AlGaAs Photonic Devices - From Quantum State Generation To Quantum Communication

AlGaAs photonic devices: from quantum state
generation to quantum communications

Claire Autebert
To cite this version:

Claire Autebert. AlGaAs photonic devices: from quantum state generation to quantum communi-
cations. Quantum Physics [quant-ph]. Université Paris 7 - Denis Diderot, 2016. English. �NNT : �.
�tel-01676987�
HAL Id: tel-01676987

https://theses.hal.science/tel-01676987
Submitted on 7 Jan 2018
HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
Université Paris Diderot - Paris 7
Laboratoire
Matériaux et Phénomènes Quantiques
École Doctorale 564 :
Physique en Île-de-France
UFR de Physique
THÈSE
présentée par
Claire AUTEBERT
pour obtenir le grade de

Docteur ès Sciences de l’Université Paris Diderot
AlGaAs photonic devices: from quantum state

generation to quantum communications
Soutenue publiquement le 14 novembre 2016,

devant la commission d’examen composée de :
M. Philippe Adam, Invité

M. Philippe Delaye, Rapporteur
Mme Sara Ducci, Directrice de thèse
M. Riad Haidar, Président
M. Steve Kolthammer, Examinateur
M. Aristide Lemaître, Invité
M. Anthony Martin, Invité
M. Fabio Sciarrino, Rapporteur
M. Carlo Sirtori, Invité
Acknowledgment
En premier lieu, je tiens à remercier Sara Ducci qui a été pour moi une excellente directrice de thèse,
tant du point de vue scientifique que du point de vue humain. Elle a su être toujours très présente,
tout en me laissant une grande liberté dans mon travail. Ces trois années de thèse ont été un plaisir
grâce à elle et j’espère avoir le bonheur de pouvoir continuer à collaborer avec elle.
Je suis reconnaissante à Philippe Delaye et Fabio Sciarrino d’avoir accepté de prendre le temps
d’être rapporteurs de cette thèse. Merci également à Philippe Adam, Riad Haidar, Steve Kolthammer,
Aristide Lemaître, Anthony Martin et Carlo Sirtori pour leur participation à mon jury de thèse.
Parce qu’un travail de thèse ne se fait souvent pas seul, je voudrais dire un mot sur les collabora-
tions qu’il m’a été possible de tisser pendant ces trois ans. Tout d’abord merci à Aristide Lemaître,
Elisabeth Galopin et Carme Gomez-Carbonell pour la croissance de mes échantillons (et merci de
m’avoir laissée venir une journée dans votre antre pour voir comment tout cela fonctionne !). Pendant
cette thèse j’ai eu l’opportunité de partir quelques semaines à l’université de Genève. Merci donc
à Natalia Bruno et Anthony Martin pour m’avoir prise en charge et surtout appris tant de choses.
Merci à Hugo Zbinden pour avoir permis à cette collaboration de voir le jour, pour son aide sur la
manip’ et sa gentillesse. Merci à Rob Thew pour l’intérêt qu’il a montré à la manip’ et les discussions
et conseils qu’il nous a donnés. Enfin, il n’est pas toujours évident de se retrouver dans un nouveau
laboratoire, loin de sa zone de confort. Je suis ainsi reconnaissante à toutes les personnes du GAP de
Genève de m’avoir si bien accueillie et permis de me sentir immédiatement comme chez moi. Merci
donc à Natalia (encore), Thiago, Ephanielle, Gianluca, Boris, et tous les autres ! Ensuite il y a eu la
collaboration avec Telecom ParisTech. Cette fois, ce sont eux qui sont venus dans mon laboratoire
et j’espère qu’ils s’y sont bien sentis. Merci à Julien Trapateau d’avoir passé tout ce temps sur les
mesures avec moi et à Adeline Orieux, Eleni Diamanti et Isabelle Zaquine pour toutes les discussions
que nous avons eues. Merci à Ivan Favero pour le prêt du laser. Enfin au laboratoire MPQ merci à
Carlo Sirtori et Maria Amanti pour les discussions sur la diode laser et le prêt du cryostat, avec la
table de manip’ qui allait avec. Merci à Pérola Milman pour son enthousiasme sur mes mesures de
HOM et les discussions sur le lien entre elles et la fonction de Wigner.
Ce manuscrit a été relu avec beaucoup de soin par David Fenby pour l’anglais. Je le remercie très
sincèrement pour ce travail. Je veux aussi remercier de nouveau Adeline Orieux pour tout le temps
passé à m’aider à préparer la soutenance et toutes ses réponses à mes nombreux mails de dernière
minute.
Merci beaucoup à Anne Servouze, qui fait toujours tout pour que tout se déroule dans les
meilleures conditions, et à Joëlle Mercier, Jocelyne Moreau et Sandrine Di Concetto du service
administratif pour leur efficacité et gentillesse. Merci aussi à Patrick Lepert et Martial Nicolas de
l’atelier méca’ et Marc Apfel de l’atelier élec’ pour toutes les ‘pièces’ qu’ils m’ont fabriquées, et en
particulier pour leur intérêt, enthousiasme et implication quand je suis arrivée avec mon idée, très
vague, de harpe laser pour la fête de la science : cela a été une réussite. Merci aux ingénieurs salle
blanche : à Christophe Manquest qui m’a initiée à ce nouveau monde, et à Pascal Filloux et Stephan
Suffit pour avoir toujours répondu à mes questions et pour les discussions qui font passer les temps
d’attente toujours plus vite. Je voudrais aussi remercier Michael Rosticher de l’ENS pour m’avoir
aidée à utiliser leur microscope électronique pour les lithographies dans la bonne humeur, même
quand j’arrivais avec un masque faux à refaire entièrement.
i
ii Acknowledgment
Un merci spécial pour Fabien Boitier, ‘mon’ post-doc. Pour le temps passé à m’apprendre tout
sur la manip’, à répondre à mes nombreuses questions, à m’obliger à faire des calculs dans lesquels
je ne voulais pas me lancer ; pour les réponses aux mails, même après avoir quitté le labo, et pour
son intérêt constant : merci.
Trois ans dans un laboratoire pour faire une thèse, c’est aussi beaucoup de collègues, qui rendent
la vie sympathique. Pour les échanges de matériels et de tuyaux, et la bonne ambiance, merci aux
doctorants et post-doc du thésarium et de l’équipe DON : Alessio, Marco, Chris, David, Mehdi,
Debora, Trung, Eduardo, Pierre, Marc, Natalia, Alice, Alok, Valerio, Oleksandr, Jonathan, Giorgio
(bon courage pour la suite !), Saverio, Qifeng, Andreas. Merci à Silvia, qui a toujours été là et
su trouver les mots quand je doutais de moi, et à Cécile pour nos rafraichissantes pauses thé et
les discussions sur tout et rien. Merci à William et Biswarup pour ces trois années de thèse faites
ensemble et tous les fou-rire qui ont été avec. Et merci à Yacine pour les nombreuses discussions que
nous avons eues, tant sur le travail que sur le reste. Merci aussi à Florent Baboux, tout nouveau
dans l’équipe, pour son aide sur la préparation de la soutenance et sa bonne humeur constante.
Et bien sûr je voudrais remercier tous les doctorants et post-doctorants du laboratoire pour tous
les bons moments passés ensemble : Charlotte, Benjamin, Roméo, Anna, Simon, Thibault, Pierre,
Giulia, Philippe, Hélène, Margaux, Benjamin, Zahra, Batiste, Tom, Andreas, Adrien, Aurianne,
Saulo, Simone, Giulia,. . .
Parce que la thèse est l’aboutissement d’une longue scolarité, je voudrais remercier l’ensemble
de mes professeurs qui m’ont donné goût aux études et aux sciences. Merci tout particulièrement à
Alain Lumediluna, professeur de physique en MP, qui m’a mise dans les rails de la recherche quand
je ne savais plus très bien quelle école d’ingénieur faire, et ce même si cela n’aidait pas les statistiques
du lycée.
De manière plus personnelle, merci à mes amis qui me soutiennent depuis toutes ces années.
Merci à ma famille et surtout à mes parents, qui m’ont inculqué la curiosité nécessaire à la recherche,
et qui, en ayant l’air de tellement s’épanouir dans leur travail de chercheur, m’ont donné l’envie de
me lancer dans cette voie. Et comme on dit en anglais, last but not least, merci Guillaume pour ton
amour, ton soutien et ta présence de tous les jours.
Contents
Introduction 1
1 Theoretical background 5
1.1 Nonlinear optical processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.1 Generalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.2 Three Wave Mixing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.1.3 Three wave-mixing in waveguides . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.1.4 Phase-matching strategies in AlGaAs . . . . . . . . . . . . . . . . . . . . . . . 15
1.2 Integrating the pump laser in a nonlinear waveguide . . . . . . . . . . . . . . . . . . . 17
1.2.1 Semiconductors lasers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.2.2 Working principle of our device . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.3 Characterization of the quantum state of photon pairs . . . . . . . . . . . . . . . . . . 21
1.3.1 Indistinguishability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.3.2 Quantum bits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.3.3 Entanglement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.4 Quantum key distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2 Fabrication 35
2.1 Heterostructure design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.2 Sample growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.3 Sample processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.3.1 Waveguide etching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.3.2 Back metallic contact deposition . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.3.3 Insulator deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.3.4 Up metallic contact deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3 Characterization of the laser emission 47

3.1 Laser emission on the Bragg mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.2 Optoelectronics characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.3 Tunability of the laser wavelength . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.4 Towards continuous wave operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4 Characterization of the nonlinear waveguide 53

4.1 Measurement of propagation losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.2 Second-harmonic generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2.1 Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2.2 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.2.3 Tunability of the phase-matching wavelength . . . . . . . . . . . . . . . . . . . 60
4.3 Spontaneous Parametric Down-Conversion . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.4 Joint Spectral Amplitude and Intensity . . . . . . . . . . . . . . . . . . . . . . . . . . 64
iii
iv CONTENTS
5 Characterization of the quantum state emitted by a passive source 69

5.1 Indistinguishability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.2 Energy-time entanglement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
6 Multi-user quantum key distribution 79

6.1 Experimental method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
6.2 Quantum key distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6.4 Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
7 An electrically injected photon-pair source at room temperature 89

7.1 Research of the working conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
7.2 Coincidences probability calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
7.2.1 Transmission of the generated photons . . . . . . . . . . . . . . . . . . . . . . . 91
7.2.2 Coincidence probabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
7.2.3 Estimating the fidelity to a Bell state from the SNR . . . . . . . . . . . . . . . 94
7.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
Conclusion 99
Appendix 103
A List of Conferences and Publications 103
Bibliography 145
Introduction
Since the establishment of the theory of quantum mechanics at the beginning of last century, a large
number of new concepts, discoveries and inventions have followed. In particular, the last thirty
years have seen the emergence of the field of quantum information, allowing to transport and treat
information using the state of quantum systems. The most publicized applications of this field are
quantum cryptography [1, 2], quantum computing [3] and quantum metrology [4]. Among the main
concepts at the basis of quantum information, we can cite the use of quantum bits (or Qubits),
which are two-state quantum mechanical system. While the classical bit can be in one or the other
of its two possible states, the qubit can be in a superposition of both states at the same time, a
property which is fundamental to quantum computing. Another important element is the no cloning
theorem, stating that an arbitrary quantum state cannot be copied; this theorem is fundamental
to achieve perfect security in quantum cryptography protocols. Finally entanglement, consisting
in quantum correlations between two or more particles, leads to many advantages for example in
security protocols, parallel computing and metrology.
In these last years, a great deal of effort has been devoted to the miniaturization of components
able to generate, manipulate and detect qubits. This has led to the development of a large panel
of approaches using different physical supports in view of future quantum information technologies;
these include ions and atoms chips [5, 6], superconductor circuits [7], single and two photon sources
and detectors [8, 9, 10] and photonics circuits to manipulate single and multiple photon states [?].
In the coming years the challenge is to pass from laboratory experiments to real world applications;
this needs to achieve miniaturized and robust components and possibly their integration in a single
chip.
Among the different approaches under investigation, some are more suitable for the implemen-
tation of quantum memories or quantum processors, while others are more suitable for quantum
communication and cryptography. For these two last applications, photons have a great advantage
over other physical systems under investigation, and it will certainly play a key role in the devel-
opment of future technologies [11, 12, 13]. Indeed, owing to their high speed, making them natural
flying Qubits, and their almost immunity against decoherence, photons are an ideal support to convey
information over long distances both in free space and fibers, in particular at wavelengths compat-
ible with telecommunication networks [14, 15]. Consequently, optics is the dominant approach in
quantum cryptography [16], teleportation [17], and entanglement distribution over long distances
[14, 18].
This is the context for the work presented in this thesis. My main goal has been the demonstration
of a miniaturized electrically driven photon-pair source, emitting in the main telecom band, and
operating at room temperature.
Two-photon sources state of the art
Numerous approaches have been used to produce photon-pair sources. Historically, atomic cascades
in Calcium have been first used to demonstrate the existence of entangled states [19, 20]; this was
the validation of a major concept of quantum mechanics, which has been followed by more and more
refined experiments close all the possible loopholes.
1
2 INTRODUCTION
With the development of the crystals for nonlinear optics, the most used process to produce
photon pairs has become spontaneous parametric down conversion (SPDC), which has been obtained
for the first time in birefringent dielectric bulk crystals [21]. One of the most striking success was
the realization by Kwiat et al., in 1995, of a polarization entangled photon source, which was widely
used afterward in quantum optics (violation of Bell’s inequality by over 100 standard deviations).
Then the use of the quasi-phase matching technique in dielectric waveguides (periodic rever-
sal of the nonlinear susceptibility) has permitted to achieve an efficiency of 10−6 pairs per pump
photon thanks to the confinement of the interacting modes [?, 22]. The most widely used sources
today are periodically poled lithium niobate (PPLN) waveguides, generating photon pairs at telecom
wavelength.
In order to maximize the collection efficiency, several groups work on the generation of two photons
states inside optical fibers. In particular, entangled photon-pairs have been generated with photonic
crystal fibers by four wave mixing [23] and with periodically poled fibers by three wave mixing [24].
More recently, four wave mixing has also been used in liquid-core fibers [25], which has the advantage
to produce Raman-free correlated photons, which is one of the main issues for spontaneous four wave
mixing in fibers. Very recently, four-wave mixing has been demonstrated in high-refractive-index
glass in microring resonator architecture [26], which is a CMOS compatible platform.
In the context of entangled photon sources, the target property of the generated quantum state
depends on the intended application: photon indistinguishability is essential to controlled logic gates
[27], quantum networking [28] and boson sampling [29, 30, 31]; entanglement is used to speed up
algorithms [32], protect encoded information [2], teleport quantum states [33], and reduce intrinsic
uncertainty in interferometric measurement [34]. For all these reasons, there is a considerable interest
in developing high-performance sources generating controlled quantum states, operating at room
temperature, easy to fabricate and compatible with the existing telecom fiber network.
In this context, semiconductor materials have a strategic position and several teams are working
using different systems and platforms. Their attractiveness comes from the high technological control
in both growth and processing, their small size, their optoelectronic capabilities, and their compati-
bility with existing technologies. Two main physical processes are used today to produce entangled
photons in semiconductor devices: the first is the radiative decay of a biexciton state in a single quan-
tum dot. Entangled photon generation has been demonstrated under both optical [35] and electrical
[36] pumping. This approach has led to bright and almost deterministic sources. However, since the
use of this process requires that the thermal energy of the system is inferior to the biexciton energy,
these sources have to be cooled to temperatures of few Kelvin, thus reducing their compactness and
integrability. The second approach is based on optical parametric conversion, the main platforms
being Silicon and AlGaAs. Despite its nondeterministic nature, this approach is the most widely
used to produce photon pairs for quantum information and communication protocols. The spectac-
ular progress in silicon photonics has recently led to the appearance of entangled photon sources
with integrated devices having different geometries (waveguides, nanowires, whispering-gallery mode
resonators, etc.) [37, 38, 39]. Since Silicon does not naturally show a second-order nonlinearity, the
physical process used in this case is four-wave mixing. Sources emitting in the telecom band at room
temperature have been produced making use of Silicon-On-Insulator (SOI) and Silica-On-Silicon
(SOS) platforms, compatible with CMOS (Complementary Metal Oxide Semiconductor) technology.
However, the indirect band gap of silicon hinders the achievement of electrically driven devices, which
have a clear advantage over optically driven ones in terms of portability, energy consumption and
integration. On the other hand, AlGaAs presents a second order nonlinearity [40] allowing to produce
photon pairs by spontaneous parametric down conversion [9, 41, 42]; moreover, its direct bandgap
allows to integrate the pump laser within the device while working at room temperature. This plat-
form is attracting the interest of several teams, and some building blocks are already available: the
electrically driven photon pairs source described in this thesis, quantum circuits able to manipulate
the photon state through electro-optics effect [?], GaAs waveguide integrated with superconducting
nanowires for highly efficient single photon detection [10]. These achievements underline the potential
INTRODUCTION 3
of this platform in realizing miniature chips containing the generation, manipulation, and detection
of quantum states of light. Note also that recent works using the third order nonlinearity of AlGaAs
have been reported to produce photon-pair sources [43].
Outline of the manuscript

The first chapter introduces the theoretical concepts relevant for my work. It describes in a first
part the tools underlying the working principle of the device. The second part explains how to
characterize quantum state emitted by a photon-pairs source and how this state can be used in a
quantum cryptography protocol.
After a brief introduction on the design and growth steps, chapter 2 describes the sample fabri-
cation procedure, including the waveguide definition and the electrical contact deposition. Chapter
3 presents the optoelectronics characterization of the device under electrical pumping. The optical
and electrical characteristics are given, as well as the laser wavelength tunability. Finally we discuss
the continuous wave operation of our device. Chapter 4 presents the results on the optical losses
measurements and the nonlinear optical characterization of the device including second harmonic
generation (allowing to identify the phase-matching resonance) and spontaneous parametric down
conversion. Finally, a method allowing high accuracy in reconstructing the biphoton frequency state
is used. This chapter also includes a comparison of two fabrication techniques of our waveguides.
Chapters 5 and 6 focus on the characterization of the quantum state generated by a passive sample
and its utilization in a quantum key distribution protocol. Chapter 5 describes the demonstration
of indistinguishability (via a Hong-Ou-Mandel experiment), and of energy-time entanglement (via a
Franson experiment). Chapter 6 describes how our source can be used for a multi-user quantum key
distribution (over up to 50 km), a core protocol for securing communication in quantum networks.
The key information is encoded using polarization entanglement; the small birefringence of AlGaAs
waveguides and the large broadband of the source emission are used to implement a simple and
robust experimental setup.
Finally, chapter 7 presents our work on the first electricallly driven photon pairs source emitting
in the telecom range and working at room temperature.
Chapter 1
Theoretical background
In the first section of this chapter, we describe the basic concepts of nonlinear optics, focusing
on the equations of three-wave mixing, and presenting the main phase-matching schemes used in
AlGaAs. Then we recall the functioning of a laser diode and we explain the working principle of
our device. The second section reviews the tools used to characterize the quality of a photon-pair
source: we first describe how to measure indistinguishability between two photons; we then explain
what entanglement is, and we conclude by showing one of its application, quantum key distribution.
1.1 Nonlinear optical processes
1.1.1 Generalities
Nonlinear optics is the branch of optics that describes the optical effects arising from the interaction
of intense optical fields with matter. Such intense fields are not available with classical light sources
and for a long time most observed optical phenomena were described by linear equations. But, as in
most physical phenomena, a linear description is only an approximation of nonlinear processes that
holds in a limited range. The demonstration of the first laser in 1960 by Maiman [44] gave rise to the
field of nonlinear optics, beginning with the first observation of frequency doubling of a ruby laser
beam in a quartz crystal by Franklin et al. [45] in 1961. Since then, the progress in laser sources and
nonlinear media has led to the discovery of a full range of new effects and phenomena, which have
given rise to many applications.
In a simple picture, matter consists of positively charged nuclei (of atoms or molecules) and sur-
rounding negatively charged electrons. Those microscopic dipoles can be assimilated with oscillators
[46]. When a weak light field is sent through the medium, the electronic cloud oscillates in a parabolic
potential at the excitation frequency. When the light field intensity increases, the potential becomes
anharmonic and higher order harmonics appear in the induced macroscopic polarization spectrum.
These higher order harmonics act as new radiation sources at new optical frequencies: this is the
phenomenon of frequency conversion.
The most rigorous approach to nonlinear optics is based on quantum electrodynamics, treating
both the medium and the optical field as quantized systems. This has considerable formal complexity.
However, in many cases, a semi-classical treatment can be sufficient; in this case the medium is
described with quantum mechanics, and light radiation with classical electromagnetic field theory.
The quantum description of matter is used to obtain the macroscopic nonlinear electric susceptibilities
of the medium, which can then be employed in Maxwell equations. This theoretical approach is
suitable for describing wave-mixing interactions, with the exception of parametric fluorescence, where
the introduction of quantum noise is necessary to generate the output fields. Hereafter, we present
both treatments as we used both in this thesis.
5
6 THEORETICAL BACKGROUND
Nonlinear polarization
In the following, only nonlinear dielectric, homogeneous and time invariant media, which have a local
response in time and space, will be considered. This means that we restrict ourselves to interactions
where the media are transparent. In this case, the electric polarization P~ (~r, t) induced in the medium
can be developed as a power series expansion of the applied optical field E(~ ~ r, t):
~NL
P
~L
P z }| {
~ r, t)2 + 0 χ(3) ...E(~
z }| {
P~ (~r, t) = 0 χ(1) · E(~
~ r, t) + 0 χ(2) : E(~ ~ r, t)3 + . . . (1.1)
| {z } | {z } | {z }
~ (1)
P ~ (2)
P ~ (3)
P
where 0 is the free space permittivity and χ(m) is a rank (m + 1) tensor which characterizes the mth-
order electrical susceptibility of the medium. The susceptibility amplitude decreases rapidly with the
nonlinear order ( χ(1) ∼ 1, χ(2) ∼10−11 m/V, χ(3) ∼10−22 m2 /V2 [46]); for a weak field the linear
termqP~L is generally sufficient to describe the light-matter interaction, through the refractive index
n = 1 + <(χ(1) ), but for a strong field, the nonlinear term P~NL becomes significant.
An applied field being a superposition of (quasi-)monochromatic waves, it is possible to write the
~ r, t) and P~ (~r, t) through their Fourier
counterpart of eq. 1.1 in the frequency domain by expressing E(~
transform. The total electric field can be decomposed in :
{êl El (~r, ωl )ei(kl · ~r−ωl t) + c.c.}

~ r, t) =
X ~
E(~ (1.2)
l
where êl is the unit vector depicting the polarization of the wave, El the wave amplitude, ~kl the wave
vector and ωl the angular frequency, while c.c. stands for complex conjugate, ensuring that E(~ ~ r, t) is
a real quantity. The Fourier transform of eq. 1.1 gives then:
P~ (ω) = P~ (1) (ω)δ[ω − ω1 ] + P~ (2) (ω)δ[ω − (ω1 + ω2 )] + P~ (3) (ω)δ[ω − (ω1 + ω2 + ω3 )] + . . . (1.3)
with
P~ (m) (ω) = 0 χ(m) (−ω; ω1 , ω2 , . . . , ωm ) : E(ω
~ 1 )E(ω
~ 2 ) . . . E(ω
~ m) (1.4)
where the frequencies coupled by the mth-order susceptibility fulfill the summation relation
m
X
ωj = ω, (1.5)
j=1
which expresses the energy conservation during the conversion process.
Propagation in a nonlinear medium

The material’s nonlinear response to an intense field can cause the medium’s polarization to develop
new frequency components not present in the incident radiation field. For a non-magnetic, isotropic
and dielectric optical medium with no free charges or currents, Maxwell’s equations give

~
~ = − ∂B

~ ×E

∇ (1.6a)


∂t
~
~ = ∂D


∇

 ~ ×H (1.6b)
∂t
where H~ = 0 c2 B~ and D~ = 0 E
~ + P~ . This leads to a wave equation describing the coupling between
the electrical field and the induced polarization:
2 2~ 2~
∇ ~ − n ∂ E = 1 ∂ PNL
~ 2E (1.7)
c2 ∂t2 0 c2 ∂t2
1.1. NONLINEAR OPTICAL PROCESSES 7
where c is the speed of light in vacuum. This last equation shows that the nonlinear response
of the polarization acts as source term in the wave equation to generate new components of the
electromagnetic field.
1.1.2 Three Wave Mixing

Second-order polarization
Since the studies reported in this thesis involve AlGaAs, which is a medium presenting second-order
nonlinear susceptibility, in the following we will focus only on the second-order term of the nonlinear
polarization, neglecting the higher-order terms. We can then express eq. 1.4 for the ith Cartesian
component of the second-order nonlinear polarization as:
(2) X (2)
Pi (ω3 ) = 0 D(2) χijk (−ω3 ; ω1 , ω2 ) · Ej (ω1 ) · Ek (ω2 ) . (1.8)
j,k
where D(2) is a degeneracy factor of value 2 for distinguishable fields and 1 for indistinguishable ones.
Symmetry relations of the nonlinear susceptibility

One can note that, with respect to reciprocal interactions, i.e. ω3 = ω1 + ω2 , ω1 = ω3 − ω2 and
ω2 = ω3 − ω1 , the second-order susceptibility tensor exhibits the following permutation symmetry
(2) (2) (2)
χijk (−ω3 ; ω1 , ω2 ) = χjki (−ω1 ; −ω2 , ω3 ) = χkij (−ω2 ; ω3 , −ω1 ) . (1.9)
Moreover, in situations far from any resonance of the medium (as is the case here), the dispersion of
the nonlinear response can be neglected, and the second-order susceptibility can then be considered
independent of the frequencies. Hence the Kleinman’s symmetry condition [47] is valid and all
permutations of (i, j, k) are equivalent.
By introducing the tensor d:
1 (2)
dijk = χijk (1.10)
2
the intrinsic permutation symmetry allows a simplification of the notation by introducing a contracted
matrix dil according to the prescription
jk 11 22 33 23/32 13/31 12/21

,
l 1 2 3 4 5 6
which reduces the nonlinear susceptibility from a 33 tensor to a 3 × 6 matrix. Then the introduction
of the Kleinman’s symmetry condition reduces the number of independent elements of dil to only 10:
 
d11 d12 d13 d14 d15 d16
d16 d22 d23 d24 d14 d12  . (1.11)
 
d15 d24 d33 d23 d13 d14
Finally, for Alx Ga1-x As compounds, the zincblende crystallographic structure imposes a cubic 4̄3m
symmetry of the crystal, leading to a unique non-zero element (d14 ) of the matrix [48]:
 
0 0 0 d14 0 0
0 0 0 0 d 0 . (1.12)
 
 14
0 0 0 0 0 d14
Overview of the three-wave mixing processes

The second-order nonlinearity allows several interactions between three photons of different angular
frequency ω1 , ω2 and ω3 (see figure 1.1), which can either be created (downward arrows) or annihilated
(upward arrows) during the process. Energy conservation (term with m = 2 of eq. 1.5) defines the
possible conversion processes:
~ω3 = ~ω1 + ~ω2 (1.13)
where ~ is the reduced Planck’s constant. As the medium is transparent, the energy transfers do not
take place on real energy levels but on virtual levels: there is no energy transfert between the fields
and the medium, although the medium supports the interaction. The different possible interactions
are:
• Sum Frequency Generation (SFG): two pump fields of frequency ω1 and ω2 interact within
the medium to produce a third field at frequency ω3 = ω1 + ω2 (figure 1.1.a).
• Second Harmonic Generation (SHG): this is the degenerate case of SFG in which the two
pump fields have the same frequency ω = ω1 = ω2 . In this case a unique pump field is needed,
which interacts with itself and the medium, to produce a field at frequency 2ω (figure 1.1.b).
• Difference Frequency Generation (DFG): two pump fields of frequency ω3 and ω2 interact
with the medium to produce a third field at frequency ω1 = ω3 − ω2 (figure 1.1.c). At the same
time, the field at frequency ω2 is amplified [Optical Parametric Amplification (OPA)] in
order to fulfill energy conservation.
• Spontaneous Parametric Down-Conversion (SPDC): a photon at frequency ω3 sponta-

neously annihilates, with the simultaneous generation of two photons at frequencies ω1 and ω2 ,
triggered by vacuum fluctuations (figure 1.1.d).
(a) SFG (b) SHG (c) DFG(/OPA) (d) SPDC
Figure 1.1 – Three wave-mixing interactions: (a) sum frequency generation; (b) second harmonic
generation; (c) difference frequency generation and optical parametric amplification; (d) spontaneous
parametric down-conversion. Solid black lines correspond to real energy levels of the medium and
dashed lines represent the states of the virtual transitions.
Three wave-mixing coupled dynamics

In three wave-mixing, the total electric field of eq. 1.2 restrains to the sum of three components with
oscillation frequencies ω1 , ω2 , and ω3 . Assuming the direction of propagation is along the z axis, it
writes:
3
~ t) =
X
E(z, {êl El ei(kl z−ωl t) + c.c.} . (1.14)
l=1
We can then substitute the electric field (eq. 1.14) and the second-order nonlinear polarization
(eq. 1.8) into the wave propagation equation (eq. 1.7). Under the slowly varying envelope approxi-
mation, this equation appears in Fourier space as a closed system of coupled differential equations:

dE1 ω1
= −i deff E2∗ (z)E3 (z)e−i∆kz


 (1.15a)
dz n c

1




 dE
2 ω 2
= −i eff 1 d E ∗ (z)E (z)e−i∆kz
3 (1.15b)


 dz n2 c
dE3 ω3


= −i deff E1 (z)E2 (z)ei∆kz (1.15c)



dz n3 c

where ni is the medium index at frequency ωi , deff is the effective quadratic susceptibility of the
three-wave-mixing, and ∆k = k3 − k2 − k1 is call the phase-mismatch. This last parameter is a key
factor in the three wave interaction.
Phase-matching consideration
Since the interacting fields are traveling waves, this is an extended, not a point, interaction. Thus
the relative phase of the interacting waves is an important factor. Eqs. 1.15 show that if the phase-
mismatch is non-zero (∆k 6= 0), the evolution of the fields amplitude along the propagation direction
contains an oscillating term, which is detrimental to the frequency conversion efficiency. On the other
hand, if the phase-mismatch is zero (∆k = 0), the imaginary exponential terms become constant and
the amplitude of the generated field can increase with the length of propagation (see figure 1.2), thanks
to the constructive interaction among the three waves. Hence, from all the possible interactions, only
those satisfying the phase-matching condition
∆k = k3 − k2 − k1 = 0 , (1.16)
will occur efficiently inside the medium. In a corpuscular picture, this condition is known as momen-
tum conservation.
Sum Frequency Generation efficiency
To continue the development of the theory, we will take SFG as an example. The procedures are
roughly the same for SHG and DFG.
In SFG, two pump fields at frequency ω1 and ω2 interact in the medium to produce a field at
frequency ω3 = ω1 + ω2 . In the absence of pump depletion1 and losses in the medium, the optical
power generated by SFG is obtained by solving eq. 1.15 and integrating the resulting wave intensity
over the effective area A of the interacting beam [48]:
8π 2 2 2 2 ∆kL

Pω3 (L) = deff L Pω 1 Pω 2 sinc (1.17)
0 cλ2ω3 nω1 nω2 nω3 A 2
where λω3 is the wavelength corresponding to the frequency ω3 , Pω1 and Pω2 are the optical powers
of the two pump fields, and L is the length of the nonlinear medium.
1
There is no pump depletion if the conversion efficiency is small enough to neglect the variations of the pump field.
This approximation is always valid in this thesis.
0
The conversion efficiency ηSFG and the normalized conversion efficiency ηSFG of the nonlinear
process are then defined as:
Pω3 ∆kL
ηSFG = 0
= ηSFG L2 sinc2 [W−1 ] (1.18a)
Pω1 Pω2 2
8π 2
0
ηSFG = d2 [W−1 cm−2 ] . (1.18b)
0 cλω3 nω1 nω2 nω3 A eff
2
0
ηSFG is a good parameter to compare performances between different devices, as it is independent
of the nonlinear medium length. On the other hand, ηSFG is a good parameter to evaluate the
performances of a particular device in terms of output power.
Once again, the impact of the phase-matching condition on the generated optical power is clear.
For ∆k 6= 0 the energy transfer between the pump fields and the generated field is periodically inverted
along the direction of propagation. As shown in figure 1.2.a, this leads to negligible conversion
efficiency in comparison with the case of perfect phase-matching, where the generated field grows
quadratically with the interaction length. Moreover, the phase-matching condition determines the
spectral bandwidth of the conversion process, as shown in figure 1.2.b.
a) b)
Figure 1.2 – a) Conversion efficiency over normalized conversion efficiency as a function of propagation
distance in a nonlinear medium for perfect phase-matching (blue curve) and for a phase mismatch
∆k = 10/L (red curve) for SFG; b) SFG normalized resonance as a function of phase mismatch.
1.1.3 Three wave-mixing in waveguides

As can be seen in eq. 1.17, the conversion efficiency is inversely proportional to the effective area
A of the interacting beam: when the optical power is concentrated in a small area, the electric
fields are more intense, which leads to a higher response from the nonlinear medium. When the
phase-matching condition is achieved, the amplitude of the generated field is also proportional to the
square of the interacting length L. Hence the use of guided-wave geometry, in which the interacting
waves can be confined over a small area and propagated over a long distance without diffraction,
can significantly enhance the intensities of the generated fields, provided that the phase-matching
condition is satisfied. Moreover, the use of waveguides allows to control the spatial mode of the
created photons; this facilitates fiber coupling, eliminates spatial walk-off, and opens the way to the
miniaturization of circuits and devices.
In the previous sections, the waves were supposed to be plane waves propagating in a nonlin-
ear medium. In the following sections, we shall see how to adapt the equations and solutions to
waveguides.
cladding
core
cladding
Figure 1.3 – Propagation of light in a waveguide.
Guided propagation in ridge waveguides

A large variety of devices has been developed to guide light (fibers, waveguides, photonic crystals,
microcavities, ...).
An optical waveguide is a physical structure that guides electromagnetic waves of the optical
spectrum by confining them in the transverse dimensions and allowing them to propagate in the
longitudinal dimension. In the picture of geometrical optics, the waves confinement in the core of the
waveguide is obtained by total reflection at the claddings surronding it (figure 1.3). A propagation
mode in a waveguide is a solution of the wave equations (the fields from multi-reflections at the
core/cladding interfaces have to constructively interfere in the waveguide). Due to this condition,
there is only a discrete set of frequencies and forms for the wave function which can propagate in the
waveguide. The mode with the lowest cutoff frequency is the basic mode of the waveguide, and its
cutoff frequency is the waveguide cutoff frequency.
In this thesis we will focus on ridge waveguides. This type of waveguides provides wave-confinement
in two dimensions (figure 1.4) and is particularly adapted to semiconductor materials. Indeed the
vertical confinement is provided by epitaxial growth of different layers having different refractive
indexes, while the lateral confinement is provided by etching a ridge structure.
TM
TE
Cladding
Cladding
Core Core
Cladding Cladding
Substrat Substrat
Figure 1.4 – Sketch of ridge waveguides. Light gray corresponds to high refractive index layers. The
lateral confinement is provided by ridge etching.
The waveguide modes are characterized by:
• a polarization direction T E (Transverse Electric, for an electric field parallel to the layers) or
T M (Transverse Magnetic, for an electric field orthogonal to the layers);
• an effective wave-vector ~keff , associated to an effective refractive index neff , that measures the
phase velocity: ~keff = ~ez neff ωc ;
• a field spatial distribution E(x, y, z) = Φ(x, y)Ez (z), where Ez (z) is the z-dependent part of
the electric field and Φ(x, y) defines the transverse intensity distribution of the mode.
The two physical phenomena we will use to confine light in our device are:
• Total internal reflection: this is a phenomenon which occurs when the refractive index of
the core is higher than that of the cladding (ncore > nclad ). Again in the picture of geometric
optics, when a wave is injected into the core, there exists a critical angle for the propagating
wave, above which the wave cannot pass through the cladding and is entirely reflected into the
core. For such modes, nneff satisfies ncore > neff > nclad ;
• Photonic bang gap: in this case the core is surrounded by regions of regularly spaced high
and low dielectric constant. Thanks to the periodic refractive index modulation, a photonic
band gap arises at certain wavelengths2 , which prevents them propagating into the cladding
region, thus forcing their transverse component to propagate into the core. For such modes,
nneff is smaller than any refractive index of the structure.
Frequency conversion in waveguides

The spatial overlap between the modal distributions and the nonlinear coefficient can then be ex-
pressed through the nonlinear overlap integral
ZZ
Γ= dxdy deff (x, y)Φ1 (x, y)Φ2 (x, y)Φ3 (x, y) , (1.19)
S
dxdy Φ2 (x, y) = 1.
RR
with normalized modal distributions S
In this case, eq. 1.17 becomes:
8π 2 2 2 2 ∆kL

Pω3 (L) = |Γ| L Pω 1 Pω2 sinc , (1.20)
0 cλ2ω3 neff eff eff
ω1 n ω2 n ω3 2
and the normalized conversion efficiency is given by
8π 2
0
ηSFG = |Γ|2 . (1.21)
0 cλ2ω3 neff eff eff
ω1 nω2 nω3
The same modifications occur in the formula for SHG and DFG.
Spontaneous Parametric Down-Conversion

If we stay in the framework of the semi-classical approximation, when a single field E3 at frequency
ω3 is sent to the nonlinear medium, eqs. 1.15 shows that the fields E1 and E2 cannot be generated and
E3 stays constant. However, experiments show that new waves do appear (the first demonstrations
of SPDC were carried out by Magde and Akhlanov [49] and Byer and coworkers [50] in 1967). To
understand this phenomenon, it is necessary to take into account vacuum fluctuations, which are set
as seed leading to the generation of new fields. This process can no longer be explained using a semi-
classical approximation but requires a quantum theoretical approach, which we develop hereafter.
In the following the input intense field will be called pump field (p) and the two generated fields
will be called signal (s) and idler (i), with for convention ωi ≤ ωs .
Quantization of the electromagnetic field: Following Ou et al. [51], we describe the two-photon
state produced by the SPDC process using a perturbation expansion in the interaction picture. The
canonical quantization of the electromagnetic field is performed by substituting the amplitudes of the
electric fields with the corresponding non-commuting creation and annihilation quantum operators:
[âk,σ , âk† 0 ,σ0 ] = δσσ0 δ(k − k 0 ) . (1.22)

2
In the same way that the periodic potential in a semiconductor crystal affects electron motion by defining allowed
and forbidden electronic energy bands.
The electric field observable can then be expressed as3
Ê(r, t) = Ê(+) (r, t) + Ê(−) (r, t) (1.23)
with
Ê(−) (r, t) = (Ê(+) (r, t))† (1.24)
and s
L X
Z
Ê(+) (r, t) = d3 k eσ Eσ (x, y)ei(kσ r−ωσ t) âk,σ , (1.25)
2π σ
where σ stands for two orthogonal polarizations perpendicular to the propagation direction z. Since
the waveguide enforces collinear propagation directions for all three fields, the wave vector can be
replaced with a scalar, and the electric field collapses to a one-dimensional integral in the z direction:
s
L X
Z
Ê(+) (r, t) = dkz eσ Eσ (x, y)ei(kσ z−ωσ t) âk,σ . (1.26)
2π σ
Then the integration over kz can be translated into a frequency integral by substituting
dω q
dkz → , â(kz ) → vg (ω)â(ω) , (1.27)
vg (ω)
the quantity vg (ω) being the group velocity of the light inside the waveguide, which is given by
vg = ∂ω
∂k . This gives the following expression for the quantized electric field operator:
s
L X ei(kσ z−ωσ t)
Z
(+)
Ê (r, t) = eσ Eσ (x, y) dω q âω,σ . (1.28)
2π σ vg (ω)
The nonlinear Hamiltonian: As previously, the medium nonlinear response is assumed to be local
and the nonlinear interaction Hamiltonian is the quantum equivalent of the classical nonlinear density
representing the interaction of the three waves, integrated over the whole interaction volume [46].
Limiting the treatment to second order nonlinearities, the Hamiltonian is given by:
Z
(2)
ĤNL (t) = d3 r P̂N L Ê(r, t)
V
0 X
Z
(2)
(1.29)
= d3 r dijk (r)Êi (r, t)Êj (r, t)Êk (r, t)
3 i,j,k
In our case, the phase-matching we are interesting in is the type II phase-matching (this will be
justified later) in which the pump beam is T E polarized and the two created idler and signal photons
have orthogonal polarizations. As the electric fields can be separated in one term Ê(+) (r, t) which
annihilates a photon and in its hermitian conjugate term Ê(−) (r, t) which creates a photon, the only
terms of eq. 1.29 that satisfy energy conservation are those in which one pump photon is annihilated,
and one signal and one idler photon are created. Moreover, a strong, undepleted pump field can
be modeled as a classical field, allowing us to drop all quantum mechanical operators acting on the
pump state, and to treat it like a classical, static object. The nonlinear Hamiltonian then reduces to
Z
(−) (−)
ĤNL (t) = 0 d3 r deff E(+)
p (r, t)Ês,T E (r, t)Êi,T M (r, t) + h.c.
ZV (1.30)
(−) (−)
+ 0 d3 r deff E(+)
p (r, t)Ês,T M (r, t)Êi,T E (r, t) + h.c.
V
3
To simplify the notation, vector quantities will now be indicated in bold rather than by an upper arrow.
with Z
E(+)
p (r, t) = Ep (x, y) dωp ei(kp z−ωp t) α(ωp ) (1.31)
where α(ωp ) is the spectrum of the pump field. Eq. 1.30 shows that two interactions can occur in
the device: one in which the signal photon is T E polarized and the idler photon T M polarized and
the other with exchanged polarizations.
Finally, as the propagation of the three waves is collinear, we can perform the z-integration over
the waveguide length L, which yields
Z L/2
L∆k i L∆k
dz ei∆kz = L sinc ( )e 2 = φPM (ωp , ωs , ωi ) (1.32)
−L/2 2
where we define φPM (ωp , ωs , ωi ) as the phase-matching function. This leads to the following expression
of the Hamiltonian:
L X
ZZ
(−)
ĤNL (t) =0 dxdy deff Ep (x, y)Ê(−)
s,σ (x, y)Êi,σ⊥ (x, y)
2π σ
ZZZ
e−i∆ωt (1.33)
× dωp dωi dωs q α(ωp )φPM (ωp , ωs , ωi )âs† âi† + h.c.
vg (ωi )vg (ωs )
with ∆ω = ωp − ωs − ωi .
Time evolution of the two-photon state: The evolution of the state of the system is derived by
solving the Schrödinger equation
1
Rt
dτ ĤNL (τ )
|Ψ(t)i = T̂ e i~ t0 |Ψ(t0 )i (1.34)
where T̂ is the time-ordering operator4 , which can be neglected for weak interactions that can be
treated perturbatively, as in our case. When a steady state is reached and the interaction is over
(we detect the state outside the source), the system does not have a memory of the initial and final
times, and so we can set t0 = −∞ and t = ∞ in eq. 1.34. Moreover, in a weak pump energy regime,
the output state is given by the first order approximation of eq. 1.34:
1 ∞
Z
|Ψ(t)i = 1 + dτ ĤNL (τ ) |0s , 0i i
i~ −∞
= |0s , 0i i + β |ψi (1.35)
with the vacuum |0s , 0i i taken as initial state, β the generation probability amplitude, and |ψi the
two-photon state.
The only time-dependentZ terms in ĤNL are the exponents containing the frequency mismatches.
This readily reduces to dτ ei∆ωτ = 2πδ(∆ω), corresponding to energy conservation. Hence it is
possible to perform the integration over the pump frequency ωp and find ωp → ωs +ωi , and so remove
the integral over ωp from the equation of the state:
X ZZ
|ψi = dωs dωi Ax,y φσ (ωs , ωi )âσ† (ωs )âσ†⊥ (ωi ) |0s , 0i i + h.c. (1.36)
σ
with
0 L
ZZ
(−)
Ax,y = q dxdy deff Ep (x, y)Ê(−)
s,σ (x, y)Êi,σ⊥ (x, y) (1.37)
iβ~ vgσ (ωs )vgσ⊥ (ωi )
4
The time-ordering operator takes into account pairs that have been already produced. These pairs stimulate then
the production of new pairs with identical characteristics (it is this phenomenon which brings SPDC into DFG).
and
φσ (ωs , ωi ) = φPM (ωs , ωi ) × α(ωs + ωi ) (1.38)
the Joint Spectral Amplitude (JSA) or biphoton wavefunction, which contain complete informa-
tion on the spectral-temporal properties of the processes. The JSA function is the product of
the spectrum of the pump field α(ωs + ωi ), which reflects energy conservation, and the phase-
matching function φPM (ωs , ωi ); it represents the joint probability amplitude of having generated
a signal photon at frequency ωs and an idler photon at frequency ωi . It is assumed to be normalized
dωi dωs |φ(ωi , ωs )|2 = 1.
RR
1.1.4 Phase-matching strategies in AlGaAs

As has been noted in previous sections, phase-matching is a real issue for an efficient frequency
conversion process. Moreover, all conversion processes must satisfy energy conservation. Those two
conditions can be written :
(
k~3 = k~2 + k~1 (1.39a)
ω3 = ω1 + ω2 . (1.39b)
Then, in dispersive materials where
ωi
ki = n(ωi ), (1.40)
c
with the refractive index n(ω) being a monotonic (increasing) function of ω, perfect phase matching
would appear in principle to be impossible. The birefringence of nonlinear crystals (for which the
refractive index depends on the polarization) was first used to get around this difficulty [52]. But
AlGaAs is an isotropic material and so not birefringent. Hence alternative strategies have to be
implemented to achieve phase-matching: those are presented in the following sections.
Quasi-phase-matching (QPM)
This method, which does not lead to exact phase-matching, consists of a periodic modulation of
the effective nonlinear coefficient of the medium (figure 1.5.a) by reversing its sign with a period of
length 2π/ |∆k|. This compensates for the phase-mismatches accumulated in adjacent sections of the
nonlinear material and leads to a constructive interaction between the different fields along the entire
crystal length L. This technique, first proposed in 1962 [53], has been used successfully in PPLN bulk
crystal [54] and waveguides [55] or PPKTP bulk crystals and waveguides [56] with a high conversion
efficiency and low losses. In AlGaAs, however, even though it is possible to achieve QPM by doing
epitaxial regrowths, this is technologically challenging. The main fabrication techniques consist in
periodically rotating the orientation of the crystal by 90◦ about the [001] crystal axes either by wafer
+
- + -
- + - + - +
(a) (b)
Figure 1.5 – Sketch representing the principle of quasi-phase-matching in (a) waveguide and (b) ring
resonator.
bonding or by orientation-patterned regrowth. Although it has been used to demonstrate an optical

parametric oscillator (OPO) [57], it still induces significant optical losses [58], which hinders it use
for quantum information.
It is also possible to achieve quasi-phase-matching using whispering gallery modes (as proposed
by Dumeige et al. in 2006 [59]). The advantage in using a ring resonator is that the periodic inversion
of the nonlinear coefficient comes naturally with the shape of the structure (figure 1.5.b). Hence there
are no optical losses induced by barriers between domains. Second-harmonic generation has been
achieved with this method in GaAs [60] and AlGaAs [61].
Form birefringence
Another technique to achieve phase-matching involves inducing an artificial birefringence in the

medium by inserting small aluminum oxide (AlOx) layers (refractive index ∼ 1.6) into the AlGaAs
waveguide (refractive index ∼ 3.2) (figure 1.6). The existence of repeated discontinuities in refractive
index along one direction breaks the original three axes symmetry of the constituent media, and
results in a linear birefringent material. In practice, AlOx is obtained by selective oxidation of
AlGaAs layers with a high aluminum content [62]. This technique allows very efficient conversions,
and has enabled the demonstration of DFG in mid-infrared (IR) [63], of SHG in near IR [64], and
of an OPO in near IR [65]; but it also induces roughness and absorption, leading to optical losses,
which are detrimental for quantum optics applications. Moreover, the insulating nature of the AlOx
layers hinders the possibility of electrically pumping the device.
Figure 1.6 – Sketch representing the principle of form birefringence phase-matching.
Counter-propagating phase-matching
This phase-matching technique is based in transverse pump geometry: one beam (corresponding to
the high frequency) impinges on the top of a waveguide, while the two other fields (corresponding
to the low frequencies) are propagating in opposite directions into the waveguide (figure 1.7.a). The
phase-matching condition in the epitaxial direction is satisfied by alternating AlGaAs layers with
different aluminum concentrations (having nonlinear coefficients as different as possible) to implement
a quasi-phase-matching scheme. The phase-matching condition can be projected (figure 1.7.b) along
the epitaxial direction (eq. 1.41a) and on the waveguide direction (eq. 1.41b). We observe that the
angle θ, with which the pump impinges onto the waveguide, provides a new degree of freedom with
which to tune the frequencies of the emitted fields:
(
k3 cos θ = kQPM (1.41a)
ω3 sin θ = ω1 n1 (ω1 ) − ω2 n2 (ω2 ) . (1.41b)
This scheme has been demonstrated for the first time in our group [66].
1.2. INTEGRATING THE PUMP LASER IN A NONLINEAR WAVEGUIDE 17
a) b)
Figure 1.7 – Sketch of the counter-propagating phase-matching scheme.
Modal PM
In modal phase-matching, the chromatic dispersion of the medium is compensated by multimode
waveguide dispersion (figure 1.8). This is the technique we use in this thesis. The interacting modes
can either be guided by total internal reflection [67, 68] or by photonic band gap effect [69, 70].
This second approach has demonstrated to be more interesting since on the one hand it avoids aging
problems of the device via the reduction of the total aluminum content, and on the other hand modes
that are confined by Bragg reflectors can have an effective index much lower than the material indexes
of the waveguide constituents giving more flexibility for device engineering.
Index
@ 1550 nm
eff
@ 1550 nm
@ 780 nm
a) b) Freq.
Figure 1.8 – Sketch representing the modal phase-matching scheme.
1.2 Integrating the pump laser in a nonlinear waveguide

