This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
3.1 Introduction............................................................................................................................... 36 3.1.1 Overview of the fibre with high nonlinearity ...................................................................... 36 3.1.2 The loss mechanisms within HNL-MOFs........................................................................... 37 3.1.3 Outline .............................................................................................................................. 39 3.2 Scattering losses ........................................................................................................................ 40 3.2.1 Introduction....................................................................................................................... 40 3.2.2 Scattering mechanisms....................................................................................................... 41 3.2.3 Cut-back loss measurement of HNL-MOFs........................................................................ 47 3.2.4 Back-scattering measurement for HNL-MOFs.................................................................... 53 3.2.5 Conclusions....................................................................................................................... 59 3.3 Reducing the OH induced losses in MOFs.................................................................................. 59 3.3.1 Mechanism........................................................................................................................ 60 3.3.2 Process considerations ....................................................................................................... 61 3.3.3 Experimental observations ................................................................................................. 66 3.3.4 Summary........................................................................................................................... 67 3.4 Conclusions............................................................................................................................... 68
Chapter.4 Small core rare-earth doped microstructured optical fibres
4.1 Introduction............................................................................................................................... 70 4.2 Fabrication of doped highly nonlinear microstructured fibres...................................................... 73 4.2.1 The ytterbium doped MOF................................................................................................. 73 4.2.2 The erbium doped MOF..................................................................................................... 75
4.3 Optical properties...................................................................................................................... 77 4.3.1 Dispersion ......................................................................................................................... 77 4.3.2 Effective mode area........................................................................................................... 78 4.3.3 Modal birefringence........................................................................................................... 79 4.4 A mode locked ytterbium doped MOF laser...............................................................................81 4.4.1 Experimental setup............................................................................................................81 4.4.2 Laser characteristics...........................................................................................................83 4.4.3 Discussion......................................................................................................................... 85 4.4.4 Summary........................................................................................................................... 87 4.5 A nonlinear amplifier based on a ytterbium doped MOF............................................................. 87 4.5.1 Experimental setup............................................................................................................ 87 4.5.2 Operating principles........................................................................................................... 88 4.5.3 Single Raman soliton generation – forward pumping configuration ....................................89 4.5.4 Backward pumping configuration ......................................................................................92 4.5.5 Multiple Raman soliton generation.....................................................................................92 4.5.6 Summary........................................................................................................................... 95 4.6 A low threshold, high efficiency erbium doped MOF laser......................................................... 95 4.6.1 Absorption characteristics.................................................................................................. 95 4.6.2 A high efficiency, low threshold laser based on erbium doped MOF...................................96 4.6.3 Summary.........................................................................................................................101 4.7 Conclusions.............................................................................................................................101
Chapter.5 Air-clad microstructured optical fibres
5.1 Introduction.............................................................................................................................104 5.1.1 The tunability of the Ytterbium doped cladding pumped fibre lasers (YDCPFLs)..............105 5.1.2 In-fibre gratings in MOFs................................................................................................108 5.1.3 Outline of this chapter......................................................................................................109 5.2 Design and fabrication............................................................................................................. 110 5.3 A cladding pumped ytterbium doped laser using air-clad MOFs............................................... 114 5.3.1 Properties of ytterbium doped air-clad MOFs................................................................... 114 5.3.2 Experimental setup..........................................................................................................116 5.3.3 A cladding pumped ytterbiund doped fibre laser with a wide tuning range........................117 5.3.4 A cladding pumped 980nm ytterbium doped fibre laser....................................................118 5.3.5 Summary......................................................................................................................... 120 5.4 Fibre Bragg grating in air-clad MOFs....................................................................................... 120 5.4.1 A GeO2-B2O3 co-doped air-clad MOF.............................................................................. 120 5.4.2 The effect of the air-silica interface.................................................................................. 121 5.4.3 Summary......................................................................................................................... 122 5.5 Conclusions............................................................................................................................. 123
Chapter.6 Large mode area microstructured optical fibres
6.1. Introduction............................................................................................................................. 124 6.2. Fabrication of large mode area microstructured optical fibres...................................................126 6.2.1. Refractive indices of silica based materials.......................................................................126 6.2.2. Evolution of the LMA-MOFs fabrication process.............................................................128 6.3. Effective mode area................................................................................................................. 134 6.3.1. Measurement via nonlinearity .......................................................................................... 135 6.3.2. Measurement via mode field diameter (MFD) ..................................................................138 6.4. Bend losses.............................................................................................................................. 141 6.4.1. Bend losses at 1550nm..................................................................................................... 142 6.4.2. Wavelength dependence of the bend losses...................................................................... 144 6.4.3. Discussion.......................................................................................................................146 6.5. Transmission losses................................................................................................................. 148 6.6. Conclusions............................................................................................................................. 150
Chapter.7 An ytterbium-doped all-glass double-clad large mode area microstructured optical fibre
7.1. Introduction............................................................................................................................. 152 7.2. Fabrication.............................................................................................................................. 155 7.2.1. Fabrication of the core..................................................................................................... 155
Chapter.8 Conclusions and future directions ................................................................... 176 Appendix.A Effective index model ................................................................................. 180 Bibliography................................................................................................................... 183 List of Publications......................................................................................................... 202
Acknowledgement
The last three years has passed so quickly just like travelling with light. I believe that this is
primarily because things have been moving very quickly in the field related to holey fibres.
Working in such an environment has been a little bit challenging, but stimulating and has also been
the greatest opportunity to learn a lot of things, from establishing methodology and time
management to practical and technical things.
To Prof.Dave Richardson, who gave this great opportunity to me, I would express my best
gratitude. This expression might sound like usual acknowledgement, but I am confident that I
would have been unable to make a better choice than doing my PhD at Southampton with him. I
would also like to thank Dr.Tanya Monro, from whom I learnt how important it is to be optimistic.
I have always tried to exercise my ingenuity in fabrication with positive thinking after interacting
with her. Dave and Tanya’s optimistic views have always provided something I could challenge
with fresh mind, possibly owing to their non-fabricator’s perspective.
I would also gratefully acknowledge many help from the people in the fabrication area, which was
indispensable for my activity i n the cleanroom. Dr.Jayanta Sahu and Paul Turner showed me their
gifted talents as fabricators. Without observing their activities, it would have been difficult for me
to make usable fibres. The successive head of the silica groups, Dr.Duncan Harwood and
Dr.Richard Williams who have always retained a better environment for silica fibre fabrication, are
also greatly appreciated.
Thanks are also to the (former and current) members of the advanced fibre technologies and
applications group, the high power fibre lasers group, and the novel fibres and waveguides group. I
would like to thank Dr.Neil Broderick, for his introduction to fibre experiments, Dr.Girberto
Brambilla for the grating experiments, and Jonathan Price for the nonlinear amplifier work,
Dr.Cyril Renaud and Romeo Selvas for the JAC fibres applications, Joanne Baggett for the work of
the large mode area fibres, and Vittoria Finazzi for the discussions of the optical properties of holey
fibres in general. The enthusiasm of many people at the ORC, who kindly dared to use/investigate
my more speculative creations, is also greatly appreciated.
I am deeply indebted to Prof.Alistair Fitt and Prof.Colin Please of the Department of Mathematics
for modelling of capillary drawing, which is described in the Chapter 2. I also grateful to Ping Hua,
Dr.Barbara Cressay, and Dr.Richard Pierce for passing their SEM skills to me, which I have used
every occasion throughout my study at the ORC. I would also like to acknowledge Dr.Eleanor
Tarbox, who kindly reading my thesis through. Her suggestions were also quite helpful.
I would like to thank those who encouraged me to go to Southampton: Dr.Marrku Oksanen, Dr.Ari
Tervonen, and Prof.Isao Endo. With their suggestions and encouragement, I decided to take this
opportunity in Southampton. I would also like to thank Prof.Minoru Obara. By working with him,
my interest in the field of optoelectronics has been triggered on.
Finally, I would like to thank to my family for their endless mental support. This thesis is dedicated
to them in token of my gratitude and heartfelt respect.
“Try to do your best, because that’s all part of the fun.” Hermann.A.Haus
DECLARATION OF AUTHORSHIP
I, Kentaro Furusawa, declare that the thesis entitled Development of rare-earth doped microstructured optical fibres and the work presented in it are my own. I confirm that:
� this work was done wholly or mainly while in candidature for a research degree at this University;
� where any part of this thesis has previously been submitted for a degree or any other
qualification at this University or any other institution, this has been clearly stated;
� where I have consulted the published work of others, this is always clearly attributed;
� where I have quoted from the work of others, the source is always given. With the exception of such quotations, this thesis is entirely my own work;
� I have acknowledged all main sources of help;
� where the thesis is based on work done by myself jointly with others, I have made clear exactly what was done by others and what I have contributed myself;
� parts of this work have been published as indicated in the list of publication. Signed: ……………………………………………………………………….. Date:…………………………………………………………………………….
Chapter.1
Introduction
1.1 Introduction
Fibre technology has already found widespread use in a variety of advanced applications in various
industries ranging from telecommunication to medical services. Since the early 70’s, the
continuous development of low loss optical fibres, birefringent fibres, photosensitive fibres, and
rare-earth doped fibres has offered many fibre based device applications providing novel, practical
and ever more sophisticated functions. Recently, a novel type of optical fibre has been introduced:
microstructured optical fibres (MOFs), which possess a transverse microstructure defined by air
holes running along their entire lengths. MOFs were initially named as photonic crystal fibres
(PCFs) and are also referred to as holey fibres (HFs).
By introducing the microstructure within the optical fibres, the guidance mechanisms can be
provided by the air holes, differing from the conventional optical fibres in which different materials
are used to provide an index boundary between the core and the cladding. When the
microstructures possess a solid core, the air holes serve as index decreasing elements leading to a
modified form of total internal reflection. On the other hand, by appropriately arranging the air
holes, it is also possible to guide light within a certain air hole by taking advantage of Bragg
scattering. Such MOFs are specifically referred to as photonic band gap fibres (PBGFs). This thesis
focuses upon the former type of MOFs.
It is interesting to note that both ideas of introducing the air holes within the fibres and of guiding
light using Bragg scattering can be found in the 1970’s[1,2]. However, due to the technical
difficulties in realising such structures and the lack of interesting optical properties discovered
within such structures, these directions had not been explored until recently.
A breakthrough was made by Knight et al., who demonstrated the first microstructured optical
fibres possessing unique optical properties, which cannot be obtained by conventional fibre designs
Chapter.1 Introduction 2
in 1996[3]. With strong demands on high performance optical fibres, the importance of MOFs has
become widely recognised during the last couple of years.
Fig.1.1.1 shows the rapidly increasing number of scientific publications related to microstructured
optical fibres. The exponential growth of the number indicates that MOFs have captured research
interests very rapidly and that a number of important technological breakthroughs are anticipated to
enable useful applications. The work presented in this thesis has been carried out during this
Fig. 1.1.1 The number of scientific publications related to MOFs during recent years.
Below, a brief review of MOFs in terms of properties and applications is followed by a summary of
the following chapters.
