Quantum Gates in Photonic Waveguides

UNIVERSITY OF COPENHAGEN
FACULTY OF SCIENCE
Master's Thesis
Kasper Ejdal Lund
Quantum Gates in Photonic Waveguides
Academic advisor: Anders S. Sørensen
Submitted: 30/09/2016
"Imagination is the Essence of Discovery"
Abstract
The field of quantum information is concerned with investigating the possibilities of using quan-
tum phenomena for computational and communication purposes by using the sometimes counter
intuitive behaviour of system described by quantum mechanics. This is a tremendous challenge,
primarily due to the quantum systems’ tendency to interact with the environment. Shielding
these systems is thus an important part of the practical implementation of quantum technologies
and this remains an active field of research within the field.
Quantum information systems are described by the qubit states |0i and |1i and can be formed in
a linear superposition of these states, which is one of the main reasons for the strength of quan-
tum information protocols. An important part in implementing quantum information protocols
is to implement a two qubit gate, which is a controlled evolution of two coupled quantum states
that are coupled together. This coupling is often mediated by some other physical entity (such as
a propagating photon in a waveguide). One such gate in the controlled phase gate, which imprint
a change of phase of e−iπ onto the state where both qubits are in |1i but otherwise leaves the
system unchanged. This thesis is concerned with investigating the possibilities of implementing a
controlled phase gate using self-assembled quantum dot embedded in a nanophotonic waveguide.
This thesis provides a proposal to implement a controlled phase gate by using a strong in-
teraction between three quantum dots. Two of these dots constitute the qubit, while the third
auxiliary emitter is used to herald the gate. It is possible to control the dynamical change of
phase of the qubits by illuminating the auxiliary quantum dot with a weak laser field. This will
project the auxiliary quantum dot into a definite state, due to the nature of the dynamics. The
primary source of error will project the auxiliary into another state; it is thus possible to check
whether the gate has succeeded ot not by measuring the state of the auxiliary emitter, giving
the gate an integrated error detection.
This thesis will report in a method to obtain a unity fidelity for this controlled phase gate and
will report on the behaviour of the success probability. In addition, errors induced by variations
in the physical parameters of the quantum dots are discussed and methods of mitigating these
errors are presented.
i
ii
Resumé
Mulighederne for at udnytte de til tider besynderlige kvantemekaniske fænomener til at udføre
beregningsmæssige protokoller og for at implementere sikre kommunikationslinjer er nogle er
de store mål inden for kvanteinformation. Dette er en besværlig opgave, da kvantemekaniske
systemer har en tendens til at koble til diverse omgivelser. Det er derfor nødvendigt at beskytte
disse systemer, og dette er en af hovedopgaverne inden for feltet.
Inden for kvanteinformation bliver fysiske tilstande beskrevet af qubit tilstandene |0i og |1i. Dis-
se systemer kan endda være i de såkaldte superpositioner, hvilket er en af grundene til styrken
ved kvantemekaniske informationsprotokoller. Implementeringen af en to-qubit gate er en vigtig
del ved udførslen af disse protokoller og er typisk udført ved at udnytte anden fysisk proces (for
eksempel en photon, der bevæger sig fra en kvantemekanisk enhed til en anden) til at koble to
systemer sammen. Disse gates er en kontrolleret tidlig udvikling af de kvantemekaniske syste-
mer. Den kontrollerede fase gate er en vigtig to-qubit gate, som kan give en ændring af fasen
af den bestemte tilstand, hvor begge qubits er i |1i, og som ellers ikke ændrer systemet. Denne
afhandling undersøger mulighederne for at implementere denne slags gate i et system bestående
af tre groede kvantedots i en nanofotonisk bølgeleder.
Denne afhandling giver et forslag til udførslen af en kontrolleret fase gate ved at bruge en stærk
kobling mellem tre kvantedots. To af disse dots bliver bruge til at lave de to qubits, mens at
den trejde bruges til at styre gaten. Det er muligt at styre udviklingen af den tidligt udviklende
ændring i fasen af de to qubits ved at belyse den såkaldte "hjælpemotor"med en svag laser,
hvorved hjælpemotoren bliver projekteret ind i en bestemt tilstand. De primære fejlkilder for
gaten vil derimod projektere hjælpemotoren hen i en anden tilstand. Det er dermed muligt at
tjekke, om hvorvidt gaten har givet det korrekte output eller ej ved at måle på tilstanden af
hjælpemotoren. Det giver dette forslag en integreret fejlsøgning.
Denne afhandling rappoterer en metode til at opnå en fidelity på 1 for den kontrollerede fase
gate og vil desuden diskutere den resulterende opførsel af sandsynligheden for succes. Diverse
variationer i de fysiske parameter kan opstå, når kvantedotsne bliver groet, og disse variationer
kan have indflydelse på præstationen af gaten. Denne afhandling vil derfor også undersøge hvor-
dan fidelitien of sandsynligheden for succesafhænger af disse variationer, og om der er muligt at
minimere eller fjerne disse fejl.
iii
Acknowledgements
I would first of all like to thank my supervisor Anders S. Sørensen for his invaluable help and
guidance throughout this project, without which this project would not have been possible.
To this end, I would also like to thank Sahand Mahmoodian for the plentiful and thoughtful
discussions during this last year and for being willing to discuss my ideas and findings. Both
have helped to steer me in the proper direction and have always been ready to give their helpful
input.
In addition, I would like to thanks the rest of the Theoretical Quantum Optics Group for letting
me tag along for the last year and it has been fascinating to listen to their presentations at the
group meetings. Special thanks go to Luca, Sasha and Julia for their help in preparation to, and
during, the DHL relay.
It has been fascinating to be part of the Quantum Optics and to get a taste of working with an
active research field. I have enjoyed the friendly atmosphere that exists amongst the different
groups, whether it has been through the cosy coffee breaks in the lounge, the friendly rivalry
that exists for the DHL relay or the Christmas dinner.
I also want to thank the USG cross fit group for all the great sessions in the gym, which has
helped me stay fit and focused for the duration of this thesis.
Lastly, I want to thank my parents and family for the support throughout the year.
iv
Contents
1 Introduction 1
2 Quantum Mechanics & Quantum Information 5

2.1 Basics of Quantum Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.1 States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.2 Observables and Measurement . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.3 Time Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.4 Non-Unitary Time Evolution & Quantum Jump Formalism . . . . . . . . 8
2.2 Quantum Computation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 DiVicenzo’s Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.2 Qubits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.3 Universal Quantum Gates . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.4 Initialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.5 A Qubit Specific Measurement Capability . . . . . . . . . . . . . . . . . . 16
2.2.6 Long Decoherence Times . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.7 Transmission of Flying Qubtis & the Ability to Interconvert Stationary
and Flying Qubits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.8 Quantum Error Correction . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3 Basics of Nanophotonic Structures 19

3.1 Nanophotonic Waveguides . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.1.1 Structural Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.1.2 Optical Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2 Quantum Dots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2.1 Basic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2.2 Effective Mass Theory & the Transition Matrix Element . . . . . . . . . . 26
3.2.3 Optical Properties of Quantum Dots . . . . . . . . . . . . . . . . . . . . . 28
3.3 Initialization of Qubit States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4 Decay Properties of Coupled Quantum Emitters 35

4.1 Defining the System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2 Decay Dynamics for Three Two-Level Quantum Emitters . . . . . . . . . . . . . 36
4.2.1 General Formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2.2 Dynamics in the Interaction Picture . . . . . . . . . . . . . . . . . . . . . 38
4.2.3 Wigner-Weisskopf Theory for a W1 PhC Waveguide . . . . . . . . . . . . 39
4.2.4 Superradiance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.3 Full Time Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
v
5 Ideal CZ-Gate 47
5.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.2 Ideal CZ Gate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.2.1 Effective Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.2.2 Solving the Effective Master Equation . . . . . . . . . . . . . . . . . . . . 50
5.2.3 Effective Phase Evolution for C 1 . . . . . . . . . . . . . . . . . . . . . 52
5.3 Criteria for the Driving Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.3.1 Satisfying the Adiabatic Condition . . . . . . . . . . . . . . . . . . . . . . 56
5.3.2 Effects of Dephasing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
6 Non-Ideal CZ Gate 61
6.1 Variable Qubit Dipole Moment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.2 Variable Qubit Dipole Moment & Position . . . . . . . . . . . . . . . . . . . . . . 69
6.3 Conditions for the Adiabatic Treatment and Dephasing . . . . . . . . . . . . . . . 75
7 Conclusion and Outlook 77

7.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
7.2 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
A Effective Operator Formalism 79

A.1 Derivation of the Effective Operators . . . . . . . . . . . . . . . . . . . . . . . . . 80
A.1.1 Projection Operator Formalism . . . . . . . . . . . . . . . . . . . . . . . . 80
A.1.2 Non-Hermitian Time Evolution in the Quantum Jump Picture . . . . . . 80
A.1.3 Perturbation Theory in the Interaction Picture . . . . . . . . . . . . . . . 80
A.1.4 Adiabatic Elimination of the Excited States . . . . . . . . . . . . . . . . . 83
B Calculation of Effective Operators 91

B.1 Two qubits in the ground state . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
B.1.1 One qubit in the ground state . . . . . . . . . . . . . . . . . . . . . . . . . 93
B.1.2 No qubits in the ground state . . . . . . . . . . . . . . . . . . . . . . . . . 94
B.1.3 Inclusion of δqj . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
B.2 Variable Qubit Dipole Moments & Positions . . . . . . . . . . . . . . . . . . . . . 96
C Additional Contour Plots 97
D Solving The Effective Master Equation 103

D.1 The Effective Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Bibliography 107
vi
List of Figures
3.1 Dispersion Diagram & Frequency Dependence of the LDOS in a PhC Waveguide 20
3.2 Spatial Suppression of the Local Density of States . . . . . . . . . . . . . . . . . . 21
3.3 Enhancement of the Purcell and β Factor in a W1 Waveguide . . . . . . . . . . . 24
3.4 Transition Rules for Excitonic States . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.5 Decay Dynamics of the Bright and Dark Exciton . . . . . . . . . . . . . . . . . . 29
3.6 Phonon Induced Broadening of the QD Emission Spectrum . . . . . . . . . . . . 31
3.7 Level Scheme of the Trion Used for Initialisation of the Qubit . . . . . . . . . . . 32
4.1 Level Structure of the Auxiliary and Qubit Quantum Dots . . . . . . . . . . . . . 37
5.1 Ideal Probability & Fidelity of a CZ Gate . . . . . . . . . . . . . . . . . . . . . . 55

5.2 Effect of Dephasing on the Optimised Fidelity . . . . . . . . . . . . . . . . . . . . 58
6.1 Optimised Fidelity and Success Probability for Variable Dipole Moments . . . . . 63
6.2 Success Probability & Fidelity Dependence of Equal Coupling Constant . . . . . 64
6.3 Success Probability & Fidelity Dependence of Different Dipole Moments . . . . . 66
6.4 Success Probability & Fidelity Dependence of Different Dipole Moments . . . . . 68
6.5 Spatial Profile for the Success Probability & Fidelity for various Detunings . . . . 70
6.6 Success Probability on Qubit Displacement . . . . . . . . . . . . . . . . . . . . . 72
6.7 Fidelity Dependence on Qubit Displacement . . . . . . . . . . . . . . . . . . . . . 74
6.8 Effects of Dephasing for the Non-Ideal Fidelity . . . . . . . . . . . . . . . . . . . 76
C.1 Contourplot for the Sucess Probability for High Dipole Moments . . . . . . . . . 98
C.2 Contourplot for the Fidelity for High Dipole Moments . . . . . . . . . . . . . . . 98
C.3 Contourplot for the Success Probability for Low Dipole Moments . . . . . . . . . 99
C.4 Contourplot for the Fidelity for Low Dipole Moments . . . . . . . . . . . . . . . . 99
C.5 First Contourplot for the Success Probability for Mixed Dipole Moments . . . . . 100
C.6 First Contourplot for the Fidelity for Mixed Dipole Moments . . . . . . . . . . . 100
C.7 Second Contourplot for the Success Probability for Mixed Dipole Moments . . . . 101
C.8 Second Contourplot for the Fidelity for Mixed Dipole Moments . . . . . . . . . . 101
C.9 Contourplot for the Success Probability for an Alternative set of Detunings . . . 102
C.10 Contourplot for the fidelity for an Alternative set of Detunings . . . . . . . . . . 102
vii
viii
Chapter 1
Introduction
The processing, storage and transmission of information have been a cornerstone of society
throughout human history. This can be traced as far back as to the early Mesopotamian civ-
ilizations (most notably the Sumerian civilization) at 3600 BC who developed the first writing
system to keep track of manufacturing and agricultural goods and for use in trading [19]. The
advent of computers (analogue versions from the early 20th century and digital in the post
war period) and modern cryptography heralded the modern information age and has allowed
several fields to rapidly advance. Science has greatly benefited from using numerical simulation
on digital computers, whereas society at large depend on modern communication methods and
security protocols (e.g. the RSA protocol).
All of this has been accomplished by classical information theory which can be readily im-
plemented using, in principle, simple physical systems (such as a voltage running through a
capacitor for a digital computer) but these implementations face several challenges. Several
problems, such as the travelling salesman and the factorization into prime numbers, are classi-
fied as computational hard. This means that the required time to solve these problems grows
exponentially, whereas the solution for easy problems grow polynomial in time. The aforemen-
tioned RSA protocol is heavily dependent on this property. Note that is has not been proved
that a classical and polynomial-time based solution for these problems do not exist.
It is possible to define a classical and completely secure encryption method. This is called the
one-time pad method: Alice wants to send a message to Bob. To do this Alice writes her message
as a sequence of binary numbers m and generates another sequence of binary number k through
an ideal random number generator. This second sequence is called the key and is added (modulo
two) to the message to scramble the information: s = m ⊕ k. Alice then sends the message to
Bob who can decrypt the message by adding the key again: m = s ⊕ k. This has proven to
be completely secure [12] but there exist certain constraints. Firstly, each key can only be used
once, otherwise it is possible for eavesdroppers to extract information from various messages.
This limits the use of this scheme for only extremely important uses. Secondly, Alive and Bob
need to have a method for sharing the secret key. The only classically key distribution method
is a trusted man with a suitcase, which may not be feasible.
Yet another example is simulation of quantum systems. Nature is, at the smallest length scales,
governed by quantum mechanics. This includes collective effects like magnetism and supercon-
ductivity. Numerical simulation has proven to be an important tool in examining classical and
small quantum systems but has difficulty in treating large quantum systems. This is due to the
exponential increase in the number of parameters with the number of particles.
1
Lastly, there is a technological problem: Moors’ law has a limit. The basic building block of
any integrated circuit is the transistor. These transistors can be made on smaller dimensions as
technology progresses, which allows integrated chips to include more and more transitions per
square inch. This increase in number of transistors was first observed in 1965 by Intel co-founder
Gordon Moore, who observed an exponential increase in the number of transistors per square
inch on integrated circuits [15]. This poses a problem when these small structures begin to be
governed by quantum mechanics. One such effect is tunneling, which allows particle to penetrate
classically impassable potential barriers.
Classical computers operate with discrete bits, which are represented as the binary numbers 0
and 1. Bits can be physically implemented by various means and the most common example is
the presence (the number 1) or absence (the number 0) of an electron in a circuit. There are
fewer electrons per square inch of the chips as the components become smaller, which causes
the probabilistic effect of tunneling to become apparent through its ability to change a bit from
0 to 1 (or vice versa). This is a non-deterministic change to the logical state of an operation,
which can lead to computational errors, as classical algorithms assumes that classical bits are
unchanged unless explicitly operated on.
This highlights the need to be able to control these quantum effects in order to mitigate the
induced errors. Alternatively, it may be possible to use quantum effects for advanced problems.
In fact, several advances has already been made in the field of Quantum Information and there
exists theoretical proposals to solve classically hard problems(e.g. Shor’s factorisation algorithm,
quantum key distribution and quantum simulation [10].
2
This thesis focuses on a specific and crucial part of implementing a quantum computer: a
heralded two-qubit controlled phase gate, where the qubits are represented by ground states in
self-assembled quantum dots embedded in a nano-photonic wave guide. This is based on a recent
proposal by Borregaard et al. [3], where the gate was implemented using three atoms in a cavity.
This resulted in a CZ-gate with the possibility of attaining a unity fidelity. In addition, the gate
has an integrated error detection, as a successful operation will project an auxiliary atom into a
definite state, will a failed operation projects it into another state. The success, or failure, of the
gate can be readily determined by measuring the state of the auxiliary atom without disturbing
the state of the qubit. This work examines the possibilities of implementing this scheme in a
system comprised of three self-assembled quantum dots embedded in a nanophotic crystal.
The main contributions of this work are:
• A theoretical description of the decay dynamics for several emitters coupled in a single-
mode wave guide (Chapter (4)).
• A detailed calculation of the derivation of the success probability and the fidelity a heralded
two-qubit controlled phase gate (Chapter (5)).
• The effects of changing various physical parameters, namely the dipole moments and po-
sitions of the qubits, on the success probability and the fidelity (Chapter (6)).
The outline of the thesis is
Chapter 1: (This chapter). This presents the motivation and objectives of this thesis.
Chapter 2: Provides a brief introduction to basic concepts of quantum mechanics and

quantum information.
Chapter 3: Contains an introduction to the physical system under consideration in thus

thesis. These systems are self-assembled quantum dots and photonic crystals.
Chapter 4: This gives a derivation of the decay dynamics of three coupled quantum
emitters embedded in a photonic crystal waveguide.
Chapter 5: The dynamics of the three coupled quantum dots are derived and are used
to implement a controlled phase gate. In addition, the success probability and fidelity is
derived for the ideal case.
Chapter 6: The effects of changing the physical parameters of the qubit states are ex-
amined and methods for mitigating the detrimental effects on the fidelity of the gate are
provided.
Chapter 7: Concludes on this work and provides an outlook for further research on this
topic.
3
4
Chapter 2
Quantum Mechanics & Quantum

Information
A system described by quantum mechanics can exhibit properties otherwise thought impossible
by the laws of classical physics. This includes effects like quantum parallelism and entanglement
and may be useful for information and computation purposes ([11]). This section will provide
a basic introduction to concepts from quantum mechanics (Section (2.1)) and to the field of
quantum information (Section (2.2)). This section is primarily based on [17] unless otherwise
noted.
2.1 Basics of Quantum Mechanics

2.1.1 States
The complete description of a physical system is contained in a state vector |ψi defined in the
Hilbert space H. This is a vector space over the complex numbers C; it maps the inner product
hφ|ψi between two states |ψi and |φi to C with the following properties:
• Positivity: hφ|ψi ≥ 0.
• Linearity: hφ| (a |ψ2 i + b |ψ2 i) = a hφ|ψ1 i + b hφ|ψ2 i.
• Skew symmetry: hφ|ψi = hψ|φi∗ .
Additionally, the Hilbert space is complete in its norm: kψk = hψ|ψi1/2 . Any state can be
expanded in the orthonormal basis states |{n}i spanned by the Hilbert space:
X
|ψi = |ni hn|ψi , (2.1)
n
where hn|ψi = cn is the expansion coefficient cn for the projection of |ψi along |ni. A normalised
state |ψ̃i can be formed from |ψi by
1
|ψ̃i = p |ψi , (2.2)
hψ|ψi
such that hψ̃|ψ̃i = 1. The norm of |ψi can be found from Eq. (2.1):
X X
hψ|ψi = |hn|ψi| = |cn |2 , (2.3)
n n
5
orthonormal basis with the inner product hn|n0 i = δn,n0 . If |ψi is a
using that |{n}i is an P
normalised state, then n |cn |2 = 1. This result is important for the statistical interpretation
of quantum mechanics introduced below.
2.1.2 Observables and Measurement

An observable is a physical entity that can be measured and is represented by a self-adjoint (this
is also denoted as hermitian) operator. An operator M̂ is a linear map that takes vectors to
vectors:
M̂ : |ψi → M̂ |ψi , M̂ (a |ψi + b |φi) = aM̂ |ψi + bM̂ |φi . (2.4)
The adjoint M̂ † of an operator is defined by:
hφ|M̂ ψi = hM̂ † φ|ψi , (2.5)
for all vectors |ψi , |φi. A hermitian operator satisfies M̂ = M̂ † , leading to the following property
∗
for M̂ : hφ|M̂ |ψi = hψ|M̂ |φi , forming a real vector space. A measurement is a process where
information about the system is gained; more specifically, information about the observable M̂
is acquired and is represented by the real numbered expectation value:
hM̂ i = hψ|M̂ |ψi . (2.6)
Any hermitian operator defined in A Hilbert space is diagonalisable, where its eigenstates form
a complete orthonormal basis in H. The operator can then be written in terms of the basis
states |{n}i and its eigenvalues Mn by:
X
M̂ = Mn |ni hn| , (2.7)
n
where M̂ |ni = Mn |ni. Writing |ψi in terms of these basis states by Eq. (2.1) causes the
expectation value to take the following form:
X
hM̂ i = Mn |cn |2 , (2.8)
n
where |cn |2 = pn is the probability to obtain the n’th eigenstate of M̂ upon repeating the same
measurement on a number of identically prepared systems. This statistical interpretation carries
over to the state written in Eq. (2.1), highlighting one of the main differences between classical
and quantum physics. According to the complimentary principle, the actual state of the system
is in a superposition of the basis states |{n}i with a probability of measuring it in the n’th
state of |cn |2 . Only after a measurement can the state be said to be in a definite state, as it is
projected into:
Mn |ni
|ψiAfter = . (2.9)
kMn |ni k
Repeated measurements will always return this state with probability 1. This interpretation of
a quantum state provide the starting point of realising the strengths of quantum information
processes and will be discussed in further detail in Section (2.2.2).
6
2.1.3 Time Evolution
The time evolution of physical systems remains important in the quantum mechanical formalism.
The description of dynamics can be particularly involved when both the quantum states, the
operators and the Hamiltonian depend on time; a variety of methods has thus been developed
to solve this problem. The main methods (or pictures, as they are often referred to) are briefly
discussed here. Note that the following assumes that there is no source of loss in the system;
section Eq. (2.1.4) introduces the necessary corrections in the the presence of loss.
The Schrödinger Picture

The treatment of time evolution is simple if the Hamiltonian and operators are independent of
time; all time dependence is thus only containing the in state vectors |ψ (t)i. The time evolution
is then governed by the Schrödinger equation:
d
i~ |ψ (t)i = Ĥ |(t)i , (2.10)
dt
with the formal solution:
|ψ (t)i = e−iĤt/~ |ψ (t = 0)i , (2.11)
where Û (t) = e−iĤt/~ is a unitary operator that evolves the state |ψ (t = 0)i into |ψ (t)i. It may
often be more useful to collect all time evolution in the operators while leaving the state vectors
stationary in time. This can be done in the Heisenberg picture;
The Heisenberg Picture

As stated above, the idea of the Heisenberg picture is to have all time evolution contained in
the operators of the system, while the Hamiltonian and state vectors are independent in time.
This can be accomplished by considering the expectation value of some operator M̂ at time t
(the factor of ~ has been omitted from the exponential function for brevity in the following):
hψ (t) |M̂ |ψ (t)i = ψ0 eiHt M̂ e−iHt ψ0 = hψ0 |M̂ (t) |ψ0 i ,

(2.12)
where |ψ0 i = |ψ (t = 0)i and M̂ (t) = eiHt M̂ e−iHt = Ô† (t) M̂ Ô (t). Ô (t) = e−iĤt is a unitary
operator that transforms the ssytem from the Schrödinger picture to the Heisenberg picture.
Note that the operator M̂ (t) is not the same operator as M̂ in this formalism. The relevant
equation of motion for M̂ (t) can be straightforwardly derived:
h i
Ṁ (t) = i Ĥ, M̂ (t) + ∂t M̂ (t) , (2.13)
resulting in the Heisenberg equation of motion for M̂ (t). Both this and the Schrödinger picture
require that the Hamiltonian is independent of time, otherwise the interaction picture has to be
used.
The Interaction Picture

This representation is introduced in order to treat the dynamics of a quantum system that is
represented by a time dependent Hamiltonian. This Hamiltonian is split into two parts: the first
is a time independent part Ĥ0 that is exactly solvable, while the second part is a time dependent
perturbative interaction V̂ (t):
Ĥ (t) = Ĥ0 + V̂ (t) . (2.14)
7
The trivial time evolution from Ĥ0 can be removed from the dynamics by only including this
part of the Hamiltonian in the unitary transformation: Ô (t) = e−iĤ0 t . This puts the time
dependence into the operators and the states:
|ψ̃ (t)i = eiĤ0 t |ψi (2.15)

M̃ (t) = eiĤ0 t M̂ e−iĤ0 t . (2.16)
The dynamics of the system are described by equations of motion for the state vectors, operators
and the density matrix:
d
i |ψ̃ (t)i = Ṽ (t) |ψ̃ (t)i (2.17)
dt
d h i
i M̃ (t) = M̃ (t) , Ĥ0 (2.18)
dt
d h i
i ρ̃ (t) = Ṽ (t) , ρ̃ (t) , (2.19)
dt
where the density matrix ρ will be introduced in Section (2.1.4). The perturbative part of the
Hamiltonian is the only reference to time in this Schrödinger equation. The time evolution of
the state |ψ̃ (t)i can thus be described by a unitary operator Û (t, t0 ):
|ψ̃ (t)i = Û t, t0 |ψ̃ t0 i ,

(2.20)
This operator is in general given by [4]
Zt
  
h i
Û t, t0 = Tt exp −i dt00 Ṽ t00  ,

(2.21)
t0
where Tt is a time ordering operator. The properties of, and relation to quantum information
concepts (i.e. quantum gates), of these unitary operators are discussed in Section (2.2.3).
2.1.4 Non-Unitary Time Evolution & Quantum Jump Formalism

The description of time evolution of quantum systems presented in Section (2.2.6) is only valid
no dissipative effects are at play between the quantum system and the outlying environment. In
this case the time evolution is governed by the master equation
h i
ρ̇ = −i Ĥ, ρ + L [ρ] , (2.22)
where ρ = |ψi hψ| is the density matrix for a pure state, Ĥ is the Hamiltonian governing the
unitary evolution and L [ρ] is the Lindblad superoperator describing the dissipative processes.
This is given by:
1 X †
L [ρ] = − L̂k L̂k ρ + ρL̂†k L̂k − 2L̂k ρL̂† , (2.23)
2
k
where the sum runs over the possible decay channels and the Lindblad operator L̂k is an operator
describing this process. As an example, a non-radiative decay √ from the excited state |ei in a
quantum dot to some ground state |0i can be written as L̂γ 0 = γ 0 |0i he|. It can be advantageous
to rearrange the terms in Eq. (2.23) into:
h i X
ρ̇ = −i ĤNH , ρ + L̂k ρL̂†k , (2.24)
k
8
where the effective Hamiltonian ĤNH is given by the original Hamiltonian with the no-jump
terms added to it:
iX †
ĤNH = Ĥ − L̂k L̂k . (2.25)
2
k
This non-hermitian Hamiltonian can be evolved directly using the Schrödinger equation, as long
as |ψi is a pure state. This has the formal solution:
∂
i |ψ (t)i = ĤNH |ψ (t)i (2.26)
∂t
− 12 L̂†k L̂k
P
|ψ (t)i = e−iĤNH t |ψ (0)i = e−iĤt e k |ψ (0)i , (2.27)
where the no-jump terms gives rise to an exponential decay of the states and assuming that
Ĥ commutes with the Lindblad operators. The effects of the jump term 2 L̂k ρL̂†k can be in-
P
k
cluded in the stochastic wavefunction approach [6] by projecting them onto |ψi at discrete times.
Solving the master equation in Eq. (2.22) gives the complete description of the dynamics of
an open quantum system. This method becomes impractical at best for large systems as open
system involve both unitary and dissipative evolution. Various effects often cause the system to
experience different time scales, which makes it possible to divide the Hilbert space into two sep-
arate subspaces; one for the rapidly evolving excited states and one for the (meta)stable ground
states. It is then advantageous to eliminate the excited states to get a simpler description of the
ground states.
The standard method is to adiabatically eliminate the excited states, provided that the time
scale of the ground states is significantly slower than that for the excited states. This method
can become rather impractical to carry out if the size of the system is large due to the large
number of elements in the density matrix. Splitting the Hamiltonian governing the system and
the interactions into subspaces for the ground states and excited states makes it possible to
obtain an effective master equation [18] involving only the dynamics of the ground states:
h i X † 1 † †
k k k k k k
ρ̇ = −i Ĥeff , ρ + L̂eff ρ L̂eff − L̂eff L̂eff ρ − ρ L̂eff L̂eff , (2.28)
2
k
with the effective Hamiltonian and Lindblad operator:

†
1 −1

−1
Ĥeff = − V̂− ĤNH + ĤNH V̂+ + Ĥg (2.29)
2
−1
L̂keff = L̂k ĤNH V+ , (2.30)
where V+ (V− ) is the perturbative (de)excitation of the system , Ĥg is the ground state Hamil-
tonian and ĤNH is a slightly modified version of the non-hermitian Hamiltonian introduced
above:
iX †
ĤNH = Ĥe − L̂k L̂k , (2.31)
2
k
where the unitary part of the non-hermitian Hamiltonian is only given in terms of the excited
state Hamiltonian Ĥe . This formalism will be used in Chapters (5) and (6) to derive the effective
operators describing the time evolution of the qubit states for the proposed controlled phase gate.
A full derivation of this formalism is given in Appendix (A).
9
2.2 Quantum Computation
2.2.1 DiVicenzo’s Criteria
There is a list of 5+2 requirements identified by David DiVicenzo in [7] that any system has to
full fill in order to build a general purpose quantum computer. The +2 denotes two additional
conditions that have to be met in order to implement a quantum communication network. These
conditions will be listed here and their implementation in two distinct physical systems will be
discussed. These systems are the ion trap [5] and self-assembled quantum dots embedded in a
nano-photonic wave guide. The former due to the already highly successful implementation and
the later due to it being the focus of this work.
2.2.2 Qubits
Information carriers in classical information theory are bits. These are typically represented by
the absence or presence of a physical entity and take the distinct values 0 and 1, respectively.
The equivalent entity for quantum information theory is the qubit and are represented by the
state vectors |0i and |1i. The most general form of a qubit is described trough the computational
basis and takes the form:
|ψi = α |0i + β |1i , (2.32)
which is a linear superposition of the two qubit states, where α and β are complex coefficients
satisfying |α|2 + |β|2 = 1 to ensure that the qubit state is normalised. The superposition of |0i
and |1i sets qubits apart from the definite valued classical bits which is further highlighted by
the probabilistic nature of a measurement on the qubit. This quantum parallelism enables the
construction of quantum algorithms that operates on several inputs simultaneously, leading to
higher computational power of a quantum computer. Only the relative phase between the two
vectors in the qubits state is physically significant; the global phase does not contribute to any
meaningful physical observable.
Several qubits are often needed for any realistic application of quantum information. The vectors
|ψi1 and |ψ2 i describing the two qubits are defined in separate two-dimensional Hilbert spaces
H1 and H2 , where the total Hilbert space H is a tensor product between the smaller Hilbert
spaces: H = H1 ⊗ H2 . The combined state for the two qubits is then a tensor product between
the individual qubit states: |ψiTwoQubits = |ψi1 ⊗ |ψi2 . If |ψij = αj |0ij + βj |1ij then the
combined qubit state takes the form:
|ψiTwoQubits = |ψi1 ⊗ |ψi2 = (α1 |0i1 + β1 |1i1 ) ⊗ (α2 |0i2 + β2 |1i2 ) (2.33)
= c00 |11i + c01 |01i + c10 |10i + c11 |11i , (2.34)
where |00i is a shorthand notation for the product state |0i1 ⊗|0i2 . This notation is used through-
out the thesis. The coefficients in Eq. (2.34) are defined as c00 = α1 α2 , c01 = α1 β2 , c10 = β1 α2
and c11 = β1 β2 . Note that it is the newly defined coefficients ckl that evolve when the two-qubit
state is evolved in time.
A simple physical example of a qubit is the spin states of any spin-1/2 particle. A spin-1/2
particle has a spin magnitude of S = 1/2 with projection mS = ±1/2. It is then possible to
define |0i ≡ |s = 1/2, mS = 1/2i = |↓i and |1i ≡ |S = 1/2, mS = −1/2i = |↑i. Other simple
examples include the excited and ground states of a two level atom or the polarization degree
of freedom of a photon. These simple systems (except from the photon polarization) are often
not useful for practical applications due to the four other DiVincenzo criteria.
10
Trapped atoms constitutes a working application of a quantum computer. This is largely due
to the well developed ability to control the state of atoms and the well documented knowledge
about the internal structure of various atoms. The majority of experimental methods for ma-
nipulating the internal states of atoms involve illuminating the atoms with lasers. The discrete
nature of the atomic energy levels ensures that any transition between two states can only be
driven if the laser is on resonance with the frequency of the transition. In addition, transition
rules dictate that only light fields with a particular polarization can drive the transition. This
polarization is defined by the combined angular momentum (i.e. the sum of orbital L, electronic
spin S and nuclear spin I angular momentum) of the two states. Illuminating the atom with a
light field with a particular energy (i.e. frequency) and polarization ensures that only transitions
between two specific levels are driven. These two levels can then be directly mapped to the two
qubit states.
Consider a number of 9 Be+ ions confined in a trap (it is advantageous to use ions due to
their charge. The charge makes it possible to use various time-varying electric fields to trap
the ions). These ions have a nuclear spin of I = 3/2 which gives a hyperfine splitting in the
level structure. The hyperfine states |0i = |F = 1, mF = ±1i and |1i = |F = 1, mF = ±2i for
the 2 S1/2 fine structure ground state constitutes a good choice for a qubits, as these states are
well defined and shielded (i.e. they have no allowed transitions that they can decay through).
Interactions between these two states can be readily implemented using Raman transitions using
|F = 2, mF = ±1i as a virtual state. Note that a chain of ions will undergo a vibrating motion
due to the Coulomb force that each ion exert on the other ions in the chain. It is thus necessary
to cool the ions into the motion ground state before any quantum information process can be
implemented. Methods for initializing this qubit (both driving into the qubit states and cooling
the vibrational motion) will be discussed in Section (2.2.4).
It is possible to use confined electronic states in quantum dots for the implementation of qubits.
The so-called trion provides a suitable physical platform for this purpose, as it has a stable
ground state consisting of the two magnetic spin states of an electron. Section (3.3) discusses
this in more depth.
Another important consideration is the need to be able to make a quantum system scalable
in order to make practical use of quantum information protocols. This is evident for the ion
trap, as it gets progressively harder to add ions into the trap and to keep them on a straight line.
This can be accomplished by generating entanglement between the ions and then teleporting
the qubit to various points in the chain [2].
The scalability of self assembled quantum dots in nanohpotonic crystals is an active area of
research but it is still a considerable challenge to attaining the same level of control of a system
with several quantum dots linked together, compared to that demonstrated for a single quan-
tum dot. One proposal involves using the quantum dots as nodes in a quantum communication
network, using the near-unity on-demand single-photon generation that these structures are
promising candidates for.
2.2.3 Universal Quantum Gates

Any logical gate, classical or quantum, is implemented by changing the physical state of a (qu)bit
by taking a set of inputs and deterministically change them to an output (or several outputs
for certain gates). This process is described through Boolean algebra. The basic classical gates
include the AND, OR and NOT gates. It can be shown [23] that any Boolean operation can be
performed using just AND and NOT gates. Furthermore, the NAND gate can implement both
of these gates; the NAND gate is thus referred to as an universal gate from which any Boolean
11
function can be performed. Note that certain classical gates (such as the AND, OR and NAND)
gates are logically irreversible as it is not possible to determine unique inputs for certain output.
As an example, take the output 0 from an OR gate. This output can be the result from inputs
of the form 00, 01 and 10.
Quantum gates perform similarly as classical gates: they take a physical entity, i.e. the qubit,
as an input, and evolves it according to some logical operation, producing the output qubit.
The evolution of the qubit |ψi must follow the laws of quantum mechanics and is governed by
the Schrödinger equation (typically represented in the interaction picture) i~∂ |ψi /∂t = Ĥ |ψi,
where Ĥ is the Hamiltonian that specifies the fields at work. This time evolution, and by
extensionquantum gate that gives rise to the evolution, is described through unitary operators
Û = exp −iĤt/~ , as long as no measurements take place and that unwanted interactions with
the environment are suppressed. Any state, starting from an initial time t0 will then evolve to
the final time t by
|ψ(t)i = Û t, t0 |ψ t0 i .

(2.35)
Unitary matrices thus provide the mathematical description of quantum gates. A matrix is said
to be unitary when Û −1 = Û † and it has the following properties:
• Û † Û = Û Û † = 1
• hφÛ † |Û ψi = hφ|ψi - The inner product is preserved when the vectors are multiplied by
the unitary matrix
• |det (U )| = 1
• Û = V D̂V ∗ - Û is diagonalizable, where D̂ is a diagonal and unitary matrix and V is an
unitary matrix.
• The columns and rows of Û form an orthonormal set of vectors
Pn Pn
• For a fixed column: 2i=1 |Uij |2 = 1 and for a fixed row: 2j=1 |Uij |2 = 1
The evolution of a quantum state can be reversed by multiplying Eq. (2.35) by Û † and using
Û † Û = 1. Any qubit gate is thus logically reversible, in contrast to the classical gates mentioned
above. There do exist logically reversible classical gates (such as the NOT and TOFFOLI gates).
The matrix representation for these gates are given by permutation matrices, whereas Pn unitary
matrices provide a sufficient representation for quantum gates. The two sums 2i=1 |Uij |2 = 1
Pn
and 2j=1 |Uij |2 = 1 ensure that any normalised initial state is evolved into a normalised final
state.
Consider the following matrix:

 
1 1 1 
H=√ (2.36)
2 1 −1
This is the so called Hadamard √ gate that changes a computational basis state in to the super-
position: H |0, 1i = (|0i ± |1i) / 2. This is the basic workhorse of quantum computing due to
its ability to turn a single input state into a superposition. Consider a situation where N qubits
are prepared in |0i and where a Hadamard gate is applied to each qubit in parallel:
N −1
2X
⊗N ⊗N 1 1
H |0i = (|0i + |1i) ⊗ (|0i + |1i) ⊗ · · · ⊗ (|0i + |1i) = |ji , (2.37)
2N/2 2N/2 j=1
12
where |ji is a number written in binary notation. This creates an equal superposition of all
integer states between 0 and 2N − 1 by loading just N qubits into the system. This makes
the Hadamard gate a key component in a number of quantum algorithms, including the Grover
search algorithm and the Deutsch-Jozsa algorithm.
It is possible to make a geometrical representation of a qubit using the Bloch sphere picture.
The qubit can be thought of a a vector pointing to the surface of a unit sphere described by
the three polar coordinates: the radius r, the polar angle θ ∈ [0, π] and the azimuthal angle
φ ∈ [0, 2π]. The chosen frame of reference is such that the North pole is located at the coor-
dinates (1, 0, 0) and the South pole at the coordinates (1, π, 0). The polar coordinates can be
related to Cartesian through the usual transformations:
x = r sin (θ) cos (φ) (2.38)

y = r sin (θ) sin (φ) (2.39)
z = r cos (θ) . (2.40)
The qubit takes the general form |ψi = cos 2θ |0i + eiφ sin 2θ |1i. The full description of the

qubit includes an addition phase eiη multiplied unto both terms in |ψi but, as mentioned before,
this phase holds no physical significance. Rotations about the three axes can be readily expressed
as Euler rotations [20]:
 
cos (θ/2) −i sin (θ/2)
Rx (θ) = exp (−iθσz /2) =   (2.41)
−i sin (θ/2) cos (θ/2)
 
cos (θ/2) − sin (θ/2)
Ry (θ) = exp (−iθσy /2) =   (2.42)
sin (θ/2) cos (θ/2)
 
e−iφ/2 0
Rz (φ) = exp (−iφσz /2) =   (2.43)
0 eiφ/2
Rph (η) = eiη 1. (2.44)
These rotations can be readily implemented using e.g. laser fields for Rabi oscillations. Note
that two full rotations are needed in order to return a qubit to its original state. Any single-
qubit quantum
gate
can be written in terms of these rotation matrices, which can be seen by
considering det Û = 1. It is sufficient to use the special unitary matrix V that is defined by

Û = Rph (η) V . This matrix has a relaxed condition: det (V ) = 1 and has the general form:
 
α −β̄
V = , (2.45)
β ᾱ
from which the determinant is readily calculated: det (V ) = |α|2 + |β|2 = 1. A general choice for
the parameters that satisfies this equation is α = eiκ cos (θ/2) and β = ei sin (θ/2). The special
unitary matrix can then be expressed in the Block sphere picture as:
 
iκ −i
e cos (θ/2) e sin (θ/2)
V =  (2.46)
i
e sin (θ/2) e −iκ cos (θ/2)
13
This matrix can, in addition, also be constructed as the product
Rz ( − κ) · Ry (θ) · Rz (− [ + κ]) (2.47)
with the geometrical interpretation that any rotation on the Bloch sphere can be decomposed
into a sequence of three rotations; one about the z-axis followed by a rotation around the y-axis
and end with another rotation about the z-axis. A simple example is the decomposition of a
rotation about the x-axis into three small rotations by:
Rx (θ) = Rz (±π/2) Ry (θ) Rz (∓π/2) . (2.48)
This has a significant implication for the application of quantum gates: any quantum gate is
represented by a unitary matrix which acts on the qubit by rotating it three times as given by
Eq. (2.47). Each of these rotations are only about a single axis and can be implemented through
standard experimental techniques, such as the effective Hamiltonian arising when performing
Raman transitions between two ground states via a virtual excited state (though it is rotations
about the x- and z axis in this case.) Another, and a more specific, example is the decomposition
of the Hadamard gate into:
H = Rz (0) · Ry (π/2) · Rz (π) · Rph (π/2) . (2.49)
Single-qubit gates are not enough for quantum computational purposes. Gates that operate
between several qubits are also needed. It is sufficient to use two-qubit gates for this; gates
between more than two qubits can be made by chaining several two-qubit and single-qubit gates
together between various qubit pairs. In fact, any kind of quantum gate can be implemented
using just a two-qubit gate and a number of single-qubit gates. These form a set of universal
quantum gates from which any quantum computation can be implemented. The fact that only
a few gates are needed for any computation greatly simplifies the task of engineering specific
interactions between several qubits.
Two-qubit gates are typically implemented by using some physical property to mediate interac-
tions between the two qubits. This can be the quantized vibrational modes that exists in a chain
of trapped ions, or it can be through the emission and absorption of photon by two quantum
dots embedded in a waveguide.
(2) (2)
The controlled phase gate Uπ and the CNOT gate UCNOT are often used to form one of these
universal sets. The superscripts denotes that these are two-qubit gates. The controlled phase
gate imparts a change of phase (this is often chosen to be such that the imprinted phase is
e−iπ = 1) of both input qubits are in state |1i but it leaves the other three input states un-
changed. The CNOT gate changes the state of the second qubit only if the first qubit is in state
|1i and leaves it unchanged otherwise. These gates are summarized in Table. (2.1). The CNOT
gate can be related to the controlled phase gate by the use of a sequence of Hadamard gates by:
UCNOT = (11 ⊗ H2 ) Uπ(2) (H1 ⊗ H2 ) ,

(2)
(2.50)
where the subscripts refer to the specific qubit that the gate operates on. This thesis is concerned
with the possibility of created a heralded controlled phase gate with an integrated error detection
by exploiting the interaction between several quantum dots embedded in a nanophotonic wave
guide.
14
CNOT Controlled Phase
Input Output Input Output
Control Target Control Target Control Target Control Target
|0i |0i |0i |0i |0i |0i |0i |0i
|0i |1i |0i |1i |0i |1i |0i |1i
|1i |0i |1i |1i |1i |0i |1i |0i
|1i |1i |1i |0i |1i |1i |1i − |1i
Table 2.1: Truth tables describing the operation of the CNOT gate and the Controlled
Phase gate
2.2.4 Initialization
Any quantum computation protocol involves the deterministic time evolution of the quantum
mechanical system of choice and the subsequent measurement of the final state. These processes
depend on the initial state of the system; consider the simple case of √ a π/2 rotation about the
ŷ-axis for the three initial states |↓iz , |↑iz and |↑↓iz ≡ (|↑iz + |↓iz ) / 2:
1
Ry (π/2) |↑iz = √ (|↑iz + |↓iz ) = |↑ix (2.51)
2
1
Ry (π/2) |↓iz = √ (|↑iz − |↓iz ) = |↓ix (2.52)
2
1
Ry (π/2) |↑↓iz = |↑iz = √ (|↑ix + |↓ix ) , (2.53)
2
where the spin states also have been decomposed into the basis states for the spin along the x̂
axis. A measurement along the ẑ-axis of the first two final states have a probability of 1/2 to
return either |↑iz or |↓iz , whereas the same measurement for the last final state is guaranteed
to return |↑iz . Changing the measurement axis to be along the x̂-axis changes the results, such
the the result for the first two states are always |↑ix and |↓ix , respectively. This simple example
highlights the need to be a to initialize the system into a well known initial state in order to
deterministically evolve the system during the computation operation.
The ground state |0i = |F = 1, mF = −1i for the ion trap introduced in Section (2.2.2) can be
prepared using the method of optical pumping. This is implemented by continuously pumping
the optical transition between the 2 S1/2 , F = 2 and the 2 P1/2 , F = 2 states. This needs to be done
using a laser with σ − polarization, such that only processes for which MExcited = MGround − 1
are induced by the laser; this ensures that the target state is not excited by the laser field. The
continuous pumping causes the undesired long lived states to be excited to some short lived ex-
cited states; subsequent spontaneous emission will relax the system into the desired state |0i in
the steady state situation. In addition, tuning the laser along the lower vibrational sideband [5]
ensures that the ions are cooled into the vibrational ground state, completing the initialization
process.
A proposed initialisation scheme for a qubit in a self-assembled quantum dot comprised of the
electronic ground state of a trion is described in Section (3.3).
The system will always have some interaction with the outlying environment, which may in-
troduce errors into any quantum computation protocol. There exists protocols that can detect
15
and correct these errors (see Section (2.2.8) for an example if one such protocol). The efficient
implementation of these protocols often needs a continuous supply of ground states in order
to function properly. The need to implement a continuous initialization process is a substan-
tial experimental challenge, making integrated error detection schemes highly desirable. The
implementation for a controlled phase gate found in this thesis has one such integrated error
detection, thereby bypassing the needed for a continuous initialization process.
2.2.5 A Qubit Specific Measurement Capability
The reliable method for obtaining the final result of a quantum computation is needed in order to
make the protocol useful; this requires being able to perform a measurement on a specific qubit.
This can be reliably implemented in the ion trap by applying a σ − polarised laser field on the
2S 2
1/2 , F = 2 to the P3/2 , F = 3 states. The resulting photon cascade from spontaneous emission
can be measured. The transition is chosen such that there is only one allowed optical transition
for spontaneous emission and that only one of the qubit states (in this example it would be the
state |1i) is illuminated by the laser. A cascade of photons will be measured the qubit is in |1i
due to repeated excitation and emission, whereas none are measured in the case where the qubit
is in |0i. Similar measurement schemes exist for probing the optical properties fo quantum dots,
namely resonance florescence and photoluminescence spectroscopy. Alternatively, the approach
used in [8] can be employed.
In addition to measure the final result of a quantum computation, several protocols in the
field of quantum information relies on the ability to perform a qubit specific measurement.
These protocols include quantum teleportation, which relies on performing Bell measurements
to imprint the information stored in the input state to the output state, or to encode additional
information into the input state, respectively.
2.2.6 Long Decoherence Times
Section (2.1.4) introduced the master equation in order to treat the non-unitary dynamics that
is introduced by the coupling of a quantum system to the environment. This can introduce
unwanted changes in the qubit states, ruining the deterministic evolution needed by quantum
information protocols. In addition, the quantum mechanical properties of the system are washed
out, leaving the qubit described by the statistical mixture ρ = |a|2 |0i h0| + |b|2 |1i h1|. In this
situation there will always be a probability of |a|2 to measure the qubit in |0i, regardless of the
measurement axis chosen.
This unwanted evolution happens on a time scale set by the decoherence time T2 , which will
be introduced in Section (3.2.3). The detrimental effects of decoherence can be minimised
by having the time of the gate operation be significantly short than the decoherence time.
This is easily achievable in the ion trap implementation, due to the long life time of hyperfine
states and their insensitivity to the electric fields that trap the ions. The decoherence processes
are more pronounced for quantum dots, as they are mesoscopic emitters embedded in a solid
state environment. These include interactions with phonons in the photonic crystal and charge
fluctuations from lattice defects. These processes are one of the main causes for errors in the
controlled phase gate described in this thesis; these errors and the possibilities of removing them
are discussed in Chapters (5) and (6). The detrimental effects of pure dephasing is also presented
in these chapters..
16
2.2.7 Transmission of Flying Qubtis & the Ability to Interconvert Stationary
and Flying Qubits
The five points outlined above constitute a sufficient set of requirements that need to be satisfied
in order to build a quantum computer. Two additional conditions has to be met in order to build
a suitable platform for quantum communication. This topic is outside of the scope of this thesis
and will only be briefly discussed. Stationary qubits, like those discussed in Section (2.2.2),
are well suited for computational purposes but are confined to a particular lab due to their
tendency of interacting with the environment. Flying qubits thus describe qubits that are well
suited for transmission purposes; photons are a particular useful choice as they interact weakly
with the environment and the transmission of photons is readily available using optical fibres
as a technological baseline. This introduces a new decoherence effect, as the probability to lose
the photon scales exponentially with distance; various schemes exist to remedy this problem
through, an example is a network of quantum repeaters. Quantum dots in a nanophotonic
crystal provide a promising platform for the efficient generation of coherent single photons due
to the new-unity observed β-factor, high achievable Purcell factors and long decoherence times.
These concepts are introduced in Section (3).
2.2.8 Quantum Error Correction

The practical realisation of the quantum gates discussed above have some risk to produce an
erroneous qubit state as the output due to e.g. the coupling to the environment. These errors are
detrimental to the performance of any computational protocol. Various methods for detecting
and correcting these errors have thus been produced in order to protect the implementation of
quantum information against these errors. One such method is Shor’s nine qubit code, which
can correct a single bit flip and a number of phase flips of a qubit state. These errors are defined
as, for a pure qubit state:
(
α |1i + β |0i , Bit flip
|ψi = α |0i + β |1i → (2.54)
Error α |0i − β |1i , Phase flip
The basic idea is to make a correction scheme similar to a classical repetition code. This is
complicated by the no-clone theorem, which forbids the possibility of encoding |ψi as |ψi |ψi |ψi.
Instead, each of the two basis states in |ψi are encoded onto eight additional qubits giving rise
to the following state:
(|000i + |111i) (|000i + |111i) (|000i + |111i)


 α

 √
2 2
|ψi → (2.55)
Encode  (|000i − |111i) (|000i − |111i) (|000i − |111i)
 +β
 √
2 2
Each parenthesis in Eq. (2.55) organises the encoded state into three blocks, each consisting of
three qubit states of the form:
|000i + |111i |000i − |111i

α √ +β √ , (2.56)
2 2
where the first qubit of each block is encoded onto two others with respect to bit flip errors. The
correction of errors then follow a simple procedure, as long as there is one bit flip in only one
of the blocks. Consider the following specific example where there is a phase flip in the second
17
block and a bit flip on the fourth qubit:
(|000i + |111i) (|010i − |101i) (|000i + |111i)

α √
2 2
(2.57)
(|000i − |111i) (|010i − |101i) (|000i − |111i)
+β √
2 2
Running the bit flip code will return the following three binary numbers: 00, 10, 00. The first
and third represents that there are no errors in the respective blocks, whereas 10 informs that
there is a bit flip error on the second qubit in block two. Correcting the error results in the
following state:
(|000i + |111i) (|000i − |111i) (|000i + |111i)

α √
2 2
(2.58)
(|000i − |111i) (|000i + |111i) (|000i − |111i)
+β √
2 2
The phase flip code is applied to this state, which yields a single binary number. The output is
in this case 10, representing the phase flip error in the second block. Correcting this error results
in the original state of Eq. (2.55), which can be decoded into |ψi = α |0i + β |1i by running the
encoding procedure in reverse.
The main drawback of this procedure is the need to load eight additional ground states in order to
implement the error correction code. This, as mentioned in Section (2.2.4), can be experimentally
cumbersome, especially if the error correction has to be implemented several times. This makes
quantum gates with integrated error detection and correction a highly valuable alternative.
18
Chapter 3
Basics of Nanophotonic Structures
The physical structures that are considered in this work are the so called self-assembled quantum
dots embedded in nanophotonic structures that have opened up new avenues for research into the
fields of quantum electrodynamics and quantum optics. Both are concerned with the interaction
between matter and light at the quantum level. It is therefore a great advantage to be able to
tailor the light-matter interaction and nanophotonic structures provide a platform that is well
suited for this task. These structures exhibit a number of advantages compared to the traditional
use of atoms for experiments in quantum optics., which includes the ability to strongly enhance
the light-matter interaction. Modern nanofabrication methods have enabled the possibility to
build photonic crystals (PhC) where the optical properties enhance the interaction between a
photon and a quantum emitter by inhibiting the coupling into the vast array of unwanted optical
modes and/or enhance the coupling to one particular preferred mode. These single-photon
emitters, the quantum dots, can be grown directly in the nanophotonic structures and, due to
the nature of solid-state systems, do not need the trapping and cooling techniques used in atomic
systems. Their optical properties can be tailored by the well developed growth methods. This
section reviews the basics of the photonic crystals in Section (3.1) and Section (3.2) provides an
overview of the concepts and properties of the epitaxially grown III-V semiconductor quantum
dots. Section (3.3) provides a proposal for the initialisation of two qubit state needed for the
implementation of the controlled phase gate presented in this thesis. This chapter is largely
based on the review article from [13].
3.1 Nanophotonic Waveguides

3.1.1 Structural Properties
A photonic crystal is an inhomogeneous dielectric material with a periodically modulated re-
fractive index. The propagation of light is strongly affected by optical Bragg scattering, if the
periodicity of the refractive is on the order of the optical wavelength. This is analogous to the
electronic Bragg scattering in crystallographic experiments in solids and is the key to the desired
optical properties employed in nanophotonic QED. These materials are highly dispersive and
can support a wealth of Bloch modes. A photonic band gap, in which no Bloch modes exist,
can open up in the band diagram for the material, provided that the Bragg scattering is suffi-
ciently powerful. The typical devices used are 2D membranes, where the propagation of light is
suppressed in two dimensions by the band gap and is confined in the third dimension through
total internal reflection at the semiconductor-air interface, similar to the reflection of light in
optical fibres. A suppression in all three dimensions would be ideal but there are substantial
technological challenges associated with the production of 3D modulated structures. In addition,
19
Figure 3.1: Dispersion diagram and frequency dependence of the LDOS in a photonic
crystal waveguide. (a) Dispersion diagram of the TE-modes of a W1 waveguide in
a photonic crystal. Three waveguide modes appear in the band gap; it is typically
the lowest frequency mode that is employed in experiments. The grey areas show
contain extended Bloch modes and leaky radiation modes. The insert shows a sketch
of the crystals with a triangular lattice with lattice constant a and hole radius r. (b)
Frequency dependence of the LDOS, normalised to that for a homogeneous medium.
The red line corresponds to a x-dipole while the y-dipole is represented by the blue
line; both dipoles are placed at the cross i the insert of (a). The dashed vertical
lines mark the emergence of the frequency region of the extended Bloch modes. The
expression for the two LDOS are given in Eqs. (3.3) and (3.4). This figure is reprinted
from [13]
2D membranes provide a convenient method for probing emitters with laser excitation from the
top of the membrane. GalliumArsenide (GaAs) has proven to be well suited for quantum optics
experiments due to the ease of growing InGaAs quantum dots in the crystals and due to the
highly advanced and precise fabrication methods. A 2D GaAs structure is shown in Fig. (3.1)
with air holes of radius r. These holes are arranged in a triangular pattern with lattice constant
a. The size and central frequency og the band gap can be controlled by modifying the values
of r and a. The typical values for the radius of the holes is r = a/3 and the structure has a
thickness of t = 2a/3.
3.1.2 Optical Properties

The local density of states, LDOS, quantifies the local light-matter interaction strength. It
specifies the number of optical states at frequency ω per frequency bandwidth and mode volume
as experienced by the emitter. The LODS and the corresponding radiative decay rate, is given
by:
20
Figure 3.2: (b) and (c) shows a spatial map in the x − y plane of the LDOS ρ of
an x- and y-dipole, respectively. The LDOS is normalised to the LDOS ρ/ρHom for
a homogeneous GaAs structure. (d) Frequency dependence of ρ/ρHom for a x-dipole
(red curve) and a y-dipole (blue curve). These dipoles are placed at the crosses in (b)
and (c). This figure is an exert from Fig. (8) from [13].
X
ρ (r0 , ω, êd ) = |êd · ûk (r0 )|2 δ (ω − ωk ) (3.1)
k
πω0
γrad (r0 , ω0 , êd ) = |d|2 ρ (r0 , ω, êd ) , (3.2)
0 ~
where r0 is the position of the emitter, k is the Bloch wave vector and êd is a unit vector
pointing in the direction of the transition dipole moment d of the quantum dot. The LDOS for
a homogeneous medium and for a W1 waveguide for the interaction with a single quantum dot
are given by:
nω 2
ρHom = (3.3)
3π 2 c3
a f (r)
ρW1 (rn , ω, ed ) = |ê (r) · êd |2 , (3.4)
πvg (r) VEff
where n is the refractive index of the waveguide, vg is the group velocity, f (r) is a dimensionless
function describing the spatial mismatch between the emitter and waveguide-mode field, VEff is
the effective mode volume and ed is a unit vector poiting along the dipole. The expression for
21
the waveguide and the various functions are explained in greater detail in Chapter (4). Fig. (3.1)
shows a calculated LDOS for as a function of frequency for a GaAs membrane. A strong sup-
pression of the LDOS in the band gap region is clearly seen, leaving any radiation of a dipole
oriented in the plane of the dipole to only be of the form of the weakly radiating non-guided
modes found above the light line of the band diagram. A suppression up to a factor of a 160
can be seen from Fig. (3.2), relative to the LDOS found from a homogeneous medium of GaAs.
This results in a strongly suppressed decay rate for these modes, as is seen from Eq. (3.2). This
large suppression the coupling to the unwanted modes eliminates one of the main sources of loss
and is the main strength of the photonic crystals. Note that any other source of loss, such as
absorption of the transmitted light, is detrimental to the functionality of the crystal.
It is also possible to enhance the strength of the coupling to a particular preferred mode, in
addition to the suppression of the coupling to the unwanted modes detailed above, to greatly
enhance the light-matter interaction. The W1 waveguide can be readily implemented in a pho-
tonic crystal by leaving out a row of holes i the Γ-K direction of a triangular lattice. This allows
a limited numbed of bands to appear in the band gap, making it possible for certain modes of
light to propagate through the waveguide. These modes are highly dispersive, which can give rise
to a strong slow-down effect og the propagating light through the group velocity vg (ω) = |∇kω |.
Losses are ideally eliminated, as the bands appear below the light line of the band diagram,
but fabrication imperfections will allow a small leakage through a vertical coupling. The band
diagram and frequency dependence of the LDOS are shown in Fig. 3.1 for a W1 waveguide of
GaAs. Three guided modes are formed inside the band gap and the slow-light effect is clearly
seen from the vanishing slopes of the dispersion curves. This effect is the most pronounced at
the edge og the guided mode, where the group velocity tends to zero, and the corresponding
group index ng = c/vg diverges. Additionally, the mode in the waveguide is confined to a small
spatial volume, which causes the electric field to have an additional component, directed along
the propagation direction. It is possible to couple the x- and y-dipoles to the odd and even
mode, respectively, by placing the emitter in a high-symmetry point. This gives the y-dipole
a wide range of possible frequencies with an enhanced LDOS; this makes the even mode the
most commonly used mode for experiments. The x-dipole, on the other hand, has a shorter
range of frequencies with several sharp peaks for the LDOS, due to a vanishing group velocity
at several frequencies. The effect of the band gap region is also seen in the frequency region
ωa/2π = (0.255, 0.26) which shows a suppressed LDOS.
The effect of slow light enhances the interaction between the guided modes and a quantum emit-
ter embedded in the waveguide. This enhancement is quantified by the Purcell factor, which for
a PhC waveguides is given by:
3 λ2 /n2