1.2.1 Semiconductors lasers
A laser is a device that emits light through a process of optical amplification based on the stimulated
emission of electromagnetic radiation (laser is an acronym for Light Amplification by Stimulated
Emission of Radiation). It differs from other sources of light in that it emits light coherently and
monochromatically. The first laser was built in 1960 by Maiman [44], based on theoretical work by
Townes and Schawlow [71]. A laser is made up of three main components:
• a gain medium, to amplify optical waves by stimulated emission;
• a pumping system, to create a population inversion in the gain medium;
• a resonant cavity, to provide an optical feedback (this selects the emitted wavelengths and
boost amplification).
Current injection
P-doped
Emitted light
Gain medium
N-doped
Cleaved facets
Figure 1.9 – Scheme of a laser diode. Figure 1.10 – Energy diagram of the quan-
tum well and the intrinsic region under forward
bias current. CB (V B): conduction (valence)
band, the filled (empty) dots represent electrons
(holes); Eg : gap energy; dQW : quantum well
width; hν: photon energy.
There exists several types of laser: gas lasers, chemical lasers, dye lasers, metal-vapor lasers, solid-
state lasers, semiconductor lasers, ...
Our work is focused on a laser diode, which is an electrically pumped semiconductor laser in
which the gain medium is a P-I-N junction. It was first demonstrated in 1962 by two groups leb by
Hall at General Electrics [72] and Nathan at IBM [73]. Diode lasers have numerous advantages: they
are very compact, they cover a large spectral band from the visible to near infra-red (IR) according
to the chosen semiconductor material, they are easy to fabricate and low cost thanks to the progress
done in semiconductor material science.
The active region of a laser diode is the intrinsic (I) region (in red in figure 1.10). When the
diode is polarized, the carriers (electrons and holes) are injected into that region from the N and P
regions respectively. The Fermi levels on the two sides of the junction are deplaced. When the bias
is high enough injection mechanism leads to a population inversion. The goal for a laser diode being
to recombine all carriers in the I region and produce light, they are fabricated using direct bandgap
semiconductors, for which spontaneous and stimulated emission are vastly more efficient [74].
The addition, in the active region, of a thin layer with a smaller bandgap energy creates a quantum
well, leading to energy quantization. The laser efficiency is then increased due to carriers confinement
in the quantum well, which induces a stepwise form of its density of energy states, and hence lower
the laser threshold. The wavelength of the light emitted is determined by the width of the well, as
well as by the bandgap of the material composing it, according to the following equation:
c
h = Eg + ∆Ec + ∆Ev (1.42)
λ
with Eg the energy gap of the well material, and ∆Ec (∆Ev ) the confinement energy of electrons
(holes) in the quantum well.
The quantum well thickness is of the order of 10 nm and for this reason it cannot represent a
dielectric guide for the laser mode. Hence two layers, having a refractive index superior to the N and
P regions, are added around the quantum well. These layers act as an optical vertical confinement
area for the created photons. This is called a separate confinement heterostructure scheme.
The cavity arises from the cleaving along the crystal planes of the end-facets, which act as
reflectors (due to the sharp refractive index difference at the semiconductor/air interface of about 3.4
1.2. INTEGRATING THE PUMP LASER IN A NONLINEAR WAVEGUIDE 19
to 1). This forms a Fabry-Perot resonator. The photons emitted by the active region are amplified
by stimulated emission and reflected at the end facets (with a probability linked to their modal
reflectivity). The laser oscillation emerges when the gain in the amplifying medium exceeds the total
cavity losses:
1 nth 1 1
gthreshold = × exp(− ) × (α + ln( )) (1.43)
Γ ntr 2L R1 R2
with α the waveguide losses arising from free carrier absorption, scattering and all parasitic losses of
the optical mode, L the cavity length, R1 and R2 the mirror reflectances, Γ the overlap representing
the integral of the intensity of the optical mode over the gain region, and nth and ntr the threshold and
transparency densities of the electrons in the quantum well respectively. This leads to a transparency
current density
ntr
Jtr = q , (1.44)
τ
with τ the life-time of the electrons in the conduction band of the quantum well [75].
1.2.2 Working principle of our device

The direct bandgap of AlGaAs is obviously of interest for the realization of electrically injected
devices. The aim of this thesis is to merge the pumping laser and the nonlinear material to create
an ultra-compact photon pair source.
To achieve such a device, we used the structure sketched in figure 1.11. The active part of the
structure is a PIN junction, in which a quantum well allows laser emission in the pump mode around
780 nm by electrical injection. Because of the selection rules, the emission from the quantum well is
T E polarized.
Figure 1.11 – Sketch of our electrically injected photon pair source.
In order to achieve SPDC in the same structure, we need to fulfill the following conservation
conditions5 :
p

ωp nT E (ωp ) = ωs nT M (ωs ) + ωi nT E (ωi ) (1.45a)


p
ωp nT E (ωp ) = ωs nT E (ωs ) + ωi nT M (ωi ) (1.45b)


ω = ω + ω . (1.45c)
p s i
5
As the pump has T E polarization, we are only interested in type II phase-matching, in which a T E polarized pump
photon generates a T E polarized photon and a T M polarized photon (T E → T E + T M ).
Among the different phase-matching strategies listed in section 1.1.5, the most promising one to
integrate monolithically the laser source and the nonlinear medium into a single device is modal phase
matching, in which the phase velocity mismatch is compensated by multimode waveguide dispersion
(figure 1.12.a). This led our team to engineer the following structure (this work has been performed
by the previous PhD student, A. Orieux): two Bragg mirrors, surrounding a core layer, provide
both photonic band gap vertical confinement for the laser mode at 780 nm (in a T E Bragg mode),
and total internal reflection cladding for the photon pair modes at telecom (1.56 µm) wavelength
(in fundamental modes with orthogonal polarization). Figure 1.12.b presents the simulated spatial
profile of the three interacting modes [76]. Lateral confinement is ensured by the etching of the
waveguide.
Pump TE SPDC TE
@1.56µm
@780nm
SPDC TM
@1.56µm
(a) Refractive index engineering. (b) Simulation of the three-interacting modes
Figure 1.12 – Modal phase-matching in the monolithic photon pairs source.
The production of orthogonally polarized paired photons is a good focus for quantum information
applications since the photons can be easily separated and manipulated through polarization. The
simulated tuning curves for the photons emitted by our device (based on [77, 78]), solutions of the
equations system 1.45, are given in figure 1.13. SPDC is possible if the laser wavelength is below the
degeneracy point. Due to the strong dispersion of the Bragg mode arising from the proximity to the
energy band gap of the waveguide core, small shifts of the laser wavelength from degeneracy produce
a large wavelength separation between the generated photons.
Figure 1.13 – Simulated tuning curves for our device.

1.3. CHARACTERIZATION OF THE QUANTUM STATE OF PHOTON PAIRS 21
1.3 Characterization of the quantum state of photon pairs

1.3.1 Indistinguishability
Two-photon interference was first demonstrated experimentally by Hong, Ou and Mandel in 1987
[79], and theoretically by Fearn and Loudon [80] the same year. It is a phenomenon that cannot
be understood in terms of classical physics, but can only be explained using quantum physics; it
shows the quantum nature of light. Two-photon interference involves a pair of photons entering
a lossless beam splitter from two separate input ports. Hence this is hardly an interferometer in
the traditional sense, i.e. where waves are first split and then recombined. Nevertheless, there is
indeed an interference effect, which has been widely used as a criterion for testing single-photon and
two-photon sources, due to its extremely simple geometry [81].
Figure 1.14 – Two possible outputs for one photon impinging on a BS.
When a photon enters a beam splitter (BS) there are two possibilities; it will either be transmitted
or reflected (as shown on figure 1.14).
coinc coinc
no no
coinc coinc
Figure 1.15 – Four possible outputs for two photons impinging on a BS.
Consider two photons, one in each input of a lossless 50/50 BS. There are four possible ways of
equal probability in which the photons can behave: both photons are transmitted (figure 1.15.a); both
are reflected (figure 1.15.b); one is transmitted while the other is reflected (figures 1.15.c and 1.15.d).
If the two impinging photons are indistinguishable, which means all their physical properties are
alike, then the first two possibilities (figures 1.15.a and .b) are indistinguishable. (There will be a
photon at each output of the BS, with no way of distinguishing which path each has taken.) This
results in amplitude addition or interference. However, because of conservation energy on the BS,
there is an overall π-phase difference for the two photons between these two possibilities6 . This phase
shift is independent of the specifics of the BS. This leads to destructive interference, and the two
possibilities completely cancel each other when their amplitudes are the same. In conclusion the two
photons will always exit the same output port, in a random manner (possibilities of figures 1.15.c
and .d) [81].
In the following, the two input ports of the BS will be labeled A and B, and the two output
ports C and D. The initial state, where a photon arrives at both input ports of the BS, is given by
†
applying the photon-creation operators âA,B on the vacuum state |0A , 0B i in the (A, B) basis
† †
|Ψiin = |1A , 1B i = âA âB |0A , 0B i . (1.46)
The effect of a BS on an electric field can be described by the unitary transformation U

! √ √ ! !
âA T − R â
→ √ √ C , (1.47)
âB R T âD
with â{A,B,C,D} the photon-annihilation operators corresponding to the different ports of the BS and
where T and R are the transmissivity and reflectivity of the BS respectively. For a lossless BS these
coefficients also satisfy
(
T +R=1 (1.48a)
√ √
T R∗ + RT ∗ = 0 . (1.48b)
Then the output state in the (C, D) basis will be
|Ψiout = = U |Ψiin
√ † √ † √ † √ †
= ( T âC − RâD )( RâC + T âD ) |0C , 0D i
√ †2 †2 † † † †
(1.49)
= [ 2T R(âC − âD ) + T âC âD − RâD âC ] |0C , 0D i
√
= 2T R(|2C , 0D i − |0C , 2D i) + (T − R) |1C , 1D i .
† † † †
Since the operators âC and âD act on different output ports their commutator vanishes: [âC , âD ] = 0.
It follows that for a 50/50 beam splitter with T = R = 1/2, the second term is zero; there is
destructive interference of the corresponding two-photon amplitudes. The output state is then
1
|Ψiout = √ (|2C , 0D i − |0C , 2D i) . (1.50)
2
The probability of having a coincidence between two detectors, one at each output, is zero, and the
probability of having both photons in one detector, either C and D, is 1/2.
The observation of this phenomenon allows to measure both the degree of indistinguishability
between two photons and their wave-packet size δt. The size of the photon wave-packet created
6
A BS is composed of two media with different refractive indexes: a reflection off the higher index medium
results in a relative phase shift of π, and a reflection off the lower index medium results in a null phase shift.
by parametric conversion is of the order of picoseconds, while the time resolution of single-photon
detectors is of the order of hundreds of picoseconds. Hence it is impossible to measure it directly with
detectors. In the two-photon interference experiment, the time resolution is much better because it
depends only on the difference in the arrival times of the photons on the BS, which is controlled
mechanically. When the two wave-packets perfectly overlap, the two photons leave the BS by the
same output port, and there is no simultaneous detection at the two output detectors; when the two
wave-packets do not overlap, the two photons leave the BS by the two output ports with probability
1/2, leading to simultaneous detection. Then the number of coincidences as a function of the time
difference shows a dip for δτ = 0 (figure 1.16). The full width at half maximum (FWHM) of the dip
is determined by the size of the photon wavepacket or coherence time of the down-converted photons
(reciprocal of the phase-matching spectral bandwidth.). The visibility (depth) of the dip gives the
degree of indistinguishability between the two photons.
Coincidence probability
δτ 0.5
δt
δt
0
0
a) b) Relative delay between photons
Figure 1.16 – a) Overlap of the two wave-packets of size δt with respect to the arrival time difference
δτ . b) Example of the dip for the HOM experiment. The FWHM gives the photon wavepacket size,
and the visibility gives the degree of indistinghuishability.
1.3.2 Quantum bits
In information science, the basic unit to encode information is the bit, which is a two-state system
that can have only two values ”0” and ”1”. The quantum mechanical analog of a classical bit, the
Qubit, is also a two-state system in which the states are indicated with the kets |0i and |1i. Contrary
to classical systems, quantum mechanics allows the qubit to be in a coherent superposition state:
|Ψi = α |0i + β |1i (1.51)
with |α|2 + |β|2 = 1. This means that the qubit is in a superposition of both states and, if we measure
the qubit, we will find it with probability |α|2 to carry the value “0” and with probability |β|2 to
carry the value “1”.
The two orthogonal states of the system |0i and |1i form a complete basis, referred to as the
computational basis, for the 2D-Hilbert space of the qubit. Any other basis can be expressed by
linear combinations of the computational basis, and all states of the qubit can be expressed in the
computational basis as
|Ψi = cos θ |0i + eiφ sin θ |1i (1.52)
where the angles θ and φ define a point on the unit sphere in three dimensional space. The state of
the qubit can therefore be visualized as a vector pointing from the origin to the unit sphere, which
is referred to as the Bloch sphere, as shown in figure 1.17.
Figure 1.17 – The Bloch sphere.
1.3.3 Entanglement
1.3.3.1 Definition
Quantum entanglement corresponds to a system of particles in which the properties of the whole
system are defined, while those of the individual particles are not, whatever the separation between
the particles. It can then be seen as the generalization of the superposition principle to multi-particle
systems.
If we consider two non-interacting systems A and B, with respective Hilbert spaces HA and HB ,
then the Hilbert space of the composite system is the tensor product HAB = HA ⊗ HB . Taking the
basis {|iiA } for HA and {|jiB } for HB , the most general pure state in HAB is of the form
X
|ψiAB = cij |iiA ⊗ |jiB . (1.53)
i,j
This state is called separable if there exist vectors {cA B A B

i , cj } such that cij = ci cj , yielding
X X
|ψiA = cA
i |iiA , |φiB = cB
j |jiB and |ψiAB = |ψiA ⊗ |ψiB . (1.54)
i j
It is called inseparable if, for at least one pair of coordinates cA B A B

i , cj , there is cij 6= ci cj . Inseparable
states are referred to as entangled states, and |ψiAB cannot be written in a state product form:
|ψiAB 6= |ψiA ⊗ |ψiB . (1.55)
For example, for the 2-dimensional Hilbert spaces HA and HB , with basis vectors {|0iA , |1iA }
and {|0iB , |1iB } respectively, the following are entangled states:
1
|Ψ+ i = √ (|0iA ⊗ |1iB + |1iA ⊗ |0iB ) (1.56a)
2
1
|Ψ− i = √ (|0iA ⊗ |1iB − |1iA ⊗ |0iB ) (1.56b)
2
+ 1
|Φ i = √ (|0iA ⊗ |0iB + |1iA ⊗ |1iB ) (1.56c)
2
− 1
|Φ i = √ (|0iA ⊗ |0iB − |1iA ⊗ |1iB ) (1.56d)
2
Moreover, they form a complete orthonormal basis in the 4-dimension Hilbert space HAB = HA ⊗HB
and are called Bell states. If the system is in one of these pure states, it is impossible to attribute to
either system A or system B a definite pure state.
It is worth noting that there is no preferred basis for an entangled state, and perfect correlation
between the two systems will remain regardless the measurement basis. For example, a change of
basis by a rotation of θ across the horizontal equator of the Bloch sphere can be noted as
|0θ i = cos θ/2 |0i + sin θ/2 |1i

(1.57)
|1θ i = sin θ/2 |0i − cos θ/2 |1i .
Then for the |Φ+ i state

1
|Φ+ i = √ (|0θ iA |0θ iB + |1θ iA |1θ iB )
2
1
= √ {(cos θ/2 |0iA + sin θ/2 |1iA )(cos θ/2 |0iB + sin θ/2 |1iB )
2 (1.58)
+ (sin θ/2 |0iA − cos θ/2 |1iA )(sin θ/2 |0iB − cos θ/2 |1iB )}
1
= √ (|0iA |0iB + |1iA |1iB ) .
2
Entanglement is a central and disconcerting concept in quantum physics, and has been the cause
of much discussion. The term was introduced in 1935 by Schrödinger. There followed a long debate
concerning the interpretation and completeness of quantum mechanics, which began with the famous
article of Einstein, Podolsky and Rosen describing what is known as the EPR paradox [82]. They
considered such behavior to be impossible because it violated the local realist view of causality, and
they argued that the accepted formulation of quantum mechanics must therefore be incomplete and
should be completed with supplementary (or hidden) variables. On the other hand, Bohr strongly
opposed this conclusion, saying that completing the quantum theory would destroy its self-consistency
[83].
To illustrate these concepts I will take the example of two observers, Alice and Bob, measuring
each one a particle of the state |Ψ+ i (eq. 1.56a). Alice will measure the particle A and Bob the
particle B. Let’s say that Alice makes a measurement in the {|0i , |1i} eigenbasis of A; there are
two possible outcomes, occurring randomly but with equal probability, thanks to the superposition
principle:
• Alice measures 0 → the state of the system collapses to |0iA |1iB → Bob measures 1.
• Alice measures 1 → the state of the system collapses to |1iA |0iB → Bob measures 0.
The result for Alice is not known before measurement; she cannot decide which state the composite
system will collapse into, and therefore she cannot transmit information to Bob by acting on her
system. But, once this measurement is made, the result for Bob is immediately known with certainty.
System B has therefore been altered by Alice performing a local measurement on system A. This
remains true even if systems A and B are spatially separated, and so when no signal, even at the
speed of light, can be conveyed between the two measurement apparatuses to ensure the "proper"
outcome of the measurement on system B. This violates the principle of locality, which states that
distant objects cannot have direct influence on one another; this is the heart of the EPR paradox.
A possible resolution of the paradox, proposed by Einstein, might be to assume that the state of
the particles contains some local hidden variables, the values of which effectively determine what the
outcomes of the measurements on the two sides will be. This means that each particle carries all the
required information with it, and nothing needs to be transmitted from one particle to the other at
the time of measurement. In this vision, the accepted quantum mechanical description, with random
measurement outcomes, must be incomplete.
However, even if a hidden variable theory can explain the behavior of the particles in one chosen
basis, it fails to explain the results when the system is measured in a different basis than the com-
putational basis. Indeed, with a local hidden variable theory, the measurements in the second basis
should be uncorrelated, while quantum mechanics still predicts correlations.
In 1964 Bell proved that one of the key assumptions in the EPR paradox, the principle of locality,
was mathematically inconsistent with the predictions of quantum theory. He was able to show that
there is an upper limit to the strength of correlations that can be produced in any theory obeying local
realism. He then showed that quantum theory predicts violations of this limit for certain entangled
systems [84]. Bell thus shifted the epistemological debate to experimental physics. Later Clauser
et al. proposed a practical implementation of Bell’s idea [85]; today Bell-type experiments are at the
heart of many experiments both in fundamental quantum mechanics and in quantum information
protocols.
A Bell-type experiment consists of a source S emitting pairs of correlated particles (see fig-
ure 1.18). Each particle enters an apparatus (a or b) that analyses the correlation feature and
ascribes a binary value (0 or 1) to the outcome. The operation of each device sets the parameter.
Then classical information about the detection of the particles (when and where they are detected) is
sent to coincidence electronics equipment that measures the number of time-correlated events. This
enable one to calculate the correlation coefficient
E(a, b) = P (1A , 1B | a, b, λ) + P (0A , 0B | a, b, λ) − P (1A , 0B | a, b, λ) − P (0A , 1B | a, b, λ) , (1.59)
where P (iA , jB | a, b, λ) is the probability of finding measurement i in system A and j in system B,

knowing a, b, and λ.
Figure 1.18 – Scheme of a Bell-type experiment.
Bell’s inequalities
In the following, we derive Bell’s inequalities in the most frequently used form, known as Clauser-
Horne-Shimony-Holt (CHSH) Bell’s inequalities. Assuming a local hidden variable theory, when
systems A and B are separated by a large (space-like) distance, the results for A can depend on a
but not on b, and similarly the results for B do not depend on a. The local hidden variable λ can
denote either a single or a set of parameters, or a set of functions; these variables can be discrete or
continuous. It is shared by both particles from their creation, and carries information on the result
of the measurement. The result on system A is then determined by a and λ, and that on system B
is determined by b and λ.
R
Noting ρ(λ) the normalized probability distribution of λ ( dλ ρ(λ) = 1), the expectation value
of the shared probability is
Z
P (iA , jB | a, b, λ) = dλ ρ(λ)P (iA | a, λ)P (jB | b, λ) . (1.60)
Then the correlation coefficient becomes

Z
E(a, b) = dλ ρ(λ) P (0A | a, λ)P (0B | b, λ) + P (1A | a, λ)P (1B | b, λ)

− P (1A | a, λ)P (0B | b, λ) − P (0A | a, λ)P (1B | b, λ)
Z (1.61)

= dλ ρ(λ) P (0A | a, λ) − P (1A | a, λ) P (0B | b, λ) − P (1B | b, λ)
| {z } | {z }
A(a,λ) B(a,λ)
with
| A(a, λ) |6 1, and | B(b, λ) |6 1 . (1.62)
Now we consider two different bases for the apparatus, {a, b} and {a0 , b0 }. We can hence calculate
Z
0
| E(a, b) ± E(a, b ) | =| dλ ρ(λ) A(a, λ)B(b, λ) ± A(a, λ)B(b0 , λ) |
Z
6 dλ ρ(λ) | A(a, λ) || B(b, λ) ± B(b0 , λ) | (1.63)
Z
6 dλ ρ(λ) | B(b, λ) ± B(b0 , λ) |
and in a same way

Z
0 0 0
| E(a , b) ∓ E(a , b ) |6 dλρ(λ) | B(b, λ) ∓ B(b0 , λ) | . (1.64)
Since | X |6 1 and | Y |6 1 imply | X ± Y | + | X ∓ Y |6 2, then for all local hidden variable models,
the correlation operator S has to satisfy the inequality:
S =| E(a, b) ± E(a, b0 ) | + | E(a0 , b) ∓ E(a0 , b0 ) |

Z
6 dλ ρ(λ) | B(b, λ) ± B(b0 , λ) | + | B(b, λ) ∓ B(b0 , λ) | (1.65)
Z
6 dλ ρ(λ) × 2 = 2 .
Predictions of quantum mechanics

From the measurement theory in quantum mechanics and, in particular, the projection postulate, we
can calculate
1 θa − θb
P (iA , iB | a, b) = cos2 ( ), (1.66a)
2 2
1 θa − θb
P (iA , jB | a, b) = sin2 ( ). (1.66b)
2 2
Hence the correlation coefficient becomes
E(a, b) = P (1A , 1B | a, b) + P (0A , 0B | a, b) − P (0A , 1B | a, b) − P (1A , 0B | a, b)

(1.67)
= cos(θa − θb )
and the Bell correlation operator
S =| E(a, b) ± E(a, b0 ) | + | E(a0 , b) ∓ E(a0 , b0 ) |

(1.68)
=| cos(θa − θb ) ± cos(θa − θb0 ) | + | cos(θa0 − θb ) ∓ cos(θa0 − θb0 ) | .
This last equation shows that, for certain settings, the Bell’s inequality (eq. 1.65) can be violated.
The largest violation corresponds to the choice of angles θa = 0, θa0 = π2 , θb = π4 , and θb0 = − π4 . For
√
these angles S = 2 2 > 2. We can therefore conclude that quantum mechanics disagrees with all
local hidden variable models, and is non-local.
Bell's limit
a) b)
Figure 1.19 – a) Angles corresponding to the highest violation of Bell’s inequality. b) Bell correlation
operator as function of the angle θ = (θa , θb ).
1.3.3.2 The experiments: the end of the story?

Since the 1980s, numerous experimental tests of the so-called Bell’s inequalities have been performed
[86, 87, 88, 89, 90, 91], all validating quantum theory without any need to add local hidden variables.
But all these tests have had at least one “loophole”, this allowing a local realist interpretation of the
experimental results unless one made a supplementary (albeit reasonable) hypothesis. Recently three
groups have independently closed the two main loopholes (the locality loophole and the detection
loophole); this conclusively is a demonstration that nature is indeed nonlocal [92, 93, 94].
Besides its interest to prove the foundations of quantum mechanics, entanglement lies at the heart
of the very active field of quantum information, which objective is to process and transport informa-
tion much more efficiently than can be achieved using classical physics. One of the most promising
results of this new field is quantum key distribution (also referred to as quantum cryptography),
which allows the creation of a secret key between two parties that can be used to encode and decode
a message [2, 95]. We will discuss this application in the last section. Other examples of quantum
information processing based on entanglement are dense coding [96, 97], which is the possibility of
sending more than one bit of classical information encoded in a single quantum bit, and teleportation
[17, 98], which is the transmission of an arbitrary quantum state from one particle to another.
1.3.3.3 Creating entangled states.

Quantum systems can become entangled through various types of interactions and observables, and
can involve different physical supports. The choice of preferred support and/or observable depends
strongly on the desired application: some are more adapted to quantum information processing,
in which the physical supports are required to interact with each other (superconducting systems
[99], cold ions and atoms [100, 101], color centers [102], ...); others are more adapted for quantum
communication and cryptography, in which long distance distribution of the entanglement is desired.
In this latter case, photonics has a big advantage over other systems; photons have a high speed,
they are (almost) decoherence-free in free space and optical fibers, and they do not interact with
each other.
In this thesis, we focus on photonics. Many observables of photons are suitable for information
coding. In the following, we describe two of the most widely used strategies to encode information
in a photon that are polarization and energy-time entanglement. We just mention that there are
various other types of photonics entanglement that are also possible such as time-bin (a pulsed
variant of energy-time)[103], frequency-bin [104], and orbital angular momentum [105]. There is also
the possibility of encoding multiple qubits onto one photon by combining several degrees of freedom
[106].
Entanglement in polarization
One of the most widely used photonic entanglement observables is polarization; this is because of its
facility in use. In this case, three conjugate bases are defined as
• {|Hi , |V i} for the horizontally and vertically polarized photons;

√ √
• {|Di = 1/ 2(|Hi + |V i), |Ai = 1/ 2(|Hi − |V i)} for the diagonal basis;
√ √
• {|Ri = 1/ 2(|Hi + i |V i), |Li = 1/ 2(|Hi − i |V i)} for the basis of circular right and circular
left polarized photons.
Polarization encoded photonic qubits are conveniently manipulated using birefringent optics,
such as half-waveplates and quarter-waveplates. This allows the generation of any state of the
form cos θ/2 |Hi + eiφ sin θ/2 |V i, with (θ, φ) ∈ R. The polarization state is then analyzed using a
polarizing beam-splitter (PBS), which projects polarization on the state H or V , followed by single
photon detectors at each of its outputs.
A drawback of polarization is that long distance distribution of entanglement requires some
precautions because the photon polarization state is rotated by propagating in optical fibers, and
because the birefringence of materials can induce systematic delay between the two polarization
states.
PBS WP WP PBS
Figure 1.20 – Scheme of the apparatus to measure polarization entanglement.
Entanglement in energy-time
This type of entanglement is based on a theoretical proposal by Franson [107] in 1989; it was first
demonstrated by Ou et al. [108] and the following year by Kwiat et al. [109].
As we have seen, parametric down-conversion in a medium is ruled by energy and momentum
conservation (eq. 1.45). Moreover, the emission of the signal and idler photons is quasi-instantaneous:
when the first photon is emitted, the creation of the second photon has to follow immediately in order
not to violate energy conservation. Energy-time entanglement exploits the simultaneity of the signal
and idler emission due to energy conservation.
In the Franson scheme, a parametric source is pumped in continuous wave (cw) regime to produce
photon pairs. The creation of the entangled state is done by the implementation of unbalanced
interferometers on each of the correlated photon paths (see figure 1.21). The interferometer path
length has to be chosen such that the implemented time delay between the two interferometer arms
∆T is much larger than the correlated photons coherence time τc to avoid single photon interference;
in this way measured interference pattern can only be explained as two-photon interference. To carry
out the analysis of the SPDC biphoton state, Alice and Bob are provided with the independent relative
phase adjustments φA and φB the photons acquire in their respective interferometers. Another
condition is that the indistinghuishability in creation time τp of both SPDC photons (i.e. the coherence
time of the pair) is much higher than the time delay ∆T introduced by the interferometers, in order
to obtain interferences between the two different paths.
Figure 1.21 – Scheme of the energy-time entanglement.
Each photon of the photon pair chooses randomly the long (l) or short (s) arm of the interfer-
ometer on its path. Measuring the coincidence in the arrival times of the photons to the detectors,
we obtain that photons arrive simultaneously at three possible detection times (−∆T, 0, ∆T ). This
correspond to the projection of the biphoton state into one of the following states: |siA |liB , |siA |siB ,
|liA |liB , |liA |siB , as can be seen on figure 1.22. The two states |siA |liB and |liA |siB appear as
two satellite peaks provided that the single photon detectors have a temporal resolution τdet high
enough to resolve ∆T . These peaks show no interference effects. On the contrary, the two biphoton
states |siA |siB and |liA |liB are indistinguishable in their detection time as long as the path length
difference in both interferometers is kept equal up to the coherence length lc of the SPDC photons.
Thus, by counting only the events of the central peak (with time delay ∆t = 0), an entangled state
is post-selected; pumping the source in cw regime hinders the possibility to distinguish between the
situation in which both photons take the short arm, and both photons take the long arm. Then the
two paths remain indistinguishable in the post-selected state.
Coincidences
Figure 1.22 – Distribution of coincidence peaks in a Franson setup. The satellite peaks correspond to
the situation in which the photons take the long-short (left) and short-long (right) path respectively.
The two indistinguishable possibilities, short-short and long-long, lead to coincidences within the
central peak.
As a result, the entangled state can be written as the coherent superposition of the two biphoton
states, and hence as a function of the relative phase φ = φA − φB each correlated photon acquires in
its respective interferometer:
1
|ΨiAB = √ (|siA |siB + eiφ |liA |liB ) . (1.69)
2
1.4. QUANTUM KEY DISTRIBUTION 31
1.4 Quantum key distribution

In quantum key distribution (QKD) the goal is to establish a secret key between two parties (Alice
and Bob) that can be used to encode and decode a message. The inviolability of the key is bound
to the laws of quantum mechanics rather than to difficult operations, in the sense of algorithm
complexity, carried out by classical computers, as it is the case in today’s classical cryptography.
Basic model for unconditionally secure cryptography

A cryptographic system is unconditionally secure when an eavesdropper (Eve) cannot obtain enough
information from the encrypted message to reconstruct the original message, even if she has infinite
computational and time resources. Figure 1.23 shows the basic model for unconditionally secure
cryptography. The sender of the message (Alice) communicates with the receiver (Bob) over a public
channel that potentially can be eavesdropped. To ensure secrecy of the communication, a secret key
has to be shared by Alice and Bob, and used to encrypt and decrypt the message.
Public channel
Message Encrypter Decrypter Message
Secure channel
Key
Alice Bob
Figure 1.23 – Schematic for unconditionally secure cryptography.
One unconditional security algorithm is the Vernam cipher7 , in which the key has to be as long
as the message, and is discarded after each transmission. This algorithm has not been used in
cryptographic systems because of the key distribution problem; it requires a trusted courier for the
exchange of a new key after each transmission. Quantum physics provides a solution to this key
distribution problem. QKD algorithms allow the exchange of secret keys between Alice and Bob
without requiring a trusted courier; security is guaranteed by the laws of quantum mechanics, these
ensuring that the key can be used afterwards to encrypt and decrypt messages as a one-time pad for
unconditionally secure cryptography.
General quantum key distribution algorithm

The first proposal for quantum cryptography was made in the 1970s by Wiesner [110], although he
published this a decade after. To formulate his theory, he considered the following fundamental rules
of quantum physics:
• The polarization of a photon cannot be measured in non-compatible bases (for example the
{|Hi , |V i} and {|Di , |Ai}) at the same time.
• Individual quantum processes cannot be distinctly described.
• It is not possible to take measurements of a quantum system without disturbing it.
• It is not possible to duplicate unknown quantum states without error.
The two last points imply that an eavesdropper (Eve) cannot eavesdrops on a key sent between Alice
and Bob without changing it, and therefore exposing herself. Furthermore, Eve cannot conceal her
presence by recreating her detected particles.
7
The Vernam cipher algorithm, also known as one-time pad, is a symmetrical stream cipher in which the message
in encoded with a logic XOR gate, in combination with an absolutely random and private key at least as long as the
message. Provided that every key is used only once, this encryption technique has been proven to be absolutely secure.
There exist a large number of methods to establish such a key, the first being proposed by Bennett
and Brassard [1] in 1984. In the following, we describe one protocol of QKD relying on quantum
entangled states, the BBM92 protocol. It offers secure communication through the generation of
absolutely secret and random keys to an eavesdropper (Eve).
BBM92 protocol
Proposed by Bennet, Brassard and Mermin [111] in 1992, this is the most commonly used entanglement-
based QKD scheme. It assumes that an entanglement source is placed somewhere between Alice and
Bob; this source does not need to be trusted, and can even be in Eve’s hands.
1. A source of entangled qubits emits repeatedly one of the entangled Bell’s states, and sends one
photon to Alice and the other to Bob via a quantum channel. This is a superposition of states;
it is impossible to know in which state each particle is until the measurement is made.
2. Measurement: Alice and Bob both receive the qubits, and perform a measurement √ in a random
basis,√either the computational one ({|0i , |1i}) or the diagonal one ({(|0i + |1i)/ 2, (− |0i +
|1i)/ 2}). The outcomes are labelled ‘0’ or ‘1’, and they form the raw key. As demonstrated in
section 1.5.1, the measurement in one or the other basis does not alter the correlations between
Alice and Bob. This automatically ensures total randomness in the measurement outcome of
the “first” qubit, which instantaneously projects the “second” qubit into the corresponding
correlated (or anticorrelated) state.
3. Basis reconciliation:
(a) Alice and Bob communicate the basis they used for each measurement via a public channel.
(b) They discard events obtained when they used incompatible bases, without actually reveal-
ing the outcomes of their measurements. This sifting process shrinks the raw key down to
50% of its original size, and into the sifted key.
If Alice and Bob chose different bases for analysis, the measurement outcomes are random and
no correlations are observed. On the other hand, if Alice and Bob chose the same basis, then
both measurement outcomes are perfectly correlated (or anti-correlated). Hence post-selecting
only the cases in which the same bases have been used allows Alice and Bob to establish
completely random, but identical (or perfectly opposed) data streams of 0s and 1s, which can
be used further as secret keys.
4. Eavesdropping detection:
(a) The individual qubits of the entangled quantum state are not defined until measured. This
permits to detect the presence of a spy in a relatively simple manner. A spy needs to per-
form some kind of measurement on the qubits, meaning that entanglement between Alice
and Bob is broken. This reduces the correlation strength and, consequently, introduces
errors between the recorded data streams. Therefore Alice and Bob compare a fraction
of their data streams; an error in the measurement outcomes reveals the presence of an
eavesdropper.
(b) If an eavesdropping is detected, Alice and Bob change the quantum channel until one is
found with absolute privacy.
For example, in a simple intercept-resend attack Eve intercepts a qubit (for example on Bob’s
side) and measures it in a random basis. (She does not know the bases Alice and Bob will
choose.) If she has chosen the same basis as Alice and Bob, her measurement leads to the right
result. However, if she choses the wrong basis, her measurements are wrong in half of the cases,
which means that the amount of information Eve can gain on the qubits is limited, such that
she will never obtain the full key. Moreover, to be as discrete as possible, Eve needs to send
back qubits to Bob, to avoid an interruption in the qubit stream. Eve prepares new qubits,
ideally identical to the measured ones, and sends them to Bob. But in the cases where she has
not chosen the same basis as Alice and Bob, she prepares qubits in this wrong basis, which will
introduces errors between the measurement outcomes for Alice and Bob (see table 1.1). In the
intercept-resend attack, the amount of error introduced by Eve is 25% of the sifted key. There
exist more sophisticated eavesdropping strategies permitting Eve to obtain more information
and/or introduce less errors between Alice and Bob, but it has been shown that this error rate
remains at least 7% [112, 113, 114]. Therefore, the comparison of a fraction of the data streams
of Alice and Bob showing an error rate less than 7% ensures the establishment of a secret key.
5. Alice and Bob can encode messages with the secret key (for example, with the Vernam cipher
algorithm) and transmit them over a public channel.
Alice + + × + × × × +
0 1 1 1 0 0 1 0
Eve × + × + + × + ×
1 1 1 1 0 0 1 1
Bob + × + + × × + ×
0 0 0 1 1 0 1 1
Table 1.1 – Intercept-resend attack. Only measurements in which Alice and Bob have chosen the
same basis, + or ×, are considered. In 50% of these cases, Eve chose the correct basis and is able
to return the correct qubit. In the other cases, she choses the wrong basis and her measurements
are uncorrelated. In half of these instances this results in an error in the comparison between the
measurements of Alice and Bob.
In practice, the BBM92 protocol requires the use of additional techniques in order to ensure the
secure establishment of a key, such as communication partner authentication, error correction and
privacy amplification [?]. These aspects are beyond the scope of this thesis. As we will see, the
objective of chapter 6 is to demonstrate how to construct a sifted key with our source and standard
telecom components.
In conclusion, quantum mechanics allows parties to transmit messages and to detect the presence
of an eavesdropper in every instance. This is because quantum principles state that the key being
transmitted does not actually exist until it is observed, and so it is hard to gain information about
it while it is traveling between users.
Chapter 2
Fabrication
This chapter describes the fabrication steps of our devices. After a brief introduction on the design
guidelines, I will describe the technique used for the growth and the processing steps I have developed
in the clean room of the University Paris Diderot.
2.1 Heterostructure design

The vertical heterostructure design has been set up during my predecessor’s theses, L. Lanco and A.
Orieux [115, 116]. We recall here the general principles. The objective is to achieve a diode laser
emitting photon pairs: the nonlinear waveguide and the pump source are integrated within the same
device, supplied by electrical injection.
The material choice was AlGaAs. This has the advantages of being a semiconductor with direct
bandgap allowing lasing emission by electrical pumping, going from the near infrared to the red at
room temperature, and of having a large second order nonlinear susceptibility, allowing an efficient
parametric conversion. Moreover, the lattice mismatch between GaAs, AlAs and their tertiary alloys
Alx Ga1-x As is very small, leading to high quality heterostructure growth with negligible constraints.
This allows to vary at will the Al molecular fraction, providing a great versatility for dispersion
engineering for the heterostructure design. Finally, Alx Ga1-x As is a very well-known material, which
permits an accurate design. In our case, we used the Gehrsitz model [78] for the optical indexes and
the Ohashi model [117] for the nonlinear effective tensor values.
The main difficulty with this device is to equalize the phase-matching at the laser wavelength
(by effective indices engineering) and to obtain the laser emission on the spatial mode corresponding
to the pump mode of the nonlinear interaction. These two constraints bending the structure in
opposing directions, a compromise must be made. The structure is composed of a quantum well (for
laser emission) embedded in a core surrounded by confining layers. These last consist of two Bragg
mirrors providing both a photonic band gap vertical confinement for the laser mode at 780 nm and
total internal reflection cladding for the photon pair modes at telecom wavelength. The constraints
and guidelines to take into account for the design are then the following:
• Maximizing the SPDC efficiency:
– maximizing the nonlinear coefficient (core layer with the smallest possible Al content);
– maximizing the nonlinear overlap integral between the different interacting modes by mode
confinement:
∗ high index contrast between the low and high index layers of the Bragg mirrors,
∗ core layer with the biggest possible Al content to confine the photon pair modes.
35
36 FABRICATION
• Laser emission:
– in the Bragg mode: the overlap between the quantum well and the Bragg mode has to
compare favourably with that of the fundamental mode at 780 nm (core layer index not
too big);
– electrical injection: efficient carrier transportation towards the quantum well (high dop-
ing);
– transparent materials at the laser wavelength (Al content superior to 25%).
• Propagation losses minimization: indeed the loss of one photon of the pair implies the pair
destruction and noise on the detectors:
– avoid guided modes leakage in the substrate (which has a higher optical index than the
mode effective index): by choosing high index contrast in the Bragg mirror, and a sufficient
number of layers.
– keep a low carrier density to minimize absorption processes.
• Robust device: layers having an Al content superior to 80% are subject to natural oxidation;
this process not only changes the optical properties of the device but also fragilize it.
Several parameters are used to design an heterostructure satisfying the different constraints: the
aluminium content; the layer thickness; the number of layers in the Bragg mirrors (which we want
minimize to keep the structure as thin as possible to reduce the growth complexity and cost). The
simulations were carried out using the transfer matrix method [77]; the final optimized structure is
shown on figure 2.1, which presents the index profile at 780 nm, the profiles of the guided modes
involved in the laser and nonlinear process, and the doping in the epitaxial direction.
QW
Substrat
Cap
High
1E20
Heart
Low
1E19
1E18
N P
1E17
1E16
Figure 2.1 – Design of the final optimized structure: index profile (black curve); guided mode intensity
profiles (dark blue curve for the pump mode at 780 nm and light blue curves for the paired photons
modes at 1550 nm); and doping profile (red curve) in the epitaxial direction.
2.2. SAMPLE GROWTH 37
2.2 Sample growth

The heterostructure growth was carried out by Molecular Beam Epitaxy (MBE) at the Laboratory
of Photonics and Nanostructures (LPN) by Aristide Lemaître, Elizabeth Galopin and Carme Gomez
Carbonell. MBE is an epitaxy technique which involves sending one or several molecular beams, under
ultra-high vacuum (10−13 bar), to a given substrate. The different elements are heated in separate
cells until they sublime, and then the gaseous elements condense on the wafer in a crystalline manner.
The content of the different materials in the epitaxied layers is controlled by the cell temperature;
the growth speed is of the order of 1 µm/h. Hence this method is expensive and slow, but it allows
high precision of the layer’s thickness, abrupt interfaces, good sample uniformity, and low impurity
levels.
In our case, we used a two inches GaAs (100) wafer, N doped (with Si atoms) for the active
samples and undoped for the passive samples. The growth uses 5 different cells: Ga, As, Al, Si (N
doping) and C (P doping). Moreover, in order to obtain a thickness gradient in the sample core, and
hence relax a constraint on the phase-matching wavelength, the wafer rotation was stopped during
the core growth.
Despite the high control during the growth, the samples obtained can differ slightly from the
nominal structure (in thickness, Al percentage, doping, ...). The phase-matching condition is very
sensitive to these parameters. To know them precisely, at the end of the growth, the multi-layer
sample was analyzed using reflectometry spectroscopy (FTIR for Fourier Transform Infrared Spec-
trometer). This allows, thanks to a simulation based on the transfer matrix method [77], a preliminary
verification on the agreement between the nominal and real structure.
2.3 Sample processing

AlGaAs benefits from a wide-spread, mature clean room technology; all the nanofabrication steps
to produce the samples took place in the clean room of the Paris Diderot University (except for the
electronic lithography, which was carried out in the clean room of the École Normale Supérieure de
Paris).
The sample fabrication took place in four major steps:
1. The etching of the waveguides, to ensure a transverse confinement of the optical modes;
2. the back metallic contact deposition;
3. the insulating layer deposition.
4. the upper metallic contact deposition for the current injection in the diode.
2.3.1 Waveguide etching

The first step of the fabrication process is the etching. The back metallic contact, if already present,
may be damaged during this step. During my thesis we used two different approaches to etch the
samples.
2.3.1.1 Etching method choice

Wet etching The substrate (partially protected by a masking material) is plunged into a solution
which attacks chemically the non-protected surface. This method is easy to implement, rapid (high
etching rate) and produces extremely smooth flanks, which leads to very low propagation losses
in the waveguides. On the other hand, it has the inconvenience of being highly isotropic (i.e. the
substrate is equally eroded in all directions), which leads to a curved flank profile. Figure 2.2.a shows
a Scanning Electron Microscope (SEM) image of a waveguide etched by this method.
38 FABRICATION
Dry etching The substrate is exposed to an ion bombardment (usually a plasma) in order to
remove layers of materials. Dry etching is hence a physical process contrary to wet etching. This
method has the advantage of etching directionally; the waveguide flanks will be straight, which gives
better confinement of the different modes in the cavity. The drawbacks are a higher complexity in
implementation with respect to wet etching, a low etching rate and a slightly higher flank roughness.
Figure 2.2.b shows a SEM image of a waveguide etched by this method.
(a) Wet etching (b) Dry etching
Figure 2.2 – SEM images of two waveguides etched with two different methods: (a) curved flanks
with wet etching and (b) straight flanks with dry etching.
2.3.1.2 Process
Sample preparation Starting from the epitaxially grown wafer, small sample chips (roughly 1 cm
by 1.5 cm) were cleaved using a diamond pen. The samples were then carefully cleaned (figure 2.3.a)
to avoid any failure propagation during the clean room process. The cleaning procedure was as
follows:
• the sample is stirred in an acetone bath (which removes organic impurities and greasy contam-
inations), then an isopropanol bath (which removes polluted acetone), and finally a deionized
water bath;
• then the sample is blow-dried with a nitrogen (N2 ) gun and put into a plasma oxygen (O2 )
cleaner (that breaks organic bonds of surface contaminants);
• finally, the sample is heated on a hotplate to dehydrate the substrate for a better resist depo-
sition.
Photoresist spin-coating The resist allows the tranfer of a desired pattern from an optical mask
to the sample. To obtain good results for the mask transfer, it is important to have a uniform resist
layer on the whole sample (figure 2.3.b). The resist deposition procedure consists in:
• adhesion promoter spin-coating1 , to enhance resist adherence to the surface;
• resist spin-coating.
The adhesion promoter and resist used depend on the considered lithography (optic or electronic):
1
several drops are deposited by means of a pipette on the sample centre, and the sample is then spun round for
several seconds.
2.3. SAMPLE PROCESSING 39
optic electronic
Promoter HMDS TiPrime
Resist S1805 ma-N 2405
The resist thickness depends on the etching deep. The thinner the layer, the more accurate is
the lithography, but the etching process may attack the resist, and hence a sufficient resist thickness
is needed to protect the pattern during the whole etching time.
After the resist deposition, a soft-baking can be necessary in order to remove all residual solvents
from the resist. Since the resists are photosensitive, any exposure to light must be avoided following
resist deposition.
Lithography
Optical lithography (for wet etching) Before optical lithography, it is necessary to check the
cleanliness of the mask used (a glass with chromium patterns) to avoid a transfer of dust or residual
resist to the sample surface. Then the mask and the sample are put in the mask aligner. The
parameters used to adjust are the pressure of the sample on the mask, the alignment between the
sample and the mask and the ultraviolet (UV) exposure time (figure 2.3.c). The exposure to the
radiation changes the resist chemical nature, making it soluble (positive resist) or insoluble (negative
resist) in the developer bath.
As the acid etching is going to under-etch the resist, due to its isotropy, it is necessary to adapt
the mask width to the depth and width of the final waveguides.
Electronic lithography (for dry etching) The e-beam lithography equipement is basically a
SEM able to move the electron beam over the surface in order to draw patterns. In our case we used
the Fixed Beam Moving Stage (FBMS) mode, in which the electron beam stays stationary while the
stage is moved to generate the pattern on the resist-covered sample using a precision piezo translation
stage. This mode allows for rapid and stitching error-free exposure of long straight waveguides. The
working parameters are:
• the voltage (20 kV),
• the diaphragm aperture (30 µm),
• and the area dose, which is the quantity of electrons per cm2 needed to expose the pattern
(165 µC cm−2 ).
The time required to expose and draw the pattern is directly linked to the employed dose and
the pattern length. In our case, the lithography took several hours.
When the e-beam draws the chosen pattern onto the substrate, the exposed resist undergoes
a crosslinking (i.e. a bonding of the resist polymer chains), while the unexposed resist remains
unchanged. This produces a much slower development rate of the exposed areas as opposed to the
unexposed ones.
Development After the exposure, the resist was developed by means of a developer (specific to
each resist). This reveals the mask pattern on the resist, which will afterwards be used as a mask for
the substrate during the etching process (figure 2.3.d). The etching time is an important parameter;
an under-developed resist will cause a wrong pattern and an over-developed resist will erode the
pattern edges. Hence it is preferable to use several short development steps and to verify the pattern
appearance each time with an optical microscope supplied with an UV filter.
Oxidation removal To obtain a better etching, the surface oxides have to be removed by quickly
immersing the sample into a hydrochloric (HCl) bath.
40 FABRICATION
(a) Epitaxied structure (b) Resist spin-coating
(c) Photolithography (selective UV insolation) (d) Resist development (removal of the insulated re-
sist)
(e) Chemical etching (f) Resist lift-off
Figure 2.3 – Steps for the chemical etching of the waveguides.

