consider the first order contribution
3.1 Quantum Master Equations 55

which can now be written as


1 t+∆t 0
Z
(1)
∆ρ̃S = dt TrR [SFR† (t0 ) + FR† (t0 )S, ρ̃S (t0 ) ⊗ ρ̃R (t0 )]
i~ t
1 t+∆t 0
Z
= dt [S, ρ̃S (t0 )] ⊗ TrR {FR† (t0 ), ρ̃R (t0 )} + H.c.
i~ t
1 t+∆t 0
Z (3.13)
= dt [S, ρ̃S (t0 )] ⊗ 2hFR† (t0 )i + H.c.
i~ t
1 t+∆t 0
Z
= dt [2ShFR† (t0 )i + 2S † hFR (t0 )i, ρ̃S (t0 )].
i~ t | {z }
HS0

With a large reservoir that doesn’t change much on the relevant time scale, hFR† (t0 )i
is essentially constant. This result can be interpreted as an additional term shifting
the coherent system dynamics, and can be neglected.
(2)
The second term, ∆ρ̃S , results in the lowest order non-trivial effects of the reser-
voir, the integrand of which we evaluate and simplify as follows:

I = TrR [H̃SR (t0 ), [H̃SR (t00 ), ρ̃(t00 )]]

= |t|2 TrR [S † F (t0 ) + F † (t0 )S, (S † F (t00 ) + F † (t00 )S)ρS ⊗ ρR − ρS ⊗ ρR (S † F (t00 ) + F † (t00 )S)]

= (S † S † ρS − 2S † ρS S † + ρS S † S † )TrR (F (t0 )F (t00 )ρR )

+ (SS † ρS − SρS S † − S † ρS S + ρS S † S)TrR (F † (t0 )F (t00 )ρR )

+ (S † SρS − SρS S † − S † ρS S + ρS SS † )TrR (F (t0 )F † (t00 )ρR )

+ (SSρS − 2SρS S + ρS SS)TrR (F † (t0 )F † (t00 )ρR ).


(3.14)

Noting that TrR (FR† (t0 )FR (t00 )ρR ) = hFR† (t0 )FR (t00 )i is a time-time correlation function
of the reservoir operators, we evaluate it as
0 00
X †
hFR† (t0 )FR (t00 )i = |t|2 hck ck0 ie−i(k −d )t +i(k0 −d )t
0
Zkk∞ (3.15)
−i(−d )(t0 −t00 )
= |t|
2
dg()f ()e ,
−∞

where ĉ†k ĉk = n̂k is the number operator, resulting in the fermionic thermal average
hc†k ck0 i = f ()δk,k0 . This delta function kills one sum over k, while the other is con-
verted to an integral via k → dg(), where g() is the density of states of the
P R
56 3 Details of Calculation

reservoir. Similarly hck c†k0 i = (1 − f ())δk,k0 , while the remaining averages vanish,
hc†k c†k0 i = hck ck0 i = 0. Rewriting the equation for ∆ρ̃S , now in terms of these two
(2)

time correlation functions results in


Z t+∆t Z t0
(2) 1
∆ρ̃S = 0
dt dt00 [hFR† (t0 )FR (t00 )i(SS † ρ̃S −S † ρ̃S S)+hFR (t0 )FR† (t00 )i(S † S ρ̃S −S ρ̃S S † )].
(i~)2 t t
(3.16)
The remaining time dependence comes only from integrating the time-time correlation
functions of the reservoir:
Z t0 Z t0 Z ∞
† 0 0 00 )
dt hFR (t )FR (t )i = |t|
00 00 2 00
dt dg()f ()e−i(−d )(t −t
t t −∞
Z ∞ Z t0
0 00
' |t| 2
dg()f () dt00 e−i(−d )(t −t ) (3.17)
−∞
|t {z }
' πδ(−d )

= π|t|2 g(d )f (d ),

where we have used a Wigner-Weisskopf approximation in bringing the time inte-


gral past the energy integral and setting t = −∞ and t0 = 0, to use the identity
1
R ∞ −it
δ() = 2π −∞
e to kill the final energy integral. We thus define the lead-dot
coupling constant as Γ = 2π
~
|t|2 g0 , and rewrite equation 3.7 in form of a master equa-
tion, commonly known as a “jump term” since it causes an abrupt change to the
wavefunction and reduced density operator of the system:
(2)
∆ρ̃S 1 1
= Γ[(d† ρd − (dd† ρ + ρdd† ))f (d ) + (dρd† − (d† dρ + ρd† d))(1 − f (d ))], (3.18)
∆t 2 2

Defining separate transition rates into and out of the dot as Γd0 = Γf (d ) and
Γ0d (1 − f (d )), and limiting ourselves to two states {|0i, |di}, we can re-write these
“jump terms” as

1 1
ρ̇ = Γd0 (L† ρL − (LL† ρ + ρLL† )) + Γ0d (LρL† − (L† Lρ + ρL† L)), (3.19)
2 2
(2)
∆ρ̃S
where we can now take ∆t
= ρ̇, and write L̂ = |0ihd| as the “jump operator” that
incoherently transitions the system from the empty state to occupied state. Here we
see that a Fermi function is present for transitions into the dot as an electron in the
3.1 Quantum Master Equations 57

lead must be present with the energy of the charge state it is moving into, and that
one minus a Fermi function is present for transitions into the lead in which case the
lead Fermi sea must have a vacancy with that energy.

At this point we can see a sufficiently simple form to these “jump terms”, that we
could easily just write them down by hand for any particular transition, then adding
on the rate of this transition. The transition rates can then be calculated via Fermi’s
golden rule to lowest order. Next we will see the simplest way in which to couple two
leads to the triple dot system of interest, and then consider another technique more
suitable for this case.

However, i working in the interaction picture we had to include a time dependance


due to the system Hamiltonian. With the two dots being tunnel coupled together, now
this system Hamiltonian must include this tunneling. In this case it is more natural
to work in the basis of the system Hamitonian’s molecular eigenstates. Next, we will
see how to write “jump terms” in which electrons tunnel from the leads directly into
these molecular eigenstates.