FPmax (ω) = ng (ω) , (3.5)
4πn Veff /a
where Veff ∼ a (λ/n)2 /3 is the effective mode volume for a W1 waveguide and λ/n is the wave-
length of the light inside the waveguide, with n a the refractive index of the waveguide. The
mode volume varies weakly over the waveguide band, which leaves the group index as the pri-
mary contributor to the Purcell factor. Values of ng ∼ 300 for a silicon based W1 waveguide
[22] has been achieved experimentally, which gives a Purcell factor of 120, whereas the group
index approaches ng ∼ 50 for GaAs waveguides [1]. Fig. (3.3) shows the spatial dependence of
the Purcell factor and the β-factor (introduced below) in both the fast light (ng = 5)- and the
slow-light (ng = 58) regime.
An additional advantage of PhC waveguide is that they are open systems, which allows a quan-
tum emitter to directly channel an emitted photon into the waveguide. This is in contrast to
22
cavity based systems, where the single photons are coupled to localized modes and needs to be
coupled out of the cavity. This makes PhC waveguides prime candidates for the use in quantum
information processes that requires the interaction between stationary and flying qubits. The
coupling between the quantum dot and the waveguide is quantified by the β-factor:
γwg
β= , (3.6)
γwg + γng + γnrad
where γwg is the emission rate into the waveguide, γng is the loss rate to non-guided modes and
γnrad is the loss rate to various non-radiative losses of the quantum dot. Recent experiments
have determined a near-unity β-factor > 0.98 [1] due to the strong suppression of γng by the
photonic band gap. The main limiting factor for the β-factor is from the intrinsic loss rate γnrad
of the quantum dot.
23
Figure 3.3: Spatial dependence of the Purcell factor and β-factor in a photonic waveg-
uide. The left panel shows the overall Purcell factor, while the center panel shows
the Purcell enhancement of emission into nonradiative modes. The panel to the right
shows the enhancement of the radiative β factor. (a) is for a x-dipole with ng = 5,
while (b) is for a y-dipole with the same group index. (c) and (d) shows plots for a
x- and y-dipole respectively with ng = 58. This figure is reprinted from [13].
24
3.2 Quantum Dots
3.2.1 Basic Properties
Self-assembled semiconductor quantum dots provide a suitable platform for an implementation
of quantum emitters into a PhC waveguide. These structures contain thousands of individual
atoms that collectively display similar optical properties to natural atoms, due to the quantum
confinement of electrons on a nanometer length scale. This confinement can be implemented by
sandwiching a lower-band gap semiconducting material between two pieces of another semicond-
cting material with a higher band gap. The regions with the lower band gap acts as a trapping
potential and strongly changes the electronic density of states (E), depending on the number of
confined dimensions:
meff p
No confinement: Bulk material − g3D = 3 2 2meff (E − E0 ) (3.7)
π ~
meff X
Confinement in 1D: Quantum well − g2D = θ (E − Ej ) (3.8)
π~2
j
√
2meff X θ (E − Ej )
Confinement in 2D: Quantum wire − g1D = p (3.9)
π~ E − Ej
j
Confinement in 3D: Quantum dot − g0D = 2δ (E − Ej ) , (3.10)
where Ej denotes discrete energy levels. The discrete structure of the electronic density of states
allows quantum dots to emit single photons when an correlated election-hole pair (also called
an exciton) recombine. An energetic electron-hole pair can be generated at the GaAs band
gap by absorption of a photon. The pair can then relax into the confinement potential of the
quantum dot through nonradiative processes. This correlated pair will eventually recombine,
thereby emitting a single photon. This emission process can be measured using photolumines-
cence spectroscopy, which reveals a sharp peak at the emission frequency, similar to that of an
atom. The electron occupancy can be directly controlled by varying the gate voltage over the
heterojunction[21]. This also makes it possible to control the interaction strength between the
quantum dot and the guided mode by enhancing the overlap of the envelope functions. These
are introduced in Section (3.2.2).
One common approach for growing quantum dots is the Stranski-Krastanov method, which
relies on the self assembly of InAs quantum dots on a GaAs surface due to the mismatch be-
tween lattice constants in the two materials. This leads to a truncated pyramidal shape due to
material intermixing and further leads to an inhomogeneous indium distribution and a spatially
varying strain. The typical height of a quantum dot is 1 − 10 nm with an in-plane size of 10 − 70
nm. This leads to aspect ratios relatively larger than unity with a quantization axis in the
growth direction. Each quantum dot needs to be tuned differently from others due to inevitable
differences in the size of quantum dots from the same growth run.
The tetrahedral bonds and zinc-blende crystal structure for InAs, GaAs and InGaAs quantum
dots results in the formation of three degenerate energy bands in the valence band. The split-off
band gets shifted to a lower energy level by including spin-orbit couplings, while the degeneracy
between the heavy- and light-hole bands gets lifted by the aspect ratio of these quantum dots.
This energy spacing is sufficiently large that the lowest energy transition is between the conduc-
tion band and the heavy-hole band. The unfilled orbital shells of Ga, In and As are 4s2 4p, 5s2 5p
and 4s2 4p3 , which results in the conduction band exhibiting s-symmetry while the valence band
shows p-symmetry. Additionally, the heterojunctions of InGaAs embedded in GaAs (as well as
GaAs in AlGaAs) have type-I energy bands alignments, which ensures that carriers are confined
25
in both the conduction and valence bands. This is crucial for the efficient interaction with light
for the formation of quantized states of elections and holes.
3.2.2 Effective Mass Theory & the Transition Matrix Element

Quantum dots can be reliably described through the effective mass theory, provided that the
energy-level spacing of carries is large compared to the confinement potential such that the
potential can be treated as a perturbation. In addition, only a small region in k-space is
assumed to contribute to the interaction and that only small carrier populations is involved
with the interaction. Lastly, the larger than unity aspect ration for the relevant quantum dots
shift the light-hole band energy far enough away from the heavy hole band energy that only the
latter is needed for a sufficient description of the quantum dot. Note that this assumption relies
heavily on the large aspect ratio and can easily break down. The wave function for an electron
in the conduction band c or valence band v can be split into three parts:
|ψc,v i = |χc,v i |uc,v i |αc,v i , (3.11)
where |χc,v i is the envelope wave function, |uc,v i is the electronic Bloch function and |αc,v i is the
spin wave function. The envelope function is obtained from the time independent effective-mass
Schrödinger equation:
2
−~ 1
∇ ∇ − Vc,v (r) |χc,v (r)i = (E − Ec,v ) |χc,v (r)i , (3.12)
2m0 me (r)
and describes the modulation of the Bloch function to give the full wave function. E is the
electron eigenenergy, Ec,v is the band edge energy, Vc,v (r) is the confinement potential, m0 is
−1
the electron rest mass and me (r) = ~2 ∂ 2E/∂ (k)2 is the effective mass. This Schrödinger
equation can be cast into a form that resembles the equation for a hydrogen atom by assum-
ing that the effective mass is isotopic (which is a reasonable approximation for type III-V
semiconductors) and using an attractive Coulombic potential between the electron and hole:
V (r) = −e2 / (4π0 |re − rh |). Here e is the elementary charge, 0 is the vacuum permittivity,
is the permittivity of the material and |re − rh | is the distance between the election and the
hole. The complicated potential profile of the unit cell has been merged into the effective mass,
which can be accurately determined from experiments. These assumptions greatly simplifies the
problem of solving the Schrödinger equation and allows to find the discrete energy levels of the
quantum dot.
The minimal coupling Hamiltonian Hc induces optical transitions and is in the Coulomb gauge
given by:
e
Hc = − p · A (r0 , t) , (3.13)
m0
where p = −i~∇ is the momentum operator and A (r0 , t) is the vector potential of the elec-
tromagnetic field and is evaluated at the position of the quantum dot. This Hamiltonian is
written in the dipole approximation, which is valid when the wavelength of light is significantly
larger than spatial extent of the quantum dot. This Hamiltonian enter into the decay rate for
spontaneous emission, which is given by Fermi’s golden rule:
πe2
Γ (ω) = |hf |p|ii|2 ρ (r0 , ω, êp ) , (3.14)
~ωm20 0
26
Figure 3.4: The transition for the lowest-energy confined states in a quantum dot.
The full blue (empty red) circles represent the electron (hole) configuration in the
conduction (valence) bands. Reprinted from [13]
where |f i and |ii represent final and initial states, respectively. The crucial component is the
transition dipole matrix element P = hψv |p|ψc i which, when using Eq. (3.11), takes the following
form:
P = hFv |Fc i huv |p|uc i hαv |αc i , (3.15)
which reveals three selection rules for optical transitions: (i) the envelope functions must have
the same parity, (ii) the Bloch functions must have opposite parity, and (iii) the electron spin
must be unchanged. These selection rules, along with the respective s- and p-symmetry of the
conduction- and valence band, makes it possible to determine the polarization of the optical
transitions. The electron pseudospin states are defined as
|↑i ≡ |uc i |↑e i = |us i |↑e i (3.16)

|↓i ≡ |uc i |↓e i = |us i |↓e i (3.17)
and the hole pseudospin states are

1
|⇑i ≡ |uv i |↑h i = − √ (|ux i + i |uy i) |↑h i (3.18)
2
1
|⇓i ≡ |uv i |↓h i = √ (|ux i − i |uy i) |↑h i , (3.19)
2
27
where |us i , |ux i and |uy i denotes functions with even parity, odd parity along x and odd parity
along y, respectively. It is easy to see that the states |⇑↑i and |⇓↓i have optically forbidden
transitions, as the relevant matrix s are given by the inter products h↑e | ↑ hi = h↓e | ↓ hi = 0.
These definitions makers it possible to evaluate the matrix element for the transition dipole; this
is simplified by assuming that the magnitude of the two odd parity Bloch function are equal and
differ only bu the direction of their circular polarization. The transition matrix elements for an
electron-hole pair is then given by:
1
hg|p| ⇑↓i = − √ hFv |Fv i hux |p|us i (êx + iêy ) (3.20)
2
1
hg|p| ⇓↑i = √ hFv |Fv i hux |p|us i (êx − iêy ) , (3.21)
2
where the inner product between the spin states have been implicitly evaluated to one. The
optical transition for an electron-hole pair is evidently circular polarized. The most general
electron-hole states are linear superpositions of the optically allowed states listened in Eqs. (3.16)
through (3.19). These bright exciton states are defined as
1
|Xb i = √ (|⇑↓i − |⇓↑i) (3.22)
2
1
|Yb i = √ (|⇑↓i + |⇓↑i) . (3.23)
2
The optical transitions for these two exciton states are linearly polarized along x and y respec-
tively, which is easily seen from Eqs. (3.20) and (3.21). The dark exciton states are defined
as:
1
|Xb i = √ (|⇑⇓i − |⇓⇑i) (3.24)
2
1
|Yb i = √ (|⇑⇓i + |⇓⇑i) , (3.25)
2
and are related to the bright states through spin-flip processes. There exist several other excitonic
states, such as the trion states (with two holes in the valence band and a single electron in the
conductance band for the positive trion or a single hole and two electrons for the negative trion)
and the biexciton, which is comprised of two correlated electron-hole pairs. The negative and
positive trion decay to a single electron and hole, respectively, while the biexciton can decay to
either of the bright exciton states. The various excitonic states, their pseudospin representation
and their allowed optical transitions are summarized in Fig. (3.4).
3.2.3 Optical Properties of Quantum Dots

The ground states of an exciton in comprised of an electron with spin and spin projection
S = 1/2, MS = ±1/2 in the conduction band and a heavy hole with angular momentum J = 3/2
and projections MJ = ±3/2 in the valence band. The light hole band can often be neglected, as
it is separated in energy from the heavy-hole band due to strain effects and the large aspect ratio
of the quantum dot. The total angular momentum of the exciton is thus Jex = ±2 for the dark
excitons of Eqs. (3.24) and (3.25), while Jex = ±1 is associated with the bright excitons defined
in Eqs. (3.22) and (3.23). The Coulomb exchange energy induce an energy-splitting between the
dark and bright states of ∆Efss ∼ 10 − 100µeV and they are coupled through spin-flip processes,
as mentioned above, at a rate of γsf . An additional effect of the exchange energy is a splitting
of the two dark states on the order of 1µeV.
28
Figure 3.5: Basic level scheme for the exciton. The bright exciton can decay through
radiative and nonradiative processes, while the dark exciton only decay nonradiatively.
Spin-flip processes between the two states can occur at the rate γsf . This is an excerpt
from Fig. 4 in [13].
The bright states can, in addition to the radiative transitions with decay rate γrad discussed
above, decay through nonradiative transitions with a rate of γnrad , while the dark states can only
decay through decay nonradiantly. The spin-flip processes typically takes place over longer time
scales than the radiative decay, and transition between |Xb,d i and |Yb,d i require the simultaneous
change of spin on both the electron and the hole. It is therefore sufficient to only include the
spin flip processes between |Xb i and |Xd i, as well as between |Yb i and |Yd i. Consider the
level scheme in Fig. (3.5) for the decay dynamics of the bright and dark exciton. The exciton
follows a bi-exponential decay in the case of non-resonant excitation and with evenly distributed
populations in the dark and bright state ρb (0) = ρd (0) = 1/2. The population of the bright
state then decay as:
ρb = Af e−γf t + As e−γs t , (3.26)
with fast and slow decay rates and amplitudes

s
2
γrad
γrad
γf,s = + γnrad + γsf ± + γsf2 (3.27)
2 4

γrad γsf
Af,z = ρb (0) 1 + ± ∓ ρd (0) . (3.28)
γf − γs γf − γs
This bi-exponential decay makes it possible to map out the radiant, nonradiant and spin-flip
decay rates for a single quantum dot through time resolved photoluminescence spectroscopy.
This must be done for each single quantum dot in an ensemble, as random processes during
fabrication can result in considerable differences in the spin-flip and nonradiative decay rates.
Note that trions and biexcitons do not have a fine structure. The carrier population will thus
follow a single-exponential decay for these excitonic states with no direct way of accessing the
spin-flip and nonradiant decay rates.
29
The mesocopic nature of quantum dots and the solid state structure of the photonic crystals that
they are embedded in gives rise to various decoherence processes, including interactions with
phonons in the crystal, interactions with charge fluctuations due to lattice defects and coupling
to spin fluctuations in the nuclei of the lattice atoms all give rise to the non-radiative decay and
to dephasing, where the processes due to dephasing can reduce the coherence of the quantum
dot without affecting the population of the excitonic states. High coherence is needed in order to
efficiently implement quantum information protocols. The simplest approach is to assume that
the dephasing processes are Markovian and can be described by a single decay rate γdp . This
decay can be phenomenologically introduced into the equations of motion for the off-diagonal
element of the master equation. The resulting lifetime can be decomposed in to a contribution
from spontaneous emission and from dephasing processes:
1 1 1
= + ∗, (3.29)
T2 2T1 T2
where T2 is the total coherence time, T1 is the inverse of the total decay rate of the emitter
−1
and T2∗ = γdp is the dephasing time. The existence of dephasing causes the coherence to decay
faster than the population and can introduce additional errors in quantum information proto-
cols. Dephasing times up to T2∗ = 0.6ns have been reported for a photonic crystal cavity [14].
The effects of charge noise can be reduced by resonantly exciting the quantum dot, whereas the
effects imposed by interactions with phonons can be reduced by cooling the quantum dot down
to near absolute zero temperatures, as can be seen from Fig. (3.6). Fig. (3.6) shows the emission
spectrum of a quantum dot coupled to a longitudinal acoustic (LA) phonon reservoir for various
temperatures. This demonstrates that LA phonons constitute a significant broadband source of
dephasing with an asymmetrical broadening of the linewidth of the emission spectrum.
The detrimental effects of the pure dephasing processes on the proposed controlled phase gate
are discussed in Sections (5) and (6).
The interaction strength is determined by the transition matrix element, as introduced above.
This in an intrinsic, immutable property of atoms but can be manipulated in quantum dots.
This is quantified by the oscillator strength f , defined as the ratio of the radiative decay rate
in a homogeneous medium and the decay rate of a classical harmonic oscillator of elementary
charge. The oscillator strength is given by:
2 Ep
f= |P|2 = |hFv |Fc i|2 , (3.30)
~ω0 m0 ~m0
where EP is the Kane energy and is of the order of EP ∼ 25eV for InAn and GaAs. It
is seen that the oscillator strength can be maximized by increasing the overlap between the
conduction band and valence band envelope functions. These can be changed by tuning the
trapping potential barrier of the quantum dot and it is thus possible to increase the light-
matter interaction directly through the emitter, in contrast to the case with atomic system,
where the light-matter interaction strength can only be increased by changing the LDOS of
electromagnetic modes that the atom can couple to. It is possible to further increase the light-
matter interaction for the quantum dot by embedding it in a W1 waveguide in a PhC, thereby
inhibiting the continuum of unwanted modes and enhancing the coupling to a desired guided
mode, as described in Section (3.1).
30
Figure 3.6: Broadening of the normalised emission spectrum of a quantum dot coupled
to a LA phonon reservoir. The strong dependence on temperature is observed with a
high degree of asymmetry in the sidebands at low temperatures. Reprinted from [13].
3.3 Initialization of Qubit States

The properties of quantum dots and waveguides embedded in photonic crystals provide a promis-
ing platform for implementing quantum information protocols. Any successful implementation
requires that the qubits are represented by long lived quantum states, as explained in Section
(2.2.2). The bright exciton described above is not a useful physical entity for this purpose,
due to the emptiness of the ground state. The biexciton is unsuited for the same reason, as
it decays to the either of the bright excitons. Instead, consider the structure of the negatively
charged trion: it consists of two electrons in a singlet with a heavy hole. The trion is denoted by
|↓↑⇑i (|↓↑⇓i) and has the spin projections MS = 3/2 (MS = −3/2). This is coupled to a ground
state consisting of an electron |↑i (|↓i) with spin projection MS = 1/2. (MS = −1/2). The
optical transition between the trion and the electron is circularly polarised (see Fig. (3.4)) with
a decay rate of γ 1D into the guided mode. It is possible to use the states |↑i and |↓i for the
electron as qubit states due to the lack for optical transitions for an electron and the trapping
of the electron in the confinement potential; they will thus be denoted as |↑i = |1i and |↓i = |0i.
These qubits can be initialised following the procedure from [8], which is highlighted below.
The four states listed above each undergo a different energy shift when an external magnetic field
is applied along the z axis. Transitions from the state |↓↑⇓i can be neglected if a σ + polarised
laser at Rabi frequency ΩR is introduced at the optical transition, as the coupling strength is
reduced by a factor of 103 for magnetic fields larger than 60mT. The laser has a detuning of
∆ω = ω0 − ωL , where ω0 is the transition frequency of the trion and ωL is the laser frequency.
The Hamiltonian describing the system is
Ĥ = ĤHyp + ĤCharge + ĤPhonon + ĤInt,rad + ĤZeeman , (3.31)
31
Figure 3.7: (a). Four-level system for the negatively charged trion. A magnetic field
is applied along the growth direction, which induces a Zeeman splitting ~ωz of the
electron and trion states. A laser with Rabi frequency ΩR̊ is introduced along the σ +
transition. ΩH and κ denote coherent and incoherent spin-flip processes, respectively.
(b) Dressed state picture formed by eliminating ΩH . The laser field has been split up
into ΩR,1 and ΩR,2 , while a weak decay rate γ̃ has been introduced. Reprinted from
[8]
where the first three terms describe coupling to the nuclear spin (i.e. the hyperfine effect),
electron gas and phonons, respectively. In addition, this scheme suppresses the decay rate for
nonradiative decay processes between |↑↓⇑i and |↓i. The interaction with the radiation field
is treated in Chapter (4) and results in dissipative evolution of the quantum dot with a decay
rate of γ 1D . Coherent and dissipative dipole-dipole interactions exist in a structure containing
several quantum dots, which may exert an influence on the initialisation scheme presented here.
In addition, the coupling to the phonons and electron gas can be treated as dissipative processes
in the Born Markov approximation, leaving ĤHyp and ĤZeeman as the only unitary processes of
Eq. (3.31).
The magnetic field is given by B = BExt + BN , where BExt = (0, 0, Bz ) is the external magnetic
field aligned along the z axis and BN is the nuclear magnetic field. The hyperfine Hamiltonian
is given by:
ĤHyp = ~ΩH σ̂x , (3.32)

q
ge µB BN,x2 2
+ BN,y
~ΩH = , (3.33)
2
where ge is the quantum dot electron g factor, µB is the Bohr magneton and BN,{x,y} is the x and
y component of the nuclei magnetic field. ΩH describes a coherent spin-flip process between the
two electronic ground states due to fluctuations in the nuclei magnetic field. This will typically
be much lower than the effect from the external magnetic field. The z component BN,z is not
included here as it leads to a Zeeman shift; it is thus included in the Hamiltonian for the Zeeman
32
effect:
ĤZeeman = ~ωz σ̂z + gh µB Bz · Jˆz , (3.34)
ωz = ge µB (Bz + BN,z ) , (3.35)
where gh is the quantum dot hole g factor, σ̂i are Pauli spin operators for the electronic ground
states and Jˆz is the z component of the hole spin operator. The spin-reservoir couplings leads to a
weak mixing between |↓i and |↑i and between the two trion states; the mixing rate κ between the
electronic state can be suppressed using sufficiently low gate voltages, whereas mixing between
the trion is neglected due to the low probability for the system to be in these excited states.
The Hamiltonian is Eq. (3.34) can be cast in a more intuitive picture by transforming to the
dressed state picture. The new dressed states are:
|0̃i ≡ cos θ |0i − sin θ |1i (3.36)
|1̃i ≡ sin θ |0i + cos θ |1i (3.37)
|ei ≡ |↓↑⇑i , (3.38)
where θ = ΩH /ωz is the mixing angle. This can be taken as θ 1 for the experimental
parameters in [8], leading to the dressed state Hamiltonian:

H̃ = ~ωz |0̃i h0̃| + ∆ω |ei he| + ΩR,0 |0̃i he| + |ei h0̃| + ΩR,1 |1̃i he| + |ei h1̃| . (3.39)
This leads to a system that is diagonal in the dressed ground states |0̃i and |1̃i. The single
laser field is in this basis represented by two laser fields, ΩR,0 and ΩR,1 , that couple the excited
state to |0̃i and |1̃i, respectively. In addition, the spontaneous emission is split into two distinct
channels with emission rates γ̃ 1D and γ̃. The relevant parameters for the interactions between
|1̃i and |ei are:
γ̃ 1D = γ 1D , ΩR,1 = ΩR , ∆ω0 = ∆ω, (3.40)
while the parameters for the |0̃i ↔ |ei subsystem are:
γ̃ = θ2 γ 1D , ΩR,0 = θΩR , ∆ω1 = ∆ω + ωz (3.41)
This creates an effective Raman transition between two dressed ground states when a resonant
laser is applied on the |1̃i ↔ |ei. The absence of an external magnetic field equalises the two
subsystems, leading to a random mixed spin: ρ0̃0̃ (t = ∞, Bz = 0) = ρ1̃1̃ (t = ∞, Bz = 0) = 1/2,
leading to a mixed state ρ = 1/2 (|0i h0| + |1i h1|) unsuited as a starting point for quantum infor-
mation processes. On the other hand, keeping the external magnetic field non-zero inhibits the
transfer rate from ρ0̃0̃ to ρ1̃1̃ , where ρ is the density matrix formed from the three dressed states
in Eqs. (3.36) through (3.38). This makes it possible to prepare the population of the spins to
be solely in ρ00 for a strong external magnetic field in the steady state, while keeping the popu-
lation of the excited state empty. Keeping the external magnetic field at a value of Bz ∼ 60mT
leads to a coherent population transfer with ρ0̃0̃ (t = ∞, Bz ∼ 60) = ρ1̃1̃ (t = ∞, Bz ∼ 60) = 1/2.
Choosing the dressed ground states as the qubit states will then make it possible to prepare the
following superposition state:
1
|ψi = √ |0̃i + |1̃i (3.42)
2
1
≈ √ (|0i + |1i) , (3.43)
2
for mixing angles θ 1. Repeating this scheme for a second qubit gives the product state:
1
|Ψi = |ψiq,1 ⊗ |ψq,2 i = (|00i + |01i + |10i + |11i) , (3.44)
2
from which the derivation of the controlled phase gate in Chapter (5) starts from.
33
34
Chapter 4
Decay Properties of Coupled Quantum

Emitters
A strong interaction between a number of quantum emitters is a necessary condition in order

to implement a two-qubit quantum gate. This interaction is typically mediated by some other
physical entity. This chapter will formalise the interaction between three self assembled quantum
dots embedded in a Q1 waveguide in a GaAs photonic crystal. The quantum dots acts as the
quantum emitters, where the interaction between them is mediated by photons propagating in
the waveguide. Section (4.1) introduces the model for the three-level quantum emitters, while
the simplified case of three coupled two-level emitters is discussed in Section (4.2) as a staring
point. The full treatment of the coupled three-level emitters is given in Section (4.3).
4.1 Defining the System

It is possible to use the dipole-dipole interaction between three quantum emitters to implement
a controlled phase gate. Two of the emitters are used to represent the actual qubit states, while
the third auxiliary emitter is used to probe the system and for implementing an integrated
error detection. This chapter will explore the decay properties of three coupled quantum dots
embedded in a W1 photonic waveguide. All three emitters couple to the waveguide through
optical transitions between the excited state and the optical ground state. In addition, a weak
laser field V̂ is directed at a forbidden transition for the auxiliary emitter. The level scheme of
the auxiliary and qubit emitters is shown in Fig. (4.1) and can be represented by the following
Hamiltonian in the dipole approximation and in a proper rotating frame:
2
X Ω
Ĥ = ~∆A |EiA hE| + ~ δn |ein he| + [|EiA hg| + |giA hE|] (4.1)
2
n=1
2
" #
(A) (A) (n) (n)
X X
+~ gk σ̂+ ak ei(∆A −ωk )t + gk σ̂+ ak ei(δn −ωk )t + H.C. (4.2)
k n=1
where ∆A = ωE − ωg − ωL is the detuning between the laser and the transition energy of the
forbidden transition for the auxiliary emitter. δn = ∆A + ωe,n − ω1,n is a detuning for the qubits
formed from the choice of rotating frame. ωL is the frequency of the laser, while ωx denotes
the frequency for level x, Ω is the strength of the laser and ak is the annihilation operator of a
photon with wave vector k and frequency ωk . The auxiliary emitter is not included in the sums
in the equations above, as it may have different optical properties from the qubit emitters. The
coupling constant gk is defined from the interaction Hamiltonian between the dipole operator
35
d̂ = d (σ̂+ + σ̂− ) and a reservoir of photonic states through the quantized electric field Ê. The
electric field is written in terms of the photonic annihilation and creation operators as:
X
Ê (r, t) = i [Ek (r) ak − H.C.] , (4.3)
k
where the electric field is expanded in the normalized mode functions uk (r) by:
r
ωk
Ek (r) = ~ uk (r) . (4.4)
2~ε0
The interaction Hamiltonian is then given by:

3
X
ĤI = − d̂(n) · Ê (r, t) (4.5)
n=1
3 Xh i
(n) (n)
X
≈~ gk σ̂+ ak + H.C. , (4.6)
n=1 k
in the dipole approximation and rotating wave approximation. The coupling constant for the
n’th emitter has been defined as:
r
(n) ωk n
gk = d ed · uk (r) , (4.7)
2~ε0
where ε0 is the vacuum permittivity, d(n) is the magnitude of the dipole moment for emitter n
and ed is a unit vector pointing in the direction of the dipole.
4.2 Decay Dynamics for Three Two-Level Quantum Emitters

This section develops a formal derivation of the decay dynamics of three two-level emitters in an
arbitrary dielectric material, whose electromagnetic properties are described though the Green
tensor G (r, ω, êd ). In addition, this section will elaborate upon Eqs. (3.1) and (3.2) that were
introduced in Chapter (3.1.2) and will expand upon the discussion presented is this chapter.
4.2.1 General Formalism

A strong interaction between several quantum emitters is needed in order to make a functioning
two-qubit quantum gate. This interaction will be modified by the optical properties by the
material, in which the emitter is embedded, as well as the properties of the emitters. This
section will describe a general theory for the spontaneous emission of three coupled emitters
embedded in an arbitrary dielectric material and will specialize this to a W1 waveguide in a
photonic crystal. In addition, the effects if super- and subradiance will be introduced.
Only the levels coupled to the radiation continuum in Fig. (4.1) will be considered in this section.
The Hamiltonian governing the system is written in the non-rotated frame:
3 3 Xh
X 1 (n)
X † 1 X (n) (n)
i
Ĥ = ĤS + ĤI = ~ω0 σ̂z(n) + ~ωk âk âk + +~ gk σ̂+ âk + H.C. , (4.8)
2 2
n=1 k n=1 k
(n) (n) (n)

where ω0 is the transition energy for emitter n and σ± and σz are the emitter (de-)excitation
and population operators.
36
|ek i |Ei
δq,k ∆A
γ 00 Ω γ 00
|0k i gk γ 1D k, γ 0 |gi gk γ 1D A, γ 0
|1k i |f i
Figure 4.1: Left panel: Level structure of the qubit quantum dot, where only the
transition between |ei and |1i couple to the waveguide with a decay rate of γ 1D . This
transition also includes a nonradiative decay with γ 0 . Another nonradiative decay
processes with a rate of γ 00 between |ei and |0i is also shown. Right Panel: Level
structure of the auxiliary quantum dot and the transitions driven by the weak laser
Ω and by the coupling to the waveguide gk . Only |f i couples to the waveguide. The
nonradiative decay rate of γ 00 will be assumed to be neglible in this work.
The (de-)excitation and population operators for the emitters satisfy the commutation relation:
h i
(n) (n)
σ̂± , σ̂z(m) = ∓2σ̂± δnm (4.9)
h i
(n) (m)
σ̂+ , σ̂− = ±σ̂z(n) δnm . (4.10)
(n) †

(n) (n) (n)
where σ̂+ = |ein hg| , σ̂− = σ̂+ and σ̂z = |ein he| − |gin hg|. The annihilation and creation
operators for the photons satisfy:
h i
ak , a†k0 = δk,k0 (4.11)
h i
[ak , ak ] = a†k , a†k = 0. (4.12)
The treatment is simplified by changing to the interaction picture using the unitary operator
Ô (t) = e−iĤS t/~ , which reduces the Schrödinger equation to:
d
i~ |ψ (t)i = H̃I |ψ (t)i , (4.13)
dt
where the interaction Hamiltonian is given by:
3 Xh i
(n) (n)
X
H̃I = ~ gk σ̂+ e−i∆k t âk + H.C. (4.14)
n=1 k
X
S+ e−∆k t âk + H.C. .

=~ (4.15)
k
37
(n) (n)
Here ∆k = ω0 − ωk . It will be assumed in this treatment that the transition frequency is
(n) (m)
the same for the three emitters, such that ∆k = ∆k . An arbitrary state of the system is
expanded in terms of the basis states:
X
|ψ (t)i = cf,k (t) |f f f i ⊗ |{k}i + (ce,1 (t) |ef f i + ce,2 (t) |f ef i + ce,3 (t) |f f ei) ⊗ |{0}i ,
k
(4.16)
where |ef f i = |ei1 ⊗ |f i2 ⊗ |f i3 is a short hand notation for the tensor product between the
three emitter states. |{k}i is the state describing a single photon with wavenumer k, while |{0}i
is the photon vacuum state.
4.2.2 Dynamics in the Interaction Picture

The right hand side of Eq. (4.13) can be written as:
X h (1) i 
(2) (3)
ĤI |ψ (t)i = gn |ef f i + gk |f ef i + gk |f f ei cf,k (t) e−i∆k t ⊗ |{0}i 


k (4.17)
(1) ∗ (2) ∗ (3) ∗
h i
i∆k t

+ gk ce,1 (t) + gk ce,2 (t) + gk ce,3 (t) e |f f f i ⊗ |{k}i

It is now possible to find equations of motion for the expansion coefficients in Eq. (4.16) by
projecting the basis states onto Eq. (4.17):
3 X
(n) ∗
X
ċf,k = −i gk ce,n (t) ei∆k t (4.18)
n=1 k
(n)
ċe,n = −gk cf,k (t) e−i∆k t , n = 1, 2, 3. (4.19)
The equations of motion for ce,n (t) can be decoupled from cf,k (t) by formally integrating
Eq. (4.18), using the initial condition cf,k (t = 0) = 0:
3 X Zt
(n) ∗
X h 0
i
dt0 ce,n t0 ei∆k t .

cf,k (t) = −i gk (4.20)
n=1 k 0
Substituting this expression into Eq. (4.19) gives an equation of motion for the expansion coef-
ficient for the n’th excited state:
3 Zt
(m) ∗
h 0
i
(n)
XX
dt0 ce,m t0 e−i∆k (t−t ) .