Etching
The etching height is a crucial parameter and has to be controlled precisely: the quantum well in
the core sample should not be etched (or surface states appear and the radiative lifetime increases at
the cost of the laser threshold) but the more profound is the etching and the more confined are the
guided transverse modes. For both etching methods, it will be the etching time that will determine
the sample height.
Wet etching To transfer the pattern from the resist mask to the heterostructure, a non-selective,
isotropic etchant solution is needed. All the layers, whatever their aluminium content, have to be
etched the same way, and the etching has to take place along the (011) or (011̄) crystalline axis, in
order to exploit the maximum non-zero optical nonlinear coefficient and a natural cleavage plane.
The solution used was an acid mixture composed, in equal proportions (1:1:1), of hydrogen bromide
(HBr), acetic acid (CH3 COOH) and potassium dichromate (K2 Cr2 O7 ).
The etching strongly depends on the temperature and the stirring: in our case, the solution
was at room temperature and the sample was slowly stirred in a way as to have the waveguides
perpendicular to the solution for about 25 second by micrometer.
Dry etching The samples were dry-etched in an Inductively Coupled Plasma - Reactive Ion Etch-
ing (ICP-RIE) system with a SiCl4 -Argon gas mixture, using a silicon (Si) wafer as a sample tray.
This method is anisotropic and dependant on the Al content. However, the layer thickness and the
Al content difference between the layers were small enough in the heterostructure so as not to be
sensible to this effect.
Resist lift-off After etching, the remaining resist was removed (figure 2.3.f) in an acetone bath,
followed by an isopropanol bath.
For optically pumped samples, the fabrication ended here. As these samples do not have a
quantum well, the etching deep was chosen to be 5 µm (until the end of the Bragg structure, as can
be seen on the SEM images 2.2) in order to obtain a better mode confinement.
Figure 2.4 – Insulation layer detachment in the case of deposition before the baking at 420 ◦ C.
42 FABRICATION
2.3.2 Back metallic contact deposition

Substrate thinning First the substrate was mechanically thinned down until a thickness of around
300 µm, in order to facilitate the final cleaving of the sample without weakened it too much (figure
2.5.a).
Ion pickling Then the substrate surface was bombarded by an ion beam to remove its upper
mono-atomic layers. This step cleans the surface and make it coarser, which allows a better metal
grip.
Metallic deposition Next the back contact is deposited in a vacuum evaporation machine (figure
2.5.b). Evaporation is a method of thin-film deposition, in which source materials are evaporated in
vacuum. This allows vapour particles to travel directly to the substrate, where they condense back
to a solid state. This contact is a stacking of 10 nm nickel (Ni), 60 nm germanium (Ge), 120 nm gold
(Au), 20 nm Ni and 200 nm Au.
Baking Last, a quick baking (420 ◦ C during 1 minute) allows the germanium atoms to migrate into
the N-doped substrate thereby creating a N-over-doped area and an ohmic contact between the metal
and the semiconductor (figure 2.5.c). This step has to take place before the insulation deposition in
order to avoid its detachment during the baking, as it is the case on the SEM figure 2.4.
(a) Substract thinning (b) Electrode deposition
(c) Ohmic baking
Figure 2.5 – Steps for the metallic back contact deposition.

2.3.3 Insulator deposition

In order that the current flows through all the device, it is essential that the metallic up contact does
not cover the flanks of the waveguide; the whole surface will therefore be covered by an insulation
layer with the exception of the top waveguides.
Nitride deposition First a 400 nm uniform layer of silicon nitride (SiN) is deposed by Plasma-
Enhanced Chemical Vapour Deposition (PECVD) on the whole surface (figure 2.6.a). PECVD is a
process used to deposit thin films from a gas state to a solid state on a substrate.
Optical lithography In order to open the insulation layer exactly on top of the waveguides, a new
optical lithography is needed: the resist was spin-coated (S1828) (figure 2.6.b) and insolated through
a mask whose pattern was carefully aligned on the previous ones (figure 2.6.c). After development,
the resist protects all the surface but the top of the guides (figure 2.6.d).
Insulation layer etching The up contact was opened by etching the insulation layer, where it is
not protected by the resist, by Reactive-Ion Etching (RIE) (figure 2.6.e). Afterwards the resist was
removed with acetone and O2 plasma (figure 2.6.f).
2.3.4 Up metallic contact deposition

1st optical lithography To be able to drive each diode independently, the insulation layer has
to be protected between each waveguide. Therefore a new optical lithography was performed: resist
(TiES reversible) was deposited by spin-coating (figure 2.7.a) and insolated a first time (figure 2.7.b).
Then the resist was baked to invert it: this produces a resist layer which is thinner at the bottom
than at the top, facilitating lift-off after metallic deposition. The resist is next insolated a second
time without any mask, and finally developed (figure 2.7.c).
Evaporation gold deposition The upper contact was deposited in the vacuum evaporation ma-
chine (figure 2.7.d): 10 nm titanium (Ti) and 400 nm gold (Au). To have a uniform metallic layer on
the waveguide flanks, the sample was turned with a 40◦ angle with respect to the atomic flow. This
deposition does not create an ohmic contact but the over-doping of the structure cap (last epitaxial
layer) allows a weak resist contact. Then the resist and the metallic residues were removed (figure
2.7.e).
2nd optical lithography For the first sample generation the fabrication ended at this stage. These
samples had a strong tendency to burn under the tip used to inject current through the sample. Hence
we decided to increase significantly the upper gold layer, which should at the same time reduce the
laser threshold. The optical lithography for this second deposition was exactly the same as for the
first one.
Electrolytic gold deposition The principle of an electrodeposition is to create an electric current

between an electrode and the sample metallic surface, immersed into a solution containing a solvent
and Au+ ions. The passage of the electric current reduces the dissolved metal cations so that they
form a coherent metal coating on the negatively polarized sample. This method has the advantage
of depositing thick gold layers in a few minutes (several µm in 20 minutes).
44 FABRICATION
(a) Insulation layer deposition (b) Photoresist spin-coating
(c) Photolithography (d) Resist development
(e) Insulator etching (f) Resist lift-off
Figure 2.6 – Steps of insulator deposition.

(a) Photoresist spin-coating (b) Photolithography
(c) Resist development (d) Electrode deposition by evaporation
(e) Resist lift-off
Figure 2.7 – Steps for the metallic up contact deposition.

46 FABRICATION
At the end of fabrication, the sample was cleaved at both ends to produce the waveguide input
and output facets. A typical length of the bars was 2 mm and it contained several waveguides. A
SEM image of the final sample is shown in figure 2.8.
Figure 2.8 – SEM image of a final sample. The electrolytic gold layer stops before the facet in order
to facilitate the cleaving.
Chapter 3
Characterization of the laser emission
This chapter describes the optoelectronics characterization of the device (wafer C4T46) under elec-
trical pumping. Since the goal of our device is to emit photon pairs under electrical injection, we
must check that the laser emission occurs on the T E Bragg mode which is phase-matched with the
fundamental T E00 and T M00 modes around 1.56 µm. To perform the optoelectronics characteriza-
tion of the device, we have evidenced its laser emission, then checked that the emitted mode is the
expected Bragg one. We have also studied its behavior as a function of temperature, injected current
and repetition rate.
3.1 Laser emission on the Bragg mode

Experimental setup
DCC
Powermeter
Waveguide
FC
Osc.
Temperature
controller OSA
Trigg.
(+) (-)
Pulsed power
supply
Figure 3.1 – Experimental setup for the laser diode characterization. CCD: InGaAs camera; FC:
fiber collimator; OSA: optical spectrum analyzer; Osc: oscilloscope.
The laser emission and characteristics are tested using the setup shown in figure 3.1. This allows
measurement of the voltage through the heterostucture as a function of the injected current and of
the power and spectral profile of the emitted light. The laser diode is injected with two tips, one
placed near the ridge on the gold contact, and the second (the mass) placed on the sample holder,
which is gold coated. The sample is glued to the holder with silver paint in order to have electrical
contact. The electrical supply is provided by a pulse generator (Agilent 8114A). The voltage and
the current passing through the diode are measured with an oscilloscope. A powermeter collects the
output power of the laser. A camera (CCD) allows verification of the spatial intensity distribution,
and an optical spectrum analyzer (OSA) is used to obtain the spectral profile of the emitted light.
47
48 CHARACTERIZATION OF THE LASER EMISSION
Laser spectrum
Two spectra of the laser diode are presented in figure 3.2. The blue and red curves show the light
emission below and above threshold respectively. The laser peak above threshold is centered at
λ = 782.61 nm. The full width at half maximum is about 280 pm, which correspond to the OSA
resolution.
4.0
Above threshold
3.5
Below threhold
3.0
Intensity (a.u.)
2.5
2.0
1.5
1.0
0.5
0.0
780.0 780.5 781.0 781.5 782.0 782.5 783.0 783.5 784.0 784.5
Laser wavelength (nm)
Figure 3.2 – Spectrum obtained below (I = 267 mA - blue curve) and above threshold (I = 557 mA -
red curve) of the sample. The acquisition has been performed for a heat-sink temperature of 22.1 ◦ C.
Spatial intensity distribution
The spatial intensity distribution of the emitted laser beam above threshold has been studied; the
recorded near field and far field distributions are reported in figures 3.3.b and c, as well as the near
field obtained with our numerical simulations (figure 3.3.a). Due to the spatial extension of the Bragg
mode, it is not easy to produce a near field image of good quality (due to spherical aberrations and
numerical aperture of the microscope objective). The far field image shows clear evidence of emission
in the Bragg mode, the three main lobes being clearly visible.
a) b) c)
Figure 3.3 – Simulated (a) and measured (b) laser mode in the near field; measured (c) far field
emission.
Polarization
The selection rules for the stimulated emission predict that the polarization of the laser light of our
device is T E. We have investigated the polarization of the laser diode beam by placing a polarizer
before the powermeter. Measurements done under threshold show that the spontaneous emission
is not polarized, while those above threshold show that the stimulated emission is T E polarized as
expected.
3.2. OPTOELECTRONICS CHARACTERISTICS 49
3.2 Optoelectronics characteristics

To characterize the device, the internal peak power and the voltage have been tested as a function
of the injected current and the temperature using a standard characterization bench (figure 3.1). In
order to explore a large range of temperatures, we have implemented an experimental setup allowing
to insert the sample in a nitrogen cooled cryostat. The current pulses employed have a duration of
60 ns and a repetition rate of 40 kHz in order to avoid unwanted thermal drifts.
Figure 3.4.a shows typical power-current-voltage (PIV) curves as a function of the temperature
of the sample under electrical pumping. At T = 291 K, the V(I) curve shows a resistance of R =
3.56 Ω and a turn-on voltage of 1.83 V; the latter being very close to the quantum well band-
gap, no current-blocking effects seem to occur at the hetero-interfaces. The P(I) curve at T =
291 K displays a threshold current of Ith = 0.274 A1 , corresponding to a threshold current density
Jth = 1.5 (±0.2) kA/cm2 for a contact surface of S = 1.9 (±0.2) × 10−4 cm2 . As the temperature
decreases from Troom = 291 K to T = 202 K, the threshold current decreases exponentially, according
to Ith ∼ eT /T0 , with T0 = 87.8 K [118], as can be seen in figure 3.4.b.
4 20
202K
227K
243K
277K
3 291K 15 P (mW)
V (V)
2 10
1 5
0 0
0 0.1 0.2 0.3 0.4 0.5
a) I (A) b)
Figure 3.4 – Optoelectronics characteristics of our device. (a) Voltage bias (red crosses) and emitted
power as a function of the injected current. The V(I) curve is measured at Troom = 291 K, while the
P(I) curves are measured in the temperature range ∆T = 202 − 291 K. (b) Measured (crosses) and
calculated (continuous line) threshold current as a function of temperature. The green band around
the calculated threshold corresponds to the uncertainty on the diode surface after the process.
These experimental results are compared with the numerical simulations developed by our team
to design the device2 . The carrier transport and recombination in the hetero-junction are calculated
using a self-consistent solution of the drift-diffusion and Poisson equations. The optical parameters of
the modeling are the following: the Bragg mode internal losses (set at α = 35 cm−1 , from the exper-
imental results obtained by second-harmonic generation in chapter 4); the overlap integral between
the Bragg mode and the quantum well (Γ = 2.45%, calculated with a commercial eigenmode solver:
Lumerical Solutions); the facets modal reflectivity for the Bragg mode (R = 0.86, calculated by
FDTD). The radiative recombination time in the quantum-well is set to τrad = 5.10−17 m3 s−1 , while
the Shockley–Read–Hall non-radiative recombination processes is τSRH = 5 ns [119]. Figure 3.4.b
shows good agreement between the experimental (crosses) and calculated (continuous line) threshold
current as a function of the temperature in the ∆T = 216 − 305 K temperature range; this agreement
is within the uncertainty related to the device contact surface S.
1
Note that the threshold current was of the order of 0.45 − 0.50 A before developing the process based on electrolytic
gold deposition.
2
All the simulations described in the following were performed by the PhD student G. Maltese.
50 CHARACTERIZATION OF THE LASER EMISSION
From the calculated internal resistance of the diode (Rint ≈ 1.5 Ω at room temperature), the
quality of the fabricated contacts is evaluated; we estimate that the contacts introduce an additional
resistance lower than R = 2.0 Ω. For higher values of injected current, the Bragg mode can be in
competition with the fundamental T E00 mode due to index anti-guiding effects coming from a carrier
density increase. This might explain the irregular behavior observed in the P(I) curves.
3.3 Tunability of the laser wavelength

As shown in the simulated tuning curves of the source presented in chapter 1, the strong dispersion
of the Bragg mode induces a great sensitivity of the signal and idler wavelengths to the pump
wavelength. The constraint to have the laser wavelength equal to the phase-matching wavelength
implies that the tunability of the laser wavelength is a key factor for the device.
Figure 7.1 shows the laser emission intensity spectra as a function of the heat-sink tempera-
ture, measured with an optical spectrum analyzer for a laser emitting on the Bragg mode at room
temperature, with an injected current of 422 mA. The general trend corresponds to the theoretical
temperature dependence of the QW band gap (0.23 nm/◦ C - dashed white line on the figure). Mode
hopping is clearly visible, with wavelength jumps of ∆λ ≈ 1.5 − 2 nm.
786
785
Wavelength (nm)
784
783
782
781
780
779
15 20 25 30 35
Temperature (°C)
Figure 3.5 – Laser emission intensity as a function of wavelength and heat sink temperature for I =
422 mA. The dashed white line shows the expected temperature variation of the QW bandgap.
These jumps correspond to more than 35 times the spacing between adjacent longitudinal modes
of the device; indeed, taking into account the laser length and the modal dispersion, the free spectral
range is given by:
λ2
∆λFSR = ≈ 40 pm, (3.1)
2L(n − λdn/dλ)
with L the device length [115]. This behavior has already been observed in AlGaAs laser structures
[119]; it seems to be explained by the presence of an additional saturable source of losses, resulting
from deep-level traps induced in the n-type doped Bragg layers by Si donors. As the traps absorption
is inversely proportional to the photon density, the standing-wave intensity profile of the dominant
longitudinal mode I(z) determines the periodicity of the optical loss pattern [120]:
α0
αtraps (z) = . (3.2)
1 + I(z)/Is
3.4. TOWARDS CONTINUOUS WAVE OPERATION 51
Then, by integrating along the whole device length, it is possible to show that the dominant longi-
tudinal mode presents smaller losses than the adjacent modes by a factor ∆α ≈ 0.05 − 0.1α0 . For
this reason, for a small shift in the gain curve, this mode will still be the lasing one; then for a larger
shift in the gain curve, a jump of several longitudinal modes will occur. In order to verify that the
observations are related to this physical process, we have estimated ∆α, calculating the overlaps
of the Bragg mode with each n-doped layer. By assuming the reflectivity R = 0.86, we find α0 =
5.54 cm−1 at room temperature, this leading to ∆α ≈ 0.28 − 0.55 cm−1 , which is compatible with
the experimental results.
Since the operational temperature of the device simultaneously sets the laser and the phase-
matching wavelength, we have checked the possibility of having an additional degree of freedom to
control the laser emission.
784.0
Wavelength (nm)
783.5
783.0
782.5
782.0
781.5
0.3 0.4 0.5 0.6 0.7
Current (A)
Figure 3.6 – Laser wavelength dependence on the injected current.
Figure 3.6 displays the laser wavelength dependence on the injected current for a heat sink
temperature of 14 ◦ C. The data demonstrate the presence of a Stark effect shifting the gain spectrum
(and thus the laser wavelength emission) of the device [121]; this effect could therefore be used for a
fine tuning in operating the device.
3.4 Towards continuous wave operation

In order to check the possibility of using the device in continuous wave operation, we have measured
the output power as a function of the duty cycle. The observation of a constant peak output power for
a duty cycle up to 66% (limit of the current pulse generator we have) indicates that a CW operation
is possible. However, a clear thermal drift arises with increasing duty cycle (the wavelength shifts
with increasing duty cycle). Better heat management must therefore be set up before operating the
device in continuous wave.
Chapter 4
Characterization of the nonlinear

waveguide
In this chapter, we describe the experiments related to the optical characterization of the waveguide
sample, the main goal being the determination of the phase-matching (PM) resonance that per-
mits photon-pair generation at telecom wavelengths. First, linear characterizations are performed
with optical losses measurement. Secondly, nonlinear characterizations are performed with second
harmonic generation (SHG) experiment, this providing a fast estimation of the PM resonance wave-
length, and spontaneous parametric down-conversion (SPDC) experiment, to achieve fine tuning of
the PM. Finally, a method allowing high accuracy in reconstruction of frequency states is used.
The results given in this chapter include also a comparison between two fabrication techniques
of our waveguides. Indeed before the beginning of my PhD work, all samples have been fabricated
using wet etching. I have set up a fabrication technique based on dry etching, allowing to get a better
confinement of the interacting fields and thus a higher conversion efficiency.
4.1 Measurement of propagation losses

In a two-photon source, a key factor is the minimization of the losses; indeed, the loss of one of the
photons of the pair results in a reduction of the number of usable pairs at the output of the source
and in a degradation of the signal to noise ratio. Different factors can cause photons losses within
the waveguide: poor etching quality, free carriers, mode leaking ...
To measure the propagation losses in the telecom range, we use a method proposed by Regener
and Sohler [122], and extended by our team to semiconductor waveguides [123]. This make use of
the relation between the waveguide attenuation coefficient and the Fabry-Perot fringes1 contrast of
the signal transmitted by the waveguide when injected with a laser whose wavelength is sweeped.
The expression for the transmitted power PT is then given by:
T 2 e−αL
PT = P0 η (4.1)
(1 − Re−αL )2 + 4Re−αL sin(φ)2
in which η is the coupling efficiency to the waveguide mode, T the end-facet mode transmissivity,
R the mode reflectivity2 , φ = 2πLneff /λ + φ0 the internal phase difference accumulated by the two
modes, with L the waveguide length, and α the attenuation coefficient of the optical waveguide.
Since the sample is a Fabry-Perot resonator, the transmitted power varies periodically with the
1
The Fabry-Perot fringes are due to the partial reflectivity at the facets induced by the refractive index contrast
between air and AlGaAs compounds.
2
The mode reflectivities are calculated by numerical simulations: R = 0.27 and 0.24 in T E00 and T M00 polarization
respectively.
53
54 CHARACTERIZATION OF THE NONLINEAR WAVEGUIDE
optical phase difference φ, and hence with the wavelength λ. By fitting the experimental data with
eq. 4.1, we can deduce the attenuation coefficient α.
PC
cw laser
DCC
λ Waveguide
Powermeter
P
FC
Temperature
controller
Figure 4.1 – Experimental setup for the propagation losses measurement. P: polarizer; FC: fiber
collimator; PC: polarization controller.
As shown in figure 4.1, the experiment is performed by coupling a continuous wave laser (Tunics),
tunable from 1.5 to 1.6 µm and having a spectral width of 2 kHz, into the waveguide via a 40x
microscope objective. The polarization is adjusted with a fibered polarization controller (PC) and
a polarizer (P). The transmitted beam is collected through a second 63x microscope objective, and
sent to a powermeter. An InGaAs CCD camera permits the observation of the transmitted optical
mode in order to control the coupling quality.
effective refractive index

coupling efficiency
attenuation coefficient
(a) T E polarization (b) T M polarization
Figure 4.2 – Transmitted power versus wavelength with incoming (a) T E and (b) T M polarization
at T = 20 ◦ C in an active sample (C4T46). The adjustement with eq. 4.1 is shown in red.
Figure 4.2 shows some typical experimental results for a T E-polarized and T M -polarized injected
beam in an active sample. The fitting of our experimental data with eq. 4.1 is shown in red. The free
parameters are the attenuation coefficient α, the phase φ0 , the coupling efficiency η and the effective
refractive index neff . We find an attenuation coefficient of α = 2.3 cm−1 for the T M polarization and
of α = 2.2 cm−1 for the T E polarization. In order to have a better understanding of the origin of
the losses, similar measurements have been performed on a different type of samples; passive samples
in which the epitaxial structure is the same except that there is no doping and no quantum well.
For these samples, different clean room processes have been tested: sample processed exactly as an
active one (an etch depth of 2.1 µm and insulation as well as electrodes depositions); samples with a
deep etching (5 µm depth), either by wet or dry etching. The results are summarized in table 4.1.
4.2. SECOND-HARMONIC GENERATION 55
Sample type Etching T E losses ( cm−1 ) T M losses ( cm−1 )
Passive wet (5 µm deep) 0.3 0.5

Passive dry (5 µm deep) 0.5 0.7
wet (2.1 µm deep)
Passive (+insulator and 0.6 0.7
electrodes)
Active wet (2.1 µm deep) 2.2 2.3
Table 4.1 – Propagation losses for the mode T E00 and T M00 at 1.56 µm for different sample types.
These values show that most of the losses in active samples are due to the doping of the structure,
which introduces free carrier absorption. The losses difference between the deeply etched passive
samples and those processed as the active one is due to mode leakage. The losses difference obtained
using the two etching processes is due to the different roughness of the waveguide flanks. We note
that all these values are the best ever reported for passive Bragg reflection waveguides. In particular,
the still low level of losses obtained with dry etching suggests that this technique could replace wet
etching in the fabrication for our future samples.
Polarization conservation
In the SPDC process, we convert a T E Bragg mode into a fundamental T E mode and a fundamental
T M mode. In order to manipulate the polarization of the emitted photons (for example, in a
polarization entanglement experiment), it is important that the waveguide conserves the polarization
of the light it generates.
This property has been checked by injecting the T unics laser beam into the waveguide, after
choosing the polarization of the incoming wave to be either T E or T M . Then the outgoing wave is
collected with a powermeter. The output polarization is analyzed via a polarizer inserted before the
detector. In both cases, the power drops to zero, which demonstrates that the waveguide maintains
its initial linear polarization.
4.2 Second-harmonic generation

A convenient way to test the nonlinear properties of the waveguide is to perform an experiment of
SHG, which is the reverse of SPDC at degeneracy; indeed, the SH signal is easier to detect since it
is generated by two pumping beams rather than by vacuum fluctuations as is the case for SPDC.
Hence it permits rapid identification of the phase-matching resonance and gives a first estimation of
the conversion efficiency.
4.2.1 Simulations
From chapter 1, the optical power generated by SH in a lossless waveguide is given by:
∆kL
0
P2ω = ηSHG L2 Pω2 sinc2 (4.2a)
2
8π 2
with 0
ηSHG = |Γ|2 . (4.2b)
0 cλ22ω (neff
ω )2 neff
2ω
But, as it has been seen in the previous section, our waveguide is not lossless. Then the expression
of the SH power, taking into account the optical losses [48] and the reflectivity of the facets, is given
0.8
SHG power (a.u.)

0.6
0.4
0.2
0
1.562 1.564 1.566 1.563 1.565 1.567 1.564 1.566
Wavelength (µm)
1
0.8
0.6
0.4
0.2
0
1.562 1.564 1.566 1.563 1.565 1.567 1.564 1.566
0.8
0.6
0.4
0.2
0
1.562 1.564 1.566 1.563 1.565 1.567 1.564 1.566
1
0.8
0.6
0.4
0.2
0
1.562 1.564 1.566 1.563 1.565 1.567 1.564 1.566
Figure 4.3 – Simulations of SH power using eq. 4.3 as function of the sample length L and the losses
α inside the waveguide.
by:
0
PSHG = ηSHG × L2 × PIR
2
× (1 − RSHG ) × (1 − RIR−T M ) × (1 − RIR−T E ) × F

(4.3)

αSHG L
α sin2 ∆kL
2 + sinh2 (αIR − 2 )2
−(αIR + SHG )L
×e 2 × 2 2 ,
∆kL αSHG L
2 + (αIR − 2 ) 2
in which L is the sample length, PIR the injected power inside the waveguide, RSHG and RIR the modal
reflectivities of the Bragg mode at 780 nm and of the fundamental modes at 1.56 µm respectively,
αIR and αSHG are the loss coefficients in the waveguide, and ∆k the phase-matching. F is given by
the product of the three Fabry-Perot functions corresponding to the reflection of the three modes on
the waveguide facets:
F = FSHG × FIR−T M × FIR−T E

1
=
(1 − RIR−T E e−αIR−T E )2 + 4RIR−T E e−αIR−T E sin2 (φT E )
1 (4.4)
× −α
(1 − RIR−T M e IR−T M ) + 4RIR−T M e−αIR−T M sin2 (φT M )
2
1
× ,
(1 − RSHG e SHG ) + 4RSHG e−αSHG sin2 (φSHG )
−α 2
with the phase φm = 4πneff 0

m L/λm + φm . This gives an additional modulation to the signal envelope
(a square sinus cardinal, characteristic of a phase matching resonance). The conversion efficiency of
the device is thus affected by the relative phases of the interacting fields. Figure 4.3 shows different
simulations of the generated normalized second harmonic power with the losses and reflectivities
(RSHG = 0.79 given by numerical simulation) taken into account (the envelop function is shown in
blue). The three columns show the numerical simulations obtained for three different values of length
keeping equal all the other parameters. We can see that the sample length is indeed an important
parameter as it can vary the generated power by more than 50%.
It is also of interest to vary the losses of the Bragg mode at 780 nm and of the fundamental modes
at the telecom wavelengths; we can see on the simulated curves that the losses of the Bragg modes
have an effect on the form of the curve (rows 2 to 4), while the only effect of the losses on the telecom
modes (rows 1 and 2) is to reduce the output power (which cannot be seen on the figure, due to the
renormalization).
4.2.2 Experiments
The SHG experiment consists in the injection a laser beam in the telecom range polarized at 45◦ in
the waveguide, and in the detection of a SH signal on a Bragg mode around 780 nm. The injection
at 45◦ provides the two fundamental waves with the orthogonal polarization needed in a type II
phase-matching.
Experimental setup
The experimental setup for the SHG experiment (represented in figure 4.4) is composed of a T unics
laser that is coupled to the sample mounted on a Peltier system to control its temperature. The
laser is connected to an Erbium amplifier (EDFA - KEOPSYS) and a fiber polarization controller
(PC) to rotate the laser light polarization to be diagonal at the input of the waveguide. The light is
then coupled to the waveguide via a 40x microscope objective; a 63x objective with a large numerical
aperture (N A = 0.95) is used to collect the generated second-harmonic signal and the transmitted
fundamental, which is filtered by a bulk high pass filter (HP F). A silicon photodiode, connected to a
lock-in amplifier (DLPCA-200 ), is used to measure the SH signal. The polarization of the generated
mode can be analyzed via a polarizer inserted before the detector. A CCD camera permits the
observation of the SH mode, in order to ensure that it is a Bragg mode.
PC
cw laser
DCC
EDFA Waveguide
λ
HPF
P
I V
Gain
FC Si
Temperature Lock-in
controller
f
Figure 4.4 – Experimental setup for the SHG measurement. P: polarizer; FC: fiber collimator; PC:
polarization controller; HP F: bulk high-pass filter.
Results
Figures 4.5.a, b and c show the measured signal for second harmonic generation as a function of the
pumping wavelength for different samples. The red lines correspond to the data, and the black lines
to the fit with eq. 4.3, where the parameters are L the length of the sample and αSHG the losses of
the Bragg mode at 780 nm (the losses of the two fundamental modes have been determined in the
previous experiment). The phase-matching resonance is clearly observed. The insert in figure 4.5.a
shows the generated second harmonic maximal output power as a function of the fundamental input
power in a log-log scale. The expected quadratic dependence (eq. 4.3) is confirmed, as can be seen
from the very good agreement with the theoretical curve (in red). The insert in figure 4.5.c shows
the far field image of the SH field, recorded with the CCD camera. The spatial intensity distribution
of the SH mode shows the three main lobes expected for the Bragg mode. The polarization of the
mode has been checked to be T E.
We can see from the results that the efficiency is strongly dependent on the type of sample.
These differences arise from differences in the overlap between the three interacting modes. The
active sample, etched only to the quantum well (and thus only to the middle of the structure), is
the less confining sample, leading to the smallest overlap and thus the smallest conversion efficiency.
The passive sample fabricated using dry etching is the best confining sample, leading to the highest
overlap and thus the highest efficiency. These results support the option of changing the clean room
process from wet to dry etching for future devices. Moreover, the fitting allows the evaluation of the
losses of the Bragg mode. We can see that the main difference is between the passive samples and
the active sample, as expected due to the doping and the quantum well absorption. The results are
summarized in table 4.2.
Sample type αBragg ( cm−1 ) η0 ( % W−1 cm−2 )
Active - wet etching 35 ± 2 50 ± 5

Passive - wet etching 4.0 ± 0.5 210 ± 10
Passive - dry etching 2.0 ± 0.5 1980 ± 30
Table 4.2 – Losses of the Bragg mode at 780 nm and normalized conversion efficiency for different
types of samples.
0.4
-64
0.35
-68
0.3
PSHG (nW)
0.25 -72
0.2 -76
0 2 4 6
0.15
0.1
0.05
0.0
1562 1563 1574 1575 1576 1577
Wavelength (nm)
(a) Active sample (wafer C4T46) and quadratic dependency in insert
18.0
16.0
14.0
PSHG (nW)
12.0
10.0
8.0
6.0
4.0
2.0
0.0
1565.5 1566.0 1566.5 1567.0 1567.5
Wavelength (nm)
(b) Passive sample (wet etching) (wafer EBW002)
60.0
50.0
PSHG (nW)
40.0
30.0
20.0 Growth
Growth axis
axis
10.0
0.0
1530.0 1530.5 1531.0 1531.5 1532.0 1532.5 1533.0
Wavelength (nm)
(c) Passive sample (dry etching) (wafer ABQ71) and far field in insert
Figure 4.5 – Typical type II SH spectrum as a function of the pumping wavelength for different types
of samples. The red lines show the data and the black lines the fitting with eq. 4.3.
4.2.3 Tunability of the phase-matching wavelength

A necessary condition for our active device to work is that the laser wavelength is equal to the phase-
matching (PM) wavelength (λlaser = λPM ). It is therefore important to have parameters with which
to tune these wavelengths. It can also be of interest to tune the PM of a passive sample, for example
to match it with that of a filter.
First we looked at the PM wavelength as a function of the waveguide width by defining different
widths during the etching process. Figure 4.6 shows the experimental results; the PM wavelength
increases as the width of the wavelength decreases, as predicted by the numerical simulations.
Figure 4.6 – Measured dependence of the PM wavelength on the waveguide width (ABQ71).
This offers a first control parameter during the clean room process. Then we have look for another
parameter with which to have fine tuning of the phase-matching wavelength during the experiments.
This led us to investigate the dependence of the PM wavelength on temperature, as had been done
with the laser wavelength. The experimental results, obtained by recording the SHG signal as a
function of the heat-sink temperature, are shown in figure 4.7; we find a thermal tuning coefficient
consistent with the Gehrsitz model [78].
Figure 4.7 – Experimental tuning of the PM resonance wavelength with the temperature (C4T46).
4.3. SPONTANEOUS PARAMETRIC DOWN-CONVERSION 61
4.3 Spontaneous Parametric Down-Conversion

Before any characterization of the quantum state emitted by our device, it is important to find its
optimal working condition. The SHG characterization allowed us to find good starting parameters
for the phase-matching, but a second experiment is needed for fine tuning. We therefore recorded the
temporal correlations (or coincidence histogram) between the paired photons at telecom wavelength.
In the remainder of this chapter we restrict ourselves to passive samples. We first present the method
we used to determine the optimal working condition; we then compare the results obtained for dry
and wet etched samples. The characterization for an active sample is different, as will be presented
in chapter 7.
Experimental setup
The experimental setup is sketched in figure 4.8. A light beam of a continuous wave laser at 780 nm
is used to excite the Bragg mode of the sample. After selecting the T E polarization with a polarizer
(P) and performing spatial shaping of the mode using a holographic mask (HM), the laser beam
is injected into the waveguide with a 63x microscope objective. Light emerging from the opposite
end is collected with a 40x microscope objective and a fiber coupler (FC). A low-pass filter (F) with
a cut-off wavelength at 1.5 µm allows to eliminate the pump light beam. Having photon pairs at
telecom wavelength, all the remaining setup is fibered; this enables us to replace, add, or remove ele-
ments without modifying the optical free space path. A tunable fibered Bragg grating (FBG) having
a full width at half-maximum (FWHM) of 10.8 nm is used to select photons around the degeneracy
wavelength. Thanks to type II phase-matching, the two photons of the pair have orthogonal polar-
izations. Hence they can be deterministically separated with a polarization controller (PC) coupled
to a polarized beam splitter (PBS). Two InGaAs single-photon avalanche photodiodes (SPAD - D)
connected to a time-to-digital converter (&) are used for coincidence counting.
Waveguide
DM F
FC FBG
HM Temperature
controller
M
FC P
FC PC
Ti:Sa laser D PBS
λ &
PC
Figure 4.8 – Sketch of the experimental setup used for coincidence searching. PC: polarization
controller; FC: fiber collimator; M: mirror; DM: dichroic mirror; P: polarizer; HM: holographic mask;
F: low-pass filter; FBG: tunable Bragg grating filter; PBS: polarized beam splitter; D: detector; &:
time-to-digital converter.
Optimal working point search
In looking for the optimal working condition, an important factor is the signal to noise ratio (SNR),
which is calculated by taking the number of true coincidences within the FWHM of the peak over
the background signal, on the same time window taken apart from the peak:
peak − noise
SNR = . (4.5)
noise
The SNR dependence on both the pump wavelength (figure 4.9.a) and the internal pump power
(figure 4.9.b) has been studied. The optimal working condition corresponds to the phase-matching
resonance of the device. For a pump wavelength above the degeneracy point, the phase-matching
condition is no longer satisfied; below degeneracy, a shift of 0.2 nm results in a variation of 100 nm
for the signal and idler wavelengths, which are consequently outside the interferential filter. Further,
the SNR increases with increasing pumping power until it reaches a maximum, after which too many
double pairs are created, leading to a degradation of the SNR. Figure 4.9 shows the measurements
done for the optimal working condition search for a wet etched passive sample, and the coincidence
histogram obtained for the best choice of parameters. The FWHM of the peak is due to the detector
jitter.
160 140
140 120
120 100
100
80
SNR
SNR
80
60
60
40 40
20 20
0 0
782.8 782.9 783.0 783.1 783.2 350 400 450 500 550 600 650 700 750 800
Pump wavelength (nm) Internal pump power (µW)
(a) Constant pump power (625 µW) (b) Constant pump wavelength (782.996 nm)
10
Coincidences counts (/s)
9
8
7
6
5
4
3
2
1
0
-10 -5 0 5 10
Time delay (ns)
(c) Coincidence histogram
Figure 4.9 – Dependence of the SNR on a) the pump wavelength and b) the pump internal power; c)
is a typical histogram at the optimal working condition for a wet etched passive sample (EBW002).
Comparison between wet- and dry-etched samples
After the PM determination with SHG and then having searched the optimal working condition
for SPDC for wet and dry etched samples, we have compared the best results for the two samples.
The results are shown in figure 4.10. For the wet etched sample (figure 4.10.a), we have obtained
a SNR value of 144, while for the dry etched sample (figure 4.10.b) we have a SNR value of 1728.
Consequently dry etched samples are a very interesting alternative to wet etched samples for fu-
ture devices. Simulations are in progress to understand these results with respect to photon-pair
production efficiency, losses reduction and coupling efficiency.
4.3. SPONTANEOUS PARAMETRIC DOWN-CONVERSION 63
10 80

9 70
8
60
7
6 50
5 40
4 30
3
20
2
1 10
0 0
-10 -5 0 5 10 -10 -5 0 5 10
Time delay (ns) Time delay (ns)
(a) Wet etching (EBW002) (b) Dry etching (ABQ71)
Figure 4.10 – Comparison of the coincidence histograms for samples processed with two different
etching technics. The resolution sampling is 164 ps.
Comparison between different detectors

The detector’s dark counts contribute to the noise, thus limiting the SNR. Hence we have tested
different types of detectors. In the following figure, two histograms of the same waveguide pumped
under the same conditions but recorded with detectors having different dark count rates are shown.
In figure 4.11.a, the detectors used are one free running idQ210 with an efficiency of 10% and dark
count rate of 2.3 kHz, and one gated idQ201 with an efficiency of 25%, triggered by the first detector.
In this case the SNR value is 20.48. In figure 4.11.b, the detectors used are two stirling-cooled free-
running [124] avalanche photodiodes with an efficiency of 20% and dark count rates of about 60 Hz.
In this case the SNR value is 82.81.
1.2 1.2

1.0 1.0
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0.0 0.0
-10 -5 0 5 10 -10 -5 0 5 10
(a) id210 and 201 detectors (b) id230 detectors
Figure 4.11 – Comparison between the coincidence histograms pumping the sample (EBW002) in
the same conditions but using different detectors. The resolution sampling is 164 ps.
4.4 Joint Spectral Amplitude and Intensity

Theory
In chapter 1 we derived the expression of the two-photon state emitted by our device. We have
seen that the joint spectral amplitude, containing all the spectro-temporal information about the
two-photon state, is given by:
φ(ωs , ωi ) = α(ωs + ωi ) × φPM (ωs , ωi ) , (4.6)
where α(ωs + ωi ) is the spectrum of the pump field, and φPM (ωs , ωi ) is the phase-matching function.
In our case we pump the device with a continuous wave laser, then α(ωs + ωi ) is given by the
bandwidth of the pumping laser. Figures 4.12.a and d show the modulus of α(ωs + ωi ) for a pump
having a central wavelength of 765.9 nm and a bandwidth of 6 × 1012 Hz. φPM (ωs , ωi ) is a cardinal
sine function, as shown in chapter 1, that drops away from the line of zero phase mismatch between
the pump and paired photons. Figures 4.12.b and e show the norm of this function for our device. The
norm of the JSA is then given by the product of these two functions, as can be seen in figures 4.12.c
and f. The φPM and α functions are both aligned along the anti-diagonal axis, hence the photon
pairs are highly anti-correlated in frequency in our sample.
1580 1580
Idler wavelength (nm)
1560 1560
1540 1540
1520 1520
1500 1500
a) b) c)
1500 1520 1540 1560 1580 1500 1520 1540 1560 1580 1500 1520 1540 1560 1580
1536 1536
1534 1534
1532 1532
1530 1530
1528 1528
d) e) f)
1528 1530 1532 1534 1536 1528 1530 1532 1534 1536 1528 1530 1532 1534 1536
Signal wavelength (nm) Signal wavelength (nm) Signal wavelength (nm)
Figure 4.12 – Numerical simulations for the (a) norm of the spectral function α(ωs + ωi ), (b) norm
of the phase-matching function φPM (ωs , ωi ) and (c) norm of the JSA for the biphotons emitted by
our device when pumped with a beam having a central wavelength of 765.9 nm and a bandwidth of
6 × 1012 Hz. (d), (e) and (f) present a zoom of the (a), (b) and (c) plots, respectively.
Effects of the facets reflectivities

Due to the refractive index contrast between air and AlGaAs, the facets of the waveguide create a
cavity for the signal and the idler modes. This cavity has a non negligible impact on the emitted pho-
tons; the original JSA is modulated by the transfer functions describing the Fabry-Perot transmission
4.4. JOINT SPECTRAL AMPLITUDE AND INTENSITY 65
for the signal and idler frequencies [125]. The JSA obtained with reflective facets is then
φFP (ωs , ωi ) = fT E (ωs )fT M (ωi )φ(ωs , ωi )

or (4.7)
φFP (ωs , ωi ) = fT M (ωs )fT E (ωi )φ(ωs , ωi ) .
This is illustrated in figure 4.13 in which the JSA obtained for non-reflective facets represented in
(a) is transformed into the JSA considering the facets reflectivity in (b). Specific frequencies for the
signal and idler photons are selected, this leading to a checkerboard pattern on top of the original
JSA.
1536 1536
1534 1534
1532 1532
1530 1530
1528 1528
a) b)
1528 1530 1532 1534 1536 1528 1530 1532 1534 1536
Signal wavelength (nm) Signal wavelength (nm)

Figure 4.13 – Numerical simulation for the norm of the JSA with (a) non reflective facets, and (b)
reflective facets.
Stimulated emission tomography: working principle

The JSA is a complex function containing the phases of α and φ. The joint spectral intensity (JSI),
which is much easier to access from the experiments, is defined as
JSI = |φ(ωs , ωi )|2 . (4.8)
This represents the probability of detecting a signal photon with frequency ωs and a corresponding
idler photon with frequency ωi for a photon pair.
There exist several methods to reconstruct the JSI. We can cite for example the single-photon
fiber spectrograph, in which the frequency information is translated into temporal information using
highly chromatic dispersion fibers [126, 127]; the group velocity dispersion of optical fibers is used
to convert the frequency components of the photons into temporal delays. Recording the arrival
times of the photons with free-running single photon detectors then enables the reconstruction of the
JSI. However, such technique uses detection in the single photon regime, and single photon detectors
suffer from low detection efficiencies and poor temporal resolution.
The stimulated counterpart of SPDC is DFG (see chapter 1), which can be described with a
classical treatment: two beams illuminate the medium (the pump beam and the seed beam), this
creating a third beam (idler) with energy and momentum characteristics determined by the conser-
vation equations. In 2013, Liscidini and Sipe [128] showed theoretically that, as the pump spectrum
and the phase-matching resonance are identical in the spontaneous and the stimulated processes, the
biphoton wavefunction serves as a response function for the generation of the idler field. They iden-
tified a relation between the number of photon pairs generated by parametric fluorescence, through
spontaneous parametric down-conversion (SPDC), and the number of photons generated through
the corresponding stimulated process (DFG). The average number of idler photons generated in a
spectral window δωi around ωi is
D E
â† (ωi ) â(ωi ) δωi ≈ 2 |B(ωs )|2 |β|2 |φ(ωs , ωi )|2 δωs δωi , (4.9)
B(ωs )
with |B(ωs )|2 the average number of photons in the stimulating field, and |β|2 a normalization
constant giving the probability that a pair is emitted by SPDC. The number of stimulated idler
photons is then proportional to the intensity of the signal field, which means that the field produced
via DFG can be intense enough to be detected and characterized with classical detectors (i.e. other
detectors that SPADs), and can still give information about the JSI as long as the seeded field does
not deplete the pump beam.
On the basis of this general result, they showed that the characterization of SPDC sources of
two-photon states in a given system can be performed solely by studying DFG, and hence classical
fields. They call this technique stimulated emission tomography (SET).
The validation of this technique was first demonstrated experimentally by Eckstein et al. in
AlGaAs waveguides [129], and then by Silverstone et al. in Si [130].
Experimental setup and results
PC
cw laser OSA
λ Waveguide
DM PC
P F
PBS
Powermeter
FC FC
HM Temperature
controller
M
FC P
FC
Ti:Sa laser
PC λ
Figure 4.14 – Sketch of the experimental setup for the JSI measurement. P: polarizer; FC: fiber
collimator; PC: polarization controller; HM: holographic mask; M: mirror; DM: dichroic mirror; F:
bulk high-pass filter with cut-off wavelength at 1.4 µm; PBS: polarized beam splitter.
For this experiment, we used the dry etched sample (wafer ABQ71), which has the highest SPDC
efficiency. The experimental setup is shown in figure 4.14. As for the coincidence setup, a continuous
wave laser (Ti:Sa M squared), having a spectral width of less than 25 kHz, at 765.915 nm is used
to excite the Bragg pump mode of the sample. Then a second tunable infrared continuous wave
laser (T unics) at around 1.53 µm is also injected to provide the seed beam. A polarizer (P) and
a polarization controller (PC) allow to select between the T E or T M polarization for the seed.
The transmitted pump is cut by a low-pass filter (F) having a cut-off wavelength at 1.4 µm. The
transmitted part of the seed and the generated idler beam (which propagate in the same direction
but with orthogonal polarization) are collected in fiber using a microscope objective and a fiber
collimator (FC). The orthogonal polarizations of the two beams allows to separate them (PBS). The
seed beam is sent to a powermeter, while the idler beam is sent to an optical spectrum analyzer (OSA
- Y okogawa6730C). This instrument can attain a resolution of 20 pm with sufficient sensitivity to
(0)
detect idler photons in the telecom band. Then, for a fixed seed wavelength λs , the DFG emission
is analyzed with the OSA by measuring the optical power at different idler wavelengths λi . Hence a
(0)
slice |φ(λs , λi )|2 of the JSI is reconstructed. By sweeping the seed laser, the complete JSI can be
determined within the tunability range of the seed.
4.4. JOINT SPECTRAL AMPLITUDE AND INTENSITY 67
A partial reconstruction of the JSI is shown in figure 4.15. The modulations due to the Fabry-
Perot of the signal and idler photons are clearly visible in the simulation and the experiment. The
jumps observed in the experimental results are due to mode hopping of the T unics laser, as can
be seen on figure 4.16. There is very good agreement between the numerical simulation of the
spontaneous process and the results of the stimulated experiment, with a higher resolution than that
achievable with single-photon detectors.
1526.8 1526.8
1526.6 1526.6
1526.4 1526.4
a) b)
1526.2 1526.2
1536.8 1537.0 1537.2 1537.4 1536.8 1537.0 1537.2 1537.4
Seed wavelength (nm) Seed wavelength (nm)

Figure 4.15 – Comparison between a) the numerical simulation, and b) the experimental JSI recon-
structed with the DFG-based method. The jumps on the numerical simulation are the modulations
due to the Fabry-Perot of the idler and signal frequencies.
1537.4
Laser real (nm)
1537.2
1537.0
1536.8
1536.8 1537.0 1537.2 1537.4
Laser request (nm)

