Upload
others
View
3
Download
1
Embed Size (px)
Citation preview
MICROMOTION IN TRAPPED
ATOM-ION SYSTEMS
LE HUY NGUYEN
B.Sc. (Hons.), NUS
A THESIS SUBMITTED FOR
THE DEGREE OF DOCTOR OF PHILOSOPHY
NUS Graduate School for Integrative Sciences and Engineering
National University of Singapore
2012
Declaration
I hereby declare that this thesis is my original work and it has been written by
me in its entirety. I have duly acknowledged all the sources of information which
have been used in the thesis.
This thesis has also not been submitted for any degree in any university previ-
ously.
Acknowledgments
I would like to express my utmost gratitude and appreciation to my supervisor,
Professor Berthold-Georg Englert, for his support, encouragement and patience.
Were it not for his guidance and advice I would not have been able to complete
this research project on time. Also of great importance is the contribution of my
collaborator Amir Kalev, who has worked with me on writing the numerical codes
and many analytical calculations. His inputs in our discussions are vital to the
completion of the project. I appreciate very much the help of Professor Murray
D. Barrett, who introduced the micromotion problem to me and provided valuable
information related to the experimental aspects of the problem. Murray also par-
ticipated in many discussions and suggested many improvements in the writing of
the manuscript. Special thank is due to Z. Idziaszek and T. Calarco for reading
through the manuscript and giving helpful comments. I wish to thank Professor
Gong Jiangbin for his interesting remark on the chaotic behaviour of the trapped
atom-ion system. I would also like to acknowledge the useful discussions and kind
hospitality of P. Zoller, P. Rabl, and S. Habraken at the Institute for Quantum
Optics and Quantum Information (IQOQI). I am grateful to Ua for drawing some
of the figures and Kean Loon for showing me how to prepare my thesis in Latex.
This work is financially supported by the NUS Graduate School for Integrative Sci-
ences and Engineering (NGS) and the Centre for Quantum Technologies (CQT).
Finally, I want to say thank to my family and Ua for their understanding and
i
ii ACKNOWLEDGMENTS
support throughout the period of my graduate study.
The main results presented in this thesis were published in Phys. Rev. A 85,
052718 (2012).
Contents
Acknowledgments i
Abstract v
List of Figures vii
List of Symbols ix
List of Abbreviations xiii
1 Introduction 1
2 Motion of a trapped ion 9
2.1 Ion micromotion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 The harmonic approximation . . . . . . . . . . . . . . . . . . . . . 13
3 Trapped atom-ion systems 15
4 Floquet formalism 21
4.1 Floquet theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2 Floquet Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.3 Resonance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
iii
iv CONTENTS
5 Micromotion effect in one dimension 35
5.1 The unperturbed system . . . . . . . . . . . . . . . . . . . . . . . . 36
5.2 Atom-ion quantum gate . . . . . . . . . . . . . . . . . . . . . . . . 47
5.3 Micromotion-induced coupling . . . . . . . . . . . . . . . . . . . . . 52
5.4 Numerical calculation of the quasienergies . . . . . . . . . . . . . . 59
5.5 Adiabatic evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.6 A scheme to bypass the micromotion effect . . . . . . . . . . . . . . 77
5.7 Excess micromotion . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.8 A more realistic 1D model . . . . . . . . . . . . . . . . . . . . . . . 86
5.9 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
6 Micromotion effect in three dimensions 93
6.1 Micromotion Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . 93
6.2 Quasi-1D trap configuration . . . . . . . . . . . . . . . . . . . . . . 96
6.3 Minimization of micromotion effect . . . . . . . . . . . . . . . . . . 102
6.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
7 Conclusions 107
A Numerov method 111
B Floquet states of a trapped ion 117
C Floquet adiabatic process 125
D Derivation of Eq. (6.26) 131
Abstract
We examine the validity of the harmonic approximation, where the radio-frequency
ion trap is treated as a harmonic trap, in the controlled collision of a trapped atom
and a single trapped ion. This is equivalent to studying the effect of the micromo-
tion since this motion must be neglected for the trapped ion to be considered as a
harmonic oscillator. By applying the transformation of Cook et al. we find that the
micromotion can be represented by two periodically oscillating operators. In order
to investigate the effect of the micromotion on the dynamics of a trapped atom-
ion system, we calculate (i) the coupling strengths of the micromotion operators
by numerical integration and (ii) the quasienergies of the system by applying the
Floquet formalism, a useful framework for studying periodic systems. It turns out
that the micromotion is not negligible when the distance between the atom trap
and the ion trap is shorter than a characteristic distance. Within this range the
energy diagram of the system changes dramatically when the micromotion is taken
into account. The system exhibits chaotic behaviour through the appearance of
numerous avoided crossings in the energy diagram when the micromotion coupling
is strong. Excitation due to the micromotion leads to undesirable consequences
for applications that are based on an adiabatic process of the trapped atom-ion
system. We suggest a simple scheme for bypassing the micromotion effect in order
to successfully implement a quantum-controlled phase gate proposed previously
and create an atom-ion macromolecule. The methods presented in this thesis are
v
vi ABSTRACT
not restricted to trapped atom-ion systems and can be readily applied to studying
the micromotion effect in any system involving a single trapped ion.
List of Figures
1.1 Adiabatic collision of a trapped atom and a trapped ion . . . . . . . 2
1.2 Two-qubit gates for quantum computing with neutral atoms in op-
tical lattices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Mesocopic molecular ions . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1 Diagram of the linear Paul trap . . . . . . . . . . . . . . . . . . . . 9
2.2 Classical motion of a trapped ion . . . . . . . . . . . . . . . . . . . 12
3.1 Trapped atom-ion system . . . . . . . . . . . . . . . . . . . . . . . 15
3.2 Atom-ion interaction potential . . . . . . . . . . . . . . . . . . . . . 17
4.1 Quasienergies of a two-level system . . . . . . . . . . . . . . . . . . 33
5.1 Eigenstates of the unperturbed Hamiltonian . . . . . . . . . . . . . 39
5.2 Energy diagram of the unperturbed system . . . . . . . . . . . . . . 44
5.3 Molecular states and vibrational states . . . . . . . . . . . . . . . . 46
5.4 Energy difference between the paralel and anti-paralel qubit combi-
nations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.5 Resonances in the unperturbed system . . . . . . . . . . . . . . . . 57
5.6 Micromotion-induced coupling strength . . . . . . . . . . . . . . . . 58
5.7 Quasienergies of the trapped atom-ion system . . . . . . . . . . . . 70
vii
viii LIST OF FIGURES
5.8 Quasienergies of the trapped atom-ion system obtained when the
dominant term of the micromotion is neglected . . . . . . . . . . . . 71
5.9 Change of the trap distance during an adiabatic evolution . . . . . 73
5.10 Evolution without the micromotion . . . . . . . . . . . . . . . . . . 74
5.11 Evolution with the micromotion . . . . . . . . . . . . . . . . . . . . 76
5.12 Coupling strength of the excess micromotion . . . . . . . . . . . . . 85
6.1 Micromotion-induced coupling strength in a quasi-1D system . . . . 102
List of Symbols1
a Ion trap parameter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
[a, b] ≡ ab− ba, commutator. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .*
a, b ≡ ab+ ba, anti-commutator. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . *
α Atomic polarizability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
b The s-wave scattering length of atom-ion collisions . . . . . . . . . . . . . . . . 38
d Distance between the centers of the atom trap and the ion trap . . . 36
dc Characteristic trap distance of trapped atom-ion collisions . . . . . . . . 42
dmm Characteristic trap distance of the micromotion effect . . . . . . . . . . . . . 56
E Eigenenergy of the unperturbed system . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Eac ac electric field associated with the excess micromotion . . . . . . . . . . . .82
Edc dc electric field associated with the excess micromotion . . . . . . . . . . . 82
ε Quasienergy of a periodic system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
γ Ion trap parameter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
H Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . *
H0 Hamiltonian of the unperturbed system . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Hex Hamiltonian of the excess micromotion . . . . . . . . . . . . . . . . . . . . . . . . . . . .82
HF Floquet Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .22
Hmm Hamiltonian of the micromotion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
la Atom harmonic oscillator length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
1The page where a given symbol are defined/introduced is listed at the rightmost column.When the definition is general, the page number is given as *.
ix
x LIST OF SYMBOLS
li Ion harmonic oscillator length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
lac Length scale associated with the excess micromotion . . . . . . . . . . . . . . 84
lrel Harmonic oscillator length of the reduced mass . . . . . . . . . . . . . . . . . . . 89
m Reduced mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
ma Atom mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
mi Ion mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .16
M Total mass of the atom-ion system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
µ Quasienergy spacing of a single trapped ion . . . . . . . . . . . . . . . . . . . . . . . 55
ω Micromotion frequency of the trapped ion . . . . . . . . . . . . . . . . . . . . . . . . .11
ω0 Secular frequency of the trapped ion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
ωa Atom trapping frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Ω Rabi frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
P Momentum operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . *
Ψ(x, t) Position wave function of the state |Ψ(t)〉 . . . . . . . . . . . . . . . . . . . . . . . . . . *
Φ(x, t) Effective wave function in the transformed picture . . . . . . . . . . . . . . . . 13
ϕs Short-range phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
q Ion trap parameter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
r Atom-ion distance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
~ra Position vector of the atom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . *
~ri Position vector of the ion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . *
ρ(x, t) Probability density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .*
R Atom-ion interaction length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .16
Ri Redefined atom-ion interaction length . . . . . . . . . . . . . . . . . . . . . . . . . . . . .36
T = 2π/ω, Period of the RF driving field . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
θ Adiabatic phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
u(t) Floquet state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
U(t2, t1) Time evolution operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
xi
X Position operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . *
Vint Atom-ion interaction potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
〈〈u|v〉〉 Scalar product of the extended Hilbert space . . . . . . . . . . . . . . . . . . . . . . 23
| . . .〉 A ket that represents a quantum state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . *
xii LIST OF SYMBOLS
List of Abbreviations
1D One-dimensional
2D Two-dimensional
3D Three-dimensional
4D Four-dimensional
ac Alternating current
AC Avoided crossing
BEC Bose-Einstein condensate
cm Center-of-mass
cp coupling
dc direct current
eff effective
ex excess
HO Harmonic oscillator
LZ Landau-Zener
mm Micromotion
ODE Ordinary differential equation
quasi-1D Quasi one-dimensional
rel Relative
rf radio-frequency
xiii
xiv LIST OF ABBREVIATIONS
Chapter 1
Introduction
The progress of quantum information and quantum computation has shown promises
of novel technologies such as quantum cryptography, quantum teleportation and
quantum computers [1]. Although the principles of quantum computation are
more or less well established, the physical realization [2, 3] of proposed applica-
tions is still a formidable task. Two of the most well-known systems for quantum
computing are subjects of atomic physics: trapped ions [4–6] and cold atoms [7].
Experimental advances in atomic physics have been one of the main driving forces
behind the rapid development in the implementation of quantum technologies in
the past decade.
This strong interplay between atomic physics and quantum computation laid
the foundation for the emergence of the recent research trend in ultracold atom-ion
interaction. As a result of the great importance of trapped ions and cold atoms,
many experimental techniques have been developed to enable the manipulation of
individual ions and atoms with extreme precision and control. Magneto-optical
traps [8] and optical lattices [9] were invented for trapping atoms, while ions can
be stored in Paul traps [10] or Penning traps [11]. Various cooling techniques such
as laser cooling and evaporative cooling [12] are then used to cool the atoms and
1
2 CHAPTER 1. INTRODUCTION
Figure 1.1: Controlled collision of a trapped atom and a single-trapped ion: (a)The particles are initially prepared in the motional ground states; (b) the traps aremoved toward each other and the atom collides with the ion when the wave functionsbegin to overlap; (c) the two particles end up in some final motional states after thecollision. While the final motional state is approximately the same as the initial statewhen the collision process is adiabatic, the internal states can be in an entangledsuperposition.
ions to the ultracold regime where collision processes are studied.
Perhaps the best way to appreciate the attractiveness of ultracold atom-ion
collisions is looking at some of the interesting proposals based on trapped atom-ion
systems. Most of these rely on the strong collisional interaction between trapped
ions and atoms. The first application considered has direct implication for quantum
computing: an atom-ion entangling gate [13, 14]. The entanglement of a trapped
atom and a trapped ion is achieved through the adiabatic collision process shown
in Fig. 1.1. If the qubits are encoded in the hyperfine binaries of the atom and the
ion, it is shown that this collision process is equivalent to the implementation of a
Controlled Phase gate in the qubit space.
This proposal is appealing because it has the potential of filling a crucial gap in
quantum computing with neutral atoms in optical lattices. It is known that one-
qubit gates can be implemented by utilizing Raman transitions; however, a two-
qubit gate between two particular atomic qubits remains a challenge. Although a
3
Figure 1.2: A trapped ion is used as a bus for implementing two-qubit gatesbetween a single pair of atomic qubits stored in an optical lattice.
Controlled Not gate was demonstrated by Mandel et al. [15] based on the method
proposed in Ref. [16], many pairs of qubits were involved. The difficulty of realizing
a two-qubit gate between a single pair of qubits can be overcome by applying the
atom-ion entangling process described above. An optical lattice with atoms stored
as qubits is shown in Fig. 1.2. The trapped ion can be used as a bus in the following
sense: We first entangle a particular atom in the optical lattice with the trapped
ion, then we move the ion trap towards another atom, carry out the entanglement
again and by doing so effectively implement a two-qubit gate between the two
atoms. This two-qubit entangling gate can be combined with single qubit gates to
realize an arbitrary quantum logic operation [17].
While the above example shows promises for quantum information processing,
the next proposal considered — the creation of a mesocopic atom-ion molecule
— is of great interest in atomic and molecular physics. Cote et al. suggested in
Ref. [18] that when an ion is immersed in a Bose-Einstein Condensate (BEC), the
atoms in the surrounding are attracted to the ion to form a large molecular ion
4 CHAPTER 1. INTRODUCTION
(see Fig. 1.3). It is predicted that the number of captured atoms can be as large
as a few hundreds for the example of a sodium ion in a BEC of sodium atoms.
Other intriguing proposals based on atom-ion systems include sympathetic
cooling of ions [19, 20] and molecules that do not allow laser cooling [21–23], as
well as a scanning microscope for probing local properties of an ultracold atomic
gas [24, 25]. The sympathetic cooling of a trapped ion by cold atoms in a BEC is
suggested as a model for continuous cooling of quantum computers [29]. Moreover,
The ultracold collisions of trapped ions with atoms may enable the investigation
of entanglement in hybrid systems and the decoherence of a particle coupled to a
quantum environment [29]. A mixture of positive ions and neutral atoms is also
predicted to exhibit interesting phenomena such as the transition from an almost
insulating to a conducting phase at ultralow temperature [26].
Figure 1.3: A trapped ion (red) is immersed in a Bose-Einstein Condensate. Theatoms (blue) in the surrounding are attracted to the ion and form a large molecularion.
Many experiments have been done on trapped atom-ion systems. Grier et
al. [27, 28] observe the charge-exchange collisions between a cloud of Yb+ ions
and a cloud of Yb atoms. Zipkes et al. study the elastic and inelastic collision
processes of a trapped Yb+ ion immersed in a BEC of Rb atoms [29, 30]; a similar
5
investigation for Ba+ ion and Rb atoms is carried out by Schmid et al. [31].
Although none of these experiments reaches the ultracold temperature regime, they
provide important information on the characteristics of cold atom-ion collisions in
traps, particularly the scattering cross section for different types of collision. The
experiments on a single trapped ion immersed in a BEC also give evidence for the
sympathetic cooling of the trapped ion.
On the theoretical aspects, a great effort has been made on studying the ultra-
cold collisions of free atoms and ions [32–35], but the situation when the particles
are trapped, which is relevant in experiments, has not been explored with compa-
rable depth. The first quantum treatment of atom-ion interaction in traps is given
by Idziaszek and his collaborators. They study the controlled collision in a trapped
atom-ion system [13, 14] and propose this system as the mean for implementing
the quantum phase gate. The system considered is composed of a single atom
trapped in a harmonic trap interacting with an ion trapped in a radio-frequency
(rf) trap.
In these studies, the rapid motion of the ion on a short time-scale — the
micromotion — is averaged out. In the literature, this procedure is referred to
as the harmonic approximation since it produces an effective harmonic motion of
the ion, as if the ion were trapped in a time-independent harmonic trap. In fact,
many of the proposals for applications of this system are derived with the aid of
the harmonic approximation. However, there is a concern about the validity of this
approximation since the kinetic energy of the micromotion can be comparable or
even much larger than that of the ion’s harmonic motion [36]. More importantly,
it is found in experiments that the ion’s micromotion plays an important role in
the dynamics of atom-ion collisions [27, 29, 31].
Although the effect of the micromotion in the collision of a trapped ion with
a cloud of atoms was considered previously. These works are classical and semi-
6 CHAPTER 1. INTRODUCTION
classical treatments and hence are valid only for collision energies well above the
ground energy [37, 38]. A quantum mechanical description is essential for under-
standing the role of micromotion in the ultracold collisions of trapped atoms and
trapped ions since quantum effect dominates at very low temperature. Motivated
by such concerns, we present here two approaches, as mentioned in the Abstract,
to studying the effect of the micromotion in trapped atom-ion systems. The first is
based on the calculation of the micromotion-induced coupling strengths, and the
second, which is exact but much more time-consuming, is the computation of the
quasienergies of the system.
For this purpose we make use of the Floquet formalism [39]. The potential
of the rf electric field used to trap ions depends periodically on time. When the
potential is periodic, Floquet theory provides a powerful tool for treating the ex-
act dynamics of the quantum system. It enables one to compute the so-called
quasienergies and quasienergy states, also referred to as Floquet states, which are
the analogs of the eigenenergies and eigenstates of a time-independent system.
Hence it offers a quantum-mechanical treatment of the micromotion problem in
any system involving a single trapped ion. By studying the exact quasienergies
and Floquet states we obtain valuable information about the role of the micro-
motion in such systems. Its effect will then be deduced by comparing the exact
dynamics, as derived by the Floquet formalism, with the approximate one of the
harmonic approximation. In particular, this comparison reflects on the validity of
various proposals for applications using controlled collisions of trapped atoms and
ions.
Although we demonstrate the method for the system of interacting trapped
atoms and ions, the formalism can be readily applied to any system in which
the trapped ion is coupled to an external time-independent subsystem. We see
later that, within the Floquet formalism, the Hamiltonian of a periodic system is
7
strikingly similar to that of a time-independent system. Therefore, provided that a
time-independent treatment exists for a given system (when the ion is considered as
a harmonic oscillator), Floquet formalism can be used to generalize this treatment
to account for the ion’s micromotion.
Since many different aspects are involved, this work is presented in steps with
increasing layers of complexity. In chapters 2 and 3 we describe the ion’s micro-
motion and the trapped atom-ion systems. Chapter 4 is on the Floquet formalism
and how we use it to study the effect of a periodically oscillating potential on
the energy structure of a quantum system. In chapter 5 we study the micro-
motion effect in one-dimensional (1D) trapped atom-ion systems by considering
the micromotion-induced coupling and computing the exact quasienergies of the
system. The quantum motion of a trapped ion in the presence of the so-called
excess micromotion is derived and the effect of this excess micromotion in trapped
atom-ion systems is considered. We also suggest a simple scheme to bypass the
micromotion effect in order to realize some interesting applications based on the
interaction of trapped atoms and ions. After studying three-dimensional (3D) sys-
tems in chapter 6, we offer conclusions and present some technical material in the
appendices.
8 CHAPTER 1. INTRODUCTION
Chapter 2
Motion of a trapped ion
The classical and quantum dynamics of a single trapped ion and its interaction
with a radiation field have been studied extensively [40–43]. Here we describe the
key points in the motion of trapped ions which are relevant to our study and then
explain the reasoning behind the harmonic approximation. The reader may refer
to Refs. [6, 42, 43] for other interesting theoretical and experimental aspects of
trapped ions.
Figure 2.1: A schematic diagram of the linear Paul trap. The oscillating potentialsare applied to the four cylindrical electrodes A, B, C, and D. The endcaps are con-nected to a dc potential to provide axial confinement. In their equilibrium positions,the trapped ions form a straight line along the axial direction (the z axis).
9
10 CHAPTER 2. MOTION OF A TRAPPED ION
2.1 Ion micromotion
Since it is not possible to use a static electric field to trap a charged particle, a
combination of a static field and an oscillating field is used to trap ions in a rf trap.
Let us denote the driving frequency by ω; the total trapping potential is given by
Vrf(x, y, z, t) =U
2
(αxx
2 + αyy2 + αzz
2)
+U ′
2cos(ωt)
(βxx
2 + βyy2 + βzz
2), (2.1)
with U,U ′ > 0. The above potential must obey Laplace’s equation, hence the
coefficients αk and βk (k = x, y, z) must satisfy the conditions
∑k
αk =∑k
βk = 0. (2.2)
In the particular case of a linear Paul trap, these coefficients are related by
αx = αy = −1
2αz < 0,
βx = −βy, βz = 0. (2.3)
From the form of the trapping potential we see that the equations of motion of
the ion along the x, y, z axes are decoupled. To simplify the presentation we will
only describe the motion along the x axis. It is straightforward to derive similar
results for the motion along the other directions. With mi denoting the ion’s mass,
the Hamiltonian for the motion along the x axis is
H(t) =P 2
2mi
+1
8miω
2X2[a+ 2q cos(ωt)], (2.4)
where X and P are the position operator and the momentum operator; a and q
are trap parameters which depend on mi, ω, the ion’s charge Q and the coefficients
2.1. ION MICROMOTION 11
of the trapping potential through the relations
a =4QUαxmiω2
, q =4QU ′βxmiω2
. (2.5)
The Hamiltonians for the motion along the y and z directions assume similar forms
with different values for the parameters a and q. In a linear Paul trap, we have as
a consequence of Eq. (2.3)
ax = ay = −1
2az,
qx = −qy, qz = 0, (2.6)
which means the oscillating part of the trapping potential only exists in the radial
directions (x, y). Since we will only consider the motion of the ion along the x
axis, we have dropped the subscript x and write the relevant trap parameters as
a and q. The typical values encountered in experiments for these trap parameters
are |q| 1 and |a| q2. In this work we assume these conditions for the trap
parameters unless stated otherwise.
Using the Hamiltonian of Eq. (2.4) and writing down the Hamilton’s equations
of motion one can derive the classical equation of motion
x(t) + [a+ 2q cos(ωt)]ω2
4x(t) = 0. (2.7)
This is a differential Mathieu equation for which the stability conditions are dis-
cussed in Ref. [44, 45]. Normally one chooses a and q such that they are within
the lowest stability region where a ≈ 0. It can be shown that the ion’s motion is
a combination of a harmonic oscillation at the secular frequency ω0 and the mi-
cromotion that oscillates at the frequency ω of the driving potential. The secular
12 CHAPTER 2. MOTION OF A TRAPPED ION
frequency is given by
ω0 =ω
2
√a+
q2
2; (2.8)
thus, ω0 ω; that is, the secular motion is much slower than the micromotion.
These conclusions are drawn from the following approximate solution of Eq. (2.7)
x(t) ≈ x0 cos(ω0t)[1 +
q
2cos(ωt)
](2.9)
with x0 an arbitrary constant [42].
The form of x(t) in Eq. (2.9) shows that the amplitude of the micromotion
goes as the small parameter q, hence it can be seen as a jiggling motion around
the overall path of the secular motion. The classical motion of the trapped ion
is demonstrated in Fig. 2.2 in which the fast oscillation along the curve comes
from the micromotion. When one averages over the short time period of the
micromotion, the cos(ωt) term in Eq. (2.9) vanishes and the resulting motion of
the ion is a harmonic oscillation. This is a classical derivation of the so-called
harmonic approximation which is the main topic of the next section.
Figure 2.2: The classical motion of the trapped ion illustrated by a plot of x(t).The fast oscillation along the curve results from the micromotion.
