DÉPARTEMENT DE PHYSIQUE DE L’ÉCOLE NORMALE SUPÉRIEURE

LABORATOIRE KASTLER BROSSEL

THÈSE DE DOCTORAT
DE L’UNIVERSITÉ PIERRE ET MARIE CURIE
Spécialité : PHYSIQUE QUANTIQUE

Study of dipole-dipole interaction between Rydberg atoms

Toward quantum simulation with
Rydberg atoms

présentée par

Thanh Long NGUYEN

pour obtenir le grade de

DOCTEUR DE L’UNIVERSITÉ PIERRE ET MARIE CURIE

Soutenue le 18/11/2016 devant le jury composé de :

Dr. Michel BRUNE Directeur de thèse

Dr. Thierry LAHAYE Rapporteur
Pr. Shannon WHITLOCK Rapporteur
Dr. Bruno LABURTHE-TOLRA Examinateur
Pr. Jonathan HOME Examinateur
Pr. Agnès MAITRE Examinateur
To my parents and my brother
To my wife and my daughter
ii
Acknowledgement
“Voici mon secret. Il est très simple : on ne voit bien qu’avec le cœur.
L’essentiel est invisible pour les yeux.”

— Antoine de Saint-Exupéry, Le petit prince

This thesis is the culmination of many years of work, carried out with the support of
so many people that I know when writing down these lines, it is impossible to give a
special thank to each and every one of them. I, thus, list here only those who directly
inluenced this thesis. If someone is not mentioned, it does not by any means imply
that he or she has played a less important role in making this thesis possible.
First and foremost, I would like to thank the permanent members of the “Cavity
Quantum Electrodynamics” group, who gave me the wonderful opportunity to par-
ticipate in one of their frontier researches at Laboratoire Kastler Brossel, where I have
been surrounded by exceptional scientists, brilliant ideas and outstanding research
equipments.
I am forever indebted to my academic supervisors: Serge Haroche, who through
his series of lectures in Collège de France and by his Nobel prize, has inspired us,
young PhD students to follow fundamental research; Jean-Michel Raimond, who keeps
surprising us with his enormous imagination and is alway available to lend a listening
ear despite his busy schedule; and Michel Brune who agreed to supervise this thesis,
and relentlessly guided and supported me throughout the process. With Michel’s in-
sights into the ield, he is a “super” problem solver and a driving force in innovative
research. I always enjoy learning from his years of expertise, experimental skills, and
above all, his endless enthusiasms on the experiment.
I have the utmost privilege of working closely with the R14 team and would like
to give extensive gratitude to my team members. To Sha Liu for showing me how
to correctly design and realize an optical system at the time of my irst arrival even
though she could stay with us only for a few short months before leaving for another
job; Raul Teixeira for those long but enjoyable discussions on physics, his very good
understanding of physics and willingness to tackle new problems every single time;
Carla Hermann for teaching me how to treat the experiment gently, make the Helium
transfer less boring and how to keep oneself optimistic even in the most depressed
situation; Clément Sayrin for returning to lead the experiment after three-year doing
post-doc in Vienna, as working with him is quite pleasant and gainful; Tigrane Cantat-
iv Acknowledgement

Moltrecht for his initial C++ and Qutip programs that helped save me a lot of time in
building the simulation and Rodrigo Cortiña for his endless fundamental questions
on quantum mechanics which helped me to understand better the formalization of
the quantum theory.
My sincere thank goes next to Sébastien Gleyzes, who leads the fountain experi-
ment – a new version of the cavity CQED experiment, and sacriies his time working
as the “Helium manager” of the group. Without him, the experiment would have
been performed at room temperature. He also helped us tremendously in equipping
the experiment, from “big” machines like frequency generators to “small” stufs like
SMA connectors, and was there every time if we have any problem in R14, either ex-
perimentally or personally. I am also grateful to him for correcting my M2 internship
report, scanning and sending it to me and yet, I have not had a chance to thank him
personally.
A warm thank you to Igor Dotsenko for always inding time for us for whatever
problem that we have, his willingness to dive into hundreds of pages of code just to
troubleshoot the experiment controls, his endless and fascinating stories about Ukraine,
and many useful advices from both him and his wife on how to raise a child.
Beside those mentioned above, I would like to extend great thanks to my wonder-
ful group members, who have made my PhD life so enjoyable and successful: Bruno
Peaudecerf, Théo Rybarczyk, Stefan Gerlich, Adrien Signoles, Adrine Facon, Mariane
Penasa, Eva-Katharina Dietsche, Dorian Grosso, Fréderic Assemat, Valentin Métillon
and Arthur Larrouy.
I am thankful and honored to have Pr. Agnès Maître, Pr. Jonathan Home, Dr.
Bruno Laburthe-Tolra, Dr. Thierry Lahaye and Pr. Shannon Whitlock in my defense
committee.
I would like to thank also the administrative service, the electronic service, the
cryogenic service as well as the mechanical workshop as their work has helped the
experiment go on smoothly.
I am deeply thankful to my parents for their love, sacriices and constant supports
throughout my life, and to my brother for taking over my responsibilities to my parents
when I am away. And most importantly, a special recognition goes to my wife for
her unwavering love, quiet patience, constant encouragement and relentless support
during the entire process that are undeniably what made this thesis happens. And
inally, I would love to dedicate this last acknowledgment to my little lovely girl who
is the pride and joy of my life.
Contents

Introduction 1

I A short review on Rydberg atoms 9

1 Properties of individual Rydberg atoms 11

1.1 From the Hydrogen atom to the Rubidium Rydberg atom 11
1.2 Electric dipole 16
1.3 Radiative lifetime 20
1.3.1 Spontaneous emission 20
1.3.2 Blackbody-radiation-induced decay 22
1.4 Rydberg atoms in external ields 23
1.4.1 Magnetic ield 23
1.4.2 Electric ield 25

2 Interacting Rydberg atoms 31

2.1 Two antennas in communication 31
2.2 A pair of atoms in the same state 32
2.3 An atom pair involving two diferent states 34

II Towards quantum simulation with low- Rydberg atoms 37

3 Superconducting atom chip experiment 39

3.1 Ultra-cold atom source 40
3.1.1 Cryostat 40
3.1.2 Atom chip 42
3.1.3 Atom imaging 44
3.1.4 Laser system 46
3.1.5 Trapping and cooling of atoms 47
3.2 Rydberg excitation 51
3.2.1 Two-photon excitation 51
3.2.2 Rydberg detection 53
vi Contents

3.3 Coherent manipulation of Rydberg atoms 54

3.3.1 Taming the stray electric ields 54
3.3.2 Coherent manipulation of the Rydberg atoms 56

4 Microwave ruler for dipole-dipole interaction 59

4.1 Two regimes of Rydberg excitation 59
4.1.1 Dipole blockade 59
4.1.2 Facilitated excitation — Rydberg aggregate 61
4.1.3 Optical spectra 63
4.2 Microwave probe of van der Waals interaction energy 65
4.2.1 Choice of levels 65
4.2.2 Experimental work 68
4.3 Preliminary interpretation 71
4.3.1 Monte Carlo simulation 71
4.3.2 Numerical result 72
4.3.3 Limitation 74
4.4 Monte Carlo - rate equations simulation 75
4.4.1 Rate equations limit of the optical Bloch equations 75
4.4.2 Algorithm 77
4.5 Experimental data simulation 79
4.5.1 Optical proile 79
4.5.2 Microwave spectra 82
4.5.3 Spatial distribution of Rydberg atoms 83
4.5.4 Discussion on the validity of the Monte Carlo – rate equations model 83

III Quantum simulation with circular Rydberg atoms – A proposal 89

Objectives 91

5 Principles of the quantum simulator 95

5.1 The quest for long-lived circular atoms 95
5.1.1 Inhibition of spontaneous emission 95
5.1.2 Suppression of blackbody radiation absorption 96
5.1.3 Background gas collisions 97
5.2 Trapping of circular atoms 97
5.2.1 Ponderomotive potential 97
5.2.2 Photo-ionization by trapping lasers 99
5.3 Two interacting circular atoms in free space 100
5.3.1 Application of a magnetic ield 102
5.3.2 Dipole-dipole interaction and its scaling laws 105
5.4 1D spin chain’s Hamiltonian 110
5.4.1 1D spin chain Hamiltonians 112
5.4.2 Choice of levels and tunable spin chain 115
 Contents vii

5.5 Deterministic chain preparation 116

5.5.1 Initial irregular chain preparation 117
5.5.2 Spin chain tailoring 120
5.6 Arbitrary chain initialization 127
5.7 Site-resolved state-selective detection 128
5.8 Efects of decoherence 128

Conclusion and perspectives 131

Appendices 137

A Electric dipole matrix element estimation 139

A.1 Radial matrix element 139
A.2 Angular matrix element 140

B Laser frequency stabilization and laser distribution 143

C Measurement of the trap dimension 145

C.1 Temperature measurement 145
C.2 Trap frequencies 145

D Fourier limited line width 147

E Two atoms: motion and spin exchange 149

Résumé en français 153

viii Contents
Introduction

At the beginning of the last century, physicists have realized the “weirdness” of the
quantum world. Since then, they have witnessed revolutionary evolutions in our un-
derstanding of the quantum theory, which can be summarized in two achievements:
the description of wave-particle duality and the understanding of superposition and
entanglement, in which no classical model can it [1–3]. Together with these concep-
tual progresses, technologies and applications have also achieved important advance-
ments. One can mention the laser and the optical revolution, the atomic clocks and
the precise time measurements or the transistor and the computer revolution [3, 4].

Figure 1 Improvement in computational power (in log scale) over the past century. Figure
extracted from [5].

Let us focus on the latter. Born in the late 1940s, transistors followed by integrated
circuits have dramatically changed the way we communicate, diagnose and compute.
2 Introduction

Assemblies of cupboard-size boxes, packed with vacuum tubes and linked by kilome-
ters of wires have been replaced by microchip processors of a stamp size, a thousand
times smaller in weight and volume, but billion times more eicient in computational
power (Figure 1) [5]. The data storage capacities have been improved in the same
manner [6]. The reach of nowadays computers spreads from daily applications to
heavy computational tasks like weather forecasting [7], probabilistic analysis [8] or
deep learning and artiicial intelligence [9, 10]. The range seems to continue increas-
ing without any limitation.

Quantum simulations

The properties of electronic devices in modern computers are ruled by quantum laws
at the microscopic scale: tunnel efects, exclusion principle,... However, the computa-
tional algorithms are still based on pre-quantum logics, i.e., classical bits, which take
a value either 0 or 1. From the early 1980s, the limitations to which traditional com-
putations have to face when dealing with quantum many body systems became clear
[11]. Let us take as an example interacting spin 1/2 particles. Due to the fact that
a spin can be hung suspended between two states: up and down, the Hilbert space
has a dimension 2� . Calculating the time evolution of the system requires exponen-
tiating matrices with 2� × 2� elements. Classical stochastic methods, such as Monte
Carlo algorithms have been developed to deal with this ‘exponential-explosion’ prob-
lem [12]. However for some quantum systems, especially fermionic and frustrated
systems, they exhibit an exponential growth of statistical errors, which results in an
exponential growth of calculation time with the number of particles [13]. A system of
more than 40 spins is frequently cited in the literature [14–17] as the standard “thresh-
old” above which classical computational methods fail.
When R. Feynman delivered his seminal lecture in 1982, he proposed a brilliant
solution to this problem [11]

“ Let the computer itself be built of quantum mechanical elements which

obey quantum mechanical laws.”

That is to have “one controllable quatum system simulate another” [11]. This idea
of quantum simulation [16, 18, 19] is very appealing in that one solves at the same time
two problems: the problem of large amounts of information storage and the problem
of large amount of information processing. In a general procedure, after preparing an
initial state, one lets the simulator evolve into the inal quantum state and measures
some quantity of interest.
How does one quantum system simulate another? One way would be to approxi-
mate the time evolution of the simulated system by applying a sequence of elementary
quantum gates on an ensemble of qubits that encodes the state of the quantum sys-
tem. This follows the circuit model for quantum computation [20] and is referred to
as digital quantum simulation – the yet-to-be-built quantum computers. Despite its uni-
versality [16], it comes at a hight cost in gate number, and thus requires very high
idelity gates. Quantum error correction methods [21] have been developed to relax
this requirement.
 Introduction 3

Another approach, analog quantum simulation, uses a quantum system to mimic

(emulate) the evolution of another, i.e., “... be an exact simulation, that the computer
will do exactly the same as nature” [11]. This approach has less stringent requirements
on resources. Up to a certain tolerance level, the analog quantum simulator can give
useful results even in the presence of errors. It is thus simpler to design an analog
quantum simulator to handle a peculiar class of problems.
In recent years, quantum simulations has become one of the most promising sub-
ields of the quantum information science. First, advanced technologies are mature
enough to enable the control of quantum systems at a very high level. The 2012 Nobel
prize in physics was awarded for “groundbreaking experimental methods that enable
measuring and manipulation of individual quantum systems” [22]. The physical real-
ization of quantum simulation is thus within reach. Second, quantum simulation, by
overcoming the limitations encountered by classical computational methods, would
be able to tackle diicult problems for quantum many-body systems. Let us take as an
example spin networks. They present fascinating properties, such as quantum trans-
port, exotic phase transition, localization, slow evolution. A complete understanding
of these systems is of paramount importance for the advancement of fundamental
science, related to physics, chemistry and even biology. It would also accelerate the
development of radically new materials with fully engineered properties. Quantum
simulations have thus the potential to shape future applications.

State of the art

Quantum simulations are attracting an increasing interest in many areas of physics

(Figure 2). They have been intensively studied both theoretically and experimentally
worldwide with cold atoms [23–26], trapped ions [27, 28], superconducting devices
[29], polaritons in cavities [30], photons in networks [31, 32], polar molecules [33],
quantum dots [34–36] and many others. Several proof-of-principle demonstrations
have been performed [15, 37–40]; however, the majority have not yet outperformed the
capacities of classical computers. Covering all these developments is out of the scope
of this thesis. Instead, we briely review here some of the most advanced systems,
which rely on trapped ions, superconducting circuits or optical lattices.

Superconducting circuits

Superconducting circuits involve macroscopic qubits that display quantum behaviors

like “artiicial atoms” [41, 42]. They can be used to test quantum physics at macro-
scopic scale or to simulate atomic physics and quantum optics. These electric-circuit
based qubits can be manipulated and measured with very high idelity. In addition,
the extremely strong coupling between them provides spin-spin interaction. This ar-
chitecture is suitable for both digital [43, 44] and analog [45, 46] quantum simulations.
Another advantage is that the artiicial atoms are quite easy to design, fabricate and
eventually wire together to realize various lattice geometries. The coherence times
of the state-of-the-art superconducting qubits can exceed 100 µs. However, quantum
simulation speed-up with a large number of qubits is still an open question [47–49].
4 Introduction

Polar
molecules
Neutral atoms
Arrays of in optical lattices
quantum dots

Quantum simulator

Arrays of
Others trapped ions

Superconducting Arrays of
circuits cavities

Nuclear spins

Figure 2 Quantum simulation has been raising interest from many areas of physics. Figure
extracted from [19].

Trapped ions

Trapped ions [50, 51] have shown to be a very promising system for quantum informa-
tion processing [52] and for quantum simulation [27, 53, 54]. Both the internal energy
levels and the vibrational modes of the ions can be used to encode quantum informa-
tion, which can be read out with almost unit eiciency. The coherence times of ion
qubits when encoded on two hyperine levels have reached in the order of seconds
(even minutes) and application of sequences with hundreds of high-idelity gates is
feasible. Trapped ions have performed the most advanced implementation of digital
quantum simulations [55, 56] to date, notably the simulation of a QED pair creation
process [57]. When operating basing on the motional modes, the coherence time is
much shorter.
Analog quantum simulation with trapped ions has also been demonstrated. Among
these demonstrations are the realization of frustrated spin systems, of entanglement
propagation [58, 59] and of many-body-localization [60]. The coupling between the in-
ternal states and the vibrational modes induced by laser driving leads to an efective
spin-spin interaction, which can be tuned from long-range to medium-range behav-
iors [61, 62]. However, most of these experiments are performed with 1D chains of
ions held in a linear Paul trap. Eforts to go to quantum simulations with 2D arrays
have resulted in the realization of ferromagnetic and anti-ferromagnetic Hamiltonians
with hundreds of ions in a Penning trap [63, 64]. The interactions, nevertheless, cannot
be tuned in the nearest neighbor regime of great interest. There are other proposals
using 2D Paul traps [65, 66] or arrays of microtraps [67, 68]; however coherence times
in these systems have not yet been measured.
 Introduction 5

Cold atoms in optical lattices

Cold atoms in optical lattices are remarkable tools for analog quantum simulation
[24–26]. The atoms, either bosons or fermions, are placed at the bottom of 1D, 2D or
3D optical potentials. Dynamics is then provided by the competition between adja-
cent site tunneling and on-site two-body interaction. The former is controlled by the
lattice potential depth, while the latter is adjusted through the Feschbach resonance
mechanism. Atoms are imaged individually with sophisticated optics [69, 70]. Their
internal state is resolved by luorescence. Cold atoms in optical lattices are lexible
systems, not only because of their capability of realizing various geometries but also
because of the tunability of the experimental parameters. Since the irst observation
of the phase transition from a superluid to a Mott insulator [37], quantum simulation
with cold atoms has become a very active ield. It can address a wide range of prob-
lems such as the Bose-Hubbard Hamiltonian [24], the many-body localization [71, 72],
or recently the artiicial gauge ields [73, 74] and the fractional quantum Hall efect [75,
76].
In spite of these successes, quantum simulation with cold atoms in optical lattices
has a drawback. The tunneling rate is relatively low (few hundred Hz) as compared
to the lifetime of the trapped atomic cloud (seconds). The evolution is thus limited
to a few hundred typical tunneling times. This is certainly a big disadvantage when
slow dynamics is involved such as in many-body localization or spin glasses. As a
solution, trapping with other type of lattices with smaller spacing has been proposed
[77, 78]. Alternatively one can use polar molecules [79, 80] or atomic species with
a large magnetic dipole moment [81, 82] to enhance the interaction strength. These
systems are, however, rather challenging to realize.

Quantum simulation with Rydberg atoms

Rydberg states are highly excited atomic states with remarkable properties [3, 83],
even quite exaggerated when compared to those of ordinary atomic states. Among
their properties, one will note particularly their large electric dipole matrix elements
and their long lifetimes, not to mention their simple internal structures and their easy
experimental manipulation. Put into numbers, the dipole matrix elements are typ-
ically of a few hundred to thousand times larger than those of ground-state atoms.
Rydberg states can last longer than hundreds of µs. In particular, circular Rydberg
states — Rydberg states with maximum angular momentum and magnetic quantum
numbers — have lifetimes in the order of tens of milliseconds. Their properties have
been successfully exploited in the framework of cavity quantum electrodynamics (cavity
QED) with a series of experiments developed at Laboratoire Kastler Brossel [84–95].
With large electric dipole matrix elements, these long-lived, highly excited atoms
when placed close together interact strongly via the van der Waals dipole-dipole inter-
action, 6 / 6 where is the inter-atomic distance. With of a few µm, the interaction
energy can be in the MHz range, and even larger for the Förster-like resonant energy
exchange interaction (scaling as 1/ 3 ) [96]. Changing the principal quantum number,
one can tune this interaction energy by orders of magnitude.
A protocol for digital quantum simulations with Rydberg atoms has been devel-
oped [97, 98]. It is based on the dipole blockade mechanism, which is a direct conse-
6 Introduction

quence of the strong Rydberg-Rydberg interaction [99]. In a small sample of cold

ground-state atoms, the van der Waals interaction is much higher than the typical
laser excitation linewidth (∼ 100 kHz). As soon as an atom is resonantly laser excited,
all the others are tuned out of resonance, resulting in the excitation of one and only one
Rydberg atom at a time [100, 101]. One can thus, realize a Controlled-NOT (CNOT)
gate, in which depending on the state of the control qubit, the state of all other qubits
is left unchanged or lipped [96, 99, 102–104].
The strong interaction between Rydberg atoms also makes them very promising
tools for analog quantum simulation of spin networks. This has renewed the interest
for these atoms, and led to many proposals for quantum simulations with Rydberg
atoms [105–108]. Among many experimental demonstrations, the observation of co-
herent excitation transport over a chain of three Rydberg atoms is quite remarkable
[109]. The Rydberg excitation dynamics of arrays with more than 20 atoms has been
recently reported [110]. Eforts to mix ground-state atoms with Rydberg atoms, i.e.,
Rydberg dressing, in optical lattices have been proposed as a solution to the slow tun-
neling problem [111–115]. However, these experiments are at 300 K and thus sufer
from the blackbody radiation, which modiies the Rydberg interaction in an uncon-
trolled way [116–118]. Going to cryogenic environment to suppress the number of
thermal photons is a diicult solution.
However, Rydberg atoms used in the above systems have low angular momentum.
Quantum simulation with low- Rydberg atoms has intrinsic limitations. First of all,
the Rydberg atoms experience mechanical forces either attractive or repulsive, due
to their strong interactions. Without being trapped, any Rydberg crystal will “melt”
down after a few tens µs. Trapping Rydberg atoms in an optical lattice [119] has been
studied. However, this trapping sufers from a strong reduction in the Rydberg life-
time due to photo-ionization by the trapping laser [120]. Second, the lifetimes of low-
Rydberg states are long compared to those of ordinary excited states, but still only of
a few hundred µs, corresponding to a few hundred of typical exchange periods. This
prevents quantum simulation with low- Rydberg atoms from studying slow dynam-
ics of arrays with tens of spins.
In such a context, this thesis studies the interaction between Rydberg atoms and
explores new schemes for quantum simulations with Rydberg atoms. Experimen-
tally, we laser-excite Rydberg atoms, either resonantly or of-resonantly, from a cold
and dense cloud of ground-state Rubidium atoms prepared using a superconducting
atom chip. Since Rydberg atoms are quite sensitive to thermal radiation, the exper-
iment is performed in a cryogenic environment, to which a superconducting chip is
perfectly adapted. This work contributes to four main achievements, which can be
summarized as follows.
Control of the electric ield and coherent manipulation of Rydberg levels in the
vicinity of the chip. The downside of having large dipole matrix elements is that Ry-
dberg atoms are extremely sensitive to stray electric ields. The inevitable slow depo-
sition of Rubidium atoms on the chip surface during the experiment creates patches of
dipoles, resulting in a temporally unstable and spatially inhomogeneous electric ield
in the vicinity of the chip [121, 122]. The Stark efect thus enlarges the transition lines
to an extent (∼ 40 MHz) that completely destroys the dipole blockade mechanism.
This has been a formidable experimental challenge for us and other teams. A solution
 Introduction 7

to this problem is inally rather simple. A thick layer (∼ 100 nm) of Rubidium cover-
ing the chip in a large area saturates the efect of slow Rubidium deposition. The ield
inhomogeneity is thus strongly suppressed, providing a good environment to study
the Rydberg-Rydberg interaction. This solution is detailed in Carla Hermann’s thesis,
and is the object of a highlight in Physical Review A [123].
Microwave probe of the van der Waals interaction energy in a cold Rydberg gas.
Being laser excited from a dense cloud, Rydberg atoms strongly interact with each
other. Microwave spectroscopy of the transition to a nearby Rydberg state allows us
to precisely probe the interaction energy distribution. We have applied this technique
to observe the mechanical expansion of an ensemble of Rydberg atoms under strong
repulsive interaction. This observation sets up a time scale during which the frozen
Rydberg gas approximation is still valid. A simple Monte Carlo model has also been
developed to explain the microwave spectra. The method is the subject of Raul Teix-
eira’s thesis and is summarized in a paper in Physical Review Letter [124].
Developing a more rigorous simulation of the Rydberg excitation process. De-
spite the fact that the simple Monte Carlo model succeeds in reconstructing the mi-
crowave spectra, it fails for the optical excitation lines and exhibits several limitations,
preventing us from studying new excitation schemes. A new model has been devel-
oped during the reconstruction of the experiment after its re-localization in our new
laboratory at Collège de France in 2014. This new model gives us a good insight into
the excitation process.
Quantum simulation with laser-trapped circular atoms. Atoms in circular Ry-
dberg states, despite their unique properties and successes with cavity QED experi-
ment, have not yet been, to our knowledge, fully exploited for quantum simulation.
After careful investigations on the interaction between them, we came up with a pro-
posal for quantum simulation. The proposed scheme aims at overcoming the limita-
tions of other systems, including those with low angular momentum Rydberg atoms.
It is based on the groundbreaking concept of laser-trapped circular Rydberg atoms,
protected from spontaneous decay and interacting strongly through a dipole-dipole
coupling. More than that, it combines the lexibility found in atomic lattices, the strong
interactions, typical of Rydberg atoms, and the exquisite control, typical of trapped
ions. The realization of such a platform would allow quantum simulation to exceed
the grasp of classical computational methods.

This manuscript is organized in three parts with ive chapters. In the irst part,
we recall the important properties of Rydberg atoms. These properties make it clear
why Rydberg atoms are very good tools for quantum simulation. The irst part also
includes elements necessary for the understanding of the two following parts. Chapter
1 concentrates on individual Rydberg atoms. It consists in the description of Rydberg
states, Rydberg’s lifetimes and their behaviors in an external electric and/or magnetic
ield. Low angular momentum Rydberg atoms and circular (high- ) Rydberg states are
presented in parallel. Considering two Rydberg atoms close to each other, chapter 2
formalizes the interaction between Rydberg atoms in free space. We discuss in details
the situation of the low Rydberg atoms, but the concept can be easily extended to
circular atoms as well.
8 Introduction

In the second part of the manuscript, we present our studies on dipole interaction
between low- Rydberg atoms. Chapter 3 describes the current experimental setup,
and the preparation of cold and dense ground-state atoms as well as the Rydberg ex-
citation. We discuss briely also how we managed to obtain an unprecedented long
coherence time with Rydberg atoms near a metallic surface. Chapter 4 presents our
new insight into the Rydberg excitation process. We recall the experimental results on
the excitation of Rydberg atoms in the strong interaction regime as well as our prelim-
inary interpretation. Then we develop a more rigorous model, the Monte Carlo – rate
equations simulation, with which the Rydberg excitation process is better described.
To end this chapter, we use the developed model to consider the preparation of a 1D
Rydberg chain for quantum transport simulations.
In the third part, we describe our proposal of a new platform — laser-trapped cir-
cular atoms — for quantum simulation. After presenting the objective and principles
of the proposal, we discuss in details the main ingredients for realizing the quan-
tum simulator with circular atoms. We show that a tunable spin Hamiltonian for a
1D chain, based on the interaction between circular atoms is feasible with a proper
choice of Rydberg states. We also present a complete experimental sequence to deter-
ministically prepare a chain with more than 40 atoms. Many technical details will be
considered, showing the realism of the proposal.
We conclude this manuscript by discussing a wide range of problems that trapped-
circular-atoms could be able to address.
 I.

A short review on Rydberg atoms

Properties of individual
Rydberg atoms 1
In this chapter, we will discuss the main properties of Rydberg atoms, which make
them very good tools for quantum simulation. They consist in their long lifetimes,
their huge dipole matrix elements and, as a consequence, their strong mutual interac-
tion. Not less important from the point of view of an experimentalist, are their easy
manipulation using available microwave technology, their easy detection using ield
ionization and their simple internal structure. In each section, we will irst discuss the
general principles. Then, we will go into details for low- states and for circular states.
The diferences between low- states and circular states can be thus easily recognized.
Examples given along this chapter are mainly for 60 state and for the circular 50
state. They are the levels chosen for our setup under technical constraints, which will
be discussed in next parts.

1.1
From the Hydrogen atom to the Rubidium Rydberg atom

Hydrogen atom approximation

Rydberg atoms are highly excited atoms, very close to the ionization limit. Most of the
time, the outer electron stays far away from the atomic core which is several orders of
magnitude smaller than the orbit of the electron* . Rydberg atoms can therefore be
very well approximated by a Hydrogen atom, where the outer electron held by the
Coulomb force orbits around an elementary charge. If we consider that the core is
ininitely massive the Hamiltonian reads

ℏ2 2 2 1
=− ∇ − ,
0
2 � 4��0
(1.1)

where � is the mass of the electron, is the elementary charge, �0 is the vacuum
electric permittivity, ℏ = ℎ/2� is the reduced Planck’s constant and is the distance of
the electron from the core. In atomic units the Hamiltonian takes the form

∇2 1
=− − .
0
2
(1.2)

*
Here we consider only akali atoms with a single valence electron
12 Properties of individual Rydberg atoms

Its eigenfunctions are well known and are written as a product of radial and angular
functions

�( , �, �) = �� ( ) (�, �) ,
��
�
(1.3)

where � and � are respectively the polar angle with respect to the quantization axis
and the azimuthal angle [83, 125]. Here � � (�, �) is a spherical harmonic function,
�

which is proportional to an associated Legendre polynomial � � (cos �)

�

(�, �) = (cos �) ,
�� ��� � ��
� �
(1.4)

where is a normalization constant. The radial part �� ( ) can be expressed using a

generalized Laguerre polynomial 2�+1
�−�−1
of degree − − 1

2 ( − − 1)! 2 2
3 �
�� ( ) = ( ) ( ) ( ),
−�/��0 2�+1
2 [( + )! ] �−�−1
(1.5)
√ 0 0 0

where 0 ≈ 5.29 × 10−11 m = 0.529 Å is the Bohr’s radius. The square root term is a
normalization factor.
A state of the atom is therefore deined by three integers , and whose values
satisfy

= 1, 2, 3, … ,
= 0, 1, 2, …, − 1 , (1.6)
� = − , − + 1, …, −1, 0, 1, …, − 1, .

They are called respectively the principal quantum number, the azimuthal quantum
number and the magnetic quantum number. Using these three quantum numbers, we
denote a state of the atom as ∣ ⟩. Rydberg atoms have very high principal quantum
number ≫ 1. The binding energy is given by the Rydberg formula:

1 1
=− =− ,
2
� 2 2 ∞ (1.7)

where
4
= �
= 0.5 a.u = 13.605693009(84) eV
8�02 ℎ2
∞ (1.8)

is the Rydberg constant [126]. To account for the inite mass of the atomic core, one
has to replace ∞ by

= ∞ .
1+ �
� �
(1.9)
�

Nevertheless the modiication is small as � ≪ .

In addition, the state of the electron spin is represented by a fourth quantum num-
ber (spin quantum number), which takes only two values, spin up = +1/2, or spin
down = −1/2.
 1.1 From the Hydrogen atom to the Rubidium Rydberg atom 13

a) b) ×10 5

1. 0

0. 5

60 ( r )
0. 0

R
0. 5

0 1000 2000 3000 4000 5000 6000 7000

r /a 0

60 ( ) for 87 Rb 60 state. b)
Figure 1.1 a) Radial probability of inding the outer electron 2

Radial part of wavefunction for the 87 Rb circular 60 state.

Circular Rydberg state

A remarkable case is when and | � | take their maximum values = | �| = − 1. From

equations (1.3) and (1.5) the wavefunction has then a simple form

1 1
�−1
�( , �, �) = (− sin � ��
) −�/��0
.
!
(1.10)
� 3
0
0
√

The orbital is a torus of radius 2 0 and thickness (dispersion) Δ / = Δ� ≈ 1/√2 ,

lying perpendicular to the quantization axis . A cut through the plane of the torus
is depicted in Figure 1.1a. In the Bohr-Sommerfeld semi-classical model, the outer
electron travels around the atomic core on eccentric elliptical trajectories depicted in
Figure 1.2. The larger , the less eccentric the orbit is. The torus somehow corresponds
to the semi-classical image of a circular orbit with a maximum angular quantum num-
ber , especially for very high , where the coninement to the Bohr orbit is tighter. The
state with = | � | = − 1 is thus given the name circular state, and denoted ∣ ⟩. The
other states are called elliptical to distinguish them from the circular one. There are
two circular states in the same manifold, corresponding to � > 0 and � < 0. In
practice, a strong electric ield is added to deine the quantization axis. The behaviors
of the two circular states in presence of an external electric ield are the same (see sub-
section 1.4.2). If not speciied otherwise, we implicitly consider the � positive case.

Fine structure

In a more complete picture, we have to take into account the interaction between the
electron spin and its motion. The interaction Hamiltonian reads

�� = �⋅�, (1.11)

where � and � are the total orbital momentum and the spin operators of the Rydberg
electron. The strength of the interaction is represented by the constant . We deine
14 Properties of individual Rydberg atoms

n = 60

l = 0 (S)

l = 20

l = 40

l = 50
l = 59

Figure 1.2 Trajectories of the outer electron as described by the Bohr-Sommefeld model, drawn
for a = 60 Hydrogen atom with = 0, 20, 40, 50 and 59. The black dot represents the position
of the atomic core.

the total angular momentum operator

�=�+�. (1.12)

It can be shown that the ive operator 0 , �2 , �2 , �2 and � – the projection of � on the
quantization axis all commute with each other and with �� . An eigenstate of the
Hamiltonian is simultaneously an eigenstate of the ive operators. In addition to ,
and , we introduce two new quantum numbers to represent an eigenstate of the atom,
the total angular momentum quantum number � and the total angular momentum
projection quantum number � . Their possible values are

1
�= ±
2 (1.13)
� = −�, −� + 1, …, � − 1, � .

Since takes only one value = 1/2 for a single valence electron, the corresponding
state of the atom can be denoted as ∣ � � ⟩. The ine structure takes into account both
the spin-orbit interaction and the correction due to relativistic efects. The corrected
binding energy is given by [127]

� �� 3
2
= + ( − ) ,
��� � 2 � + 1/2 4
(1.14)

where �� = 2 /4��0 ℏ ≈ 1/137 is the ine structure constant. For ≈ , and thus � ≈ ,
the correction term is approximately �2� /4 2 times smaller than � . As an example, it
is about 7.6 kHz for = 50, much smaller than a typical interaction with an external
electric or magnetic ield (a few MHz). We thus ignore the ine structure for high-
Rydberg states, including the circular states.
One can go further with the hyperine structure description in which the inter-
action couples the electron angular momentum with the nuclear spin. However for
 1.1 From the Hydrogen atom to the Rubidium Rydberg atom 15

Rydberg atoms, the coupling is typically much weaker than the interaction with ex-
ternal ields. For instance, the hyperine shift of 60 1/2 state is of about 200 kHz [128].
We therfore neglect the hyperine structure in the following.

Quantum defect

So far, we have treated the atomic core as a single point charge. This assumption holds
as long as the outer electron is far away from the nucleus. As illustrated in Figure 1.2,
the eccentricity tells us how much the electron gets close to the core. All the description
above in terms of the Hydrogen atom is valid for large Rydberg states. Circular state
is an extreme case where the hydrogenic approximation is nearly perfect.
However for low- Rydberg states, one has to take into account the non-negligible
penetration of the outer electron into the electronic cloud of the core. The electron
sees the deviation of the ionic core potential from that of the Hydrogen atom. It feels
a deeper Coulomb potential due to a smaller screening of the nucleus charge from
the inner electronic cloud. In addition, the presence of the outer electron close to the
core leads to a deformation of the inner electronic cloud. One needs to include higher
order correction terms (dipole, quadrupole...) to account for the polarization of the
atomic core. These efects lower the energy of the Rydberg state. It is interesting that
the energy is still described by a modiied Rydberg formula

1
=− ,
2( − �� )2
�� (1.15)

where �� is called the quantum defect.

A more precise correction as suggested in [83] takes into account the spin orbit
interaction and expresses the quantum defect as a power series of 1/( − ���,0 ):

���,2 ���,4 ���,6

���� = ���,0 + + + +⋯.
( − ���,0 )2 ( − ���,0 )4 ( − ���,0 )6
(1.16)

The coeicients are determined experimentally using precise microwave spectroscopy

for neighboring Rydberg transitions [129, 130] or by observing electromagnetically in-
duced transparency with help of a calibrated frequency comb [131]. Only the irst two
terms (modiied Rydberg - Ritz parameters) are suicient to give a precision better
than a hundred kHz. The measured quantum defects for 85 Rb and 87 Rb are summa-
rized in Table 1.1. Up to our knowledge, there are no precise measurements for 87 Rb
so far other than for and levels. However, using the quantum defects of 85 Rb still
yields good results as the electronic structure for the two Rubidium isotopes are quite
similar. The only diference comes from the mass diference, which slightly modiies
� . This approximation can be veriied for and levels as depicted in Table 1.1. No
experimental measurements of the quantum defects have been reported for > 4. As
discussed by T. Gallagher in [83], the non-penetrating high-angular-momentum states
have quantum defects scaling as −5 . Knowing that ��,0 = 0.004, we can roughly esti-
mate ��≥4 as

4
5
��≥4 = ��,0 ( ) . (1.17)
16 Properties of individual Rydberg atoms

For very high- Rydberg states, the quantum defects drop rapidly to zero as expected.
Letting ∗ be the principal quantum number corrected by the quantum defect, the
binding energy takes again the Rydberg formula form

1 1
=− ∞ = − ∞�� ∗2 .
�∗
1+ � ( − ���� ) 2 1+
�
(1.18)
� �

Table 1.1 Quantum defects for 85 Rb extracted from [129, 130] and for 87 Rb extracted from
[131]

85 Rb 87 Rb

���,0 ���,2 ���,0 ���,2

1/2 3.131 180 4(10) 0.1784(6) 3.131 180 7(8) 0.1787(2)

1/2 2.654 884 9(10) 0.2900(6)
3/2 2.641 673 7(10) 0.2950(7)
3/2 1.348 091 71(40) −0.602 86(26) 1.348 091 8(11) −0.6054(4)
5/2 1.346 465 72(30) −0.596 00(18) 1.346 462 2(11) −0.5940(4)
5/2 0.016 519 2(9) −0.085(9)
7/2 0.016 543 7(7) −0.086(7)
0.004 00(9)
, >4 0.004(4/ )5

Figure 1.3 shows the energy level diagram for Rydberg levels with ≈ 60. The
quantum defects break the degeneracy of , , and levels in the same manifold.
For > 3, the states are almost degenerate and can be well approximated by a Hydro-
gen atom. Remarkably the transition between the Rydberg levels is in the microwave
range. This allows us to use available microwave technologies in experiments with
Rydberg atoms.

1.2
Electric dipole

In a simple picture, the electron is held loosely around the core by the Coulomb force.
Any electric perturbation can lead to a deformation of the electron orbit, resulting
in a large induced dipole moment. In the language of quantum mechanics, Rydberg
atoms have huge matrix elements of the dipole operator, which correspond to a large
coupling between Rydberg states induced by external (DC or AC) electric ields. The
knowledge of the electric dipole helps to understand other properties of the Rydberg
atoms such as radiative lifetime, Stark efect and the dipole-dipole interaction between
them. The dipole matrix element between two levels ∣ � ⟩ and ∣
′ ′ ′ ⟩ is
�

⟨ ′ ′ ′∣�∣
� �⟩ =− ⟨ ′ ′ ′∣�∣
� �⟩ , (1.19)
 114.2184 63P3/2
113.8240 63P1/2

102.9011 60F7/2 103.2819 60G ... 103.4038 103.4038 103.4038 103.4038 103.4038 103.4038 60C
99.3932 63S1/2 102.9004 60F5/2
92.7630 61D5/2
92.7127 61D3/2
83.5368 62P3/2
83.1211 62P1/2

71.6350 59F7/2 72.0355 59G ... 72.1637 72.1637 72.1637 72.1637 72.1637 59C
67.9454 62S1/2 71.6342 59F5/2
60.9712 60D5/2
60.9182 60D3/2
51.2635 61P3/2
50.8260 61P1/2

38.7373 58F7/2 39.1589 ... 39.2938 39.2938 39.2938 39.2938 58C

58G
34.8532 61S1/2 38.7365 58F5/2

27.5107 59D5/2
27.4550 59D3/2
17.2874 60P3/2
16.8267 60P1/2

4.0925 57F7/2 4.5366 ... 4.6788 4.6788 4.6788 57C

57G
0 60S1/2 4.0917 57F5/2

-7.7371 58D5/2
-7.7959 58D3/2

-18.5132 59P3/2
-18.9989 59P1/2

-32.4253 56F7/2 -31.9570 ... -31.8071 -31.8071 56C

56G
-36.7414 59S1/2 -32.4262 56F5/2

-44.9020 57D5/2
-44.9639 57D3/2

-56.2715 58P3/2
-56.7840 58P1/2

-70.9537 55F7/2 -70.4594 ... -70.3012 55C

55G
-75.5100 -70.9546 55F5/2
58S1/2

-84.1254 56D5/2

1.2 Electric dipole

-84.1909 56D3/2

-96.1327 57P3/2
-96.6740 57P1/2

l
0 1 2 3 4 54 55 56 57 58 59

Figure 1.3 Energy diagram for Rydberg states around = 60. The energy is given in GHz and is referenced to the 60 1/2 level.

17
18 Properties of individual Rydberg atoms

where � = − � is the dipole operator. A convenient way to calculate this dipole cou-
pling is to use the representation of the coordinates , and in terms of the spherical
harmonics

1 3
1 (�, �) = sin � ,
−1 −��
2 √ 2�
1 3
1 (�, �) = cos � ,
0
2√�
(1.20)

1 3
1 (�, �) =− sin � ,
+1 −��
2 √ 2�
i.e.,

4� −1 − +1
= 1 1
,
√ 3 √2
−1 +
4� +1
= 1 1
�,
√ 3
(1.21)
√2
4�
= 0 .
√ 3 1

The calculation reduces to the evaluation of the following components

⟨ ′∣
1∣ �⟩ =⟨ ′ ′∣ ∣ ⟩⟨ ′ ′∣
1∣ �⟩ ,
′ ′ � �
� � (1.22)

where = 0 or ±1. For low- Rydberg atoms, equation (1.22) is written generalized to

⟨ ′ ′ �′ ′∣
1∣ � �⟩ =⟨ ′ ′ �′ ′∣ ∣ �⟩ ⟨ ′ �′ ′∣
1∣ � �⟩
� �
� � � (1.23)

in order to take the ine structure into account. The integrations are therefore sepa-
rated into a product of a radial part and an angular part.

Selection rules

The analytic forms of the angular parts in equation (1.22) and (1.23) can be derived as
shown in Appendix A. They are nonzero if , ′ and 1 satisfy a triangular relation. In
addition + ′ has to be an odd number. This restrains to

|′− |=1. (1.24)

Furthermore the momentum projection numbers have to fulill

� (�) = ′
� (�) + . (1.25)

These conditions deine selection rules for the dipole transitions. Those transitions
correspond to the absorption or emission of a � polarized ( = 0) photon or of a � ±
polarized ( = ±1) photon. They are called dipole allowed transitions.
 1.2 Electric dipole 19

Table 1.2 Scaling laws for dipole coupling between two Rydberg levels ∣ ⟩ = ∣ ⟩ and
∣ ′ ′ ′ ⟩.