ċe,n = − gk gk (4.21)
k m=1 0
This term describes interactions between the two emitters that are mediated by photons prop-
agating in the medium. It is possible to write Eq. (4.21) in terms of theqlocal density of states
(n) (n) ωk
by introducing the expression for the coupling constants: gk = d 2~0 ed · uk (xn ) and
R∞
multiplying by 1 = dωδ (ω − ωk ). The LDOS is given by:
−∞
X 2
ed · uk (xm ) u∗k (xm ) · ed δ ω − ωkR,L =

ρ (xn , xm , ω, ed ) = ed Im [G (xn , xm ; ω)] ed ,
πω
k
(4.22)
38
and was introduced in Section (3.1). The LDOS has been written in terms of Green’s tensor
G (xn , xm ; ω) projected along the orientation of the dipole emitter [16]. Eq. (4.21) can then be
written in the form:
Z∞ Zt
d(n) d(m) 0
dt0 ce,m t0 e−i∆k (t−t )

ċe,n =− dωωρ (xn , xm , ω, ed ) (4.23)
2~0
−∞ 0
3 Z∞
X
dt0 ce,m t0 Ke xn , xm , t − t0 ,

=− (4.24)
m=1 0
where the memory kernel Ke (xn , xm , t − t0 ) has been introduced:

Z∞
0 d(n) d(m) 0
dωe−i∆k (t−t ) ωρ (xn , xm , ω, ed )

Ke x n , xm , t − t = (4.25)
2~0
−∞
This function expresses how previous times t0 can contribute to cne (t). Eq. (4.23) is a crucial
result that relates the dynamics quantum emitter to the photonic environment for an arbitrary
homogeneous, dielectric medium. The imaginary part of Green’s tensor determines the LDOS,
which is the sole contributor to the dynamics of the emitter.
4.2.3 Wigner-Weisskopf Theory for a W1 PhC Waveguide

Wigner-Weisskopf theory constitutes a simple, special case of the formalism outlined above,
0
where ωρ (xn , xm , ω, ed ) is assumed to change slowly compared to e−i∆k (t−t ) . The memory
kernel then becomes:
Z∞
0 d(n) d(m) 0
dωe−i∆k (t−t )

Ke xn , xm , t − t ≈ ω0 ρ (xn , xm , ω0 , ed ) (4.26)
2~0
−∞
πd(n) d(m)
ω0 ρ (xn , xm , ω0 , ed ) δ t − t0

= (4.27)
0 ~
= γnm δ t − t0

(4.28)
The singular behaviour of the memory kernel implies that the reservoir of radiation modes is
memoryless. This is also know as the Markoff approximation. In addition, the decay rate γnm
to the radiation reservoir has been defined as:
πd(n) d(m) d(n) d(m)
γnm ≡ ω0 ρ (xn , xm , ω0 , ed ) = 2 ed · Im [G (xn , xm , ω0 , ed )] · ed . (4.29)
0 ~ 0 ~
The time integral in Eq. (4.24) can then be evaluated:
3 Z ∞
X
dt0 ce,m t0 Ke xn , xm , t − t0

ċe,n = − (4.30)
m=1 0
3
X Z∞
dt0 ce,m t0 δ t − t0

=− γnm (4.31)
m=1 0
3
X γnm
=− ce,m (t) , (4.32)
2
m=1
39
where γnn is the standard decay rate of the n’th emitter into a continuum of radiation modes,
while γnm , n 6= m is an additional decay that arises from the dipole-dipole interaction and is
mediated by the radiation reservoir. Both of these decay rates depend only on the magnitude of
the dipole moment dn,m and on the photonic environment through the LDOS. Both of the prop-
erties can be readily tailored for a self-assembled quantum for and photonic crystals, respectively.
This formalism can be specialized to the case where the dielectric material is a W1 waveg-
uide, as those discussed in Section (3.1) using the procedure from [24]. The basis functions for
a single band in the dispersion relation of the photonic waveguide are Bloch modes of the form:
r
a
uk (x) = bk (x) eikx , (4.33)
L
where a is the lattice constant, L is the length of the waveguide and bk (x) is a function that
is periodic in the direction of the waveguide. It is often a sufficient assumption to only consider
a single band due to the frequency spacing of the three bands being larger than the transition
energy between the conduction band and the lowest valence band. Green’s tensor can then be
expressed as:
X u∗k (xm ) · uk (xn )
G (xn , xm ; ω) = ω2 (4.34)
ωk2 − ω 2
k
Z
a ω
≈ dωk bk (xn ) b∗k (xm ) eik(xn −xm ) , (4.35)
4πLvg ωk − ω − i
1
dk and where ωk2 − ω 2 − i ≈
P R
where the sum has been converted into an integral by = 2π
k
(ωk − ω − i)2 ≈ 2ω (ωk − ω − i) when ωk is near the pole ωk ∼ ω. The slowdown of light is
described by the group index ng and the group velocity vg = dωk /dk = cng . The small parameter
has been introduced to keep the integral convergent, which can be solved by considering the
two different cases of xn − xm < 0 and xn − xm > 0 yielding:
iaω

GW1 (xn , xm ; ω) = bk (xn ) b∗k (xm ) eik(xn −xm ) Θ (xn − xm ) 

2vg 
(4.36)
iaω ∗ −ik(xn −xm )
+ b (xn ) bk (xm ) e Θ (xm − xn )


2vg k
which defines the forward- and backward propagating modes through the Heaviside function Θ.
The LDOS can then be written as:
( )
a bk (xn ) b∗k (xm ) Θ (xn − xm )
ρW1 (xn , xm , ω, ed ) = cos [k (xn − xm )] · (4.37)
2πvg +b∗k (xn ) bk (xm ) Θ (xm − xn )
This equation can be rewritten into


fR (xn , xm ) Θ (xn − xm ) 


 p 
 
a Veff,R ε (xn ) ε (xm ) 

ρW1 (xn , xm , ω, ed ) = cos [k (xn − xm )] · , (4.38)
2πvg  fL (xn , xm ) Θ (xm − xn ) 
+

 p 

Veff,L ε (xn ) ε (xm )

where
ε (xn ) ε (xm )bk (xn ) b∗k (xm ) Veff,R [e∗d · ek (xn ) · e∗k (xm ) · ed ]
p
fR (xn , xm ) = (4.39)
40
is a dimensionless function describing the spatial mismatch between the emitter field maximum
and that of the waveguide mode. The subscripts R and L refer to the forwards- and backwards
propagating mode, respectively. It additionally takes the alignment of the dipole with the Bloch
∗
hp ed and ek (xm ). The effective
mode into account through the product of the unit vectors i mode
−1
volume per unit cell is defined through Veff,R = max ε (xn ) ε (xm )bk (xn ) bk (xm ) for the
forward propagating mode. The equivalent definition for the mode volume for the backward
propagating mode is then straightforward to write down. The LDOS for a diagonal element
takes the simpler form:
a f (xn )
ρW1 (xn , xn , ω, ed ) = |ek (xn ) · ed |2 , (4.40)
2πvg VEff ε (xn )
where the alignment of the dipole and waveguide Bloch mode unit vectors have been explicitly
extracted from f (xn ).
The decay rates for radiative decay into the two guides modes can then be calculated:
nm πd(n) d(m)
γR,L = ω0 ρW 1 (xn , xm , ω0 , ed ) (4.41)
0 ~
ad(n) d(m)
= ω0 cos [k (xn − xm )]R,L (4.42)
20 ~vg
⇒ γ nm = γ0nm cos [k (xn − xm )] [Θ (xn − xm ) + Θ (xm − xn )] , (4.43)
where the sum over Bloch mode functions have been omitted for brevity. The dissipative single
dipole decay rate γ nn has a constant value set by the magnitude of the dipole moment, while the
decay rate from dipole-dipole interactions display spatial oscillations with a maximum amplitude
of
ad(n) d(m)
γ0nm = ω0 . (4.44)
20 ~vg
The oscillations can be seen as an effect of interference between the radiation fields and will be
discussed in further detail in Section (4.2.4).
Eqs. (4.38) and (4.43) reveal the primary methods for enhancing the light-matter though the
W1 waveguide: the LDOS increases with small mode volumes and for slow light. Both of
these requirements are met for photonic waveguides, as was discussed in Section (3.1.2). The
enhancement of the optical LDOS will then enhance the spontaneous decay rate into the preferred
mode for both the forwards and backwards propagating mode. This is quantified by the Purcell
factor by taking the ratio of the LDOS for one of the modes and the LDOS for a homogeneous
medium and was introduced in Section (3.1.2). It is printed here again for convenience:
3 λ2 /n2

FP (ω) = ng (ω) . (4.45)
4πn Veff /a
This enhancement is shown in Fig. (3.3). This also show the enhancement of the β−factor for
a single quantum dot in a W1 waveguide:
P nm
γ
n,m
β= P , (4.46)
γ nm + γng + γ 0
n,m
where γng is the decay rate into non-guided modes and γ 0 is the nonradiative decay rate. The
sum over the various γ nm gives the total decay rate into the guided modes. It is possible to obtain
41
values exceeding β > 98% [1] for the case of a single quantum dot in a waveguide, demonstrating
that these systems provides a promising platform for implementing single photon sources. The
β-factor is primarily limited by the nonradiative decay processes with γ 0 . The decay rates into
the non-guided modes will thus be set to zero in this thesis. The cooperativity C of the system
is defined as the ratio between the decay rate γ 1D (in the case of a single quantum dot in the
system) into the guided mode and the nonradiative decay:
γ 1D
C≡ , (4.47)
γ0
and can be related to the β-factor by
1
C= 1 . (4.48)
1− β
4.2.4 Superradiance
It is now possible to complete the equations of motion in Eq. (4.32) by summing over the
three emitters. This is done for the case where the emitters are embedded in a W1 waveguide
introduced in the precious section. The decay dynamics can be expressed as the following matrix
equation for the expansion coefficients of the excited states ce,n of the three emitters:
   
ċe,1 ce,1 (t)
   
ċe,2  = Ĥint · ce,2 (t) (4.49)
   
   
ċe,3 ce,3 (t)
   
γ 11 (x1 , x1 ) γL12 (x1 , x2 ) γL13 (x1 , x3 ) ce,1 (t)
1   
·
 21 22 23
= γR c (t) , (4.50)
  
(x2 , x1 ) γ (x2 , x2 ) γL (x2 , x3 )
2   e,2 
 
γR31 (x , x ) γ 32 (x , x ) γ 33 (x , x ) ce,3 (t)
3 1 R 3 2 3 3
where the matrix is represented by Ĥint . This thesis considers two distinct cases: the first is
when all matrix elements in Eq. (4.50) are equal, whereas the second is where two of the emitters
are equal (say emitter 2 and 3) and different from the third. These cases can be related to the
eigenvalues of the matrix, as described below. It is also assumed that the each emitter’s self-
(nn) (nn)
interaction couples identically to the two modes (i.e. γR = γL ). This matrix can then be
written as (while also absorbing the factor of 1/2 into the matrix elements:)
 
γA γAq γAq
 
(1)
Ĥint = γAq γq γqq  , (4.51)
 
 
γAq γqq γq
where γA and γq are the decay rates of the single auxiliary emitter and γq is the decay rate of
either of the two qubit emitters. γAq and γqq is the two-body decay rates between the auxiliary
and qubit emitters; and between the two qubit emitters, respectively. Diagonalization of the
matrix in Eq. (4.51) yields the three following eigenvalues:
γA + γq + γqq 1
q
γ± = ± (γA − γq − γqq )2 + 8γAq
2 (4.52)
2 2
γ0 = γq − γqq , (4.53)
42
with the corresponding eigenvectors:
1
|ψ± i = √ (|ef f i + |f ef i ± |f f ei) (4.54)
3
|ψi0 = N (A+ |ef f i + A− |f ef i) , (4.55)
where the coefficients A± for |ψ0 i are given at the end of this section. Notably, the eigenvalue γ+
is linked to the eigenstate that is a symmetric superposition of the excited states, while γ− and
γ0 are associated with the two non-symmetric superpositions. N is a normalization constant.
Further inside is gained by considering the ideal limit where all decay rates are equal: γ nm = γA .
The eigenstates and eigenvalues are then given by:
1
γ+ = 3γA , |ψ+ i = √ (|ef f i + |f ef i + |f f ei) (4.56)
3
1
γ− = 0, |ψ− i = (−2 |ef f i + |f ef i + |f f ei) (4.57)
2
1
γ0 = 0, |ψ0 i = √ (|ef f i − |f ef i) . (4.58)
2
The effect of super- and subradiance is evident from Eqs. (4.52) through (4.58): It is possible
to project the excited states into either the symmetric state or the two anti-symmetric states
by varying the optical properties of the emitters. This can primarily be done in two distinct
manners: the decay rate of the three emitters γA and γq can tailored, typically by varying the
gate voltage over the heterostructure, see Section(3.2.2). In addition, the value of the two-body
decay rates γAq and γqq can be changed by varying the distance between the three emitters. The
last effect induces interference between the radiation emitted from the quantum dots with an
oscillation set by γnm ∼ cos [k (xn − xm )]. This has a frequency ωk = vg k set by the guided mode
and amplitude by the relative values of the magnitude of the dipole moments of the auxiliary
and qubit emitters.
The coefficients for the eigenvectors are given by:
2γAq
A± = q (4.59)
−γA +γq +γqq
2 ± 12 (γA − γq − γqq )2 + 8γAq
2
4.3 Full Time Evolution

The discussion presented in Section (4.2) shows a simplified version of the full dynamics by
considering the coupling between a number of two-level emitters to a single electromagnetic
mode. The method used is cumbersome to apply when the external drive and nonradiative
decay is included and a more detailed method is needed. The Hamiltonian in Eqs. (4.1) and
(4.2) has been written in the rotating frame defined by the unitary operator Ô (t) = Ô2 (t) Ô1 (t),
where

Ô1 (t) = exp −iĤg t (4.60)
 
2
X
Ô2 (t) = exp i (∆A |Ei hE|) + δnq |einq he| , (4.61)
nq =1
43
where the sum runs over the two qubit emitters. The rotated Hamiltonian can then be written
as:
H̃ = Ĥe + Ṽ + H̃int (t) (4.62)
3
X Ω
Ĥe = ∆n |ein he| , V̂ = (|EiA hg| + |gi hE|) (4.63)
n
2
3 X h i
(n) (n)
X
H̃int (t) = gk σ̂+ ak ei(∆n −ωk )t + H.C. , (4.64)
n=1 k
where the sum is over all three emitters. Note that ∆n can take the form ∆A or δnq depending
on whether the terms in the sum is for the auxiliary- or on e of the qubit emitters, respectively.
The nonradiative decay γ 0 is governed by the Lindblad superoperator
3
γ 0 X h (n) (n) (n) (n) (n) (n)
i
L [ρ] = − ρσ̂+ σ̂− + σ̂+ σ̂− ρ − 2σ̂− ρσ̂+ . (4.65)
2
n=1
The commutation relations for the emitters are written here again for convenience:
h i
(n) (n)
σ̂± , σ̂z(m) = ∓2σ̂± δnm (4.66)
h i
(n) (m)
σ̂+ , σ̂− = ±σ̂z(n) δnm . (4.67)
and the commutation relation for the electromagnetic creation- and annihilation operators are:
h i
ak , a†k0 = δk,k0 (4.68)
h i
[ak , ak ] = a†k , a†k = 0. (4.69)
The equation of motion for the annihilation operator (and by extension, the creation operator)
can be found from Heisenberg’s equation of motion:
h i 3
(n)
(gkn )∗ σ̂− e−i∆n,k t ,
X
ȧk = i Ĥ, ak = −i (4.70)
n=1
(n) (n)
where ∆n,k = ∆n − ωk . The equations of motion for the coherences σ̂+ and population σ̂z
can be found in a similar manner:
(n) (n)
(gkn )∗ a†k σ̂z(n) e−i∆n,k t
X
σ̇+ = −2i∆n σ̂+ − i (4.71)
k
Xh (n)
i
σ̇z(n) = −2i gkn σ̂+ ak ei∆n,k t + H.C. (4.72)
k
These equations can be simplified by eliminating the field operators. This is done by formally
integrating Eq. (4.70) and substituting the result into Eqs. (4.71) through (4.72):
3 X Zt
(n) ∗
h 0
i
(n) (n) (m) (m)
X
dt0 σ̂+ t0 ei(∆m t −∆n t) σ̂z(n)

σ̇+ = 2i∆n σ̂+ − gk gk (4.73)
m=1 k 0
Z∞ Zt " 3
#
(n) (m)
X
0 0 0
σ̂z(n)

= 2i∆n σ̂+ − dω dt σ̂+ t Ke xn , xn , t − t (4.74)
−∞ 0 m=1
(4.75)
44
 
 3 Z∞ Zt
X (n)
h (m) 
i
dt0 Ke xn , xn , t − t0 σ̂− t0 
 
σ̂+ dω

 

 

 
 m=1 −∞ 0

σ̇z(n) = −2 , (4.76)
 3 Z∞ Zt h 
 X (m) i (n) 
dt0 Ke∗ xn , xn , t − t0 σ̂+ t0 σ̂− 
 



 + dω 


 m=1 
−∞ 0
where the memory kernel Ke (xn , xn , t − t0 ) has been introduced as in Section (4.2.3).
Note that there is no contribution from the freely propagating input field due to the choice of
rotating frame. These equations show that the evolution of the coherences and populations are
non-local, i.e. the evolution at time t depend on the value of the coherences and populations
at an earlier time t0 and is defined by the form of memory kernel. The system will be assumed
to be Markovian in the rest of the thesis (i.e. dωKe (t − t0 ) ∼ δ (t − t0 )) in order to simplify
R
the treatment but non-Markovian processes can be prevalent and should be considered in future
treatments.
The evolution described by Eqs. (4.74) and (4.3) are governed by the Hamiltonian and Lindblad
superoperator in the Markovian approximation:
3 3
X ΩδnA X (n) (m)
Ĥ = ∆n |Ein hE| + [|Ein hg| + |gin hE|] + J nm σ̂+ σ̂− (4.77)
2
n=1 n,m=1
3
X γ0δ n,m + γ nm h
(n) (m) (n) (m) (n) (m)
i
L [ρ] = ρσ̂+ σ̂− + σ̂+ σ̂− ρ − 2σ̂− ρσ̂+ , (4.78)
2
n,m=1
where the coherent and dissipative dipole-dipole interaction are given by
1 ∗ ad(n) d(m)
J nm = dn · Re [G (xn , xm ; ω0 )] · dm ∝ − ω0 sin [k (xn − xm )] (4.79)
~ε0 2~ε0 vg
2 ∗ ad(n) d(m)
γ nm = dn · Im [G (xn , xm ; ω0 )] · dm ∝ ω0 cos [k (xn − xm )] , (4.80)
~ε0 ~ε0 vg
respectively. The exact expression for Green’s function for a W1 waveguide in a photonic crystal
is given by Eq. (4.36) and includes the coupling to the forwards- and backwards propagating
guided modes. The evolution can be described by the non-hermitian Hamiltonian ĤNH in-
troduced in Section (2.1.4) if the jump terms in L [ρ] can be neglected. The non-hermitian
Hamiltonian is then given by [9]:
3
X γ 0 δn,m + iJ nm − γ nm (n) (m)
ĤNH = Ĥe − i σ̂+ σ̂− (4.81)
2
n,m=1
3
X γ 0 δn,m + γ0nm e−ik(xn −xm ) (n) (m)
= Ĥe − i σ̂+ σ̂− , (4.82)
2
n,m=1
where γ0nm is equal to the maximum magnitude of γ nm , as defined in Eq. (4.44) and where Ĥe is
the same as in Eq. (4.63). The exponential factor in the interaction term in Eq. (4.82) describes
how an emitter at site n picks up a change in phase upon being excited by a photon emitted into
45
the waveguide by an emitter at site m. This change in phase depends on the distance travelled
by the photon and is dominated by dissipative dipole-dipole interactions for xn − xm = nπ while
it is dominated by coherent interactions for xn − xm = kπ/2 for k ∈ N.
The non-hermitian Hamiltonian in Eq. (4.82) provides a sufficient description of the dynamics of
the presented system for deriving the proposed controlled phase gate. It is, however, necessary
to use another approximation to reduce the complexity of the equation of motion that can be
derived from this Hamiltonian. This is done by performing adiabatic elimination of the excited
states of the system. The formal framework for this method is presented in Appendix (A). The
derivation of the success probability and fidelity of the CZ gate is presented in Chapter (5) for
the ideal case where the matrix elements γ nm are equal, while Chapter (6) treats the case where
the matrix elements can take different values as the dipole moments and positions for the qubit
emitters are varied.
46
Chapter 5
Ideal CZ-Gate
This chapter describes the implementation of the controlled phase gate proposed in [3] in a
nanophotonic environment in the ideal case and how to judge its performance. Section (5.1)
discusses the preliminary considerations, while Section (5.2) provides the description of the gate
operation and the derivation of the success probability and the fidelity of the gate. This section
also discusses how to optimise the fidelity. Lastly, Section (5.3) provides the criteria needed to
ensure that the adiabatic treatment remains valid.
5.1 Preliminaries
The system described in Chapter (4) can be used to implement a controlled phase gate for the
two-qubit state |ψi. This state, and the ideal output state |Ψi, have the forms
1
|ψi = |00iq + |01iq + |10iq + |11iq (5.1)
2
1
|Ψi = |00iq + |01iq + |10iq − |11iq , (5.2)
2
and can be implemented using the |1i = |S = 1/2, MS = 1/2i and |0i = |S = 1/2, MS = −1/2i
electron ground states of the trion conduction band as the qubit states, as described in Sec-
tions (2.2.2) and (3.3). A third emitter quantum dot is included in this description; this is
the auxiliary emitter and is used to drive the system and to implement an integrated error
detection. The auxiliary emitter is also implemented using the ground states of the trion state:
|f i = |S = 1/2, MS = 1/2i and |gi = |S = 1/2, MS = −1/2i.
The level scheme is shown in Fig. (4.1) and the time evolution of |ψi is governed by the non-
hermitian Hamiltonian ĤNH that takes both coherent and dissipative evolution of the quantum
states into account. The non-hermitian Hamiltonian derived in Section (4.3) is for the case
where both qubit emitters are coupled to the waveguide.
h11E| he1f | h1ef |

 
i 1D + γ 0 − 2i γAq
1D − 2i γAq
1D

|11Ei ∆A − 2 γA
 
− 2i γAq
1D i
γq1D + γ0 − 2i γqq
1D

ĤNH = |e1f i  δq1 − (5.3)

n=2  2 1 
− 2i γAq
1D − 2i γqq
1D i
γq1D 0

|1ef i δq1 − 2 2
+γ
The braket notation on the outside of the matrix indicates how to write the matrix in the Dirac
notation. These kets describes the combined state of the tqo qubit emitters and the auxiliary
47
quantum dot: |ψiq1 ⊗ |ψiq2 ⊗ |ψiA It is possible that only one of the qubit emitters is coupled
to the waveguide. The non-hermitian Hamiltonian for each of these cases is:
 h10E| he0f | 
i 1D 0 − 2i γAq
1D

|10Ei ∆A − γA
2 +γ
ĤNH = (5.4)
 
n=1,q1 |e0f i − 2i γAq
1D δq1 − 2i γq1D
1
+ γ0
 h01E| h0ef | 
i 1D γ0 − 2i γAq
1D

|01Ei ∆A − γA
2 +
ĤNH = (5.5)
 
n=1,q2 |0ef i − 2i γAq
1D δq2 − i
2 γq1D
2
+ γ0
These two Hamiltonians may be different from each other, depending on the values of the
parameters of the quantum dot. The non-Hermitian takes the following form, when none of the
qubits interact with the guided mode:

i 1D 0

ĤNH = ∆A − γ +γ |00Ei h00E| (5.6)

n=0 2 A
These Hamiltonians can be combined into one single Hamiltonian by:

 
 ĤNH n=2
0


 
Ĥ
 
NH
n=1,q1
 
 
(5.7)
 
ĤNH
 
 n=1,q2 
 
0
 
 
 
ĤNH

n=1
Here Γ0A,q = γA,q + γ 0 . γA,q is the single-emitter decay rate for the auxiliary and qubit emitters
while γAq and γqq are the two-emitter decay rates between the auxiliary- and qubit emitters,
and between the two qubit emitters respectively. These rates are for decay into the guided
mode in the waveguide, while γ 0 is the decay rate for nonradiative decay. Only nonradiative
decay processes along the optical transition is considered in the following. ∆A and δq1 are the
detuning of the auxiliary and qubit emitters. The evolution of the system can be simplified
using adiabatic elimination, which results in the effective master equation with a set of effective
operators [18]:
h i 1 X † † †
k l k l k l
ρ̇ = −i Ĥeff , ρ (t) − L̂eff L̂eff ρ (t) + ρ (t) L̂eff L̂eff − 2L̂eff ρ (t) L̂eff (5.8)
2
k,l
†
1 −1

−1
Ĥeff = − V̂− ĤNH + ĤNH V̂+ + Ĥg (5.9)
2
−1
L̂keff = L̂k ĤNH V̂+ . (5.10)
This chapter will describe the mathematical foundation for the implementation of the controlled
phase gate and will quantify the effectiveness of the implementation through the success prob-
ability PS and the Fidelity F of the gate, while Chapter (6) will discus the effect of lifting the
approximations that lead to the ideal case.
48
5.2 Ideal CZ Gate
The ideal case involves three main approximations:
• The quantum dots are placed exactly in the regions of maximal constructive interference
• The coupling constants of the three quantum dots are equal
• The resonance frequency of the three dots are equal
The spatial phase factors can then be set to 1 and all decay rates can be set to γ 1D using
these three approximations. This leads to the simple case discussed in this chapter and greatly
simplifies the mathematical treatment. In addition, it is assumed that nonradiative decay pro-
cesses are suppressed along the |EiA ↔ |fiA transition for the auxiliary emitter and along the
|eiqk ↔ |0iqk transitions for the qubit emitters. The approximation also sets the two single-qubit

non-hermitian Hamiltonians to be equal: ĤNH = ĤNH . In addition, the detuning

n=1,q1 n=1,q2
of the qubits will be set to zero: δq,1 = δ1,2 = 0. The effective operators for the four different
blocks will be derived separately in the following section.
5.2.1 Effective Operators

The matrix representation of the non-hermitian Hamiltonian is given by Eq. (5.7) and is used to
derive the effective operators for the system. More details about the calculations can be found
in Appendix. (B). This results in the following effective operators, where the sum is over the
number n of qubits in the ground state:
2
X Ω2 ∆A (nC + 1)2
Ĥeff = − |giA hg| ⊗ P̂n (5.11)
γ0 γ0
2
n=0 2 ∆γA0 (nC + 1)2 + ([n + 1] C + 1)2
2
X
= ∆n |giA hg| ⊗ P̂n (5.12)
n=0
X2
√ Ω 2 ∆γA0 (nC + 1) + i ([n + 1] C + 1)
L̂A
eff = C + 1 √ 0 (nC + 1) 2 |fiA hg| ⊗ P̂n (5.13)
n=0
γ 2 ∆A (nC + 1)2 + ([n + 1] C + 1)2
γ0
2
X
= rA,n |fiA hg| ⊗ P̂n (5.14)
n=0
q
X2 √ Ω 2 ∆γA0 (nC + 1) + i ([n + 1] C + 1)
L̂effj = − C√ 0C 2 |fiA hg| ⊗ P̂n (5.15)
n=1
γ 2 ∆A (nC + 1)2 + ([n + 1] C + 1)2
γ0
2
X
= rqj ,n |fiA hg| ⊗ P̂n (5.16)
n=1
where rqj ,0 = 0, C ≡ γ 1D /γ 0 is the cooperativity and P̂n are projection operators that project
onto the qubit ground states. The only dynamical change stems from Ĥeff , which project the
auxiliary quantum dot into |giA and introduces an effective AC Stark shift of the ground states,
while the parameters rA,n and rqj ,n describe the effective decay processes. Free evolution of the
system while applying the weak laser field V̂ then imprints a phase shift of ∆n on the state
|giA ⊗ P̂n . This dynamical phase shift can be used to implement a controlled phase gate with
49
the truth table from Tab. (2.1) by, naively, applying the laser field for a time of tg = |∆π2 | . The
implementation the CZ gate will be examined more closely in Sec. (5.2.3).
The only effect of any effective decay process is to project the auxiliary emitter into |fiA ⊗ P̂n ,
as can be read off from Eqs. (5.13) and (5.15). These errors are detectable through measuring
the state of the auxiliary emitter using the methods mentioned discussed in Section (2.2.5) after
the gate operation has been implemented. The effects of these decay processes on the fidelity
can be removed by conditioning the gate on measuring the auxiliary emitter in |giA , though
they will still have a detrimental effect on the success probability. This, and additional errors,
will be discussed in further detail in Section (5.2.3).
Expressions for the success probability and fidelity can be derived by solving the effective master
equation and comparing the ideal density matrix ρI = |Ψi hΨ| to the one describing the actual
output state ρq (tg ) = |ψ (tg )iq hψ (tg )|. The success probability can be found by tracing over
the density matrix for the qubits at time tg :
2
X n o
PS = Tr {ρq (tg )} = Tr ρq (tg ) P̂n , (5.17)
n=0
and the fidelity is found by projecting the ideal output state |Ψi onto ρ̃ (tg ) (tg ) and normalizing
with PS :
1
F = hg| hΨ| ρ̃ (tg ) ρq (tg ) |Ψi |giA . (5.18)
PS A
Both the fidelity and success probability in conditioned upon measuring the auxiliary emitter in
|giA . The expressions for the success probability and the fidelity are derived in Section (5.2.3)
while the density matrix ρq (t) is found by solving the effective master equation.
5.2.2 Solving the Effective Master Equation

The effective master equation can be solved by plugging the coefficients from Eqs. (5.11) through
(5.15) into Eq. (5.8) and then project states onto the equation to obtain the equations of motion
for the sixteen matrix elements of ρ (t). The gate operation is only concerned with the evolution
of the qubit states, so the density matrix used in the following in the reduced density matrix
found by tracing over the auxiliary emitter:
ρ = TrA {ρFull } , (5.19)
where ρFull = |Ψiq |ψiA q hΨ| A hψ|. The jump terms in the effective master equation are all
proportional to the population of the excited state; these states are emptied by the initialisation
procedure, making possible to neglect the evolution imposed by the jump terms in the following
treatment.
The details of solving the effective master equation can be found in Appendix. (D). Luckily, all
the differential equations are decoupled, thereby making the system solvable. The equation of
motion for any matrix element ρxy (t) is given by:
1 n o
ρ̇xy = i (∆y − ∆x ) ρxy (t) − (Γx + Γy ) ρxy (t) , x= 00 , 01 , 10 , 11 , (5.20)
2 n=0 n=1 n=1 n=2
where the notation underneath each element in the definition of x and y indicates the number
of qubits in |1iq . This differential equation has the straightforward solution:
1
ρyx (t) = ρyx (t = 0) ei(∆y −∆x )t e− 2 (Γx +Γy )t . (5.21)
50
Here Γn denote the decay rate of the combined system of three emitters when n qubits are in
state |1i. These decay rates are defined by the coefficients of the dissipative terms in the effective
master equation and take the following form:
2
Γ0 ≡ r0A (5.22)
2 2 ∗ ∗
Γ1 ≡ r1A + |r1q | + r1A (r1q ) + r1q r1A
2 2 ∗
= r1A + |r1q | + r1A (r1q ) (5.23)
2 2 ∗ ∗
Γ2 ≡ r2A + 4 |r2q | + 2r2A (r2q ) + 2r2q r2A
2 2 ∗
= r2A + 4 |r2q | + 4r2A (r2q ) , (5.24)
as the cross terms for each n are equal in the ideal case. These three expressions can be collected
together as:
2
Γn = rnA + n2 |rnq |2 + 2n · rnA (rnq )∗ . (5.25)
It can be readily seen from Eq. (5.21) that the decay rate of the four diagonal elements is given by
Eqs. (5.22) through (5.24). The non-diagonal elements have a time-dependent change in phase
and a decay rate that is constructed from the decay rates of the diagonal elements. Plugging
in the expressions for the Lindblad coefficients from Eqs. (5.13) and (5.15)
p yields the
√ following
expressions for the decay rates (assuming that the detuning is ∆E ∝ γ γ = Cγ 0 ). The
1D 0
three different terms in Eqs. (5.23) and (5.24) are given by, where the C 1 limit is done for
the terms in the square root of Eq. (5.30) for brevity:
A 2 Ω2
rn>0 = (C + 1) (nC + 1)2
(5.26)
γ0
2
2 ∆γA0(nC + 1)2 + ([n + 1] C + 1)2
Ω2 n2 C 3 + 2n + n2 C 2 + (1 + 2n) C + 1