3.1.3 Lead-Dot Coupling: FGR Transition Rates

For the full triple dot system we are interested in, we have the four states {|0i, |ai, |bi, |ci}
coupled to left and right leads as shown in figure 2.1. Now we can just write down
three sets of these jump terms, resulting in the following master equation terms:

X 1
ρ̇ = {Γi0 [fˆi† ρfˆi − (fˆi fˆi† ρ + ρfˆi fˆi† )]
2
i={a,b,c} (3.20)
1
+ Γ0i [fˆi ρfˆi† − (fˆi† fˆi ρ + ρfˆi† fˆi )]},
2
58 3 Details of Calculation

where Γi0 and Γ0i are transition rates from the |0i state to the ith state and visa versa,
respectively. FGR then gives, for example
2π X
Γa0 = |hf |HLa |ii|2
~ i
2π X X
= |ha|hleadf | wk (fa† ck + c†k fa )|leadi i|0i|2 δ(f − i )
~ i k
2π X
= |ha|fa† |0i|2 |w (hFS|c†k )ck |FSi|2 δ((a + F S − k ) − (0 + F S ))
~ k (3.21)
2π 2 X
= |w| f (k )δ(k − a )
~ k
2π 2 ∞
Z
= |w| dk g(k )f (k )δ(k − a )
~ −∞
2π 2
= |w| g0 f (a ),
~
where we define the lead-dot coupling constant Γ = 2π
~
|w|2 g0 as before, now assuming
that the lead density of states is a constant over the energy range of the dot states.
Here we have used |ii = |0i ⊗ |leadi i and |f i = |ai ⊗ |leadf i as the initial and
final states of the system and lead, with |leadi i = |FSi and |leadf i = ĉk |FSi, where
|FSi is a filled Fermi sea. The initial and final energies are thus i = 0 + F S and
f = a + F S − k . Now to include the chemical potential of the left lead we replace
the energy of the empty state 0 = 0 with the chemical potential of the left lead
giving the FRG tunnel rate Γa0 = Γf (a − µL /e). The FGR tunnel rate into dot b
is similarly Γb0 = Γf (b − µL /e), while the the rate into dot c, Γc0 = Γf (c − µR /e),
has the chemical potential of the right lead. An applied bias voltage is the result in
an offset in the left and right chemical potentials, defined as eVbias = µL − µR for a
symmetric bias in the forward direction.
Finally, assuming we solve our master equation for the dot state probabilities,
X
we can write the current most generally as IL = e (Γi0L ρ00 − Γ0iL ρii ) = IR =
i={a,b}
e(Γ0cR ρcc − Γc0R ρ00 ) = I. For our triple dot system in Coulomb blockade regime,
the high bias voltage limit is reached when |i |  µα /e < ECi . In this case rates
moving electrons in the leftward direction are reduced to zero, Γ0a = Γ0b = Γc0 = 0,
while the right moving rates become Γa0 = Γb0 = Γ0c = Γ. This is the limit used
3.1 Quantum Master Equations 59

by Michaelis et al.[40, 41], as for a triple dot system with inter-dot tunneling much
greater than lead-dot tunneling, this model breaks down at low bias voltages as will
be discussed in section 3.3. In this case the current expression simplifies further to
X
I = e Γi0L ρ00 = eΓ0c ρcc , but the zero-bias conductance cannot be calculated.
i={a,b}
This problem can be overcome by perturbatively tunneling the lead electrons directly
into the molecular eigenstates of the strongly coupled dot system, a perturbation that
in fact goes to higher order that tunneling into the dot state basis.

The problem with writing generalizing the the jump terms derived for one dot to
three dots, is that using in the interaction picture the time evolution from H0 must
now include tunneling between the dots. To include this inter-dot tunneling, it is
natural to work in the eigenstate basis of the system Hamiltonian. In order to write
these lead-dot coupling master equation terms for tunneling into the eigenstate basis,
we can simply write

1
{Γi0 [fˆ˜i† ρfˆ˜i − (fˆ˜i fˆ˜i† ρ + ρfˆ˜i fˆ˜i† )]
X
ρ̇ =
2
i={1,2,3} (3.22)
1
+ Γ0i [fˆ˜i ρfˆ˜i† − (fˆ˜i† fˆ˜i ρ + ρfˆ˜i† fˆ˜i )]},
2

where the sum is now over the three molecular eigenstates, and the jump operators
used now become ˜fˆ i = |0ihEi |, and |Ei i = ai |ai + bi |bi + ci |ci. The lead tunneling
rates into the ith eigenstate again follow from Fermi’s golden rule giving, for the left
60 3 Details of Calculation

lead into the ground state for instance,

2π X
Γ10L = (|hf |HLa |ii|2 + |hf |HLb |ii|2 )
~ i
2π X X
= (|hE1 |hleadf | wk (fa† ck + c†k fa )|leadi i|0i|2
~ i k
X
+ |hE1 |hleadf | wk (fb† ck + c†k fb )|leadi i|0i|2 )δ(f − i )
k
2π X
= (|ha|fa† |0i|2 + |ha|fb† |0i|2 ) |w(hFS|c†k )ck |FSi|2 δ((E1 + F S − k ) − (µL /e + F S ))
~
Z ∞k
2π 2
= |w| (hE1 |ai + hE1 |bi) dk g(k )f (k )δ(k − E1 + µL /e)
~ −∞

= Γ(|a1 |2 + |b1 |2 )f (E1 − µL /e).


(3.23)

Each eigenstate now couples to both left and right leads making the total rate into
eigenstate i, Γi0 = Γi0L + Γi0R , where the rate Γ10R = Γ|c1 |2 f (E1 − µR /e) is simi-
larly calculated with HR . In this novel approach, the Fermi functions in the tran-
sitions to the molecular eigenstates are functions of the eigenstate energies rather
than those of the individual charge states. This causes current resonances when
a molecular state energy is degenerate with the leads, not when the three electro-
static charge energies are degenerate favoring sequential tunneling. This has recently
been observed experimentally for strongly coupled dot systems and triple dot systems
where two strongly coupled dots are contrasted with two weakly coupled dots[28]. Fi-
X
nally, we again calculate the current via IL = e (Γi0L ρ00 − Γ0iL ρii ) = IR =
i={1,2,3}
X
e (Γ0iR ρii − Γi0R ρ00 ) = I, where now it is clear that each eigenstate can con-
i={1,2,3}
tribute to both IL and IR . In this case the zero-bias conductance can be calculated
via the formula g = ∂I/∂Vbias |Vbias =0 . It makes no difference whether a symmetric
bias or not as we will be taking the bias voltage will be taked to zero anyways, so for
simplicity we use µL = Vbias and µR = 0. Now, rather than “jump terms” coupling
the system to metalic lead contacts, we will further consider the effects of coupling
the dots to a bosonic bath and the dephasing induced.
3.1 Quantum Master Equations 61