′ ′ ′ Transition Scaling law

−1 −1 �− 3/2

+1 +1 +1 �+ 2

+1 +1 � 3/2

+1 +1 −1 �−
+1 −1 −1 �− 2

Numerov method

In order to estimate the radial part of the matrix element in (1.22), one needs to know
the radial wavefunctions of each state. They are numerically obtained using the Nu-
merov method. Here, we describe the main idea of the calculation.
The interaction of the outer electron with the ionic core only happens at short dis-
tances. Rydberg atom can be considered as an electron in a pure Coulomb potential
with an energy given by the Rydberg formula modiied by the quantum defects. The
corresponding Schrödinger equation for the radial wavefunction �� reads

�2 2 � �� 2 ( + 1)
��
+ + [2 + − ] =0,
� 2 � �∗ 2 �� (1.26)

where �∗ is given by (1.18). This equation can be solved numerically (see Appendix A
for more details).
The integral starts far from the outer classical turning point of the electron orbit,
at a typical distance = 2 ( + 15) 0 from the atomic core, where we know that
the wavefunction exponentially decays to zero. A rough estimate of the initial val-
ues of the radial wavefunction thus quickly converges to the physical solution in the
next steps of iteration. The integral is carried out inwards towards = 0. Due to the
nonzero quantum defect, the wavefunction diverges at small . The integral has to be
stopped as soon as it encounters the inner turning point and starts to diverge. The
contribution at short distances to the dipole matrix element is nevertheless small. We
still have a good estimation.
Figure 1.4a plots the radial part of the wavefunction for 87 Rb 60 1/2 state. The
corresponding wavefunction of Hydrogen is also shown. A phase shift due to the
interaction with the 87 Rb ionic core is clearly observed. Figure 1.4b illustrates the
distribution of the outer electron which gives an intuitive image of the 60 1/2 orbital.
The wavefunctions for high- Rydberg states are approximately those of Hydrogen
atom. Their analytical forms are given in (1.5). However, in all numerical simulations,
we use the Numerov method to estimate the dipole matrix elements for all .

Scaling laws

Understanding the scaling laws for the dipole matrix elements is very important. For
low- Rydberg states, the dependence of the dipole matrix element is determined by
20 Properties of individual Rydberg atoms

a) ×10 5 b)
1
1. 0 H 60S 1/2
87
Rb 60S 1/2

0. 5
R60S ( r )

0. 0

0. 5

0 1000 2000 3000 4000 5000 6000 7000 8000 9000

r /a 0

Figure 1.4 a) Radial wavefunctions 601/2 ( )for 87 Rb 60 1/2 state (blue) and Hydrogen 60 1/2 .
b) Probability of inding the Rydberg electron 2 601/2 ( ).

the radial overlap of the two sub-levels. In the case of transitions between neighboring
levels, it is proportional to size of the orbit radius. The dipole matrix element thus
scales as 2 . However, going to circular state, the dependency of the angular part on
has to be taken into account. One can prove that the dipole connecting a circular state
with the nearest elliptical state in the same manifold scales as 3/2 while that with the
nearest manifold scales as 2 , 3/2 and 2 for � + , � and � − transition respectively.
These scalings are summarized in Table 1.2
To give an order of magnitude, the dipole matrix element corresponding to the
transition between 60 1/2 , � = 1/2 and 60 3/2 , � = 3/2 states is 2127 0 and that
of the 60 to 61 transition is 2556 0 . In comparison, the dipole matrix element for
the 2 line transition 5 1/2 − 5 3/2 is of about 4.2 0 . Those of Rydberg atoms are
obviously huge numbers. As a direct consequence, Rydberg atoms strongly couple to
microwave radiation. In other words, using available microwave technology, one can
easily manipulate transitions between Rydberg levels.

1.3
Radiative lifetime

1.3.1
Spontaneous emission

The long lifetime of Rydberg atoms can be classically interpreted as the result of the
very low acceleration of the outer electron on its large orbit. The Larmor formula in the
classical electrodynamics framework [132] predicts the radiated power of an electron
orbiting around its nucleus on a circular orbit to be

2 2 2
=
3 4��0 3
(1.27)
 1.3 Radiative lifetime 21

3
x10
2.5
30

59P
25

2.0 60P
A, B (s ) 20
-1
A, B ×103 (s-1)

15

10
1.5
5

0
20 30 40 50 60 70 80 90
1.0
nP

A 60S-nP
5P B 60S-nP at 300 K
0.5 B 60S-nP at 4 K
6P

0.0
10 20 30 40 50 60 70 80 90
nP

Figure 1.5 Contribution of individual transitions from 60 to other levels to the and
Einstein coeicients for 87 Rb. is calculated for = 300 K and = 4.2 K temperatures. Inset:
a zoom showing the background radiation inluence on the Rydberg lifetime.

where is the acceleration given by the Coulomb force

1 2
= .
4��0 ( 2
0)
2
(1.28)
�

In a quantum mechanical description, the lifetime of the Rydberg state is due to the
transition of the outer electron to inner states induced by modes of the electromagnetic
ield vacuum, i.e., spontaneous emission. The decay rate from a state � to a state can
be expressed by the Einstein coeicient:

2���3 4�� ���3

= | �� | = ∣⟨� ∣ ∣ ⟩∣
2 2
3�0 3ℎ 3
�� 2
(1.29)

where, �� is the dipole coupling and ��� /(2�) is the transition frequency. The spon-
taneous decay rate of the level � is calculated by summing over all possible decay chan-
nels.

Low Rydberg atoms

For a ∣ ⟩ state the only possible decay channels are to ∣ ′ ⟩ states. Figure 1.5 shows
the individual values of �� , numerically obtained from the evaluation of the dipole
matrix elements for the 60 Rubidium state. The spontaneous emission is dominated
by optical transitions to the lowest lying levels. The frequency of these short wave-
22 Properties of individual Rydberg atoms

length transitions does not change that much with . It turns out that the dependency
of �� on is determined from the dipole coupling, which depends on the radial over-
lap of the Rydberg wavefunction with the ground state wavefunction. Its value difers
signiicantly from zero only for short distances ≪ 0 . From equation (1.5) one inds
that the radial overlap of the wavefunctions is proportional to −3/2 . The spontaneous
decay rate of Rydberg atom thus scales as 3 . Summing up all the decay channels
yield = 0.004µs− 1 corresponding to a lifetime of 244 µs for the 60 state.

Circular atoms

The situation is quite diferent for circular atoms. The only possible spontaneous de-
cay channel is the transition to the next lower circular state by emission of a � + photon.
This explains why it takes a much longer time for a circular state to decay. For the 50
state, for instance, the strength of the dipole transition to the 49 state is 1706 0 . Us-
ing equation (1.29), this corresponds to a lifetime of about 28.6 ms. A classical calcula-
tion with equation (1.27) also leads to the same result [3]. It is not a surprise because,
as already discussed, the circular state is a perfect approximation of a classical Bohr
atom. The coupling between two nearby circular states scales as 2 while the transition
frequency scales as −3 . As a result, the lifetime of circular atoms scales as 5 .

1.3.2
Blackbody-radiation-induced decay
In the presence of blackbody radiation, in addition to the spontaneous emission, the
stimulated decay rate is enhanced by absorption or stimulated emission of a photon
due to stimulation. The stimulated decay rate is larger than the spontaneous emission
rate by a factor (�).
̄ The enhanced Einstein’s coeicients is

�� = (�)
̄ �� , (1.30)

where (�)
̄ is the mean number of photon per mode in free space at thermodynamic
equilibrium. It is given at a inite temperature by

1
(�)
̄ = ,
ℏ�/�� � −1
(1.31)

where � is the Boltzmann constant. The total decay rate for a state |�⟩ is then the sum
of the spontaneous emission and the stimulated decay rates

1
Γ� = =∑ �� [1 + (�)]
̄ +∑ (�)
̄ ,
�� �� (1.32)
� <� � >�

where �� is the radiative lifetime of the � state. The two summations run over all lower
lying and higher lying levels ∣ ⟩ which satisfy the selection rules described in sec-
tion 1.2.
The contributions from the stimulated transitions are also shown in Figure 1.5 at
room temperature (300 K) and at liquid helium temperature (4.2 K) for the 60 1/2 state.
The blackbody radiation slightly reduces the lifetime of the 60 Rubidium state, from
244 µs at 0 K to 240 µs at 4.2 K, and even further down to 99 µs at the room temperature.
 1.4 Rydberg atoms in external ields 23

In order to keep the Rydberg atoms long-lived, it is therefore necessary to put them in
a cryogenic environment.
The circular states are not an exception to this rule. The blackbody radiation not
only ampliies the decay rate to the lower circular state, it also opens new channels for
the stimulated transitions to higher lying levels, leading to a reduction of the circular
atom’s lifetime. This again requires to cool down the system in order to fully exploit
the long life of the circular states. We will further discuss on this in section 5.1.

1.4
Rydberg atoms in external ields

On the one hand, Rydberg atoms are quite sensitive to stray electric ields due to their
huge electric dipole matrix elements. On the other hand, in experiments where Ryd-
berg atoms are excited from a magnetically trapped cold atomic cloud, they are im-
mediately exposed to a strong magnetic ield. It is therefore important to understand
how Rydberg atoms behave under the presence of external ields. In the following, we
treat the Stark and Zeeman efects separately for the sake of simplicity.

1.4.1
Magnetic ield
The Zeeman Hamiltonian that describes the interaction of a Rydberg atom with an
external magnetic ield � reads
��
= −� ⋅ � = ( �� + � �) ⋅�,
�
ℏ
(1.33)

where �� is the Bohr magneton. The orbital factor � = 1 and the spin factor � ≈ 2
(an about 0.1% correction comes from quantum electrodynamics). Let’s choose the
quantization axis along �.
Depending on the relative strength of the spin-orbit interaction with respect to the
coupling with the magnetic ield, we can distinguish three situations: the anomalous
Zeeman efect for a weak magnetic ield, the Paschen-Back efect for a strong magnetic
ield, and the intermediate case.

1.4.1.1 Paschen-Back efect

When the external magnetic ield is strong enough, the angular momentum � and the
spin � are decoupled. Each of them precesses independently around the magnetic
ield. The interaction energy is thus simply given by

� = �� ( � + � �) � , (1.34)

where � is the component of �. In a strong magnetic ield, the ine structure is

modiied to

�� = � � (1.35)

since both � and � are polarized along .

24 Properties of individual Rydberg atoms

A dipole transition does not lip the electron spin. The spin part in the above equa-
tion does not change the transition spectrum. Therefore, we can neglect the spin part.
Equation (1.34) simpliies to

� = �� � � . (1.36)

1.4.1.2 Zeeman efect

In a weak magnetic ield, we treat irst the coupling of � and � into � = � + �, then take
the interaction with the magnetic ield as a perturbation. The Zeeman Hamiltonian
can be rewritten as
�� �⋅� �⋅�
= �( + )⋅�
�
ℏ �
�2
�
�2
(1.37)

A perturbative calculation yields the energy correction as

�(� + 1) − ( + 1) + ( + 1)
= �� = �� [ +( − �) ]
� � � � � �
2�(� + 1) � � (1.38)

where the term inside the square bracket is called the Landé- factor �. For a single
outer electron, is equal to 1/2.

1.4.1.3 Breit-Rabi formula

In general, one can be in an intermediate case where the Zeeman shift is of the same
order as the spin orbit splitting. One has to treat the two interactions at the same
time. The Breit-Rabi formula describes the magnetic ield dependence for � = 1/2 and
includes the hyperine structure. We can follow the same derivation for the spin-orbit
coupling only and obtain [133, page 20].

Δ �� Δ 4 �
� (�, �) =− + �� ± 1+ +
�� 2
2(2 + 1) � �
2 √ 2 +1
(1.39)

where Δ �� is the ine structure splitting and = ( � − 1)�� � /Δ �� is called the

ield strength parameter. The ± sign corresponds to sub-level with � = ± 1/2. This
formula includes both the weak ield Zeeman efect and Paschen-Back efect, and can
be applied for all and �.
Let us take = 60 multiplicity in a magnetic ield of 10 G as an example. The ine
structure level splitting can be calculated from the quantum defects listed in Table 1.1.
They are 460 MHz for the level and 53 MHz for the level, corresponding to =
0.03 and = 0.26 respectively. Therefore we can treat < 2 states in the weak ield
regime but, for the state, we enter the intermediate case. For ≥ 3, ≥ 19, the spin
is decoupled from the angular momentum. Equation (1.36) then well describes the
situation.
 1.4 Rydberg atoms in external ields 25

1.4.2

Electric ield

In an external static electric ield �, the Rydberg levels are coupled by the electric ield
via dipole transitions, leading to energy shifts. This is known as the Stark efect. The
interaction Hamiltonian reads

� = −� ⋅ � . (1.40)

Let us irst consider a Hydrogen atom in an electric ield. We also neglect for now
the ine structure. The electric ield breaks the spherical symmetry. is no longer a
good quantum number. However, the system is still cylindrically symmetric around
the direction of the electric ield that we choose as the quantization axis . The an-
gular projection quantum number � remains a good quantum number. The Stark
Hamiltonian can be expressed as

4�
=− = = 1,
0
√ 3
� � (1.41)

where � = − is the component of the dipole operator. Equation (1.21) has been
used to obtain the last equality. The electric ield thus couples only states with the
same � .

Hydrogen atom – Circular state

For high- Rydberg levels, and especially circular states, both the ine structure and
the quantum defects are negligible to a very good approximation. The Stark efect
lifts the degeneracy of the manifold. Even a small electric ield is enough to mix the
levels. A straightforward way to deal with the problem is to change to the parabolic
coordinates. The Schrödinger equation becomes separable and analytically solvable.
In addition to and , parabolic quantum numbers 1 and 2 are introduced. They
are non negative integers and related to and by

= 1 + 2 +| |+1. (1.42)

For convenience let’s note

= 2 − 1 = −2 1 −| |−1. (1.43)

For ixed and , there are − | | possible values of ranging from − − | | + 1 to

− | | − 1. It is enough to specify , and in order to deine an eigenstate. Thus
we use a ket ∣ ⟩ to represent the corresponding state. Its energy can be analytically
expressed using the perturbation calculation [125]

= (0) + (1) + (2) + ... (1.44)

26 Properties of individual Rydberg atoms

with
1
(0) =−
2 2
3
(1) =
2
(1.45)
1
(2) =− [17 2 −9 2 + 19 − 3 2 ] 4 2 ,
16
where = |�|. All quantities here are in atomic units.

Energy k =4
k =3
k =2 k =2
k =1 k =1
n = 51 k =0 k =0 k =0

k = ‒1 k = ‒1
k = ‒2 k = ‒2
k = ‒3 circular state
k = ‒4 n = 51

k =3
k =2
k =1 k =1
n = 50 k =0 k =0

k = ‒1 k = ‒1
k = ‒2
k = ‒3
circular state
n = 50

46 47 48 49 50 m
Figure 1.6 Energy levels for = 50 and = 51 states in an electric ield. The diagram only
shows states with ≥ 46.

Figure 1.6 illustrates the energy levels for high- states with = 50 and = 51.
From equation (1.45), one notes that the energy shift of a < 0 state is the same as
that of the corresponding > 0 state. Thus we can consider only the case ≥ 0.
The circular state ∣ , = − 1, = − 1⟩ in the spherical representation is un-
changed to the irst order in the presence of an electric ield, and is the same as the
circular state ∣ , = 0, = − 1⟩ in the parabolic representation. It does not have a
linear Stark shift. Instead, the second-order Stark efect quadratically lowers down its
energy level. For instance the ∣50 ⟩ state is shifted -2.03 MHz/(V/cm)2 .
States with | | ≠ −1 in the parabolic coordinates are linear combinations of states
with the same but diferent . They are also called elliptical states to distinguish them
with respect to the circular one. As an example, the two neighbors of the circular
state, ∣ , = − 1, = − 2⟩ and ∣ , = − 2, = − 2⟩, combine either symmetri-
 1.4 Rydberg atoms in external ields 27

cally into ∣ , = +1, = − 2⟩ state or anti-symmetrically into ∣ , = −1, = − 2⟩

state. For convenience, we denote them ∣ + ⟩ and | − ⟩ respectively. They are explic-
itly given by

∣ +⟩ = ∣ , = +1, = − 2⟩
1
= (∣ , = − 2, = − 2⟩ + ∣ , = − 1, = − 2⟩) ,
(1.46)
√2
and

| −⟩ = ∣ , = −1, = − 2⟩
1
= (∣ , = − 2, = − 2⟩ − ∣ , = − 1, = − 2⟩) .
(1.47)
√2
According to (1.45), elliptical states are linearly shifted by the Stark efect.

Low Rydberg atoms

a) b)
58, l > 3 60P3/2
17
Relative energy (GHz)

20
60P3/2
16
15 60P1/2 60P1/2

15
10
Relative energy (GHz)

57, l > 3
14
5 0 1 2 3 4 5 6 7
Electric field F (V/cm)
57F c)
0 0.0
60S1/2
Relative energy (GHz)

60S1/2
0.2
5
58D5/2 0.4
58D3/2
10 0.6

0.8
15
56, l > 3 1.0
0 1 2 3 4 5 6 7 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Electric field F (V/cm)
Electric field F (V/cm)

Figure 1.7 a) Stark map of 87 Rb for � = +1/2 states around 60 . The S, P, D and F levels are
outside the manifold due to their quantum defects. Levels with > 3 are quasi degenerate
and are linearly shifted by the electric ield. Zooms around b) 60P and c) 60S states are shown
on the right. The corresponding regions are marked by dashed rectangles in a). Under strong
electric ield, complex level anti-crossings are observed. The dashed red line in c) is a parabolic
it for 0 ≤ < 0.5 V/cm.

For low- Rydberg atoms, the quantum defects lift the degeneracy with the rest of
the multiplicity. The parabolic coordinate transformation is no longer necessary. For
a weak electric ield, i.e., when the Stark shift is much smaller than the level separa-
tion, one can treat the problem directly using the perturbation theory. Noting that the
dipole operator does not couple a state to itself, the Stark efect acts as a second-order
28 Properties of individual Rydberg atoms

Table 1.3 Properties of Rydberg atoms and their dependence.

Property low- state circular state

Binding energy −2 −2

Fine structure interval −3 0

Orbital radius 2 2

Dipole moment ⟨ ∣ ∣ ± 1⟩ 2 3/2

Radiative lifetime 3 5

Electric polarizability 7 6

van der Waals coeicient∗ 11 6

* see chapter 2 and section 5.3.

perturbation. One thus expects a quadratic shift of the energy levels

Δ � =� 2 (1.48)

where � is the electric polarizability and given by

|⟨ � � ∣ ∣ ′ ′ �′ ′ ⟩ |2
�
�= ∑
−
(1.49)
∣�′ �′ �′ ⟩≠∣� � �⟩ ��� �′ � ′ � ′

The summation is extended to all levels satisfying the selection rules. However, the
main contribution comes from levels closest in energy, which have a larger radial over-
lap. The lowest , i.e., ∣ ⟩ state for example, couples mainly to | ⟩ and | − 1 ⟩ levels.
Due to the efect of quantum defects, �� − �−1� > �� − �� > 0 (cf. Figure 1.3). As
a result, � < 0. The Stark efect lowers the energy. The numerator scales as 4 (square
of the dipole matrix element), while the energy diferences in the denominator scale
as −3 for adjacent Rydberg levels. As a result, the Stark shift scales as 7 .
The Stark diagram for levels around 60 with � = +1/2 is shown in Figure 1.7.
It is obtained by numerical diagonalization of the full Stark Hamiltonian. The electric
polarizability for the 60 level is found to be −89.9 MHz/(V/cm)2 . Electric ields
larger than 1 V/cm are enough to cause complex state mixing. Stronger quadratic
shifts for 60 and 58 states are also observed. The 57 state is slightly shifted out
of the manifold of ≥ 4. Inside, the levels are quasi degenerate and thus are linearly
shifted by the electric ield.

Summary

In this chapter, we briely reviewed the major properties of individual Rydberg atoms
both for low- and high- states. High- states are approximated by a Hydrogen atom,
especially for the case of circular states. For low- Rydberg states, the quantum defect
theory describes the energy correction due to the interaction with the alkali ionic core.
The quantum defects remove the degeneracy of the , , and states from the rest
of the manifold.
We studied the radiative lifetime of Rydberg atoms as well as their behaviors in an
 1.4 Rydberg atoms in external ields 29

external electric or magnetic ield. Throughout this chapter, numerical calculations

with the 60 state and the circular 50 state were used as examples. These two states
lie at the focus of this thesis. The scaling laws for the properties of Rydberg atoms
were discussed and are summarized in Table 1.3.
30 Properties of individual Rydberg atoms
Interacting Rydberg
atoms 2
2.1
Two antennas in communication

Having huge dipole matrix elements, Rydberg atoms strongly interact with each other.
This dipole-dipole interaction has been intensively studied and applied in the ield of
quantum information processing. It leads to non-trivial phenomena and is at the heart
of this thesis. The dipole interaction operator is the combination of dipole operators
acting on each atom

1 � �
�� ( )= [�1 ⋅ �2 − 3(�1 ⋅ )(�2 ⋅ )]
4��0 3
2 � �
(2.1)
= [�1 ⋅ �2 − 3(�1 ⋅ )(�2 ⋅ )] ,
4��0 3

where = |�| is the distance between the two atoms. The indices indicate which atom
the operator acts on. Here, the distance between the two atoms is treated classically
assuming that the spatial spread of the atomic wave-packet for each atom is much
smaller than the distance between them.
The interaction can be seen as resulting from the simultaneous exchange of virtual
photons between the two atoms. Strictly speaking one has to take into account in prin-
ciple the retardation efect accounting for the propagation of the photons. However,
the typical spacing between the Rydberg atoms in these experiments ranges from a
few to several hundreds micrometers. It is much smaller than the photon wavelength,
in the mm range. The retardation efect can be neglected and thus equation (2.1) is
valid.
For simplicity, we choose the axis connecting the two atoms as the quantization
axis and express the position operators �1 and �2 in terms of spherical harmonics (cf.
section 1.2). The dipole interaction operator can then be rewritten as
2
�� ( )=− ( 1−1 1+1 + +2 1 1)
+1 −1 0 0
3�0 3 1 2 1 1 (2.2)

From this expression, it is clear that the interaction operator preserves the total mag-
netic quantum number = 1 + 2 .
32 Interacting Rydberg atoms

In practice, we use a numerical approach to calculate the interaction between two

Rydberg atoms. The Hilbert space is truncated to important terms due to limited
memory and computation power. The interaction Hamiltonian is then constructed
and directly diagonalized to ind the interaction energy for each inter-atomic distance
. In order to discuss the physics of the interaction, we will make use of perturbation
theory. In the following let’s limit the discussions on the case = 0. In section 5.3, we
will discuss in details the dipole interaction between two circular atoms, which leads
to interesting efects.

2.2
A pair of atoms in the same state

For a pair of two atoms in the same state denoted | ⟩, the dipole interaction operator
in general acts as an second-order perturbation, coupling to intermediate pair states
∣ ⟩. The resulting interaction energy has the form

⟨ ∣ �� ∣⟩⟨ ∣ �� ∣ ⟩
=∑ = ,
6,�−�
��
2 �− �− 6
(2.3)
∣��⟩ �

corresponding to a van der Waals interaction with 6,�−� is the van der Waals coei-
cient. In the above equation, � is the energy of the single Rydberg atom �.
The situation is bit diferent when one of the coupled state ∣ ⟩ is nearly degenerate,
i.e., ⟨ ∣ �� ∣ ⟩ / 3 ≫ ∣2 � − � − � ∣. If | ⟩ = ∣ ⟩ , he relevant subspace consists of
two states | ⟩ and | ⟩. The corresponding Hamiltonian reads

| ⟩⟩ | ⟩
| ⟩⎛ 2 �
⟨�� ∣ ��� ∣ ��⟩
= ⎜ ⎞
⎟
⎜ ⟨�� ∣ ��� ∣ ��⟩ ⎟.
�3
| ⟩⎝ 2
��−�� (2.4)
3 � � ⎠

Its eigenstates are a symmetric combination (| ⟩ + | ⟩)/√2 and an anti-symmetric

combination (| ⟩ − | ⟩)/√2 of the two levels. The corresponding energy shifts are
given by

⟨ ∣ �� ∣ ⟩
Δ =± =± ,
3,�−�
�� 3 3
(2.5)

where 3,�−� = ⟨ ∣ �� ∣ ⟩ and the plus sign corresponds to the antisymmetric com-
bination of | ⟩ and | ⟩.
If | ⟩ ≠ ∣ ⟩, there are three nearly degenerate states | ⟩, ∣ ⟩ and ∣ ⟩. The symmet-
ric and anti-symmetric combinations of ∣ ⟩ and ∣ ⟩ are ∣ ⟩ = (∣ ⟩ + ∣ ⟩)/√2 and
∣ ⟩ = (∣ ⟩ − ∣ ⟩)/√2. One notes that | ⟩ does not couple to ∣ ⟩ by �� . The rele-
vant subspace thus consists of two states | ⟩ and ∣ ⟩. Similarly to the case of | ⟩ = ∣ ⟩,
one gets two eigenstates (| ⟩ ∓ ∣ ⟩)/√2 whose energy shifts are

⟨ ∣ �� ∣ ⟩
Δ =± =± ,
3,�−�
�� 3 3
(2.6)
 2.3 A pair of atoms in the same state 33

15 10 3
10 2 60S - 60S interaction
1 .0
10 1 VdW fit
0 .9

Interaction energy (GHz)

10 0
1 0 -1

10 1 0 -2
1 0 -3 0 .8
Interaction energy (GHz)

1 0 -4
1 0 -5
1 0 -6
0 .7
1 0 -7
5 1 0 -8 0 .6
1 0 -9
-1 0
10
1 0 -1 1 0 0 .5
10 10 1 10 2
R (µm) 60S1/2 60S1/2
0 0 .4
60P3/2 59P3/2
0 .3
60P1/2 59P1/2

5 60P1/2 59P3/2 60P3/2 59P1/2 0 .2

0 .1

10
10 0 10 1 10 2
R (μm)
Figure 2.1 The energy shift of ∣60 ; 60 ⟩ pair state due to their van der Waals interaction. Also
shown are the nearby pair states. The color code represent the square of the projection on the
unperturbed ∣60 ; 60 ⟩ state. Inset plots the energy shift in log-log scale. At short distance,
∣60 ; 60 ⟩ strongly couples to |60 ; 59 ⟩ states, the interaction gradually changes to a resonant
dipole interaction.

where 3,�−� = √2 ⟨ ∣ �� ∣ ⟩. These special situations are known as Rydberg Förster

resonance [104].

− interaction

In the case of the ∣ ; ⟩ state, the dominant term in (2.3) is the coupling with the
| ; − 1 ⟩ pair state. Since �� − �−1� > �� − �� (cf. Figure 1.3), the denomi-
nator in (2.3) is positive. The − interaction is thus always repulsive. Numerical
calculation for the 60 − 60 pair gives 6,60�−60� = 137.6(1) GHz.µm6 . Figure 2.1
depicts the numerically computed energy shift for the ∣60 ; 60 ⟩ pair state. At dis-
tances larger than 3 µm, the interaction is well itted by a van der Waals potential. At
shorter distances, the dipole coupling is as strong as the energy diference of the two
pair states ∣ ; ⟩ and | ; − 1 ⟩ (∼ 2 GHz). The interaction gradually changes to
1/ 3 behavior. From equation (2.3), one can show that the van der Waals interaction
scales as 11 .
34 Interacting Rydberg atoms

2.3
An atom pair involving two diferent states

For an atom pair in two diferent states and , there are two degenerate pair states | ⟩
and | ⟩. According to the selection rules, ⟨ ∣ ef ∣ ⟩ = ⟨ ∣ ef ∣ ⟩ = 0. Formally,
one writes an efective Hamiltonian ef [134] for a two level approximation including
second-order coupling to other states. Its matrix elements are

�� =⟨ ∣ ef ∣ ⟩=⟨ ∣ ef ∣ ⟩ , (2.7)

and

�� =⟨ ∣ ef ∣ ⟩=⟨ ∣ ef ∣ ⟩ . (2.8)

The interaction Hamiltonian is expressed as

| ⟩ | ⟩
| ⟩
= ( �� ��
).
| ⟩
(2.9)
�� ��
ef

The diagonal terms are the direct interaction of a pair state to itself, which is a second-
order perturbation through the coupling to intermediate pair states ∣ ⟩. They thus
correspond to van der Waals interactions

⟨ ∣ �� ∣
⟩⟨ ∣ �� ∣ ⟩
=∑ = ,
6,�−�
�+ �− −
�� 6
(2.10)
∣��⟩ � �

where 6,�−� is the corresponding van der Waals coeicient.

a) b)
e f f e
a b b a a b b a

c d d c c d d c
Figure 2.2 Scheme for calculating the dipole interaction between and states. The direct
(diagonal) terms are represented by the dashed blue arrows and the exchange (of diagonal)
terms by the solid red arrows. The direct interaction is the second-order perturbation coupling
to intermediate pair states of and . The exchange interaction can be either a a) irst-order or
b) higher order perturbation. In b) the pair state ∣ ⟩ is an intermediate state for the coupling
between ∣ ⟩ and ∣ ⟩.

The of-diagonal term �� corresponds to an interaction where the two atoms ex-
change their excitations. When the − transition is dipole allowed, e.g., a − ′
transition, the exchange interaction is a direct coupling of ∣ ⟩ and ∣ ⟩ (Figure 2.2a.
It varies as a 1/ 3 dipole potential. Otherwise, depending on the dipole coupling
 2.3 An atom pair involving two diferent states 35

between | ⟩ and ∣ ⟩, the exchange interaction can be an indirect coupling of second

(scaling as 1/ 6 ) or higher order as demonstrated in Figure 2.2 b.
In general the direct interaction shifts the energy of the two levels in the same way
while the exchange interaction breaks the degeneracy, splitting them into symmetric
and antisymmetric combinations of the two levels. Their total energy shifts are

Δ �� = �� ± �� , (2.11)

where the minus sign corresponds to the symmetric combination.

− ′ interaction

2.0 0.500
10 1

10 0
0.475
Interaction energy (GHz)

10 -1
1.5
Interaction energy (GHz)

10 -2 0.450
10 -3

1.0 10 -4 0.425
Direct int.
10 -5 Exchange int.
10 -6
1/R 3 fit 0.400
0.5 -7
1/R 6 fit
10
1 2 3 4 5 10 20 0.375
R (μm)

0.0 0.350

0.325
0.5
0.300
1 2 3 4 5 10 20
R (μm)

Figure 2.3 Interaction energy for 60 − 60 3/2 pair of atoms. The lower branch corresponds
to the symmetric level while the upper one corresponds to the anti-symmetric level. The color
code shows the projection on the initial pair state ∣ ; ′ ⟩ squared. The dashed line represents
the direct energy shift. The inset shows in log-log scale the direct and the exchange interaction.

This is a rather simple case. The exchange interaction directly couples ∣ ; ′ ⟩ and
∣ ′ ; ⟩ state by a resonant (irst-order) dipole interaction

=⟨ ; ∣ �� ∣ ; ⟩= ,
′ ′ 3,��−�′ �
��−�′ � 3
(2.12)

where 3,��−�′ � is a proportionality coeicient. Its dependency originates from the

dipole matrix elements, which scale as 2 each, leading to a scaling as 4 .
The direct interaction comes from the coupling to the intermediate levels ∣ ″ ; ‴ ⟩:
∣ ; ′ ⟩ ↔ ∣ ″ ; ‴ ⟩ ↔ ∣ ; ′ ⟩. From (2.10) one can easily ind that it scales as 11 .
Equation (2.11) takes the form

Δ = ± .
6,��−�′ � 3,��−�′ �
��,��−�′ � 6 3
(2.13)
36 Interacting Rydberg atoms

Figure 2.3 depicts the numerical result for 60 − 60 3/2 pair. It is calculated from
a subspace of about 500 pair states. The color code represents the probability of being
in ∣60 ; 60 3/2 ⟩ pair state. At long distances it corresponds to 0.5 for each branch as
expected. Getting closer, the color is degraded, indicating the contamination by other
levels. Fitting the energy shift for each branch with the potential described in (2.13)
yields 6,60�−60�3/2 = 7.976(1) GHz.µm6 and 3,60�−60�3/2 = 4.411(0) GHz.�m3 . The
van der Waals terms is shown by the dashed curve on the graph. Inset plots in log-log
scale the mean energy and half the energy diference of the two branches, correspond-
ing to the direct and exchange interactions respectively. A 1/ 6 and a 1/ 3 behaviors
are clearly observed even down to distances smaller than 2 �m. At distance larger than
4 �m, the van der Waals shift is more than two orders of magnitude smaller than the
dipole shift, and thus can be ignored.

Summary

In this chapter, we established the dipole-dipole interaction between a pair of Rydberg

atoms under diferent situations. We particularly focused on the case where one of the
atoms is in an state. However, the concepts discussed here can be extended to the
general case.
For an atom pair in diferent states, the exchange interaction combines symmet-
rically or anti-symmetrically the two pair states ∣ ; ′ ( )⟩ and ∣ ′ ( ); ⟩, lifting
their degeneracy. The level splitting is proportional to either 1/ 3 ( ′ state) or 1/ 6
( ′ state) depending on whether the transition between the two states is dipole al-
lowed or not. Two Rydberg atoms at distance larger than 3 µm interact with each
other by a van der Waals potential. We also performed numerical calculations for the
interaction of a 60 Rydberg atom with another one in the 60 nearby state.
 II.
Towards quantum simulation
with low- Rydberg atoms
Superconducting atom
chip experiment 3
The prerequisite in most of quantum simulation experiments is to preserve the coher-
ence of the system for a long time. Building a quantum simulator with Rydberg atoms
on a superconducting atom chip imposes two main requirements: a high-level control
of the stray electric ields and the use of cold/ultra cold atoms. The former comes from
the high sensitivity of Rydberg atoms to electric ield. The latter is meant to minimize
the atomic motion, which might eventually map into time-dependent Stark and/or
Zeeman shifts. In addition, a cold and dense atomic cloud helps to enter the strong
Rydberg-Rydberg interaction regime.

Our experiment is dedicated to study the excitation of strongly interacting Rydberg

atoms out of a cold Rubidium cloud near a superconducting atom chip. This is a rather
complex experiment, in which we implement a superconducting atom chip with laser
cooling and magnetic trapping techniques for 87 Rb inside a cryostat. The Rydberg
excitation and detection are also performed in the cryogenic environment. The ex-
periment is designed carefully in such a way that every component we put inside the
cryostat has to consume a minimum amount of liquid Helium. The list includes the
atom chip, the superconducting coils creating necessary bias magnetic ields for atom
trapping and cooling, the electrodes for ield-ionization system and the ion-counter
channeltron. We recall here the main features of the experimental setup, which is
divided into two parts: the preparation of ultra-cold atoms and the Rydberg excita-
tion/detection. More details on the chip fabrication and characterization can be found
in Raul Celistrino’s thesis [135].

At the end of this chapter, we briely represent our irst investigations to demon-
strate that Rydberg atoms are fully compatible with the atom chip, i.e., coherent ma-
nipulation of Rydberg atoms is feasible in the vicinity of the chip with a good control
of stray electric ields. The experiment is performed with a low Rydberg density such
that Rydberg-Rydberg interaction is negligible. This is the topic of Carla Hermann’s
thesis, where one can ind a full description of the experiment [136].
40 Superconducting atom chip experiment

3.1
Ultra-cold atom source

3.1.1
Cryostat

As discussed in section 1.3, the Rydberg atom’s lifetime strongly depends on the en-
vironment temperature. Of course, short lifetimes might be not a problem for certain
experiments where all interesting dynamics happen on a short time scale. However,
in order to fully exploit the long lifetime property of the Rydberg atoms, putting them
in a cryogenic environment is necessary. In addition, the operation of the chip under
non-superconducting state would turn the chip itself into a hot body right at the po-
sition of the Rydberg atoms due to the Joule heating efect. A superconducting atom
chip suppresses this efect.
Our experiment uses a cryostat, which is schematically depicted in Figure 3.1. The
heart of the experiment, where all the science happens is kept at 4.2 K temperature by
a 4 He shield, which is a copper shell in direct contact with a liquid 4 He reservoir. It is
in turn thermally shielded from 300 K blackbody radiation by an intermediate stage at
liquid nitrogen temperature, 77 K. The cryogenic stages are irmly mounted inside an
external cylindrical shell, which provides a good vacuum. A layer of lead, which be-
comes superconducting at a temperature lower than 7 K, is installed around the inner
wall of the He shield. It screens any external magnetic ield luctuation. Moreover, fast
variations of the bias magnetic ields during the experiment are necessary. Without
the superconducting lead shield, eddy currents would be induced on the outer copper
thermal shields, preventing fast variations of the bias magnetic ields.
The He shield, with a few liters volume is suiciently large to accommodate the
chip mount, the magnetic bias coils as well as the Rydberg detection system, which
are shown in Figure 3.1. The chip wires and the bias coils are all superconducting to
prevent heating and power dissipation. Several windows are opened on the walls of
the cryostat (and the thermal shields) for optical access. Both laser and imaging sys-
tems are outside the cryostat. The size of the windows limits the numerical aperture
for the collection lenses, but large windows expose the Rydberg atoms directly to the
300 K blackbody radiation. Therefore the windows are chosen to be 6 cm in diameter
as a compromise.
On the one hand, the use of the cryostat sets up some constraints on optical access
as well as on the complexity of experimental operation. On the other hand, it helps
to have a high vacuum without much efort. Residual gases are strongly adsorbed
on cold surfaces. A pressure smaller than 10−10 mbar* can be obtained without time
consuming baking of the system. Under such a pressure, the losses due to background
gas collisions are strongly reduced and the trapping lifetime is thus improved, of the
order of minutes [138, 139].

*
Expected to be lower than 10−13 mbar [137] but no direct measurement is available with our current
setup.
 3.1 Ultra-cold atom source 41

a) b)

Liq. N2

Bottom of
N2 reservoir Bottom of
To turbo 4He reservoir
pump External
cylinder

Liq. Channeltron Bias coils

4He
Electrodes
Chip mount

z
y
By bias coil x
QUAD coil

c)

4
He shield

N2 shield

Bz bias coils

Bx bias coils

Optical
windows
z
Atomic beam
y
x from 2D-MOT
Chip position

Figure 3.1 Sketch of the cryogenic setup. a) A vertical cut shows the construction of the cryo-
stat. The liquid 4 He and liquid Nitrogen reservoirs are indicated by fake colors. A slow atomic
beam from a 2D-MOT is injected inside the cryostat from the bottom. b) Schematic view of
the heart of the experiment. The chip faces the direction. The bias coils (dark green) and
detection electrodes (blue) are also represented. The Helium and Nitrogen shields as well as
the exterior shell are not shown. c) A closer view shows only the bias coils (green) and the
“QUAD” coil (purple), which generate the quadrupole magnetic ield for an on-chip MOT.
The chip position is marked by the blue rectangle covering the � bias coil and the QUAD coil.
The red zone indicates the place where the Rubidium atoms are trapped. The current low in
the QUAD coil is marked by the arrow.
42 Superconducting atom chip experiment

3.1.2
Atom chip

K L J
IZ IU
IU
100 μm
1400 μm

300 μm
70 μm 20 μm
z
x
y 1415 μm
IZ

100 μm

M H G

Figure 3.2 Scheme of the superconducting atom chip. The letters label the current in-
put/output pads on the chip. The fake colors are to distinguish between the chip wires: green
for the U-shaped wire, orange for the Z-shaped wire and blue for the RF wire. Directions of
the currents are also indicated.

The atom chip used in our current experiment is rather simple and is depicted
in Figure 3.2. The chip operation is based on three wires: a U-shaped wire (LJ), a
Z-shaped wire (LG) and a straight wire (KM) as depicted in Figure 3.2.
The U-shaped or Z-shaped wire is a simpliied version of a H-shaped wire, made
up of a straight current crossing two parallel currents (Figure 3.3a). By passing a cur-
rent through the Z-shaped wire, the magnetic ield created by the segment along the
direction, in combination with a bias ield along the direction � forms a quadrupole
ield in the plane (Figure 3.3b). The currents in the two parallel arms low in
the same direction. They create a magnetic ield in the direction, which exhibits
a nonzero minimum near the center of the quadrupole ield (marked by the yellow
dots, Figure 3.3c). The total magnetic ield forms an Iofe-Pritchard magnetic trap.
Similarly, the U-shaped wire also has a quadrupole ield in the plane. The cur-
rents in the two parallel arms low in counter directions. As a result, the total magnetic
ield has a zero minimum near the center of the quadrupole ield (Figure 3.3d). Such
a magnetic ield is suitable for a 3D magneto-optical trap (mirror MOT).
Numerical calculations of the ields created by our particular chip can be found
in Raul Celistrino’s thesis [135]. For either the U-shaped or the Z-shaped wire, the
distance to the chip of the trap center is approximately given by

�0
= .
0
2�
(3.1)
�
 3.1 Ultra-cold atom source 43

a)
I I I

I z I
x
y

y y y
b)

z z z
+ =
y y
z z
I x
I x

c) y d) y
I

x x
y
x
z
I
I I

Figure 3.3 Magnetic ield created by the chip wires. a) The ( )-shaped wire is a simpliied
version of a H-shaped wire consisting of a straight current along the direction and a pair of
parallel currents along the direction. b) A quadrupole ield is formed by the straight current
in superposition with a bias ield. The pair of currents can be either c) in the same direction or
d) in opposite directions. In modulus, the total ield can accordingly have a zero or non zero
minimum at the positions marked by yellow dots. The wires under consideration are shown
in red while the corresponding ( )-shaped wire is shown in green. In b) we plot only the
Z-shaped wire but the situation is the same for the U-shaped wire.
44 Superconducting atom chip experiment

σ‒ σ+

σ‒

σ‒ σ+ x y

σ‒
Figure 3.4 Mirror MOT. Two counter-propagating beams are sent parallel to the chip while
the other two hit the chip at 45 degrees. The relection of the latter adds up two beams with
inverted helicities, thus correctly restoring the six-beam coniguration.

while the gradient of the quadrupole ield at the trap center is

2� 2 �0
∣ ′ ( 0 )∣ = �
= .
�0 2� 2
(3.2)
0

This is an important feature of the trap geometry. The closer it is to the chip, the tighter
the coninement in the plane. Therefore at short distances from the chip, the trap is
elongated along the direction, taking a cigar shape.
The KM wire is used for generating the radio frequency during the RF-induced
evaporative cooling of the atoms in the magnetic trap.
All the chip wires are 2 µm thick, made of Niobium (Nb) deposited on a silicon
substrate. Nb wires at 4.2 K are superconducting. The critical current is measured
to be 3.6 A for the thin Z-shaped wire, while it is up to 7.5 A for the thick U-shaped
wire. The fabrication of the chip has been developed and realized in our group. A
detailed discussion can be found in [135]. During the cooling of the chip from room
temperature, the earth and stray magnetic ields are compensated to prevent the chip
from trapping residual ields. Otherwise, magnetic vortices can be formed, leading to
a deformation of the magnetic trap, and even to the formation of uncontrolled on-chip
local magnetic traps.
The chip is covered with a 200 nm thick layer of gold. A special coniguration of
four cooling laser beams making use of the high relectivity of the chip gold surface
allows us to restore the standard six-beam coniguration of a 3D-MOT (Figure 3.4).
Such a MOT is called a mirror-MOT.