= 0 2 (5.27)
γ
2 ∆γA0 (nC + 1)2 + ([n + 1] C + 1)2
q Ω2 C3
r = (5.28)
n>0
γ0
2
2 2
2 ∆γA0
(nC + 1) + ([n + 1] C + 1)
p
A q ∗ Ω2 (C + 1) CC (nC + 1)
rn>0 rn>0 = − 0 2 (5.29)
γ
2 ∆γA0 (nC + 1)2 + ([n + 1] C + 1)2
Ω 2 nC 3 + C 2
∼
= − 0 2 . (5.30)
C1 γ
2 ∆γA0 (nC + 1)2 + ([n + 1] C + 1)2
51
The three decay rates can then be calculated:
Ω2 n2 C 2 + (1 + 2n) C + 1
Γn>0 = (5.31)
γ0
2
2 ∆γA0 (nC + 1)2 + ([n + 1] C + 1)2
Ω2 n2 C 2
→ (5.32)
C1 γ 0
2
∆A
4n2 γ0 C2
Ω2 1
= 0 2 (5.33)
γ
4 ∆γA0
Ω2 (C + 1)
Γ0 = (5.34)
γ0
2
2 ∆γA0 + [C + 1]2
Ω2 1
→ . (5.35)
C1 γ 0 C
It is seen from Eq. (5.33) that the decay rate for a system comprised of the auxiliary quantum
dot coupled to either one of or both the qubits dots depends on the detuning of the weak
perturbative driving field. Note that the effective decay rate is the same for any n > 0. The
choice of detuning has a pronounced effect on the fidelity and success probability of the gate
operation and is the most direct manner of reducing errors. The decay rate for n = 2 and n = 1
are equal in the ideal case.
5.2.3 Effective Phase Evolution for C 1

The coefficients for the effective Hamiltonian is given by Eq. (5.11) and can be greatly simplified
by considering the limit where γ 1D γ 0 ⇒ C 1 and by only include the dominant term in
the denominator. Choosing the same form of detuning as in Section. (5.2.2) causes the first term
in the denominator to scale as C 3 for n > 0. The energy shift can then be expressed as:
Ω2 (∆A /γ 0 ) (nC + 1)2

∆n>0 → −
C1 γ 0 (∆A /γ 0 )2 (nC + 1)2
Ω2 1
=− (5.36)
γ 0 4 (∆A /γ 0 )
Note that the induced energy shift is the same for any n > 0. The second term in the denominator
dominates for n = 0:
Ω2 (∆A /γ 0 )
∆0 = − (5.37)
γ 0 (C + 1)2
The scaling causes ∆0 to vanish significantly faster than ∆n>0 , which ensures that the ground
state remain unshifted under any gate operation. Choosing a gate time of tg = |∆πn | will not
enable the implementation of a controlled phase gate, as the remaining three qubit states acquires
an identical phase shift. This can be mitigated by employing the following single-qubit rotations
on all four states at the end of the gate operation (each vertical line of arrows represents the
effect of a single qubit rotation. The first line of arrows thus represent the rotation that removed
the dynamical change of phase from |00iw , while the second i for removing the phase change
52
from |10iq and |01iq ):
e−i∆0 t |00iq → |00iq

e−i∆1 t |01iq → e−i(∆1 −∆0 )t |01iq → |01iq
e−i∆1 t |10iq → e−i(∆1 −∆0 )t |10iq → |10iq
e−i∆2 t |11iq → e−i(∆2 −∆0 )t |11iq → e−i(∆2 −2∆1 +∆0 )t |11iq .
Choosing a gate time of

√
π 4πγ 0 C
tg = ≈ (5.38)
|∆2 − 2∆1 + ∆0 | Ω2
thus ensures the correct phase evolution of state |11iq while keeping the three remaining terms
unshifted, thereby completing the gate operation for the controlled phase gate. The success of
the gate is, as explained above, conditioned on measuring the auxiliary quantum dot in state
|giA . The success probability PS can then be calculated as a trace over the density matrix ρq
describing the two qubits:
2
X n o X2 n o
PS = Tr {ρq (tg )} = Tr ρn (tg ) P̂n = Tr e−Γn tg ρn (t = 0) P̂n , (5.39)
n=0 n=0
where the projection operator has been explicitly extracted to keep track of the effects from
either both, only one or none qubits in the ground state. It will be assumed that the initial
values of all the matrix elements in the density operator is ρxy (t = 0) = 14 . The detuning is
√
initially chosen to take the value ∆A = Cγ 0 . The energy shifts and decay rates are then given
by:
Ω2 1 Ω2 1
∆0 = − Γ0 = (5.40)
γ 0 3/2 γ0 C
Ω2 1 Ω2 1
∆n>0 =− 0 √ Γn>0 = 0 , (5.41)
γ C γ 4C
where Γ1 and Γ2 will both be denoted as Γ2 in the following. The success probability defined in
Eq. (5.17) can then be calculated:
1
PS = (exp [−Γ0 tg ] + 3 exp [−Γ2 tg ]) (5.42)
4
1 4π π
= exp − √ + 3 exp − √ (5.43)
4 C C
7π
≈1− √ (5.44)
4 C
The fidelity of the gate is defined as the projection of the ideal output state |Ψi projected onto
the actual output density matrix ρ̃q (tg ) after the gate operation and is normalized by the success
probability:
1
F = hg| hΨ| ρ̃q (tg ) |Ψi |giA , (5.45)
PS A
The density operator in Eqs. (5.42) and (5.45) is the solution to the effective master equation
and its matrix elements are given by Eq. (5.21).
53
The fidelity is at an arbitrary time t given by Eq. (6.15) and will simplify to the following
expression in the simple, ideal case considered here:
F0 (t) + 2F1 (t) + F2 (t)
F (t) = 1 , (5.46)
4 (exp [−Γ0 t] + 2 exp [−Γ2 t])
where the functions Fn (t) are shorthand functions that collect the contribution to the fidelity
from the matrix elements of ρ (t). The subscript n correspond to the contribution when to when
n qubits are in |1iq :
1 −Γ0 t 1 0 1

F0 (t) = e + 2e− 2 (Γ0 +Γ2 )t − e−i∆2 t e− 2 (Γ0 +Γ2 )t (5.47)
16
1 − 1 (Γ2 +Γ0 )t 0

F1 (t) = e 2 + 2e−Γ2 t − e−i∆2 t e−Γ2 t (5.48)
16
1 i∆02 t − 1 (Γ2 +Γ0 )t 0

F2 (t) = −e e 2 − 2ei∆2 t e−Γ2 t + e−Γ2 t . (5.49)
16
Note that the term 2F1 (t) in Eq. (5.46) will be split into two terms in the general case due to
possible difference between the decay rates of either of the qubits. Adding the function above
then gives:
h i
e−Γ0 t + 5e−Γ2 t + 4e− 21 (Γ0 +Γ2 )t − 2e−i∆02 t 2e−Γ2 t + e− 21 (Γ0 +Γ2 )t
1
F (tg ) = (5.50)
4 e−Γ0 t + 3e−Γ2 t
Here ∆02 = ∆2 − ∆1,1 − ∆1,2 is the phase shift of the |11iq state after the single-qubit rotations
have been performed. All the complex exponential functions evaluate as −1 at the chosen gate
time tg = π/ |∆02 |. The fidelity simplifies to the following expression in the ideal case:
1
1 e−Γ0 tg + 6e− 2 (Γ0 +Γ2 )tg + 9e−Γ2 tg
F (tg ) = (5.51)
4 e−Γ0 tg + 3e−Γ2 tg
1
!
1 1 + e 2 (Γ0 −Γ2 )tg
= 1+3 1 (5.52)
4 cosh 2 (Γ0 − Γ2 ) tg + e 2 (Γ0 −Γ2 )tg
1
" 2 #!
1 1 1
≈ 1+3 1− (Γ0 − Γ2 ) tg (5.53)
4 4 2
27π 2 1
=1− . (5.54)
64 C
√
The success probability and the fidelity of the gate with detunings ∆E = Cγ 0 and δ1 = δ2 = 0
have thus been found as:
7π 1
PS (tg ) = 1 − √ (5.55)
4 C
27π 2 1
F (tg ) = 1 − . (5.56)
64 C
The effect of these errors on the success probability and fidelity are shown in Fig. (5.1), demon-
strating the importance of being able to obtain a high cooperativity to overcome the influence
of the various errors that can occur during the gate operation.
These errors can be explained by the different values for Γ0 and Γn>0 . This changes the relative
weight of the qubit states in Eq. (5.50) leading to the errors in the fidelity of order O C −1
derived above. Additional errors will arise from higher order corrections from the energy shifts
54
0.9
0.8 1.00
0.7
Succes Probability PS
0.6 0.95
Fidelity F
0.5
0.4
0.90
0.3 PS, Not Optimised
PS, Optimised
0.2 F, Not Optimised
F, Optimised 0.85
0.1 1
10 102 103
Cooperativity C
Figure 5.1: Plot of the probability (solid lines) and the fidelity (dashed lines)of √the
0
√ non-optimised case with ∆A = γ C
CZ gate as a function of cooperativity. Both the
(blue lines) and the optimised case ∆A = γ 0 C/2 (green lines) are included. It is
clearly seen that the errors due to the uneven weights of the qubit basis states can
be removed with a proper choice of detuning, at the cost of lowering the success
probability.
√
∆n . The energy shift used above all scale as ∆n>0 ∼ Ω2 / 4γ C while the higher order cor-
rections are of order O Ω2 /C 3/2 . This leadsto an uneven phase evolution dependent on the

numberof qubits in state |1i of order O C −1 . These errors in the fidelity will then scales as
O C −2 .
Any errors for the gate operation that stem from the effective decay processes through Eqs. (5.13)
and (5.15) are, as described in Section (5.2.1), detectable upon measuring the state of the aux-
iliary quantum dot. These errors will not affect the fidelity if the gate is conditioned upon
measuring the auxiliary emitter in |giA but the success probability will be negatively affected,
as the system is projected into some erroneous state. Any gate operation that has failed in this
manner will be detected by the measurement on the auxiliary emitter. Discarding the present
gate operation and repeating the process until a successful measurement is performed makes it
possible to produce the desired output state with a high fidelity. The errors will thus only affect
the bit rate through the sub-unity probability.
The dominant source of errors for the gate operation thus comes from the difference in the
rate Γn of detectable errors, due to the O C −1 scaling of the infidelity. These errors can be

mitigated by standardizing the decay rates such that Γ0 = Γ1 = Γ2 ≡ Γ. This can be done by
choosing a specific form of the detuning for the auxiliary emitter ∆A while keeping δ1 and δ2 at
zero. The success probability is then given by:
55
PS (tg ) = e−Γtg (5.57)
(5.58)
It is then easily from Eq. (5.50) that this leads to the optimised case where the fidelity becomes:
4e−Γtg
F (tg ) = = 1. (5.59)
e−Γtg
√
Setting ∆A = γ 0 C/2 is sufficient to eliminate the errors due to the different relative weights
of the qubit states. The standardized decay rate and gate time are in this case given by:
Ω2 1 γ0 √
Γ0 = Γn>0 = , tg = 2π C (5.60)
γ0 C Ω2
with the success probability and fidelity:

2π
−√ 2π
PS (tg ) = e−Γ0 tg = e C ≈1− √ (5.61)
C
F (tg ) = 1. (5.62)
The optimised success probability and fidelity are included in Fig (5.1). Note that the this choice
of detuning results in a lower success probability than in the non optimised case, leading to a
lower bit rate for the gate operation than in the non-optimised case.
The dynamics of the ground states of the three quantum dots, that are described by the effective
operators in Eqs. (5.11) through (5.15), completely mimic the dynamics of the three atoms in
Borregaard et al. in [3]. In addition, the behaviour of the success probability and fidelity of
the CZ gate follows the same dependence on the cooperativity for the non-optimising choice
of detuning and it is possible to optimise the fidelity in a manner similar to that employed in
Borregaard’s proposal with the same integrated error detection scheme. A brief discussion about
the advantages and disadvantages of both systems is presented at the end of Chapter (6) when
the effects of lifting the approximations listed in Section (5.2) have been examined.
5.3 Criteria for the Driving Field

5.3.1 Satisfying the Adiabatic Condition
It is necessary to put certain constraints of the drive field in order for the adiabatic elimination
to be valid. This can be satisfied by setting ∆2n /Ω2 1 [3], giving the following constraint on
Ω:
√
Ω 4π∆A = 2πγ 0 C, . (5.63)
√
using ∆A = γ 0 C/2. This condition can be satisfied for a driving of the form
√
Ω ∼ a2πγ 0 C, a 1, (5.64)
where the value of a sets the size of the errors induced from non-adiabatic effects. Numerical
treatment of the master equation has shown that setting a = 0.25 is sufficient to reduce the
errors induced by the adiabatic treatment.
In addition to this, the effect of the driving when no qubits needs to be taken into consideration.
56
In this case, the driving field and the coupling to the waveguide creates a Raman transition
between |giA and |fiA governed by the Hamiltonian:

gk |EiA hf| ak + |fiA hE| a†k
X
Ĥ = ∆0 |EiA hE| + Ω (|EiA hg| + |giA hE|) + (5.65)
k
" # " #
gk |fiA a†k
X X
= ∆0 |EiA hE| + |EiA Ω A hg| + gk A hf|ak + Ω |giA + A hE|. (5.66)
k k
The excited state couple to the bright state |ψiB = N (Ω |giA + g |fiA ) but adiabatic evolution
makes it possible for the system to enter the dark state:
!
X
|ψiD = N gk |g, 0iA − Ω |f, 1iA (5.67)
k
= cos θ |g, 0iA − sin θ |f, 1iA , (5.68)
with the mixing angle

√
Ω a2πγ 0 C
θ=P = P , (5.69)
gk gk
k k
where the number in the state vectors represent the number of photons in the waveguide. The
dark state is an eigenstate of the Hamiltonian in Eq. (5.65) with an eigenvalue of zero. It is
possible to project thus emitter into |ψiD during the adiabatic evolution, unless the mixing angle
is sufficiently small to eliminate |fiA from |ψiD :
Ω X
θ=P 1⇒Ω gk (5.70)
gk k
k
√
aγ 0 C
⇒ P 1. (5.71)
gk
k
This condition can be met by suppressing the nonradiative decay processes(preferentially only
to the guided mode) and by choosing a low value for a.
5.3.2 Effects of Dephasing

Dephasing constitutes another potential source of errors by letting the coherences decay with-
out affecting the population of the qubit states. This effect can readily be incorporated into
the formalism for the CZ gate by phenomenologically including a dephasing rate γdp into the
equations of motion for the off-diagonal elements of the density matrix:
1
ρ̇xy = i (∆y − ∆x ) ρxy (t) − (Γx + Γy ) ρxy (t) − γdp ρxy (t) , n 6= m. (5.72)
2
This can be readily solved:
1
ρyx (t) = ρyx (t = 0) ei(∆y −∆x )t e− 2 (Γx +Γy +γdp )t , x 6= y. (5.73)
57
1.00
0.95
0.90
0.85
0.80
Fidelity F
0.75
0.70
0.65 γdp/γ = 1.0

0.60 γdp/γ = 0.1
γdp/γ = 0.01
0.55 1
10 102 103
Cooperativity C
Figure 5.2: The fidelity plotted against the cooeprativity in the presence of pure
dephasing processes. The strong dependence on the dephasing rate compared to the
nonradiative decay rate is observed.
This does not have an effect on the success probability, but dephasing processes have a substantial
the fidelity, in the optimized case:
1 e−Γt + 3e−(Γ+γdp )t
F = (5.74)
4 e−Γt
1 3
= + e−γdp t (5.75)
4 4 " √ #
1 3 2πγ 0 γdp C
= + exp − . (5.76)
t=tg 4 4 Ω2
Eq. (5.76) shows that dephasing can lead to a fidelity of F = 1/4 as C 1. The detrimental
−1
effect of dephasing can be neglected if the dephasing time tdp = γdp is longer than the gate
time. This gives the following constraint:
tg γdp 1 (5.77)
γdp 1
⇒ √ 1, (5.78)
2πγ 0 a2 C
when using Eq. (5.64) for the drive field Ω in order to reduced non-adiabatic errors. Eq. (5.78)
shows that there exists a trade off between the non-adiabatic errors and the errors induced by
pure dephasing processes. Reducing the non-adiabatic errors will increase the dephasing errors
and vice versa. This behaviour is shown in Fig. (5.2). It may be possible to choose a more
intricate drive in order to reduce both of these detrimental effects but this is beyond the scope
of this work.
58
It is thus possible to implement a controlled two qubit phase gate by performing the single-qubit
rotations at the end of a driving pulse with a duration of tg = π/ |∆02 | on the auxiliary emitter.
√
Choosing a detuning between laser frequency and the transition frequency of ∆A = γ 0 C en-
sures that the gate operation can be performed with a unity fidelity and a sub-unity probability
set by the cooperativity.
Conditioning the gate operation on measuring the auxiliary emitter in |giA will eliminate the ef-
fects of dissipative errors due to the high fidelity. Any other error will only have a negative effect
on the bit rate, due to having to repeat the gate operation until a desired outcome is obtained by
reinitializing the system upon detecting a failed gate operation. This integrated error detection
removes the need for using more complicated techniques, such as entanglement purification and
quantum error correction, and may be directly implemented in quantum circuits. In addition it
is possible to completely eliminate the disruptive effect of dephasing by satisfying the criteria
outlined in Section (5.3).
Additional errors will be introduced when the approximations outlined in Section (5.2) are re-
laxed. These, and the possibilities for mitigating them, are discussed in Chapter (6).
59
60
Chapter 6
Non-Ideal CZ Gate
Chapter (5) has introduced the possibility of implementing a controlled phase gate using the
superradiative effect of the dipole-dipole interaction between three coupled quantum dots. In
addition, a number of errors has been identified and mitigated by conditioning the gate on
measuring the auxiliary emitter in state |giA and by standardizing the decay rates. A number
of assumptions has been made in order to simplify the treatment:
• The detuning δq of the two qubits were set to zero
• The three emitters were assumed to have the same coupling constant gk
• Three three emitters were placed with a perfect displacement, such that no subradient
effects were present.
The last two assumptions will generally not hold for self-assembled quantum dots due to the
random nature of the growth process and will reintroduce errors to the relative weights of the
qubit states. The effect of these errors will be discussed in this chapter and the possibility of
mitigating them by letting δq be non-zero will be explored. This is done in Sections (6.1) and
(6.2), while Section (6.3) discusses the detrimental effects of dephasing processes and the criteria
that has to be fulfilled in order to reduce errors from non-adiabatic effects. In this chapter, the
ideal case will refer to the case explored in Chapter (5) while the optimized case will refer to
any system where it is possible to remove part of the infidelity by changing the detuning of the
emitters.
6.1 Variable Qubit Dipole Moment

Each quantum dot in the chain can have a dipole moment that is different from that of the other
two emitters due to variations in the vertical and in-plane composition of the quantum dot. A
simple way of treating this is to parametrise the magnitude of the dipole moments of the two
qubits in terms of the one from the auxiliary emitter:
dq1 = κdA , dq2 = ηdA , (6.1)
leading to the following non-hermitian Hamiltonian for the case of both qubits interacting with
the waveguide:
h11E| he1f | h1ef |
 
i 1D + γ 0 − 2i κγA
1D − 2i ηγA
1D

|11Ei ∆A − 2 γA
− 2i κγA
1D i 1D + γ 0
κ2 γA − 2i κηγA
1D

ĤNH = |e1f i  δq − (6.2)
 
2

n=2
− 2i ηγA
1D − 2i κηγA
1D i
η 2 γA
1D γ0

|1ef i δq − 2 +
61
where γA1D is the decay rate in to the guided mode for the auxiliary quantum dot and the detuning
of the two qubit emitters has been set to the same value of δq . It is straightforward to write up the
non-Hermitian for the case where n = 1 and n = 0. Note that the two 2×2 matrices representing
the ĤNH for n = 1 will now be different due to the different parametrization constants. This
splits Γ1 into two different decay rates, one for each qubit. The decay rates are in this general
case given by summing up the contributions from the various Lindblad coefficients from the
dissipative part of the effective master equation:
2
Γ0 ≡ r0A (6.3)
A 2 q 2 ∗
A q q A ∗

Γ1,1 ≡ r1,1 + r + r1,1
1,1 r1,1 + r1,1 r1,1 (6.4)
A 2 q 2 ∗
A q q A ∗

Γ1,2 ≡ r1,2 + r + r1,2
1,2 r1,2 + r1,2 r1,2 (6.5)
A 2 q 2 q 2 ∗ ∗ 
r2 + r + r + rq rq + r q q
2,2 r2,1

2,1 2,1 2,1 2,2

Γ2 ≡ ∗ ∗ ∗ ∗  (6.6)
+r2A rq 2,1 + r2A rq 2,1 + rq r2A + rq r2A 
2,1 2,2
where the letter in the subscript denotes whether the Lindblad coefficient comes from the Aux-
iliary or Qubit emitter, the first number denotes the number of qubits in |1iq and, in the case of
only one connected qubit, the second number determines which particular qubit is contributing.
The jump terms from the effective Hamiltonian does not contribute in this case due to the ini-
tialisation of the three quantum dots in the electronic ground states. The effective Hamiltonian,
Lindblad operators and decay rates now takes the form:
2
" #
Ω2 ˜ A δ̃ 2 + fH,n (κ, η) C 2 δ̃q + gH,n (κ, η) C 2 ∆
4∆ Ã
q
X
Ĥeff = ∆0 P̂0 − P̂ ⊗ |giA hg|
γ 0 4δ̃q2 C 2 + hn (κ, η) ∆ ˜ 2 C 2 + kn (κ, η) C 2 n
˜ A δ̃q C 2 + jn (κ, η) ∆
n=1 A
(6.7)
2
X
= ∆n |giA hg| ⊗ P̂n (6.8)
n=0
X2
L̂A
eff = rnA (C, κ, η) |fiA hE| ⊗ P̂n (6.9)
n=0
2
q q
X
L̂effj = rnj (C, κ, η) |fiA hE| ⊗ P̂n (6.10)
n=1
Ω2 4δ̃q2 C + fΓ,n (κ, η) C 3 + gΓ,n (κ, η) C 2

Γn>0 = (6.11)
γ 0 4δ̃q2 C 2 + hn (κ, η) ∆ ˜ 2 C 2 + kn (κ, η) C 2
˜ A δ̃q C 2 + jn (κ, η) ∆
A
Ω2 1
Γ0 = , (6.12)
γ0 C
where ∆0 = −Ω2 / γ 0 C −3/2 . The dimensionless detunings ∆ ˜ A = ∆A /γ 0 and δ̃q = δq /γ 0 has

been introduced to ease the notation. The end of Section (5.2.3) described errors in the fidelity
due to the omission of the O (C)−3/2 terms. It is necessary to include these terms in the expres-
sions above in order to include the effects of δq when optimising the decay rates.
Note that there is a separate term for each of the two qubits for n = 1 in the sum. The
various functions in Eqs. (6.7) and (6.11) are defined in Appendix (B.2) and depend only on the
parametrization parameters κ and η. It has been assumed that the two qubits are tuned in the
62
0.9 Solid line - PS, Dashed Line - F 1.06
0.8
1.04
0.7
0.6 1.02
Probability PS
0.5
Fidelity F
1.00
0.4
0.3 0.98
0.2
= 0.8, η = 0.8 0.96
0.1 = 1.0, η = 1.0
= 1.5, η = 1.5
0.0 1 0.94
10 102 103
Coorporativity C
Figure 6.1: The success probability (solid lines) and fidelity (dashed lines) plotted
against the cooperativity C for various values of κ = η. This has been plotted for
the case of complete optimisation from Eqs. (6.26) and (6.27), leading to a fidelity
of F = 1 for all three cases. Note the strong fall off in the success probability when
κ 6= 1.
same manner, δq1 = δq2 . The effective Hamiltonian and Lindblad operators above show that
the different values of κ and η only change the values of the energy shift and decay rates, while
leaving the essential dynamics of the system unchanged. The gate operation can thus be imple-
mented using the same procedure as described in Chapter (5) by letting the auxiliary emitter be
illuminated by a weak laser drive for a duration of tg = π/ |∆0A |. Here ∆0A = ∆2 −∆1,1 −∆1,2 +∆0
is the dynamical change in phase of the state |11i1 after the sequence of single-qubit rotations
detailed in chapter (5) has been applied at the end of the gate operation.
The variation in the coupling constants change the decay rates, which reintroduces the infidelity
from the different weights of the qubit state, as can be seen in the general expression for the
fidelity in Eq. (6.15). Fig. (6.2) shows that this leads to a rapid drop off in the fidelity and √
the
the unoptimised case where ∆A = γ 0 C
success probability for κ, η 6= 1. This is plotted for both √
and for the simple optimisation scheme where ∆A = γ 0 C/2 and δq = 0, highlighting that the
detunings derived in Chapter (5) are insufficient to reduce the errors in the non-ideal cases.
Success Probability & Fidelity

The success probability and fidelity are defined as in Chapter (5):
2
X n o X2 n o
PS = Tr {ρq (tg )} = Tr ρn (tg ) P̂n = Tr e−Γn tg ρn (t = 0) P̂n , (6.13)
n=0 n=0
1
F = hg| hΨ| ρ̃ (tg ) |Ψi |giA , (6.14)
PS A
63
0.7
0.6 Unoptimised
0.5
Simple Optimisation
Complete Optimisation
Probability PS
0.4
0.3
0.2
0.1
0.0
0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
1.0
0.8
Fidelity F
0.6
0.4
0.2
0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Figure 6.2: The success probability and fidelity plotted against the parametrization
constant
√ κ in three distinct cases. The first is the unoptimised case p where ∆A =
γ 0 C and δq√= 0. The second is the simple case where ∆A = γ 0 C − 1/2/2 and
δq = −7γ 0 /8 C leading to the fidelity being optimized only for the ideal case where
κ = 1. The last is the completely optimised case where ∆A and δq are given by
Eqs. (6.26) and (6.27), respectively. This optimises the fidelity for any value of κ and
leads to a slightly higher success probability. All cases are plotted for a cooperativity
of C = 100. Note the high success probability and the peaks in the fidelity for the
unoptimised case.
64
where |Ψi = |00iq + |01iq + |10iq − |11iq is the ideal output of a CZ gate, up to a normalization
and ρn (t = 0) = 1/4 is the initial value of the diagonal elements in the density matrix. The
success probability then becomes a sum over four exponential functions, one for each Γn , whereas
the fidelity at an arbitrary time t is given by:
F0 (t) + F1,1 (t) + F1,2 (t) + F2 (t)

F (t) = 1 . (6.15)
4 (exp [−Γ0 t] + exp [−Γ1,1 t] + exp [−Γ1,2 t] + exp [−Γ2 t])
The functions Fn (t) are shorthand functions that collect the contribution to the fidelity from
the n’th column from the density matrix:
F0 (t) = q h00|ρ (t) |00iq + q h00|ρ (t) |01iq + q h00|ρ (t) |10iq − q h00|ρ (t) |11iq (6.16)
1 −Γ0 t 1 1 0 1

= e + e− 2 (Γ0 +Γ1,1 )t + e− 2 (Γ0 +Γ1,2 )t − e−i∆2 t e− 2 (Γ0 +Γ2 )t (6.17)
16
F1,1 (t) = q h01|ρ (t) |00iq + q h01|ρ (t) |01iq + q h01|ρ (t) |10iq − q h01|ρ (t) |11iq (6.18)
1 − 1 (Γ1,1 +Γ0 )t 1 0 1

= e 2 + e−Γ1,1 t + e− 2 (Γ1,1 +Γ1,2 )t − e−i∆2 t e− 2 (Γ1,1 +Γ2 )t (6.19)
16
F1,2 (t) = q h10|ρ (t) |00iq + q h10|ρ (t) |01iq + q h10|ρ (t) |10iq − q h10|ρ (t) |11iq (6.20)
1 − 1 (Γ1,2 +Γ0 )t 1 0 1

= e 2 + e− 2 (Γ1,2 +Γ1,1 )t + e−Γ1,2 t − e−i∆2 t e− 2 (Γ1,2 +Γ2 )t (6.21)
16
F2 (t) = − q h11|ρ (t) |00iq − q h11|ρ (t) |01iq − q h11|ρ (t) |10iq + q h11|ρ (t) |11iq (6.22)
1 i∆02 t − 1 (Γ2 +Γ0 )t 0 1 0 1

= −e e 2 − ei∆2 t e− 2 (Γ2 +Γ1,1 )t − ei∆2 t e− 2 (Γ2 +Γ1,2 )t + e−Γ2 t , (6.23)
16
after having projected |giA onto the density matrix. It is, as in the ideal case, possible to
minimize this effect by choosing specific values for ∆A and δq . Consider the simplified case of
κ = η where the coupling constants of the two qubits are equal. The decay rates for n > 0 are
then:
Ω2 4δq2 C + n2 −κ7/2 + κ4 + κ3 C 3 + n2 κ4 + 2n −κ3/2 + κ2 C 2

Γn>0 = 0 (6.24)
γ 4C 2 [δq + κ2 ∆A ]2 + [nκ2 + 1]2 C 2
Ω2 1
Γ0 = . (6.25)
γ0 C
Note that, in this approximation, the decay rate for a single qubit in |1i1 is the same regardless
which qubit is connected to the waveguide. It is possible to set Γ2 and Γ1 equal to Γ0 by choosing
γ0
q
∆A = 2 −2κ7/2 + κ4 + κ3 C 2 + κ4 (C − 1) + 1/2 (6.26)
2κ
5/2 − κ4 C − 4κ5/2 − 3√κ

02 4 κ
δq = γ . (6.27)
16κ5/2 ∆A
This eliminates the errors in the fidelity that stems from the different weights of the four qubit
states due to the uneven decays. The effects on the fidelity and the success probability for the
controlled phase gate are shown in Fig. (6.1) for different values of κ = η. A unity fidelity is
observed for all three cases but the success probability experiences a significant drop off as the
dipole moments of the two qubits are varied. This heralds the emergence of the subradiant
effects discussed in Section (4.2.4), leading to a weaker interaction between the auxiliary emitter
and the two qubits. The subradiant effects are enhanced even further when the position of the
two qubits is varied in Section (6.2).
65
Optimised for = η = 1
1.4 0.54
0.48
1.2 0.42
0.36
1.0 0.30
η
0.24
0.8 0.18
0.12
0.6 0.06
0.6 0.8 1.0 1.2 1.4
Optimised for = η
1.4 0.54
0.48
1.2 0.42
0.36
1.0 0.30
η
0.24
0.8 0.18
0.12
0.6 0.06
0.6 0.8 1.0 1.2 1.4
Figure 6.3: Contours of the success probability against the parametrization param-
eters of the dipole moments of the qubit emitters. This is done for the simply and
completely optimised case, respectively. Fig. (6.2) is a cross section along the green
diagonal line where κ = η. Note the enhancement of the success probability for the
fully optimised case (lower panel) along the line where η = 1.
Setting κ = 1 in Eqs. (6.26) and (6.27) results in the following expressions
γ0 p