2.2. THE HARMONIC APPROXIMATION 13
2.2 The harmonic approximation
Cook, Shankland, and Wells provide a quantum-mechanical derivation of the har-
monic approximation in Ref. [46]. They start with the Schrodinger equation for
the Hamiltonian of Eq. (2.4)
i~∂
∂tΨ(x, t) =
− ~2
2mi
∂2
∂x2+
1
8miω
2x2[a+ 2q cos(ωt)]
Ψ(x, t), (2.10)
and come up with a clever observation. They first ignore all the time-independent
terms in the Hamiltonian, the solution is then straightforward:
Ψ(x, t) = exp
[− i
4~miqωx
2 sin(ωt)
]Ψ(x, 0). (2.11)
Cook et al. intuitively guess that the main effect of the oscillating term is to intro-
duce the above exponential phase to the wave function; thus they are motivated
to write the wave function of the ion as
Ψ(x, t) = exp
[− i
4~miqωx
2 sin(ωt)
]Φ(x, t) (2.12)
and insert this into the Schrodinger equation. It follows that the effective wave
function Φ(x, t) obeys the modified Schrodinger equation
i~∂
∂tΦ(x, t) =
[− ~2
2mi
∂2
∂x2+
1
2miω
20x
2 −mi(γω0)2x2 cos(2ωt)
+ 2i~γω0
(x∂
∂x+
1
2
)sin(ωt)
]Φ(x, t), (2.13)
where ω0 is the secular frequency mentioned above and the factor γ is
γ =1√
2(
1 + 2aq2
) . (2.14)
14 CHAPTER 2. MOTION OF A TRAPPED ION
Cook et al. argue that most of the fast time dependence of Ψ(x, t) is contained
in the exponential factor and hence Φ(x, t) may be treated as a slowly varying
function of time. Therefore, one may take in Eq. (2.13) the time average over
the short time interval 2π/ω of the micromotion; after that we are left with the
well-known Schrodinger equation for a harmonic oscillator. This is the quantum-
mechanical basis for the harmonic approximation. Equation (2.13) tells us that
this approximation is valid only when the time-dependent terms have little effect
on the unperturbed wave function. As we see later, while this is all right for a
single trapped ion, it may not hold when the ion is coupled to an external system.
The Hamiltonian associated with the effective wave function Φ(x, t) can be read
off from Eq. (2.13),
Heff(t) =P 2
2mi
+1
2miω
20X
2 +Hmm(t), (2.15)
where
Hmm(t) = −mi(γω0)2X2 cos(2ωt)− γω0X,P sin(ωt) (2.16)
accounts for the time-dependent contribution of the micromotion. Here X,P is
the anti commutator of the position and momentum operators. The form of Heff ,
where the time-independent part of the secular motion and the time-dependent one
of the micromotion are separated, makes it more convenient to work with when
one wishes to study the effect of the micromotion. In the next chapter we make
use of this transformation to investigate the trapped atom-ion system.
Chapter 3
Trapped atom-ion systems
The system is composed of an atom in a harmonic trap interacting with an ion
in a rf trap as shown in Fig. 3.1. To simplify the problem we first consider the
one-dimensional system in which the atom and ion are confined to moving along
only one axis, say the x axis. The realistic three-dimensional system is discussed
in chapter 6 below. Let us choose the coordinate origin at the center of the atom
trap. When the trap centers are separated by a distance d, the Hamiltonian of the
system is
H(t) = Ha +Hi(t) + Vint, (3.1)
Figure 3.1: A trapped atom-ion system. O is the center of the atom trap which ischosen as the coordinate origin; ~ra and ~ri are the position vectors of the atom andion, respectively; and ~d is the position vector of the center of the ion trap.
15
16 CHAPTER 3. TRAPPED ATOM-ION SYSTEMS
where Ha is the Hamiltonian of an atom in an atom trap, Hi(t) is the Hamiltonian
of an ion in a rf trap, and Vint is the interaction potential between the atom and
the ion. More specifically, we have
H(t) =P 2a
2ma
+1
2maω
2aX
2a +
P 2i
2mi
+1
8miω
2(Xi − d)2[a+ 2q cos(ωt)] + Vint. (3.2)
In the above equation, the indices a and i label the atom and ion, respectively.
The properties of the interaction potential Vint are well-documented in Ref. [13].
When the distance r between the atom and the ion is large, this potential is the
polarization potential which has the form Vint(r) ≈ −(αe2)/(2r4), where α is the
polarizability of the atom. This long-range form results from the attractive force
between the ion’s charge and the atom’s induced dipole. The characteristic length
of the atom-ion interaction is defined by R =√mαe2/~2, where m is the reduced
mass of the atom-ion system. This length indicates the range within which the
atom-ion interaction potential is larger than the quantum kinetic energy ~2/2mr2.
A list of the theoretical calculations and experimental measurements for the
atomic polarizabilities of various elements is given in Ref. [47]. In table 3.1 are the
ground-state polarizabilities [in units of (4πε0)−1 × 10−24 cm3] of a few commonly
used elements. The most accurate result for Rubidium was obtained with an
atom interferometer and is α = (4πε0)−1 × 47.24 × 10−24 cm3 [47, 48]; hence, the
interaction length R of the 135Ba+ and 87Rb system is around 5550 Bohr radii
(293 nm).
atom Na Ca Rb Cs Ybα 24.11 22.8 47.24 59.42 20.9
Table 3.1: Dipole polarizabilities for ground state atoms.
The behavior of the interaction potential at short distances is very different
from its long-range form. The short-range potential has a repulsive core which is
17
non-central in general and depends on the electronic configurations of the atom and
ion (see Fig. 3.2). This spin dependence of the short-range potential is key to the
implementation of a quantum phase gate proposed in Ref. [14] (here “spin” means
the binary alternative of two hyperfine states). Idziaszek et al. [13] show how to
take into account the effect of the short-range potential by utilizing quantum defect
theory where the short-range potential can be characterized by a single quantum
defect parameter called the short-range phase. The basic idea is to replace the
potential Vint with its long-range form while imposing a specific boundary condition
on the wave function at a small distance rmin R. In addition, this distance rmin
must be sufficiently larger than the length scale set by the short-range potential,
which is a few Bohr radii.
Figure 3.2: Atom-ion interaction potential. The long range form follows the inversepower-fourth law.
To investigate the micromotion effect, we follow Cook et al. and make the
transformation
Ψ(xa, xi, t)= exp
[− i
4~miqω(xi − d)2sin(ωt)
]Φ(xa, xi, t). (3.3)
The resulting effective Hamiltonian for the effective wave function Φ(xa, xi, t) is
18 CHAPTER 3. TRAPPED ATOM-ION SYSTEMS
the sum of a time-independent Hamiltonian and an oscillating term Hmm(t) that
represents the ion’s micromotion,
H(t) =P 2a
2ma
+1
2maω
2aX
2a +
P 2i
2mi
+1
2miω
20(Xi − d)2 − αe2
2(Xi −Xa)4+Hmm(t),
(3.4)
with
Hmm(t) =−mi(γω0)2(Xi − d)2 cos(2ωt)− γω0Xi − d, Pi sin(ωt). (3.5)
In the above we already replaced the interaction potential by its long-range form,
which is given by an inverse fourth-power law. The analytical solution of the
Schrodinger equation with this polarization potential can be expressed in terms of
the modified Mathieu functions [33, 49, 50]. The extension of such an attractive
singular potential to the origin results in unphysical wave functions and energy
spectrum [51–53]. However, an orthonormal set of eigenfunctions and a discrete
energy spectrum for the bound states are obtained if the potential is cut off at a
sufficiently small distance at which the wave functions share a fixed short-range
phase [51]. This cut-off distance and the short-range phase are determined by the
repulsive core of the atom-ion interaction potential. Any calculation of the wave
function must take into account the boundary condition dictated by the short-
range phase. The cut-off mechanism and the short-range phase are explained in
more detail in Sec. 5.1.
In the harmonic approximation, Hmm(t) is neglected and we only need to deal
with a time-independent system. This approximate approach has been studied
previously [13, 14]. Our purpose is to investigate how the micromotion affects the
unperturbed system; thus, we have to work with the full Hamiltonian of Eq. (3.4).
We are particularly interested in how the state of the system evolves in an adiabatic
19
process where the trap distance is changed slowly in time as such a process is vital
for the implementation of the quantum phase gate proposed in Ref. [14].
Although the micromotion is represented by an oscillating term similar to
an electromagnetic field, it is an intrinsic property of the system and cannot be
switched on or off. Therefore, time-dependent perturbation theory which empha-
sizes the transitions between unperturbed states is not a suitable approach to the
micromotion problem, which is the reason why we need to consider the Floquet
formalism.
20 CHAPTER 3. TRAPPED ATOM-ION SYSTEMS
Chapter 4
Floquet formalism
The Floquet formalism for a quantum system with a Hamiltonian periodic in time
was introduced by Shirley [54] and has been developed in great depth for studying
various periodic systems in atomic and molecular physics [39, 55–61]. Thus, we
have an advanced mathematical framework ready in our hands to investigate the
exact quantum dynamics of the trapped atom-ion system. Within the Floquet
formalism, time-periodic systems are similar in structure to time-independent sys-
tems. Therefore, it is convenient to use this formalism to generalize the theoretical
works which were done in the harmonic approximation for studying the effect of
the ion’s micromotion. In this chapter we outline the key points which are impor-
tant for our study and also explain how we apply them to address the micromotion
problem in general.
4.1 Floquet theory
The Hamiltonian of Eq. (3.4) satisfies the periodicity condition H(t+ T ) = H(t)
with T = 2π/ω. According to the Floquet theorem, the Schrodinger equation for
such a periodic system,
i~∂
∂tΨ(t) = H(t)Ψ(t), (4.1)
21
22 CHAPTER 4. FLOQUET FORMALISM
adopts a special class of solutions called the Floquet solutions which can be ex-
pressed in terms of the quasienergy ε and the Floquet wave function u(t) as
Ψ(t) = exp
(− i~εt
)u(t), (4.2)
where u(t) is a function of both time and space coordinates and is periodic in time;
that is,
u(t+ T ) = u(t). (4.3)
A substitution of Ψ(t) from Eq. (4.2) to Eq. (4.1) yields
[H(t)− i~ ∂
∂t
]u(t) = εu(t). (4.4)
Thus, the quasienergy and the Floquet wave function are respectively the eigen-
vector and eigenvalue of the operator
HF(t) = H(t)− i~ ∂∂t, (4.5)
which is called the Floquet Hamiltonian. One observes from Eq. (4.2) that quasiener-
gies and Floquet states are to a periodic system what eigenenergies and eigenstates
are to a time-independent system.
The solutions to the Floquet eigenvalue equation (4.4) has the following im-
portant Brillouin-zone-like structure: If u(t) is a Floquet state with quasienergy
ε, then u(t) exp(ikωt) is also a Floquet state with quasienergy ε + k~ω for any
integer k. These sates are physically equivalent because they belong to a unique
wave function, inasmuch as
Ψ(t) = u(t)e−i~ εt =
[u(t)eikωt
]e−
i~ (ε+k~ω)t. (4.6)
4.1. FLOQUET THEORY 23
In other words, Floquet states and quasienergies come in congruent classes modulo
ω, each containing an infinite number of equivalent members. It is convenient to
denote the Floquet states and their corresponding quasienergies by un,k and εn,k
where the index n indicates physically different classes and k different members in
a class. Since for each quasienergy ε there are equivalent quasienergies ε+k~ω, it is
always possible to reduce an arbitrary quasienergy to a single zone [E−~ω/2, E+
~ω/2] for an arbitrary energy E. As there can be only one member from each class
εn,k in a single zone, we need not worry about the redundancy of the physically
equivalent states if we restrict our study of the quasienergies to one zone.
Sambe [62] introduces the extended Hilbert space for all the square-integrable
periodic functions with period T in which the scalar product is defined as
〈〈u(~r, t), v(~r, t)〉〉 =1
T
∫ T
0
dt 〈u(~r, t), v(~r, t)〉 . (4.7)
This scalar product is a 4D generalization of the normal scalar product in 3D.
Time and space can be treated on equal footings in the extended Hilbert space.
Since the Floquet states are constant vectors in this space, an advantage of working
with it is that methods developed for time-independent quantum mechanics, such
as the Rayleigh-Schrodinger perturbation method and the variational principle can
be readily generalized for the Floquet states. The generalization is straightforward
thanks to the similarity between the Floquet eigenvalue equation (4.4) and the
time-independent Schrodinger equation.
The final properties of the Floquet states we need to mention are the orthonor-
mality
〈〈un,k, um,j〉〉 = δn,mδk,j, (4.8)
and the completeness, ∑n,k
|un,k〉〉〈〈un,k| = I, (4.9)
24 CHAPTER 4. FLOQUET FORMALISM
which comes from the fact that any square-integrable and T -periodic wave function
ψ(t) can be expanded as
ψ(t) =∑n,k
un,k(t)〈〈un,k, ψ〉〉, (4.10)
which is just a generalized Fourier series of ψ(t).
Recall that without the micromotion the trapped atom-ion system has a time-
independent Hamiltonian with well-documented eigenenergies and eigenstates [13,
14]. If the micromotion is included, the system is a periodic one that possesses
well-defined quasienergies and Floquet states. When the strength of the micromo-
tion Hmm(t) in Eq. (3.4) is reduced to zero, these quasienergies and Floquet states
must approach the eigenenergies and eigenstates obtained by the harmonic approx-
imation. The difference between the exact quasienergies (Floquet states) and the
approximate eigenenergies (eigenstates) tells us how important the micromotion
effect is.
To be more specific, let us consider the Hamiltonian of the trapped atom-ion
system which can be written as
H(t) = H0 +Hmm(t), (4.11)
where H0 is the unperturbed, time-independent part with eigenenergies En and
eigenstates φn. When the micromotion is taken into account these states will
transform to some Floquet states un(t) with quasienergies εn. An obvious way to
quantify the micromotion effect is to calculate the energy difference
δE = |εn − En| (4.12)
and the deviation between the exact, oscillating probability density ρn(t) = |un(t)|2
4.2. FLOQUET HAMILTONIAN 25
and the approximate, static density ρ(0)n = |φn|2. There are many ways to measure
the difference between these probability densities. An example is
δρn =1
T
∫dt
∫dx∣∣ρn(x, t)− ρ(0)
n (x)∣∣ , (4.13)
which is the average over time and space of the absolute difference between ρn(x, t)
and ρ(0)n (x). We see later that, for our particular problem, it suffices to calculate the
energy difference; however, the oscillation of the wave function may have important
effects in general.
4.2 Floquet Hamiltonian
In this section we show how to obtain the Floquet states and quasienergies by
the Floquet Hamiltonian method. A different but not less important approach
— the time-propagator method — is discussed later in Sec. 5.4. While both the
time-propagator and Floquet Hamiltonian methods are useful for the numerical
computation of quasienergies and Floquet states, the latter has the advantage
that it allows analytical studies based on perturbation theory.
Let us consider a system described by the Hamiltonian
H(t) = H0 + V cos(ωt), (4.14)
where V is a time-independent operator. The corresponding Floquet Hamiltonian
for this system is
HF(t) = H0 − i~∂
∂t+ V cos(ωt). (4.15)
Suppose we know the eigenenergy En and eigenstates |n〉 of the unperturbed Hamil-
tonian H0. Then the quasienergies and Floquet states of the unperturbed Floquet
26 CHAPTER 4. FLOQUET FORMALISM
Hamiltonian
H(0)F = H0 − i~
∂
∂t(4.16)
are simply
ε(0)n,k = En + k~ω,
u(0)n,k = |n〉 eikωt, (4.17)
for any integer k. We may now obtain the matrix elements of the full Floquet
Hamiltonian HF by utilizing the scalar product of the extended Hilbert space,
〈〈u(0)n,k|HF |u(0)
m,j〉〉 = (En + k~ω) δn,mδk,j +1
2〈n|V |m〉 (δk,j+1 + δk,j−1) . (4.18)
Diagonalizing the above infinite matrix gives the exact quasienergies and Floquet
states of the periodic system. In practice one needs to truncate the Floquet Hamil-
tonian matrix by choosing n = 1, . . . , Ne and k = −Nf , . . . , Nf for some sufficiently
large numbers Ne and Nf . Then, the linear size of the Floquet Hamiltonian matrix
is N = Ne(2Nf + 1).
Besides the numerical diagonalization of the Floquet Hamiltonian, we can also
use perturbation methods [62] to find the approximate quasienergies and Flo-
quet states when the oscillating potential is weak. According to the Rayleigh-
Schrodinger perturbation method generalized for Floquet states, the lowest-order
corrections for the quasienergies and Floquet states are
ε(1)n,k = 〈〈u(0)
n,k|V (t) |u(0)n,k〉〉,
u(1)n,k =
∑m,j
m,j6=n,k
〈〈u(0)m,j|V (t) |u(0)
n,k〉〉ε
(0)n,k − ε
(0)m,j
u(0)m,j, (4.19)
4.2. FLOQUET HAMILTONIAN 27
and
ε(2)n,k =
∑m,j
m,j6=n,k
∣∣∣〈〈u(0)m,j|V (t) |u(0)
n,k〉〉∣∣∣2
ε(0)n,k − ε
(0)m,j
. (4.20)
Upon inserting V (t) = V cosωt and ε(0)m,j = Em+ j~ω into the above equations and
evaluating the scalar products, we arrive at
ε(1)n,k = 0,
u(1)n,k = eikωt
∑m
|m〉〈m|V |n〉∆Enmcos(ωt)+ i~ω sin(ωt)
∆E2nm − (~ω)2
, (4.21)
and
ε(2)n,k =
∑m
1
2|〈m|V |n〉|2 ∆Enm
∆E2nm − (~ω)2
, (4.22)
where
∆Enm = En − Em. (4.23)
The fact that the energy shift is the same for all values of k implies that the
Rayleigh-Schrodinger perturbation method preserves the ω-modulo structure of
each class of quasienergies. It is clear from the above equation that this perturba-
tion approach is valid only when
|〈m|V |n〉|2 |(En − Em)2 − (~ω)2|, (4.24)
which requires that no resonance exists between |n〉 and any other state |m〉, that
is, |En −Em| 6= ~ω. When there are two eigenstates close to resonance, one needs
to use the almost-degenerate perturbation method which is discussed in the next
section.
28 CHAPTER 4. FLOQUET FORMALISM
4.3 Resonance
To demonstrate clearly the important role of resonance in quasienergy structures,
we first discuss a two-level system whose Hamiltonian has the form given in
Eq. (4.14) with
H0 =~ω0
2
1 0
0 −1
, (4.25)
and
V = ~η
0 1
1 0
, (4.26)
where the coupling constant η is small enough so that perturbation theory is ap-
plicable. Furthermore, we allow the frequency ω0 to be varied from zero to well
beyond the frequency ω of the oscillating field. This system is exactly solvable but
we choose perturbation methods as they are still useful when one goes to multilevel
systems. The eigenenergy diagram and the corresponding quasienergy diagram of
the unperturbed system is shown in Fig. 4.1(a) on page 33, where each quasienergy
level is indicated by its double index n, k. Note that resonances appear as ap-
parent crossings in the unperturbed quasienergy diagram, here drawn in four zones
from −2ω to 2ω.
Now we turn to the calculation of the quasienergies of the full system. To gain
insight we first use the perturbation method to find the approximate quasienergies
and Floquet states. Far from resonance, when
∣∣∣ω0
ω− 1∣∣∣ & 1/2,
one may use the Rayleigh-Schrodinger perturbation theory to obtain
εn,k ≈ ε(0)n,k + ε
(2)n,k = (−1)n+1~ω0
2
(1 +
η2
ω20 − ω2
), (4.27)
4.3. RESONANCE 29
and
un,k ≈ u(0)n,k + u
(1)n,k = |n〉 eikωt + η |n〉 eikωt (−1)n+1ω0 cos(ωt) + iω sin(ωt)
ω20 − ω2
, (4.28)
with n = 1, 2 and n = [3− (−1)n] /2. Thus, the energy shift is of the order
δε = εn,k − ε(0)n,k ∼ ~ω0
( ηω
)2
, (4.29)
which is very small if η ω. The small periodic correction to the Floquet state
means that this state is represented by a wave function that undergoes small oscil-
lations around the mean value which is the unperturbed wave function. Therefore,
the unperturbed wave function is a good approximation to the Floquet state in
this far off-resonance region.
In the opposite extreme, where the system is close to resonance when ω0/ω ≈ 1,
one needs to use the almost-degenerate perturbation method [62]. We start with
computing the Floquet Hamiltonian using Eq. (4.18). If one arranges the double
index n, k such that n goes through 1, 2 before each change in k, the Floquet
Hamiltonian has this infinite block-tridiagonal structure:
−2 −1 0 1 j = 2
. . . . . . .
−2 . H0−2~ωI V2
0 0 0 .
−1 . V2
H0−~ωI V2
0 0 .
0 . 0 V2
H0V2
0 .
1 . 0 0 V2
H0+~ωI V2
.
k = 2 . 0 0 0 V2
H0+2~ωI .
. . . . . . .
(4.30)
30 CHAPTER 4. FLOQUET FORMALISM
which contains the 2× 2 identity matrix I and the 2× 2 zero matrix 0. Although
the structure of the Floquet Hamiltonian strongly suggests a recursive scheme for
computing the eigenvalues and eigenvectors, no such procedure is known.
Close to resonance we have ε(0)1,k−1 ≈ ε
(0)2,k. The sub matrix corresponding to
u(0)1,k−1 and u
(0)2,k can be read off from the full Floquet Hamiltonian and is found to
be
H ′F =
ε(0)1,k−1
12~η
12~η ε
(0)2,k
. (4.31)
When the coupling between these two Floquet states is much stronger than the
coupling to other states, which is usually true for two states in resonance, the
quasienergies and Floquet states are approximately given by the eigenvalue and
eigenvector of the sub matrix. Therefore, at exact resonance where ε(0)1,k−1 = ε
(0)2,k = ε,
the quasienergies and Floquet states are
ε1,k−12,k
≈ ε∓ ~η2,
u1,k−12,k
≈ eikωt√2
(|1〉 e−iωt ∓ |2〉
). (4.32)
Because of the coupling, the two levels are pushed apart by an amount δε ≈ ~η.
Comparing this result with that for the far off-resonance region of Eqs. (4.27) and
(4.28), we observe two important differences: First, the energy shift at resonance is
much larger than its far off-resonance value; and second, while for the off-resonance
regime the time-dependent part of the wave function is only a small oscillation
around the dominant time-independent part, these parts possess comparable am-
plitudes at resonance. Thus, the wave function oscillates much more strongly when
resonance happens and there is little resemblance between the unperturbed wave
function and the real one. We conclude that the oscillating field has an important
4.3. RESONANCE 31
effect on the quasienergy structure and the characteristics of the wave function if
there is resonance between any two levels of the unperturbed system.
For the intermediate values of ω0/ω, numerical diagonalization of the Floquet
Hamiltonian is needed to obtain the quasienergies. The result rapidly converges
with the number Nf of the Floquet modes. In our computation we choose Nf = 20,
which is verified to be sufficiently large. The result is shown in Fig. 4.1(b) on
page 33. The energy shift is indeed maximum at resonance as predicted by the
perturbation methods.
The oscillating potential has another important effect when the parameter ω0
is varied slowly in an adiabatic process. As can be seen in Fig. 4.1(b), there is an
apparent crossing of the unperturbed quasienergy lines at resonance and hence the
system started from the branch (1) will continue to move to the same state (1) in
the opposite end of the crossing. However, in the real situation when the oscillating
potential is accounted for, the two curves are pushed apart by the coupling and we
have an avoided crossing. Thus, the system will move from state (1) to state (2)
in an adiabatic process. Therefore, the dynamics of an adiabatic process passing
through a resonance changes completely when the oscillating potential is taken
into account.
We must stress that the adiabatic approximation for Floquet states is slightly
different from that for the time-independent eigenstates. The mathematical frame-
work of the Floquet adiabatic process for short laser pulses is described in Ref. [56].
In Appendix C we modify these formulations a little so that it suits a more general
problem. A particularly useful result is the Landau-Zener formula for a Floquet
adiabatic process which gives the approximate transition probability of a system
when it passes through an avoided crossing. For an avoided crossing with asymp-
totic slope difference α and energy spacing δε as shown in Fig. 4.1(b), the Landau-
32 CHAPTER 4. FLOQUET FORMALISM
Zener transition probability is
PLZ = exp
[−π
4
(δε/~)2
ω0 tan(α/2)
], (4.33)
with ω0 denoting the time derivative of ω0, which is assumed to remain constant.