3.1.3
Atom imaging

The cryostat reduces our freedom in coniguring the imaging system. Here we record
the image of the atomic cloud both from the front and the size directions as schemat-
 3.1 Ultra-cold atom source 45

Cryostat - external shell

Optical window Chip

~7°
L = lens
L M = mirror
Side probe
laser
M

M
Front probe
laser M
L L

Reflected cloud
Chip

~7° CCD

CCD PhotonMax
Side probe
laser
MicroMax

Figure 3.5 Scheme of the imaging system. The probe beam can be sent either nearly perpen-
dicular (front probe) or parallel (side probe) to the chip surface. The relected beams from the
chip surface are collected on CCD cameras. The objective lenses are placed as close to the chip
as possible. The atomic cloud casts a shadow on the relected beam. For the side probe, the
relection of the chip surface adds another image of the cloud as illustrated in the inset.

Figure 3.6 A cold cloud of Rubidium atom after 16.5 ms time of light imaged by the side
probe beam. We get two images: one is the direct image (right) and the other originates from
the relection on the chip.

ically depicted in Figure 3.5. The probe beams are sent either (nearly) perpendicular
to the chip surface (front imaging), or at an angle of about 7° from the chip surface
(side imaging). The objective lenses are installed right at the external cylinder of the
cryostat in a way such that they do not disturb other laser beams but still allow us to
collect as much as possible the light scattered from the atomic cloud .
46 Superconducting atom chip experiment

The atomic cloud can be imaged either by collecting its luorescent light or by mea-
suring the change of the probe beam intensity due to the absorption of the cloud.
However, the latter method gives a more precise estimation of the number of atoms.
Well below the saturation of the atomic transition, the optical density of the cloud
at a point ( � , � ) is found by taking the natural logarithm of the transmission

� ( �, �)
( �, �) = − ln ,
�( �, �)
(3.3)

where � ( � , � ) and � ( � , � ) are the light intensities of the probe beam at point ( � , � )
with and without the absorption due to atoms. These intensities are measured directly
with a CCD camera. Here we deine the coordinates ( � , � , � ) such that � is the
propagation direction of the probe beam. The Beer-Lambert ’s law relates the
with the column density (̄ � , � ) = ∫ ( � , � , � ) d � , where ( � , � , � ) is the atomic
density

( �, �) = �0 (̄ � , �) . (3.4)

Thus

1 � ( �, �)
(̄ � , �) = ∫ ( �, �, �) d � =− ln .
�0 �( �, �)
(3.5)

In the above equation, �0 is the resonant scattering cross-section. Cold atoms in the
magnetic trap are prepared in the 5 1/2 , = 2, � = +2 state. The probe beam is
adjusted to drive the � + cycling transition, i.e., between the 5 1/2 , = 2, � = +2 and
the 5 3/2 , ′ = 3, � = +3 states. The corresponding cross-section is given by

3�2
�0 = ,
2�
(3.6)

where � = 780 nm is the transition wavelength.

In practice, the probe polarization is not perfectly circular. In addition, the inter-
ference between the probe beam and its relection from the chip surface modulates
the light intensity. Due to these efects, the apparent cross-section �0 is reduced by a
factor �abs , which is experimentally calibrated [135, 140].
For the side imaging, �abs is measured to be about 2.06 ± 0.1. As shown in Fig-
ure 3.6, we get in general two images of the cloud: one from the direct image of the
cloud and the other corresponding to the relection on the chip surface. Half of the
distance between the two images is equal to the distance of the atomic cloud from the
chip surface.

3.1.4
Laser system

A laser system is mandatory for trapping and cooling the atoms. In our experiment,
we implement standard techniques for trapping and cooling 87 Rb atoms. A good con-
trol of the power, frequency and polarization of the lasers determines the success of the
 3.1 Ultra-cold atom source 47

> 20 MHz
F'= 3
5 P3/2 266.7 MHz
F'= 2
156.9 MHz
F'= 1
72.2 MHz

λ = 780.24 nm
F'= 0
D2 line

Zeeman pumper
Cooling beam
Repumper
5 S1/2 F' = 2

Probe
F'= 1
λ = 794.98 nm
D1 line

F= 2
5 S1/2
≈ 6.834 GHz
F= 1

Figure 3.7 Hyperine structure of 87 Rb. The cycling transitions are shown for the cooling laser,
the repumper, the optical Zeeman pumper as well as for the probe beams.

experiment. The cooling laser comes from a commercial TOPTICA Master Oscillator
Power Ampliier (MOPA-TA 110) laser. It is frequency-ofset locked to the excitation
red laser, which in turn is frequency-stabilized to a Fabry-Perot cavity. The cavity is
locked to a transition line of 87 Rb (see Appendix B for optical scheme). The spectrum
linewidth of the laser is of a few tens of kHz. The cooling cycle makes use of the D2
line = 2 − ′ = 3 transition. A repumper is required to bring back the atoms scat-
tered out of the cooling cycle. Part of the cooling laser is extracted and frequency ofset
to serve as probe beams or to perform Zeeman optical pumping. The corresponding
transitions are shown in Figure 3.7. All the laser beams are guided to the cryostat by
coupling to polarization maintaining ibers. Appendix B gives more details on the
laser frequency stabilization scheme and laser distribution.

3.1.5
Trapping and cooling of atoms
Figure 3.8 demonstrates the timing of a typical cooling and magnetic trapping se-
quence where each parameter is well controlled and optimized. It consists in two main
stages: trapping and cooling the atoms in on-chip mirror MOTs followed by trapping
in a magnetic trap, where RF evaporative cooling allows us to reach a sub-µK temper-
ature.

On-chip mirror MOTs

1. Loading a QUAD-MOT. In a typical sequence of the experiment, Rubidium atoms

are irst cooled down in a 2D-MOT. They forms a slow beam of atoms propagating up
to a few mm in front of the chip. The atoms are then captured in an on-chip mirror
 1 2 3 4 5 6 7
On-chip mirror MOTs Magnetic trap
U-MOT far U-MOT close Magnetic trap ON
QUAD-MOT U-MOT Optical Magnetic trap Magnetic trap Rydberg Absorption
RF evaporative cooling
loading comp. Molasses pumping compression decompression excitation imaging

4A
0A 0A
IQ
5A
0.5 A
0A 0A 0A
IU
2A 1.5 A 3.3 A
0A
IZ 0 A
45.8 G
7.3 G
9.16 G 3.3 G 27 mG 13.7 G
B0z 0G
1.8 G
5.4 G
0.21 G 0.76 G 0.29 G 0G 0G
B0x 75 mG
Superconducting atom chip experiment

–0.65 G 0G 0G 0G 0G
B0y –2.61 G
–2.45 G –1.63 G
δ –3.3 Γ
–9 Γ
Imaster 27 mW/cm2
0 mW/cm2

2.4 s 20 ms 20 ms 0.9 ms 0.9 ms 150 ms 0-5s 300 ms 0-1s 120 ms

Figure 3.8 A typical experimental sequence. The timing and values of each parameters are carefully optimized as shown. Each step is numbered and
discussed in the main text.
48
 3.1 Ultra-cold atom source 49

MOT.
A large magnetic ield gradient is necessary to eiciently capture the atoms [141].
According to equation (3.2), one would increase the current . This is obtained by
using the lower part of a rectangular superconducting coil – the QUAD coil – to serve
as a big U-shaped wire (Figure 3.1 c). The current is thus multiplied by the number of
turns in the coil. About 108 atoms at about 400 µK are trapped in the QUAD-MOT.

2. U-MOT. The trap is then transfered to another MOT created by the U-shaped wire
(U-MOT far). A higher magnetic ield gradient, and thus a larger restoring light force,
allows us to eiciently cool down the atoms. We therefore reduce the current in the
U-shaped wire to increase the ield gradient. In doing so, we also compress the size
of the atomic cloud and approaching the chip (U-MOT close). A temperature of about
40 µK is obtained.

3. Optical molasses. At this stage, the currents in the bias coils are adjusted to precisely
cancel the residual magnetic ield. In the meantime, the cooling laser power is gradu-
ally reduced and the cooling laser detuning is increased to far of resonance. This 3D
optical molasses technique allows us to form a viscous coninement of the atoms and
further cool them down to about 13 µK. The atoms are about 700 µm away from the
chip.

Iofe Pritchard magnetic trap

The cooling with a MOT is intrinsically based on the optical transition cycles of the
atoms, and is thus limited by the recoil temperature. To further cool the atoms, we
implement the magnetic ield trapping in complement with the RF evaporative cool-
ing.

4. Optical Zeeman pumping. The magnetic trap is a weak-ield seeker trap. The trapping
potential is given by

� = � � �� ∣ ( , , )∣ , (3.7)

where � is the hyperine-structure Landé factor. Thus 5 1/2 , = 2, � = +1 and

� = +2 are the two trapped states. We aim to trap the latter. An intermediate stage
optical pumping is introduced to bring the atoms into the Zeeman sub-level 5 1/2 , =
2, � = +2. It increases the number of atoms transferred into the magnetic trap by a
factor ∼ 3.

5. Magnetic trap. Note that the magnetic ield coniguration of the U-MOT cannot be
used for magnetic trapping due to the presence of a zero ield at the center of the trap,
leading to spin-lip Majorana losses [142, 143]. The use of the Z-shaped wire ensures
that the ield at the trap center points along the direction, removing the zero ield at
trap center. Furthermore, the bottom of the trap is risen up by a bias ield to prevent
atomic transitions to un-trapped states induced by low-frequency noises near the chip
[138, 144].
50 Superconducting atom chip experiment

Energy

RF RF

RF RF

Position

Figure 3.9 RF-induced evaporative cooling. The RF knife induces the transition to un-trapped
states, through the � = +1 state. Gradually lowering the RF knife kicks out the hottest atoms,
truncating the hot tail of the Boltzmann distribution. As a result, the temperature decreases
after thermalization.

Figure 3.10 Rubidium cloud after 16.5 ms time-of-light for diferent temperatures. The tem-
perature is controlled by the inal value of the evaporative cooling RF. The images reveal di-
rectly the momentum distribution. The three images correspond to a thermal cloud, a BEC,
and a quasi pure BEC respectively

6. Evaporative cooling. For evaporating cooling, a RF knife is used as demonstrated in

Figure 3.9b to remove the hottest atoms out of the trap, carrying with them a signiicant
amount of kinetic energy, that helps to increase the phase space density. An important
point is that the re-thermalization of the trapped cloud should be faster than the trap
truncation [145]. Otherwise one removes also the “useful” cold atoms. In order to
enter this runaway regime, we adiabatically compress the trap after the transfer in
order to increase the collision rate. The evaporative cooling occurs at about 100 µm
away from the chip. At the end of the evaporation, we can manage to bring the atomic
cloud down to its quantum degeneracy (Bose-Einstein condensate BEC) [146, 147] .

7. Decompressed magnetic trap. By changing the currents in the Z-shaped wire as well
as in the bias coils, we can decompress and move the trap to the position of interest
for the Rydberg excitation. A typical trap used throughout this thesis is about 210 µm
 3.2 Rydberg excitation 51

mj = +1/2
mj = –1/2
60S1/2
480 nm, σ –

Ωb

δ 5P3/2
mF' = +3
F' = 3
mF' = 0
mF' = –3 Ωr
780 nm, σ+

5S1/2
mF = +2 F=2
mF = 0
mF = –2

Figure 3.11 Rydberg excitation scheme. Trapped atoms in the 5 1/2 , � = +2 are brought to
the Rydberg state 60 1/2 , � = +1/2 by two-photon transition, detuned by � from the interme-
diate level 5 3/2 , � = +3. The red laser is � + polarized while the blue laser is � − polarized.

away from the chip surface. The trap frequencies are measured to be 47 Hz, 244 Hz
and 262 Hz along the , and directions respectively (see Appendix C).

Finally the atomic cloud is imaged by absorption imaging. Figure 3.10 shows im-
ages of three clouds taken for diferent inal heights of the RF knife, with a 16.5 ms
time-of-light. They correspond to a thermal cloud, a BEC and a quasi-pure BEC re-
spectively. The cloud in the last image contains about 30 000 atoms at 450 µm from
the chip.

3.2
Rydberg excitation

3.2.1
Two-photon excitation
From a cold atomic cloud trapped on the atom chip, Rubidium atoms are brought
to their Rydberg states by laser excitation. In the presence of the magnetic ield, the
dipole-dipole interaction between the Rydberg atoms generally strongly depends on
the orientation of the atoms with respect to the magnetic ield. However, this angular
dependence for the 60 − 60 interaction is almost negligible thanks to the isotropy of
the 60 orbital. To keep things simple, we concentrate on the 60 1/2 , � = +1/2 level.
The transition from the trapped state 5 1/2 , = 2, � = +2 to 60 1/2 , � = +1/2
is accomplished by absorbing two photons: one red photon at a 780 nm wavelength,
and one blue photon at a 480 nm wavelength, going through the level 5 3/2 , ′ =
3, � = +3 with a detuning of �. The two lasers are sent parallel to the chip surface,
focused on the atomic cloud with 150 µm and 22 µm waists for the red and blue lasers,
52 Superconducting atom chip experiment

respectively, as illustrated in Figure 3.12. Here, the quantization axis is deined by the
magnetic ield at the bottom of the magnetic trap, oriented along the axis. The red
laser is � + polarized while the blue laser is � − polarized. Under this coniguration,
the 60 1/2 , � = −1/2 state is not excited. The excitation path is schematically shown
in Figure 3.11.

It is crucial to avoid sending the blue laser on the chip’s Nb wires. This strongly
focused and powerful blue laser makes the Nb wires locally transit to the normal state,
forming local heating spots. The transition then quickly spreads out. Eventually, the
chip is no longer superconducting and the experiment is interrupted. Another con-
sequence is the photo-voltaic efect, by which the blue photons rip out electrons from
the chip surface. Some of these electrons will end up on nearby dielectric surfaces,
building up stray electric ields. These are unwanted efects.

The red laser is detuned � = +2�×540 MHz from the intermediate level 5 3/2 , ′ =
3, � = +3 . Its power is typically � = 50 µW. One inds the corresponding Rabi
frequency Ω� = 2� × 40 MHz. The spontaneous scattering rate of the red photons via
transitions to the intermediate state, whose lifetime is about 26 ns [148] , is given by

1 Ω2� Γ 1 Ω2� Γ
Γ� = ≈ ,
2 �2 + Γ2 + Ω2� 2 �2
(3.8)

where Γ = 2� × 6.065 MHz is the natural linewidth of the 5 3/2 state. The scattering
rate is thus, estimated to be 0.1 photon per microsecond. The spontaneous scattering
of a photon on average gives the cloud a kick in the direction of the red laser, pushing
the cloud out of trap center. As a result, the cloud starts to oscillate and heats up. This
is an important factor that limits the power of the red laser, the duration of excitation
as well as the total number of Rydberg excitation laser pulses that can be sent on a
single cloud.

Due to the relatively weak dipole transition between the 5 state and a Rydberg
state, the blue laser must be much more powerful. We measured about 8 mW at the
entrance of the cryostat. Taking into account the transmission of about 80% for a win-
dow† , the power is 4 mW at the position of the cloud, after crossing three windows
(Figure 3.1). The corresponding Rabi frequency is calculated to be Ω� = 2� ×7.9 MHz.

From the fact that Ω� , Ω� ≪ �, one can adiabatically eliminate the intermediate
level and end up with an efective two-level system. The two-photon coupling strength
of the ground state to the Rydberg state, i.e., the corresponding efective two-photon
Rabi frequency can be expressed as

Ω� Ω�
Ω= .
2�
(3.9)

With the given parameters, one inds Ω = 2� × 280 kHz.

 3.3 Rydberg excitation 53

Chip Chip
a) 480 nm b)
z 780 nm z
x x
y I1
I2 y

D1/D2
Channeltron

S Channeltron

D1
S I2
I1
D2

Figure 3.12 Scheme of the Rydberg detection setup for a) top view and b) axonometric view.
I1 (dark green) and I2 (light green)are ionization electrodes, D1 (blue) and D2 (blue) are ion
delectors and S (purple) is the electric shield. The channeltron is mounted inside the chan-
neltron house. The trajectory of the ions is represented by the dashed lines. Excitation lasers
are also shown in a).

3.2.2
Rydberg detection

Rydberg atoms are very close to the ionization threshold and are very sensitive to
electric ields. They are thus quite easy to detect using ield-ionization technique. The
scheme of the detection is depicted in Figure 3.12. A voltage ion is applied onto the
electrodes I1 and I2 while the chip is kept grounded to ionize the Rydberg atoms. The
obtained ions are then accelerated and guided with help of two delector electrodes
D1 and D2 to a channeltron. As soon as an ion hits the channeltron, it creates a signal
sent to a discriminator, where we can count ions one by one. The channeltron operates
under a high voltage of -3000 V. An additional electrode screens the stray ield from
the channeltron. To improve the performance of the channeltron, it is kept ”warm” at
about 42 K.
Each Rydberg state is ionized at a diferent voltage. If one ramps down ion in
time from ∼ 0 V to a negative value, depending on the state of the initial Rydberg
atom, the corresponding ion arrives at the channeltron at diferent times. Thus we can
distinguish the Rydberg states as depicted in 3.13. The voltage ramp is designed so that
it gives the best discrimination between the states of interest. Deining appropriate
temporal windows as shown by the red and blue dashed lines in Figure 3.13, we can
state-selectively detect the Rydberg atoms. The detection eiciency is measured to be
up to 90% ± 10%.
Another advantage of the coniguration shown in Figure 3.12 is that, during the
Rydberg excitation, one can apply an appropriate voltage on the two electrodes I1
and I2 to compensate for the stray ields perpendicular to the chip surface at the atom
position. The required voltage for the ield compensation is typically less than 1 V.

†
The windows are not yet coated for the blue laser.
54 Superconducting atom chip experiment

a) V 200 μs
ion
0
~30 V
60S1/2
57S1/2

150 V

b) 1.4
60S1/2 57S1/2
without microwave
1.2
π pulse 60S - 57S
1.0
Count per μs

0.8

0.6

0.4

0.2

0.0
50 100 150
Arrival time (μs)

Figure 3.13 State-selective detection of Rydberg atoms. a) A typical ionization voltage ramp
in our experiment. Rydberg atoms in 60 1/2 and 57 1/2 are ionized at diferent times marked
by red and blue ovals respectively. b) The arrival time of the corresponding ions fall in deined
temporal windows, allowing us to state-selectively detect the Rydberg atoms. The atoms are
prepared in the 60 1/2 state. The 57 1/2 is experimentally populated by transferring atoms
from the 60 1/2 state with a � microwave pulse. The detections with and without the mi-
crowave pulse correspond to the detections of 60 1/2 and 57 1/2 atoms respectively.

3.3
Coherent manipulation of Rydberg atoms

Thanks to their huge electric dipole matrix elements, Rydberg atoms are very easy
to manipulate by making use of the Stark efect and microwave transition. However
at the same time, Rydberg atoms are also extremely sensitive to stray electric ields,
which will potentially wash out any coherences between the Rydberg atoms. To build
a quantum system from Rydberg atoms, the very irst requirement is thus to reduce
the inluence of the stray ields.

3.3.1
Taming the stray electric ields

During the experiment, some Rubidium atoms inevitably stick to the chip front gold
surface. The direct deposition of Rubidium atoms on the gold layer forms huge elec-
 3.3 Coherent manipulation of Rydberg atoms 55

a) b)
Chip Chip

I1/I2 Rb dispenser I1/I2

Figure 3.14 The electric ield close to the chip a) before and b) after the Rubidium coating of
the chip. The electrodes I1 and I2 are schematically shown. Rubidium is represented in red
color, either as a patch or as a homogeneous layer.

3.5
Experimental data
3.0 Gaussian fit

2.5
Count (a.u.)

2.0

1.5

1.0

0.5

0.0

-2 -1 0 1 2
Relative frequency (MHz)

Figure 3.15 Optical transition with a dilute cloud in a magnetic trap (blue dots). Rydberg-
Rydberg interaction is expected to be negligible. The excitation duration is 10 µs. A it with a
Gaussian proile (red solid line) gives the FWHM of 579 kHz. The error bars are 1 standard
deviation of the statistical error.

tric dipoles due to the big mismatch in the work functions of Rubidium and gold‡ .
These Rubidium patches on the chip surface keep building up, leading to very inho-
mogeneous and unstable stray electric ields right at the position of the atoms.
As a solution, two Rubidium dispensers were installed inside the cryostat to cover
the chip surface with a thick enough layer of Rubidium (∼ 80 nm). The idea is to
perform a fast and controlled Rubidium deposition on a large area so that the later
adsorbed cold atoms do not afect the electric ield. The Rubidium coating also helps
to cover some small dust dielectric particles that could have stuck to the chip surface,
and reduce their contribution to the stray ields. The coating of the chip was per-
formed under cryogenic temperature conditions. Figure 3.14 illustrates qualitatively
the structure of the electric ield before and after the Rubidium coating.
As a result, we obtain a narrow and stable optical transition. Figure 3.15 shows an
optical spectrum recorded with a dilute cloud after the Rubidium coating. It is well
itted by a Gaussian proile; the FWHM is 579 kHz, which is mainly due to the spectral
‡
Work function of Rubidium is 2.26 eV [149] and that of gold is 5.30 − 5.45 eV [150].
56 Superconducting atom chip experiment

width of our laser. More details on the ield characterization can be found in [136].

3.3.2
Coherent manipulation of the Rydberg atoms
In order to access the coherence time, we irst perform a spectroscopy measurement
using the 60 1/2 , � = +1/2−61 1/2 , � = +1/2 two-photon transition. This transition
is relatively insensitive to the diferential Stark efect (-10.9 MHz/(V/cm)2 ), and has no
diferential linear Zeeman shift. About 0.3 Rydberg atoms are excited out of a dilute
ground-state cloud. The driven microwave pulse lasts 300 µs. Figure 3.16 represents
the measured spectrum. It exhibits a quite narrow Lorentzian proile with a 6.6 kHz
full width half maximum (FWHM).
The transverse coherence time 2 is measured using the Hahn spin-echo technique.
Its principle is depicted in Figure 3.17a illustrating the evolution of the spins on a
Bloch sphere during this sequence. We apply a � microwave pulse (duration 0.6 µs) in
between two Ramsey �/2 microwave pulses (0.3 µs). The irst �/2 pulse prepares the
atoms in a superposition of the 60 1/2 and 61 1/2 states. They evolve freely during /2.
At this time, the � pulse mirrors the spins through the plane. The free evolution
in the second half time compensates for the dephasing accumulated during the irst
half. The coherence of the superposition is best recovered at the refocusing time . It is
then projected on the 60 1/2 state by the second �/2 pulse. We detune the microwave
pulses 70 kHz from the atomic transition and scan the (time) position of the second
�/2 pulse around to observe the revival of the Ramsey fringes. Figure 3.17 plots
the contrast of the fringes probed for diferent total durations . It is well itted by
a Gaussian, allowing us to deduce a coherence lifetime (at 1/ contrast) of 2 = 631
µs. Knowing that the lifetime of the corresponding Rydberg levels is measured to be
about 1 = 210 µs, the coherence time is suiciently long for many applications of our
system.

Summary

Throughout this chapter, the experimental setup has been briely presented. This
cryogenic setup allows us to prepare a cold atomic cloud at a sub-µK temperature.
Techniques for Rydberg excitation and detection have been also discussed.
Covering the chip with a Rubidium layer provides a very good homogeneous and
stable electric ield in the vicinity of the chip surface. As a result we obtain a rather
narrow optical line as well as the longest coherence time measured so far (630 µs) with
Rydberg atoms near a chip. This sets a very good playground to study the Rydberg
interactions.
 3.3 Coherent manipulation of Rydberg atoms 57

a) b) 0.7

61S1/2 0.6
0.5
Δν = 6.6 kHz

Transfer
0.4
2x 17.4265 GHz 0.3
0.2

60S1/2 0.1
0.0
34.85313 34.85316 34.85319 34.85322
Microwave frequency x2 (GHz)

Figure 3.16 Microwave spectroscopy on the Rydberg transition 60 1/2 − 61 1/2 . a) Energy di-
agram of the two-photon process b) Microwave spectrum taken with a 300 µs long microwave
pulse. The dots are experimental data with statistical error bars (1 s.d.). They are itted by a
Lorentzian shape (solid red line) with a 6.6 kHz width.

a)
z z z z z

y y y y y
x x x x x

π/2 T/2 π T/2 π/2 t

0.7
b)
0.6
0.5
Contrast

0.4
0.3
0.2
0.1
0.0
0 100 200 300 400 500 600 700 800
T (μs)

Figure 3.17 Coherent manipulation of the Rydberg transition 60 1/2 − 61 1/2 . a) Principle
of the spin-echo represented on Bloch sphere. ∣ ⟩ and | ⟩ denote the states 60 1/2 and 61 1/2
respectively. b) Contrast of the spin-echo sequence as a function of its total duration . The
dots are experimental with statistical error bars (1 s.d.) and the solid red line is a Gaussian it.
58 Superconducting atom chip experiment
 4
Microwave ruler for
dipole-dipole
interaction

This experiment is devoted to the study of the Rydberg excitation of a cold atomic
cloud with strong dipole interaction. It is explained in details in Raul Teixeira’s thesis
[135], in which a simple quantitative analysis was presented. Here, we develop a new,
more quantitative model, based on a rigorous simulation of rate equations. In the
irst section, we describe the optical transition from the ground state to the Rydberg
60 1/2 state for diferent laser detunings. One can distinguish two regimes: the dipole
blockade regime, where the Rydberg excitation is suppressed at short inter-atomic
distances, and the anti-blockade regime, where the Rydberg excitation is favored for
speciic inter-atomic distances. In the second section, we use microwave spectroscopy
to directly measure the interaction energy of the created Rydberg cloud. In the third
section, we recall our preliminary interpretation of the experimental results. In the
two last sections, we concentrate on a reined numerical model which allows us to
eiciently reproduce the experimental results.

4.1
Two regimes of Rydberg excitation

Beneiting from advanced improvement of laser trapping and cooling techniques, we

record the spectral line for the Rydberg excitation in a regime of strong Rydberg-
Rydberg interaction. Let us irst discuss two regimes of excitation.

4.1.1
Dipole blockade
The mechanism of “dipole blockade” is based on the dipole-dipole interaction be-
tween Rydberg atoms. Figure 4.1a illustrates the idea by considering a pair of atoms
under optical excitation. The laser is tuned on resonance to excite a single Rydberg
atom ∣ , ⟩ or ∣ , ⟩ out of two ground-state atoms ∣ , ⟩. The corresponding frequency
hardly depends on the atom separation because of the weak interaction between a
Rydberg atom and a ground-state atom. However, the Rydberg-Rydberg interaction
strongly shifts the Rydberg pair level | , ⟩ at short distances, which makes the excita-
tion with the same laser from the singly excited state to the doubly excited state out of
60 Microwave ruler for dipole-dipole interaction

a) b)
Vdd(R)

γ
|r, r〉
Rb

|g, r⟩
|r, g⟩

|g, g〉

Rb R

Figure 4.1 Dipole blockade mechanism. a) At short distance, the van der Waals interaction
pushes the excitation of the second Rydberg atom out of resonance. The red shaded circles in
b) represent the blockade volumes, where no more than one Rydberg atom can be excited.

resonance. The detuning is exactly equal to the van der Waals interaction energy �� * .
Let us deine the blockade radius as the distance at which the van der Waals shift is
equal to the excitation linewidth �/2

6
6
= �/2 (4.1)

thus,
1/6
=( 6
) ,
�
�/2
(4.2)

where 6 is the interaction strength. At distances smaller than � , the probability of

having two excited atoms is strongly reduced. The excitation of the second Rydberg
atom is blocked. There are two degenerate states of one-Rydberg-excitation ∣ , ⟩ and
∣ , ⟩. The symmetric combination of the two states | ⟩ = (∣ , ⟩ + ∣ , ⟩)/√2 is called
the collective Dicke state of two atoms. The laser couples equally ∣ , ⟩ to either ∣ , ⟩
or ∣ , ⟩ with the efective two-photon Rabi frequency Ω. Thus the strength of the
coupling between the ground state and the collective Dicke state | ⟩ is Ω√2, enhanced
by a factor √2. This enhancement factor was experimentally observed using two atoms
in optical tweezers [151]. The authors recorded the Rabi oscillation of the probability
to have one Rydberg excitation when one atom is trapped and when both atoms are
illuminated with the same excitation laser. In the latter case, the frequency of the
Rabi oscillation is about √2 times faster than that in the former case as represented in
Figure 4.2.
The concept is extended to more than two atoms and depicted in Figure 4.1b. Due
to the blockade efect, one cannot ind more than one Rydberg atom inside a sphere of
blockade radius � . The sphere is called the blockade volume. The stronger the van
We explicitly consider the van der Waals interaction, but the mechanism can be applied for the res-
*

onant dipole interaction as well.

 4.1 Two regimes of Rydberg excitation 61

1.0
R = 3.6 µm
0.8
Excitation probability
0.6

0.4

0.2

0
0 40 80 120 160 200 240 280 320
Duration of the excitation (ns)

Figure 4.2 Experimental demonstration of the collective excitation with two atoms. Each atom
is trapped in an optical tweezers. They are separated by 3.6 µm. The red circles represent the
probability to excite the irst atom when the second one is absent. The blue squares represent
the probability to excite only one atom when the two atoms are trapped and are exposed to the
same excitation pulse. The solid lines are its to the data, yielding Rabi frequencies of 7.0 ± 0.2
MHz and 9.7 ± 0.2 MHz respectively. Their ratio is 1.38 ± 0.03, close to the value √2. Figure
from [151].

der Waals interaction over the excitation linewidth, the more pronounced the block-
ade efect. Let � be the number of atoms inside a blockade volume. Due to the
blockade efect these atoms get correlated. Similarly to the case of two atoms, the cou-
pling between the ground state and the singly excited state, i.e., the collective Dicke
state, is Ω√ � . A very simple model consists in describing the blockaded mesoscopic
ensemble of � atoms as a single “super-atom” with a √ � times larger electric dipole
moment. This super-atom model has been recently used to successfully explain exper-
imental results in [117] and [152].

4.1.2
Facilitated excitation — Rydberg aggregate

In the Rydberg blockade regime, the interaction pushes the excitation out of resonance
at short distances. Note that the van der Waals interaction between two Rydberg atoms
in the same state is repulsive. If one blue detunes the excitation (Δ > 0), the interaction
energy now compensates the excitation mismatch. One can thus recover the resonant
excitation regime.
The transition of the pair of atoms can be accomplished in two successive steps: an
of resonant excitation of a irst atom followed by a resonant excitation of the second
one. The second step requires that the atom separation � satisies

Δ= �� ( � ) = 6
6
(4.3)
�
62 Microwave ruler for dipole-dipole interaction

a) Vdd(R) b) c)

γ Rf

Δ {
|r, r〉

|r, g〉 {Δ d) e)
|g, r〉

|g, g〉

Rf R

Figure 4.3 Facilitated excitation of Rydberg aggregate via a sequential excitation process. a)
The energy diagram for a pair of atoms as a function of the distance shows that the facilitated
efect happens only for distance � deined by the laser detuning Δ. b–e) Temporal evolution
of a Rydberg aggregate out of a ground atom ensemble. The blue points represent the ground
atoms and the red points the Rydberg atoms. The green bands represent the facilitated region
where atoms (shown by dark green points) can be resonantly excited.

or
1/6
=( 6
) .
�
Δ
(4.4)

The excitation of the second Rydberg atom is thus facilitated by the presence of the
irst one. Moreover the position of the second atom is controlled by the laser detuning
Δ ((4.4)). Due to the inite linewidth of the excitation, this requirement is a little bit
relaxed into � ± � � , where � � = 6 � �/Δ. It is represented by a green strip in
Figure 4.3a.
For an ensemble of atoms, any of them has more or less the same chance of being
the irst atom to be excited to the Rydberg state. However as soon as we get the irst
Rydberg atom, the excitation becomes more favorable for atoms that satisfy (4.3), i.e.,
at distance � to the already excited atom. These atoms are represented by the green
points lying on the light green spherical shell of radius � — the facilitated volume—
in Figure 4.3b.
The resonant facilitated excitation continues adding up Rydberg atoms as long as
there is an atom � satisfying

Δ= ∑ 6
, (4.5)
Ryd �≠� ��

where the summation runs over all already excited Rydberg atom �. Here, �� denotes
the distance from the atom � to the atom �. Writing down this equation, we assumed
that the van der Waals interaction energy can be added up to a good approximation
 4.1 Two regimes of Rydberg excitation 63

from all individual Rydberg pairs [153]. As a result, the dipole blockade is broken; a
strongly correlated Rydberg aggregate quickly grows around the irst Rydberg atom
“seed”. The process is demonstrated in Figure 4.3b–e. The relative distance between
the atoms is controlled by equation (4.5), or in another word, the laser detuning Δ.

4.1.3
Optical spectra

Choice of the atomic density

In order to experimentally observe the facilitated excitation, we record the optical

spectra of the Rydberg excitation with a cold cloud. A crucial technical point is that for
a given detuning, the avalanche excitation process requires a dense cloud to maintain
the conditions described in (4.3) and (4.5).
On the other hand, the nearly free Rydberg electron is scattered by the ground-
state atoms. In a irst approximation, the interaction energy � between the Rydberg
electron and the ground-state atoms is proportional to the atomic density � ̄ [154]

2�ℏ2
� = �
�,̄ (4.6)
�

where the interaction strength is characterized by the scattering length � , which is

independent of the principal quantum number and equal to −16.1 0 for 87 Rb [154].
The minus sign represents an attractive interaction � < 0. The interaction energy
is about −0.1 MHz per 1012 cm−3 . For a BEC with a typical density of 1013 cm−3 , the
interaction energy � is 1 MHz. As a result, there are two competing interactions of
the same order of strength, the repulsive van der Waals interaction, which shifts up
the energy levels and the electron–ground-state atom interaction, which lowers down
the energy levels. One inds that the blockade radius is efectively reduced. This is not
a favorable situation to study the Rydberg interaction.
As a compromise on the two conlicting requirements on the atomic density, we
choose a thermal cloud in a magnetic trap rather than a BEC. The cloud has about
10 000 atoms, held 210 µm away from the chip. The atoms are cooled to around 500
nK, just above quantum degeneracy. The cloud takes a quasi 1D cigar-shape elongated
along the direction. The extensions in each direction follow a Gaussian proile with
the width (at −1/2 ) given by �� = 23.2 µm , �� = 4.5 µm and �� = 4.2 µm (see Ap-
pendix C). The peak density is about 1.4 × 1012 cm−3 , corresponding to a maximum
red-shift of 150 kHz. We can thus neglect � in the next of discussion.

Result

The optical excitation line is displayed in Figure 4.4 for diferent laser durations. A
broadening to the blue side, i.e., high frequency, is observed, which is a strong sig-
nature of the Rydberg-Rydberg interactions (compared to that of a dilute cloud in
Figure 3.15). Using � = 579 kHz found from Figure 3.15, one estimates a blockade
radius of about 8.8 µm. Comparing this value to the inter-atomic distance of 0.9 µm at
the center of the trap, we are deinitely in the strong blockade regime.
64 Microwave ruler for dipole-dipole interaction

2 µs
400
20 µs
50 µs
300 100 µs
Count

200

100

-10 -5 0 5 10 15 20 25 30
Relative detuning (MHz)

Figure 4.4 Optical spectra for excitation durations of 2 µs, 20 µs, 50 µs and 100 µs. The peaks
at 20 MHz are due to a residual sideband modulation of the blue laser, which is required for
the Pound-Drever-Hall locking scheme [155]. The error bars denote one standard deviation of
the statistical error.

450
0 MHz
400
2 MHz
350 4.7 MHz
300 10 MHz

250
Count

200
150
100
50
0
0 20 40 60 80 100
Excitation duration (µs)

Figure 4.5 Number of Rydberg atoms as a function of the excitation duration. Data extracted
from Figure 4.4.
 4.2 Microwave probe of van der Waals interaction energy 65

Figure 4.5 plots the number of Rydberg atoms, extracted from Figure 4.4, as a func-
tion of the excitation duration for Δ = 0, 2 and 10 MHz. One inds that the excitation
starts slowly for large detunings, and then speeds up at later times. This can be inter-
preted as the time needed to of-resonantly create the initial “seed” necessary for the
facilitated growth of the Rydberg aggregate. One also notes a change in the number
of Rydberg atoms in the irst few microseconds at zero-detuning. This corresponds
to the saturation of the Rydberg excitation when the cloud is illed with super-atoms.
Further excitation of Rydberg atoms is possible at the border of the cloud but with
slower dynamics. The decay of a Rydberg atom due to its inite lifetime, or the motion
of the Rydberg atoms due to, for instance, the repulsive van der Waals force between
them can open a gap between the super-atoms, where a new Rydberg atom can be ex-
cited. This explains why the number of Rydberg atoms continues to increase instead
of staying constant in the context of the frozen Rydberg gas approximation commonly
used.
The formation of Rydberg aggregates was also observed with cesium atoms in a
vapor cell [156], with Rubidium atoms in a MOT [157] or in a dipole trap [158]. Notably,
in the latter reference, the authors observe also a strong broadening to the blue side
when they systematically increase the density of the cloud.

4.2
Microwave probe of van der Waals interaction energy

Microwave spectroscopy can be used as a probe of the energy distribution in a Rydberg

ensemble. This will be a useful tool to measure the regularity of a Rydberg chain used
for quantum simulation. In the following, we describe the principle of the method as
well as experimental results.
The idea is quite simple. Diferent Rydberg levels are shifted diferently by the
van der Waals interaction. The transition frequencies between two Rydberg levels are
thus changed correspondingly. Probing these shifts by microwave spectroscopy will
directly give us the van der Waals interaction energy. However, one has to carefully
design the measurement to extract useful information.

4.2.1
Choice of levels

The laser excitation prepares an ensemble of the 60 1/2 Rydberg atoms. The question
to ask now is “which Rydberg transition should we use?”.
The strong attractive branch of the resonant 60 − ′ dipole interaction would
trigger a fast Penning ionization process, in which the surrounding 60 atoms collapse
on the ′ atoms and get ionized. The spectrum will not be useful. Thus this pair
of levels is not relevant for our purpose. A 60 − ′ two-photon excitation with a
narrow spectral line is a good option† . Let us irst consider the 60 − ′ dipole-dipole
interaction.
†
The momentum quantum number � = � = 1/2 and its projection �� is conserved (no spin lip). We
thus omitted them for convenience.
66 Microwave ruler for dipole-dipole interaction

Table 4.1 van der Waals coeicients for diferent pairs of 60 − ′ atoms.

6,60�−�′ � (GHz.µm6 ) 6,60�−�′ � (GHz.µm6 )

60 − 63 −89.26 0.61
60 − 62 −411.36 14.96
60 − 61 292.25 248.60
60 − 60 137.62 _
60 − 59 245.13 209.55
60 − 58 −209.26 7.77
60 − 57 −43.67 0.30

60 − ′ interaction

The dipole operator does not directly couple ∣60 ⟩ to ∣ ′ ⟩ due to the selection rules.
The exchange interaction, as well as the direct interaction, results from a second-order
perturbation. It takes the form of a van der Waals potential. The efective interaction
Hamiltonian in (2.9) is rewritten as

|60 ; ′ ⟩ | ′ ; 60 ⟩
|60 ; ′ ⟩⎛ 6,60�−�′ � ⎞
⎜
6,60�−�′ �
⎟ 1
= ⎜ ⎟
⎜
⎜ ⎟ 6 ,
⎟ (4.7)
| ′ ; 60 ⟩ ⎝ 6,60�−�′ � ⎠
ef
6,60�−�′ �

where 6,60�−�′ � and 6,60�−�′ � are respectively the van der Waals coeicients of the
direct and the exchange interactions. As a result, the degenerate pair states ∣60 ; ⟩
and ∣ ; 60 ⟩ are split into a symmetric and an anti-symmetric combination of the two
bare states. Each of them is shifted ( 6,60�−�� ± 6,60�−�� )/ 6 .
Table 4.1 lists the van der Waals coeicients for the interaction of a 60 atom with
another atom in nearby ′ level. The exchange interaction drops much faster than the
direct interaction when the diference between and ′ increases. For ∣ − ′ ∣ ≥ 3, the
exchange interaction is two orders of magnitude smaller than the direct interaction. It
can thus be neglected. Interestingly, the direct interaction changes from repulsive to
attractive when ∣ − ′ ∣ ≥ 2.

− ′ transition

Let us now consider a pair of two 60 atoms at a distance undergoing such a tran-
sition. There are four pair states involved: ∣60 ; 60 ⟩, ∣60 ; ⟩, ∣ ; 60 ⟩ and ∣ ; ⟩.
However, weak microwave pulses couple ∣60 ; 60 ⟩ to ∣ ; ⟩ only to the second or-
der. We thus neglect the doubly excited state ∣ ; ⟩.
Due to the level splitting, we get in general two excitation lines, separated by
2 6,60�−�� / 6 . If one adds another Rydberg atom, we would have three interacting
pairs, resulting in an even more complex spectrum. One thus has diiculty to relate
each line to its corresponding transition, which is necessary to deduce the interaction
energy. The situation is even worse when more Rydberg atoms are involved.
If 6,60�−�� ≪ 6,60�−�� , the situation is much simpler as the two lines merge into
 4.2 Microwave probe of van der Waals interaction energy 67

a) b)
j k

j k

quasi two level system

Figure 4.6 a) The van der Waals interactions lead to an energy shift for a Rydberg pair state as
well as to a level splitting. The 60 − 57 pair is however quasi degenerate, leading to a quasi
two-level system for the corresponding transition. b) The shift of microwave photon energy
bringing an atom � from the 60 to the 57 state is proportional to the initial interaction energy
of the atom �. The doubly excited state ∣ ; ⟩ is not shown as it is rarely populated under
weak excitation.

one. According to table 4.1, this should happen for either the 60 − 57 or 60 − 63
transition, where the exchange interaction is two orders of magnitude smaller than the
direct interaction. In the following, we focus on the 60 − 57 transition but similar
results could be obtained with the 60 − 63 transition. Figure 4.6a summarizes the
idea.

For two noninteracting atoms, we obtain a single two-photon transition line cen-
tered at the frequency �0 /2 = 58.229 GHz, where ℎ�0 is the energy separation of
the 60 − 57 levels. In the presence of interactions, the excitation line is shifted to
[�0 + Δ�( )]/2, with Δ�( ) given by

−
Δ�( ) = =� ,
6,60�−60� 6,60�−57� 6,60�−60�
6 6
(4.8)

where � ≈ 1.317.

