Tapered Glass-Fibre Nanospike Optomechanics

Tapered Glass-Fibre Nanospike
Optomechanics
Optomechanik mit verjüngten
Glasfaser-Nanospikes
Der Naturwissenschaftlichen Fakultät

der Friedrich-Alexander-Universität Erlangen-Nürnberg
zur
Erlangung des Doktorgrades Dr. rer. nat.
vorgelegt von
Riccardo Pennetta
aus Bari, Italien
Als Dissertation genehmigt
von der Naturwissenschaftlichen Fakultät
der Friedrich-Alexander-Universität Erlangen-Nürnberg
Tag der mündlichen Prüfung: 19. Februar 2019
Vorsitzender des Promotionsorgans: Prof. Dr. Georg Kreimer
Gutachter: Prof. Philip St.J. Russell, D. Phil

Prof. Dr. Gustavo Wiederhecker
ABSTRACT
This thesis investigates the optomechanical properties of glass-fibre nanospikes and

introduces various applications of these fibre-based mechanical resonators.
A nanospike is a tapered fibre tip, sub-wavelength in diameter, whose profile
has been engineered to convey adiabatic (therefore low-loss and single-mode) optical
guidance and at the same time mechanical resonances with quality factors greater
than 100,000 at room temperature in vacuum.
Even during early measurements, these excellent mechanical properties un-
veiled the extraordinary sensitivity of the nanospike to the surrounding environment.
The anomalous behaviour of its resonance frequency at low pressure was exploited
as a probe of molecular gas dynamics and revealed that under vacuum even weak
optical absorption suffices to transform the nanospike into a microscopic Knudsen
pump, through a mechanism related to the singular features of Crookes radiometers.
In the second part of the thesis a novel optomechanical configuration, in which
a nanospike is inserted in the hollow-core of a photonic crystal fibre, is presented.
Adiabatic evolution of the optical mode together with optomechanical trapping of
the nanospike into the core center permit lens-less, reflection-free, self-aligned and
self-stabilized coupling from a single-mode fibre to the hollow core.
Finally, coupling the nanospike to an ultra-high-Q whispering gallery mode
(WGM) resonator allowed for the first time the observation of dissipative optome-
chanical cooling of a free-standing waveguide coupled to a WGM, confirming a the-
oretical prediction published a decade ago. The remarkable consequence of the
suppression of the nanospike Brownian motion is a strong stabilization of its cou-
pling to the WGM, surpassing the limits set by thermodynamics, which promises
direct impact on the numerous applications of WGM resonators.
iii
ZUSAMMENFASSUNG
Diese Thesis untersucht die optomechanischen Eigenschaften und verschiedenen An-

wendungen von Glasfaser-Nanospikes.
Ein Nanospike ist das verjüngte Ende einer optischen Einmodenfasern und
hat einen Durchmesser, der kleiner ist als die Wellenlänge des Lichts das er leitet.
Das Profil des Nanospikes wurde derart konstruiert, dass gleichzeitig adiabatische
(d.h. verlustarme und einzel-modale) optische Leitung und mechanische Resonanzen
mit Güten von mehr als 100,000 bei Raumtemperatur im Vakuum erreicht werden
können.
Bereits bei ersten Messungen offenbarten diese exzellenten mechanischen Eigen-
schaften die außerordentliche Sensibilität des Nanospikes gegenüber seiner Umge-
bung. Das ungewöhnliche Verhalten seiner Resonanzfrequenz bei niedrigem Umge-
bungsdruck wurde zur Untersuchung molekularer Gasdynamik genutzt und zeigte,
dass unter Vakuum sogar schwache Lichtabsorption ausreicht um den Nanospike
durch einen dem Crookes Radiometer ähnlichen Mechanismus in eine mikroskopis-
che Knudsen-Pumpe zu verwandeln.
Im zweiten Teil der Thesis wird eine neuartige optomechanische Konfiguration
vorgestellt, bei der ein Nanospike in den hohlen Kern einer photonischen Kristall-
faser eingeführt wird. Die adiabatische Entwicklung der optischen Mode zusammen
mit dem optomechanischem Einfangen des Nanospikes im Zentrum des Faserkerns
ermöglicht eine linsenfreie, reflexionslose, sich selbst ausrichtende und stabilisierende
Kopplung der Einmodenfaser mit der photonischen Kristallfaser.
Abschließend wird beschrieben, wie die Kopplung des Nanospikes mit einem
Flüstergalerieresonator von extrem hoher Güte zum ersten Mal passive optomech-
anische Kühlung eines freistehenden Wellenleiters gekoppelt mit einem Flüsterga-
v
lerieresonator ermöglicht. Diese Beobachtung bestätigt eine vor einer Dekade veröf-
fentliche theoretische Vorhersage. Die bemerkenswerte Konsequenz aus der Unter-
drückung der Brownschen Molekularbewegung des Nanospikes ist die starke Sta-
bilisierung seiner Kopplung zum Resonator, die die Grenzen der Thermodynamik
übertrifft, was einen direkten Einfluss auf die zahlreichen Anwendungen von Flüster-
galerieresonatoren verspricht.
vi
CONTENTS
Introduction 1
1 Tapered Glass-Fibre Nanospikes 7

1.1 Optical properties of tapered fibres . . . . . . . . . . . . . . . . . . . 7
1.1.1 Adiabaticity criteria . . . . . . . . . . . . . . . . . . . . . . . 8
1.1.2 Calculation of the propagation constants . . . . . . . . . . . . 10
1.1.3 Calculation of the maximum tapering angle for adiabatic prop-
agation of light . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.2 Mechanical properties of glass-fibre nanospikes . . . . . . . . . . . . . 12
1.2.1 Calculation of the resonance frequencies and relative eigen-
modes of a silica nanospike . . . . . . . . . . . . . . . . . . . . 12
1.2.2 Measurement of the mechanical resonances of a nanospike . . 14
1.2.3 Understanding the high Q-factor . . . . . . . . . . . . . . . . 18
1.3 Fabrication techniques . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.3.1 The tapering rig . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.3.2 Fabrication of the shortest adiabatic nanospike . . . . . . . . . 22
1.3.3 HF etching of silica nanospikes . . . . . . . . . . . . . . . . . 25
1.4 Optical characterization of the taper profile . . . . . . . . . . . . . . 25
2 Glass-fibre nanospike as optically driven Knudsen pump 29

2.1 The experimental observation . . . . . . . . . . . . . . . . . . . . . . 30
2.2 Simulation of the temperature distribution along the microspike at
low pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.2.1 Thermal gradients and thermal creep flow . . . . . . . . . . . 32
vii
Contents
2.2.2 Competing between thermal creep and thermal edge flow . . . 35

2.3 Interaction between the vibrating nanospike and the Knudsen flow . . 35
2.4 Temperature dependence of the resonance frequency . . . . . . . . . . 37
2.5 Calculation of the total frequency shift . . . . . . . . . . . . . . . . . 38
2.6 Perturbation of the Knudsen flow . . . . . . . . . . . . . . . . . . . . 39
3 Optomechanical interaction of a nanospike inserted into an hollow core

photonic crystal fibres 41
3.1 Self-alignment of a nanospike by optomechanical back-action in HC-
PCF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.1.1 The main idea . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.1.2 Calculation of the optical forces . . . . . . . . . . . . . . . . . 45
3.1.3 The experimental configuration . . . . . . . . . . . . . . . . . 48
3.1.4 Optical spring inside HC-PCF . . . . . . . . . . . . . . . . . . 49
3.1.5 Self-aligned high-efficient coupling into HC-PCF . . . . . . . . 50
3.2 Nanospike based gas-cell . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.2.1 Fabrication techniques . . . . . . . . . . . . . . . . . . . . . . 54
3.2.2 Device performance . . . . . . . . . . . . . . . . . . . . . . . . 56
3.2.3 Theoretical consideration on the gas cell design . . . . . . . . 58
3.2.4 Observation of Raman scattering and molecular modulation
in H2 -filled HC-PCF . . . . . . . . . . . . . . . . . . . . . . . 59
4 Optomechanical cooling of a nanospike coupled to a whispering-gallery-

mode bottle resonator 63
4.1 Whispering gallery mode resonators . . . . . . . . . . . . . . . . . . . 65
4.1.1 Experimental techniques for light coupling to WGM resonators 66
4.1.2 Light coupling between a single mode waveguide and a WGM
resonator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.2 WGM bottle resonators . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.2.1 Fabrication and characterization of bottle resonators . . . . . 70
4.3 Observation of optomechanical cooling and self-stabilization of the
nanospike . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.3.1 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . 71
4.3.2 Coupling light to bottle resonator using a nanospike and mea-
surement of the mode profile . . . . . . . . . . . . . . . . . . . 72
4.3.3 Dispersive and dissipative coupling . . . . . . . . . . . . . . . 74
4.3.4 Optomechanical cooling of the nanospike . . . . . . . . . . . . 76
viii
Contents
4.3.5 Self-stabilization of the coupling with the WGM resonator . . 77

4.3.6 Experimental evidence of the features of the dissipative cooling 80
4.4 Theory of a dissipative and dispersive optomechanical system . . . . 82
4.4.1 Relation between optical damping and effective temperature . 84
5 Conclusions and outlook 87
Bibliography 93
List of publications 107
Acknowledgements 110
Acknowledgements 111
ix
INTRODUCTION
Light carries momentum and is able to exert optical forces through radiation pres-
sure. This curious property of the electromagnetic field fostered the development
of a new branch of physics named optomechanics, whose aim is to investigate the
mechanical interaction of light and matter.
ω
A single photon carries a linear momentum equal to p = , where is the re-
c
duced Planck constant, ω is the photon frequency and c is the speed of light. Even at
the first sight, the presence of the speed of light at the denominator suggests that the
optical forces are usually weak, which is the reason why "moving objects using light"
evades our daily experience and in our imagination still belongs to science-fiction.
Nevertheless, the first speculation about the existence of optical forces dates back
837 a.d., year in which Chinese astronomers reported that the tail of the Halley’s
comet always points away from the Sun [1]. The same phenomenon was observed in
Europe in 1531 from the Italian astronomer Fracastoro in his Homocentrica [2] and
independently from his German colleague Apian [3]. In 1619 Kepler in his attempt
to explain the observation suggested that light should be able to exert "some kind
of pressure" on the comet tail. A theoretical understanding of the problem required
more than 200 years, when in 1873 Maxwell deduced radiation pressure from his
theory of electromagnetism [4]. At the beginning of the twentieth century, the ad-
vent of quantum mechanics opened new questions about the momentum of light
at the single photon level, motivating the work of Compton, that in 1922 with his
Nobel prize awarded experiment observed the discrete momentum transfer between
a X-ray photon and an electron.
The year 1970, in which Ashkin observed the acceleration and trapping of
microscopic particles using radiation pressure [5], marks the birth of modern op-
1
Ch. 0 Introduction
tomechanics. His idea of focusing a laser beam to manipulate small objects cul-
minated in 1986 with the invention of the optical tweezers [6], which today, after
more than 30 years, assume a pivotal role in fundamental and applied physics as
well as in biology and medicine. In particular Ashkin recognized that a dielectric
particle in a non-uniform laser field experiences optical forces directed towards the
high intensity region. This negative light pressure, as Ashkin referred to it in Ref.
[6], or gradient force has been recently investigated in the context of guided optics
by Povinelli et al. [7]. Their theoretical analysis suggested that the optomechanical
interaction between two closely spaced waveguides, which arises from the overlap of
their optically guided modes, determines optical forces whose sign can be reversed
by changing the relative optical phase between the two waveguides. The theoreti-
cal predictions were experimentally confirmed a few year later by Li et al., which
succeeded in the fabrication of two free-standing silicon nano-waveguides only few
hundreds of nanometers apart [8] and about 10 μm long. Since in this configura-
tion the mechanical deflections of the waveguides strongly influences the effective
refractive index of the guided modes, the presence of the optical forces results in a
power-dependent optical phase shift, which can be considered as an optomechanical
variant of the Kerr effect [9]. A similar system is the dual-nanoweb fibre, which con-
sists of two glass planar-waveguides 440 nm thick and separated by a 550 nm air gap,
embedded in a fibre capillary [10]. The superior optical properties of silica compared
to silicon, permitted in this case to extend the optomechanical interaction-length up
to meters, allowing for instance the generation of optomechanical frequency combs,
whose spacing matches the mechanical resonant frequency of the webs [11].
All the systems mentioned till now involve structures with features on the
micro/nano-scale and instinctively one might wonder whether similar phenomena
could be observed on a larger scale, maybe closer to our everyday experience. Indeed,
the up-scaling of the optical forces is one of the biggest challenges of optomechanics.
After careful thinking, three different ideas come to mind, each of which is briefly
discussed below and motivated one of the experiments described in this Thesis.
Since the optical force is proportional to the power of the light source, employ-
ing strong or tightly focused laser beams appears to be the natural solution of the
problem. Unfortunately, increasing the optical power often determines a rise of the
temperature of the optomechanical system due to residual absorption of the laser
light, which can be detrimental even for highly transparent materials (e.g. fused
silica). Considering that most of the experiments in optomechanics are performed
at low pressure to avoid viscous damping and provide better isolation from envi-
2
ronmental disturbance, this issue has recently received great attention because the
absence of air molecules determines less efficient heat dissipation. For example melt-
ing and burning of levitated microparticles caused by the tight light focusing under
high vacuum has been recently reported [12, 13].
An alternative consists in enhancing the optical forces by carefully engineer-
ing the optomechanical interaction. For instance, the use of structured light was
demonstrated to increase more than one order of magnitude the trapping stiffness of
tweezered microbeads by exploiting the particle Mie resonances in order to achieve a
more efficient momentum transfer [14]. Similar concepts have been applied to metal-
lic nanoparticles and nanostructures, which support plasmonic resonances [15, 16].
In traditional optical tweezers, the trapped object only cause minimal changes in
spatial structure of the optical beam. The so-called self-induced back-action tech-
nique suggests the idea of creating a trap in which the particle plays an active role
by strongly affecting the local electromagnetic field. The consequence is a drastic
enhancement of the optical forces [17]. In the first experimental implementation a
polystyrene nanoparticle was trapped in a metallic nanoaperture in such a way that
movements of the nanoparticle were able to modify the resonance frequency on the
nanoaperture. Regarding waveguide-based optical trapping, as optical forces are in-
versely proportional to the group velocity [7], using slow-light waveguides has been
observed to provide one more direction to harness the stiffness of optical trapping
[18].
Finally, the use of optical cavities was shown to be a very effective way to
circumvent the intrinsic weakness of radiation pressure. Already in 1983 Dorsel et
al. demonstrated optical bi-stability due to the radiation pressure induced change
of the length of a Fabry-Perot cavity with a moving mirror [19], pioneering the
field of cavity optomechanics [20]. The improvement in the fabrication techniques
of micro/nano optomechanical structures attracted the attention of the scientific
community towards two kinds of optical cavities, namely whispering gallery modes
resonators and photonic-crystal cavities [21]. In such structures almost perfect over-
lap between the light and the mechanical modes allowed in recent years remarkable
achievements including self-sustaining mechanical oscillations [22, 23], synchroniza-
tion of optomechanical resonators [24] and optical cooling to the quantum ground
state of mesoscopic mechanical systems [25, 26].
Currently optomechanics seems to favour two distinct directions. The enor-
mous effort to engineer harmonic oscillators with excellent optical and mechanical
properties opened the chance of ultra-precise measurements of physical quantities,
3
Ch. 0 Introduction
while, on the other side, optomechanics represents one of the best candidates to
test fundamental laws of physics, as for example quantum mechanical behaviour of
macroscopic objects. The reader will experience this double soul of optomechanics
also in the experiments presented in the following Chapters.
The main aim of this Thesis is to introduce a novel optomechanical system: a
glass-fibre nanospike. Fashioned by tapering single-mode fibres, a nanospike exhibits
unique optical and mechanical properties and, at the same time, subwavelength in
diameter, it guides a considerable part of the light wave in the evanescent field, allow-
ing easy optical coupling to other waveguides or whispering gallery mode resonators.
Following the ideas described above, three different experimental configurations are
analysed and discussed. A free-standing nanospike is proved to be a sensitive probe
of optothermal effects in low vacuum, where air heat-dissipation becomes negligible.
Inserting the nanospike in a hollow waveguide constitutes both a new method for
enhancing the optical forces via dynamic back-action and a new efficient interface
between solid core and hollow core fibres. Finally, coupling the nanospike to an
optical cavity allowed the observation of optomechanical cooling of a 7 mm long
mechanical resonator with a low resonance frequency (only 2.5 kHz) by exploiting
a recently proposed novel mechanism for the optomechanical interaction.
Structure of this Thesis
Chapter 1 examines the optical and mechanical properties of glass-fibre nanospikes.

In particular, an experimental study demonstrates how to tailor the profile of nanospikes
in order to achieve adiabatic optical guidance and at the same time high mechanical
Q-factor. The fabrication and the inspection techniques of glass-fibre nanospikes are
also deeply discussed.
Chapter 2 analyzes the pressure-dependent response of a free-standing nanospike.
The experimental observation of an anomalous shift of the resonance frequency of
the nanospike for pressure below 1 mbar led to a systematic investigation of the
heating caused by residual absorption of the light under vacuum condition, in which
thermal dissipation through the surrounding gas is very inefficient. In particular, a
temperature gradient caused by the tapered geometry of the nanospike was found
to act as a microscopic Knudsen pump, driving a gas flow toward the nanospike
tip, where the temperature is the highest. The mechanical interaction between the
vibrating nanospike and the molecular flow determines a small shift in the resonance
frequency of the nanospike, observable because of its very high mechanical Q-factor.
This experiment offers novel insights for understanding the behaviour of weakly ab-
4
sorbing optomechanical systems (e.g. optically levitated particles or free-standing
waveguides) at low gas pressure. The results presented in this Chapter have been
published in the journal paper [27].
Chapter 3 introduces a new optomechanical system in which a nanospike is
inserted in the hollow core of a photonic-crystal fibre (PCF). The optical mode of
the nanospike spreads out into the surrounding space as the nanospike diameter falls
below the wavelength, adiabatically evolving into mode of the nanospike plus hollow
core. As a consequence of this optical interaction, the nanospike starts to feel the
presence of the hollow core and becomes optomechanically trapped at its centre.
This configuration offers a novel way to couple light from single-mode fibres into
hollow core PCFs, in which the presence of the optical forces allows self-alignment
and self-stabilization of the optical system. In the second part of this Chapter, these
concepts are used to fabricate a compact nanospike based gas-cell only a few cen-
timeters long with potential applications in non-linear optics and gas spectroscopy.
To demonstrate its capabilities, this device was used to observe stimulated Raman
scattering and molecular modulation in a hollow core PCF. The results presented
in this Chapter have been published in the journal papers [28, 29].
Chapter 4 explores the optomechanical interaction between a nanospike and
a bottle resonator, demonstrating for the first time the possibility to passively cool
a free standing waveguide coupled to a whispering gallery mode (WGM) resonator.
This configuration, first proposed in 2009 [30, 31], differs from the standard setups
used in cavity optomechanics because movements of the nanospike modulates not
only the resonance frequency of the WGM cavity (dispersive coupling) but also its
decay rate (dissipative coupling). In this context, dissipative cooling has been ham-
pered by the inadequate mechanical properties of the commonly used free standing
waveguides. Given its very high mechanical Q-factor, the use of a nanospike over-
comes the issue, allowing cooling of its fundamental mechanical mode down to a
minimum temperature of 1.8 K for a launched optical power of only 250 μW. The
results presented in this Chapter have been reported in the ArXiv paper [32].
Finally in the Conclusions and Outlook the results reported in the previous
Chapters are discussed and possible future research directions are presented.
5
CHAPTER 1
TAPERED GLASS-FIBRE NANOSPIKES
A glass-fibre nanospike is nothing more than a tapered optical fibre, or, to be more
precise, just half of it. A micrograph of a typical nanospike is shown in Fig. 1.1.
Despite a length which usually exceeds several millimetres, we use the prefix nano
because most of the applications described in this Thesis require a final tip di-
ameter as small as a few hundreds nanometers. As thoroughly discussed later in
this chapter, properly designed glass-fibre nanospikes inherit the excellent optical
properties of fibre tapers and support mechanical resonances with quality factors
exceeding 100,000 under low gas pressure. This unique combination appears as an
ideal platform for optomechanical experiments.
1.1 Optical properties of tapered fibres

A tapered fibre is usually produced by slowly pulling an optical fibre while it is heated
with a flame or a laser. It consists of two conical regions in which the diameter of the
fibre varies and a waist with constant diameter up to several centimetres long (see
Fig. 1.2). During the propagation in the tapered structure, when the core diameter
becomes too small to efficiently confine the optical mode, light escapes from the fibre
core and starts to be guided at the glass-air interface. If the fibre is further tapered

Fig. 1.1: Optical micrograph of a glass-fibre nanospike.
7
Ch. 1 TAPERED GLASS-FIBRE NANOSPIKES
and the cladding diameter falls below the optical wavelength, the mode spreads out
in the surrounding air in the form of an evanescent wave [33, 34].
Since tapering perturbs the longitudinal invariance of an optical fibre, coupling
of light from the fundamental into higher-order or radiation modes might occur. A
taper is defined to be adiabatic if the power loss from the fundamental mode is
negligible [33]. By looking at Fig. 1.2 it is easy to imagine that large tapering
angles Ω(z) are most likely to increase optical losses. Naively, one could argue that
engineering adiabatic guidance is just a matter of keeping the tapering angle small
enough or, in other words, to fabricate very long transitions regions. Unfortunately,
from a practical perspective, this is often undesirable because very long tapers are
fragile, difficult to handle and more sensitive to external perturbations. This leads
us to the complicated problem of how to fabricate the shortest taper transitions,
which still allows adiabatic propagation of light.
1.1.1 Adiabaticity criteria

The need for adiabaticity criteria has been the key motivation behind the seminal
work of J.D. Love et al. [33], which already in 1991 proposed two different methods
to design relatively short and low-loss fibre tapers. The first of the adiabaticity
criterion is called length-scale criterion and it is based on the intuitive argument
that the local taper length-scale has to be longer than the coupling length between
the fundamental mode and the first higher-order mode. With reference to Fig. 1.2
the local taper length-scale zt can be defined as the long cathetus of the triangle
with base coincident with the local core radius and apex angle corresponding to
the local taper angle Ω(z) = tan−1 (dρ/dz) ≈ dρ/dz, where z is the longitudinal
coordinate and ρ = ρ(z) is the local core radius. The approximation above holds
because in practice the tapering angles are usually very small (less than 0.1 rad).
A reasonable estimation of the local coupling length between the two modes is the
intermodal beat-length zb = 2π/(βF M (z) − βHOM (z)), where βF M and βHOM are
the local propagation constants of the fundamental mode and the first higher order
mode respectively. Therefore, by imposing zt = zb it is possible to determine an
approximate criterion to distinguish between adiabatic and lossy tapering angles:
ρ(z)(βF M (z) − βHOM (z))