3.1.4 Bath-Dot Coupling: Dephasing and Eigenstate Transitions


Considering again a single dot now coupled to a bosonic bath, we can now follow the
same derivation of the lead-dot “jump terms” using the Hamiltonian in equation 2.4
for a single dot. In this case the interaction Hamiltonian takes the form
X
HSR = (λn̂d )(âj + â†j )
j (3.24)
† †
= F S + S F,
X
where now S = nd , and F = (âj + â†j ). Substituting S into equation 3.16 now
j
results in
Z t+∆t Z t0
(2) 1
∆ρ̃S = 0
dt dt00 [hF † (t0 )F (t00 )i(nd n†d ρ̃S −n†d ρ̃S nd )+hF (t0 )F † (t00 )i(n†d nd ρ̃S −nd ρ̃S n†d )],
(i~)2 t t
(3.25)
where now knowing that the time-time correlation functions will just result in tran-
sition rates that can be calculated via FGR, we can immediately write down a the
relevant master equation “jump terms” for the system state transitions described
above. Writing nd = d† d = |dih0|0ihd| = |dihd|, we see that the resulting transitions
are from the charge state back to the same charge state. Generalizing this result back
to three dots while ignoring inter-dot tunneling, results in the dephasing jump terms
used by Michaelis and Emary[40, 41]:
X 1
ρ̇ = Λ[|jihj|ρ|jihj| − (|jihj|ρ + ρ|jihj|), (3.26)
2
j∈{a,b,c}

where Λ is currently an arbitrary dephasing rate which would be calculated via FGR.
These “jump terms” result in ρ̇aa = ρ̇bb = ρ̇cc = ρ̇00 = 0 and ρ̇ij = −Λρij for
i, j ∈ {a, b, c}. This gives an exponential decay in coherence between any of the two
dots, localizing the electron in a particular dot rather than a molecular eigenstate.
However, now that we have written our coherent evolution and lead-dot tunneling
in the eigenstate basis, it will be convenient to consider the effects of this charge
dephasing on the molecular eigenstates.
As we wish to analyze our triple dot system in the regime that inter-dot tunneling is
much stronger than the bath coupling, we do not wish to neglect inter-dot tunneling in
62 3 Details of Calculation

the derivation of the bath induced “jump terms”. In performing a elementary change
of basis from {|ai, |bi, |ci} to {|E1 i, |E2 i, |E3 i}, we find this dephasing to actually
induce incoherent transitions between the molecular eigenstates. Knowing that we
were interested in the molecular eigenstates, we could have just started with FGR to
see what it does to these eigenstates. From this we find the transition rate from the
j th molecular eigenstate to the k th eigenstate,
2π X
Γkj = |hf |He−p |ii|2 δ(Ef − Ei )
~ i,f
2π X
= |λhk|(n̂a + n̂b + n̂c )|ji|2
~ i,f
X
⊗ |hbathf |(â−q + â†q )ρbath |bathi i|2 δ(Ef − Ei )
q
2π 2 X
= |λ| |hk|n̂m |ji|2 (3.27)
~
m={a,b,c}
X
⊗ |hbathf |ρbath |bathf i|2 δ(Ef − Ei )
f
Z
2π 2 X
= |λ| |hk|n̂m |ji|2
dεg(ε)nB (ε)δ(ε − ~ωkj )
~
m={a,b,c}
2π 2 X
= |λ| g(~ωkj )nB (~ωkj ) |hk|mi|2 |hj|mi|2 ,
~
m={a,b,c}

where ~ωkj = Ek − Ej , the initial and final states are defined as |ii ≡ |ji ⊗ |bathi i
and |f i ≡ |ki ⊗ |bathf i, and g() is the bath density of states, which we take to be
linear for simple Ohmic dissipation. The bath coupling constant is therefor defined as
Λ= 2π
~
|λ|2 g(~ωkj ). The sum of matrix elements in the above expression says that for
a transition between states k and j, those eigenstates must overlap with the localized
dots actually interacting with the bath. Finally, we can write the additional master
equation jump terms as
X 1
ρ̇ = Γij [L̂ij ρL̂†ij − (L̂†ij L̂ij ρ + ρL̂†ij L̂ij )], (3.28)
i6=j
2

where L̂ij = |Ei ihEj | is the jump operator that takes an electron from the j th to
ith eigenstate. We then define an effective dephasing rate from the relaxation rates
3.1 Quantum Master Equations 63

for the fist two eigenstates as Γφ = (Γ12 + Γ21 )/2, where the dephasing rate has a
clear temperature dependance and an energy cost associated with scattering between
molecular states. The dephasing “jump terms” (equation 3.26) used by Emary et al.
[41] overlook this energy cost, as they do not induce transitions in the system state
or its energy. Transitions between non-degenerate molecular eigenstates do cause a
change in the system energy, which must be made up for by the bath, either giving
or receiving the necessary energy. This highlights the connection between charge
dephasing and energy relaxation, showing that Emary’s dephasing model is in fact
valid only in the high bath temperature limit where any amount of energy can be
absorbed or emitted with equal probability. We now have all the elements necessary
to write the complete master equation of our triple dot system, which we will now
attempt to solve.

3.1.5 Solving A Quantum Master Equation


We can now gather together the master equation components of equations 3.19,3.22,
and 3.28, writing the complete master equation for our triple dot system as
i X ˆ˜† ˆ˜
ρ̇ = − i (fi fi ρ − ρfˆ˜i† fˆ˜i )
~ i
1
{Γi0 [f˜ˆi† ρf˜ˆi − (f˜ˆi f˜ˆi† ρ + ρfˆ˜i fˆ˜i† )]
X
+
2
i={1,2,3}
(3.29)
1
+ Γ0i [fˆ˜i ρfˆ˜i† − (fˆ˜i† fˆ˜i ρ + ρfˆ˜i† fˆ˜i )]}
2
X 1
+ †
Γij [L̂ij ρL̂ij − (L̂†ij L̂ij ρ + ρL̂†ij L̂ij )]
2
i6=j={1,2,3}

where i is the ith eigenstate energy, Γi0 and Γ0i are the tunnel rates into and out of the
eigenstates via the leads, and Γij ’s are the transition transition rates from the j th to
ith eigenstate. The creation and annihilation operators for the molecular eigenstates
are defined as fˆi ≡ |0ihEi |, and the eigenstate transition operators as L̂ij ≡ |Ei ihEj |.
In order to now solve our master equation, we project our quantum operators
onto the molecular eigenstate basis, generating a set of coupled differential equations
ρ̇ij = hi|ρ̇|ji, where i, j ∈ {0, E1 , E2 , E3 }. In the eigenstate basis, the differential
64 3 Details of Calculation

equations for the diagonal elements of the density operator decouple from the off-
diagonal elements, resulting in a system of four coupled differential equations rather
than sixteen necessary to characterize the current. Solving for the steady state current
through the system in turn results in solving the following system of equations:

ρ˙00 = −(Γ10 + Γ20 + Γ30 )ρ00 + Γ01 ρ11 + Γ02 ρ22 + Γ03 ρ33 = 0

ρ˙11 = Γ10 ρ00 − (Γ01 + Γ21 + Γ31 )ρ11 + Γ12 ρ22 + Γ13 ρ33 = 0
(3.30)
ρ˙22 = Γ20 ρ00 + Γ21 ρ11 − (Γ02 + Γ12 + Γ32 )ρ22 + Γ23 ρ33 = 0

ρ˙33 = Γ30 ρ00 + Γ31 ρ11 + Γ32 ρ22 − (Γ03 + Γ13 + Γ23 )ρ33 = 0.