1.2 Properties of MOFs
1.2.1 Single mode operation
The first interesting property found in MOFs was ‘endlessly single mode’ guidance[3] in fibres
whose cladding consists of a hexagonally packed array of air holes. This cladding structure can be
characterised by the diameter d and the pitch Λ of the air holes, as shown in Fig.1.2.1. The MOF
cladding thus contains discrete and periodic changes of refractive index profile across the fibre
cross section, prohibiting simple waveguide analyses. Birks et al. introduced the effective index
model[4,5], where the complex air-silica cladding structure is regarded as a medium with an
effective refractive index neff and then the entire structure is approximated as a step index fibre, to
which an exact analytical solution exists. With this significant simplification, they demonstrated
that the effective V value (normalised frequency) 243556 7789 : ; ;<#= > = ?.@: ; ; AAB CDE can always be
Chapter.1 Introduction 3
below a certain value; an appropriately designed MOF can thus always be single-moded
(Veff<2.405) regardless of the operating wavelength λ.
Fig. 1.2.1 SEM of the core region of the index guiding MOF with an air hole diameter d and a hole to hole spacing Λ (These parameters range approximately from 1µm to 10µm.)
This interesting wavelength dependence of Veff originates from the fact that neff approaches nsilica as
the wavelength reduces. Although there used to be long lasting discussions about the definition of
the Veff value for MOFs, the fact that the cladding index displays a strong wavelength dependence
was confirmed by experimentally characterising the far-field diameter and the beam divergence[6].
This robust single mode operation uniquely possible using MOFs led to the first large mode area
(LMA) MOFs[7]. Chapter 6 and 7 of this thesis are devoted to exploring this direction.
Since waveguide properties are determined by the structural scale with respect to operating
wavelengths, terahertz pulse propagation within a polymer MOF has recently been reported[8].
MOFs based on new materials are currently being developed[8-11], good transparency in infrared
spectral range should be obtainable using these materials. An important fact here is that it is
possible to realise waveguides using a single material: rather than solving all the issues related to
material combinations to realise the waveguides. This flexibility may be very important for some
cases. Thus, the introduction of these new materials to MOFs offers an alternative route for
realising practical fibres in various wavelength ranges.
1.2.2 Group velocity dispersion (GVD)
The fact that the cladding refractive index is strongly wavelength dependent also leads to unusual
dispersion properties. While the effective index model provides some important insights to MOFs
(and more detailed methods for deriving the indices of the space filling modes have been
developed[12,13]), a number of numerical methods have been proposed for accurately calculating
the optical properties[14-31].
Chapter.1 Introduction 4
Thanks to the earlier development of the numerical models, some interesting features of the MOFs
were discovered. Mogilevstev et al. showed that MOFs can exhibit interesting dispersion properties
such as the zero dispersion wavelength (ZDW) below 1.28µm[32]. This is not obtainable using
conventional fibres with single mode operation, and the first demonstration of supercontinuum
generation took advantage of this feature combined with the high peak powers of the ultrashort
laser pulses from a Ti:sapphire laser[33]. In addition, interferometric dispersion characterisation
confirmed anomalous dispersion around 800nm in such fibres[34-36].
It has been shown that the MOFs with a very small core and large air holes can be modelled by
approximating the structure by a silica rod. Such fibres can exhibit a large normal dispersion at
1550nm[37], providing an alternative approach for the dispersion compensation in existing
telecommunication systems. Fibres with controlled dispersion characteristics around 1550nm have
been developed for the last couple of years[38]. Chapter 3 discusses the loss properties of such
MOFs.
These dispersion properties of MOFs imply that it is possible to flatten the dispersion profile over a
wider spectral range than possible using conventional fibres. Indeed, it has been shown that a very
small dispersion over a 400nm bandwidth covering both the 1.3 and 1.55µm telecommunication
bands is possible by optimising the hole size and spacing of the holey cladding[39]. Such a fibre
has recently been realised demonstrating ±1.2ps/nm/km over a 600nm spectral range from 1µm to
1.6µm[40].
1.2.3 Nonlinearity
Fibre nonlinearity can most conveniently be characterised by the effective mode area Aeff since the
effective nonlinearity per unit length � �� � ��� ��� �� is inversely proportional to the effective
mode area[41], where n2 and λ are the nonlinear index coefficient and the wavelength. Although
the intrinsic n2 of the silica based fibre materials is not high, the tight modal confinement
achievable within MOFs leads to high effective nonlinearity compared with conventional fibres.
This tight modal confinement is very sensitive to the fibre structure and therefore the waveguide
dispersion becomes strongly dependent on the operating wavelengths. In other words, the
interesting dispersion properties are often naturally accompanied by high nonlinearity, and which
has led to numerous applications described in Section 1.3.
Chapter.1 Introduction 5
In order to increase the effective nonlinearity per unit length, new glass materials have been
exploited[9]. Owing to high intrinsic nonlinearity of high index glasses[42], the effective
nonlinearity γ achieved in a MOF now exceeds 500W-1km-1 at 1550nm[43].
On the other hand, at the other extreme, very low nonlinearity can also be achieved by an
appropriate choice of the structural parameters owing to the inherently low nonlinearity of high
silica glass. As described, combined with the robust single mode operation, a novel opportunity for
the development of the large mode area fibres has been found and we have demonstrated γ as low
as 0.1W-1km-1 at 1550nm[44]. Thus, MOFs can play a role in extending both the high and low
nonlinearity extremities possible within optical fibres.
1.3 Applications
1.3.1 Nonlinear devices
The combined action of the high nonlinearity and the unusual dispersion properties of MOFs offers
new opportunities for developing useful nonlinear optical devices. The importance of the
nonlinearity was pointed out and characterised in ref.[45], where the effective mode area estimated
from the nonlinear phase shift data agreed well with the numerical simulations, assuming the
intrinsic nonlinearity n2~2.3x10-20 m2/W.
The high effective nonlinearity allows us to readily access various nonlinear optical effects and the
following nonlinear devices have been demonstrated by many groups around the world.
�
All optical switching[46-48] �
Raman amplification[49] �
Parametric devices[50-52] �
Soliton generation and squeezing[53-61] �
Supercontinuum (SC) generation[62-84]
Due to the possibilities of reduced power requirements and/or device lengths for these MOF-based
devices, the first three examples could be very attractive for signal processing for future optical
networks. Soliton effects that can be realised over a broad spectral range are also useful for
ultrashort pulse applications. In particular, utilisation of soliton self frequency shift (SSFS), with
moderate power requirements, allows us to access wavelength ranges where ultrashort pulses were
difficult to obtain previously. As a result, the demonstrated tuning range for fibre sources is now
from 800nm to 1.7µm[53-55,57]. Recently, soliton squeezing has been demonstrated using self
phase modulation within MOF[59-61]. Owing to the reduced soliton energy, this offers ease of
Chapter.1 Introduction 6
implementation in generating bright entangled states as required for many quantum optics
applications experiments.
The combination of an ultrashort pulse oscillator and a highly nonlinear MOF has become a
popular tool for generating broadband SC because of the modest power requirements. Under the
appropriate conditions, the spectrum broadens over more than an octave with well defined mode
spacing without loosing the coherence of the pump. By taking a beat of the octave frequency, the
caesium clock can then be extended directly to the optical domain by controlling the repetition rate
of the mode-locked oscillator[87-90] using a phase lock loop. This has enabled extremely accurate
spectroscopic measurements[91,92]. This strategy has also been used for stabilising the carrier
envelope of the ultrashort pulses[93-96] over a long period of time. This approach may pave the
way for accurate control over ultrafast phenomena in an attosecond regime[97].
Broadband SC sources have been used for optical coherence tomography (OCT) with a depth
resolution as short as 2µm around a 1.3µm wavelength[98] and ~0.75µm at a 0.72µm
wavelength[99]. The extremely broadband radiation available from this type of SC sources may
allow us to add spectroscopic diagnostics into OCT.
Driven by such a wide range of applications, a large body of work related to SC generation in
MOFs has been reported, including both theoretical[64-66,73,82-84] and experimental work[33,62-
84]. The reasonable agreement between the theory and the experiment indicates that now the SC
process can be qualitatively explained as a combined action of various nonlinear optical effects,
depending on the pump wavelength and the dispersion properties of MOFs.
It has become common knowledge that use of short pump pulses and a short section (~1cm) of
MOF is essential to obtain flatter spectra[85], in order to take advantage of the dominant effect of
self-phase-modulation, and is also important to retain a high degree of coherence[82,86]. Then,
these conditions automatically require the coincidence between the pump wavelength and ZDW of
the fibre. Thus, the design of the MOF with appropriate dispersion characteristics for a given pump
source is becoming very important.
1.3.2 Optical devices
By taking advantage of the interesting optical properties of MOFs, a variety of other optical devices
have also been reported, demonstrating unique opportunities as follows.
Fig. 2.3.2 ID and OD of the capillaries at different ud for various uf (2, 4 and 8 mm/min.). The dots correspond to the experimental data, while the curves are calculated by use of the analytic model. The experiment was performed at 1950 °C.
Table. 2.3.1 Fitting parameters used for the data, and comparison of the extrapolated characteristic temperatures with the published data in ref.[138].
O - P'() '+* ,�-�Q R S�T�UVO - / '() '+* ,�-�Q R S�T�UVW - O '() '+* ,�-�Q R S�T�UVW -&'() '+* ,�-�Q R S�T�UVO - P'() '+* ,�-�Q T�XY�VO - / '() '+* ,�-�Q T�XY�VW - O '() '+* ,�-�Q T�XY�VW -&'() '+* ,�-�Q T�XY�V
(a)
(b)
Chapter.2 Fabrication of microstructured optical fibres 22
2.3.5 Practical issues for capillary drawing
The above discussions were focused upon the control of the draw parameters. However, in practice,
other issues such as surface treatment and dimensional homogeneity along the length are also very
important for assembling a good quality preform. Here, the author discusses the choice of the
material first, and then processing that improves the quality.
Low qualit y grade silica tubes (HLQ-210) were often used for initial trials. However, the default
uniformity of the diameter per unit length is typically measured to be ± ~5%/m. Although it is
possible to adjust ud so that the outer diameter is always within 10µm, this may lead to a variation in
wall thickness. The yield of the capillaries with sufficient uniformity is ~60% after screening,
during which the inner and the outer diameters are characterised by the microscope and the
diameter gauge, respectively, and the capillaries that satisfy the acceptable range (<1%) are chosen.
Although callipers were initially used to measure O.D., it is recommended to use diameter gauges
since the head of the callipers may scratch the surface. Note that diameter gauges can now measure
I.D. and O.D. simultaneously and non-invasively. Note that the uniformity of the low quality silica
tubes can be improved by circularising the preform tube on a glass lathe by applying the optimum
pressure for a given temperature and by collapsing the tube slightly. In this case, the uniformity can
be imporoved to be ~1%/m for both HLQ-210 and Vycor® tubes, improving the total yield of the
capillaries.
Suprasil® F300 tubes possess better initial uniformity, where the diameter deviation is typ.
±~0.2%/m. However, due to the limited accuracy of the draw parameters, some transverse
deformation induced through the drawing process degrades the uniformity of capillaries to the level
of ~0.5%/m. Nevertheless, more than 80% of the drawn capillaries can still be used after screening.
Therefore, the yield can be improved by using high qualit y silica tubes.
Next, the surface quality is discussed as it directly affects the optical losses as shown in Chapter 3.
When the glass tubes are shipped, they contain some surface scratches (particularly in the case of
low quality tubes), and these are known to be the cause of imperfection or scattering losses for
conventional MCVD fibres[139]. These surface defects can be eliminated by fire-polishing tubes
followed by pre-baking. Although the capillary drawing process itself involves heating a glass tube,
the capillaries have to be pulled at reasonably low temperatures and using a high feed speed, in order
to preserve the geometry and to ensure the stability. This makes it difficult to improve the surface
quality during the drawing process. Thus, use of glass tubes with good surface quality is essential.