Ω(z) = F (1.1)
2π
where F is an adimensional parameter between 0 and 1, known as "adiabaticity
factor", phenomenologically introduced to quantify how far the considered taper
8
Optical properties of tapered fibres 1.1
Fig. 1.2: Sketch of a fibre taper, ρ(z) is the taper radius and Ω(z) represents the local
tapering angle.
is from the adiabaticity criterion. Smaller values of F reduce the mode coupling.
Experimentally we found that F ≈ 0.5 is enough to produce taper with transmission
higher than 95%, in agreement with reported values [35]. In summary the length-
scale criterion succeeds in producing an upper bound for the local taper angle and
ensures adiabatic guidance along the taper in a remarkably elegant and effective
fashion, nevertheless it does not provide a quantitative estimate of the light coupled
to higher-order modes.
The second adiabaticity criterion is the weak power transfer criterion and
proposes to limit the tapering angle by imposing an upper boundary to the power
loss from the fundamental mode. The latter can be calculated by solving the set
of coupled equations that relate the amplitudes of the modes propagating along the
taper. Even though more accurate, this method requires not only the knowledge
of the propagation constants but also of the field distribution at each cross section
along the tapered fibre and can be computationally very demanding. This criterion
was not exploited in this thesis and more information can be found in Refs. [33, 36].
A physical interpretation of intermodal coupling in the transition regions can

be given by considering that, during the propagation along the tapered region, the
field distribution is unable to adapt to rapid variations of the fundamental local-
mode. Recently K. Harrington et al. proposed and experimentally demonstrated a
new fibre design named "endlessly adiabatic fibre", in which a logarithmic refractive
index profile determines a mode-field distribution independent from the fibre diam-
eter. Such a fibre can be tapered adiabatically no matter how short the transition
region is [37]. Adiabatic evolution of light has also been studied in mode involved
configurations such as null couplers [38] and photonic lanterns [39].
9
1.1.2 Calculation of the propagation constants
As discussed above, the key for the fabrication of adiabatic optical transitions is the
precise knowledge of the propagation constants of the modes in a tapered fibre as a
function of the taper radius. Since in a tapered fibre optical coupling can occur only
between modes with the same azimuthal symmetry, in this chapter the discussion
will be limited to the fundamental fibre mode HE11 and the first high-order mode
with the right symmetry HE12 . The calculation of the optical modes of a step index
fibre with infinite cladding is a textbook problem [36, 40] and analytical solutions
are readily available. This simplified model of an optical fibre works well at the
beginning of the taper, when light is guided by the core-cladding index difference,
and at its final part, in which only the cladding-air interface becomes relevant.
Generally, in the transition region, in which the core mode starts to spread into
the cladding, light is able to "feel" the presence of all the three layers from which a
real optical fibre is made of (e.g. core, cladding and air) and a more sophisticated
model is required. In particular, as proved later on, the transition region is the
most critical if tapers with minimal length need to be fabricated, because here
the propagation constants of the fundamental and the high-order mode are the
closest. Therefore, it is important to consider the complete analytical solutions of
a "three-layer" step index fibre. Even if conceptually the solution of the problem
does not differ from the two-layer case, the presence of an extra layer lengthens
the calculation and a complete analysis of the matter goes beyond the scope of this
Chapter. A recent review of this topic can be found in Ref. [41]. Briefly, for the
hybrid modes, as long as the taper diameter is large enough and nef f > ncladding ,
the weak guidance approximation holds and the simplified solution presented in
Ref. [42] can be adopted. When the nef f falls below ncladding , strong guidance
at the cladding-air interface requires the full vectorial solution published in Refs.
[43, 44].
Fig. 1.3 shows the calculated effective indices for the HE11 and HE12 modes
as a function of the taper radius across the transition region, in which the two-layer
model fails, for a single-mode fibre SMF-28 at a wavelength of 1550 nm. The dark
blue dashed lines represents the solution calculated using the two-layer model for
core-cladding and cladding-air guidance of the HE11 mode. The solid light-blue line
is the complete solution obtained from the three-layer model, which nicely joins the
two above mentioned solutions. The dashed orange line illustrates the effective index
of the HE12 mode for cladding-air guidance, while the solid orange line represents its
counterpart calculated from the three-layer model. The blue and orange circles are
10
Optical properties of tapered fibres 1.1
Fig. 1.3: Effective index nef f for the HE11 and HE12 modes of the tapered fibre as a
function of the taper diameter for a SMF-28 optical fibre at a wavelength of 1550 nm.
calculation performed using FEM and the agreement with the analytical solution of
the three-layer model is excellent.
1.1.3 Calculation of the maximum tapering angle for adiabatic

propagation of light
We are now in the position to estimate the maximum tapering angle as a function
of the taper diameter, using the above described length scale criterion (Eq. 1.1 ).
Once again we consider a SMF-28 single mode fibre at a wavelength of 1550 nm.
The results are shown in Fig. 1.4, in which the solid blue line was estimated from
the analytical solutions of a three-layer fibre, while the orange one was obtained
with FEM simulations. The two calculations agree very well. The yellow shaded
area highlights the region of non-adiabatic propagation of light, in other words,
tapers fabricated with tapering angles lying in this region will experience significant
coupling to high-order modes and high optical losses. It is interesting to notice that
the tapering angle shows a clear minimum at a diameter of about 60 μm, because
here the propagation constants of the fundamental mode and the first high-order
11
Fig. 1.4: Tapering angle allowed from the adiabatic criterion as a function of the taper
diameter for a SMF-28 optical fibre at a wavelength of 1550 nm. The yellow shaded area
highlights the region of non-adiabatic (and therefore lossy) propagation of light. The solid
blue line was calculated using the analytical model of a 3-layer fibre, while the orange line
is a result of FEM simulations.
modes are the closest. This happens exactly in the transition region, when the core
starts to become too small to efficiently confine the light field.
1.2 Mechanical properties of glass-fibre nanospikes

While large effort has been devoted the study of the optical properties of tapered
fibres, their mechanical ones are still not well explored. Here, we perform a theo-
retical and experimental analysis of the mechanical resonances of silica nanospikes
and we investigate the origin of their high mechanical Q-factor.
1.2.1 Calculation of the resonance frequencies and relative

eigenmodes of a silica nanospike
From the mechanical perspective, silica nanospikes can be modelled as single-side
clamped beams with non-uniform circular cross-section. The problem, which is
well-known in the field of mechanical engineering, requires the solution of the Euler-
Bernoulli beam equation [45]:

∂2 ∂ 2 w(z, t) ∂ 2 w(z, t)
Y I(z) + ρA(z) =0 (1.2)
∂z 2 ∂z 2 ∂t2
12
Mechanical properties of glass-fibre nanospikes 1.2
where z is the longitudinal coordinate along the nanospike, w(z, t) represents the
nanospike deflection, Y is the Young’s modulus (in case of silica Y = 73.1 GPa), ρ
is the density (for silica ρ = 2203 Kg/m3 ), A(z) is the area of the nanospike cross
section and I represents the second moment of inertia and it is defined as:

π
x2 dx dy = R(z)4 (1.3)
A 4
where R(z) is the local nanospike radius. This result is valid only for circular cross-
section. Interestingly Euler and Bernoulli formulated their beam theory around
the year 1750, more than a century before Eq. 1.2 found practical applications in
mechanical and structural engineering [46].
As usual, to identify the eigenmodes of the system we are interested in solutions

which oscillate harmonically in time: w(z, t) = w(z)eiωt . So that Eq. 1.2 can be
rewritten as:
∂2 ∂ 2 w(z)
Y I(z) − ω 2 ρA(z)w(z) = 0 (1.4)
∂z 2 ∂z 2
Since the nanospike is clamped on one side both the deflection w(z) and its derivative
vanish for z = 0. At the free end of the nanospike for z = L we require the bending
moment (∝ ∂ 2 w(z)/∂z 2 ) and the shear force (∝ ∂ 3 w(z)/∂z 3 ) equal to zero. The
boundary conditions can be summarized as:

∂w(z) ∂ 2 w(z) ∂ 3 w(z)
w(z) = 0, = 0, = 0, =0 (1.5)
∂z z=0 ∂z 2 z=L ∂z 3 z=L
z=0
Solutions of a fourth order differential equation are never straightforward and simple-
to-handle analytical expressions are only possible in the case of a uniform cross-
section [45], but unfortunately this does not correspond to a realistic experimental
situation. On the other end, fibre tapers with exponential profiles are often desired,
because their fabrication does not require modulation of the heat-length during the
tapering process. If the radius of the nanospike follows R(z) = R0 e−z/zd , where zd
is the characteristic exponential decay length, Eq. 1.4 can be rewritten as:

d4 d4 d4
− 8 + 16 − q e2zn w(zn ) = 0 (1.6)
dzn4 dzn4 dzn4
where zn = z/z d is a new dimensionless coordinate and q = 4ω 2 ρ/(Y R02 ) is a di-

mensionless constant. Eq. 1.6 can be numerically solved by imposing the following
13
boundary conditions:

∂w(zn ) ∂ 2 w(zn )
w(zn ) = 0, = 0, = 0, w(zn ) =1
∂zn zn =zn0 ∂zn2 zn =znL
zn =zn0 zn =znL
(1.7)
where zn0 is the fixed end of the nanospike and znL the free end, at which a unit
amplitude is required. Once the solution is found, it is possible to find the eigenfre-
quency by additionally requiring that ∂ 3 w(z)/∂z 3 |z=znL = 0, which is necessary for
no shear force at the nanospike endface. The calculated resonance frequencies and
the corresponding eigenmodes for a nanospike with initial radius R0 = 62 μm, tip
radius R(z = L)=0.5 μm and zd = 1 mm are shown in Fig. 1.5. The results agree
extremely well with solutions obtained by finite element modelling of the structure.
Even numerical solutions of Eq. 1.4 start to become challenging, if more com-
plicated profiles are considered (e.g. the shortest adiabatic taper discussed above).
In this case finite element modelling was usually exploited to estimate the mechan-
ical eigenmodes of the nanospike, even though, because of the large aspect ratio of
the nanospikes, the computation can be very time consuming.
1.2.2 Measurement of the mechanical resonances of a nanospike
To characterize the mechanical properties of the nanospikes, we used the experimen-

tal setup shown in Fig. 1.6(a). The nanospike was placed in a vacuum chamber and
laser light with a wavelength of 1064 nm was launched into the untapered end of the
SMF. Light coming from the tip of the nanospike was then collected and launched
into a second piece of SMF using a telescope. Since the coupling to the SMF is very
sensitive to the launching conditions, even tiny movements of the tip produced a
change in the transmission. The maximum sensitivity could be achieved by slightly
misaligning the SMF in one direction (e.g. in the horizontal plane). In this case,
displacements of the tip towards left or right respectively increased or decreased the
coupling into the SMF, providing a linear relation between the optical transmission
and the position of the tip for small tip displacements (below 1μm). The system
was calibrated by varying the relative position between the tip and the SMF and
measuring the change in transmission. A balanced photodiode was introduced to
suppress the laser noise and eliminate the DC component of the signal.
14
Fig. 1.5: First four eigenfrequencies and eigenmodes for a nanospike with exponential
profile. The solid blue line illustrate the eigenmode found by solving the Euler-Bernoulli
equation, while the orange dashed line results from FEM calculations.
15
Fig. 1.6: (a) Schematic of the experimental setup, including a vacuum chamber (VC), sin-
gle mode fibre (SMF) and a balanced photodiode (BPD). (b) Mechanical power spectrum
of the nanospike at atmospheric pressure. The dashed lines represents the FEM simulated
mechanical eigenfrequencies. Inset: FEM simulation of the mechanical eigenmodes of the
taper and relative frequencies.
Measurements at ambient pressure
The very first measurement of a nanospike mechanical power spectrum we performed

is shown in Fig. 1.6(b). This experiment was conducted at ambient pressure and,
due to significant air damping, it was necessary to excite the movement of the
tip using a piezoelectric actuator. Three clear peaks appear at frequencies of ≈ 9
kHz, ≈ 20 kHz and ≈ 35 kHz. Finite element modelling (FEM) of the nanospike
mechanical eigenmodes revealed that the observed resonances correspond to the
three first flexural modes of the nanospike and their shape is drawn as insets in Fig.
1.6(b). The calculated eigenfrequencies are reported as dashed lines in Fig. 1.6(b)
and agree very well with the experimental values.
The Q-factor of the resonance, defined as the ratio between energy stored in the
resonator and the energy dissipated per cycle, was estimated as Q = Ω/Γ, where Ω
is the angular frequency of the vibration and Γ/2π is the linewidth of the resonance.
At ambient pressure it was found to be rather low (only few tens) because of air
damping.
16
Fig. 1.7: (a) Mechanical power spectrum of the nanospike driven by Brownian motion in
the vicinity of the first mechanical eigenmode. (b) Linewidth Γ and Q-factor of the first
mechanical eigenmode of the nanospike measured as a function of the gas pressure.
Measurements in high vacuum
To unveil the real potential of silica nanospikes, the measurement of the mechanical
spectrum was repeated under vacuum conditions. In the following we will present
the spectra collected with a nanospike 4 mm long and with a final diameter of
3.9 μm. The same nanospike was used to obtain the experimental results reported
in Chapter 2. Comparable results were obtained using other samples with similar
characteristics. Fig. 1.7(a) plots the thermally driven mechanical spectrum recorded
at 10−5 mbar and 300 K for frequencies close to the fundamental flexural resonance
of the nanospike. The resonant frequency is Ω/2π ≈ 4340 Hz, in good agreement
with the value of 4338 Hz predicted by FEM.
To our great surprise, in this experiment we measured a mechanical linewidth

of Γ/2π = 20 mHz, which corresponds to a lifetime τ = 50 s and a Q-factor of
Ω/Γ = 217,000. This ultranarrow linewidth made it possible to distinguish two
resonance peaks only 0.6 Hz apart. We attribute them to a non perfectly circular
nanospike cross section, resulting in a splitting of the frequency of the fundamental
mechanical mode in orthogonal directions. Fig. 1.7(b) depicts the linewidth as a
function of gas pressure. Above 10−3 mbar it is dominated by air viscous damping
and at lower pressures it saturates to 20 mHz, a value determined by material and
clamping losses.
17
1.2.3 Understanding the high Q-factor
The total phonon decay rate Γ of the nanospike can be expressed as the sum
Γmaterial + Γclamping + Γair , where Γmaterial is caused by material losses in the glass
(e.g. material and surface imperfections), Γclamping accounts for leakages of phonons
into the supporting fibre and Γair is determined by air viscous damping and can
be suppressed by decreasing the pressure below 10−3 mbar. While material losses
represent a fundamental limitation, it should be possible to reduce the clamping
losses by tailoring the shape of the nanospike [47]. To quantify the clamping losses
we performed FEM of the mechanical modes of a clamped nanospike with different
taper profiles. The real experimental situation was considered, in which the sample
is fixed into the vacuum chamber using a vacuum-compatible resin epoxy. The ge-
ometry implemented in the simulation is sketched in Fig. 1.8(a). The acoustic waves
propagating into the substrate are absorbed by introducing a perfectly matched layer
(PML), the outer surface of which is set to be a fixed boundary. The displacement
in the vertical direction is depicted in Fig. 1.8(b) for the first mechanical mode. As
expected the acoustic energy is mostly confined at the very end of the tip, but if the
color scale is modified in order to emphasize small displacements, the origin of the
clamping losses starts to become clear. Due to the finite stiffness of the boundary
at the clamping of the taper, part of the acoustic energy flows into the substrate
and is eventually lost (absorbed by the PML in the simulation). The Q-factor in
this situation could be estimated as Q = Re(f )/2Im(f ), where Re(f ) and Im(f )
are the real and imaginary part of the calculated eigenfrequency.
Different nanospike profiles were modelled using the following expression for
the taper radius:
ρ0 (e−ζL − e−ζz ) + ρT (e−ζL − e−ζz )

ρ(z) = (1.8)
e−ζL − 1
in which at z = 0 (beginning of the taper) ρ = ρ0 (i.e. the untapered fibre radius),

at z = L (at the tip) ρ = ρT and ζ is a constant related with taper shape. The inset
in Fig. 1.9 shows ρ(z) for different values of ζ (ζ = 0 corresponds to a linear profile).
Fig. 1.9 plots the simulated clamping losses as a function of ζ. In the same plot,
material losses were introduced to reproduce the experimentally observed saturation
in the total losses and their value is consistent with previous reports [48]. As ζ is
increased, or in other words for sharper taper profile, the clamping losses decrease
strongly because of better confinement of phonons nanospike and eventually for
ζ greater than ≈ 0.6mm−1 material losses dominate. The light-blue shaded area
18
Fig. 1.8: (a) Sketch of the geometry implemented in the FEM calculation. (b) Normalized
vertical displacement for the first eigenmode of the structure. (c) Same as (b), but with
the color scaled modified in order to to emphasize small displacements. (d) Zoom-in of
area included in the dashed rectangle in (c). The white arrows indicate the direction of
motion.
19
Fig. 1.9: Finite-element modelling of clamping loss of the nanospike as a function of taper
profile. The dots are measured data points, and the arrow marks the parameters of the
nanospike used in the experiments described in Chapter 2.
indicates the region in which the nanospike is optically nonadiabatic (ζ > 1.1mm−1 .
This demonstrates that there is a well-defined window in which minimum mechanical
damping can be achieved without affecting the optical properties of the nanospike.
In particular, the nanospike used in the experiments reported in Chapter 2 lies
within this region (marked with an arrow in Fig. 1.9). Regarding the material
losses, the classical work of Startin et al. [49] identified fused silica as the material
with the highest intrinsic Q among glasses at room temperature, while later other
groups proposed post-processing techniques such as vacuum annealing treatment to
further decrease the dissipation [50], offering a possible way to improve the Q-factor
by developing new fabrication methods.
1.3 Fabrication techniques

As mentioned before, to fashion nanospikes with excellent optical properties a precise
control over the transition regions is required. The theoretical framework for the
fabrication of fibre tapers has been established by T. Birks et al. in 1992 [51]. Their
analysis starts by considering an optical fibre gently pulled from both ends, while a
section of its length L is heated up to uniform temperature. The particular value of
the temperature of the glass is irrelevant as long as it is high enough to soften the
glass without melting it. Considering that the mass of the glass has to be conserved
20
Fabrication techniques 1.3
Fig. 1.10: (a) At t=0 the section P Q of length L0 is heated. (b) During tapering P and
Q are separated by a distance x. The section P Q is now also equal to 2z0 + L.
during the stretching process, with the help of Fig. 1.10, it is possible to derive the
so-called "volume law":
drw rw
=− (1.9)
dx 2L
where rw is the radius of the waist after an extension x and a hot-zone length L.
In the same way, comparing the the total length PQ of the tapered fibre with the
initial value of PQ, we can obtain the "distance law":
2z0 + L = x + L0 (1.10)
where z0 is the instantaneous length of the transition region. Given an arbitrary

taper profile, which can be for example calculated using the adiabatic criterion (see
Eq. 1.1), Eqs. 1.9 and 1.10 can be integrated to calculate how the length of the
hot-zone L has to be varied as a function of the elongated distance x to obtain the
desired taper. In our setup, we always set the fibre elongation rate to be uniform over
time (i.e. x(t) = velong t, where velong is the elongation speed), therefore eventually
the hot-zone length can be calculated as a function of time L = L(t).
From the practical point of view, engineering a heat source with controllable
length is not as hard as one might expect. Already in Ref. [51] the author proposed
to use a point-like heat source consisting of a small flame, quickly scanned under-
neath the fibre during the elongation process. We underline that for this approach
to be successful, the scanning speed has to be much higher than the elongation rate.
Alternatively a focused CO2 laser beam can be used to heat up the glass above the
softening point [52]. The main advantage of this technique lies in the waist diameter
of a focus CO2 laser beam (about 100 μm), which is significantly smaller than the
typical flame diameter (from 0.5 mm to 1 mm). On the other hand, since the heating
of the glass is based on the absorption of light, successful tapering requires careful
adjustments of the laser power during the fabrication due to the reduced light-glass
overlap when the taper diameter decreases[53].
21
Fig. 1.11: (a) 3D rendering of the experimental setup used to fabricate the fibre tapers.
(b, c) Sketch of the procedure used to fabricate the glass-tapers.
1.3.1 The tapering rig

A sketch of the rig used for fibre tapering is shown in Fig. 1.11. A stripped and
cleaned fibre is placed onto two motorized stages using standard fibre clamps. Two
others motorized stages are used to control the position of a oxybutane flame and
to quickly scan it underneath the fibre during the fabrication process. In most of
our experiments the elongation speed was on the order of 0.3 mm/s, while the flame
speed was about 10 times higher.
It is possible to control the size of the flame by changing the gas mixture or
by properly choosing the diameter of the stainless steel nozzle used to deliver the
gas mixture. Big flames, with diameter of several millimetre are the best choice to
fabricate long tapers with uniform waist diameter, because of the higher uniformity
of the temperature during the fabrication. On the other hand, smaller flames, sub-
millimeter in size, obtained using a nozzle with diameter of 150 μm and higher
oxygen concentration, produced fibre tapers with profile closer to the theoretical
predictions. Unfortunately, decreasing the diameter of the nozzle also increased the
speed of the gas molecules. As a consequence when the taper diameter becomes
submicron, the gas flow can mechanically push and bend the glass strand, making
the final taper diameter hardly predictable. A good compromise was obtained using
a nozzle diameter of 400 μm and a corresponding flame diameter of ≈ 1 mm.
1.3.2 Fabrication of the shortest adiabatic nanospike