We can then map these four diagonal elements into a vector ρ~˙ = (ρ̇00 , ρ̇11 , ρ̇22 , ρ̇33 ),
which now results in the usual matrix representation of a system of differential equa-
tions, ρ~˙ = M~
ρ, where ρ is now a vector and M is a matrix. As we are only interested
in the steady state current of our device, we need only solve for the nullspace of M.
In the regime of zero-bias conductance, where we have seen that only the first two
molecular eigenstates contribute to the current resonance as discussed in section 2.2
(see figure 2.3), we can actually solve the above system analytically in the limits of
no relaxation and strong relaxation. With the aid of Maple, one can easily find the
solutions

ρ̄00 = Γ01 Γ02 + Γ01 Γ12 + Γ02 Γ21

ρ̄11 = Γ10 Γ02 + Γ10 Γ12 + Γ20 Γ12 (3.31)

ρ̄22 = Γ01 Γ20 + Γ10 Γ21 + Γ20 Γ21 ,

which when properly normalized are given by ρii = ρ̄ii /ρ̄ for i ∈ {1, 2, 3}, with
ρ̄ = ρ̄00 + ρ̄11 + ρ̄22 . We then calculate the current through the left lead-dot junction
as

I = ρ00 (Γ10L + Γ20L ) − ρ11 Γ01L − ρ22 Γ02L

= [(Γ01 Γ10L − Γ10 Γ01L )(Γ02 + Γ12 ) + (Γ02 Γ20L − Γ20 Γ02L )(Γ01 + Γ21 ) (3.32)

+ (Γ01 Γ20L − Γ20 Γ01L )Γ12 + (Γ02 Γ10L − Γ10 Γ02L )Γ21 ]/ρ̄,
where we have used the fact that Γ0i = Γ0iL + Γ0iR and similarly Γi0 = Γi0L + Γi0R .
Setting Γ12 = Γ21 = 0, four of the final six terms vanish, and the current can be
3.1 Quantum Master Equations 65

reduced to
T1L T1R T2L T2R
INo Rel = [f (E1 −µL /e)−f (E1 −µR /e)]Γ02 + [f (E2 −µL /e)−f (E2 −µR /e)]Γ01 ,
ρ̄ ρ̄
(3.33)
where we have used the fact that (Γ0i Γi0L − Γi0 Γ0iL ) = TiL TiR [f (Ei − µL /e) − f (Ei −
µR /e)] for i ∈ {1, 2}, with Γ0iL = TiL [1 − f (e1 − µL /e)] and Γi0L = TiL f (E1 − µL /e).
The zero-bias conductance can then be most easily calculated by setting µL = eVbias
and µR = 0 and then using g = ∂I/∂Vbias |Vbias =0 . This results in the conductance
equation 2.7 presented in section 2.2.
Solving equation 3.31 in limit that the eigenstate relaxation rates are much larger
than the lead-dot tunneling rates requires a bit more work. First we simplify the
current equation to

I = T1L T1R [f (E1 − µL /e) − f (E1 − µR /e)](Γ02 + Γ12 )/ρ̄

+ T2L T2R [f (E2 − µL /e) − f (E2 − µR /e)](Γ10 + Γ21 )/ρ̄

+ T2L T1R [f (E2 − µL /e) − f (E1 − µR /e)eβ(E1 −E2 )


(3.34)
+ f (E2 − µL /e)f (E1 − µR /e)(eβ(E1 −E2 ) − 1)]Γ12 /ρ̄

+ T1L T2R [f (E1 − µL /e)eβ(E1 −E2 ) − f (E2 − µR /e)

+ f (E1 − µL /e)f (E2 − µR /e)(1 − eβ(E1 −E2 ) )]Γ12 /ρ̄.


Finally, to take the strong relaxation limit, we make the simplification
ρ̄ (Γ01 Γ02 + Γ10 Γ02 + Γ01 Γ20 )
= +Γ01 + Γ10 + Γ20 + (Γ02 + Γ20 + Γ10 )eβ(E1 −E2 )
Γ12 Γ12
| {z }
→0

' T1 [1 + f (E1 )eβ(E1 −E2 ) ] + T2 [f (E2 ) − eβ(E1 −E2 ) ],


(3.35)

where Ti = TiL + TiR for i ∈ {1, 2} and β = 1/kB Tbath . Also using the fact that
Γ21 /Γ12 = eβ(E1 −E2 ) , and taking derivatives with respect to voltage, the equation 3.34
can be reduced to equation 2.9. As the bias voltage will be taken to zero, chemical
potentials in equation 3.35 have been prematurely omitted.
Although it may appear that the derivation of the conductance equations is com-
plete, the price we pay for working in the molecular eigenstate basis is having to
66 3 Details of Calculation

actually find analytic equations for them. This is particularly simple and illustrative
in the limit if zero-bias conductance where only two eigenstate solutions are needed.

3.2 Analytic Diagonalization


In order to diagonalize the system Hamiltonian given in line one of equation 2.2, we
first write it as a matrix in the basis of the three dot states {|ai, |bi, |ci}, giving
 
0
 t t
 a 
 0
Hsys =  t b t  , (3.36)

 
t∗ t∗ c

We can begin by diagonalizing the strongly coupled a and b dots (ie the 2 × 2 sub-
matrix in the top-left corner), resulting in
 
+ 0 X
 
Hsys =  0 − Y  , (3.37)
 
 
∗ ∗
X Y c

where the + and − are the energies of the |ψ+ i and |ψ− i molecular eigenstates
formed by the a and b dots and defined by equation 2.5. Now, in the limit that
t0  t, also with t0  kB T in the zero-bias conductance regime, we can throw out
the higher energy |ψ− i state, resulting in another 2 × 2 matrix to diagonalize:
 
+ X
Hsys =  , (3.38)

X c

where the effective coupling X (as shown earlier in equation 2.6) is given by the
matrix element of the system Hamiltonian given by line one of equation 2.2, coupling
the states |ψ+ i and |ci,