When α and δε are known from the energy diagram, the above formula tells us how
slow (fast) we need to change the parameter ω0 to obtain an adiabatic (diabatic)
process.
The conclusion that the effect of a weak oscillating field is most important at
resonances is also true for a multilevel system. One may use in a similar way the
perturbation methods to prove that the energy shift and the oscillation amplitude
of a Floquet state un,k are largest when there is a resonance between the unper-
turbed level u(0)n,k and another level u
(0)m,k−1. The split of these levels due to their
coupling is
δε ≈ |〈n|V |m〉| . (4.34)
Thus, a simple way to know whether the exact quasienergies structure differs by
a great amount from the approximate unperturbed eigenenergies is to look for
resonances in the unperturbed energy diagram and then calculate the coupling
strength |〈n|V |m〉| between the resonating levels.
4.3. RESONANCE 33
Figure 4.1: (a) Eigenenergies and unperturbed quasienergies. The solid lines in-dicate the two eigenenergy levels, the dashed ones show the zone structure of theunperturbed quasienergies. (b) A comparison of the exact quasienergies, representedby the dotted curves, with the unperturbed quasienergies. A typical avoided crossingcaused by the oscillating field is marked by AC.
34 CHAPTER 4. FLOQUET FORMALISM
4.4 Summary
A periodic quantum system possesses quasienergies and Floquet states which are
the analogs of the eigenenergies and eigenstates of a time-independent system.
These quasienergies and Floquet states can be found by diagonalizing a Floquet
Hamiltonian constructed in an extended Hilbert space. In the context of trapped
atom-ion systems, the micromotion effect can be revealed by a comparison of the
exact quasienergies and Floquet states with the approximate eigenenergies and
and eigenstates obtained in the harmonic approximation. The analysis of the two-
level system and perturbation theory demonstrate that the effect of an oscillating
potential is most important when there are resonances between any two eigenstates
of the unperturbed system. The oscillating potential results in three important
changes when one goes from the approximate to the exact picture: (i) There is
a shift in the energy of the system, (ii) a significant oscillating part is added to
the wave function and (iii) a lot of new avoided crossings appear in the energy
diagram.
Chapter 5
Micromotion effect in one
dimension
We now show how we use the Floquet formalism to evaluate the exact quasienergies
of the trapped atom-ion system and from there deduce the effect of the micromo-
tion. As a starting point we assume that the atom is fixed at the center of its
trap. This corresponds to the situation when the atom is tightly trapped, that
is, the atom trapping frequency is very large. We explain later why most of the
conclusions about the micromotion effect in this simplified model also hold for a
more realistic system.
In view of the above assumption, the presence of the atom only results in
an additional potential in the Hamiltonian of the trapped ion. Therefore, the
Hamiltonian of the 1D system given in Eq. (3.4) is simplified to
H(t) = H0 +Hmm(t), (5.1)
35
36 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
with
H0 =P 2i
2mi
+1
2miω
20(Xi − d)2 − αe2
2X4i
, (5.2)
and
Hmm(t) = −mi(γω0)2(Xi − d)2 cos(2ωt)− γω0Xi − d, Pi sin(ωt). (5.3)
For the purpose of numerical calculation it is convenient to write the above Hamil-
tonian in the following dimensionless form:
H(τ)
~ω0
=1
2
(liPi~
)2
+1
2
(Xi − dli
)2
− 1
2
(Ri/li)2
(Xi/li)4
− γ2
(Xi − dli
)2
cos
(2ω
ω0
τ
)− γ
Xi − dli
,liPi~
sin
(ω
ω0
τ
), (5.4)
where Ri =√miαe2/~2 is the interaction length redefined for this simplified sys-
tem,
li =
√~
miω0
(5.5)
is the harmonic oscillator length, and τ = ω0t a new time variable. In short, all
lengths are scaled by li and energies by ~ω0.
5.1 The unperturbed system
Before we compute the exact quasienergies of the system, it is worth examining
whether resonances, as described in Sec. 4.3, exist in our system. First we cal-
culate the eigenenergies of the unperturbed Hamiltonian H0. Following Idziaszek
et al. [13] we compute, at zero trap distance d = 0, the eigenenergies En(0) and
eigenstates |n(0)〉 of the unperturbed Hamiltonian H0(d = 0) using the Numerov
method [64–67]. The states |n(0)〉 are then used as a basis to diagonalize H0(d)
5.1. THE UNPERTURBED SYSTEM 37
for positive values of the trap distance.
Let us consider the time-independent Schrodinger equation for the unperturbed
system at zero trap distance
[− ~2
2mi
∂2
∂x2i
+1
2miω
20x
2i −
αe2
2x4i
− E]
Ψ(xi) = 0, (5.6)
with E an eigenenergy. To account for the short-range potential, we need to fix
the solution to the short-range form described in Ref. [13]. Before going into a
detailed description of this idea, let us first define the characteristic distances
r1 =
(αe2
miω20
)1/6
= R1/3i l
2/3i (5.7)
where the atom-ion interaction potential and the trapping potential of Eq. (5.6)
are equal, and
r2 =
(αe2
2E
)1/4
= (Rili)1/2
(~ω0
2E
)1/4
(5.8)
where the atom-ion interaction potential is the same as the eigenenergy E.
At short distances, where |xi| minr1, r2, the atom-ion interaction po-
tential, which goes as 1/x4i , is much larger than the trapping potential and the
eigenenergy. Hence we can neglect the trapping potential and the eigenenergy E
in Eq. (5.6); the resulting differential equation is
[~2
2mi
∂2
∂x2i
+αe2
2x4i
]Ψ(xi) = 0, (5.9)
for which the solution is
Ψ(xi) ∝ xi sin
(Ri
xi+ ϕs
)(5.10)
with some constant phase ϕs; this is the asymptotic form of the wave function
38 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
in the range |xi| minr1, r2. Of course this asymptotic form is only valid
for distances considerably larger than the range of the short-range interaction
potential, otherwise we cannot use the long-range form 1/x4i for the interaction
potential in Eq. (5.9). To be precise, if we denote the scale of the short-range
interaction potential by r0, then the asymptotic form given in Eq. (5.10) is valid
for r0 |xi| minr1, r2. For Alkali atoms, r0 is typically comparable to the
Bohr radius a0.
The short-range phase ϕs in the asymptotic wave function is determined by
the properties of the short-range potential. When one wishes to find the wave
function only at distances much larger than r0, the short-range phase is all one
needs to know. This is the main point of the quantum defect theory where one
can avoid the complicated short-range interaction by cutting off the wave function
at a sufficiently small distance and imposing a boundary condition on it. By
considering the scattering of a free atom and a free ion, Idziaszek et al. show that
the short-range phase can be related to the s-wave scattering length b by [13]
cot(ϕs) = − b
Ri
. (5.11)
Once we know the scattering length b and hence the short-range phase, we may find
a distance rmin satisfying r0 rmin minr1, r2 such that the wave function
given in Eq. (5.10) vanishes at rmin. The boundary condition Ψ(rmin) = 0 is then
used to carry out the Numerov computation (details of this numerical method are
provided in Appendix. A.4). In general there are two possible values for the short-
range phase of an 1D system: the even and odd short-range phases [13]. In our
study we assume a single value for these two phases.
For our numerical calculation we consider the 135Ba+ ion and 87Rb atom system
with Ri = 8927a0. The chosen trapping parameters are γ = 1/√
2 (a = 0),
5.1. THE UNPERTURBED SYSTEM 39
Figure 5.1: Wave functions for the eigenstates of the unperturbed system at zerotrap distance. The red horizontal lines indicate the corresponding eigenenergy foreach eigenstate. The black curve is the plot of the total potential in Eq. (5.6).
ω0 = 2π × 100 kHz, ω = 2π × 1.27 MHz; hence, li = 516 a0 and Ri/li ≈ 17.2. The
scattering length b is assumed to be 0.9Ri, from which we obtain rmin = 0.135 li.
Although the secular frequency and the micromotion frequency stated above are
unrealistically small, we choose these values to obtain plots with better clarity. We
repeated our calculation with more realistic values of the frequencies and it did
not result in any qualitative change in what we are going to describe about the
micromotion effect. More about this in Sec. 5.8.
Here we restrict our attention to the situation when the ion trap is placed
on the right of the atom trap. In a 1D model the ion cannot penetrate to the
left side of the atom, and as a result we are only concerned with wave functions
which vanish for xi < 0. The wave functions of a few eigenstates, together with
their eigenenergies, of the unperturbed Hamiltonian H0 at d = 0 are plotted in
40 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
Fig. 5.1. When xi → rmin, the wave functions oscillate rapidly as predicted by the
asymptotic form given in Eq. (5.10). Using this asymptotic form one may work out
quite easily that the oscillation period of the wave functions at a small distance xi
is
∆xi ≈ 2πx2i
Ri
. (5.12)
This fast oscillation poses certain difficulties in the numerical calculation of the
wave functions, as one must choose a step size sufficiently smaller than 2πr2min/R
2i
to obtain a reasonably good result.
In our calculation we choose rmax = 55 li and the boundary conditions are
Ψ(rmax) = Ψ(rmin) = 0. (5.13)
Since the wave functions obtained will be used to carry out further computations,
the results must be highly accurate. For this reason we choose a large number
of grid points, Ng = 2 × 106, and the number of digits is set to 25 to make sure
that round-off error plays no significant role. The eigenenergies are determined by
the bisection procedure, which is part of the Numerov calculation, to an accuracy
of 10−6 × ~ω0. And for a given eigenenergy, the Numerov method generates the
corresponding eigenfunction with an accumulated error of order O(h4) where h =
(rmax − rmin)/Ng is the step size.
Using the Numerov method we obtain, for the unperturbed Hamiltonian at
zero trap distance H0(d = 0), a total of 253 eigenstates corresponding to all the
eigenenergies in the interval [−5000 ~ω0, 500 ~ω0] and use them as a basis to diagno-
lize H0(d) at positive trap distances. The eigenenergy diagram of the unperturbed
Hamiltonian is shown in Fig. 5.2 on page 44. At a large trap distance the spectrum
is almost identical to that of a harmonic oscillator as it should be [see Fig. 5.2(a)].
At small distances we have a spectrum of vibrational states (E > 0) where the
5.1. THE UNPERTURBED SYSTEM 41
ion is localized in the ion trap and molecular states (E < 0) where it is bounded
to the atom. Note that the energy gaps between adjacent molecular levels shown
in Fig. 5.2(b) are much larger than the ones between the vibrational levels. The
molecular states at large trap distances correspond to the situations when we have
the traps far apart but the atom and ion are close to each other. These states are in
general not the initial states in which we prepare our system. The properties of the
vibrational as well as molecular states are discussed thoroughly in Refs. [13, 14].
We also observe that, for large n, the eigenenergies En(0) for d = 0 obey the
relation
En+1(0)− En(0) ≈ 2~ω0, (5.14)
which is a feature of the energy spectrum of a spherical harmonic oscillator with no
angular momentum. This can be understood as follows: When the ion is in a high-
energy state it spends most of its time in the region far away from the origin where
the trapping potential dominates. For this state, the atom-ion interaction is a very
small perturbation, and hence the ion behaves like a spherical harmonic oscillator
with no angular momentum (as can be seen from the form of the Hamiltonian of
Eq. (5.6)). This explains the large-n form of the eigenenergies En(0) in Eq. (5.14).
In the proposed experiment for implementing the quantum phase gate [14],
one starts with the atom and ion cooled to the ground states at a large trap
distance and then adiabatically moves the ion trap closer to the atom trap. Thus,
we are mainly interested in how the asymptotic ground level changes as the trap
distance decreases. From now on when we mention ground level we mean the
energy curve that asymptotically coincides with the ground harmonic oscillator
level at a very large trap distance, that is, the curve marked by |0〉 in Fig. 5.2(a).
The excited energy curves above the ground level are indicated by |n〉 with positive
integers n, while the lower molecular energy curves are marked by negative integers.
One notices that, when the trap distance is around some characteristic distance
42 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
dc ' 5 li the energy of the asymptotic ground state begins to decrease rapidly.
This is an indication that the atom-ion interaction is dominant within the range
[0, dc]. Furthermore, the energy diagram exhibits an avoided crossing where the
ion changes from its vibrational state to the molecular state around dc. Since
the interaction potential must be comparable to the trapping potential at dc, we
may estimate this characteristic distance as follows: Suppose the ion is half-way
between the atom trap and the ion trap, after equating the two potentials we have
dc ' 2R1/3i l
2/3i ≈ 5.2 li, (5.15)
which is close to the value obtained numerically. While this good agreement may
not hold for different values of Ri and li, the above estimation should yield a correct
order of magnitude for the ratio dc/li.
To see the difference between a molecular state and a vibrational state we plot
the wave functions of the ground state |0〉 and the molecular state |−1〉, together
with the total potential
V (xi) =1
2miω
20(xi − d)2 − αe2
2x4i
, (5.16)
at two trap distances d = 7 li and d = 5 li in Fig. 5.3. For d = 7 li, which is relatively
large, the potential has a large barrier around xi = 3 li [see Fig. 5.3(a) on page 46].
It also possesses a nearly harmonic well centered at around xi ' d = 7 li due to
the fact that the trapping potential is dominant at large distances. This potential
well supports many vibration modes, the lowest one being the ground state whose
wave function is described by the red curve. One would see immediately that this
wave function is basically the same as that of the ground harmonic oscillator state.
Once it is in this ground state, the ion is located near the center of the ion trap.
In contrast, the wave function of the molecular state |−1〉 is squeezed to the deep
5.1. THE UNPERTURBED SYSTEM 43
well on the left of the potential barrier, which means that the ion is bounded to
the atom when it is in a molecular state.
44 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
Figure 5.2: The unperturbed eigenenergies of (a) the vibrational levels and (b)the low lying molecular levels. The energy gaps between adjacent molecular levelsare very large compared to those between the vibrational levels which appear as acontinuum in (b).
5.1. THE UNPERTURBED SYSTEM 45
When the trap distance decreases, the atom-ion interaction becomes stronger
and it has greater impact on the wave function of the ion. At d = 5 li, the
potential barrier is now reduced to a significantly small value and the ion is now
able to tunnel through to the region close to the atom [see Fig. 5.3(b)]. As a result,
the wave function of the ground state is distorted and there is now a similarity
between this wave function and the wave function of the molecular state |−1〉 for
small xi. This is not a surprise since all the eigenfunctions possess the same short-
range form given in Eq. (5.10). For the wave functions of the ground state and
excited states, this short-range part is non-negligible only when the trap distance
is sufficiently small, as shown by the fact that the short-range part of the ground-
state wave function is too small to be observed in Fig. 5.3(a). The transition in the
magnitude of the short-range part of the ground-state wave function begins to take
place around the avoided crossing at d = dc shown in Fig. 5.2(a). This transition
also indicates how the ion changes its state, as the trap distance is reduced to
zero, from the asymptotic ground harmonic-oscillator state to a molecular state
for which the ion is bounded to the atom.
Another important remark is that while the overlap, or more precisely the
Franck-Condon factor, between the wave function of the ground state |0〉 and that
of the molecular state |−1〉 is negligible at d = 7 li, it is much larger at the shorter
trap distance d = 5 li. Therefore, if we want to make a transition from the ground
state to one of the molecular states lying below, we need to move the traps to a
distance small enough so that the Franck-Condon factor is sufficiently large. This
is central to the experimental scheme we propose in Sec. 5.6 for bypassing the
micromotion effect.
46 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
Figure 5.3: The wave functions of the ground state |0〉 (red) and the molecularstate |−1〉 (blue) plotted together with the potential (black) at (a) d = 7 li and (b)d = 5 li.
5.2. ATOM-ION QUANTUM GATE 47
5.2 Atom-ion quantum gate
Now let us describe the adiabatic quantum phase gate proposed by Doerk et al.
in Ref. [14]. This phase gate is realized by an adiabatic collision process of the
trapped atom and the trapped ion as illustrated in Fig. 1.1 on page 2. Initially the
atom and the ion, in their respective traps, are prepared in their ground states at
a very large trap distance. Hence, there is essentially no interaction and the initial
state of the system is simply the tensor product of the particles’ states. Next, the
traps are brought closer to each other before being separated back to the initial
trap distance. The process is carried out slowly enough so that the evolution of
the system can be considered as adiabatic.
We may describe such an adiabatic process on the eigenenergy diagram shown
in Fig. 5.4(a) on page 51. At the beginning, the atom and the ion are in their
ground states and are kept at a very large trap distance; this initial state can
be marked by the right bar on the ground state energy curve. When the traps
are brought closer to each other in an adiabatic process, the system will always
settle in the ground state, and therefore it will start to move to the left along the
ground state energy curve. The state of the system at the shortest trap distance
is represented by the left bar. In the second phase, when the traps are separated
back to their initial positions, the system must move along the ground energy curve
and go back to its initial position. As a consequence, a system which starts in the
ground state |φ0〉 at the initial time ti will mostly end in the same state at the final
time tf provided that the process is adiabatic. We use the word “mostly” because
this is only an approximation; in reality there is always a small probability that
the system transits to other states. In the adiabatic approximation we have
|φ0〉 → |φ0〉 eiθ. (5.17)
48 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
The only change in the wave function is the accumulated phase θ which, according
to the adiabatic theorem, is given by
θ = −1
~
∫ tf
ti
dtE0
(d(t)
), (5.18)
where E0
(d(t)
)is the eigenenergy of the ground state expressed as a function of the
trap distance d(t). We will explain shortly that this eigenenergy depends on the
spin of the atom and ion, hence the phase θ must be spin dependent. Here we use
the term “spin” for the two internal states chosen as the |0〉 and |1〉 qubit states.
Each qubit combination of |00〉 , |01〉 , |10〉 , |11〉 specifies a collision channel.
To see how the eigenenergy for the motional degree of freedom depends on the
qubit state of the atom and the ion, let us consider again the short-range phase
ϕs mentioned in Eq. (5.10). This phase is determined by the characteristics of the
short-range potential, which in turn depends on the internal states of the atom
and the ion. Therefore, the short-range phases for different qubit states must be
different. Assuming the coupling between the channels is weak, each channel is
associated with an effective short-range phase [14]. Let us consider a hypothetical
situation where the value of the scattering lengths is b1 = 0.9Ri for the channels
|01〉 , |10〉 and b2 = 0.95Ri for the channels |00〉 , |11〉. In reality each channel
should possess a different effective scattering length but this simplified model is
sufficient to capture the mechanism of the quantum phase gate. A more rigorous
treatment is given in Ref. [14].
When we calculate the eigenenergy numerically by the Numerov procedure
as described in Sec. 5.1, we obtain different motional ground-state eigenenergies
for the “parallel” (|00〉 , |11〉) and “anti-parallel” (|01〉 , |10〉) qubit combinations.
These eigenenergies are shown in Fig. 5.4(b). Although we observe no difference
at large trap distances, the ground energy curves for the parallel and anti-parallel
5.2. ATOM-ION QUANTUM GATE 49
states begin to deviate as the trap distance is closer to dc. As we see below, this
small energy difference is essential for the adiabatic quantum phase gate.
While we consider only the motional degree of freedom in Eq. (5.17), in the
analysis of the quantum gate we must take into account both the qubit degree of
freedom and the motional degree of freedom. Initially the system is prepared in
the ground motional state |φ0〉. In the qubit subspace, single qubit gates can be
used to prepare the system in a separable superposition
|χ(ti)〉 =∑k
ck |k〉 , (5.19)
with k = 00, 01, 10, 11, and c1c4 = c2c3. If there is no channel-mixing, the adiabatic
collision process turns the initial state
|Ψ(ti)〉 = |φ0〉∑k
ck |k〉 (5.20)
to
|Ψ(tf )〉 = |φ0〉∑k
|k〉 eiθk . (5.21)
Each of the state |k〉 gets a different phase θk since the accumulated phase in
an adiabatic process depend on the energy of the motional ground state which
in turn depends on the qubit state. The final state in the qubit subspace is ob-
viously |χ(tf )〉 =∑
k |k〉 eiθk , and thus the adiabatic process is equivalent to the
implementation of the operator
U1 = diageiθ00 , eiθ01 , eiθ10 , eiθ11 (5.22)
in the qubit subspace. Next, by applying the following two single-qubit gates
U2 = 1⊕(|0〉〈0| e−iθ00 + |1〉〈1| e−iθ01
), (5.23)
50 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
and
U3 =(|0〉〈0|+ |1〉〈1| ei(θ00−θ10)
)⊕ 1, (5.24)
in sequence, one may verify that the total operation is the control phase gate
U = U3U2U1 = diag1, 1, 1, eiθ, (5.25)
where θ is the phase difference
θ = θ00 + θ11 − θ01 − θ10. (5.26)
Such a gate can be used to entangle the qubit state of the atom and that of the
ion. As can be seen in Eq. (5.18), these phases depend on the function d(t), thus
one can control how the trap distance is changed with time to obtain the desired
phase θ.
The energy difference seen in Fig. 5.4(b) is significant only in the left of the
avoided crossing shown in Fig. 5.4(a). Since the phase difference θ is proportional
to the energy difference, a fast gate is achieved only if the minimum trap distance
of the adiabatic collision process is at least close to the distance dc of the avoided
crossing, which means dmin ≈ 5.2 li.
What we described in the first two sections of this chapter are based on the
harmonic approximation where the perturbing Hamiltonian Hmm(t) of the micro-
motion is neglected. In the following sections we present the main results of our
work on the influence of the micromotion on the energy spectrum of the trapped
atom-ion system, and how it affects specific applications like this adiabatic quan-
tum phase gate.
5.2. ATOM-ION QUANTUM GATE 51
Figure 5.4: (a) The adiabatic collision process represented in the energy diagram,and (b) the motional ground eigenenergies for the paralel qubit state (red) and theanti-paralel qubit state (black) of the atom-ion system.
52 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
5.3 Micromotion-induced coupling
Having obtained the eigenenergy diagram of the unperturbed system, we now
turn our attention to the exact system which is described by the time-dependent
Hamiltonian in Eq. (5.1). As demonstrated for the two-level system in Sec. 4.3,
in order to understand how the oscillating micromotion affects the energy of the
ground level we need to look for resonances between this level and the excited
states of the unperturbed system. In Eq. (5.3), the micromotion is represented by
two terms with the first oscillating at twice the frequency of the second; hence,
there are two types of resonance which we call the 2ω resonance and ω resonance.
Since in our system ω = 12.7ω0, which is not an integral multiple of ω0, there is
no resonance at a large trap distance (we see shortly that resonances at large trap
distances are not important anyway). However, as the trap distance is reduced
and the energy levels begin to deviate from their asymptotic values, the ground
level inevitably comes into resonance with a number of excited levels at different
values of the trap distance. To find where the resonances appear, we plot the
eigenenergies En together with En − ω in Fig. 5.5(a) and En − 2ω in Fig. 5.5(b)
on page 57, which show that the ground level passes a total number of ten ω-
resonances and fifteen 2ω-resonances with various excited levels as d goes from
0 to 7 li; these resonances appear as crossings and are marked by the letters An
and Bn. When the micromotion is included we expect the energy levels to be
pushed apart by the micromotion-induced coupling and these apparent crossings
to become avoided crossings.
Next we need to consider the coupling strengths at the resonances. As men-
tioned in the previous section, the shift in the quasienergy of the ground level can
be inferred from the coupling strength |〈0|V |n〉|, where |n〉 is the excited state in
resonance with the ground level. As can be seen in Eq. (5.3), the micromotion-
5.3. MICROMOTION-INDUCED COUPLING 53
induced coupling is represented by the operators
V1 = −mi(γω0)2(Xi − d)2,
V2 = −γω0Xi − d, Pi. (5.27)
Therefore, to study its effect, the quantities |〈0|V1,2 |n〉| need to be calculated at
various trap distances. Since
〈0|V2 |n〉 =En − E0
iγ~ω0
〈0|V1 |n〉 , (5.28)
the second coupling strength is much larger than the first for sufficiently large n
and thus has more important effects.