As a an example of the techniques described above, Fig. 1.12 shows the details of
the fabrication of the shortest nanospike allowed by the adiabatic criterion using a
standard single-mode fibre SMF-28 at a wavelength of 1550 nm. The final taper
diameter was targeted to be 1 μm. The taper profile was calculated using Eq. 1.1
22
Fabrication techniques 1.3
where the adiabaticity parameter F was set to 0.4 and it is represented with an
orange solid line in Fig. 1.12(b). Higher values of F usually determines a lower
optical transmission and decrease the fabrication success rate. For instance with
F=0.7 we found very hard to obtain optical transmission higher than 85% − 90%.
Solving Eqs. 1.9 and 1.10 the hot-zone length L was calculated as a function of
time and the result as a orange dashed line in Fig. 1.12(a). Here, the solid blue
line illustrates the actual movements of the oxybutane flame. As mentioned before,
for very small tapering radius the adiabatic tapering angle becomes very large. To
account for this, the hot-zone length starts to decrease at T = 17 s and after T = 35
s it drops below 1 mm. To account for the finite size of the flame the hot-zone length
is fixed to 1 mm for the remainder of the process, determining an exponential profile
for the final part of the taper (yellow shaded area in Fig. 1.12(a)). The fabricated
taper radius is shown in Fig. 1.12(c) (blue dots) and agrees excellently with the
theoretical prediction. The inset in Fig. 1.12(c) emphasizes the transition from
shortest adiabatic to exponential profile.
Very valuable information can be obtained by measuring the transmission dur-

ing the tapering process. As depicted in Fig. 1.12(b)), after a slight decrease of the
transmission due to residual coupling to high order modes, oscillations start to ap-
pear at a frequency which increases over time. Thorough studies (see for example
[54, 55]) proved that these oscillation are due to beating among the different optical
modes in the taper and their frequency, which depends on the local value of the
propagation constants, contains information about the instantaneous taper radius.
Interestingly these oscillations disappear after T = 45 s, meaning the taper is now
single mode and its diameter is smaller than the one predicted by the cut-off con-
dition. In this measurement the final transmission was about 98%. Generally using
these parameters transmission well above than 95% could be routinely obtained. To
further check the adiabaticity of the nanospike, Fig. 1.13 shows the measured taper-
ing angle as a function of the taper radius. The blue dots are obtained directly from
the data in Fig. 1.12(b), while the solid blue line was calculated by differentiating a
polynomial fit of the same data. The orange shaded area indicates the lossy region,
calculated setting the adiabaticity factor F = 1, while the light-blue line represents
the design (F = 0.4). The data points are in the adiabatic region for the entire
taper length and they nicely reproduce the minimum in the design for tapering radii
around 30 μm. Below 10 μm radius, the experimental tapering angle deviates from
the design and starts to decrease because the exponential profile of the final part of
the nanospike.
23
Fig. 1.12: (a) Flame position as a function of time. (b) Taper profile for adiabatic
tapers at 1550 nm. Optical transmission measured during the tapering process. The fast
oscillation are due to beating between different fibre modes. At T = 48 s the fibre taper
becomes single mode. Optical transmission well above 95% could be routinely obtained
with our fabrication procedure.
Fig. 1.13: Experimentally measured tapering angle as a function of the taper radius. The
blue dots are data points obtained by differentiating the data in Fig. 1.12c, while the solid
blue line was obtained by differentiating a polynomial fit of the same data. The orange
line represent the calculated adiabatic criterion with F=1 for the considered fibre, while
the solid light-blue line is the designed angle (abiabatic criterion with F=0.4).
24
Optical characterization of the taper profile 1.4
1.3.3 HF etching of silica nanospikes

Some applications of silica nanospikes described in this thesis require final tip diam-
eter smaller than 200 nm, for which fabrication based on thermal tapering is often
not sufficient to ensure good mechanical properties. Indeed, as mentioned before,
because of the finite size of the flame used during the fabrication process, the ta-
pering angle experimentally obtainable for diameter below a few μm is very small.
Therefore, direct tapering to diameter between 100 nm and 400 nm results in very
long nanospikes, which are difficult to handle and with very low resonant frequencies,
easily disturbed by laboratory noise. To overcome this problem, we used chemical
etching with hydrofluoric acid (HF) to further tailor the nanospike shape and final
diameter close to its tip. Alternatively, it was recently reported that rapid pulling
of the optical fibre during tapering allows fabrication of short subwavelength tapers
[56].
Fig. 1.14(a) shows a sketch of the experimental procedure used for wet-etching
of the nanospikes. A silica nanospike with a final diameter of a few μ m is fabricated
by thermal tapering and then its final part (about 1 mm) is immersed into the HF
acid for a several minutes. The typical spike shape is readily obtained thanks to
the surface tension of the liquid. Fig. 1.14(c) shows the measured tip diameter
as a function of the etching time for a HF concentration of 20%. This approach
allowed the fabrication of very short adiabatic tapers with final diameters as small
as 150 nm and high mechanical stiffness. Alternatively the nanospike can be placed
very close to the surface of the liquid, without being inserted in it, and etched from
evaporating HF molecules (Fig. 1.14(b)). This technique usually requires higher HF
concentrations and is experimentally more challenging, but, since the nanospike is
never in physical contact with the liquid, the samples are cleaner and have a better
surface quality. We underline that the tapering angle allowed by the adiabatic
criterion for fibre diameter around 1 μm is so large that the optical properties of
the nanospike are very easily preserved after the chemical etching.
1.4 Optical characterization of the taper profile

After developing the methodology needed to control the taper profile during the
fabrication process, it raised naturally the need to measure precisely the taper profile
and compare it with the prediction. Because of the huge aspect ratio of fibre tapers,
it is impractical to manually check them under a microscope. A more efficient
solution consists in mounting the fibre taper on a motorized stage, scanning the
25
Fig. 1.14: HF etching of silica tip. (a) Sketch of liquid etching procedure. Surface
tension is responsible for the typical spike shape at the tip end. (b) Sketch of vapour
etching procedure. (c) Measured tip diameter as a function of the etching time in the case
of liquid etching and with a HF concentration of 20%.
Fig. 1.15: Optical measurement of the taper profile. (a) Raw image of the taper collected
with a microscope. (b) Post-processed image. (c) Reconstructed taper diameter.
26
Optical characterization of the taper profile 1.4
Fig. 1.16: Optical characterization of a fibre taper. The orange line represent the intended
designed, while the blue dots are experimental data.
taper under the microscope and periodically collecting pictures of its profile. After
calibrating the microscope magnification, all the pictures are then post-processed
using an image recognition Matlab-based code, in order to detect the boundaries of
the fibre taper and hence its local diameter (see Fig. 1.15). Eventually the complete
taper profile can be completely reconstructed by merging together the data obtained
from the different pictures. This approach allowed the characterization of tapers,
whose length is only limited by the traveling range of the motorized stage, with
diameter resolution of about 1 μm.
To test the limits of our tapering rig and to assess the quality of the measure-
ment scheme, we designed the taper profile depicted as an orange line is Fig. 1.16, in
which a series of very sharp drops in the taper radius were intended to challenge the
fabrication technique. The measured diameter (blue dots) showed excellent agree-
ment with the theoretical prediction and, as highlighted in the inset of Fig. 1.16,
sudden variations of just a few microns in the tapering radius could be accurately
reproduced.
27
Conclusion of this chapter

In this chapter the optical and mechanical properties of glass-fibre nanospikes have
been investigated. By properly engineering the taper profile, adiabatic propagation
of light and high mechanical Q-factor can be achieved simultaneously. To fashion
our samples we used the heat brush technique, occasionally combined with HF-
etching when final diameters below ≈ 400 nm were required. The result is the
routine fabrication of glass-fibre nanospikes with single-mode optical transmission
exceeding 95% and flexural resonances with mechanical Q-factor above 100,000.
28
CHAPTER 2
GLASS-FIBRE NANOSPIKE AS OPTICALLY
DRIVEN KNUDSEN PUMP
Mechanical resonators with ultra-narrow linewidth and optical readout are ideal de-
vices for ultra-sensitive measurements [20, 57]. In the last 10 years these systems
have found applications for measuring physical quantities such as displacement [58],
force [59], electric field [60], magnetic torque [61] and temperature [12, 62]. To sup-
press air damping and provide better isolation from environmental disturbance, most
of these experiments were performed at low gas pressure. Under these conditions,
the absence of conductive or convective heat dissipation through the surrounding
gas, together with the relative inefficiency of radiative heat loss, determine large rises
in temperature, which can be as high as the melting point of fused silica [12, 13],
even for very small optical absorption. For mild thermal heating, one might at first
think that the low temperature increase will have little if any effect.
The experiment described in this Chapter proves that even a moderate amount
of heating caused by residual absorption of the laser light can noticeably affect the
mechanical behaviour of a glass-fibre nanospike at low gas pressure. In particular, a
temperature gradient along the nanospike causes it to act as a microscopic Knudsen
pump, driving a flow of gas molecules towards the tip where the temperature is the
highest. The consequence is that an additional optothermal Knudsen stiffness can
act back on the nanospike and influence its mechanical motion. A temperature gra-
dient of few K/mm was able to shift the nanospike resonant frequency by ≈ 0.03%,
clearly resolvable because of its ultra-narrow mechanical linewidth, as a consequence
of momentum exchange between the vibrating nanospike and the flowing molecules.
This effect was observed to be strongest in the Knudsen regime, when the mean free
29
Ch. 2 GLASS-FIBRE NANOSPIKE AS OPTICALLY DRIVEN KNUDSEN PUMP
path of the air molecules is comparable to the dimensions of the vacuum chamber.
We think that this system might offer a novel means of monitoring the behaviour of
weakly absorbing optomechanical systems at low gas pressure.
The results presented in this Chapter have been published in Ref. [27].
2.1 The experimental observation
The experimental results reported in this Chapter were collected using a nanospike
4.3 mm long with a tip diameter of 3.9 μm. It had a resonance frequency Ω/2π
= 4340 Hz (see Chapter 1, Fig. 1.7) and a Q-factor of 2.17 105 . The optical and
mechanical properties of this particular nanospike have already been discussed in
Chapter 1 and the experimental setup used is sketched in Fig. 1.6.
Gases at low pressure are usually characterized by the Knudsen number (Kn),
defined as the ratio between the mean free path and a characteristic length-scale of
considered system, as in our case the width of the vacuum chamber (L = 2 cm in
our setup). This experiment investigates the range 0.01 < Kn < 1000: large values
of Kn characterize the free-molecular regime, in which gas dynamics are dominated
by collisions with the vacuum chamber rather than intermolecular collisions as in
the continuum regime (high pressure, low Knudsen number).
Fig. 2.1 plots the measured resonance frequency fR of the nanospike as a

function of gas pressure for different values of laser power. As a reference, the
upper axis reports the corresponding value of Kn. The results show that fR first
increases and then decreases with pressure, reaching a maximum at around 10−3
mbar when Kn ≈ 1. This non-monotonic behaviour becomes more pronounced at
higher laser power. At fixed gas pressure, fR was found to be a linear function
of laser power (see Fig. 2.2). The observation is in sharp contrast with what
predicted by gaseous damping [63, 64] and thermal effects [65], since in both cases fR
increases monotonically as the pressure is lowered. In particular, gaseous damping
is independent of laser power and is, anyway, negligible in the regime studied here,
while thermal effects are driven by the increase in the Young’s modulus of silica with
temperature [65], which increases the nanospike stiffness and raises fR .
30
Simulation of the temperature distribution along the microspike at low pressure 2.2
Fig. 2.1: (a) Measured and (b) simulated mechanical resonant frequency of the nanospike
as a function of gas pressure for five different laser powers.
2.2 Simulation of the temperature distribution along

the microspike at low pressure
To disclose the physical mechanisms behind this anomalous pressure dependence of
fR , finite element modelling (FEM) was used to simulate the thermodynamics of the
system at low pressure, including the interaction between the nanospike and the gas
molecules, which required to solve the Navier-Stokes equations with slip boundary
conditions [66].
In the nanospike there are three mechanisms of heat dissipation: convection,
radiation and thermal conduction both into the fibre base and through the air. At
high pressure, convection and conduction in air dominate over radiative heat loss
and conduction along the fibre. As the pressure is reduced, however, the thermal
conductivity of air falls [67]. In particular, as specified in [68], for high values of Kn
the heat conductivity is no longer a property of the gas alone, but depends on Kn
itself. In the simulations the heat conductivity κ of air at low pressure was estimated
as follows:
κ0
κ(Kn) = (2.1)
1 + Kn a/(ρRS λ)
where κ0 = 0.026 W/(m K) is the thermal conductivity of air at ambient pressure
(Kn → 0), RS = 287.1 J/(Kg K−1 ) is the specific gas constant and a is a constant
equal to 7.6 10−5 N/(m K). The same is true for the air viscosity, which can be
expressed as [69]:
η0
η(Kn) = (2.2)
1 + (β1 + β2 e−β3 /Kn )Kn
where η0 is the viscosity of air at ambient pressure and β1 = 1.231, β2 = 0.469
31
Fig. 2.2: (a) Resonant frequency of the nanospike as a function of laser input power at
a pressure of 10−5 mbar. (b) Power spectra corresponding to the data points in (a). (c)
Frequency shift measured as a function of the nanospike temperature, giving a slope of
0.071 Hz/K.
and β3 = 1.178. Interestingly, at very low pressure, even though the nanospike is
connected to the supporting fibre, calculations indicate that radiative heat loss is the
main cooling mechanism (thermal conduction is proportional to taper area, which
is only a few μm2 in the proximity of the tip). The only unknown parameter in the
simulations is the absorption coefficient of the silica glass after thermal tapering. At
the lowest pressure used (10−5 mbar), since the molecular density is extremely small,
we expect the shift in fR to be caused purely by heating. The measured dependence
between frequency shift and laser power (Fig. 2.2(a)) was therefore used to estimate
the absorption coefficient, resulting in α = 0.34 m−1 . We attribute the significant
increase in absorption coefficient compared to pure silica to the diffusion of water
and other impurities during the tapering process.
The results of the simulations are shown is Fig. 2.3. As expected the av-
erage temperature of the nanospike and thus the thermal frequency shift increase
monotonically as the pressure is reduced.
2.2.1 Thermal gradients and thermal creep flow

The temperature distribution along the nanospike at several gas pressure is de-
picted in Fig. 2.3(a). Interestingly, the above discussed average temperature rise is
accompanied by a highly non-uniform temperature distribution along the nanospike
(z axis), with a maximum at the tip. This temperature gradient drives the gas
32
Simulation of the temperature distribution along the microspike at low pressure 2.2
Fig. 2.3: (a) Simulated temperature shift along the nanospike at different gas pressures.
(b) Simulated maximum temperature increase as a function of the gas pressure. (c) Ve-
locity of thermal creep flow at the surface of the nanospike at z = 4 mm. The orange line
plots Eq. (1), while the blue dashed line is the result of FEM simulations.
molecules along the surface of the nanospike towards higher temperatures, creating
a collective gas flow in the vacuum chamber within a layer of thickness comparable
to the mean free path. Commonly referred as to thermal creep flow [70], this effect
has been recognized as the cause of the singular behaviour of Crookes radiometers
and is the working mechanism of Knudsen pumps [71, 72]. Thermal creep has also
found applications in the field of optical tweezers as an optothermal particle trap in
gas-filled hollow-core photonic crystal fibres [73].
Fig. 2.4 illustrates the calculated flow pattern at 10−3 mbar (Kn ≈ 1). Here,
the black solid curves represent the average trajectories of the gas molecules (the
streamlines) and the red arrows mark the local velocity directions. The mean molec-
ular velocity can reach some m/s in the vicinity of the nanospike (see inset of Fig.
2.4). This average velocity along the gas-solid surface can also be analytically cal-
culated using the following formula originally introduced by Maxwell [74, 75]:
η ∂T
vcreep = (2.3)
ρ T ∂z
where η is the gas viscosity, ρ is the density, and T is the temperature. Fig. 2.3(c)
plots vcreep versus pressure at z = 4 mm on the nanopsike surface, showing excellent
agreement between Eq. 2.3 and the full FEM analysis.
33
Fig. 2.4: Numerically simulated Knudsen flow inside the chamber at 10−3 mbar. The
black solid curves represent the trajectories of the gas molecules (the flow) and the red
arrows show the average direction of the molecular velocity. Inset: zoom-in of the region
close to the tip.
34
Interaction between the vibrating nanospike and the Knudsen flow 2.3
2.2.2 Competing between thermal creep and thermal edge flow

A gas flow can also be excited also when the temperature distribution is uniform
if the solid-gas interface has points where the curvature changes sharply, producing
the so-called thermal edge flow [76, 77]. Simulations showed that both thermal creep
flow and thermal edge flow are simultaneously present in the nanospike. As shown
in Fig. 2.3, this happens because the temperature gradient has opposite signs before
and after the tip. When T increases from the base to the tip, molecules flow towards
the tip (see Fig. 2.5). Beyond the end of the tip there is a negative temperature
gradient, which creates thermal edge flow and drives molecules back towards the tip.
Although these two effects compete, flow along the nanospike will dominate at low
pressure when the mean free path becomes longer. Fig. 2.5 shows the simulated gas
flow at three different pressures and as expected, it is possible to see that already at
10−3 mbar thermal creep overcomes the thermal edge flow, reversing the direction
of the gas circulation inside the vacuum chamber.
2.3 Interaction between the vibrating nanospike and

the Knudsen flow
The interaction with the flowing molecules determines a change in the nanospike
stiffness. Moving away from its equilibrium position, the nanospike locally perturbs
in the gas flow, deflecting molecules from their trajectories. The result is a net
transfer of momentum to the nanospike. Since the molecules flow towards the tip of
the nanospike, this effect creates a restoring force that pushes the nanospike back
towards its equilibrium position. For small tip deflections, calculations predict that
this restoring force is proportional to the displacement and thus effectively increases
fR . In the simulations the gas was once again modelled using the Navier-Stokes
equations.
The features of the described process can be more easily highlighted in the
simplified geometry in Fig. 2.6, in which a slant elongated rod is placed in a vertical
gas flow. Here, the blue arrows represent the local force density, while the red arrows
indicates the direction of the mean gas velocity. The interaction between the gas and
the rod bends the streamlines and cause the appearance of a force in the direction
opposite to the rod deflection.
The purple dash-dotted curve in Fig. 2.7 illustrates the resulting Knudsen stiff-
ness, calculated by FEM simulations of the nanospike-flow interaction, as a function
35
Fig. 2.5: Calculated gas flow at a pressure of (a) 10−1 mbar, (b) 10−2 mbar, (c) 10−3
mbar and (d) 10−4 mbar.
36
Temperature dependence of the resonance frequency 2.4
Fig. 2.6: Sketch of the mechanical forces due to a gas flow acting of an slant elongated
rod. The blue arrows represent the local force density, while the red arrows indicates the
direction of the mean gas velocity.
of the gas pressure. The typical trend observed in the experiment is reproduced.
Intuitively, one can argue at high pressure, efficient thermal dissipation through the
surrounding air suppresses the temperature gradient and slows down the gas flow,
while at low pressure the temperature gradient saturates while the molecular den-
sity (and thus the momentum transfer) falls. The consequence is a maximum in the
Knudsen stiffness at Kn ≈ 1, as observed in the experiments.
2.4 Temperature dependence of the resonance

frequency
As mentioned earlier, the Young’s modulus E of silica glass increases with temper-
ature [65] and as the resonance frequency is proportional to the square root of E, it
is possible to write:
dfR dE
= (2.4)
fR 2E
For silica glass E is to good accuracy linearly proportional to temperature T from
room temperature up to ≈ 1200 K, or in other words dE = b dT , where b is a constant
We measured the coefficient b for the nanospike by monitoring fR while uniformly
increasing the temperature of the vacuum chamber at very low laser power (<1 mW)
to avoid laser heating. The result is shown in Fig. 2.2(b). The measured values is
b = 7.4 10−3 GPa/K, which is slightly less than the reported value in the literature
for bulk silica glass, which equals 11.0 10−3 GPa/K [65]. The discrepancy may be
37
Fig. 2.7: Calculated contributions from pure thermal (dark-blue dashed line) and Knudsen
(purple dotted line) effects to the resonant frequency of the nanospike as a function of gas
pressure. The solid blue curve is the sum of two effects.
due to flame post-processing of the glass, which results in high water concentration.
2.5 Calculation of the total frequency shift

As in the case of a simple harmonic oscillator, the resonance frequency is propor-
tional to the square root of the stiffness. A small addition δk to the stiffness k0
implies an increase of the frequency, approximately equal to fR δk/(2k0 ). We can
therefore write:
fR = f0 + Δfth + ΔfKn (2.5)
where f0 is the resonant frequency at room temperature in vacuum, Δfth results

from the temperature dependence of Young’s modulus, and ΔfKn is the Knudsen
contribution. The simulated overall frequency shift for a launched power of 66 mW
is plotted in Fig. 4 (solid blue curve) as the sum of Δfth and ΔfKn . The frequency
shift is small compared to f0 and is proportional to the optical power, as measured in
the experiments (see Fig. 2.2). As shown in Fig. 2.1, the simulation reproduces well
the observed behaviour at different power levels. Quantitatively, the predicted and
measured frequency shifts agree well, while the gas pressure at which the Knudsen
effect reaches a maximum is somewhat higher in the experiment (see Fig. 2.2).
We attribute this discrepancy to the fact that the Navier-Stokes equations are not
precisely correct in the free molecular regime (Kn > 10).
38
Perturbation of the Knudsen flow 2.6
Fig. 2.8: (a) Frequency shift as a function of insertion length Lins in a glass capillary
with internal diameter 100 μm. Inset: sketch of the experimental setup. The circles
are experimental data points and the full orange line is the numerical simulation. (b)
Simulated gas flow around the microspike at Lins = 0. The black solid lines plot the
trajectories of the gas molecules.
2.6 Perturbation of the Knudsen flow

To further test the validity of this physical picture, the nanospike was gradually
inserted axially into a glass capillary with an inner diameter of 100 μm, so as to
perturb the gas flow around the nanospike (see inset in Fig. 2.8(a)). The mechani-
cal motion of the nanospike was monitored by measuring the position of the beam
emerging from the capillary using a quadrant photodiode (QPD). In the experiment
fR was found to fall as the insertion length Lins was increased from -0.5 to 0.3 mm
(Fig. 2.8(a)), where positive values of Lins mean that the nanospike was actually
inserted in the capillary. The measurement was performed at a pressure of 10−3
mbar, when the Knudsen stiffness is maximum. In this configuration, because of the
large final diamater of the nanospike and the large inner diameter of the capillary,
the optical forces described in Ref. [28] and subject of Chapter 3 are completely neg-
ligible. As the capillary does not physically or optically interact with the nanospike,
the drop in fR can only be caused by a change in the molecular flow around the tip.
The simulation in Fig. 2.8(b) shows the disruption of the streamlines of gas flow,
resulting in suppression of the Knudsen effect.
Conclusions of this Chapter