The pre-baking process greatly impacts the quality of the surface after fire-polishing. A glass tube
was heated up to ~1700°C on the MCVD lathe. The oxy-hydrogen burner carriage was traversed
approximately 5~8 passes with a speed of 150mm/min., depending on the surface condition of the
Chapter.2 Fabrication of microstructured optical fibres 23
tube and the oxygen flow inside the tube (~500cc/min.). If the temperature is too high (~1750°C), air
can be trapped within the scratch, forming bubbles that are diffcult to eliminate. If the temperature is
too low (1600°C), no changes occur to the surface since the visocity of the glass is too high. The
scratch dimensions vary depending on the tubes and their handling prior to the pre-baking process.
Therefore, a choice of an appropriate temperature is important. After prebaking the tube, the inner
surface of the tube can be etched off (~100µm at 1950°C) using SF6 and can then be fire-polished
(2050°C), through which process the presence of the defects/bubbles can readily be assessed.
Using appropriately prebaked tubes, no scattered light is observed at the bottom of the furnace when
the capillaries are drawn from the tower. However, the prebaking process incorporates a significant
amount of water into the glass due to the oxy-hydrogen burner heat source. Therefore, this process
may better be performed on the fibre drawing tower or using a dry heat source such as a plasma
torch[140] or a furnace. Recently, a technique that allows for mechanical polishing of silica tubes
(both outer and inner walls) to a high quality has been developed[141]. Such a low temperature
process is particularly attractive for obtaining good quality preform tubes since diffusion of the
water impurity is negligible in addition to the improved homogeneity.
After drawing capillaries, they are cut into suitable lengths for the preform assembly stage.
However, the cut end may possess sharp edges that can cause additional scratches when the
capillaries are stacked to form a preform. Therefore, the edges are trimmed by using a flame burner.
Although the impact of trimming is unknown, a lot of scatches as well as tiny silica particles were
observed within the assembled preforms without trimming. In contrast, they are perfectly
eliminated by the use of trimmed capillaries. Mechanical trimming is not recommended as it
generates micron-size silica powder that is difficult to remove due to static charges on a silica
surface.
2.3.6 Summary
The capillary drawing process has been discussed. The mathematical model for the process and the
derivation of its solution have been qualitatively described. It has been proven that it is difficult to
use numerical solvers based on the FEM method for analysing the practically important information
such as capillary dimensions with the desired accuracy. In contrast, the analytical solution provides
useful physical insights to qualitatively understand the process mechanism, which will be
demonstrated in the following sections. Finally, practical aspects of the capillary drawing process
have been presented. These form foundations for understanding the fabrication process and realising
a wide range of different MOFs.
Chapter.2 Fabrication of microstructured optical fibres 24
2.4 Preform fabrication
Preform fabrication involves preparing capillaries and stacking them into a jacket tube. In this
section, practical issues related to the preform preparation are described, in order to neatly prepare
the structure as it has a direct influence on the fibre performance. The cleaning of capillaries and the
jacket tube is first discussed, and preform assembling is then presented.
2.4.1 Cleaning
Cleaning the capillaries is important in order to avoid contaminants that can be incorporated into
MOFs. Due to the elevated temperature in the furnace, any residue within a preform can be burnt and
some elements can diffuse into the glass, which leads to extra losses. In reality, however, it is
difficult to clean the inner wall of the capillaries because of their small inner diameter (<<1mm).
Consequently, only the outer surface of the capillaries is cleaned using acetone and iso-propanol
(IPA). Acetone removes most contaminants, while IPA removes acetone and then naturally
evaporates. Note that the evaporation of alcohol leads to condensation of water from the atmosphere.
When the furnace itself is contaminated, a colouring of the capillary surface is observed, primarily
due to the carbon inside the furnace. In this case, diluted hydrofluoric acid (~5%) can be used to
etch off the capillary surface by ~1µm prior to the normal cleaning procedure. Note that
over-etching leads to degradation of the surface, which is clearly observable since it loses its
specularity[142]. Inorganic contaminants such as iron that are trapped on the silica surface can also
be removed by using this process[143]. This contamination occurs only when the furnace element
is about to be exhausted and can be prevented by replacing the furnace element and cleaning the
inside of the furnace.
Both acetone and IPA have to be perfectly removed by drying capillaries since they decompose into
a powder form of carbon at the high temperature in the furnace, which not only contaminates the
glass, but may also cause instabilities by blocking the air holes during the drawing. In order to dry the
capillaries after cleaning, dried gas flow from the MCVD system can be used to purge the
capillaries. Once the capillaries are dried, one end may be sealed using a burner. By heating
capillaries, scratches and contaminants within the capillaries are illuminated by the black-body
radiation, and which allows us to select clean capillaries. Care must be taken so that the diameter of
the sealed end is not greater than the capillary diameter since overheating leads to globule formation,
which prohibits us from neatly staking capillaries into a jacket tube.
The jacket tube can be cleaned in a similar manner using acetone and IPA. Pre-baking a tube not only
improves the surface quality but also helps to clean it. Although a tube has to be dismantled from the
Chapter.2 Fabrication of microstructured optical fibres 25
lathe after pre-baking, the jacket tube should be separated by using a burner rather than by cutting the
fused joint using a saw with a diamond coated blade. Cutting with a diamond coated dice generates
silica powders, which not only contaminate the surface, but also scratch the tube despite the use of
lubricant. Note that pre-baking the jacket tube is very important for obtaining a sufficiently strong
fibre, particularly when it is pulled at low temperatures.
If the preform is pulled into a cane for the two step drawing approach, the jacket tube can be etched
in diluted hydrofluoric acid prior to cleaning. This allows the elimination of adsorbed contaminants
from the surface, which are generally difficult to remove by using both acetone and IPA, as
described in relation to the capillary cleaning.
2.4.2 Preform assembly
General guidelines for the preform assembling are given here. The structural dimensions that are
related to the final fibre structure, which determines its optical properties, are excluded as they vary
substantially, depending on the target structures and the drawing strategies.
a) Choice of materials
In order to retain the air-silica structure during the fibre drawing process, there are two options in
terms of material combinations as follows.
�
Sealed silica capillaries + a silica jacket �
Unsealed silica capillaries + a Vycor® jacket
The key difference is whether to use sealed capillaries or not. For many MOFs, it is important to be
able to collapse interstitial holes and gaps between the jacket and capillaries, and this can be
accomplished by using sealed capillaries. Furthermore, this allows us to fabricate purely
single-material fibres (Fig.2.4.1 left). On the other hand, the interstitials can intentionally be left
open using a combination of unsealed silica capillaries and a Vycor® jacket (Fig.2.4.1 right). The
reason for this is explained below.
Eq.(2.8), discussed in Section 2.3, tells us that a tube with larger dimensions experiences less
collapse. Since this argument directly applies to the individual elements within a preform assembly,
it is clear that the capillaries are more collapsed than the jacket if both are made of the same material.
Thus, the structure cannot be maintained during the fibre drawing/caning because the capillaries
can freely move within the jacket tube. This issue can be overcome by pressurising the individual
capillaries so that the inner structure is less collapsed than the jacket tube, and which can be
Chapter.2 Fabrication of microstructured optical fibres 26
accomplished by sealing the capillary ends. Then, the inner structure can be fitted to the jacket,
maintaining the structure. Otherwise, it is impossible to pull a single material structure without
sealing the capillaries, which was experimentally confirmed.
Fig. 2.4.1 Two types of possible preforms in terms of material combinations. An all silica fibre made of sealed capillaries (left) and a fibre made of unsealed capillaries using a Vycor® jacket (right).
Vycor® tubes contain ~4% of boron oxide, and this results in significantly lower viscosity than that of
silica for a given temperature. This leads to significant collapse of a Vycor® jacket, despite the large
initial dimensions, allowing the jacket tube to be fitted to the unsealed silica capillary bundle. Note
that the mechanical force from the jacket is negligibly small due to its significantly lower viscosity.
With the combination of the silica capillaries with a Vycor® jacket, it is not necessary to seal
capillaries. Below, preforms made of this combination are referred to as a composite preform.
Depending on the possible material combinations, the draw conditions can vary significantly. In the
case of unsealed capillaries, the viscosity and the surface tension of the glass material primarily
determine the evolution of the structure during the draw.
On the other hand, the balance between the pressure within the air holes and the mechanical forces
within the glass are the primary factors for the resultant geometry in the case of the preform with
the sealed capillaries. The pressure within the capillaries is inversely proportional to the length of
dc
Vycor jacket
Λc Silica rod (core)
Silica jacket Sealed silica capillaries
Silica capillaries (Unsealed)
Silica rods
Chapter.2 Fabrication of microstructured optical fibres 27
the remaining preform, by ignoring the pressure decay inside the capillaries. Since most of the
pressures are cancelled out by the adjacent air holes, the net pressure imposed on the jacket tube is
not determined by the total volume of the air holes but by the dimensions of a single air hole.
Therefore, the pressure built up within the preform can well be compensated by either use of thick
enough jacket or by incorporating a large number of capillaries (>~50) to reduce the dimensions of
the air holes. On the other hand, when the air hole diameters of the capillaries within the preform
are small (<1mm), the pressure decay along the length of the capillaries can become significant due
to the temperature gradient along the preform length. This effect can be relaxed by sufficiently
warming up the preform before pulling it. Care must be taken when sealed silica capillaries are
used together with a Vycor® jacket tube. In this instance, the tube must be sufficiently thick.
Otherwise, the increasing pressure within capillaries can readily blow up the entire preform in this
case due to the small viscosity of the jacket tube.
b) Assembly
There are two important considerations for preform assembly. One consideration is the absolute
dimensions of the preform. The other issue is related to the practical care that must be taken during
the preform assembly.
The preform dimensions are determined by the capillary dimensions and the number of air holes
required within the structure. Although in general, the use of larger capillaries (~2mm) makes
assembly easier, there is a limit for the maximum possible preform dimensions because of either of
the following two reasons. One is the bore of the furnace, and the other is the thermal gradient
within the preform (the transverse temperature distribution) when it is within the furnace. Consider
that 2mm capillaries are stacked to form 8 rings of air holes. The I.D. of the jacket has to be 34mm.
Since the furnace bore used for the work in this thesis is 35mm, the jacket thickness must be as thin
as 0.5mm, which will cause a problem because of the possible expansion (or explosion) of the
preform due to the residual pressure when the sealed capillaries are used. The number of capillaries
can be increased by using smaller capillaries. However, it becomes increasingly difficult to
assemble them as their dimensions are reduced.
The preform dimensions are more severely restricted by the temperature gradient within the preform
in the case of single material preforms. Both large diameter preforms and ones that contain more air
holes are significantly affected by the temperature difference between the vicinity of the core and
that of the jacket tube, the latter of which is hotter and tends to be unstable. This is more apparent
when a vacuum is applied within the preform (see Section.2.5) and the resultant pressure is too low,
as the slight difference in viscosity leads to a significant transverse deformation through the
drawing. As a result, the air holes near the jacket tube collapse more than those near the centre.
Chapter.2 Fabrication of microstructured optical fibres 28
If a thermal gradient exists along the length of the preform, the deformation induced by this effect
can be recovered by continuously pulling over a long length until the temperature distribution
within the preform reaches a steady state. However, any deformations caused by the transverse
thermal gradient cannot be recovered. A large bore furnace would help solve this issue since the
magnitude of the thermal gradient is primarily determined by the dimensions of the furnace
element.