Let Xi = Xi − d; Eq. (5.28) can be derived from the commutation relation[Xi, Pi
]= i~. We have
2mi
[X2i , H0
]=[X2i , P
2i
]= XiXiPiPi − PiPiXiXi, (5.29)
and from the commutation relation we see that
PiPiXiXi = Pi(XiPi − i~)Xi
=(XiPi − i~
) (XiPi − i~
)− i~PiXi
= XiPiXiPi − i~(2XiPi + PiXi). (5.30)
Therefore,
2mi
[X2i , H0
]= XiXiPiPi − XiPiXiPi + i~(2XiPi + PiXi)
= Xi(XiPi − PiXi)Pi + i~(2XiPi + PiXi)
= 2i~(XiPi + PiXi), (5.31)
54 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
or
Xi − d, Pi =mi
i~[(Xi − d)2, H0
], (5.32)
which yields
V2 = (iγ~ω0)−1 [V1, H0] , (5.33)
and Eq. (5.28) follows since |0〉 and |5〉 are eigenstates of H0.
Figures 5.6(a) and 5.6(b) on page 58 show the collective behavior of many
coupling strengths, obtained by numerical integration [68], at four different values
of the trap distance d = 9 li, 5.5 li, 5.3 li and d = 5.1 li. As can be seen in
Fig. 5.2(a), the energy gap between the ground level |0〉 and the third excited level
|3〉 is at most 11ω0 at d = 0. Because this gap is still smaller than the micromotion
frequency, we know that there is no resonance between the ground level and the
first three excited levels. On the other hand, the energy gap between the ground
level and the 26th excited level |26〉 is too large for resonance. Since the ground
level can only come into resonance with those excited levels |n〉 with 4 ≤ n ≤ 25,
only the coupling strengths in that range are important. We see in Figs. 5.6(a)
and 5.6(b) that when d = 9 li, the elements |〈0|V1,2 |n〉| are extremely small, which
is due to the fact that the atom-ion interaction is very weak at this distance, and
hence the unperturbed states are almost identical to the eigenstates of a harmonic
oscillator. For these states we see immediately that |〈0HO|V1,2 |nHO〉| = 0 for n > 2.
As a result, there is no possible resonance with nonzero coupling strength, and we
observe in Sec. 4.3 that the energy shift caused by the micromotion in this case
must be very small. In fact, the exact quasienergy spectrum of a single trapped
ion was found by Glauber [40] to be
εn = (n+1
2)~µ, (5.34)
5.3. MICROMOTION-INDUCED COUPLING 55
where µ differs from the secular frequency ω0 by
µ− ω0
ω0
'(ω0
ω
)2
(5.35)
when |a| q2 (see Appendix B). This number is very small, so the harmonic
approximation works relatively well for trapped ions as long as the external inter-
action is absent or relatively weak compared with the trapping potential.
The situation changes completely when the ion is subjected to a significant
external interaction. At d = 5.5 li, where the ion trap just crosses to the strong
interaction region, the micromotion-induced coupling begins to increase rapidly
and reach a remarkably large value at d = 5.3 li. This is due to the now strong
enough interaction potential which changes considerably the shape of the unper-
turbed eigenstates. To see how the coupling strengths change with trap distance
we plot a representative one between the ground level and the fifth excited level,
that is, |〈0|V1,2 |5〉|, in Fig. 5.6(c). In addition to the sharp rise at d ' 5.5 li, we
see that the coupling due to V1 is indeed much weaker than the one due to V2,
which is in agreement with Eq. (5.28).
Since at a small distance the coupling strength of the second term is comparable
to the secular frequency ω0 while the spacing of the unperturbed energy spectrum
is of the order ω0, we expect the energy diagram to change completely when the
oscillating terms of the micromotion are included. More importantly, the fact that
the strength of the micromotion-induced coupling increases sharply at a certain
value of trap distance means that it can be used to forecast at what characteristic
range dmm the micromotion effect becomes significant and must be taken into
account. For our particular system we may predict that the micromotion effect
begins to kick in at around dmm ' 5.5 li.
In the limit of a very large ratio ω/ω0, for instance ω/ω0 = 100, resonance
is possible only with a large excitation number n. As shown in Fig. 5.6(d), the
56 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
coupling strength of V1 for large n is essentially zero at all trap distances and
therefore can be neglected. On the other hand, Fig. 5.6(b) shows that the coupling
strength of V2 within the range d ∈ [0, 5.5 li] only decreases slowly with n because
it is proportional to the energy difference [see Eq. (5.28)]. Even for n as large as
100 we have |〈0|V2 |100〉| ' 0.1~ω0 at d = 5.3 li which is not negligible. Moreover,
the coupling strength of V2 for the highly excited state n = 100 exhibits a sudden
rise at the same distance dmm ' 5.5 li. We verify that the sharp increase at
dmm ' 5.5 li also appears for all n ∈ [3, 100], which implies that the characteristic
distance where the micromotion effect becomes important is quite insensitive of the
frequency of the trapping potential when the ratio ω/ω0 is in the interval [10, 100],
which contains most of the realistic values for current rf traps.
5.3. MICROMOTION-INDUCED COUPLING 57
Figure 5.5: A plot of (a) the eigenenergies En together with En − ω shows theω resonances and (b) the eigenenergies En together with En − 2ω shows the 2ωresonances. These resonances appear as crossings.
58 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
Figure 5.6: Coupling strengths for (a) V1 and (b) V2 at d = 9 li (diamonds),d = 5.5 li (circles), d = 5.3 li (squares) and d = 5.1 li (crosses). The couplingstrengths between the ground state and (c) an intermediate excited state n = 5 and(d) a very high excited state n = 100 of V1 (diamonds) and V2 (circles) are alsoplotted as functions of the trap distance.
5.4. NUMERICAL CALCULATION OF THE QUASIENERGIES 59
5.4 Numerical calculation of the quasienergies
In this section we obtain, by numerical calculation, the quasienergy diagram, from
which we then infer the importance of the micromotion effect. There are two
widely used methods for computing the quasienergies of a periodic system: Floquet
Hamiltonian and time-propagator methods [39]. The Floquet Hamiltonian method
is simply the numerical diagonalization of the Floquet Hamiltonian as described
in Sec. 4.2. For our system, it can be expressed in terms of H0 and Hmm(t) given
in Eqs. (5.2) and (5.3) as
HF = H0 +Hmm(t)− i~ ∂∂t. (5.36)
We use the Floquet states of the unperturbed Floquet Hamiltonian at zero trap
distance H(0)F (d = 0), which are
u(0)n,k(0) = |n(0)〉 eikωt, (5.37)
as the basis for evaluating the matrix elements of the Floquet Hamiltonian HF at
an arbitrary trap distance
〈〈u(0)n,k(0)|HF |u(0)
m,j(0)〉〉 =1
T
∫ T
0
dte−ikωt 〈n(0)|HF |m(0)〉 eijωt. (5.38)
If one arranges the double indices n, k such that n runs through all the
eigenstates before each change in the Floquet mode k, the Floquet Hamiltonian
matrix has an infinite block-pentadiagonal structure. With I denoting the identity
matrix and 0 the zero matrix, the Floquet Hamiltonian matrix is
60 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
−2 −1 0 1 j = 2
. . . . . . .
−2 . H0−2~ωI iV22
V12
0 0 .
−1 . −iV22
H0−~ωI iV22
V12
0 .
0 . V12
−iV22
H0iV22
V12
.
1 . 0 V12
−iV22
H0+~ωI iV22
.
k = 2 . 0 0 V12
−iV22
H0+2~ωI .
. . . . . . .
, (5.39)
where H0, V1, and V2 are matrices whose elements are
(H0)n,m =[En(0) +miω
20d
2/2]δnm −miω
20d 〈n(0)|Xi |m(0)〉 ,
(V1)n,m =−mi(γω0)2 〈n(0)| (Xi − d)2 |m(0)〉 ,
(iV2)n,m = (γ~ω0)−1 [Em(0)− En(0)] (V1)n,m . (5.40)
Therefore, the Floquet Hamiltonian matrix can be obtained from the matrix el-
ements of Xi and X2i at d = 0. Since we already obtained En(0) and |n(0)〉 by
the Numerov method in Sec. 5.1, the Floquet Hamiltonian matrix elements can be
evaluated by numerical integration.
In practice there are two types of truncation one needs to make. The first
comes from the number Ne of the eigenstates |n(0)〉 and the second comes from
the number Nf of the Floquet mode k. To generate the basis u(0)n,k(0) we choose all
the eigenstates |n(0)〉 in the energy range [−5000ω0, 300ω0] and k = −10, . . . , 10.
In total we have Ne = 150 eigenstates and Nf = 21 Floquet modes; hence, the
linear dimension of our Floquet Hamiltonian matrix is NeNf = 3150. To obtain a
5.4. NUMERICAL CALCULATION OF THE QUASIENERGIES 61
quasienergy diagram, we diagonalize this matrix for trap distances from 0 to 10 li
in step of ∆d = 0.002 li.
We also use the time-propagator method to calculate the quasienergies of the
system. This method is based on Floquet’s theorem which states that the time
evolution operator satisfying the Schrodinger equation
i~∂
∂tU(t, 0) = H(t)U(t, 0) (5.41)
has the periodicity property [63]
U(t+ T, 0) = U(t, 0)U(T, 0), (5.42)
where U(T, 0) is a unitary operator whose eigenvalues λn are related to the quasiener-
gies εn by
λn = e−i~ εnT . (5.43)
So, we need to obtain U(T, 0) and diagonalize it to get the quasienergies. First we
set U(0, 0) = I and compute U(T, 0) by using the propagating scheme
U(t+ ∆t, t) = exp
[−i∫ t+∆t
t
dt′H(t′)
]+O(∆t3), (5.44)
which is a second-order method. This simple propagator is the lowest order of the
numerical integrator based on the Magnus series [69]. The error bound of this class
of numerical propagators is discussed in Refs. [70, 71]. Perhaps a much faster and
more accurate numerical method to accomplish the task is the fourth-order method
based on the interpolatory Magnus integrator in which the evolution operator from
time t1 = t to time t2 = t+ ∆t is approximated by [71]
U(t2, t1) = e−iA +O(∆t5). (5.45)
62 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
The operator A is given by
A =∆t
2[H(1) +H(2)] +
√3
12(∆t)2 i [H(1)H(2)−H(2)H(1)] , (5.46)
where
H(j) = H(t+ cj∆t), j = 1, 2, (5.47)
with c1,2 = 1/2 ∓√
3/6. This fourth-order method can be very useful for prop-
agating systems in an adiabatic process (with long duration) as it helps reduce
considerably the number of time steps. One advantage of the Magnus method is
that it preserves the unitarity of the evolution operator. Magnus methods are also
known to give better performance than classical methods (such as Runge–Kutta)
of the same order when the system is highly oscillatory [71, 72]. The reader may
refer to Ref. [73] for an extensive review on the Magnus series and its associated
numerical methods.
The matrix exponential e−iA can be computed using the Eigenvalue, Cheby-
shev, Pade approximation or Krylov space methods [74, 75]. Most of these methods
are available in the Expokit package which is the most extensive collection of al-
gorithms for computing the matrix exponential [76]. The source code is available
in Fortran and Matlab.
In this work the quasienergies are obtained by the second-order propagator. We
use a total number of Ne = 150 eigenstates in the energy range [−5000ω0, 300ω0]
as a basis to compute the elements of the matrix H(t) and choose ∆t = 10−3T as
the time step. The computation is carried out for trap distances within the interval
[0, 10 li] in step of ∆d = 0.002 li. In both the Floquet Hamiltonian and the time-
propagator methods, we increased Ne up to 253 and repeated all the calculations
to verify that we obtained sufficient accuracies. Here we present the results for
Ne = 150 as the quasienergy plot is clearer with smaller Ne.
5.4. NUMERICAL CALCULATION OF THE QUASIENERGIES 63
We apply both the Floquet Hamiltonian and the time-propagator methods to
find the quasienergies and observe very good agreement in the results. However,
while the Floquet Hamiltonian method is more insightful, the time-propagator
method is found to give more accurate results for the quasienergies of the trapped
atom-ion system for equal CPU time. The inaccuracy of the former approach is
mainly due to the difficulty in choosing the most accurate member of each con-
gruent class of quasienergies. In general, it is not convenient to use the Floquet
Hamiltonian method when the range of the energy spectrum involved is consider-
ably larger than the frequency of the oscillating field because of the overlapping
between different but equivalent energy spectra. The quasienergy diagrams shown
in Figs. 5.7 and 5.8 on pages 70 and 71 are obtained by the time-propagator
method.
As can be seen in Fig. 5.7, for large trap distances, the quasienergy of the ground
Floquet state, marked by u0, agrees quite well with the eigenenergy obtained by
the harmonic approximation shown in Fig. 5.2. The densely distributed curves
with steep slopes are quasienergies of the Floquet states that correspond to highly
excited unperturbed eigenstates. Since all the quasienergies are returned in the
zone [−~ω/2, ~ω/2], which is an intrinsic property of the Floquet Hamiltonian
as well as time-propagator method, these highly excited quasienergy levels are
projected down and they appear to cross the ground level. However, at large
trap distances, say d & 5.7 li, the interactions between the ground level and those
states are extremely weak and hence the avoided crossings, if they really exist, have
very small gaps. Therefore, it is not hard to move the trap distance fast enough to
diabatically cross these possible weak avoided crossings so that the system remains
in a single Floquet state, which is the ground level in this case.
As the trap distance gets smaller than a characteristic distance dmm ' 5.7 li, the
quasienergy begins to differ greatly from the eigenenergy obtained by the harmonic
64 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
approximation. This is due to the sharp increase in the micromotion-induced
coupling discussed in Sec. 5.3. Within this range, numerous avoided crossings with
observable gaps occur in the quasienergy diagram, many of which do not appear
in the diagram obtained by the harmonic approximation shown in Fig. 5.2. The
reason is that, at a small trap distance, the coupling of the second term shown in
Fig. 5.6 is so strong that it changes the energy diagram completely.
A really interesting observation is that the appearance of the new avoided
crossings in the region d ≤ dmm may have strong connection with the transition to
chaos of the classical atom-ion system [58, 78]. The chaotic behaviour is due to the
nonlinear force associated with the polarization potential −1/r4. The nonlinear
dynamics of the classical systems of the trapped atom and trapped ion as two
harmonic oscillators coupled by the charge-dipole force has not been explored so
far. The Floquet theory introduced in the previous chapter is also a powerful tool
for studying classical and quantum chaos [58, 78, 79].
What does the dramatic change in the energy diagram mean for the implemen-
tation of the quantum phase gate? As we slowly change the trap distance passing
through dmm, the large number of new avoided crossings of various sizes will lead
to branching from the initial ground level to different levels [unless near-perfect
experimental control of d(t) is carried out]. The branching of the wave function
would result in a poor fidelity F of the phase gate, as we show in Appendix C that
1−F ∝ pe, where pe is the small probability of the ion being in an excited Floquet
state at the end of the adiabatic process. Moreover, even if one can control an
adiabatic passage of a micromotion-induced avoided crossing, for example the one
marked by AC in Fig. 5.7(b), the duration it takes must be quite long since the
energy gap of this avoided crossing is quite small compared with the energy gaps
of the original avoided crossings in the approximate eigenenergy picture shown
in Fig. 5.2(a). This long duration would lead to a slow gate speed and hence
5.4. NUMERICAL CALCULATION OF THE QUASIENERGIES 65
traversing the micromotion-induced avoided crossings is not desirable.
The effect of the micromotion may be more severe in reality. The steep energy
curve that form one branch of the avoided crossing AC belong to a high energy level.
There is a considerable probability that the system transits to this high energy
state when it crosses the avoided crossing. We see later that such an excitation
due to micromotion-induced avoided crossings also happen in the realistic system
where the atom is allowed to move. In this situation, micromotion excitation
may provide the atom enough kinetic energy to escape the atom trap because the
depth of the atom trapping potential can be quite small compared with the ion’s
secular frequency. This results in possible atom lost from the trap, which implies
a complete failure of the quantum gate. Micromotion excitations may also prevent
the formation of the weakly-bounded molecular ions predicted in Ref. [18].
The obvious solution is moving the ion trap toward the atom trap until it
reaches the characteristic distance dmm, and then stopping for a while with the
hope that the ion picks up enough phase difference for the phase gate. However,
it is likely that the energy difference between different qubit states [14] at dmm
is very small and a full phase of π is not obtainable in a reasonably short time.
This energy difference must be determined by experiments because it depends on
the short-range phase of the atom-ion interaction potential which is not known.
In short, whether stopping at the minimum distance dmm is good enough cannot
be answered with certainty by a purely theoretical calculation despite the strong
suggestion that it is not.
There is a straightforward experimental procedure to detect the characteristic
distance dmm of the micromotion effect. One prepares the ion in the ground Floquet
state in the ion trap initially placed very far from the atom trap. The ion trap is
then slowly brought closer to the atom trap up to some distance dmin and back to
the initial position. If dmin > dmm, one expects the ion to come back to its initial
66 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
ground state; on the other hand, when dmin < dmm, branching would occur and
only a fraction of the ion’s wave function is in the ground state.
As indicated in Fig. 5.7(a), while the vibrational energy levels are affected
by the micromotion at small trap distances, the energy curves for the low-lying
molecular states, for instance the u−3 level, stay in good agreement with the re-
sult obtained by the harmonic approximation shown in Fig. 5.2(b) for all trap
distances. Note that the quasienergy curve for the u−3 level goes out at the top
and comes back at the bottom of the picture. The fact that the molecular Floquet
states have smooth energy curves without any avoided crossing of observable size
means that the micromotion does not greatly affect these states. The reason is
that the energy gap between a low unperturbed molecular state and other states,
as shown in Fig. 5.2(b), are much larger than the micromotion frequency ω and
hence the oscillating terms of the micromotion cannot couple the unperturbed
molecular state to other states. In other words, the unperturbed molecular state
does not come into resonance when the trap distance is changed from 10 li to 0,
and as a result the oscillating terms have little effect on the energy structure of
these states. Consequently, the wave function of these Floquet molecular states
must consist of a dominant time-independent part which is just the unperturbed
wave function and a small time-dependent part that oscillates at the frequency
ω. As we show in the next section, the fact that these molecular states are al-
most unaffected by the micromotion offers a way to implement the quantum phase
gate despite the difficulties caused by the micromotion that we discussed in the
preceding paragraphs.
To compare the effect of the two oscillating terms of the micromotion, we
calculate the quasienergies of the system when only the first term V1 is included.
The result is shown in Fig. 5.8; the energy curve of the ground level is similar to the
result obtained by the harmonic approximation except that new avoided crossings
5.4. NUMERICAL CALCULATION OF THE QUASIENERGIES 67
appear at what was the 2ω resonances in the approximate energy diagram shown
in Fig. 5.5(b). This structure can be explained by the lowest-order perturbation
theory described in Sec. 4.3. It confirms the observation we made in Sec. 5.3 that
the first term only has a weak effect. Although they seem to be insignificant,
the new avoided crossings lead to a completely different dynamics of the system
in an adiabatic process. This means an oscillating term, even if relatively weak,
must still be taken into account if it results in resonances between the unperturbed
energy levels.
We see in Figs. 5.7(a) and 5.8 that most of the dramatic change of the energy
structure is caused by the second term V2 of the micromotion. This coupling is
so strong that perturbation methods are no longer good enough to account for its
influence. Because V2 contains the momentum operator P which is related to the
velocity of the micromotion, the intuition that the micromotion must be important
because its classical velocity is large is correct after all. In the limit of very large
ratio ω/ω0, we may neglect the first term V1 in the micromotion Hamiltonian for
the reasons explained in the last paragraph of Sec. 5.3. In the same paragraph we
also showed why the characteristic distance dmm should not change when the ratio
ω/ω0 increases up to 100. Since V1 can be ignored when ω/ω0 ' 100 , numerical
computation of the quasienergies is considerably simpler in this limit.
Last but not least, we must stress an important point regarding the numerical
calculation of the quasienergies of a trapped atom-ion system. One may notice
that the Floquet Hamiltonian and time-propagator method can be used with the
original Hamiltonian given in Eq. (3.2). The reason we always work in the trans-
formed picture is that the transformation of Cook et al. reduces the magnitude of
the time-dependent terms in the Hamiltonian by a factor of ω0/ω and hence much
faster numerical convergence is achieved. In fact, given the same size of the Floquet
Hamiltonian matrix, the accuracy of the quasienergies obtained when working with
68 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
the original Hamiltonian is much poorer than that of the quasienergies obtained
in the transformed picture.
Remarks on the numerical calculation of the quasienergies
When the micromotion is neglected, the energies of the system are the eigenen-
ergies En of the unperturbed Hamiltonian. If the micromotion Hamiltonian is
increased from 0 to Hmm(t), these eigenenergies change to the exact quasiener-
gies εn. Thus it is possible to assign to each En a corresponding quasienergy εn:
En → εn. When N eigenstates from |n1〉 to |n2〉 of the unperturbed system are
used as a basis to calculate the quasienergies, it is known that the quasienergies
lying near the boundary of the basis, e.g. the εn for which n ∼ n1 or n ∼ n2,
are not accurate. For our calculation in which 253 states are used, the results for
εn with n = [−5 . . . 90] are sufficiently accurate. Thus, we are confident that the
ground state quasienergy ε0, and the quasienergy levels that interact significantly
with it, are obtained with high enough accuracy for us to make definite statements
about the micromotion effect in the adiabatic process.
However, there is a remarkable difficulty in the reading of the numerical result.
Since all the quasienergies are projected down to one zone, it is not easy to identify
the accurate quasienergy levels from the inaccurate levels. In Fig. 5.7(a) and 5.7(b),
for instance, we can only identify the ground quasienergy level for trap distances
larger than 5 li. In this region we can still see the trace of the ground level by a
comparison with the ground eigenenergy E0 of the unperturbed system. Beyond
this trap distance, the ground quasienergy level disintegrates completely and we
no longer know which levels are accurate. There is a procedure to eliminate the
inaccurate levels from the figure but it is tedious and not necessary since the
knowledge of the ground level up to d = 5 li is good enough for us to understand
the effect of the micromotion.
5.4. NUMERICAL CALCULATION OF THE QUASIENERGIES 69
The molecular states, on the other hand, are virtually not affected by the
micromotion and thus there are no difficulties in identifying them for the entire
range of trap distances. We simply compare the quasienergies of these molecular
levels with the corresponding eigenenergies of the unperturbed system to find their
index numbers n.
70 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
Figure 5.7: (a) Quasienergies of the trapped atom-ion system plotted in the firstzone [−~ω/2, ~ω/2]. While the vibrational energy levels are affected by the micro-motion at small trap distances, the energy curves for the low-lying molecular statesstay in good agreement with the results obtained by the harmonic approximation.(b) An enlarged plot of the ground level Floquet state. The micromotion-inducedavoided crossings appear and the energy curve begins to disintegrate around thecharacteristics distance dmm ' 5.7 li.
5.4. NUMERICAL CALCULATION OF THE QUASIENERGIES 71
Figure 5.8: Quasienergy levels obtained when only the first term V1 of the mi-cromotion is included, the energy curve of the ground level is similar to the resultobtained by the harmonic approximation except that the new avoided crossings ap-pear at what was the 2ω resonances in the approximate energy diagram shown inFig. 5.5.
72 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
5.5 Adiabatic evolution
The effect of the micromotion excitation on the adiabatic quantum gate may be
seen directly from the evolution of the wave function during a long duration. In
Fig. 5.9 the trap distance is changed by five harmonic oscillator lengths (which is
around 150 nm) during a period of τ = 160 µs. The trap distance as a function of
time follows a linear dependence from the initial value d0 = 10 li to d1 = 6 li, then
continues with a parabola up to the final value dmin = 5 li before reverting back
to d0 along a symmetric curve. The slope of the parabola and that of the straight
line are equal at d1. Also, the slope of the parabola vanishes at the minimum trap
distance.
In the harmonic approximation, the adiabatic condition requires that the speed
of the trap |d| is much smaller than the speed of the ion’s secular motion ω0 li. As
we see in a short while, the process shown in Fig. 5.9 is nearly adiabatic in the
harmonic approximation but is no longer so when the micromotion is included.