In this chapter, glass-fibre nanospikes were shown to be effective probes of optother-
mal and optomechanical effects at low gas pressures. The very small linewidth of the
39
fundamental mechanical mode allowed the resolution of tiny shifts in the resonance
frequency caused by the interaction between the nanospike and the thermally excited
molecular flow. Such a high sensitivity to the surrounding environment may help
to disclose the mystery behind the anomalous behaviour of laser tweezered particles
under high vacuum conditions, when residual laser absorption creates temperature
gradients.
40
CHAPTER 3
OPTOMECHANICAL INTERACTION OF A
NANOSPIKE INSERTED INTO AN HOLLOW
CORE PHOTONIC CRYSTAL FIBRES
The pioneering work of Ashkin on optical levitation demonstrated that focused laser
beams can exert forces large enough to manipulate micrometer sized objects [5, 6].
Mainly motivated by the capability of optical tweezers as ultra-sensitive probes and
their numerous applications in biology and medicine, several research groups tried to
extend this idea to nano-scaled particles and waveguides. However, since the optical
forces scale linearly with the particle volume, stable optical trapping was difficult
to achieve [78]. Several approaches were proposed to enhance the trapping stiffness,
including the use of plasmonic resonances in metallic particles [16] or phase shaping
of the laser beam for improved light-matter momentum transfer [79]. Particularly
successful has been the self-induced back-action mechanism, in which the optical
forces on a particle are strengthened by feedback from a planar photonic crystal
cavity [80] or a nanoaperture in a metallic film [17].
In this Chapter we introduce a novel optomechanical system consisting of a
glass-fibre nanospike inserted in a hollow-core photonic crystal fibre (HC-PCF) (see
Fig. 3.1). Laser light, launched into nanospike adiabatically evolves into a mode that
extends strongly into the surrounding space as the nanospike diameter falls below
the optical wavelength, eventually becoming comparable to the diameter of the
hollow core. At this point, the nanospike starts to perceive the optical confinement
provided by the photonic crystal cladding as its movements modify the shape of
the optical mode. The consequence is the appearance of a strong optomechanical
41
Ch. 3 OPTOMECHANICAL INTERACTION OF A NANOSPIKE INSERTED . . .
force, caused by the interplay and back-action between nanospike and HC-PCF,
which traps the nanospike at the core centre. At low gas pressure, the excellent
mechanical properties of the nanospike allowed direct measurements of these forces
through the optical spring effect.
The results here presented demonstrate that the combined action of adiabatic
following of light and optomechanical trapping permit self-stabilized, self-aligned
and Fresnel-reflection free coupling between a single-mode fibre (SMF) and a HC-
PCF with an experimentally measured efficiency of 87.8%.
Even if the actual optical coupling between the two waveguides takes place
over just a few tens of microns, systematic characterization of the proposed ap-
proach required the use of a large vacuum chamber and high resolution actuators.
This bulky set-up is unlikely to find direct practical applications. Our efforts to
dramatically simplify the degree of control needed for efficient coupling led to the
design and fabrication of a compact gas-cell, which serves as an interface between
SMF and HC-PCF. This device is just a few centimetre long, but exhibits all the ad-
vantageous properties of the novel light coupling mechanism and at the same time,
being sealed, allows operation either in vacuum or at high pressure.
The results presented in this chapter have been published in Ref. [28, 29].
3.1 Self-alignment of a nanospike by optomechanical

back-action in HC-PCF
3.1.1 The main idea
As described in Chapter 1, during the propagation along the nanospike, light evolves
adiabatically from the core mode of the SMF fibre to the fundamental mode of a
glass-air waveguide. The tapering angle for the nanospike used in the experiment
is shown as a function the taper diameter in Fig. 3.2(b), in which the boundaries
of the adiabaticity region were calculated using the length-scale criterion (see Eq.
1.1) [51]. When the nanospike diameter drops below the optical wavelength (1064
nm), light spreads out from the glass and starts to be guided in the evanescent field
[34]. Under these conditions, the mode field diameter (MFD) dramatically depends
on the nanospike diameter and can exceeding several tens of μm for tip diameters
below 200 nm (see Fig. 3.5(a)). If the nanospike is inserted into a HC-PCF a second
phase of adiabatic following occurs. When the MFD becomes comparable with the
HC-PCF diameter, the optical mode begins to feel the presence of the hollow core
42
Self-alignment of a nanospike by optomechanical back-action in HC-PCF 3.1
Fig. 3.1: Glass-fibre nanospike couple to a HC-PCF. Left inset: optical micrograph of
a silica nanospike inserted into HC-PCF (upper panel) and SEM of the final section of
the nanospike (lower panel); Right inset: SEM of the HC-PCF structure and measured
near-field profile of the mode excited by the nanospike
and eventually it is captured and guided by the HC-PCF. At this stage, the length-
scale criterion can be used once again to estimate the conditions for an adiabatic
transition, by calculating the propagation constants of the fundamental and the first
higher-order mode of the nanospike plus hollow core structure as a function of the
nanospike diameter. The result is shown in Fig. 3.2(b, right panel), for different
values of the nanospike offsets δ from the core centre. Adiabaticity was satisfied,
provided that δ < 3 μm.
The optical mode of the nanospike plus hollow core can be reasonably ap-
proximated with that of a glass capillary containing a glass strand, whose diameter
equals that of the nanospike. The FEM simulated evolution of optical mode over
the 50 μm insertion length in the HC-HCF is illustrated in Fig. 3.2(a), in which
the white dashed line indicates the local MFD. Fig. 3.4(c) shows the Poynting vec-
tor distribution for nanospike displacements from the core centre δ of 1 μm and 3
μm. It is possible to notice that movements of the nanospike strongly perturb the
optical mode, leading to the appearance of optical restoring forces which traps the
nanospike at the core centre.
43
Fig. 3.2: (a) Simulated adiabatic evolution of the nanospike mode (z component of Poynt-
ing vector is plotted) over the 50 μm insertion length, with the nanospike placed at the
core centre. The gray-shaded area represents the core wall and the dashed curves indicate
the local mode field diameter (MFD). (b) Left panel: measured local taper angle versus
diameter for the whole pretaper and nanospike (blue open-circles) before insertion into
the HC-PCF. The solid black line represents the adiabaticity criterion in free space. Right
panel: local taper angle versus diameter close to the tip of the nanospike when it is inserted
50 μm into the HC-PCF. The solid curves show the adiabaticity criteria for the nanospike
plus hollow core structure with different offsets δ of the nanospike from the core centre.
The joined circles show the actual taper angle versus diameter. Adiabaticity is violated
for δ 3 μm at a taper diameter of about 160 nm
44
3.1.2 Calculation of the optical forces

To unveil the relevance of the optomechanical interaction in the considered system,
it is crucial to estimate the magnitude of the optical force acting on the nanospike
when it is inserted in the HC-PCF and compare it to the mechanical restoring force
of the nanospike. In the following two different methods to calculate optical forces
are discussed and their results compared.
Maxwell stress tensor approach
Optical forces can be rigorously calculated using the so-called Maxwell stress tensor
formalism, which is based on the conservation of the total linear momentum in
classical electrodynamics. The latter states that [81]:
dpEM (r, t) dpM E (r, t)

+ = ∇ · T(r, t) (3.1)
dt dt
where pEM = (E × H)/c2 is the electromagnetic momentum density, pM E is the

mechanical momentum density, and T is the Maxwell stress tensor (MST), whose
components are given by [82]:
1
Tαβ = (Eα Dβ + Hα Bβ − (E · D + H · B)δαβ ) (3.2)
2
where δαβ is the Kronecker delta. For harmonically oscillating electromagnetic fields,
the time derivative of the electromagnetic momentum averaged over one cycle is zero
(< dpEM /dt >= 0), therefore, the optical force, as stated by Newton’s second law,
can be calculated by integrating the time derivative of the mechanical momentum
density (Eq. 3.1) over an arbitrary volume V which contains the optical system of
interest:
dpM E (r, t)
Fopt = dV = ∇ · T(r, t)dV (3.3)
V dt V
The term inside the integral in Eq. 3.3 plays the role of force density and using
Maxwell’s equations can be expressed as:
1 1 ∂
fopt = ∇ · T(r, t) = ρE + j × B − E 2 ∇ − H 2 ∇μ + (D × B) (3.4)
2 2 ∂t
Since we deal with glass waveguides surrounded by air or vacuum, we can restrict the
calculation to the special case of an isotropic, dielectric and non magnetic medium
without free charges or currents (ρ = 0, j = 0 and μ = μ0 ). In addition we observe
that the last term in Eq. 3.4 vanishes when time averages are considered. Hence,
45
the force density can be expressed in a remarkably simple form:
1
fopt = − E 2 ∇ (3.5)
2
Since the dielectric constant is constant inside the glass waveguide and in vacuum,
but presents a discontinuity at the interfaces between these two materials, its gra-
dient produces a Dirac δ centred on the afore mentioned interfaced, that simply the
integration over the volume V , to the surface integral:

1 1 2
Fopt = − E 2 ∇ dV = E Δn dΣ (3.6)
V 2 Σ 2
where Σ in our case is the nanospike surface, Δ = glass − 0 > 0 and n is a unity
vector normal to the surface towards the low index material. Eq. 3.6 demonstrates
that the optical force is localized at the boundary between the two different ma-
terial, it is always normal to the interface and points from the high to low index
medium. Unfortunately, in general the integral in Eq. 3.6 is not properly defined,
because the normal component of the electric field E⊥ is discontinuous at the glass-
vacuum interface (only E is continuous). This apparent complication can be easily
solved by repeating the calculation using the normal component of the displacement
D⊥ = E⊥ , which does not present discontinuities. The correct expression for the
calculation of the optical force is therefore [83, 84]:
1 2 2
1
Fopt = E Δ−D⊥ Δ(−1 ) n dΣ = < E2 >t Δ− < D⊥ 2
>t Δ(−1 ) n dΣ
2 Σ 4 Σ
(3.7)
−1 −1 −1
where Δ( ) = (glass − 0 ). In Eq. 3.7 all the field are continuous and the integral
is well defined. In the second equality, the additional factor 1/2 results from the
time average (indicated as < >t ) of the electric and displacement fields.
The MST approach is a very effective technique for the calculation of the
optical force, but since it depends on the integral of the fields components, it offers
little physical insights about how the optical forces are related to the parameters of
the physical system (e.g. the position of the nanospike in the hollow core).
Response theory
An alternatively approach based on energy considerations was first proposed by

Povinelli et al. in Refs. [7, 85]. The main idea is to derive the force as spatial
derivative of an optomechanical potential Uopt , which depends of the number of
46
photons stored in the system and their optical frequency. In the case the optical
force between two dielectric waveguides evanescently coupled, Fopt can be estimated
as [7, 86]:

dUopt (q) ω(q)
Fopt (q) = − = −N (3.8)
dq k dq k
where Uopt indicates the energy stored in the considered optical eigenmode, N is
the photon number, is the Planck constant and ω is the angular frequency of the
eigenmode. In Eq. 3.8 q represent a generalized coordinate of the system (e.g. the
distance between the two waveguides) and the derivative is calculated for a fixed
eigenmode wave.
More recently, Rakich et al. [87] using a similar approach, named by the
authors Response Theory of Optical Forces, concluded that the knowledge of the
phase and the amplitude response of an optomechanical system suffices to quantify
the mechanical action of the light field. In particular for a linear lossless media, the
optical force can be computed as:
dUopt (P, q) φ(q)

Fopt (q) = − = Φ(P ) (3.9)
dq dq
where Φ is the photon flux through the waveguide, P is the optical power and φ
represent the phase response of the optomechanical system. In the case of two
coupled dielectric waveguides it is possible to show that [87]:
P ω
Φ(P ) = , φ(q) = nef f (q)L (3.10)
ω c
where nef f is the effective index of the propagating mode and L is the length of the
waveguide. By substituting these relations in Eq. 3.9, we obtain a very convenient
expression to estimate the optical forces:
P ∂nef f (q)
Fopt (q) = L (3.11)
c ∂q
Under the specified assumptions, Eq. 3.11, which only requires the knowledge of
nef f as a function of the system coordinate q, provides the same results as the full
calculation performed using the MST, for which, instead, all the components of the
electromagnetic field are needed.
47
Optical forces acting on a nanospike inserted in a HC-PCF
The two approaches presented above were use to estimate the optical forces acting
on a nanospike inserted in a HC-PCF. Starting from the Maxwell stress method, we
calculated the field distribution in the system via finite element modelling, using the
local mode approximation [36], which is certainly valid if the optical propagation is
adiabatic. Integration of the MST around the nanospike allowed us to obtain the
local force per unit length. The total restoring force Fopt was then found by further
integration over the insertion length (≈ 50 μm). The dependence of Fopt on the
offset δ from the core centre for a launched power of 1 W is shown in Fig. 3.4(a)
(open circles). Fopt was also estimated using the response theory. In this case, using
finite element modelling, nef f was calculated as a function of δ and the result is
reported as an orange dashed curve in Fig. 3.4(a). Since the nanospike diameter in
not constant along the fibre axis, the total force was calculated by integration over
the insertion length:
P L ∂nef f (z, δ)
Fopt (q) = dz (3.12)
c 0 ∂δ
The result is shown in Fig. 3.4(a) as a blue dashed line and agrees with the one
obtained using the MST.
For comparison, the red line in Fig. 3.4(a) illustrates the optical force exerted
on the nanospike by a fixed Gaussian beam, whose waist matches to one of the fun-
damental mode of the HC-PCF. The maximum trapping force and optical stiffness
in the HC-PCF are an order of magnitude stronger than in conventional optical
tweezers. The reason lies in the different trapping mechanism in this two cases. In
particular, as shown in Fig. 3.4(c), the optical mode of the nanospike plus hollow
core is very sensitive to the nanospike displacement, causing an enhancement of the
optical force.
3.1.3 The experimental configuration

Fig. 3.1 shows a sketch of the experimental configuration. The nanospike was
fashioned by tapering standard SMF-980 down to waist diameter of 400 nm, which
was then further reduced to 150 nm by HF etching (see Chapter 1). The left inset
of Fig. 3.1(lower panel) depicts a scanning electron micrograph of the nanospike
tip. The HC-PCF exploited in the experiment was a photonic bandgap fibre with
a core diameter of 12.1 μm (right inset of Fig. 3.1). The setup used to perform
the experimental measurement is depicted in Fig. 3.3. Both the nanospike and a
30 cm long piece of HC-PCF (coiled with a diameter of 9 cm) were placed in a
48
Fig. 3.3: Sketch of the experimental setup used to insert a nanospike into a HC-PCF;
PD, photodiode; PC, polarization controller; NS, nanospike; QPD, quadrant photodiode
vacuum chamber to control the surrounding gas pressure. A stepper motor with a
step resolution of 50 nm was used to adjust the position of the nanospike and CCD1
was used to image the system inside the vacuum chamber (see left inset in Fig.
3.1(upper panel)). The light coming from laser at wavelength of 1064 nm, whose
power was monitored using a 99:1 coupler (photodiode PD1), was launched into
the nanospike and collected with the HC-PCF. The photodiode PD2 was used to
measure the backreflections. A polarization controller (PC) was used to optimize for
maximum transmission. The near-field profile from the endface of the HC-PCF was
imaged onto a CCD camera (CCD2) using a lens. A quadrant photodiode (QPD)
was use to monitor the mechanical motion of the nanospike and its response was
calibrated measuring the output signal after displacing the nanospike by a known
value. Experimentally, a linear relation was found for displacements below ≈ 3 μm
from the core centre. The mechanical spectrum of the nanospike was then estimated
as Fourier transform of the QPD output signal. The mechanical properties of the
nanospike were characterized by launching a low power (10 μmW ) optical probe
when the nanospike was placed just in front of the endface of the HC-PCF, without
being inserted into it. The intrinsic resonance frequency of the nanospike was 1016
Hz and the its mechanical Q-factor 1.26 105 .
3.1.4 Optical spring inside HC-PCF
Similar to the optical spring effect measured in optical tweezers setups [88], the above
calculated optical force Fopt will act as a additional restoring force and thus modify
the resonance frequency of the nanospike. Experimentally Fopt can be estimated
by measuring the mechanical resonance of the nanospike as a function of the laser
power. For a simple harmonic oscillator model, the observed resonant frequency fR
49
is proportional to the square root of the stiffness:

(km + kopt )
fR = (3.13)
mef f
where km represent the mechanical stiffness of the nanospike, kopt = −∂F/∂δ is the
additional optical stiffness and mef f is the effective mass of the nanospike, which
can be calculated using FEM after measuring the nanospike geometric profile. Since
kopt is proportional to the optical power P , it is possible to write fR2 + fm
2
∝ P,
where fm = km /mef f is the intrinsic resonant frequency of the tip.
The optical forces Fopt was characterized by inserting the nanospike ≈ 50 μm
into the HC-PCF and measuring the resonant frequency as a function of the laser
power. The experiment was performed at a pressure of 4 10−1 mbar to ensure
a high enough mechanical Q-factor. Fig. 3.4(d) shows the mechanical spectrum
of the nanospike for several values of the laser power. The large frequency shift
observed proves very strong optomechanical interaction. As expected fR2 + fm 2
was
found to be proportional to the laser power (Fig. 3.4(b), green squares). Here
the dashed line was calculating without fitting parameters from FEM simulations.
The measurement was repeated displacing the nanospike from the core centre of
1 and 2 μm (orange circles and purple triangles respectively in Fig. 3.4(b)). For
low launched power the measurement agrees nicely with the simulations, however at
higher power level, since the optical stiffness becomes comparable to the mechanical
one, the nanospike equilibrium position is pulled toward the centre, causing an
increase of Fopt (see Fig. 3.4(a)).
Finally, the optical stiffness was measured at a pressure of 3 10−4 mbar, at
which the high Q-factor of the nanospike could be fully exploited. The result is
shown in Fig. 3.4(e). Given the very narrow linewidth of the mechanical resonance
(≈ 8 mHz), the system was sensitive to a minimum shifts in the stiffness of 8
attoN/μm.
3.1.5 Self-aligned high-efficient coupling into HC-PCF

Insertion of the nanospike in the HC-PCF also provides an elegant way of cou-
pling light from SMF to HC-PCF with high efficiency, while the optical forces pro-
vide automatic stabilization of the optical coupling against external perturbations.
Fig. 3.5(a) clarifies how the coupling efficiency depends of the final diameter of the
nanospike. Here the calculated fundamental MFD at the nanospike endface, placed
50
Fig. 3.4: Calculation of the optomechanical force and optical spring effect. (a) Simulated
optical trapping force for 1 W of power plotted against the nanospike offset from the core
centre, calculated by integrating the Maxwell stress tensor (blue open-circles) and using
response theory (blue dashed line). The red solid curve shows the trapping force when
a focused Gaussian beam, with waist matching the MFD of the HC-PCF, is used. The
orange dashed line plots the calculated effective mode index of the fundamental mode
of a hollow core with a 150 nm glass strand placed inside. (b) Scaling of measured and
simulated values of f 2 R − f 2 m versus power for different values of base offset Delta from
fibre axis. (c) Simulated Poynting vector distributions of the supermode at δ = 1 μm
and 3 μm for a hollow core containing a 150 nm glass strand. The bottom figures show
the zoom-in around the strand. (d), (e) Measured Brownian motion spectra at (d) 0.4
mbar and (e) 3.2 10−4 mbar for several different power levels. The solid curves are fits to
Lorentzian lineshapes.
51
in free space (blue curve) and inside the HC-PCF (orange curve), is shown against
the nanospike tip diameter. At a nanospike end-diameter of about 190 nm, the free-
space mode already has an MFD equal to the MFD of the HC-PCF, indicating that
light will be captured and guided by the HC-PCF when the nanospike is inserted
into it. Below this critical point, the MFD of the nanospike plus HC-PCF slowly
approaches the MFD of the HC-PCF, ensuring perfect mode matching and near-
unity coupling efficiency regardless of the particular HC-PCF core diameter. In Fig.
3.5(a) the purple dot-dashed curve was obtained by calculating the overlap integral
between the mode of the nanospike plus HC-PCF and the one of the HC-PCF, with
the purple points corresponding to measured results from different samples.
The launch efficiency was measured at atmospheric pressure with the nanospike
placed at the core centre. The measured total transmission from the SMF to the
output end of the HC-PCF was 81%, which, considering the loss of the taper (0.31
dB) and of the HC-PCF (0.13 dB/m), corresponds to a launch efficiency of 87.8%.
The result is not far from the theoretical prediction of 88.3% as reported in in Fig.
3.5(a). The measured mode profile is depicted in Fig. 3.1(right inset) and shows
no contaminations from high order modes. The backreflection was measured to
be 0.05%, which is significantly lower than the 4% Fresnel reflection from a silica-
air interface. The total transmission was found to drop by ≈ 10% when the laser
polarization was adjusted to the orthogonal polarization state, most likely because
of polarization-dependent loss in the taper transition and in the HC-PCF.
Fig. 3.5(b) investigates the optical transmission as a function of the launched
laser power. In this plot, the error bars are the standard deviation of the time-
domain signal collected over 200 s normalized to the input power. When the
nanospike was placed in the core centre (blue curve) a clear reduction of the er-
ror bars proves stabilization of the nanospike position and therefore of the optical
coupling due to the stronger trapping force. This self-stabilization is more clearly
underlined in the inset of Fig. 3.5(b), which compares the time-domain traces at
low (1 mW) and high (450 mW) power. When the power is increased above 100
mW, fluctuations in the transmission are dominated by intensity noise of the laser
source.
When the nanospike was intentionally offset up to 4 μm from the core centre,
the transmission improved with increasing power (Fig. 3.5(b)), indicating that the
nanospike was pulled toward the core centre by the optical forces. At around 200
mW the transmission saturated to a value about 3% lower than when the nanospike
was placed in the core centre, perhaps because of the slight bending of the entire
52
Nanospike based gas-cell 3.2
Fig. 3.5: (a) Simulated mode-field diameter for a nanospike in free space (blue) and in the
centre of the hollow core (orange) for a core diameter of 12.1 μm. For tip diameters below
190 nm, the fundamental mode is guided mainly by the HC-PCF (gray-shaded area). The
purple curve shows how the coupling efficiency varies with the final tip diameter when
the nanospike is centred in the core. The solid purple dots show the coupling efficiencies
measured for nanospikes with different final tip diameters. (b) Self-alignment and self-
stabilization measured at atmospheric pressure, when the base of the nanospike is offset
from the centre by 0, 1, 2, 3 and 4 μm. Inset: transmitted signal recorded over 200 s at 450
mW (dark green) and 1 mW (bright green) when the nanospike is at the core centre. The
gray curve is the laser output monitored simultaneously (via PD1, scaled for comparison)
at 450 mW, showing that the noise on the transmitted signal is caused by the laser.
taper.
3.2 Nanospike based gas-cell

In contrast with conventional optical fibres, HC-PCFs, being hollow, can be filled
with gases or liquids [89, 90]. Already in the early days, this property was exploited
to increase light-matter interaction lengths thousands of times with respect to the
Rayleigh limit [91]. The suppression of beam diffraction and the consequent en-
hancement of non-linear light scattering gave HC-PCFs a leading role in ultrafast
and quantum optics [92, 93], atomic physics [94–96] and gas spectroscopy [97]. Gen-
erally these experiments employ macroscopic gas cells or vacuum chambers, whose
dimensions are limited by the size of the optical and gas components and are hardly
scalable. Moreover, external perturbations can strongly influence the system align-
ment, which becomes critical at high power levels, because of the increased overlap
between incident beam and the fibre microstructure.
All-fibre based gas cells were fabricated by direct fusion-splicing solid-core
53
single-mode fibres (SMF) to HC-PCF [98]. The resulting devices were indeed much
more compact and stable compared to free-space arrangements, but efficient cou-
pling, with 1-2 dB insertion loss, could only be achieved if the mode field diameters
(MFD) of the two fibres matched [99]. For example, considering the MFD of a
standard SMF, which is about 10 μm, low-loss coupling can only be obtained using
HC-PCFs with core diameters of ≈ 14.5 μm. Different core diameters will experience
a lossy and highly multimode optical coupling. Furthermore, the presence of a sharp
glass-air interface causes unavoidable back-reflections, which can be detrimental in
applications at high-power levels. [100]. Finally, broadband-guiding HC-PCFs with
thin walls (e.g. kagome-style or single-ring HC-PCF designed for guidance in the
ultraviolet [101]) are not suitable for fusion splicing, since even modest arc powers
easily damage the microstructure and result in severe optical losses.
We soon realized that the techniques described earlier in this Chapter solve
many of these challenges by offering automatic MFD-matching for any hollow-core
diameter, together with almost complete suppression of Fresnel reflections and op-
tomechanical self-alignment. Therefore we decided to face the problem of dramati-
cally reducing the size of the setup in Fig. 3.3 and design a fully integrated all-fibre
coupled device that can operate as a compact in-line gas cell. The result was a
compact gas cell only a few cm long (see Fig. 3.6 and Fig. 3.7(a,b)) which allows
operation in high gas pressure or ultra high vacuum. To prove the potential of
the device for non-linear light-gas interactions, stimulated Raman scattering and
molecular modulation of light were demonstrated in a H2 -filled HC-PCF. Other po-
tential applications of the device include the delivery of high-power laser light and
optofluidic experiments in liquid-filled HC-PCFs [102, 103].
3.2.1 Fabrication techniques