∆A = C − 1/2

2

(6.28)
7γ 0
δq = − √



8 C
which optimises the fidelity for the simple case discussed in Chapter (5) when the terms of order
−3/2

O C are included in the expressions for ∆n and Γn . This makes it possible to remove the
errors due to the uneven change in phase resulting from these higher order corrections and from
the different weights
of the qubit states simultaneously. Even higher corrections still exist; terms
of order O C −5/2 lead to errors in the fidelity of order O C −3 . More advanced methods a
required to remove these errors, but the effects of these errors can be neglected for any realistic
values of the cooperativity in W1 waveguides.
The detunings given by Eqs. (6.26) and (6.27) ensures that the fidelity will always be F = 1
66
when both qubits have the same dipole moment. They are not sufficient to keep this unity
fidelity in the general case where κ 6= η. It has not been possible to find a method to optimise
this case but there are several possible avenues for tackling this problem. These are discussed
in Chapter (7). Using the detunings in Eqs. (6.26) and (6.27) does result in a near unit fidelity
and high success probabilities for a wide range of values for κ and η. Fig. (6.3) and Fig. (6.4)
show contours of the success probability and fidelity, respectively, as a function of κ and η for
the fully and simply optimised cases. Both sets of detunings have a wide range of values for the
parametrization constants allowing for a near unity fidelity (albeit the range is larger for the
fully optimised case). The simply optimised case has a limit range of possible parametrization
values that result in a maximised success probability, in contrast to the higher range for the
fully optimised case.
There are three distinct choices of detunings with various effects on the fidelity and success
probability, as is shown by Fig. (6.2). These choices are:
√
1. Surprisingly, using the completely unoptimised detunings ∆A = γ 0 C and δq = 0 for
specific values of κ (κ ∼ 0.7 and κ ∼ 1.45) will lead to F ∼ 1 and a success probability
comparable to that obtained from the completely optimised case. Values between these
two has the highest success probability but with a high but sub-unity values for the fidelity.
2. The fully optimised case of Eqs. (6.26) and (6.27) that keeps a constant F = 1 for any
value of κ = η. The success probability rapidly falls off as the strength of the qubit
dipole moment is changed but this can be remedied by increasing the cooperativity C (see
Fig. (6.1)). This choice is also useful for the situation where κ 6= η with a high fidelity for
a wide range of values for the two parametrization constants. This is shown in Fig. (6.3).
It also has a larger freedom in the possible values κ and η that lead to a maximised success
probability, compared to the simply optimised case.
3. The detunings in Eq. (6.28) lead to F = 1 for κ = η = 1 with a rapid drop off in the
fidelity and a success probability that is strictly lower than for the completely optimised
case. This set of detunings result in a narrower range of values for κ and η in order to
obtain a high success probability, as is seen from Fig. (6.3).
67
Optimised for = η = 1
1.4 1.0
0.9
1.2 0.8
0.7
1.0 0.6
η
0.5
0.8 0.4
0.3
0.6 0.2
0.6 0.8 1.0 1.2 1.4
Optimised for = η
1.4 1.0
0.9
1.2 0.8
0.7
1.0 0.6
η
0.5
0.8 0.4
0.3
0.6 0.2
0.6 0.8 1.0 1.2 1.4
Figure 6.4: Contours of the fidelity against the parametrization parameters of the
dipole moments of the qubit emitters. This is done for the simply and completely
optimised case, respectively. Fig. (6.2) is a cross section along the green diagonal line
where κ = η. Note the wide range of values for κ and η with a unity or near-unit
fidelity in the optimised case (lower panel).
68
6.2 Variable Qubit Dipole Moment & Position
Quantum dots grown using the Stranski-Kastranov will in general be distributed randomly along
the photonic crystal, leading to an imperfect spatial separation between the three emitters. This
results in the introduction of destructive interference between the radiation fields for the three
emitters, thereby lowering the interaction between them. It can, in principle, be possible to map
the distribution of the quantum dots (e.g. using atomic-force micrography) in order to find three
quantum dots with a mutual separation that satisfies the condition for constructive interference.
Alternatively, it may be possible to build a bending waveguide between three randomly chosen
quantum dots, thereby tailoring the spatial separation. Both of these methods may be infeasible
to carry out in practise; the first relies on producing a lucky sample while the bending waveguide
may be impractical to etch in practise. Is is therefore important to examine the effects of an
imperfect position of the qubit quantum dots.
This section will discuss the negative effects of having three emitters embedded along a straight
line in a W1 waveguide with an imperfect separation. The difference between the position xk −xl
of any two emitters has to fulfil:
k (xk − xl ) = 2πm, m ∈ N, (6.29)
in order to obtain full constructive interference between the radiation fields. Here k is the
magnitude of the wave vector of the propagating photon. As a reminder, the total decay rate
due to the dipole-dipole interaction is given by summing the contribution from the coherent
(J nm from Eq. (4.79)) and dissipative (γ nm from Eq. (4.80)) interactions:
ad(k) d(l)
Γkl = J kl + iγ kl = γ0kl e−ik(xk −xl ) , γ0kl = ω0 , (6.30)
~ε0 vg
where the complex exponential function describes the change in phase of the k’th quantum dots
upon absorbing a photon emitted by the l’th quantum dot. If the requirement in Eq. (6.29)
is fulfilled, then the dissipative dipole-dipole has the only contribution to the dynamics of the
ground state. Varying the positions such the the requirement is broken will enhance the contri-
bution from the coherent dipole-dipole interaction, which will lower the success probability, as
will be shown below.
The origin will be placed at the position of the auxiliary quantum dot, while the positions of
each emitter will be written as:
xqk = x0,qk + δxqk , (6.31)
where x0,qk is the ideal position of the k’th quantum dot and δxqk is the displaced distance to the
actual position. This makes it possible to write the exponential functions in Eq. (6.30) in terms
of only the displacements δxqk . There are two cases pertaining to the position of he auxiliary
emitter, depending on whether it is placed at the end of the chain or in the middle between the
qubit emitters; both of these cases has been considered in the treatment below. In addition, the
dipole moment of the two qubits are assumed to be different from that of the auxiliary emitter
in the same manner as in Section (6):
dq1 = κdA , dq2 = ηdA . (6.32)
69
0.7
0.6
Unoptimised
Simple Optimisation
0.5 Complete Optimisation
Probability PS
0.4
0.3
0.2
0.1
0.0 π
−4 −π8 0 π
8
π
4
1.0
0.8
Fidelity F
0.6
0.4
0.2 π
−4 −π8 0 π
8
π
4
kδx
Figure 6.5: The success probability and fidelity plotted against the simultaneous
displacement δxq of both qubit quantum dots. This has been done with the detunings
from the unoptimised and the simply optimised case previously discussed and for the
completely optimised case from Eqs. (6.41) and (6.42). Note the unity fidelity for the
completely optimised case and the low success probabilities for kδx & π/8.
This leads to the following non hermitian Hamiltonian for n = 2:
h11E| he1f | h1ef |

 
i 1D + γ 0 − 2i κγA1D eikδq,1 − 2i ηγA
1D eikδq,2

|11Ei ∆A − γA 2
 
i 1D −ikδq,1 δq − 2i κ2 γA 1D + γ 0 − 2i κηγA1D eik(δxq,2 −δxq,1 ) 

ĤNH = |e1f i 
 − 2 κγA e

n=2 
1D e−ikδq,2
− 2i ηγA − 2i κηγA 1D eik(δxq,1 −δxq,2 ) δq − 2i η 2 γA 1D + γ 0

|1ef i
(6.33)
where both qubits have the same detuning δq . This results in the familiar effective Hamiltonian:
2
X
Ĥeff = ∆n |giA hg| ⊗ P̂n , (6.34)
n=0
with the coefficients
70
˜ A δ̃ 2 + bH,n (δxq,1 , δxq,2 , κ, η) C 3 + fH,n (κ, η) C 2 δ̃q + gH,n (κ, η) C 2 ∆
Ã

Ω2 4∆ q
∆n>0 = − 0
γ ˜ A δ̃q C 2 + jn (κ, η) ∆
4δ̃q2 C 2 + hn (κ, η) ∆ ˜ 2 C 2 + kn (κ, η) C 2
A
(6.35)
Ã
Ω2 ∆
∆0 = − 0 2 . (6.36)
γ C
Note that bH,n (C, δxq,1 = 0, δxq,2 = 0, κ, η) = 0, which recovers the Hamiltonian from Eq. (6.34),
while the remaining functions in Eq. (6.35) are the same as in Section (6.1) and can be found in
Appendix (B.2). The dynamics resulting from the effective Lindblad operators remain unchanged
but the Lindblad coefficients are modified due to the imperfect qubit positions:
2
X
L̂A
eff = rnA (C, δxq,1 , δxq,2 , κ, η) |fiA hE| ⊗ P̂n (6.37)
n=0
2
q q
X
L̂effj = rnj (C, δxq,1 , δxq,2 , κ, η) |fiA hE| ⊗ P̂n . (6.38)
n=1
Placing the auxiliary emitter in the central position in the chain results in the same expressions
as those listed above; the dynamics of the three emitters are thus unchanged when permuting the
position of the auxiliary emitter. The position of the auxiliary emitter can become important by
implementing the following modifications to the theory: this treatment assume that the dipole
moments couple with the same strength to the forwards- and backwards propagating photon
mode. The auxiliary quantum dot will only interact with one mode if it is placed at the end of
the chain, while it interacts with both modes when placed in the middle. This can give rise to
different dynamics, if its dipole moment couple differently to the two modes. The same holds
for the qubits, which may exhibit more complicated dynamics depending on whether they are
placed next to each other or are separated by the auxiliary emitter. Secondly, this treatment does
not take loss of the propagating photon into account. Such a loss (which can, for instance, be
caused if total internal reflection is not complete) may affect the dynamics differently, depending
the distance that the photon has to propagate, in order to mediate the interaction between the
two qubits. These two effects provide a natural extension for a further treatment of the proposal.
The imperfect qubit position will, just like the modified qubit dipole moments, only change
the magnitude of the Hamiltonian- and Lindblad coefficients, but will leave the essential dy-
namics of the interaction unchanged. The implementation of the CZ-gate then follows the same
form as in the previous sections. The modified decay rates are:
Ω2 4δ̃q2 C + fΓ,n (δxq,1 , δxq,2 , κ, η) C 3 + gΓ,n (δxq,1 , δxq,2 , κ, η) C 2

Γn>0 = (6.39)
γ 0 4δ̃q2 C 2 + hn (κ, η) ∆˜ A δ̃q C 2 + jn (κ, η) ∆
˜ 2 C 2 + kn (κ, η) C 2
A
Ω2 1
Γ0 = , (6.40)
γ0 C
where the functions fΓ,n (δxq,1 , δxq,2 , κ, η) and fΓ,n (δxq,1 , δxq,2 , κ, η) are listed in Appendix
(B.2).
The success probability and fidelity is calculated from Eq. (6.13) and (6.15), respectively, and are
greatly inhibited by varying the position of the two qubit emitters, as can be seen in Fig. (6.5)
for the different optimisation schemes employed. The success probability rapidly drops off for
all three optimisations and reaches a minimum value at kδxq,n ≈ π/8.
71
π PS, Simply Optimised
4
0.54
π 0.48
8 0.42
0.36
kδxq,2
0 0.30
0.24
−π8 0.18
0.12
0.06
−π4
π PS, Completely Optimised
4
0.54
π 0.48
8 0.42
0.36
kδxq,2
0 0.30
0.24
−π8 0.18
0.12
0.06
−π4 π
−4 −π8 0 π
8
π
4
δxq,1
Figure 6.6: Contour plot of the success probability against the displacement of the
two
0
√ show the simply optimised case with ∆A =
p qubit emitters. The top0 panel
γ C − 1/2/2 and δq = −7γ /8 C, whereas the bottom panel show the situation
for the completely optimised case from Eqs. (6.41) and (6.42). The green line show
the diagonal where δxq,1 = δxq,2 , along which the optimisation has been carried out.
The success probability remains high for displacements kδx . π/8.
It is possible to maximise the fidelity by setting:
γ0
q
∆A,n = 2 −2κ7/2 cos (kδxq,n ) + κ4 + κ3 C 2 + κ4 (C − 1) + 1/2 (6.41)
2κ
γ0
κ3/2 cos (kδxq,n ) − κ2 C + 4κ2 + 3

8 4
δq = q , (6.42)
κ3/2 sin (kδxq,n ) C 2 − −2κ7/2 cos (kδxq,n ) + κ4 + κ3 C 2 + κ4 (C − 1) + 1/2
which optimises the fidelity to F = 1 for the special case where both qubits are displaced by
δxq,n and share the same dipole moment through η = κ. The resulting success probability and
fidelity are shown in Figs. (6.6) and (6.7), respectively, for a optimisation along the displacement
of the first qubit and for the case where κ = η = 1. The fidelity is maximised along the diagonal
line of δxq,1 = δxq,2 and remains high within a displacement range of kδxq,n . π/8, but drops
off when the displacements are outside of this range. The success probability, on the other hand,
is only optimised along the chosen optimisation displacement of δxq,1 and has an unchanged and
limited range of kδxq,n . π/8 along the displacement of the other qubit. Evaluating Eqs. (6.41)
72
and (6.42) for δxq,2 will maximise the probability along the direction of the displacement of the
second qubit, allowing for the ability to tailor the bit rate of the gate, depending on the spatial
distribution of the quantum dots. Appendix (C) shows additional contour plots similar to those
presented in this section. These are drawn for various values of κ and η and shows that the
fidelity with the completely optimised detunings is very robust when the magnitude of the dipole
moment of the qubits is changed. This robustness does not carry over to the success probability,
which sees lower values for all the different sets of parametrisation parameters. The shape of
the success probability profile is also heavily dependent on the values of κ and η but the general
features remain the same: The probability is enhanced along the chosen qubit displacement for
the detunings. It is thus possible to rotate each contour plot of the success probability by ninety
degrees by evaluating ∆A and δq along δxq,2 , as is seen from Figs. (C.9) and (C.10). This makes
it possible to tailor the success probability depending on the exact distribution of the three
quantum dots.
Fig. (6.5) shows the cross section of the contour plots along the highlighted diagonal line where
δxq,1 = δxq,2 √for the aforementioned sets of detunings. This situation of the unoptimised case
with ∆A = γ 0 C and δq = 0 is also shown; This has the same general features as Fig. (6.2) for
the case where only the qubit dipole moments were changed: The unoptimised and the simply
optimised cases have a limited range with a unity or near-unity fidelity with a rapid drop-off in
the success probability for all three sets of detunings. The maximum obtainable success proba-
bility changes between the three sets depending on the magnitude of the qubit displacement.
Both the success probability and the fidelity can attain high values for displacement that limited
by range kδxq,n . π/8, with variations depending on the optimisation scheme (i.e. F is limited
for both displacements for the simple optimisation, but is only limited along the displacement
that is not optimised for the complete optimisation). This range corresponds to a maximal
allowed displacement of δxq,n ∼ 18nm for a refractive index of n = 3.5 and a wavelength of
λ = 1000nm for the emitted light from an InAs quantum dot [13]. The typical in-plane size
of quantum dots are typically in the range of 10 − 70nm, making the allowed displacement
small compared to the in-plane size of the quantum dots. This may even allow small struc-
tural variations in the quantum dots themselves to reduce the success probability and fidelity;
the treatment of these effects are beyond the scope of the work but can provide an interesting
staring point for further treatment.
73
π F, Simply Optimised
4
1.0
π
8 0.8
0.6
kδxq,2
0
0.4
−π8 0.2
0.0
−π4
π F, Completely Optimised
4
1.0
π
8 0.8
0.6
kδxq,2
0
0.4
−π8 0.2
0.0
−π4 π
−4 −π8 0 π
8
π
4
kδxq,1
Figure 6.7: Contour plot of the fidelity against the displacement of thep two qubit
0
emitters. The top √ panel show the simply optimised case with ∆A = γ C − 1/2/2
and δq = −7γ 0 /8 C, whereas the bottom panel show the situation for the completely
optimised case from Eqs. (6.41) and (6.42). The green line show the diagonal where
δxq,1 = δxq,2 , along which the optimisation has been carried out.
74
6.3 Conditions for the Adiabatic Treatment and Dephasing
The conditions for satisfying the validity of the adiabatic treatment and the suppression of
dephasing processes can be carried out for the simplified case where both qubits are have the
same dipole moment κ = η and are displaced by the same amount from their ideal positions
δxq,1 = δxq,2 ≡ δx. This derivation for a gate time tg written only to leading order in C, where
Ω2 1
∆n ∼ − , (6.43)
γ 0 4∆
Ã
leaving all the displacement- and κ dependence in the detuning ∆A of the auxiliary quantum
dot. Discarding the lowest order term in the argument for ∆A yields:
r
γ0C 7/2 cos (kδx) + κ4 + κ3 +
κ4
∆A ≈ −2κ , (6.44)
C1 2κ2 C
where the zeroth-order term has been discarded by rescaling the time scale. This results in a
gate time that grows linear with the cooperativity:
π γ0C
tg = = 2f −1 (δx, κ) π , (6.45)
|∆2 − ∆1,1 − ∆1,2 | Ω2
where f (δx, κ) is the constant of proportionality:

r
1 κ4
f (δx, κ) = 2 −2κ7/2 cos (kδx) + κ4 + κ3 + . (6.46)
2κ C
This is in contrast to the
√ gate time in the ideal case that scaled as the square root of the
cooperativity: tg,I = 2π C.
The new scaling of the gate time requires a new choice for the drive field Ω in order to satisfy the
adiabatic condition and to suppress the detrimental effect of dephasing processes. The adiabatic
condition needs to satisfy:
∆2n
1 ⇒ Ω 4∆A = 2f (δx, κ) γ 0 C, (6.47)
Ω2
which can be satisfied by choosing a drive of:
Ω ∼ 2πaf −1 (δx, κ) γ 0 C, a 4π. (6.48)
The value of a can be set to a = 0.25, as in Section (5.3.2) and is modified by the qubit
displacement δx the qubit dipole moment κ through the function f (δx, κ).
The effects of dephasing processes can be seen from Eq. (5.76): the maximum obtainable fidelity
is significantly reduced. The aforementioned equation is rewritten here for convenience:
1 3 −γdp tg
FOpt = + e , (6.49)
4 4
where γdp is the dephasing rate. These effects can be minimised by setting the gate time to be
−1
shorter than the dephasing time γdp tg :
γdp f (δx, κ)
1, (6.50)
4πγ 0 a2 C
75
1.00
0.98
0.96
Fidelity F
0.94
0.92
γdp/γ = 1.0, = 1.5
0.90 γdp/γ = 0.1, = 1.5
γdp/γ = 1.0, = 0.8
γdp/γ = 0.1, = 0.8
0.88 π
−4 −π8 0 π
8
π
4
k ·δx
Figure 6.8: The fidelity plotted against kδx for C = 100 for different values of κ and
γdp /γ 0 . Note how the dephasing induced errors decrease when the qubits are moved
away from the auxiliary emitter.
leading to the same trade off between reducing the non-adiabatic errors or reducing detrimental
effects from pure dephasing as in Section (5.3.2). This effect is shown in Fig. (6.8), with a
smaller error when the qubits are moved in the same direction and when the dipole moments of
the qubits are smaller than that of the emitter.
The detunings presented in Eqs. (6.41) and (6.42) gives the optimal choice due to the pos-
sibility to obtain a unit to near-unity fidelity for a wide range of values for the qubit dipole
moment and position. In addition, setting the driving field as in Eq. (6.48) ensures that non-
adiabatic errors are suppressed, even though this enhanced the impact on the gate operation
from pure dephasing processes. The conditions outlined in this section provide the means to
maximise the success probability of the gate through the simple optimisation schemes outlined
in this section. Possibilities for enhancing these optimisation schemes exist, such that it may
be possible to obtain a unity fidelity for all values of the aforementioned parameters. These
possibilities are listed in Chapter (7).
76
Chapter 7
Conclusion and Outlook
7.1 Conclusion
The main subject of this thesis has been to examine the possibilities of implementing the con-
trolled phase gate suggested by Borregaard et. al. [3] in a solid state environment. The basic
idea has been to use the strong dipole-dipole interaction between three self-assembled quantum
dots embedded in a nanophotonic waveguide. This has been done in order to control the dy-
namics of the three quantum emitters thereby producing the desired output state.
The dynamics of the physical system has been derived by adiabatically eliminating the excited
states of the system when the auxiliary emitter is illuminated by a weak laser, yielding effec-
tive operators for the evolution of the ground states. This resulted in an effective Hamiltonian
and a set of Lindblad operators describing the unitary and dissipative evolution, respectively.
The evolution of these operators results in a controlled phase gate functionally identical to that
proposed by Borregaard et al. This has been achieved by conditioning the gate upon measuring
a specific ground state in the auxiliary emitter, as this emitter can only enter this state upon
a successful gate operation. This eliminates detectable errors stemming from the dissipative
evolution of the system and highlights one of the main strengths of this proposal: the gate has
an integrated error detection, eliminating the need of implementing involved external quantum
error correction protocols.
The dynamics of the gate are described through the effective Hamiltonian. This thesis has shown
that this induces an effective AC Stark shift of the ground states of the system. This resulted in a
dynamical phase change of the ground states that was used to implement the proposed CZ gate,
when combined with a number of single qubit rotations. The success probability and fidelity of
the gate were derived in order to quantify the effectiveness of the gate; both of these have been
show to converge towards unity as the cooperativity of the system is increased. In addition, the
ability to attain unity fidelity by standardising the effective decay rates of the system has been
demonstrated. This standardisation results in the elimination of undetectable errors in the gate,
namely errors resulting in the qubit states being weighted differently in time. This has been
achieved by choosing a specific detuning of the laser field illuminating the auxiliary emitter. The
conditions that are needed to be fulfilled in order to reduce the effect of non-adiabatic evolution
has also been derived and the detrimental effects of pure dephasing processes has been described.
The effects of the emerging subradiance were examined by changing the physical parameters
of the qubits, namely the dipole moments and their positions. Re-standardising the decay rates
along the simple conditions of equal qubit dipole moments and positions has shown that this
77
proposal is reasonably robust against changes in both of the aforementioned parameters, with
the ability to attain a unity to near-unity fidelity for a wide range of qubit dipole moments
and positions. This robustness is limited to only the fidelity; the success probability undergoes
a large drop off as the aforementioned parameters are changed. This is most pronounced for
changes in the qubit position, demonstrating the large effects of the subradiant features of this
proposal.
7.2 Outlook
There are several possible avenues for a further investigations into the proposal presented in this
thesis. One could look into the permutation invariance. This can be implemented by letting the
various emitter dipole moments couple differently to the backwards- and forwards propagating
mode in the waveguide. In addition, the effects of photon loss in the waveguide may have an
effect on the quality of the gate. Both of these additions may lead to changes depending on
whether the auxiliary quantum dot is placed at the center or at either end of the quantum dot
chain and can be implemented through the non-hermitian Hamiltonians governing the system.
The resulting equations may be hard to treat analytically with the chosen laser scheme, leading
to the next point.
The results for the non-ideal case in this thesis are derived for a simplified case where the
two qubits are varied equally when compared to the auxiliary quantum dot. The effects of these
quasi-complete optimisations on the completely non-ideal case have then been discussed. The
main limitation for a full optimisation has, as previously mentioned, been computational com-
plexity in the chosen method. One way to remedy this is to implement a numerical treatment of
the completely non-ideal case in order to find a set of detunings that can standardise the decay
rates. Alternatively, an analytical result for the completely non-ideal case may be obtained by
introducing several perturbing lasers fields. One possible scheme is to use two lasers applied in
succession of each other, where the first is nearly on resonance with the first qubit, while the
second laser is nearly on resonance with the other qubit. This will result in separate effective
operators for the two laser fields; the complete description of the dynamics can be achieved by
a simple addition of these operators due to the nature of the formalised adiabatic elimination.
This may lead to the ability of tailoring the coupling between the auxiliary quantum dot and
either of the qubit dots independently, ultimately leading to a simplified scheme for standardis-
ing the decay rates.
Dephasing processes induce significant errors on the fidelity and are enhanced when the non-
adiabatic errors are eliminated. The trade off between dephasing and non-adiabatic errors puts
huge constraints on the usefulness of this proposal; it is then important to examine the possi-
bilities of mitigating, or even removing, these errors and on how to simultaneously reduce both
errors. Building upon this, a more complete description of dephasing can be carried out, as only
pure dephasing has been considered in this work. Dephasing effects from phonons constitute
are significant obstacle for experiment using quantum dots and photonic crystals. These effects
may have a strong and detrimental effect on the controlled phase gate of this thesis and should
thus de examined. In addition, the effects of dephasing of the excited states also need to be
examined due to their dependence of the driving of the auxiliary emitter.
78
Appendix A
Effective Operator Formalism
This section will provide a derivation of the effective master equation and operators introduced
in Section (2.1.4). The results and approach follows the procedure from [18]. The open system
considered in this section can be divided into two distinct subspaces, which are connected by
perturbative couplings. In addition, the system is considered to be Markovian. This makes is
possible to describe the evolution of the density operator ρ through a master equation of the
Lindblad form:
h i X 1 †
ρ̇ = −i Ĥ, ρ + L̂k ρL̂†k − L̂k L̂k ρ − ρL̂†k L̂k , (A.1)
2
k
where Ĥ is the Hamiltonian of the system and where each Lindblad operator L̂k represents a
source of decay. This can be reduced to an effective master equation by combining perturbation
theory and adiabatic elimination fo the excited states. This reduced master equation only
involves the ground state manifold:
h i X † 1 † †
k k k k k k
ρ̇ = −i Ĥeff , ρ + L̂eff ρ L̂eff − L̂eff L̂eff ρ − ρ L̂eff L̂eff , (A.2)
2
k
with the effective Hamiltonian and Lindblad operator:

†
1 −1

−1
Ĥeff = − V̂− ĤNH + ĤNH V̂+ + Ĥg (A.3)
2
−1
L̂keff = L̂k ĤNH V+ , (A.4)
where V+ (V− ) is the perturbative (de)excitation of the system and Ĥg is the ground state
Hamiltonian. ĤNH is the non-hermitian Hamiltonian of the quantum jump formalism:
iX †
ĤNH = Ĥe − L̂k L̂k , (A.5)
2
k
with Ĥe being the excited state Hamiltonian. This formalism can be extended to systems with
non-perturbative ground state Hamiltonians and containing several systems. The generalized
effective Hamiltonian and Lindblad operator:

1X (f,l) −1 (f,l)(t)

Ĥeff = − V̂− (t) ĤNH V+ + H.C. + Ĥg (A.6)
2
f,l
X (f,l) −1 (f,l)
L̂keff = L̂k ĤNH V+ (t) , (A.7)
f,l
79
where the sum over l runs over the different diagonal, non-perturbative ground states and f
runs over the different perturbations or coupling fields.
A.1 Derivation of the Effective Operators

This section focuses on the derivation of the effective Hamiltonians and Lindblad used through-
out this theses The derivation follows from [18].
A.1.1 Projection Operator Formalism

The projection operator formalism is used to structure the Hamiltonian into two separate sub-
spaces. The total Hilbert space can be separated into a subspace for the ground states and a
subspace for the excited state by the operators Pg and Pe , respectively. The projection operators
fulfil Pg + Pe = I and Pg Pe = 0. The Hamiltonian is divided by:
Ĥ = Ĥg + Ĥe + V̂+ + V̂− , (A.8)
where Ĥe = Pg ĤPg , Ĥe = Pe ĤPe , V+ = Pe ĤPg and V̂− = Pg ĤPe . Note that V̂+ = V̂−† and
that V̂ = V̂+ + V̂− .
The excited states decay to the stable ground states such that the Lindblad operators can be
written as L̂k = Pg L̂k Pe . There is no transition form the ground state manifold to the excited
state manifold from spontaneous emission, which can be represented by writing Pe L̂k Pg = 0.
A.1.2 Non-Hermitian Time Evolution in the Quantum Jump Picture

The unitary and dissipative dynamics can be combined into a singe non-hermitian Hamilto-
nian ĤN H in Eq. (A.5). This can be introduced into the master equation by noting that
P k l †
L̂
k eff L̂eff = 2i ĤNH − Ĥ e = 2i ĤNH + Ĥ g + V̂ − Ĥ . Substituting this into the mas-
ter equation given by Eq. (A.1) yields:
h i X † h i h i
ρ̇ = −i Ĥ, ρ + L̂keff ρ L̂keff − i ĤNH + Ĥg + V̂ − Ĥ ρ − ρ ĤNH + Ĥg + V̂ − Ĥ
k
(A.9)
X † h i h i
= L̂keff ρ L̂keff −i ĤNH + Ĥg + V̂ ρ − ρ ĤNH + Ĥg + V̂ . (A.10)
k
The dynamics of the decaying excited states are mainly governed by the non-Hermitian Hamil-
tonian ĤNH for the case where the ground state interactions Ĥg are much weaker than Ĥe .
A.1.3 Perturbation Theory in the Interaction Picture

The couplings between the subspaces V̂± are assumed to be sufficiently weak in order to be
described as perturbations of the evolution governed by the unperturbed Hamiltonian defined as
Ĥ0 = Ĥg + ĤNH . The perturbation theory is performed on the density matrix ρ in the interaction
picture introduced in Section (2.1.3). The change to the interaction picture is performed using
the operator:

Ô (t) = exp −iĤ0 t = exp −i Ĥg + ĤNH t . (A.11)
80
This makes the change to the interaction picture possible for the Hamiltonian Ĥ, the density
operator ρ and the Lindblad operator L̂keff . This is done using the transformed wavefunction
|ψ̃ (t)i, which is defined by:
|ψ (t)i = Ô (t) |ψ̃ (t)i . (A.12)
The time-dependent Schrödinger equation can then be transformed by:

d
i [|ψ (t)i] = Ĥ |ψ (t)i (A.13)
dt
d h i
⇒i Ô (t) |ψ̃ (t)i = Ĥ 0 Ô (t) |ψ̃ (t)i (A.14)
dt
d 0 d
⇒ iÔ (t) |ψ̃ (t)i = Ĥ Ô (t) − i Ô (t) |ψ̃ (t)i (A.15)
dt dt

d d
⇒ i |ψ̃ (t)i = Ô−1 (t) Ĥ 0 Ô (t) − iÔ−1 (t) Ô (t) |ψ̃ (t)i (A.16)
dt dt

d −1 0 d h
−1
i
⇒ i |ψ̃ (t)i = Ô (t) Ĥ Ô (t) + i Ô (t) Ô (t) |ψ̃ (t)i , (A.17)
dt dt
where Ĥ 0 = Ĥ0 + V̂ . The change from Eq. (A.16) to (A.17) has been done using the following
trick:
d1 d h −1 i d h −1 i d
0= = Ô (t) Ô (t) = Ô (t) Ô (t) + Ô−1 (t) Ô (t) (A.18)
dt dt dt dt
d d h i
⇒ Ô−1 (t) Ô (t) = − Ô−1 (t) Ô (t) , (A.19)
dt dt
where 1 is the identity operator. The transformed interaction Hamiltonian Ṽ (t) from Eq. (A.16)
is defined by:
d
Ṽ (t) = Ô−1 (t) Ĥ0 + V̂ Ô (t) − iÔ−1 (t) Ô (t) (A.20)
dt