For the purpose of comparision, the evolution of the wave function is done
in two different cases: (i) in the harmonic approximation when the micromotion
Hamiltonian is excluded and (ii) in the full picture when the micromotion is taken
into account. Time-propagation in the latter situation is quite tricky: The time
step must be small enough compared with the short period T of the micromotion
which is in turn very small compared with the evolution duration τ . Thus, we
end up with a large number of time steps. In this case the fourth-order method
based on the Magnus propagator described in the previous section is extremely
useful since it helps reduce enormously the number of time steps required for each
period of the micromotion. Another important note is that the error of the result
is proportional to the norm of the Hamiltonian matrix; hence, the evolution is
done in the interaction picture of the unperturbed Hamiltonian H0 to exploit the
small norm of the Hamiltonian matrix HI in the interaction picture.
5.5. ADIABATIC EVOLUTION 73
Figure 5.9: Change of the trap distance during the controlled collision process.
In the fist case (without the micromotion), the system is prepared in the ground
state |Φ0〉 at the initial trap distance d0 [see the energy diagram of Fig. 5.2(a) on
page 44] and then the ion trap is moved toward the atom located at x = 0. Note
that the minimum trap distance dmin of the process in consideration is at the
location of the avoided crossing where the system crosses to the strong interaction
region. In Fig. 5.10, the magnitude of the wave function is plotted at times t =
0, τ/2, τ which corresponds to trap distance values d = d0, dmin, d0, respectively.
We observe that the system is in a vibrational state initially but it transits to a
molecular state (signalled by the sharp oscillation near the origin) at around dmin
before reverting back to the vibrational state. The final wave function |Φf〉 is very
similar to the initial one. Numerical integration shows that the overlap |〈Φ0|Φf〉|
is 0.999, which corresponds to a gate fidelity of F ≈ 0.999 (see Appendix C). This
fidelity can be improved by increasing the duration or optimizing the function d(t)
[14].
In the second case when the micromotion is considered, the system is prepared
in the ground Floquet state at the initial trap distance d0. This ground Floquet
74 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
Figure 5.10: The evolution of the system in the harmonic approximation. Themagnitude of the wave function is plotted at (a) t = 0, d = 10 li, (b)t = τ/2, d = 5 liand (c) t = τ/2, d = 10 li.
5.5. ADIABATIC EVOLUTION 75
wave function is a Gaussian wave packet u0(t) that oscillates periodically in time
(see Appendix B). Initially, the system moves along the ground quasienergy curve
in Fig. 5.7 on page 70. Again the magnitude of the wave function is plotted at
times t = 0, τ/2, τ . Note that the minimum trap distance is smaller than the
micromotion distance dmm so we expect the micromotion excitation to play an
important role. The result is shown in Fig. 5.11. The final wave function uf (t)
is now very different from the initial one. The sharp oscillation in the final wave
function indicates excitation to high energy levels. We observe that these sharp
oscillations occur only when the trap distance d < dmm, that is, when the system
crosses to the chaotic region. Although a good fraction of the wave function is still
localized in the ion trap, the ion now possesses enough kinetic energy to travel to a
very large distance. The overlap of the wave functions |〈u0(τ)|uf (τ)〉| is only 0.590
in this case and thus the fidelity is F ≈ 0.511. This very poor fidelity demonstrates
the negative effect of the micromotion excitation. A final remark is that, in a real
system, the atom will be lost if it gains enough kinetic energy through micromotion
excitation and scatters to a large distance from the center of the atom trap. This
micromotion-induced atom lost were indeed observed in experiments [27, 29, 31].
76 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
Figure 5.11: The evolution of the system when the micromotion is included. Themagnitude of the wave function is plotted at (a) t = 0, d = 10 li, (b)t = τ/2, d = 5 liand (c) t = τ/2, d = 10 li.
5.6. A SCHEME TO BYPASS THE MICROMOTION EFFECT 77
5.6 A scheme to bypass the micromotion effect
We already showed that an adiabatic process using the ground level may not be
practically possible because of the micromotion excitation. However, the fact that
the low-lying molecular states are almost unaffected by the micromotion means
that we can bypass the micromotion effect by making a transition from the ground
level to one of the molecular levels at a distance larger than dmm before the mi-
cromotion becomes important. More specifically, we may prepare the ion in the
ground Floquet state at a large trap distance and then move it adiabatically to-
ward the atom trap. When the trap distance reaches some value d1 which is slightly
larger than dmm, we make the transition from the ground level u0 to a molecular
level lying below whose wave function has a sufficient overlap with u0. This tran-
sition can be realized by applying an electromagnetic field ~E(t). When the system
is in this Floquet molecular state it will be “protected” from the micromotion and
we may continue the adiabatic process up until d = 0. This modified adiabatic
process can be used to implement the quantum controlled phase gate provided that
the transition strengths for the different qubit states are equal at d1.
In principle, we need to investigate the transition between two Floquet states:
u0 and a Floquet molecular state, say the one lying two levels below the ground
level, u−3. Nevertheless, when d > d1 we know that the micromotion has little
effect and the dominant parts of these Floquet states are given by the unperturbed
eigenstates |0〉 and |−3〉 shown in Fig. 5.2; hence, it is possible to use the states
|0〉 and |−3〉 to study the transition between the exact Floquet states without
encountering too much error.
Suppose the one-dimensional system described in Eq. (5.1) is subjected to an
electromagnetic field,
~E(t) = E0 cos(ωf t)~ex. (5.48)
78 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
The coupling energy resulting from the interaction of this field with the atom’s
induced dipole pa = 4πε0αe/x2i and the ion’s charge e is
Vcp =4πε0αe
X2i
E(t)− eXiE(t). (5.49)
If we choose the distance at transition to be d1 = 6 li, the frequency ωf of the
electromagnetic field must satisfy the resonance condition
~ωf = E0(6 li)− E−3(6 li), (5.50)
and a quick glance at Fig. 5.2(b) reveals that
ωf ≈ 62ω0 = 2π × 6.2MHz. (5.51)
We need to evaluate the matrix element |〈0|Vcp |−3〉| to know whether the coupling
is strong enough to make a transition in practice. At d1 the first term in the
potential Vcp is negligible compared with the second term; thus, we only need to
compute the element |〈0|Xi |−3〉|. This gives
|〈0|Xi |−3〉| ≈ 1.3× 10−4 li. (5.52)
The Rabi frequency of the transition is
Ω =eE0
~|〈0|Xi |−3〉|, (5.53)
and for an electric field strength of 2 Vcm−1 we obtain a Rabi frequency of around
1 MHz which corresponds to a sufficiently strong transition. The numerical values
required for the frequency ωf and the field strength are both reasonable; thus, it is
practically possible to carry out a transition from u0 to u−3 at d1. A transition from
5.7. EXCESS MICROMOTION 79
a vibrational state to a molecular state may also be realized by Raman transitions.
But a full understanding of Raman transitions between the Floquet states of a
periodic system requires further studies.
Another interesting application of the scheme above is the creation of an atom-
ion macro molecule. At the end of the process when the atom and ion are in the
state u−3 at d = 0, it is possible to make a transition to a much lower molecular
state. After that we may turn off the atom trap and what is left is a bounded atom-
ion molecule stored in the ion trap. The same procedure between the resulting ion
and another atom can be repeated to create a larger molecule as long as the stability
condition for the ion trap is not violated. Such an atom-ion macromolecule may
well be of interest in its own right.
5.7 Excess micromotion
In practice, there are two types of micromotion in an ion trap. The one we de-
scribed so far is the intrinsic micromotion which comes from the driving potential.
The other type of micromotion is the excess micromotion which comes from im-
perfections in experimental setups. In Ref. [36], Berkeland et al. give a detailed
description of the excess micromotion and its unwanted effects in high-resolution
spectroscopy of an ion in a linear Paul trap. Ideally, the excess micromotion must
be eliminated. Here we investigate a realistic situation where a small amount of
the excess micromotion is left in the motion of a trapped ion.
Although the quantum dynamics of a trapped ion is well understood [42], its
motion under the influence of the excess micromotion has been described only with
a classical treatment [36] . In Appendix B we generalize Glauber’s approach [40]
to describe the quantum motion of a trapped ion in the presence of the excess
micromotion. The energy spectrum and wave functions of the ion are obtained.
80 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
Basically, the excess micromotion is caused by a uniform static electric field
~Edc or a phase difference ϕac between the ac potentials of the ion trap’s electrodes.
This phase shift results in an oscillating electric field ~Eac sin (ωt), where ω is the
frequency of the driving potential. With Rt denoting the radius of the trap’s
cylindrical electrodes, the amplitude Eac is related to the phase difference ϕac by
[36]
Eac 'mi
eqω2Rtϕac. (5.54)
Suppose ~Edc and ~Eac points along the x axis; the Schrodinger equation for the
motion along the x axis of a single trapped ion is now
i~∂
∂tΨ(x, t) =
− ~2
2mi
∂2
∂x2+
1
8miω
2x2[a+ 2q cos(ωt)]
− eEdcx− eEacx sin(ωt)
Ψ(x, t). (5.55)
The term proportional to x2 in the potential describes the driving potential in
the ideal case when there is no excess micromotion. This potential leads to the
intrinsic micromotion whose effect has been discussed in previous sections. The
other terms which are linear in x describe the excess micromotion.
To investigate the effect of the excess micromotion, we first follow Cook et al.
[46] and ignore all the time-independent operators in the Schrodinger equation.
The solution is then, obviously,
Ψ(x, t) = Ψ(x, 0) exp− i
~
[q4miωx
2 sin(ωt) +eEac
ωx cos(ωt)
]. (5.56)
The phase factor here reduces to the phase derived by Cook et al. in Eq. (2.12)
when Eac vanishes. Again, one may argue that the main effect of the time-
dependent terms in the potential is to introduce this oscillating phase to the wave
5.7. EXCESS MICROMOTION 81
function. Therefore, we are motivated to write the wave function in Eq. (5.55) as
Ψ(x, t) = exp− i
~
[q4miωx
2 sin(ωt) +eEac
ωx cos(ωt)
]Φ(x, t) (5.57)
with the expectation that the time-dependent part of Φ(x, t) is sufficiently small.
A substitution of the above expression into Eq. (5.55) yields the following equation
for Φ(x, t):
i~∂
∂tΦ(x, t) =
− ~2
2mi
∂2
∂x2+
1
2mω2
0x2 − eEdcx−mi(γω0)2x2 cos(2ωt)
+ 2i~γω0
(x∂
∂x+
1
2
)sin(ωt+
γω0
ωeEacx sin(2ωt)
+ i~eEacmiω
cos(ωt)∂
∂x+
1
2mi
(eEac
ω
)2
[cos(ωt)]2
Φ(x, t). (5.58)
The last term gives rise to a time-dependent global phase in the wave func-
tion with no physical meaning and hence can be discarded. Consequently, the
Hamiltonian of a single trapped ion in the transformed picture is
H(t) = H0 +Hmm(t) +Hex(t), (5.59)
where
H0 =P 2
2mi
+1
2miω
20X
2 − eEdcX (5.60)
is the unperturbed Hamiltonian, Hmm(t) is the Hamiltonian of the intrinsic micro-
motion which has the same form as in Eq. (2.16), and Hex(t), the Hamiltonian of
the excess micromotion, is
Hex(t) =γω0
ωeEacX sin(2ωt)− eEac
P
miωcos(ωt). (5.61)
82 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
In Appendix B we demonstrate that the addition of the excess micromotion
does not lead to any change in the quasienergies of a single trapped ion. In other
words, the ion possesses a harmonic oscillator-like quasienergy spectrum as given
in Eq. (5.34), where µ is a function of ω, a, q and does not depend on Edc and Eac.
However, the excess micromotion does lead to significant changes in the structure
of the Floquet states.
Now let us work out the effect of the excess micromotion for the one-dimensional
trapped atom-ion system in which the atom is fixed at the center of the atom trap
as described in Fig. 3.1. In this case, the presence of the atom results only in
an additional interaction potential in the Hamiltonian given in Eq. (5.59). Let us
choose the location of the atom as the coordinate origin, then the Hamiltonian of
the system is
H(t) =H0 +Hmm(t) +Hex(t), (5.62)
where
H0 =P 2i
2mi
+1
2miω
20(Xi − d)2 − eEdc(Xi − d)− αe2
2X4i
, (5.63)
Hmm(t) =−mi(γω0)2(Xi − d)2 cos(2ωt)− γω0Xi − d, Pi sin(ωt), (5.64)
and
Hex(t) =γω0
ωeEac(Xi − d) sin(2ωt)− eEac
Pimiω
cos(ωt). (5.65)
Thus, the contribution of the field Edc is to change the structure of the unperturbed
eigenstates. Apart from an additive constant, the unperturbed Hamiltonian H0
5.7. EXCESS MICROMOTION 83
can be expressed as
H0 =P 2i
2mi
+1
2miω
20
(Xi − d−
eEdc
miω20
)2
− αe2
2X4i
. (5.66)
Thus, the effect of the dc electric field amounts to shifting the trap distance by
δd =eEdc
miω20
. (5.67)
For a small value of Edc, say, 0.01 Vm−1, with the values of mi and ω0 as given in
the numerical calculation carried out in Sec. 5.1, the distance shift is δd ≈ 0.7 li.
Here the distance shift is already significant and it can be even larger when either
the electric field Edc increases or the secular frequency ω0 decreases. Thus, care
must be exercised when such a uniform electric field exists in the trap.
It is advantageous to make the change of variable d′ = d+δd to bring the form
of H0 to that of the unperturbed Hamiltonian we considered in Eq. (5.2). Then
the eigenenergies and eigenstates of H0 at various values of d′ can be readily taken
from the calculation done in Sec. 5.1. With this transformation, the micromotion
Hamiltonian can be written as the sum H ′mm(t)+Hac(t)+ Hdc(t), where the forms
of H ′mm(t) and Hac(t) are, respectively, those of Hmm(t) and Hex(t) with d replaced
with d′, and Hdc(t) is
Hdc(t) = −2mi(γω0)2δd(Xi − d′) cos(2ωt)− 2γω0(δd)Pi sin(ωt). (5.68)
So, by the change of variable we may now view the unperturbed system as
unaffected while the dc field effectively results in two additional oscillating terms in
the Hamiltonian. The fact that a dc electric field leads to not only a displacement
of the ion but also an additional micromotion is not a surprise because an off-
centered ion is subjected to a stronger force of the driving rf field.
84 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
To compare the magnitude of the excess micromotion caused by the ac field
Hac(t) and the intrinsic micromotion H ′mm(t), we introduce the length scale lac
associated with the force eEac by
eEac = miωω0lac (5.69)
and write the Hamiltonians of the excess micromotion in the following dimension-
less forms:
Hac(t)
~ω0
=lac
li
[γXi − d′
lisin(2ωt)− liPi
~cos(ωt)
](5.70)
and
Hdc(t)
~ω0
= −2γδd
li
[γXi − d′
licos(2ωt) +
liPi~
sin(ωt)
]. (5.71)
By a comparison with the dimensionless form of H ′mm(t) given in Eq. (5.4) (with
d replaced by d′), we see that the relative strength of the excess micromotion is
described by the factors lac/li and δd/li. From Eqs. (5.54) and (5.69) we obtain
lac ' Rtϕac. For a millimeter-sized trap with electrode radius Rt ' 1 mm, a phase
shift as small as ϕac ' 10−2 degrees yields lac/li ' 6. Thus, the effect of the ac
excess micromotion is comparable to that of the intrinsic micromotion even when
the phase shift is very small. To avoid further complications due to the excess
micromotion we have to keep this phase shift and the electric field Edc as small as
possible.
In circumstances when δd/li and lac/li are not negligible, the numerical methods
described in Sec. 5.4 can be readily applied to obtain the quasienergy diagram of
the system. However, for a qualitative picture, it suffices to compute the coupling
strength as described in Sec. 5.3. When the excess micromotion is considered,
we need to calculate six instead of two coupling strengths. Figure. 5.12 shows
5.7. EXCESS MICROMOTION 85
Figure 5.12: A plot of the coupling strength |〈0|V3,4 |5〉| as a function of trapdistance.
the representative coupling strength |〈0|V3,4 |5〉|, plotted against the original trap
distance d, for the operators
V3 =γω0
ωeEac(Xi − d),
V4 = −eEacPimiω
, (5.72)
of the ac excess micromotion Hamiltonian Hac(t); the two operators of Hdc(t) only
differ from V3,4 by the factor 2γδd/lac. There is a sharp rise at around dmm ' 4.8 li,
which differs from the corresponding value dmm ' 5.3 li obtained in Sec. 5.3 by
δd = 0.7 li, as it should be. We observe that the coupling strengths, within the
range of resonance, of all the operators in H ′mm(t), Hac(t), and Hdc(t) also exhibit
a sudden increase at dmm ' 4.8 li.
We may now describe the key features of the quasienergy diagram. At large
trap distances, that is, d > dmm, the interaction potential can be neglected and
the quasienergy of the system is essentially the energy of a single trapped ion.
Recall that the excess micromotion does not change the quasienergy of a single
trapped ion, the asymptotic quasienergies of the system must still be given by
86 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
Eq. (5.34). When the trap distance decreases we expect resonances to occur and
micromotion-induced avoided crossings to appear in the quasienergy diagram. As
a result, the qualitative behavior of the quasienergy curves shown in Fig. 5.7 still
holds when excess micromotion of reasonable magnitudes exists in the system.
The only difference expected in the quasienergy diagram is a shift of δd in the
micromotion characteristic distance dmm.
5.8 A more realistic 1D model
The simplified model of a fixed atom interacting with a trapped ion considered
in the previous sections helps us understand the quasienergy structure of a more
realistic system. In fact, all of the conclusions we obtained from the simplified
model are still qualitatively true when we allow the trapped atom to move. The
Hamiltonian of this system is then given in Eq. (3.4). Let us consider the case
when ωa = ω0. By switching the coordinates to that of the relative motion
X = Xi −Xa, (5.73)
and the shifted coordinate of the center-of-mass motion
Xcm =maXa +mi (Xi − d)
ma +mi
, (5.74)
we may write the Hamiltonian of the system as
H(t) = Hcm(t) +Hrel(t) +H1(t), (5.75)
5.8. A MORE REALISTIC 1D MODEL 87
with the center-of-mass Hamiltonian
Hcm(t) =P 2
cm
2M+
1
2Mω2
0X2cm
− mi
M
[M(γω0)2X2
cm cos(2ωt) + γω0Xcm, Pcm sin(ωt)], (5.76)
the relative Hamiltonian
Hrel(t) =P 2
2m+
1
2mω2
0 (X − d)2 − αe2
2X4
− m
mi
[m(γω0)2 (X − d)2 cos(2ωt) + γω0X − d, P sin(ωt)
], (5.77)
and the coupling term
H1(t) =− 2m(γω0)2Xcm (X − d) cos(2ωt)
− 2γω0
[XcmP +
m
M(X − d)Pcm
]sin(ωt), (5.78)
where M = ma +mi is the total mass and m = mami
ma+miis the reduced mass of the
atom-ion system.
Without the micromotion, the Hamiltonians for the center-of-mass mode and
relative mode are decoupled. It is the term H1(t) caused by the micromotion that
couples these modes. It is clear that the Hamiltonian of the relative motion Hrel(t)
is, apart from some multiplicative constants, the Hamiltonian of the simplified
model we considered in Eq. (5.1). This was our motivation for the simplification
in the first place.
The unperturbed eigenstates of the system are the product states
|ncm, n〉 = |ncm〉 |n〉 , (5.79)
88 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
and the unperturbed eigenenergies are the sums
Encm,n =
(ncm +
1
2
)~ω0 + En. (5.80)
where |n〉 and En are the unperturbed eigenstates and eigenenergies of the relative
Hamiltonian as discussed in Sec. 5.1. The only difference is the replacement of the
ion mass mi by the reduced mass m and hence there is no qualitative change in
the eigenenergy diagram as shown in Fig. 5.2.
Now let us analyze how the quasienergy of the ground Floquet state looks when
the trap distance is varied in an adiabatic process. The asymptotic quasienergies
of the system are the sums of the eigenenergies of the trapped atom,
Ea =
(na +
1
2
)~ω0, (5.81)
and the quasienergy of a single trapped ion given in Eq. (5.34). Furthermore,
the ground Floquet state of the system is the product |0a〉 |u0(t)〉 of the ground
harmonic oscillator state of the atom and the ground Floquet state of the ion.
The asymptotic form of the ground Floquet state and its quasienergy are well ap-
proximated by the unperturbed eigenstates |ncm = 0, n = 0〉 and their eigenvalues
E0,0.
When we decrease the trap distance, the micromotion will become important
when the ground unperturbed state |0, 0〉 comes into resonance with an excited
state |ncm, n〉, that is,
Encm,n − E0,0 = ~ω. (5.82)
If the coupling strength |〈0, 0|V |ncm, n〉| of the oscillating terms in H(t) is suffi-
ciently large, micromotion-induced avoided crossings of observable size will occur
along the quasienergy curve of the ground Floquet state.
5.8. A MORE REALISTIC 1D MODEL 89
We need to compute the coupling strengths to understand the impact of the
micromotion. For the two oscillating terms in Hcm(t) they are, apart from the
factor mi/M of order one,
M(γω0)2∣∣〈0, 0|X2
cm |ncm, n〉∣∣ =
γ2
2~ω0
(√2δncm,2 + δncm,0
)δn,0, (5.83)
and
γω0 |〈0, 0| Xcm, Pcm |ncm, n〉| =γ
2ncm~ω0
(√2δncm,2 + δncm,0
)δn,0. (5.84)
These couplings are nonvanishing only if n = 0; that is, resonances must occur in
the subspace of the center-of-mass unperturbed eigenstates for them to be of any
importance. Moreover, as long as ω & 10ω0 these resonances are only possible if
ncm > 2, which implies that the coupling strengths described above always vanish
at resonances. Hence, we conclude that the two oscillating terms in Hcm(t) have
negligible effect. This is not a surprise when we notice that Hcm(t) is just the
Hamiltonian of a single trapped ion for which the micromotion is known to have
little effect in the energy structure of the first few states.
Similarly, for the two oscillating terms in Hrel(t), the coupling strengths are
nonvanishing only for resonances in the subspace of the relative-mode unperturbed
eigenstates. Therefore, these couplings reduce to the ones we discussed in Sec. 5.3.
By repeating the numerical calculation shown in that section with mi replaced by
m, we find that these coupling strengths are significant only when d . 4.8 lrel with
lrel =
√~
mω0
≈ 824 a0. (5.85)
Finally, the coupling strengths for the oscillating terms in H1(t) are nonvan-
ishing only if ncm = 1, which corresponds to the the following sideband resonance
90 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
in the subspace of the unperturbed eigenstates of the relative mode
En − E0 = ~ (ω − ω0) . (5.86)
Furthermore, when ncm = 1, the coupling strengths for the two terms in H1(t)
reduces, apart from multiplicative constants of order one, to those considered in
Sec. 5.7 and shown in Fig. 5.12. Again, a repeat of the numerical calculation with
mi replaced with m shows that these coupling strengths are significant only when
d . 4.8 lrel.
Since all the coupling strengths at possible resonances are either vanishing or
only significant at trap distances smaller than a characteristics distance dmm '
4.8lrel, all of the conclusions we derived about the micromotion effect in the pre-
vious sections are still true for the more realistic model where the atom is allowed
to move. Moreover, one could see the scheme mentioned in Sec. 5.6 is still pos-
sible since the gap between the low lying unperturbed molecular states of the
relative Hamiltonian is much larger than the micromotion frequency and hence
these states are protected from resonances and the resulting micromotion-induced
avoided crossings.
So the qualitative behavior of the micromotion effect is very much the same
for the model considered here, which corresponds to ω0 = ωa, as it is for the
simplified model of Sec. 5.3, which corresponds to ω0 ωa. Thus, we expect all
the main features of the micromotion effect to occur for ω0 . ωa. These features
include resonances, micromotion-induced avoided crossings, and the existence of
the characteristic distance dmm. In the other regime where ω0 ωa, which means
a very large trapping frequency for the ion trap, the ion must be tightly bounded
to the center of its trap; so it is possible for the atom trap to come really close to
the ion without sensing a very small amplitude of the micromotion. This fact can
5.8. A MORE REALISTIC 1D MODEL 91
be seen easily by considering the extreme limit of ω0 →∞, for which the trapped
ion can be treated as a fixed ion at the center of its trap. In this case, we retrieve
the simplified model described in Sec. 5.3 with the atom and the ion interchanging
their roles. It is obvious that the micromotion does not enter the Hamiltonian for
this model and the exact energy diagrams must look similar to the ones shown
in Fig. 5.2. These observations imply that, for applications such as the quantum
controlled phase gate which require the atom trap and the ion trap to be as close
as possible, it is better to use a configuration with a large secular frequency ω0.