Figure 4.16 – Spectrum of the used T unics laser, showing mode hopping.
Chapter 5
Characterization of the quantum state

emitted by a passive source
This chapter describes the characterization of the biphoton state emitted by a passive sample (wafer
EBW002). For these experiments I had the opportunity to work in the laboratory of the ‘Group
of Applied Physics-Quantum Technologies’ in Geneva, which is one of the leader groups in quan-
tum communications. In our experiments, we have demonstrated the indistinguishability and the
energy-time entanglement in the same device emitting photon-pairs at telecom wavelength and room
temperature.
For these experiments, a 5.5 µm wide and 5 µm deep waveguide defined by wet etching was used.
The optical propagation losses in the telecom range are 0.3 and 0.5 cm−1 for the T E00 and T M00
modes, respectively. The identification of the optimal working condition was carried out with the
same procedure described in the previous chapter. The sample temperature was stabilized at 20.1 ◦ C
to generate paired photons at 1566 nm.
5.1 Indistinguishability
As described in chapter 1, the indistinguishability of two photons is demonstrated by a two-photon
interferometer (or HOM experiment). This section describes the setup implementation and the
experimental results.
Experimental setup
In this experiment, as presented in figure 5.1, the light beam of a continuous wave laser (Ti:Sa),
after spatial shaping in the Bragg mode by a holographic mask (HM) and polarization selection by
a polarizer (P), is coupled into the waveguide via a 63X microscope objective. The transmitted
beam is collected with a 40X objective, and sent to a fiber coupler (FC) after filtering the residual
pump beam (F: low-pass filter with a cutoff wavelength at 1.5 µm). Then the photons of the pair
are filtered by a fibered tunable Bragg grating (FBG), and separated by a polarized beam splitter
(PBS) coupled to a polarization controller (PC). The polarizations of the two photons are set to be
the same by means of fibered polarization controllers. Finally the two photons are recombined on
a 50/50 beam splitter (BS) before being sent to the single-photon detectors (D - idQ230 [124]) and
a time-to-digital converter (&) allows time-correlation measurement. The variation of the relative
arrival time of the two photons is done through an optical delay line inserted on one of the two arms.
The pump power coupled inside the waveguide (625 µW) and wavelength (around 783 nm) are set
to maximize the coincidence counts.
69
70 CHARACTERIZATION OF THE QUANTUM STATE EMITTED BY A PASSIVE SOURCE
Waveguide
DM F
FC FBG
HM Temperature
controller
M
FC P
FC BS 50/50 PC
Ti:Sa laser D PBS
λ &
PC
optical delay line HOM

Figure 5.1 – Sketch of the experimental setup for the HOM interference. FC: fiber collimator; P:
polarizer; HM: holographic mask; M: mirror; DM: dichroic mirror; PBS: polarized beam splitter;
BS 50 / 50: beam splitter 50/50; PC: polarization controller; F: low-pass filter; FBG: tunable fiber
Bragg grating filter; D: detector; &: time-to-digital converter.
Filtering stage
700
600
Transmitted Power (µ W)
500
400
300
200
100
0
1535 1540 1545 1550 1555 1560 1565 1570 1575
Wavelength (nm)
Figure 5.2 – Transmission spectrum around 1.554 µm of the tunable filtered Bragg grating. The full
width at half maximum is 10.8 nm.
The shape of the HOM dip is given by the convolution of the wave packets of the two photons arriving
at the 50/50 beam splitter (BS). When the signal and idler photons are filtered, the coherence length
of the photons and consequently the dip-shape are determined by the filter. Here a flat-top FBG
filter is used (as can be seen in its transmission spectrum in figure 5.2); its Fourier transform in time
space is a cardinal sine function. Hence the full width at half maximum (FWHM) of the dip, which
gives the coherence length of the photons, should be equal to that of the filter, and the coincidence
rate as a function of the photons arrival on the 50/50 BS should be given by the expression
δλc

Nc (δt) = A 1 − V × sinc 2πδt (5.1)
λ2
with Nc the coincidence rate as a function of the time delay δt between the two photons, A the
coincidence rate apart from the bunching region, V the visibility (giving the quality of the indistin-
guishability), and δλ the FWHM of the spectral intensity.
5.1. INDISTINGUISHABILITY 71
Reflectivity of the facets
Due to the index contrast between AlGaAs and air, the modal reflectivity for the T E00 and T M00
modes at 1.56 µm is 24%; this directly affects the maximal visibility of the two-photon interference.
Indeed, coincidence events can be due not only to two photons directly transmitted by the facets,
but also to one photon directly transmitted and one photon having experienced two reflections (or
any multiple of two) before leaving the waveguide. In latter case, the path difference between the
two photons is different from that in the former case: one of the photons exits with a delay (∆τ )
corresponding to two times the propagation time through the waveguide (or to a number of even
reflections multiplied by the propagation times through the waveguide). This temporal delay (∼
43 ps) is smaller than the detector jitter (around 200 ps); hence the coincidence event will not be
distinguished from one in which the two photons are directly transmitted. But the temporal delay is
larger than the photon temporal coherence; therefore these photons do not interfere [131], and there
will be a reduction in visibility. Table 5.1 shows the possible situations of photon emission from the
device due to facets’ reflectivity.
Relative Measured
Photons path Delay
probability coincidence
1 1 0
R 0
R 0
R2 1 δτ
R2 0
R2 1 −δτ
... ... ... ...
Table 5.1 – Possible events due to the effet of the facets reflectivities on the photon pairs generated
within the waveguide. We have considered interference terms up to the second order.
Taking into account this effect, the maximum achievable visibility is given by:
1
Vmax = = 90.5% (5.2)
R2
1+ (ηT M + ηT E )
1−R
where ηT M and ηT E correspond to the transmission efficiencies for the two polarization modes in the
sample, and R = 24%.
Search of the zero of the interferometer
Waveguide
DM F
FBG
HM Temperature
controller
M
FC P
Ti:Sa laser BS 50/50

LED Powermeter PBS
PC λ
optical delay line
Figure 5.3 – Sketch of the experimental setup for the search of the zero of the interferometer.
In order to equilibrate the two arms of the interferometer, we used a low temporal coherence light
source. We have then injected a light-emitting diode (LED) through the entire apparatus (first
through the waveguide, before passing through the interferometer) and observed the interferogram,
recorded in a powermeter placed after the 50/50 BS, (figure 5.3), as a function of the optical delay
path. The result is shown on figure 5.4; we observe pronounced interference fringes around 193.75 ps,
corresponding to the zero of the interferometer. There are also secondary (localized at 150.5 ps and
237.19 ps) and tertiary peaks (localized at 107.19 ps), due to the facets’ reflectivity. The positions of
the side peaks are in good agreement with the length of the waveguide and the discussion done in the
previous section. Once the zero of the interferometer is identified, we come back to the experimental
setup shown in figure 5.1.
0.80 1.00
0.70 0.90
0.80
intensity (a.u.)
intensity (a.u.)
0.60
0.70
0.50 0.60
0.50
0.40
0.40
0.30 0.30
0.20 0.20
0 50 100 150 200 250 300 350 400 186 188 190 192 194 196 198 200 202 204
delay (ps) delay (ps)
(a) Complete Spectrum (b) Zoom around the central high peak
Figure 5.4 – Interference fringes observed with a powermeter at the output of the interferometer
injected with a LED when scanning the optical delay line.
Experimental results
Figure 5.5 shows the results of the HOM experiment. As expected, a dip in the coincidence counts
between the two detectors is observed. The uncertainty associated with each point has been calculated
using standard squared root deviation. The curve corresponds to the fit of the data with eq. 5.1,
using V and δλ as fitting parameters: the obtained net (raw)1 visibility is Vnet(raw) = 89.0 ± 2.8%
1
In the net visibility, the noise coincidences have been removed.
5.1. INDISTINGUISHABILITY 73
(86.1 ± 2.7%), which is in excellent agreement with the maximum visibility of 90.5% calculated in
eq. 5.2. Note that standard telecom anti-reflection coating on the facets of AlGaAs waveguides would
allow to reach transmittivities of almost 100%; this kind of treatment applied to our device would
therefore increase the indistiguishability of the emitted photons. The fit allows also to recover the
coherence length of the photons; we obtain ∆λ = 10.7±0.2 nm, which is in very good agreement with
the FWHM of the FBG filter (10.8 nm). Moreover, the cardinal sine features due to the rectangular
filter are clearly visible.
Coincidence counts (/3min)
600
coincidences
noise
500
400
300
200
100
0
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0
Temporal delay (ps)
Figure 5.5 – Results of the HOM experiments. The line shows the fit of the data with eq. 5.1. The
obtained net visibility is Vnet = 89.0 ± 2.8%.
Interpretation of our results in terms of the chronocyclic Wigner function

The characterization and engineering of the state produced by our source involve variables with dis-
crete (e.g. polarization) and continuous (e.g. frequency-time) spectrum. While for discrete variables
the density matrix is the most suited tool, for continuous variables a more intuitive description is
provided by the Wigner function [132]. Indeed, this is the formalism adopted by the quantum optics
community to represent the quadratures of the field [133] and by the ultra-fast optics community to
describe the time and frequency properties of ultra short pulses of light [?]. The Wigner function
is a real-valued quasi-probability distribution representing the state (/the pulse) as a function of
conjugated variables X and P (/ω and t). It is useful to provide a clear visual picture of squeezing
and chirping . The Wigner function can exhibit negative values, indicating non classical properties
of a state, for example as is the case of a Schrödinger cat-like state. Nevertheless by integrating the
Wigner function along one quadrature, the marginal over the conjugated variable is retrieved, which
is a density probability. For ultra-short pulses, this would be the spectrum or the temporal profile
of the pulse.
In the perspective of characterizing and engineering the state produced by our source, we have
developed a tool in collaboration with the theoreticians of the team to reconstruct the Wigner function
of a state produced by SPDC in the frequency-time space: in 2013 Douce et al. [134] showed that the
interferogram measured in a modified HOM experiment is directly proportional to the chronocyclic
Wigner function, which is an alternative description to the joint spectral amplitude φ of the state,
defined in chapter 1 (section 1.1.3). Indeed, it is possible to switch from one description to the other
via the relation [135]:
1 ωs0 ω0 ω0 ω0
ZZ
0 0
W (ωs , ts , ωi , ti ) = dωs0 dωi0 e−i(ωs ts +ωi ti ) φ(ωs + , ωi + i )φ∗ (ωs − s , ωi − i ) (5.3)
2π 2 2 2 2
in which ωs and ωi are the signal and idler photon frequencies, and ts and ti are their conjugate
variables representing the arrival times of the photons on the 50/50 BS.
As we have seen in chapter 1, the HOM experiment is a two-photon interference experiment,
revealing the quantum nature of photon pairs. In the case of indistinguishable photons, a dip is
observed at the output; if the two photons have different arrival times ts and ti , a delay τ = (ts − ti )
must be applied on the idler arm to get a dip of full visibility. In the same manner, if the photons
are not degenerate, a frequency shift Ω = ωs − ωi must be applied to compensate this detuning
and restore the indistinguishability. Ref [134] shows that a HOM experiment, in which these two
parameters can be set independently, allows reconstruction of a subset W− (ωs − ωi , ts − ti ) of the
full chronocyclic Wigner function (i.e. the part which depends on the arrival time and the frequency
differences between the two photons) (figure 5.6). The coincidence probability Pc (Ω, τ ) is then given
by [127, 136]
1 1
Pc (Ω = ωs − ωi , τ = ts − ti ) = − W− (ωs − ωi , ts − ti ) (5.4)
2 2
with
ts − ti
ZZ
W− (ωs − ωi , ) = d(ωs + ωi )d(ts + ti )W (ωs , ωi , ts , ti ) (5.5)
2
δω δω ∗ δω δω −iδω(ts −ti )/2
ZZ
= d(ωs + ωi )dδωφ(ωs + , ωi − )φ (ωs − , ωi + )e .
4 4 4 4
Figure 5.6 – Modified HOM experiment. Displacement in time and frequency are realized in each
arm of the interferometer, allowing to reconstruct the W− part of the Wigner function.
The Wigner function corresponding to the state generated by our device in the previous exper-
imental conditions (10.8 nm spectral filtering, 24% facet reflectivity) is shown in figure 5.7. The
secondary (and even the tertiary) interferences due to the facet reflectivity are clearly visible. A
standard two-photon interference experiment, as the one we have performed, collects the data cor-
responding to a slice of this function for zero frequency detuning (Ω = 0) using the relation given
by eq. 5.4. Figure 5.8 shows the numerical data presented in figure 5.7 along the line Ω = 0. The
red points correspond to the experimental data; we observe an excellent agreement between the
experiment and the Wigner function reconstructed from the JSA. This validation of the method,
proposed by Douce et al., is a first step towards a complete characterization of the state generated
by the devices developed in our team and by the community in general. Indeed, negative values
of the biphoton Wigner function is a sufficient condition to prove the presence of non-Gaussian en-
tanglement in the system; this reconstruction technique could thus constitute an interesting tool to
engineer quantum states.
5.2. ENERGY-TIME ENTANGLEMENT 75
1.0
10
0.8
5
0.6
(nm)
0 0.4
0.2
-5
0.0
-10
-0.2
-100 -50 0 50 100
(ps)
Figure 5.7 – Simulated chronocyclic Wigner function of the state generated by our device in the
experimental conditions given in the previous section (filter FWHM: 10.8 nm; facets’ reflectivity:
24%). The Ω axis is given in units of Λ = − 8πc
ω2
Ω.
p
Normalized coincidence rate
Normalized coincidence rate
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2
0.2
0.1
0.1
0.0
0.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5
-100 -50 0 50 100
(ps)
(ps)
(a) Slice at Ω = 0 (b) Zoom around τ = 0
Figure 5.8 – Slice of the simulated chronocyclic Wigner function in figure 5.7 at zero frequency
detuning. The points in figure (b) represent the experimental data.
5.2 Energy-time entanglement

Entangled photons are widely used in quantum communication. For such applications, it is essential
to have sources able to produce high quality entangled state (ideally Bell states). Even if, as we have
seen, the first experiments were done on entangled states in polarization, energy-time entanglement
is a very convenient format of entanglement, because it can be easily manipulated with integrated
circuits, and can be preserved over long distances in standard optical fibers.
Working conditions
As discussed in chapter 1, there are three essential conditions to be fulfilled in a Franson-type

experiment:
1 3
z }| { z }| {
∆t τc , τdet ∆T τp (5.6)
| {z }
2
1. ∆t τc : The difference between the two interferometers ∆t has to be smaller than the
coherence time τc of the signal and idler photons. In our case, the coherence time of the
generated photons is given by the FBG filter (τc = 0.7 ps as shown in figure 5.5). Since making
two interferometers this close is not simple, we chose to send the two photons into the same
interferometer before separating them. This configuration, which we called ‘Franson Replié’
[137], avoids the problem of the interferometers mismatch.
2. τc , τdet ∆T : The coherence time of the two generated photons τc has to be much smaller than
the time difference ∆T between the two arms of the interferometer; this is necessary to avoid
one photon interference. Moreover, the jitter of the detector τdet , has also to be much smaller
than ∆T , to suppress cases in which the two photons have taken different arms. Experimentally,
it is the jitter of the detector, which is around 200 ps, that defines the minimal interferometer
size. We selected the path difference to be such that ∆T = ∆L/c = 2.5 ns.
3. ∆T τp : To have interference, the coherence time of the pair has to be larger than the
time corresponding to the length difference between the two arms of the interferometer. The
coherence time of the pair is the same as that of the initial pump photons; we therefore chose
a laser T OP T ICA (DL100) as pumping source since it has a coherence time of 1 µs, allowing
to satisfy the condition.
Experimental setup
The first part of the experimental setup, sketched in figure 5.9, is the same as that discussed for
the HOM experiement (figure 5.1). A continuous pump laser (this time the T OP T ICA) is injected,
after beam shaping by the HM, into the waveguide by means of a microscope objective. The twin
photons are collected through a second microscope objective, and sent to a standard telecom fiber
after filtering the pump beam (F). The photon pairs are then filtered by the FBG before being
directed into the unbalanced interferometer. Finally, the two photons of the pair are split by a
polarized beam splitter (PBS), and sent to the two SPAD detectors (D - idQ230).
Waveguide
DM F
FC FBG
Franson FM
L
HM Temperature BS 50/50
controller
M
FC P φ shift FM
PC S
FC
Ti:Sa laser
Toptica laser D PBS
λλ &
PC
Figure 5.9 – Sketch of the experimental setup for the Franson experiment. BS 50/50: beam splitter
50/50, FM: Faraday mirror, l: long arm and s: short arm
5.2. ENERGY-TIME ENTANGLEMENT 77
Comments on the interferometer

The interferometer is entirely fibered and contains a 50/50 beam splitter. The optical path difference
between the long and short arms ∆L is such that ∆L/c = ∆T = 2.5 ns. The temperature is stabilized
(so as to have a constant phase) and a piezo actuator is used to control the relative phase φ between
the two arms.
The birefringence existing in the fiber due to asymmetries and stresses can be detrimental to the
indiscernability between the different paths of the interferometer. To counter this effect, we used
Faraday mirrors (FM) at the end of each arm. These mirrors are made up of Faraday rotators,
whose length is adjusted to turn the polarization of 45 degrees, placed just before usual mirrors.
This eliminates all birefringent effects in the interferometer by ensuring that each incident photon,
which can have any polarization state, is transmitted with an exact orthogonal polarization with
respect to its entrance polarization.
Results
0.03
Coincidences (/s)
a) b)
0.02
0.01
0.00
-10 -5 0 5 10 -5 0 5 10
Time delay (ns)
Figure 5.10 – Coincidence histograms measured at the end of the setup for two different phase settings
chosen to minimize (a) and maximize (b) the central peak.
As shown in figures 5.10.a and b, the recorded coincidence histograms have the three expected
coincidence peaks described in chapter 1. The left and right peaks correspond to a situation in
which one photon goes through the short arm (s) and the other through the long arm (l) of the
interferometer. The central peak corresponds to situations in which both photons pass through the
short arm and in which both photons pass through the long arm. So this peak results from the
quantum interference of the following post selected state :
1
|Φi = √ [|sii |sis + e−i2φ |lii |lis ] . (5.7)
2
The two values of φ for the measurements reported in figures 5.10.a and b have been chosen
to minimize and maximize the central peak height, respectively. The height of the satellite peaks
is independent of the phase. The measurement accuracy is 200 ps, dominated by the jitter of the
detectors. The total number of coincidence counts in the central peak as a function of the phase φ
is plotted in Figure 5.11. The trend is well fitted by a sinusoidal curve having a net (raw) visibility
of 95.6 ± 3.7% (91.5 ± 3.6%). If we assume that this visibility is independent on the chosen basis, we
can estimate the corresponding fidelity with the formula
1+V
F = = 97.8 ± 1.9% . (5.8)
2
This result is close to a perfectly entangled photons state, where the visibility of the interference
fringe is 100% when the phase φ is varied.
In conclusion, we have demonstrated that our device, working at room temperature, generates
photon pairs linked by a high degree of entanglement and able to interfere on a beam splitter with
high visibility. These are essential characteristics to make our device a good candidate for quantum
communication technologies.
central peak
80
Coincidences (/10min)
noise
70
60
50
40
30
20
10
0
0 5 10 15 20
Phase shift (a.u.)
Figure 5.11 – Coincidence count rate of the central peak as function of the phase φ. The internal
estimated power is 273 µW. The solid curve shows the fitting of the data, giving a visibility Vnet =
95.6 ± 3.7%.
Chapter 6
Multi-user quantum key distribution

Quantum key distribution is a core protocol for securing communications in quantum networks.
In this chapter, we exploit the possibility to generate polarization entangled photon pairs without
the use of any off-chip compensation or interferometric scheme, as well as the large bandwidth of
the emitted paired photons in the telecom band, to demonstrate QKD. With respect to real world
applications of quantum information technologies, the combination of miniature quantum resources
(here the AlGaAs source emitting broadband polarization entangled photons) with existing fiber
networks (here standard telecom components for wavelength division multiplexing) is a crucial issue.
The simple experimental setup used in this chapter implements the BBM92 QKD protocol [111],
which is the entanglement-based1 generalization of the famous BB84 protocol [1]. In this protocol,
two distant users, Alice and Bob, share entangled photon pairs in a Bell state generated by a source
that can be situated in either Alice’s or Bob’s site, or in between, for exemple at a server node in
a telecommunication network. The photon pairs emitted by the source are directly fed into a dense
wavelength division multiplexer (DWDM), and secret keys are distributed between several pairs of
users, separated by up to 50 km of standard optical fiber, communicating through different channels.
This work has been performed in collaboration with the Quantum Information team of the labo-
ratory ‘Traitement et Communication de l’Information’ of Telecom ParisTech. All the measurements
reported here are related to wafer EBW002.
6.1 Experimental method

Experimental setup
A schematic of the experimental setup is shown in figure 6.1. The T E Bragg mode of the sample
is excited by using the light beam of a continuous-wave (cw) laser (Ti:Sa) that is first spatially
shaped with a holographic mask (HM) and then focused to the sample through a 63X microscope
objective. Light emerging from the opposite end of the sample is collected with a 40X microscope
objective, and sent to a fiber coupler (FC) followed by a polarization controller (PC), after filtering
out the pump wavelength with a lowpass filter (F). The entangled photon pairs are then sent to four
different pairs of users through the output channels of an 8-channel dense wavelength demultiplexer
(DWDM). The photons of a given pair of users then exit the central quantum server node of the
setup and either directly enter Alice’s and Bob’s sites or first propagate through a 25 km spool of
standard single mode fiber, this leading to a 50 km total fiber link between the users. The stations
of Alice and Bob contain a polarization controller, which is required because the DWDM is a non-
polarization-maintaining device, followed by a polarization analysis setup which includes a half-wave
plate (HWP), a fibered polarization beam-splitter (PBS) and an InGaAs free-running single-photon
detector (D - idQ220). Finally, the detectors of Alice and Bob are connected to a time-to-digital
converter (&) in order to perform time-correlation measurements.
1
Note that, although QKD can be implemented with single photons or coherent light pulses [138], entanglement-
based QKD [111, 2] has attracted a great interest to achieve device independence [113], thereby removing security
loopholes due to imperfections in the components.
79
80 MULTI-USER QUANTUM KEY DISTRIBUTION
Bob 28
Bob 29
Quantum server node Waveguide
DM
DWDM 26
b
Bo
HM P F
Temperature FC
Al
Ti:Sa laser
Alice 22
Alice 21
ic
controller
e
λ
24
M
Alice 23 Bob 27
PC PC
HWP
PBS & PBS
HWP
D D
Figure 6.1 – Sketch of the multi-user entangled-based quantum key distribution setup. HM: holo-
graphic mask; P: polarizer; DM: dichroic mirror; M: mirror; F: lowpass filter; FC: fiber collimator;
DWDM: dense wavelength demultiplexer; PC: polarization controller; HWP: half-wave plate; PBS:
polarized beam splitter; D: detector; &: time-to-digital converter.
Dense wavelength demultiplexer

In fiber-optical communications, wavelength-division multiplexing (WDM) is a technique that allows
the multiplexing of a number of optical signals into a single optical fiber, or, reciprocally, their
demultiplexing into several fibers by using different wavelengths. This technique, as well as the
active optical routing, is widely used in classical telecommunications [139]. It has been shown to be
compatible with entangled photon distribution [140, 141, 142, 143, 144, 145], and hence to enable
the reduction of the required number of sources in a network. Indeed, with DWDM it is possible to
distribute simultaneously entangled photon pairs emitted in a large band to several pairs of users,
using only one source [143, 146].
The DWDM used in the experiment belongs to the 100 GHz International Telecommunication
Union (ITU) grid (100 GHz channel width and 100 GHz channel separation). The choice of an
OptoLink dielectric thin film DWDM resulted from an analysis of the performance of various demul-
tiplexer technologies with respect to their effect on the distribution of polarization entangled photon
pairs previously made by our colleagues of the Telecom laboratory [143].
Measurement
The experimental measurement is performed by placing one of the analyzers (Alice) in the basis
{H, V } and then {D, A}, and then measuring the variation of coincidences as a function of the
analysis angle of the second analyzer (Bob).
6.2 Quantum key distribution

Multi-user polarization entanglement
As seen in chapter 1, the wavelengths of the emitted photon pairs are determined by energy and
phase-matching conservation:
~ωp = ~ωi + ~ωs



 (6.1a)

n TEBragg (ωp ) ωp = n TE00 (ωi ) ωi + n TM00 (ωs ) ωs (6.1b)

n
TEBragg (ωp ) ωp = n TM00 (ωi ) ωi + n TE00 (ωs ) ωs , (6.1c)

in which ωi and ωs (with ωi ≤ ωs ) are the down-converted photon frequencies, ωp is the pump
frequency, and ni (with i = T EBragg , T E00 , T M00 ) are the effective indexes of the three guided
modes.
TE00
TE00 wavelength (nm)
TM00
TM00 wavelength (nm)
Figure 6.2 – Simulated tuning curves of the type-II phase matching at T = 20 ◦ C. The white dotted
line shows the chosen pump wavelength, and the curves on the right plot show the spectral intensity
of the paired photon modes at this wavelength. There is a clear overlap between the two spectral
emissions, leading to broad-band polarization entanglement.
The simulated tuning curves for a 2.1 mm sample pumped by a cw laser at a temperature of
20 ◦ C are shown in figure 6.2. For a given pump wavelength, the pairs emitted over the entire region
of spectral overlap between the T E00 and T M00 modes are produced in the polarization entangled
Bell state
1
|Ψ+i = √ (|Hii |V is + eiφ |V ii |His ) (6.2)
2
in which the states |Hi and |V i correspond to the polarization modes T E00 and T M00 , respectively,
and i (s) corresponds to the longer (shorter) wavelength photon of the pair. Due to the very small
birefringence (∆n ≈ 7 × 10−3 ) of the AlGaAs waveguide, the group velocity mismatch between the
photons of a pair is so small that no off-chip compensation of the walk-off is required to obtain the
state of eq. 6.2 for about 30 nm around the degeneracy wavelength. This is a key feature of the
source; it enables the direct use of the generated pairs, and therefore the easy integration of the
device in simple and robust architectures for practical applications.
Moreover, when a narrowband pump laser is used (experimentally, the laser used had a linewidth
smaller than 50 kHz), the strong large-band anti-correlation between the frequencies of the photons of
a given pair makes them exit through symmetric frequency channels with respect to the degeneracy
frequency ωp /2. Hence the energy/frequency entanglement generated by SPDC in the source is
transfered to an entanglement in optical channels of the DWDM. Consequently, the working point
of the device is set such that the phase-matching resonance corresponds to the central channel of
the DWDM used in the QKD setup. Experimentally, the targeted central channel is channel 25 of
the ITU grid, while channels 21-24 and 26-29 are sent to Alice and Bob, respectively. The optimum
working point is found at a temperature T = 19.5 ◦ C by adjusting the sample temperature with a
Peltier cooler while monitoring the signal to noise ratio and the emission wavelength.
Quantum key distribution protocol
In the BBM92 protocol (see chapter 1), Alice and Bob, upon reception of the entangled photon
pairs, both locally measure their respective photon, choosing randomly and √ independently among√the
computational basis {|Hi , |V i} or the diagonal basis {|Di = (|Hi + |V i)/ 2, |Ai = (|Hi − |V i)/ 2},
Rraw . After a basis reconciliation step Rsift and post-processing procedures, including error correction
(e and f (e)) and privacy amplification (performed through a classical communication channel) they
can extract a secret key from their coincidence measurements in identical bases Rkey .
Note that, for practical reasons, coincidences were measured separately in the eight measurement
configurations in which identical bases are chosen by Alice and Bob. The full implementation of
the random basis choice could be achieved passively by means of a 50-50 beam-splitter routing the
photon to a polarization beam-splitter aligned either with the computational or the diagonal basis,
and four single-photon detectors for each user. Only a single detector per user is required if temporal
multiplexing of the four projected polarization states can be implemented after the polarization
beam-splitters [144].
Security guarantees
Theoretically, the information security of the key distribution process is guaranteed as long as the
entanglement shared by Alice and Bob is sufficiently high. Quantitatively, from the visibility of their
coincidence measurements, Alice and Bob can estimate the quantum bit error rate (QBER) e of their
quantum communication link [147]:
1 − Vtot
e= , (6.3)
2
in which Vtot is the total entanglement visibility in both the computational and the diagonal bases:
Cmax − Cmin
Vtot = , (6.4)
Cmax + Cmin
with Cmax and Cmin the maximum and minimum number of coincidences observed in both bases
respectively.
From the value of QBER and the total rate of coincidence counts, Rraw , measured in all possible
combinations of both bases, Alice and Bob can estimate a lower bound for their secret key rate,
Rkey [148]:
1
Rkey ≥ Rraw (1 − f (e)H2 (e) − H2 (e)) , (6.5)
2
in which the 1/2 prefactor is the basis reconciliation (or sifting) factor that accounts for the fact
that only the measurements in which Alice and Bob choose the same basis can be used for the key,
which happens half of the time; f (e) corresponds to the efficiency of the error-correction code (in
general f (e) ≥ 1 and the Shannon limit is f (e) = 1); H2 is the binary entropy function, H2 (x) =
−x log2 (x)−(1−x) log2 (1−x). The bound of eq. 6.5 is derived against general coherent attacks [148],
this providing maximal security guarantees.
6.3. RESULTS 83
6.3 Results
Zero distance between Alice and Bob
QKD experiments were first performed with the entangled photons entering directly the stations
of Alice and Bob, for all four symmetric channel pairs (21-29; 22-28; 23-27; 24-26) of the DWDM,
corresponding to four different pairs of users sharing a secret key.
700
0° Basis 45° Basis
Coincidences (/3 min)
600
500
400
300
200
100
0
0 50 100 150 200 250 300 350
Angle (°)
Figure 6.3 – Visibility curve in the natural and diagonal polarization basis for ITU channel pair
23-27.
Figure 6.3 displays the complete visibility curves for the 23-27 channel pair when the Alice
measurement basis is set to the natural (0◦ ) or diagonal (45◦ ) basis, and the Bob measurement basis
varies. Figure 6.4 shows the measured coincidence histograms corresponding to the eight possible
projective measurements obtained when Alice and Bob make the same basis choice, also for the
23-27 channel pair. From these measurements, it is possible in all cases to estimate the sifted key
generation rate, Rsift = 12 Rraw , by calculating from the obtained data:
CpHH + CpHV + CpV V + CpV H + CpDD + CpDA + CpAA + CpAD

Rsift = , (6.6)
τ
in which each of the eight terms is obtained by adding together the number of coincidences measured
in the five bins corresponding to the coincidence peak in figure 6.4 accumulated over a time τ = 3 min.
It is also possible to calculate the false coincidence rate as:
C0HH + C0HV + C0V V + C0V H + C0DD + C0DA + C0AA + C0AD
Rfalse = 2 , (6.7)
τ
in which each of the terms is obtained from the mean value of the number of coincidences measured
in five bins outside the coincidence peak.
Finally, the maximum and minimum number of coincidences in both bases are given by:

Cmax = C HV + C V H + C DD + C AA (6.8a)
p p p p
 Cmin = CpHH + CpV V + CpDA + CpAD . (6.8b)
These are inserted in eq. 6.4 to calculate the total entanglement visibility Vtot , which leads to
the calculation of the QBER (eq. 6.3). Finally, taking standard values for f (e) [147, 149], eq. 6.5
Computational basis
250 250
coincidences (3 min)
200 200
150 150
100 100
50 50
0 0
-3 -2 -1 0 1 2 3 -2 -1 0 1 2 3
250 250
200 200
150 150
100 100
50 50
0 0
-3 -2 -1 0 1 2 3 -2 -1 0 1 2 3
Diagonal basis
250 250
200 200
150 150
100 100
50 50
0 0
-3 -2 -1 0 1 2 3 -2 -1 0 1 2 3
250 250
200 200
150 150
100 100
50 50
0 0
-3 -2 -1 0 1 2 3 -2 -1 0 1 2 3
Figure 6.4 – Coincidence histograms of the eight different projective measurements, corresponding
to an identical choice of basis by Alice and Bob, for channels 23 and 27 centered at 1558.98 nm and
1555.75 nm, respectively. The data were accumulated during 180 s with a sampling resolution of
164 ps.
6.3. RESULTS 85
can be used to estimate the secret key rate, Rkey . For the 23-27 channel pair, we obtain Rsift =
13.8 ± 0.3 bit/s, Vtot = 86.7% ± 1.0%, e = 0.066 ± 0.005, and Rkey = 3.28 ± 0.08 bit/s. The results
for all pairs of users are summarized in Table 6.1.
channel fibre link Vtot Rsift QBER Rkey

pair distance (%) (bit/s) (%) (bit/s)
21–29 0 km 90.0 ± 1.0 9.91 ± 0.23 5.0 ± 0.5 3.79 ± 0.10

22–28 0 km 83.5 ± 2.1 13.0 ± 0.27 8.2 ± 1.0 1.31 ± 0.09
0 km 86.7 ± 1.0 13.8 ± 0.3 6.6 ± 0.5 3.28 ± 0.08
23–27
50 km 86.1 ± 2.9 1.01 ± 0.06 6.9 ± 1.5 0.21 ± 0.13
24–26 0 km 87.9 ± 1.4 6.9 ± 0.2 6.1 ± 0.7 1.95 ± 0.08
Table 6.1 – Measured and estimated values for the QKD protocol for four different DWDM channel
pairs centered on channel 25. Errors are calculated assuming Poisson statistics for the coincidence
counts. Finite-size effects have not been considered in these results.
50 km distance between Alice and Bob

In order to assess the performance of the system over long distances, the experiments were repeated
(for the 23-27 channel pair) with a spool of 25 km standard single-mode fibre inserted after the
DWDM in both user paths, this leading to a total distance of 50 km between Alice and Bob. The
measured coincidence histograms in this case are shown in figure 6.5. Note that the FWHM of
the coincidence peak is broader than for 0 km due to fiber dispersion. The C ij terms are obtained
by adding together the number of coincidences measured in the seven bins corresponding to the
coincidence peak, accumulated over a time τ = 5 min. In this configuration, the QBER increases
slightly to e = 0.069 ± 0.015, and the secret key rate we obtain is Rkey = 0.21 ± 0.13 bit/s. These
results are also included in Table 6.1.
Note that in this long-distance case, the optimized state presents a π−shift compared to the state
of eq. 6.2: |Ψ−i = √12 (|Hii |V is − |V ii |His ). Then the terms CpDD + CpAA and CpDA + CpAD need to
be interchanged in eqs. 6.8a and b.
Long-distance performance
The theoretical model [149] for secret key generation as a function of distance is calculated with the
security bound of [148] (eq. 6.5). In particular, the error rate is given by the expression:
1
2 pfalse
+ b ptrue
e= , (6.9)
ptrue + pfalse
in which ptrue and pfalse are the probabilities of true coincidences (originating from photons of the same
pair) and false coincidences (from noise and lost photons) respectively, and b is a parameter accounting
for a systematic polarization measurement error. The true and false coincidence probabilities in the
case of a SPDC source in the CW pumping regime can be expressed as [150]:
ptrue = µ η 2 (6.10)
2 2 2
pfalse = µ η + 8 µ η d + 16 d , (6.11)
in which µ is the mean number of photons per pulse (in the cw regime, we can define an effective
pulse duration and an effective repetition rate, frep ); d = 4.4 × 10−6 is the noise count probability
Computational basis
25 25
Coincidences (5 min)
20 20
15 15
10 10
5 5
0 0
-3 -2 -1 0 1 2 3 -2 -1 0 1 2 3
25 25
20 20
15 15
10 10
5 5
0 0
-3 -2 -1 0 1 2 3 -2 -1 0 1 2 3
Diagonal basis
25 25
20 20
15 15
10 10
5 5
0 0
-3 -2 -1 0 1 2 3 -2 -1 0 1 2 3
25 25
20 20
15 15
10 10
5 5
0 0
-3 -2 -1 0 1 2 3 -2 -1 0 1 2 3
Figure 6.5 – Coincidence histograms of the eight different projective measurements, corresponding to
an identical choice of basis by Alice and Bob, for the 23-27 channel pair, for a 50 km distance between
Alice and Bob. The data were accumulated during 300 s with a sampling resolution of 164 ps.
6.4. PERSPECTIVES 87
in the single-photon detector coincidence window, which combines the dark count probability of the
detector (≈ 2 × 10−6 ) and the spurious light count probability (luminescence noise and unfiltered
ambient light); and η is the global detection efficiency of the setup for each channel:
αL
η = ηcoll ηdet 10− 10 , (6.12)
in which ηcoll = 5% is the collection efficiency of the setup, ηdet = 20% is the detection efficiency
of the single-photon detectors, and α = 0.22 dB/km is the propagation loss coefficient of each fibre
spool of length L.
From the sifted and false coincidence rates (eqs. 6.6 and 6.7), measured for zero distance and for
L = 25 km, and using Rsift = ptrue frep and Rfalse = pfalse frep , we estimated the SPDC generation
probability to be µ = 0.0035 for an effective repetition rate frep = 78 MHz. Finally, from the visibility
Vtot measured at zero distance, we estimated a polarization measurement error b = 0.06, using the
b) ptrue
relation Vtot = (1−2
ptrue +pfalse (from eqs. 6.3 and 6.9).
The ultimate long-distance performance of the QKD system is illustrated in figure 6.6, in which
the two experimental points of the secret key generation rate and the QBER as a function of the
distance for the 23-27 channel pair are shown, together with the fit derived from the theoretical
model taking into account all experimental parameters. We see that secret key generation is possible
for a maximum distance of around 80 km under our experimental conditions.
Figure 6.6 – Evolution of the secret key generation rate and the QBER as a function of the distance
between Alice and Bob for the 23-27 channel pair. The theoretical fits are based on the model of
[149], and take into account the quantum efficiency and dark counts of the single-photon detectors,
the collection efficiency of the setup, standard loss of single-mode fibers, and a baseline polarization
measurement error.
6.4 Perspectives
These first multi-user entanglement-based QKD results obtained with our semiconductor chip source
are very promising for quantum communication protocol implementations based on particularly sim-
ple experimental setup configurations. Further progress is possible in several ways:
• The collection efficiency is presently limited by the semiconductor facet reflectivity (25%) and
the mode mismatch between the semiconductor waveguide and the single-mode fiber (25%
coupling efficiency). The use of an antireflection coating and of a laser-diode-to-fiber-coupler

for the coupling of the emitted photons to the fiber could improve the collection efficiency by
a factor 4.2, reaching 21% instead of the current value of 5%.
• The use of commercially available superconducting detectors with a quantum efficiency of 87%
and a dark count rate of 10 count/s would increase the detection efficiency by a factor 4.3,
from 20% to 87%, and remove the influence of dark counts, leaving only spurious light related
to false coincidences.
• The polarization-maintaining fibers of the polarization analysis setups at Alice’s and Bob’s
stations introduced polarization-mode dispersion (PMD), which limited the maximum possible
visibility. They consist of a fibered half-wave plate and a polarization beam-splitter, with
Lf ≈ 3 m, of panda-style polarization-maintaining fibers. This introduces a differential group
λL
delay τPMD = cLbf ≈ 3 ps (with Lb = 5 mm the polarisation beat length of the panda fiber).
The temporal overlap between two orthogonal polarization modes was thus limited to
Z
η= dtfA (t)fB (τPMD − t) ≈ 84%
in which fA (t) and fB (t) are the normalized temporal transmission functions of the DWDM
channels. As a result, the visibility of the measured entangled state is upper bounded by
max = 1+η ≈ 92% [143]. Replacing these fibers by standard single-mode fibers could reduce
Vtot 2
the polarization error rate b from the current value of 6% to 2.5%.
Taking into account all these factors, the performance of the setup could reach a secret key generation
rate of 2.8 kbit/s in the same pumping conditions at zero distance, and could allow key distribution at
a distance of up to 230 km between users. The setup can also be readily adapted to perform random
basis choice, as described previously, and active optical switches for the reconfigurable selection of
user pairs. Moreover, even though the experiment was performed on only 4 pairs of user, figure 6.2
shows that the device could be used over about 30 nm around the degeneracy, this resulting in 16
pairs of simultaneous users.
Finally, the results on the dry etched sample presented in chapter 4, showed an improvement
of the SNR by a factor greater than 10 thanks to the increase in the parametric down-conversion
efficiency. Such samples should allow a higher secret key generation rate under the same experimental
conditions.
Chapter 7
An electrically injected photon-pair

source at room temperature
In chapter 4 we have presented our results on the nonlinear properties of the waveguide, and in chapter
3 those on the laser diode characteristics. We have seen that the nonlinear waveguide converts two
photons at 1.56 µm in the T E00 and T M00 modes, into a T E polarized photon at 780 nm in the
fundamental Bragg mode by SHG. We have also seen that under electrical pumping, the waveguide
emits laser radiation in the fundamental T E polarized Bragg mode. In this chapter we present our
work on the generation of photon pair sources under electrical injection exploiting internal parametric
down-conversion within the diode laser; the key point is to obtain the laser emission to the wavelength
corresponding to the phase-matching for SPDC.
For these experiments, we used a 6 µm wide and 2 µm deep waveguide defined by wet etching
(wafer C4T46).
7.1 Research of the working conditions

Coarse tuning
The fine tuning parameter for both phase-matching and laser emission will be the temperature.
Figure 7.1 shows the evolution of the SH peak wavelength (crosses) and the laser wavelength (dashed
line) as a function of the temperature. The shaded area represents the FWHM of the phase-matching
bandwidth. We observe that these tunability curves intersect in the explored temperature range at
around 25 ◦ C.
Experimental setup
In order to confirm the existence of a working region of the device and to demonstrate the emission
of photon pairs in the telecom range, time-correlation measurements were performed under electrical
injection. The setup used is shown in figure 7.2. The waveguide is electrically injected with current
pulses. The generated photons at telecom wavelengths are collected through a 40X microscope
objective and focused into a fiber via a fiber coupler (FC). The residual pump beam is removed using
a bulk filter (F) with a cut-off frequency at 1.5 µm. Then the two photons of the pair pass through a
tunable interferential filter (FBG) and are deterministically separated by a fibered polarizing beam
splitter (PBS) coupled to a polarization controller (PC). The detectors (D) are two gated single-
photon avalanche photodiodes (idQ201). A delay generator synchronizes the pump pulses with the
detector gates. A time-to-digital converter (&) is used to analyze the time correlations between
detected photons. Finally, an oscilloscope (Osc.) allows to verify the electrical properties of the
device while operating.
89
90 AN ELECTRICALLY INJECTED PHOTON-PAIR SOURCE AT ROOM TEMPERATURE
788
Wavelength (nm)
787
786
785
784
783
15 20 25 30 35
Temperature (°C)
Figure 7.1 – Research of the working condition: the crosses represent the data for the SH experiment,
with a linear fit in red. The shaded area indicates the FWHM of the phase-matching bandwidth.
The dashed line shows the variation of the quantum well bandgap, inferred from the laser wavelength
versus the temperature curve.
Waveguide
F
FC FBG
Osc.
Temperature
controller
Trigg.
(+) (-)
PC
Pulsed power
D PBS
supply &
Trigger
Figure 7.2 – Experimental setup for the detection of coincidences in electrically injected devices. F:
filter with cut-off wavelength at 1.5 µm; FC: fiber coupler; FBG: fibered tunable Bragg grating; PC:
polarization controller; PBS: polarized beam splitter; D: single-photon detectors; &: time-to-digital
converter; Osc: oscilloscope.
First results
Figure 7.3 shows a histogram of the detection time delays between T E and T M polarized photons
at 25 ◦ C using an interferential filter of spectral bandwidth 12 nm. The sharp peak emerging at zero
delay from the background is a clear evidence of pair production. The full width at half maximum
of the peak is due to the jitter of the detectors, and the noise is due to both the dark counts of
the detectors and the optical noise coming from the luminescence of the diode. The detected SPDC
signal is optimized by tuning the temperature.
The Signal to Noise Ratio (SNR) is evaluated by taking the number of true coincidences within
the FWHM of the peak over the background signal on the same time window. The data presented
in figure 7.3 give a SNR of 5.39.
7.2. COINCIDENCES PROBABILITY CALCULATION 91
0.9

0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-50 -40 -30 -20 -10 0 10 20 30 40 50
Time delay (ns)
Figure 7.3 – Time-correlation histogram of the T E and T M photons at 1570 nm. The sample is
electrically injected with current pulses having an intensity of 709 mA, a duration of 120 ns and a
repetition rate of 5 kHz. The data were collected over 600 s with a sampling resolution of 162 ps.
The detector quantum efficiency was set to 25%.
7.2 Coincidences probability calculation
Two important figures of merit of a photon-pair source for quantum information are its brightness
and the signal to noise ratio in the coincidence measurement; they depend both on the intrinsic
properties of the source and on the setup used to characterize it. In this section we analyze the
different parameters of our system affecting the probability of obtaining a true coincidence and the
SNR. This analysis helped us in the optimization of our measurements.
7.2.1 Transmission of the generated photons
The probability of getting a true coincidences on the detectors as well as the SNR depend on the
transmission probability of the photons through the setup.
Spectral filtering
The filters on both the T E and T M photon paths are approximated by a rectangular gate function1
(ideal filter); the transmission is considered zero outside the filter pass band and constant inside (see
1
The transmission spectrum of our experiment filter is close to such an idler filter, as can be seen in figure 2 of
chapter 5.
figure 7.4). They are centered at νT E/T M with a bandpass ∆νTfilter filter
E/T M and a transmission TT E/T M .
∆νTfilter
E/T M
-
T = TTfilter
E/T M
T =0
νT E/T M
Figure 7.4 – Spectral transmission of an ideal filter.
We define x = min(∆νTfilter
E /∆ν
SPDC , ∆ν filter /∆ν SPDC ) the filter spectral bandwidth over the
TM
emission spectral bandwidth of the SPDC at full width at half maximum. In our case, we work at
degeneracy to ensure that both photons of the pair stay in the spectral range where the InGaAs
single-photon detectors are sensitive. Consequently, in our experiment, the two filters are one unique
filter.
Losses
The total losses, occurring from the emission of the pair inside the waveguide to its detection, have
to be taken into account in the calculation of the total transmission.
The waveguide transmission TTwg E/T M for the two polarization modes is estimated taking into account:
• the propagation and absorption losses at 1.56 µm inside the waveguide, αT E/T M ;
• the transmission through the waveguide facet TTfacet

E/T M .
The setup transmission TTsetup

E/T M is estimated taking into account:
• the objective transmission T obj ;
• the filter transmission TTfilter

E/T M ;
• the free space losses (propagation, mirrors);
• the coupling into fiber via the fiber collimator;
• the connections between the fibers;
• the fiber losses;
• the PBS losses (due to non-perfect transmission/reflection).
Detectors
The single-photon detectors are two InGaAs avalanche photodiodes with quantum efficiency η idqA/B
idqA/B
and dark count probability pdark . The dark counts are due to random thermally-generated carriers
that fire the avalanche process.
Total transmission and detection

We define TTtot tot
E (resp. TT M ), the total transmission probability and detection efficiency of a T E
photon (resp. T M ), i.e. the probability for a photon created by SPDC in the waveguide to be
detected by the single-photon detector. This is given by
wg setup
TTtot
E/T M = TT E/T M × TT E/T M × η
idqA/B
. (7.1)
In a semiconductor photon-pair source, there can be noise due to the luminescence of the device.
This is detrimental to the signal to noise ratio and therefore the detection probability of these photons
has to be taken into account in the calculation. The luminescence photons are not polarized. There
is then a 50% probability that they take one path or the other at the polarized beam splitter. Hence
the total transmission probability of a luminescence photon, Tlum tot , is given by
tot TTtot T tot

Tlum = Tlum × η idq ≈ E
≈ TM . (7.2)
2 2
7.2.2 Coincidence probabilities

There exist three types of coincidences that can occur between the events detected by the two
detectors A and B:
• true coincidences: the two detected photons come from a pair generated by SPDC;
• false coincidences due to double pairs: the two detected photons come from two different,
incoherent pairs emitted within the same pump impulsion;
• noise coincidences: these are linked to the detector dark counts and the optical noise (mainly
due to the luminescence of the diode). They include the coincidences between a count of noise
(optical or dark count) and a photon from a pair, and coincidences between two counts of
noises.
Probability of true coincidences

Type II phase-matching allows a deterministic separation of the photons of the pair at the PBS; the
two photons of the pair are separated and sent each to one detector. Moreover, energy conservation
implies that if a photon has a wavelength allowing it to pass through the filter, its paired photon also
passes through it.
We define the probability of the following events:
→ creation of a pair inside the waveguide: p0
→ transmission of the pair through the filter bandpass: p0 x
→ detection of the T E photon of the pair in detector A: pidqA tot

T E = p0 xTT E
→ detection of the T M photon of the pair in detector B: pidqB tot

T M = p0 xTT M
→ detection of a true coincidence2 :

pidq tot tot
CT = p0 xTT E TT M (7.3)
2
This probability is given without including the probability of having true coincidences coming from multiple pair
creation, which is negligible under our pumping condition.
Probability of false coincidences due to double pairs