As the preform dimensions increase, it typically takes more time to stabilise the draw process due to
the increased glass volume. Therefore, it is always better to use relatively smaller preform
dimensions (<20mm) with a sufficiently long length (>20cm) in all cases. The smallest capillary
dimensions that have been used for stacking to date are ~500µm. This allows us to reduce the
preform dimensions significantly so that no problems associated with the thermal gradient are
present. However, as the number of the air holes increases, this issue may have to be studied more
fully along with the temperature dynamics.
Additional care that should be taken during the preform assembly includes minimising accidental
scratches and breaks, and aligning the capillary bundle without any twists. The former can mostly
be avoided by trimming the ends of the capillaries as described in Section 2.3. The capillary bundle
needs to be held at the top end, to which the sealed ends are aligned in the same position so that there
is no significant pressure difference between the capillaries during the draw. This can be done by
slightly collapsing (<1% in diameter) one end of the jacket tube and inserting the last two or three
capillaries from the top end. When the cladding region is shaped, additional rods are inserted near
the jacket tube (see Fig.2.4.1 left), and which can conveniently be used to fix the stack to the jacket.
These additional rods are also useful for preventing any unexpected collapse of air holes near the
jacket tube that may occur due to the thermal gradient.
c) drop
The drop is a silica rod attached on the bottom of the preform, as discussed. For MOFs, the
temperature required for the drop is typically higher than the draw temperatures. In the case of
preforms with sealed capillaries, it is possible that pressure will build up within the air holes due to
the heat, and so there is a possibility of blowing up the preform before starting to draw it. There are
many options for preventing this accident from happening and some examples are given below.
First, the initial neckdown should be on the preform (not at the drop itself) since the heat capacity of
the MOF preform is much smaller. This allows us to start drawing fibres at lower temperatures
(~-100°C). Second, a large enough drop should be used for the given preform dimensions to apply
sufficient tension to counteract the increased internal pressure. Third, the preform should be warmed
Chapter.2 Fabrication of microstructured optical fibres 29
up prior to pulling if the capillaries are sealed. Alternatively, a vacuum may be applied to the inside
of the preform to reduce the pressure, although this leads to substantially different draw conditions.
The internal pressure also depends sensitively on the water trapped inside the preform. Therefore, it
is imperative to warm up the preform prior to making a joint so that no water condensation occurs.
d) handle
In order to clamp the preform at the top of the tower, a handle needs to be attached to the preform.
The diameter of the handle tube should ideally be the same as that of the jacket tube, since it provides
a smooth transition in diameter. The argon flow within the furnace can greatly influence the stability
of the draw[144], so a smooth transition at the joint is important since it does not require any changes
of the iris aperture, thereby minimising the instability . The position of the joint should well be
separated from the capillaries (>3cm). Otherwise, if the jacket is overheated, the capillaries can be
badly deformed due to strong surface tension.
2.5 Caning and Fibre drawing
An appropriately prepared preform can be drawn into either a cane or a fibre on the fibre draw tower.
Depending upon the final dimensions of the structure in the fibre, either a one-step or a two-step
drawing approach is used. In the former case, the preform is directly drawn into a fibre. In the latter
case, the microstructured cane is drawn first from a preform, and is then pulled again by jacketing
with another tube. This approach is particularly effective for obtaining the small structures with Λ of
less than 2µm, the reason for which can be understood by considering the maximum draw ratio
required to achieve the small scale structures.
The smallest capillary dimensions, which can practically be stacked to form a preform, are limited
to ~500µm, as discussed in Section 2.4. Because of the initial capillary dimensions, a volume
reduction ratio η has to be ~10-7 to achieve Λ~1.5µm, for instance. However, as can be understood
from eq.(2.8), uf must be fast enough to preserve the structure. This leads to the requirement for
extremely fast draw speeds (ud >>100m/min.), which are not possible due to the restrictions on the
height of the fibre draw tower used for the work described in this thesis, as discussed in Section 2.2.
Hence, a two step approach has to be taken.
In this section, general guidelines for determining the draw parameters are presented. First, the
one–step fibre drawing process is discussed. Then, the two-step approach is discussed together with
the caning process.
Chapter.2 Fabrication of microstructured optical fibres 30
2.5.1 Single-step fibre drawing
Fibres with relatively large scale structures can directly be pulled from preforms on the fibre draw
tower. Comparing the capillary drawing and the one-step fibre drawing in terms of eq. (2.8); the
one-step fibre drawing involves fibres with Λ>~3µm, the volume reduction ratio η ranging from
10-4 to 10-5, compared with ~10-3 in the capillary drawing. Furthermore, the minimum structural
dimensions within the fibre preform (~dc) are also an order of magnitude smaller (100~1000µm)
than those of the glass tubes (~10mm) used in the capillary drawing. Given that uf is not
significantly different in these two processes, eq.(2.8) suggests that more than 10 times higher
viscosity is required in one-step drawing to obtain the same collapse ratio, and this corresponds to
~200°C lower temperature. Therefore, it is understood that the temperature for the one-step fibre
drawing has to be significantly lower than the temperature required for the capillary drawing.
The above argument is reasonably applicable for composite preforms as their dynamics are
primarily determined by the unsealed silica capillaries. This in turn gives a straightforward way of
determining the fibre draw parameter by substituting (Λ/Λc) for (φ/φc) in eq.(2.1). Indeed, without
sealing, it has experimentally been observed that the temperature has to be 150°C lower than the drop
temperature in order to retain the structure with a high fraction of air (Fig.2.4.1 right) using uf
~4mm/min., which is comparable to the speed used in the capillary drawing. Thus, the draw
parameters can be well predicted by considering the individual capillaries within the preform.
On the other hand, for a single material preform, the final fibre diameter sensitively depends upon
the balance between the pressure built up within the capillaries and the other mechanical forces due
to the viscosity and the surface tension of the glass, as discussed. Since the pressure depends on the
size of the air holes within the preform, the optimum draw parameters must be tuned depending on
the preform design by monitoring the fibre cross section.
In general, owing to the presence of the pressure, the fibre can be pulled at higher temperature than
that used for the composite preform. However, it is also true that the structure in the fibre is more
sensitive to both uf and the temperature. The draw parameters can also be estimated by considering
the ratio (Λ/Λc)2, as the same as the composite preform. Drawing at higher temperatures or with uf
that is too low (<2mm/min.) may result in instabilities, where collapse and expansion of the air holes
occurs irregularly along the length of the fibre because the viscosity of the glass is too low. In
practice, this can be observed as diameter fluctuations. On the other hand, when temperatures are
too low, a structure is somewhat blown up. Since the collapse ratio of the jacket is inherently
smaller than that of the capillaries, a reduced temperature allows its collapse ratio to be almost
unity. This allows the capillaries within the preform to expand to fill the interstitials. Further
reductions in the temperature can result in the onset of interstitials.
Chapter.2 Fabrication of microstructured optical fibres 31
As a result, the practically useful temperature range for stable fibre drawing can be very narrow for
single material preforms. For example, the optimum temperature range was as narrow as 10°C for
the preforms which contained 8 rings of air holes of dc/Λc~0.2 with a jacket thickness of ~3Λc. This
is because the small i nner diameter of the capillaries (~100µm) led to the onset of the instabilities
at lower temperature, whilst the interstitials open up at higher temperature. Thus, in general, when
either Λc or dc/Λc is small, the temperature range tends to be narrow.
To widen the stable drawing temperature range, either a thicker jacket whose ratio between I.D.
and O.D is more than dc/Λc can be used or a well controlled pressure, lower than atmospheric
pressure, can be applied inside the preform. However, in the former case, the fibre diameter change
becomes insensitive to the inner structure, making it difficult to observe the stability of the
structure along the length. In addition, when d/Λ is required to be <0.5, the fibre can become
unacceptably thick.
When a preform is pulled over a length of a couple of centimetres, the temperature may need to be
slightly adjusted (~10°C), so that the balance between the forces and the internal pressure within the
individual capillary is retained. When a lower pressure than the atmospheric pressure is applied
inside the preform, the pressure can alternatively be adjusted. Note that the required pressure
changes are slow compared with the other diameter fluctuation mechanisms. Therefore, in principle,
feed back control could be implemented to control the furnace temperature or the pressure inside the
preform, in order to obtain a uniform structure over a longer length.
2.5.2 Two-step drawing
The total volume reduction factor, a product of two steps as ηt =η1 η2, varies from 10-6 to 10-7 in the
two-step approach, where the subscript number corresponds to each step. The problem then
becomes how to allocate the volume reduction factor for each step that provides optimal structural
control in achieving the small scale structure in the final fibre.
As discussed, a smaller η permits a smaller collapse ratio, thereby better preserving the structure.
On the other hand, the collapse ratio is also inversely proportional to the initial structural
dimensions. Therefore, it is intuitive that η2 should be as small as possible since the default
structural scale involved in the second step is small. However, the maximum value of η is limited
to be ~10-4, assuming that the lower limit of uf is ~2mm/min. whilst the highest ud is ~60m/min. as
described. Therefore, η1 must be at least 10-2, which corresponds to an order of magnitude
reduction in scale at the first step. Because η2 is comparable to that of the single step approach, this
Chapter.2 Fabrication of microstructured optical fibres 32
suggests that an order of magnitude higher viscosity is required to retain the same collapse ratio at
the second stage. In practice, this means that the fibre has to be pulled more than 250°C below the
drop temperature, which is impractical because of the high tension that would be required. (The
fibres pulled with the two step approach used tension >120cN that is beyond the measurement
range of the available tension gauge. Therefore, it is impossible to specify the actual values here.)
This implies that it is necessary to seal the cane to preserve the structure, and it is difficult to
reliably pull the fibre with a small η2 value.
Fig. 2.5.1 SEM photographs of early two-step drawn fibres.
In order to cope with this issue, some techniques were developed. One is to use additional
capillaries for the fibre preform so that the cane is thermally isolated as shown in Fig.2.5.1 (a). This
method also allows us to reduce the fibre tension because of the small heat capacity due to the
reduced silica volume in the region surrounding the sealed cane. However, this is accompanied by
the reduced mechanical strength of the fibre, which makes use of this fibre type impractical. To
reinforce the fibre, a solid Vycor® jacket was also used (Fig.2.5.1 (b)). Although this approach also
greatly reduces the fibre tension while improving the mechanical strength, preparation of the jacket
is very difficult because of the poor qualit y of the starting tubes, whose circularity needs to be
improved on the lathe prior to stacking. Furthermore, a Vycor® tube is not a sufficiently transparent
glass to allow for the ultimate low loss fibres, given the diffusion length of OH ions during the
fibre drawing process, as described in Chapter 3. Single material fibres, ideally all silica fibres, are
desired.
In order to reduce the tension on the fibre during the second stage of the drawing process, there are
two options. One is to use a small preform (<<10mm), since the preform is more uniformly heated
for the given furnace bore. Another option is to reduce ud. These options can be applied by
decreasing η1 to ~10-3, or by correspondingly enlarging η2 to ~10-3. By drawing small canes, it is
possible to use ud as slow as ~10m/min. where the fibre tension is low enough to allow us to pull at
(a) (b)
Chapter.2 Fabrication of microstructured optical fibres 33
low temperatures. Although the initial dimensions at the second step are smaller than 100µm, and
correspondingly the temperature has to be reduced (~250°C below the drop temperature), it has
become possible to reliably draw truly single material MOFs with extremely small dimensions,
combined with the use of high qualit y (pre-baked) jacket tubes (see Fig.2.5.2).
Fig. 2.5.2 SEM photographs of all silica MOFs pulled using two-step approach.