The precise condition for the trapped ion to be considered as a fixed ion, mean-
ing micromotion would be of no importance, requires the value of the minimum
trap distance involved in each specific problem. This minimum trap distance is
around dc = 2R1/3a l
2/3a for the adiabatic quantum phase gate [see Eq. (5.15)]. At
this trap distance, the shortest atom-ion distance is roughly dc− la− li. For the ion
to be treated as a fixed ion, this distance must be much larger than the harmonic
oscillator length li of the ion, which yields
2liRa
+laRa
2
(laRa
)2/3
. (5.87)
After interchanging the subscripts i and a, one obtains the condition for the
trapped atom to be treated as a fixed atom.
For the numerical calculations in this section we use ω0 = ωa = 2π × 100 kHz
and ω = 2π × 1.27 MHz, which are unrealistically small for an ion trap. We
repeated our calculation of the coupling strengths for the more reasonable values
ω0 = ωa = 2π × 1 MHz and ω = 2π × 12.7 MHz and observed the familiar sharp
rising at dmm ≈ 6.1 lrel. Hence, the micromotion effect, as we describe it, is still
valid for these values of the secular frequency and the micromotion frequency. For
typical trapping frequencies encountered in experiments, say ωa = 2π × 100 kHz
92 CHAPTER 5. MICROMOTION EFFECT IN ONE DIMENSION
and ω0 = 2π × 1 MHz, we have li ≈ 0.5 la. Since li does not differ greatly from la,
the geometry of this realistic system is closer to that of a symmetric model with
ω0 = ωa, which is discussed in this section, than it is to a model with ω0 ωa or
ω0 ωa.
5.9 Summary
The short-range behavior of the −1/r4 atom-ion interaction potential can be mod-
elled by a single quantum defect parameter — the short-range phase. With this
phase one may derive a boundary condition for the Numerov computation of the
eigenenergies and eigenstates of the unperturbed system. The effect of the mi-
cromotion can be inferred from the coupling strengths of the two operators repre-
senting the micromotion. The fact that these coupling strengths increase sharply
at the characteristic trap distance dmm means that the micromotion effect is im-
portant when the trap distance is smaller than dmm. This is confirmed by a full
numerical computation of the quasiernergies using the time-propagator method
and the Floquet Hamiltonian method. The micromotion poses various difficulties
for applications based on an adiabatic process. However, there is a simple scheme,
which relied on a transition to a molecular state, for bypassing the micromotion
effect.
Chapter 6
Micromotion effect in three
dimensions
6.1 Micromotion Hamiltonian
Again we consider the system illustrated in Fig. 3.1, but now the atom is stored
in a 3D harmonic trap and the ion is confined in a linear Paul trap [36]. Let us
denote the components of the position vectors by ra,k for the atom and ri,k for the
ion (k = x, y, z). The driving potential used to confine an ion in a 3D rf trap is
Vi(t) =∑k
1
2miWk(t)r
2i,k, (6.1)
where
Wk(t) =ω2
4[ak + 2qk cos(ωt)] , (6.2)
and the parameters ak and qk obey Eq. (2.6). The axial trapping frequency of the
Paul trap is ωi,z = ω√a/2 and the radial secular frequencies are
ωi,x = ωi,y = ω0, (6.3)
93
94 CHAPTER 6. MICROMOTION EFFECT IN THREE DIMENSIONS
where ω0 is given in Eq. (2.8). By taking into account that the trapping potential
for the atom is
Va =∑k
1
2ma ω
2a,k r
2a,k, (6.4)
one can show that the transformed Hamiltonian for the trapped atom-ion system
is the sum
H(t) = Ha +Hi(t) + Vint (6.5)
of the Hamiltonian of a single trapped atom,
Ha =∑k
(P 2a,k
2ma
+1
2ma ω
2a,k r
2a,k
), (6.6)
the transformed Hamiltonian of a single trapped ion,
Hi(t) =∑k
[P 2i,k
2mi
+1
2mi ω
2i,k (ri,k − d)2
]+Hmm(t) (6.7)
with
Hmm(t) =−mi(γω0)2(Xi − d)2 cos(2ωt)− γω0Xi − d, Pi,x sin(ωt)
−mi(γω0)2(Yi − d)2 cos(2ωt) + γω0Yi − d, Pi,y, (6.8)
and the interaction potential,
Vint = − αe2
2(~ri − ~ra)4. (6.9)
At large trap distances, Vint can be ignored so that the motions of the atom
and ion are decoupled and the quasienergies of the system are simply the sums of
the atom eigenenergies
Ea =∑k
(na,k +
1
2
)~ωa,k, (6.10)
6.1. MICROMOTION HAMILTONIAN 95
and the ion’s quasienergies, as mentioned in Eq. (5.34),
εi =∑k=x,y
(ni,k +
1
2
)~µk +
(ni,z +
1
2
)~ωi,z. (6.11)
This asymptotic energy spectrum can be well described in the harmonic approxi-
mation. However, we know that at small trap distances the interaction potential
will distort the unperturbed eigenstates and eigenenergies so that they come into
resonance with each other. This is when the exact quasienergies differ greatly
from the approximate eigenenergies and micromotion-induced avoided crossings
appear in the quasienergy diagram. Idziaszek et al. investigated the unperturbed
Hamiltonian for the cases of quasi-1D elongated traps and spherically symmetric
traps in Refs. [13, 14] and the energy spectrum they obtained is very similar to
that shown in Fig. 5.2. Therefore, we expect micromotion-induced couplings to ex-
hibit the same sudden rise at some characteristic distance dmm as in Fig. 5.6. The
quasinergy diagram of the 3D system must also look similar to the one obtained in
Fig. 5.7, where the energy spectrum changes completely and micromotion-induced
avoided crossings appear at trap distances smaller than dmm. Thus, all the quali-
tative statements we made for the simplified system considered in Sec. 5.4 should
carry over to a realistic 3D system. The values of parameters of interest like the
characteristic distance dmm cannot be computed with high accuracies due to the
lack of knowledge about the exact value of the short-range phase and must be
determined by experiments.
Because of the asymmetrical geometry of the Paul trap and the atom trap, the
dynamics of the system in an adiabatic process depends on the direction along
which we move the traps. From the form of the driving potential Vi(t) in Eq. (6.1)
and the micromotion Hamiltonian Hmm(t), we notice that the micromotion occurs
only in the radial directions. Along the axial direction the ion behaves like a
96 CHAPTER 6. MICROMOTION EFFECT IN THREE DIMENSIONS
harmonic oscillator. Therefore, when the atom and ion are trapped very tightly in
the radial directions, it is likely that the micromotion effect is weaker if we move
the traps along the axial direction. However, even then the micromotion cannot be
ignored at small trap distances because the interaction potential Vint clearly couples
the motion along the radial directions to the motion along the axial direction and
thus it is possible for the micromotion to be propagated to the motion along the
axial direction. This transfer of the micromotion is negligible for relatively large
trap distances but it must be significant at small trap distances, especially when
the system transforms to a molecular state for which the atom and ion are close
to each other.
6.2 Quasi-1D trap configuration
For a better understanding of the micromotion propagation discussed above, let
us consider a quasi-1D system with the following elongated geometry
ωa,x, ωa,y, ω0 ωa,z, ωi,z, (6.12)
and
ωa,z = ωi,z. (6.13)
With these assumptions, from now on we may write the axial trapping frequency
for both the atom and the ion as ωz. It can be seen easily that the radial harmonic
oscillator lengths are much smaller than the axial harmonic oscillator length, that
is,
la,x, la,y, li la,z, li,z, (6.14)
6.2. QUASI-1D TRAP CONFIGURATION 97
with
la,x =
√~
maωa,x, la,y =
√~
maωa,y, la,z =
√~
maωz, (6.15)
and
li =
√~
miω0
, li,z =
√~
miωz. (6.16)
Suppose we prepare the ion and the atom in their respective ground states and
move the traps along the axial direction in an adiabatic process, the micromotion
effect can be understood by the following self-consistent approach. At large trap
distances, where d & lz, we have
|yi − ya| ' |xi − xa| |zi − za|, (6.17)
which is a consequence of Eq. (6.14). Thus, the interaction potential can be ap-
proximated by its Taylor series expansion up to the first order,
Vint = − αe2
2(~ri − ~ra)4
= −αe2
2
[(xi − xa)2 + (yi − ya)2 + (zi − za)2
]−2
= − αe2
2(zi − za)4
[1 +
(xi − xa)2 + (yi − ya)2
(zi − za)2
]−2
≈ − αe2
2(zi − za)4+ αe2 (xi − xa)2 + (yi − ya)2
(zi − za)6+ . . . . (6.18)
So the leading term in the expansion that couples the motion along the radial
directions to the motion along the axial direction is
Vcp = αe2 (xi − xa)2 + (yi − ya)2
(zi − za)6. (6.19)
98 CHAPTER 6. MICROMOTION EFFECT IN THREE DIMENSIONS
At large trap distances, the motion along the x and y axes is almost undis-
turbed by the interaction potential, thus the state for the motion along the radial
directions is approximately the ground harmonic-oscillator state for the atom and
the single-trapped-ion ground Floquet state, which is described by Glauber in
Ref. [40], for the ion. Thus, the wave function for the radial degrees of freedom is
the product state
Φxy = |0x〉 |0y〉 |u0x〉 |u0y〉 (6.20)
where |0x,y〉 (|u0x,0y〉) denotes the ground harmonic oscillator states (ground Flo-
quet states) of the atom (ion) for the x and y directions; respectively. We would
like to remind the reader that the states |u0x,0y〉 are time dependent. Next, we
may approximate the total wave function as
Ψ(~ra, ~ri, t) ≈ ΦxyΦ(za, zi, t). (6.21)
Then, the effective potential for the motion along the axial direction can be ob-
tained by tracing out the radial degrees of freedom
Veff(zi, za, t) = 〈Φxy|Vint |Φxy〉 ≈ −αe2
2(zi − za)4+αe2L(t)2
(zi − za)6(6.22)
with
L(t)2 = 〈Φxy| (xi − xa)2 |Φxy〉+ 〈Φxy| (yi − ya)2 |Φxy〉 . (6.23)
From the well-known properties of the quantum harmonic oscillator we have
〈0x|xa |0x〉 = 0,
〈0y| ya |0y〉 = 0,
〈0x|x2a |0x〉 = l2a,x/2,
〈0y| y2a |0y〉 = l2a,y/2. (6.24)
6.2. QUASI-1D TRAP CONFIGURATION 99
In addition, by using the wave function of the ground Floquet state of a single-
trapped ion, we show in Appendix D that
〈u0x|xi |u0x〉 = 0,
〈u0y| yi |u0y〉 = 0, (6.25)
and more importantly
〈u0x|x2i |u0x〉+ 〈u0y| y2
i |u0y〉 ≈ l2i
[1 +
3
16q2 cos(2ωt)
]. (6.26)
Hence, we obtain
L(t)2 ≈ 1
2
(l2a,x + l2a,y + 2l2i
)+
3q2
16l2i cos(2ωt), (6.27)
and the effective potential becomes
Veff(zi, za, t) = − αe2
2(zi − za)4+αe2
(l2a,x + l2a,y + 2l2i
)2(zi − za)6
+3q2
16
αe2l2i(zi − za)6
cos(2ωt).
(6.28)
The second term in the effective potential is just a small contribution in the
time-independent part of the potential; the inclusion of this term leads to only
insignificant changes in the wave function of the unperturbed system (we use the
term “unperturbed” to indicate that all the time-dependent terms in the Hamil-
tonian are neglected). Since we are only interested in the oscillating term that
represents the micromotion, we may neglect the second term and write the effec-
tive Hamiltonian for the motion along the axial direction as
Heff =P 2a,z
2ma
+1
2maω
2zz
2a +
P 2i,z
2mi
+1
2miω
2z(zi − d)2
− αe2
2(zi − za)4+
3q2
16
αe2l2i(zi − za)6
cos(2ωt). (6.29)
100 CHAPTER 6. MICROMOTION EFFECT IN THREE DIMENSIONS
Now we introduce the shifted center-of-mass coordinate
zcm =maza +mi(zi − d)
ma +mi
, (6.30)
and the relative coordinate
z = zi − za. (6.31)
Then, the effective Hamiltonian can be written as a sum of the center-of-mass
Hamiltonian
Hcm =P 2
cm
2M+
1
2Mω2
zz2cm (6.32)
with the total mass M , and the relative Hamiltonian
Hrel =P 2
2m+
1
2mω2
z(z − d)2 − αe2
2z4+
3
16
αe2q2l2iz6
cos(2ωt) (6.33)
with the reduced mass m. Since the center-of-mass Hamiltonian is independent
of the trap distance, we only need to consider the relative motion for which the
Hamiltonian can be expressed in the dimensionless form
Hrel
~ωz=
1
2
(lzP
~
)2
+1
2
(z
lz
)2
− 1
2
(R/lz)2
(z/lz)4+
3q2
16
(lilz
)2(R/lz)
2
(z/lz)6cos(2ωt) (6.34)
with lz =√~/mωz the harmonic oscillator length for the reduced mass. Since
|q| 1 and li lz, the time-dependent term that represents the micromotion
is indeed a small perturbation as long as z is not too small compared with the
characteristic length R1/3l2/3z encountered earlier in Eq. (5.15).
The effect of the oscillating term due to the micromotion can be studied by
a calculation of the micromotion-induced coupling strengths similar to what we
did in Sec. 5.3. Again we consider the 135Ba+ ion and 87Rb atom system with
ω0 = 2π × 1 MHz, ωz = 2π × 50 kHz, and q = 0.1. The representative coupling
6.2. QUASI-1D TRAP CONFIGURATION 101
strength between the ground state and the fifth excited state for the operator
Vmm =3q2
16
(lilz
)2(R/lz)
2
(z/lz)6(~ωz) (6.35)
of the micromotion, which is denoted by V(0,5)
mm , is shown in Fig. 6.1. Since our
estimation is only valid when z li, which is true for sufficiently large trap
distances, we only plot the coupling strength up to the minimum trap distance
dmin = 2 lz. This minimum trap distance should be small enough to get a sufficient
phase difference for the implementation of the adiabatic quantum phase gate.
We observe a characteristic sharp rising at a certain trap distance (d ≈ 4.5 lz),
which is similar to what we see for the micromotion-induced coupling discussed
in Sec. 5.3. However, the micromotion-induced coupling strengths in this case
are much smaller because of the factor q2 (li/lz)2 in the micromotion operator of
Eq. (6.35). Thus, we expect the quasienergy diagram of the system to be similar
to the one shown in Fig 5.8 but with much smaller micromotion-induced avoided
crossings. These small avoided crossings may be passed easily in an diabatic process
and hence it is likely that, for the described quasi-1D trap, the micromotion has
little effect on the implementation of an adiabatic quantum phase gate.
At trap distances that are considerably smaller than lz, the quasi-1D geom-
etry is no longer satisfied and the micromotion effect must become significant.
Therefore, in the application of creating the atom-ion macromolecule in which it
is necessary to reach the zero trap distance, one still needs to utilize the scheme of
Sec. 5.6 to bypass the micromotion effect. Although our analysis for the quasi-1D
geometric system is based on a rough estimation, it gives a hint that the micromo-
tion effect can be significantly reduced if we (i) use a linear Paul trap for the ion
with a radial trapping frequency much larger than the axial trapping frequency,
and (ii) move the traps along the axial direction in the adiabatic process. But we
102 CHAPTER 6. MICROMOTION EFFECT IN THREE DIMENSIONS
Figure 6.1: A plot of the coupling strength V(0,5)
mm as a function of trap distance.
must stress that, in reality, the axial direction cannot be identified with certainty
in experiments, and as a result there is always some amount of micromotion along
the direction identified as the “axial direction” in the experiments. In spite of this
imperfection, it is reasonable to expect that the micromotion effect is weaker when
the trap movement is carried out along the experimental axial direction.
6.3 Minimization of micromotion effect
Let us collect all the information we have so far to derive what an experimentalist
can do to minimize the effect of the micromotion on an adiabatic collision of a 3D
trapped atom-ion system. We will assume that the ion is trapped in a linear Paul
trap with radial trapping frequency (or secular frequency) ω0 and axial trapping
frequency ωz, and the traps are moved along the axial direction (the z axis) of
6.3. MINIMIZATION OF MICROMOTION EFFECT 103
both traps. It would become clear that our arguments can be easily applied to
obtain similar requirements for different trap configurations.
In the last paragraph of Sec. 5.3 we see that the micromotion effect in a 1D
system should be weaker if the micromotion frequency is much larger than the
secular frequency. This is also true in a 3D system because a large micromotion
frequency means that the oscillating terms of the micromotion only couple the
ground level of the system with very high excited levels. Since the micromotion-
induced coupling strength decreases with increasing energy gap, the micromotion
effect must be weak if the micromotion frequency is sufficiently large compared
with the secular frequency. So the first condition which should be realized to
minimize the micromotion effect is that the ratio
g1 =ω0
ω(6.36)
is as small as possible.
The second condition is based on the argument presented in the last two para-
graphs of Sec. 5.8 and it requires the knowledge of the minimum trap distance
dmin of the adiabatic process. When the trap distance equals dmin, an estimate of
the minimum atom-ion distance is dmin − la,z − li,z assuming dmin > la,z + li,z. If
the amplitude of the micromotion in the radial directions, which goes as qli [see
Eq. (2.9)], is very small compared with the minimum atom-ion distance, the micro-
motion would play little role in the interaction of the atom and the ion. Therefore,
the second condition is that the ratio
g2 =qli
dmin − la,z − li,z(6.37)
is as small as possible. If the excess micromotion is present the amplitude of the
total micromotion is bounded by q (li + δd+ lac) (as can be seen in Appendix B)
104 CHAPTER 6. MICROMOTION EFFECT IN THREE DIMENSIONS
and the factor g2 need to be modified a little.
The final requirement is already inferred from the calculation in Sec. 6.2, which
shows that the micromotion effect is indeed very weak in a quasi-1D trap con-
figuration with radial trapping frequencies much larger than the axial trapping
frequencies. This corresponds to the condition that the ratios
g3 =ωzω0
, (6.38)
and
g4 =ωa,z
minωa,x, ωa,y(6.39)
are as small as possible. To conclude, the ideal experimental configuration for
minimizing the micromotion effect is
g1, g2, g3, g4 1. (6.40)
Our analysis of the micromotion from Sec. 5.3 to the current section implies
that applications based on an adiabatic process, such as the creation of the atom-
ion macro molecule and the quantum phase gate, are realizable as long as we
can implement the following steps: (i) preparing the trapped ion in its ground
Floquet state, (ii) moving the atom trap adiabatically toward the ion trap and (iii)
transferring the system to a low lying molecular state at a trap distance d1 > dmm
before continuing with the adiabatic process until d ' 0. Requirement (i) raises the
question about whether the micromotion has any effect on the cooling of a trapped
ion to its ground state. The laser cooling process of trapped ions when the intrinsic
micromotion is taken into account is discussed in Refs. [42, 80], and the influence of
the excess micromotion is studied in Ref. [36]. Even though the micromotion may
result in unwanted heating of the ion, cooling can still be achieved by choosing the
6.4. SUMMARY 105
appropriate value for the laser frequency. Provided that the amount of the excess
micromotion is not too large, the micromotion is not a fundamental obstacle for
the task of cooling a trapped ion to its ground Floquet state.
6.4 Summary
In a 3D system where the linear Paul trap is used for the ion, the micromotion only
exist in the radial directions. Although the structure of a 3D system is far more
complicated, the qualitative behavior of the micromotion effect, as we described for
1D systems in chapter 5, is expected to carry over to 3D systems. In particular we
predict the existence of the micromotion characteristic distance dmm beyond which
the harmonic approximation no longer works. With a little thought one would see
that the scheme we propose to bypass the micromotion effect in chapter 5 is also
applicable to 3D systems. The study of a quasi-1D system, together with previous
results, shows that certain experimental configurations can be implemented to
minimize the impact of the micromotion. In general the micromotion effect is
significantly reduced if (i) the micromotion frequency is much larger than the
radial secular frequency of the ion, (ii) the radial frequencies are much larger than
the axial frequencies for both traps, and (iii) the adiabatic process must be carried
out along the axial direction.
We would like to point out that our analysis of the trapped atom-ion system
relies on two assumptions. First, the possible charge-exchange collisions between
the atom and the ion, which cause a loss of the ion from the ion trap, must be
negligible. This may be satisfied by choosing appropriate chemical elements for the
atom and ion. This helps suppress the charge-exchange collisions [29] so for the
elastic collisions to dominate the process. The second assumption is that the height
of the potential well of the atom trap is sufficiently large, so the atom does not
106 CHAPTER 6. MICROMOTION EFFECT IN THREE DIMENSIONS
oscillate out of the confinement region during its interaction with the ion. There
are experimental situations in which this condition does not hold and the atom
is lost as a result. An improved calculation, which takes into account the finite
height of the atom trap’s potential well, is important to understand the quantum
dynamics of trapped atom-ion collisions and thus requires further studies.
Chapter 7
Conclusions
In this work we demonstrate that, while the micromotion can be neglected for
single trapped ions, it must be taken into account when the ion strongly interacts
with an external system. In the context of trapped atom-ion interaction, the
effect of the micromotion is most important when there is resonance between two
eigenstates of the approximate, or unperturbed, Hamiltonian. This happens when
the energy separation is close to the micromotion frequency and the coupling that
arises from the micromotion between the two states is significant. The micromotion
leads to not only a shift in energy but also numerous new avoided crossings in the
energy diagram which is a quantum signature of strong chaotic behaviour. The
magnitude of these effects can be inferred by (i) a calculation of the micromotion-
induced coupling strengths and (ii) an exact computation of the quasienergies of the
system. Although the second method gives much more detail, the first method is
much faster yet sufficiently accurate to predict the main features of the quasienergy
diagram. When exact quasienergies are required, the Floquet Hamiltonian method
and the time-propagator method can be used. For the particular problem of the
micromotion effect in trapped atom-ion systems, the time-propagator method is
more accurate.
107
108 CHAPTER 7. CONCLUSIONS
Our results show that the approximation that the ion trap is a harmonic
trap breaks down at trap distances smaller than a characteristic value dmm. The
quasienergy diagram in this region is far more complicated than the approximate
eigenenergy diagram, and thus the behavior of the system in an adiabatic process
may change completely when the micromotion is included. This complication due
to the presence of the micromotion may seriously reduce the fidelity of the atom-
ion quantum gate proposed in Ref. [14]. However, the effect of the micromotion
on the realization of the quantum gate is weaker when the ion’s secular frequency
is much larger than the atom’s trapping frequency. If a linear Paul trap is used
for the ion, the ideal configuration is the one for which the micromotion frequency
is as large as possible compared with the radial trapping frequency of the Paul
trap, and for both traps the radial frequencies should be much larger than the
axial trapping frequencies; in addition, the traps must be moved along the axial
direction in the adiabatic collision process. These conditions for the minimization
of the micromotion effect are expressed mathematically in Sec. 6.3.
For situations in which the micromotion efffect is unavoidable and short trap
distances, we suggest a scheme for overcoming this difficulty by utilizing a transi-
tion from the initial ground Floquet state to a low-lying molecular state before the
onset of the micromotion effect. With the help of this scheme, the quantum gate
can be safely implemented by the following modified adiabatic process: (i) The
atom and the ion are initially prepared in their ground states at a very large trap
distance, (ii) the traps are moved toward each other until the trap distance reaches
a value d1 which must be larger than the micromotion characteristic distance dmm,
(iii) an additional electromagnetic field is used to make a transition of the system
from the ground level to a low-lying molecular level, and (iv) the adiabatic process
is now continued until the trap distance reaches a value small enough to achieve the
required phase difference for the quantum gate. This modified adiabatic process
109
can also be applied to bound atoms to a trapped ion for the creation of an atom-ion
macromolecule. Such an object may indeed be of interest in its own right.