The probabilistic nature of a parametric process implies that for each pulse we have a certain prob-
ability to create 0 pairs, 1 pair, 2 pairs, and so on. Here we limit the discussion to the impact of
double pairs generation on the accidental coincidences.
→ creation of two pairs inside the waveguide having wavelengths passing through the filter band-
pass: (p0 x)2
→ detection of a coincidence from two photons belonging to different pairs:

1
pidq 2 tot tot tot tot
CA = (p0 x) TT E TT M (2 − TT E )(2 − TT M ) (7.4)
2
This probability is very low with respect to the probabilities of true and noise coincidences (it is of
the order of 10−3 times less than the probability of having a true coincidence). It will therefore be
neglected in the following.
Probability of noise coincidences

→ creation of a luminescence photon in the waveguide: plum

idqA/B
→ detection of this photon on one of the two detectors: plum tot
= plum xTlumA/B
idqA/B
→ detection of a dark count: pdark
→ detection of a pair due to noise:
pidq idqA idqA idqB idqB idqB idqB idqA idqA idqA idqB
CN = (plum + pdark ) × (plum + pdark ) + (plum + pdark ) × pT E + (plum + pdark ) × pT M
| {z } | {z }
two 0 noise0 counts one 0 noise0 count and one 0 true0 count
(7.5)
7.2.3 Estimating the fidelity to a Bell state from the SNR

In the previous sections, we have identified the different parameters that enter in the calculation of
the true coincidence probability and the SNR. These parameters are p0 , plum , x, T tot , and pdark . The
production rate of true pairs p0 is adjustable only by varying the pump power. But an increase in
the pump power also increases plum and the probability of double pairs, which are both detrimental
for the SNR, and hence for the fidelity. Indeed, assuming that the source emits a Werner state [151],
which is the usual assumption when the noise is not polarized as in our case, the associated density
matrix is ρ̂W = P |φ+ i hφ+ | + (1 − P )/4 × 1 with P = SNR/(2 + SNR). This leads to a maximal
fidelity estimate F to the Bell state |φ+ i that can be produced in the device
SNR
1 + 3P 1 + 3 2+SNR
F = = . (7.6)
4 4
Filter bandwidth influence

The device emits photon pairs on a very large spectral bandwidth (120 nm) at degeneracy. But energy
conservation ensures that, if one photon of the pair in emitted inside the filter bandwidth, the second
is also emitted inside it. This explains why the parameter x, which is equal to the filter spectral
bandwidth over the FWHM of the emission spectral bandwidth of the SPDC, appears only once in
eq. 7.3. On the other hand, the luminescence of the diode is also broad band. The photons due to
luminescence, however, are uncorrelated, which means that the filtering of one of these photons does
not affect the filtering of the others. Consequently there will be terms in eq. 7.5 in which parameter
x will be squared (when both detections are due to luminescence or photons from one pair) and
terms in which x appear once (when one of the two counts is a dark count). Hence the presence of
a filter will have a bigger impact on the number of ‘noise’ coincidences than on the number of ‘true’
coincidences, and will increase the SNR. Figure 7.5 shows the calculated SNR as a function of the
parameter x. We can see that it increases as x decreases until a maximum is reached. The decrease
after this optimal point is due to the dark counts of the detectors, which are not related to the filter
and hence are constant; these dominate the ‘noise’ probability term of the SNR for small values of
x, while the ‘true’ probability term continues to decrease.
45
40
35
30
SNR
25
20
15
10
5
0
0 1 2 3 4 5 6 7 8 9 10
Figure 7.5 – SNR as a function of the filter bandwidth.
We observe that a narrower filter than the one used (12 nm bandwidth, which gives x = 10%),
should increases the SNR, and hence the fidelity.
Detector influence
The two detector parameters that appear are:
• the quantum efficiency η idq (which apprears in the T tot parameter);
• the dark count probability pdark .
The dark counts are exponentially proportional to the quantum efficiency of the detector, as can
been seen from the experimental results in figure 7.6.
55
50
Dark count rate (Hz)
45
40
35
30
25
20
15
10
10 15 20 25
Quantum efficiency (%)
Figure 7.6 – Detector dark count rate as a function of quantum efficiency, with a trigger frequency
of 10 kHz.
The blue and red curves in figure 7.5 show the SNR as a function of the filter bandwidth for
detector efficiencies of 25% and 20% respectively, taking into account the corresponding dark counts.
We can see that when the efficiency inscreases, even so the signal increases (more photon pairs are
detected), the dark count and the luminescence noise together increase more rapidly, leading to a
descrease in the SNR.
Losses influence
The blue and green curves in figure 7.5 show the SNR as a function of the filter bandwidth for two
different levels of total transmission of the photons through the setup (T tot and T tot × 2). We can
see that an increase of the overall transmission in the setup should increase the SNR.
7.3 Results
Taking into account all these considerations, we have thus decided to replace the interferential filter
with a FWHM of 12 nm with a filter with a FWHM of 1.2 nm, which is close to the ideal situation
represented in figure 7.5. We have also chosen to reduce the quantum efficiency of the detectors from
25% to 20%. A higher repetition rate (40 kHz instead of 5 kHz as previously) then allows a reduction
in the integration time of the measurement.
First, we have recorded the zero delay time coincidence value as a function of the filter wavelength.
This allows us to find the optimal working point of the filter. Figure 7.7 reports the corresponding
results.
7.3. RESULTS 97
0.50
Coincidence counts (/s)

true coincidences
noise coincidences
0.40
0.30
0.20
0.10
0.00
1560 1561 1562 1563 1564
Filter wavelength (nm)

Figure 7.7 – Coincidence counts versus the interferential filter central wavelength (FWHM 1.2 nm).
The sample is electrically injected with current pulses having a duration of 60 ns and a repetition
rate of 40 kHz. The data were accumulated during 600 s with a sampling resolution of 162 ps. The
detector quantum efficiency was set to 20%.
25 1.0
20 0.9
15 0.8 fidelity
SNR
10 0.7
5 0.6
0 0.5
250 300 350 400 450 500 550 600 650 700 750
Current (mA)
Figure 7.8 – SNR and fidelity to a Bell state as a function of the injected current. The sample is
electrically injected with current pulses having a duration of 60 ns and a repetition rate of 40 kHz.
The data were accumulated during 600 s with a sampling resolution of 162 ps. The detector quantum
efficiency was set to 20%.
Figure 7.8 shows the SNR and the estimated fidelity as a function of the injected current. Three
regimes are clearly visible: under the laser threshold, no ‘true coincidences’ peak emerges from
the background noise; for a current value between 450 and 600 mA, we observe the expected linear
dependency of the SPDC peak with the injected current; for current values above 600 mA, the SPDC
signal drops, probably because of thermal effects within the device having a significant influence on
the laser operation and the phase-matching curve.
Figure 7.9 shows the best histogram of the detection time delays between T E and T M polarized
photons. The sample was electrically injected with current pulses having an intensity of 601 mA, a
duration of 60 ns, and a repetition rate of 40 kHz. The data were accumulated during 300 s. The
SNR is 21.33, this leading to an estimate of the maximal fidelity to the Bell state |φ+ i, F ∼ 94%,
which is compliant for future experimental violation of Bell’s inequality.
0.25
0.20
0.15
0.10
0.05
0.00
-50 -40 -30 -20 -10 0 10 20 30 40 50
Time delay (ns)

Figure 7.9 – Time-correlation histogram. The sample is electrically injected with current pulses
having an intensity of 601 mA, a duration of 60 ns and a repetition rate of 40 kHz. The data were
accumulated during 300 s with a sampling resolution of 162 ps. The detector quantum efficiency was
set to 20%.
From these data, and taking into account the overall transmission along the optical path (T tot =
2%), we can estimate that the internal generation efficiency of the device is ≈ 7 × 10−11 pairs
per injected electron above the threshold. This value corresponds to a SPDC efficiency ∼ 10−9
pairs/pump photon. These results are in agreement with the SH generation efficiency letting expect
∼ 6×10−9 pairs/pump photon, and consistent with the numerical simulation on an undoped structure
giving ∼ 1.8 × 10−8 pairs/pump photon for a 2 mm-long waveguide.
Conclusion
Since its emergence, about thirty years ago, the field of quantum information has behold tremendous
advances. After a large number of experimental demonstrations of the fundamental concepts and
protocols of quantum information, we witness nowadays a rise of quantum information technologies
developing different physical supports, some of them having the potential to be hybridized. In
particular, the community is working to build reliable, compact, versatile and efficient quantum
components able to generate controlled quantum states.
The work of this PhD thesis is a contribution to this field. It is focused on an AlGaAs photon-
pair source, presenting many advantages in view of real world quantum information technologies.
First, its operation at room temperature allows to get ride of cryogenic systems; second, the emission
of photons at the telecom wavelengths makes it compatible with existing fibered networks for long
distance communications; finally, the use of III-V semiconductor materials presents a strong potential
for miniaturization, thanks to a direct bandgap allowing the integration of the pumping laser inside
the nonlinear medium, leading to a fully integrated device much more compact and easier to use.
The first part of the manuscript is dedicated to the presentation of the fabrication of the device
and its characterization. The processing developed to fabricate the device is described and two
etching techniques are compared. The optoelectronics characterization of the device is presented and
its behavior has been studied as a function of temperature and injected current. The optical quality
of the device is assessed through optical losses measurements and through the characterization of its
nonlinear behavior (second harmonic generation, parametric down-conversion, stimulated emission
tomography).
The second part of the manuscript is devoted to the characterization of the quantum state emit-
ted by the device. In the framework of a collaboration with the researchers of the Group of Ap-
plied Physics-Quantum Technologies in Geneva, we have demonstrated high entanglement level in
energy-time and high indistinguishability of the paired photons. We have then exploited the small
birefringence of our waveguides to produce polarization entangled states without any off-chip com-
pensation. This property, together with the broadband emission of our source, has been exploited
to implement a multi-user quantum key distribution protocol with a simple and robust experimental
setup. This work has been done in a collaboration with the team of ‘Quantum cryptography’ of
Telecom ParisTech. This experiment has allowed to distribute quantum keys with a low quantum bit
error rate for several pairs of users separated by up to 50 km of standard optical fiber. Finally, we
have demonstrated the first electrically driven device emitting photon-pairs at telecom wavelengths
and operating at room temperature. These results open the way towards real world large scale
photonic-circuit-based quantum communications.
The next step, on a short term, will consist to characterize the quantum state of the photons
emitted by the electrically driven photon-pair source. A new sample has been grown by our colleagues
at C2N after careful calibration of layer thicknesses and alloys composition; it will be processed and
tested in the next weeks. The work done on the ICP etching recipe, has proved to enhance the
conversion efficiency in the passive sample; we will build on these results to fabricate a new generation
of more efficient electrically driven devices.
On a longer term, the team will work on the optimization of the design of the source; indeed the
99
sample design has not been changed during my PhD thesis and now that a working sample has been
demonstrated, there are several possibilities to improve it. We can play for example on the number
of quantum wells; adding one or two quantum wells could allow to lower the laser threshold thus
leading to samples with a lower level of luminescence noise. Another possibility is to separate the
lasing region and the nonlinear region in order to separate the management of the ‘laser performances’
and the nonlinear ones. We could imaging for example to etch two distributed bragg reflectors to
define the laser cavity in a region of the waveguide and to do the nonlinear conversion in another
region.
Another direction that our group is following is to take benefit of the recent progress on the
hybrid integration of III-V on Si, to combine the advantages of these two material platforms: on
one hand the source developed in this PhD thesis is compatible with electrical injection; on the
other hand, Si-based quantum photonic circuits, compatible with CMOS technology, have already
demonstrated their potential to manipulate with success quantum states of two or more photons. To
reach this goal, two collaborations have already started: a work with the ‘Centre de Nanosciences
et de Nanotechnologies’ is focused on the adhesive bonding of III-V sources on Si circuits (Silicon
On Insulator (SOI) platform). The structure of the passive samples presented in this work has been
adapted to be bonded onto a Si circuit. A second collaboration has been established with the team of
I. Walmsley at the University of Oxford to couple our devices to Silica on Silicon circuits developed
for quantum metrology.
Appendix
101
Appendix A
List of Conferences and Publications
Communications
The person who presented the work at the conference is indicated by an asterisk*.
Invited talks
• C. Autebert, G. Boucher, J. Belhassen, G. Maltese, Y. Halioua, and S. Ducci*,
« AlGaAs photonic devices: from quantum state generation to quantum communications »,
QCMC, Singapore, July 2016
• C. Autebert, G. Boucher, G. Maltese, J. Belhassen, Y. Halioua, and S. Ducci*,

« Biphoton Engineering and Quantum Communications with AlGaAs devices », Photonics
North, Québec city, May 2016
• C. Autebert, G. Boucher, Y. Halioua, Q. Yao, A. Lemaître, C. Gomez, I. Favero,

G. Leo, and S. Ducci*, « Quantum state engineering and measurement with AlGaAs devices
», SPIE Photonic West, San Francisco, February 2016
• G. Boucher, C. Autebert, Y. Halioua, A. Lemaître, C. Gomez, I. Favero, G. Leo,

and S. Ducci*, « Photon pair sources in AlGaAs: from electrical injection to quantum state
engineering », Single Photon Workshop, Geneva, July 2015
• G. Boucher, C. Autebert, F. Boitier, I. Favero, G. Leo, and S. Ducci*, « Photon pair

sources in AlGaAs : from electrical injection to quantum state engineering », SPIE Photonics
West, San Francisco, February 2015
• F. Boitier, A. Orieux, C. Autebert, A. Lemaître, E. Galopin, C. Manquest, C.

Sirtori, I. Favero, G. Leo, and S. Ducci*, « A photon-pair-emitting laser diode », SPIE
Photonics Europe, Brussels, April 2014
• F. Boitier, C. Autebert, G. Boucher, A. Orieux, A. Eckstein, E. Galopin, A.

Lemaître, C. Manquest, I. Favero, G. Leo, and S. Ducci*, « A photon-pair-emitting
laser diode », SPIE Photonic West, San Francisco, February 2014
Talks
• C. Autebert*, J. Trapateau, A. Orieux, A. Lemaître, C. Gomez-Carbonell, E.
Diamanti, I. Zaquine, and S. Ducci, « Multi-user quantum key distribution with a semi-
conductor source of entangled photon pairs », Frontiers in Optics, Rochester, October 2016
103
104 LIST OF CONFERENCES AND PUBLICATIONS
• C. Autebert, J. Trapateau, A. Orieux*, A. Lemaître, C. Gomez-Carbonell, E.

conductor source of entangled photon pairs », Young IQIS, Rome, September 2016
• C. Autebert*, J. Trapateau, A. Orieux, A. Lemaître, C. Gomez-Carbonell, E.

conductor source of entangled photon pairs », SPSS Workshop, Oxford, September 2016

conductor source of entangled photon pairs », QuPa Days, Paris, July 2016
• C. Autebert*, N. Bruno, A. Martin, A. Lemaître, C. Gomez, I. Favero, G. Leo,

H. Zbinden and S. Ducci, « Integrated AlGaAs source of highly indistinguishable and
energy-time entangled photons », GdR IQFA - 6th, Palaiseau, November 2015
• Y. Halioua*, C. Autebert, G. Boucher, Q. Yao, J. Belhassen, A. Lemaître, C.

Gomez, I. Favero, G. Leo, and S. Ducci, « AlGaAs devices generating non-classical
states of light », QLIMS, Barcelona, November 2015

energy-time entangled photons », Frontiers in Optics, San Jose, October 2015
• C. Autebert*, G. Boucher, F. Boitier, A. Eckstein, A. Lemaître, I. Favero, G.

Leo, and S. Ducci, « Photon pair sources in AlGaAs : from electrical pumping to quantum
state engineering », PICQUE Workshop, Oxford, January 2015
• G. Boucher*, C. Autebert, F. Boitier, A. Eckstein, A. Lemaître, C. Manquest,

C. Sirtori, I. Favero, G. Leo, and S. Ducci, « Integrated AlGaAs sources of quantum
correlated photon pairs », ECOC, Nice, September 2014
• G. Boucher, C. Autebert, F. Boitier*, A. Eckstein, A. Orieux, A. Lemaître, E.

Galopin, C. Manquest, C. Sirtori, I. Favero, G. Leo, and S. Ducci, « Photon pair
sources in AlGaAs : from state engineering to electrical injection », Quantum – VII workshop,
Turin, Mai 2014
• F. Boitier, A. Orieux, C. Autebert*, A. Lemaître, E. Galopin, C. Manquest, C.

Sirtori, I. Favero, G. Leo, and S. Ducci, « A laser diode for integrated photon pair
generation at telecom wavelength », QIM, Berlin, March 2014
Posters
conductor source of entangled photon pairs », IQIS, Rome, September 2016
• C. Autebert, J. Trapateau, A. Orieux, A. Lemaître, C. Gomez-Carbonell, E. Dia-

manti*, I. Zaquine, and S. Ducci, « Multi-user quantum key distribution with a semi-
conductor source of entangled photon pairs », QCrypt, Washington, September 2016
• C. Autebert, Y, Halioua, G. Maltese*, A. Lemaître, C. Gomez-Carbonell, M.

Amanti, C. Sirtori, and S. Ducci, « Etude d’un pointeur laser à photons jumeaux »,
COLOQ’15, Bordeaux, July 2016
105
• C. Autebert, J. Trapateau*, A. Orieux, A. Lemaître, C. Gomez-Carbonell, E.

Diamanti, I. Zaquine, and S. Ducci, « Distribution de clés quantiques multi-utilisateurs
avec une source semi-conductrice », COLOQ’15, Bordeaux, July 2016

energy-time entangled photons », Single Photon Workshop, Geneva, July 2015
• C. Autebert*, F. Boitier, Y. Halioua, A. Lemaître, I. Favero, G. Leo, and S. Ducci,

« Electrically injected source of photon pairs at room temperature: device performances and
entanglement generation », CLEO Europe, Munich, June 2015
• F. Boitier, A. Orieux, C. Autebert*, G. Boucher, A. Eckstein, E. Galopin, A.

Lemaître, C. Manquest, I. Favero, G. Leo, and S. Ducci, « A photon-pair-emitting
laser diode », GDR IQFA - 4th, Paris, November 2013
• F. Boitier*, C. Autebert, G. Boucher, A. Orieux, A. Eckstein, E. Galopin, A.

Lemaître, C. Manquest, I. Favero, G. Leo, and S. Ducci, « Vers une diode laser
émettrice de photons jumeaux », COLOQ’13, Paris, July 2013
Seminars
• C. Autebert, « A Photon Pair Source in AlGaAs », PhD Seminar of the MPQ laboratory,
Paris, 26 February 2016.
• C. Autebert, « Photon pair sources in AlGaAs: from electrical pumping to quantum state
engineering », Group of Applied Physics - Optics, University of Geneva, Geneva, 24 November
2014.
106 LIST OF CONFERENCES AND PUBLICATIONS
Publications
• C. Autebert, A. Lemaître, and S. Ducci, « Highly efficient Parametric Down Conversion
in a monolithic AlGaAs waveguide », In preparation
• C. Autebert, J. Trapateau, A. Orieux, A. Lemaître, C. Gomez Carbonell, E. Dia-

manti, I. Zaquine and S. Ducci, « Multi-user Quantum Key Distribution with entangled
photons from an AlGaAs chip », to appear in Quantum Science and Technologies
• C. Autebert, G. Maltese, Y. Halioua, F. Boitier, A. Lemaître, M.Amanti, C. Sir-

tori and S. Ducci, « Electrically injected twin photon emitting lasers at room temperature
», Technologies, 4, 3 p24 (2016)
• C. Autebert, N. Bruno, A. Martin, A. Lemaître, C. Gomez Carbonell, I. Favero,

G. Leo, H. Zbinden and S. Ducci, « Integrated AlGaAs source of highly indistinguishable
and energy-time entangled photons », Optica 3, 2 p143 (2016)
• C. Autebert, G. Boucher, F. Boitier, A. Eckstein, I. Favero, G. Leo, and S. Ducci,

« Photon pair sources in AlGaAs: from electrical injection to quantum state engineering »,
Journal of Modern Optics, p1–7 (2015)
• F. Boitier, A. Orieux, C. Autebert, A. Lemaître, E. Galopin, C. Manquest, C.

Sirtori, I. Favero, G. Leo, and S. Ducci, « Electrically Injected Photon-pair Source at
Room Temperature », Physical Review Letter 112, 183901 (2014)
Awards
• October 2016, Bourse L’Oréal-UNESCO, For Women in Science, France
• October 2015, Emil Wolf Prize, of the Optical American Society Foundation for Outstanding
Student Paper Competition, conference Frontiers in optics, San Jose (USA)
• July 2015, Best poster prize, conference Single Quantum Workshop, Geneva (Switzerland)
• June 2014, Best poster prize, MPQ laboratory day, Paris (France)
Articles
In the following pages we include the papers reporting on the work presented in this doctoral thesis.
Selected for a Viewpoint in Physics
week ending
PRL 112, 183901 (2014) PHYSICAL REVIEW LETTERS 9 MAY 2014
Electrically Injected Photon-Pair Source at Room Temperature

Fabien Boitier,1 Adeline Orieux,1,* Claire Autebert,1 Aristide Lemaître,2 Elisabeth Galopin,2 Christophe Manquest,1
Carlo Sirtori,1 Ivan Favero,1 Giuseppe Leo,1 and Sara Ducci1,†
1
Laboratoire Matériaux et Phénomènes Quantiques, Université Paris Diderot, Sorbonne Paris Cité, CNRS-UMR 7162,
Case courrier 7021, 75205 Paris Cedex 13, France
2
Laboratoire de Photonique et Nanostructures, CNRS-UPR20, Route de Nozay, 91460 Marcoussis, France
(Received 20 December 2013; revised manuscript received 21 February 2014; published 7 May 2014)
One of the main challenges for future quantum information technologies is the miniaturization and
integration of high performance components in a single chip. In this context, electrically driven sources
of nonclassical states of light have a clear advantage over optically driven ones. Here we demonstrate the
first electrically driven semiconductor source of photon pairs working at room temperature and telecom
wavelengths. The device is based on type-II intracavity spontaneous parametric down-conversion in an
AlGaAs laser diode and generates pairs at 1.57 μm. Time-correlation measurements of the emitted pairs
give an internal generation efficiency of 7 × 10−11 pairs=injected electron. The capability of our platform
to support the generation, manipulation, and detection of photons opens the way to the demonstration of
massively parallel systems for complex quantum operations.
DOI: 10.1103/PhysRevLett.112.183901 PACS numbers: 42.65.Lm, 03.67.Bg, 42.55.Px, 42.82.-m
Photons have a peculiar advantage in the development of electrical injection. In order to deal with the isotropic
quantum information technologies [1–3], since they behave structure of this crystal, several solutions have been pro-
naturally as flying qubits presenting a high speed trans- posed to achieve nonlinear optical conversion in AlGaAs
mission over long distances and are almost immune to waveguides [14–18]; among these, modal phase matching,
decoherence [4,5]. The intrinsic scalability and reliability in which the phase velocity mismatch is compensated
of integrated photonic circuits has recently given rise to a by multimode waveguide dispersion, is one of the most
new generation of devices for quantum communication, promising to monolithically integrate the laser source and
computation, and metrology [6]. Nevertheless even if great the nonlinear medium into a single device [19,20]. In this
progress has been made in the manipulation [7,8] and scheme, the interacting modes can either be confined by
detection [9] of a nonclassical state of light on chip, a homogeneous claddings [21] or by the photonic band gap
complete integration of the light source in the photonic [22], this latter option avoids aging problems via the
circuitry remains one of the main challenges on the way reduction of the total aluminum content.
towards large scale applications; such devices would have a In this Letter we present an electrically injected AlGaAs
clear advantage over optically driven ones in terms of porta- device that emits photon pairs at telecom wavelengths
bility, energy consumption, and integration. Semiconductor and operates at room temperature. Our device, shown in
materials are ideal to achieve extremely compact and Fig. 1(a), has been engineered for simultaneous lasing
massively parallel devices: concerning photon-pair sources, around 785 nm and efficient type-II internal SPDC with
the biexciton cascade of a quantum dot has been used to photon pairs around 1.57 μm. Two Bragg mirrors provide
demonstrate an entangled-light-emitting diode at a wave- both a photonic band gap vertical confinement for the laser
length of 890 nm [10]. However, even if the use of a single mode—a transverse electric Bragg (TEB) mode—and total
emitter guarantees a deterministic emission, these devices internal reflection claddings for the photon-pair modes
operate at cryogenic temperatures, greatly limiting their (one TE00 and one TM00 ). The nonlinear process is possible
potential for applications. thanks to the interaction of the TEB pump mode and the
Optical parametric conversion offers an alternative two twin photon modes verifying the equations of energy
approach. Despite its nondeterministic nature, this process conservation and type-II phase matching:
is the most widely used to produce photon pairs for
quantum information and communications protocols. Up ℏωTEB ¼ ℏωTE00 þℏωTM00 ;
to now, entangled photon pairs have been generated by nTEB ðωTEB ÞωTEB ¼ nTE00 ðωTE00 ÞωTE00 þnTM00 ðωTM00 ÞωTM00 ;
optical pumping in passive semiconductor waveguides by
exploiting four-wave mixing in silicon [11] or spontaneous where ωi and ni (with i ¼ TEB, TE00 , TM00 ) are the
parametric down conversion (SPDC) in aluminium gallium angular optical frequency and the effective index of the ith
arsenide (AlGaAs) [12,13]. Thanks to its direct band gap, mode. The simulated tuning curves based on Refs. [23,24],
the latter platform presents an evident interest for the solutions of the above system, are shown in Fig. 1(b).
0031-9007=14=112(18)=183901(5) 183901-1 © 2014 American Physical Society

PHYSICAL REVIEW LETTERS week ending
PRL 112, 183901 (2014) 9 MAY 2014
Figure 2(a) shows the internal peak power and voltage

characteristics of the device as a function of the injected
current. The device is mounted epi-side up on a copper heat
sink; the temperature can be tuned between 15 °C and 40 °C
with a standard Peltier module. In order to avoid unwanted
thermal drifts, we employ current pulses of duration 120 ns
and the repetition rate is set to 10 kHz. The laser internal
peak power is evaluated by taking into account the modal
reflectivity of the TEB mode (79%), numerically simulated
by 2 D FDTD. We observe that the turn-on voltage is
(a) ∼1.6 V, which is very close to the QW band gap
(∼1.58 eV), thus meaning that no current-blocking effects
occur at the heterointerfaces. The threshold current is
around 420 mA, corresponding to a threshold current
density of 3.3 kA=cm2 . This value is higher than state-
of-the-art laser diodes in this spectral range [25], probably
because of the crudely optimized doping of the Bragg
mirrors. The spatial intensity distribution of the laser beam
is studied by imaging the output facet; the recorded near-
field distribution is reported in Fig. 2(b) together with the
(b)
FIG. 1 (color online). Working principle of the device. (a) Sche-

matic view of the source. The laser light emitted by the quantum
well is converted into telecom photon pairs by intracavity
spontaneous parametric down-conversion. (b) Simulated tuning
curves of the type-II phase matching at T ¼ 20 °C. Energy
conservation imposes pair generation either on the thick or the
thin branches of the curves. (a) (b)
Because of the strong dispersion of the TEB mode arising
from the proximity to the energy band gap of the waveguide
core, small shifts of the laser wavelength from degeneracy
produce a large wavelength separation between the gen-
erated photons. For this reason, taking into account the
sensitivity range of our single-photon avalanche photo-
diodes, our spectral window to detect the two photons of
each pair is limited to the region of frequency degeneracy.
The sample was grown by molecular beam epitaxy on a
(100) n-doped GaAs substrate. It consists of an n-doped
six-period Al0.80 Ga0.20 As=Al0.25 Ga0.75 As Bragg reflector
(lower mirror), a 298 nm Al0.45 Ga0.55 As core with an (c)
8.5 nm Al0.11 Ga0.89 As quantum well (QW) in the middle,
and a p-doped six-period Al0.25 Ga0.75 As=Al0.80 Ga0.20 As FIG. 2 (color online). Laser operation. (a) Voltage (cross) and
Bragg reflector (upper mirror). The Bragg reflectors are internal optical power (circle) versus current. Measurements are
gradually doped from 1 × 10−17 cm−3 to 2 × 10−18 cm−3 . performed with a current pulse duration of 120 ns and a repetition
A 230 nm GaAs cap layer (2 × 10−19 cm−3 , p-doped) rate of 10 kHz for a heat-sink temperature of 19 °C. The solid
protects the structure and facilitates the upper contact. line is a linear fit for current values above laser threshold.
Waveguides are fabricated using wet chemical etching to The laser diode has an electrical resistivity of 3.1Ω, a turn-on
voltage of 1.6 V, a laser threshold of 0.420 A, and an efficiency
define 5.5–6 μmm wide and 2 μm deep ridges along the
of 132 mW=A. (b) Measured (top) and simulated (bottom)
(011) crystalline axis, in order to exploit the maximum near-field emission of the laser mode. (c) Normalized laser
nonzero nonlinear coefficient and a natural cleavage plane. emission intensity as a function of wavelength and heat-sink
Processing is completed by sample thinning and contact temperature measured with a fibered optical spectrum analyzer.
metallization with Au alloys. Samples are cleaved into The dashed line shows the expected temperature variation of the
2 mm long stripes. QW band gap.
183901-2
PRL 112, 183901 (2014) 9 MAY 2014
corresponding numerical simulation, showing a clear

evidence of emission on the TEB mode. Figure 2(c)
displays the laser emission intensity spectra as a function
of heat-sink temperature, for an injected current of 650 mA.
Apart from the longitudinal mode hopping—typical of
the laser diode—the general trend corresponds to the theo-
retical temperature dependence of the QW band gap
(0.23 nm=°C).
Optical propagation losses in the waveguide, a key issue
for photon sources intended for quantum information, are
measured via a standard Fabry-Perot technique [26]: the
values obtained for the TE00 and TM00 modes in the (a)
telecom range are around 2 cm−1 . Similar measurements
on an undoped waveguide give a value of 0:1 cm−1 ; the
losses on the active device are mainly attributed to doping.
The nonlinear optical properties of the sample are first
explored through a second harmonic (SH) generation
measurement performed without electrical injection. An
input beam at the fundamental wavelength is polarized at
45° and is injected in the waveguide in order to couple the
TE and TM modes simultaneously. Figure 3(a) shows clear
growth of the SH power for an input beam wavelength
around 1.57 μm at T ¼ 19 °C; the inset shows the expected
quadratic dependence of the SH power with the funda- (b)
mental power. The observed modulation as a function
of the input wavelength is due to Fabry-Perot interferences FIG. 3 (color online). Second harmonic generation. (a) SH
spectrum as a function of the fundamental wavelength at
between the waveguide facets. The solid curve results T ¼ 19 °C. The curve is a fit taking into account propagation
from a fit taking into account propagation losses and losses and modal facet reflectivities. The inset shows the peak SH
modal reflectivities of the three interacting modes [27]. power as a function of the fundamental beam power. The solid
The inferred internal SH generation efficiency is line shows the expected squared power law function. (b) SH peak
∼35% W−1 cm−2 and the FWHM of the phase-matching wavelength versus temperature. The solid line is a linear fit of
bandwidth is ∼0.6 nm Figure 3(b) reports the variation of the experimental data whereas the shaded area indicates the
the SH peak wavelength with temperature. The comparison FWHM of the phase-matching bandwidth. The experimental
between these data and those of Fig. 2(c) shows that the slope of 0.09 nm=K is consistent with the theoretical slope of our
tunability curves of the laser emission and of the SH signal numerical modeling (0.07 nm=K). The dashed line reports the
expected variation of the QW band gap presented in Fig. 2(c).
intersect in the explored temperature range.
In order to confirm the existence of a working region
of the device and to demonstrate the emission of photon that obtained in a completely passive device based on the
pairs around 1.57 μm, time-correlation measurements are same kind of phase matching [17].
performed under electrical injection [see Fig. 4(a)]. The The signal to noise ratio (SNR) is evaluated by taking the
detected SPDC signal is optimized by tuning the temper- number of true coincidences within the FWHM of the peak
ature. Figure 4(b) shows a histogram of the detection time over the background signal on the same time window; data
delays between TE and TM polarized photons at presented in Fig. 4(b) give a SNR of 13.5, mainly limited
T ¼ 25 °C. The sharp peak emerging from the background by the luminescence noise of the device. In this respect,
is a clear evidence of pair production. From these data, an optimization work leading to a smaller laser threshold
taking into account the overall transmission along the will be beneficial to reduce spurious luminescence and,
optical path, we can estimate that the internal generation thus, to increase the SNR. Our result enables us to estimate
efficiency of the device is ∼7 × 10−11 pairs=injected elec- the fidelity F to the Bell state jψ þ i that can be produced
tron above the threshold [28]. This value corresponds to a with our device. Assuming that the source emits a Werner
SPDC efficiency ∼10−9 pairs=pump photon: these results state [29,30]—which is reasonable since the noise is
are in agreement with our SH generation efficiency, letting not polarized—the associated density matrix is ρ̂W ¼
expect ∼6 × 10−9 pairs=pump photon, and consistent with Pjψ þ ihψ þ j þ ð1 − PÞ=4 × 1 with P ¼ SNR=ð2 þ SNRÞ.
our numerical simulation on an undoped structure giving This leads to a maximal fidelity estimation F to
∼1.8 × 10−8 pairs=pump photon for a 2 mm-long wave- jψ þ i ¼ ð1 þ 3PÞ=4 ∼ 90%, which is compliant with a
guide. Note here that such efficiency compare well with future experimental violation of Bell’s inequality.
183901-3
PRL 112, 183901 (2014) 9 MAY 2014
No. ANR-11-LABX-086 for financial support in the

framework of the Program No. ANR-11-IDEX-0005-02.
A. O. and C. A. acknowledge the Délégation Générale de
l’Armement (DGA) for financial support. We acknowledge
G. Boucher and A. Eckstein for help with the experimental
setup and A. Andronico for discussions on numerical
(a) simulations. S. D. and C. S. are members of Institut
Universitaire de France.
*
Present address: Dipartimento di Fisica, Sapienza Univer-
sità di Roma, Piazzale Aldo Moro, 5, I-00185 Roma, Italy.
†
sara.ducci@univ‑paris‑diderot.fr
[1] T. D. Ladd, F. Jelezko, R. Laflamme, Y. Nakamura, C. R.
Monroe, and J. L. O’Brien, Nature (London) 464, 45
(2010).
[2] N. Gisin and R. Thew, Nat. Photonics 1, 165 (2007).
[3] M. D. Eisaman, J. Fan, A. Migdall, and S. V. Polyakov,
Rev. Sci. Instrum. 82, 071101 (2011).
(b)
[4] R. Ursin, F. Tiefenbacher, T. Schmitt-Manderbach, H.
Weier, T. Scheidl, M. Lindenthal, B. Blauensteiner, T.
FIG. 4 (color online). Coincidence measurement. (a) Experi- Jennewein, J. Perdigues, P. Trojek, B. Omer, M. Furst,
mental setup. The emitted pairs, collected through a 63 × M. Meyenburg, J. Rarity, Z. Sodnik, C. Barbieri, H.
microscope objective, are focused into a fibered 1.2 nm-FWHM Weinfurter, and A. Zeilinger, Nat. Phys. 3, 481 (2007).
interferential filter centered at 1.57 μm then sent in a fibered [5] D. Stucki, C. Barreiro, S. Fasel, J.-D. Gautier, O. Gay,
polarizing beam splitter (PBS). The polarization controller before N. Gisin, R. Thew, Y. Thoma, P. Trinkler, F. Vannel, and
the interferential filter is adjusted to match the TE=TM axes of H. Zbinden, Opt. Express 17, 13326 (2009).
the sample and the PBS outputs. The emerging TE and TM [6] J. L. O’Brien, A. Furusawa, and J. Vučković, Nat. Photonics
photons are detected with two InGaAs single-photon avalanche 3, 687 (2009).
photodiodes having 20% detection efficiency and 50 ns gate, [7] J. C. F. Matthews, A. Politi, A. Stefanov, and J. L. O’Brien,
synchronized with the current pulses. A time-to-digital converter Nat. Photonics 3, 346 (2009).
is used to analyze the time correlations between detected photons. [8] J. E. Kennard, J. P. Hadden, L. Marseglia, I. Aharonovich,
(b) Time-correlation histogram of TE=TM photons around S. Castelletto, B. R. Patton, A. Politi, J. C. F. Matthews,
1.57 μm at T ¼ 25 °C. The sample is electrically injected with A. G. Sinclair, B. C. Gibson, S. Prawer, J. G. Rarity, and
current pulses having an intensity of 700 mA, a duration of 60 ns J. L. O’Brien, Phys. Rev. Lett. 111, 213603 (2013).
and a repetition rate of 10 kHz. The data were accumulated during [9] J. P. Sprengers, A. Gaggero, D. Sahin, S. Jahanmirinejad,
1200 s with a sampling resolution of 162 ps. The inset shows a G. Frucci, F. Mattioli, R. Leoni, J. Beetz, M. Lermer,
zoom of the sharp central peak. M. Kamp, S. Hofling, R. Sanjines, and A. Fiore, Appl. Phys.
Lett. 99, 181110 (2011).
These results open the way towards large scale photonic- [10] C. L. Salter, R. M. Stevenson, I. Farrer, C. A. Nicoll,
circuit-based quantum computation. Indeed, one applica- D. A. Ritchie, and A. J. Shields, Nature (London) 465,
tion of this source could be the controlled on-chip electrical 594 (2010).
injection of an arbitrary number of heralded single photons [11] N. Matsuda, H. Le Jeannic, H. Fukuda, T. Tsuchizawa, W. J.
Munro, K. Shimizu, K. Yamada, Y. Tokura, and H. Takesue,
or photon pairs on an arbitrary number of input modes
Sci. Rep. 2, 817 (2012).
of an integrated photonic circuit. This could be achieved [12] A. Orieux, A. Eckstein, A. Lemaître, P. Filloux, I. Favero,
by fabricating a monolithic device consisting of equally G. Leo, T. Coudreau, A. Keller, P. Milman, and S. Ducci,
spaced laser diodes independently injected through a Phys. Rev. Lett. 110, 160502 (2013).
control electronics, which is allowed by the mature III-V [13] A. Vallés, M. Hendrych, J. Svozilík, R. Machulka, P.
technology. Interfacing it with multiport reconfigurable Abolghasem, D. Kang, B. J. Bijlani, A. S. Helmy, and
circuits [1] would allow practical medium size reconfig- J. P. Torres, Opt. Express 21, 10841 (2013).
urable on-chip quantum photonics computation, such as [14] A. Fiore, V. Berger, E. Rosencher, P. Bravetti, and J. Nagle,
boson sampling [31–34] and multiple photon quantum Nature (London) 391, 463 (1998).
walks, allowing medium size optical simulations [35]. [15] X. Yu, L. Scaccabarozzi, A. C. Lin, M. M. Fejer, and
J. S. Harris, J. Cryst. Growth 301–302, 163 (2007).
This work was partly supported by Région Ile-de-France [16] L. Lanco, S. Ducci, J.-P. Likforman, X. Marcadet, J. A. W.
in the framework of C’Nano IdF with the TWILIGHT van Houwelingen, H. Zbinden, G. Leo, and V. Berger,
project. F. B. acknowledges the Labex SEAM Program Phys. Rev. Lett. 97, 173901 (2006).
183901-4
PRL 112, 183901 (2014) 9 MAY 2014
[17] R. Horn, P. Abolghasem, B. J. Bijlani, D. Kang, [26] A. De Rossi, V. Ortiz, M. Calligaro, L. Lanco, S. Ducci,
A. S. Helmy, and G. Weihs, Phys. Rev. Lett. 108, 153605 V. Berger, and I. Sagnes, J. Appl. Phys. 97, 073105 (2005).
(2012). [27] R. L. Sutherland, Handbook of Nonlinear Optics, edited
[18] J. P. Van der Ziel, R. C. Miller, R. A. Logan, W. A. by J. B. Thompson, Optical Engineering (Marcel Dekker,
Nordland, and R. M. Mikulyak, Appl. Phys. Lett. 25, 238 New York, 1996).
(1974). [28] See Supplemental Material at http://link.aps.org/
[19] A. Orieux, C. E. Rodrigues de Souza, A. Lemaître, supplemental/10.1103/PhysRevLett.112.183901 for addi-
E. Galopin, C. Manquest, I. Favero, G. Leo, and S. Ducci, tionnal photon correlation measurements.
in Conference on Lasers and Electro-Optics 2012 (Optical [29] R. F. Werner, Phys. Rev. A 40, 4277 (1989).
Society of America, Washington, DC, 2012), p. JW4A.114. [30] M. Barbieri, F. De Martini, G. Di Nepi, and P. Mataloni,
[20] B. J. Bijlani, P. Abolghasem, and A. S. Helmy, Appl. Phys. Phys. Rev. Lett. 92, 177901 (2004).
Lett. 103, 091103 (2013). [31] A. Crespi, R. Osellame, R. Ramponi, D. J. Brod, E. F.
[21] A. De Rossi, V. Ortiz, M. Calligaro, B. Vinter, J. Nagle, Galvão, N. Spagnolo, C. Vitelli, E. Maiorino, P. Mataloni,
S. Ducci, and V. Berger, Semicond. Sci. Technol. 19, L99 and F. Sciarrino, Nat. Photonics 7, 545 (2013).
(2004). [32] M. Tillmann, B. Dakić, R. Heilmann, S. Nolte, A. Szameit,
[22] B. J. Bijlani and A. S. Helmy, Opt. Lett. 34, 3734 (2009). and P. Walther, Nat. Photonics 7, 540 (2013).
[23] J. Chilwell and I. Hodgkinson, J. Opt. Soc. Am. A 1, 742 [33] M. A. Broome, A. Fedrizzi, S. Rahimi-Keshari, J. Dove,
(1984). S. Aaronson, T. C. Ralph, and A. G. White, Science 339,
[24] S. Gehrsitz, F. K. Reinhart, C. Gourgon, N. Herres, 794 (2013).
A. Vonlanthen, and H. Sigg, J. Appl. Phys. 87, 7825 [34] J. B. Spring, B. J. Metcalf, P. C. Humphreys, W. Steven
(2000). Kolthammer, X.-M. Jin, M. Barbieri, A. Datta, N. Thomas-
[25] L. A. Coldren, S. W. Corzine, and M. L. Mashanovitch, Peter, N. K. Langford, D. Kundys, J. C. Gates, B. J. Smith,
Diode Lasers and Photonic Integrated Circuits, Microwave P. G. R. Smith, and I. A. Walmsley, Science 339, 798
and Optical Engineering, Vol. 218 (Wiley, Hoboken, NY, (2013).
2012). [35] A. Aspuru-Guzik and P. Walther, Nat. Phys. 8, 285 (2012).
183901-5
SUPPLEMENTARY INFORMATION :
An electrically injected photon-pair source at room temperature
Fabien Boitier,1 Adeline Orieux,1 Claire Autebert,1 Aristide Lemaı̂tre,2 Elisabeth

Galopin,2 Christophe Manquest, Carlo Sirtori, Ivan Favero, Giuseppe Leo, and Sara Ducci
1
Laboratoire Matériaux et Phénomènes Quantiques, Université Paris Diderot, Sorbonne
Paris Cité, CNRS-UMR 7162, Case courrier 7021, 75205 Paris Cedex 13, France
2
PHOTON CORRELATION MEASUREMENTS ear dependence of the number of created photon pairs
ON LARGE TIME DELAYS AND SPDC on the laser power (and consequently on the current in-
DEPENDENCE ON THE INJECTED CURRENT jected into the device). Figure ?? reports the zero delay
time coincidence value as a function of injected current.
The characterization of the time correlations between Three regimes are clearly visible: below the laser thresh-
the photons emitted by our sources has been completed old no peak emerges from the background noise (see in-
through additional measurements performed on a device set in Fig. ??). For a current value between 570 mA and
on the same wafer. This device has a laser threshold of 780 mA a SPDC peak appears and increases linearly with
570 mA and presents an optimal SPDC efficiency at T = the injected current. The slope of the linear fit is 0.49
27◦ C. The sample is electrically injected with current coincidences s−1 A−1 . For current values above 800 mA
pulses having a duration of 60 ns and a repetition rate the SPDC signal drops probably due to thermal effects
of 40 kHz. This last value is higher than that used in within the device having important effects on the laser
Fig. 4 of the paper in order to explore a larger range operation and the phase matching curve. A deeper anal-
of time delays and reduce the integration time of our ysis on the data obtained for i= 786 mA allows us to
measurements. estimate a minimum value of gT2 E/T M (0). Removing the
In order to compare the zero delay coincidence value accidental coincidences flux φacc.dark involving the dark
with that corresponding to adjacent current pulses we counts of the detectors (predominantly dark-dark and
have measured the second order correlation function dark-luminescence events ∼ 8.7 × 10−3 Hz), we obtain
gT2 E/T M (τ ) on a time delay scale of 60 microseconds; the :
results are presented in Figure ??. The strong bunching
(2) φM ax − φacc.dark
observed for τ = 0 shows the high probability that a TE- gT E/T M (0) = ≈ 11.3
polarized photon and a TM-polarized one are detected φNoise − φacc.dark
within the same time bin (160 ps). The periodic signal where φNoise is the background coincidence flux (triangle)
on either side of the zero delay time is predominantly a and φMax the peak coincidence value. Work is currently
result of accidental pairs produced by the luminescence under progress to optimize the devices in order to reduce
of the sample; the value of these peaks corresponds well their laser threshold and improve the SPDC signal-to-
to the height of the background noise clearly visible in noise ratio.
Fig. 4.
Zero delay coincidence counts (s-1)
0.20 0.14
0.18 0.12 786 mA

Coincidence counts (s-1)
0.16
0.10
0.14
0.12 0.08
0.10 0.06 504 mA

0.08
0.04
0.06 -40 -20 0 20 40
0.04 0.02 Delay (ns)
0.02 0.00
0.00 0.2 0.3 0.4 0.5 0.6 0.7 0.8
-30 -20 -10 0 10 20 30 Current (A)
TE/TM photon detection time delay (µs)
FIG. S2. Zero delay time coincidence value as a function
FIG. S1. Photon correlation measurements on several cycles of the applied current. Accidental coincidences have been re-
of current pulses. moved. The insets show two coincidence histograms obtained
below and above the laser threshold.
As a final test of SPDC, we have checked the lin-
This article was downloaded by: [Sara Ducci]
On: 20 January 2015, At: 13:04
Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK
Journal of Modern Optics

Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/tmop20
Photon pair sources in AlGaAs: from electrical

injection to quantum state engineering
a a a a a a a
C. Autebert , G. Boucher , F. Boitier , A. Eckstein , I. Favero , G. Leo & S. Ducci
a
Laboratoire Matériaux et Phénomènes Quantiques, Université Paris Diderot, Paris Cedex
13, France.
Published online: 15 Jan 2015.
Click for updates
To cite this article: C. Autebert, G. Boucher, F. Boitier, A. Eckstein, I. Favero, G. Leo & S. Ducci (2015): Photon
pair sources in AlGaAs: from electrical injection to quantum state engineering, Journal of Modern Optics, DOI:
10.1080/09500340.2014.1000412
To link to this article: http://dx.doi.org/10.1080/09500340.2014.1000412
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Journal of Modern Optics, 2015
http://dx.doi.org/10.1080/09500340.2014.1000412
Photon pair sources in AlGaAs: from electrical injection to quantum state engineering
C. Autebert, G. Boucher, F. Boitier, A. Eckstein1 , I. Favero, G. Leo and S. Ducci∗
Laboratoire Matériaux et Phénomènes Quantiques, Université Paris Diderot, Paris Cedex 13, France
(Received 27 October 2014; accepted 11 December 2014)
Integrated quantum photonics is a very active field of quantum information, communication, and processing. One of
the main challenges to achieve massively parallel systems for complex operations is the generation, manipulation, and
detection of many qubits within the same chip. Here, we present our last achievements on AlGaAs quantum photonic
devices emitting nonclassical states of light at room temperature by spontaneous parametric down conversion (SPDC).
The choice of this platform combines the advantages of a mature fabrication technology, a high nonlinear coefficient, a
SPDC wavelength in the C-telecom band and the possibility of electrical injection.
Keywords: integrated quantum photonics; semiconductor devices; electrical injection; parametric down conversion;
entanglement; frequency correlations
Downloaded by [Sara Ducci] at 13:04 20 January 2015
1. Introduction nanowires, whispering-gallery mode resonators,etc.) [3–5]