For convenience during the fabrication of the device and insertion process, the
nanospike should be as short as possible, because long nanospikes (> 15 mm) are
mechanically highly compliant and easily attach to the wall of the HC-PCF. An abi-
aticity factor F = 0.5 (see Chapter 1) was found to be a good compromise between
short transition lengths and single mode, low-loss guidance (transmission >95%).
The blue solid curve in Fig. 3.8 shows the profile a nanospike adiabatic at a wave-
length of 1550 nm and the blue dots mark the measured taper diameters. In the same
plot, the solid orange curve represent the calculated local effective index (nef f ) of the
fundamental mode. When the taper diameter falls below 5 μm, nef f suddenly drops
dramatically from 1.45 to 1, indicating that the optical mode spreads out signifi-
54
Fig. 3.6: Photograph of a nanospike gas-cell
cantly into the surrounding air. The consequence is an almost complete suppression
of the Fresnel reflection. Subsequently, the nanospike was inserted into a glass cap-
illary C1 with an inner diameter (ID) of 250 μm and an outer diameter (OD) of
1600 μm and the HC-PCF was inserted into capillary C2 (ID 320 μm, OD 1400 μm)
(Fig. 3.7(d)). The nanospike/C1 bundle was then inserted into capillary C3 (ID
2000 μm, OD 2400 μm). Vacuum compatible epoxy was used to secure the junctions
(Fig. 3.7(e)). The wall thickness of the outer capillary C3 (400 μm) was selected to
be just thick enough to withstand high pressure while allowing a clear view of the
device in an optical microscope. The HC-PCF/C2 bundle was then inserted into
C3 and placed facing the nanospike (Fig. 3.7(f)). To precisely control the relative
position of the nanospike and the HC-PCF, the latter was mounted on a stage with
3 translational and 2 rotational axes. Fine alignment of the system and insertion
of the nanospike into the hollow core were performed under a microscope with 10x
magnification. The optical coupling was optimized by monitoring the output mode
profile with a CCD camera and the total transmitted power with a power-meter.
Note that precise matching of the capillary diameters ensures pre-alignment of the
system to within few tens of microns. This becomes important particularly in the
case of small core HC-PCF (core diameter < 15 μm), since excessive angular or
positional misalignment can cause the nanospike to attach to the core wall during
the insertion process. This does not normally damage the nanospike or affect its
55
Fig. 3.7: (a) Photograph of the gas cell beside a one-Eurocent coin. (b) Optical micro-
graph of the central portion of the gas cell containing a nanospike inserted into the HC
PCF. (c)-(f) The fabrication procedure of the gas cell. The black arrows indicate the inser-
tion direction of the capillaries, and the red arrows show the positions where the vacuum
epoxy is applied.
guidance properties, but, after extracting and discharging the nanospike, the fab-
rication procedure has to repeated from the beginning. Eventually the system was
secured by gluing C3 on the HC-PCF. During the ≈20 minute curing time the mode
profile and the transmission of the system were monitored and optimized to correct
for possible misalignment due to mechanical stress exerted by the hardening of the
glue. This operation was performed at low optical power (<5 mW) to avoid the in-
crease in the curing time caused by absorption of the scattered light by the vacuum
epoxy. Once the glue is set, fine adjustments of the alignment were no longer pos-
sible, nevertheless optomechanical self-alignment could still provide close to perfect
positioning of the nanospike at core centre.
3.2.2 Device performance

The measured transmission of one of the devices is depicted in Fig. 3.9 as a function
of the input power at a wavelength of 1550 nm. The final nanospike diameter was
≈ 350 nm and the HC-PCF core diameter 19 μm. The total transmission of ≈ 59%
was observed to increase to ≈ 69%, for a launched laser power of 1 W because of the
optomechanical self-alignment. The inset in Fig. 3.9 shows an image of the near-
field mode profile at the output of the HC-PCF, which has no sign of contamination
by HOMs. The measured back-reflection was < 0.05%. Other devices, fabricated
using the same procedure, featured similar performance. We point out that in a
real application, a strong pump beam could be used to optomechanically align the
56
Fig. 3.8: (a) Taper diameter and effective index nef f as a function of taper length. The
blue solid line shows the designed adiabatic profile and the dots show the experimental
measurements. (b) Final tip diameter required for optimal light coupling and U opt (see
text) as a function of the hollow core diameter normalized to the wavelength. Optome-
chanical self-alignment is less pronounced in the gray-shaded region.
nanospike at the centre of the core so as to optimize the coupling of a secondary

probe beam in another spectral window.
The proposed approach permits efficiently coupling of light from SMF into HC-
PCFs with significantly different core diameters by careful choice of final nanospike
diameter. As a rule of thumb, the optimal tip diameter for efficient coupling can be
estimated by matching the MFD at the end of the nanospike to the one of the HC-
PCF. Fig. 3.8(b) (blue curve, left axis) plots the calculated optimal tip diameter as a
function of HC-PCF core diameter, normalized to the laser wavelength λ. In general,
nanospikes with smaller tip diameters are necessary for larger HC-PCF cores. As
an example, Fig. 3.9(b) shows the measured mode profile in a single-ring HC-PCF
with a core diameter of 29 μm. In this case the measured total transmission was
≈ 48%, with the transmission loss mainly caused by the mismatch between the final
tip diameter (≈ 350 nm) and its optimal value (≈ 250 nm), see Fig. 3.8(b). The
resulting non-adiabaticity determines coupling into high loss higher-order modes,
reducing the overall transmission. As discussed in Chapter 1 chemical etching of
the nanospike could be used to reduce the tip diameter, but unfortunately in this
case the procedure was found to increase the fabrication failure rate, because etching
increases the fragility of the nanospike tip.
To test the sealing of the gas cell, high-pressure hydrogen gas (H2 ) was injected
from the other end of the HC-PCF. After closing the gas line, the pressure inside the
device was monitored. As plot in Fig. 3.9(c) the gas cell was able to hold a pressure
higher than 10 bar for more than 5 hours. The same sample was then evacuated
using a turbo-pump and the minimum pressure reached was 3 10−7 mbar.
57
Fig. 3.9: (a) Transmission of the system (including the losses of the the nanospike and
the HC PCF) as a function of launched power. Upper left inset, near-field mode profile
measured at the output end of the HC PCF (scale bar 10 μm). Lower right inset, sketch
of the experimental setup. PM, power meter. (b) Measured mode profile at the output
of a single-ring HC-PCF with core diameter 29 μm. (c) Measured pressure in the gas cell
over time.
3.2.3 Theoretical consideration on the gas cell design
To characterize the strength of the optomechanical self-alignment effect at different

core sizes under optimal coupling condition, we calculated the work done by the
optical forces (Uopt ) to displace the tip of the nanospike by one wavelength in the
vicinity of the core centre:
λ
1
Uopt = kopt xdx = kopt λ2 (3.14)
0 2
where kopt is the optical stiffness [28] and x represents a coordinate orthogonal to
the fibre axis. In this calculation the insertion length was set to 100 μm and the
optical power to 1 W. The result was found to be independent of the wavelength λ
and is shown in Fig. 3.8 (the orange line, right axis). The strength of the optical
restoring force decreases when the core size is increased. Considering that the typical
mechanical stiffness of a nanospike is km ≈ 1pN/μm, the light-grey shaded area
highlights the region in which kopt < km , indicating that here the optomechanical
self-alignment is less effective.
58
3.2.4 Observation of Raman scattering and molecular

modulation in H2 -filled HC-PCF
To prove the functionality of the gas cell, we used it to observe stimulated Raman
scattering (SRS) in H2 and molecular modulation inside the HC-PCF [91, 104].
The experimental setup is sketched in Fig. 3.10(c). A laser beam with a central
wavelength of 1030 nm, a pulse duration of 1.8 ns and a repetition rate of 2 kHz
was launched into the gas cell through ≈ 1 m of SMF. The HC-PCF used was a
photonic bandgap fibre with a core diameter of 10 μm and about 9 m long. The H2
gas was injected from the other end of the HC-PCF using a traditional gas cell. The
transmitted light was collected using a 5x NIR-objective and delivered to an optical
spectrum analyser (OSA). The experiment was performed at frequencies separated
by H2 pressure of 10 bar.
The output spectrum recorded at a pulse energy of 1.5 μ J is shown in Fig.

3.10(a). The first (S1) and second Stokes (S2) and the first anti-Stokes (AS1) lines
appeared at a distance ΩR /2π = 18 THz from the pump, which correspond to the
rotational frequency of ortho-hydrogen. A tiny peak (S1p) was also detectable at
11 THz from the pump, which equals the rotational frequency of para-hydrogen.
A more complicated optical spectrum was observed, when the input pulse energy
was increased to 3 μ J (Fig. 3.10(b)). During the propagation, the pump beam is
scattered by the H2 molecules and generates the Stokes signal. The beating of these
two wavelengths creates a synchronized motion of the H2 molecules, usually referred
to as coherence wave (Cw) [104, 105], that can be exploited to efficiently scatter
other wavelengths if phase-matching condition is fullfilled. This contributes to the
additional sidebands around S1p at frequency spacings corresponding to ΩR /2π
(green arrows). Given the relatively small frequency range considered here, the
phase matching condition is roughly met. The appreance of other peaks with broader
linewidth is attributed to nonlinear interaction in the SMF before the gas cell (see
the reference spectrum in Fig. 3.10(d) collected with an evacuated HC-PCF). In
particular, four-wave mixing in the normal dispersion regime is responsible for the
peaks labeled as A and A’ and Raman scattering in silica generates peak B at 13.2
THz from the pump [106]. Peak A is again scattered by Cw generating sidebands
(orange arrows). Note that the main peaks generated from silica are ≈ 30 dB weaker
than the Raman lines generated by the H2 gas.
59
Fig. 3.10: (a) Sketch of the experimental setup to observe stimulated Raman scattering
and molecular modulation in H 2 -filled HC-PCF. (b) Optical spectrum recorded with an
input pulse energy of 1.5 μ J and (c) 3.0 μ J. The H 2 pressure during the measurement
was about 10 bar. (d) Spectrum recorded with an input pulse energy of 3 μ J and with
the HC-PCF evacuated.
60
Conclusions of this Chapter

This Chapter introduced a new optomechanical system consisting of a glass-fibre
nanospike inserted in a HC-PCF. Adiabatic propagation of the optical field first
along the nanospike first and then in the HC-PCF results in highly efficient single-
mode optical coupling, free from Fresnel back-reflections, suitable for HC-PCFs with
arbitrary shape and core diameters. In addition, the optomechanical interaction
between the two waveguides gives rise to strong optical forces which cause self-
alignment and stabilization of the nanospike in the centre of the hollow core. The
high mechanical Q-factor of the nanospike made it possible to measure the above
mentioned optical forces with high precision by means of the optical spring effect.
Furthermore we showed that the entire system can be scaled, sealed and used as a
compact gas cell and we proved its potential application in nonlinear light-matter
interactions.
61
CHAPTER 4
OPTOMECHANICAL COOLING OF A
NANOSPIKE COUPLED TO A
WHISPERING-GALLERY-MODE BOTTLE
RESONATOR
In 1997 Knight et al. demonstrated light coupling into a whispering gallery mode
(WGM) resonator using a tapered optical fibre [107]. Adiabatic guidance of light, low
loss and large evanescent field, together with the complete compatibility with fibre-
based optical components were shown to be key elements for efficient, stable and
practically convenient coupling. There is no doubt that today this represents one of
the most common applications of fibre tapers. Of course, since glass-fibre nanospikes
inherit directly the excellent optical properties of traditional fibre tapers with the
additional benefit of high mechanical Q-factors, we found intriguing the possibility
of optomechanically coupling a nanospike to a WGM optical microcavity. Indeed,
as extensively investigated in the rapidly growing field of cavity optomechanics,
coupling a harmonic oscillator to an optical cavity provides an elegant and powerful
means of tailoring its mechanical response [20].
One of the prominent results achieved in this field is so-called optomechanical
cooling, in which the coupling with the cavity is exploited to transfer energy from
the mechanical motion to the light field. This unidirectional flow of energy results
in the effective cooling of the mechanical oscillator. More precisely, the optical
coupling to the cavity manifests itself as an additional optical damping (also known
as cold damping) acting on the mechanical oscillator, which, in sharp contrast with
traditional damping, does not increase the thermal noise.
63
Ch. 4 OPTOMECHANICAL COOLING OF A NANOSPIKE COUPLED TO A . . .
In the most common experimental configurations the optomechanical coupling

has been dispersive, which means that the motion of the harmonic oscillator changes
the cavity resonance frequency. Under these conditions, when the pump laser is red-
detuned from the cavity resonance, cooling is indeed very efficient in the sideband-
resolved regime, in which the mechanical frequency Ω is much larger than the cavity
decay rate γ (Ω/γ » 1). Since 2011, when Teufel et al. and independently Chan et
al. experimentally demonstrated optomechanical cooling of mesoscopic mechanical
oscillators to their quantum ground state [25, 26], the access to the sideband-resolved
regime seemed to offer an unprecedented opportunity for observing quantum me-
chanical effects on the macroscale. Increasing the mass or the size of the harmonic
oscillator, however, implies a decrease of its resonant frequency, limiting the effi-
ciency of sideband-resolved cooling approximately to the MHz range. Nevertheless,
numerous optomechanical systems operate at lower resonance frequencies, as for in-
stance the mirrors used in LIGO [108], ultracold atomic gases [109] and suspended
micro-mirrors [58].
To overcome this limitation several schemes have been proposed over the past
decade [30, 110–112]. In particular it has been predicted [30, 31] that dissipative
optomechanical coupling, in which the motion of the mechanical oscillator modulates
the cavity decay rate, could provide an alternative cooling mechanism for any value
of Ω/γ. So far, dissipative cooling has been reported only in one specific system
consisting of a silicon nitride membrane placed in a Michelson-Sagnac interferometer
[113]. In 2009 Li et al. showed that dissipative coupling is also a key element of
the simpler optical system consisting of a free-standing waveguide (as for example
a nanospike) coupled to a WGM resonator. Despite great efforts to engineer such a
configuration [114–116], the limited mechanical properties of the selected waveguides
have prohibited demonstration of dissipative cooling.
In this Chapter we experimentally demonstrate passive cooling of the me-

chanical motion of a nanospike via dissipative optomechanical interactions with the
neighbouring WGM resonator. When launching ≈ 250 μW laser power at an opti-
cal frequency close to the WGM resonance, cooling of the fundamental mechanical
mode of the nanospike from room temperature down to 1.8 K was measured. Si-
multaneous cooling of the first high order mechanical mode was also observed. The
result is strong suppression of Brownian nanospike motion and demonstration of
self-stabilisation of the waveguide-WGM coupling. Furthermore, given the low me-
chanical frequency of the nanospike (around 2.5 kHz), the results illustrate the ca-
pabilities of dissipative optomechanical coupling for achieving efficient cooling even
64
Whispering gallery mode resonators 4.1
for Ω/γ as low as 10−4 . To the best of our knowledge, this system has the lowest
mechanical frequency that could be passively cooled so far and represents the first
practical application of dissipative optomechanics. We think our result will find di-
rect impact on the numerous applications of WGM resonators, including nonlinear
optics [117], atom physics [118], optomechanics [119, 120] and sensing [121, 122].
4.1 Whispering gallery mode resonators
The phenomena of the whispering gallery, of which there is a good

and accessible example in St. Paul’s cathedral, indicate that sonorous
vibrations have a tendency to cling to a concave surface.
So begins the classical paper from Lord Rayleigh "The problem of the whispering
gallery" [123] (1910), which provides the first explanation in terms of propagating
waves of a curios phenomenon observed in the St. Paul’s cathedral in London: a
whisper on one side of the dome can be heard by on the other side of the dome by
placing the ear close to the wall. Lord Rayleigh proposed that the sound waves,
repeatedly reflected from the concave wall of the dome, could propagate along the
surface and subsequently interfere, producing the spatial pattern that today we
refer to as whispering gallery modes. The gap between acoustic wave to light wave
is relatively short and soon Mie realized that similar features characterize the light
scattered by spherical objects with diameter comparable to light wavelength [124].
In such small structures, these Mie resonances have very small Q-factor and for
decades their potential as optical cavities remained concealed.
The first breakthrough in the field of WGM resonator dates 1989, when Bra-
ginsky et al. reported the fabrication of fused silica microspheres, supporting WGMs
with ultra-high Q-factors in the order of 109 [125]. Today WGMs of several shapes
and sizes, made from different materials have been reported that display remarkale
high Q-factor, including microspheres [125], microbubbles [126, 127], bottle res-
onators [128, 129], toroids [130], wedge-resonators [131] and ring resonators [132].
These families of WGM resonators found applications in different fields of linear and
non-linear optics (for a comprehensive review refer to Refs. [20, 117]).
65
4.1.1 Experimental techniques for light coupling to WGM

resonators
WGM cavities support ultra high-Q resonances because radiation loss into the far
field are very small. Consequently, direct coupling of light from free space is ex-
tremely inefficient and more sophisticated techniques are usually necessary.
Prism coupling
Historically, the first technique proposed to couple light into WGMs exploited a
prism placed close to the surface of the WGM resonator [125]. Making use of total
internal reflection of light at the prism-air interface, it is possible to produce an
evanescent wave, whose overlap with the WGM allows to light to be coupled into
the optical resonator. The coupling strength can be adjusted by modifying the
distance between the prism and the resonator and, in addition, by changing the
angle of incidence on the internal surface of the prism, it is possible to match the
propagation constant of the evanescent field with the one of the desired WGM. This
method can be very efficient and mechanically stable, but it is bulky and often
requires complex free space optics.
Fibre taper coupling
Introduced by Knight and co-workers [107], taper coupling is the most commonly
used technique to couple light not only in WGM resonators, but also in photonic
crystal waveguides and cavities [133]. When the diameter of a fibre taper becomes
comparable with the wavelength, a considerable fraction of the optical energy is
guided in the evanescent field. If the taper is placed close enough to a WGM
resonator, light can be coupled into the optical cavity. By properly choosing the
diameter of the tapered fibre, the propagation constant of the fundamental fibre
mode and the WGM can be matched. We observe that, due to the tapered profile,
fine tuning of the fibre propagation constant is possible by simply moving the taper
along the longitudinal direction. Precise adjustments of the coupling position in the
other two directions is also straightforward and simply requires a 3 axis translation
stage. This technique can be extremely efficient and has been shown to offer almost
ideal single mode to single mode coupling characteristics [134]. Since tapering a
single mode fibre can only decrease the propagation constant of the fundamental
mode, it is not possible to use traditional taper coupling with WGM fabricated
from materials with refractive index higher than silica (nsilica =1.45), unless the
66
Whispering gallery mode resonators 4.1
taper itself is fashioned from high index glass [135].
Free-space coupling
If the diameter of the WGM resonator is reduced below 30 μm, the quality factor
of the WGM decreases dramatically because of radiation losses [136–138]. Under
these circumstances it is possible to couple light into the WGM by simply focusing
a laser beam towards the structure. Even though the coupling efficiency is hard
to quantify, several recent experiments have shown interesting applications of this
technique [62, 139].
4.1.2 Light coupling between a single mode waveguide and a

WGM resonator
In the experiments described in this Chapter, light is coupled to WGM resonators
using the nanospike as a bus waveguide. Here, the goal is to provide a theoretical de-
scription of the coupling mechanism and to understand the link between the optical
field launched into the nanospike and the optical power stored into the cavity.
Our analysis starts by considering a general coupler as the one shown in Fig.
4.1(a). Here τ and ρ represent respectively the transmission and coupling coefficients
for the complex field amplitudes. By using reciprocity it is possible to derive impor-
tant relations between these two coefficients. The first step consists in multiplying
the fields in the diagram (a) by τ ∗ in order to obtain (b). On the other hand, flipping
the coupler (which is perfectly symmetric) and multiplying by ρ∗ gives (d). We can
reverse the fields after taking the complex conjugate and the result is reported in
(c). Finally, combining (c) and (d) we can derive (e), that, when compared to (b)
implies that:
|ρ|2 + |τ |2 = 1 and τ ∗ ρ + τ ρ∗ = 0 (4.1)
The first of these relations ensures power conservation, while the second requires
τ ∗ ρ be imaginary. Since for zero coupling τ needs to be real-valued, ρ must be
purely imaginary. For this reason, in the following the notation |ρ| = κ is adopted.
The case of a waveguide coupled to a ring cavity in depicted in Fig. 4.1(f) and is
obtained by simply connecting one of the output ports of the coupler to one of its
input ports.
The special case of a waveguide coupled to a WGM is illustrated in Fig. 4.2,
where sin and sout represent the input and output waveguide fields, while a1 and
a2 are the cavity fields before and after the coupling point. We note that all the
67
Fig. 4.1: Couplers and reciprocity. By using reciprocity it is possible to derive important
relations between the coupling coefficients τ and ρ.
fields are normalized such that their module squared represents the optical power.
Following the discussion above, the entire system can be simply described by the
matrix relation:
⎛ ⎞ ⎛ ⎞⎛ ⎞
s
⎝ out ⎠ = ⎝
τ −iκ ⎠ ⎝ sin ⎠
(4.2)
a1 iκ τ a2
In the stationary state, self-consistency of the fields after the propagation
around the cavity requires:
a2 = e(iβ−α)L a1 (4.3)
where β is the propagation constant of the cavity field, α the amplitude loss rate
and L the cavity length. From Eq. 4.2 the field a1 equals:
iκsin
a1 = iκsin + τ a2 = iκsin + τ e(iβ−α)L a1 ⇒ a1 = √
1 − 1 − exp((iβ − α)L)
κ2
(4.4)
√
where the relation τ = 1 − k has been used. If the laser frequency ωL is close
2
to one of the mth cavity resonance, with resonance frequency ωm , it is possible to

write:
βL = ωm Ln0 /c + (ωL − ωm )L/vG = 2mπ + θL/vG (4.5)
where n0 is the modal index, vG the group velocity and θ the detuning between the
cavity resonance and the laser frequency. For small detuning and cavity loss, Eq.
68
WGM bottle resonators 4.2
Fig. 4.2: Sketch of a single mode waveguide coupled to a WGM resonator.
4.4 can be approximated to:
i κsin i κsin /T
a1 = 2 (4.6)
κ2 /2 + αL − iθL/vG (κ /2 + αL)/T − iθ
where T = L/vG is the round-trip time. The energy stored in the cavity at frequency
ωL is given by:
Jc = |ec |2 = L|a1 |2 /vG (4.7)
If we indicate the external and intrinsic cavity energy decay rates respectively as
γe = κ2 /T and γi = 2αL/T we can finally obtain the link between the energy stored
in the cavity |ec |2 and the launched optical power |sin |2 :
|sin |2 γext
|ec |2 (4.8)
(γext + γi )2 /4 + θ2
4.2 WGM bottle resonators

Among the large variety of WGM resonator investigated in the last 20 years, we
decided to exploit the so-called bottle resonators because they can provide ultrahigh
Q-factor and their fabrication procedure, based on tapered fibres, fits very well our
experimental capabilities. First demonstrated experimentally in Ref. [52], the bottle
resonators are highly prolate dielectric resonators with a cylindrical symmetry. Their
optical properties have been recently theoretically and experimentally investigated
in Refs. [128, 129, 140].
69
Fig. 4.3: Fabrication of bottle resonators using a fibre splicer. For (a) to (c): the upper
panel is a sketch of the system, while the lower panel shows micrographs collected with
the imaging system integrated into the fibre splicer.
4.2.1 Fabrication and characterization of bottle resonators
The fabrication of bottle resonators involves a two steps process. First a SMF is
thermally tapered to create a uniform waist diameter of a few tens of μm and then
the taper is placed in an arc-splicer. Here an electric discharge locally heats up the
waist, while two sides of the taper are pushed towards each other (see Fig. 4.3(a)).
Surface tension causes the formation of the prolate shape of the bottle resonators.
Fig. 4.3 shows the images collected with the imaging system integrated in the arc-
splicer during the fabrication. Tuning the arc power and its duration allows precise
control of the resonator diameter. The curvature can also be tuned by properly
choosing the waist diameter to which the fibre is initially tapered. Optical Q-factors
in the range of 107 - 108 could be routinely obtained using this technique.
A typical microscope picture of a bottle resonator is shown in Fig. 4.4(a).