= Ô−1 (t) Ĥ0 + V̂ Ô (t) − iÔ−1 (t) −iĤ0 Ô (t) (A.21)
= Ô−1 (t) V̂ Ô (t) . (A.22)
The Hermitian conjugate is easily calculated:

†
Ṽ † = Ô† (t) Ṽ Ô−1 (t) . (A.23)
Note that this interaction Hamiltonian is not Hermitian, due to the non-unitary nature of Ô (t).
The density operator ρ is transformed as:
ρ = |ψ (t)i hψ (t)| = Ô (t) |ψ̃ (t)i hψ̃ (t)| Ô† (t) = Ô (t) ρ̃ (t) Ô† (t) (A.24)
−1
⇒ ρ̃ (t) = Ô−1 (t) ρ Ô† (t) . (A.25)
The Lindblad operator has the same form as V̂− , so the transformation to the interaction picture
follows the same form as for V̂ :
L̃k (t) = Ô−1 (t) L̃k (t) Ô (t) (A.26)

†
L̃†k (t) = Ô† (t) L̃†k (t) Ô−1 (t) (A.27)
81
The time derivative of the transformed density matrix is given by:
−1
d −1

†
ρ̃˙ = −i Ô (t) ρ Ô (t) (A.28)
dt
−1 −1 −1
= Ȯ (t) ρ Ô† (t) + Ô−1 (t) ρ Ȯ† (t) + Ô−1 (t) ρ̇ Ô† (t) (A.29)
−1 −1
= iĤ0 Ô−1 (t) ρ Ô† (t) + Ô−1 (t) ρ −iĤ0† Ô† (t) (A.30)
" #
h i h i −1
†
X
−1
+ Ô (t) L̃k (t) ρL̃k (t) − i Ĥ0 + V̂ ρ − ρ Ĥ0 + V̂ Ô† (t) , (A.31)
k
where the derivatives in the first two terms in Eq. (A.29) have been explicitly carried out and
the full expression for the master equation from Eq. (A.1) have been substituted into the third
term. The Ĥ0 terms cancel, which leaves:
h i −1 −1
L̃k (t) ρL̃†k (t) Ô† (t)
X
ρ̃˙ = −iÔ−1 (t) V̂ ρ − ρV̂ Ô† (t) + Ô−1 (t) (A.32)
k
h i −1 −1
−1 −1 † †
= −iÔ (t) V̂ Ô (t) Ô (t) ρ − ρ Ô (t) Ô (t) V̂ Ô† (t) (A.33)
h i † −1
L̃k (t) Ô (t) Ô (t) ρ Ô (t) Ô (t) L̃†k (t) Ô† (t)
X
−1 −1 −1 †
+ Ô (t) (A.34)
k
h i X
= −i Ṽ (t) ρ̃ (t) − ρ̃ (t) Ṽ † (t) + L̃k (t) ρ̃ (t) L̃†k (t) , (A.35)
k
†
Eq. (A.32) by 1 = Ô (t) Ô−1 (t) = Ô† (t) Ô−1 (t) a number of times in order to simplify the
expression. These factors of 1 are shown in Eqs. (A.33) ans (A.34). This makes it possible to
identify the operators ρ̃, Ṽ (t) , Ṽ † (t) , L̃k (t) and L̃†k (t), which yields Eq. (A.35).
The perturbative expansion of ρ̃ is written in terms of the small parameter :

1 (0)
ρ̃ (t) ≈ = ρ̃ (t) + ρ̃(1) (t) + 2 ρ̃(2) (t) + . . . (A.36)
N
Plugging Eq. (A.36) into Eq. (A.35) results in a recursive expression for the reduced density
operator by using that Ṽ is a perturbative (de)excitation: Ṽ ∝ . This gives:
h i X
ρ̃˙ (n) (t) = −i Ṽ (t) ρ̃(n−1) (t) − ρ̃(n−1) (t) Ṽ † (t) + L̃k (t) ρ̃(n) (t) L̃†k (t) . (A.37)
k
The first three orders of the recursive reduced density operator is thus:
L̃k (t) ρ̃(0) (t) L̃†k (t)

X
ρ̃˙ (0) (t) = (A.38)
k
h i X
ρ̃˙ (1) (t) = −i Ṽ (t) ρ̃(0) (t) − ρ̃(0) (t) Ṽ † (t) + L̃k (t) ρ̃(1) (t) L̃†k (t) (A.39)
k
h i X
ρ̃˙ (2) (t) = −i Ṽ (t) ρ̃(1) (t) − ρ̃(1) (t) Ṽ † (t) + L̃k (t) ρ̃(2) (t) L̃†k (t) . (A.40)
k
Decay processes can be neglected for orders n ≤ 1 in the absence of initial excitations; the
zeroth order is trivially in the ground state and attains no change due to spontaneous emission
82
events. This ensures that the first-order terms only evolves according to the drive terms; this
evolution allows the higher order terms to have contributions from decay processes. This reduces
Eqs. (A.38) through (A.40) to:
ρ̃˙ (0) (t) = 0 (A.41)

h i
ρ̃˙ (1) (t) = −i Ṽ (t) ρ̃(0) (t) − ρ̃(0) (t) Ṽ † (t) (A.42)
h i X
ρ̃˙ (2) (t) = −i Ṽ (t) ρ̃(1) (t) − ρ̃(1) (t) Ṽ † (t) + L̃k (t) ρ̃(2) (t) L̃†k (t) . (A.43)
k
The projection operator formalism to separate the ground state evolution from the evolution of
the excited states. The terms Pg Ṽ (t) ρ̃(0) (t) in the equation for the first order terms will vanish,
as they pick out non-physical de-excitations from the ground state. The ground state evolution
is then given by:
Pg ρ̃˙ (0) (t) Pg = Pg ρ̃˙ (1) (t) Pg = 0 (A.44)

h i
Pg L̃k (t) Pe ρ̃(2) (t) Pe L̃†k (t) Pg ,
X
Pg ρ̃˙ (2) (t) Pg = −iPg Ṽ (t) ρ̃(1) (t) − ρ̃(1) (t) Ṽ † (t) Pg +
k
(A.45)
where the Lindblad operator has been written in the form L̃k = Pg L̃k Pe in the last line. The
evolution of the first order terms vanishes due to the lack of initial excitations. The identities
Pg (Pg + Pe ) = Pg · 1 ⇒ Pg Pg = Pg have also been used. The ground states are thus connected
by unitary and dissipative processes of second order. Tvolution of the excited states can be
simplified in a similar manner:
Pe ρ̃˙ (0) (t) Pe = Pe ρ̃˙ (1) (t) Pe = 0 (A.46)

h i
Pe ρ̃˙ (2) (t) Pe = −iPg Ṽ (t) ρ̃(1) (t) − ρ̃(1) (t) Ṽ † (t) Pg , (A.47)
where the dissipative feeding term have not been included for the second-order evolution, as
the excited states do not gain population from decay/spontaneous emission. The excited states
dynamics are thus only driven by the interaction Hamiltonian Ṽ (t), while interactions between
the two subspaces are given by the terms Pe ρ̃˙ (1) (t) Pg and Pg ρ̃˙ (1) (t) Pg .
A.1.4 Adiabatic Elimination of the Excited States

It is, in principle, possible to solve the second-order equations, but the calculations can be rather
involved, even after having separated the various subspaces. This process can be simplified as
the decaying excited states are essentially unpopulated. The evolution of these states can thus
be adiabatically eliminated:
Pe ρ̃˙ (2) (t) Pe ≈ 0. (A.48)
The dynamics of the second order term in Eq. (A.40) are approximated by the ground states of
Eq. (A.43). An effective equation for Pe ρ̃(2) (t) Pe is acquired by formally integrating Eq. (A.47):
Zt h i
(2)
dt0 Ṽ t0 ρ̃(1) t0 − ρ̃(1) t0 Ṽ t0 Pe .

Pe ρ̃ (t) = −iPe (A.49)
0
83
Integrating Eq. (A.38) gives an expression for ρ̃(1) (t):
Zt h i
(1)
dt0 Ṽ t0 ρ̃(0) t0 − ρ̃(0) t0 Ṽ t0 .

ρ̃ (t) = −i (A.50)
0
This can be substituted into Eq. (A.49):

t0

Z t Z h
(2) 0 00 0 00 (0) 00 0 (0) 00 †

Pe ρ̃ (t) Pe = −Pe dt dt Ṽ t Ṽ t ρ̃ t − Ṽ t ρ̃ t Ṽ (t) 

0 0
i 
− Ṽ t00 ρ̃(0) t00 Ṽ † t0 + ρ̃(0) t00 Ṽ † t00 Ṽ † t0 Pe 

(A.51)
Z t Zt0 h
0
i
dt00 Ṽ t0 ρ̃(0) t00 Ṽ † (t) + Ṽ t00 ρ̃(0) t00 Ṽ † t0 ,

= Pe dt
0
0
(A.52)
where the two terms of the form Pe Ṽ (t0 ) Ṽ (t00 ) ρ̃(0) (t00 ) Pe have been neglected, as Ṽ (t0 ) Ṽ (t00 )
effectively is an excitation followed by an de-excitation of ρ̃(0) (t). ρ̃(0) (t) lives in the ground
state subspace and has therefore no contribution to the excited state subspace. The two terms
in Eq. (A.52) are to be substituted into the last term in Eq. (A.43). For the first two terms in
Eq. (A.43):
h i
Pg ρ̃˙ (2) (t) Pg = −iPg Ṽ (t) ρ̃(1) (t) − ρ̃(1) (t) Ṽ † (t) Pg (A.53)

FirstTerms
 

 

  t
  


 Z h i 

0 0 (0) 0 (0) 0 † 0 
 
 Pg Ṽ (t) −i d t Ṽ t ρ̃ t − ρ̃ t Ṽ t Pg 

  

= −i 0
 
t

   


 Z h i 

0 0 (0) 0 (0) 0 † 0 †

 +Pg −i d t Ṽ t ρ̃ t − ρ̃
 

  t Ṽ t  Ṽ (t)Pg 

 
0
(A.54)
Zt h Zt
i h i
dt0 Ṽ t0 ρ̃(0) t0 Pg − Pg dt0 ρ̃(0) (t) Ṽ † t0 Pg ,

= −Pg Ṽ (t)
0 0
(A.55)
where terms of the form Pg Ṽ (t) ρ̃(0) (t) Ṽ † (t) Pg have been neglected, as these terms describes
(de)excitations between the sub spaces and therefore do not contribute to the ground state
evolution. Combining Eqs. (A.55) and (A.52) yields:
Zt h i Zt h i
Pg ρ̃˙ (2) (t) Pg = − Pg Ṽ (t) dt0 Ṽ t0 ρ̃(0) t0 Pg − Pg dt0 ρ̃(0) (t) Ṽ † t0 Pg

0 0
Z t Zt0 h i
dt00 Ṽ t0 ρ̃(0) t00 Ṽ † (t) + Ṽ t00 ρ̃(0) t00 Ṽ † t0 Pe L̃†k (t) Pg .
X
dt0

+ Pg L̃k (t) P e
k 0
0
(A.56)
84
The four terms in Eq. (A.56) are either Hamiltonian-like or Lindblad-like. They have the fol-
lowing form:
Zt h i
dt0 Ṽ t0 ρ̃(0) t0 Pg

I1 = Pg Ṽ (t) (A.57)
0
Zt Zt0
1 h i
dt0 dt00 Ṽ t0 ρ̃(0) t00 Ṽ † t00 .

I2 = Pe (A.58)
2
0
These integrals can be solved by assuming that the interaction between the ground state sub-
spaces to be perturbative, such that the ground state evolution is negligible as compared to the
evolution of the excited states. The contribution to the evolution of the ground state subspace
from Ô (t) can be found by:
h i
Pg Ô (t) Pg = Pg exp −i Ĥg + ĤNH t Pg (A.59)
h i h i2
= Pg 1Pg + Pg −i Ĥg + ĤNH t Pg + Pg − Ĥg + ĤNH t2 Pg + . . . (A.60)
2
≈ Pg 1 − iĤNH t + −iĤNH t + . . . Pg (A.61)
= Pg 1Pg . (A.62)
It is thus possible to set Ô (t) Pg ' Pg for a perturbative ground state interaction. I1 then
simplifies to:
Zt h i
dt0 Ṽ t0 ρ̃(0) t0 Pg

I1 = Pg Ṽ (t) (A.63)
0
Zt h i
= Pg Ô−1 (t) V Ô (t) dt0 Ô−1 t0 V Ô t0 ρ̃(0) t0 Pg

(A.64)
0
Zt h i
dt0 Ô−1 t0 V Pg ρ̃(0) (t)

' Pg V Ô (t) (A.65)
0
Zt h i
dt0 Ô−1 t0 (Pg + Pe ) V Pg ρ̃(0) (t)

= Pg V (Pe + Pg ) Ô (t) (A.66)
0
Zt h i
dt0 Ô−1 t0 V̂+ ρ̃(0) (t) Pg

= Pg V̂− Ô (t) (A.67)
0
−1 t
Ô−1 t0 V̂+ ρ̃(0) (t) Pg

= Pg V̂− exp −iĤNH t iĤNH t exp iĤNH t (A.68)
0
−1
' Pg V̂− iĤNH V̂+ ρ̃(0) (t) Pg , (A.69)
where the interaction Hamiltonian Ṽ (t) have been transformed back to the Schrödinger picture.
In addition, Pg V Pe = Pe V Pg = 0 have been used to simplify the expression. The density
operator for the ground state thus evolves slowly and to second order in V . The factors of Ṽ (t)
85
and Ṽ (t0 ) will cause the interaction to have features of fourth order. This can be overcome by
assuming that ρ̃(0) (t0 ) ≈ ρ̃(0) (t).
The t0 = 0 term have been dropped at the last equation,
as it is detuned with respect to the
t = t term due to the time dependence of exp iĤNH t . The term arising from t0 = 0 can
0
be neglected due to its fast evolution, assuming that the evolution of the ground states is slow
compared to that due to ĤNH . The second integral of type I1 is treated in the same manner
and gives the Hermitian conjugate of Eq. (A.69):
†
−1
I1,2 = Pg ρ̃(0) (t) V̂− −iĤNH V̂+ Pg . (A.70)
The factor ρ̃(0) (t00 ) can be approximated as ρ̃(0) (t) in I2 using the same argument as described
above. The integrals in I2 can then be separated as follows:
Zt Zt0
1 h i
dt0 dt00 Ṽ t0 ρ̃(0) (t) Ṽ † t00

I2 ' Pe (A.71)
2
0
Zt Zt
1 h i (0) h i
= Pe 00
dt Ṽ t ρ̃ (t) dt0 Ṽ † t0 Pe (A.72)
2
0 0
Zt Zt
1 h
−1 0
0
0
i (0) 0 † 0
−1 0 †
= Pe dt Ô t V Ô t ρ̃ (t) dt Ô t V Ô t Pe (A.73)
2
0 0
Zt Zt
1 h
0
−1 0
0
i (0) 0 † 0

−1 0
†
= Pe dt Ô t V (Pg + Pe ) Ô t ρ̃ (t) dt Ô t (Pe + Pg ) V Ô t Pe
2
0 0
(A.74)
Zt
†
Z
1 0
h
−1 0
i (0) 0 −1 0
' dt Ô t V̂+ ρ̃ (t) V̂− dt Ô t (A.75)
2
0 0
1 −1 †
−1
' iĤNH V̂+ ρ̃(0) (t) V̂− −iĤNH . (A.76)
2
The second integral of type I2 yields the same result. Eqs. (A.69), (A.70) and (A.76) are
substituted into Eq. (A.56), which yields:
−1 † 
−1
Pg ρ̃˙ (t) Pg = − Pg V̂− iĤNH
(2) (0) (0)
ρ̃ (t) + ρ̃ (t) −iĤNH Pg V̂+




−1 † (A.77)
L̃†k
X
(0) −1
+ Pg L̃k (t) iĤNH V̂+ ρ̃ (t) V̂− −iĤNH (t) Pg 



k
The following relations are defined for a shorthand notation:

−1
A (t) = −Pg V̂− iĤNH ρ̃(0) (t) V̂+ Pg (A.78)
−1
†
B (t) = −Pg V̂− ρ̃(0) (t) iĤNH V̂+ Pg (A.79)
−1 †
V̂+ ρ̃ (t) V̂− −iĤNH L̃†k (t) Pg
X
(0) −1
C (t) = Pg L̃k (t) iĤNH (A.80)
k
86
It is useful to change the remaining operators back to the Schrödinger picture. The terms on
the left hand side and the terms on the right hand side are treated separately separately for
brevity. The left hand side of Eq. (A.77) is:
 −1 
 dÔ−1 (t) (2)
−1 d Ô† (t)
ρ Ô† (t) + Ô−1 (t) ρ(2)

˙ (2)
Pg ρ̃ (t) Pg = Pg 

dt dt  Pg

(A.81)
 −1 
+Ô−1 (t) ρ̇(2) Ô† (t)

= i Ĥg ρ(0) − ρ(0) Ĥg + ρ̇(2) , (A.82)
where the relations Ô (t) Pg ' Pg and that Pg Ĥ0 Pg = Pg Ĥg Pg + Pg ĤNH Pg = Ĥg are needed in
order to simplify the expression.
The first term on the right hand side of Eq. (A.77) is changed to the Schrödinger picture using
−1
Ô (t) Pg ' Pg . It is useful to note that that Ô (t) commutes with V̂− ĤNH V̂+ :
−1
−1
V̂− ĤNH V̂+ ∝ |gi he| · Ĥe − L̂†k L̂ · |ei hg| ∝ |gi he|ei he|ei |gi = |gi hg| . (A.83)
Lindblad operators
h are proportional i to |gi he|, which effectively gives ĤNH ∝ |ei he|. It is then
−1
easily seen that Ô (t) , V̂− ĤNH V̂+ = 0. The change to the Schrödinger picture can then be
carried out:
−1
−1
A (t) = iPg V̂− ĤNH V̂+ Ô−1 (t) ρ(0) Ô† (t) Pg (A.84)
−1
−1
= iPg Ô−1 (t) V̂− ĤNH V̂+ ρ(0) Ô† (t) Pg (A.85)
−1
' iPg V̂− ĤNH V̂+ ρ(0) Pg . (A.86)
This expression is simplified by adding zero and splitting each term into two halves:
† †
−1 −1 −1
A (t) = iPg V̂− ĤNH V̂+ ρ(0) Pg + iPg V̂− ĤNH V̂+ ρ(0) Pg − iPg V̂− ĤNH V̂+ ρ(0) Pg (A.87)
† †
i −1

−1 (0) i −1

−1
= Pg V̂− ĤNH + ĤNH V̂+ ρ Pg + Pg V̂− ĤNH − ĤNH V̂+ ρ(0) Pg (A.88)
2 2
†
0 (0) i −1

−1
= − iPg Ĥeff ρ Pg + Pg V̂− ĤNH − ĤNH V̂+ ρ(0) Pg , (A.89)
2
87
†
0 ≡ − 1 V̂ −1 −1
where Ĥeff 2 − ĤNH + ĤNH V̂+ . The last term in Eq. (A.89) is rewritten by multi-
plying by 1. In the following, any factor in a bracket multiplies to 1:
† † † h i † h i
−1 −1 −1 −1 −1 −1 −1
iV̂− ĤNH − ĤNH V̂+ = V̂− ĤNH ĤNH ĤNH ĤNH ĤNH − ĤNH ĤNH ĤNH V̂+
(A.90)
−1 † −1
−1 −1 −1 −1 −1
= iV̂− ĤNH · ĤNH · ĤNH − ĤNH · ĤNH · ĤNH V̂+
(A.91)
" ! !#
−1 iX † iX †
−1 −1
= iV̂− ĤNH Ĥe + L̂k L̂k − Ĥe − L̂k L̂k ĤNH V̂+
2 2
k k
(A.92)
−1
L̂†k L̂k ĤNH
X
−1 −1
=− V̂− ĤNH V̂+ (A.93)
k
X †
=− L̂keff L̂keff , (A.94)
k
where the effective Lindblad operator is defined:

X
−1
L̂keff ≡ V̂− ĤNH L̂k . (A.95)
k
Eq. (A.89) can then be written as:
0 1 X k † k (0)
A (t) = −iĤeff ρ(0) − L̂eff L̂eff ρ . (A.96)
2
k
Note that the time dependence vanishes. The second term on the right hand side of Eq. (A.77)
is the Hermition conjugate of Eq. (A.96):
0 1 X (0) k † k
B (t) = iρ(0) Ĥeff − ρ L̂eff L̂eff . (A.97)
2
k
The third term on the right hand side of Eq. hA.77 is rewritten iby changing L̃keff (t) and ρ̃(0) (t)
−1
to the Schrödinger picture and by using that Ô (t) , V̂− ĤNH V̂+ = 0. This gives:
−1 † † †
Ô† (t) L̂†k Ô−1 (t) Pg
X
−1
C (t) = Pg Ô−1 (t) L̂k Ô (t) iĤNH V̂+ Ô−1 (t) ρ(0) Ô−1 (t) V̂− −iĤNH
k
(A.98)
−1 †
L̂†k Pg
X
−1
' Pg L̂k iĤNH V̂+ ρ(0) V̂− −iĤNH (A.99)
k
X †
= L̂keff ρ(0) L̂keff , (A.100)
k
where the same definition for the effective Lindblad operator has been used as above. Eqs. (A.96),
(A.97) and (A.100) for the right hand side of Eq. A.77 can now be collected together. An effective
88
equation for the time evolution for the reduced density operator of second order can be found
by using Eq. (??) for the left hand side:
 h i X † 
0


 −i Ĥeff , ρ(0) + L̂keff ρ(0) L̂keff 


h i  
k
i Ĥg , ρ(0) + ρ̇(2) = † (A.101)
1 X 1 X k † k 
 −2 L̂keff L̂keff ρ(0) − ρ(0) L̂eff L̂eff 


 
2 
k k
This gives an effective equation of motion for the ground states

h i X † 1 † †
(2) (0) k (0) k k k (0) (0) k k
ρ̇ = −i Heff , ρ + L̂eff ρ L̂eff − L̂eff L̂eff ρ + ρ L̂eff L̂eff , (A.102)
2
k
where the effective Hamiltonian is defined as:

†
0 1 −1

−1
Ĥeff = Ĥeff + Ĥg = − V̂− ĤNH + ĤNH V̂+ + Ĥg . (A.103)
2
Eq. (A.101) is of the standard Lindblad form with the effective Hamiltonian Ĥeff and effective
Lindblad operator L̂keff . These Lindblad operators describe second-order decay processes. Each
consist of a weak excitation V̂+ , an evolution between the excited states through ĤNH and
subsequent decay L̂k .
89
90
Appendix B
Calculation of Effective Operators
The system considered in the thesis is governed by the non-hermitian Hamiltonian ĤNH that take
both coherent and dissipative evolution of the quantum states into account. This Hamiltonian
can be split into four different non-Hermitian Hamiltonians, depending on the number of qubits
in state |1iq :
h11E| he1f | h1ef |

 
1D i
γ0 − 2i γAq
1D − 2i γAq
1D

|11Ei ∆A − γA 2 +
 
− 2i γAq
1D δq1 − 2i γq1D γ0 − 2i γqq
1D

ĤNH = |e1f i  + (B.1)

1
n=2  
− 2i γAq
1D i 1D i
γq1D γ0

|1ef i − 2 γqq δq1 − 2 2
+
 h10E| he0f | 
1D i
γ0 − 2i γAq
1D

|10Ei ∆A − γA 2 +
ĤNH = (B.2)
 
n=1,q1 |e0f i − 2i γAq
1D δq1 − i
2 γq1D
1
+ γ0
 h01E| h0ef | 
1D i 0 − 2i γAq
1D

|01Ei ∆A − γA 2 +γ
ĤNH = (B.3)
 
n=1,q2 |0ef i − 2i γAq
1D δq2 − 2i γq1D
2
+ γ0

i 1D 0

ĤNH = ∆A − γ +γ |00Ei h00E| (B.4)

n=0 2 A
and includes the cases where either two, one or no qubit states are optically connected to the
waveguide. Here γA,q0 is the decay rate into the wave gudie for the auxiliary emiiter or for a qubit
emitter, respectively. γ 0 is the nonradiative decay rate while γAq and γqq are the two-emitter
decay rates between the auxiliary- and qubit emitters, and between the two qubit emitters
respectively. These rates are for decay into the guided mode in the waveguide ∆A and δq are
the detuning of the auxiliary and qubit emitters. The evolution of the system can be simplified
using adiabatic elimination, which results in the effective master equation:
h i X † 1 X † †
ρ̇ = −i Ĥeff , ρ (t) + L̂keff ρ (t) L̂leff − L̂keff L̂leff ρ (t) + ρ (t) L̂keff L̂leff
2
k,l k,l
(B.5)
†
1 −1 −1

Ĥeff = − V̂− ĤNH + ĤNH V̂+ + Ĥg (B.6)
2
−1
L̂keff = L̂k ĤNH V̂+ . (B.7)
91
B.1 Two qubits in the ground state
The calculations for deriving the effective operators in Eqs. (B.6) and (B.7) are shown below in
the ideal case where the coupling constants of the three emitters are equal, the detuning of the
two qubit emitters are δq1 = 0 and the emitters are placed in the constructive interference zones
to maximize the effects of superradiance. This causes all the decay rates to the guided mode in
Eqs. (B.1) trough (B.4) to be equal to γA and will be denote γ 1D throughout this section.
The inverse of ĤNH can then be calculated, where the each element is given by:

(−1)k+l · det ĤNH,k,l
−1
ĤNH [k, l] = , (B.8)
det ĤNH
where det is the determinant and ĤNH,k,l is the complimentary matrix to the matrix element
ĤNH [k, l]. The only relevant contribution to the dynamics of the gate comes from three matrix
elements in the first row contribute, as these are the terms that V̂+ picks out. These terms can
be formally extracted using the projection operator P̂2 = |11iq h11|. The three terms are:
2 2γ 1D + γ 0

−1
ĤNH,j [1, 1] = |EiA hE| ⊗ |11iq h11| (B.9)
2∆A [2γ 1D + γ 0 ] − i [3γ 1D γ 0 + γ 02 ]
2 γ 1D + γ 0 2∆A 2γ 1D + γ 0 + i 3γ 1D γ 0 + γ 02

= |EiA hE| ⊗ |11iq h11| (B.10)
4∆2A [2γ 1D + γ 0 ]2 + [3γ 1D γ 0 + γ 02 ]2
(B.11)
h i
∆A
1 2 (2C + 1) 2 γ 0 (2C + 1) + i (3C + 1)
= 0 2 |EiA hE| ⊗ |11iq h11| (B.12)
γ

2 ∆γA0 [2C + 1]2 + [3C + 1]2
−1 2γ 1D
ĤNH,j [2, 1] = − |fi hE| ⊗ |e1iq h11| (B.13)
2∆A [2γ 1D + γ ] − i [3γ 1D γ 0 + γ 02 ] A
(B.14)
h i
∆A
1 2C 2 γ 0 (2C + 1) + i (3C + 1)
= − 0 2 |fiA hE| ⊗ |e1iq h11| (B.15)
γ
2 ∆γA0 [2C + 1]2 + [3C + 1]2
−1 2γ 1D
ĤNH,j [3, 1] = − |fi hE| ⊗ |1eiq h11| (B.16)
2∆A [2γ 1D + γ 0 ] − i [3γ 1D γ 0 + γ 02 ] A
(B.17)
h i
∆A
1 2C 2 γ 0 (2C + 1) + i (3C + 1)
=− 0 2 |fiA hE| ⊗ |1eiq h11| , (B.18)
γ
2 ∆γA0 [2C + 1]2 + [3C + 1]2
where the cooperativity C = γ 1D /γ 0 has been introduced. The cooperativity is related to the
β−factor introduced in Section (3.1.2) by:
1
C= . (B.19)
β −1 −1
92
Only Eq. (B.12) is needed to calculate the effective Hamiltonian:
†
1 −1

−1
Ĥeff,2 = − V̂− ĤNH + ĤNH V̂+ (B.20)
2
= ∆2 |giA hg| ⊗ P̂2 , (B.21)
where the energy shift ∆2 is defined as:
Ω2 ∆A (2C + 1)2
∆2 = − . (B.22)
γ0 γ0
2
2 2
2 ∆γA0 (2C + 1) + (3C + 1)
The effective Lindblad operators are readily calculated:

L̂A,2
eff = L̂
A,2 −1
ĤNH V̂+ = rA,2 |fiA hg| ⊗ P̂2 (B.23)
q ,2 −1
L̂effj = L̂qj ,2 ĤNH V̂+ = rqj ,2 |fiA hg| ⊗ P̂2 , j = 1, 2, (B.24)
where the Lindblad coefficients for the decay of the auxiliary emitter rA,2 and the qubits rqj ,2
are given by:
h i
∆A
q
Ω (2C + 1) 2 γ 0 (2C + 1) + i (3C + 1)
rA,2 = γ 1D + γ 0 0 2 (B.25)
γ

∆A 2 2
2 γ0 [2C + 1] + [3C + 1]
h i
∆A
q
Ω C 2 γ 0 (2C + 1) + i (3C + 1)
rqj ,2 = − γ 1D 0 2 . (B.26)
γ
2 ∆γA0 [2C + 1]2 + [3C + 1]2
Note that rq1 ,2 = rq2 ,2 in this ideal case.
B.1.1 One qubit in the ground state

The two 2 × 2 sub matrices in Eqs. (B.2) ans (B.3) represent the case when only the first or
the second qubit is in the optical ground state. These two sub matrices are, in the ideal case
considered here, equal. The two terms that are picked out by V̂+ are:
2 γ 1D + γ 0

−1
ĤNH,j [k, k] = |EiA hE| ⊗ P̂1j (B.27)
2∆A [γ 1D + γ 0 ] − i [γ 1D γ 0 + γ 02 ]
2 γ 1D + γ 0 2∆A /γ 0 γ 1D + γ 0 + i 2γ 1D + γ 0

= |EiA hE| ⊗ P̂1j (B.28)
4∆2A [γ 1D + γ 0 ]2 + (2γ 1D + γ 0 )2
1 2 (C + 1) [2∆A /γ 0 (C + 1) + i (2C + 1)]

= |EiA hE| ⊗ P̂1j (B.29)
γ0
2
∆A 2 2
2 γ0 [C + 1] + [2C + 1]
−1 2γ 1D
ĤNH,j [k + 1, k] = − |fi hE| ⊗ |{ej }iqj h{1j }| (B.30)
2∆A [γ 1D + γ 0 ] − i [γ 1D γ 0 + γ 02 ] A
1 2C [(2∆A /γ 0 ) (C + 1) + i (2C + 1)]