The above result can be generalized for an ensemble of Ryd interacting Rydberg
atoms. Initially all of them are in the 60 state. We assume that the microwave pulse
can excite at most one atom (�) at a time to the 57 state. The total interaction energy
of the atom before and after the excitation respectively read

�Ryd
1
�,60� = 6,60�−60� ∑ 6
, (4.9)
�≠� ��
68 Microwave ruler for dipole-dipole interaction

and
�Ryd
1
�,57� = 6,60�−57� ∑ 6
, (4.10)
�≠� ��

where we merely add up the interaction energies of the atom � with every individual
atom , at a distance �� away. The shift of the microwave frequency �� from the
non-interacting case is given by

�,60� − �,57� �,60�

�� = =� .
ℎ ℎ
(4.11)

The probability (�� ) of having an atom whose frequency shift is �� is directly the
van der Waals energy distribution ( �,60� )

�,60�
(�� ) = (� )= ( �,60� ) .
ℎ
(4.12)

The microwave spectrum, which probes (�� ), takes the shape of the Rydberg en-
semble’s interaction energy histogram within a known scaling factor .

4.2.2
Experimental work
4.2.2.1 Microwave spectra

A microwave probe is used to extract information on the interaction energy distribu-

tion of Rydberg aggregates created by a 2 µs-long laser ield. The corresponding opti-
cal excitation spectrum in Figure 4.4 is shown again in Figure 4.7. We use three laser
detunings Δ = 0, 1 and 2 MHz. A microwave pulse is applied just after the laser excita-
tion so that the energy distribution is not yet modiied due to the atomic motion‡ . For
the same reason, the microwave pulse duration is set to 1 µs. The microwave power
is adjusted on a dilute Rydberg cloud to a � pulse. The timing of the experiment is
schematically depicted in Figure 4.7a. The experiment sequence is repeated 10 times
on the same atomic cloud, with a 3 ms time interval, without noticeable heating efect.

The fraction of atoms transferred into the 57 state is plotted as a function of the
scaled microwave frequency shift Δ�/� in Figure 4.7c. Less than 3 atoms are excited
in 57 , and thus the interaction between them can be neglected. Strictly speaking,
some of the 57 atoms undergo Penning ionization with surrounding 60 atoms due
to their attractive interaction. However, this is a weak attraction. Most atoms are not
yet ionized until their detection. The microwave spectrum is therefore not altered by
the Penning ionization process.
In Figure 4.7c, the horizontal axis gives directly the interaction energy of the
atoms. According to (4.9), the nearest neighbor of an atom whose interaction energy is
, can not be closer than � = ( 6,60�−60� / ) . At Δ = 0, the microwave spectrum is
1/6

slightly shifted by about the laser linewidth �. Noting that = 2 MHz corresponding
‡
A short delay of 0.5 µs is introduced to avoid any overlap of the laser and the microwave pulses.
 4.2 Microwave probe of van der Waals interaction energy 69

a) Ionization
Δtμ
Δtl

60S - 57S
5S - 60S

tl tμ td t
b) c)
0.04 0 MHz NRyd = 66
80
0.02

0.00

60S - 57S transfer rate

60
0.06
1 MHz NRyd = 59
0.04
Count

40 0.02
0.00
-0.02
20 0.04 2 MHz NRyd = 41
0.02
0.00
0 -0.02
-4 -2 0 2 4 6 8 10 -2 0 2 4 6 8 10
Laser detuning (MHz)

Figure 4.7 Microwave probe of van der Waals energy. a) Experimental sequence showing the
timing (not to scale) of the Rydberg ensemble preparation laser pulse, the microwave probe
and the detection. b) Optical excitation spectrum. The dashed lines marks the detunings of 0,
1 and 2 MHz. The error bars are one standard deviation of the statistical error. c) Microwave
spectra at 0, 1 and 2 MHz detunings. The small peak at 5.8 MHz in b) is due to a small fraction
of atoms trapped in 5 , = 2, � = +1 state which is populated during the RF evaporation
stage. The error bars denote one standard error of the mean.

to � = 6.4 µm (cf. Table 4.1), the microwave spectrum shows no Rydberg pair closer
than this distance due to the blockade efect. When we detune the laser to the blue, Δ >
0, the facilitated excitation correspondingly shifts the spectrum. A larger detuning
gives a higher interaction energy to the Rydberg cloud. The distance between the
atoms is expected to be smaller. We thus somehow compact the Rydberg ensemble.
The long tails to the blue side observed in Figure 4.7c is due to the interaction
energy accumulated by the already excited Rydberg atoms during the aggregate for-
mation as illustrated in Figure 4.8. The development of a Rydberg cluster around a
Rydberg atom gradually gives to this atom an interaction energy larger than Δ. It also
has more neighbors and eventually becomes an atom in the cloud bulk.

4.2.2.2 Time evolution of the Rydberg cloud

Under the repulsive interaction energy given by Δ > 0, the 60 Rydberg atoms repel
each other, making the whole cloud expand. The initial interaction energy is converted
into kinetic energy. Probing the van der Waals interaction energy of the 60 cloud at
diferent time delays Δ after the laser excitation allows us to observe this expansion.
The timing and results are represented in Figure 4.9 for Δ = 1 MHz and Δ = 2 MHz.
As time goes, the spectrum gets narrower and approaches that of a dilute cloud
where the van der Waals interaction is negligible. The atoms are cooled to 500 nK
70 Microwave ruler for dipole-dipole interaction

a) b) c)

d) e) f)

Figure 4.8 Illustration of the interaction energy accumulation during the growth of a Rydberg
aggregate from a) to e). The blue spheres are ground state atoms and the red spheres are the
Rydberg atoms. The thickness of the solid lines connecting Rydberg atoms show the strength
of the corresponding interaction. The size of the red sphere represents the total interaction
energy of the corresponding atom. The dashed circle around a ground-state atom indicates
which atom will be excited in the next step.

a) Ionization
Δtμ
Δtl
waiting time Δt
60S - 57S
5S - 60S

tl tμ td t

b) c)
1.0 1.5

Δ = 1 MHz Δ = 2 MHz
1.4 0.5 µs
0.9 0.5 µs
1.3 5 µs
10 µs
10 µs
0.8 20 µs 1.2
20 µs
30 µs
1.1 30 µs
0.7 50 µs
100 µs 1.0 50 µs
60s to 57s transfer rate

60s to 57s transfer rate

0.6 0.9 100 µs

0.8
0.5
0.7
0.4 0.6
0.5
0.3
0.4
0.2 0.3
0.2
0.1
0.1
0.0 0.0

-4 -2 0 2 4 6 8 10 12 -4 -2 0 2 4 6 8 10 12

Figure 4.9 Expansion of a repulsive Rydberg ensemble probed by microwave spectroscopy. a)

Timing of the experimental sequence and the results for b) 1MHz and c) 2MHz. The spectra
are vertically shifted for a better visibility.
 4.3 Preliminary interpretation 71

corresponding to about 10 kHz kinetic energy, a hundred times smaller than the in-
teraction energy; therefore thermal motion plays no signiicant role in the observed
spectra. The recoil velocity§ due to the absorption of a 480 nm and a 780 nm photon
during the excitation only makes the Rydberg cloud globally drift, and thus gives no
change in the relative distance between the Rydberg atoms. The evolution of the mi-
crowave spectrum therefore results from the expansion of the Rydberg cloud in the
strong interaction regime.
The motion of the Rydberg atoms is not negligible after ∼ 5 µs for Δ = 2 MHz. The
commonly used frozen Rydberg gas approximation is no longer valid. This thus sets
a time limit over which one can perform quantum simulations with a Rydberg crystal,
i.e., a regularly prepared array of Rydberg atoms.

4.3
Preliminary interpretation

The time-dependent Schrödinger equation for 10 000 strongly correlated atoms cannot
be exactly integrated. Instead, quantum Monte Carlo simulation is more relevant and
is widely used for such a problem. In the thesis of Raul Teixeira, we used a preliminary
simulation to explain the experimental results. Here, we recall this simple excitation
model as well as its outcome. We discuss also the limitations of this model.

4.3.1
Monte Carlo simulation

A C++ simulation program was built based on a simple algorithm. It is a model of

incoherent excitation. To run the simulation, we need to provide the following pa-
rameters:

• The number of ground-state atoms and the spatial distribution of the cloud,
i.e., a Gaussian proile (Appendix C).

• The intensity distribution of the excitation lasers (red and blue lasers).

• The laser excitation spectrum. We approximate it by a normalized Voigt pro-

ile Voigt (Δ) of 0.9 MHz FWHM, which is a convolution of a 720 kHz FWHM
Gaussian proile and a 360 kHz FWHM Lorentzian proile.

• The laser detuning Δ.

• The total number of Rydberg atoms at the end of the excitation Ryd , which is
an integer close to the experimental results obtained for Δ. Speciically, it is 66
atoms for zero-detuning, 59 for Δ = 1 MHz and 41 for Δ = 2 MHz (Figure 4.7).

The simulation can be summarized as follows

§
The recoil velocity is of about 15 µm/ms, corresponding to a recoil energy of 26 kHz.
72 Microwave ruler for dipole-dipole interaction

Algorithm

(a) Initialization. We begin the simulation by drawing the positions of ground-

state atoms according to the spatial distribution of the atomic cloud.

(b) Calculation of excitation probability. We choose randomly a ground-state atom

� and calculate its excitation detuning taking into account the interaction with the al-
ready excited Rydberg atom(s) �

Δ� = Δ − ∑ 6
6
. (4.13)
Ryd �≠� ��

If there is no other Rydberg atom, then Δ� = Δ. Based on the linewidth of the excitation,
we determine the excitation probability � of the atom �. This probability is then
Voigt

weighted by the intensity distribution of excitation lasers (relative to that at the beam
center) to take into account the laser proiles. This gives the inal probability � for the
atom � to be excited.

(c) State update. Next, we decide whether the atom � gets excited or not according
to � calculated in step (b). For this, a random real number between 0 and 1 is drawn.
The atom � is excited if � > .

(d) Loop of the Monte Carlo sequence. The steps (b) and (c) are now repeated until
the total number of Rydberg atoms reaches Ryd .

The state lipping of an atom is considered as incoherent in this model. The atomic
cloud is assumed to be frozen during the excitation. The simulation explicitly de-
scribes only a sequential incoherent one-photon excitation process¶ .
Another important remark is that the model has no timescale. Thus it cannot pre-
dict the inal number of Rydberg atoms. Instead, an empirical number Ryd , obtained
from the experiment is required to stop the simulation.

4.3.2
Numerical result
The mechanical expansion of a Rydberg gas due to the repulsive van der Waals inter-
action after the excitation can be easily calculated by integrating Newton’s second law
of motion. The inite lifetime of the 60 state can also be included in the calculation.
Together with the Monte Carlo simulation, this allows us to generate microwave spec-
tra at diferent delays Δ , represented by the solid lines in Figure 4.11. Although the
model is quite simple, the resulting microwave spectra agree fairly with the experi-
ment.
We aim at reproducing the microwave spectra in subsection 4.2.2. For a ixed de-
tuning Δ, from the numerically generated Rydberg ensemble with the above program,
we calculate the van der Waals interaction energy for each Rydberg atom. After 50 to
200 Monte Carlo realizations, we can construct a histogram of the interaction energy.
¶
By one-photon excitation, we mean an efective two-photon process, simultaneously absorbing a red
and a blue photons.
 4.3 Preliminary interpretation 73

0.04
0 MHz

60S - 57S transfer rate

0.00

1 MHz
0.04

0.00

2 MHz
0.04

0.00

-2 0 2 4 6 8 10

Figure 4.10 Monte Carlo simulation results (solid lines) superposed with the experimental
data (points) for the van der Waals energy measurement of a Rydberg ensemble. The laser
detuning is set to 0 MHz (squares), 1 MHz (circles) and 2 MHz (diamonds). The error bars
indicate one standard error of the mean.

a) b)
1.0

Δ = 1 MHz Δ = 2 MHz
delay 1.4 delay
0.9
100 µs
0.8 100 µs 1.2

0.7 50 µs
50 µs 1.0
60s to 57s transfer rate

60s to 57s transfer rate

0.6
30 µs
0.8
0.5 30 µs
20 µs
0.4 0.6
20 µs
0.3 10 µs
0.4
0.2 10 µs
5 µs
0.2
0.1
0.5 µs
0.0 0.5 µs
0.0
-4 -2 0 2 4 6 8 10 12 -4 -2 0 2 4 6 8 10 12

Figure 4.11 Monte Carlo simulation results (solid lines) superposed with the experimental
data (points) for the expansion of a repulsive Rydberg ensemble. The laser detuning is set to
a) 1 MHz and b) 2 MHz.

The inal microwave spectrum is a convolution of this histogram with the microwave
pulse Fourier transform limited linewidth (see Appendix D). The obtained results
are plotted as the solid lines in Figure 4.10. They are vertically scaled to best it the
experimental data.
74 Microwave ruler for dipole-dipole interaction

4.3.3
Limitation
Although the simple Monte Carlo model can reproduce rather well the microwave
spectra, it fails in explaining the optical spectra. This comes from the fact that the
model lacks an excitation dynamics, i.e., a real physical time scale. As a workaround,
one can relate the number of iterations to the excitation duration. A reasonable as-
sumption is a linear correlation between them. We thus keep the number of iterations
the same for all laser detunings and for each excitation duration. The number of iter-
ations is chosen so that the simulation best its the high frequency tail of the optical
spectrum. The assumption of a linear correlation between the number of iterations
and the excitation duration allows us to include the mechanical motion of the Ryd-
berg atoms due to the repulsive van der Waals force between them. The simulation
provides us the solid lines in Figure 4.12. There are two problems. First the used
number of iterations and the excitation durations exhibit no clear linear dependency.
Second the resulting curves does not it the optical spectra at low frequencies.
450
100 μs 20000 iterations
400
50 μs 10000 iterations
20 μs 3000 iterations
350
2 μs 500 iterations
300

250
Count

200

150

100

50

0
-10 0 10 20 30
Relative detuning (MHz)
Figure 4.12 Reproduction of optical spectra with Monte Carlo simulation. For each excitation
duration, the number of iterations is the same for all laser detunings so that the simulation best
reproduces the high frequency tail of the spectrum. The calculation includes the mechanical
expansion of the repulsive Rydberg ensemble during the excitation.

The lack of a physical time scale coaxes us to provide the inal Rydberg atom num-
ber, Ryd , a parameter coming from the experimental results. Feeding the experimen-
tal outcome at the input of the simulation, this phenomenological approach makes the
model unable to predict results for diferent sets of excitation parameters. We thus
cannot use the model to test new excitation schemes.
Another limitation is the exclusion of the de-excitation of Rydberg atoms. If the
number of ground state atoms in resonance with the excitation is large, the de-excitation
of a Rydberg atom is immediately compensated by the excitation of another Rydberg
 4.4 Monte Carlo - rate equations simulation 75

atom. This is no longer true in the case of large detunings. The facilitated excitation
searches for atoms at very short distances to an already excited Rydberg atom, a con-
dition hardly fulilled with a cloud that is not dense enough. Even for a dense cloud, a
similar situation happens at later times when big Rydberg aggregates already ill the
cloud. One has to look for atoms at the border of the cloud, where the density is much
lower.

4.4
Monte Carlo - rate equations simulation

Another model is needed to overcome the limitations of the previous one. M. Wei-
demüller and coworkers obtain similar optical spectra [158]. They explain the results
by integrating the rate equations. However, they do not include the atomic motion,
resulting from the strong interaction between the Rydberg atoms, which is a must-to-
have in a model to simulate the dynamics of long excitations at high detunings.
In this section, we present our approach, solving the rigorous rate equations by a
Monte Carlo simulation — the Monte Carlo - rate equations simulation. Before en-
tering into details, we irst revisit the optical Bloch equations in the strong dephasing
regime.

4.4.1
Rate equations limit of the optical Bloch equations

Let us consider a two-level atom in a laser ield. The density matrix of the system is

� = ��� | ⟩⟨ | + ��� ∣ ⟩⟨ ∣ + ��� ∣ ⟩⟨ ∣ + ��� ∣ ⟩⟨ ∣ , (4.14)

where ��� and ��� are the populations of the excited state and the ground state respec-
tively. The coherence between the two states is ��� = �∗�� . Under the transformation to
the rotating frame deined by the unitary operator = exp(�� | ⟩⟨ |), where � is the
laser frequency, the populations are unchanged while the coherences take the form

�̃�� = �̃∗�� = ��� −��� . (4.15)

The time evolution of the density matrix elements in the rotating frame after dis-
carding fast oscillating terms, i.e., using the rotating wave approximation, are de-
scribed by the optical Bloch equations

Ω 1
�� ��� = � (�̃�� − �̃�� ) − ���
2 �
(4.16a)

� Ω
�� �̃�� = −( − �Δ)�̃�� + � (��� − ��� ) ,
2 2
(4.16b)

where Ω is the Rabi frequency, � is the lifetime of the corresponding Rydberg state
and � is the dephasing rate.
In the limit of strong dephasing, � ≫ Ω ≫ � −1 and ≳ � −1 , the coherences quickly
76 Microwave ruler for dipole-dipole interaction

reach their steady state values, i.e., �� �̃�� ⟶ 0. Extracting �̃�� (and �̃�� ) from (4.16b)
under this assumption and substituting into (4.16a), we obtain

(Ω/2)2 � 1
�� ��� = − (��� − ��� ) − ��� .
(�/2) + Δ
2 2 �
(4.17)

This equation describes the time evolution of the excited state depending on the rel-
ative population between the excited and the ground states. One can immediately
identify it as the rate equations limit of the optical Bloch equations, i.e., the Einstein’s
coeicients model. The problem of atomic excitation is reduced to a classical stochastic
process in which the atom gets excited with a rate

(Ω/2)2 �
Γ= ,
(�/2)2 + Δ2
(4.18)

and de-excited with rate

Γd = Γ + � −1 . (4.19)

Extending the model to an ensemble of atoms, the detuning of the �-th atom is mod-
iied due to the van der Waals interaction with surrounding Rydberg atoms as given
by (4.13). The corresponding rates of excitation and de-excitation become

(Ω/2)2 �
Γ� = 2
(4.20)
(�/2)2 + (Δ − ∑Ryd �≠� �66 )
���

and

Γd,� = Γ� + � −1 (4.21)

respectively.

A question arises: how can the dephasing rate � be estimated? Equation (4.17) is
valid for a dilute Rydberg cloud where the van der Waals interactions are negligible.
For long-lived Rydberg state, � −1 ≪ Γ, one can neglect the decay of the Rydberg state.
The solution when all the atoms are initially in the ground state is found to be

1
��� = (1 − −2Γ � ) .
2
(4.22)

It is approximately given by

(Ω/2)2 �
��� = à =
(�/2) + Δ2
2
(4.23)

at short time scale, Γ ≪ 1. The excitation spectrum is thus a Lorentzian proile with
a FWHM given by �. Figure 3.15 gives � = 579 kHz in our case.
 4.4 Monte Carlo - rate equations simulation 77

4.4.2
Algorithm
We have all the ingredients to build a Monte Carlo simulation. For each time interval
d , it calculates the excitation rate for each atom and decides whether the atom get
excited or not. The simulation iterates over the excitation duration. The parameters
required by the simulation consist of

• The number of ground-state atoms and the spatial distribution of the cloud,
i.e., a Gaussian proile (Appendix C).

• The peak two-photon Rabi frequency Ω0 at the center of excitation lasers.

• The intensity distribution of the excitation lasers.

• The laser excitation linewidth �.

• The laser detuning Δ.

• The iteration time step d and the total excitation duration Δ .

In the following we describe the algorithm of the simulation.

Algorithm

(a) Initialization. The simulation starts by drawing the positions of ground-

state atoms according to the spatial distribution of the atomic cloud. The time is set
to zero.

(b) Calculation of transition probability for each atom. We calculate the transition
probability � of atom � during time interval d , taking into account the interaction
with surrounding already excited Rydberg atoms. This probability is calculated by
multiplying the excitation rate Γ� for a ground-state atom, or the de-excitation rate Γd,�
for a Rydberg atom with the time step d . These rates are given by (4.20) and (4.21)
respectively. The transition is driven with a Rabi frequency Ω� , estimated using the
peak Rabi frequency Ω0 and the laser proiles. Similarly, we carry out the calculation
for all atoms in the cloud.

(c) State update. Next, we determine whether the state of the atom � is lipped or
not according to � calculated in step (b). We draw a random real number between
0 and 1. The atom � get excited or de-excited if � > . This step is performed for all
atoms.

(d) Mechanical motion. We calculate the displacements of the Rydberg atoms dur-
ing d due to the van der Waals forces between them.

(e) Loop of the Monte Carlo sequence. The time is incremented by d . The steps (b)
and (c) are repeated until reaches the total excitation duration.

Running the simulation, it can arrive a situation in which two or more atoms have the
more or less the same chance of being excited in an iteration. If these atoms are close
to each other, the excitation of an atom will modify the lipping rate of the others. One
78 Microwave ruler for dipole-dipole interaction

thus has to choose the time step d small enough such that only one atom is lipped in
an iteration ∑� Γ� × d < 1. Take the case of a resonant excitation of atoms initially in
the ground state as an example. The excitation rate for each of them is approximately
given by Γres = Ω20 /�. The excitation of one atom will suppress the excitation of all
other � − 1 atoms in a blockade sphere. Thus d should be chosen at least smaller
than 1/( � Γres ). As a consequence, the simulation take quite a long time to inish.
We replace the step (c) by another step (c*) described below to speed up the sim-
ulation. The idea is to create a list of candidates that can be excited. The excitation of
each candidate is decided considering the states of the other candidates.

(c*) State update. We compare the transition probability � of the atom � with a
real number randomly generated between 0 and 1. If � > and the atom � is in
the Rydberg state, it is then de-excited. If � > and the atom � is currently in the
ground state, it is then added to a list of candidates ℒ. The criterion is applied for
all atoms. We now have a full list of candidates ℒ but not yet new Rydberg atoms.
Each of candidate in the list ℒ has more or less the same chance of being excited. We
successively consider each candidate � in the list in a random order. The chance for it
to be excited is given by a Lorentzian

(�/2)
2

� = 2
, (4.24)
(�/2) + (∑�≠� ��6 �� )
2 ℒ
��

where the summation runs over the list ℒ. The “Rydberg kronecker” �� takes a value
0 or 1 depending on whether the candidate � is in the ground state or Rydberg state
respectively. By comparing � to a random real number � drawn between 0 and 1,
we decide whether the candidate � gets excited ( � > � ) or not. After updating the
state of the candidate �, we move on with the next candidate.

Equation (4.24) prevents one from exciting two candidates in the same time step at
a distance shorter than a blockade radius. If the candidate � is far from the other
candidates, � ≈ 1. It gets excited no matter what the states of the other candidates
are.
The same argument applies for the de-excitation of already excited Rydberg atoms.
However, the number of atoms in the Rydberg state in a blockade volume is much
smaller than that of atoms in the ground state. One can choose a time step d such
that it is small enough for the probability of having two Rydberg atoms decay in an
iteration to be negligible, but still large enough to speed up the program. Thus in the
step (c*), it is not necessary to create a list of de-excitation candidates and follow a
procedure similar to that of the excitation process.

Choice of time step

Decreasing the time step d will yield a better precision. However, the simulation
becomes time-consuming. We want to choose d such that the simulation has a rea-
sonable precision with an acceptable computation time. Especially, we want to test
whether the step (c*) speeds up the simulation without loss of precision. We perform
for each d = 1 ns and d = 50 ns a hundred realizations. The atomic cloud proile
 4.5 Experimental data simulation 79

a) b)
70
Δ = 0 MHz 50 Δ = 2 MHz
60

50 40
Count

Count
40 30
30
20
20

dt = 1 ns 10 dt = 1 ns
10
dt = 50 ns dt = 50 ns
0 0
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
Time (μs) Time (μs)

Figure 4.13 The time evolution of the Rydberg excitations averaged over 100 realizations of
the Monte Carlo -rate equations model withd = 1 ns (blue) and d = 50 ns (green) . The
shaded areas correspond to one standard deviation.

and the excitation parameters are those that allow us to reproduce the experimental
data as will be discussed in the next section. Figure 4.13 shows the numerical results
for Δ = 0 MHz and Δ = 2 MHz. The discrepancy in the number of Rydberg atoms
obtained with d = 1 ns and d = 50 ns is much smaller than the standard deviations.
With d = 1 ns, we observed almost no iteration with more than one candidate in
a blockade sphere (< 1%). This gives us conidence to use the procedure (c*) in the
model. From now on, we ix d = 50 ns, and thus save a factor of about 50 on the
calculation time.

4.5
Experimental data simulation

4.5.1
Optical proile

In addition to the ground-state atoms distribution and the excitation laser beam pro-
iles, the new algorithm requires two more parameters: the Rabi frequency Ω and the
dephasing rate �. In our attempt to reproduce the experimental optical spectra, we
vary Ω, starting from its measured value. However, in order to best it the experimen-
tal data, we have to increase the temperature of the cloud with the excitation duration.
This relects a non-negligible heating efect in our measurements. Nevertheless, the
heating mechanism is not clear. We thus approximate the heating efect by raising the
average temperature. Figure 4.14 represents the numerical results (solid lines), super-
posed with the experimental data (points). Table 4.2 summarizes the parameters used
in the simulation. We show as well the parameter used in our preliminary simulation.
Compared to Figure 4.12, we now obtain a much better it. We do not just capture
the shapes of the optical spectra but more than that, within the error bar the experi-
mental data is reproduced. This is what the simple Monte Carlo model is incapable
80 Microwave ruler for dipole-dipole interaction

2 μs 2.0

Temperature (μK)
400 20 μs
1.5
50 μs
100 μs 1.0
300
0 20 40 60 80 100
Excitation duration (μs)
Count

200

100

-10 -5 0 5 10 15 20 25 30
Relative detuning (MHz)

Figure 4.14 Optical spectra for diferent excitation durations. The dots are experimental data
while the solid lines are obtained from the Monte Carlo - rate equations simulation. The error
bars denote one standard deviation of the statistical error. For each excitation duration, the
cloud temperature is adjusted for the best result: 0.7 µK, 0.7 µK, 1.1 µK and 1.9 µK for 2 µs,
20 µs, 50 µs and 100 µs respectively as shown in the inset.

Table 4.2 Parameters used to reproduce the experimental optical spectra.

Exp. value MC sim. MC - rate equ. sim.

Trap frequency � (Hz) 47 47 47
Trap frequency � (Hz) 244 244 47
Trap frequency � (Hz) 262 262 262
Temperature (µK) 0.5 ± 0.15 0.5 0.7⋆
Number of atoms 12 000 ± 2000 10 000 12 000
Blue laser waist (µm) 22 22 22
Red laser waist (µm) 150 ∞ ∞
Dephasing - linewidth(kHz) 579 900 500
Peak Rabi frequency (kHz) 170 _ 76⋆
Number of Rydberg atoms† 41 41 _

⋆
Fit parameters.
†
Values for ∆ = 2 MHz.

of. A time scale is also naturally included in the new model, instead of a “fake”time
scale based on the iteration number - duration relation.
The diference between the simulation and the experimental values in the Rabi
frequency can be attributed to the low transmission of the optical windows, resulting
from the fog-like deposition of residual gases or grease on the cold optical surfaces.
The transmission of the optical windows were measured at room temperature at which
this deposition disappears. The same fogging phenomenon was observed on other
cryogenic experiments [159].
The simulation also igures out a non-negligible heating efect in the experiment
 4.5 Experimental data simulation 81

Mandel Q factor 2

-1
-5 0 5 10
Realative laser detuning (MHz)

Figure 4.15 Mandel -factor as a function of the excitation detuning. The results are obtained
using the Monte Carlo - rate equations simulation. The laser pulse is ixed to 2 µs. The gray
shaded region corresponds to a sub-Poissonian statistics.

Figure 4.14. This heating may originate from technical noises such as the luctuations
of currents that lead to luctuations of the trap center and the trap frequency. A quan-
titative investigation is thus necessary.

Mandel -factor

Another useful information that one can extract from the simulation is the luctuation
of the number of Rydberg atoms excited out of the cold atomic cloud in a ixed exci-
tation duration. Technically we use the Mandel -factor to characterize whether the
luctuation obeys a Poissonian distribution. It is deined as
2
⟨ 2 ⟩−⟨ ⟩
= −1.
Ryd Ryd

⟨ ⟩
(4.25)
Ryd

A Poissonian distribution yields = 0 while a deterministic source of Rydberg atoms

gives = −1, i.e., zero variance. In between, a negative value of indicates a sub-
Poissonian statistics. A super-Poissonian distribution corresponds to a positive .
With a 2 µs excitation, the calculated value of is plotted in Figure 4.15 for difer-
ent laser detunings. As a result of the strong Rydberg blockade at zero detuning, we
enter in a deep sub-Poissonian region. On the blue size, the -factor quickly grows
up, becoming a super-Poissonian statistics. The statistics luctuation of one atom ex-
citation is ampliied by the facilitated development of the Rydberg aggregates. This
behavior of the -factor characterizes the strong Rydberg interaction regime and was
experimentally observed in [158]. A comparison of Figure 4.15 with experimental data
would strengthen the validity of the Monte Carlo - rate equations model. Neverthe-
less, the measurement has not yet been performed in our particular situation.
82 Microwave ruler for dipole-dipole interaction

4.5.2
Microwave spectra

Running the simulation with the parameters given in the previous section to generate
Rydberg clusters after 2 µs of excitation, and then following the same procedure as
in subsection 4.3.2, we obtain Figure 4.16 and 4.17 for the microwave spectra. A very
good agreement between the simulation and the experimental data is observed.
Although our irst preliminary numerical results with a very simple Monte Carlo

a) b)
0.04 0 MHz
80
0.02
Experiment
Simulation 0.00

60S - 57S transfer rate

60
0.06
1 MHz
0.04
Count

40
0.02
0.00
-0.02
20 0.04 2 MHz
0.02
0.00
0
-0.02
-4 -2 0 2 4 6 8 10 -2 0 2 4 6 8 10
Laser detuning (MHz)

Figure 4.16 Monte Carlo - rate equations simulation results (solid lines) supeposed with the
experimental data (points). a) Optical proile of 2 µs excitation duration and b) the microwave
spectra probing the interaction energy. The dashed lines in a) indicate the corresponding laser
detuning for the microwave spectra in b). The dashed line in b) represents the zero interaction
point. The error bars indicate one standard deviation in a) and one standard error of the mean
in b).

a) b)
1.0

Δ = 1 MHz Δ = 2 MHz
delay 1.6
delay
0.9
100 µs
0.8 100 µs 1.4

0.7 50 µs
1.2
60s to 57s transfer rate

60s to 57s transfer rate

50 µs
0.6
1.0 30 µs
0.5 30 µs
0.8 20 µs
0.4
20 µs 0.6
0.3 10 µs
0.2 10 µs 0.4
5 µs
0.1 0.2
0.5 µs
0.0 0.5 µs
0.0
-4 -2 0 2 4 6 8 10 12 -4 -2 0 2 4 6 8 10 12

Figure 4.17 Monte Carlo - rate equations simulation results (solid lines) supeposed with the
experimental data (points) for the expansion of a repulsive Rydberg ensemble. The laser de-
tuning is set to a) 1 MHz and b) 2 MHz.
 4.5 Experimental data simulation 83

model cannot explain the optical proile, it can reproduce quite well the microwave
spectra. This conirms the fact that Rydberg clusters are developed from one or several
initial “seeds” by sequential one-photon process.
The mismatch between the numerical result and the experimental data in Fig-
ure 4.16 for Δ = 2 MHz is interpreted as an incorrect positioning of the laser detuning
Δ when taking the data for this curve. The same measurement performed on another
cloud with a slightly diferent geometry yield a good agreement with the numerical
calculation. Taking again the data at this detuning would conirm whether there was
a mistake or another mechanism underlining.

4.5.3
Spatial distribution of Rydberg atoms

The simulation allow us to get informations that are diicult to access experimentally.
Rydberg atoms are hard to image optically and the spatial distribution of the Rydberg
cluster, for us at least, cannot be directly measured. Here, using the simulations, we
extract the spatial distribution of the generated Rydberg cluster. The results are rep-
resented in Figure 4.18 for zero detuning and in Figure 4.19 for 2 MHz detuning. The
Rydberg blockade efect forces to search for Rydberg atoms at the edge of the cloud,
resulting in a larger Rydberg cluster as compared to the initial ground atomic cloud.
Blue detuning the laser somehow compacts the Rydberg cluster. These results are
consistent with those observed by microwave spectra.
More interestingly, Figure 4.20 plots the temporal evolution of the transverse and
the longitudinal sizes during the free expansion of a Rydberg cluster. Comparing to
the “expansion” of a pair of atoms whose initial energy is Δ, the peak van der Waals
interaction energy of the Rydberg cluster (the dashed line), one inds that the dynamics
of the free expansion is slowed down. A bulk atom has to wait for the edge atoms
to get away before it can move. Moreover, the expansion is found to be anisotropic.
The microwave spectra reveal the hydrodynamics expansion of a Rydberg cluster in a
regime of strong van der Waals interactions.

4.5.4
Discussion on the validity of the Monte Carlo – rate equations model

The Monte Carlo – rate equations simulation includes the spatial distribution of the
cold atomic cloud, the laser intensity proiles, the inite lifetime of the Rydberg state,
the mechanical repulsion between the Rydberg atoms and more importantly the dy-
namics of the excitation. However, there are several efects that this model does not
take into account.
First, the model describes only a sequential one-photon excitation process. The
resonant excitation of an atom pair can also be achieved by simultaneous absorption
of two photons‖ . Figure 4.21 illustrates the two excitation processes. If Δ ≫ �, one

‖
For the 5� − 60� transition, this corresponds to the absorption of four photons: two red and two
blue photons.
84 Microwave ruler for dipole-dipole interaction

a)

20 μm

1.0 Ground atoms

b) c) d) Gaussian fit
Normalized distribution

0.8 Simulation

0.6

0.4

0.2

0.0
-120 -80 -40 0 40 80 120 -30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30
Position x(μm) Position y(μm) Position z(μm)

Figure 4.18 Spatial distribution of the Rydberg ensemble for 0 MHz laser detuning. a) A
realization of the simulation. The blue dots are the ground state atoms and the red dots are
the Rydberg atoms. The radius of the circles centered on each Rydberg atom is equal to the
bloackade radius � . b)-d) Statistical spatial distribution of the Rydberg ensemble along the
, and respectively. A discrepancy from a Gaussian it (dashed lines) is observed at the
lower edges and is attributed to the Rydberg interactions. The distributions of the ground
state atoms are also shown in blue solid lines.

a)

20 μm

1.0 Ground atoms

b) c) d) Gaussian fit
Normalized distribution

0.8 Simulation

0.6

0.4

0.2

0.0
-120 -80 -40 0 40 80 120 -30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30
Position x (μm) Position y (μm) Position z (μm)

Figure 4.19 The same as Figure 4.18 for 2 MHz laser detuning.
 4.5 Experimental data simulation 85

30

25

σx
σy
20
σx,y,z (µm)

15 σz
Pair at 2 MHz
10

0 50 100 150 200

time (μs)

Figure 4.20 The statistical widths of the Rydberg ensemble as functions of the expansion time.
Shown in dashed line is the distance of a pair of Rydberg atoms initially placed at 6.4 µm,
corresponding to an interaction energy of 2 MHz.

a) b) |r, r⟩
{
Δ {
|r, r⟩
{ Vdd
Δ
Δ { { Vdd

Δ { Δ {
|g, r⟩ |g, r⟩

|g, g⟩ |g, g⟩

Figure 4.21 Illustration by energy diagrams of the facilitated excitation where the laser de-
tuning compensates for the van der Waals shift. The excitation can be either a) a sequential
single-photon process or b) a simultaneous multi-photon process. The blue spheres stand for
ground state atoms and the red spheres for Rydberg atoms.

ind the rate to simultaneously excite a pair of atoms

(Ω2 /2Δ)2 Ω2 � 2
Γ(2) = = Γ 2 [1 + ( ) ]
2� 2� 2Δ
(4.26)

by adiabatically eliminating the intermediate one-excitation state. The strong dephas-

ing assumption � ≫ Ω leads to Γ(2) ≪ Γ. One can generalize for the simultaneous
-photon excitation Γ(�) ≪ … ≪ Γ(2) ≪ Γ. Thus, the sequential one-photon excita-
tion is the dominant process.
Second, this is a model of an incoherent excitation process. The solution of the
rate equations is valid if ≫ � −1 and � ≫ Ω. From Table 4.2, � ≈ 6.5Ω. The latter
condition is more or less satisied. The former requires that ≫ 2 µs. The model can
86 Microwave ruler for dipole-dipole interaction

not reproduce the coherent evolution of the excitation of an atom at a time shorter than
this value. However, if there are many atoms inside a blockade volume, the collective
state decays faster than � −1 . This explains why we obtain a good it to the optical
spectrum with 2 µs excitation duration.

Summary

Under the strong Rydberg blockade regime of a dense atomic cloud, the optical spectra
are broadened to the blue as a result of the facilitated excitation, in which Rydberg
atoms are resonantly excited at distances deined by the laser detuning. The repulsive
interaction energy acquired during the excitation causes the Rydberg cloud to expand.
The van der Waals interaction energy of the Rydberg ensemble is directly revealed via
microwave spectra.
We have developed a simple Monte Carlo simulation that allows us to explain the
measured microwave spectra. Better results were obtained with the Monte Carlo -
rate equations simulation, where the dynamics of the excitation is included. Both
optical and microwave spectra were reproduced with a reasonable agreement. This
encourages us in using this model to explore new routes towards quantum simulation
of many-body systems.

Perspective

We aimed at performing quantum transport simulations with a 1D chain of Rydberg

atoms. The irst requirement is to prepare a regular Rydberg chain. Our microwave
tool probing the interaction energy can reveal information on the regularity of a Ryd-
berg chain.

a) b)

Δ 2Δ E Δ 2Δ E

Figure 4.22 Chain regularity probed by microwave spectroscopy. Red spheres represent Ry-
dberg atoms while the blue spheres represent a missing site (gap). The height of the red bar
under a atom indicates the relative interaction energy of the atom. The graphs at the bottom
show the expected microwave spectra.

The principle of the technique is depicted in Figure 4.22. If the van der Waals
interaction energy of two adjacent atoms is Δ, then each “bulk” atom has twice this
 4.5 Experimental data simulation 87

energy due to the contributions from it two neighbors. As a result, one expects two
peaks in the microwave spectrum, one at Δ corresponding to atoms at the ends, and
another at 2Δ for the “bulk” atoms. The ratio of the peak height (or area) is equal
to the number of “bulk” atoms over the number of atoms at the ends. If some sites
are empty forming � sub-chains, the ratio is found to be ( � − 2 � )/(2 � ), where
� is the total Rydberg atoms detected by the channeltron. One thus immediately
identiies whether a gap exists or not from the microwave spectra. Moreover, the width
of each microwave peak gives the interaction energy dispersion, which maps into the
dispersion of the inter-atomic distance. In other words, it represents the disorder of
the chain.
But, how to prepare a regular Rydberg chain?. An idea is to make use of the fa-
cilitated excitation. By blue detuning the excitation laser, one expects to control the
distance between Rydberg atoms excited out of a 1D ground-state atomic cloud and
resonantly develop a chain from an initial “seed”.
However, for the same atomic density, the number of atoms in a facilitated volume
is much smaller for a 1D than a 3D cloud. Thus, the facilitated excitation hardly occurs.
One would increase the number of atoms in the 1D cloud. Nevertheless, by doing so,
one also increases the scattering of the Rydberg electron by the ground-state atoms
as discussed in subsection 4.1.3. As a result, the energy level is lowered, and thus,
destroying the facilitated regime.
A more eicient method to prepare a regular Rydberg chain is turning to laser-
trapped-circular-atoms, which will be the central part of our proposal for quantum sim-
ulation.
88 Microwave ruler for dipole-dipole interaction
 III.
Quantum simulation with circular
Rydberg atoms – A proposal
Objectives

The phase transitions of quantum magnetic systems are of long-lasting interest, both
from the experimental and theoretical points of view. Among many models used in
the study of these problems is the anisotropic Heisenberg “XXZ” model. It describes
at a microscopic level an anisotropic anti-ferromagnetic spin-1/2 chain in a magnetic
ield. The spin coupling along the longitudinal direction � difers from that along the
transverse direction � = � = . The Hamiltonian of the system is expressed in terms
of Pauli matrices as

⎡ � + � � � � + Δ� � � � ) − ℎ ∑ � � − ℎ ∑ � � ⎤ ,
��� = ⎢∑ (��� ��+1 � �+1 � �+1 � � � � ⎥ (III)
⎣ � � � ⎦

where � � is the spin operator along the � direction. The dimensionless parameters ℎ�
and ℎ� represent the interaction with the longitudinal and transversal magnetic ields,
respectively. The anisotropy parameter Δ is deined as � / .

2 Heisenberg points
NZ
Néel phase
1 Luttinger liquid
Δ = JZ/J

NY Ferromagnetic
0
NY Y-Néel phase
-1
PX PX X-polarized phase
NZ Z-Néel phase
-2 Ferromagentic phase
0 0.5 1 1.5 2 2.5
hX

Figure III.1 Qualitative ground-state phase diagram of the model when Δ = 0. Figure ex-
tracted from [160].

The magnetic properties of the system depend on the direction of the applied mag-
netic ield. In the absence of a transverse ield, the model can be exactly solved by the
Bethe ansatz [161]. A transverse ield, however, breaks the symmetry and makes the
system behave diferently. The dynamics of the system, for instance, quenches, are
diicult to study numerically, since it involves a Hilbert space of very large dimen-
sion.
The ground state has been studied theoretically using various approximation meth-
ods [162–164]. Figure III.1 quantitatively sketches the phase diagram in the presence of
92 Objectives

a transverse ield only, in the thermodynamic limit [165–167]. The horizontal axis rep-
resents the dimensionless transverse ield ℎ� . The system features at least two gapped
Néel phases with a staggered order parameter ( � and � ), a ferromagnetic domain
( ), a gapless Luttinger liquid phase for | � | < and ℎ� = 0. One can explore the phase
transition in this diagram by adiabatic variation of the anisotropy parameters and/or
the transverse ield, starting from an easily accessed ground state, for instance in the
� region where all spins polarized along the direction.
In order to simulate this “XXZ” Hamiltonian, a quantum simulator has to satisfy
several requirements. First, the spins have to be long-lived, which allows for thou-
sands of typical exchange times to happen. This is required for an adiabatic transition
from one phase to another. Second, a defect-free, regular chain of spins is necessary.
Third, one has to be able to apply a “magnetic ield” on the spin chain and to tune the
interaction between the spins. And last but not least, one needs a scheme to read-out
the inal state of the system.

z
x
y

Figure III.2 Artist’s view of the laser-trapped circular Rydberg atom chain inside a
spontaneous- emission-inhibiting capacitor. The circular atoms (red) are placed at the min-
imum intensity of a CO2 laser optical lattice (green) superposed with a hollow Laguerre-
Gaussian beam (blue).