In the field of integrated quantum photonics, semiconductor Since Silicon doesn’t present naturally a second-order non-
materials have a strategic position to achieve extremely linearity, the main physical process used in this case is
compact and massively parallel devices. In these last years, four-wave mixing. Sources emitting in the telecom band at
many efforts have been devoted to the development of room temperature have thus been demonstrated as making
biphoton sources of photonic circuits to manipulate cor- use of the SOI (Silicon-On-Insulator) platform, compatible
related photon pairs and of chips including several func- with the CMOS technology. However, the indirect band gap
tionalities (generation and manipulation or manipulation of silicon hinders the demonstration of electrically driven
and detection). In the context of entangled photon sources, devices. Conversely, the parametric down-conversion in
the efforts are focused on three main types of systems: direct band gap semiconductors is a process allowing to
quantum dots, silicon waveguides/resonators, and AlGaAs achieve devices working at room temperature, keeping the
waveguides/resonators. Quantum dots-based photon pair door open to the integration of the pump laser and the
sources exploit the radioactive decay of a biexciton state; nonlinear medium. Several advances have recently been
entangled photon generation has been demonstrated both reported by exploiting the process of three-wave mixing
under optical [1] and electrical [2] pumping. This approach on the AlGaAs platform which combines high conversion
allows obtaining almost deterministic sources taking into efficiencies (χ 2 of GaAs around 110 pm/V at 1550 nm) with
account that a quantum dot can emit a photon pair per the emission of photons in the near infrared. Since this
exciting pulse, and the recent progress in the efficiency material is not birefringent, different solutions have been
of photon pairs extraction. However, since the exploitation proposed to satisfy the phase-matching condition among
of this physical process requires that the thermal energy which form birefringence (consisting in inducing an arti-
of the system be inferior to the biexciton energy, these ficial birefringence by alternating GaAs and AlOx layers
sources have to be cooled to typical temperatures of some [6]), quasi-phase-matching (by fabricating devices present-
Kelvin. Optical parametric conversion offers an alterna- ing a periodical inversion of the nonlinearity [7]), modal
tive approach. Despite its nondeterministic nature, this pro- phase-matching (in which phase velocity mismatch is com-
cess is the most widely used to produce photon pairs for pensated by multimode waveguide dispersion [8,9]), and
quantum information and communications protocols. The counterpropagating phase-matching (with a pump beam
spectacular progress made in silicon photonics has recently transverse to the generated photons [10]). In this paper, we
led to the demonstration of entangled photon sources with address two main aspects: the pair production by
integrated devices having different geometries (waveguides, electrical injection, leading to extremely compact sources
∗ Corresponding author. Email: [email protected]

1 Present address: Clarendon Laboratory, University of Oxford U.K
© 2015 Taylor & Francis

2 C. Autebert et al.
directly compatible with the existing telecom network, and induces a great sensibility of the signal and idler wavelength
the biphoton state engineering with a device in a transverse on the pump wavelength. This means that the spectral sen-
pump configuration, opening the way to dense coding of sitivity range of the single-photon avalanche detectors will
information. limit the window, in which both photons of a pair can be
detected; the tunability of the laser wavelength will thus
be a key issue for our device. The internal peak power and
2. Photon pair production by electrical injection at the voltage characteristics of the device as a function of the
room temperature injected current are tested with a standard characterization
As mentioned before, the direct band gap of AlGaAs bench allowing a temperature control between 15 and 40◦ C
presents an evident interest for the realization of electri- with a Peltier module. In the first devices, in order to avoid
cally injected devices. Among the above mentioned phase- unwanted thermal drifts, we have employed current pulses
matching strategies, modal phase matching is the most of duration 120 ns and a repetition rate of 10 KHz. Figure 4
promising to monolithically integrate the laser source and (a) reports the laser emission intensity spectra as a function
the nonlinear medium into a single device [11,12]. In this of heat sink temperature, for an injected current of 422 mA;
scheme, the interacting modes can either be confined by the general trend corresponds to the theoretical temperature
homogeneous claddings [13] or by photonic band gap [14], dependence of the QW band gap (0.23 nm/◦ C) plotted in
this latter option avoiding aging problems via the reduction the same graph. Since the operation temperature of the
in the total aluminum content. device sets at the same time the laser and the phase-matching
Our device consists of a multilayer AlGaAs ridge waveg- wavelength [15], we have checked the possibility to have an
uide: two Bragg mirrors provide both a photonic band gap additional degree of freedom to control the laser emission;
confinement for a TE Bragg mode at 780 nm and total inter- Figure 4 (b) displays the laser wavelength dependence on
nal reflection claddings for TE and TM modes at 1.56 µm. the injected current for a heat sink temperature of 14◦ C.
The structure has been optimized for efficient type-II SPDC The data demonstrate the presence of a stark effect shifting
and lasing on the TE Bragg mode at room temperature. The the gain spectrum (and thus the laser wavelength emission)
spatial profile of the three interacting modes is shown in of the device [16]; this effect, presenting much shorter re-
Figure 1. The details of the epitaxial structure can be found sponse times than a temperature tuning, could indeed be
in [15]. exploited to set the working point of our devices.
The sample was grown by molecular beam epitaxy on a In order to check the possibility to drive the device in
(100) n-doped GaAs substrate. Waveguides are fabricated continuous wave operation, we have measured the output
using wet chemical etching to define 5.5–6 µm wide and power as a function of the duty cycle; Figure 5 reports our
2 µm deep ridges along the (011) crystalline axis, in order results for a device working at T = 20◦ C. The observation
to exploit the maximum nonzero nonlinear coefficient and a of a constant peak output power for a duty cycle up to
natural cleavage plane. Processing is completed by sample 66% indicates that a CW operation is possible, showing
thinning and contact metallization with Au alloys; finally, the technological maturity of the device.
samples are cleaved into 2 mm long stripes. Figure 2 shows The demonstration of photon pair emission in the
the main processing steps used to fabricate the devices. telecom range is done by performing time correlation
The simulated tuning curves of the source are given in measurements under electrical injection in pulsed regime;
Figure 3; the strong dispersion of the Bragg mode arising the emerging TE and TM photons are detected with two
from the proximity to the energy band gap of the core InGaAs single-photon avalanche photodiodes having 20%
detection efficiency and 50 ns gate, synchronized with the
current pulses. A time-to-digital converter is used to analyze
the time correlations between detected photons. Figure 6
reports the zero delay time coincidence value as a func-
tion of injected current for a typical device; for a current
value below the laser threshold, no ‘true coincidences’ peak
emerges from noise. For a current value between 570 and
780 mA, a peak appears and increases linearly with the
injected current as expected from the typical diode behavior
of the device.
Taking into account the overall transmission along the
optical path, the internal generation of the device is
Figure 1. Simulated intensity profiles of the TE Bragg mode ∼ 7×10−11 pairs per injected electron above the threshold,
(pump beam at 775 nm), and the TE and TM fundamental modes corresponding to a SPDC efficiency ∼ 10−9 pairs/pump
(twin photons modes in the telecom range). (The colour version photon. The maximum signal-to-noise ratio (SNR) mea-
of this figure is included in the online version of the journal.) sured up to now in our devices is 15, mainly limited by the
Journal of Modern Optics 3
Figure 2. Main processing steps for the fabrication of the electrically pumped device: (a) ridge definition by wet etching; (b) lower contact
deposition; (c) insulator deposition for independent addressing of each source; (d) upper contact deposition. (The colour version of this
figure is included in the online version of the journal.)
Figure 3. Numerical simulation of the type-II phase-matching resonance giving the twin photons wavelength as a function of the pump
wavelength. (The colour version of this figure is included in the online version of the journal.)
luminescence noise; an optimization work on both sample the specific application (linear optics quantum computa-
design and technological processing is under progress to tion, clocks synchronization, long-distance communication
reduce spurious luminescence. Assuming that the source in optical fibers, etc.). Indeed, their existence can either
emits a Werner state, our result allows to estimate the fidelity be an advantage for protocols based on hyperentangled
F to the Bell state |ψ+ that can be produced with the states, or a drawback when, for example, the frequency of
present value of SNR: F would be given by (1 + 3P)/4 a photon must not reveal that of its twin. In this section, we
with P = S N R/(2 + S N R) leading to a value ∼ 91%, present a source, based on a transverse pump configuration,
which is compliant with future experimental violation of allowing for a great versatility in the quantum state of the
Bell’s inequality. emitted biphoton. This device is based on the scheme of
Figure 7: a pump beam at 775 nm impinging on top of a
multilayer AlGaAs waveguide with an incidence angle θ
3. Quantum state engineering in a transverse pump
generates by SPDC two orthogonally polarized signal/idler
configuration
guided modes around 1.55 µm. Two Bragg mirrors provide
The control of frequency correlations of photon pairs is a vertical microcavity for the pump beam increasing the
a key issue for quantum information processing since conversion efficiency of the device [17,18].
different kinds of correlations are needed depending on
Figure 4. (a) Laser emission intensity as a function of wavelength and heat sink temperature. The dashed line shows the theoretical
temperature variation of the QW band gap. (b) Laser wavelength dependence on the injected current demonstrating the presence of stark
effect. (The colour version of this figure is included in the online version of the journal.)
Figure 6. Zero delay time coincidence value as a function of the

applied current. (The colour version of this figure is included in
the online version of the journal.)
By simultaneously pumping the device with the angle

+θdeg and the angle −θdeg and filtering out the additional
Figure 5. Laser output power as a function of the duty cycle of frequency nondegenerate photons, we create the superposi-
the injected current having a value of 755 mA. tion state [19]
As a consequence of the opposite propagation directions |(θdeg , −θdeg )

for signal and idler modes, two simultaneously 1 †
dωs dωi A(1) ωs , ωi , θdeg â H (ωs ) b̂V (ωi )
†
=√
phase-matched processes occur: one where the signal (which 2
we will consider copropagating with the horizontal com- †
+ A(2) ωs , ωi , −θdeg âV (ωs ) b̂ H (ωi ) |0
†
ponent of the pump beam) is horizontally polarized and (1)
the idler (which we will consider counterpropagating with
the horizontal component of the pump beam) is vertically with
1
polarized (interaction 1), and the other where the signal is A( j) (ωs , ωi , θ ) = ovl f M (ωs + ωi ) α(ωs + ωi )
vertically polarized and the idler horizontally (interaction Nj
2). The frequencies of the emitted fields are fixed by the en- 1L
2
ergy conservation of the interacting modes ; Figure 8 reports × dz ei k j(ωs ,ωi ,θ )z S(z, θ ) (2)
the maximum values of the signal and idler wavelengths as − 12 L
a function of the angle of incidence of the pump beam. The Here, N j is a normalization constant, ovl is the nonlinear
black slashed lines correspond to pump beams at +θdeg overlap integral between the interacting modes, L is the
and −θdeg such that each process generates a frequency length of the illuminated region on the top of the waveguide,
degenerate photon pair. f M (ωs + ωi ) is a function describing the Fabry–Perot effect
Figure 7. Operation principle of the device: a pump beam

impinging on top of a waveguide with an angle θ produces
twin photons via two possible interactions. Phase-matching is Figure 9. Sketch of the joint spectral amplitude obtained by
obtained automatically along the waveguide direction and through pumping the device in the vicinity of +θdeg . The dimensions
a periodic modulation of the waveguide core in the epitaxial of the spectral distributions are determined by the spatial and
direction. (The colour version of this figure is included in the spectral properties of the pump beam having a waist w p and a
online version of the journal.) pulse duration τ p . (The colour version of this figure is included in
the online version of the journal.)
photon pair emerging in the modes â H (ωs ) and b̂V (ωi )

( j = 1), or âV (ωs ) and b̂ H (ωi ) ( j = 2).
In Figure 9, we have sketched the joint spectral ampli-
tudes corresponding to the two interactions, when the device
is pumped in the vicinity of +θdeg with a pulsed beam.
A complete reconstruction of the produced polarization
state when the device is simultaneously pumped at ±θdeg
has been done by a state tomography [20,21]. If only po-
larization measurements are performed and frequencies are
not measured, it is convenient to work with the reduced
state obtained by taking the trace over frequencies. We
can, thus, calculate the reduced state and evaluated how
its purity is linked to the parameters of the system [19]. In
particular, we can link the concurrence C , a commonly used
Figure 8. Simulated signal and idler tuning curves as a function of entanglement estimator, to the joint spectral amplitudes of
the angle of incidence of the pump beam. Polarization entangled the two interactions by C = dωs dωi A(1) ωs , ωi , θdeg
photons can be generated by simultaneously pumping interaction A (2)∗ ωs , ωi , −θdeg .
1 (solid line) at +θdeg and 2 (dashed line) at −θdeg . (The colour A maximally entangled state will be obtained only if the
version of this figure is included in the online version of the
two interactions are excited with the same strength and if the
journal.)
joint spectral amplitudes characterizing interaction 1 and 2
perfectly overlap. Figure 10 illustrates this point with two
examples; case (a) corresponds to the experimental situation
of the microcavity on the pump beam, α is the spectral am- of [21], in which the two pumping beams are obtained by
ω−ω p
plitude of the pump beam given by α p (ω) = sech ω , passing a Gaussian pump beam through the center of a
with ω p the pump central frequency and ω its bandwidth, Fresnel biprism. The angle of each prism can be chosen
k j (ωs , ωi , θ ) are the phase-mismatch functions and such that it deviates the pump beam entering the first face
S(z, θ ) is the transverse spatial distribution of the pump with a normal incidence by θdeg . Case (b) corresponds to
beam. a situation in which the two incident pumping beams are
A( j) (ωs , ωi , θ ) are usually called joint spectral ampli- obtained by producing two replicas of a Gaussian beam with
tudes and give the probability amplitude of an emitted a specific type of grating designed to diffract the incident
(a) (b)

Figure 10. Squared absolute value of the difference between A(1) ωs , ωi , θdeg and A(2)∗ ωs , ωi , −θdeg using a biprism (a) or a lateral
shear grating (b) to produce the two pumping beams to generate polarization entanglement. The concurrence value is 0.86 for case (a) and
0.97 for case (b). (The colour version of this figure is included in the online version of the journal.)
intensity only into its first order with an angle ±θdeg . For joint spectral amplitude given above modifies in:
each case, concurrence is calculated : the use of the grating
allows to almost ’erase’ the which-path information in the A( j) (ωs , ωi , θ ) → A( j) (ωs , ωi , θ ) f TE (ωi ) f TM (ωs )
frequencies of the photons, and obtain a very high degree
of entanglement (C = 0.97).
Another interesting feature of this device is its versatility where f TE (ωi ) and f TM (ωs ) describe the effect of the
in the control of the degree of frequency correlations be- reflection on the waveguide facets for the generated TE
tween the two photons of the pair by an appropriate choice and TM polarized photons. In order to experimentally
of the spatial and spectral pump beam profiles [22,23]. characterize frequency correlations, we have recently
Taking into account, also the effect of the facets’ reflec- demonstrated a technique based on difference frequency
tivities, which is around 25–30% due to the index contrast generation [25,26]. When the structure is stimulated with a
between semiconductor and air [24], the expression of the CW signal beam, while it is subject to the same pumping
scheme of interest for SPDC, the biphoton wavefunction
relevant for the spontaneous experiment plays the role of
the response function of the structure that characterizes the
idler generated. Thus, by scanning the full spectrum of the
signal photons, it is possible to obtain the joint spectral
density that one would derive in a SPDC experiment. This
method allows reconstructing the joint spectral density of
a biphoton state with an unprecedented resolution (20 pm)
giving access to features that were theoretically predicted
but never observed (Figure 11). This streamlined technique
has the potential to speed up design and testing of massively
parallel integrated nonlinear sources and opens the way to a
new generation of experiments to explore hitherto unstudied
aspects of nonclassical states of light [27].
All these results demonstrate that III-V semiconductors
are a viable platform to develop integrated components
for future applications in quantum communications and
information. Note that the small birefringence associated
Figure 11. High resolution joint spectral intensity of the biphoton to these devices together with small footprint avoid the
state emitted by our device pumped with a beam having a waist necessity of additional steps (such as walk off compensa-
of 0.22 mm and a pulse duration of 3.5 ps. The observed intensity tion or interferometric schemes) to convert the correlated
modulation is due to the multiple reflections of the twin photons on
the waveguide facets. (The colour version of this figure is included photons into highly entangled states thus, facilitating the
in the online version of the journal.) implementation of large-scale architectures.
Acknowledgements [12] Bijlani, B.J.; Abolghasem, P.; Helmy, A.S. Appl. Phys. Lett.
C.A. acknowledges the Délégation Générale de l’Armement (DGA) 2013, 103, 091103-091105.
for financial support. F.B. acknowledges the Labex SEAM Pro- [13] Rossi, A.D.; Ortiz, V.; Calligaro, M.; Vinter, B.; Nagle,
gram N◦ANR-11-LABX-086 for financial support in the frame- J.; Ducci, S.; Berger, V. Semicond. Sci. Technol. 2004, 19,
work of the Program N◦ANR-11-IDEX-0005-02. This work was L99–L102.
partly supported by the French Brazilian ANR HIDE project. [14] Bijlani, B.J.; Helmy, A.S. Opt. Lett. 2009, 34, 3734–3736.
S. Ducci acknowledges the Institut Universitaire de France. [15] Boitier, F.; Orieux, A.; Autebert, C.; Lemaître, A.; Galopin,
E.; Manquest, C.; Sirtori, C.; Favero, I.; Leo, G.; Ducci, S.
Phys. Rev. Lett. 2014, 112, 183901–183904.
[16] Le Thomas, N.; Pelekanos, N.; Hatzopoulos, Z. Appl. Phys.
References Lett. 2003, 83, 1304-1306.
[1] Dousse, A.; Suffczyski, J.; Beveratos, A.; Krebs, O.; [17] Lanco, L.; Ducci, S.; Likforman, J.P.; Marcadet, X.; van
Lemaître, A.; Sagnes, I.; Bloch, J.; Voisin, P.; et al. Nature Houwelingen, J.A.W.; Zbinden, H.; Leo, G.; Berger, V. Phys.
London. 2010, 466, 217–220. Rev. Lett. 2006, 97, 173901–173904.
[2] Salter, C.L.; Stevenson, R.M.; Farrer, I.; Nicoll, C.A.; [18] Orieux, A.; Caillet, X.; Lemaître, A.; Filloux, P.; Favero, I.;
Ritchie, D.A.; Shields, A.J. Nature London. 2010, 465, Leo, G.; Ducci, S. J. Opt. Soc. Am. B: Opt. Phys. 2011, 28,
594–597. 45–51.
[3] Takesue, H.; Matsuda, N.; Kuramochi, E.; Munro, W.J.; [19] Boucher, G.; Eckstein, A.; Orieux, A.; Favero, I.; Leo, G.;
Notomi, M. Nat. Commun. 2013, 4, 2725-1–2725-7. Coudreau, T.; Keller, A.; Milman, P.; Ducci, S. Phys. Rev. A.
[4] Azzini, S.; Grassani, D.; Strain, M.; Sorel, M.; Helt, L.; Sipe, 2014, 89, 033815-1–033815-8.
J.; Liscidini, M.; Galli, M.; Bajoni, D. Opt. Express. 2012, [20] James, D.F.V.; Kwiat, P.G.; Munro, W.J.; White, A.G. Phys.
20, 23100–23107. Rev. A. 2001, 64, 052312-1–052312-15.
[5] O’Brien, J.L.; Furusawa, A.; Vučković, J. Nat. Photonics. [21] Orieux, A.; Eckstein, A.; Lemaître, A.; Filloux, P.; Favero,
2009, 3, 687–695. I.; Leo, G.; Coudreau, T.; Keller, A.; Milman, P.; Ducci, S.
[6] Fiore, A.; Berger, V.; Rosencher, E.; Bravetti, P.; Nagle, J. Phys. Rev. Lett. 2013, 110, 160502–160505.
Nature London. 1998, 391, 463–466. [22] Walton, Z.D.; Sergienko, A.V.; Saleh, B.E.A.; et al. Phys.
[7] Vodopyanov, K.L.; Levi, O.; Kuo, P.S.; Pinguet, T.J.; Harris, Rev. A. 2004, 70, 052317-1–052317-5.
J.S.; Fejer, M.M.; Gerard, B.; Becouarn, L.; Lallier, E. Opt. [23] Caillet, X.; Berger, V.; Leo, G.; Ducci, S. J. Mod. Opt. 2009,
Lett. 2004, 29, 1912–1914. 56, 232–239.
[8] Ducci, S.; Lanco, L.; Berger, V.; de Rossi, A.; Ortiz, V.; [24] Jeronimo-Moreno, Y.; Rodriguez-Benavides, S.; U’Ren,
Calligaro, M. Appl. Phys. Lett. 2004, 84, 2974–2976. A.B. Laser Phys. 2010, 20, 1221–1233.
[9] Helmy, A. Opt. Express. 2006, 14, 1243–1252. [25] Liscidini, M.; Sipe, J.E. Phys. Rev. Lett. 2013, 111, 193602–
[10] Rossi, A.D.; Berger, V. Phys. Rev. Lett. 2002, 88, 043901-1– 193605.
043901-4. [26] Eckstein, A.; Boucher, G.; Lemaître, A.; Filloux, P.; Favero,
[11] Orieux, A.; de Souza, C.E.R.; Lemaître, A.; Galopin, E.; I.; Leo, G.; Sipe, J.E.; Liscidini, M.; et al. Laser Photon. Rev.
Manquest, C.; Favero, I.; Leo, G.; Ducci, S.A. A Laser 2014, 8, L76–L80.
diode for Integrated Photon Pair Generation at Telecom [27] Douce, T.; Eckstein, A.; Walborn, S.P.; Khoury, A.Z.; Ducci,
Wavelength. Conference on Lasers and Electro-Optics, San S.; Keller, A.; Coudreau, T.; Milman, P. Sci. Rep. 2013, 3,
José, CA, 2012. pp JW4A.114. 3530-1–3530-6.
Letter Vol. 3, No. 2 / February 2016 / Optica 143
Integrated AlGaAs source of highly

indistinguishable and energy-time
entangled photons
CLAIRE AUTEBERT,1 NATALIA BRUNO,2 ANTHONY MARTIN,2 ARISTIDE LEMAITRE,3 CARMEN GOMEZ CARBONELL,3
IVAN FAVERO,1 GIUSEPPE LEO,1 HUGO ZBINDEN,2 AND SARA DUCCI1,*
1
Laboratoire Matériaux et Phénomènes Quantiques, Université Paris Diderot, Sorbonne Paris Cité, CNRS-UMR 7162, Case courrier 7021,
75205 Paris Cedex 13, France
2
Group of Applied Physics, University of Geneva, Geneva, Switzerland
3
*Corresponding author: sara.ducci@univ‑paris‑diderot.fr
Received 7 July 2015; revised 18 December 2015; accepted 19 December 2015 (Doc. ID 250898); published 2 February 2016
The generation of nonclassical states of light in miniature application; photons indistinguishability is at the heart of
chips is a crucial step toward practical implementations of fu- controlled logic gates [14], quantum networking [15] and boson
ture quantum technologies. Semiconductor materials are ideal sampling [16]. Entanglement is used to speed up algorithms [17],
for achieving extremely compact and massively parallel sys- protect encoded information [18], teleport quantum states [19],
tems and several platforms are currently under development. and reduce intrinsic uncertainty in interferometry measurement
In this context, spontaneous parametric downconversion in [20]. For these reasons, sources producing different quantum
AlGaAs devices combines the advantages of room temperature states while maintaining a high degree of compactness and inte-
operation, possibility of electrical injection, and emission in gratability are highly desirable. GaAs and its material derivatives
the telecom band. Here we report on a chip-based AlGaAs like AlGaAs present a strong case for the miniaturization of differ-
source, producing indistinguishable and energy-time en- ent quantum components in the same chip [21]. Its direct
tangled photons with a brightness of 7.2 × 106 pairs∕s and bandgap has already led to the monolithic integration of the pri-
mary laser source and the nonlinear medium into a single device
a signal-to-noise ratio of 141 12. Indistinguishability
emitting photon pairs under electrical injection at room temper-
between the photons is demonstrated via a Hong–Ou–
ature [22]. Moreover, GaAs strong electro-optical Pockels effect
Mandel experiment with a visibility of 89 3%, mainly lim-
enables a fast control and manipulation of the generated photons,
ited by the reflectivity of the chip facets, while energy-time
as recently demonstrated [23]. On the front of on-chip single
entanglement is tested via a Franson interferometer leading photon detection as well, high-efficiency superconducting nano-
to a visibility of 96 4. © 2016 Optical Society of America wire single-photon detectors have been integrated with GaAs
OCIS codes: (270.0270) Quantum optics; (190.4390) Nonlinear optics, waveguides [24]. All of these achievements consolidate the poten-
integrated optics; (130.5990) Semiconductors. tial of this platform to realize miniature chips containing the gen-
eration, manipulation, and detection of quantum states of light.
http://dx.doi.org/10.1364/OPTICA.3.000143 In this Letter, we present an AlGaAs ridge waveguide produc-
ing highly indistinguishable and entangled photon pairs at tele-
com wavelengths and room temperature. Our device has been
Photonics is playing a key role in the development of future quan- optimized for efficient type-II spontaneous parametric downcon-
tum technologies. Thanks to their propagation speed and long version (SPDC); two Bragg mirrors provide both a photonic
coherence time, photons are the most promising support for bandgap confinement for a transverse electric (TE) Bragg mode
quantum communications [1]. Moreover, integrated photonics at 783 nm [25,26] and total internal reflection claddings for TE00
technology enables the development of on-chip platforms for de- and TM00 modes at 1.56 μm. The sample is grown by molecular
manding applications of quantum simulation [2], computation beam epitaxy on a (100) GaAs substrate. It consists of a 6-period
[3], and metrology [4] and allows critical problems to be solved Al0.80 Ga0.20 As∕Al0.25 Ga0.75 As Bragg reflector (lower mirror), a
in terms of scalability and reliability [5]. In this context, semicon- 298 nm Al0.45 Ga0.55 As core, and a 6-period Al0.25 Ga0.75 As∕
ductor materials may play a central role; in particular, the gener- Al0.80 Ga0.20 As Bragg reflector (upper mirror). Waveguides are
ation of entangled photon pairs has been demonstrated using fabricated using wet chemical etching to define 5.5–6 μm wide
biexciton cascade in quantum dots [6,7] and nonlinear interaction and 5 μm deep ridges along the (011) crystalline axis in order to
in AlGaAs [8–10] or silicon-based devices [11–13]. The target exploit the maximum nonzero optical nonlinear coefficient and a
property of the generated quantum state depends on the intended natural cleavage plane. Optical propagation losses in the telecom
2334-2536/16/020143-04$15/0$15.00 © 2016 Optical Society of America

range of 0.3 and 0.5 cm−1 for the TE00 and TM00 modes, respec-
tively, are measured via a standard Fabry–Perot technique. For
this, we use a cw tunable laser (Tunics) to measure the transmit-
ted signal at the output of the sample as a function of the wave-
length [27]. This result is the state of the art for AlGaAs and of the
same order of that measured in Si-based platforms.
The ability of our device to produce indistinguishable photons
is tested through a Hong–Ou–Mandel (HOM) experiment [28];
in this type of measurement, two indistinguishable photons
enter a 50/50 beam splitter at the same time. The destructive
quantum interference makes them exit the beam splitter through
the same output, thus inducing a dip in the coincidence histo-
gram. Figure 1 shows a sketch of the experimental setup used for
this experiment. The light beam of a cw Ti:Sa laser is used to Fig. 2. Coincidence histogram of TE/TM photons passing through
excite the Bragg mode of the sample; after a spatial shaping with the FBG filter centered at 1.566 μm at T 20.1°C. The data were
a holographic mask, it is injected into the waveguide with a micro- accumulated during 100 s with a sampling resolution of 164 ps.
scope objective. Light emerging from the opposite end is collected
with a second microscope objective, a fiber coupler, and filtered
with a tunable fibered Bragg grating (FBG) having a full width at single-photon avalanche photodiodes connected to a time-to-
half-maximum (FWHM) of 10.8 nm. digital converter (TDC) are used for coincidence counting [29].
The optimum working point of our source is determined by Figure 2 shows a histogram of the recorded detection time delays
measuring the temporal correlations between the TE and trans- at temperature T 20.1°C for an internal estimated pump
verse magnetic (TM) photons. Two Stirling-cooled free running power of 625 μW in the guided mode. The coincidence to ac-
cidental ratio (CAR), an important figure of merit for a photon
pair source, is calculated by taking the number of true coinciden-
ces within the FWHM of the peak over the background signal, on
the same time window taken apart from the peak. The depend-
ence of the CAR on both the pump wavelength and the internal
pump power has been studied; a maximum value of the CAR of
141 12 is obtained for a pump wavelength around 783 nm and
an internal pump power around 625 μW, leading to a brightness
of 7.2 × 106 pairs∕s. This working point corresponds to the
phase-matching resonance of the device. For a pump wavelength
above the degeneracy point, the phase-matching condition is no
longer satisfied and below degeneracy a shift of 0.2 nm results in a
variation of 100 nm for the signal and idler wavelengths, which
are, by consequence, outside the interferential filter. The CAR
value, limited by the detector’s dark counts, is the maximum ever
obtained on a semiconductor waveguide, to our knowledge; this is
due to the low value of optical losses of our sample and the low
level of noise of the detectors employed in this work.
After the identification of the optimum working point, we
proceeded to the HOM measurement. For this we used a polari-
zation controller to align the polarizations of the photons and we
inserted an interferometer with an optical delay line on one of the
two arms, followed by a 50/50 beam splitter before the detectors.
Figure 3 reports the dip observed in the coincidence counts as a
function of the optical path length difference between the two
arms of the interferometer. This dip is a clear signature of the
destructive quantum interference between the two photons.
The shape of the dip is given by the convolution of the two wave
packets arriving at the 50/50 beam splitter. Since the signal/idler
Fig. 1. Sketch of the experimental setup used for (a) the HOM experi- photons are filtered with a rectangular FBG filter, we obtained a
ment and for (b) the Franson experiment. The pumping scheme and the
fit to the experimental data with the theoretical expression of the
collection of the photon pairs is common to the two experiments. For the
HOM experiment the photon pairs are deterministically separated with a
HOM dip,

polarizing beam splitter and then recombined in a 50/50 beam splitter. δλc
An optical delay line allows for varying of the relative arrival time of the N c A 1 − V × sinc 2πδt 2 ; (1)
λ
two photons. For the Franson experiment, the two entangled photons are
directed into an unbalanced interferometer. A piezo actuator is used to where N c is the coincidence rate, A is the coincidence rate apart
control the relative phase ϕ between its short and long arm. from the bunching region, V is the visibility, δt is the time delay,
the detectors, Δt ΔL∕c is the time difference between the

two paths of the interferometer, and τp is the pump laser coher-
ence time. As shown in Fig. 3, the coherence time of the photons
is 0.7 ps. We have thus chosen a path-length difference of the
interferometer of 2.5 ns, which is also much smaller than the
1 μs of coherence time of the cw laser pump (TOPTICA DL
100) and bigger than the timing jitter of the detectors (200 ps).
As shown in Figs. 4(a) and 4(b) the recorded histogram has
three coincidence peaks; the left and right peaks correspond to
a situation when one photon goes through the short arm (s)
and the other through the long arm (1) of the interferometer.
The middle peak results from the interference between the state
where both photons pass through the short arm and the one
Fig. 3. Results of the HOM-type experiment. The uncertainty asso- where both photons pass through the long arm. This peak results
ciated with each point has been calculated using a standard square root from the quantum interference of the post-selected state,
deviation. The line shows the fit of the data with Eq. (1). The obtained 1
net visibility is V net 89.0 2.8%. jΦi pffiffiffi jsii jsis e −i2ϕ jl ii jl is ; (3)
2
where the indices s and i stand for signal and idler.
For perfectly entangled photons, the total number of coinci-
and δλ is the FWHM spectral intensity. The two fitting param-
dences in the central peak exhibits an interference fringe of unity
eters are δλ and V ; for the first one we obtain 10.7 0, 2 nm,
visibility when the phase ϕ is varied. The two values of ϕ for the
which is in very good agreement with the FBG filter FWHM,
measurements reported in Figs. 4(a) and 4(b) have been chosen to
while for the second one we obtain a net (raw) visibility of 89.0
minimize and maximize the central peak height, respectively. The
2.8% (86.1 2.7%). This result is an unprecedented value in a
height of the satellite peaks is independent of the phase. The mea-
semiconductor waveguide; the limitation to the visibility in our
surement accuracy is 200 ps, dominated by the jitter of the de-
experiment can be attributed to the reflectivity R of our wave-
tectors. The coincidence counts corresponding to the central peak
guide facets, which is around 24% for the TE and TM modes.
as a function of ϕ are plotted in Fig. 4(c). The trend is well fitted
Thus, a coincidence event can not only be due to two photons
by a sinusoidal curve having a net (raw) visibility of 95.6 3.7%
directly transmitted by the facets, but also to one photon directly
(91.5 3.6%), which corresponds to a fidelity of 97.8 1.9%.
transmitted and one photon having experienced two reflections
In conclusion, we have demonstrated an AlGaAs device work-
before leaving the waveguide. Since in the latter case the path dif-
ing at room temperature with proven compliancy with electrical
ference for the two photons is not the same as for the former case,
pumping [22], generating photon pairs linked by a high degree of
these photons do not contribute to the dip [30]. In this case, the
maximum achievable visibility is given by
1 (a) (b)
V 90.5%; (2)
1 R ηTM ηTE
2
where ηTM and ηTE correspond to the transmission efficiency for

the two polarization modes in the sample. This expression is in
excellent agreement with our experimental results. Standard tele-
com antireflection coating on the facets of AlGaAs waveguides
allows to reach transmittivities of almost 100%; this kind of treat-
ment applied to our device would thus increase the indistiguish- (c)
ability of the emitted photons.
Among different possible entangled states, we have chosen to
produce energy-time entangled photons; this is a very convenient
format of entanglement, as it can be easily manipulated with in-
tegrated circuits and can be preserved over long distances in stan-
dard optical fibers [31]. By pumping the device with a cw laser,
the photon pairs are emitted simultaneously, but their emission
time is undetermined within the coherence time of the pump
laser. This lack of information leads to energy-time entangled
pairs, as first pointed out by Franson [32]. We have thus imple-
mented the experimental setup sketched in Fig. 1(b). Before pass-
ing through the polarizing beam splitter, the two entangled
Fig. 4. Results of the Franson experiment. (a) and (b) are histograms
photons are directed into an unbalanced interferometer. A piezo of the coincidence rate for two different phase settings, chosen to min-
actuator is used to control the relative phase ϕ between its short imize and maximize the central peak; (c) is two photon interference; the
and long arm. An essential condition to fulfill in a Franson type coincidence count rate of the central peak is plotted as a function of the
experiment is that τc ; τdet ≪ Δt ≪ τp , where τc is the coher- phase ϕ. The internal estimated pump power is 273 μW. The line show
ence time of the signal and idler photons, τdet is the jitter of the fitting curve used to estimate the visibility V net 95.6 3.7%.
entanglement and able to interfere on a beam splitter with high 9. R. T. Horn, P. Kolenderski, D. Kang, P. Abolghasem, C. Scarcella, A.
Della Frera, A. Tosi, L. G. Helt, S. V. Zhukovsky, J. E. Sipe, G.
visibility. The only solid state source featuring both of the last two
Weihs, A. S. Helmy, and T. Jennewein, Sci. Rep. 3, 2314 (2013).
properties demonstrated up to now is based on single quantum 10. P. Sarrafi, E. Y. Zhu, B. M. Holmes, D. C. Hutchings, S. Aitchison, and
dot technology [33]; with respect to that result, our device L. Qian, Opt. Lett. 39, 5188 (2014).
presents the advantage of working at room temperature and dis- 11. K. Harada, H. Takesue, H. Fukuda, T. Tsuchizawa, T. Watanabe,
playing higher values of HOM and Bell visibilities. The present K. Yamada, Y. Tokura, and S. Ichi Itabashi, Opt. Express 16, 20368
results constitute a step toward on-chip large scale photonic (2008).
12. H. Takesue, N. Matsuda, E. Kuramochi, and M. Notomi, Sci. Rep. 4,
circuit-based quantum computation and simulation. They set 3913 (2014).
the AlGaAs platform in an advantageous position for the develop- 13. D. Grassani, S. Azzini, M. Liscidini, M. Galli, M. J. Strain, M. Sorel, J. E.
ment of monolithic complex architectures exploiting photon pair Sipe, and D. Bajoni, Optica 2, 88 (2015).
generation by electrical injection and fast quantum state manipu- 14. A. Politi, J. C. F. Matthews, and J. L. O’Brien, Science 325, 1221
lation via the electro-optics effect. Moreover, the recent progress (2009).
15. H. J. Kimble, Nature 453, 1023 (2008).
on hybrid integration of III-V compound semiconductors onto 16. J. B. Spring, B. J. Metcalf, P. C. Humphreys, W. S. Kolthammer, X.-M.
silicon-on-insulator substrates allows for combining the advan- Jin, M. Barbieri, A. Datta, N. Thomas-Peter, N. K. Langford, D. Kundys,
tages of our device to a nearly complete suite of silicon photonics J. C. Gates, B. J. Smith, P. G. R. Smith, and I. A. Walmsley, Science 339,
components. These include filters, (de)multiplexers, splitters, 798 (2013).
interferometers, and photodetectors for the fabrication of novel 17. E. Knill, Nature 463, 441 (2010).
18. A. K. Ekert, Phys. Rev. Lett. 67, 661 (1991).
generations of CMOS compatible chips for quantum information 19. J. A. W. van Houwelingen, N. Brunner, A. Beveratos, H. Zbinden, and
technologies. N. Gisin, Phys. Rev. Lett. 96, 130502 (2006).
20. P. C. Humphreys, M. Barbieri, A. Datta, and I. A. Walmsley, Phys. Rev.
Funding. Agence Nationale de la Recherche (ANR-14-CE26- Lett. 111, 070403 (2013).
21. K. D. Jöns, U. Rengstl, M. Oster, F. Hargart, M. Heldmaier, S. Bounouar,
0029-01); Direction Générale de l’Armement (DGA); French
S. Bounouar, S. M. Ulrich, M. Jetter, and P. Michler, J. Phys. D 48,
RENATECH; Institut Universitaire de France; Région Ile-de- 085101 (2015).
France in the framework of DIM Nano-K; Swiss National 22. F. Boitier, A. Orieux, C. Autebert, A. Lemaître, E. Galopin, C. Manquest,
Science Foundation NCCR-QSIT. C. Sirtori, I. Favero, G. Leo, and S. Ducci, Phys. Rev. Lett. 112, 183901
(2014).
23. J. Wang, A. Santamato, P. Jiang, D. Bonneau, E. Engin, J. W.
Acknowledgment. The authors thank B. Korzh for help
Silverstone, M. Lermer, J. Beetz, M. Kamp, S. Höfling, M. G. Tanner,
with the Stirling-cooled, free-running single-photon avalanche C. M. Natarajan, R. H. Hadfield, S. N. Dorenbos, V. Zwiller, J. L.
photodiodes and Carlos Eduardo R. de Souza for the fabrication O’Brien, and M. Thompson, Opt. Commun. 327, 49 (2014).
of the holographic mask. 24. J. P. Sprengers, A. Gaggero, D. Sahin, S. Jahanmirinejad, G. Frucci, F.
Mattioli, R. Leoni, J. Beetz, M. Lermer, M. Kamp, S. Hfling, R. Sanjines,
and A. Fiore, Appl. Phys. Lett. 99, 2013 (2011).
REFERENCES 25. P. Yeh and A. Yariv, Opt. Commun. 19, 427 (1976).
1. N. Gisin and R. T. Thew, Nat. Photonics 1, 165 (2007). 26. A. S. Helmy, Opt. Express 14, 1243 (2006).
2. A. Aspuru-Guzik and P. Walther, Nat. Phys. 8, 285 (2012). 27. A. De Rossi, V. Ortiz, M. Calligaro, L. Lanco, S. Ducci, V. Berger, and
3. A. Crespi, R. Ramponi, R. Osellame, L. Sansoni, I. Bongioanni, F. I. Sagnes, J. Appl. Phys. 97, 073105 (2005).
Sciarrino, G. Vallone, and P. Mataloni, Nat. Commun. 2, 566 (2011). 28. C. Hong, Z. Ou, and L. Mandel, Phys. Rev. Lett. 59, 2044 (1987).
4. V. Giovannetti, S. Lloyd, and L. Maccone, Nat. Photonics 5, 222 (2011). 29. B. Korzh, N. Walenta, T. Lunghi, N. Gisin, and H. Zbinden, Appl. Phys.
5. S. Tanzilli, A. Martin, F. Kaiser, M. P. De Micheli, O. Alibart, and D. B. Lett. 104, 081108 (2014).
Ostrowsky, Laser Photon. Rev. 6, 115 (2012). 30. X. Caillet, A. Orieux, A. Lemaître, P. Filloux, I. Favero, G. Leo, and
6. R. J. Young, R. M. Stevenson, P. Atkinson, K. Cooper, D. A. Ritchie, and S. Ducci, Opt. Express 18, 9967 (2010).
A. J. Shields, New J. Phys. 8, 29 (2006). 31. B. Korzh, C. C. W. Lim, R. Houlmann, N. Gisin, M. J. Li, D. Nolan,
7. A. Dousse, J. Suffczyński, A. Beveratos, O. Krebs, A. Lemaître, I. B. Sanguinetti, R. T. Thew, and H. Zbinden, Nat. Photonics 9, 163
Sagnes, J. Bloch, P. Voisin, and P. Senellart, Nature 466, 217 (2010). (2015).
8. A. Orieux, A. Eckstein, A. Lemaître, P. Filloux, I. Favero, G. Leo, T. 32. J. Franson, Phys. Rev. Lett. 62, 2205 (1989).
Coudreau, A. Keller, P. Milman, and S. Ducci, Phys. Rev. Lett. 110, 33. M. Müller, S. Bounouar, K. D. Jöns, M. Glässl, and P. Michler, Nat.
160502 (2013). Photonics 8, 224 (2014).
technologies
Article
Electrically Injected Twin Photon Emitting Lasers at
Room Temperature
Claire Autebert 1 , Giorgio Maltese 1 , Yacine Halioua 1 , Fabien Boitier 1,2 , Aristide Lemaître 3 ,
Maria Amanti 1 , Carlo Sirtori 1 and Sara Ducci 1, *
1 Laboratoire Matériaux et Phénomènes Quantiques, Université Paris Diderot, Sorbonne Paris Cité,
Paris 75205, France; [email protected] (C.A.);
[email protected] (G.M.); [email protected] (Y.H.);
[email protected] (F.B.); [email protected] (M.A.);
[email protected] (C.S.)
2 Nokia Bell Labs, Route de Villejust, Nozay 91620, France
3 Centre de Nanosciences et de Nanotechnologies, CNRS/Université Paris Sud, Marcoussis 91460, France;
[email protected]
* Correspondence: [email protected]; Tel.: +33-1-5727-6225
Academic Editor: Stephan Reitzenstein

Received: 29 June 2016; Accepted: 8 August 2016; Published: 18 August 2016
Abstract: On-chip generation, manipulation and detection of nonclassical states of light are some of
the major issues for quantum information technologies. In this context, the maturity and versatility
of semiconductor platforms are important assets towards the realization of ultra-compact devices.
In this paper we present our work on the design and study of an electrically injected AlGaAs photon
pair source working at room temperature. The device is characterized through its performances
as a function of temperature and injected current. Finally we discuss the impact of the device’s
properties on the generated quantum state. These results are very promising for the demonstration
of electrically injected entangled photon sources at room temperature and let us envision the use of
III-V semiconductors for a widespread diffusion of quantum communication technologies.
Keywords: entanglement production; semiconductor laser; integrated quantum optics
1. Introduction
Integrated quantum photonics is a very active field of quantum information science.
In particular, the maturity of semiconductor technology offers a huge potential to build ultra-compact
devices including generation, manipulation and detection of many quantum bits. In these last
years, spectacular progress has been done on different material platforms, such as AlGaAs,
silicon-on-insulator [1,2], silica-on-silicon [3,4], as well as on their hybrid integration, e.g., with
superconducting detectors [5]. In this context, the direct band gap and electro-optics effect
characterizing III-V semiconductors are important assets for the achievement of electrically injected,
tunable, integrated quantum photonic devices. While quantum dots-based sources working at
cryogenic temperatures have allowed us to reach unprecedented levels of brightness and to tune
the degree of photon indistinguishability [6], spontaneous parametric down-conversion (SPDC) in
AlGaAs has allowed us to produce entanglement on various degrees of freedom with devices working
at room temperature and telecom wavelength (see Section 3).
Recently, the utilization of this platform has led to the demonstration of a completely
integrated electrically injected device consisting of a quantum-well laser emitting photons at 780 nm
that are converted into telecom-wavelength photon pairs by internal spontaneous parametric
down-conversion [7].
Technologies 2016, 4, 24; doi:10.3390/technologies4030024 www.mdpi.com/journal/technologies

Technologies 2016, 4, 24 2 of 7
In this paper we study the behavior of this device as a function of temperature and of the
In this
injected paper and
current we study the behavior
we discuss of thisofdevice
the impact as a function
the device’s of temperature
properties and of thequantum
on the generated injected
current
state. and we discuss the impact of the device’s properties on the generated quantum state.
2. Results
2. Results
2.1.
2.1. Design
Design and
and Fabrication
Fabrication
The
Thechallenge
challengein in designing
designing suchsucha device lies inlies
a device simultaneously
in simultaneouslyaddressing the electrical
addressing injection
the electrical
of the laser and the efficient down-conversion of laser photons in photon
injection of the laser and the efficient down-conversion of laser photons in photon pairs. pairs.
In
Inthe
thenonlinear
nonlinearprocess,
process,thetheenergy
energyand andmomentum
momentumof ofthe
thephotons
photonsare areconserved—nothing
conserved—nothingisis
exchanged with the crystal. While energy conservation is straightforward,
exchanged with the crystal. While energy conservation is straightforward, conserving conserving the momentum the
requires
momentum requires compensating for the phase velocity mismatch of the three interacting which
compensating for the phase velocity mismatch of the three interacting photons, photons,is
usually
which is done by using
usually donebirefringent crystals. However,
by using birefringent crystals.even if GaAseven
However, present a highpresent
if GaAs second-order
a high
nonlinearity, which is beneficial for achieving efficient frequency conversion
second-order nonlinearity, which is beneficial for achieving efficient frequency conversion processes (χ(2)processes
around
110
(χ(2)pm/V
around at 110
1550pm/V
nm [8]), this material
at 1550 nm [8]), lacks natural birefringence.
this material Among different
lacks natural birefringence. techniques
Among to
different
achieve
techniques phase-matching, the recent development
to achieve phase-matching, of Bragg reflection
the recent development of Bragg waveguides [9,10] represents
reflection waveguides [9,10]
arepresents
breakthrough, since this strategy allows us to reduce the total aluminum
a breakthrough, since this strategy allows us to reduce the total aluminum content (thus avoiding
content (thus
aging problems of the device) and increases the flexibility of effective index
avoiding aging problems of the device) and increases the flexibility of effective index engineering. engineering. In ourIn
case, two Bragg mirrors provide both a photonic band gap vertical
our case, two Bragg mirrors provide both a photonic band gap vertical confinement for the laser confinement for the laser
mode
mode (a (atransverse
transverse electric
electric Bragg
Bragg mode)
mode) andandtotal
totalinternal
internalreflection
reflectioncladdings
claddingsfor forthe
thephoton-pairs
photon-pairs
modes
modes (one TE00 and one TM00). The device’s design is dictated by trade-offs that must optimize
(one TE 00 and one TM 00 ). The device’s design is dictated by trade-offs that must optimize
electrical
electricaltransport,
transport,waveguiding
waveguiding andandnonlinear interaction
nonlinear at theatsame
interaction the time
same[11].time Figure
[11]. 1a,b
Figurepresent,
1a,b
respectively, the simulated intensity profile of the laser Bragg mode and the
present, respectively, the simulated intensity profile of the laser Bragg mode and the refractive index refractive index and
doping
and dopingprofiles of theofresulting
profiles design.
the resulting design.
(a) (b) (c)