By applying the same image recognition algorithm described in Chapter 1, that
we developed to characterize fibre tapers, the local diameter along the longitudinal
coordinate could be reconstructed (see Fig. 4.4(b)). In the proximity of the bottle
equator, the obtained data fit well the parabola: D(z) = D0 (1 − 1/2(Δk z)2 ), where
D0 is the maximum bottle diameter and Δk is the bottle curvature. With our
technique we fabricated resonators with D0 raging from ≈ 10μm to a few hundred
μm and Δk between 0.005-0.025 μm−1 . Similar properties have been shown in Ref.
[129], in which bottle resonators were fabricated using focused a CO2 laser.
70
Observation of optomechanical cooling and self-stabilization of the nanospike 4.3
Fig. 4.4: (a) Top: optical micrograph of a bottle resonator. Bottom: image post-processed
using the techniques described in 1. (b) Reconstructed profile of the bottle resonator.
Inset: the region with larger diameter fits well a parabolic function, which allows precise
estimation of the resonator diameter and surface curvature.
4.3 Observation of optomechanical cooling and

self-stabilization of the nanospike
4.3.1 Experimental setup
The experimental setup is sketched in Fig. 4.5. The nanospike was fabricated by
tapering standard single mode fibre SMF-980 and the taper profile was engineered to
yield single-mode adiabatic guidance of light at 1150 nm (pump laser) and 1064 nm
(probe 1), while preserving high mechanical stiffness [27]. The resulting nanospike
was ≈8 mm long with a tip diameter of ≈700 nm. The fundamental mechanical
mode of the nanospike had a resonant frequency of 2.5 kHz and a Q-factor of 1.2
105 . The WGM bottle-resonator used had a diameter of 46 μm and an optical Q-
factor of 4.1 107 under critical coupling conditions. The experiment was conducted
in vacuum (10−5 mbar) to eliminate air damping. The vacuum chamber (see Fig.
4.6(b)) was assembled using a standard 6 port ISO-CF 63 cube. Custom flanges
and stainless steel holders were designed to permit optical access into the chamber
and to mount five vacuum compatible stepper motors, which permitted fine-tuning
of both the relative position and the orientation between the nanospike and the
WGM resonator. The single mode fibre, at the end of which the nanospike was
connected, was fed-through into the vacuum chamber using a Swagelok-type 1/16
inch connector and a PTFE ferrule with a 300 μm hole in the middle [141]. Two
microscopes were used to image the system from the side (Fig. 4.5(b)) and from
71
Fig. 4.5: (a) Sketch of the experimental set-up. NS, nanospike; BR, Bottle resonator,
QPD, quadrant photodiode. (b, c) Optical micrograph of the nanospike coupled to a
WGM resonator with a diameter of 46 μm; (b) side-view and (c) top view.
the top (Fig. 4.5(c)). Because of the low transmission of the pump laser in the
vicinity of the cavity resonance, a secondary weak (≈ 500 μW) and non-resonant
probe laser was launched to image the mechanical movements of the nanospike tip
on a quadrant photodiode (QPD). This permitted two-dimensional reconstruction
of the motion of the nanospike with nm-scale spatial resolution.
4.3.2 Coupling light to bottle resonator using a nanospike and

measurement of the mode profile
Fig. 4.7(a) shows a microscope picture of a bottle resonator when laser light (wave-
length of 1150 nm and ≈ 100 μm optical power) launched through the nanospike
was locked to one of the optical resonances. We found that even the tiny losses
of these resonators were enough to directly image the mode profile on a sensitive
infra-red camera with a long exposure time. The mode in this particular case is
72
Fig. 4.6: (a) Photograph of the experimental setup used to observe optomechanical cooling
of tha nanospike. (b) Photograph of the inner part of the vacuum chamber.
73
characterized by 6 intensity maxima and is only visible on the right edge of the bot-
tle resonator. This happens because the nanospike, whose position is marked with
a dashed-dotted line, was placed behind the bottle resonator and therefore only on
this side of the resonator light coupled to the WGM travels towards the camera.
Fig. 4.7(b) depicts a comparison between the experimental measurement (zoom-in
of Fig. 4.7(a), replotted with a blue-to-red colormap) and the time averaged Poynt-
ing vector distribution of the corresponding mode calculated using finite element
methods, obtained after careful measurement of the bottle resonator diameter and
curvature.
The same procedure permitted to visualize profiles with 2, 3 and 4 lobes, which
are illustrated in Fig. 4.7(b) together with their simulated counterparts. For all of
them an optical Q-factor exceeding 107 was measured. We note that the fringes in
the experimental data are not due the presence of high orders intensity maxima in
the radial direction, but are artefacts due to the imaging system.
In general, light coupling requires the nanospike to be placed in close to one of

the intensity maxima, but the highest efficiency was reached when coupling through
one of the two brighter spots located at the edge of the mode pattern.
4.3.3 Dispersive and dissipative coupling
To confirm the presence of dissipative optomechanical interactions in the system,

we measured the frequency and the linewidth of a cavity resonance as a function of
nanospike distance for a bottle resonator with a diameter of 55 μm. As shown in
Fig. 4.8 (a), both the frequency and the linewidth of the cavity are found to depend
on the nanospike position and the experimental data fit well to an exponential func-
tion. By testing several resonators, we observed that the frequency shift can vary
substantially among different modes, occasionally reaching values so small that they
could be hardly measured. On the other hand the broadening of the linewidth did
not vary significantly and was always found to be dominant. Fig. 4.8(b) illustrates
the measured cavity resonances and the relative Lorentzian fits for three particular
values of the distance between the nanospike and the bottle resonator, which corre-
spond to the situation of under-coupling (645 nm), critical-coupling (414 nm) and
over-coupling (240 nm).
74
Fig. 4.7: (a) Micrograph of a bottle resonator when the laser light is locked into resonance.
At high power level, even the tiny optical leakages of this resonator were enough to image
the optical mode using a microscope objective. The unidirectional propagation of light in
the WGM makes the mode visible only on one side of the resonator. Indeed on the right
side, light propagates towards the imaging system. The horizontal dashed line indicates
the position of the nanospike. (b,c) Images of some of the optical mode experimentally
observed. Each mode is compared with its FEM simulated counterpart.
75
Fig. 4.8: (a) Measured cavity resonant frequency shift (right axis) and linewidth (left axis)
as a function of the nanospike-WGM resonator distance for a resonator with a diameter of
55 μm. The solid lines are exponential fits of the data. The light blue shaded area marks
the over-coupled regime. (b) Measured cavity resonances and relative Lorentzian fits for
nanospike-WGM distance of 645 nm (under-coupling), 414 nm (critical-coupling) and 240
nm (over-coupling)
4.3.4 Optomechanical cooling of the nanospike

When the nanospike was placed in the over-coupled regime with the pump wave-
length set very close to, but blue-detuned from, the cavity resonance, efficient op-
tomechanical cooling was observed. The laser detuning was stabilized using the
thermo-optical nonlinearity of glass. This technique, known as thermal self-locking
[142], permits very clean measurements without the need for electronic feedback to
stabilize the laser wavelength (as for instance in the Pound-Drever-Hall stabiliza-
tion scheme [143]), but at the same time it prevents measurements with red-detuned
light because of thermo-optical instabilities. The measured mechanical spectra in
the vicinity of the fundamental nanospike resonance are depicted in Fig. 4.9(a) for
increasing values of pump power. A significant drop in the amplitude of the mechan-
ical resonance, accompanied by linewidth broadening, was observed. The effective
temperature (Tef f ) of this mechanical mode of the nanospike was estimated by in-
tegrating the area underneath the power spectra [20]. As shown in Fig. 4.9(b), an
increase of three orders of magnitude in the mechanical linewidth could be measured
at only 250 μW pump power, with a minimum Tef f value of 1.8 K. When the pump
power was increased above 250 μW, the cooling efficiency saturated, probably due
to an increase in the ambient temperature induced by residual absorption of the
pump light.
These measurements refer to vibrations of the nanospike orthogonal to the
76
surface of the bottle resonator. The presence of residual electrical charges on the
resonator surface and on the nanospike breaks the degeneracy of its mechanical
modes and causes a frequency shift of about 50 Hz for vibrations orthogonal and
parallel to the BR surface. We emphasise that Coulomb interactions are conservative
and cannot modify the effective temperature of the nanospike. On the other hand, it
gave us the chance to independently monitor the behaviour of the vibrations parallel
to the surface. Given the symmetry of the optical modes, we expect in this case a
weaker optomechanical coupling, nonetheless this degree of freedom could still be
cooled. The inset of Fig. 4.9(a) compares the power spectra at launched pump
powers of 0 μW and 250 μW, while the inset in Fig. 4.9(b) shows the measured
linewidth and effective temperature for the same mechanical mode. The minimum
effective temperature achieved was 68 K, which confirmed a reduced optomechanical
interaction for this degree of freedom.
Since the resonance frequencies of the nanospike mechanical modes are a few
order of magnitudes smaller than the cavity linewidth, the cooling should not differ
substantially for higher order mechanical modes [20]. Fig. 4.10 shows the measured
power spectra for first high order flexural mode supported by the nanospike for
several pump powers, from the same data set as presented in Fig. 4.9. At zero
pump power, this mode had a resonance frequency of 6.45 kHz and a Q-factor of
1843 (about two orders of magnitude lower than the fundamental mode). The clear
trend observed when increasing the power of the pump laser confirmed our intuition
and proves simultaneous cooling of multiple mechanical modes. Since the first high
order mechanical mode had a lower mechanical Q-factor, the minimum achievable
Tef f was 118 K (see Fig. 4.10). We point out that multimode cooling is very
difficult to achieve in the sideband-resolved regime, which requires exact matching
of the laser detuning and the mechanical resonance frequency.
4.3.5 Self-stabilization of the coupling with the WGM resonator

At room temperature, in the absence of optomechanical cooling, thermal motion
of the nanospike causes fluctuations of tens of nm in its distance from the WGM
resonator. This causes random fluctuations in the frequency and the linewidth of the
WGM, which can exceed several MHz at critical coupling (see Fig. 4.8(a)). For the
presented system, since the optomechanical cooling involves all principal degrees of
freedom of nanospike motion, substantial stabilization of its coupling to the WGM
resonator could be achieved. Fig. 4.11 plots the displacement of the nanospike (after
calibrating the response of the QPD) recorded over 100 ms for pump powers of 0
77
Fig. 4.9: (a) Measured mechanical power spectrum in the vicinity of the fundamental
(flexural) nanospike mode as a function of the pump power. Inset: measured power spec-
trum for vibration parallel to the surface (see text) with the pump power as a parameter.
The solid lines are Lorentzian fits. (b) Mechanical linewidth (left axis) and inferred effec-
tive temperature T eff (right axis) as a function of the pump power. The dashed lines are
guides for the eye. Inset: same measurement as in (b) but for vibration parallel to the
surface.
78
Fig. 4.10: Measured mechanical power spectrum in the vicinity of the first high order
(flexural) nanospike mode with the pump power as a parameter. The solid lines are
Lorentzian fits. Inset: linewidth (left axis) and effective temperature T eff (right axis) of
the same mechanical mode as a function of the pump power.
μW, 10 μW and 250 μW. The panel at the right-hand side compares histograms
for data collected over 100 s. The reduction of the thermal noise can be clearly
observed. At low power, the Brownian motion of the nanospike has a mean-square
displacement (MSD) of 530 nm2 , in agreement with estimates from the equipartition
theorem. When the pump power is increased to 250 μW, the value of the MSD drops
significantly to 37 nm2 (Fig. 4.11), which corresponds to a suppression factor of 11.6
dB.
To further explore the effect of stabilization, a second probe laser at a wave-

length of 1550 nm (probe 2 in Fig. 4.5) was introduced into the setup and its
wavelength was scanned across one of the cavity resonances with and without the
laser cooling. Fig. 4.12(a) and 4.12(b) shows 50 consecutive measurements of the
resonance for a Tef f of 300 K and 6.7 K, revealing a clear overall increase in the
system stability. In particular, the laser cooling significantly reduced the measured
standard deviation of the minimum transmission from σ300K = 0.0117 at room tem-
perature to σ6.7K = 0.0025 after cooling to 6.7 K (see Fig. 4.12(c) and Fig. 4.12(d)).
79
Fig. 4.11: Self-stabilized coupling to the WGM bottle resonator. (a) Temporal displace-
ment of the nanospike for different launching pump powers. Right panel: histogram plots
of the nanospike displacements. (b) Mean-squared displacements of the nanospike as a
function of pump power.
4.3.6 Experimental evidence of the features of the dissipative

cooling
To highlight the features of dissipative optomechanical interactions, the nanospike

mechanical linewidth and Tef f were measured as a function of laser detuning (see
Fig. 4.13(a)). In this measurement the nanospike was placed in the over-coupled
regime and the pump power was 40 μW. In sharp contrast to traditional disper-
sive coupling, both cooling and amplification were possible on the blue side of the
cavity resonance. The measured mechanical linewidths and Tef f agree reasonably
well with theoretical predictions (solid lines), whose derivation is discussed in the
next session. The shaded area in Fig. 4.13(a) corresponds to the predicted unsta-
ble regime of optomechanical heating. To observe more clearly the self-oscillation
regime, experiments were conducted at higher pump power (60 μW) in the under-
coupled regime, when the greater distance between nanospike and WGM resonator
permitted measurement of larger mechanical displacements. The result is shown in
Fig. 4.13(b). Over the 0-20 s time interval the laser detuning was slowly reduced
from 5.0 MHz to 3.5 MHz (upper axis) and thereafter kept constant. Strong expo-
nential amplification of the nanospike motion was observed. Fig. 4.13(c) shows a
zoom-in of Fig. 4.13(b), comparing the cases of thermally driven (upper panel) and
coherently amplified motion (lower panel).
80
Fig. 4.12: (a, b) 50 consecutive measurements of a cavity resonance observed with the
second probe laser (1550 nm) when Tef f equals 300 K (a) and 6.7 K (b). (c) Minimum
transmission recorded in (a) and (b) as a function of the measurement number; the blue
dots correspond to Tef f = 300 K, while the orange dots to Tef f = 6.7 K. The apparent
oscillations in the experimental data result from the short total acquisition time (1 s),
which was much smaller than the lifetime of the fundamental mechanical mode (≈ 30 s).
(d) Position of the nanospike collected over 1.5 ms for Tef f = 300 K (blue line) and Tef f
= 6.7 K (orange line).
81
Fig. 4.13: (a) Mechanical linewidth (left axis) and inferred effective temperature T eff as
a function of the pump laser detuning. The dots are data points, while the solid lines are
fits of the theoretical model (see Methods). The yellow shaded indicates the theoretically
predicted instability region. (b) Position of the nanospike (blue line) as a function of time
when the laser detuning is reduced from 5.0 MHz to 3.5 MHz. After 20 seconds (marked
by a dashed line) the laser detuning is fixed at 3.5 MHz. The orange line is an exponential
fit of the oscillation envelop. (c) Zoom-in of (b), where the nanospike is thermally driven
(upper panel) and coherently excited (lower panel). Note the increase in the scale of the
y axis between the two panels.
4.4 Theory of a dissipative and dispersive

optomechanical system
The dynamics of a harmonic resonator both dispersively and dissipatively coupled

to an optical cavity was for the first time described theoretically in Ref. [30] and
then analyzed in more detail in Refs. [31, 144, 145]. The Hamiltonian describing
such a system needs to account for both types of coupling and can be written as:
Ĥ = ωC â† â + Ωb̂† b̂ + Ĥγ + ĤΓ + Ĥint (4.9)
in which the first two terms describe the Hamiltonian of the optical cavity and
mechanical resonator, Ĥγ + ĤΓ express the damping and driving of the cavity
and the mechanical resonator respectively and the last term Ĥint is the interaction
Hamiltonian which describes the optomechanical coupling. In the regime of weak
coupling it is possible to understand the effects of the cavity backaction onto the
mechanical resonators without solving the equation of motion using the so-called
quantum noise approach, which only requires knowledge of the spectral density of
the force operator F̂ acting on the system [20]. A complete description of this
method is beyond the scope of this thesis and more details can be found in [20, 146].
If we consider small fluctuations of the optical field around the steady state solution
82
Theory of a dissipative and dispersive optomechanical system 4.4
(i.e. â = ā + δâ) the force operator assumes the following expression [30, 145]:
F̂ 1 dHint i
=− = −θ1 (ā∗ δâ) − γ1 √ ā∗ ξˆ − sin ∗ δâ + h.c. (4.10)
dx 2 γext
where h.c. indicates the hermitian conjugate, sin is the driving field entering the
cavity from the waveguide and ξˆ the correspondent vacuum noise, γext is the decay
rate of the cavity field into the waveguide and the two parameters θ1 and γ1 account
for the dispersive and dissipative parts of the optomechanical coupling respectively
and are defined by the relations:

∂ωC
θ(x) = ωL − ωC + x = ωL − (ωC + θ1 x) (4.11)
∂x
∂γext
x = γext,0 + γ1 x
γext (x) = γext,0 + (4.12)
∂x
Here θ is the laser detuning and γext is the decay rate in the external waveguide.
It is worth to mention that the second term in the force operator is directly pro-
portional to the dissipative coupling coefficient γ1 and it is absent in the case of
pure dispersive coupling. The spectral density of the force operator is defined as:
∞
SF F = < F̂ (t)F̂ (0) > eiωt dt and in this case it can be written as:
−∞
2
|ā| γext [γ1 (ω + 2θ) − 2θ1 γext ]2 + γi (γ1 θ − 2θ1 γext )2 + γ12 γ 2 /4
SF F =
2γext (θ + ω)2 + (γe xt + γi )2 /4
(4.13)
With this in hand, it is possible to calculate the optically induced shift in the
mechanical linewidth and mechanical resonance frequency as:
1
Γopt = [SF F (Ω) − SF F (−Ω)] (4.14)
2mef f Ω
∞
1 1 1 dω
δΩ = SF F (ω) − (4.15)
2mef f Ω Ω − ω Ω + ω 2π
−∞
The solid blue line in Fig. 4.13 is a fit of Eq. 4.14 with fitting parameters θ1 /2π =
50 kHz/nm and γ1 /2π = - 600 kHz/nm. We point out that this theoretical picture
could qualitatively reproduce all the features observed in the experiment, but failed
in quantitatively predicting the observed mechanical frequency shift, a matter that
will require further study.
83
4.4.1 Relation between optical damping and effective

temperature
The connection between the additional optical damping and the effective temper-
ature was first discussed in Ref. [147]. The temporal dynamics of a harmonic
oscillator under the action of the thermal Langevin force can be written as:
ẍ(t) + Γẋ(t) + Ω2 x(t) = Fth (t)/mef f (4.16)
where the Langevin force Fth (t) can be characterized through its correlator:
< F (t)th F (0)th >= 2mef f ΓkB T δ(t) (4.17)
If we now assume that the harmonic oscillator also feels the action of the optical
force Fopt , whose only effect is to change the mechanical damping of the system (i.e.
Fopt = −mef f Γopt ẋ(t)), Eq. 4.16 needs to be rewritten as:
ẍ(t) + (Γ + Γopt )ẋ(t) + Ω2 x(t) = Fth (t)/mef f (4.18)
where, of course, the value of the correlator of Fth still depends only on Γ and not
on Γopt . We underline that the cooling mechanism originates from the subtle nature
of the additional optical damping Γopt which increases the dissipation of the system
without introducing any further fluctuation. At this point, it is easy to understand
that the dynamics described by Eq. 4.18 is analogous to the one of a harmonic
oscillator with intrinsic damping equal to Γ + Γopt , but at a temperature Tef f given
by:
Γ
Tef f = T (4.19)
Γ + Γopt
This expression has been shown to provide very accurate description of many exper-
imental situations, including the results discussed in this Chapter (see Fig. 4.13),
however its validity is limited to the classical regime, in which the shot noise raising
from fluctuations in the laser field can be ignored [20].
Conclusions of this chapter