We propose hereby a scheme to realize such a quantum simulator by turning to

laser-trapped circular atoms as depicted in Figure III.2. The key ingredient of the pro-
posal is an ensemble of extremely long-lived circular Rydberg atoms laser-trapped
inside a spontaneous-emission inhibiting capacitor. A chain of tens of atoms can be
deterministically prepared. The interaction between the atoms allows us to realize the
“XXZ” Hamiltonian. Moreover, this Hamiltonian is tunable over a wide range by mi-
crowave dressing and electric ield adjustment. Finally, the atoms can be individually
state-selectively detected. By overcoming the bottlenecks of other systems, including
the quantum simulation with low- Rydberg atoms, this new approach would allow
us to address a wide range of spin networks problems: from the quantum transport
over a 1D chain to the phase transitions of the “XXZ” spin model, and even problems
beyond the grasp of classical computation methods.
In section 5.1, we show that the spontaneous decay can be eiciently inhibited, by
 Objectives 93

placing the atoms between the two parallel plane plates of a capacitor, with a spacing
below the cut-of for the radiated wavelength (Figure III.2). In a cryogenic environ-
ment below 1 K, circular states are preserved for durations in the minute range, lead-
ing to a trapping time of a few seconds without a single loss for a chain with tens of
atoms.
In section 5.2, we describe the trapping of circular atoms, using the ponderomo-
tive force acting on the nearly free Rydberg electron. For low- Rydberg states, photo-
ionization in the laser ield strongly reduces the lifetime. However, this efect drops
exponentially with , down to negligible values for circular states. A 1 µm-wavelength
Laguerre-Gaussian beam in combination with a co-propagating 10 µm-wavelength
(CO2 laser) standing wave forms a 1D lattice, in which the circular atoms are trapped
in intensity minima, regularly separated by 5 µm (Figure III.2). Trapping of circular
atoms will be discussed in section 5.2.
In section 5.3, we study the interaction between two circular atoms, and then in sec-
tion 5.4, we derive the spin Hamiltonian. By encoding the spin state on the 48
and 50 circular levels, the van der Waals interaction between the atoms is in a good
range to realize a spin Hamiltonian. The spin coupling is directly related to the
direct and the exchange dipole-dipole interactions. Moreover, the direct interaction
between the circular atoms depends strongly on the electric ield while the exchange
interaction does not. Their relative strength is thus tunable through the electric ield.
The “transverse magnetic ield” is obtained via a resonant microwave dressing, which
is under full experimental control. As a result, the spin Hamiltonian, can be tuned over
the whole range of the phase diagram in Figure III.1, by merely varying the dressing
strength and the static electric ield. This is a unique feature of quantum simulation
with trapped circular atoms.
In section 5.5, we propose an innovative approach to the deterministic preparation
of a vacancy-free, regular chain with more than 40 atoms. It is a variant of evaporative
cooling, successively removing atoms, one by one, out of a chain in a controlled way.
The mechanism is based on the repulsive van der Waals interactions, that strongly
push on the atoms at the two ends of a compressed chain.
In section 5.6 and section 5.7, we respectively represent methods to initialize a spin
chain and to detect the spin states individually. Finally, we discuss some decoherence
efects that present in our proposed system.
94 Objectives
Principles of the
quantum simulator 5
5.1
The quest for long-lived circular atoms

A circular atom, as discussed in section 1.3, can decay only toward the next lower
circular state, emitting a � + -polarized photon. In the presence of blackbody radiation,
the circular sate can absorb a photon and get excited to a higher Rydberg state. The
former process can be eiciently inhibited by means of a capacitor, while the latter
can be suppressed by going to a cryogenic environment. Collisions with the residual
background gas can also reduce the lifetime of a circular atom. We shows here that
we can manage to reach a 100 s lifetime.

5.1.1
Inhibition of spontaneous emission

a) b)
40
... 50C
σ+ -10 dB
35 -20
54.26 GHz -30
-40
... 49C 30 -50
-60
π σ+ -70
-80
57.69 GHz 25 -90
a (mm)

-100
20
... 48C
σ+ 15
61.41 GHz
10

... 47C 5
m 1.0 1.5 2.0 2.5
46 47 48 49 D (mm)

Figure 5.1 a) Level scheme. The horizontal axis corresponds to the magnetic quantum number
. b) Γ/Γ48 (log scale) as a function of the capacitor geometry: spacing and size . The thick
solid black line indicates a 40 dB inhibition. The open red triangle shows the chosen operation
point.
96 Principles of the quantum simulator

Despite the fact that circular atoms are long-lived, the intrinsic spontaneous emis-
sion lifetime (฀25 ms for the 48 circular state) limits the duration over which one can
run quantum simulations with a large number of atoms. A chain of 40 circular atoms
will decay in less than 650 µs, corresponding to about 45 exchange times only. How-
ever, spontaneous emission can be inhibited using the following method.
We place the circular chain in between a plane-parallel capacitor as depicted in
Figure III.2. We also apply an electric ield between the two capacitor plates. This
electric ield deines the quantization axis , stabilizing the circular orbit plane [168]
parallel to the capacitor plates. A circular state decays by emitting a � + polarized
photon w.r.t to the quantization axis (Figure 5.1a). The emitted electric ield is
thus parallel to the capacitor plates. As the electric ield must cancel at the surface
of each plate, in an ideal ininitely large capacitor, there should be no emission mode
in which the circular state can decay, as long as the separation of the two plates is
smaller than half the corresponding wavelength: < �/2 = 2.5 mm for the 48 . The
spontaneous emission rate thus drops to zero [169]. Using the same principle, Klepner
et. al. experimentally observed that the natural lifetime of a circular state is enhanced
by a factor of more than 20 compared to that in free space [170]. The experiment was
performed in 1985.
In order to determine the capacitor geometry, we have performed a classical sim-
ulation with a square capacitor made of gold, cooled to 1 K. Each plate is a square
with size . We model a circular atom by an antenna (dipole) placed inside the capaci-
tor, and calculate the radiated powers with and without the presence of the capacitor.
The ratio of the two powers directly provides us the spontaneous-emission inhibition
factor, i.e., the ratio of the residual spontaneous emission rate Γ over the natural spon-
taneous emission rate �� . The calculation is carried out using the CST-Studio suite.
The result for 48 state is represented in Figure 5.1b.
We choose to use a capacitor made with two square 13 mm-wide plates sepa-
rated by 2 mm (open red triangle in Figure 5.1b). A suppression better than 50 dB
is achieved. It corresponds to a single atom lifetime of 2500 s. Note that the inhibition
is even stronger for the 50 state since the emission wavelength is larger (5.5 mm). This
geometry, as will be clear later, allows a large enough optical access for the trapping
laser beams.

5.1.2
Suppression of blackbody radiation absorption

While a �-polarized ield is eiciently inhibited by the capacitor, a �-polarized ield

is not inhibited due to diferent boundary conditions. On the contrary, the capaci-
tor structure leads to a higher mode density, resulting in a slight enhancement in the
stimulated absorption rate[169]. At a inite temperature, blackbody radiation induces
upwards transition from a circular state into the elliptical states of the + mani-
folds (Figure 5.1a). The dipole matrix element on these transitions drops rapidly with
, as the radial overlap of the wavefunctions decreases. We thus consider only the
= 1 case. In free space, the stimulated absorption rate is given by (1.30). At a tem-
perature = 0.4 K (the base temperature of a standard 3 He cryostat), this corresponds
to a lifetime of 1260 s for the 48 and ∼ 720 s for the 50 state. Taking into account the
 5.2 Trapping of circular atoms 97

decay rate enhancement due to the capacitor structure, we estimate an enhancement

factor of less than 2 in the absorption rate[169]. We thus still reach lifetimes longer
than 350 s at 0.4 K.

5.1.3
Background gas collisions

2 a0
r=n

Figure 5.2 Toroidal orbital of a circular atom.

Another efect that may conspire to reduce the circular state lifetime is that of the
collisions with the background gas. The main residual gas in a cryogenic setup is
Helium. De Prunele calculated the state-transfer cross-section ��,20 to be about 3200
atomic unit for the = 20 circular state [171]. The collision mainly results from the
interaction between the Helium atoms and the nearly free Rydberg electron. Assum-
ing that ��,� is proportional to the volume of the orbital — a torus of radius = 2 0 ,
and thickness Δ = 2 0 /√2 (Figure 5.2) — the cross section scales as 5 . We thus,
estimate ��,50 for = 50 to be 313 000 20 . This value is a bit smaller for = 48, and thus
less dramatic. The circular atom is lost after a time ��,50 = 1/(��,50 �� �� ), where ��
and �� are the density and velocity of Helium atoms at the cryogenic temperature 0.4
K. For a pressure smaller than 10−15 mbar, which is realistic at cryogenic temperatures
at about 0.4 K [137, 172], ��,50 exceeds 1200 s.
By means of spontaneous-emission inhibition and suppression of blackbody radia-
tion absorption at a cryogenic temperature = 0.4 K and ultra-high-vacuum ∼ 10−15
mbar, a circular atom can live more than 250 s. As a result, a chain with 40 atoms
should last longer than 10 s without loss of a single atom. We could drive up to
1 000 000 typical exchange periods (∼ 100 kHz) with just a single chain. The study
of slow dynamics of a 1D spin chain would be made possible.

5.2
Trapping of circular atoms

5.2.1
Ponderomotive potential

A method for trapping circular atoms in an electric ield gradient has been proposed
in our group in 2004 [169, 173]. In this scheme, the transition line is broadened due to
the Stark efect. In order to preserve the coherence between two circular states, one can
98 Principles of the quantum simulator

a) b) MHz

10
25
25

10
20
20

5
5

z (μm)
15
15
z (μm)

0
0

10
10
-5

-5
-10

y 55

-10
z x
-10 -5 0 5 10
y (μm) 00
-10 -5 0 5 10
x (μm)

Figure 5.3 Laser trapping ponderomotive potential. a) Transverse proile of a Laguerre-

Gaussian beam of 7 µm waist and 2 W in power. b) The total trapping potential is produced
by superposing a Laguerre-Gauss beam with a standing wave CO2 laser optical lattice.

microwave dress one of them on a transition to an auxiliary elliptical state such that the
diferential Stark shift of the two circular states is canceled. Extending this complex
scheme to an array of circular atoms is not yet clear and requires a good control of the
Stark efect.
We propose here, an alternative method to trap circular atoms, based on the pon-
deromotive force acting on the nearly free Rydberg electron. The interaction energy is
given by
2 2
ℰ= �
,
4 � ��
2
(5.1)

where �� is the laser angular frequency, and � is the electron mass. Note that the
atomic core also experiences a similar force. Nevertheless, it is several thousand times
heavier than the electron, and thus the corresponding force is negligible.
Knowing that the laser intensity is related to the electric ield amplitude by

1
= �0 2 ,
2 � (5.2)

the ponderomotive potential can be expressed in terms of the laser intensity as

2
ℰ= ,
2 � �0 ��2
(5.3)

that is, proportional to the laser intensity. Acting on the Rydberg electron, this poten-
tial efectively acts on the atom as a whole. As a result, the atom seeks low intensity
regions.
The transverse proile of a Laguerre-Gaussian laser beam is circularly symmetric.
It is expressed in cylindrical coordinates ( , �, ), using a generalized Laguerre poly-
nomial �� (2 2 / ( )2 ), where ( ) is the beam radius at . Accordingly, a Laguerre-
Gaussian beam is called in the mode ( , ). A Laguerre-Gaussian beam mode (0, 0) , for
 5.2 Trapping of circular atoms 99

instance, corresponds to the well known Gaussian beam. The next mode, a Laguerre-
Gaussian beam mode (0, 1) is a hollow beam. Its transverse proile takes a donut
shape, exhibiting a zero of intensity at the beam center as depicted in Figure 5.3a.
Such a long tube of light provides a coninement of the atoms in a 1D geometry.
More speciically, we create a Laguerre-Gaussian beam mode (0, 1) of 7 µm-waist
using a1 µm-wavelength laser* . This beam aligns the atoms along the direction,
between the two capacitor plates as shown in Figure III.2. The transverse ponderomo-
tive potential provided by the Laguerre-Gaussian beam with a 2 W-power is plotted
in Figure 5.3a. The transverse trapping frequencies are �� = �� = 2� × 24 kHz. The
beam diameter at the edges of the capacitor is about 630 µm. The clipping due to the
capacitor (2 mm spacing) is thus negligible (∼ 10−8 %), avoiding power dissipation.
We also send two counter-propagating 10 µm-wavelength CO2 laser beams to form
an optical lattice, superposed with the Laguerre-Gaussian beam (Figure III.2). Atoms
are trapped at the intensity minima of the standing wave. An inter-site spacing of
5 µm is achieved. The beam waist and the power are adjusted so that we obtain an
isotropic trapping potential in three dimensions. A waist of 80 µm and a power of
70 mW provide the required trapping frequency �� = 2� × 24 kHz. The resulting
trapping potential is represented in Figure 5.3b. The beam diameter at the edges of
the capacitor is 570 µm. The beams are thus negligibly clipped by the capacitor plates.

5.2.2
Photo-ionization by trapping lasers
One might worry about the photo-ionization efect due to the trapping lasers which
has been found to be prohibitive for low angular momentum Rydberg states [120, 174].
From [175], an analytical hydrogenic expression for the photo-ionization cross-section
��� of a ∣ ⟩Rydberg state is

44 �� 3 �� 3
��� = [ 2/3 (
2 )+ 1/3 (
2 )] ,
9 3 �� 3 3
(5.4)

where � ( ) is the modiied Bessel function of the second kind, which vanishes rapidly
when increasing . Here, �� is written in atomic units. For the case of circular atoms,
the loss of atoms due to the photo-ionization is predicted to be negligible (��� ∼
10−137 20 for 48 and 10−155 20 for 50 ) as compared to that due to the background
gas collision discussed previously.
We can think of the problem in terms of the wavefunction overlap. A photon cou-
ples the Rydberg electron to the continuum. If the photon energy is much larger than
the Rydberg binding energy, we can approximate the continuum wavefunction by a
plane wave of wave vector �� . The photo-ionization cross-section can be expressed in
terms of the transition matrix element as
2
��� ∝ ∣⟨Ryd ∣ ∣ −��� � ⟩∣ ,
−��� �
(5.5)

where ∣Ryd ⟩ is the Rydberg wavefunction and �� is the wave vector of the laser. The
*
A Laguerre-Gaussian beam mode (0,1) is obtained by sending a Gaussian beam through a vortex lens
or a spatial light modulator (SLM) , appropriately modifying the spatial phase of the Gaussian beam.
100 Principles of the quantum simulator

wavelength of the continuum state is about a few Bohr radii while that of the laser is
−��� �
about 1 µm. As a result, is almost constant. With this dipole approximation, the
above equation simpliies to

��� ∝ ∣⟨Ryd ∣ −��� � ⟩∣2 , (5.6)

which is the wavefunction overlap between the Rydberg state and the continuum. The
wavefunction of high- Rydberg state varies slowly with (c.f Figure 1.1 for circular
state) while that of the continuum quickly oscillates. The mismatch of the two wave-
functions thus, makes ��� ≈ 0; the photo-ionization is negligibly small.
On the contrary, the wavefunction of low- Rydberg state oscillates much faster
when approaching he atomic core (c.f Figure 1.4), becoming comparable to the wave-
function of the continuum. As a result, the photo-ionization cross-section for low- Ry-
dberg state is signiicantly larger. This argument agrees well with the observations in
[174] and[120]. In the former, the authors calculated the photo-ionization cross-section
and ind it to exponentially drop with 1 ≤ ≤ 7, while in the latter, an enhancement of
the photo-ionization rate is experimentally measured when shining more light close
to the nucleus (for = 2).

5.3
Two interacting circular atoms in free space

The simulation of the Hamiltonian in (III) is based on the dipole–dipole interaction

between circular Rydberg atoms. The theoretical discussion for two interacting atoms
in low angular momentum Rydberg states, given in chapter 2, can be generalized and
extended to circular states. In this section, we will go into details for special cases that
are relevant to our proposal. We will consider the scaling laws with that will help
us to pick up appropriate states to realize the quantum simulator.

As discussed previously, an electric ield � across the spontaneous-emission in-

hibition capacitor is necessary to maintain the circular orbit parallel to the capacitor
plates. This thus sets up a coniguration in which the inter-atomic axis , i.e., the
trapping laser axis, is perpendicular to the electric ield � (Figure III.2).
Before going into details, it is useful to remind here the energy of a circular state
∣ ⟩ and of the two next elliptical states | ⟩ and | − ⟩, taking into account the Stark
shifts
1 1
=− − 4 (8 2 + 18 + 10) 2
��
2 2 16
1
(5.7)
= + ��� 2 ,
2 2
3 1 1
=− ± − 4 (8 2 + 36 − 20) 2
��±
2 2 2 16
1 3
(5.8)
=− 2 ± + ��� 2 .
2 2
In the above equations, we have introduced the coeicients ��� and ��� for conve-
 5.3 Two interacting circular atoms in free space 101

nience. We denote also the = − 3 elliptical states in the same manifold as that of
a circular state ∣ ⟩

∣ +⟩ = ∣ , = +2, = − 3⟩
∣ 0⟩ = ∣ , = 0, = − 3⟩ (5.9)
| −⟩ = ∣ , = −2, = − 3⟩ .

The level scheme is depicted in Figure 5.4a (c.f Figure 1.6). Table 5.1 summarizes the
scaling laws of the dipole matrix element coupling ∣ ⟩ and a nearby level.

a) Energy b)

circular state

m
n-4 n-3 n-2 n-1

Figure 5.4 a) Level scheme for high- states in an electric ield. b) Level scheme for atomic pair
states. The ∣ ; ⟩ and ∣ + ; − ⟩ are quasi-degenerate. The dipole-dipole interaction thus,
mixes the circular state with non-circular states. Pair states shown in gray are not coupled to
the ∣ ; ⟩ state.

Table 5.1 Scaling laws for dipole coupling between a circular state ∣ ⟩ to a nearby state. The
last column shows the corresponding dipole coupling 50 for = 50.

∣ ⟩↔ Transition Scaling law 50 ( . 0)

∣ +⟩ �− 3/2 372
| −⟩ �+ 3/2 372
∣( ± 1) ⟩ �+ 2 1776
∣( + 1) + ⟩ � 3/2 177
|( + 1) − ⟩ � 3/2 177
∣( + 1) + ⟩ �− 2 7
∣( + 1) 0 ⟩ �− 2 26
|( + 1) − ⟩ �− 2 7

The equation (2.2) gives the dipole-dipole interaction Hamiltonian when choosing
a quantization axis along the inter-atomic axis. The interaction preserves the total
magnetic quantum number. However, the presence of a transverse electric ield breaks
the cylindrical symmetry of the system. With a typical ield of a few V/cm, the Stark
efect is, in general, dominant. We thus choose the quantization axis along �, i.e., along
102 Principles of the quantum simulator

the direction in Figure III.2. The dipole-dipole interaction Hamiltonian reads

2 1 3
= 1 2
[ 0 0 + ( +1 −1
+ −1 +1 ) − ( +1 +1
+ −1 −1 )] , (5.10)
��
3�0 3 1 1 2 1 1 1 1 2 1 1 1 1

where the notations are the same as those in (2.2). The two last terms indicate that
the total magnetic quantum number is no longer preserved. Figure 5.4b extends the
level scheme of single atomic state to pair states. The circular pair state ∣ ; ⟩ is
quasi-resonantly coupled by the dipole-dipole interaction to the elliptical pair state
∣ + ; − ⟩, which is in turn coupled to the pair state ∣ +; − ⟩, and so on (blue

arrows in Figure 5.4b). The circular state is thus, strongly mixed with non-circular
states.
The structure of the capacitor can eiciently inhibit only the emission of a � ± -
polarized photon of wavelength larger than twice the capacitor spacing. It cannot
suppress the decay of an elliptical state towards the next lower manifold by emitting
a �-polarized photon, not to mention the transitions to low lying state with shorter
wavelengths. The dipole-dipole coupling, by mixing with non-circular states, thus
leads to a strong reduction in the lifetime of the circular state.

5.3.1
Application of a magnetic ield

In order to prevent the detrimental level mixing with non-circular states, we have to
apply a magnetic ield � to Zeeman shift the levels, remove any degeneracies of the
circular pair state with the non-circular ones. The question now is: at which strength
and in which direction do we apply the magnetic ield? In the following, we assume
that the Stark efect is dominant over the dipole-dipole coupling and the Zeeman ef-
fect. We thus keep the quantization axis along . The Zeeman Hamiltonian for a
circular state reads
��
= �� ⋅�
�
ℏ
��
= �( �+ �+ �) ,
(5.11)
ℏ � � �

where the spin part is omitted since the spin is not lipped by the dipole-dipole inter-
action. The Zeeman Hamiltonian can be expressed as

�� + −
= �(
+ −
+ + −
+ � �) ,
�
ℏ 2 �
2� � (5.12)

by deining two ladder operators

+ = � +� �

= −� ,
(5.13)
− � �

Accordingly, � and � , via the ladder operators ± , couple a circular state ∣ ⟩ to

the two next elliptical states ∣ ± ⟩. As a result, the pair state ∣ ; ⟩ is mixed with
the non-circular ∣ + ; − ⟩ (resonant coupling). This, again, leads to a reduction in
 5.3 Two interacting circular atoms in free space 103

a) b)
1
10 1.0
0
10 Simulation
-1
10 0.9
Interaction energy (GHz)

-2
10
-3
10 0.8
-4
10
-5 0.7
10
-6
10
-7
10 0.6
-8
10
10
-9 0.5
-10
10
-11
10 0.4 0 1 2
0 1 2 10 10 10
10 10 10
R (μm) R (μm)

Figure 5.5 a) Interaction energy (log scale) for two atoms in 50 state as a function of the
distance. The solid line is a it with a summation of a van der Waals and a resonant dipole
potentials 6 / 6 + 3 / 3 to ≥ 4 µm. b) Probability of the bare ∣50 ; 50 ⟩ state in the inal
state. The dashed lines in a) and b) mark = 5 µm position.

the lifetime of the circular state. We thus keep only , i.e., � perpendicular to the
capacitor plates.
We now have to ind the magnitude of � such that the mixing with the ∣ + ; − ⟩ is
negligible. We have successfully used the full-Hamiltonian-diagonalization technique
to numerically calculate the dipole-dipole interaction between low angular momen-
tum Rydberg atoms. We extend this technique for two atoms in the ∣ ⟩ circular state.
We irst create a Hilbert space of pair states with several cutofs. The diference in the
magnetic quantum number of an atom in a pair from that of the ∣ ⟩ circular state is
not larger than 2: Δ < 2 . Another criteria is that an atom in a pair is less than 3
manifold away from the ∣ ⟩ state (Δ < 3), since the transition dipole matrix element
quickly decreases with Δ . The resulting Hilbert space consists of 361 pair states.
We then construct the full Hamiltonian, including the Stark efect, the Zeeman ef-
fect and the dipole-dipole interaction for = 50 with = 1 mT and = 2 V/cm. By
diagonalizing this Hamiltonian, we deduce the interaction energy of the two circular
atoms at distance . We also carefully check that, relaxing the cutofs, i.e., changing
to Δ < 3 (1225 pair states), or Δ < 4 (961 pair states) does not yield a signiicantly
diferent result (< 150 Hz at = 5 µm)† . A 1 mT magnetic ield along , corre-
sponding to a diferential Zeeman shift of 14 MHz between two successive states, is
found to be enough to suppress the detrimental mixing efect. This value will be used
throughout the rest of this thesis.
Figure 5.5a represents the interaction energy between two circular 50 atoms as
a function of the inter-atomic distance . It is well itted with a summation of a van
der Waals 6 / 6 and a dipole 3 / 3 potential with 6 = 4.23 GHz.µm6 and 3 =
3.95 × 10−5 GHz.µm3 (red line).
The dipole interaction is interpreted as the interaction between two induced dipoles,
which result from the displacement of the circular orbit with respect to the atomic core

A further relax of the cutofs would result in a Hilbert space of more than 2000 pair states, which is
†

out of the performance of our desktop computer.

104 Principles of the quantum simulator

a) b)
z ... 50C
σ+
F
x
y π
54.26 GHz
+ B +
... 49C
‒ ‒
σ+

57.69 GHz
... 48C
m

Figure 5.6 a) The deformations of the electronic orbitals induced by external electric ield,
leads to two dipoles aligned in the same direction.b) Level scheme for spontaneous emission
of the ∣50 ± ⟩ states.

in the electric ield (Figure 5.6a). From (1.45), one can deduce the induced dipole

1
= (4 2 + 9 + 5) 4 ,
4
i (5.14)

and ind the corresponding interaction coeicient

2
= ,
4��0
i
i3,�−� (5.15)

which is 3.95 × 10−5 GHz.µm3 at = 2 V/cm for = 50, in good agreement with the
numerical result. Since the two induced dipoles align to the electric ield, they repel
each other.

For a separation larger than 3 µm, the induced dipole interaction is smaller than
1.5 kHz and thus can be neglected. However for higher and/or , it grows quickly
as 12 2 , according to (5.14) and (5.15), and becomes more signiicant. Therefore one
should be careful when neglecting this induced dipole interaction. At the distance of
interest = 5 µm, the interaction of two circular atoms can be approximated by a van
der Waals interaction. The dipole-dipole interaction acts as a second order perturba-
tion by coupling to intermediate pair states.

Figure 5.5b plots the contribution from the bare state ∣50 ; 50 ⟩ to the resulting
level. The contamination from non-circular states at 5 µm is about 0.65%, mainly from
the elliptic (∣50 + ; 50 − ⟩ + ∣50 − ; 50 + ⟩)/√2 state. As depicted in Figure 5.6b, the
spontaneous emission path to the ∣49 ± ⟩ by the emission of a � + -polarized photon is
inhibited since the radiated wavelength is below the cutof of the capacitor. The two
possible decay channels are transitions to the ∣49 ⟩ and ∣48 ⟩ states. The decay rates
are about 1/1.4 s−1 for the � transition and 1/0.3 s−1 for the � + transition. As a result,
the lifetime of the circular state is reduced to about 70 s, which is still in the minute
range.
 5.3 Two interacting circular atoms in free space 105

5.3.2
Dipole-dipole interaction and its scaling laws
In this section, we will igure out the scaling laws of the interaction between two circu-
lar atoms of diferent . On the one hand, this will help us in choosing a pair of levels,
on which we will encode the spin state, but on the other hand, this is a cross check of
the numerical simulation.
In the presence of the magnetic ield along , equations (5.7) and (5.8) are modi-
ied into
1
=− + ��� 2 + ( − 1)�� ,
2
�� 2 �

1 3
(5.16)
=− ± + ��� 2 + ( − 2)�� .
��±
2 2 2 �

5.3.2.1 − interaction

The interaction acts as a second order perturbation by coupling to intermediate pair

states ∣ ; ⟩ , where | ⟩ and ∣ ⟩ are single atom Rydberg states. The interaction is given
by

∣⟨ ; ∣ �� ∣ ; ⟩∣
2
Δ =∑ ,
��
��
(5.17)
∣�;�⟩

where Δ�� = � + � −2 �� is the corresponding detuning.

Figure 5.7 Level scheme for pair states around the ∣ ; ⟩ level. Pair states that do not couple
to the ∣ ; ⟩ state by the dipole-dipole operator are shown in gray.

From (5.10), the total magnetic quantum number is either unchanged or changed
by two units. The closest intermediate pair states in terms of energy that satisfy this
condition are ∣ + ; + ⟩, | − ; − ⟩ and (∣ + ; − ⟩ + ∣ − ; + ⟩)/√2 as shown in
106 Principles of the quantum simulator

Figure 5.7. Note that the anti-symmetric state (∣ + ; − ⟩ − ∣ − ; + ⟩)/√2 does not
couple to ∣ ; ⟩. According to table 5.1, the dipole coupling between a circular state
∣ ⟩ and its next elliptical neighbors ∣ ± ⟩ scales as 3/2 . As a result, all the numerators
in (5.17) for the listed pair states scale as 6 . Let us now consider the denominator for
each term.

1. ∣ ±; ± ⟩: the detuning reads

Δ± = ±3 + 2(��� − ��� ) 2 − 2�� � ≈ ±3 , (5.18)

as the linear Stark shifts are dominant. The couplings of ∣ + ; + ⟩ and | − ; − ⟩

to ∣ ; ⟩ are therefore of the same strength but of opposite signs. They cancel
each other.

2. ∣ +; −⟩ +∣ −; + ⟩: the detuning is given by

9
Δ = 2(��� − ��� ) 2 − 2�� ≈− 5 2 − 2�� + �( 7) .
�
4 � (5.19)

Note that , the 6 terms cancel when taking the diference ��� − ��� .

The van der Waals coeicient of two ∣ ⟩ atoms can, therefore, be expressed as
6
= ,
+1
6,��−�� 5 2
(5.20)

where and are proportionality coeicients.

In addition, there is a small correction to the interaction energy, coming from the
coupling with the ∣( − 1) ; ( + 1) ⟩ state. The quadratic Stark shifts and the Zee-
man shifts are negligible compared to the energy separation between the two mani-
folds. The detuning is approximately

1 1 2 3 2 −1
Δ= + − 2 = ,
2( − 1) 2 2( + 1) 2 2 2( 2 − 1)2
(5.21)

which scales as 1/ 4 . Since the dipole coupling between two neighboring circular
states scales as 2 , the corresponding van der Waals interaction thus varies as 12 ,

where is a proportionality coeicient. The van der Waals coeicient of the total
interaction is modiied into
6
= + 12 .
+1
6,��−�� 5 2
(5.22)

Figure 5.8 plots the van der Waals coeicient 6,��−�� determined numerically as
a function of the principal quantum number = 45 − 61 for diferent values of .
The solid lines are it curves with equation (5.22) where , , are global it parame-
ters, i.e., shared between curves. We can observe a very good agreement between the
analytic expression and the simulation.
Now, for two atoms in two diferent circular states ∣ ⟩ and ( + ) , where ≠ 0,
 5.3 Two interacting circular atoms in free space 107

Interaction energy (kHz)

10 640
2 V/cm
3 V/cm
4 V/cm
5 V/cm
6 V/cm
7 V/cm
8 V/cm
9 V/cm
1 64
45 47 49 51 53 55 57 59 61
n

Figure 5.8 Van der Waals coeicients 6 as function of for a | , ⟩ pair. The corresponding
interaction at = 5 µm is shown on the right axis. Solid lines are itting curves using equation
(5.22).

as discussed in section 2.3, one can write the efective interaction Hamiltonian as

⎛
⎜
��−(�+�)� ��−(�+�)� ⎞
⎟
=⎜
⎜
⎜
⎟
⎟,
⎟ (5.23)
⎝ ⎠
ef
��−(�+�)� ��−(�+�)�

where ��−(�+�)� and ��−(�+�)� are respectively the direct and the exchange inter-
actions.

5.3.2.2 − ( + 1) interaction

Similarly to the case of the − ′ interaction (see section 2.3), one inds that for a
pair made up of an ∣ ⟩ atom and an ∣( + 1) ⟩ atom, the exchange interaction varies
as 1/ 3 while the direct interaction takes the form of a van der Waals interaction. The
dependency of the exchange interaction comes only from the corresponding dipole
matrix elements, thus it scales as 4 . Figure 5.9a plots the van der Waals coeicient of
the exchange interaction as a function of for diferent values of . The solid line is a
it with a 4 function.
The scaling laws of the direct interaction can be derived with an argument similar
to that used for the − interaction case. The main contributions come from the
coupling to the pair states ∣ + ; ( + 1) + ⟩, | − ; ( + 1) − ⟩ and the symmetric com-
bination ∣ + ; ( + 1) − ⟩ + ∣ − ; ( + 1) + ⟩. Adding up a small correction from the
∣( − 1) ; ( + 2) ⟩ level, one can express the dependency of the direct interaction on
and as
′ 6
= + ′ 12 ,
+ +1
6,��−(�+1)� ′ 5 2 ′
(5.24)

where ′, ′, ′ and ′ are proportionality coeicients. Figure 5.9b shows the numeri-
108 Principles of the quantum simulator

a) b)
3.5 224

3 192

Interaction energy ( kHz)

10 640
2.5 160

Interaction energy ( kHz)

2 128

1.5 96

1 64
1 64
45 47 49 51 53 55 57 59 61 45 47 49 51 53 55 57 59 61
n n

Figure 5.9 Interaction of a − ( + 1) pair at diferent electric ields . a) The exchange

interaction exhibits a 1/ 3 dependence. The dipole coeicient is independent of and scales
as 4 . The solid line is a it with a power law 4 . b) The van der Waals coeicient calculated
from simulations for the direct interaction. The solid lines are its with equation (5.24).

cally calculated van der Waals coeicients for = 45 to 61 at diferent values of . The
solid lines are it curves by equation (5.24) with ′ , ′ , ′ and ′ as it parameters. A
good agreement is observed for ≥ 48. The small deviations at high around = 45
can be attributed to the weak coupling to ∣( + 1) 0,± , ⟩ levels.

5.3.2.3 − ( + 2) interaction

Both the exchange interaction and the direct interaction are second order couplings
through intermediate states ∣ ; ⟩. Respectively, they are given by

⟨ ; ( + 2) ∣ �� ∣ ; ⟩⟨ ; ∣ �� ∣ ( + 2) ; ⟩
Δ =∑ =ℎ
��� 6,��−(�+2)�
�� �� 6
(5.25)
∣�;�⟩

and

∣⟨ ; ( + 2) ∣ �� ∣ ; ⟩∣
2
Δ =∑ =ℎ ,
��� 6,��−(�+2)�
�� �� 6
(5.26)
∣�;�⟩

where Δ�� = � + � − �� − (�+2)� . The strengths of the interactions are given by

the corresponding van der Waals coeicients 6,��−(�+2)� and 6,��−(�+2)� .
Let us have a look irst at the exchange interaction. Due to the selection rules, the
only nonzero term stems from the coupling to ∣( + 1) ; ( + 1) ⟩ state. One can eas-
ily deduce that it scales as 12 and almost does not depend on . Figure 5.10 shows the
numerical results for the exchange coeicient 6,��−(�+2)� for diferent ∣ ; ( + 2) ⟩
states and diferent electric ields.
On the contrary, there are several intermediate states contributing to the direct in-
teraction. Table 5.2 lists some of the main contributions as well as their corresponding
scaling laws. As can be seen from the fourth and the two last rows of table 5.2, for high
and/or , the Stark shifts can compensate for the level spacing (Δ = 0), recovering
 5.3 Two interacting circular atoms in free space 109

a) b)

10 640

Exchange energy (kHz)

(n+1)C (n+1)C

1 64

nC (n+2)C (n+2)C nC

45 47 49 51 53 55 57 59 61
n

Figure 5.10 a) Level scheme for the exchange interaction between two atoms in the and
( + 2) states. b) Exchange van der Waals coeicient as a function of for a − ( + 2)
pair. The solid line is a it with a 12 power law.

Table 5.2 Scaling laws for dipole coupling between a ∣ ; ( + 2) ⟩ pair state and a nearby
state.

∣ ; ( + 2) ⟩ ฀ Numerator Detuning
∣ +; ( + 2) + ⟩ 6

| −; ( + 2) − ⟩ 6 −
∣ + ; ( + 2) − ⟩ 6 −12 − 9 5 2 − 8��
∣ − ; ( + 2) + ⟩ 6 +12 − 9 5 2 − 8��
∣( + 1) ; ( + 1) ⟩ 8 −4

∣( − 1) ; ( + 3) ⟩ 8 −4

∣( + 1) − ; ( + 1) ⟩ 8 1/ − 3
4

∣( − 1) ; ( + 3) + ⟩ 8 −1/ 4 + 3

the resonant dipole-dipole coupling which scales as 1/ 3 . The pair state ∣ ; ( + 2) ⟩

is mixed with non-circular states, resulting in a reduction in the lifetime of the circular
states. To observe the mixing efect, we consider the direct interaction for a pair at a
5 µm distance. Assuming a power law for the interaction �� ∝ 1/ � , we numerically
calculate the efective power � at 0 = 5 µm. Figure 5.11 represents the results. With
smaller than 55 and in a moderate ield up to 7 V/cm, the direct interaction is well
approximated by a van der Waals interaction.

5.3.2.4 −( + ) interaction with >2

To complete the discussion, let us discuss the −( + ) interaction, with > 2. The
selection rules allow only the order of perturbation to be non zero for the exchange
interaction of a − ′ pair where = ′ − . Figure 5.12 illustrates the coupling path
that leads to the exchange of excitations. As a result, the exchange interaction varies
as 1/ 3� and quickly drops when increasing the diference of and ′ . One expects
110 Principles of the quantum simulator

a)
0.6 2 V/cm 3 V/cm
4 V/cm 5 V/cm

Interaction energy (MHz)

0.5
6 V/cm 7 V/cm
8 V/cm 9 V/cm
0.4

0.3

0.2

0.1
b)
6

5
α

45 47 49 51 53 55 57 59 61

n
Figure 5.11 a) Direct interaction energy of a − ( + 2) pair at 5 µm. b) Efective power law
� obtained from numerical derivation shows the deviation from a van der Waals interaction
(� = 6) to a resonant dipole interaction (� = 3).

a 4(2�−1) dependency. Figure 5.13a and Figure 5.14a plot the exchange interaction
coeicient for = 3 and = 4 as a function of at diferent values of . Power
laws of 20 and 28 are found, respectively. The exchange interaction coeicient of the
47 −50 pair is 42 MHz.µm9 while that of 46 −50 is 1.2 MHz.µm12 . At 5 µm, these
correspond to 21 Hz and 5 mHz respectively. They are thus negligible.
The diagonal terms have a quite complex behavior. The Stark efect changes the
detuning in the denominators of the second order coupling to nearby pair states. Sim-
ilarly to the − ( + 2) case, with some values of , the interaction can recover
resonant dipole coupling, scaling as 1/ 3 . This leads to a strong reduction in the life-
time of the circular states of interest. Figure 5.13 and 5.14 represent the numerical
results for − ( + 3) and − ( + 4) pairs at a 5 µm distance, together with the
corresponding efective power �.

5.4
1D spin chain’s Hamiltonian

We now discuss how to construct the spin Hamiltonian in (III) with trapped circular
atoms. We encode the spin states |↑⟩ and |↓⟩ onto two circular levels. A −( +1 )
pair would lead to a very strong exchange interaction (few MHz for ∼ 50). On the
one hand, it overwhelms the direct van der Waals interaction between two atoms in
(few tens of kHz). This would limit the accessible range of Hamiltonians. On the
 5.4 1D spin chain’s Hamiltonian 111

a) b)

(n+1)C (n'-1)C (n'-1)C (n+1)C (n+1)C (n'-1)C (n'-1)C (n+1)C

nC n'C n'C nC nC n'C n'C nC

Figure 5.12 Exchange interaction as a multi-photon process shown for ′ − is an a) even or

b) an odd number.

a) 10 5000 b)
2 V/cm 2

Interaction energy (MHz)

3 V/cm
4 V/cm 1
5 V/cm
6 V/cm
Exchange energy (Hz)

0
1 7 V/cm 500
8 V/cm
2 V/cm 6 V/cm
9 V/cm -1
3 V/cm 7 V/cm
4 V/cm 8 V/cm
-2 5 V/cm 9 V/cm

0.1 50 c) 6

5
α

3
0.01 5
45 47 49 51 53 55 57 59 61 45 47 49 51 53 55 57 59 61
n

Figure 5.13 a) Coeicient 9 of the exchange interaction between two atoms in the and
( + 3) states. The dotted line is a it with a 20 power law and the solid line is that with the
next order correction 28 . The corresponding energy at a 5 µm separation is shown on the right
axis. b) Direct interaction energy of the same pair at a 5 µm separation. c) The corresponding
efective power law �.

4
2 V/cm 5 V/cm 8 V/cm
Interaction energy (MHz)

2 V/cm 3 3 V/cm 6 V/cm 9 V/cm

3 V/cm 4 V/cm 7 V/cm
1 4100 2
4 V/cm
5 V/cm 1
6 V/cm
Exchange energy (mHz)

0
7 V/cm
0.1 8 V/cm 410 -1
9 V/cm
-2

-3

0.01 41
6

5
α

4
1E-3 4.1
3

45 47 49 51 53 55 57 59 61 45 47 49 51 53 55 57 59 61
n n

Figure 5.14 The same as Figure 5.13 for two atoms in the and ( + 4) states. The exchange
interaction scales as 28 . Dotted line is the corresponding it. The solid line is obtained when
adding a 32 correction term.
112 Principles of the quantum simulator

other hand, the strong exchange interaction, similarly to the − interaction dis-
cussed in chapter 2, results in two branches: an attractive branch, corresponding to a
symmetric combination of the two spins |+⟩ = (|↑⟩ + |↓⟩)/√2, and a repulsive branch,
corresponding to an anti-symmetric combination |−⟩ = (|↑⟩ − |↓⟩)/√2. Since the in-
teraction varies as 1/ � (� = 3 or 6), if it is strong enough, it makes the two atoms
move closer or away from each other, depending on whether the atoms are in the |+⟩
or |−⟩ state. Starting from the |↑⟩ |↓⟩ state, for instance, which is a superposition state
of |+⟩ and |−⟩. The time evolution of the two spins gives rise to an entanglement of the
spin with the atomic motion in the lattice, which alternatively washes out the coher-
ence of the spins. One needs to weaken the exchange interaction or tighten the trap
so that the induced atomic motion is much smaller than the extension of the atomic
wavefunction.Appendix E discusses the problem in more details.
A − ′ pair where | ′ − | > 2, on the contrary, would lead to a very weak
exchange interaction. The system can be considered as a repulsive ensemble of parti-
cles without any excitation exchange: a classical Newton’s cradle. A choice of and
( + 2) is thus the most relevant. Both the exchange and the direct interactions vary
as 1/ 6 and are of comparable magnitudes.

5.4.1
1D spin chain Hamiltonians

5.4.1.1 Two-atom Hamiltonian

Let us irst write down the Hamiltonian for an atom pair. They are separated by a
distance = 5 µm.
In the basis {∣ ; ⟩ , ∣ ; ( + 2) ⟩ , ∣( + 2) ; ⟩ , ∣( + 2) ; ( + 2) ⟩} the ef-
fective interaction Hamiltonian reads

0 0 0
⎛ 6,��−�� ⎞
ℎ⎜
⎜ 0 0 ⎟
⎟
= 6⎜
⎜
⎜
6,��−(�+2)� 6,��−(�+2)� ⎟
⎟
⎟ .
⎜
⎜ 0 0 ⎟
⎟
(5.27)
6,��−(�+2)� 6,��−(�+2)�
ef

⎝ 0 0 0 6,(�+2)�−(�+2)� ⎠

Using the Pauli matrices � � , where � = , or , we deine the spin operators for each
atom acting on the tensor product space as follows

�1� = � � ⊗ �
�2� = � ⊗ � � ,
(5.28)

where the index indicates which spin is manipulated, and

1 0
�=( )
0 1
(5.29)

is the identity matrix.