Although this work has a strong emphasis on the specific application of the
atom-ion quantum gate, its main contribution is the methods for predicting the
micromotion effect and computing the energies and wave functions of the trapped
atom-ion system when the micromotion is included. By comparing the full results
with those obtained using the harmonic approximation, one should also be able to
figure out the validity of other proposals in atom-ion interaction.
110 CHAPTER 7. CONCLUSIONS
Appendix A
Numerov method
The Numerov method is very efficient for finding the solutions of linear second-
order ordinary differential equations (ODE) with no first-order derivative term.
The one-dimensional Schrodinger equation
[d2
dx2+
2m
~2[E − V (x)]
]Ψ(x) = 0 (A.1)
belongs to this class of ODEs. Bound state wave functions can be obtained as
follows: One first chooses two positions xmin, xmax at the boundary where the wave
function vanishes, that is,
Ψ (xmin) = Ψ (xmax) = 0. (A.2)
The interval [xmin, xmax] is then divided into N + 1 grid points xn, n = 0, 1, . . . , N .
Let Ψn = Ψ(xn) and
Tn =h2
6
m
~2[V (xn)− E] (A.3)
111
112 APPENDIX A. NUMEROV METHOD
with the step size h not to be confused with the Planck constant, the Numerov
formula is the three-term recurrence relation
(1− Tn+1)Ψn+1 − (2 + 10Tn)Ψn + (1− Tn−1)Ψn−1 = 0 (A.4)
with a truncation error of order O(h6) [65]. This equation can be written as
Fn+1 − UnFn + Fn−1 = 0, (A.5)
where
Fn = (1− Tn)Ψn (A.6)
and
Un =2 + 10Tn1− Tn
. (A.7)
Now let us define the ratio
Rn =Fn+1
Fn(A.8)
and divide all terms in Eq. (A.5) by Fn to arrive at
Rn = Un −1
Rn−1
(A.9)
which is the forward iteration formula. Since Ψ0 = 0, we have F0 = 0 and hence
R0 = ∞. The above iteration can then be used to obtain the ratio Rn at all
grid points. However, the forward iteration alone is not useful in practice since
in general the wave function does not converge to the correct boundary condition
ΨN = 0. This can be overcome with the help of the reverse iteration formula. Let
Rn =Fn−1
Fn=
1
Rn−1
, (A.10)
113
the forward iteration formula can be rearranged to yield
Rn = Un −1
Rn+1
. (A.11)
As ΨN = 0, we have FN = 0 and RN =∞. The reverse iteration can now be used
to compute the ratio Rn at all grid points.
In practice, the eigenenergies and eigenfunctions are computed by the following
procedure. One starts with a trial energy Et and first iterate in the reverse direction
using Eq.(A.11) from xN to the position xM where the wave function reaches
its first local maximum. Since ΨM is a local maximum, we have approximately
RM = 1, thus the iteration is stopped when RM ≤ 1 (in the interval [xN , xM ], the
ratio R decreases from ∞ to 1). Next, we iterate in the forward direction using
Eq. (A.9) from x0 to xM and compute all the ratios Rn in that interval. If the trial
energy Et is equal to an exact eigenenergy E, Eq. (A.10) shows that
1
RM+1
−RM = 0, (A.12)
otherwise we observe a gap
D(Et) =1
RM+1
−RM (A.13)
which is a function of the trial energy Et. When Et is in a close vicinity of E we
have
D(Et) > 0 if Et > E,
D(Et) < 0 if Et < E,(A.14)
that is, the slope of D(Et) is positive at the eigenenergy E [64]. This can be
understood by observing that in a small interval including xM and no zeros of the
wave function, the quantity U , and hence the ratio R decreases when the trial
114 APPENDIX A. NUMEROV METHOD
energy Et increases. As a result the gap D(Et) increases with increasing Et.
In order to obtain all the eigenenergies in a desired energy range [Emin, Emax] we
first scan through the region with the energy step Estep. The trial energies are then
Et,n = Emin + nEstep, n = 0, 1, . . . The energy step must be small enough so that
there is no more than one eigenenergy in any interval [Et,n, Et,n+1]. We calculate
all the gaps D(Et,n), and if D(Et,n) < 0 while D(Et,n+1) > 0 there must be an
eigenenergy En in the interval [Et,n, Et,n+1]. After all the intervals that contain
an eigenenergy are recorded, we use a bisection procedure with the gap D(Et) as
an indicator to rapidly converge the upper and lower energies to the eigenenergy.
With Eu and El denoting the upper and lower energies of the interval, the code
below gives the eigenenergy with the accuracy ε.
while Eu − El > ε
Et =Eu + El
2,
compute D(Et),
if D(Et) > 0
Eu = Et,
else
El = Et,
end.
Now that the eigenenergies are known, the eigenfunctions can be evaluated.
First we use the reverse and forward iteration to obtain all the ratios R and R.
Recall that both iterations are stopped at xM where RM = R−1M+1 = 1. Since each
eigenfunction may contain an arbitrary multiplicative constant, it can be chosen
such that the function F defined in Eq. (A.6) satisfies FM = 1. Since the ratios
115
R and R are known, the values Fn at all grid points can be obtained from FM by
an iteration based on Eqs. (A.8) for n ≤ M and (A.10) for n > M . Finally, the
eigenfunctions at all the grid points are computed from Eq. (A.6)
Ψn =Fn
1− Tn, (A.15)
and numerical integration is then used for normalization. The accumulated error
in the eigenfunction Ψn is shown to be of order O(h4) [66].
116 APPENDIX A. NUMEROV METHOD
Appendix B
Floquet states of a trapped ion
Here we consider a single ion confined in a rf trap with both the intrinsic micro-
motion and the excess micromotion. The Floquet state of an ion with only the
intrinsic micromotion is given by Glauber in Ref. [40]. Here we apply his elegant
proof to derive the Floquet state of a trapped ion in the presence of the excess
micromotion. The potential for the motion along the x axis is
Vx(t) =1
8miω
2 [a+ 2q cosωt]X2 − e [Edc + Eac sin(ωt)]X, (B.1)
and the equation of motion is an inhomogeneous differential equation:
X(t) + [a+ 2q cos(ωt)]ω2
4X(t) = F0 + F1 sin(ωt), (B.2)
where F0 = eEdc/mi, and F1 = eEac/mi.
The homogeneous equation is the Mathieu equation which possesses a special
Floquet solution f(t) satisfying the initial condition
f(0) = 1, f(0) = iν. (B.3)
117
118 APPENDIX B. FLOQUET STATES OF A TRAPPED ION
This special solution has the expression
f(t) = eiµtϕ(t), (B.4)
where µ is called the Floquet exponent which depends on a, q, ω and ϕ(t) is a
periodic function whose Fourier series is
ϕ(t) =∑n
Cneinωt. (B.5)
The coefficients Cn and the Floquet critical exponent µ can be found by simple
numerical procedures [84]. The value of ν can be then computed from the relation
ν = µ+ ω∑n
nCn. (B.6)
When |a|, |q| 1, the lowest-order approximation in a and q gives µ ≈ ω0, with
ω0 defined in Eq. (2.8), and [42]
ϕ(t) ≈ 1 + (q/2) cos(ωt)
1 + q/2. (B.7)
If f(t) is the solution, f(t)∗ must also be a solution, and a general homogeneous
solution has the form
xh = Af(t) +Bf(t)∗. (B.8)
When |a|, |q| 1 we have µ ω so the Floquet exponent cannot be an integer
multiple of ω. This means that f(t) and f(t)∗ are not periodic functions and hence
no homogeneous solution is periodic.
We now show that the inhomogeneous equation (B.2) has one and only one
periodic solution when |a|, |q| 1. It is easy to see that there is not more than
one periodic solution: Assume xp(t) is a particular periodic solution; any other
119
solution x′p(t) is the combination of xp(t) and a homogeneous solution, that is,
x′p(t) = xp(t) + Af(t) +Bf(t)∗. (B.9)
Since f(t) and f(t)∗ are not periodic, it is impossible for x′p(t) to be periodic
unless A = B = 0, which results in x′p(t) = xp(t). Thus, we conclude that the
inhomogeneous equation has no more than one periodic solution.
We prove the existence of the unique periodic solution by explicitly constructing
it by the variation-of-parameters method [83]. We look for a periodic solution of
the form
xp(t) = k(t)f(t) + k(t)∗f(t)∗ (B.10)
subject to the condition
k(t)f(t) + k(t)∗f(t)∗ = 0. (B.11)
By making use of the time-independent Wronskian,
f(t)f(t)∗ − f(t)∗f(t) = 2iν, (B.12)
we obtain xp(t) = ReΦ(t)ϕ(t)∗, where ϕ(t) is the function given in Eq. (B.5)
and
Φ(t) =∞∑
n=−∞
Dneinωt, (B.13)
with
Dn =2CnF0 + iF1(Cn+1 − Cn−1)
2ν(µ+ nω). (B.14)
In the lowest-order approximation we have
xp(t) ≈ δd[1 +
q
2cos(ωt)
]− ω0
ωlac sin(ωt), (B.15)
120 APPENDIX B. FLOQUET STATES OF A TRAPPED ION
where δd and lac are defined in Sec. 5.7 and ω0 is defined in Eq. (2.8). It is
obvious that xp(t+ T ) = xp(t) and so we have found the unique periodic solution
of Eq. (B.2).
Glauber’s elegant approach can be used to find the quasienergies and Floquet
states of the trapped ion. First we introduce a new position operator that depends
explicitly on time,
X1(t) = X(t)− xp(t), (B.16)
so that X1(t) satisfies the homogeneous differential equation. The Wronskian
f(t)X1(t)− f(t)X1(t) (B.17)
is constant in time and thus the operator
A = i
√mi
2~ν
[f(t)X1(t)− f(t)X1(t)
](B.18)
is also a constant of motion, and it possesses constant eigenkets. The operator A
and its adjoint are the analogs of the ladder operators for the quantum harmonic
oscillator, and one can show that [A,A†] = 1. Let us define the states
A |0〉 = 0,
|n〉 =
(A†)n
√n!|0〉 , (B.19)
and demonstrate that they are indeed the Floquet states of the trapped ion.
Although the eigenket |0〉 is time independent, its representing wave function
〈x, t|0〉 is not. We have 〈x, t|A |0〉 = 0, which yields
~imi
f(t)∂
∂x− f(t) [x− xp(t)]− xp(t)f(t)
〈x, t|0〉 = 0, (B.20)
121
The solution in its normalized form is
〈x, t|0〉 =(miν
π~
)1/4 eis(x,t)
[f(t)]1/2exp
imi
2~f(t)
f(t)[x− xp(t)]2
, (B.21)
with s(x, t) = mixp(t)x/~. By applying the operator A† repeatedly to the state |0〉
we obtain
〈x, t|n〉 =1√2nn!
(miν
π~
)1/4 eis(x,t)
[f(t)]1/2
[f(t)∗
f(t)
]n/2Hn
([miν
~|f(t)|2
]1/2
[x− xp(t)]
)
× exp
imi
2~f(t)
f(t)[x− xp(t)]2
, (B.22)
where Hn is the Hermite polynomial of order n. Upon inserting f(t) from Eq (B.4)
to the above expression, we get
〈x, t|n〉 = exp
[−i(n+
1
2
)µt
]un(t), (B.23)
with the periodic functions
un(t) =1√2nn!
(miν
π~
)1/4 eis(x,t)
[ϕ(t)]1/2
[ϕ(t)∗
ϕ(t)
]n/2Hn
([miν
~|f(t)|2
]1/2
[x− xp(t)]
)
× exp
−miµ
2~
[1− iϕ(t)
µϕ(t)
][x− xp(t)]2
. (B.24)
The structures of the wave functions 〈x, t|n〉 suggest they are the Floquet states
with quasienergies
εn =
(n+
1
2
)~µ. (B.25)
Since µ is independent of Edc and Eac, the quasienergies of a single trapped ion
are indeed not affected by the excess micromotion.
In situations where the excess micromotion vanishes, we have xp(t) = 0 and the
wave functions of the Floquet states |n〉 reduce to the forms obtained by Glauber
122 APPENDIX B. FLOQUET STATES OF A TRAPPED ION
in Ref. [40] as expected. A large amount of the excess micromotion, which results
in a large amplitude of xp(t), will lead to strong oscillations in the wave functions
un(t).
Let us now work out the kinetic energy of the trapped ion when it is prepared
in the state |n〉. Using Eqs. (B.16) and (B.18), we have
P (t) = miX(t) =
√~mi
2ν
(f(t)∗A+ f(t)A†
)+mixp(t), (B.26)
and the kinetic energy associated with the state |n〉 is
〈n|P (t)2 |n〉2mi
=1
2
(n+
1
2
)~|f(t)|2
ν+mi[xp(t)]
2
2. (B.27)
The mean kinetic energy is defined as the time-averaged value 〈P (t)2〉/2mi taken
over one period 2π/ω of the micromotion. After inserting f(t) from Eqs. (B.4) and
(B.5) to the above equation and taking the time average we arrive at
〈n|P (t)2 |n〉2mi
=1
2
(n+
1
2
)~ν
∞∑n=−∞
C2n(µ+ nω)2 +
mixp(t)2
2, (B.28)
and when |a| q2 1 we have
〈n|P (t)2 |n〉2mi
≈(n+
1
2
)~ω0 +
1
2miω
20
[(δd)2 + l2ac/2
]. (B.29)
For the ground state we have
〈P (t)2〉2mi
≈ ~ω0
2+
1
2miω
20
[(δd)2 + l2ac/2
]. (B.30)
The cooling of a trapped ion to its ground Floquet state and the final value of its
mean kinetic energy are studied in Ref. [80] (without the excess micromotion). We
see that the excess micromotion leads to an increase in the mean kinetic energy
123
of the ground state, but this does not mean a worse cooling limit if we are only
interested in the population of the ground state.
124 APPENDIX B. FLOQUET STATES OF A TRAPPED ION
Appendix C
Floquet adiabatic process
Before we discuss the adiabatic quantum phase gate in the Floquet picture, we
need to understand how a quantum system evolves in a Floquet adiabatic process
[56]. We start with the Schrodinger equation
i~∂
∂tΨ(t) = H
(λ(t), t
)Ψ(t), (C.1)
where the parameter λ, which is the trap distance in the specific case of the trapped
atom-ion system, is varied slowly in time. At each moment we have the instanta-
neous Floquet states which satisfy the Floquet eigenvalue equation
HF (λ, t)un,k(λ, t) = εn,k(λ)un,k(λ, t). (C.2)
To study the adiabatic evolution, one needs to employ a trick called the (t, t′)
method [81] in which we change the implicit time parameter t in λ(t) to a new
time variable t′ and introduce a higher-dimensional wave function Ψ(t, t′) with the
requirement that this wave function satisfies the “extended” Schrodinger equation
i~∂
∂t′Ψ(t, t′) = HF
(λ(t′), t
)Ψ(t, t′). (C.3)
125
126 APPENDIX C. FLOQUET ADIABATIC PROCESS
When the initial condition of Ψ(t = 0) is transformed to that of Ψ(t, t′ = 0) in an
appropriate way [56], we have
Ψ(t) = Ψ(t, t′ = t). (C.4)
An advantage of using the extended wave function Ψ(t, t′) is the similarity of
Eq. (C.3) with the Schrodinger equation encountered in the normal adiabatic pro-
cess for which the Hamiltonian only depends on time implicitly through the pa-
rameter λ, and thus standard perturbation techniques can be used. In this picture
the variable t is treated as a fourth dimension of space and hence can be dropped
from now on.
We expresses the wave function Ψ(t′) in terms of the instantaneous Floquet
states,
Ψ(t′) =∑n,k
cn,k(t′)un,k
(λ(t′)
)× exp
[− i~
∫ t′
0
dτ εn,k(λ(τ)
)], (C.5)
and insert it into Eq. (C.3) to obtain the differential equations for the coefficients
cn,k(t′) which in turn can be solved approximately by perturbation methods. If
the system starts in a single Floquet state, say u0,0, then first-order perturbation
theory gives us
cn,k(t′)=−
∫ t′
0
dτ∂λ(τ)
∂τ
⟨⟨un,k
(λ(τ)
) ∣∣∣∣ ∂∂λ∣∣∣∣u0,0
(λ(τ)
)⟩⟩×exp
i
~
∫ τ
0
dτ ′[εn,k(λ(τ ′)
)− ε0,0
(λ(τ ′)
)]. (C.6)
Recall that all the Floquet states un,k correspond to a unique physical state un,
to find the transition amplitude to the physical state un at the real time t we need
127
to sum over all values of k and then replace t′ with t
cn(t) =∞∑
k=−∞
cn,k(t′ = t). (C.7)
For the system to almost stay in a single Floquet state at any moment in time, we
need |cn(t)| 1 for all n, which is the general requirement for a Floquet adiabatic
process.
Let us now consider the Floquet adiabatic process used to implement the quan-
tum phase gate as considered in Sec. 5.4. The adiabatic theorem states that if the
initial motional state is a Floquet state un(d(ti), ti) of the system, the final motional
state at the end of the process is
φ(tf ) = un(d(tf ), tf
)exp
[− i~
∫ tf
ti
dτ εn(d(τ)
)], (C.8)
where it is understood that the quasienergy of the ground state is set to zero
at all trap distances. Since the instantaneous quasienergy εn(d) depends on the
interaction potential which in turn depends on the spin of the atom and the ion, we
know that the phase accumulated in an adiabatic process must be spin dependent
[14].
In the spin subspace, one can create, by single qubit gates, a separable state
|χ(ti)〉 =∑k
|k〉 ck, (C.9)
with k = 00, 01, 10, 11, and c1c4 = c2c3. Again “spin” means the hyperfine levels of
the atom and the ion which are chosen as the qubit states. The Floquet adiabatic
process turns the initial state,
|Ψ(ti)〉 = u0(d(ti), ti)∑k
|k〉 ck, (C.10)
128 APPENDIX C. FLOQUET ADIABATIC PROCESS
to
|Ψ(tf )〉 = u0(d(tf ), tf )∑k
|k〉 ckeiθk . (C.11)
Each of the spin terms |k〉 accumulates a different phase θk since the phase in an
adiabatic process is spin dependent. The final spin state is |χ(tf )〉 =∑
k |k〉 ckeiθk ,
and the adiabatic process is equivalent to the application of a quantum phase gate
U = diageiθk in the spin subspace.
In a realistic situation there must be a small fraction of the motional state
being transferred to excited Floquet levels at the end of the adiabatic process. We
consider the simplified situation when only one dominant excited Floquet state ue
is occupied with a small probability pe. If we first ignore the spin degree of freedom
and write the amplitude of the ground state and the excited state in their polar
forms, the final motional state, apart from a global phase, is
φ(tf ) =√p0u0
(d(tf ), tf
)eiθ +
√peue
(d(tf ), tf
)eiθ
′, (C.12)
with p0 + pe = 1 and pe 1.
For the initial state given in Eq. (C.10), because of the spin-dependence of the
phases θ and θ′, the final total wave function is more complicated,
|Ψ(tf )〉 =∑k
|ϕk〉 |k〉 ckeiθk , (C.13)
where
|ϕk〉 =√p0u0
(d(tf ), tf
)+√peue
(d(tf ), tf
)eiαk , (C.14)
with αk = θ′k − θk. Note that in the above we have ignored a very small spin-
dependence in the amplitude√p0 and
√pe.
A comparison of the realistic final wave function |Ψ(tf )〉 and the ideal wave
function |Ψ(tf )〉 shows that when branching occurs the motional state is entangled
129
to the spin state, and the adiabatic process no longer results in a pure state in the
spin subspace. The density matrix for the mixed state of the spin subspace can
be obtained by taking the partial trace of the total density matrix |Ψ(tf )〉 〈Ψ(tf )|;
the result is
ρ(tf ) =∑k,j
|k〉 ckλkjei(θk−θj)c∗j 〈j| , (C.15)
where
λkj = 〈ϕj|ϕk〉 = 1− pe[1− ei(αk−αj)
]. (C.16)
The gate fidelity is defined as [82]
F = min
√〈χ(tf )| ρ(tf ) |χ(tf )〉
, (C.17)
where the minimization is done over all possible values of ck satisfying the con-
straint∑
k |ck|2 = 1 .
Now we introduce the 4× 4 symmetric matrix M whose elements are
Mkj = 1− cos(αk − αj), (C.18)
and the column V with the components Vj = |cj|2. It follows that
〈χ(tf )| ρ(tf ) |χ(tf )〉 = 1− peV TMV. (C.19)
With AT denoting the constant row (1, 1, 1, 1), the constraint can be written as
V TA = 1. (C.20)
To evaluate the maximum value of V TMV we consider the variation
δ(V TMV
)= δV TMV + V TMδV = 0
δV TMδV = 0, (C.21)
130 APPENDIX C. FLOQUET ADIABATIC PROCESS
where the last step follows from the fact that δV TMV = V TMδV as M is sym-
metric. The constraint of Eq. (C.20) requires that
δV TA = 0. (C.22)
Thus we must have
MδV = κA, (C.23)
with κ a constant. Upon inserting V = κM−1A to the constraint of Eq. (C.20) we
obtain κ =(ATM−1A
)−1, and the extremal value of V TMV , which can be verified
easily to be the maximum value, is
maxV TMV
=(κATM−1
)M(κM−1A
)=(ATM−1A
)−1. (C.24)
Therefore, the gate fidelity is
F =
√1− pe (ATM−1A)−1 ≈ 1− pe
2
(ATM−1A
)−1. (C.25)
There is no simple analytic form for ATM−1A; however, using the fact thatMkj = 0
for k = j and Mkj ≤ 2 for k 6= j, one can show that V TMV ≤ 32
and derive a
lower bound for the gate fidelity
F ≥√
1− 3
2pe ≈ 1− 3
4pe. (C.26)
In short, we see that 1 − F ∝ pe. Thus, the gate fidelity decreases linearly with
the occupancy of the excited Floquet state.
Appendix D
Derivation of Eq. (6.26)
We are now in the position to prove Eq. (6.26), which is central to the description of
the propagation of the micromotion from the radial directions to the axial direction
in a quasi-1D system. In a linear Paul trap, the classical equation of motion along
the x axis is the homogeneous Mathieu equation
x(t) + [a+ 2q cos(ωt)]ω2
4x(t) = 0. (D.1)
We first need to obtain the periodic function ϕ(t) of the special Floquet solution,
which is introduced in Eq. (B.4), up to the second order in q. In the following
steps we consider the typical experimental situation when |a| q2 and |q| 1; so
the parameter a can be treated as O(q3). The notation O is used to indicate the
terms that are time independent, that is, they may depend on a and q but not t.