Glass-fibre nanospikes allowed for the first time observation of optomechanical pas-
sive cooling of an optical waveguide evanescently coupled to a WGM resonator, in
a regime very difficult to access using traditional techniques. Compared to active
84
Theory of a dissipative and dispersive optomechanical system 4.4
cooling schemes [114], this method does not require accurate knowledge of the posi-
tion and velocity of the mechanical resonator, which becomes critical in the regime
of high optomechanical damping and therefore it is independent from the detection
noise. In addition we showed that cooling of the nanospike motion is very effective
in stabilising the coupling to an optical cavity well beyond the limits set by thermo-
dynamics. Moreover, dissipative cooling reaches its maximum efficiency close to the
cavity resonance at high intracavity power, potentially allowing other experiments
to be performed at the same time. We underline that the approach here presented
is general and may be applied to any type of optomechanical system with a high
enough mechanical Q-factor, allowing efficient optical cooling of low-frequency me-
chanical oscillators.
85
CHAPTER 5
CONCLUSIONS AND OUTLOOK
Adiabatic guidance of light and low mechanical dissipation makes glass-fibre nanospikes
a rich and versatile platform for investigating optomechanical interactions.
The experiment described in Chapter 2 shows that a free-standing nanospike
provides a highly sensitive and convenient probe of optothermal and optomechanical
effects at low gas pressures, for instance, gas pumps based on the Knudsen effect.
The large evanescent field of subwavelength nanospikes is ideal to couple them
to other types of waveguides. Chapter 3 analyzed in detail the optomechanical in-
teraction in a novel configuration consisting of a nanospike inserted into the hollow
core of a photonic crystal fibre (PCF). The results presented in this thesis demon-
strate the remarkable properties of this new approach for efficient light coupling into
hollow core PCF, in which the presence of optical forces, that trap the nanospike
in the core centre, determine self-alignment and self-stabilization of the optical sys-
tem, while adiabatic propagation of light almost completely suppresses the Fresnel
back-reflection. We further showed that the technique is not far from routine labo-
ratory applications by scaling the entire system to a length of a few centimetres and
fabricating a device, which acts as a universal interface between solid and hollow
core fibres.
Evanescently coupling a nanospike to a bottle resonator permitted for the
first time the observation of passive optomechanical cooling of an optical waveguide
coupled to a whispering gallery mode (WGM) optical cavity. Being very far from
the sidebands-resolved regime, the experiment illustrated in Chapter 4 sets a new
benchmark on the lowest mechanical frequency that can be passively cooled and
operates in a regime difficult to access in traditional cavity optomechanics. Cooling
87
Ch. 5 CONCLUSIONS AND OUTLOOK
Fig. 5.1: Nanospike array fabricated on the endface of a microscructured fibre: (a) SEM
image of the end face of a three-core fibre after HF-etching; (b) zoom in collected at angle
of 30 degrees with respect to the fibre endface to highlight the three free-standing fibre
cores.
of the nanospike motion also resulted in a significant stabilization to its coupling

to the optical cavity, which we believe it is going to be highly beneficial for a wide
range of sensitive experiments based on WGM cavities.
In the following, possible future directions, which could further extend the
already numerous applications of glass-fibre nanospikes, are presented and discussed.
Optomechanical interaction in nanospike arrays
Rapid progress in the micro/nano fabrication techniques have brought the attention
of the scientific community towards the study of the collective dynamics of systems
consisting of many optically coupled mechanical oscillators. After the pioneering
theoretical proposal of Heinrich et al. [148] in 2011, the groups of H. Tang and
M. Lipson reported respectively on the synchronization of two and seven resonators
coupled only by the optical field [24, 149, 150]. Nevertheless newly exciting phe-
nomena such as self-organization and pattern formation in resonators arrays have
yet to be proved experimentally.
Our attempt to extend the ideas described in this thesis to a system consisting
of "multiple nanospikes" is depicted in Fig. 5.1. By HF-etching a silica multicore
fibre, we managed to fashion three free-standing nanospike on the fibre end face.
Since the distance between the nanospike is comparable to the optical wavelength,
optomechanical interaction should be possible through evanescent field coupling.
Since the multicore fibre was fabricated using the standard stack and draw technique,
the number of nanospikes can be arbitrarly increased in order to create a "forest" of
88
optomechanically coupled nanospikes. Recently, my colleague Z. Wang improved the
etching technique and demonstrated the fabrication of 7 nanospikes on the endface
of a soft-glass fibre [151].
Another possibility to optically couple several nanospike comes from the in-
trinsic "multi-port" nature of bottle resonators [118]. As shown in Chapter 4, the
optical mode of a bottle resonator consists of several intensity maxima, each of
which represent a possible coupling point between the cavity and a nanospike. As a
consequence, cavity-enhanced optomechanical coupling between different nanospikes
should be possible by simply coupling them to the same bottle resonator as to observe
collective phenomena like synchronization or simultaneous optomechanical cooling.
Multimode adiabatic coupling into HC-PCF using a nanospike
A null-coupler is a device fabricated by tapering together single-mode fibres with

dissimilar diameters [38]. If adiabatic propagation of light in ensured, launching
a laser into the fibre with the smaller diameter causes light to be coupled to the
high-order mode of the taper waist. Such a device, which can be fabricated using
our tapering rig, could be used to extend the idea of nanospike-based adiabatic
coupling into hollow core fibres to higher-order modes. The challenge for such an
experiment is that different modes are going to couple to the hollow-core mode
at different nanospike diameters, therefore a successful demonstration required a
dramatic increase of insertion length of the nanospike into the hollow core from
the 50 μm demonstrated in [28] to several hundreds of μm. Another possibility
would be to implement a technique, alternative to HF-etching, to fabricate even
sharper nanospikes. It is worth mentioning that currently very pure high-order mode
excitation requires side-coupling through the photonic crystal cladding, a technique
hardly exceeding an efficiency of 10−4 [152].
Enhancement of the cooling efficiency of a nanospike coupled to a WGM

resonator
In the experiment described in Chapter 4, we decided to exploit bottle resonators

because their fabrication technique, based on tapered fibres, was an easy exten-
sion of the methodology developed to fashion glass-fibre nanospikes. Even if our
resonators exhibit an excellent optical Q-factor (between 107 and 108 ), their per-
formance is still inferior to what has been achieved using other types of whispering
gallery mode resonators [117]. Increasing the optical Q-factor by a few orders of
magnitude should strongly benefit the cooling efficiency and performing the experi-
89
Ch. 5 CONCLUSIONS AND OUTLOOK
Fig. 5.2: Tapering a kagome PCF: (a) SEM picture of the original (untapered) core of a
kagome fibre; (b) same as (a) but after tapering of the fibre. The kagome-style cladding
is nicely preserved while the core diameter is reduced by a factor ≈ 4.
ment at cryogenic temperature might make it possible to reach phonon occupancies

close to the quantum ground state.
Tapering of microscructured fibres
The techniques developed to precisely control the fibre profile during the tapering
process have been so far only applied to standard single-mode fibre. It is intriguing
the possibility to extend these concepts to microstructured fibres. Fig. 5.2 shows
a SEM picture of the end face of (a) untapered and (b) tapered hollow core PCF
with a kagome cladding, fabricated using the tapering rig described in Chapter 1.
Any difference between the two pictures, apart from the size of the microstructure,
is hardly noticeable. Of course adiabatic following of light can be engineered in
a similar way as for single-mode fibres, in order to achieve low loss, single-mode
guidance along the tapered hollow core. As a possible application of such fibres we
can look back to the experiment described in Chapter 3, in which a nanospike is
optomechanically trapped at the core centre of a PCF. Adiabatically tapering the
PCF just before the coupling point should determine stronger optical forces and
more efficient coupling, while at the same time using nanospike with larger final tip
diameter, which are shorter and more mechanically robust.
Tapered hollow core fibres might also find applications in other fields of physics
like non-linear optics and ultrafast photonics, as the scaling of the microstructure
and the consequent reduction of the core-wall thickness extends the fibre guidance
90
into the ultraviolet region [92].
Sensing with free standing nanospikes
Glass-fibre nanospikes demonstrated an exceptional sensitivity to the surrounding

environment. If, on the one hand, the optical properties of the nanospike and its pe-
culiar shape seem to be made for near field measurements, it is also relatively simple
to couple mechanical degrees of freedom to other physical quantities of interest. For
instance Wu et al. demonstrated torque magnometry by coating a photonic crystal
cavity with a thin layer of permalloy [61], effectively using a silicon-made microcav-
ity to measure magnetic fields. Considering the noteworthy mechanical properties
of the nanospike, the application of similar concepts could open new opportunities
for a novel types of fibre-based sensors.
91
BIBLIOGRAPHY
[1] J. C. Brandt and R. D. Chapman, Introduction to comets (Cambridge Uni-

versity Press, 2004).
[2] G. Fracastoro, Homocentrica (Venezia, 1538).
[3] P. Apian, Astronomicum Caesareum (Ingolstadt, 1540).
[4] J. C. Maxwell, “A treatise on electricity and magnetism. volume 1,” Clarendon,

Oxford (1873).
[5] A. Ashkin, “Acceleration and trapping of particles by radiation pressure,”

Physical Review Letters 24, 156 (1970).
[6] A. Ashkin, J. M. Dziedzic, J. Bjorkholm, and S. Chu, “Observation of a single-

beam gradient force optical trap for dielectric particles,” Optics Letters 11,
288–290 (1986).
[7] M. L. Povinelli, M. Lončar, M. Ibanescu et al., “Evanescent-wave bonding

between optical waveguides,” Optics Letters 30, 3042–3044 (2005).
[8] M. Li, W. Pernice, and H. Tang, “Tunable bipolar optical interactions between
guided lightwaves,” Nature Photonics 3, 464 (2009).
[9] W. Pernice, M. Li, and H. Tang, “A mechanical kerr effect in deformable

photonic media,” Applied Physics Letters 95, 123507 (2009).
[10] A. Butsch, M. Kang, T. Euser et al., “Optomechanical nonlinearity in dual-

nanoweb structure suspended inside capillary fiber,” Physical Review Letters
109, 183904 (2012).
93
Bibliography
[11] J. Koehler, R. Noskov, A. Sukhorukov et al., “Resolving the mystery of

milliwatt-threshold opto-mechanical self-oscillation in dual-nanoweb fiber,”
APL Photonics 1, 056101 (2016).
[12] J. Millen, T. Deesuwan, P. Barker, and J. Anders, “Nanoscale tempera-

ture measurements using non-equilibrium brownian dynamics of a levitated
nanosphere,” Nature Nanotechnology 9, 425 (2014).
[13] A. Rahman, A. Frangeskou, M. Kim et al., “Burning and graphitization of op-

tically levitated nanodiamonds in vacuum,” Scientific Reports 6, 21633 (2016).
[14] M. A. Taylor, M. Waleed, A. B. Stilgoe, H. Rubinsztein-Dunlop, and W. P.

Bowen, “Enhanced optical trapping via structured scattering,” Nature Pho-
tonics 9, 669–673 (2015).
[15] M. Pelton, M. Liu, H. Y. Kim et al., “Optical trapping and alignment of single
gold nanorods by using plasmon resonances,” Optics Letters 31, 2075–2077
(2006).
[16] M. L. Juan, M. Righini, and R. Quidant, “Plasmon nano-optical tweezers,”

Nature Photonics 5, 349 (2011).
[17] M. L. Juan, R. Gordon, Y. Pang, F. Eftekhari, and R. Quidant, “Self-induced

back-action optical trapping of dielectric nanoparticles,” Nature Physics 5,
915 (2009).
[18] M. G. Scullion, Y. Arita, T. F. Krauss, and K. Dholakia, “Enhancement of

optical forces using slow light in a photonic crystal waveguide,” Optica 2,
816–821 (2015).
[19] A. Dorsel, J. D. McCullen, P. Meystre, E. Vignes, and H. Walther, “Optical

bistability and mirror confinement induced by radiation pressure,” Physical
Review Letters 51, 1550 (1983).
[20] M. Aspelmeyer, T. J. Kippenberg, and F. Marquardt, “Cavity optomechan-

ics,” Reviews of Modern Physics 86, 1391 (2014).
[21] M. Eichenfield, J. Chan, R. M. Camacho, K. J. Vahala, and O. Painter, “Op-

tomechanical crystals,” Nature 462, 78 (2009).
94
Bibliography
[22] T. Kippenberg, H. Rokhsari, T. Carmon, A. Scherer, and K. Vahala, “Analysis

of radiation-pressure induced mechanical oscillation of an optical microcavity,”
[23] T. Carmon, H. Rokhsari, L. Yang, T. J. Kippenberg, and K. J. Vahala, “Tem-

poral behavior of radiation-pressure-induced vibrations of an optical micro-
cavity phonon mode,” Physical Review Letters 94, 223902 (2005).
[24] M. Zhang, G. S. Wiederhecker, S. Manipatruni et al., “Synchronization of

micromechanical oscillators using light,” Physical Review Letters 109, 233906
(2012).
[25] J. Teufel, T. Donner, D. Li et al., “Sideband cooling of micromechanical motion

to the quantum ground state,” Nature 475, 359 (2011).
[26] J. Chan, T. M. Alegre, A. H. Safavi-Naeini et al., “Laser cooling of a nanome-

chanical oscillator into its quantum ground state,” Nature 478, 89 (2011).
[27] R. Pennetta, S. Xie, and P. St.J. Russell, “Tapered glass-fiber microspike:

High-q flexural wave resonator and optically driven knudsen pump,” Physical
Review Letters 117, 273901 (2016).
[28] S. Xie, R. Pennetta, and P. St.J. Russell, “Self-alignment of glass fiber

nanospike by optomechanical back-action in hollow-core photonic crystal
fiber,” Optica 3, 277–282 (2016).
[29] R. Pennetta, S. Xie, F. Lenahan et al., “Fresnel-reflection-free self-aligning

nanospike interface between a step-index fiber and a hollow-core photonic-
crystal-fiber gas cell,” Physical Review Applied 8, 014014 (2017).
[30] F. Elste, S. Girvin, and A. Clerk, “Quantum noise interference and backac-
tion cooling in cavity nanomechanics,” Physical Review Letters 102, 207209
(2009).
[31] M. Li, W. H. Pernice, and H. X. Tang, “Reactive cavity optical force on

microdisk-coupled nanomechanical beam waveguides,” Physical Review Let-
ters 103, 223901 (2009).
[32] R. Pennetta, S. Xie, R. Zeltner, and P. St.J. Russell, “Optomechanical cooling

of a glass-fibre nanospike evanescently coupled to a whispering-gallery-mode
bottle resonator,” arXiv preprint arXiv:1804.09115 (2018).
95
Bibliography
[33] J. Love, W. Henry, W. Stewart et al., “Tapered single-mode fibres and devices.
part 1: Adiabaticity criteria,” IEE Proceedings J (Optoelectronics) 138, 343–
354 (1991).
[34] L. Tong, J. Lou, and E. Mazur, “Single-mode guiding properties of

subwavelength-diameter silica and silicon wire waveguides,” Optics Express
12, 1025–1035 (2004).
[35] R. Nagai and T. Aoki, “Ultra-low-loss tapered optical fibers with minimal
lengths,” Optics Express 22, 28427–28436 (2014).
[36] A. W. Snyder and J. Love, Optical waveguide theory (Springer Science & Busi-
ness Media, 2012).
[37] K. Harrington, S. Yerolatsitis, D. Van Ras, D. Haynes, and T. Birks, “End-

lessly adiabatic fiber with a logarithmic refractive index distribution,” Optica
4, 1526–1533 (2017).
[38] T. A. Birks, S. G. Farwell, P. St.J. Russell, and C. N. Pannell, “Four-port

fiber frequency shifter with a null taper coupler,” Optics Letters 19, 1964–
1966 (1994).
[39] T. A. Birks, I. Gris-Sánchez, S. Yerolatsitis, S. Leon-Saval, and R. R. Thomson,

“The photonic lantern,” Advances in Optics and Photonics 7, 107–167 (2015).
[40] K. Okamoto, Fundamentals of optical waveguides (Academic press, 2010).
[41] K. Karapetyan, “Single optical microfibre-based modal interferometer,” PhD

Thesis, Rheinischen Friedrich-Wilhelms-Universitaet Bonn (2012).
[42] M. Monerie, “Propagation in doubly clad single-mode fibers,” IEEE Transac-

tions on Microwave Theory and Techniques 30, 381–388 (1982).
[43] T. Erdogan, “Cladding-mode resonances in short-and long-period fiber grating

filters,” JOSA A 14, 1760–1773 (1997).
[44] T. Erdogan, “Cladding-mode resonances in short-and long-period fiber grating

filters: errata,” JOSA A 17, 2113–2113 (2000).
[45] S. S. Rao and F. F. Yap, Mechanical vibrations, vol. 4 (Prentice Hall Upper
Saddle River, 2011).
96
Bibliography
[46] S. M. Han, H. Benaroya, and T. Wei, “Dynamics of transversely vibrating

beams using four engineering theories,” Journal of Sound and Vibration 225,
935–988 (1999).
[47] G. D. Cole, I. Wilson-Rae, K. Werbach, M. R. Vanner, and M. Aspelmeyer,

“Phonon-tunnelling dissipation in mechanical resonators,” Nature Communi-
cations 2, 231 (2011).
[48] D. B. Fraser, “Acoustic loss of vitreous silica at elevated temperatures,” Jour-

nal of Applied Physics 41, 6–11 (1970).
[49] W. J. Startin, M. A. Beilby, and P. R. Saulson, “Mechanical quality factors

of fused silica resonators,” Review of Scientific Instruments 69, 3681–3689
(1998).
[50] A. Ageev, B. C. Palmer, A. De Felice, S. D. Penn, and P. R. Saulson, “Very

high quality factor measured in annealed fused silica,” Classical and Quantum
Gravity 21, 3887 (2004).
[51] T. A. Birks and Y. W. Li, “The shape of fiber tapers,” Journal of Lightwave
Technology 10, 432–438 (1992).
[52] G. Kakarantzas, T. E. Dimmick, T. A. Birks, R. L. Roux, and P. S. J. Russell,

“Miniature all-fiber devices based on co2 laser microstructuring of tapered
fibers,” Opt. Lett. 26, 1137–1139 (2001).
[53] T. E. Dimmick, G. Kakarantzas, T. A. Birks, and P. St.J. Russel, “Carbon

dioxide laser fabrication of fused-fiber couplers and tapers,” Applied Optics
38, 6845–6848 (1999).
[54] F. Orucevic, V. Lefèvre-Seguin, and J. Hare, “Transmittance and near-field

characterization of sub-wavelength tapered optical fibers,” Optics Express 15,
13624–13629 (2007).
[55] J. E. Hoffman, F. K. Fatemi, G. Beadie, S. L. Rolston, and L. A. Orozco,

“Rayleigh scattering in an optical nanofiber as a probe of higher-order mode
propagation,” Optica 2, 416–423 (2015).
[56] T. G. Tiecke, K. P. Nayak, J. D. Thompson et al., “Efficient fiber-optical

interface for nanophotonic devices,” Optica 2, 70–75 (2015).
97
Bibliography
[57] P. R. Saulson, “Thermal noise in mechanical experiments,” Physical Review

D 42, 2437 (1990).
[58] M. Vanner, J. Hofer, G. Cole, and M. Aspelmeyer, “Cooling-by-measurement

and mechanical state tomography via pulsed optomechanics,” Nature commu-
nications 4, 2295 (2013).
[59] G. Ranjit, D. P. Atherton, J. H. Stutz, M. Cunningham, and A. A. Geraci,

“Attonewton force detection using microspheres in a dual-beam optical trap
in high vacuum,” Physical Review A 91, 051805 (2015).
[60] D. Bykov, O. Schmidt, T. Euser, and P. St.J. Russell, “Flying particle sensors
in hollow-core photonic crystal fibre,” Nature Photonics 9, 461 (2015).
[61] M. Wu, N. L.-Y. Wu, T. Firdous et al., “Nanocavity optomechanical torque

magnetometry and radiofrequency susceptometry,” Nature Nanotechnology
12, 127 (2017).
[62] R. Zeltner, R. Pennetta, S. Xie, and P. St.J. Russell, “Flying particle micro-
laser and temperature sensor in hollow-core photonic crystal fiber,” Optics
Letters 43, 1479–1482 (2018).
[63] M. Christen, “Air and gas damping of quartz tuning forks,” Sensors and Ac-
tuators 4, 555–564 (1983).
[64] K. Kokubun, M. Hirata, M. Ono, H. Murakami, and Y. Toda, “Frequency

dependence of a quartz oscillator on gas pressure,” Journal of Vacuum Science
& Technology A: Vacuum, Surfaces, and Films 3, 2184–2187 (1985).
[65] S. Spinner, “Elastic moduli of glasses at elevated temperatures by a dynamic

method,” Journal of the American Ceramic Society 39, 113–118 (1956).
[66] E. J. Davis and G. Schweiger, The airborne microparticle: its physics, chem-
istry, optics, and transport phenomena (Springer Science & Business Media,
2012).
[67] A. Trowbridge, “Thermal conductivity of air at low pressures,” Physical Re-

view 2, 58 (1913).
[68] E. Lifschitz and L. Pitajewski, “Theoretical physics vol. 10, landau and lifc-
shitz,” (1983).
98
Bibliography
[69] D. Hutchins, M. Harper, and R. Felder, “Slip correction measurements for solid
spherical particles by modulated dynamic light scattering,” Aerosol Science
and Technology 22, 202–218 (1995).
[70] B. Annis, “Thermal creep in gases,” The Journal of Chemical Physics 57,
2898–2905 (1972).
[71] M. Knudsen, “Die gesetze der molekularströmung und der inneren rei-
bungsströmung der gase durch röhren,” Annalen der Physik 333, 75–130
(1909).
[72] S. Vargo, E. Muntz, G. Shiflett, and W. Tang, “Knudsen compressor as a

micro-and macroscale vacuum pump without moving parts or fluids,” Journal
of Vacuum Science & Technology A: Vacuum, Surfaces, and Films 17, 2308–
2313 (1999).
[73] O. Schmidt, M. Garbos, T. Euser, and P. St.J. Russell, “Reconfigurable op-

tothermal microparticle trap in air-filled hollow-core photonic crystal fiber,”
[74] J. C. Maxwell, “Vii. on stresses in rarified gases arising from inequalities of

temperature,” Philosophical Transactions of the royal society of London 170,
231–256 (1879).
[75] D. A. Lockerby, J. M. Reese, D. R. Emerson, and R. W. Barber, “Velocity

boundary condition at solid walls in rarefied gas calculations,” Physical Review
E 70, 017303 (2004).
[76] A. Passian, R. Warmack, T. Ferrell, and T. Thundat, “Thermal transpiration

at the microscale: a crookes cantilever,” Physical Review Letters 90, 124503
(2003).
[77] T. Zhu and W. Ye, “Origin of knudsen forces on heated microbeams,” Physical
Review E 82, 036308 (2010).
[78] O. M. Maragò, P. H. Jones, P. G. Gucciardi, G. Volpe, and A. C. Ferrari, “Op-

tical trapping and manipulation of nanostructures,” Nature Nanotechnology
8, 807 (2013).
[79] D. Phillips, M. Padgett, S. Hanna et al., “Shape-induced force fields in optical

trapping,” Nature Photonics 8, 400 (2014).
99
Bibliography
[80] N. Descharmes, U. P. Dharanipathy, Z. Diao, M. Tonin, and R. Houdré, “Ob-

servation of backaction and self-induced trapping in a planar hollow photonic
crystal cavity,” Physical Review Letters 110, 123601 (2013).
[81] J. D. Jackson, Classical electrodynamics (Wiley, New York, NY, 1999), 3rd
ed.
[82] I. Brevik, “Experiments in phenomenological electrodynamics and the electro-

magnetic energy-momentum tensor,” Physics Reports 52, 133–201 (1979).
[83] S. G. Johnson, M. Ibanescu, M. A. Skorobogatiy et al., “Perturbation theory

for maxwell’s equations with shifting material boundaries,” Physical Review
E 65, 066611 (2002).
[84] W. Qiu, P. T. Rakich, H. Shin et al., “Stimulated brillouin scattering in

nanoscale silicon step-index waveguides: a general framework of selection rules
and calculating sbs gain,” Optics Express 21, 31402–31419 (2013).
[85] M. L. Povinelli, S. G. Johnson, M. Lončar et al., “High-q enhancement of at-

tractive and repulsive optical forces between coupled whispering-gallery-mode
resonators,” Optics Express 13, 8286–8295 (2005).
[86] J. R. Rodrigues and V. R. Almeida, “Optical forces through the effective

refractive index,” Optics Letters 42, 4371–4374 (2017).
[87] P. T. Rakich, M. A. Popović, and Z. Wang, “General treatment of optical forces

and potentials in mechanically variable photonic systems,” Optics Express 17,
18116–18135 (2009).
[88] J. Gieseler, B. Deutsch, R. Quidant, and L. Novotny, “Subkelvin parametric

feedback cooling of a laser-trapped nanoparticle,” Physical Review Letters
109, 103603 (2012).
[89] P. St.J. Russell, “Photonic crystal fibers,” Science 299, 358–362 (2003).
[90] P. St.J. Russell, “Photonic-crystal fibers,” Journal of Lightwave Technology