Figure1.1.(a)
Figure (a)Simulated
Simulatedintensity
intensity profile
profile for for
the the Bragg
Bragg modemode within
within the waveguide;
the waveguide; (b) Refractive
(b) Refractive index
and doping profile of the device; the Bragg mode has a higher overlap with the quantum well thanwell
index and doping profile of the device; the Bragg mode has a higher overlap with the quantum the
thanmode;
TE the TE(c)
00 mode; (c) SEM image of the sample showing the facet of the sample and the upper
SEM image of the sample showing the facet of the sample and the upper electrode with
00
electrode
the with the thick
thick electrolytic electrolytic layer.
layer.
The sample under study has been grown by molecular beam epitaxy on a (100) GaAs substrate.
The sample under study has been grown by molecular beam epitaxy on a (100) GaAs substrate.
It consists of two Al0.8Ga0.2As/Al0.25Ga0.75As Bragg reflectors of six periods each surrounding a 365 nm
It consists of two Al0.8 Ga0.2 As/Al0.25 Ga0.75 As Bragg reflectors of six periods each surrounding a
Al0.45Ga0.55As core. A 8.5 nm Al0.11Ga0.89As quantum well is embedded into the core to ensure a gain
365 nm Al0.45 Ga0.55 As core. A 8.5 nm Al0.11 Ga0.89 As quantum well is embedded into the core to ensure
peaked at 780 nm (see Figure 1a). The top and bottom Bragg reflectors are, respectively, gradually
a gain peaked at 780 nm (see Figure 1a). The top and bottom Bragg reflectors are, respectively, gradually
p-doped and n-doped, from 1 × 1017 to 2 × 1018, in order to ensure efficient electrical injection of the
p-doped and n-doped, from 1 × 1017 to 2 × 1018 , in order to ensure efficient electrical injection of the
device. An additional highly doped cap layer is added to protect the device and facilitate electrical
device. An additional highly doped cap layer is added to protect the device and facilitate electrical
injection from the upper contact [12]. The device is processed in a standard waveguide geometry, as
injection from the upper contact [12]. The device is processed in a standard waveguide geometry, as
presented in Figure 1c: 5–6 µm wide and 2 µm deep waveguides are defined by wet etching of the
presented in Figure 1c: 5–6 µm wide and 2 µm deep waveguides are defined by wet etching of the
top Bragg reflector. A SiN layer is used for electrical insulation and Ti/Au and Ni/Ge/Au/Ni/Au
top Bragg reflector. A SiN layer is used for electrical insulation and Ti/Au and Ni/Ge/Au/Ni/Au
layer sequences are used for the top and bottom contact, respectively. A thick electrolytic Au layer is
As the temperature decreases from Troom = 291 K to T = 202 K, the threshold current decreases
exponentially as shown in Figure 2b, according to Ith ~ eT/T0, with T0 = 87.8 K [13].
These experimental results are compared with the numerical simulations we have developed to
design our device; the carrier transport and recombination in the hetero-junction are calculated
using a self-consistent solution of the drift-diffusion and Poisson equations. The optical parameters
of our modeling are the Bragg mode internal losses, whose value α = 35 cm−1 is extracted from the
experimental results obtained by a second-harmonic generation measurement [14]; the overlap
layer sequences
integral between are the
used for the
Bragg modetop and
andbottom contact,well
the quantum respectively.
Γ = 2.45%, A calculated
thick electrolytic
with aAu layer is
commercial
used for the top
eigenmode contact,
solver [15]; helping
and the heat facetsmanagement. Samples
modal reflectivity for are
the cleaved into approximately
Bragg mode 2 mmby
R = 0.86, calculated
long stripes.
Finite-Difference Time-Domain (FDTD). The radiative recombination time in the quantum well is set
to τrad = 5 × 10−17 m3·s−1, while the Shockley-Read-Hall non-radiative recombination processes is τSRH =
2.2.
5 nsOptoelectronic Characterization
[16]. The waveguide length and width are L = 2 mm and w = 6 µm, respectively, while the
doping
Figureprofile and aluminum
2a reports the typicalcontents correspond to the
power-current-voltage (PIV)nominal
curves values of the epitaxial
as a function structure.
of the temperature
of one of our samples under electrical pumping; the pulse duration is 60 ns at a repetitionthreshold
Figure 2b shows the good agreement between the experimental and the calculated rate of
40current as a function
kHz. These of the
curves attest fortemperature in the 216–305
the lasing emission. K range,
We observe thatwithin
at T = the
291uncertainty of theshows
K, the I-V curve device
contact surface
a resistance of R =S3.56
= 1.9Ω(±and0.2)a ×turn-on
10 cmvoltage
−4 2 . of 1.83 V, which is very close to the quantum well band
gap, thus meaning that no current-blocking effectsofoccur
From the calculated internal resistance the diode (Rint ≈ 1.5 Ω at room
at the hetero-interfaces. The P-Itemperature)
curve displays we
evaluate the
a threshold quality
current of Iof the fabricated contacts, which introduce an additional resistance lower than 2
th = 0.274 A, corresponding to threshold density current Jth = 1.5 ± 0.2 kA/cm
2.0 Ω. surface of S = 1.9 (± 0.2) × 10−4 cm2 .
forR a= contact
(a) (b)
Figure
Figure 2. 2. Laser
Laser operation.
operation. (a)(a) Voltage
Voltage biasbias (crosses)
(crosses) andand emitted
emitted power
power versus
versus injected
injected current.
current. TheThe
I-V curve is measured at T room = 291 K, while I-P for T is in the range 202–291 K. At room temperature
I-V curve is measured at Troom = 291 K, while I-P for T is in the range 202–291 K. At room temperature
thelaser
the laserhashasa threshold
a threshold ofof IthIth= =0.274
0.274A,A,a aturn
turnononvoltage
voltageofof1.81.8V;V;(b)(b) Measured(crosses)
Measured (crosses)and
and
calculated (continuous line) threshold current as a function of temperature.
calculated (continuous line) threshold current as a function of temperature. The error band around The error band around
the
the calculated
calculated threshold
threshold corresponds
corresponds to to
thethe uncertainty
uncertainty of of
thethe diode
diode surface
surface after
after the
the wet
wet etching
etching andand
cleaving processes.
cleaving processes.
The spatial intensity distribution of the emitted laser beam above the threshold is studied by
As the temperature decreases from Troom = 291 K to T = 202 K, the threshold current decreases
imaging the output facet; the recorded near-field and far-field distributions are reported in Figure
exponentially as shown in Figure 2b, according to Ith ~eT/T0 , with T0 = 87.8 K [13].
These experimental results are compared with the numerical simulations we have developed
to design our device; the carrier transport and recombination in the hetero-junction are calculated
using a self-consistent solution of the drift-diffusion and Poisson equations. The optical parameters
of our modeling are the Bragg mode internal losses, whose value α = 35 cm−1 is extracted from
the experimental results obtained by a second-harmonic generation measurement [14]; the overlap
integral between the Bragg mode and the quantum well Γ = 2.45%, calculated with a commercial
eigenmode solver [15]; and the facets modal reflectivity for the Bragg mode R = 0.86, calculated by
Finite-Difference Time-Domain (FDTD). The radiative recombination time in the quantum well is set
to τrad = 5 × 10−17 m3 ·s−1 , while the Shockley-Read-Hall non-radiative recombination processes is
τSRH = 5 ns [16]. The waveguide length and width are L = 2 mm and w = 6 µm, respectively, while the
doping profile and aluminum contents correspond to the nominal values of the epitaxial structure.
Figure 2b shows the good agreement between the experimental and the calculated threshold
current as a function of the temperature in the 216–305 K range, within the uncertainty of the device
contact surface S = 1.9 (± 0.2) × 10−4 cm2 .
From the calculated internal resistance of the diode (Rint ≈ 1.5 Ω at room temperature) we evaluate
the quality of the fabricated contacts, which introduce an additional resistance lower than R = 2.0 Ω.
3a, Technologies
showing a2016, clear
4, 24evidence of emission on the Bragg mode. For higher values of injected current4 of 7
the Bragg mode can be in competition with the fundamental TE00 mode due to index anti-guiding
effects coming from a carrier density increase. This might explain the irregular behavior observed in
The spatial
the I-P curve; furtherintensity
studies on distribution of the
this point are emitted laser beam above the threshold is studied by
in progress.
imaging
Figure 3b displays the laser emission intensity and
the output facet; the recorded near-field far-field
spectra distributions
as a function are reported
of heat-sink in Figure 3a,
temperature,
showing
measured a clear
with evidenceSpectrum
an Optical of emission on the Bragg
Analyzer mode.emitting
for a laser For higher values
on the of injected
Bragg mode atcurrent
room the
Bragg mode can be in competition with the fundamental
temperature. The general trend corresponds to the theoretical TE mode due to index anti-guiding
00 temperature dependence of the effects
quantum well band gap (0.23 nm/°C); mode hopping is clearly visible with wavelength jumps of I-P
coming from a carrier density increase. This might explain the irregular behavior observed in the
curve; further
approximately Δλstudies
≈ 1.5–2on nm.this point are in progress.
(a) (b)
Figure 3. (a)
Figure 3. Measured near- near-
(a) Measured and far-field emissions
and far-field of a laser
emissions of aemitting on the on
laser emitting Bragg
the mode;
Bragg (b)
mode;
Normalized laser emission intensity as a function of wavelength and heat-sink temperature.
(b) Normalized laser emission intensity as a function of wavelength and heat-sink temperature.
We notice that these jumps correspond to more than 35 times the spacing between adjacent
Figure 3b displays the laser emission intensity spectra as a function of heat-sink temperature,
longitudinal modes of our device; indeed, taking into account the laser length and the modal
measured with an Optical2 Spectrum Analyzer−1 for a laser emitting on the Bragg mode at room
dispersion, we find ΔλFSR = λ /2L × (n − λ dn/dλ) ≈ 40 pm [17]. This behavior has already been
temperature. The general trend corresponds to the theoretical temperature dependence of the quantum
observed in AlGaAs laser structures [18]; it seems to be explained by the presence of an additional
well band gap (0.23 nm/◦ C); mode hopping is clearly visible with wavelength jumps of approximately
saturable source of losses, resulting from deep-level traps induced in the n-type doped Bragg layers
∆λ ≈ 1.5–2 nm.
by Si donors. As the traps’ absorption is inversely proportional to the photon density, the
We notice that these jumps correspond to more than 35 times the spacing between adjacent
standing-wave intensity profile of the dominant longitudinal mode I(z) determines the periodicity of
longitudinal modes oftraps
our device; indeed, taking into account the laser length and the modal dispersion,
the optical loss pattern2 α (z) = α0/(1 + I(z)/Is−) 1[19]. By integrating along the whole device length, it is
we find ∆λFSR = λ /2L × (n − λ dn/dλ) ≈ 40 pm [17]. This behavior has already been observed
possible to show that the dominant longitudinal mode presents smaller losses than the adjacent
in AlGaAs laser structures [18]; it seems to be explained by the presence of an additional saturable
modes by a factor Δα ≈ 0.05 − 0.1α0. In order to verify that our observations are related to this
source of losses, resulting from deep-level traps induced in the n-type doped Bragg layers by Si donors.
physical process, we have estimated Δα, calculating the overlaps of the Bragg mode with each
As the traps’ absorption is inversely proportional to the photon density, the standing-wave intensity
n-doped layer. By assuming the reflectivity R = 0.86, we find α0 = 5.54 cm−1 at room temperature
profile of the dominant longitudinal mode I(z) determines the periodicity of the optical loss pattern
leading to Δα ≈ 0.28–0.55 cm−1, which is compatible with our experimental results (Figure 3b).
αtraps (z) = α0 /(1 + I(z)/Is ) [19]. By integrating along the whole device length, it is possible to show
The demonstration of the photon pair emission in the telecom range has been done by
that the dominant longitudinal mode presents smaller losses than the adjacent modes by a factor
performing time correlation measurements under electrical injection in a pulsed regime at 601 mA.
∆α ≈ 0.05 − 0.1α0 . In order to verify that our observations are related to this physical process, we
The emerging TE and TM photon pairs, corresponding to a type-II SPDC, are detected with two
have estimated ∆α, calculating the overlaps of the Bragg mode with each n-doped layer. By assuming
InGaAs single-photon avalanche photodiodes (IdQ201) having 20% detection efficiency and a 50 ns
the reflectivity R = 0.86, we find α = 5.54 cm−1 at room temperature leading to ∆α ≈ 0.28–0.55 cm−1 ,
gate, synchronized with the current 0pulses. A time-to-digital converter is used to analyze the time
which is compatible with our experimental results (Figure 3b).
correlations between cross-polarized photons, separated by a polarization beam splitter. Figure 4
The demonstration of the photon pair emission in the telecom range has been done by performing
shows a histogram of the detection time delays at T = 25 °C. The sharp peak emerging from the
time correlation measurements under electrical injection in a pulsed regime at 601 mA. The emerging
background is the evidence that photons are produced by pairs. From these data, taking into account
TE and TM photon pairs, corresponding to a type-II SPDC, are detected with two InGaAs single-photon
the optical losses along the optical path, we estimate an internal generation efficiency of the device of
avalanche photodiodes (IdQ201) having 20% detection efficiency and a 50 ns gate, synchronized with
∼10−10 pairs/injected electron.
the current pulses. A time-to-digital converter is used to analyze the time correlations between
cross-polarized photons, separated by a polarization beam splitter. Figure 4 shows a histogram of the
detection time delays at T = 25 ◦ C. The sharp peak emerging from the background is the evidence that
photons are produced by pairs. From these data, taking into account the optical losses along the optical
path, we estimate an internal generation efficiency of the device of ∼10−10 pairs/injected electron.
Technologies 2016,4,4,2424
Technologies2016, 55ofof77
0.25
(/s)
0.25
counts(/s)
0.20
Coincidencescounts
0.20
0.15
0.15
Coincidences
0.10
0.10
0.05
0.05
0.00
0.00-50 -40 -30 -20 -10 0 10 20 30 40 50
-50 -40 -30 -20 -10 0 10 20 30 40 50
Time delay (ns)
Time delay (ns)
Figure 4. Time-correlationhistogram
histogram ofTE/TM
TE/TM photonsaround
around 1.57µm µm atTT==1515◦°C. This sampleisis
Figure4.4. Time-correlation
Figure Time-correlation histogramof of TE/TM photons
photons around 1.57
1.57 µmat C. This
at T = 15 °C. This sample
sample is
electrically injectedwith
electrically with currentpulses
pulses of601
601 mA,aaduration
duration of60
60 nsand
and a repetitionrate
rate of 40
40 kHz.
electricallyinjected
injected withcurrent
current pulsesofof 601 mA,
mA, a duration of
of 60 ns
ns and aa repetition
repetition rate of
of 40 kHz.
kHz.
The
Thedata
datawere
wereaccumulated
accumulatedduring
during600
600s swith
withaasampling
samplingresolution
resolutionofof162
The data were accumulated during 600 s with a sampling resolution of 162 ps.
162ps.
ps.
3. Discussion
3. Discussion and and Conclusions
and Conclusions
Conclusions
3. Discussion
In thissection
In section wediscuss
discuss theimpact
impact ofthe the performancesof of ourdevices
devices onthe the qualityof of the
In this
this section we
we discuss the the impact of of the performances
performances of our our devices on on the quality
quality of the
the
quantum
quantum state
state of
of the
the emitted
emitted photons
photons and
and we
we give
give some
some perspectives.
perspectives.
quantum state of the emitted photons and we give some perspectives.
The signal-to-noiseratio
The ratio (SNR) is evaluated by taking the number of true coincidences within
Thesignal-to-noise
signal-to-noise ratio(SNR)(SNR)isisevaluated
evaluated byby
taking
takingthethe
number
numberof true coincidences
of true coincidenceswithin the
within
the
Full Full
Width Width
Half Half
MaximumMaximum(FWHM) (FWHM)
of the of
peakthe peak
over theover the
backgroundbackground
signal on signal
the on
same the
timesame time
window;
the Full Width Half Maximum (FWHM) of the peak over the background signal on the same time
window;
Figure Figure 5 reports
5 reports the SNR value as a function of the injected current for the same sample
window; Figurethe SNR value
5 reports the as
SNRa function
value asofathe injected
function ofcurrent for thecurrent
the injected same sample
for the characterized
same sample
characterized
in Figure 4. Forin in Figure 4.
a current For a current
value value below the laser threshold, no “true coincidence” peak
characterized Figure 4. For below the value
a current laser threshold,
below theno “true
laser coincidence”
threshold, peakcoincidence”
no “true emerges frompeakthe
emerges
noise. For from
a the
current noise.
value For a current
between 450 value
mA between
and 600 450
mA, mA
the and
SNR 600 mA,
increases the SNR
linearly increases
with the linearly
injected
emerges from the noise. For a current value between 450 mA and 600 mA, the SNR increases linearly
with the injected
current current as expected from behavior
the typicalofdiode behavior of the device.values
At high current
with theasinjected
expected from as
current the typical diode
expected from the the device.
typical diode At of
behavior high
thecurrent
device. At highthe SNR
current
values
saturates, the SNR saturates, probably due to thermal effects within the device having an impact on the
values theprobably due to probably
SNR saturates, thermal effects
due towithin
thermal theeffects
device having
within theandevice
impact on thean
having laser operation
impact on the
laser
and operation and the phase-matching curve.
laserthe phase-matching
operation curve.
and the phase-matching curve.
25 1.0
25 1.0
20 0.9
20 0.9
15 0.8
fidelity
SNR
15 0.8
fidelity
SNR
10 0.7
10 0.7
5 0.6
5 0.6
0 0.5
0 0.5
250 300 350 400 450 500 550 600
250 300 350 400 450 500 550 600
Current (mA)
Current (mA)
Figure 5. Signal-to-noise ratio and fidelity to a Bell state as a function of the current for one of
Figure 5. Signal-to-noise ratio and fidelity to a Bell state as a function of the current for one of our
Figure
our 5. Signal-to-noise ratio and fidelity to a Bell state as a function of the current for one of our
devices.
devices.
devices.
Assuming
Assumingthat thatthe
thesource
sourceemits
emitsaaWerner
Wernerstate,
state,we
wecan
canestimate
estimatethethefidelity
fidelityFFto
tothe
themaximally
maximally
Assuming that the source emits a Werner state, we can estimate the fidelity F to the maximally
entangled
entangled Bell state |ψ++>: F = (1 + 3P)/4 with P = SNR/(2 + SNR). The value of F as a function of
Bell state |ψ >: F = (1 + 3P)/4 with P = SNR/(2 + SNR). The value of F as a function ofthe
the
entangled Bell state |ψ+>: F = (1 + 3P)/4 with P = SNR/(2 + SNR). The value of F as a function of the
current
currentisisalso
alsogiven
givenininFigure 5: 5:
Figure weweobserve
observethat forfor
that an an
injected current
injected of 600
current mA,mA,
of 600 F should reachreach
F should the
current is also given in Figure 5: we observe that for an injected current of 600 mA, F should reach
value of 93%,
the value which
of 93%, is compliant
which withwith
is compliant an experimental
an experimentalviolation of Bell’s
violation inequality.
of Bell’s inequality.
the value of 93%, which is compliant with an experimental violation of Bell’s inequality.
Recently,
Recently, we have improved the fabrication method of our devices, passing from
we have improved the fabrication method of our devices, passing from wet
wet to
to dry
dry
Recently, we have improved the fabrication method of our devices, passing from wet to dry
etching
etching to
todefine
definethe
thewaveguides;
waveguides; this
thishas
hasallowed
allowedincreasing
increasingthetheconfinement
confinement and andoverlap
overlap of
ofthe
the
etching to define the waveguides; this has allowed increasing the confinement and overlap of the
interacting
interactingmodes
modes(and
(andconsequently
consequently the
the nonlinear
nonlinear process
process efficiency)
efficiency) while
whilekeeping
keeping aalow
lowlevel
levelofof
interacting modes (and consequently the nonlinear process efficiency) while keeping a low level of
optical losses. We have thus reached a SNR value of 1730, which potentially would bring us to the
condition of achieving a fidelity value of 99% [20].
In order to check the possibility to drive the device in continuous wave (CW) operation, we have
measured the output power as a function of the duty cycle: the observation of a constant peak output
power for a duty cycle up to 66% indicates that a CW operation is possible [12].
In parallel, quantum optics experiments on the photon pairs generated by passive samples
based on the same phase-matching technique have allowed us to demonstrate high values of
indistinguishability [21,22] and entanglement on polarization [23] and energy-time [22]. The quality
of the produced quantum state as well as its compatibility with the telecom network have also
allowed to use these sources in multi-user quantum key distribution protocols using standard dense
wavelength division multiplexers [24]. This approach seems also to be promising for the control of
the frequency correlations of the emitted biphoton state; for example, original designs of Alx Ga1−x N
devices have been proposed in which quasi-phase-matching of the waveguide core is used to achieve
phase-matching at the desired wavelength, while the control of waveguide dispersion is used to control
the frequency correlation between the generated photons [25].
In conclusion, Bragg reflection waveguides emerge as an extremely attractive and versatile
platform on which to perform scalable photonics-based quantum information tasks. Their ability
to be electrically driven at room temperature lets us envision the use of III-V semiconductors for a
widespread diffusion of quantum technologies.
Acknowledgments: This work has been funded by the Agence Nationale de la Recherche (ANR-14-CE26-0029-01);
the Direction Générale de l’Armement (DGA); the French RENATECH; the Institut Universitaire de France.
The authors thank A. Orieux for the design of the sample.
Author Contributions: Claire Autebert, Yacine Halioua, Fabien Boitier did the experiments, Giorgio Maltese,
Maria Amanti, Carlo Sirtori contributed to the numerical simulations and interpretation, Aristide Lemaitre did
the sample growth; Sara Ducci supervised all the stages of the work and wrote the paper.
Conflicts of Interest: The authors declare no conflict of interest.
References
1. Grassani, D.; Azzini, S.; Liscidini, M.; Galli, M.; Strain, M.J.; Sorel, M.; Sipe, J.E.; Bajoni, D. MicRometer-scale
integrated silicon source of time-energy entangled photons. Optica 2015, 2, 88–94. [CrossRef]
2. Matthews, J.C.F.; Politi, A.; Stefanov, A.; O’Brien, J.L. Manipulation of multiphoton entanglement in
waveguide quantum circuits. Nat. Photonics 2009, 3, 346–350. [CrossRef]
3. Spagnolo, N.; Vitelli, C.; Bentivegna, M.; Brod, D.J.; Crespi, A.; Flamini, F.; Giacomini, S.; Milani, G.;
Ramponi, R.; Mataloni, P.; et al. Experimental validation of photonic boson sampling. Nat. Photonics 2014, 8,
615–620. [CrossRef]
4. Spring, J.B.; Metcalf, B.J.; Humphreys, P.C.; Kolthammer, W.S.; Jin, X.M.; Barbieri, M.; Datta, A.;
Thomas-Peter, N.; Langford, N.K.; Kundys, D.; et al. Boson sampling on a photonic chip. Science 2013, 339,
798–799. [CrossRef] [PubMed]
5. Sprengers, J.P.; Gaggero, A.; Sahin, D.; Jahanmirinejad, S.; Frucci, G.; Mattioli, F.; Leoni, R.; Beetz, J.;
Lermer, M.; Kamp, M.; et al. Waveguide superconducting single-photon detectors for integrated quantum
photonic circuits. Appl. Phys. Lett. 2011, 99, 181110. [CrossRef]
6. Somaschi, N.; Giesz, V.; de Santis, L.; Loredo, J.C.; Almeida, M.P.; Hornecker, G.; Portalupi, S.L.; Grange, T.;
Antón, C.; Demory, J.; et al. Near-optimal single-photon sources in the solid state. Nat. Photonics 2016, 10,
7. Boitier, F.; Orieux, A.; Autebert, C.; Lemaitre, A.; Galopin, E.; Manquest, C.; Sirtori, C.; Favero, I.; Leo, G.;
Ducci, S. Electrically injected photon-pair source at room temperature. Phys. Rev. Lett. 2014, 112, 183901.
[CrossRef] [PubMed]
8. Shoji, I.; Kondo, T.; Ito, R. Second-order nonlinear susceptibilities of various dielectric and semiconductor
materials. Opt. Quantum Electron. 2002, 34, 797–833. [CrossRef]
9. Yeh, P.; Yariv, A. Bragg reflection waveguides. Opt. Commun. 1976, 19, 427–430. [CrossRef]
10. Helmy, A.; Bijlani, B.; Abolghasem, P. Phase matching in monolithic Bragg reflection waveguides. Opt. Lett.
2007, 32, 2399–2401. [CrossRef] [PubMed]
11. Rossi, A.D.; Ortiz, V.; Calligaro, M.; Vinter, B.; Nagle, J.; Ducci, S.; Berger, V. A third-order-mode laser diode
for quantum communication. Semicond. Sci. Technol. 2004, 19, L99–L102. [CrossRef]
12. Autebert, C.; Boucher, G.; Boitier, F.; Eckstein, A.; Favero, I.; Leo, G.; Ducci, S. Photon pair sources in algaas:
From electrical injection to quantum state engineering. J. Mod. Opt. 2015, 62, 1739–1745. [CrossRef]
13. Svelto, O.; Hanna, D.C. Principles of Lasers, 5th ed.; Springer: Berlin, Gemany, 2012.
14. Ravaro, M.; Guillotel, E.; Le Dû, M.; Manquest, C.; Marcadet, X.; Ducci, S.; Berger, V.; Leo, G. Nonlinear
measurement of mid-infrared absorption in AlOx waveguides. Appl. Phy. Lett. 2008, 92, 151111. [CrossRef]
15. Lumerical Solutions, I. Available online: http://www.lumerical.com/tcad-products/fdtd (accessed on
25 June 2016).
16. Marvin, D.C.; Moss, S.C.; Halle, L.F. Analysis of transient photoluminescence measurements on GaAs and
AlGaAs double heterostructures. J. Appl. Phys. 1992, 72, 1970–1984. [CrossRef]
17. Lanco, L. Sources Semiconductrices de Photons Jumeaux. Ph.D. Thesis, Université Paris Diderot, Paris,
France, 2006.
18. Chinone, N.; Kuroda, T.; Ohtoshi, T.; Takahashi, T.; Kajimura, T. Mode-hopping noise in index-guided
semiconductor lasers and its reduction by saturable absorbers. IEEE J. Quantum Electron. 1985, 21, 1264–1270.
[CrossRef]
19. Copeland, J. Single-mode stabilization by traps in semiconductor lasers. IEEE J. Quantum Electron. 1980, 16,
20. Autebert, C.; Lemaître, A.; Ducci, S. Highly efficient parametric down conversion in a monolithic AlGaAs
waveguide. 2016, in preparation.
21. Günthner, T.; Pressl, B.; Laiho, K.; Geßler, J.; Höfling, S.; Kamp, M.; Schneider, C.; Weihs, G. Broadband
indistinguishability from bright parametric downconversion in a semiconductor waveguide. J. Opt. 2015, 17,
125201. [CrossRef]
22. Autebert, C.; Bruno, N.; Martin, A.; Lemaitre, A.; Carbonell, C.G.; Favero, I.; Leo, G.; Zbinden, H.; Ducci, S.
Integrated algaas source of highly indistinguishable and energy-time entangled photons. Optica 2016, 3,
23. Horn, R.T.; Kolenderski, P.; Kang, D.; Abolghasem, P.; Scarcella, C.; Della Frera, A.; Tosi, A.; Helt, L.G.;
Zhukovsky, S.V.; Sipe, J.E.; et al. Inherent polarization entanglement generated from a monolithic
semiconductor chip. Sci. Rep. 2013, 3, 2314. [CrossRef] [PubMed]
24. Autebert, C.; Trapateau, J.; Orieux, A.; Lemaitre, A.; Carbonell, C.G.; Diamanti, E.; Zaquine, I.; Ducci, S.
Multi-User Quantum Key Distribution with Entangled Photons from an AlGaAs Chip. 2016. Available
online: http://arxiv.org/abs/1607.01693 (accessed on 6 July 2016).
25. Svozilyk, J.; Hendrych, M.; Helmy, A.S.; Torres, J.P. Generation of paired photons in a quantum separable
state in Bragg reflection waveguides. Opt. Express 2011, 19, 3115–3123. [CrossRef] [PubMed]
© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).
Multi-user quantum key distribution with entangled
photons from an AlGaAs chip
C Autebert1 , J Trapateau2 , A Orieux2 , A Lemaı̂tre3 , C
arXiv:1607.01693v1 [quant-ph] 6 Jul 2016
Gomez-Carbonell3 , E Diamanti2 , I Zaquine2 and S Ducci1

1
Laboratoire Matériaux et Phénomènes Quantiques, Université Paris Diderot,
Sorbonne Paris Cité, CNRS-UMR 7162, 75205 Paris Cedex 13, France
2
Laboratoire Traitement et Communication de l’Information, CNRS, Télécom
ParisTech, Université Paris-Saclay, 75013, Paris, France
3
Centre de Nanosciences et de Nanotechnologies, CNRS/Université Paris Sud, UMR
9001, 91460 Marcoussis, France
E-mail: [email protected]
July 2016
Abstract. In view of real world applications of quantum information technologies,

the combination of miniature quantum resources with existing fibre networks is
a crucial issue. Among such resources, on-chip entangled photon sources play
a central role for applications spanning quantum communications, computing and
metrology. Here, we use a semiconductor source of entangled photons operating at
room temperature in conjunction with standard telecom components to demonstrate
multi-user quantum key distribution, a core protocol for securing communications
in quantum networks. The source consists of an AlGaAs chip emitting polarization
entangled photon pairs over a large bandwidth in the main telecom band around 1550
nm without the use of any off-chip compensation or interferometric scheme; the photon
pairs are directly launched into a dense wavelength division multiplexer (DWDM)
and secret keys are distributed between several pairs of users communicating through
different channels. We achieve a visibility measured after the DWDM of 87% and show
long-distance key distribution using a 50-km standard telecom fibre link between two
network users. These results illustrate a promising route to practical, resource-efficient
implementations adapted to quantum network infrastructures.
1. Introduction
Entangled photon pairs are an important resource in quantum information including

quantum communications and computing [1, 2, 3]. In particular for quantum
communications, photons are the most suitable information carriers because of their
high transmission speed and their robustness to decoherence. To enable the wide use
of entangled photon pairs in future quantum telecommunication networks, practical
sources of photonic entanglement combined with techniques allowing to reduce the
required quantum resources are highly desirable. The motivation is therefore strong
Multi-user quantum key distribution with entangled photons from an AlGaAs chip 2
to develop compact, high-performance sources operating at room temperature that

can be easily fabricated and integrated into telecom fiber networks. In this context,
semiconductor materials play a central role; indeed, great progress has been reported in
last years on quantum devices based, for instance, on the Silicon and AlGaAs integration
platforms [4]. The latter is particularly appealing for effective miniaturization, thanks to
the direct bandgap and strong electro-optical effect exhibited by AlGaAs. Moreover, the
small birefringence of AlGaAs waveguides has led to the demonstration of polarization
entangled photon states without any off-chip compensation [5, 6, 7], which provides
a significant advantage towards achieving stable and scalable architectures with a
minimum amount of components.
At the same time, techniques widely used in classical telecommunications [8], such
as dense wavelength division multiplexing (DWDM) and active optical routing, have
been shown to be compatible with entangled photon distribution [9, 10, 11, 12, 13, 14],
and hence to effectively reduce the number of required sources in a network. Indeed, with
the use of DWDM it is possible to distribute entangled photon pairs emitted in a large
band to several user pairs at the same time, while using only one source [12, 15]. Such
schemes can then be made fully reconfigurable by adding active optical switches allowing
to distribute an entangled photon pair between any pair of users in the network; this
technique has been used recently with PPLN waveguides [10, 14] and with a periodically
poled fiber [13].
In this work, we bring together such elements to demonstrate in practice one
of the most important quantum communication protocols, namely quantum key
distribution (QKD), in a network setting. We show in particular the distribution of
quantum secret keys between multiple pairs of users using an AlGaAs source emitting
broadband entangled photons and standard telecom components for wavelength division
multiplexing. It is important to remark that although QKD can be implemented
with single photons or coherent light pulses [2], entanglement-based QKD [16, 17] is
necessary to achieve device independence [18], hence removing security loopholes due to
imperfections of practical components. In our implementation, the key information
is encoded on polarization entangled photons [13, 19], which are readily produced
by our source. This allows us to use a simple experimental setup that does not
necessitate stabilized interferometers required for systems based on time-bin or energy-
time entanglement [14]. With this setup we obtain high visibility and low quantum bit
error rate results for several pairs of users separated by up to 50 km of standard optical
fiber.
2. Experimental setup and methods
The schematic of our experimental setup is shown in Fig. 1. This setup implements
the BBM92 QKD protocol [17], which is the entanglement based generalization of the
standard BB84 protocol [20]. In this protocol, two distant users, Alice and Bob, share
entangled photon pairs in a Bell state generated by a source that can be situated in either
B29
quantum server node AlGaAs
B28
waveguide
MO1 MO2 B26
HM BB
DWDM A24
CW Ti:sa FC
laser
temperature A22
P controler F
A21
fiber links
Alice 23 Bob 27
SPAD TDC SPAD
PC HWP PBS PBS HWP PC
Figure 1: Sketch of the multi-user entangled-based quantum key distribution setup. HM:
holographic mask; BB: beam block; P: polarizer; MO1, MO2: microscope objectives; F:
longpass spectral filter; FC: fiber coupler; DWDM: dense wavelength demultiplexer; PC:
polarization controller; HWP: half-wave plate; PBS: fibered polarizion beam splitter;
SPAD: single-photon avalanche photodiode; TDC: time-to-digital converter.
Alice’s or Bob’s site or in the middle, for instance at a server node in a telecommunication
network. This node is the central element of our setup and includes the entangled photon
source and a demultiplexing device, as shown in Fig. 1.
The source consists of a multilayer AlGaAs ridge waveguide (see Fig. 2(a)), designed
to emit photon pairs through type-II spontaneous parametric down conversion (SPDC)
(see Supplementary Information for more details on the sample). The modes involved in
the nonlinear process are a TE Bragg mode at 779 nm for the pump [21, 22, 23] and TE00
and TM00 modes around the degeneracy wavelength of 1.56 µm for the photon pair. The
TE Bragg mode of the sample is excited by using the light beam of a continuous-wave
(CW) Ti:sapphire laser that is first spatially shaped with a holographic mask (HM) and
then focused to the sample through a microscope objective (MO1). The wavelengths
of the emitted photon pairs are determined by the following energy conservation and
phase-matching relations:
~ωp = ~ωA + ~ωB (1)
n TEBragg (ωp ) ωp = n TE00 (ωA ) ωA + n TM00 (ωB ) ωB (2)
n TEBragg (ωp ) ωp = n TM00 (ωA ) ωA + n TE00 (ωB ) ωB , (3)
where ωA and ωB (with ωA ≤ ωB ) are the optical angular frequencies of the down-
converted photons, ωp is the pump frequency, and ni (with i = TEBragg , TE00 , TM00 )
are the effective indices of the three guided modes. The simulated tuning curves for
a 2.1 mm-long sample pumped by a CW laser at a temperature of 20◦ C are shown in
Fig. 2(b); in these simulations we have modeled the refractive index of the materials
using Refs [24, 25]. For a given pump wavelength, the pairs emitted over the entire region
of spectral overlap between the TE00 and TM00 modes are produced in the polarization
entangled Bell state
1
|Ψi = √ (|01i + |10i)A,B (4)
2
where the logical states |0i and |1i correspond to the polarization modes TE00 and
TM00 , respectively, and A (B) corresponds to the longer (shorter) wavelength photon
of the pair. Note that due to the very small birefringence (∆n ≈ 7 × 10−3 ) of the
AlGaAs waveguide, the group velocity mismatch between the photons of a pair is so
small that no off-chip compensation of the walk-off is required to obtain the state of
Eq. (4) [5, 6]. This is a key aspect that enables the direct use of the generated pairs
and therefore the easy integration of our device in simple and robust architectures for
practical applications.
(a) (b)
Figure 2: (a) Scanning Electron Microscope image of the sample: the upper and lower
Bragg mirrors as well as the waveguide core are clearly visible. (b) Simulated tuning
curves of the type-II phase matching at T = 20◦ C.
As shown in Fig. 1, light emerging from the opposite end of the sample is collected
with a second microscope objective (MO2) and sent to a fibre coupler (FC), after filtering
out the pump wavelength with a longpass spectral filter (F). The entangled photon pairs
are then sent to four different pairs of users through the output channels of an 8-channel
dense wavelength demultiplexer (DWDM) belonging to the 100 GHz International
Telecommunication Union (ITU) grid (100 GHz channel width and 100 GHz channel
separation). We remark that when a narrowband pump laser is used, the strong anti-
correlation between the frequencies of the photons of a given pair makes them exit
through symmetric frequency channels with respect to the degeneracy frequency ωp /2.
Hence, the working point of our device is set such that the phase-matching resonance
corresponds to the central channel of the DWDM (OptoLink) used in our QKD setup;
in particular, the temperature of the sample is adjusted with a Peltier cooler while
monitoring the coincidence to accidental (CAR) ratio and the emission wavelength,
which allows us to find an optimum working point at a temperature T = 19.5◦ C. The
targeted central channel is channel 25 of the ITU grid, while channels 21-24 and 26-29
are sent to Alice and Bob, respectively. The choice of the OptoLink dielectric thin
film DWDM was the result of an analysis of the performance of various demultiplexer
technologies with respect in particular to their effect on the distribution of polarization
entangled photon pairs [12].
The photons of a given pair of users then exit the central node of our setup and
either directly enter Alice’s and Bob’s sites or first propagate through a 25-km spool
of standard single mode fibre each, leading to a 50-km total fiber link between the
users. The stations of Alice and Bob contain a polarization controller (PC), which is
required because our DWDM is a non polarization-maintaining device, followed by a
polarization analysis setup including a half-wave plate (HWP), a fibred polarization
beam-splitter (PBS) and an InGaAs free-running single-photon detector (IdQuantique
ID220). Finally, the detectors of Alice and Bob are connected to a time-to-digital
converter (TDC) in order to perform coincidence measurements.
These setups for Alice and Bob partly realize the required steps for the
BBM92 protocol, where upon reception of the entangled photon pairs, the two
communicating parties both locally measure their respective photon, choosing randomly
and independently among the computational basis {|0i, |1i} or the diagonal basis
√
{|+i, |−i} (with |±i = (|0i ± |1i)/ 2). After a basis reconciliation step and post-
processing procedures, including error correction and privacy amplification, performed
through a classical communication channel, they can extract a secret key from their
coincidence measurements in identical bases. We note here that for practical reasons
in our experiment we measured coincidences separately in the eight measurement
configurations where identical bases are chosen by Alice and Bob. The full
implementation of the random basis choice could be achieved passively by means of
a 50-50 beam-splitter routing the photon to a polarization beam-splitter aligned either
with the computational or the diagonal basis, and four single-photon detectors for each
user. Additionally, only a single detector per user is required if temporal multiplexing of
the four projected polarization states can be implemented after the polarization beam-
splitters [13].
At a high level, the information-theoretic security of the key distribution process
is guaranteed as long as the entanglement shared by Alice and Bob is sufficiently high.
More specifically, from the visibility of their coincidence measurements, Alice and Bob
can estimate the quantum bit error rate (QBER), e, of their quantum communication
link [26]:
1 − Vtot
e= , (5)
2
where Vtot is the total entanglement visibility in both the computational and the diagonal
bases:
Cmax − Cmin
Vtot = , (6)
Cmax + Cmin
where Cmax and Cmin are respectively the maximum and minimum number of
coincidences observed in both bases.
From this value of QBER and the total rate of coincidence counts, Rraw , measured
in all possible combinations of both bases, Alice and Bob can then estimate a lower
bound for their secret key rate, Rkey [27]:
1
Rkey ≥ Rraw (1 − f (e)H2 (e) − H2 (e)) , (7)
2
where the 1/2 prefactor is the basis reconciliation (or sifting) factor that accounts for
the fact that only the measurements where Alice and Bob choose the same basis can be
used for the key, which happens half of the time; the quantity f (e) corresponds to the
efficiency of the error-correction code, where in general f (e) ≥ 1 and the Shannon limit is
f (e) = 1; and H2 is the binary entropy function, H2 (x) = −x log2 (x)−(1−x) log2 (1−x).
The bound of Eq. (7) is derived against general coherent attacks [27], hence providing
maximal security guarantees.
3. Results
Using the experimental setup and techniques described above, we first performed QKD
experiments with the entangled photons entering directly the stations of Alice and Bob,
for all four symmetric channel pairs, namely 21-29; 22-28; 23-27; 24-26, of the DWDM,
corresponding to four different pairs of users sharing a secret key.
In Fig. 3 we show the measured coincidence histograms corresponding to the eight
possible projective measurements obtained when Alice and Bob make the same basis
choice, for the 23-27 channel pair. From these measurements, we can estimate in
all cases the sifted key generation rate, Rsift = 21 Rraw , and the total entanglement
visibility, Vtot (see Supplementary Information for details). The latter leads to the
calculation of the QBER (Eq. (5)) and finally taking standard values for f (e) [26, 28],
we can use Eq. (7) to estimate the secret key rate, Rkey . For the 23-27 channel pair,
we obtain Rsift = 13.8 ± 0.3 bit/s, Vtot = 86.7% ± 1.0%, e = 0.066 ± 0.005, and
Rkey = 3.28 ± 0.08 bit/s. The results for all different pairs of users are summarized
in Table 1.
In order to assess the performance of our system over long distances, we repeat the
experiments, for the 23-27 channel pair, with a spool of 25-km standard single-mode
fibre inserted after the DWDM in both user paths, leading to a total distance of 50 km
between Alice and Bob. The measured coincidence histograms in this case are shown in
the Supplementary Information. In this configuration, the QBER increases slightly to
e = 0.069 ± 0.015, and the secret key rate we obtain is Rkey = 0.21 ± 0.13 bit/s. These
results are also included in Table 1.
(a) (b)
Figure 3: Coincidence histograms of the eight different projective measurements,

corresponding to an identical choice of basis by Alice and Bob, used to estimate the
QBER and the secret key generation rate for channels 23 and 27 centered at 1558.98 nm
and 1555.75 nm, respectively, for zero distance between Alice and Bob. Data was
accumulated during 180 s with a sampling resolution of 164 ps. (a) both Alice and Bob
choose the computational basis; (b) they both choose the diagonal basis.
channel fibre link Vtot Rsift QBER Rkey

pair distance (%) (bit/s) (%) (bit/s)
21–29 0 km 90.0 ± 1.0 9.91 ± 0.23 5.0 ± 0.5 3.79 ± 0.10
22–28 0 km 83.5 ± 2.1 13.0 ± 0.27 8.2 ± 1.0 1.31 ± 0.09
0 km 86.7 ± 1.0 13.8 ± 0.3 6.6 ± 0.5 3.28 ± 0.08
23–27
50 km 86.1 ± 2.9 1.01 ± 0.06 6.9 ± 1.5 0.21 ± 0.13
24–26 0 km 87.9 ± 1.4 6.9 ± 0.2 6.1 ± 0.7 1.95 ± 0.08
Table 1: Measured and estimated quantities for the QKD protocol for four different
DWDM channel pairs centered on channel 25. Errors are calculated assuming Poisson
statistics for the coincidence counts. Finite-size effects have not been considered in these
results.
The ultimate long-distance performance of our QKD system is illustrated in Fig. 4,

where we show the two experimental points of the secret key generation rate and the
QBER as a function of the distance for the 23-27 channel pair, as well as a theoretical fit
derived from a model of our experiment taking into account all experimental parameters;
this model is based on Ref [28] and is detailed in the Supplementary Information. We
see that secret key generation is possible for a maximal distance of around 80 km in our
experimental conditions.
Figure 4: Evolution of the secret key generation rate and the QBER as a function of
the distance between Alice and Bob for the 23-27 channel pair. The theoretical fits are
based on the model of Ref [28] and take into account the quantum efficiency and dark
counts of the single-photon detectors, the collection efficiency of the setup, standard
loss of single-mode fibers and a baseline polarization measurement error; the values of
these experimental parameters are provided in the Supplementary Information.
4. Discussion
In Table 2, we provide representative results obtained in entanglement-based QKD

experiments over standard single-mode fiber links in the telecom C band together with
some of our results. A direct comparison between the various experiments in not
straightforward as the experimental conditions may vary or may be described with
respect to different parameters, such as optimal pump power [19], mean photon number
per detection window per channel pair [14], etc. It is clear, however, that these first
multi-user entanglement-based QKD results, obtained with a semiconductor chip source,
are very promising for quantum communication protocol implementations based on a
particularly simple experimental configuration.
Further progress is possible in multiple ways. The design of the source can be
optimized and the global impact of this improvement on our QKD system can be
readily evaluated: the collection efficiency is presently limited by the semiconductor
facet reflectivity (25%) and the mode mismatch between the semiconductor waveguide
and the single-mode fiber (25% coupling efficiency); the use of an antireflection coating
and of a laser-diode-to-fiber-coupler for the coupling of the emitted photons to the
fiber could improve the collection efficiency by a factor 4.2, reaching 21% instead of
the current value of 5%. Moreover, the use of commercially available superconducting
detectors with a quantum efficiency of 87% and a dark count rate of 10 count/s (see for
instance Ref [29]) would increase the detection efficiency by a factor 4.3, from 20% to
87%, and remove the influence of dark counts, leaving only spurious light related false
Ref [19] [13] [14] this work

source PPKTP crystal poled fiber PPLN WG AlGaAs WG
type I type II type 0 type II
qubit type polarization polarization energy-time polarization
pumping regime CW 81.6 MHz CW CW
SPAD ID201† ID220 ID220 & ID230 ID220
DWDM grid 3 nm†† 200 GHz 100 GHz 100 GHz
distance (km) 0 16††† 40 150 0 50
QBER 2.3% 3.5% 2.0% 6.0%‡ 6.6% 6.9%
Rkey (bits/s) 12×103 2×103 20 0.3‡ 3.3 0.21
Table 2: Relevant state-of-the art entanglement-based QKD experiments over standard

single-mode fiber links in the telecom C band.
†: Bob’s photon only (Alice’s photon at 810 nm was detected close to the source by a
silicium SPAD). ††: no DWDM was used, the photons were filtered with a 3 nm bandpass
filter. † † †: installed fiber link in Vienna. ‡: estimated from reported coincidence rates
and visibilities.
coincidences. Finally, the polarization-maintaining fibers of our polarization analysis

setups introduced polarization-mode dispersion (PMD), which limited the maximum
possible visibility to 92% (see Supplementary Information). Replacing these fibers by
standard single-mode fibers could reduce the polarization error rate from the current
value of 6% to 2.5%. Taking into account all these elements, the performance of the
setup proposed in this work could reach a secret key generation rate of 2.8 kbit/s in the
same pumping conditions at zero distance and could allow key distribution at a distance
of up to 230 km between users.
As far as the semiconductor chip is concerned, the fabrication of highly efficient low
loss waveguides is currently under intensive investigation; in particular the development
of inductively coupled plasma (ICP) processes to etch straight flank waveguides while
keeping low optical losses would increase the confinement of the interacting optical fields
and consequently the efficiency of the parametric down conversion. Encouraging results
have recently been obtained on our devices leading to a CAR on the order of 1700 [30].
In conclusion, we have implemented, for the first time, a setup allowing multi-
user QKD using a semiconductor source of entangled photons and standard telecom
components. We have performed entanglement-based quantum key distribution between
users at a distance of 50 km with a secret key rate of 0.21 bit/s. The compliance of
our source with electrical pumping [31], the possible improvement in its design together
with the continuous progress in detector fabrication techniques make our approach a
promising candidate for real-world quantum communications.
Acknowledgments
C. Autebert acknowledges the Délégation Générale de l’Armement (DGA) for funding.