Let us define the constant β from the Floquet critical exponent µ discussed in
Appendix B by the relation
µ = βω
2. (D.2)
131
132 APPENDIX D. DERIVATION OF EQ. (6.26)
It is known that [42], to the lowest order in a and q,
β ≈√a+
q2
2. (D.3)
Another important result is that the constant β does not depend on the sign of q
[45], that is,
β(a, q) = β(a,−q). (D.4)
The coefficients Cn in the Fourier expansion of the periodic function ϕ(t), as
introduced in Eq. (B.5), satisfy the recurrence relation [45]
Cn+1 −DnCn + Cn−1 = 0,
(D.5)
with
Dn = [(2n+ β)2 − a]/q. (D.6)
For later purposes we rewrite this relation as
Cn+1
Cn−Dn +
Cn−1
Cn= 0. (D.7)
Since we are only interested in terms up to the second order in q, we may set
C±3 = 0 because these coefficients are of the order O(q3). Using Eq. (D.7) we have,
for n = 2,
C1
C2
= D2, (D.8)
and n = 1,
C0
C1
−D1 +C2
C1
= 0, (D.9)
133
which can be solved quite easily to obtain
C1 = C0D2
D1D2 − 1, (D.10)
and
C2 = C01
D1D2 − 1. (D.11)
Similarly, by letting n = −2 and then n = −1 in the recurrence relation, one will
be able to obtain
C−1 = C0D−2
D−1D−2 − 1, (D.12)
and
C−2 = C01
D−1D−2 − 1. (D.13)
In view of the fact that β is of the first order in q, we have
D2
D1D2 − 1=
q[(β + 4)2 − a]
[(β + 4)2 − a][(β + 2)2 − a]− q2
=q
(β + 2)2+ O(q3)
=q
4(1− β) + O(q3), (D.14)
and
1
D1D2 − 1=
q2
[(β + 4)2 − a][(β + 2)2 − a]− q2
=q2
4222+ O(q3)
=q2
64+ O(q3). (D.15)
Thus
C1 =q
4(1− β)C0 + O(q3), (D.16)
134 APPENDIX D. DERIVATION OF EQ. (6.26)
and
C2 =q2
64C0 + O(q3). (D.17)
In essentially the same way we obtain
C−1 =q
4(1 + β)C0 + O(q3), (D.18)
and
C−2 =q2
64C0 + O(q3). (D.19)
The condition f(0) = 1 leads to
∑n
Cn = 0, (D.20)
which requires
C0
[1 +
q
2+q2
32+ O(q3)
]= 1. (D.21)
Hence we arrive at
C0 =1
1 + q/2 + q2/32+ O(q3). (D.22)
From the results just found for the coefficients Cn we have
ϕ(t) =1 + (q/2) [cos(ωt)− iqβ sin(ωt)] + (q2/32) cos(2ωt)
1 + q/2 + q2/32+O(q3). (D.23)
Notice that the terms indicated by the notation O, unlike those represented by O,
can be time dependent. For later calculations we compute
|ϕ(t)|2 =1 + (q2/8) + q cos(ωt) + (3q2/16) cos(2ωt)
1 + q + 5q2/16+O(q3). (D.24)
Having obtained the periodic function ϕ(t) of the special Floquet solution up
to the second order in q, we may now calculate the matrix element 〈u0x|x2i |u0x〉
135
of Eq. (6.26). By switching to the Heisenberg picture we may write this matrix
element as 〈0|xi(t)2 |0〉, where the state |0〉 is defined in Eq. (B.19). In the absence
of the excess micromotion, we may set xp(t) = 0 in Eq. (B.16). Using this equation
and Eq. (B.18) we have
X(t) =
√~
2νmi
[f(t)∗A+ f(t)A†
]. (D.25)
From the well-known relations for the ladder operators
A |n〉 =√n |n− 1〉 ,
A† |n〉 =√n+ 1 |n+ 1〉 , (D.26)
one can show that
〈0|xi(t)2 |0〉 =~
2miν|f(t)|2 =
~2miν
|ϕ(t)|2, (D.27)
and using the result of Eq. (D.24) one would have
〈0|xi(t)2 |0〉 =~
2miν
1 + q2/8 + q cos(ωt) + (3q2/16) cos(2ωt)
1 + q + 5q2/16+O(q3). (D.28)
It can be seen from Eq. (B.6) and Eq. (D.2) that
ν =ω
2
[β +
∑n
2nCn
]. (D.29)
Upon inserting the values found for the coefficients Cn we get
ν =βω
2
[1− C0q + O(q2)
]= µ
[1− q + O(q2)
]. (D.30)
136 APPENDIX D. DERIVATION OF EQ. (6.26)
The results of Eq. (D.28) and Eq. (D.30) yield
〈0|xi(t)2 |0〉 =~
2miµ
1 + q2/8 + q cos(ωt) + (3q2/16) cos(2ωt)
[1 + q + 5q2/16][1− q + O(q2)
] +O(q3)
=~
2miµ
1 + q2/8 + q cos(ωt) + (3q2/16) cos(2ωt)[1 + O(q2)
] +O(q3)
=~
2miµ
[1 + O(q2) + q cos(ωt) +
3
16q2 cos(2ωt)
]+O(q3). (D.31)
Notice that in the above expression we keep the exact value for the Floquet critical
exponent µ, which is itself a function of the trap parameters a and q. For the
calculation of the similar matrix element for the motion along the y direction, it is
necessary to express the dependence on the trap parameters explicitly; hence we
will write µ as µ(a, q) from this point onward. The matrix element 〈0|xi(t)2 |0〉
should also be written as 〈0(a, q)|xi(t)2 |0(a, q)〉 to reflect the dependence of the
state |0〉 on the traps parameter. So, we rewrite Eq. (D.31) as
〈0(a, q)|xi(t)2 |0(a, q)〉 =~
2miµ(a, q)
×[1 + O(q2) + q cos(ωt) +
3
16q2 cos(2ωt)
]+O(q3).
(D.32)
Although the time-independent constants in O(q2) are not determined in our cal-
culation, it will be clear later that these constants contribute to a small time-
independent perturbation that can be ignored. Only the oscillating terms need to
be considered since they result in the micromotion effect.
Recall that for a linear Paul trap we have qx = −qy, and ax = ay. The motion
along the y direction of an ion in a linear Paul trap is also described by the Mathieu
equation for the motion along the x direction, but with the parameter q replaced
with −q. Therefore, by a simple change of the sign of q in Eq. (D.32), we can derive
137
the corresponding matrix element for the motion along the y direction, which is
〈0(a,−q)| yi(t)2 |0(a,−q)〉 =~
2miµ(a,−q)
×[1 + O(q2)− q cos(ωt) +
3
16q2 cos(2ωt)
]+O(q3).
(D.33)
As mentioned in Eq. (D.4), the constant β(a, q), and hence the Floquet critical
exponent µ(a, q), does not depend on the sign of q, that is,
µ(a, q) = µ(a,−q). (D.34)
Upon replacing µ(a,−q) in Eq. (D.33) with µ(a, q) and adding the resulting equa-
tion to Eq. (D.32) we obtain
〈0(a, q)|xi(t)2 |0(a, q)〉+ 〈0(a,−q)| yi(t)2 |0(a,−q)〉
=~
miµ(a, q)
[1 + O(q2) +
3
16q2 cos(2ωt)
]+O(q3). (D.35)
As the time-independent terms in the bracket are dominated by the factor 1, the
terms in O(q2) are negligible. Furthermore, since µ(a, q) ≈ ω0 with ω0 the secular
frequency of the motion along the radial direction, we finally arrive at
〈0(a, q)|xi(t)2 |0(a, q)〉+ 〈0(a,−q)| yi(t)2 |0(a,−q)〉 ≈ l2i
[1 +
3
16q2 cos(2ωt)
].
(D.36)
which is indeed Eq. (6.26) written in the Heisenberg picture.
138 APPENDIX D. DERIVATION OF EQ. (6.26)
Bibliography
[1] C.H. Bennett and D.P. DiVincenzo, Quantum information and computation,
Nature 404, 247 (2000). doi:10.1038/35005001.
[2] D.P. DiVicenzo, The physical implementation of quantum computation, in
Scalable Quantum Computers: Paving the Way to Realization, edited by
S.L. Braunstein, H.K. Lo and P. Kok (Wiley-VCH, 2001), p. 1. arXiv:quant-
ph/0002077.
[3] T. D. Ladd, F. Jelezko, R. Laflamme, Y. Nakamura, C. Mon-
roe, and J. L. OBrien, Quantum computers, Nature 464, 45 (2010).
doi:10.1038/nature08812.
[4] A. Steane, The Ion Trap Quantum Information Processor, Appl. Phys. B64,
623 (1997). doi:10.1007/s003400050225.
[5] J.I. Cirac and P. Zoller, Quantum computations with cold trapped ions, Phys.
Rev. Lett. 74, 4091 (1995). doi:10.1103/PhysRevLett.74.4091.
[6] H. Haffner, C.F. Roos, and R. Blatt, Quantum computing with trapped ions,
Phys. Rep. 469, 155 (2008). doi:10.1016/j.physrep.2008.09.003.
[7] G.K. Brennen, C.M. Caves, P.S. Jessen, and I.H. Deutsch, Quan-
tum logic gates in optical lattices, Phys. Rev. Lett. 82, 1060 (1999).
doi:10.1103/PhysRevLett.82.1060.
139
140 BIBLIOGRAPHY
[8] E.L. Raab, A. Cable, S. Chu, and D.E. Pritchard, Trapping of neutral
Sodium atoms with radiation pressure, Phys. Rev. Lett. 59, 2631 (1987).
doi:10.1103/PhysRevLett.59.2631.
[9] I. Bloch, Ultracold quantum gases in optical lattices, Nature Physics 1, 23
(2005). doi:10.1038/nphys138.
[10] W. Paul, Electromagnetic traps for charged and neutral particles, Rev. Mod.
Phys. 62, 531 (1990). doi:10.1103/RevModPhys.62.531.
[11] L.S. Brown and G. Gabrielse, Geonium theory: Physics of a single
electron or ion in a Penning trap, Rev. Mod. Phys. 58, 233 (1987).
doi:10.1103/RevModPhys.58.233.
[12] H.J. Metcalf and P.v.d Straten, Laser Cooling and Trapping (Springer 1999).
[13] Z. Idziaszek, T. Calarco, and P. Zoller, Controlled collisions of a single atom
and an ion guided by movable trapping potentials, Phys. Rev. A76, 033409
(2007). doi:10.1103/PhysRevA.76.033409.
[14] H. Doerk, Z. Idziaszek, and T. Calarco, Atom-ion quantum gate, Phys. Rev.
A81, 012708 (2010). doi:10.1103/PhysRevA.81.012708.
[15] O. Mandel, M. Greiner, A. Widera, T. Rom, T.W. Hansch and I. Bloch, Con-
trolled collisions for multi-particle entanglement of optically trapped atoms,
Nature 425, 937 (2003). doi:10.1038/nature02008.
[16] D. Jaksch, H.-J. Briegel, J. I. Cirac, C. W. Gardiner, and P. Zoller, Entangle-
ment of atoms via cold controlled collisions, Phys. Rev. Lett. 82, 1975 (1999).
doi:10.1103/PhysRevLett.82.1975.
[17] M.J. Bremner, C.M. Dawson, J.L. Dodd, A. Gilchrist, A.W. Harrow, D. Mor-
timer, M.A. Nielsen, and T.J. Osborne, Practical scheme for quantum compu-
BIBLIOGRAPHY 141
tation with any two-qubit entangling gate, Phys. Rev. Lett. 89, 247902 (2002).
doi:10.1103/PhysRevLett.89.247902.
[18] R. Cote, V. Kharchenko, and M.D. Lukin, Mesoscopic molecular ions
in Bose-Einstein condensates, Phys. Rev. Lett. 89, 093001 (2002).
doi:10.1103/PhysRevLett.89.093001.
[19] O.P. Makarov, R. Cote, H. Michels, and W.W. Smith, Radiative charge-
transfer lifetime of the excited state of NaCa+, Phys. Rev. A67, 042705
(2003). doi:10.1103/PhysRevA.67.042705.
[20] M. Krych, W. Skomorowski, F. Pawowski, R. Moszynski, and Z. Idzi-
aszek, Sympathetic cooling of the Ba+ ion by collisions with ultra-
cold Rb atoms: Theoretical prospects, Phys. Rev. A83, 032723 (2011).
doi:10.1103/PhysRevA.83.032723.
[21] W.W. Smith, O.P. Makarov, and J. Lin, Cold ion-neutral collisions in a hybrid
trap, J. Mod. Opt. 52, 2253 (2005). doi:10.1080/09500340500275850.
[22] Z. Idziaszek, T. Calarco, and P. Zoller, Ion-assisted ground-state cool-
ing of a trapped polar molecule, Phys. Rev. A83, 053413 (2011).
doi:10.1103/PhysRevA.83.053413.
[23] K. Mølhave and M. Drewsen, Formation of translationally cold MgH+
and MgD+ molecules in an ion trap, Phys. Rev. A62, 011401(R) (2000).
doi:10.1103/PhysRevA.62.011401.
[24] C. Kollath, M. Khl, and T. Ciamarchi, Scanning tunneling mi-
croscopy for ultracold atoms, Phys. Rev. A76, 063602 (2007).
doi:10.1103/PhysRevA.76.063602.
142 BIBLIOGRAPHY
[25] Y. Sherkunov, B. Muzykantskii, N. dAmbrumenil, and B. D. Simons, Probing
ultracold Fermi atoms with a single ion, Phys. Rev. A79, 023604 (2009).
doi:10.1103/PhysRevA.79.023604.
[26] R. Cote, From classical mobility to hopping conductivity: Charge
hopping in an ultracold gas, Phys. Rev. Lett. 85, 5316 (2000).
doi:10.1103/PhysRevLett.85.5316.
[27] A.T. Grier, M. Cetina, F. Orucevic, and V. Vuletic, Observation of cold col-
lisions between trapped ions and trapped atoms, Phys. Rev. Lett. 102, 223201
(2009). doi:10.1103/PhysRevLett.102.223201.
[28] M. Cetina, A. Grier, J. Campbell, I. Chuang, and V. Vuletic, Bright source
of cold ions for surface-electrode traps, Phys. Rev. A76, 041401(R) (2007).
doi:10.1103/PhysRevA.76.041401.
[29] C. Zipkes, S. Palzer, C. Sias, and M. Kohl, A trapped single ion inside a
Bose-Einstein condensate, Nature 464, 388 (2010). doi:10.1038/nature08865.
[30] C. Zipkes, S. Palzer, L. Ratschbacher, C. Sias, and M. Kohl, Cold
heteronuclear atom-ion collisions, Phys. Rev. Lett. 105, 133201 (2010).
doi:10.1103/PhysRevLett.105.133201.
[31] S. Schmid, A. Harter, and J.H. Denschlag, Dynamics of a cold trapped
ion in a Bose-Einstein condensate, Phys. Rev. Lett. 105, 133202 (2010).
doi:10.1103/PhysRevLett.105.133202.
[32] Z. Idziaszek, T. Calarco, P.S. Julienne, and A. Simoni, Quantum the-
ory of ultracold atom-ion collisions, Phys. Rev. A79, 010702 (2009).
doi:10.1103/PhysRevA.79.010702.
BIBLIOGRAPHY 143
[33] Z. Idziaszek, A. Simoni, T. Calarco, and P.S. Julienne, Multichannel quantum-
defect theory for ultracold atom-ion collisions, New J. Phys. 13, 083005 (2011).
doi:10.1088/1367-2630/13/8/083005.
[34] R. Cote and A. Dalgarno, Ultracold atom-ion collisions, Phys. Rev. A62,
012709 (2000). doi:10.1103/PhysRevA.62.012709.
[35] B. Gao, Universal properties in ultracold ion-atom interactions, Phys. Rev.
Lett. 104, 213201 (2010). doi:10.1103/PhysRevLett.104.213201.
[36] D.J. Berkeland, J.D. Miller, J.C. Bergquist, W.M. Itano, and D.J. Wineland,
Minimization of ion micromotion in a Paul trap, J. Appl. Phys. 83, 5025,
(1998). doi:10.1063/1.367318.
[37] R.G. DeVoe, Power-law distributions for a trapped ion interacting
with a classical buffer gas, Phys. Rev. Lett. 102, 063001 (2009).
doi:10.1103/PhysRevLett.102.063001.
[38] C. Zipkes, L. Ratschbacher, C. Sias, and M. Kohl, Kinetics of a single
trapped ion in an ultracold buffer gas, New J. Phys. 13, 053020 (2011).
doi:10.1088/1367-2630/13/5/053020.
[39] S.H. Chu, Recent developments in semiclassical Floquet theories for multipho-
ton process, in Advances in Atomic and Molecular Physics 21, edited by D.R.
Bates and B. Bederson (Academic Press, 1985), p. 197.
[40] R.J. Glauber, Quantum Theory of Optical Coherence: Selected Papers and
Lectures (Wiley-VCH, 2007), p. 577.
[41] L.S. Brown, Quantum motion in a Paul trap, Phys. Rev. Lett. 66, 527 (1991).
doi:10.1103/PhysRevLett.66.527.
144 BIBLIOGRAPHY
[42] D. Leibfried, R. Blatt, C. Monroe, and D. Wineland, Quantum dy-
namics of single trapped ions, Rev. Mod. Phys. 75, 281 (2003).
doi:10.1103/RevModPhys.75.281.
[43] D.J. Wineland, C. Monroe, W.M. Itano, D. Leibfried, B.E. King, and
D.M. Meekhof, Experimental issues in coherent quantum-state manipula-
tion of trapped atomic ions, J. Res. Natl. Inst. Stand. Tech. 103, 259 (1998).
arXiv:quant-ph/9710025.
[44] N.W. McLachlan, Theory and Applications of Mathieu Functions (Clarendon,
Oxford, 1947).
[45] Z.X. Wang, D.R. Guo, Special Functions (World Scientific, 1989), p. 610.
[46] R.J. Cook, D.G. Shankland, and A.L. Wells, Quantum theory of parti-
cle motion in a rapidly oscillating field, Phys. Rev. A31, 564 (1985).
doi:10.1103/PhysRevA.31.564.
[47] T.M. Miller, Atomic and Molecular Polarizabilities, in CRC Handbook of
Chemistry and Physics, 93rd Ed., edited by W.M. Haynes (CRC Press, 2012),
p. 10-188.
[48] W.F. Holmgren, M.C. Revelle, V.P.A. Lonij, and A.D. Cronin, Absolute and
ratio measurements of the polarizability of Na, K, and Rb with an atom inter-
ferometer, Phys. Rev. A81, 053607 (2010). doi:10.1103/PhysRevA.81.053607.
[49] R.M. Spector, Exact solution of the Schrodinger equation for inverse fourth-
power potential, J. Math. Phys. 5, 1185 (1964). doi:10.1063/1.1704224.
[50] E. Vogt and G.H. Wannier, Scattering of ions by polarization forces, Phys.
Rev. 95, 1190 (1954). doi:10.1103/PhysRev.95.1190.
BIBLIOGRAPHY 145
[51] K.M. Case, Singular potentials, Phys. Rev. 80, 797 (1950).
doi:10.1103/PhysRev.80.797.
[52] W.M. Frank, D.J. Land, and R.M. Spector, Singular potentials, Rev. Mod.
Phys. 43, 36 (1971). doi:10.1103/RevModPhys.43.36.
[53] R.M. Spector, Properties of the wave function for singular potentials, J. Math.
Phys. 8, 2357 (1967). doi:10.1063/1.1705167.
[54] J.H. Shirley, Solution of the Schrodinger equation with a Hamiltonian periodic
in time, Phys. Rev. 138, B 979 (1965). doi:10.1103/PhysRev.138.B979.
[55] S.H. Chu and D.A. Telnov, Beyond the Floquet theorem: Generalized
Floquet formalisms and quasienergy methods for atomic and molecular
multiphoton processes in intense laser fields, Phys. Rep. 390, 1 (2004).
doi:10.1016/j.physrep.2003.10.001.
[56] K. Drese and M. Holthaus, Floquet theory for short laser pulses, Eur. Phys.
J. D5, 119 (1999). doi:10.1007/s100530050236.
[57] S. Guerin and H.R. Jauslin, Control of quantum dynamics by laser pulses:
adiabatic Floquet theory, Advances in Chemical Physics 125, 147 (2003).
doi:10.1002/0471428027.ch3.
[58] W. Li and L.E. Reichl, Floquet scattering through a time-periodic potential,
Phys. Rev. B60, 15732 (1999). doi:10.1103/PhysRevB.60.15732.
[59] P.G. Burke, P. Francken and C.J. Joachain, R-matrix-Floquet theory of
multiphoton processes, J. Phys. B: At. Mol. Opt. Phys. 24, 761 (1991).
doi:10.1088/0953-4075/24/4/005.
[60] M. Grifoni and P. Hanggi, Driven quantum tunneling, Phys. Rep. 304, 229
(1998). doi:10.1016/S0370-1573(98)00022-2.
146 BIBLIOGRAPHY
[61] T.O. Levante, M. Baldus, B.H. Meier, and R.R. Ernst, Formal-
ized quantum mechanical Floquet theory and its application to sample
spinning in nuclear magnetic resonance, Mol. Phys. 86, 1195 (1995).
doi:10.1080/00268979500102671.
[62] H. Sambe, Steady states and quasienergies of a quantum-mechanical
system in an oscillating field, Phys. Rev. A7, 2203 (1973).
doi:10.1103/PhysRevA.7.2203.
[63] W.R. Salzman, Quantum mechanics of systems periodic in time, Phys. Rev.
A10, 461 (1974). doi:10.1103/PhysRevA.10.461.
[64] B.R. Johnson, New numerical methods applied to solving the one-dimensional
eigenvalue problem , J. Chem. Phys. 67, 4086 (1977). doi:10.1063/1.435384.
[65] J.M. Blatt, Practical points concerning the solution of the Schrodinger equa-
tion, J. Comp. Phys. 1, 382 (1967). doi:10.1016/0021-9991(67)90046-0.
[66] I.H. Sloan, Errors in the Numerov and Runge-Kutta methods, J. Comp. Phys.
2, 414 (1968). doi:10.1016/0021-9991(68)90047-8.
[67] B.V. Numerov, A method of extrapolation of perturba-
tions, Mon. Not. R. Astron. Soc. 84, 592 (1924). url:
http://adsabs.harvard.edu/abs/1924MNRAS..84..592N.
[68] W.H. Press, S.A. Teukolsky, W.T. Vetterling, and B.P. Flannery, Numerical
recipes: The art of scientific computing, 3rd Ed. (Cambridge University Press,
2007), p.158.
[69] W. Magnus, On the exponential solution of differential equations
for a linear operator, Comm. Pure Appl. Math. 7, 649 (1954).
doi:10.1002/cpa.3160070404.
BIBLIOGRAPHY 147
[70] A. Iserles and S.P. Norsett, On the solution of linear differential equa-
tions in Lie groups, Phil. Trans. R. Soc. Lond. A357, 983 (1999).
doi:10.1098/rsta.1999.0362.
[71] M. Hochbruck and C. Lubich, On Magnus integrators for time-
dependent Schrodinger equations, SIAM J. Numer. Anal. 41, 945 (2003).
doi:10.1137/S0036142902403875.
[72] A. Iserles, Think globally, act locally: Solving highly-oscillatory ordinary dif-
ferential equations, Appl. Numer. Math. 43, 145 (2002). doi:10.1016/S0168-
9274(02)00122-8.
[73] S. Blanes, F. Casas, J.A. Oteo, and J. Ros, The Magnus ex-
pansion and some of its applications, Phys. Rep. 470, 151 (2009).
doi:10.1016/j.physrep.2008.11.001.
[74] C. Moler and C.F. Van Loan, Nineteen dubious ways to compute the ex-
ponential of a matrix, twenty-five years later, SIAM Rev. 45, 3 (2003).
doi:10.1137/S00361445024180.
[75] G.H. Golub and C.F. Van Loan, Matrix Computation, 3rd Ed. (The Johns
Hopkins University Press, 1996), p. 555.
[76] R.B. Sidje, Expokit software. url: http://www.maths.uq.edu.au/expokit.
[77] R.B. Sidje, Expokit: Software package for computing matrix exponentials,
ACM Trans. Math. Software 24, 130 (1998). doi:10.1145/285861.285868.
[78] L.E. Reichl, The transition to chaos: Conservative classical systems and quan-
tum manifestations, 2nd Ed. (Springer 2004).
[79] F. Haake, Quantum signatures of chaos, 3rd Ed. (Springer 2010).
148 BIBLIOGRAPHY
[80] J.I. Cirac, L.J. Garay, R. Blatt, A.S. Parkins, and P. Zoller, Laser cooling
of trapped ions: The influence of micromotion, Phys. Rev. A49, 421 (1994).
doi:10.1103/PhysRevA.49.421.
[81] U. Peskin and N. Moiseyev, The solution of the time dependent Schrodinger
equation by the (t, t) method: Theory, computational algorithm and applica-
tions, J. Chem. Phys. 99, 4590 (1993). doi:10.1063/1.466058.
[82] M.A. Nielsen and I.L. Chuang, Quantum Computation and Quantum Infor-
mation (Cambridge University Press, 2000), p. 418.
[83] K.F. Riley, M.P. Hobson and S.J. Bence, Mathematical methods for Physics
and Engineering, 3rd Ed. (Cambridge University Press, 2006), p. 508.
[84] M. Abramowitz and I.A. Stegun, Handbook of Mathematical Functions, 10th
Ed. (Dover Publications, 1965), p. 727.