 5.4 1D spin chain’s Hamiltonian 113

The efective interaction Hamiltonian can be decomposed into

�� �
=� �⊗2 + (� + �2� ) + � �1 �2
� � + (�1� �2� + �1� �2� ) ,
ℎ 2 1
ef
(5.30)

where
+2 +
� =
6,��−�� 6,��−(�+2)� 6,(�+2)�−(�+2)�
4 6
−
�� =
6,��−�� 6,(�+2)�−(�+2)�
2 6
−2 +
(5.31)
=
6,��−�� 6,��−(�+2)� 6,(�+2)�−(�+2)�
4 6
�

= .
6,��−(�+2)�
2 6
The irst term in (5.30) just redeines the energy origin and thus can be omitted. The
interaction Hamiltonian is thus rewritten as
�� �
= (� + �2� ) + � �1 �2
� � + (�1� �2� + �1� �2� ) .
ℎ 2 1
ef
(5.32)

5.4.1.2 Chain Hamiltonian

We now extend the above Hamiltonian to a 1 chain of atoms regularly separated

by . We keep only the nearest neighbor interaction terms since the contribution from
the next nearest neighbor is already 64 times smaller than that from the nearest one.
The interaction Hamiltonian for the spin chain is generalized from the above two-atom
Hamiltonian

�� � �−1
= (�1 + ��
� ) + �� ∑ � � +
ℎ 2 �
ef

�=2
�−1 �−1
(5.33)
� ∑ ��� ��+1
� + ∑ (� � � � + � � � � ) .
� �+1 � �+1
�=1 �=1

Here we deine the spin operators ��� acting on the spin number � as

1 � �
↓ ↓ ↓
��� =�⊗⋯⊗ �� ⊗⋯⊗ � , (5.34)

where � = , or . The irst term in (5.33) comes from the “border” efect of the
chain. It is clear that the interaction energy of an atom at an end of the chain, having
only one neighbor, is half that of a “bulk” atom, which interacts with its two neighbors
(c.f chapter 4).
Let �0 be the transition frequency from to ( +2) levels without any interaction.
The Hamiltonian for a non-interacting chain is given by

�0 � �
0
= ∑� .
ℎ 2 �=1 �
(5.35)
114 Principles of the quantum simulator

The total Hamiltonian is thus written as a summation of (5.35) and (5.33)

�0 � �
= ∑ � + ef .
ℎ 2 �=1 � ℎ
(5.36)

Rearranging the above equation, we obtain

�0 + �� � � �−1
= � ) + ( 0 + �� ) ∑ � � +
(�1 + ��
ℎ 2 2 �=2
�

�−1 �−1
(5.37)
� ∑ ��� ��+1
� + ∑ (��� ��+1
� + ��� ��+1
� ) .
�=1 �=1

At a distance = 5 µm, � and typically range from several kHz up to a few MHz
while �0 is more than 50 GHz. The atomic energy is thus by far dominant, making
the ground state trivial: all the atoms are in the lower state. However, when in-
troducing one or more excitations, the system is out of equilibrium. The excitations
propagates over the chain. The study of quantum transport is thus feasible.

5.4.1.3 Adding some “spices” — Rydberg microwave dressing

The situation is much more appealing when adding a classical ield at a frequency
� ≈ �0 /2 to dress the two states, via a two-photon transition. The coupling strength is
represented by the efective two-photon Rabi frequency Ω. The Hamiltonian is mod-
iied to

�0 + �� � � �−1 �−1
= (�1 + ��
� ) + ( 0 + �� ) ∑ � � + ∑ ��� ��+1
� +
ℎ 2 2 �=2
� �
�=1
�−1 �
(5.38)
∑ (��� ��+1
� + ��� ��+1
� ) + Ω cos(4�� ) ∑ ��� ,
�=1 �=1

where the driving phase is set to zero explicitly without loss of generality.

We change to the rotating frame deined by the unitary transformation

⎛ ⎞
�
= exp ⎜
⎜�2�� ∑ ��� ⎟
⎟. (5.39)
⎝ �=1 ⎠

The new Hamiltonian is given by

̃ = † + �ℏ(�� ) † . (5.40)

Under this transformation, ���(�) is replaced by cos(4�� )���(�) ∓ sin(4�� )���(�)

 5.4 1D spin chain’s Hamiltonian 115

while ��� is unchanged. We have

Δ′ � Δ �−1 �
= (�1 + ��
�) + ∑� +
̃
ℎ 2 2 �=2 �
�−1 �−1
Ω �
(5.41)
+ ∑ ��� ��+1
� + ∑ (��� ��+1
� + ��� ��+1
� ) + ∑ ��� ,
�
�=1 �=1
2 �=1

with Δ = 2� − (�0 + 2�� ) and Δ′ = 2� − (�0 + �� ). In the derivation of the last

term, we have used the rotating wave approximation to drop out fast oscillating terms.
Rearranging the above equation, we obtain

⎡
�−1
= ⎢ ∑ (��� ��+1
� + ���� + �
��� ��+1
� )+
̃
ℎ � �+1
⎣ �=1
Δ � � Ω � � ⎤ �� �
(5.42)
∑ � + ∑ �� ⎥ − (�1 + ��
�) ,
2 �=1 � 2 �=1 2
⎦

where the last term accounts for the “border” efect. Compared to (III), the above
equation represents the general form of a -spin Hamiltonian for a spin chain in
an external magnetic ield. Here, and � correspond to the transversal and longitu-
dinal spin couplings respectively. The anisotropic parameter is given directly by � / .
The longitudinal and the transverse “magnetic ields” are given by the microwave de-
tuning Δ/(2 ) and the dressing strength Ω/(2 ) respectively.

5.4.2
Choice of levels and tunable spin chain

2.5

2.0

1.5

1.0

0.5

0.0

-0.5

-1.0

-1.5

-2.0

-2.5

2 3 4 5 6 7

Figure 5.15 The ratio � / as a function of the electric ield for the 48 and 50 states. The
dotted lines indicate the isotropic Heisenberg model. The central solid line corresponds to the
pure XX model.
116 Principles of the quantum simulator

According to section 5.3, � and scale diferently when changing the principal
quantum number . As a result, we can choose a range of values in which � is of the
same order of magnitude as ( ∼ 50). Moreover, for a given , the direct interaction
terms do depend on the electric ield while the exchange interaction and thus is
nearly constant. This gives us the possibility of tuning the � / ratio. Figure 5.15 plots
this ratio for a 48 − 50 pair as a function of . For a variation of between 2 and 7
V/cm, � varies between 2.3 and −2.3 . We can thus tune our spin chain from a purely
transverse Hamiltonian ( � = 0, only spin coupling in the transverse direction ) to
an isotropic Heisenberg model ( � = � = � = ) by a mere voltage control. In
addition, Ω and Δ (Δ′ ) are fully controlled in the experiment and can be varied over a
wide range (a few MHz) in a rather short time scale (less than a µs). In other words,
we have the possibility to explore the whole range of the phase diagram depicted in
Figure III.1 (resonant dressing Δ = 0) and beyond.
For such a lexibility, we will encode the spin states on the 48 and 50 levels. The
diference between Δ and Δ′ is about 33 kHz. On the one hand, this diference allows
us to easily address the two atoms at the ends of the chain using a microwave pulse.
On the other hand, with a chain with tens of atoms, what happens inside the chain
would be less afected by the boundary condition efect. This “border” efect will be
neglected in the next sections.

5.5
Deterministic chain preparation

The Hamiltonian described by (5.42) is valid provided that the spin chain is gap-less,
i.e., has no empty site. In this section, we propose a protocol to deterministically load
the optical lattice with unit illing. The principle is based on a variant of evaporative
cooling as described below
We start with a dilute and large sample of circular atoms held in the Laguerre-
Gaussian light-tube. With two “plug” focused laser beams crossing the hollow trap
beam, we create two barriers, one higher than the other, conining the atoms longitu-
dinally as demonstrated in Figure 5.16. We slowly reduce the distance between the
two “plug” beams. The atoms are thus squeezed, building up the repulsive van der
Waals forces between them. As soon as the energy of the last atom at the weak barrier
side exceeds the height of the barrier, it is expelled and escapes along the Laguerre-
Gauss beam. By compressing further the chain, we remove more and more atoms.
The number of atoms left in the trap is determined by the inal barrier distance.
As soon as the number of atoms of interest is reached, we rise up the heights of
the two barriers to prevent further losses of atoms. By adjusting the separation of the
two barriers, we match the inter-atomic distance with the optical lattice spacing, and
transfer all the trapped atoms into the lattice, thereby achieving a unit illing factor.
In the following, we discuss in details our proposal of a full experimental sequence.
It consists in two main steps as illustrated in Figure 5.17:

1. Preparing a long irregular chain of circular atoms,

2. Ordering the chain by evaporative cooling and transferring it into the trapping
standing-wave potential.
 5.5 Deterministic chain preparation 117

C S D
z
x
y

Figure 5.16 Sketch of the proposed experimental setup. Cold atoms are prepared with the
atom chip , and transferred into the science zone where they are excited to the Rydberg
state, circularized and trapped in the hollow Laguerre-Gaussian beam (blue). Two plug beams
(red) compress the chain, removing atoms one by one until the inal number of atoms of in-
terest reached. At the end of the sequence, the state of the circular atoms is read out at the
detection zone .

5.5.1
Initial irregular chain preparation

In order to prepare a long and dilute chain of circular atoms, we irst excite low angu-
lar momentum Rydberg atoms out of a long and cold ground-state cloud in the dipole
blockade regime, and then transfer them into the circular state by a circularization pro-
cess [94].

Laser excitation of a Rydberg atom chain

We start with a cigar shaped ground state rubidium cloud at a sub-µK temperature
prepared with the atom chip setup. For instance, a BEC of about 30 000 atoms can be
experimentally prepared (Figure 3.10). We adiabatically transfer and let the atoms ex-
pand into a red-detuned dipole trap made of a 1 µm-wavelength Gaussian beam. The
use of the dipole trap enables an elongated cloud while maintaining tight transverse
coninements. This dipole trap also serves as an optical tweezer to move the cloud
from the atom chip into the “science capacitor” (Figure 5.16). Further cooling in
the dipole trap can be an option. We assume that about 2000 atoms are held in the
dipole trap, forming a cloud of ∼ 1 mm long.
We now turn of the dipole trap and apply a 10 µs-long laser pulse to bring the
atoms into the 50 Rydberg state in the dipole blockade regime (zero detuning). The
positions of the excited Rydberg atoms are simulated using the Monte Carlo – rate
equations model described in section 4.4. About 100 Rydberg atoms are excited, sep-
arated by (9 ± 3) µm. At this separation, the van der Waals interaction between the
atoms is negligible.
118 Principles of the quantum simulator

Figure 5.17 Deterministic chain preparation. It consists in two main step: 1) Irregular circular
chain preparation and 2) Tailoring the chain using evaporative cooling and transferring into
optical lattice. Circular atoms are represented by the red spheres. The two barriers (green)
compress the chain, removing atoms one by one until reaching the desired number of atoms.
Then the chain is transferred into the optical lattice (blue).

Circularization of the Rydberg atoms

Now, we apply a voltage across the capacitor, creating a homogeneous electric , along
. This electric ield, as discussed previously, deines a quantization axis and lifts the
degeneracy of the manifold . The levels, shifted by the linear Stark efect, make up
an open “umbrella”, whose tip is at the circular state as depicted in Figure 5.18a. The
lowest Stark levels of each , i.e., the ∣ , , − 1 − ⟩ levels, forms a staircase (blue
lines). Neighboring steps are separated by

3
ℎ�� ≈ ,
2
(5.43)

which is about 100 MHz/(V/cm) for = 50. A � + -polarized radio-frequency couples

all the levels of the staircase. An appropriate adjustment of the RF power and the
electric ield will adiabatically transfer the atoms from the 50 state ( = 0), climbing
the staircase into the 50 circular state.
An intuitive way to explain the adiabatic passage is to use the dressed atom model
to describe the atom-light interaction. Starting from the 50 state with -RF photons,
by successively absorbing a photon, the atom gets excited to the circular state. We de-
note an intermediate atom-light state in the process by ∣50, = , − ⟩, where is the
number of photons absorbed. The 50 and 50 states correspond to |50, = 0, ⟩ and
|50, = 49, − 49⟩ respectively. In a very weak RF ield, the states are not coupled.
Each atom-light state is shifted from the circular state by (49 − ) × ℎ(�� − ��� ), where
��� is the RF frequency. The levels are represented in Figure 5.18b by the dashed blue
lines as a function of the electric ield . At �� = ��� , the system is degenerate. With
 5.5 Deterministic chain preparation 119

a) m=0 m=1 m=2 m=3 m=46 m=47 m=48 m=49

k=3
k=2
k=1
k=0 k=0
k = ‒1
k = ‒2
Stark k = ‒3
k = ‒45 circular level
switching k = ‒46 n = 50
k = ‒46 k = ‒47
k = ‒48
k = ‒49 radio frequency
transition

Laser fast adiabatic passage

excitation

b)

resonance

⟵ ⟵ ⟵ ⟵ ⟵
⟵
⟵
⟵
circular level ⟵
Energy

. . . . .

n = 50 ⟵
⟵
⟵
F
c) m=2 m=3 m=46 m=47 =48 m=49

k=3
k=2
k=1
k=0 k=0
k = ‒1
Stark k = ‒2
Microwave
switching k = ‒3
transition circular level
k = ‒46 m n = 50
k = ‒46

k = ‒47
radio frequency
transition

Laser fast adiabatic passage

excitation

Figure 5.18 a) Level scheme of the circularization process for Hydrogen atom. b) Principle of
the adiabatic passage. c) Level scheme of the circularization process for 87 Rb atom.
120 Principles of the quantum simulator

a higher power, the RF couples the levels, leading to an anti-crossing between them,
resulting in the red lines in Figure 5.18b. Starting from the 50 state in a high electric
ield, by adiabatically ramping down the electric ield while keeping ��� constant, one
ends up in the circular state.
Due to the quantum defect of Rubidium, the few irst steps of the staircase are
much lower in energy (few tens to hundreds GHz), and thus are not coupled by the
RF. The 50 , = 2 is the lowest state that can be involved in the adiabatic passage
as depicted in Figure 5.18c (c.f Figure 1.7). A two-photon microwave �-pulse transfer-
ring the atoms into the 50 state, followed by another one-photon microwave pulse
transferring the atoms into the 50 , = 2 (∣ = −46, = 2⟩) state at the beginning of
the adiabatic passage is necessary. Alternatively, one can adjust the laser to directly
excite the 50 instead of the 50 state, and then transfer them into the 50 , = 2 state
(Figure 5.18c).
Technically, the � + -polarized RF ield is generated by applying voltages with syn-
chronized phases and amplitudes on the four electrodes near the edges of the capacitor
plates (green patches on Figure 5.16). The whole circularization process lasts a few µs
and reaches a very high eiciency [94, 95]. Even if we cannot transfer all the atoms into
the circular state, atoms in the elliptical states quickly decay and become untrapped
in the next step of the experiment.
This step ends by applying a short pulse of a strong 780 nm wavelength laser to
push out the ground state atoms. We are left with about 100 circular atoms aligned
along . Due to the weak interaction between the atoms (∼ 250 kHz), we assume
that, so far, the atoms have not moved.

5.5.2
Spin chain tailoring

In this step, we trap the circular atoms on a 1D chain using the Laguerre-Gaussian
beam in combination with the two “plug” beams. We, irst, compress and evapora-
tively cool the chain of circular atoms until obtaining the number of atoms of interest.
Here, we aim at a chain with 40 circular atoms. We then adjust the distance between
the atoms and transfer them into the standing-wave optical lattice. This step requires
an appropriate control of the laser beams (position, power and waist) with a good
timing.
Taking as the input the positions of the Rydberg atoms, obtained from the excita-
tion model in the previous step, we simulate the classical motion of the atoms, looking
for the optimal laser beam parameters. The simulation takes into account the van der
Waals interaction between the atoms‡ . To keep things simple, we neglect the trans-
verse motion and concentrate on the axial motion of the atoms only.
A remark on the simulation is that the problem involves many nonlinearly inter-
acting bodies. As a result, the system is chaotic, i.e., a small deviation in the initial
state leads to an appreciable diference in the inal result, especially when the system
evolves over a long time. We thus need a high precision numerical integrator. In addi-
tion, numerical methods for solving the equations of motion usually slightly modify
the total energy, resulting in artiicial excitation or damping of the system after a large
‡
We calculate the interaction between all the atoms, not just the nearest neighbors.
 5.5 Deterministic chain preparation 121

number of iterations. A symplectic integrator helps to suppress this numerical efect.

We thus, use a recently developed sixth-order Runge-Kutta-Nyström method, follow-
ing the scheme described in [176].
Optimizing the laser parameters using the simulation, we successfully load 40 cir-
cular atoms into the optical lattice with unit illing factor. The main criterion for the
optimization throughout the process is to keep the atoms close to the motional ground
state, and thus to try to suppress any heating efect as much as possible. Figure 5.19
shows a typical trajectory of the atoms over the process, while Figure 5.20 represents
the time variation of each laser beam. The time evolution of the average kinetic and
potential energies per trapped atom is plotted in Figure 5.21. In the following, we
discuss in details the result. For clarity, we divide this step into four sub-steps:

(a) Turning on the trap,

(b) Atom evaporation,

(c) Matching the distance,

(d) Transfer into the optical lattice.

a) 1D trap of circular atoms: 0–100 ms

After the circularization of the atoms (the time origin is set to 0 ms), we turn on the
Laguerre-Gauss beam. The repulsive van der Waals forces between the atoms, even
weak, will give to the atoms a velocity. If one slowly switches on the beam, the atoms
have enough time to drift out of the trap (Figure 5.22a). Abruptly switching on a tight
trap, on the contrary, will directly give high potential energies to the atoms that are
away from the trap center (Figure 5.22b). As a result, the atoms oscillate, rising the
temperature. Optimally, we abruptly turn on the trap that is just high enough to cap-
ture all the atoms, then adiabatically ramp up the laser power, thus the trap depth and
trap frequency, to the inal value (2 W, corresponding to a 24 kHz transverse trapping
frequency, Figure 5.20a) as depicted in Figure 5.22c.
In the meantime, we turn on in a similar manner the “plugs”, made of two 1 µm-
wavelength Gaussian beams of 30 µm waist, propagating along , crossing the Laguerre-
Gaussian beam (Figure 5.16). These two plug beams form two barriers conining the
atoms along . One of the barriers is of 4 MHz height, slightly higher than the other,
of 3 MHz. We quickly compress the chain by reducing the distance between the
two “plugs” from its initial value of 1 mm down to 0.5 mm as shown in Figure 5.20a.
This fast compression helps to save time without signiicantly modifying the loading
eiciency into the optical lattice later.
As can be seen in Figure 5.21, when abruptly turning on the trap, the kinetic energy
of the atoms slightly increases. However, some of the atoms can easily escape out
of the trap at this initial times, taking away with them a part of the van der Waals
interaction energy. As the result, the remaining atoms are cooled down. The kinetic
energy is decreased. As soon as the trap is tight and deep enough, there is no more
atom escape. The chain is compressed, building up the interaction energy, and thus
also the kinetic energy of the atoms.
 a b c d

400

200
Position x (μm)

-200
Principles of the quantum simulator

-400

0 200 400 600 800 1000 1200

Time (ms)
Figure 5.19 A typical trajectory of the atoms during the chain preparation. The process is divided into four steps: a) Turning on of the 1D trap, b) atom
evaporation, c) matching the distance and d) loading of the optical lattice.
122
 5.5 Deterministic chain preparation 123

a b c d
1000

Barrier height (MHz) Barrier separation L (μm) 800 900 950 1000
214
600 212

210
400
208

200

30
4

Barrier waist (μm)

20
2
Low barrier
High barrier
Waist 10
0

Trap frequency (kHz)

20
20
Trap depth (MHz)

trap frequency - 1D lattice

15
trap frequency - LG beam
10 trap depth - 1D lattice 10
5 trap depth - LG beam

0 0

0 200 400 600 800 1000 1200

Time (ms)

Figure 5.20 Time variation of the laser parameters during a chain preparation. Top: the dis-
tance between the two “plug” beams. Middle: the heights of the two “plug” beams are shown
on the left axis and their waists are on the right axis. Bottom: the trap depth (left axis) and
trap frequencies (right axis) of the Laguerre-Gaussian (LG) and the optical lattice beams.

a b c d
100
Kinetic engery ×1000, interaction energy (MHz)

Number of trapped atom

0.8 Average kinetic energy x1000
Average potential energy
80
Trapped atom number

0.6
60

0.4
40

0.2
20

0.0 0
0 200 400 600 800 1000 1200
Time (ms)

Figure 5.21 Evolution of the average kinetic (red) and van der Waals potential (blue) energies
(per trapped atom) during a sequence preparing a chain of 40 circular atoms. The number of
trapped atom is also shown on the right axis.
124 Principles of the quantum simulator

a) Slow

b) Abrupt

c) Hybrid

Figure 5.22 Turning on of the Laguerre-Gaussian trapping potential (blue lines). a) Slow
switching on of the trap potential leads to a loss of atom. b) Abruptly switching on the trap
potential gives the atom a high energy, resulting in signiicant heating efect. c) An optimized
hybrid method to switching on the trap. The trap is turned on to capture the atom, and then
adiabatically ramped up to its inal value. The dotted open circles and the red illed circles
respectively represent the positions of the atoms initially and when the trap reaches its inal
potential (blue lines).

b) Atom “evaporation”: 100–896 ms

This is the main step to prepare a deterministic number of atoms. We now slowly
reduce the distance between the two “plug” beams. As mentioned previously, the
atoms keep being compressed, building up the repulsive van der Waals forces between
them. The last atom at the weak “plug” beam is expelled out of the trap as soon as its
energy exceeds the height of the barrier. As observed in Figure 5.23a, which is a zoom
in of Figure 5.19, the atoms are successively removed one after another in a more or
less regular rate.
Assuming that the atoms quickly attain their equilibrium during the atom evapo-
ration, for a distance of the two “plug” beams, there is a maximum number of atoms
��� ( ) that can be trapped in between. As a result, by slowly reducing , an atom is
removed as soon as the number of trapped atoms exceeds ��� ( ). In another word,
the number of atoms left is ��� ( ), which is determined by the inal distance of
the two “plug” beams. Figure 5.23b plots the number of trapped atoms, averaged over
100 trajectories, as a function of . It takes the shape of a staircase.
However, the atoms have a inite velocity. The escape of the atoms can thus, occur
earlier or later than the moment when ��� ( ) decreases by one unit. This efect
blurs out the staircase. When more atoms are evaporated, the remaining atom chain is
signiicantly cooled down. The staircase becomes sharper. The cooling efect is clearly
observed on Figure 5.21b as a drop of the average potential energy (blue line) at every
atom ejection, as well as a global diminution of the average kinetic energy (red line).
Here, the thermalization of the chain is possible thanks to the many-body collisions
under the van der Waals interaction. Note that, although the van der Waals potential
 5.5 Deterministic chain preparation 125

0
a)
Position x (μm)

-100

500 600 700 800 900

Time (ms)
b)

50

40
Average atom number

30

42
20
41

40
10 39

38

0 200 210 220

0 50 100 150 200 250 300

L (μm)
Figure 5.23 a) A zoom in of Figure 5.19 during the atom evaporation. b) Number of atoms
left as a function of the distance between two barriers. A chain with up to 43 atoms can be
obtained with 100% rate of success.
126 Principles of the quantum simulator

drops as 1/ 6 , the interaction with the next nearest neighbor is still a few kHz, of the
same order of magnitude as the kinetic energy of the atoms.
Figure 5.23b also shows that one can control the number of trapped atoms from 1
to 43 with 100% of success rate. Aiming at a chain of 40 circular atoms, we stop the
compression as soon as reaches 208 µm. The mean inter-site distance is (4.8 ± 0.4
)µm.

c) Site matching: 896–1000 ms

This step prepares the transfer of the atomic chain into the optical lattice described
previously. The power of the weak barrier is increased up to the same as that of the
high barrier (4 MHz), to prevent further atom loss.
a)

Smooth Evaporation Stiff plug

b) 20
Position x (μm)

-20
850 900 950 1000
Time (ms)

Figure 5.24 a) A large waist helps to smoothen the ejection of atoms during the evaporation
while a small waist if more appropriate to ine adjust the position of the atoms. b) The positions
of the atoms during the site-matching adjustment.

It is now, important to reduce the waist of the barriers. A large waist during the
“evaporation” helps to smoothen the escape of atoms, thus maintaining some adia-
baticity as depicted in Figure 5.24a. However, we want to push only the atoms at the
two ends without disturbing the positions of the neighboring ones. A smaller waist is
better. We thus, linearly reduce the waists of the “plug” beams from 30 µm to 10 µm
 5.7 Arbitrary chain initialization 127

within ∼ 100 µs.

In the meantime, we slightly increase (to ∼ 214 µm) so that the inter-atomic dis-
tance matches the inter-site spacing of the lattice, i.e., 5 µm. Figure 5.24b plots the
positions of some atoms during this step. As the atoms get apart, from 4.8 µm to
5.0 µm, the van der Waals interaction energy is reduced as represented in Figure 5.21c.

d) Loading of the optical lattice

At this moment, the power of the CO2 laser is ramped up slowly in 100 ms, putting
each circular atom in an individual harmonic well. The inal state of a deterministic
chain of 40 atoms is reached when the optical lattice has the desired trap frequency (24
kHz). Despite an increase of the temperature during the compression of the lattice,
the inal temperature is smaller than 1.5 µK, corresponding to 15 kHz of average ki-
netic energy as observed in Figure 5.21. The residual oscillation amplitude is 100 nm.
This corresponds to about 4 oscillation quanta per mode, knowing that the spatial ex-
tension of the ground state wavefunction is Δ 0 = 50 nm. We reach the quantum limit
for describing the atomic motion.
Even once the chain is prepared, one should keep the two barriers during the rest
of the experiment. They help to compensate for the lack of van der Waals forces acting
on one side of the last atoms of the chain, putting their equilibrium positions back to
the bottom of the harmonic well.

5.6
Arbitrary chain initialization

As discussed previously, the phase diagram in Figure III.1 can be explored preparing
the simulator in the region � and let it adiabatically evolve into the exotic phase of
interest. In the region � , all the spin are polarized along the direction. This state
is obtained by applying a microwave pulse on the chain prepared above to globally
rotate all the spins from the direction (50 ) to the direction. Thus a local state
preparation is not necessary.
However, an arbitrary chain initialization would allow us to study quantum trans-
port in diferent situations. With trapped circular atoms, individual atom address-
ing is possible by locally changing the atomic spacing. Two transverse focused laser
beams along are used as optical tweezers, slightly displacing the two neighbors of
the atom of interest. Its van der Waals energy is accordingly shifted by a few MHz.
The microwave transition to the 48 state is thus shifted by the same amount. A few
µs microwave pulse will lip the spin of this atom without disturbing the others. The
process is repeated where necessary to obtain an arbitrary initialization of the chain.
The atoms at the two ends can be selectively addressed without the optical tweezers,
since their van der Waals energy is diferent from that of the “bulk” atoms.
During the initialization process, it might be necessary to stop the exchange inter-
action. Instead of 48 state, we prepare the corresponding atoms in the 46 state. The
exchange interaction between the 50 and 46 is below 5 mHz at a distance of 5 µm.
At the end, a hard microwave pulse brings the 46 atoms to the 48 state.
128 Principles of the quantum simulator

5.7
Site-resolved state-selective detection

The simulation result is read out by detecting the inal state of the individual atoms.
We need to irst freeze the evolution of the spin chain at a chosen time by halting the
exchange interaction. This is done by applying fast, strong microwave pulses, trans-
ferring the 48 state towards the 46 state, with which the exchange interaction of
the 50 is negligible§ . The two barriers are again set to uneven heights. One resumes
the compression of the chain. The atoms are evaporated one after the other. They
are guided to the region shown in Figure III.2 where the atoms can be state selec-
tively detected by ield ionization. Note that, during the simulation, the electric ield
applied on the spontaneous-emission inhibition capacitor is varied between 2 and
7 V/cm. If one transfers the 48 atoms to the 46 state at an electric ield of about 4
V/cm, the strong mixing of the ∣46 ; 50 ⟩ state with nearby elliptical pair states will
strongly reduce the lifetimes of the 46 and 50 states (Figure 5.14). To avoid this
detrimental mixing, one has to lower the electric ield to ∼ 2 V/cm before sending
atoms to the 46 state. The repulsive interactions between the 46 − 46 , 46 − 50
and 50 − 50 atoms are of the same order of magnitude, thus preventing the atoms
from getting too close during the compression, which can lead to the ionization of the
atoms.
The above scheme directly measures �� of each atoms. Additional spin-spin cor-
relation functions can be constructed by applying a hard pulse on the 48 − 50 tran-
sition before stopping the exchange interaction.

5.8
Efects of decoherence

For a quantum simulator, the longer coherence time, the better. We discuss here three
main sources of decoherence: the Stark efect, the average of the trapping potential
due to a inite site of the atoms and the spin-motion entanglement.

Stark efect

As discussed in section 3.3, stray ields due to contact potential between Rubidium and
the metallic surfaces of the the capacitor could be a problem. A study of the electric
ield in the vicinity of the atom chip, represented in Carla Hermann’s thesis, estimates
a residual electric ield smaller than 0.05 V/cm, with a gradient of about 0.4 V/cm2 at
∼ 1 mm away from the metallic surfaces. The diferential Stark shift of the 48 − 50
levels is of 560 kHz/(V/cm2 ). Taking into account the ∼ 200 µm extension of the
chain with 40 atoms, we estimate a maximum diference in the 48 − 50 transition
frequency between the last atoms at the two ends to be about 50 Hz. At a distance
of 5 µm, the exchange frequency (4 ) is about 70 kHz. The Stark broadening is thus
negligible at the scale of the exchange frequency.
§
Alternatively, one can send the 50� atoms to the 52� state. The 48� − 52� exchange interaction is
also negligible (<15 mHz).
 5.8 Efects of decoherence 129

Average of trapping potential

The ponderomotive energy acts on the electron as a nearly free charge. It does not
depend on the principal quantum number , at least for a uniform intensity. However,
the trapping potential (created by the transverse Laguerre-Gaussian beam and the
standing-wave optical lattice) is averaged over the electronic orbital. The radius of the
circular orbital is 2 0 ≈ 130 nm for the 50 and ≈ 120 nm for the 48 states. The
50 and 48 atoms, due to their diferent sizes, experience diferent potentials. This
leads to a shift in the 48 − 50 transition frequency, which depends on the position
of the atoms. The spatial extension of the atoms thus gives rise to a broadening of the
transition line. We numerically calculate and plot this shift as a function of the atom
position (note that the trap is isotropic) in Figure 5.25. The simulation of the chain
preparation estimates the residual oscillation of the atoms in the optical lattice to be
about 100 nm (see above). This corresponds to a line broadening of about 8 Hz, which
is negligible at the scale of the exchange frequency. Note also that, the decoherence

ΔU (kHz)
- 6.0

- 6.2

- 6.4

- 6.6

- 1.0 - 0.5
Δr (μm) 0.5 1.0

Figure 5.25 The frequency shift of the 48 − 50 transition as a function of the atom position.
The extension of the atom is represented by a red shaded Gaussian proile.

due to the Stark efect and the averaging of the trapping potential can be suppressed
using the Hahn spin-echo technique (see section 3.3).

Spin-motion entanglement

The entanglement between the motional and spin (atomic) states, induced by the ex-
change interaction, can lead to a collapse of the exchange oscillation. Let us consider
two atoms in two adjacent sites. Initially they are in the 50 state and at the bottom of
the trap. Suddenly lipping an atom to the 48 state makes the excitation (48 state)
transfer back and forth between the two atoms. Using the formula (E.25) derived in
Appendix E, Figure 5.26 plots the probability of detecting the irst atom in the 48
state, as a function of time. The modulation of the exchange oscillation is thus small.
The efect can be further reduced by tighten the trap. The spin-motion entanglement
with more than two atoms would be an interesting problem to study.
130 Principles of the quantum simulator

1.0

P48C, 50C (t)

0.5

0.0

0 20 40 60 80 100 120
Time (µs)

Figure 5.26 Probability of being in the 50 state for a pair of atoms initially in the ∣50 ; 48 ⟩
state, calculated using (E.25). The dashed line show the envelop of the oscillation.

Summary

We have presented a full experimental scheme to realize a quantum simulator of the

XXZ model for a 1D spin chain. It is based on the groundbreaking concept of trap-
ping circular Rydberg atoms. Extremely long-lived spin chains are provided using
spontaneous emission inhibition in a cryogenic environment. We also proposed an
innovative method to prepare deterministically a chain with up to 40 atoms, pushing
the capacity of the simulator out of the grasp of classical computation. Arbitrary chain
initialization and site-resolved state-selective detection are also proven to be realistic,
allowing the proposed system to address much larger range of 1D physics.
Conclusion and perspectives
“Ce qui embellit le désert, dit le petit prince, c’est qu’il cache
un puits quelque part...”

— Antoine de Saint-Exupéry, Le petit prince

Towards quantum simulation with low angular momentum Rydberg atoms

During this PhD work, we were interested in building quantum simulators based on
Rydberg atoms. Our irst direction was to study the physics of Rydberg excitation
in a ultra-cold atomic cloud. With the available experimental setup, we trapped and
cooled 87 Rb atoms down to sub µK temperature using a superconducting atom chip.
The excited Rydberg atoms are protected from blackbody radiation. Coating the chip
with a layer of rubidium provided us with a good control of the electric ield at the
chip vicinity. We observed unprecedented long coherence times for Rydberg atoms
near a metallic surface.
The study of the Rydberg excitation under strong blockade regime led us to several
important results. First we developed a novel tool to precisely measure the distribu-
tion energy of an interacting Rydberg ensemble. This technique is based on probing
the shift of the microwave transition frequency towards a nearby Rydberg level. We
successfully observed the hydrodynamic expansion of a Rydberg cloud under the re-
pulsive van der Waals forces between the atoms, and thus measured the limit of the
frozen Rydberg gas approximation.
With the obtained experimental data, we used numerical approaches to have an
insight into the Rydberg excitation process. The irst simple Monte Carlo model was
not powerful enough. We developed another more rigorous model that allowed us to
reproduce most of the observed experimental data, both the optical excitation proiles
and the microwave spectra. Note that the excitation was limited to the incoherent
lipping of the atoms. An improvement in the power of the blue excitation laser would
allow us to exploit the coherent collective behavior of the atomic cloud. This is one
interesting direction to follow.
Being fascinated by quantum simulations, we investigated quantum transport in a
1D chain of low angular momentum number Rydberg atoms. A careful consideration
pointed out that a regular chain, even as short as of a few atoms, is diicult to prepare
based on the facilitated excitation efect. However, these studies led us to the new
concept of building quantum simulators based on circular Rydberg atoms.
132 Conclusion and perspectives

Towards quantum simulation with a 1D chain of trapped circular atoms

Elaborating on the unique properties of circular Rydberg atoms, we proposed a new

platform for quantum simulation. It is based on a groundbreaking concept: the use
of laser-trapped circular Rydberg atoms. We have performed numerical simulations
and examined many technical details to make the proposal realistic. Its main features
can be summarized as follows

1. Extremely long-lived circular atoms, with lifetimes in the minutes range, by the
inhibition of spontaneous emission in combination with protection from black-
body radiations.

2. Strong interaction between the circular atoms and tuneability of the interactions.
The corresponding spin Hamiltonian is fully tunable over a wide range by
microwave dressing and electric ield variation.

3. Innovative deterministic preparation of a chain based on the repulsive interac-

tion between the atoms. A long chain up to 40 atoms is reachable.

4. Laser-trapping by ponderomotive force acting on the Rydberg electron. Atom

positions are thus well controlled at sub-µm precision.

5. Detection of any spin observable. The chain evaporation together with ield ion-
ization allows site-resolved state-selective detection of all spins.

In addition, the ability of arbitrary initialization of the spin chain enlarges the ad-
dressable range of quantum simulations with a trapped circular atoms. It opens a new
avenue for quantum simulations. An experimental realization of the proposed system
would lay a landmark in the evolution of quantum simulators.
In the following we discuss our short-term targets towards quantum simulations
as well as some extensions and several physical domains that can be simulated using
trapped circular atoms.

Beyond the grasp of the classical computation methods

Upgrades of the 1D system

Active cooling of a spin chain

From an experimental point of view, heating of the atoms is technically inevitable. In

our proposed system, it may come from the luctuations of the trapping-laser powers,
which are translated into the fast variations of trap frequencies. It may also come from
the micro-vibrations of lenses and mirrors which inally shake the trapped atoms. In
addition, if the adiabaticity throughout the deterministic chain preparation is some-
how violated, we will end up with a “hot” atom chain. Thermal motion of the atoms
could wash out coherence between them. Thus, an active cooling mechanism would
help to suppress these bothersome efects.
One way to include cooling into our system is to maintain a source of cold atoms in
low angular momentum Rydberg state at the two ends of the inal chain. These atoms
can be excited from two cloud of ultra cold ground state atoms trapped in two optical
 Conclusion and perspectives 133

tweezers. The dipole-dipole interaction between the 50 chain atoms and the short-
lived low Rydberg atoms will transfer kinetic energy out of the chain. After a few
cycles of Rydberg excitation, the whole chain can be cooled down close to its ground
state of motion. A further study is necessary to determine which low- Rydberg level
is relevant.

2D network of spins

The use of laser-trapped atoms makes it rather easy to extend the system to 2D ge-
ometries. A straightforward method is to prepare several 1D chains and bring them
together into a 2D coniguration. The 1D standing-wave along is shared between
the chains. At the end, one obtains a 2D square lattice, either parallel or perpendicular
to the quantization axis . The latter leads to an isotropic Rydberg interaction while
the former is an anisotropic interacting lattice.
A hexagonal lattice can also be achieved. Circular atoms are irst conined in be-
tween two laser light sheets. Four plug beams are added up, forming a 2D square
trap. One of the barrier is weaker than the others. Reducing the distances between
the four beams will induce the atom evaporation. The inal number of trapped atoms
should determined by the inal relative positions of the beams. The atoms self-order
into a 2D Wigner crystal similar to that in 2D Penning ion trap [64]. The crystal is then
transfered into a hexagonal optical lattice [177].

Quantum simulation with trapped circular atoms

We sketch here several speciic problems that are addressable with trapped circular
atoms. They can range from few-particle physics, which are numerically solvable to
benchmark the system, to many-body dynamics beyond the grasp of classical compu-
tation power.

Quantum transport along a spin chain

As soon as a regular chain of atoms can be realized, site-resolved state-selective de-

tection of the spins would make the study of quantum transport possible. Due to the
border efect, one or both ends of the chain is microwave promoted into the 48 state.
Exchange interaction with the neighboring 50 atoms will propagate the excitation
along the chain.
Local addressing of the initial spin states allows one to excite an atom in the middle
of the chain into the 48 state. The probability amplitudes of being excited symmetri-
cally propagates to the two ends, bounce and return to the initial position, where they
interfere and make the excitation revive. Figure 3a represents a transport of an 48
excitation over a perfectly regular chain with 41 atoms as an example. Initially all the
atoms are in the 50 state. One lips the 21th atom to the 48 state and calculates the
excitation probability for each atom (color scale).
Furthermore, one can prepare a chain with several excited atoms. Quantum trans-
port with multiple excitations in a long chain is not easy to predict with classical com-
putation. The strong 48 − 48 interaction between the excitations would lead to
non-trivial quantum transport.
134 Conclusion and perspectives

5 a)
10
15
20 1 .0
25 0 .9
30 0 .8
0 .7
35
0 .6
40
0 .5
5 b) 0 .4
0 .3
10
0 .2
15
0 .1
20 0 .0
25
30
35
40
0 20 40 60 80 100
c) Time (μs)

Figure 3 Excitation transport on a 1D chain with 41 atoms. a) The atoms are regularly sepa-
rated by 5 µm. b) The deviation of the atom from the trap center is a Gaussian proile of 0.5 µm
width. Initially all the atoms are prepared in the 50 state, but the 21th one in the 48 state.

Disorder and localisation

Imprinting a laser speckle pattern, one can introduce disorder into the chain. Tuning
the relative strength of the disorder and the interaction will allow us to realize the
phase transition from conducting state (non-localized state) to isolating state (localized
state) as the disorder increases. Figure 3b represents as an example the time evolution
of a chain with 41 atoms, whose positions deviate from the trap center by an amount
randomly sorted from a Gaussian proile of 0.5 µm width. The excitation is “trapped”
in the middle of the chain. Our model would provide a tool to study the many-body
localization from the point of view of level statistics, which can be measured using our
microwave tool (see chapter 4). Signatures of chaoticity of the Hamiltonian can also
be looked for.

Slow dynamics and quenches

The phase diagram of the spin Hamiltonian is of great interest. Its structure has
been theoretically predicted as shown in Figure III.1. With trapped circular atoms, by
adiabatically varying Ω and/or � , we will systematically reconstruct the phase dia-
gram. A comparison between experimental results and theoretical prediction would
give us an insight into the nearest neighbor spin chain.
Moreover, taking advantage of the long-lived chain and the fully controlled Hamil-
tonian, for instance, by slowly modifying the spin chain state, in a time scale much
 Conclusion and perspectives 135

longer than 1/(4 ), one could investigate adiabatic processes in the quantum mechan-
ical regime. It would provide insights into the limitations of adiabatic quantum com-
puting [178].
For sudden changes of the Hamiltonian parameters, the evolution of the system
after the quench is rather complex since it involves highly excited states of the inal
Hamiltonian [179]. Predictions using classical computation methods are diicult tasks
since they involves a large portion of the Hilbert space. Studying this non-trivial dy-
namics requires a lexible quantum simulator such as trapped circular atoms.

Spin dynamics and atomic motion

In the described system, atomic motion has negligible inluence on the spin states.
However strong spin-motion entanglement can be created by either loosening the trap
or using resonant exchange interaction between and ( + 1) states. By coupling
the spin chain with a phonon bath, interesting behavior of the spin chain dynamics in
the presence of dissipation can be studied with controlled parameters. The transport
of entanglement, beyond that of electronic excitation, would be possible in a similar
context to that in [180, 181].

2D physics

S=1
J
a)
J⟂

J
b)
J
J2

Figure 4 Two diferent mappings to realize an efective = 1 chain from two coupled = 1/2
Heisenberg chains. Figure extracted from [182].