Since the cane must be sealed to prevent the structure within the cane from collapsing, care must be
taken at the second step to achieve good structural control. The hole-to-hole spacing Λc within the
cane can be used as a guide to determine the draw parameters by comparing it to the target
parameters Λ.
The collapse ratio of the jacket is nearly zero at a low temperature. Thus, the cane must be tightly
fitted within the jacket. Otherwise, the sealed cane can expand during drawing while preserving the
glass volume according to eq.(2.1). This results in a larger fraction of air in the fibre than in the
original cane (Fig.2.5.2 (a)). However, because of the small cane dimensions (<1mm), it is difficult
to prepare a jacket tube so that it perfectly fits the cane. To overcome this problem, the fibre is
pulled by using a relatively loose jacket (more than ~300µm greater than the cane diameter), and
applying a lower pressure (~300mbar) inside the preform to intentionally collapse the jacket so as to
allow the jacket to fit onto the cane, as shown in Fig.2.5.3. Note that the accuracy of the pressure
gauge reading can be poor due to possible leaks and to the long length (typically >1m) of the
preform. The resultant structure is shown in Fig.2.5.2 (b).
Under this condition, the expansion of the cane structure is sensitively determined by the pressure
and the temperature for a given thickness of the jacket for a given value of the draw speed (i.e.
tension). In other words, the balance between these three factors determines the cladding structure.
The pressure, at which the jacket tube fits the cane, varies from 400 to 700mbar depending on the
thickness of the jacket tube. Naturally, use of thick jacket tubes requires a smaller pressure inside
the preform at a given temperature.
(a) (b)
Chapter.2 Fabrication of microstructured optical fibres 34
Fig. 2.5.3 Schematic of the pressure control inside the preform for the two-step approach.
Although the jacket fits the cane well by applying the low pressure, the cane structure can still be
expanded depending on the jacket thickness, particularly when the pressure is too low. The reason
may be attributed to the fact that the cane starts expanding before the jacket shrinks because of the
low pressure. Thus, there is an optimum pressure, where the jacket is sufficiently collapsed while
suppressing the expansion of the cane, and which can be found by accurately controlling the
regulator valve. From the above observations, the range of this optimum pressure is expected to be
very narrow, depending on the preform (the initial air hole diameter within the cane and the
thickness of the jacket tube). Fixing the draw speed is also important since a slight difference in
draw speed modifies the effect of the surface tension of the jacket tube. Therefore, the expansion of
the cladding structure is also sensitively affected by the draw speed or the applied tension (i.e.
when the tension is too high, the cladding structure tends to expand.)
A systematic study of the pressure control (and the effect of tension) is ongoing. The structures
fabricated so far have been controlled primarily by increasing the jacket thickness to suppress the
expansion (compare the fibre diameters in Fig.2.5.2 (a) and (b)) by fixing the draw speed. This has
allowed us to obtain similar d/Λ to dc/Λc (the difference is <10%) for structural ranges of
~0.2<d/Λ<~0.9 and Λ<2.5µm. By improving the pressure control, more accurate control over the
structural dimensions can be obtained without significant restrictions in fibre diameters (and thus
jacket thickness).
Pressure gauge
Argon supply
Vacuum pump
Microstructured cane (sealed & fused)
Jacket tube
Valve
(Bubblers) Regulator
valve
Chapter.2 Fabrication of microstructured optical fibres 35
2.5.3 Summary
General guidelines for determining the parameters for fibre drawing and caning have been presented.
It has been shown that the analytical model for capillary drawing can be applied for the individual
components within the preform, and can conveniently be used to consider the draw parameters.
Sealing capillaries or canes has allowed us to realise single material fibres, which provides more
ideal realisations of the unique characteristics of MOFs as demonstrated in Chapter 5. For the
two-step approach, it has been found that the key issue depends on relaxing the fibre tension at low
temperatures, and that precise pressure control is required for more accurately controlling the
cladding structure.
2.6 Conclusions
The fabrication technology of MOFs has been discussed and the steps taken to refine the process and
to realise a wide range of different structures of MOFs have been described.
First, a mathematical model for a capillary drawing process has been developed. Experiments have
been carried out to examine the validity of the model. It has been demonstrated that the model can
predict the draw parameters required to obtain the desired dimensions of capillaries for given
dimensions of a tube. Although a numerical approach was also taken to obtain solutions to the model,
it has been proven to be impractical primarily due to the intensive computational tasks involved.
Practical aspects of capillary drawing such as characteristics of raw materials and surface quality
improvement have also been discussed, where the importance of the use of high quality silica tubes
and of a pre-baking process has been emphasised.
Secondly, general guidelines for the preform preparation, caning, and fibre drawing have been
presented. It has been shown that physical insights obtained from the capillary drawing model can be
conveniently used for determining the draw parameters. The dynamics involved in the fibre
drawing were qualitatively discussed based upon the observations for both single and two-step
approaches. It has been shown that the single step approach possesses certain advantages, and that
the use of a high speed fibre drawing system will allow the fabrication of fibres with any structural
dimensions in a single step. Despite the difficulties involved in the two-step approach, it has been
demonstrated that, by optimising the draw parameters to relax the fibre tension and by improving the
quality of the jacket tube, it is possible to reliably fabricate MOFs with a wide range of structures.
Chapter.3
Transmission properties of
highly nonlinear microstructured optical fibres
3.1 Introduction
3.1.1 Overview of the fibre with high nonlinearity
Recent advances in wavelength division multiplexing (WDM) and time division multiplexing
(TDM) technologies have made it possible to transmit more than 1Tbit/sec through a single mode
optical fibre over several tens of kilometres or more[145,146]. The control of optical nonlinear
effects is one of the key enabling technologies for these ultrahigh bit-rate systems, allowing for
ultrafast signal processing, which is difficult to perform electronically, to be performed directly in
the optical domain.
In these systems, conventional fibre types have often been used as an optical nonlinear medium.
Although the nonlinearity of silica-based optical fibres is not particularly high relative to many
other materials, the long interaction lengths that can be achieved by virtue of the low transmission
losses of the fibres, combined with their potential low cost, makes fibres very attractive options as
nonlinear media.
However, due to the inherently low nonlinearity of the high silica fibres, kilometre scale lengths are
typically required to take advantage of the nonlinearity of these conventional fibres. Such a long
length requirement leads to several limitations such as accumulated chromatic dispersion, which
typically limits the useful bandwidth of such devices. For this reason, highly nonlinear dispersion
shifted fibres (HNL-DSFs) have been developed, which possess a high nonlinearity (γ~20W-1km-1)
combined with a small amount of dispersion (D<2ps/nm/km) by achieving high NA (~0.4) while
retaining compatibility with conventional fibres[147].
Chapter.3 Transmission properties of small core microstructured optical fibres 37
MOFs provide a simple route for realising highly nonlinear fibres relative to conventional HNL-
DSF, since it is possible to fabricate fibres with very small effective mode areas due to the high
index contrast between the silica and air[33], and we specifically refer to such highly nonlinear
MOFs as HNL-MOFs. In addition, high nonlinearity in MOFs may be accompanied by interesting
dispersion properties[32,37,38], which can potentially be tailored over wider spectral ranges than
conventional fibres. To date, it is commonly accepted that higher γ can be achieved within HNL-
MOFs than in HNL-DSFs and that the fibre based nonlinear device lengths can significantly be
shortened using HNL-MOFs. The highest nonlinearity theoretically achievable within pure silica
based HNL-MOFs is predicted to be γ~52W-1km-1 at 1550nm (i.e. Aeff~1.7µm2 assuming n2~2.2x10-
20m2/W)[148]. By using a highly germanium doped core, a further factor of two increase in
nonlinearity could be anticipated[41].
However, there are currently several serious drawbacks in silica based HNL-MOFs, compared with
HNL-DSFs: interface issues and losses. The losses are in particular a severe factor since the useful
fibre nonlinearity is determined by the product of the nonlinearity and the effective device length.
Apart from dispersion properties, which make the most impact for some applications, one possible
approach to evaluate the fibre for nonlinear device applications, in which the phase-matching
condition is automatically satisfied, is to define the figure of merit (FOM: γ/α) since the effective
length (Leff=(1-exp(-αL))/α) is ultimately limited by the attenuation length (i.e. α-1). HNL-DSFs
exhibit FOM=25~30dB-1W-1[147]. Given the current loss levels of HNL-MOFs (10~100dB/km),
their FOMs are <3dB-1W-1. Thus, if the loss level can be reduced by a factor of ten or more, and the
interface issues can be solved, then HNL-MOFs will be in the position to compete with or even
improve upon the performance of HNL-DSFs. Thus, it is of paramount importance to reduce the
current loss levels of HNL-MOFs. These parameters are summarised in Table.3.1.1.
Table. 3.1.1 A comparison of the fibre parameters between HNL-DSF and HNL-MOF (typical).
Thus, for given parameters d, Λ, λ, and the parameters for roughness (i.e. δf and Lc), it is possible to
roughly estimate the resulting losses due to the backward coupling. Note that this model only
accounts for coupling to the backward propagating fundamental mode. Also note that δf and Lc
must be characterised by some means, as discussed. By taking a derivative with respect to Lc, it is
readily understood that αb possesses a peak value at Lc=1/2β, which is of the order of 100nm.
Fig.3.2.2 shows the dependence on Λ for the structure with d/Λ=0.9. aeff is approximated to be EFG[23]. In Appendix A, a more reliable approach is discussed in order to obtain an adequate
value of aeff. The roughness parameters δf=5nm and Lc=100nm are tentatively used. Given the
observed roughness in planar waveguides[162], δf is most likely overestimated while Lc is most
Fig. 3.2.12 Measured back-scattered signal and predicted trends for different αRcl.
Chapter.3 Transmission properties of small core microstructured optical fibres 58
It was found that most of the data points fall between 1000~2000dBµm4/km. A comparison to the
value obtained from Fig.3.2.11 ( �� ���
�� ~5700dBµm4/km) indicates that the majority of the losses are
not reflective components: the backscattered power is weaker than anticipated from the attenuation
data. Therefore, the reason for the large Rayleigh scattering coefficient estimated from Fig.3.2.11 is
possibly due to the substantial forward scattering losses despite the high NA designs of the fibres.
Also, the reasonable quality of fitting in Fig.3.2.11 suggests that the forward scattering losses also
exhibit a strong dependence on the structural dimensions similar to the Rayleigh scattering losses.
It is difficult to predict the loss spectra of the forward scattering losses αf without any knowledge of
perturbation spectra. However, since the modes confined within small cores can result in deeper
modal penetration into the cladding in this regime, they possess greater modal overlap with
radiation modes, leading to stronger modal coupling to radiation modes. This suggests that the
strength of the modal coupling (and thus the forward scattering losses αf) is also strongly dependent
on the core dimensions and increases as the core dimensions are reduced. Therefore, the structural
dependence of the forward scattering losses can be similar to that of the Rayleigh scattering losses.
As discussed, the fibre with high d/Λ>0.9 has a cut-off about Λ~1.1µm and the fibres examined
here are therefore mostly single mode fibres. From Fig.3.2.1, the perturbation length scale is below
10µm since λ/Λ~1.5. Given this length scale, it is hard to consider any reasons for the perturbation
since fibres should be smooth over this length scale. One possible cause of the forward scattering
losses can be micro-fractures of the holey cladding. Even when their length scales are greater than
the beat length zc, a perturbation strength given by these micro-fractures can be so significant that
the forward scattering can still be induced by them. In fact, the author observed a loss increase after
re-winding. The fibres used for this experiment were all rewound whilst the fibres used in the
previous section (i.e. Fig.3.2.5) were immediately characterised soon after drawing and without
rewinding.