=− |fiA hE| ⊗ |{ej }iqj h{1j }|, (B.31)
γ0
2
2 ∆γA0 [C + 1]2 + [2C + 1]2
93
where P̂1j is the projection operator when only qubit j is in the grounds state, |{ej }iqj (|{1}iqj )
is the excited state (optical ground state) for qubit j with the other qubit in state |0i. The
effective Hamiltonian is in this case given by:
Ĥeff,1j = ∆1 |giA hg| ⊗ P̂1,j , (B.32)
with the coefficient:
Ω2 ∆A (C + 1)2
∆1 = − . (B.33)
γ0 γ0
2
2 ∆γA0 [C + 1]2 + [2C + 1]2
The Lindblad operators are given by:
L̂A,1j
eff = L̂
A,1j −1
ĤNH V̂+ = rA,1j |fiA hg| ⊗ P̂1j (B.34)
q ,1 −1
L̂effj = L̂qj ,1 ĤNH V̂+ = rqj ,1 |fiA hg| ⊗ P̂1j , (B.35)
with coefficients
Ω (C + 1) [2∆A /γ 0 (C + 1) + i (2C + 1)]
q
rA,1 = γ 1D + γ 0 0 2 (B.36)
γ

2 ∆γA0 [C + 1]2 + [2C + 1]2
Ω C [(2∆A /γ 0 ) (C + 1) + i (2C + 1)]
q
rqj,1 = − γ 1D 0 2 . (B.37)
γ
2 ∆γA0 [C + 1]2 + [2C + 1]2
B.1.2 No qubits in the ground state

This case is particularly simple, as it only involves the single element in Eq. (B.4). The inverse
is then
−1 2
ĤNH,j [1, 1] = |EiA hE| ⊗ P̂0 (B.38)
2∆A − i [γ 1D + γ 0 ]
2 2∆A + i γ 1D + γ 0

= |EiA hE| ⊗ P̂0 (B.39)
4∆2A + [γ 1D + γ 0 ]2
∆A
2 2 γ 0 + i [C + 1]
= 0 2 |EiA hE| ⊗ P̂0 . (B.40)
γ
2 ∆γA0 + [C + 1]2
The corresponding matrix element for the effective Hamiltonian is:
Ĥeff,0 = ∆0 |giA hg| ⊗ P̂0 , (B.41)
where the coefficient is defined as:

Ω2 ∆A 1
∆0 = − . (B.42)
γ0 γ0
2
2 ∆γA0 + (C + 1)2
The single Lindblad operator is in this case given by:
L̂A,0
eff = L̂
A,0 −1
ĤNH V̂+ = rA,0 |fiA hg| ⊗ P̂0 , (B.43)
94
where
∆A
Ω 2 γ 0 + i (C + 1)
q
rA,0 = γ 1D + γ 0 0 2 . (B.44)
γ ∆A 2
2 γ0 + [C + 1]
These effective operators derived above can be written in a compact by summing over the number
n of qubits in the ground state:
2
X Ω2 ∆A (nC + 1)2
Ĥeff = − 0 0 2 |giA hg| ⊗ P̂n (B.45)
γ γ
n=0 2 ∆γA0 (nC + 1)2 + ([n + 1] C + 1)2
2
X
= ∆n |giA hg| ⊗ P̂n (B.46)
n=0
2
X √ Ω 2 ∆γA0 (nC + 1) + i ([n + 1] C + 1)
L̂A
eff = C + 1 √ 0 (nC + 1) 2 |fiA hg| ⊗ P̂n (B.47)
n=0
γ 2 ∆A (nC + 1)2 + ([n + 1] C + 1)2
γ0
2
X
= rA,n |fiA hg| ⊗ P̂n (B.48)
n=0
q
X2 √ Ω 2 ∆γA0 (nC + 1) + i ([n + 1] C + 1)
L̂effj = − C√ 0C 2 |fiA hg| ⊗ P̂n (B.49)
n=1
γ 2 ∆A (nC + 1)2 + ([n + 1] C + 1)2
γ0
2
X
= rqj ,n |fiA hg| ⊗ P̂n (B.50)
n=1
B.1.3 Inclusion of δqj

q
The coefficients for the effective operators for Ĥeff , L̂A j
eff and L̂eff for a non-zero qubit detuning
δqj will be listed below under the assumption that δq1 = δq2 . The calculations follow the same
procedure as in the preceding section. Only the operators for n = 2 and n = 1 will be shown,
as the n = 0 case is unchanged:
h i
Ω2 ˜ 2 ˜ 2
− γ 0 ∆A (nC + 1) + δ̃q 4∆A δq + nC
∆n>0 = 2 h i (B.51)
˜ 2 2 ˜ 2 ˜ 2 2 2
2∆A (nC + 1) + ([n + 1] C + 1) + 8n∆a δ̃q C + 16∆A + 4 (C + 1) δ̃q
√ h h i i
Ω √C+1 2 + 4i (C + 1) δ̃ 2
γ 0 (nC + 1) · α n (C) + 2 δ̃ q 4 δ̃q + nC q
rA,n>0 = 2 h i (B.52)
˜ 2 2 ˜ 2 ˜ 2 2 2
2∆A (nC + 1) + ([n + 1] C + 1) + 8n∆a δ̃q C + 16∆A + 4 (C + 1) δ̃q
√ h i h i
− Ω√γCC
0
˜ A (nC + 1)2 + i ([n + 1] C + 1) + 2δ̃q C 1 + C + 2i∆
2∆ Ã
rqj ,n = 2 h i (B.53)
˜ 2 2 ˜ 2 ˜ 2 2 2
2∆A (nC + 1) + ([n + 1] C + 1) + 8n∆a δ̃q C + 16∆A + 4 (C + 1) δ̃q
where αn in Eq. (B.52) is defined as:
˜ A (nC + 1) + i ([n + 1] C + 1)
αn (C, n) = 2∆ (B.54)
95
The Lindblad coefficients in Eqs. (B.52) and (B.53) result in the following effective decay rates:
h i
2 4C δ̃ 2 + nC
Ω q
Γn>0 = 0 2 i .
γ
h
˜ 2 2 ˜ ˜ 2 2
2
2∆A (nC + 1) + ([n + 1] C + 1) + 8n∆a δ̃q C + 16∆A + 4 (C + 1) δ̃q2
(B.55)
B.2 Variable Qubit Dipole Moments & Positions

The coefficients will be listed below for a non-zero qubit detuning δqj and where the dipole
moments for the qubit emitters differ from the auxiliary emitter by:
dq1 = κdA , dq2 = ηdA , (B.56)
and where the n’th qubit can be displaced from its ideal position by δxq,n . This results in the
following expressions for the energy shifts ∆n and decay rates Γn , for n > 0:
˜ A δ̃q2 + bH,n (δxq,1 , δxq,2 , κ, η) C 3 + fH,n (κ, η) C 2 δ̃q + gH,n (κ, η) C 2 ∆
Ã

Ω2 4∆
∆n>0 = − 0
γ ˜ A δ̃q C 2 + jn (κ, η) ∆
4δ̃q2 C 2 + hn (κ, η) ∆ ˜ 2 C 2 + kn (κ, η) C 2
A
(B.57)
Ω2 4δ̃q2 C + fΓ,n (δxq,1 , δxq,2 , κ, η) C 3 + gΓ,n (δxq,1 , δxq,2 , κ, η) C 2
Γn>0 = . (B.58)
γ 0 4δ̃q2 C 2 + hn (κ, η) ∆˜ A δ̃q C 2 + jn (κ, η) ∆
˜ 2 C 2 + kn (κ, η) C 2
A
These expressions are used in Chapters (6.1) (for the case where δxq,1 = δxq,2 = 0) and (6.2) in
order to describe the dynamics of the ground states. The various functions in the numerator for
Eqs. (B.57) and (B.58) are given by:
bH,n (δxq,1 , δxq,2 , κ, η) = 4κ2 η 2 cos2 (δxq,1 − δxq,2 ) + sin (δxq,1 − δxq,2 ) sin (δxq,1 + δxq,2 ) − 1

(B.59)
fH,n (κ, η) = −κ2 − η 2 (B.60)
2 2 2

gH,n (κ, η) = − κ + η (B.61)
h i
2
gΓ,n (δxq,1 , δxq,2 , κ, η) = −2 κ3/2 cos (δxq,1 ) + η 3/2 cos (δxq,2 ) + κ2 + η 2 + κ2 + η 2

(B.62)
with the last function defined as:

2
κ2 + η 2 + κ3 + η 3 + 2 (κη)3/2 cos (δxq,1 − δxq,2 )
 

 

 
 3/2 2 



 −2κ η [− cos (δxq,1 − δx q,2 ) cos (δxq,2 ) + sin (δxq,1 − δxq,2 ) sin (δxq,2 )]



3/2 2 2

fΓ,n (δxq,1 , δxq,2 , κ, η) = −2κ 2η cos (δxq,1 ) + κ cos (δxq,1 )
 
3/2 2
 
−2η κ [− cos (δx − δx ) cos (δx ) + sin (δx − δx ) sin (δx )]
 
q,1 q,2 q,1 q,1 q,2 q,1

 


 

3/2 2 2
 
−2η 2κ cos (δxq,2 ) + η cos (δxq,2 )
(B.63)
The shared functions in the denominator of Eqs. (B.57) and (B.58) are defined as:
hn (κ, η) = 8 κ2 + η 2

(B.64)
2
jn (κ, η) = 4 κ2 + η 2 (B.65)
2
kn (κ, η) = κ2 + η 2 + 2 κ2 + η 2 + 1.

(B.66)
96
Appendix C
Additional Contour Plots
This section provides additional plots of the contours for the success probability and fidelity as a
function of the displacements of the two qubits. These are plotted for a variety of combinations
of qubit dipole moments by changing the value of κ and η. The detunings used for the simply
optimised case are:
r
γ0 1
∆A = C− (C.1)
2 2
7γ 0
δq = − √ , (C.2)
8 C
while the detunings employed in the complete optimisation are:
γ0
q
∆A,n = −2κ7/2 cos (kδxq,n ) + κ4 + κ3 C 2 + κ4 (C − 1) + 1/2 (C.3)
2κ2
γ0
κ3/2 cos (kδxq,n ) − κ2 C + 4κ2 + 3

8 4
δq = q (C.4)
κ3/2 sin (kδxq,n ) C 2 − −2κ7/2 cos (kδxq,n ) + κ4 + κ3 C 2 + κ4 (C − 1) + 1/2
97
4
0.54
π 0.48
8 0.42
0.36
kδxq,2
0 0.30
0.24
−π8 0.18
0.12
0.06
−π4
4
0.54
π 0.48
8 0.42
0.36
kδxq,2
0 0.30
0.24
−π8 0.18
0.12
0.06
−π4 π
−4 −π8 0 π
8
π
4
δxq,1
Figure C.1: Plot of the contours for the success probability as a function of the
displacement of the two qubits for κ = η = 1.3. Note the massively lowered success
probability compared to the case where κ = η = 1
4
1.0
π
8 0.8
0.6
kδxq,2
0
0.4
−π8 0.2
0.0
−π4
4
1.0
π
8 0.8
0.6
kδxq,2
0
0.4
−π8 0.2
0.0
−π4 π
−4 −π8 0 π
8
π
4
kδxq,1
Figure C.2: Plot of the contours for the fidelity as a function of the displacement of
the two qubits for κ = η = 1.3.
98
4
0.54
π 0.48
8 0.42
0.36
kδxq,2
0 0.30
0.24
−π8 0.18
0.12
0.06
−π4
4
0.54
π 0.48
8 0.42
0.36
kδxq,2
0 0.30
0.24
−π8 0.18
0.12
0.06
−π4 π
−4 −π8 0 π
8
π
4
δxq,1
displacement of the two qubits for κ = η = 0.75.
4
1.0
π
8 0.8
0.6
kδxq,2
0
0.4
−π8 0.2
0.0
−π4
4
1.0
π
8 0.8
0.6
kδxq,2
0
0.4
−π8 0.2
0.0
−π4 π
−4 −π8 0 π
8
π
4
kδxq,1
the two qubits for κ = η = 0.75.
99
4
0.54
π 0.48
8 0.42
0.36
kδxq,2
0 0.30
0.24
−π8 0.18
0.12
0.06
−π4
4
0.54
π 0.48
8 0.42
0.36
kδxq,2
0 0.30
0.24
−π8 0.18
0.12
0.06
−π4 π
−4 −π8 0 π
8
π
4
δxq,1
displacement of the two qubits for κ = 1.3 and η = 0.75.
4
1.0
π
8 0.8
0.6
kδxq,2
0
0.4
−π8 0.2
0.0
−π4
4
1.0
π
8 0.8
0.6
kδxq,2
0
0.4
−π8 0.2
0.0
−π4 π
−4 −π8 0 π
8
π
4
kδxq,1
the two qubits for κ = 1.3 and η = 0.75.
100
4
0.54
π 0.48
8 0.42
kδxq,2 0.36
0 0.30
0.24
−π8 0.18
0.12
0.06
−π4
4
0.54
π 0.48
8 0.42
0.36
kδxq,2
0 0.30
0.24
−π8 0.18
0.12
0.06
−π4 π
−4 −π8 0 π
8
π
4
δxq,1
displacement of the two qubits for κ = 0.75 and η = 1.3.
4
1.0
π
8 0.8
0.6
kδxq,2
0
0.4
−π8 0.2
0.0
−π4
4
1.0
π
8 0.8
0.6
kδxq,2
0
0.4
−π8 0.2
0.0
−π4 π
−4 −π8 0 π
8
π
4
kδxq,1
the two qubits for κ = 0.75 and η = 1.3.
101
4
0.54
π 0.48
8 0.42
kδxq,2 0.36
0 0.30
0.24
−π8 0.18
0.12
0.06
−π4
4
0.54
π 0.48
8 0.42
0.36
kδxq,2
0 0.30
0.24
−π8 0.18
0.12
0.06
−π4 π
−4 −π8 0 π
8
π
4
δxq,1
displacement of the two qubits for κ = η = 1. The completely optimising detunings
are here evaluated along δxqw , thereby maximising the success probability along this
displacement.
4
1.0
π
8 0.8
0.6
kδxq,2
0
0.4
−π8 0.2
0.0
−π4
4
1.0
π
8 0.8
0.6
kδxq,2
0
0.4
−π8 0.2
0.0
−π4 π
−4 −π8 0 π
8
π
4
kδxq,1
Figure C.10: Same as Fig. (C.10), but for the fidelity.
102
Appendix D
Solving The Effective Master Equation
The dynamics of an open quantum mechanical system can be described well by the master
equation for the density operator:
h i X 1 Xh † i
ρ̇ = i Ĥ, ρ + L̂k ρL̂†l − L̂k L̂k ρ − ρL̂†k L̂k , (D.1)
2
k,l k,l
where Ĥ is the Hamiltonian for the system and L̂k is the Lindblad operator describing a source
of decay. This assumes that the dynamics of the system is Markovian. The system of equations
that can be extracted from Eq. (D.1) will often be large and can be cumbersome to solve. These
systems of equations can be simplified by adiabatically eliminating rapidly evolving excited
states, resulting in an effective master equation:
†
h i X
k k 1 X k † l
k
†
l
ρ̇ = i Ĥeff , ρ + L̂eff ρL̂eff − L̂eff L̂eff ρ − ρ L̂eff L̂eff , (D.2)
2
k,l k,l
where the effective operators are given by:

†
1 −1

−1
Ĥeff = − V̂− ĤNH + ĤNH V̂+ + Ĥg (D.3)
2
−1
L̂keff = L̂k ĤNH V̂+ . (D.4)
Here V̂+ (V̂− ) is a weak (de)excitation laser field and ĤNH = Ĥe − 2i L̂†k L̂l is the non-Hermitian
P
k,l
Hamiltonian of the quantum jump formalism, with Ĥe being the Hamiltonian of the excited state
manifold and Ĥg for the ground state manifold. The following sections will show the derivation
of the system of equations describing the dynamics of the system discussed in this work.
D.1 The Effective Operators

The effective operators are derived in Appendix (B) and have the general form:
2
X
Ĥeff = ∆n |giA hg| ⊗ P̂n (D.5)
n=0
X2
L̂A
eff = rA,n |fiA hg| ⊗ P̂n (D.6)
n=0
2
q
X
L̂effj = rqj ,n |fiA hg| ⊗ P̂n , (D.7)
n=1
103
where P n projects onto the states where there are n qubits in state |1i:
P̂2 = |11iq h11| (D.8)

P̂1,1 = |01iq h01| (D.9)
P̂1,2 = |10iq h10| (D.10)
P̂0 = |00iq h00| . (D.11)
The gate in question is conditioned on measuring the state of the auxiliary quantum dot to be in
state |giA , so the second term in Eq. (D.2) will not contribute to the success probability or the
fidelity of the gate and can thus be omitted from the following discussion. An arbitrary input
state for the qubits of the form |ψ (t)i = c00 (t) |00iq + c01 (t) |01iq + c10 (t) |10iq + c11 (t) |11iq
P m
is considered for the following. The density matrix is given by ρ = |ψi hψ| = ρn (t) |mi hn|,
mn
∗
where ρmn (t) = cm (t) cn (t) with m, n = {00, 01, 10, 11}. The coherent part of the effective
master equation is given by:
h i 2
X
−i Ĥeff , ρ = − i ∆n P̂n ρ − ρP̂n (D.12)
n=0
= − i∆0 ρ00 00 00 00
00 |00iq h00| + ρ01 |00iq h01| + ρ10 |00iq h10| + ρ11 |00iq h11| (D.13)
−ρ00 01 10 11

00 |00iq h00| − ρ00 |01iq h00| − ρ00 |10iq h00| − ρ00 |11iq h00| (D.14)
− i∆1,1 ρ01 01 01 01
−ρ00 01 10 11

− i∆1,2 ρ10 10 10 10
−ρ00 01 10 11

− i∆2 ρ11 11 11 11
−ρ00 01 10 11

11 |00iq h11| − ρ11 |01iq h11| − ρ11 |10iq h11| − ρ11 |11iq h11| . (D.20)
All possible no-jump terms in the dissipative part of Eq. (D.2) are listed here for convenience
{} denotes that there is an additional term where A and qk have been interchanged. All off the
following expressions are multiplied by ⊗ |giA hg|. The dissipative no-jump terms for n = 2 are
given by, where k, l denotes a term corresponding to either the first or the second qubit:
† A 2 11
L̂Eff
A2 L̂Eff ρ00 |11i h00| + ρ11
A2 ρ = r2 q
11 11
01 |11iq h01| + ρ10 |11iq h10| + ρ11 |11iq h11| (D.21)
† A 2 00
ρ L̂Eff
A2 L̂Eff ρ11 |00i h11| + ρ01
A2 = r2 q
10 11
11 |01iq h11| + ρ11 |10iq h11| + ρ11 |11iq h11| (D.22)
† ∗
q q
L̂Eff
qk 2 L̂Eff
ql 2 ρ = r2,k r2,l ρ11 11 11 11
† ∗
q q
ρ L̂Eff
qk 2 L̂Eff
ql 2 = r2,k r2,l ρ00 01 10 11
† ∗
q
L̂Eff
A2 L̂Eff
qk 2 ρ = r2A r2,k ρ11 11 11 11
† ∗
q
ρ L̂A2 L̂qk 2 = r2A r2,k
Eff Eff
ρ00 01 10 11
104
The dissipative terms when only the first qubit is optically connected to the waveguide are given by
† A 2 10
L̂Eff
A1,1 L̂Eff
A1,1 ρ = r1,1
ρ00 |10i h00| + ρ10
q
10 10
01 |10iq h01| + ρ10 |10iq h10| + ρ11 |10iq h11| (D.27)
†
A 2 00
ρ L̂Eff
A1,1 L̂Eff
A1,1 = r1,1
ρ10 |00i h10| + ρ01
q
10 11
10 |01iq h10| + ρ10 |10iq h10| + ρ10 |11iq h10| (D.28)
† q 2 10
L̂Eff
q1,1 L̂Eff
q1,1 ρ = r1,1
ρ00 |10i h00| + ρ10
q
10 10
01 |10iq h01| + ρ10 |10iq h10| + ρ11 |10iq h11| (D.29)
† q 2 00
ρ L̂Eff
q1,1 L̂Eff
q1,1 = r1,1
ρ10 |00i h10| + ρ01
q 10 |01i q h10| + ρ 10
10 |10iq h10| + ρ 11
10 |11i q h10| (D.30)
†
q ∗
L̂Eff
A1,1 L̂Eff
q1,1 ρ
A
= r1,1 r1,1 ρ10
00 |10iq h00| + ρ10
01 |10iq h01| + ρ 10
10 |10iq h10| + ρ10
11 |10i q h11|
(D.31)
†
q ∗
ρ L̂Eff L̂Eff A
ρ00 01 10 11

A1,1 q1,1 = r1,1 r1,1 10 |00iq h10| + ρ10 |01iq h10| + ρ10 |10iq h10| + ρ10 |11iq h10|
(D.32)
The terms corresponding to the case where only the second qubit is connected can be written down by
A A q q
setting r1,1 → r1,2 and r1,1 → r1,2 . The contributions to the effective master equation for the case where
n = 0 are:
† A 2 00
L̂Eff
A0 L̂Eff ρ00 |00i h00| + ρ00
A0 ρ = r0 q 01 |00iq h01| + ρ 00
10 |00i q h10| + ρ 00
11 |00iq h11| ⊗ |giA hg| (D.33)
†
A 2 00
ρ L̂Eff
A0 L̂Eff ρ00 |00i h00| + ρ01
A0 = r0 q
10 11
00 |01iq h00| + ρ00 |10iq h00| + ρ00 |11iq h00| ⊗ |giA hg| (D.34)
Any jump term in Eq. (D.2) for any of the four non-hermitian Hamiltonians take the following form:
∗
L̂k ρL̂†l ∝ rnk rnl ρee . (D.35)
The initialization scheme presented in Section (3.3) ensures that the population of the excited states of
the quantum dots is nearly empty. The effects of the jump terms can thus be neglected in this work.
The Lindblad coefficients can be used to define the following decay rates:
2
Γ0 ≡ r0A (D.36)
A ∗
A 2 q 2 A q ∗ q
Γ1,1 ≡ r1,1 + r1,1 + r1,1 r1,1 + r1,1 r1,1 (D.37)
A 2 q 2 A q ∗ q A
∗
Γ1,2 ≡ r1,2 + r1,2 + r1,2 r1,2 + r1,2 r1,2 (D.38)
A 2 q 2 q 2 ∗ ∗
r2 + r + r + rq rq q q
)
2,1 2,1 2,1 2,2 + r2,2 r2,1
Γ2 ≡ ∗ ∗ ∗ ∗ (D.39)
q q q q
+r2A r2,1 + r2A r2,1 r2A + r2,2 r2A

+ r2,1
The equations of motion for the system can now be found by projecting onto the effective master equation,
which results in sixteen differential equations of the form:
1 n o
ρ̇xy = i (∆y − ∆x ) ρxy (t) − (Γx + Γy ) ρxy (t) , x, y = 00 , 01 , 10 , 11 , (D.40)
2 n=0 n=1 n=1 n=2
where the notation underneath each element in the definition of x and y indicates the number of qubits
in |1iq . This equation has a straightforward solution:
1
ρyx (t) = ρyx (t = 0) ei(∆y −∆x )t e− 2 (Γx +Γy )t . (D.41)
The off-diagonal elements has has a dynamical change in phase set by the energy shifts ∆n and are
exponentially decaying set by various sums of the decay rates in Eqs. (D.37) through (D.39), while the
only time evolution for the diagonal elements is an exponential decay set by Γn .
105
106
Bibliography
[1] M. Arcari et al. Near-Unity Coupling Efficiency of a Quantum Emitter to a Photonic

Crystal Waveguide. Phys. Rev. Lett., 113:093603, Aug 2014.
[2] M. D. Barrett et al. Deterministic Quantum Teleportation of Atomic Qubits. Nature,

249:737–739, Jun 2004.
[3] J. Borregaard, P. Kómár, E. M. Kessler, A. S. Sørensen, and M. D. Lukin. Heralded

Quantum Gates with Integrated Error Detection in Optical Cavities. Phys. Rev. Lett.,
114:110502, Mar 2015.
[4] H. Bruus and K. Flensberg. Many-body quantum theory in condensed matter physics - an
introduction. Oxford University Press, 2004.
[5] J. I. Cirac and P. Zoller. Quantum Computations with Cold Trapped Ions. Phys. Rev.
Lett., 74:4091–4094, May 1995.
[6] J. Dalibard, Y. Castin, and K. Mølmer. Wave-function approach to dissipative processes in

quantum optics. Phys. Rev. Lett., 68:580–583, Feb 1992.
[7] D. P. DiVincenzo. The Physical Implementation of Quantum Computation. Fortschritte

der Physik, 48(9-11):771–783, 2000.
[8] J. Dreiser, M. Atatüre, C. Galland, T. Müller, A. Badolato, and A. Imamoglu. Optical

investigations of quantum dot spin dynamics as a function of external electric and magnetic
fields. Phys. Rev. B, 77:075317, Feb 2008.
[9] H. T. Dung, L. Knöll, and D.-G. Welsch. Resonant dipole-dipole interaction in the presence
of dispersing and absorbing surroundings. Phys. Rev. A, 66:063810, Dec 2002.
[10] J. Eisert and M. M. Wolf. Quantum computing. In Handbook of nature-inspired and

innovative computing, pages 253–286. Springer US, 2006.
[11] R. P. Feynman. Simulating Physics with Computers. International Journal of Theoretical

Physics, 21:467–488, June 1982.
[12] N. Gisin, G. Ribordy, W. Tittel, and H. Zbinden. Quantum cryptography. Rev. Mod.
Phys., 74:145–195, Mar 2002.
[13] P. Lodahl, S. Mahmoodian, and S. Stobbe. Interfacing single photons and single quantum
dots with photonic nanostructures. Rev. Mod. Phys., 87:347–400, May 2015.
[14] K. H. Madsen, S. Ates, J. Liu, A. Javadi, S. M. Albrecht, I. Yeo, S. Stobbe, and P. Lo-
dahl. Efficient out-coupling of high-purity single photons from a coherent quantum dot in
a photonic-crystal cavity. Phys. Rev. B, 90:155303, Oct 2014.
107
[15] G. E. Moore. Cramming more components onto integrated circuits. Electronics, 38, April
1965.
[16] L. Novotny and B. Hecht. Principles of Nano-Optics. Cambridge University Press, 2006.
[17] J. Preskill. Lecture notes for physics 229: Quantum information and computation, 1998.
[18] F. Reiter and A. S. Sørensen. Effective operator formalism for open quantum systems.
Phys. Rev. A, 85:032111, Mar 2012.
[19] R. Rudgley. The Lost Civilizations of the Stone Age. A Touchstone Book. Free Press, 2000.
[20] J. Sakurai and J. Napolitano. Modern Quantum Mechanics. Addison-Wesley, 2011.
[21] J. M. Smith, P. A. Dalgarno, R. J. Warburton, A. O. Govorov, K. Karrai, B. D. Gerardot,

and P. M. Petroff. Voltage Control of the Spin Dynamics of an Exciton in a Semiconductor
Quantum Dot. Phys. Rev. Lett., 94:197402, May 2005.
[22] Y. A. Vlasov et al. Active control of slow light on a chip with photonic crystal waveguides.
Nature, 438:65–69, Nov 2005.
[23] C. P. Williams. Quantum Gates, pages 51–122. Springer London, London, 2011.
[24] P. Yao, V. Manga Rao, and S. Hughes. On-chip single photon sources using planar photonic
crystals and single quantum dots. Laser & Photonics Reviews, 4(4):499–516, 2010.
108

Quantum Gates in Photonic Waveguides

Uploaded by

Copyright:

Available Formats

Quantum Gates in Photonic Waveguides

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Quantum Gates in Photonic Waveguides

Uploaded by

Copyright:

Available Formats

UNIVERSITY OF COPENHAGEN

Quantum Gates in Photonic Waveguides

Academic advisor: Anders S. Sørensen

2 Quantum Mechanics & Quantum Information 5

3 Basics of Nanophotonic Structures 19

4 Decay Properties of Coupled Quantum Emitters 35

7 Conclusion and Outlook 77

A Effective Operator Formalism 79

B Calculation of Effective Operators 91

C Additional Contour Plots 97

D Solving The Effective Master Equation 103

4.1 Level Structure of the Auxiliary and Qubit Quantum Dots . . . . . . . . . . . . . 37

5.1 Ideal Probability & Fidelity of a CZ Gate . . . . . . . . . . . . . . . . . . . . . . 55

The outline of the thesis is

Chapter 2: Provides a brief introduction to basic concepts of quantum mechanics and

Chapter 3: Contains an introduction to the physical system under consideration in thus

Quantum Mechanics & Quantum

2.1 Basics of Quantum Mechanics

• Linearity: hφ| (a |ψ2 i + b |ψ2 i) = a hφ|ψ1 i + b hφ|ψ2 i.

• Skew symmetry: hφ|ψi = hψ|φi∗ .

2.1.2 Observables and Measurement

M̂ : |ψi → M̂ |ψi , M̂ (a |ψi + b |φi) = aM̂ |ψi + bM̂ |φi . (2.4)

The adjoint M̂ † of an operator is defined by:

hφ|M̂ ψi = hM̂ † φ|ψi , (2.5)

hM̂ i = hψ|M̂ |ψi . (2.6)

The Schrödinger Picture

|ψ (t)i = e−iĤt/~ |ψ (t = 0)i , (2.11)

The Heisenberg Picture

hψ (t) |M̂ |ψ (t)i = ψ0 eiHt M̂ e−iHt ψ0 = hψ0 |M̂ (t) |ψ0 i ,

The Interaction Picture

Ĥ (t) = Ĥ0 + V̂ (t) . (2.14)

|ψ̃ (t)i = eiĤ0 t |ψi (2.15)

2.1.4 Non-Unitary Time Evolution & Quantum Jump Formalism

with the effective Hamiltonian and Lindblad operator:

|ψi = α |0i + β |1i , (2.32)

2.2.3 Universal Quantum Gates

Consider the following matrix:

x = r sin (θ) cos (φ) (2.38)

Rz ( − κ) · Ry (θ) · Rz (− [ + κ]) (2.47)

Rx (θ) = Rz (±π/2) Ry (θ) Rz (∓π/2) . (2.48)

H = Rz (0) · Ry (π/2) · Rz (π) · Rph (π/2) . (2.49)

UCNOT = (11 ⊗ H2 ) Uπ(2) (H1 ⊗ H2 ) ,

2.2.5 A Qubit Specific Measurement Capability

2.2.6 Long Decoherence Times

2.2.8 Quantum Error Correction

(|000i + |111i) (|000i + |111i) (|000i + |111i)

|000i + |111i |000i − |111i

(|000i + |111i) (|010i − |101i) (|000i + |111i)

(|000i + |111i) (|000i − |111i) (|000i + |111i)

Basics of Nanophotonic Structures

3.1 Nanophotonic Waveguides

3.1.2 Optical Properties

Confinement in 3D: Quantum dot − g0D = 2δ (E − Ej ) , (3.10)

3.2.2 Effective Mass Theory & the Transition Matrix Element

|ψc,v i = |χc,v i |uc,v i |αc,v i , (3.11)

P = hFv |Fc i huv |p|uc i hαv |αc i , (3.15)

|↑i ≡ |uc i |↑e i = |us i |↑e i (3.16)

and the hole pseudospin states are

3.2.3 Optical Properties of Quantum Dots

ρb = Af e−γf t + As e−γs t , (3.26)

Rz ( − κ) · Ry (θ) · Rz (− [ + κ]) (2.47)

5.2.3 Effective Phase Evolution for C 1

Ω ∼ 2πaf −1 (δx, κ) γ 0 C, a 4π. (6.48)

The perturbative expansion of ρ̃ is written in terms of the small parameter :