The realization of 2D-spin networks would open more possibilities to simulate

problems where modelization and numerical approaches are challenging. As an ex-
ample, a square lattice with properly tuned anisotropic interactions would lead to a
situation in which anti-ferromagnetic chains are ferromagnetically coupled (Figure 4).
Such a 2D lattice is mapped on to an equivalent spin-one chain in the Haldane phase
[182]. Its non-trivial topological order has not yet been probed from many-body mea-
surements. The discovery of this topological phase was awarded the Nobel prize in
2016.
136 Conclusion and perspectives

Realization of irst building blocks for quantum simulation with trapped

circular atoms

With these ideas in mind, we will step by step conduct proof-of-principle experiments
to demonstrate the main features of quantum simulation with trapped circular atoms.
Installing atom-circularization electrodes inside the present setup, we would be able
to prepare an ensemble of 50 atoms. Successfully trapping them with a Laguerre-
Gaussian 1 µm-wavelength laser beam will encourage us with the proposed scheme.
The demonstration of extremely long-lived circular Rydberg atoms would be then
possible by placing the trapped ensemble of 50 atoms in between capacitor plates.
An extra modiication of the current cryostat in order to operate with liquid 3 He is nec-
essary. At this point, a long coherence time for the 48 −50 transition will be probed.
If successful, these long-lived strong interacting qubits by themselves will ind their
applications in quantum information processing such as quantum gates based on the
dipole blockade mechanism extended to circular atoms.
Adding two plug beams will allow us to prepare a deterministic chain of circular
atoms. If the protocol works, even with a short chain, it provide us a tool to bench-
mark our quantum simulator. Microwave spectroscopy probing of the van der Waals
interaction energy of the chain under diferent static electric ields will prove the tune-
ability of the spin Hamiltonian.
A site-resolved detection of the spin states if successfully implemented, will be the
inal validation of our proposal. We would be able to simulate the quantum transport
with one or multiple excitations as well as explore the phase transition of the
spin Hamiltonian. Furthermore, we would be able to push quantum simulations with
trapped circular atoms beyond the grasp of classical computation methods.
Appendices
Electric dipole matrix
element estimation A
The electric dipole matrix element can be written as in (1.23) and (1.22), as a product
of a radial part ⟨ ′ ′ ′ ∣ ∣ ⟩ and an angular part ⟨ ′ (�′ ) ′�(�) ∣ 1 ∣ (�) �(�) ⟩.
�
Here
we show how to calculate each of them using numerical methods.

A.1
Radial matrix element

We rewrite the Schrödinger equation (1.26) for the radial wavefunction of an electron
in the Coulomb potential

�2 2� 2 ( + 1)
+ + [2 + − ] =0
� 2 � �∗ 2
(A.1)

We follow the procedure of Zimmerman et al. [183] by making the substitutions

= ln( ) , (A.2)

and

( ) = √( ) (A.3)

into (A.1). We obtain

d2
= ( ) ( ),
d 2
(A.4)

where

1
2
( )=2 2� (− −� − )+( + ) .
�∗
2
(A.5)

Equation (A.4) can be numerically solved by the Numerov algorithm [184]. We need
to provide �∗ , and two initial guesses of explicitly. Knowing the radial wavefunc-
tions, one can obtain the radial matrix element with no diiculty.
140 Appendix A: Electric dipole matrix element estimation

A.2

Angular matrix element

The angular matrix element is developed into an analytical form using a Clebsch Gor-
dan coeicient and the Wigner-Eckart theorem [185]:

⟨′ ′∣
1∣ �⟩ = ⟨ ′‖ 1‖ ⟩⟨ ′ ′∣ 1 ⟩
�
� � �
′ 1
= ⟨ ′‖ 1‖ ⟩ (−1)� −�� √2 ′ + 1 ( ) ,
′ ′ (A.6)
− ′
� �

where ⟨ ′ ‖ 1 ‖ ⟩ is the reduced matrix element and the Clebsch Gordan coeicient is
written in terms of Wigner 3-� symbol. For low states, one has to include the ine
structure. The angular matrix element is written as

′ + 1⎛
�′ 1 � ⎞
⟨ ′ �′ ′∣
1∣ � � ⟩ = ⟨� ‖ 1 ‖�⟩ (−1) 2� ⎜ ⎟ .
� ′ �′ −�′�

⎝−
� ′
�⎠
(A.7)
�
√

Factoring out the � and �′ dependence of the reduced matrix element into a Wigner 6-�
symbol, and a reduced matrix element that depends only on and ′ , we get

⟨�′ ∥ 1 ∥ �⟩ = ⟨ ′ ′ �′ ∥ 1∥ �⟩
� �

�′ 1 �
= ���′ √2� + 1√2 ′ + 1(−1)1+� +�+� { ′} ⟨
′∥
1∥ ⟩ ,
′ ′ � (A.8)
′

where = ′ = 1/2. Knowing that

3 1 ′
⟨ ′∥ 1∥ ⟩ = (−1)1−� √2 + 1( ),
�
√ 4� 0 0 0
(A.9)

The angular matrix elements (A.6) and (A.7) can be expressed using (A.8) and (A.9) as

3
⟨′ ′∣
1∣ �⟩ = (−1)1−�+� −�� √(2 + 1)(2 + 1)×
� ′ ′
′
�
√ 4�
′ 1 1 ′
(A.10)
×( )( )
− �′
� 0 0 0

and

3
⟨ ′ �′ ′| �| � �⟩ = (−1) (2� + 1)(2�′ + 1)(2 + 1)(2 ′ + 1)×
1+�′ +�′ +�−�′�
√ 4� √
� 1

� 1 �′ ⎛ �′ 1 � ⎞ 1 ′
(A.11)
×{ ′ }⎜ ⎟( ) .
1/2 ⎝− � �⎠ 0 0 0
′
 A.2 Angular matrix element 141

The Wigner 3-� symbol is estimated using Racah formula [186, 187]

( ) =(−1)�−�−� √Δ( )×
� � �
× √( + �)! ( − �)! ( + �)! ( − �)! ( + �)! ( − �)!× (A.12)
(−1)�
×∑ ,
�
()

where

( ) = ! ( − + + �)! ( − + − �)! ( + − − )! ( − − �)! ( − + �)! . (A.13)

The sum runs over all integers for which the factorial in ( ) all have non-negative
arguments. Similarly for the Wigner 6-� symbols

(−1)� ( + 1)!
{ } = √Δ( )Δ( ��)Δ(� )Δ(�� ) × ∑ ,
� � � ()
(A.14)
�

where

( ) = ( − − − )! ( − − � − �)! ( − � − − �)! ( − � − � − )! ×
× ( + + � + � − )! ( + + � + � − )! ( + + � + � − )! .
(A.15)

The sum also runs over all integers for which the factorial in ( ) all have non negative
arguments.
In both formulae, Δ( ) is a triangle coeicient given by

( + − )! ( − + )! (− + + )!
Δ( )= .
( + + + 1)!
(A.16)

To speed up the numerical estimation of the factorials, one can use Stirling’s approxi-
mation of the factorial for large arguments

ln !≈ ln − + ln(√2� ) . (A.17)
142 Appendix A: Electric dipole matrix element estimation
 B
Laser frequency
stabilization and laser
distribution

960 nm
960 nm
diode laser
Stabilization of
Stabilization of FP 960 nm laser on FP Frequency
on saturated absorption doubling cavity
spectroscopy
FP cavity

+195 MHz

+200 MHz 480 nm

+283 MHz
Rb cell - 200 MHz

480 nm
excitation laser

Optical fiber
AOM
Mirror
Half-wave plate
780 nm Stabilization of
+500 MHz

excitation laser 780 nm laser on FP Quarter-wave plate Dichroic mirror

Polarized beam splitter Photodiode

EOM (phase modulator)
780 nm EOM (polarization modulator)
diode laser Frequency
reference for
MOPA

Figure B.1 Scheme of laser frequency stabilization for the 780 nm and 480 nm wavelength
excitation lasers.

Figure B.1 depicts the scheme used to stabilize the lasers. The 780 nm-wavelength
excitation laser is a Toptica DL Pro laser. The 480 nm-wavelength laser is provided by
a Toptica TA-SHG 110, which consists in a Master oscillator power ampliier (MOPA)
at 960 nm wavelength, followed by a frequency doubling cavity. Both the 780 nm and
the 960 nm laser are frequency stabilized to a relection peak of a Fabry-Perrot cavity,
144 Appendix B: Laser frequency stabilization and laser distribution

whose length is stabilized to the crossover of the = 2 − ′ = 2 and = 2 − ′ = 3

transitions. In order to lock the laser frequencies, modulations at 20 MHz are applied
on the phase of the 780 nm laser using an EOM, and on the current though the 960 nm
laser diode (frequency modulation). Note that the resulting sidebands of the 960 nm
laser is iltered out by the phase matching condition of the frequency doubling cavity.
As depicted in Figure B.2, the lasers necessary for the trapping and cooling of 87 Rb
atoms are extracted from a Toptica TA-110 780 nm-wavelength MOPA. The 780 nm
MOPA is ∼ 160 MHz frequency-ofset from the 780 nm excitation laser. Another laser
whose frequency is stabilized to the = 1 − ′ = 2 transition, is used to repump the
atoms that are scattered out of the cooling cycle into the = 1 state.
2D-MOT 3D-MOT Front Side Optical
Offset-frequency
pumpers pumpers probe probe pumper
stabilization

Frequency
reference for
MOPA
780 nm
MOPA
diode laser
780 nm

+80 MHz +100 MHz -133 MHz

+80 MHz

Half-wave plate AOM

Rb cell
Quarter-wave plate
Photodiode
Mirror
Stabilization repumper laser Optical fiber
on saturated absorption spectroscopy
Polarized beam splitter

Figure B.2 Distribution of lasers for cold atom experiment. The superposition with repump-
ing lasers (gray) is also shown.
Measurement of the
trap dimension C
Our imaging systems have a limited spatial resolution of about ∼ 10 µm. The dimen-
sions of a dense atomic cloud is, in general, below this limit. A direct in situ imaging
cannot give a good enough estimation. Another approach is making use of the relation

��2 ��2 = � , (C.1)

where �� and �� are the trap frequency and the trap width ( −1/2 position) along the �
direction. We need thus to measure the cloud temperature and the trap frequencies.

C.1
Temperature measurement

The temperature is measured using the time-of-light technique, which maps the mo-
mentum distribution into a spatial distribution according to

��2 ( ) = ��2 (0) + Δ 2 = � 2 (0)

� � + � 2 , (C.2)

where is the free fall duration from rest. Measuring the trap size at diferent times
allows us to extract the temperature . For the trap described in chapter 4, the mea-
surement gives = 500 ± 150 nK.

C.2
Trap frequencies

Giving a small kick to the atomic cloud (for example instantaneously moving the trap
center and then going back), makes the cloud oscillate around the trap bottom. The
oscillation frequencies are the trap frequencies along the three main axes, given that
the oscillation amplitudes are small. Experimentally, the position of the cloud after
the kick is sampled as a function of time. Its Fourier transform reveals the main trap
146 Appendix C: Measurement of the trap dimension

60
60
Horizontal center (μm)
40

Vertical center (μm)

40
20
20

0
0

-20
-20

-40 -40

-60 -60
0.2 0.3 0.4 0.5 0.6 0.2 0.3 0.4 0.5 0.6

Time (s) Time (s)

2.5 2.5

Horizontal center Vertical center

2.0
244.4 2.0
262.2
Amplitude (μm)

Amplitude (μm)
1.5 1.5

244.4
1.0 262.2 1.0

46.7 46.7
0.5 0.5

0.0 0.0

0 50 100 150 200 250 300 0 50 100 150 200 250 300

Frequency (Hz) Frequency (Hz)

Figure C.1 Trap frequency measurement. Top: the position of the atomic cloud as a function
of time after the kick along the horizontal and the vertical direction (relative to the camera).
Lower: The corresponding Fourier transform of the cloud position. The probe beam is sent
along the direction.

frequencies as shown in Figure C.1. They are

�� = 2� × 46.7 Hz
�� = 2� × 244.4 Hz (C.3)
�� = 2� × 262.2 Hz .

Plugging into equation (C.1) we obtain the trap dimension of

�� = 2� × 23.6 µm
�� = 2� × 4.5 µm (C.4)
�� = 2� × 4.2 µm .
Fourier limited line
width D
1.0

0.8

0.6
Pf

0.4

0.2

0.0
−6 −4 −2 0 2 4 6

Figure D.1 Probability of being excited into state ∣ ⟩ of an atom initially in state |�⟩ as a function
of reduced detuning Δ̃ � .

Let’s consider an atom initially in a state |�⟩ undergoes a transition by microwave

to a state ∣ ⟩. The microwave frequency is detuned by Δ� away the resonance. The
population in the inal level oscillates as a function of time

Ω2� 1
�( ) = sin2 ( Ω2 + Δ�2 )
Ω2� + Δ�2 2√ �
(D.1)

and is known as Rabi lopping. The coupling strength is represented by the Rabi fre-
quency Ω� .
The excitation duration Δ � is chosen so that it makes a � pulse for the resonant
excitation
�
Δ = .
�
Ω�
(D.2)
148 Appendix D: Fourier limited line width

Substituting Ω� from (D.2) into (D.1), we obtain

1 �
= sin2 ( 1 + Δ̃ �2 ) ,
�
1 + Δ̃ �2 2√
(D.3)

where Δ̃ � = Δ� Δ � /�. Equation (D.3) is similar to (but not identical to) a cardinal sine
function squared and is plotted in Figure D.1. Accordingly, the excited state ∣ ⟩ can be
populated with a nonzero probability for near-resonant excitations. Therefore, a inite
excitation duration leads to a inite spectral line width. It is of about 2�/Δ � .
For an ensemble of Rydberg atoms initially in the |�⟩ state, the van der Waals inter-
action shifts the transition frequency of each atom correspondingly. Let (�� ) be the
energy distribution, i.e., the number of atoms whose transition frequency is �� . The
average number of atoms being excited to the ∣ ⟩ state coupled by a microwave pulse
of frequency � is then given by

� (�) = ∑ (�� ) � (� − �� ) . (D.4)

��

If the energy distribution is a continuous function, the number of Rydberg atoms

whose transition frequency is between �� and �� + d�� is written as (�� )d�� . The aver-
age number of atoms being excited is expressed by an integral

� (�) =∫ (�� ) � (� − �� )d�� . (D.5)

��

Equations (D.4) and (D.5) are the convolution of the energy distribution (�� ) and the
spectral broadening function � of the microwave pulse.
Two atoms: motion
and spin exchange E
Let us discuss the simplest case of two atoms in two lattice sites 0 = 5 µm apart. They
are in two diferent states ∣ ⟩ and | ⟩. We consider only a one-dimensional motion in
the trap, along the two atom axis . The trap frequency in this direction is �� . As
discussed in section 2.3, the efective interaction Hamiltonian of the system is

| ⟩ | ⟩
| ⟩⎛ �� ⎞
= ⎜ ��
⎟= �� Σ� + �� Σ� ,
| ⟩⎝ ⎠
(E.1)
�� ��
ef

where Σ� = ∣ ⟩⟨ ∣ + ∣ ⟩⟨ ∣ = � and Σ� = ∣ ⟩⟨ ∣ + ∣ ⟩⟨ ∣ are respectively the

- and the -Pauli matrices in the space of pair states. The direct and the exchange
interactions are �� and �� respectively. The exchange interaction symmetrically and
anti-symmetrically combines the two bare states into |±⟩ = (∣ ⟩ ± ∣ ⟩)/√2. Their
energies are

± = �� ∓ �� . (E.2)

The direct interaction globally shifts these two levels and thus plays no role in the
dynamics of the system. Since �� and �� depend on the distance between the two
atoms, when the two atoms are in the |±⟩ state, they are symmetrically displaced from
the trap center by an amount 0 < 1 = − 2 = � ∓ � , where � and � are respectively
the displacement due to the direct and the exchange forces. One can see that � only
slightly modiies the trap center of each atom. By redeining the energy origin and 0 ,
we can neglect the direct interaction.
The motion of the atoms is described by two harmonic oscillators

��� = ℏ�� ( †
1 1 + † )
2 2 , (E.3)

where � and †� are the annihilation and creation operators acting on the atom � re-
spectively. The total Hamiltonian is thus

= ℏ�� ( †
1 1 + † )
2 2 + �� � . (E.4)

�/
�,
We assume that the exchange interaction varies where � is a proportionality
150 Appendix E: Two atoms: motion and spin exchange

coeicient. For small displacements � ≪ 0, one can write

�
�� ≈ � [1
�
− ( 2 − 1 )]
0 0
Ω0 �
=ℏ [1 − ( − 1 )] ,
2 2 (E.5)
0

where we introduce the exchange frequency Ω0 = 2 � /(ℏ �0 ). Let be the mass of

one atom. Using � = 0 ( � + � ), where 0 = √ℏ/(2 �� ) is the oscillation ground
†

state extension, we inally get

Ω0
= ℏ�� ( † + † ) +ℏ [1 − �( − + ℎ. )]Σ� ,
1 1 2 2 2 2 1 (E.6)

where we deine the analog of a Lamb-Dicke parameter for ion traps

�=� 0
. (E.7)

If the two atoms are in the |±⟩ state, the displacement of the atoms can be approx-
imated as ∣±�⟩ ∣∓�⟩ describing products of coherent states with opposite real ampli-
tudes. In this notation, the irst ket refers to the motion of the irst atom. One can
estimate � using the variational method, minimizing the average value of in the
state |+⟩ ∣�⟩ ∣−�⟩. One gets

Ω0 �
�= .
2��
(E.8)

We now perform a unitary transformation deined by

= 1 (�1 Σ� ) 2 (Σ� ) , (E.9)

where the modiied displacement operators � are

� (�� � ) = .
(�� �†� −��∗ �� )Σ�
(E.10)

† (� Σ ) =
Since the -Pauli matrix commutes with all motional operators, � � � � (−�� Σ� ).
The new Hamiltonian reads

̃ = 1 (−�1 Σ� ) 2 (−�2 Σ� ) 1 (�1 Σ� ) 2 (�2 Σ� ) . (E.11)

According to the Baker-Hausdorf lemma, we have

� (−�� Σ� ) � � (�� Σ� ) = � + � � Σ� , (E.12)

� (−�� Σ� ) � � (�� Σ� ) = + ��∗ Σ� ,
† †
� (E.13)
(E.14)

and

� (−�� Σ� ) � � � (�� Σ� ) = + �� Σ� + ��∗ Σ� + |�� |2 ,

† † †
� � � � (E.15)
 Appendix E: Two atoms: motion and spin exchange 151

where we have used Σ2� = �. We get

̃ =ℏ�� ( Ω0
† + Σ + ℏ�� (|�1 |2 + |�2 |2 )+
† ) +ℏ
1 1 2 � 2 2

Ω � Ω � Ω �
(E.16)
ℏ�� [(�1 + 0 )Σ� †1 + (�2 + 0 )Σ� †2 + 0 (�1 + �2 ) + ℎ. ] .
2�� 2�� 2��

When choosing

Ω0 �
�1 = �2 = −� = − ,
2��
(E.17)

within irrelevant constants, the Hamiltonian simpliies to

̃ = ℏ�� ( Ω0
† + † ) +ℏ Σ .
1 1 2 2 2 �
(E.18)

This Hamiltonian describes two atoms undergoing an exchange, together with two
uncoupled harmonic oscillators.

We now consider the time evolution of the system from a simple initial condition
|Ψ⟩ (0) = ∣ ⟩ |0⟩ |0⟩, i.e., the two atoms in | ⟩ and ∣ ⟩ at the trap centers. We can de-
compose |Ψ⟩ (0) as

1
|Ψ⟩ (0) = (|+⟩ |0⟩ |0⟩ + |−⟩ |0⟩ |0⟩) . (E.19)
√2
The corresponding initial unitary transformed state is

1
|Ψ⟩(0)
̃ = (|+⟩ ∣�⟩ ∣−�⟩ + |−⟩ ∣−�⟩ ∣�⟩) , (E.20)
√2

which is a two oscillators Schrödinger cat. The time evolution of the system under ̃
is trivial and is

̃ )= 1 (
|Ψ⟩( |+⟩ ∣� ⟩ ∣−� ⟩+ |−⟩ ∣−� ⟩ ∣� ⟩) . (E.21)
Ω0 � Ω0 �
� 2
−� �2� � −� �2� � −� 2
−� �2� � −� �2� �

√2
We now perform the inverse unitary transformation and inally obtain

1
|Ψ⟩ ( ) = ( |+⟩ ∣−� ⟩ ∣� ⟩ + |−⟩ ∣� ⟩ ∣−� ⟩) ,
Ω0 � Ω0 �
� 2
−� 2 (E.22)
√2
with

� = �(1 − ).
−� �2� �
(E.23)

The motional and the atomic states are thus generally entangled. The entanglement
disappears periodically at the trap frequency, when the cat amplitude � cancels. One
152 Appendix E: Two atoms: motion and spin exchange

1.0

0.8

β = 0.1
0.6 0.2
Contrast

0.5
0.8
0.4

0.2

0.0
0 1 2 3 4 5
ωxt/2π

Figure E.1 Contrast of the exchange oscillation as a function of time for diferent value of �

can rewrite |Ψ⟩ ( ) as

1
|Ψ⟩ ( ) = ∣ ⟩( ∣−� ⟩ ∣� ⟩ − ∣� ⟩ ∣−� ⟩)+
Ω0 � Ω0 �
� −�
2
2 2

1
(E.24)
∣ ⟩( ∣−� ⟩ ∣� ⟩ + ∣� ⟩ ∣−� ⟩) .
Ω0 � Ω0 �
� −�
2
2 2

The probability of inding the irst atom being excited is

1 1
�� ( ) = (1 − ∣⟨� ∣ −� ⟩∣ cos Ω0 ) = (1 − cos Ω0 ) .
2 −4|� |2
2 2
(E.25)

The contrast of the exchange oscillation is thus

= −8�2 (1−cos �� �)
. (E.26)

Figure E.1 plots the contrast of the exchange oscillation as a function of time. In the
limit of a small � amplitude (Ω0 ≪ �� /� weak exchange and tight trap), the contrast
is slightly modulated

≈ 1 − 8�2 (1 − cos �� ) . (E.27)

For large �, the contrast exponentially decays in a short time scale. Equation (E.26)
can be approximated by

≈ −4�2 ��2 �2
. (E.28)
Résumé en français

Simulation quantique et état de l’art

Les transitions de phase des systèmes magnétiques quantiques ont toujours suscité un
grand intérêt à la fois du point de vue expérimental et théorique. Plusieurs modèles
de réseaux de spins en interaction ont été développés pour décrire ces systèmes à une
échelle microscopique. Cependant, les méthodes de calcul classiques, malgré leurs
récents développements, ont des diicultés à traiter ce type de problèmes à -corps
fortement corrélés. La simulation de la dynamique d’un réseau de spins en interac-
tion demande de calculer l’exponentielle de matrices à 2� éléments. Quand ≥ 40,
le problème est hors de portée de nos super-ordinateurs actuels [14–17] .
Richard Feynman, pendant ses cours des années 1980, a proposé de transcrire
le système de spins d’intérêt sur un autre qui conserve les propriétés du système
d’intérêt mais est plus facile à contrôler et étudier expérimentalement [11]. De façon
générale, l’on prépare le système dans un état initial et le laisse évoluer vers l’état
quantique inal. On mesure ensuite quelque quantité d’intérêt.
La simulation quantique devient un domaine très actif. Elle est étudiée théorique-
ment et expérimentalement dans le monde entier. Plusieurs plateformes ont été dévelop-
pées : des ions piégés [27, 28], des qubits supraconducteurs [29], des atomes neutres
en cavité ou dans des réseaux optiques [23–26], des boîtes quantiques [34–36], des
molécules polaires [33] et beaucoup d’autres.
Les qubits supraconducteurs peuvent être manipulés avec une très haute idél-
ité [41, 42]. Ce sont des systèmes très prometteurs pour la simulation quantique
numérique [43, 44] ou analogique [45, 46]. Néanmoins, ces systèmes soufrent d’une
forte décohérence [49].
Les ions piégés ont permis la mise en œuvre la plus avancée de simulations quan-
tiques numériques [55, 56], notamment la simulation de la création d’une paire de par-
ticules en l’électrodynamique quantique [57]. Des simulations quantiques analogiques
avec des ions piégés ont aussi été menées, avec la réalisation de systèmes de spins
frustrés, la propagation de l’intrication [58, 59] ou la localisation à -corps [60], par
exemple. Toutefois, la plupart de ces réalisations ont été efectuées avec une chaîne 1D
et une interaction de longue ou moyenne portée [61, 62]. Pour passer à la géométrie
2D, il a été proposé d’utiliser une matrice de micro-pièges [67, 68] ou un piège Paul à
2D [65, 66], mais la cohérence de ces systèmes n’a pas encore caractérisée.
Les atomes froids en réseaux optiques sont des outils remarquables pour la simula-
154 Résumé en français

tion quantique analogique [24–26]. Les atomes sont placés dans des puits de potentiel
optiques à 1D, 2D ou 3D. La dynamique du système vient de la compétition entre
l’efet tunnel d’un site à son voisin et l’interaction de deux atomes dans le même site.
Ce sont des systèmes très lexibles qui permettent de réaliser diférentes géométries
et de contrôler de nombreux paramètres expérimentaux. Cependant, le taux tunnel
est faible (quelques centaines de Hz) par rapport au temps de vie des atomes piégés
(quelques secondes). L’évolution du système est donc limitée à quelques centaines
de périodes tunnel. C’est un gros désavantage lorsque l’on cherche à simuler la dy-
namique lente de verres de spins ou la localisation à corps. Passer à des molécules
polaires [79, 80] ou des atomes avec un grand moment dipolaire magnétique [81, 82]
n’est pas aisé.

Atomes de Rydberg et simulation quantique

Les atomes dans des états de Rydberg sont des atomes très excités (nombre quantique
principal élevé), très proches du seuil d’ionisation [3, 83]. Ils ont des propriétés re-
marquables, et même très exagérées par rapport à celles des atomes dans des états
“ordinaires”. Parmi ces propriétés, on peut citer notamment leurs très grands élé-
ments de matrice dipolaire, leurs très long temps de vie et malgré tout une structure
interne relativement simple. Les éléments de matrice dipolaire des atomes de Ryd-
berg sont typiquement mille fois ceux des atomes fondamentaux. L’état de Rydberg
peut vivre plus qu’une centaine de microsecondes. En particulier, l’état de Rydberg
circulaire, c’est à dire l’état de Rydberg de nombres quantique orbital et magnétique
maximaux, a un temps de vie de l’ordre de la dizaine de millisecondes. Ses propriétés
ont été exploitées par une série d’expériences d’électrodynamique quantique en cavité
réalisées dans notre équipe du Laboratoire Kastler Brossel [84–95].
Avec leurs grands éléments de matrice dipolaire, les atomes de Rydberg inter-
agissent fortement entre eux. L’interaction prend la forme d’un potentiel de van der
Waals 6 / 6 , où est la distance entre deux atomes. Si est de quelques microns,
l’énergie d’interaction peut entrer dans la gamme des MHz, et même plus encore pour
l’interaction d’échange résonant de Föster (qui varie comme 1/ 3 ). On peut changer
cette énergie d’interaction par des ordres de grandeur en modiiant le nombre quan-
tique principal.
Un protocole pour la simulation quantique numérique avec atomes de Rydberg
a été proposé [97, 98]. Il est basé sur le mécanisme de blocage dipolaire qui est la
conséquence directe de l’interaction forte entre les atomes de Rydberg [99]. Dans un
petit échantillon froids d’atomes fondamentaux, l’interaction de van der Waals est plus
importante que la largeur de l’excitation (∼ 100 kHz). Dès qu’un atome est excité, tous
les autres sont mis hors résonance. Par conséquent, nous avons un et seulement un
atome de Rydberg à la fois. C’est au cœur du principe de la réalisation d’une porte
quantique CNOT. Selon l’état du qubit de contrôle, l’état de tous les autres qubits est
inchangé ou est inversé [96, 99, 102–104].
Grâce à cette forte interaction, les atomes de Rydberg sont aussi des outils très
prometteurs pour la simulation quantique analogique de réseaux de spins. L’intérêt
pour ces atomes a récemment explosé, menant à de nombreuses propositions pour la
simulation quantique avec des atomes de Rydberg [105–108]. Parmi plusieurs réalisa-
 Résumé en français 155

tions récentes, l’observation du transport cohérent d’une excitation sur une chaîne de
trois atomes de Rydberg est très remarquable [109]. La dynamique d’une excitation
sur une matrice avec plus de 20 atomes a été observée [110]. Pour résoudre le prob-
lème du faible taux tunnel dans les systèmes d’atomes froids dans réseaux optiques, il
est possible de mélanger les atomes fondamentaux avec des atomes de Rydberg, c.-à-
d. de les habiller avec des états de Rydberg [111–115]. Toutefois, ces expériences sont
opérées à 300 K et soufrent du rayonnement du corps noir qui modiie l’interaction
entre les atomes de Rydberg de façon imprévue [116–118]. Passer dans un environ-
nement cryogénique est une tâche ardue.
Jusqu’à maintenant, les atomes de Rydberg utilisés dans les expériences ci-dessus
ont toujours des moments angulaires faibles. La simulation quantique avec des atomes
de Rydberg de faible sont limitées. D’abord, les atomes de Rydberg ressentent des
forces mécaniques fortes, répulsives ou attractives, à cause de leurs interactions fortes.
Sans piégeage, n’importe quel cristal de Rydberg fond en quelques dizaines de mi-
crosecondes. Des techniques de piégeage des atomes de Rydberg dans des réseaux
optiques ont été étudiées [119]. Néanmoins, le temps de vie des états de Rydberg est
fortement raccourci à cause de la photoionisation par le laser de piégeage [120]. Par
ailleurs, si le temps de vie des atomes de Rydberg est long par rapport à celui des états
excités ordinaires, il n’est que de quelques centaines de microsecondes, ce qui corre-
spond à une centaine de périodes d’échange. Donc la simulation de la dynamique
lente d’un réseau de spins est hors de portée.
Dans ce contexte, cette thèse étudie l’interaction entre des atomes de Rydberg et
explore de nouveaux plans pour réaliser des simulations quantiques avec des atomes
de Rydberg. Expérimentalement, nous excitons les atomes vers des états de Rydberg
par laser, de façon résonante ou désaccordée. Les atomes fondamentaux de rubid-
ium sont piégés et refroidis dans un nuage dense, préparé en utilisant une puce à
atome supraconductrice. Parce que les atomes de Rydberg sont très sensibles au ray-
onnement thermique, l’expérience est efectuée dans un environnement cryogénique
auquel la puce supraconductrice est parfaitement adaptée. Les travaux de cette thèse
ont contribué aux quatre réussites principales résumées ci-dessous.
Contrôle du champ électrique parasite et manipulation cohérent de niveaux de
Rydberg au voisinage de la puce. L’inconvénient des grands éléments de matrice
dipolaire est que les atomes de Rydberg sont extrêmement sensibles aux champs élec-
triques parasites. Le dépôt lent d’atomes de rubidium sur la surface d’or de la puce
est inévitable pendant une séquence expérimentale. Cela crée des taches de dipôles,
créant un champ électrique inhomogène et instable au voisinage de la puce. Par ef-
fet Stark, le proil de l’excitation s’en trouve élargi (∼ 40 MHz). L’efet de blocage
dipolaire est donc détruit complètement. Résoudre ce problème était un déi majeur
pour nous ainsi que pour d’autre groupes. Une solution, plutôt simple, consiste à
couvrir la puce d’une couche épaisse de rubidium sur une large zone, saturant ainsi
l’efet du dépôt lent de rubidium. L’inhomogénéité du champ électrique est alors bien
supprimée. Nous disposons donc d’un bon environnement pour étudier l’interaction
entre atomes de Rydberg. Ces travaux sont détaillés dans la thèse de Carla Hermann
et ont fait l’objet d’un hightlight dans Physical Review A [123].
Sondage micro-onde de l’énergie d’interaction van der Waals dans un nuage
froid de Rydberg. Étant excités par laser à partir d’un nuage dense, les atomes de
156 Résumé en français

Rydberg interagissent fortement entre eux. Le spectre micro-onde de la transition

vers un niveau de Rydberg voisin nous permet de sonder précisément la distribution
d’énergie d’interaction. Le changement de la fréquence micro-onde de la transition
de chaque atome est proportionnel à son énergie d’interaction initiale. Nous avons
utilisé cette technique ain d’observer l’expansion mécanique d’un ensemble d’atomes
de Rydberg sous l’efet de la forte interaction répulsive. Cette observation révèle la
limite de l’approximation du gaz gelé de Rydberg. Un modèle simple Monte Carlo a
été développé et nous a permis d’expliquer les spectres micro-ondes mesurés. Cette
méthode fait l’objet de la thèse de Raul Teixeira et est résumée dans un article paru
dans Physical Review Letters [124].
Développement d’une simulation plus rigoureuse du processus de l’excitation
des atomes de Rydberg. Bien que le modèle simple Monte Carlo réussisse à reconstru-
ire les spectres micro-ondes mesurés, il échoue lorsqu’il s’agit des transitions optiques.
Le problème est que cette simulation manque n’inclut aucun échelle de temps. Il nous
faut lui fournir le nombre inal d’atomes de Rydberg obtenu lors des expériences.
Cette approche phénoménologique nous empêche d’explorer de nouvelles techniques
d’excitation. En revisitant les équations de Bloch optiques, nous obtenons le taux
d’excitation (et de désexcitation) pour chaque atome. Ces taux ne dépendent que du
taux de déphasage, de la fréquence de Rabi d’excitation et de la présence éventuelle
d’autres atomes de Rydberg à proximité. Le problème de l’excitation atomique est
donc réduit à un processus stochastique classique, qui peut être calculé par une sim-
ulation Monte Carlo. Une échelle de temps apparaît naturellement dans la simula-
tion. Nous n’avons plus besoin de savoir le nombre d’atomes de Rydberg à l’avance.
Cette nouvelle méthode plus rigoureuse nous éclaire sur le processus d’excitation. Elle
prend en compte les paramètres de l’excitation, de la distribution spatiale du nuage,
du temps de vie ini et du mouvement des atomes de Rydberg pendant l’excitation
à cause de l’interaction forte entre eux. Elle nous permet de reconstruire les spectres
optiques ainsi que les spectres micro-ondes. Elle nous permet aussi de tester de nou-
velles techniques d’excitation, en particulier l’excitation de Rydberg dans un nuage
1D pour produire une chaîne d’atomes pour la simulation quantique.
Simulation quantique avec atomes de Rydberg circulaires piégés par laser. Les
atomes de Rydberg circulaires, malgré leurs propriétés uniques et leur exploitation
fructueuse dans les expériences d’électrodynamique quantique en cavité, n’ont pas en-
core été exploités pour réaliser des simulations quantiques. Nous étudions l’interaction
entre atomes circulaires et proposons une méthode pour réaliser des simulations quan-
tiques d’une chaîne 1D de spins avec ce type d’atomes.
Le schéma du simulateur quantique proposé est représenté sur la igure ci-dessous.
L’ingrédient principal de la proposition est un ensemble d’atomes de Rydberg circu-
laire piégés entre les deux plaques d’un condensateur. Quand la distance entre les
deux plaques est plus petite que la demi-longueur d’onde émise, l’émission spontanée
est inhibée. A une température au dessous de 1 K, les temps de vie des états circulaires
entrent dans la gamme des minutes. Le temps de piégeage d’une chaîne d’une dizaine
d’atomes est alors de quelques secondes.
Les atomes circulaires sont piégés par la force pondéromotrice agissant sur l’électron
de valence , presque libre. Pour les états de Rydberg de moment angulaire faible, l’efet
de photoionisation réduit fortement le temps de vie. Néanmoins, cet efet diminue
 Résumé en français 157

z
x
y

Schéma du simulateur quantique avec atomes de Rydberg circulaire. Les atomes circulairs
(rouge) sont piégés par laser entre deux plaques d’un condensateur inhibant l’émission spon-
tanée. Un réseau optique créé par un laser CO2 (vert) est supeposé avec un faiseau creux de
Laguerre-Gauss (bleu).

exponentiellement avec et devient négligeable pour les états circulaires. Un laser

Laguerre-Gauss de longueur d’onde 1 µm, en combinaison avec un onde stationnaire
créée par un laser CO2 de longueur d’onde 10 µm, forme un réseau 1D. Les atomes
sont piégés aux minima d’intensité et régulièrement espacés de 5 µm.
L’interaction de van der Waals entre les atomes permet de réaliser le Hamiltonien
de spin où les couplages spin-spin correspondent à l’interaction directe (termes
) et à l’interaction d’échange (termes ) entre les atomes. Nous encodons l’état du
spin sur deux niveaux circulaires 50 et 48 , pour lesquels l’interaction directe et
l’interaction d’échange sont du même ordre de grandeur. La transition entre ces deux
niveaux est habillée par un champ micro-onde qui joue le rôle d’un champ magnétique
dans ce modèle. Cet Hamiltonien est largement accordable via un champ élec-
trique appliqué et l’habillage micro-onde, d’un Hamiltonien anti-ferromagnétique à
un Hamiltonien ferromagnétique en passant par un Hamiltonien purement . C’est
une caractéristique unique de cette proposition.
De plus, nous proposons une méthode novatrice pour préparer de façon déter-
ministe une chaîne de plus de 40 atomes. C’est une variante du refroidissement par
évaporation. Nous enlevons les atomes un par un de la chaîne d’une manière bien
contrôlée. Cette technique est basée sur l’interaction répulsive entre les atomes, qui
appuie fortement sur les atomes aux bouts de la chaîne comprimée. Elle peut être
aussi appliquée aux géométries 2D ou même 3D. Des méthodes d’initialisation de la
chaîne et de la détection de l’état de chaque atome sont aussi développées.
La proposition vise à surmonter les diicultés des autres systèmes de simulation
quantique, incluant ceux avec des atomes de Rydberg de moment angulaire faible.
Elle combine la lexibilité des atomes neutres dans les réseaux optiques, l’interaction
forte typique des atomes de Rydberg et le contrôle exquis des ions piégés. La réalisa-
tion d’un tel système permettrait de simuler de façon quantique des systèmes hors de
portée des méthodes classiques de calcul.
158 Résumé en français
References

[1] J. J. N. J. J. Sakurai, Modern Quantum Mechanics (ADDISON WESLEY PUB CO

INC, 2010), 550 Seiten.
[2] C. Cohen-Tannoudji, Quantum Mechanics (John Wiley and Sons Ltd, 2006), 1524
Seiten.
[3] S. Haroche and J. Raimond, Exploring the Quantum: Atoms, Cavities, and Photons,
Oxford Graduate Texts (OUP Oxford, 2006).
[4] J. P. Dowling and G. J. Milburn, “Quantum technology: the second quantum
revolution”, Philosophical Transactions of the Royal Society A: Mathematical,
Physical and Engineering Sciences 361, 1655 (2003).
[5] J. Koomey, S. Berard, M. Sanchez, and H. Wong, “Implications of Historical
Trends in the Electrical Eiciency of Computing”, IEEE Annals of the History
of Computing 33, 46 (2011).
[6] Sadik C. Esener, Mark H. Kryder, William D. Doyle, Marvin Keshner, Ma-
sud Mansuripur, and David A. Thompson, WTEC Panel report on The Future
of Data Storage Technologies, tech. rep. (International Technology Research Insti-
tute, 1999).
[7] G.-R. HOFFMANN, “Weather forecasting and parallel processing: a view from
ECMWF”, International Journal of High Speed Computing 05, 63 (1993).
[8] R. R. Joshi, “A new heuristic algorithm for probabilistic optimization”, Com-
puters & Operations Research 24, 687 (1997).
[9] Y. Chen, T. Luo, S. Liu, S. Zhang, L. He, J. Wang, L. Li, T. Chen, Z. Xu, N. Sun,
and O. Temam, “DaDianNao: A Machine-Learning Supercomputer”, in 2014
47th Annual IEEE/ACM International Symposium on Microarchitecture (Dec.
2014).
[10] E. Sykes and A. Mirkovic, “A Fully Parallel and Scalable Implementation of a
Hopield Neural Network on the SHARC-NET Supercomputer”, in 19th Inter-
national Symposium on High Performance Computing Systems and Applica-
tions (HPCS’05) ().
[11] R. P. Feynman, “Simulating physics with computers”, International Journal of
Theoretical Physics 21, 467 (1982).
[12] M. Suzuki, Quantum Monte Carlo Methods in Condensed Matter Physics (WORLD
SCIENTIFIC PUB CO INC, 1993), 368 Seiten.
160 References

[13] M. Troyer and U.-J. Wiese, “Computational Complexity and Fundamental Lim-
itations to Fermionic Quantum Monte Carlo Simulations”, Phys. Rev. Lett. 94,
170201 (2005).
[14] J. I. Cirac and P. Zoller, “How to Manipulate Cold Atoms”, Science 301, 176
(2003).
[15] A. Friedenauer, H. Schmitz, J. T. Glueckert, D. Porras, and T. Schaetz, “Simu-
lating a quantum magnet with trapped ions”, Nat Phys 4, 757 (2008).
[16] Lloyd, “Universal Quantum Simulators”, eng, Science 273, 1073 (1996).
[17] K. D. Raedt, K. Michielsen, H. D. Raedt, B. Trieu, G. Arnold, M. Richter, T.
Lippert, H. Watanabe, and N. Ito, “Massively parallel quantum computer sim-
ulator”, Computer Physics Communications 176, 121 (2007).
[18] I. Buluta and F. Nori, “Quantum simulators.”, eng, Science 326, 108 (2009).
[19] I. M. Georgescu, S. Ashhab, and F. Nori, “Quantum Simulation”, Rev. Mod.
Phys. 86, 153 (2014).
[20] M. Nielsen and I. Chuang, Quantum Computation and Quantum Information, Cam-
bridge Series on Information and the Natural Sciences (Cambridge University
Press, 2000).
[21] S. J. Devitt, W. J. Munro, and K. Nemoto, “Quantum error correction for begin-
ners”, Reports on Progress in Physics 76, 076001 (2013).
[22] Nobelprize.org. Nobel Media AB 2014., The 2012 Nobel Prize in Physics - Press
Release, http://www.nobelprize.org/nobel_prizes/physics/laureates/
2012/press.htm, Web. 25 Sep 2016.
[23] D. Jaksch and P. Zoller, “The cold atom Hubbard toolbox”, Annals of Physics
315, 52 (2005).
[24] M. Lewenstein, A. Sanpera, V. Ahuinger, B. Damski, A. Sen(De), and U. Sen,
“Ultracold atomic gases in optical lattices: mimicking condensed matter physics
and beyond”, Advances in Physics 56, 243 (2007).
[25] I. Bloch, J. Dalibard, and S. Nascimbène, “Quantum simulations with ultracold
quantum gases”, Nature Physics 8, 267 (2012).
[26] I. Bloch, J. Dalibard, and W. Zwerger, “Many-body physics with ultracold gases”,
Reviews of Modern Physics 80, 885 (2008).
[27] R. Blatt and C. F. Roos, “Quantum simulations with trapped ions”, Nature
Physics 8, 277 (2012).
[28] C. Schneider, D. Porras, and T. Schaetz, “Experimental quantum simulations
of many-body physics with trapped ions”, Reports on Progress in Physics 75,
024401 (2012).
[29] A. A. Houck, H. E. Türeci, and J. Koch, “On-chip quantum simulation with
superconducting circuits”, Nature Physics 8, 292 (2012).
[30] D. Tanese, E. Gurevich, F. Baboux, T. Jacqmin, A. Lemaître, E. Galopin, I. Sagnes,
A. Amo, J. Bloch, and E. Akkermans, “Fractal energy spectrum of a polariton
gas in a Fibonacci quasiperiodic potential.”, eng, Phys Rev Lett 112, 146404
(2014).
 References 161