X = h+ |Hsys |ci


θ θ
= (cos ha| + sin hb|)(|t|eiφ d†c da + |t|d†c db + H.c.)|ci (3.39)
2 2
θ −iφ θ
= |t| cos e + |t| sin ,
2 2
3.2 Analytic Diagonalization 67

where for simplicity here, the phase difference in the two paths is put all in the a-to-c
path, which is allowed by the gauge invariance of the magnetic fields vector potential.
The magnitude of this coupling is then given by

|X|2 = X ∗ X = |t|2 (1 + sin θ cos 2πΦ/Φ0 ), (3.40)

as shown in section 2.2. From equation 2.5 in section 2.2 we saw that θ is the mixing
angle for the |ψ+ i and |ψ− i molecular states, and that the two are most evenly
delocalized between dots a and b when θ = π/2, and in turn a = b . This is required
in order for X to reach both its maximum and minimum, which occur when φ = 0
and φ = π respectively. This is due to constructive or deconstructive interference
of the two paths, which requires both paths to contribute equally for the maximum
effect.
In order to now diagonalize equation 3.38 in the same way as the previous 2 × 2
sub-matix, we write X in complex polar form X = |X|eiα , resulting in the relevant
ground state and first excited state of our system:
θ δ θ δ δ
|E1 i = cos cos eiα/2 |ai + sin cos eiα/2 |bi + sin e−iα/2 |ci
2 2 2 2 2
θ δ iα/2 θ δ iα/2 δ
|E2 i = − cos sin e |ai − sin sin e |bi + cos e−iα/2 |ci
2 2 2 2 2
1 1p (3.41)
E1,2 = (+ + c ) ∓ (+ − c )2 + 4|X|2
2 2
2|X| ImX
tan δ ≡ , tan α ≡ .
c − + ReX
These are the analytic molecular eigenstate equations for the full system’s ground
and first excited states, as used in equation 2.8. From these equations we see that
these molecular eigenstates are most evenly delocalized across the system when both
θ = π/2 and δ = π/2 which occurs when both a = b and + = c . This is where the
triple point had been shifted too in figure 2.2, plus a further shift of amplitude |X|
which can now be made clear. Tuning the system such that + = c = 0 , we still find
an energy difference between |E1 i and |E2 i resulting in the current maximum when
E1 = 0. From the energy different E2 − E1 = 2|X|, and the symmetry of the two
states energy anti-crossing, we see this shift of |X|. This results in |X|, the effective
68 3 Details of Calculation

coupling of |ψ+i and |ci, directly controlling the size of the anti-crossing and range of
c over which the ground state will be delocalized. Counterintuitively, as this coupling
gets smaller and the energy difference between ground and excited state approaches
zero, the current peak amplitude is increased as the ground and excited states both
contribute currents over a narrower range. This creates the sharper, taller resonance
peak seen just before the current shuts off in figure 2.5.
We can now use equation 3.41 to calculate the matrix elements as shown in section
2.2 equation 2.8, where we now have

T1L T1R = (|ha|E1 i|2 + |hb|E1 i|2 )|hc|E1 i|2


θ θ β β
= (cos2 + sin2 ) cos2 sin2
| 2 {z 2} 2 2
=1 (3.42)
1
= sin2 β
4
|X|
= |X| ,
4|X|2 + (∆E)2

where in fact T1L T1R = T2L T2R as stated in equation 2.8.


Now by diagonalizing the strongly coupled a and b dots first, we have essentially
changed the picture from two spatially separate dots with equal energy levels, and
converted them to a single dot with two energy levels separated by 2|t0 |. In a regime
where t0  Γ, kB T , we can throw out this second state on the left dot, reducing the
model to an equivalent double dot system, where each dot has a single level, which
are coupled by two possible transport paths.

3.2.1 Equivalent 2-path Double Dot


After diagonalizing the strongly coupled dots a and b to be one large dot A, and
keeping only its ground state, our triple dot system is reduced to that of a double
dot system with two paths connecting the dots as shown in figure 3.1. This can be
represented by the following Hamiltonian

Hef f = (tt + tb )d†c dA + (tt + tb )∗ d†A dc , (3.43)


3.2 Analytic Diagonalization 69

t1

µL Γ A Φ c
Γ µ
R

t2
Figure 3.1: Schematic of a double dot system connected via two tunnel couplings representing two
spatially separated paths enclosing a magnetic flux Φ

which takes the matrix form of


 

eA X̃
Hef f =  , (3.44)
X̃ ec

where we use the basis {|Ai, |ci} and have denoted the effective coupling (tt +tb ) = X̃.
This effective coupling can then be calculated to be

|X̃|2 = ||tt |eiφ/2 + |tb |e−iφ/2 |2

= |tt |2 + |tb |2 + 2(tt · tb ) cos φ (3.45)

= 2|t|2 (1 + cos φ),

where we have taken tt = |t|eiφ/2 and tb = |t|e−iφ/2 . Aside from the extra pre-factor
of 2, we see that if the tunnel coupling of the two paths is equivalent, we recover
the effective coupling in the triple dot system with θ = π/2. This is because in the
double-dot system, the large left dot A has no internal structure, while the dots a
and b in the triple dot system do. This clearly gives maximal coupling at zero field
and zero coupling at φ = π, so long as |tt | = |tb | = t. But, supposing that |tt | = t and
|tb | = t + , we find that

|X̃|2 = t2 + (t + )2 + 2t(t + ) cos φ


(3.46)
= 2 + 2t(t + )(1 + cos φ).

which reduces to X̃ =  at φ = π, no longer vanishing to zero. Let us compare this


to the triple dot coupling X for a small detuning of eb − ea = ˜. In this case we can
70 3 Details of Calculation

expand and simplify the expression sin θ as follows:


2t0
sin θ = √
4t02 + ˜2
= 2t0 (4t02 + ˜2 )−1/2
(3.47)
˜2
= (1 + 02 )−1/2
4t
≈ 1 − 2 ,
where 2 ≡ ˜2 /8t2ab , and 2  1. This gives the triple dots effective coupling to be

|X|2 = T (1 + (1 − 2 ) cos φ). (3.48)

We have now seen that when the energies of dots a and b are equal (θ = π/2), our
system is essentially reduced to that of a double dot system with two transmission
paths of equal transmission. In fact, the detuning of ea and eb is analogous to a
detuning of the two transmission amplitudes in the double dot system. This is a
parameter not easily tunable in most double dot fabrications.
In the next section on other master equations, we consider an effective double dot
model, in which the left dot is actually the molecular eigenstate |ψ+ i. This model
will contain both the two-path double dot structure, as well as the additional internal
double dot structure of the left dot, but be solvable exactly.

3.3 Other Master Equations


In this section we will consider two other possible master equations that could describe
our system. Although the failure at low biases of the master equation in the charge
state basis, as used by Michaelis and Emary[40, 41] in the high bias, could not be
fully explained, even an effective double dot master equation which is solvable exactly
is shown to fail in the low bias regime. Each of these models however is in agreement
with the main results of this thesis at high bias voltages.