It is necessary to further improve the measurement accuracy in order to confirm the above
discussions. For instance, by characterising the loss distribution along the length, it should possible
to identify the approximate positions of any imperfections/micro-fractures that lead to radiation
losses. The necessary system improvements are discussed below. First, the coherence noise must be
suppressed. Although we tried to use time-gated ASE sources, it was difficult to obtain the short
pulse duration relevant for the available short lengths of the fibres in order to achieve high
resolution. Possible solutions to this problem would be to use a multimode laser, or to frequency
modulate the DFB lasers with RF frequency to generate sidebands.
Chapter.3 Transmission properties of small core microstructured optical fibres 59
Second, the high effective nonlinearity of the fibre also limits the fibre length since the peak power
can be reduced due to self phase modulation. To cope with this issue, use of a 95/5 coupler rather
than a 3dB coupler could help since they allow us to launch less pump power without reducing the
OTDR signal. Finally, the fibre alignment was so critical due to the extremely small core
dimensions that the averaging time was limited. To improve the quality of the OTDR traces, longer
time averaging (~104 frames in typical OTDR systems[169]) are necessary. Thus, a robust
motorised stage or fusion splicing will be required to improve these measurements further.
3.2.5 Conclusions
The scattering losses within HNL-MOFs were studied. The various possible scattering loss
mechanisms within HNL-MOFs were discussed. Cut-back measurements using a white light source
showed a significant increase in loss with reduced structural dimensions. By using a simple model,
it has been shown that the wavelength dependence of the Rayleigh scattering is modified due to the
significant change in modal confinement at long wavelengths and that λ-2 fitting can be used for a
limited range of spectrum whilst the conventional λ-4 fitting can be used at short wavelength.
Furthermore, it has been found that the Rayleigh scattering is sensitively affected by the wet
etching process used for some fibres.
Back scattering measurements were also performed for HNL-MOFs. It was found that the amount
of the backscattering is smaller than expected from the total losses of the fibres. This implies that
there are substantial radiation losses. At the same time, the losses due to backward scattering were
found to be relatively small, compared to the Rayleigh scattering.
Due to the issues related to the stability of the measurement combined with the large coherence
noise present in the OTDR traces, the measurement was limited in terms of its accuracy. Further
study is required to fully understand the loss mechanisms within HNL-MOFs. Other approaches
such as distributed Brillouin scattering measurements are ongoing using a commercial OTDR
system, in which the same measurement can also be performed. Along with more understanding of
the loss mechanisms, it is expected that important information for improving the fabrication
technology could be obtained.
3.3 Reducing the OH induced losses in MOFs
This section discusses the incorporation of water content that occurs throughout the fabrication
process. It is of paramount importance to reduce the OH induced losses in order to widen the usable
bandwidth around 1550nm, where the intrinsic losses of silica are the lowest[130]. The mechanism
of OH incorporation is first reviewed in Section 3.3.1. Then, possible improvements in the
Chapter.3 Transmission properties of small core microstructured optical fibres 60
fabrication process are discussed in Section 3.3.2. The experimental results from the process
modification are presented in Section 3.3.3. Finally, a summary is given.
3.3.1 Mechanism
Surface chemistry allows us to understand the mechanisms of OH incorporation at an air-silica
surface[170]. The evolution of the surface states at elevated temperatures is schematically shown in
Fig.3.3.1. At room temperature, the water content in the atmosphere can be physically adsorbed or
chemically bound at the surface. When the glass is heated to 150°C, physically adsorbed water
molecules are removed through evaporation. The chemically bonded Si-OH groups start to
dehydrate at about 400°C, as a result of pairing of neighbouring hydroxyl molecules. However,
there can still remain some isolated OH ions, and which start diffusing into the glass as the
temperature is increased.
Fig. 3.3.1 Surface characteristics of silica at different temperatures.
Therefore, simple heating alone does not allow for complete removal of hydroxyl ions from the
surface, and may indeed increase the OH content in the glass. Indeed, it has been reported that
consolidation of VAD preforms in dry atmosphere (without dehydration) results in 5~30ppm of
OH content since the soot contains some hydroxyl ions as a result of flame hydrolysis[132].
Similarly, the observed OH content for the author’s MOFs without dehydration was ~30ppm
despite the use of high quality synthetic silica tubes (i.e. F300/F320 see Table.6.2.1) that contain
only ~0.2ppm of OH content by default.
Next, the length scale over which OH ions can diffuse into the glass is considered. The diffusion
length of the OH species is defined as ������ �� , where D is the diffusion coefficient and T is
the time period over which heat is applied to the glass[140]. Thus, in the drawing process, the
diffusion length depends on the feed rate of the tube and the drawing temperature. The feed speed
range typically used for MOF fabrication is 2~10mm/min. and the hot zone length of the furnace is
~30mm. Therefore, T lies within the range of 180~900sec. Using D~10-7cm2/sec at 2000°C[140],
Si
OH
Si Si
O
H H
O H H
O
Si
OH
Si Si
O
Isolated OH ion
Si
OH
Si Si
O
H H
O H H
O
At room temperature 150� C >400� C
Chapter.3 Transmission properties of small core microstructured optical fibres 61
Ld is 42~94µm, the lower limit of which is comparable to the values reported for conventional
fibres[130].
In MOFs, the nearest air-silica interface from the core is located at a distance of ~Λ/2. This length
scale is much shorter than the diffusion length. This implies that the distribution of the OH ions
across the fibre core becomes almost uniform after drawing a preform. Therefore, all surfaces
within MOF preforms must be free from hydroxyl ions. Furthermore, holey claddings typically
span only ~20µm in HNL-MOFs where the two-step drawing approach is used (see Chapter 2).
Since the distance between the outer surface of the microstructured cane and core is of the order of
10µm, care must also be taken to eliminate the OH ions at the interface between the jacket tube and
the microstructured cane particularly when lower feed speeds and/or small preform dimensions are
used. Note that this value could be an underestimate, since the diffusion process can actually
commence far from the hot zone of the furnace.
Understanding the above mechanism suggests two approaches to eliminate hydroxyl ions: either to
dehydrate the glass prior to heating, or to eliminate an appropriate thickness of the glass surface
after heating. Note that the glasses are heated up to high temperatures (>1500°C) several times
throughout the entire MOF fabrication process. The examples of the heat treatment include pre-
baking, capillary drawing, sealing, caning, and fibre drawing. Thus, removal of the surface layer, in
which high proportion of OH ions are present, apparently complicates the fabrication process. It is
also apparent that the two-step approach (see Chapter 2) is more prone to suffer from hydroxyl
contaminations because of the small scale structures and the additional drawing process involved.
Below, possible improvements and precautions related to the fabrication process are described
following the categories used in Chapter 2 (capillary drawing, caning, and fibre drawing). The
author notes that it is difficult to examine in isolation the effects of the capillary preparation and
caning as their contamination cannot be directly assessed. After considering the process,
experimental observations for a range of fibres prepared by different procedures are described.
3.3.2 Process considerations
A) Capillary preparation
There are three principal possibilities for incorporating hydroxyl ions.
1) From the burner to the outer wall of a tube during the pre-baking process
2) From the top and bottom ends of the tube during the drawing
3) From the outer wall of the surface that is exposed to the furnace element.
Chapter.3 Transmission properties of small core microstructured optical fibres 62
The importance of the pre-baking process was discussed in Chapter 2. An oxygen-hydrogen burner
generates a high water content as a consequence of combustion. Due to the high temperature
(~1700°C, where D~10-8cm2/sec) used in the pre-baking process, diffusion also occurs, and the
resulting diffusion length is ~50µm, assuming the heating time to be 30 minutes. In addition, the
OH ions can be diffused further into the glass at the drawing process. As a consequence, more than
~10µm thickness of the outer wall will contain high water content for ~1mm capillaries after the
tube is drawn.
The removal of the surface layer of silica tubes can be performed either before or after the capillary
drawing by using hydrofluoric acid. There is a trade-off between the etching depth and the surface
quality in the etching process and it was described in Section 3.2 that over-etching of capillaries
results in increased scattering losses. Thus, the amount of etching that can be used without loss
penalties is actually limited for silica capillaries. On the other hand, the surface quality can be
somewhat recovered through the capillary drawing process by applying high temperatures and low
feed speeds although the collapse involved such parameters needs to be taken into account. Thus,
etching a glass tube before capillary drawing (but after the pre-baking process) seems a better
option than the capillary etching. However, due to the large amount of glass required for capillary
drawing, a large scale etching environment must be prepared, and which had not been available for
most of author’s PhD period.
There are a couple of possibilities for minimising the incorporation of water before drawing into
capillaries. Two examples are described below. First, the burner used in the pre-baking process can
be substituted by either a plasma torch[140] or a furnace. By incorporating these devices, the OH
incorporation can be reduced. The second possibilit y is to use mechanical polishing, instead of pre-
baking[142]. This allows us to eliminate a heat source for improving the surface qualit y of the glass
tube. Thus, the original low OH content of the glass tube is essentially preserved until it i s drawn
into capillaries.
In reality, it is difficult to dehydrate the outside of the tube since it cannot be isolated from the
atmosphere. Therefore, post-etching may be necessary for the outer surface[189]. For the inner
surface of the tube, dehydration can be performed although the pulling end of the capillary has to
be open unless sealing is performed while drawing capillaries. Primarily for this reason, the
dehydration of the tube has not yet been applied. To date, preform tubes have been purged with
argon during drawing followed by applying a vacuum as an alternative approach to minimise the
OH incorporation.
Chapter.3 Transmission properties of small core microstructured optical fibres 63
B) Caning
The drawn capillaries are exposed to the atmosphere and are typically sealed using a torch.
Therefore, the assembled preform must be dehydrated before caning. The dehydration process of
the assembled preform is described below.
There have been several different reagents that have been used for dehydrating silica
materials[130]. These include brominating, fluorinating, and chlorinating reagents. The chemical
reaction strength is proportional to the electronegativity (i.e. Br < Cl < F). However, the chemical
reaction of fluorine is so strong that it is accompanied by reactions with silicon atoms on the
surface, resulting in etching. Furthermore, fluorine is made available by heating up SF6 gas (SF6-
>SF4+F2), and this increases the local pressure due to the molecular decomposition, making it
difficult to apply to microstructured materials. Chlorine (Cl2) and thionyl chloride (SOCl2) have
been frequently used for conventional fibre fabrication. Electronics grade gas is available for Cl2,
while an ordinary bubbling purification can be done for SOCl2. The author used Cl2 to dehydrate
the preform due to the limited availability of SOCl2. The chemical reaction of chlorine is described
as follows.
2Si-OH + Cl2 � 2Si=O + 2HCl
Hence isolated hydroxyl ions can be halogenated, thereby generating hydrochloric acid. The
chemical equilibrium can be shifted toward the right hand side by continuously flowing fresh
chlorine gas while applying heat. By this means, the hydroxyl concentration can exponentially be
reduced over time. The time scale required for the dehydration is dependent on the temperature and
gas flow rates, which are discussed later. Furthermore, it is important to remove all the
hydrochloric acid components after the dehydration. This step is referred to as dechlorination
process.