[31] A. Aspuru-Guzik and P. Walther, “Photonic quantum simulators”, Nature Physics

8, 285 (2012).
[32] I. Carusotto and C. Ciuti, “Quantum luids of light”, Rev. Mod. Phys. 85, 299
(2013).
[33] H.-P. Büchler, G. Pupillo, A. Micheli, and P. Zoller, “Condensed Matter Physics
with Cold Polar Molecules”, in Cold Molecules (Informa UK Limited, June 2009).
[34] J. Cai, A. Retzker, F. Jelezko, and M. B. Plenio, “A large-scale quantum simulator
on a diamond surface at room temperature”, Nature Physics 9, 168 (2013).
[35] E. Manousakis, “A Quantum-Dot Array as Model for Copper-Oxide Supercon-
ductors: A Dedicated Quantum Simulator for the Many-Fermion Problem”,
Journal of Low Temperature Physics 126, 1501 (2002).
[36] T. Byrnes, P. Recher, N. Y. Kim, S. Utsunomiya, and Y. Yamamoto, “Quantum
Simulator for the Hubbard Model with Long-Range Coulomb Interactions Us-
ing Surface Acoustic Waves”, Phys. Rev. Lett. 99, 016405 (2007).
[37] M. Greiner, O. Mandel, T. Esslinger, T. W. Hänsch, and I. Bloch, “Quantum
phase transition from a superluid to a Mott insulator in a gas of ultracold
atoms”, Nature 415, 39 (2002).
[38] D. Leibfried, B. DeMarco, V. Meyer, M. Rowe, A. Ben-Kish, J. Britton, W. M.
Itano, B. Jelenković, C. Langer, T. Rosenband, and D. J. Wineland, “Trapped-Ion
Quantum Simulator: Experimental Application to Nonlinear Interferometers”,
Phys. Rev. Lett. 89, 247901 (2002).
[39] R. Gerritsma, G. Kirchmair, F. Zähringer, E. Solano, R. Blatt, and C. F. Roos,
“Quantum simulation of the Dirac equation”, Nature 463, 68 (2010).
[40] K. Kim, M.-S. Chang, S. Korenblit, R. Islam, E. E. Edwards, J. K. Freericks,
G.-D. Lin, L.-M. Duan, and C. Monroe, “Quantum simulation of frustrated
Ising spins with trapped ions”, Nature 465, 590 (2010).
[41] J. Clarke and F. K. Wilhelm, “Superconducting quantum bits.”, eng, Nature 453,
1031 (2008).
[42] J. Q. You and F. Nori, “Superconducting circuits and quantum information”,
Phys. Today 58, 42 (2005).
[43] R. Barends, L. Lamata, J. Kelly, L. García-Álvarez, A. Fowler, A. Megrant, E.
Jefrey, T. White, D. Sank, J. Mutus, B. Campbell, Y. Chen, Z. Chen, B. Chiaro,
A. Dunsworth, I.-C. Hoi, C. Neill, P. J. J. O’Malley, C. Quintana, P. Roushan, A.
Vainsencher, J. Wenner, E. Solano, and J. M. Martinis, “Digital quantum simu-
lation of fermionic models with a superconducting circuit.”, eng, Nat Commun
6, 7654 (2015).
[44] R. Barends, A. Shabani, L. Lamata, J. Kelly, A. Mezzacapo, U. Las Heras, R.
Babbush, A. Fowler, B. Campbell, Y. Chen, Z. Chen, B. Chiaro, A. Dunsworth, E.
Jefrey, E. Lucero, A. Megrant, J. Mutus, M. Neeley, C. Neill, P. J. J. O’Malley, C.
Quintana, P. Roushan, D. Sank, A. Vainsencher, J. Wenner, T. White, E. Solano,
H. Neven, and J. M. Martinis, “Digitized adiabatic quantum computing with a
superconducting circuit.”, eng, Nature 534, 222 (2016).
162 References

[45] C. Eichler, J. Mlynek, J. Butscher, P. Kurpiers, K. Hammerer, T. J. Osborne, and

A. Wallraf, “Exploring Interacting Quantum Many-Body Systems by Experi-
mentally Creating Continuous Matrix Product States in Superconducting Cir-
cuits”, Phys. Rev. X 5, 041044 (2015).
[46] C. Neill, P. Roushan, M. Fang, Y. Chen, M. Kolodrubetz, Z. Chen, A. Megrant,
R. Barends, B. Campbell, B. Chiaro, A. Dunsworth, E. Jefrey, J. Kelly, J. Mutus,
P. J. J. O’Malley, C. Quintana, D. Sank, A. Vainsencher, J. Wenner, T. C. White, A.
Polkovnikov, and J. M. Martinis, “Ergodic dynamics and thermalization in an
isolated quantum system”, Nature Physics advance online publication (2016)
10.1038/nphys3830.
[47] S. Boixo, T. F. Rønnow, S. V. Isakov, Z. Wang, D. Wecker, D. A. Lidar, J. M.
Martinis, and M. Troyer, “Evidence for quantum annealing with more than one
hundred qubits”, Nat Phys 10, 218 (2014).
[48] B. Heim, T. F. Rønnow, S. V. Isakov, and M. Troyer, “Quantum versus classical
annealing of Ising spin glasses.”, eng, Science 348, 215 (2015).
[49] G. Ithier, E. Collin, P. Joyez, P. J. Meeson, D. Vion, D. Esteve, F. Chiarello, A.
Shnirman, Y. Makhlin, J. Schriel, and G. Schön, “Decoherence in a supercon-
ducting quantum bit circuit”, Phys. Rev. B 72, 134519 (2005).
[50] D. J. Wineland, “Superposition, entanglement, and raising Schrödinger’s cat
(Nobel Lecture).”, eng, Angew Chem Int Ed Engl 52, 10179 (2013).
[51] D. Wineland, C. Monroe, W. Itano, D. Leibfried, B. King, and D. Meekhof, “Ex-
perimental issues in coherent quantum-state manipulation of trapped atomic
ions”, J. Res. Natl. Inst. Stand. Technol. 103, 259 (1998).
[52] D. Leibfried, R. Blatt, C. Monroe, and D. Wineland, “Quantum dynamics of
single trapped ions”, Rev. Mod. Phys. 75, 281 (2003).
[53] C. Senko, P. Richerme, J. Smith, A. Lee, I. Cohen, A. Retzker, and C. Monroe,
“Realization of a Quantum Integer-Spin Chain with Controllable Interactions”,
Phys. Rev. X 5, 021026 (2015).
[54] P. Jurcevic, P. Hauke, C. Maier, C. Hempel, B. Lanyon, R. Blatt, and C. Roos,
“Spectroscopy of Interacting Quasiparticles in Trapped Ions.”, eng, Phys Rev
Lett 115, 100501 (2015).
[55] P. Schindler, M. Müller, D. Nigg, J. T. Barreiro, E. A. Martinez, M. Hennrich,
T. Monz, S. Diehl, P. Zoller, and R. Blatt, “Quantum simulation of dynamical
maps with trapped ions”, Nat Phys 9, 361 (2013).
[56] B. Lanyon, C. Hempel, D. Nigg, M. Müller, R. Gerritsma, F. Zähringer, P. Schindler,
J. Barreiro, M. Rambach, G. Kirchmair, M. Hennrich, P. Zoller, R. Blatt, and C.
Roos, “Universal digital quantum simulation with trapped ions.”, eng, Science
334, 57 (2011).
[57] E. A. Martinez, C. A. Muschik, P. Schindler, D. Nigg, A. Erhard, M. Heyl, P.
Hauke, M. Dalmonte, T. Monz, P. Zoller, and R. Blatt, “Real-time dynamics of
lattice gauge theories with a few-qubit quantum computer.”, eng, Nature 534,
516 (2016).
 References 163

[58] P. Jurcevic, B. P. Lanyon, P. Hauke, C. Hempel, P. Zoller, R. Blatt, and C. F.

Roos, “Quasiparticle engineering and entanglement propagation in a quantum
many-body system”, Nature 511, 202 (2014).
[59] P. Richerme, Z.-X. Gong, A. Lee, C. Senko, J. Smith, M. Foss-Feig, S. Micha-
lakis, A. V. Gorshkov, and C. Monroe, “Non-local propagation of correlations
in quantum systems with long-range interactions.”, eng, Nature 511, 198 (2014).
[60] J. Smith, A. Lee, P. Richerme, B. Neyenhuis, P. W. Hess, P. Hauke, M. Heyl, D. A.
Huse, and C. Monroe, “Many-body localization in a quantum simulator with
programmable random disorder”, Nature Physics advance online publication
(2016) 10.1038/nphys3783.
[61] K. Kim, M.-S. Chang, R. Islam, S. Korenblit, L.-M. Duan, and C. Monroe, “En-
tanglement and tunable spin-spin couplings between trapped ions using mul-
tiple transverse modes.”, eng, Phys Rev Lett 103, 120502 (2009).
[62] R. Islam, C. Senko, W. Campbell, S. Korenblit, J. Smith, A. Lee, E. Edwards,
C.-C. J. Wang, J. Freericks, and C. Monroe, “Emergence and frustration of mag-
netism with variable-range interactions in a quantum simulator.”, eng, Science
340, 583 (2013).
[63] J. W. Britton, B. C. Sawyer, A. C. Keith, C.-C. J. Wang, J. K. Freericks, H. Uys, M. J.
Biercuk, and J. J. Bollinger, “Engineered two-dimensional Ising interactions in
a trapped-ion quantum simulator with hundreds of spins.”, eng, Nature 484,
489 (2012).
[64] J. G. Bohnet, B. C. Sawyer, J. W. Britton, M. L. Wall, A. M. Rey, M. Foss-Feig,
and J. J. Bollinger, “Quantum spin dynamics and entanglement generation with
hundreds of trapped ions.”, eng, Science 352, 1297 (2016).
[65] P. Richerme, “2D ion crystals in radiofrequency traps for quantum simulation”,
ArXiv e-prints (2016).
[66] M. Kumph, M. Brownnutt, and R. Blatt, “Two-dimensional arrays of radio-
frequency ion traps with addressable interactions”, New Journal of Physics 13,
073043 (2011).
[67] R. Schmied, J. H. Wesenberg, and D. Leibfried, “Optimal surface-electrode trap
lattices for quantum simulation with trapped ions.”, eng, Phys Rev Lett 102,
233002 (2009).
[68] M. Mielenz, H. Kalis, M. Wittemer, F. Hakelberg, U. Warring, R. Schmied, M.
Blain, P. Maunz, D. L. Moehring, D. Leibfried, and T. Schaetz, “Arrays of indi-
vidually controlled ions suitable for two-dimensional quantum simulations.”,
eng, Nat Commun 7, ncomms11839 (2016).
[69] J. F. Sherson, C. Weitenberg, M. Endres, M. Cheneau, I. Bloch, and S. Kuhr,
“Single-atom-resolved luorescence imaging of an atomic Mott insulator.”, eng,
Nature 467, 68 (2010).
164 References

[70] S. Kuhr, W. Alt, D. Schrader, I. Dotsenko, Y. Miroshnychenko, W. Rosenfeld,

M. Khudaverdyan, V. Gomer, A. Rauschenbeutel, and D. Meschede, “Erratum:
Coherence Properties and Quantum State Transportation in an Optical Con-
veyor Belt [Phys. Rev. Lett. 91 , 213002 (2003)]”, Phys. Rev. Lett. 95, 099901
(2005).
[71] M. Friesdorf, A. Werner, W. Brown, V. Scholz, and J. Eisert, “Many-Body Lo-
calization Implies that Eigenvectors are Matrix-Product States.”, eng, Phys Rev
Lett 114, 170505 (2015).
[72] D. J. Luitz, N. Lalorencie, and F. Alet, “Many-body localization edge in the
random-ield Heisenberg chain”, Phys. Rev. B 91, 081103 (2015).
[73] J. Dalibard, F. Gerbier, G. Juzeli ūnas, and P. Öhberg, “Colloquium : Artiicial
gauge potentials for neutral atoms”, Rev. Mod. Phys. 83, 1523 (2011).
[74] Y.-J. Lin, K. Jiménez-GarcÃa, and I. Spielman, “Spin-orbit-coupled Bose-Einstein
condensates.”, eng, Nature 471, 83 (2011).
[75] B. Stuhl, H.-I. Lu, L. Aycock, D. Genkina, and I. Spielman, “Visualizing edge
states with an atomic Bose gas in the quantum Hall regime.”, eng, Science 349,
1514 (2015).
[76] M. Mancini, G. Pagano, G. Cappellini, L. Livi, M. Rider, J. Catani, C. Sias, P.
Zoller, M. Inguscio, M. Dalmonte, and L. Fallani, “Observation of chiral edge
states with neutral fermions in synthetic Hall ribbons.”, eng, Science 349, 1510
(2015).
[77] O. Romero-Isart, C. Navau, A. Sanchez, P. Zoller, and J. Cirac, “Superconduct-
ing vortex lattices for ultracold atoms.”, eng, Phys Rev Lett 111, 145304 (2013).
[78] A. González-Tudela, C.-L. Hung, D. E. Chang, J. I. Cirac, and H. J. Kimble, “Sub-
wavelength vacuum lattices and atom–atom interactions in two-dimensional
photonic crystals”, Nature Photonics 9, 320 (2015).
[79] B. Yan, S. A. Moses, B. Gadway, J. P. Covey, K. R. A. Hazzard, A. M. Rey, D. S.
Jin, and J. Ye, “Observation of dipolar spin-exchange interactions with lattice-
conined polar molecules.”, eng, Nature 501, 521 (2013).
[80] D. DeMille, “Quantum computation with trapped polar molecules.”, eng, Phys
Rev Lett 88, 067901 (2002).
[81] S. Baier, M. Mark, D. Petter, K. Aikawa, L. Chomaz, Z. Cai, M. Baranov, P.
Zoller, and F. Ferlaino, “Extended Bose-Hubbard models with ultracold mag-
netic atoms.”, eng, Science 352, 201 (2016).
[82] T. Lahaye, C. Menotti, L. Santos, M. Lewenstein, and T. Pfau, “The physics of
dipolar bosonic quantum gases”, Reports on Progress in Physics 72, 126401
(2009).
[83] T. F. Gallagher, Rydberg atoms, Vol. 3 (Cambridge University Press, 2005).
[84] M. Brune, F. Schmidt-Kaler, A. Maali, J. Dreyer, E. Hagley, J. M. Raimond, and
S. Haroche, “Quantum Rabi Oscillation: A Direct Test of Field Quantization in
a Cavity”, Phys. Rev. Lett. 76, 1800 (1996).
 References 165

[85] M. Brune, E. Hagley, J. Dreyer, X. Maître, A. Maali, C. Wunderlich, J. M. Rai-

mond, and S. Haroche, “Observing the Progressive Decoherence of the “Meter”
in a Quantum Measurement”, Phys. Rev. Lett. 77, 4887 (1996).
[86] J. M. Raimond, M. Brune, and S. Haroche, “Manipulating quantum entangle-
ment with atoms and photons in a cavity”, Reviews of Modern Physics 73, 565
(2001).
[87] S. Haroche, G. Nogues, A. Rauschenbeutel, S. Osnaghi, M. Brune, and J. M. Rai-
mond, “Seeing a single photon without destroying it”, Nature 400, 239 (1999).
[88] P. Maioli, T. Meunier, S. Gleyzes, A. Aufeves, G. Nogues, M. Brune, J. M. Rai-
mond, and S. Haroche, “Nondestructive Rydberg Atom Counting with Meso-
scopic Fields in a Cavity”, Phys. Rev. Lett. 94, 113601 (2005).
[89] S. Gleyzes, S. Kuhr, C. Guerlin, J. Bernu, S. Deléglise, U. B. Hof, M. Brune,
J.-M. Raimond, and S. Haroche, “Quantum jumps of light recording the birth
and death of a photon in a cavity”, Nature 446, 297 (2007).
[90] C. Guerlin, J. Bernu, S. Deléglise, C. Sayrin, S. Gleyzes, S. Kuhr, M. Brune, J.-M.
Raimond, and S. Haroche, “Progressive ield-state collapse and quantum non-
demolition photon counting”, Nature 448, 889 (2007).
[91] S. Deléglise, I. Dotsenko, C. Sayrin, J. Bernu, M. Brune, J.-M. Raimond, and S.
Haroche, “Reconstruction of non-classical cavity ield states with snapshots of
their decoherence.”, eng, Nature 455, 510 (2008).
[92] C. Sayrin, I. Dotsenko, X. Zhou, B. Peaudecerf, T. Rybarczyk, S. Gleyzes, P.
Rouchon, M. Mirrahimi, H. Amini, M. Brune, J.-M. Raimond, and S. Haroche,
“Real-time quantum feedback prepares and stabilizes photon number states.”,
eng, Nature 477, 73 (2011).
[93] X. Zhou, I. Dotsenko, B. Peaudecerf, T. Rybarczyk, C. Sayrin, S. Gleyzes, J. M.
Raimond, M. Brune, and S. Haroche, “Field Locked to a Fock State by Quantum
Feedback with Single Photon Corrections”, Phys. Rev. Lett. 108, 243602 (2012).
[94] A. Signoles, A. Facon, D. Grosso, I. Dotsenko, S. Haroche, J.-M. Raimond, M.
Brune, and S. Gleyzes, “Conined quantum Zeno dynamics of a watched atomic
arrow”, Nature Physics, 715 (2014).
[95] A. Facon, E.-K. Dietsche, D. Grosso, S. Haroche, J.-M. Raimond, M. Brune, and
S. Gleyzes, “A sensitive electrometer based on a Rydberg atom in a Schrödinger-
cat state”, Nature 535, 262 (2016).
[96] M. Safman, T. G. Walker, and K. Mølmer, “Quantum information with Rydberg
atoms”, Rev. Mod. Phys. 82, 2313 (2010).
[97] H. Weimer, M. Müller, H. P. Büchler, and I. Lesanovsky, “Digital quantum sim-
ulation with Rydberg atoms”, Quantum Inf Process 10, 885 (2011).
[98] H. Weimer, M. Müller, I. Lesanovsky, P. Zoller, and H. P. Büchler, “A Rydberg
quantum simulator”, Nat Phys 6, 382 (2010).
[99] M. D. Lukin, M. Fleischhauer, R. Cote, L. M. Duan, D. Jaksch, J. I. Cirac, and P.
Zoller, “Dipole Blockade and Quantum Information Processing in Mesoscopic
Atomic Ensembles”, Phys. Rev. Lett. 87, 037901 (2001).
166 References

[100] Y. Dudin and A. Kuzmich, “Strongly interacting Rydberg excitations of a cold

atomic gas.”, eng, Science 336, 887 (2012).
[101] D. Barredo, S. Ravets, H. Labuhn, L. Béguin, A. Vernier, F. Nogrette, T. La-
haye, and A. Browaeys, “Demonstration of a strong Rydberg blockade in three-
atom systems with anisotropic interactions.”, eng, Phys Rev Lett 112, 183002
(2014).
[102] T. Wilk, A. Gaëtan, C. Evellin, J. Wolters, Y. Miroshnychenko, P. Grangier, and
A. Browaeys, “Entanglement of two individual neutral atoms using Rydberg
blockade.”, eng, Phys Rev Lett 104, 010502 (2010).
[103] L. Isenhower, E. Urban, X. Zhang, A. Gill, T. Henage, T. Johnson, T. Walker,
and M. Safman, “Demonstration of a neutral atom controlled-NOT quantum
gate.”, eng, Phys Rev Lett 104, 010503 (2010).
[104] S. Ravets, H. Labuhn, D. Barredo, L. Béguin, T. Lahaye, and A. Browaeys, “Co-
herent dipole–dipole coupling between two single Rydberg atoms at an electrically-
tuned Förster resonance”, Nat Phys 10, 914 (2014).
[105] D. W. Schönleber, A. Eisfeld, M. Genkin, S. Whitlock, and S. Wüster, “Quan-
tum simulation of energy transport with embedded Rydberg aggregates.”, eng,
Phys Rev Lett 114, 123005 (2015).
[106] I. Lesanovsky, “Liquid ground state, gap, and excited states of a strongly cor-
related spin chain.”, eng, Phys Rev Lett 108, 105301 (2012).
[107] A. Dauphin, M. Müller, and M. A. Martin-Delgado, “Rydberg-atom quantum
simulation and Chern-number characterization of a topological Mott insula-
tor”, Phys. Rev. A 86, 053618 (2012).
[108] J. P. Hague, S. Downes, C. MacCormick, and P. E. Kornilovitch, “Cold Rydberg
Atoms for Quantum Simulation of Exotic Condensed Matter Interactions”, J
Supercond Nov Magn 27, 937 (2013).
[109] D. Barredo, H. Labuhn, S. Ravets, T. Lahaye, A. Browaeys, and C. S. Adams,
“Coherent Excitation Transfer in a Spin Chain of Three Rydberg Atoms”, Phys.
Rev. Lett. 114, 113002 (2015).
[110] H. Labuhn, D. Barredo, S. Ravets, S. de Léséleuc, T. Macrì, T. Lahaye, and A.
Browaeys, “Tunable two-dimensional arrays of single Rydberg atoms for real-
izing quantum Ising models”, Nature 534, 667 (2016).
[111] J. E. Johnson and S. L. Rolston, “Interactions between Rydberg-dressed atoms”,
Phys. Rev. A 82, 033412 (2010).
[112] J. Zeiher, R. van Bijnen, P. Schauß, S. Hild, J.-y. Choi, T. Pohl, I. Bloch, and C.
Gross, “Many-body interferometry of a Rydberg-dressed spin lattice”, Nature
Physics advance online publication (2016) 10.1038/nphys3835.
[113] R. M. W. van Bijnen and T. Pohl, “Quantum Magnetism and Topological Or-
dering via Rydberg Dressing near Förster Resonances”, Phys. Rev. Lett. 114,
243002 (2015).
[114] A. W. Glaetzle, M. Dalmonte, R. Nath, C. Gross, I. Bloch, and P. Zoller, “Design-
ing Frustrated Quantum Magnets with Laser-Dressed Rydberg Atoms”, Phys.
Rev. Lett. 114, 173002 (2015).
 References 167

[115] T. Macrì and T. Pohl, “Rydberg dressing of atoms in optical lattices”, Phys. Rev.
A 89, 011402 (2014).
[116] E. A. Goldschmidt, T. Boulier, R. C. Brown, S. B. Koller, J. T. Young, A. V. Gor-
shkov, S. L. Rolston, and J. V. Porto, “Anomalous Broadening in Driven Dissi-
pative Rydberg Systems”, Phys. Rev. Lett. 116, 113001 (2016).
[117] J. Zeiher, P. Schauß, S. Hild, T. Macrì, I. Bloch, and C. Gross, “Microscopic Char-
acterization of Scalable Coherent Rydberg Superatoms”, Phys. Rev. X 5, 031015
(2015).
[118] J. A. Aman, B. J. DeSalvo, F. B. Dunning, T. C. Killian, S. Yoshida, and J. Burgdör-
fer, “Trap losses induced by near-resonant Rydberg dressing of cold atomic
gases”, Phys. Rev. A 93, 043425 (2016).
[119] S. E. Anderson, K. C. Younge, and G. Raithel, “Trapping Rydberg Atoms in an
Optical Lattice”, Phys. Rev. Lett. 107, 263001 (2011).
[120] S. E. Anderson and G. Raithel, “Ionization of Rydberg atoms by standing-wave
light ields”, Nature Communications 4, 2967 (2013).
[121] J. M. McGuirk, D. M. Harber, J. M. Obrecht, and E. A. Cornell, “Alkali-metal ad-
sorbate polarization on conducting and insulating surfaces probed with Bose-
Einstein condensates”, Phys. Rev. A 69, 062905 (2004).
[122] K. S. Chan, M. Siercke, C. Hufnagel, and R. Dumke, “Adsorbate Electric Fields
on a Cryogenic Atom Chip”, Phys. Rev. Lett. 112, 026101 (2014).
[123] C. Hermann-Avigliano, R. C. Teixeira, T. L. Nguyen, T. Cantat-Moltrecht, G.
Nogues, I. Dotsenko, S. Gleyzes, J. M. Raimond, S. Haroche, and M. Brune,
“Long coherence times for Rydberg qubits on a superconducting atom chip”,
Phys. Rev. A 90, 040502 (2014).
[124] R. C. Teixeira, C. Hermann-Avigliano, T. L. Nguyen, T. Cantat-Moltrecht, J. M.
Raimond, S. Haroche, S. Gleyzes, and M. Brune, “Microwaves Probe Dipole
Blockade and van der Waals Forces in a Cold Rydberg Gas”, Phys. Rev. Lett.
115, 013001 (2015).
[125] H. A. Bethe and E. E. Salpeter, Quantum mechanics of one-and two-electron atoms
(Springer Science & Business Media, 2012).
[126] P. J. Mohr, D. B. Newell, and B. N. Taylor, “CODATA Recommended Values of
the Fundamental Physical Constants: 2014”, ArXiv e-prints (2015).
[127] L. P. P. V. B. Berestetskii, E. M. Lifshitz, Quantum Electrodynamics (Elsevier Sci-
ence & Technology, 1996), 670 Seiten.
[128] A. Tauschinsky, R. Newell, H. B. van Linden van den Heuvell, and R. J. C.
Spreeuw, “Measurement of 87 Rb Rydberg-state hyperine splitting in a room-
temperature vapor cell”, Phys. Rev. A 87, 042522 (2013).
[129] W. Li, I. Mourachko, M. W. Noel, and T. F. Gallagher, “Millimeter-wave spec-
troscopy of cold Rb Rydberg atoms in a magneto-optical trap: Quantum defects
of the ns , np , and nd series”, Phys. Rev. A 67, 052502 (2003).
[130] J. Han, Y. Jamil, D. V. L. Norum, P. J. Tanner, and T. F. Gallagher, “Rb quantum
defects from millimeter-wave spectroscopy of cold 85 Rb Rydberg atoms”, Phys.
Rev. A 74, 054502 (2006).
168 References

[131] M. Mack, F. Karlewski, H. Hattermann, S. Höckh, F. Jessen, D. Cano, and J.

Fortágh, “Measurement of absolute transition frequencies of 87 Rb to and
Rydberg states by means of electromagnetically induced transparency”, Phys.
Rev. A 83, 052515 (2011).
[132] J. D. Jackson, Classical Electrodynamics (JOHN WILEY & SONS INC, 1998), 832
pages.
[133] S. Svanberg, Atomic and molecular spectroscopy: basic aspects and practical applica-
tions, Vol. 6 (Springer Science & Business Media, 2012).
[134] C. Cohen-Tannoudji, J. Dupont-Roc, and G. Grynberg, Atom-Photon Interactions:
Basic Processes and Applications, A Wiley-Interscience publication (Wiley, 1998).
[135] R. Celistrino Teixeira, “Efets mécaniques de l’interaction dipolaire des atomes
de Rydberg sondés par spectroscopie microonde ”, PhD thesis (Université Pierre
et Marie Curie - Paris VI, Sept. 2014).
[136] C. H. Hermann Avigliano, “Towards deterministic preparation of single Ryd-
berg atoms and applications to quantum information processing”, PhD thesis
(Université Pierre et Marie Curie - Paris VI, Nov. 2014).
[137] M. Diederich, H. Häfner, N. Hermanspahn, M. Immel, H. Kluge, R. Ley, R.
Mann, W. Quint, S. Stahl, and G. Werth, “Observing a single hydrogen-like ion
in a Penning trap at T = 4 K”, Hyperine Interactions 115, 185 (1998).
[138] Nogues, G., Roux, C., Nirrengarten, T., Lupaşcu, A., Emmert, A., Brune, M.,
Raimond, J.-M., Haroche, S., Plaçais, B., and Grefet, J.-J., “Efect of vortices on
the spin-lip lifetime of atoms in superconducting atom-chips”, EPL 87, 13002
(2009).
[139] A. Emmert, A. Lupaşcu, G. Nogues, M. Brune, J.-M. Raimond, and S. Haroche,
“Measurement of the trapping lifetime close to a cold metallic surface on a cryo-
genic atom-chip”, The European Physical Journal D 51, 173 (2009).
[140] G. Reinaudi, T. Lahaye, Z. Wang, and D. Guéry-Odelin, “Strong saturation ab-
sorption imaging of dense clouds of ultracold atoms”, Opt. Lett. 32, 3143 (2007).
[141] M. Anwar, D. V. Magalhães, S. T. Müller, M. Faisal, M. Nawaz, and M. Ahmed,
“Revisiting the capture velocity of a cesium magneto-optical trap: model, sim-
ulation and experiment”, Laser Physics 24, 125502 (2014).
[142] E. Majorana, “Atomi orientati in campo magnetico variabile”, Il Nuovo Ci-
mento (1924-1942) 9, 43 (1932).
[143] D. M. Brink and C. V. Sukumar, “Majorana spin-lip transitions in a magnetic
trap”, Phys. Rev. A 74, 035401 (2006).
[144] T. Nirrengarten, “Piégeage magnétiques d’atomes de Rudibium au voisinage
d’une surface supraconductrice ”, PhD thesis (Université Pierre et Marie Curie
- Paris VI, 2007).
[145] W. Ketterle and N. V. Druten, “Evaporative Cooling of Trapped Atoms”, in
Advances In Atomic, Molecular, and Optical Physics, Vol. 37, edited by B. Beder-
son and H. Walther, Advances In Atomic, Molecular, and Optical Physics (Aca-
demic Press, 1996), pp. 181–236.
 References 169

[146] F. Dalfovo, S. Giorgini, L. P. Pitaevskii, and S. Stringari, “Theory of Bose-Einstein

condensation in trapped gases”, Rev. Mod. Phys. 71, 463 (1999).
[147] C. Roux, A. Emmert, A. Lupascu, T. Nirrengarten, G. Nogues, M. Brune, J.-M.
Raimond, and S. Haroche, “Bose-Einstein condensation on a superconducting
atom chip”, EPL (Europhysics Letters) 81, 56004 (2008).
[148] R. F. Gutterres, C. Amiot, A. Fioretti, C. Gabbanini, M. Mazzoni, and O. Dulieu,
“Determination of the 87 Rb 5 state dipole matrix element and radiative life-
time from the photoassociation spectroscopy of the Rb2 0− � ( 3/2 ) long-range
state”, Phys. Rev. A 66, 024502 (2002).
[149] T. W. Hall and C. H. B. Mee, “The work function of rubidium”, physica status
solidi (a) 21, 109 (1974).
[150] W. Sachtler, G. Dorgelo, and A. Holscher, “The work function of gold”, Surface
Science 5, 221 (1966).
[151] A. Gaetan, Y. Miroshnychenko, T. Wilk, A. Chotia, M. Viteau, D. Comparat, P.
Pillet, A. Browaeys, and P. Grangier, “Observation of collective excitation of two
individual atoms in the Rydberg blockade regime”, Nat Phys 5, 115 (2009).
[152] T. M. Weber, M. Höning, T. Niederprüm, T. Manthey, O. Thomas, V. Guarrera,
M. Fleischhauer, G. Barontini, and H. Ott, “Mesoscopic Rydberg-blockaded en-
sembles in the superatom regime and beyond”, Nat Phys 11, 157 (2015).
[153] B. M. Axilrod and E. Teller, “Interaction of the van der Waals Type Between
Three Atoms”, The Journal of Chemical Physics 11, 299 (1943).
[154] J. B. Balewski, A. T. Krupp, A. Gaj, D. Peter, H. P. Büchler, R. Löw, S. Hoferberth,
and T. Pfau, “Coupling a single electron to a Bose-Einstein condensate”, Nature
502, 664 (2013).
[155] E. D. Black, “An introduction to Pound–Drever–Hall laser frequency stabiliza-
tion”, American Journal of Physics 69, 79 (2001).
[156] A. Urvoy, F. Ripka, I. Lesanovsky, D. Booth, J. P. Shafer, T. Pfau, and R. Löw,
“Strongly Correlated Growth of Rydberg Aggregates in a Vapor Cell”, Phys.
Rev. Lett. 114, 203002 (2015).
[157] M. Viteau, P. Huillery, M. G. Bason, N. Malossi, D. Ciampini, O. Morsch, E.
Arimondo, D. Comparat, and P. Pillet, “Cooperative Excitation and Many-Body
Interactions in a Cold Rydberg Gas”, Phys. Rev. Lett. 109, 053002 (2012).
[158] H. Schempp, G. Günter, M. Robert-de-Saint-Vincent, C. S. Hofmann, D. Breyel,
A. Komnik, D. W. Schönleber, M. Gärttner, J. Evers, S. Whitlock, and M. Wei-
demüller, “Full Counting Statistics of Laser Excited Rydberg Aggregates in a
One-Dimensional Geometry”, Phys. Rev. Lett. 112, 013002 (2014).
[159] F. M. Leupold, “Bang-bang Control of a Trapped-Ion Oscillator”, PhD thesis (.
ETH-Zürich, 2015).
[160] D. V. Dmitriev, V. Y. Krivnov, A. A. Ovchinnikov, and A. Langari, “One-dimensional
anisotropic Heisenberg model in the transverse magnetic ield”, Journal of Ex-
perimental and Theoretical Physics 95, 538 (2002).
170 References

[161] C. N. Yang and C. P. Yang, “One-Dimensional Chain of Anisotropic Spin-Spin

Interactions. I. Proof of Bethe’s Hypothesis for Ground State in a Finite System”,
Phys. Rev. 150, 321 (1966).
[162] D. V. Dmitriev and V. Y. Krivnov, “Anisotropic Heisenberg chain in coexisting
transverse and longitudinal magnetic ields”, Physical Review B 70 (2004) 10.
1103/physrevb.70.144414.
[163] F. Capraro and C. Gros, “The spin-1/2 anisotropic Heisenberg-chain in longitu-
dinal and transversal magnetic ields: a DMRG study”, The European Physical
Journal B 29, 35 (2002).
[164] A. A. Ovchinnikov, D. V. Dmitriev, V. Y. Krivnov, and V. O. Cheranovskii, “An-
tiferromagnetic Ising chain in a mixed transverse and longitudinal magnetic
ield”, Physical Review B 68 (2003) 10.1103/physrevb.68.214406.
[165] Y. Hieida, K. Okunishi, and Y. Akutsu, “Anisotropic antiferromagnetic spin
chains in a transverse ield: Reentrant behavior of the staggered magnetiza-
tion”, Physical Review B 64 (2001) 10.1103/physrevb.64.224422.
[166] D. V. Dmitriev, V. Y. Krivnov, and A. A. Ovchinnikov, “Gap generation in the
XXZ model in a transverse magnetic ield”, Physical Review B 65 (2002) 10 .
1103/physrevb.65.172409.
[167] J.-S. Caux, F. H. L. Essler, and U. Löw, “Dynamical structure factor of the anisotropic
Heisenberg chain in a transverse ield”, Physical Review B 68 (2003) 10.1103/
physrevb.68.134431.
[168] M. Gross and J. Liang, “Is a circular Rydberg atom stable in a vanishing electric
ield?”, Phys. Rev. Lett. 57, 3160 (1986).
[169] J. Mozley, P. Hyail, G. Nogues, M. Brune, J.-M. Raimond, and S. Haroche,
“Trapping and coherent manipulation of a Rydberg atom on amicrofabricated
device: a proposal”, The European Physical Journal D - Atomic, Molecular, Op-
tical and Plasma Physics 35, 43 (2005).
[170] R. G. Hulet, E. S. Hilfer, and D. Kleppner, “Inhibited Spontaneous Emission by
a Rydberg Atom”, Phys. Rev. Lett. 55, 2137 (1985).
[171] E. de Prunelé, “Orientation efects in thermal collisions between “circular” –
Rydberg-state atoms and ground-state helium”, Phys. Rev. A 31, 3593 (1985).
[172] G. Gabrielse, X. Fei, L. A. Orozco, R. L. Tjoelker, J. Haas, H. Kalinowsky, T. A.
Trainor, and W. Kells, “Thousandfold improvement in the measured antiproton
mass”, Phys. Rev. Lett. 65, 1317 (1990).
[173] P. Hyail, J. Mozley, A. Perrin, J. Tailleur, G. Nogues, M. Brune, J. M. Raimond,
and S. Haroche, “Coherence-Preserving Trap Architecture for Long-Term Con-
trol of Giant Ryberg Atoms”, Physical Review Letters 93 (2004) 10.1103/physrevlett.
93.103001.
[174] M. Safman and T. G. Walker, “Analysis of a quantum logic device based on
dipole-dipole interactions of optically trapped Rydberg atoms”, Phys. Rev. A
72, 022347 (2005).
 References 171

[175] I. I. Beterov, D. B. Tretyakov, I. I. Ryabtsev, A. Ekers, and N. N. Bezuglov, “Ion-

ization of sodium and rubidium , , and Rydberg atoms by blackbody
radiation”, Phys. Rev. A 75, 052720 (2007).
[176] S. Blanes and P. Moan, “Practical symplectic partitioned Runge–Kutta and Runge–
Kutta–Nyström methods”, Journal of Computational and Applied Mathemat-
ics 142, 313 (2002).
[177] P. Soltan-Panahi, J. Struck, P. Hauke, A. Bick, W. Plenkers, G. Meineke, C. Becker,
P. Windpassinger, M. Lewenstein, and K. Sengstock, “Multi-component quan-
tum gases in spin-dependent hexagonal lattices”, Nat Phys 7, 434 (2011).
[178] B. Tamir and E. Cohen, “Notes on Adiabatic Quantum Computers”, ArXiv e-
prints (2015).
[179] P. Calabrese and J. Cardy, “Time Dependence of Correlation Functions Follow-
ing a Quantum Quench”, Phys. Rev. Lett. 96, 136801 (2006).
[180] S. Möbius, S. Wüster, C. Ates, A. Eisfeld, and J. M. Rost, “Adiabatic entangle-
ment transport in Rydberg aggregates”, Journal of Physics B: Atomic, Molecu-
lar and Optical Physics 44, 184011 (2011).
[181] S. Wüster, C. Ates, A. Eisfeld, and J. M. Rost, “Newton’s Cradle and Entan-
glement Transport in a Flexible Rydberg Chain”, Phys. Rev. Lett. 105, 053004
(2010).
[182] S. R. White, “Equivalence of the antiferromagnetic Heisenberg ladder to a sin-
gle S =1 chain”, Phys. Rev. B 53, 52 (1996).
[183] M. L. Zimmerman, M. G. Littman, M. M. Kash, and D. Kleppner, “Stark struc-
ture of the Rydberg states of alkali-metal atoms”, Phys. Rev. A 20, 2251 (1979).
[184] S. Hovanessian, Computational mathematics in engineering, Practice Counts (Lex-
ington Books, 1976).
[185] D. Brink and G. Satchler, Angular Momentum, Oxford science publications (Claren-
don Press, 1993).
[186] G. Racah, “Theory of Complex Spectra. II”, Phys. Rev. 62, 438 (1942).
[187] U. Albert Messiah, A. Messiah, Quantum Mechanics (ELSEVIER SCIENCE &
TECHNOLOGY, Jan. 11, 1981), 1152 Seiten.
Study of dipole-dipole interaction between Rydberg atoms
Toward quantum simulation with Rydberg atoms

Abstract: Quantum simulation ofers a highly promising way to understand large corre-
lated quantum systems, and many experimental platforms are now being developed. Ry-
dberg atoms are especially appealing thanks to their strong and short-range dipole-dipole
interaction.
In our setup, we prepare and manipulate ensembles of Rydberg atoms excited from an
ultracold atomic cloud magnetically trapped above a superconducting chip. The dynamics
of the Rydberg excitation can be controlled through the laser excitation process. The many-
body atomic interaction energy spectrum is then directly measured through microwave
spectroscopy. This thesis develops a rigorous Monte Carlo model that provides an insight
into the excitation process. Using this model, we discuss a possibility to explore quantum
simulations of energy transport in a 1D chain of low angular momentum Rydberg atoms.
Furthermore, we propose an innovative platform for quantum simulations. It relies on
a groundbreaking approach, based on laser-trapped ensemble of extremely long-lived,
strongly interacting circular Rydberg atoms. We present intensive numerical results as
well as discuss a wide range of problems that can be addressed with the proposed model.

Keywords: quantum simulation, Rydberg atoms, circular atoms, dipole-dipole interaction,

atom chip, microwave spectroscopy

Étude de l’interaction dipolaire entre atomes de Rydberg

Vers la simulation quantique de chaînes de spin

Résumé : La simulation quantique ofre un moyen très prometteur pour comprendre les
systèmes quantiques corrélés macroscopiques. De nombreuses plateformes expérimen-
tales sont en cours d’élaboration. Les atomes de Rydberg sont particulièrement intéres-
sants grâce à leur forte interaction dipolaire de cours portée.
Dans notre manip, nous préparons et manipulons des ensembles d’atomes de Rydberg ex-
cités à partir d’un nuage atomique ultra-froid piégé magnétiquement sur une puce à atome
supraconductrice. La dynamique de l’excitation est contrôlée par le processus d’excitation
du laser. Le spectre d’énergie d’interaction atomique des corps est mesuré directment
par spectroscopie micro-onde. Dans cette thèse, nous développons un modèle Monte
Carlo rigoureux qui nous éclaire sur le processus d’excitation. En utilisant ce modèle, nous
discutons de la possibilité de réaliser des simulations quantiques du transport d’énergie
sur une chaîne 1D d’atomes de Rydberg de faible moment angulaire.
De plus, nous proposons une plateforme innovante pour la réalisation de simulations
quantiques. Elle repose sur une approche révolutionnaire basée sur un ensemble d’atomes
de Rydberg dont le temps de vie est extrêmement long, qui interagissent fortement et qui
sont piégés par laser. Nous présentons les résultats de simulations numériques et nous
discutons du large éventail de problèmes qui peuvent être traités avec le modèle proposé.

Mots-clés : simulation quantique, atomes de Rydberg, atomes circulaires, interaction

dipolaire, puce à atome, spectroscopie micro-ondre