3.3.1 The Analogous 2-Path Double Dot Master Equation


In writing an effective 2-path double dot master equation, we will have a left and
right each coupled to a left and right lead, but we will still write these new dot states
3.3 Other Master Equations 71

as |L >= cos 2θ |a > + sin 2θ |b > and |R >= |c > in terms of the actual triple dot
system. This preserves the original double dot structure in effective left dot, making
this model completely analogous to that used in this thesis. This double dot system
is now governed by the Hamiltonian
X
HLR = α c†α cα + Xc†L cR + X ∗ c†R cL , (3.49)
α=L,R

resulting in the quantum master equation


−i
ρ̇ = [L (nL ρ − ρnL ) + R (nR ρ − ρnR ) + X(c†L cR ρ − ρc†L cR ) + X ∗ (c†R cL ρ − ρc†R cL )]
~
X 1 1
+ {Γα0 [c†α ρcα − (cα c†α ρ + ρcα c†α )] + Γ0α [cα ρc†α − (c†α cα ρ + ρc†α cα )]}.
α=L,R
2 2

(3.50)
The steady state conductance requires the solution to the following set of equations:

ρ̇00 = −(ΓL0 + ΓR0 )ρ00 + Γ0L ρLL + Γ0R ρRR = 0


−i
ρ̇LL = (XρRL − X ∗ ρLR ) + ΓL0 ρ00 − Γ0L ρLL = 0
~
i (3.51)
ρ̇RR = (XρRL − X ρLR ) + ΓR0 ρ00 − Γ0R ρRR = 0

~
−i 1
ρ̇LR = [(L − R )ρLR + X(ρRR − ρLL )] − (Γ0R + Γ0L )ρLR = 0,
~ 2
where the off diagonal elements of the density matrix no longer decouple as is the
basis of the entire systems eigenstates. The last of these equations with its Hermitian
conjugate gives the following equation,
i 1 (Γ0L + Γ0R )
− (XρRL − X ∗ ρLR ) = − 2 |X|2 1 (ρLL − ρRR ), (3.52)
~ ~ 4
(Γ0L + Γ0R )2 + ~12 (L − R )2
where (Γ0L + Γ0R ) ≡ Γoff . We can then define a new rate Γ̃, where
|X|2 Γoff
Γ̃ = 2 Γoff 2 , (3.53)
~ ( 2 ) + ( ∆
~
)2
which is a Lorentzian of width Γoff , with the additional scaling of |X|2 /~2 . Substitut-
ing back into the middle two equations above, we now get

−Γ̃(ρLL − ρRR ) + ΓL0 ρ00 − Γ0L ρL L = ρ̇LL = 0

Γ̃(ρLL − ρRR ) + ΓR0 ρ00 − Γ0R ρR R = ρ̇RR = 0 (3.54)

,
72 3 Details of Calculation

where it now appears that Γ̃ = ΓLR = ΓRL is now a classical tunnel rate connecting
the left and right dots. These equations are equivalent to the double dot classical
master equations that we shown in section 1.2.1. These will in fact give the same
current equation but with Γ̃ instead of tab . These two equations combined with the
fact that ρ00 = 1 − ρLL − ρRR , give the following solution:

ρLL = (Γ̃Γon + ΓL0 Γ0R )/D

ρRR = (Γ̃Γon + ΓR0 Γ0L )/D


(3.55)
ρ00 = (Γ̃Γoff + Γ0L Γ0R )/D

D = Γ̃(ΓL + ΓR + ΓL0 + ΓR0 ) + (ΓL ΓR − ΓL0 ΓR0 )

The current is then given by the following formula,

I = e(ΓL0 ρ00 − Γ0L ρLL )


Γ̃
=e (Γ0R ΓL0 − ΓR0 Γ0L ) (3.56)
D
Γ̃
= e γ 2 TL TR [f (L − µL ) − f (R − µR )],
D
where the tunnel rates have been set to be:

ΓL0 = γTL f (L − µL )

Γ0L = γTL [1 − f (L − µL )]


(3.57)
ΓR0 = γTR f (R − µR )

Γ0R = γTR [1 − f (R − µR )],

where γ is the lead coupling, and TL ,TR are the matrix elements. As this model is
equivalent to a double dot model, it is known to be pathalogical at low bias voltages.
For instance the current does not go to zero as the bias voltage does. Zero-bias
conductance, in turn cannot be calculated in this case.

3.3.2 Tunneling into Charge States


Using only the charge state basis, we could write a master equation which now has
terms for inter-dot tunneling, as well as the simple lead coupling jump terms which
3.3 Other Master Equations 73

neglect this tunneling. This results in


−i X i X
ρ̇ = i (ni ρ − ρni ) − tij (d†i dj ρ − ρd†i dj )
~ i=a,b,c ~ i6=j=a,b,c
X 1
+ Γi0 [d†i ρdi − (di d†i ρ + ρdi d†i )] (3.58)
i=a,b,c
2
X 1
+ Γ0i [di ρd†i − (d†i di ρ + ρd†i di )],
i=a,b,c
2

where the Hamiltonian still only couples the left lead to dots a and b, and the right
lead to c. This is essentially the master equation used by Emary et al.[41] with no
dephasing. In this case we start off allowing lead tunneling in both directions. In
projecting this master equation on the charge state basis in order to begin solving, one
no longer finds the differential equations for the diagonal elements to decouple from
the off-diagonal elements. This is due to the inter-dot tunneling terms now explicit
in our master equaiton. In fact, solving the off-diagonal equations and reducing
the system to the four diagonal equations is equivalent to diagonalizing the system
Hamiltonian. These equations can of course be solved numerically quite easily, and
the current can again be calculated in two ways:

IL = ρoo (Γoa + Γob ) − (ρaa Γao + ρbb Γbo )


(3.59)
IR = ρcc Γoc − ρoo Γco .

Sensible results however, are only found in the high bias voltage limit of one way
tunneling where the leftward direction are reduced to zero, Γ0a = Γ0b = Γc0 = 0,
while the right moving rates become Γa0 = Γb0 = Γ0c = Γ. In this limit, the results
of Emary can in fact be reproduced.
In conclusion, we will now summarize the results of this thesis and future directions
it could be taken.
4
Conclusions

In summary, this thesis has presented a novel modeling technique in the treatment of
electron transport through a triple quantum dot ring. The model was inspired by the
experiments of Gaudreau et al.[25, 26], from which we have explicitly used experimen-
tally measured parameters. This produced a regime in which strong electron-electron
interactions allowed only single electron transport, putting the system in a Coulomb
blockade regime. This can be dealt with easily with a master equation technique,
but is entirely neglected in the work of Delgado et al.[33]. Furthermore, experimental
data from NRC demanded a model in which inter-dot tunneling was much stronger
than lead-dot tunneling. Although Michaelis and Emary do use master equations
to model a similar triple dot ring, they completely neglect inter-dot tunneling while
treating the lead-dot tunneling. This results in a master equation that fails for low
bias voltages. Our model on the other hand uses a novel technique of perturba-
tively coupling the leads directly to the molecular eigenstates of the system, which
on the other hand does give reasonable results at low bias voltages. We can also
reproduce the dephasing jump terms as used by Emary et al.[41], via a perturbative
dot-bath coupling which again neglects inter-dot tunneling. But with the desire to
include strong inter-dot tunneling, these terms too were adjusted via a treatment in
the molecular eigenstate basis. Finally, we were able to find simple analytic solutions
for the zero-bias conductance in the regime that t0  t.