We also acknowledge financial support by the Agence Nationale de la Recherche (ANR-
14-CE26-0029-01, and projects COMB and QRYPTOS), the Région Ile-de-France in
the framework of DIM Nano-K (project QUIN), the French network RENATECH, the
City of Paris (project CiQWii), and the France-USA Partner University Fund (project
CRYSP). S. Ducci is a member of the Institut Universitaire de France.
Supplementary Information
Source details. The sample consists of a 6-period Al0.80 Ga0.20 As/Al0.25 Ga0.75 As
Bragg reflector (lower cladding), a 298 nm Al0.45 Ga0.55 As core and a 6-period
Al0.25 Ga0.75 As/Al0.80 Ga0.20 As Bragg reflector (upper cladding). It is grown by molecular
beam epitaxy on a (100) GaAs substrate and waveguides are fabricated using wet chem-
ical etching to define 5.5-6 µm wide and 5 µm deep ridges along the (011) crystalline
axis, in order to use the maximum non-zero optical nonlinear coefficient and a natural
cleavage plane.
Measured coincidence data and analysis. The measured coincidence histograms

for the 50-km experiment using the 23-27 DWDM channel pair are shown in Fig. 5.
(a) (b)
Figure 5: Coincidence histograms of the eight different projective measurements used

to estimate the QBER and the secret key generation rate for the 23-27 channel pair, for
a 50 km distance between Alice and Bob. Data was accumulated during 300 s with a
sampling resolution of 164 ps. (a) both Alice and Bob choose the computational basis;
(b) they both choose the diagonal basis.
The sifted key generation rate is calculated from the obtained data as follows:
Rsift = (Cp00 + Cp01 + Cp11 + Cp10 + Cp++ + Cp+− + Cp−− + Cp−+ )/τ, (8)
where each of the eight terms is obtained from adding together the number of coinci-
dences measured in the five (seven) bins corresponding to the coincidence peak of Fig. 3
(Fig. 5) accumulated over a time τ = 3 min (5 min).
It is also possible to calculate the false coincidence rate as follows:

Rfalse = 2(C000 + C001 + C011 + C010 + C0++ + C0+− + C0−− + C0−+ )/τ, (9)
where each of the terms is obtained from the mean value of the number of coincidences
measured in five (seven) bins outside the coincidence peak of Fig. 3 (Fig. 5). As explained
below, these values were used to estimate the SPDC generation probability of our setup.
Finally, the maximum and minimum number of coincidences in both bases are given by:
Cmax = Cp01 + Cp10 + Cp++ + Cp−− (10)
Cmin = Cp00 + Cp11 + Cp+− + Cp−+ , (11)
which are inserted in Eq. (6) to calculate Vtot . (Note that Eqs. (10) and (11) only apply
if the state seen by Alice and Bob is that of Eq. (4). If the state has a π phase shift
instead: |Ψi = √12 (|01i − |10i)A,B , as was the case for the experiment with the two
spools of 25 km, the terms Cp++ + Cp−− and Cp+− + Cp−+ need to be interchanged.)
Theoretical model for secret key generation as a function of distance. The

fitting curves in Fig. 4 were calculated following the model of Ref [28], with the security
bound of Ref [27] (Eq. (7)). In particular, the error rate is given by the expression:
1
p
2 false
+ b ptrue
e= , (12)
ptrue + pfalse
where ptrue and pfalse are the probabilities of true coincidences (originating from photons
of the same pair) and accidental coincidences (from noise and lost photons) respectively,
and b is a parameter accounting for a systematic polarization measurement error. The
true and false coincidence probabilities in the case of a SPDC source in the CW pumping
regime can be expressed as [32]:
ptrue = µ η 2 (13)
2 2 2
pfalse = µ η + 8 µ η d + 16 d , (14)
where µ is the mean number of photons per pulse (in the CW regime, we can define
an effective pulse duration and an effective repetition rate, frep ); d = 4.4 × 10−6 is
the noise count probability in the single-photon detector coincidence window, which
contains the dark count probability of the detector (≈ 2 × 10−6 ) and the spurious light
count probability (luminescence noise and unfiltered ambient light); and η is the global
detection efficiency of the setup for each channel:
αL
η = ηcoll ηdet 10− 10 , (15)
where ηcoll = 5% is the collection efficiency of the setup, ηdet = 20% is the detection
efficiency of the single-photon detectors and α = 0.22 dB/km is the propagation loss
coefficient of each fibre spool of length L.
From the sifted and false coincidence rates, Eqs. (8) and (9), respectively, measured for
zero distance and for L = 25 km, and using Rsift = ptrue frep and Rfalse = pfalse frep , we
estimated the SPDC generation probability to be µ = 0.0035 for an effective repetition
rate frep = 78 MHz. Finally, from the visibility Vtot measured at zero distance, we es-
timated a polarization measurement error b = 0.06, using the relation Vtot = (1−2 b) ptrue
ptrue +pfalse
(see Eqs. (5) and (12)).
Limitation of the visibility. The polarization analysis setup at both Alice’s and
Bob’s stations consists in a fibered half-wave plate and polarization beam-splitter with
Lf ≈ 3 m of panda-style polarization-maintaining fibers. This introduces a differential
λL
group delay τPMD = cLbf ≈ 3 ps (with Lb = 5 mm the polarisation beat length of
the panda fiber). The temporal overlap between two orthogonal polarization modes
R
was thus limited to η = fA (t)fB (τPMD − t)dt ≈ 84% (where fA (t) and fB (t) are the
normalized temporal transmission functions of the DWDM channels). As a result, the
max
visibility of the measured entangled state is upper bounded by Vtot = 1+η
2
≈ 92% [12].
References
[1] Horodecki R, Horodecki P, Horodecki M and Horodecki K 2009 Rev. Mod. Phys. 81 865
[2] Scarani V, Bechmann-Pasquinucci H, Cerf N J, Dusek M, Lütkenhaus N and Peev M 2009 Rev.
Mod. Phys. 81 1301
[3] Brunner N, Cavalcanti D, Pironio S, Scarani V and Wehner S 2014 Rev. Mod. Phys. 86 419
[4] Orieux A and Diamanti E 2016 arXiv preprint arxiv:1606.00903
[5] Orieux A, Eckstein A, Lemaı̂tre A, Filloux P, Favero I, Leo G, Coudreau T, Keller A, Milman P
and Ducci S 2013 Phys. Rev. Lett. 110 160502
[6] Horn R T, Kolenderski P, Kang D, Scarcella C, Frera A D, Tosi A, Helt L G, Zhukovsky S V, Sipe
J E, Weihs G, Helmy A S and Jennewein T 2013 Sci. Rep. 3 2314
[7] Kang D, Anirban A and Helmy A S 2015 arXiv preprint arxiv:1511.00903
[8] Winzer P J 2015 Optics and Photonics News 26 28–35
[9] Lim H C, Yoshizawa A, Tsuchida H and Kikuchi K 2008 Opt. Express 16 16052–16057
[10] Herbauts I, Blauensteiner B, Poppe A, Jennewein T and Hübel H 2013 Opt. Express 21 29013–
29024
[11] Ghalbouni J, Agha I, Frey R, Diamanti E and Zaquine I 2013 Opt. Lett. 38 34
[12] Trapateau J, Ghalbouni J, Orieux A, Diamanti E and Zaquine I 2015 J. Appl. Phys. 118 143106
[13] Zhu E Y, Corbari C, Gladyshev A V, Kazansky P G, Lo H K and Qian L 2015 arXiv preprint
arxiv:1506.03896
[14] Aktas D, Fedrici B, Kaiser F, Lunghi T, Labonté L and Tanzilli S 2016 Laser and Photonics
Reviews 10 347
[15] Ciurana A, Martin V, Martinez-Mateo J, Schrenk B, Peev M and Poppe A 2015 IEEE Journal of
Selected Topics in Quantum Electronics 21 6400212
[16] Ekert A 1991 Phys. Rev. Lett. 67 661
[17] Bennett C H, Brassard G and Mermin N D 1992 Phys. Rev. Lett. 68 557–559
[18] Acı̀n A, Brunner N, Gisin N, Massar S, Pironio S and Scarani V 2007 Phys. Rev. Lett. 98 230501
[19] Treiber A, Poppe A, Hentschel M, Ferrini D, Lorünser T, Querasser E, Matyus T, Hübel H and
Zeilinger A 2009 New J. Phys. 11 045013
[20] Bennett C H and Brassard G 1984 Proc. IEEE International Conference on Computers, Systems
and Signal Processing 175 8
[21] Yeh P and Yariv A 1976 Opt. Commun. 19 427–430
[22] Helmy A S 2006 Opt. Express 14 1243
[23] Autebert C, Bruno N, Martin A, Lemaı̂tre A, Carbonell C G, Favero I, Leo G, Zbinden H and
Ducci S 2016 Optica 3 143–146
[24] Chilwell J and Hodgkinson I 1984 J. Opt. Soc. Am. A 1 742–753
[25] Gehrsitz S, Reinhart F K, Gourgon C, Herres N, Vonlanthen A and Sigg H 2000 J. Appl. Phys.
87 7825–7837
[26] Scheidl T, Ursin R, Fedrizzi A, Ramelow S, Ma X S, Herbst T, Prevedel R, Ratschbacher L, Kofler
J, Jennewein T and Zeilinger A 2009 New J. Phys. 11 085002
[27] Ma X F, Fung C H F and Lo H K 2007 Phys. Rev. A 76 012307
[28] Waks E, Zeevi A and Yamamoto Y 2002 Phys. Rev. A 65 052310
[29] Http://www.quantumopus.com/
[30] Autebert C and Ducci S 2016 in preparation
[31] Boitier F, Orieux A, Autebert C, Lemaı̂tre A, Galopin E, Manquest C, Sirtori C, Favero I, Leo G
and Ducci S 2014 Phys. Rev. Lett. 112 183901
[32] Takesue H and Shimizu K 2010 Opt. Comm. 283 276–287
Bibliography
[1] C. H. Bennett and G. Brassard, “Quantum cryptography: Public key distribution and coin
tossing”, in Proceedings of IEEE International Conference on Computers, Systems and Signal
Processing, vol. 175, p. 8, New York, 1984.
[2] A. K. Ekert, “Quantum cryptography based on Bell’s theorem”, Physical Review Letters,
vol. 67, pp. 661–663, 1991.
[3] R. P. Feynman, “Simulating physics with computers”, International Journal of Theoretical

Physics, vol. 21, no. 6, pp. 467–488, 1982.
[4] V. Giovannetti, S. Lloyd and L. Maccone, “Quantum metrology”, Physical Review Letters,
vol. 96, no. 1, p. 010401, 2006.
[5] D. Stick, W. K. Hensinger, S. Olmschenk, M. J. Madsen, K. Schwab and C. Monroe,

“Ion trap in a semiconductor chip”, Nature Physics, vol. 2, pp. 36–39, Jan. 2006.
[6] W. Hansel, P. Hommelhoff, T. W. Hansch and J. Reichel, “Bose-Einstein condensation

on a microelectronic chip”, Nature, London, vol. 413, pp. 498–501, Oct. 2001.
[7] E. Lucero, R. Barends, Y. Chen, J. Kelly, M. Mariantoni, A. Megrant, P. O’Malley,

D. Sank, A. Vainsencher, J. Wenner, T. White, Y. Yin, A. N. Cleland and J. M. Mar-
tinis, “Computing prime factors with a Josephson phase qubit quantum processor”, Nature
Physics, vol. 8, pp. 719–723, Oct. 2012.
[8] S. Laurent, S. Varoutsis, L. Le Gratiet, A. Lemaître, I. Sagnes, F. Raineri, A. Lev-

enson, I. Robert-Philip and I. Abram, “Indistinguishable single photons from a single-
quantum dot in a two-dimensional photonic crystal cavity”, Applied Physics Letters, vol. 87,
no. 16, p. 163107, 2005.
[9] A. Orieux, A. Eckstein, A. Lemaître, P. Filloux, I. Favero, G. Leo, T. Coudreau,

A. Keller, P. Milman and S. Ducci, “Direct bell states generation on a III-V semiconductor
chip at room temperature”, Physical Review Letters, vol. 110, pp. 160502–5, Apr. 2013.
[10] J. P. Sprengers, A. Gaggero, D. Sahin, S. Jahanmirinejad, G. Frucci, F. Matti-

oli, R. Leoni, J. Beetz, M. Lermer, M. Kamp, S. Hofling, R. Sanjines and A. Fiore,
“Waveguide superconducting single-photon detectors for integrated quantum photonic circuits”,
Applied Physics Letters, vol. 99, no. 18, pp. 181110–181110, 2011.
[11] N. Gisin and R. Thew, “Quantum communication”, Nature Photonics, vol. 1, no. 3, pp. 165–
171, 2007.
[12] T. D. Ladd, F. Jelezko, R. Laflamme, Y. Nakamura, C. R. Monroe and J. L. O’Brien,

“Quantum computers”, Nature, London, vol. 464, no. 7285, pp. 45–53, 2010.
145
146 BIBLIOGRAPHY
[13] M. D. Eisaman, J. Fan, A. Migdall and S. V. Polyakov, “Invited review article: Single-
photon sources and detectors”, Review of Scientific Instruments, vol. 82, no. 7, pp. 071101–25,
2011.
[14] R. Ursin, F. Tiefenbacher, T. Schmitt-Manderbach, H. Weier, T. Scheidl, M. Lin-

denthal, B. Blauensteiner, T. Jennewein, J. Perdigues, P. Trojek, B. Omer,
M. Furst, M. Meyenburg, J. Rarity, Z. Sodnik, C. Barbieri, H. Weinfurter and
A. Zeilinger, “Entanglement-based quantum communication over 144 km”, Nature Physics,
vol. 3, no. 7, pp. 481–486, 2007.
[15] D. Stucki, C. Barreiro, S. Fasel, J.-D. Gautier, O. Gay, N. Gisin, R. Thew, Y. Thoma,
P. Trinkler, F. Vannel and H. Zbinden, “Continuous high speed coherent one-way quantum
key distribution”, Optics Express, vol. 17, no. 16, pp. 13326–13334, 2009.
[16] C. Kurtsiefer, P. Zarda, M. Halder, H. Weinfurter, P. Gorman, P. Tapster and

J. Rarity, “Quantum cryptography: A step towards global key distribution”, Nature, London,
vol. 419, no. 6906, pp. 450–450, 2002.
[17] C. H. Bennett, G. Brassard, C. Crépeau, R. Jozsa, A. Peres and W. K. Wootters,

“Teleporting an unknown quantum state via dual classical and Einstein-Podolsky-Rosen chan-
nels”, Physical Review Letters, vol. 70, pp. 1895–1899, 1993.
[18] A. Fedrizzi, R. Ursin, T. Herbst, M. Nespoli, R. Prevedel, T. Scheidl, F. Tiefen-

bacher, T. Jennewein and A. Zeilinger, “High-fidelity transmission of entanglement over
a high-loss free-space channel”, Nature Physics, vol. 5, no. 6, pp. 389–392, 2009.
[19] J. F. Clauser and A. Shimony, “Bell’s theorem. experimental tests and implications”, Reports
on Progress in Physics, vol. 41, no. 12, p. 1881, 1978.
[20] A. Aspect, P. Grangier and G. Roger, “Experimental realization of Einstein-Podolsky-

Rosen-Bohm gedankenexperiment: a new violation of Bell’s inequalities”, Physical Review Let-
ters, vol. 49, pp. 91–94, 1982.
[21] D. C. Burnham and D. L. Weinberg, “Observation of simultaneity in parametric production

of optical photon pairs”, Physical Review Letters, vol. 25, pp. 84–87, 1970.
[22] K. Sanaka, K. Kawahara and T. Kuga, “New high-efficiency source of photon pairs for
engineering quantum entanglement”, Physical Review Letters, vol. 86, pp. 5620–5623, 2001.
[23] J. Fulconis, O. Alibart, J. L. O’Brien, W. J. Wadsworth and J. G. Rarity, “Nonclassical

interference and entanglement generation using a photonic crystal fiber pair photon source”,
Physical Review Letters, vol. 99, p. 120501, Sep. 2007.
[24] E. Y. Zhu, Z. Tang, L. Qian, L. G. Helt, M. Liscidini, J. E. Sipe, C. Corbari, A. Cana-

gasabey, M. Ibsen and P. G. Kazansky, “Direct generation of polarization-entangled photon
pairs in a poled fiber”, Physical Review Letters, vol. 108, no. 21, p. 213902, 2012.
[25] M. Barbier, I. Zaquine and P. Delaye, “Spontaneous four-wave mixing in liquid-core fibers:
towards fibered raman-free correlated photon sources”, New Journal of Physics, vol. 17, no. 5,
p. 053031, 2015.
[26] C. Reimer, M. Kues, P. Roztocki, B. Wetzel, F. Grazioso, B. E. Little, S. T. Chu,

T. Johnston, Y. Bromberg, L. Caspani et al., “Generation of multiphoton entangled quan-
tum states by means of integrated frequency combs”, Science, vol. 351, no. 6278, pp. 1176–1180,
2016.
BIBLIOGRAPHY 147
[27] A. Politi, J. C. F. Matthews and J. L. O’Brien, “Shor’s quantum factoring algorithm on

a photonic chip”, Science, vol. 325, p. 1221, 2009.
[28] H. J. Kimble, “The quantum internet”, Nature, vol. 453, no. 7198, pp. 1023–1030, 2008.
[29] J. B. Spring, B. J. Metcalf, P. C. Humphreys, W. Steven Kolthammer, X.-M. Jin,

M. Barbieri, A. Datta, N. Thomas-Peter, N. K. Langford, D. Kundys, J. C. Gates,
B. J. Smith, P. G. R. Smith and I. A. Walmsley, “Boson sampling on a photonic chip”,
Science, vol. 339, no. 6121, pp. 798–801, 2013.
[30] A. Lund, A. Laing, S. Rahimi-Keshari, T. Rudolph, J. O’Brien and T. Ralph, “Boson

sampling from a Gaussian state”, Physical review letters, vol. 113, no. 10, p. 100502, 2014.
[31] M. Bentivegna, N. Spagnolo, C. Vitelli, F. Flamini, N. Viggianiello, L. Latmiral,

P. Mataloni, D. J. Brod, E. F. Galvão, A. Crespi et al., “Experimental scattershot boson
sampling”, Science advances, vol. 1, no. 3, p. e1400255, 2015.
[32] E. Knill, “Physics: quantum computing”, Nature, vol. 463, no. 7280, pp. 441–443, 2010.
[33] J. Van Houwelingen, N. Brunner, A. Beveratos, H. Zbinden and N. Gisin, “Quan-

tum teleportation with a three-bell-state analyzer”, Physical Review Letters, vol. 96, no. 13,
p. 130502, 2006.
[34] P. C. Humphreys, M. Barbieri, A. Datta and I. A. Walmsley, “Quantum enhanced mul-

tiple phase estimation”, Physical Review Letters, vol. 111, p. 070403, Aug. 2013.
[35] A. Dousse, J. Suffczyński, A. Beveratos, O. Krebs, A. Lemaître, I. Sagnes, J. Bloch,

P. Voisin and P. Senellart, “Ultrabright source of entangled photon pairs”, Nature, London,
vol. 466, pp. 217–220, 2010.
[36] C. L. Salter, R. M. Stevenson, I. Farrer, C. A. Nicoll, D. A. Ritchie and A. J. Shields,

“An entangled-light-emitting diode”, Nature, London, vol. 465, no. 7298, pp. 594–597, 2010.
[37] K.-i. Harada, H. Takesue, H. Fukuda, T. Tsuchizawa, T. Watanabe, K. Yamada,

Y. Tokura and S.-i. Itabashi, “Generation of high-purity entangled photon pairs using silicon
wire waveguide”, Optics Express, vol. 16, no. 25, pp. 20368–20373, 2008.
[38] H. Takesue, N. Matsuda, E. Kuramochi and M. Notomi, “Entangled photons from on-chip
slow light”, Scientific reports, vol. 4, 2014.
[39] D. Grassani, S. Azzini, M. Liscidini, M. Galli, M. J. Strain, M. Sorel, J. E. Sipe and

D. Bajoni, “Micrometer-scale integrated silicon source of time-energy entangled photons”,
Optica, vol. 2, pp. 88–94, Feb. 2015.
[40] I. Shoji, T. Kondo and R. Ito, “Second-order nonlinear susceptibilities of various dielectric
and semiconductor materials”, Optical and Quantum Electronics, vol. 34, no. 8, pp. 797–833,
2002.
[41] R. T. Horn, P. Kolenderski, D. Kang, P. Abolghasem, C. Scarcella, A. D. Fr-

era, A. Tosi, L. G. Helt, S. V. Zhukovsky, J. E. Sipe, G. Weihs, A. S. Helmy and
T. Jennewein, “Inherent polarization entanglement generated from a monolithic semiconduc-
tor chip”, Scientific Reports, vol. 3, p. 2314, Jul. 2013.
[42] P. Sarrafi, E. Y. Zhu, B. M. Holmes, D. C. Hutchings, S. Aitchison and L. Qian,

“High-visibility two-photon interference of frequency–time entangled photons generated in a
quasi-phase-matched AlGaAs waveguide”, Optics letters, vol. 39, no. 17, pp. 5188–5191, 2014.
148 BIBLIOGRAPHY
[43] P. Kultavewuti, E. Y. Zhu, L. Qian, V. Pusino, M. Sorel and J. S. Aitchison, “Correlated

photon pair generation in AlGaAs nanowaveguides via spontaneous four-wave mixing”, Optics
Express, vol. 24, no. 4, pp. 3365–3376, 2016.
[44] T. H. Maiman, “Stimulated optical radiation in ruby”, Nature, vol. 187, pp. 493–494, Aug.
1960.
[45] P. Franken, A. Hill, C. e. Peters and G. Weinreich, “Generation of optical harmonics”,

Physical Review Letters, vol. 7, no. 4, p. 118, 1961.
[46] R. W. Boyd, Nonlinear Optics. San Diego: Academic Press, third edition éd., 2008.
[47] D. Kleinman, “Nonlinear dielectric polarization in optical media”, Physical Review, vol. 126,
no. 6, p. 1977, 1962.
[48] R. L. Sutherland, Handbook of Nonlinear Optics. Optical Science and Engineering, New
York: CRC Press, 2003.
[49] D. Magde and H. Mahr, “Study in ammonium dihydrogen phosphate of spontaneous para-
metric interaction tunable from 4400 to 16 000 å”, Physical Review Letters, vol. 18, no. 21,
p. 905, 1967.
[50] R. L. Byer and S. E. Harris, “Power and bandwidth of spontaneous parametric emission”,
Physical Review, vol. 168, no. 3, pp. 1064–1068, 1968.
[51] Z. Y. Ou, L. J. Wang and L. Mandel, “Vacuum effects on interference in two-photon down
conversion”, Physical Review A, vol. 40, pp. 1428–1435, Aug. 1989.
[52] J. Giordmaine, “Mixing of light beams in crystals”, Physical Review Letters, vol. 8, no. 1,
p. 19, 1962.
[53] J. Armstrong, N. Bloembergen, J. Ducuing and P. Pershan, “Interactions between light

waves in a nonlinear dielectric”, Physical Review, vol. 127, no. 6, p. 1918, 1962.
[54] H. Guillet de Chatellus, A. V. Sergienko, B. E. Saleh, M. C. Teich and G. Di Giuseppe,

“Non-collinear and non-degenerate polarization-entangled photon generation via concurrent
type-i parametric downconversion in ppln”, Optics Express, vol. 14, no. 21, pp. 10060–10072,
2006.
[55] M. Halder, A. Beveratos, R. T. Thew, C. Jorel, H. Zbinden and N. Gisin, “High

coherence photon pair source for quantum communication”, New Journal of Physics, vol. 10,
no. 2, p. 023027, 2008.
[56] M. Fiorentino, S. M. Spillane, R. G. Beausoleil, T. D. Roberts, P. Battle and M. W.

Munro, “Spontaneous parametric down-conversion in periodically poled ktp waveguides and
bulk crystals”, Optics Express, vol. 15, no. 12, pp. 7479–7488, 2007.
[57] K. L. Vodopyanov, O. Levi, P. S. Kuo, T. J. Pinguet, J. S. Harris, M. M. Fejer,

B. Gerard, L. Becouarn and E. Lallier, “Optical parametric oscillation in quasi-phase-
matched GaAs”, Optics Letters, vol. 29, no. 16, pp. 1912–1914, 2004.
[58] J. Ota, W. Narita, I. Ohta, T. Matsushita and T. Kondo, “Fabrication of periodically-

inverted AlGaAs waveguides for quasi-phase-matched wavelength conversion at 1.55 µm”,
Japanese Journal of Applied Physics, vol. 48, no. 4, p. 04C110, 2009.
[59] Y. Dumeige and P. Feron, “Whispering-gallery-mode analysis of phase-matched doubly res-

onant second-harmonic generation”, Physical Review A, vol. 74, no. 6, p. 063804, 2006.
BIBLIOGRAPHY 149
[60] P. S. Kuo, J. Bravo-Abad and G. S. Solomon, “Second-harmonic generation using-quasi-

phasematching in a GaAs whispering-gallery-mode microcavity”, Nature communications,
vol. 5, 2014.
[61] S. Mariani, A. Andronico, A. Lemaître, I. Favero, S. Ducci and G. Leo, “Second-

harmonic generation in AlGaAs microdisks in the telecom range”, Optics Letters, vol. 39,
pp. 3062–3065, May 2014.
[62] J. M. Dallesasse, J. N. Holonyak, A. R. Sugg, T. A. Richard and N. El-Zein, “Hy-

drolyzation oxidation of alx ga1−x as − alas − GaAs quantum well heterostructures and super-
lattices”, Applied Physics Letters, vol. 57, no. 26, pp. 2844–2846, 1990.
[63] P. Bravetti, A. Fiore, V. Berger, E. Rosencher, J. Nagle and O. Gauthier-Lafaye,

“5.2–5.6-µm source tunable by frequency conversion in a GaAs-based waveguide”, Optics letters,
vol. 23, no. 5, pp. 331–333, 1998.
[64] A. Fiore, S. Janz, L. Delobel, P. van der Meer, P. Bravetti, V. Berger, E. Rosencher
and J. Nagle, “Second-harmonic generation at l= 1.6 micro m in AlGaAs/al2o3 waveguides
using birefringence phase matching”, nTM, vol. 2, p. 1, 1998.
[65] M. Savanier, C. Ozanam, L. Lanco, X. Lafosse, A. Andronico, I. Favero, S. Ducci

and G. Leo, “Near-infrared optical parametric oscillator in a III-V semiconductor waveguide”,
Applied Physics Letters, vol. 103, p. 261105, Dec. 2013.
[66] L. Lanco, S. Ducci, J.-P. Likforman, X. Marcadet, J. A. W. van Houwelingen,

H. Zbinden, G. Leo and V. Berger, “Semiconductor waveguide source of counterpropagating
twin photons”, Physical Review Letters, vol. 97, no. 17, p. 173901, 2006.
[67] K. Moutzouris, S. Venugopal Rao, M. Ebrahimzadeh, A. De Rossi, M. Calligaro,

V. Ortiz and V. Berger, “Second-harmonic generation through optimized modal phase
matching in semiconductor waveguides”, Applied Physics Letters, vol. 83, no. 4, pp. 620–622,
2003.
[68] S. Ducci, L. Lanco, V. Berger, A. de Rossi, V. Ortiz and M. Calligaro, “Continuous-

wave second-harmonic generation in modal phase matched semiconductor waveguides”, Applied
Physics Letters, vol. 84, no. 16, pp. 2974–2976, 2004.
[69] P. Yeh and A. Yariv, “Bragg reflection waveguides”, Optics Communications, vol. 19, no. 3,
pp. 427–430, 1976.
[70] A. Helmy, “Phase matching using Bragg reflection waveguides for monolithic nonlinear optics
applications”, Optics Express, vol. 14, no. 3, pp. 1243–1252, 2006.
[71] A. L. Schawlow and C. H. Townes, “Infrared and optical masers”, Physical Review, vol. 112,
no. 6, p. 1940, 1958.
[72] R. N. Hall, G. Fenner, J. Kingsley, T. Soltys and R. Carlson, “Coherent light emission
from GaAs junctions”, Physical Review Letters, vol. 9, no. 9, p. 366, 1962.
[73] M. I. Nathan, W. P. Dumke, G. Burns, F. H. Dill Jr and G. Lasher, “Stimulated emission

of radiation from GaAs pn junctions”, Applied Physics Letters, vol. 1, pp. 62–64, 1962.
[74] L. A. Coldren, S. W. Corzine and M. L. Mashanovitch, Diode lasers and photonic inte-
grated circuits, vol. 218. John Wiley & Sons, 2012.
[75] E. Rosencher and B. Vinter, Optoelectronics. Cambridge: Cambridge University Press,

2002.
150 BIBLIOGRAPHY
[76] A. D. Rossi, V. Ortiz, M. Calligaro, B. Vinter, J. Nagle, S. Ducci and V. Berger,

“A third-order-mode laser diode for quantum communication”, Semiconductor Science and
Technology, vol. 19, no. 10, pp. L99–L102, 2004.
[77] J. Chilwell and I. Hodgkinson, “Thin-films field-transfer matrix theory of planar multilayer
waveguides and reflection from prism-loaded waveguides”, Journal of the Optical Society of
America A: Optics, Image Science, and Vision, vol. 1, no. 7, pp. 742–753, 1984.
[78] S. Gehrsitz, F. K. Reinhart, C. Gourgon, N. Herres, A. Vonlanthen and H. Sigg, “The

refractive index of Alx Ga1−x As below the band gap: Accurate determination and empirical
modeling”, Journal of Applied Physics, vol. 87, pp. 7825–7837, 2000.
[79] C. K. Hong, Z. Y. Ou and L. Mandel, “Measurement of subpicosecond time intervals between

two photons by interference”, Physical Review Letters, vol. 59, pp. 2044–2046, 1987.
[80] H. Fearn and R. Loudon, “Quantum theory of the lossless beam splitter”, Optics communi-
cations, vol. 64, no. 6, pp. 485–490, 1987.
[81] Z. Y. J. Ou, Multi-photon quantum interference. Springer Verlag, 2007.
[82] A. Einstein, B. Podolsky and N. Rosen, “Can quantum-mechanical description of physical

reality be considered complete?”, Physical Review, vol. 47, pp. 777–780, May 1935.
[83] N. Bohr, “Can quantum-mechanical description of physical reality be considered complete?

(2)”, Physical Review, vol. 48, pp. 696–702, Oct. 1935.
[84] J. S. Bell, “On the Einstein-Podolsky-Rosen paradox”, Physics, vol. 1, no. 3, pp. 195–200,
1964.
[85] J. F. Clauser, M. A. Horne, A. Shimony and R. A. Holt, “Proposed experiment to test

local hidden-variable theories”, Physical Review Letters, vol. 23, pp. 880–884, Oct. 1969.
[86] S. J. Freedman and J. F. Clauser, “Experimental test of local hidden-variable theories”,

Physical Review Letters, vol. 28, pp. 938–941, Apr. 1972.
[87] A. Aspect, P. Grangier and G. Roger, “Experimental tests of realistic local theories via
Bell’s theorem”, Physical Review Letters, vol. 47, pp. 460–463, 1981.
[88] Z. Y. Ou and L. Mandel, “Violation of bell’s inequality and classical probability in a two-
photon correlation experiment”, Physical Review Letters, vol. 61, pp. 50–53, Jul. 1988.
[89] J. G. Rarity and P. R. Tapster, “Experimental violation of bell’s inequality based on phase
and momentum”, Physical Review Letters, vol. 64, pp. 2495–2498, May 1990.
[90] P. G. Kwiat, K. Mattle, H. Weinfurter, A. Zeilinger, A. V. Sergienko and Y. Shih,

“New high-intensity source of polarization-entangled photon pairs”, Physical Review Letters,
vol. 75, pp. 4337–4341, Dec. 1995.
[91] W. Tittel, J. Brendel, H. Zbinden and N. Gisin, “Violation of bell inequalities by photons
more than 10 km apart”, Physical Review Letters, vol. 81, pp. 3563–3566, Oct. 1998.
[92] B. Hensen, H. Bernien, A. Dréau, A. Reiserer, N. Kalb, M. Blok, J. Ruitenberg,

R. Vermeulen, R. Schouten, C. Abellán et al., “Loophole-free bell inequality violation
using electron spins separated by 1.3 kilometres”, Nature, vol. 526, no. 7575, pp. 682–686,
2015.
BIBLIOGRAPHY 151
[93] L. K. Shalm, E. Meyer-Scott, B. G. Christensen, P. Bierhorst, M. A. Wayne, M. J.

Stevens, T. Gerrits, S. Glancy, D. R. Hamel, M. S. Allman et al., “Strong loophole-free
test of local realism”, Physical Review Letters, vol. 115, no. 25, p. 250402, 2015.
[94] M. Giustina, M. A. Versteegh, S. Wengerowsky, J. Handsteiner, A. Hochrainer,

K. Phelan, F. Steinlechner, J. Kofler, J.-Å. Larsson, C. Abellán et al., “Significant-
loophole-free test of bell’s theorem with entangled photons”, Physical Review Letters, vol. 115,
no. 25, p. 250401, 2015.
[95] C. H. Bennett, “Quantum cryptography using any two nonorthogonal states”, Physical Review
Letters, vol. 68, no. 21, p. 3121, 1992.
[96] C. H. Bennett and S. J. Wiesner, “Communication via one-and two-particle operators on

Einstein-podolsky-rosen states”, Physical Review Letters, vol. 69, no. 20, p. 2881, 1992.
[97] K. Mattle, H. Weinfurter, P. G. Kwiat and A. Zeilinger, “Dense coding in experimental

quantum communication”, Physical Review Letters, vol. 76, no. 25, p. 4656, 1996.
[98] D. Bouwmeester, J.-W. Pan, K. Mattle, M. Eibl, H. Weinfurter and A. Zeilinger,

“Experimental quantum teleportation”, Nature, London, vol. 390, pp. 575–579, 1997.
[99] Y. Makhlin, G. Scöhn and A. Shnirman, “Josephson-junction qubits with controlled cou-
plings”, nature, vol. 398, no. 6725, pp. 305–307, 1999.
[100] D. Leibfried, B. DeMarco, V. Meyer, D. Lucas, M. Barrett, J. Britton, W. Itano,

B. Jelenković, C. Langer, T. Rosenband et al., “Experimental demonstration of a robust,
high-fidelity geometric two ion-qubit phase gate”, Nature, vol. 422, no. 6930, pp. 412–415, 2003.
[101] J. Beugnon, C. Tuchendler, H. Marion, A. Gaëtan, Y. Miroshnychenko, Y. R. P.

Sortais, A. M. Lance, M. P. A. Jones, G. Messin, A. Browaeys and P. Grangier, “Two-
dimensional transport and transfer of a single atomic qubit in optical tweezers”, Nature Physics,
vol. 3, pp. 696–699, Oct. 2007.
[102] P. Neumann, N. Mizuochi, F. Rempp, P. Hemmer, H. Watanabe, S. Yamasaki,

V. Jacques, T. Gaebel, F. Jelezko and J. Wrachtrup, “Multipartite entanglement among
single spins in diamond”, science, vol. 320, no. 5881, pp. 1326–1329, 2008.
[103] J. Brendel, N. Gisin, W. Tittel and H. Zbinden, “Pulsed energy-time entangled twin-
photon source for quantum communication”, Physical Review Letters, vol. 82, pp. 2594–2597,
1999.
[104] L. Olislager, J. Cussey, A. T. Nguyen, P. Emplit, S. Massar, J.-M. Merolla and K. P.

Huy, “Frequency-bin entangled photons”, Physical Review A, vol. 82, p. 013804, Jul. 2010.
[105] A. Mair, A. Vaziri, G. Weihs and A. Zeilinger, “Entanglement of the orbital angular
momentum states of photons”, Nature, London, vol. 412, no. 6844, pp. 313–316, 2001.
[106] G. Vallone, R. Ceccarelli, F. De Martini and P. Mataloni, “Hyperentanglement of two

photons in three degrees of freedom”, Physical Review A, vol. 79, no. 3, p. 030301, 2009.
[107] J. D. Franson, “Bell inequality for position and time”, Physical Review Letters, vol. 62,
pp. 2205–2208, May 1989.
[108] Z. Ou, X. Zou, L. Wang and L. Mandel, “Observation of nonlocal interference in separated
photon channels”, Physical Review Letters, vol. 65, no. 3, p. 321, 1990.
152 BIBLIOGRAPHY
[109] P. G. Kwiat, W. A. Vareka, C. K. Hong, H. Nathel and R. Y. Chiao, “Correlated two-

photon interference in a dual-beam michelson interferometer”, Physical Review A, vol. 41, no. 5,
pp. 2910–2913, 1990.
[110] S. Wiesner, “Conjugate coding”, ACM Sigact News, vol. 15, no. 1, pp. 78–88, 1983.
[111] C. H. Bennett, G. Brassard and N. D. Mermin, “Quantum cryptography without bell’s

theorem”, Physical Review Letters, vol. 68, pp. 557–559, Feb. 1992.
[112] H. Inamori, N. Lütkenhaus and D. Mayers, “Unconditional security of practical quantum

key distribution”, The European Physical Journal D, vol. 41, no. 3, pp. 599–627, 2007.
[113] A. Acín, N. Brunner, N. Gisin, S. Massar, S. Pironio and V. Scarani, “Device-

independent security of quantum cryptography against collective attacks”, Physical Review
Letters, vol. 98, no. 23, p. 230501, 2007.
[114] S. Pironio, A. Acin, N. Brunner, N. Gisin, S. Massar and V. Scarani, “Device-

independent quantum key distribution secure against collective attacks”, New Journal of
Physics, vol. 11, no. 4, p. 045021, 2009.
[115] L. Lanco, Sources semiconductrices de photons jumeaux. Ph.D. dissertation, Université Paris-
Diderot, 2006.
[116] A. Orieux, Sources semiconductrices d’états à deux photons à température ambiante. Ph.D.
dissertation, Université Paris-Diderot, 2012.
[117] M. Ohashi, T. Kondo, R. Ito, S. Fukatsu, Y. Shiraki, K. Kumata and S. S. Kano,

“Determination of quadratic nonlinear optical coefficient of Alx Ga1−x As system by the method
of reflected second harmonics”, Journal of Applied Physics, vol. 74, no. 1, pp. 596–601, 1993.
[118] D. H. O. Svelto, Principles of lasers. Springer, 2012.
[119] N. Chinone, T. Kuroda, T. Ohtoshi, T. Takahashi and T. Kajimura, “Mode-hopping

noise in index-guided semiconductor lasers and its reduction by saturable absorbers”, IEEE
journal of quantum electronics, vol. 21, no. 8, pp. 1264–1270, 1985.
[120] J. Copeland, “Single-mode stabilization by traps in semiconductor lasers”, IEEE Journal of

Quantum Electronics, vol. 16, no. 7, pp. 721–727, 1980.
[121] N. Le Thomas, N. Pelekanos and Z. Hatzopoulos, “Tunable laser diodes by stark effect”,
Applied physics letters, vol. 83, no. 7, pp. 1304–1306, 2003.
[122] R. Regener and W. Sohler, “Loss in low-finesse ti: Linbo3 optical waveguide resonators”,
Applied Physics B, vol. 36, no. 3, pp. 143–147, 1985.
[123] A. D. Rossi, V. Ortiz, M. Calligaro, L. Lanco, S. Ducci, V. Berger and I. Sagnes,

“Measuring propagation loss in a multimode semiconductor waveguide”, Journal of Applied
Physics, vol. 97, no. 7, pp. 073105–7, 2005.
[124] B. Korzh, N. Walenta, T. Lunghi, N. Gisin and H. Zbinden, “Free-running InGaAs single
photon detector with 1 dark count per second at 10% efficiency”, Applied Physics Letters,
vol. 104, no. 8, p. 081108, 2014.
[125] Y. Jeronimo-Moreno, S. Rodriguez-Benavides and A. B. U’Ren, “Theory of cavity-

enhanced spontaneous parametric downconversion”, Laser physics, vol. 20, no. 5, pp. 1221–
1233, 2010.
BIBLIOGRAPHY 153
[126] M. Avenhaus, A. Eckstein, P. J. Mosley and C. Silberhorn, “Fiber-assisted single-photon

spectrograph”, Optics Letters, vol. 34, no. 18, pp. 2873–2875, 2009.
[127] G. Boucher, Biphoton Frequency-Correlations Engineering and Measurement with a Semi-

conductor Microcavity. Ph.D. dissertation, Université Paris-Diderot, 2016.
[128] M. Liscidini and J. E. Sipe, “Stimulated emission tomography”, Physical Review Letters,
vol. 111, pp. 193602–5, Nov. 2013.
[129] A. Eckstein, G. Boucher, A. Lemaître, P. Filloux, I. Favero, G. Leo, J. E. Sipe,

M. Liscidini and S. Ducci, “High-resolution spectral characterization of two photon states
via classical measurements”, Laser & Photonics Reviews, vol. 8, pp. L76–L80, Jun. 2014.
[130] J. W. Silverstone, R. Santagati, D. Bonneau, M. J. Strain, M. Sorel, J. L. O’Brien

and M. G. Thompson, “Qubit entanglement on a silicon photonic chip”, Oct. 2014.
[131] X. Caillet, A. Orieux, A. Lemaître, P. Filloux, I. Favero, G. Leo and S. Ducci,

“Two-photon interference with a semiconductor integrated source at room temperature”, Optics
Express, vol. 18, no. 10, pp. 9967–9975, 2010.
[132] E. Wigner, “On the quantum correction for thermodynamic equilibrium”, Physical Review,
vol. 40, pp. 749–759, Jun. 1932.
[133] U. Leonhardt, Measuring the quantum state of light, vol. 22. Cambridge university press,
1997.
[134] T. Douce, A. Eckstein, S. P. Walborn, A. Z. Khoury, S. Ducci, A. Keller,

T. Coudreau and P. Milman, “Direct measurement of the biphoton wigner function through
two-photon interference”, Scientific Reports, vol. 3, p. 3530, Dec. 2013.
[135] W. P. Grice, Interference and Indistinguishability in Ultrafast Spontaneous Parametric. Ph.D.

dissertation, University of Rochester, 1997.
[136] G. Boucher, T. Douce, D. Bresteau, S. P. Walborn, A. Keller, T. Coudreau, S. Ducci

and P. Milman, “Toolbox for continuous-variable entanglement production and measurement
using spontaneous parametric down-conversion”, Physical Review A, vol. 92, p. 023804, Aug.
2015.
[137] R. T. Thew, S. Tanzilli, W. Tittel, H. Zbinden and N. Gisin, “Experimental investigation

of the robustness of partially entangled qubits over 11 km”, Physical Review A, vol. 66, no. 6,
p. 062304, 2002.
[138] V. Scarani, H. Bechmann-Pasquinucci, N. J. Cerf, M. Dušek, N. Lütkenhaus and

M. Peev, “The security of practical quantum key distribution”, Reviews of modern physics,
vol. 81, no. 3, p. 1301, 2009.
[139] P. J. Winzer, “Scaling optical fiber networks: challenges and solutions”, Optics and Photonics
News, vol. 26, no. 3, pp. 28–35, 2015.
[140] H. C. Lim, A. Yoshizawa, H. Tsuchida and K. Kikuchi, “Broadband source of telecom-

band polarization-entangled photon-pairs for wavelength-multiplexed entanglement distribu-
tion”, Optics Express, vol. 16, no. 20, pp. 16052–16057, 2008.
[141] I. Herbauts, B. Blauensteiner, A. Poppe, T. Jennewein and H. Huebel, “Demonstration

of active routing of entanglement in a multi-user network”, Optics Express, vol. 21, no. 23,
pp. 29013–29024, 2013.
154 BIBLIOGRAPHY
[142] J. Ghalbouni, I. Agha, R. Frey, E. Diamanti and I. Zaquine, “Experimental wavelength-

division-multiplexed photon-pair distribution”, Optics Letters, vol. 38, pp. 34–36, Jan. 2013.
[143] J. Trapateau, J. Ghalbouni, A. Orieux, E. Diamanti and I. Zaquine, “Multi-user dis-

tribution of polarization entangled photon pairs”, Journal of Applied Physics, vol. 118, no. 14,
p. 143106, 2015.
[144] E. Zhu, C. Corbari, A. Gladyshev, P. Kazansky, H. Lo and L. Qian, “Multi-party

agile quantum key distribution network with a broadband fiber-based entangled source”, arXiv
preprint arXiv:1506.03896, 2015.
[145] D. Aktas, B. Fedrici, F. Kaiser, T. Lunghi, L. Labonté and T. Sébastien, “Entan-

glement distribution over 150 km in wavelength division multiplexed channels for quantum
cryptography”, Laser & Photon. Rev., vol. 10, pp. 451–457, 2016.
[146] A. Ciurana, V. Martin, J. Martinez-Mateo, B. Schrenk, M. Peev and A. Poppe,

“Entanglement distribution in optical networks”, IEEE Journal of Selected Topics in Quantum
Electronics, vol. 21, no. 3, pp. 37–48, 2015.
[147] T. Scheidl, R. Ursin, A. Fedrizzi, S. Ramelow, X.-S. Ma, T. Herbst, R. Prevedel,

L. Ratschbacher, J. Kofler, T. Jennewein et al., “Feasibility of 300 km quantum key
distribution with entangled states”, New Journal of Physics, vol. 11, no. 8, p. 085002, 2009.
[148] X. Ma, C.-H. F. Fung and H.-K. Lo, “Quantum key distribution with entangled photon
sources”, Physical Review A, vol. 76, no. 1, p. 012307, 2007.
[149] E. Waks, A. Zeevi and Y. Yamamoto, “Security of quantum key distribution with entangled
photons against individual attacks”, Physical Review A, vol. 65, no. 5, p. 052310, 2002.
[150] H. Takesue, K.-i. Harada, K. Tamaki, H. Fukuda, T. Tsuchizawa, T. Watanabe, K. Ya-

mada and S.-i. Itabashi, “Long-distance entanglement-based quantum key distribution exper-
iment using practical detectors”, Optics Express, vol. 18, no. 16, pp. 16777–16787, 2010.
[151] R. F. Werner, “Quantum states with Einstein-Podolsky-Rosen correlations admitting a

hidden-variable model”, Physical Review A, vol. 40, no. 8, pp. 4277–4281, 1989.

AlGaAs Photonic Devices - From Quantum State Generation To Quantum Communication

Transféré par

Droits d'auteur :

Formats disponibles

AlGaAs Photonic Devices - From Quantum State Generation To Quantum Communication

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

AlGaAs Photonic Devices - From Quantum State Generation To Quantum Communication

Transféré par

Droits d'auteur :

Formats disponibles

AlGaAs photonic devices: from quantum state

generation to quantum communications

To cite this version:

HAL Id: tel-01676987

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

pour obtenir le grade de

AlGaAs photonic devices: from quantum state

Soutenue publiquement le 14 novembre 2016,

M. Philippe Adam, Invité

3 Characterization of the laser emission 47

4 Characterization of the nonlinear waveguide 53

5 Characterization of the quantum state emitted by a passive source 69

6 Multi-user quantum key distribution 79

7 An electrically injected photon-pair source at room temperature 89

A List of Conferences and Publications 103

Two-photon sources state of the art

Outline of the manuscript

1.1 Nonlinear optical processes

{êl El (~r, ωl )ei(kl · ~r−ωl t) + c.c.}

which expresses the energy conservation during the conversion process.

Propagation in a nonlinear medium

1.1.2 Three Wave Mixing

Symmetry relations of the nonlinear susceptibility

jk 11 22 33 23/32 13/31 12/21

Overview of the three-wave mixing processes

• Spontaneous Parametric Down-Conversion (SPDC): a photon at frequency ω3 sponta-

(a) SFG (b) SHG (c) DFG(/OPA) (d) SPDC

Three wave-mixing coupled dynamics

Sum Frequency Generation efficiency

1.1.3 Three wave-mixing in waveguides

Figure 1.3 – Propagation of light in a waveguide.

Guided propagation in ridge waveguides

The waveguide modes are characterized by:

Frequency conversion in waveguides

Spontaneous Parametric Down-Conversion

[âk,σ , âk† 0 ,σ0 ] = δσσ0 δ(k − k 0 ) . (1.22)

The electric field observable can then be expressed as3

Ê(r, t) = Ê(+) (r, t) + Ê(−) (r, t) (1.23)

1.1.4 Phase-matching strategies in AlGaAs

bonding or by orientation-patterned regrowth. Although it has been used to demonstrate an optical

Another technique to achieve phase-matching involves inducing an artificial birefringence in the

Figure 1.6 – Sketch representing the principle of form birefringence phase-matching.

Figure 1.7 – Sketch of the counter-propagating phase-matching scheme.

Figure 1.8 – Sketch representing the modal phase-matching scheme.

1.2 Integrating the pump laser in a nonlinear waveguide

1.2.2 Working principle of our device

Figure 1.11 – Sketch of our electrically injected photon pair source.

(a) Refractive index engineering. (b) Simulation of the three-interacting modes

Figure 1.12 – Modal phase-matching in the monolithic photon pairs source.

Figure 1.13 – Simulated tuning curves for our device.

1.3 Characterization of the quantum state of photon pairs

The effect of a BS on an electric field can be described by the unitary transformation U

1.3.2 Quantum bits

|Ψi = α |0i + β |1i (1.51)

Figure 1.17 – The Bloch sphere.

This state is called separable if there exist vectors {cA B A B

It is called inseparable if, for at least one pair of coordinates cA B A B

|ψiAB 6= |ψiA ⊗ |ψiB . (1.55)

|0θ i = cos θ/2 |0i + sin θ/2 |1i