24, 4729–4749 (2006).
[91] F. Benabid, J. C. Knight, G. Antonopoulos, and P. St.J. Russell, “Stimulated

raman scattering in hydrogen-filled hollow-core photonic crystal fiber,” Science
298, 399–402 (2002).
100
Bibliography
[92] P. St.J. Russell, P. Hölzer, W. Chang, A. Abdolvand, and J. Travers, “Hollow-

core photonic crystal fibres for gas-based nonlinear optics,” Nature Photonics
8, 278 (2014).
[93] M. Sprague, P. Michelberger, T. Champion et al., “Broadband single-photon-

level memory in a hollow-core photonic crystal fibre,” Nature Photonics 8, 287
(2014).
[94] G. Epple, K. Kleinbach, T. Euser et al., “Rydberg atoms in hollow-core pho-

tonic crystal fibres,” Nature Communications 5, 4132 (2014).
[95] M. Bajcsy, S. Hofferberth, V. Balic et al., “Efficient all-optical switching using

slow light within a hollow fiber,” Physical Review Letters 102, 203902 (2009).
[96] P. Londero, V. Venkataraman, A. R. Bhagwat, A. D. Slepkov, and A. L. Gaeta,

“Ultralow-power four-wave mixing with rb in a hollow-core photonic band-gap
fiber,” Physical Review Letters 103, 043602 (2009).
[97] T. Ritari, J. Tuominen, H. Ludvigsen et al., “Gas sensing using air-guiding

photonic bandgap fibers,” Optics Express 12, 4080–4087 (2004).
[98] F. Benabid, F. Couny, J. Knight, T. Birks, and P. St.J. Russell, “Compact,

stable and efficient all-fibre gas cells using hollow-core photonic crystal fibres,”
Nature 434, 488 (2005).
[99] L. Xiao, M. Demokan, W. Jin, Y. Wang, and C.-L. Zhao, “Fusion splicing
photonic crystal fibers and conventional single-mode fibers: microhole collapse
effect,” Journal of Lightwave Technology 25, 3563–3574 (2007).
[100] J. Travers, A. Rulkov, B. Cumberland, S. Popov, and J. Taylor, “Visible

supercontinuum generation in photonic crystal fibers with a 400w continuous
wave fiber laser,” Optics Express 16, 14435–14447 (2008).
[101] F. Gebert, M. Frosz, T. Weiss et al., “Damage-free single-mode transmission

of deep-uv light in hollow-core pcf,” Optics Express 22, 15388–15396 (2014).
[102] R. Zeltner, S. Xie, R. Pennetta, and P. St.J. Russell, “Broadband, lensless, and
optomechanically stabilized coupling into microfluidic hollow-core photonic
crystal fiber using glass nanospike,” ACS Photonics 4, 378–383 (2017).
101
Bibliography
[103] A. M. Cubillas, S. Unterkofler, T. G. Euser et al., “Photonic crystal fibres for

chemical sensing and photochemistry,” Chemical Society Reviews 42, 8629–
8648 (2013).
[104] S. Bauerschmidt, D. Novoa, A. Abdolvand, and P. St.J. Russell, “Broadband-

tunable lp 01 mode frequency shifting by raman coherence waves in a h 2-filled
hollow-core photonic crystal fiber,” Optica 2, 536–539 (2015).
[105] M. Mridha, D. Novoa, S. Bauerschmidt, A. Abdolvand, and P. St.J. Russell,

“Generation of a vacuum ultraviolet to visible raman frequency comb in h
2-filled kagomé photonic crystal fiber,” Optics Letters 41, 2811–2814 (2016).
[106] R. H. Stolen, C. Lee, and R. Jain, “Development of the stimulated raman

spectrum in single-mode silica fibers,” JOSA B 1, 652–657 (1984).
[107] J. C. Knight, G. Cheung, F. Jacques, and T. Birks, “Phase-matched excitation

of whispering-gallery-mode resonances by a fiber taper,” Optics Letters 22,
1129–1131 (1997).
[108] C. L. Mueller, M. A. Arain, G. Ciani et al., “The advanced ligo input optics,”
Review of Scientific Instruments 87, 014502 (2016).
[109] K. W. Murch, K. L. Moore, S. Gupta, and D. M. Stamper-Kurn, “Observation

of quantum-measurement backaction with an ultracold atomic gas,” Nature
Physics 4, 561 (2008).
[110] X. Wang, S. Vinjanampathy, F. W. Strauch, and K. Jacobs, “Ultraefficient

cooling of resonators: Beating sideband cooling with quantum control,” Phys-
ical Review Letters 107, 177204 (2011).
[111] S. Machnes, J. Cerrillo, M. Aspelmeyer et al., “Pulsed laser cooling for cavity
optomechanical resonators,” Physical Review Letters 108, 153601 (2012).
[112] T. Ojanen and K. Børkje, “Ground-state cooling of mechanical motion in the

unresolved sideband regime by use of optomechanically induced transparency,”
Physical Review A 90, 013824 (2014).
[113] A. Sawadsky, H. Kaufer, R. M. Nia et al., “Observation of generalized optome-

chanical coupling and cooling on cavity resonance,” Physical Review Letters
114, 043601 (2015).
102
Bibliography
[114] Y. L. Li, J. Millen, and P. Barker, “Simultaneous cooling of coupled mechani-

cal oscillators using whispering gallery mode resonances,” Optics Express 24,
1392–1401 (2016).
[115] R. Madugani, Y. Yang, J. M. Ward, V. H. Le, and S. Nic Chormaic, “Optome-

chanical transduction and characterization of a silica microsphere pendulum
via evanescent light,” Applied Physics Letters 106, 241101 (2015).
[116] J. Huang, Y. Li, L. Chin et al., “A dissipative self-sustained optomechanical

resonator on a silicon chip,” Applied Physics Letters 112, 051104 (2018).
[117] G. Lin, A. Coillet, and Y. K. Chembo, “Nonlinear photonics with high-q

whispering-gallery-mode resonators,” Advances in Optics and Photonics 9,
828–890 (2017).
[118] M. Scheucher, A. Hilico, E. Will, J. Volz, and A. Rauschenbeutel, “Quantum

optical circulator controlled by a single chirally coupled atom,” Science (2016).
[119] M. Eichenfield, C. P. Michael, R. Perahia, and O. Painter, “Actuation of micro-

optomechanical systems via cavity-enhanced optical dipole forces,” Nature
Photonics 1, 416 (2007).
[120] A. Schliesser, P. Del’Haye, N. Nooshi, K. Vahala, and T. Kippenberg, “Radi-

ation pressure cooling of a micromechanical oscillator using dynamical back-
action,” Physical Review Letters 97, 243905 (2006).
[121] A. G. Krause, M. Winger, T. D. Blasius, Q. Lin, and O. Painter, “A high-

resolution microchip optomechanical accelerometer,” Nature Photonics 6, 768
(2012).
[122] M. R. Foreman, J. D. Swaim, and F. Vollmer, “Whispering gallery mode

sensors,” Advances in optics and photonics 7, 168–240 (2015).
[123] L. Rayleigh, “Cxii. the problem of the whispering gallery,” The London, Edin-
burgh, and Dublin Philosophical Magazine and Journal of Science 20, 1001–
1004 (1910).
[124] G. Mie, “Beitraege zur optik trueber medien, speziell kolloidaler metalloesun-
gen,” Annalen der Physik 330, 377–445 (1908).
103
Bibliography
[125] V. Braginsky, M. Gorodetsky, and V. Ilchenko, “Quality-factor and nonlinear

properties of optical whispering-gallery modes,” Physics Letters A 137, 393–
397 (1989).
[126] M. Sumetsky, Y. Dulashko, and R. Windeler, “Optical microbubble res-

onator,” Optics Letters 35, 898–900 (2010).
[127] S. Berneschi, D. Farnesi, F. Cosi et al., “High q silica microbubble resonators

fabricated by arc discharge,” Optics Letters 36, 3521–3523 (2011).
[128] M. Sumetsky, “Whispering-gallery-bottle microcavities: the three-dimensional

etalon,” Optics Letters 29, 8–10 (2004).
[129] M. Pöllinger, D. O’Shea, F. Warken, and A. Rauschenbeutel, “Ultrahigh-q tun-

able whispering-gallery-mode microresonator,” Physical Review Letters 103,
053901 (2009).
[130] D. Armani, T. Kippenberg, S. Spillane, and K. Vahala, “Ultra-high-q toroid

microcavity on a chip,” Nature 421, 925 (2003).
[131] H. Lee, T. Chen, J. Li et al., “Chemically etched ultrahigh-q wedge-resonator

on a silicon chip,” Nature Photonics 6, 369 (2012).
[132] A. R. Johnson, Y. Okawachi, J. S. Levy et al., “Chip-based frequency combs

with sub-100 ghz repetition rates,” Optics Letters 37, 875–877 (2012).
[133] C. Michael, M. Borselli, T. Johnson, C. Chrystal, and O. Painter, “An opti-

cal fiber-taper probe for wafer-scale microphotonic device characterization,”
Optics Express 15, 4745–4752 (2007).
[134] S. Spillane, T. Kippenberg, O. Painter, and K. Vahala, “Ideality in a fiber-

taper-coupled microresonator system for application to cavity quantum elec-
trodynamics,” Physical Review Letters 91, 043902 (2003).
[135] O. Aktaş and M. Bayındır, “Tapered nanoscale chalcogenide fibers directly

drawn from bulk glasses as optical couplers for high-index resonators,” Applied
Optics 56, 385–390 (2017).
[136] T. Kippenberg, S. Spillane, and K. Vahala, “Demonstration of ultra-high-q

small mode volume toroid microcavities on a chip,” Applied Physics Letters
85, 6113–6115 (2004).
104
Bibliography
[137] D. O’Shea, C. Junge, M. Pöllinger, A. Vogler, and A. Rauschenbeutel, “All-

optical switching and strong coupling using tunable whispering-gallery-mode
microresonators,” Applied Physics B 105, 129 (2011).
[138] J. Buck and H. Kimble, “Optimal sizes of dielectric microspheres for cavity
qed with strong coupling,” Physical Review A 67, 033806 (2003).
[139] M. Aas, A. Jonáš, and A. Kiraz, “Lasing in optically manipulated, dye-doped

emulsion microdroplets,” Optics Communications 290, 183–187 (2013).
[140] Y. Louyer, D. Meschede, and A. Rauschenbeutel, “Tunable whispering-gallery-

mode resonators for cavity quantum electrodynamics,” Physical Review A 72,
031801 (2005).
[141] E. R. Abraham and E. A. Cornell, “Teflon feedthrough for coupling optical

fibers into ultrahigh vacuum systems,” Applied Optics 37, 1762–1763 (1998).
[142] T. Carmon, L. Yang, and K. J. Vahala, “Dynamical thermal behavior and

thermal self-stability of microcavities,” Optics Express 12, 4742–4750 (2004).
[143] E. D. Black, “An introduction to pound–drever–hall laser frequency stabiliza-

tion,” American Journal of Physics 69, 79–87 (2001).
[144] T. Weiss, C. Bruder, and A. Nunnenkamp, “Strong-coupling effects in dissi-

patively coupled optomechanical systems,” New journal of physics 15, 045017
(2013).
[145] T. Weiss and A. Nunnenkamp, “Quantum limit of laser cooling in dispersively

and dissipatively coupled optomechanical systems,” Physical Review A 88,
023850 (2013).
[146] A. A. Clerk, M. H. Devoret, S. M. Girvin, F. Marquardt, and R. J. Schoelkopf,

“Introduction to quantum noise, measurement, and amplification,” Reviews of
Modern Physics 82, 1155 (2010).
[147] P.-F. Cohadon, A. Heidmann, and M. Pinard, “Cooling of a mirror by radia-

tion pressure,” Physical Review Letters 83, 3174 (1999).
[148] G. Heinrich, M. Ludwig, J. Qian, B. Kubala, and F. Marquardt, “Collective

dynamics in optomechanical arrays,” Physical Review Letters 107, 043603
(2011).
105
Bibliography
[149] M. Bagheri, M. Poot, L. Fan, F. Marquardt, and H. X. Tang, “Photonic cavity

synchronization of nanomechanical oscillators,” Physical Review Letters 111,
213902 (2013).
[150] M. Zhang, S. Shah, J. Cardenas, and M. Lipson, “Synchronization and phase

noise reduction in micromechanical oscillator arrays coupled through light,”
[151] Z. Wang, S. Xie, X. Jiang et al., “Optomechanically coupled glass nanospike

array on the endface of a multicore fiber,” in “CLEO: Science and Innovations,”
(Optical Society of America, 2018), pp. SF3J–1.
[152] B. M. Trabold, D. Novoa, A. Abdolvand, and P. St.J. Russell, “Selective ex-

citation of higher order modes in hollow-core pcf via prism-coupling,” Optics
Letters 39, 3736–3739 (2014).
106
LIST OF PUBLICATIONS
Journal papers
1. R. Pennetta, S. Xie, R. Zeltner and P. St.J. Russell, "Optomechanical cooling
of a glass-fibre nanospike evanescently coupled to a whispering-gallery-mode
bottle resonator", arXiv preprint arXiv:1804.09115 (2018).
2. J. Hammer, A. Cavanna, R. Pennetta, M. V. Chekhova, P. St.J. Russell and

N. Y. Joly, "Dispersion tuning in sub-micron tapers for third-harmonic and
photon triplet generation," Opt. Lett. 43, 2320-2323 (2018).
3. R. Zeltner, R. Pennetta, S. Xie, and P. St.J. Russell, "Flying particle micro-

laser and temperature sensor in hollow-core photonic crystal fiber," Opt. Lett.
43, 1479-1482 (2018).
4. R. Pennetta, S. Xie, F. Lenahan, M. Mridha, D. Novoa and P. St.J. Russell.

"Fresnel-Reflection-Free Self-Aligning Nanospike Interface between a Step-Index
Fiber and a Hollow-Core Photonic-Crystal-Fiber Gas Cell" Physical Review
Applied 8, 014014 (2017).
5. R. Zeltner, S. Xie, R. Pennetta, and P. St.J. Russell, "Broadband, Lens-

less, and Optomechanically Stabilized Coupling into Microfluidic Hollow-Core
Photonic Crystal Fiber Using Glass Nanospike." ACS Photonics 4, 2 (2017).
6. R. Pennetta, S. Xie, and P. St.J. Russell. "Tapered Glass-Fiber Microspike:

High-Q Flexural Wave Resonator and Optically Driven Knudsen Pump." Phys-
ical Review Letters 117, 27 (2016).
107
List of publications
7. S. Xie, R. Pennetta, and P. St. J. Russell, "Self-alignment of glass fiber

nanospike by optomechanical back-action in hollow-core photonic crystal fiber,"
Optica 3, 277-282 (2016).
8. V. Spagnolo, P. Patimisco, R. Pennetta, A. Sampaolo, G. Scamarcio, M.S.

Vitiello and F.K. Tittel, "THz Quartz-enhanced photoacoustic sensor for H2 S
trace gas detection," Opt. Express 23, 7574-7582 (2015).
Conference contributions
1. R. Zeltner, S. Xie, R. Pennetta, and P. St.J. Russell, "Whispering-Gallery-
Mode Temperature Sensing with Flying Dye-Doped Particle in Hollow-Core
PCF", Optical Sensors, paper SeM4E. 3, Zurich, Switzerland (2018)
2. J. Hammer, R. Pennetta, P. St.J. Russell, and N. Y. Joly, "Pressure-Tunable

Third Harmonic Generation in Tapered Solid-Core Fiber", Conference on Lasers
and Electro-Optics, paper FW4E.7, San Jose (2018)
3. Z. Wang, S. Xie, X. Jiang, R. Pennetta, J. R. Koehler, and P. St.J. Russell,

"Optomechanically coupled glass nanospike array on the endface of a multicore
fiber", Conference on Lasers and Electro-Optics, paper SF3J.1, San Jose (2018)
4. R. Pennetta, S. Xie, F. Lenahan, M. K. Mridha, D. Novoa, and P. St.J.

Russell, "Fresnel-Reflection-Free Self-Aligning Nanospike Interface between a
Step-Index Fibre and a Hollow-Core Photonic-Crystal-Fibre Gas Cell", DPG
Spring meeting, paper Q 30.4, Erlangen (2018)
5. R. Pennetta, S. Xie, R. Zeltner, P. St.J. Russell, "Dissipative optomechanical

cooling of a glass-fiber nanospike coupled to a bottle resonator", International
Congress on Ultrasound, paper 6ICU 32(1), 045028, Honolulu, USA (2017)
6. R. Zeltner, S. Xie, R. Pennetta, and P. St.J. Russell, "Broadband Optome-

chanically Stabilized Coupling to Liquid-Filled Hollow-Core Fiber Using Silica
Nanospike", Optical Trapping Applications, paper OtW4D-2, San Diego, USA
(2017)
7. S. Xie, R. Pennetta, R. Noskov, P.St.J. Russell, "Optically driven self-oscillations

of a silica nanospike at low gas pressures." Optical Trapping and Optical Mi-
cromanipulation XIII, San Diego (2016)
108
8. R. Pennetta, S. Xie, R. Noskov, P.St.J. Russell, "Adiabatic silica microspike
with high mechanical Q-factor", Conference on Lasers and Electro-Optics, pa-
per JTh2A.107, San Jose (2016)
9. S. Xie, R. Pennetta, D.S. Bykov, T.G. Euser, P.S.J. Russell, "Optomechani-

cal self-stabilization and hysteresis of a free-standing silica nanospike inside a
hollow-core photonic crystal fibre", The European Conference on Lasers and
Electro-Optics, paper PD.A.9 (post-deadline paper), Munich (2015)
109
ACKNOWLEDGEMENTS
The last four years have been a hard time, not quite alike my early expectations. I
think learning how to deal with failure and frustrations is what PhD is really about.
Many of the obstacles on my way were just too tough for me and if I managed to
overcome some of them it’s just because of the tremendous help received from a
number of people, with whom I am deeply indebted.
I am grateful to my supervisors, which thought me how to be a researcher.
I thank all the students and postdocs of the group: the time spent with you in
scientific and "non-strictly scientific" activities was the most valuable of my PhD. I
am clearly not made for bureaucracy, therefore I would like to thank the secretary
and all the administration for making my life easier.
I thank all the amazing friends that in different way shared this long journey
with me, because without you my life would be vain as a Sunday evening.
A special thank goes to my family for the endless support they gave me every
single day of the last 27 years.
111

Tapered Glass-Fibre Nanospike Optomechanics

Hochgeladen von

Dokumentinformationen

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Tapered Glass-Fibre Nanospike Optomechanics

Hochgeladen von

Copyright:

Verfügbare Formate

Tapered Glass-Fibre Nanospike

Der Naturwissenschaftlichen Fakultät

Tag der mündlichen Prüfung: 19. Februar 2019

Vorsitzender des Promotionsorgans: Prof. Dr. Georg Kreimer

Gutachter: Prof. Philip St.J. Russell, D. Phil

This thesis investigates the optomechanical properties of glass-ﬁbre nanospikes and

Diese Thesis untersucht die optomechanischen Eigenschaften und verschiedenen An-

1 Tapered Glass-Fibre Nanospikes 7

2 Glass-ﬁbre nanospike as optically driven Knudsen pump 29

2.2.2 Competing between thermal creep and thermal edge ﬂow . . . 35

3 Optomechanical interaction of a nanospike inserted into an hollow core

4 Optomechanical cooling of a nanospike coupled to a whispering-gallery-

4.3.5 Self-stabilization of the coupling with the WGM resonator . . 77

5 Conclusions and outlook 87

List of publications 107

Structure of this Thesis

Chapter 1 examines the optical and mechanical properties of glass-ﬁbre nanospikes.

1.1 Optical properties of tapered ﬁbres

Fig. 1.1: Optical micrograph of a glass-ﬁbre nanospike.

1.1.1 Adiabaticity criteria

ρ(z)(βF M (z) − βHOM (z))

A physical interpretation of intermodal coupling in the transition regions can

1.1.2 Calculation of the propagation constants

1.1.3 Calculation of the maximum tapering angle for adiabatic

1.2 Mechanical properties of glass-ﬁbre nanospikes

1.2.1 Calculation of the resonance frequencies and relative

As usual, to identify the eigenmodes of the system we are interested in solutions

where zn = z/z d is a new dimensionless coordinate and q = 4ω 2 ρ/(Y R02 ) is a di-

1.2.2 Measurement of the mechanical resonances of a nanospike

To characterize the mechanical properties of the nanospikes, we used the experimen-

Measurements at ambient pressure

The very ﬁrst measurement of a nanospike mechanical power spectrum we performed

Measurements in high vacuum

To our great surprise, in this experiment we measured a mechanical linewidth

1.2.3 Understanding the high Q-factor

ρ0 (e−ζL − e−ζz ) + ρT (e−ζL − e−ζz )

in which at z = 0 (beginning of the taper) ρ = ρ0 (i.e. the untapered ﬁbre radius),

1.3 Fabrication techniques

where z0 is the instantaneous length of the transition region. Given an arbitrary

1.3.1 The tapering rig

1.3.2 Fabrication of the shortest adiabatic nanospike

Very valuable information can be obtained by measuring the transmission dur-

1.3.3 HF etching of silica nanospikes

1.4 Optical characterization of the taper proﬁle

Conclusion of this chapter

2.1 The experimental observation

Fig. 2.1 plots the measured resonance frequency fR of the nanospike as a

2.2 Simulation of the temperature distribution along

2.2.1 Thermal gradients and thermal creep ﬂow

2.2.2 Competing between thermal creep and thermal edge ﬂow

2.3 Interaction between the vibrating nanospike and

2.4 Temperature dependence of the resonance

2.5 Calculation of the total frequency shift

where f0 is the resonant frequency at room temperature in vacuum, Δfth results

2.6 Perturbation of the Knudsen ﬂow

Conclusions of this Chapter

3.1 Self-alignment of a nanospike by optomechanical

3.1.2 Calculation of the optical forces

Maxwell stress tensor approach

dpEM (r, t) dpM E (r, t)

where pEM = (E × H)/c2 is the electromagnetic momentum density, pM E is the

the force density can be expressed in a remarkably simple form:

An alternatively approach based on energy considerations was ﬁrst proposed by

dUopt (P, q) φ(q)