This approach does however require the analytic diagonalization of the system
Hamiltonian in order to be used. This can been done analytically in the illustrative
limit of t0  t for low bias voltages, as it is found that only the two lowest energy

74
75

eigenstates contribute. In this limit, the two left a, b dots can be diagonalized first
giving a bonding and anti-bonding pair of a, b states. In some cases we can then
neglect one of these states and couple the remaining left dot state to dot c and the
right. The analytic expressions found in this way give rise to the analogous picture in
which two dots are connected by two transport path with a relative phase difference.
In this case an effective coupling combining both possible paths can either go to zero if
the paths destructively interfere, or be doubled if they constructively interfere. When
the current is blocked by destructive interference of the possible tunneling paths,
we say the electron is trapped in a “dark state”. In the limit of no bath coupling,
this effective dot coupling is found to set the width of the resonance, even for low
bias. This makes the resonance width a measure of the inter-dot tunneling and a
signature of “dark state” physics, unaffected by temperature. We were also able to
see the effects of thermal bath coupling on the molecular eigenstates formed by the
diagonalization preformed. Strong thermal bath coupling was found to destroy the
“dark state” as expected. However, resonances sharper than temperature were found
to have a surprising robustness to this dephasing. In the limit that t0 = t however,
analytic expressions are not as easily available and it is not clear whether a perfect
dark state will still be formed.

Examining the Aharonov-Bohm oscillations across the sharp resonances found, we


did not find any cases of the phase flipping. This is due to the approximation that
t0  t. In a regime in which both ψ− and ψ+ states contribute on either side of the
resonance, the AB phase would in fact flip. This should be investigated further in
future work with this model. The AB oscillations in this limit did still show some
interesting behavior. Even in the strong bath coupling limit in which there is no sharp
resonance or “dark state” blocking, there are still AB oscillations. This is due to the
energy difference in the two lowest energy states being sensitive to the magnetic flux.
This causes the resonance to oscillate between short and fat or taller and narrower,
with a periodicity of Φ0 as would be expected. Although in the limit we have explored
the phase of the oscillations did not flip, the “dark state” did cause additional sharp
76 4 Conclusions

dips in the conductance. These sharp dips were found to double the characteristic
frequency of the AB oscillations, resulting in a Φ0 /2 periodicity. Further work with
this model could continue to explore these AB oscillations in the high bias regime as
well.
As our model is valid for both high and low bias voltages, we were able to discover
an new negative differential resistance mechanism not possible in double dot systems.
In this case we found that if the excited state ψ− was tuned to be a “dark state” it
would block the current, but not until a sufficiently high bias voltage is reach that an
electron can reach the trapping state. This effect was not present for reversed bias
voltages however, giving the high bias rectification behavior found by Michaelis et
al.[40]. We have thus corrected the explanation given by Michaelis, showing that in
fact both molecular states ψ+ and ψ− are necessary for the rectification mechanism.
This rectification is thus a direct manifestation of the asymmetry in the geometry of
the system and lead-dot coupling. With a firm understanding of the various effects
seen in triple dots not possible in double dots, future work should easily be gener-
alized to include multiple electrons in the system at a time. This would require a
more complete treatment including spin, but would certainly give rise to novel effects
potentially usefully in any sort of coherent qubit state manipulations.
Bibliography

[1] Kastner, M., Artificial Atoms, Phys. Today 4601, 24(1993).

[2] Gregor Hackenbroich, Phase coherent transmission through interacting mesoscopic


systems, arXiv:cond-mat/0006361v1, (2000).

[3] A. Yacoby, M. Heiblum, D. Mahalu, and H. Shtrikman, Coherence and Phase


Sensitive Measurements in a Quantum Dot. Phys. Rev. Lett., Vol. 74, No. 20
4047-4050 (1995).

[4] R. Schuster, E. Buks, M. Heiblum, D. Mahalu, V. Umansky, and H. Shtrikman,


Phase measurement in a quantum dot via a double-slit interference experiment.
Nature, 1997 vol. 385 (6615) pp. 417-420.

[5] E. Buks, R. Schuster, M. Heiblum, D. Mahalu, and V. Umansky, Dephasing in


electron interference by a ’which-path’ detector. Nature (London), 1998 vol. 391
(6670) pp. 871-874.

[6] J. R. Petta , A. C. Johnson, C. M. Marcus, M. P. Hanson, and A. C. Gossard,


Manipulation of a Single Charge in a Double Quantum Dot, Phys. Rev. Lett. 93,
186802 (2004).

[7] M. Bayer, P. Hawrylak, K. Hinzer, S. Fafard, M. Korkusinski, Z. R. Wasilewski, O.


Stern, A. Forchel, Coupling and Entangling of Quantum Dot States in Quantum
Dot Molecules, Science 291, 451(2001).

[8] Alexander W. Holleitner, Robert H. Blick, Andreas K. Huttel, Karl Eberl, Jorg
P. Kotthaus, Probing and Controlling the Bonds of an Artificial Molecule, Science
297,70(2002).

77
78 BIBLIOGRAPHY

[9] A. K. Httel, S. Ludwig, H. Lorenz, K. Eberl, and J. P. Kotthaus, Direct control


of the tunnel splitting in a one-electron double quantum dot, Phys. Rev. B 72,
081310(2005).

[10] W.G van der Wiel, S. De Franceschi, J.M. Elzerman, T. Fujisawa, S. Tarucha,
L.P. Kouwenhoven, Electron transport through double quantum dots Reviews of
Modern Physics, Vol. 75 (1) pp. 1-22 (2002).

[11] Averin, D.V., and K.K. Likharev,Coulomb Blockade of Single-Electron Tunnel-


ing, and Coherent Oscillations in Small Tunnel Junctions J. Low Temp. Phys.
62, 345, 1986.

[12] Grabert,H., and M.H.Devoret, 1992, Eds., Single Charge Tunneling: Coulomb
Blockade Phenomena in Nanostructures,(Plenum Press and NATO Scientific Af-
fairs Division, NY/Lond.)

[13] Y.Aharanov and D. Bohm, Significance of Electromagnetic Potentials in The


Quantum Theory, Phys. Rev. 115, 485 (1959).

[14] L. DiCarlo, H. J. Lynch, A. C. Johnson, L. I. Childress, K. Crockett, C. M.


Marcus, M. P. Hanson, and A. C. Gossard, Differential Charge Sensing and Charge
Delocalization in a Tunable Double Quantum Dot, Phys. Rev. Lett. 99, 226801
(2004).