In order to perform the dehydration process, the MCVD method was adapted by mounting a MOF
preform (either cane or fibre preform) on a glass lathe and by connecting it to the gas delivery of
the MCVD system (see Fig.3.3.2). The gas flow consists of oxygen, chlorine, and helium with the
ratio of 10:2:1. Oxygen is a carrier gas and helps remove any organic impurities through
combustion. Helium helps to improve thermal conductivity of the gas, since the central part of a
MOF preform is thermally well-isolated. The total gas flow rate was always fixed to be 260cc/min.,
and this was practically limited by the head-stock pressure built up within the preform. This is due
to the small amount of the interstitials within the preform, through which the gas can flow to the
extract. With this flow rate, the head-stock pressure was always below 10mbar even when the heat
was applied to the preform. Note that the slow flow rate is also important to reduce the thermal
gradient within the preform.
Chapter.3 Transmission properties of small core microstructured optical fibres 64
The ratio of the constituent gas mixture was set to be the same as the values used for MCVD
preforms[171]. However, the temperature range used in the MCVD method is completely different
(~2200°C), since the dehydration process is usually applied at the final collapse stage in the MCVD
method. Although the relative amount of the chlorine may be increased to accelerate the
dehydration process, this approach was prevented due to possible risks of leakage.
Heat was applied by a travelling burner. The temperature was ~1000°C, which was the lower limit
of the pyrometer detection, and the burner was slowly scanned (25mm/min.) along the preform for
~1hour, depending on the structure. At the initial stage of dehydration, the evaporation of the water
content from the surface was clearly observable as the preform became more transparent. The time
duration of the process was set by considering the time scale used for both the VAD soot preforms
(~several hours) and the MCVD (sintered) preforms (~10minutes), and can be optimised in the
future. Note that the surface area inside the MOF preform is much greater than the glass tube (used
in MCVD) whereas it is much less than the VAD soot preforms.
Dechlorination was performed by flowing dry oxygen (~500cc/min.) for more than 24hours.
Completion of the process was checked using the chemical sensor (detection limit ~0.2ppm) near
the extract. Although this process is necessary, it involves some OH incorporation since the extract
side of the end of the preform is left open. This could be prevented by facilitating a vacuum
channel near the headstock (see Fig.3.3.2 (B)) so that the dehydration and the dechlorination
process could be held in an enclosed environment. Use of a furnace type heat source is also
advantageous since the uniformity of the process along the length can be improved without any
diffusion of OH ions from outside the preform.
Fig. 3.3.2 Schematic of dehydration process for MOF preforms. (A) a method used for the present work, and (B) a proposed method for future improvement.
Vacuum
Cl2 +O2+ He
Furnace
Cl2 +O2+ He Extract
~1000�
C
Rotary seal
Burner
(A)
(B)
Chapter.3 Transmission properties of small core microstructured optical fibres 65
The preform was then pulled on the tower by applying an appropriate internal pressure after
evacuating the preform as shown in Fig.2.5.3. Caning again involves similar limitations to those of
the capillary drawing: hydroxyl ions can be incorporated from the outside of the preform and the
draw end. The former can be compensated by post-etching. For the latter, a technique that allows
for in situ sealing (sealing the draw end of the cane while it is being pulled) may need to be
developed, as discussed in the previous section.
C) Fibre drawing
MOF preforms contain two components: canes and jacket tubes. Therefore, the canes and the inner
walls of the jacket tubes need to be dehydrated.
Because of the caning process, the outside of the cane contains a thin OH-rich layer of the order of
>10µm, similarly to that of the capillaries as described at the beginning of this section. Since the
holey cladding within the small scale MOFs typically spans ~20µm or less within a final fibre, it is
smaller than the possible diffusion length. Therefore, hydroxyl ions within the thin OH-rich layer
can diffuse into the core during the fibre drawing. Therefore, removal of this layer using
hydrofluoric acid is important.
Since the pull end of the cane is open to the atmosphere, the inside of the cane can also become wet
again after the caning process. Therefore, dehydration should ideally be performed before pulling
fibres for the inside of the cane. However, it turned out to be difficult to dehydrate inside the cane
since its inlet dimensions are typically too small (~100µm) to apply sufficient pressure. Although
the cane ends can be left open for a long time under the chlorine atmosphere, it is difficult to
monitor the progression of the dehydration and/or dechlorination processes because of the small
dimensions. When this approach was examined, the author observed the onset of hydrogen
absorption peaks and associated huge increases in background losses in the drawn fibre. This
probably resulted from insufficient dechlorination.
Therefore, it has so far only been possible to perform the dehydration for the outside of the cane
and the inside of the jacket tube. The dehydration process is the same as that explained in the cane
preform. The cane is inserted from the downstream of the pre-baked jacket tube and is then dried or
dehydrated.
Chapter.3 Transmission properties of small core microstructured optical fibres 66
3.3.3 Experimental observations
Based upon above discussions, a series of experiments was carried out using dehydrated canes
(DHLP00/01 O.D.~1.8mm) and by applying different fibre drawing procedures. These canes
consisted of a F300 rod and F320 capillaries (dc/Λc~0.8) with a F300 jacket tube. The difference
between these glasses (i.e. F300 and F320) is described in Section 6.2. For DHLP01, the outer
surfaces of the capillaries were etched (~10µm) using diluted hydrofluoric acid (~5%) prior to the
stacking. The numbers of rings of air holes were 6 and 8 for DHLP00 and DHLP01, respectively.
Fig. 4.3.1 Calculated wavelength dependence of dispersion and effective mode area for the two orthogonal polarisation axes. (dashed: x-polarised mode, solid: y-polarised mode.) (Courtesy T.M.Monro)
4.3.2 Effective mode area
The effective mode area Aeff of Yb-HNL-00 is also calculated from the model and is nearly linearly
increasing with the wavelength, as shown in Fig.4.3.1 (right). The values at the laser wavelengths
are 2.3~2.4µm2 for both polarisation axes, with the x-polarised mode being slightly smaller.
Interestingly, in the range λ<1.25µm, the x-polarised mode has a smaller Aeff than the y-polarised
mode while the opposite occurs when λ>1.25µm, although the difference is smaller than 5%. This
can be attributed to the small interstitials that are diametrically located around the core.
In a perfect hexagonal structure, the modal degeneracy (the equivalence of the propagation
constants for the two orthogonal polarisation modes) has been numerically confirmed while the
field distribution of the two orthogonal modes has been found to be substantially different[191,192].
This suggests that the direction of the modal fields is strongly affected by the air-silica boundary,
particularly when the core dimensions are small. Indeed, it has been numerically found that the
wavelength dependence of the field distribution is more pronounced when small scale structures
are incorporated and when their air silica interfaces are orthogonal to the field direction[193].
In the present case, the x-polarised mode results in a smaller effective mode area than the
y-polarised mode at short wavelengths. This is because the x-polarised modal field can decay more
strongly into the small interstitials at short wavelengths. On the other hand, when the air hole
dimensions of these interstitials become large enough with respect to the wavelength, the
x-polarised modal field extends more deeply into the interstitials and eventually tunnel through
them. The continuity condition tells us that the electric (or magnetic) field components parallel to
the boundary are only continuous across the boundary. Since the interstitials are elliptical, the
Chapter.4 Small core rare-earth doped microstructured optical fibres 79
x-polarised mode can more readily flip its field direction across the air silica boundary than the
y-polarised mode as long as the operating wavelength is short enough to resolve the structure. This
results in a smaller mode area for the x-polarised mode at the short wavelengths.
Notice that the wavelength, at which the effective mode area of the x-polarised mode becomes
greater than the y-polarised mode, roughly coincides with the turning point of the dispersion curve
(see the dashed curve in Fig.4.3.1.) and that the x-polarised mode exhibits stronger wavelength
dependent dispersion. This can be understood by considering the relatively stronger wavelength
dependence of the effective mode area (thus the modal confinement) of the x-polarised mode,
which reflects the strong wavelength dependence of the transverse wave vector and thus the
propagation constant. Therefore, there is a correlation between the dispersion curves and the
effective mode areas.
The predicted Aeff values are more than 10 times smaller than those of conventional single mode
fibres in this wavelength range. Therefore, the effective nonlinearity γ is expected to be ~10 times
greater. From the physical dimensions, the geometric core area is ~3.06µm2. Hence, it is
understood that the mode is tightly confined within the core and that a good modal overlap with the
doped section (~2.14µm2) is obtained.
4.3.3 Modal birefringence
Birefringence of fibres in general originates from the following two factors. One is form
birefringence that is induced by any asymmetry of the core structure. The other is stress-induced
birefringence, and this is widely used for conventional polarisation preserving fibre types. It is
generally known that the former dominates the latter when the index difference between the core
and cladding is more than 2% in conventional step index fibre[173]. The large index contrast
between air and silica therefore implies that the addition of any geometric asymmetry within small
core MOFs results in large values of birefringence in such fibres[173,63].
In ref.[194], high birefringence has been induced in the MOF by placing different sizes of air holes
around the core. The reported value for the beat length was 0.4mm at 1530nm. However, this
approach has a potential disadvantage since the guidance becomes weaker in the direction where
the smaller air holes are arranged. Such structures may thus suffer from greater confinement
losses[150] and may possess an orientation dependent bend loss. An alternative approach is to use
an elliptically shaped core (see Fig.4.2.1) which should be less prone to such issues but be able to
provide similarly high values of birefringence.
Chapter.4 Small core rare-earth doped microstructured optical fibres 80
Fig. 4.3.2 Experimental setup for the modal birefringence measurement (above), and the measured beat
spectrum around 1545nm for 1.1m length of the fibre (bottom). (Courtesy P.Petropoulos)
To confirm this we measured the birefringence of Yb-HNL-00 (again see Fig.4.2.1) around 1.55µm
using an ASE source with a pair of polarizers, as shown in Fig.4.3.2. When both axes are equally
excited by incident polarised broadband light, the polarisation state evolves along the length of the
fibre as a result of the relative difference in propagation constants of the modes. By inserting a
polarizer at the output spectral beating is obtained across the transmitted spectrum. The period of
this spectral beat gives information on the fibre beat length as follows.
The beat length LB is defined by[41]
���� ����� ���� � �, (4.1)
where β and n are propagation constants and refractive indices for the different axes, respectively.
B is generally referred to as the modal birefringence. The measured beat spectrum is a result of the
wavelength dependent phase difference, which is accumulated by propagation along the fibre. The
phase difference φ at a fixed wavelength λ is given by
� ��� � �� ���� ���� . (4.2)
Differentiating with respective to λ and substituting ∆φ into 2π leads to
0.4nm
Pol.
Pol.
ASE source OSA
Chapter.4 Small core rare-earth doped microstructured optical fibres 81
����������
�� �� ���� � , (4.3)
If we assume that LB is proportional to λk. Then, the LB is given by
�� ������ �� . (4.4)
Further, k=1 can be used if the wavelength range of the measurement is sufficiently narrow. By
putting the values, ∆λ=0.4nm, L=1.1m, and λ=1545nm; LB~0.3mm (B~5x10-3) is obtained, which
is shorter than the value reported in ref.[194,195] and is approximately 5 times shorter than that of
conventional high-birefringence fibres.
4.4 A mode locked ytterbium doped MOF laser
This section describes a mode locked laser based on the fabricated ytterbium doped MOF. First, the
experimental setup is presented, together with the principle of laser operation. The observed laser
characteristics are then presented and discussed. Finally, a brief summary is given.