[15] M. Pioro-Ladrire, M. R. Abolfath, P. Zawadzki, J. Lapointe, S. A. Studenikin,


A. S. Sachrajda, and P. Hawrylak, Charge sensing of an articial H2+ molecule in
lateral quantum dots, Phys. Rev. B 72, 125307(2005).

[16] M. A. Topinka, B. J. LeRoy, S. E. J. Shaw, E. J. Heller, R. M. Westervelt, K.


D. Maranowski, A. C. Gossard, Imaging Coherent Electron Flow from a Quantum
Point Contactce, Science 298, 2323(2000).

[17] J. M. Elzerman, R. Hanson, J. S. Greidanus, L. H. Willems van Beveren, S. De


Franceschi, L. M. K. Vandersypen, S. Tarucha, and L. P. Kouwenhoven, Few-
BIBLIOGRAPHY 79

electron quantum dot circuit with integrated charge read out, Phys. Rev. B 67,
161308 (2003).

[18] Loss, D., and D.P.DiVincenzo,Quantum Computation with Quantum


Dots,1998,Phys. Rev. A 57, 120.

[19] Burkard, g., Loss, D., and D.P.DiVincenzo,Coupled Quantum Dots as Quantum
Gates, Phys. Rev. B 59, 2070 (1999).

[20] Zandardi,P. and F.Rossi, Quantum Information in Semiconductors: Noiseless


Encoding in a Quantum-Dot Array, 1998,Phys.Rev.Lett. 81,4572.

[21] A. Barenco, D. Deutsch, and A. Ekert, Conditional Quantum Dynamics and


Logic Gates, Phys. Rev. Lett. 74, 4083 (1995).

[22] J. Gorman, D. G. Hasko, and D. A. Williams, Charge-Qubit Operation of an


Isolated Double Quantum Dot, Phys. Rev. Lett. 95, 090502 (2005).

[23] Brum, J.A., and P. Hawrylak, Superlattices Microstructures 22,431.

[24] T. Fujisawa, T. Hayashi, and Y.Hirayama, Controlled decoherence of a charge


qubit in a double quantum dot, J. Vac. Sci. Technol. B 22(4), 2035 (2004).

[25] L. Gaudreau, A.S. Sachradja, S. Studenikin, P. Zawadzki, A. Kam, and J. La-


pointe, Coherent Transport Through a Quadruple Point in a Few Electron Triple
Dot. Arxiv preprint cond-mat/0611488 (2006).

[26] L. Gaudreau, S. A. Studenikin, A. S. Sachrajda, P. Zawadzki, A. Kam, J. La-


pointe, M. Korkusinski, and P. Hawrylak, Stability Diagram of a Few-Electron
Triple Dot. Phys. Rev. Lett. (2006).

[27] K. Grove-Rasmussen, H.I. Jorgensen, T. Hayashi, P.E. Lindelof, T. Fujisawa, A


triple quantum dot in a single-wall carbon nanotube. Nano Letters, Vol. 8, No. 4
1055-1060 (2008).
80 BIBLIOGRAPHY

[28] M.C. Rogge and R.J. Haug, Two Path Transport Measurements on a Triple
Quantum Dot. Arxiv preprint arXiv:0707.2058 (2007).

[29] M.C. Rogge and R.J. Haug, Non-invasive detection of molecular bonds in quan-
tum dots, arXiv:0807.5095v1(2008).

[30] L. G. Mourokh, and A. Y. Smirnov, Negative differential conductivity and popu-


lation inversion in the double-dot system connected to three terminals, Phys. Rev.
B 72, 033310 (2005).

[31] K. Ono, D. G. Austing, Y. Tokura, S. Tarucha, Current Rectication by


Pauli Exclusion in a Weakly Coupled Double Quantum Dot System, Science
297,1313(2002).

[32] T. Brandes and B. Kramer, Spontaneous Emmision of Phonons by Coupled


Quantum Dots. Phys. Rev. Lett. Vol. 83 (15) pp. 3021-3024 (1999).

[33] F. Delgado, and P. Hawrylak, Theory of electronic transport through a triple


quantum dot in the presence of magnetic eld, arXiv:0712.0624v (2008).

[34] Many-Body Quantum Theory in Condensed Matter Physics - An Introduction,


By H. Bruus and K. Flensburg. Published by Oxford University Press, 2004.

[35] Quantum Optics, By M. O. Scully, M. S. Zubairy. Published by Cambridge Uni-


versity Press, 1997.

[36] Nazarov,Yu.V, 1993, Physica B 189,57.

[37] Stoof, T.H.,1997, Ph.D. Thesis (Delft University f Technology, The Netherlands)

[38] J. N. Pederson, B. Lassen, A. Wacker, and M. H. Hettler, Coherent transport


through an interacting double quantum dot: Beyond sequential tunneling. Phys.
Rev. B 75, 235314 (2007).
BIBLIOGRAPHY 81

[39] I. Djuric, B.Dong, and H.L.Cui, Theoretical investigations for shot noise in cor-
related resonant tunneling through a quantum coupled system J.Appl. Phys. 99,
063710, 2006.

[40] B. Michaelis, C. Emary, and C. W. J. Beenakker, All-electronic coherent popu-


lation trapping in quantum dots. Europhys. Lett. 73(5), pp. 677-683 (2006).

[41] Clive Emary, Dark States in the magnetotransport through triple quantum dots.
Phys. Rev. B 76, 245319 (2007).

[42] G. Alzetta, A. Gozzini, L. Moi, and G. Orriols, Nuovo Cimento B 36, 5(1976).

[43] R.M. Whitley and C.R. Stroud, Double Optical Resonance, Phys. Rev. A 14,
1498(1976).

[44] Ji Yang, Y. Chung, D. Sprinzak, M. Heiblum, D. Mahalm, and H. Shtrikman,


An Electronic Mach-Zehnder Interferometer, Nature, 422, 415 (2003).

[45] D. Yu. Sharvin, and Yu. V. Sharvin, Pis’ma Zh.Teor.Ehsp.Fiz. 34, 285 (1981).

[46] V. Chandrasekhar, M.J. Rooks, S. Wind, and P.E. Prober,Observation of


Aharonov-Bohm Electron Interference Effects with Periods h/e and h/2e in In-
dividual Micron-Size, Normal Metal Rings Phys. Rev. Lett. 55,1610(1985).

[47] S. A. Gurvitz and Ya. S. Prager, Microscopic derivation of rate equations for
quantum transport, Phys. Rev. B 55, 15932(1996).

[48] U. Harbola, M. Esposito, and S. Mukamel, Quantum master equation for electron
transport through quantum dots and single molecules, Phys. Rev. B 74, 235309
(2006).

Вам также может понравиться