4.4.1 Experimental setup
Fig.4.4.1 shows the experimental setup of the mode-locked laser which was based on a simple
Fabry-Perot cavity. The mode-locking mechanism employed relied upon frequency shifted
feedback into the cavity[196]. The cavity contained ~1m of Yb-HNL-00 and which corresponded
to approximately one absorption length at the pump wavelength of 966nm. The pump laser itself
was a single mode laser diode based MOPA capable of delivering up to 300mW. The pump
wavelength was detuned from the ytterbium absorption peak at 976nm so that the high nonlinearity,
anomalous dispersion and reasonably high CW efficiency could be achieved.
Coupling of the pump beam, and the signal beam into and out of the fibre was accomplished using
an appropriate choice of aspheric, achromatic lenses (f=4.5mm, NA=0.45) at both ends of the fibre
since the effective NA of the fibre was predicted to be ~0.4. Laser output was extracted from the
pump end of the cavity using a dichroic mirror (HT at 980nm and HR at 1030nm), that was angled
at 20% relative to the pump beam. Note that the required feedback from the pump end of the cavity
was provided by the ~4% Fresnel reflection from the cleaved fibre end.
Chapter.4 Small core rare-earth doped microstructured optical fibres 82
Fig. 4.4.1 Experimental setup for the ytterbium mode-locked holey fibre laser (left) and the
operation principle (right).
An acoustic optic tunable filter (AOTF) with a 3nm bandwidth (FWHM) was inserted in the cavity
and acted as both a frequency shifter and a polarizer. The AOTF could be tuned over the entire gain
bandwidth of the ytterbium transition by tuning the frequency of the RF drive by ±1MHz, about the
operating frequency of 110MHz. Thus, the AOTF provided a ~220 MHz frequency downshift per
cavity round-trip. The transmission at the central wavelength was ~90%; by maximising the
transmission using a λ/2 plate placed between the AOTF and the fibre. Combined with the high
birefringence of the fibre, this allowed us to achieve single polarisation operation of the laser at low
output powers.
The frequency shift feed back mode-locking can qualitatively be explained as follows. Under CW
operation, spectral components are continuously shifted out of the filter pass band of the AOTF as
the light makes repeated passes of the cavity, resulting in a high loss. By contrast, under pulsed
operation, new frequency components are generated via self-phase modulation, which reduces the
effective round trip loss of the cavity due to the filter. Pulsed operation is thus favoured and a
stable pulse forms within the cavity for which the effects of the amplification, SPM,
frequency-shifting and filtering per round trip balance[197]. This form of mode-locking is similar
to the idea of sliding guiding filters that was developed for soliton transmission systems[198].
Combining the above approach with nonlinear polarisation evolution (NPE) to provide saturable
absorber action, pulse durations as short as 68fs have now been obtained ������������ ��������������������pulse cavities based on conventional ytterbium doped fibre (still with bulk dispersion compensating
elements)[183]. In the anomalous regime, ~1ps pulses with a pulse energy >1nJ have been obtained
at 1.56µm using NPE and soliton effects in an erbium doped LMA fibre without the need for any
dispersion compensating components[199].
f
Gain spectrum Pulse spectrum
∆f ~220MHz
MOPA @966nm
DM
HR
Yb3+HF
AOTF
Output
L1
L2
RF
λ/2
4% Fresnel reflection
Chapter.4 Small core rare-earth doped microstructured optical fibres 83
4.4.2 Laser characteristics
Fig.4.4.2 shows the laser output characteristics obtained at 1038nm. By assuming a coupling
efficiency of 50%, the CW laser threshold and the slope efficiency with respect to the absorbed
pump power are estimated to be ~7.7mW and ~63%, respectively. These values are relatively poor
compared with conventional fibres, implying a substantial amount of background loss. Although it
was difficult to accurately characterise the transmission losses of this fibre, the background loss at
1.5µm was roughly estimated to be ~1dB/m from the cut-back measurement using a white light
source. Nevertheless, the output could be scaled up to ~100mW. Given the fact that the pump beam
diameter was greater than the clear aperture of the focusing lens, and that the fibre outer cladding
comprised relatively thick support strands that can also guide light, it is likely that an element of
pumping to the system was provided by light propagating in the cladding. However, the relative
contribution of the different pumping mechanisms to the overall laser performance was difficult to
quantify in practice.
Interestingly, the fibre often spontaneously broke when the pump radiation exceeded a certain level
(>250mW). This indicates that this fibre type possesses limited power handling capability because
of the increased thermal isolation of the core, due to the large amount of air in the cladding and the
small glass volume associated with the core. However, ten times less average power is required to
obtain the same level of nonlinearity in this fibre relative to the same length of conventional fibre
and hence the fibre is thermally capable of handling the power over the operation range of interest.
Fig. 4.6.1 Comparison of absorption spectra between the conventional fibre (EDF) and Er-HNL-00 (EDHF).
The ratios between the two peaks at the same wavelengths provide an estimate of the relative
modal overlap factors (assuming that there is no excitation of high order modes at 980nm in the
conventional fibre variant). The ratios are ~0.47 at 1550nm and ~0.69 at 980nm. Given the fact that
the high order modes are slightly excited in the conventional fibres and that high order modes
suffer less absorption, the ratio at 980nm could in fact be slightly larger than indicated by these
values. The relatively good modal overlap at 980nm compared with that at 1550nm results from the
strong wavelength dependent modal confinement in MOFs. This is favoured from a device
perspective since it compensates for the smaller doped area in this MOF allowing a shorter fibre
length to be used than might otherwise be needed.
4.6.2 A high efficiency, low threshold laser based on erbium doped MOF
Continuous wave laser operation was accomplished using a Fabry-Perot cavity. The set up is
shown in Fig.4.6.2. A fibre pigtailed single mode laser diode was used as a pump source and was
Chapter.4 Small core rare-earth doped microstructured optical fibres 97
coupled into the MOF via an aspheric lens. A 980nm optical isolator was used to isolate the laser
from the pump and a 45° angled dichroic mirror was placed between the lens and isolator to extract
the laser signal.
Fig. 4.6.2 Experimental setup for the erbium doped MOF laser.
The fibre end facets were cleaved normal to the fibre axis using a commercial mechanical cleaver
(Sumitomo:FCH-9). The cavity was closed by a high reflector at 1550nm and the ~4% Fresnel
reflection at the pumped end. The maximum pump coupling efficiency was ~50% using a lens pair
with f=12.5mm and f=3mm, and the coupling efficiency was found to be extremely sensitive to the
quality of the cleaved end of the MOF. The slope efficiency was optimised by changing the fibre
length
Fig. 4.6.3 Output characteristics of the erbium doped MOF laser at 1535nm using the 3.4m length. (left: near the threshold, and right: over the entire pump range).
Fig.4.6.3 shows the laser output obtained from a 3.4m length of Er-HNL-00, through which ~90%
of the pump power was absorbed. The laser wavelength was 1535nm and the quantum efficiency at
this wavelength is ~63.8%. A slope efficiency of 57.3% with respect to the absorbed pump power
and a threshold of 0.55mW were estimated from this data. (Note that the accuracy of the laser
output power measurement close to threshold was compromised by the stability of the laser diode
photosensitivity, than conventional fibres without increasing the NA. This is because boron doping
decreases the refractive index. Thus, GeO2-B2O3 co-doped MOFs are anticipated to possess three
advantages: high photosensitivity, better temperature stability , and increased environmental
stability.
A GeO2-B2O3 co-doped preform (HD514: NA~0.15) was stretched on the lathe and mechanically
polished to form a hexagonal rod. The diagonal distance of this rod was ~2.5mm, whilst the core
diameter was ~0.7mm. The other parameters used in the preform and the fibre drawing are
summarised in Table.5.4.1. The cut-off wavelength for the core in the final drawn fibre was set to
be 1370nm. The relatively high temperature used in the draw of this fibre enabled high speed
drawing, which resulted in relatively expanded cladding air holes. The cladding material was made
of F300 glass to ensure good transparency at UV wavelengths down to well below 240nm. The
fabricated fibres are shown in Fig.5.4.1.
Table. 5.4.1 Parameters of the preform elements, the fibre drawing, and the fibre dimensions. *: the diagonal distance of the hexagon, **: compared with the drop temperature.
Preform parameters:
Capillary Jacket Preform
I.D O.D. I.D. O.D. core diameter O.D.*
2.2mm 2.5mm 7.5mm 14mm 0.7mm 2.5mm
Draw parameters and the resultant structural dimensions:
vf vd Temperature** core diameter IC diameter Fibre O.D.
2mm/min. 49.5m/min. � 40°C 7µm 25µm 100µm
Fig. 5.4.1 A SEM photograph of GeO2-B2O3 co-doped air-clad MOF (HD514_JAC).
5.4.2 The effect of the air-sili ca interface A 6mm long grating was written by using a fixed phase mask and irradiation with pulses from a
KrF excimer laser providing a fluence of 0.5J/cm2 at 20Hz for five minutes. No care was taken
concerning the orientation of the fibre during the grating fabrication. The grating was characterised
using a white light source. The output was collected by butt-coupling the fibre to SMF, in order to
selectively characterise the core mode. The measured reflection spectrum is shown in Fig.5.4.2 (a).
A peak reflectivity of ~5dB was obtained. Note that the background ‘beat’ in the transmission
spectrum is due to the high order modes within the IC. The half width was measured to be ~0.5nm,
which is comparable to that obtained for gratings written in conventional fibres using the same
equipment under the same conditions of UV exposure.
In order to investigate the impact of the air-silica interfaces for UV writing, the air holes within
HD514_JAC were intentionally collapsed by slightly tapering the fibre at a low temperature, which
resulted in the core and fibre diameters of 4µm and 60µm, respectively. Then, the grating was
inscribed using the same dose, and was characterised using the same method described above. The
observed reflection of 3dB is somewhat smaller than in the original fibre due to the relatively poor
modal overlap with the photosensitive region due to the reduced core dimension, as shown in
Fig.5.4.2(b). However, this implies that there is no significant intensity attenuation of the UV li ght
through the air-silica interface for this type of MOF, (Note that the total number of air-silica
interfaces in the original MOF is three versus one for the collapsed fibre). Note that the different
Bragg wavelength is due to the use of a different phase mask in this instance although its alignment
was ensured by examining grating inscription within a SMF prior to the experiment.
Fig. 5.4.2 Transmission spectra from the grating inscribed in the air-clad MOF (a) and that from the grating inscribed in the ‘collapsed’ air-clad MOF (b).
5.4.3 Summary
The author has described the initial trials for fabricating fibre Bragg gratings within air-clad MOFs.
The GeO2-B2O3 co-doped air-clad MOF has been fabricated and a fibre Bragg grating with a
reasonable strength (~5dB) at 1550nm was successfully inscribed, without hydrogen loading, using
a fixed phase mask and a KrF excimer laser. Comparison with gratings produced in a conventional
fibre counterpart, which was prepared by slightly tapering the fabricated fibre and collapsing the air
holes, indicates that the air-cladding within this MOF, which comprises a single ring of air holes
with a structural scale of ~20µm, has had little effect on the quality of the grating writing process.
Unfortunately, due to failure of the excimer laser half-way through this study it was impossible to
study the detailed characteristics of the MOFs gratings with respect to the dose. In addition,
experiments could not be carried out on the temperature or environmental stability. The laser has
just recently been replaced and the investigations will be resumed in due course.
5.5 Conclusions
Two types of novel air-clad microstructured optical fibres have been developed. It has been shown
that, based upon a crude effective index model, high IC NAs in excess of 0.5 can be achieved. The
design criteria for CPFs have been discussed, and it has been shown that, although the NA can be
increased by decreasing the structural dimensions within the air hole cladding, the smallest inner
cladding diameter that can be used in practice will ultimately be determined by the pump
brightness.
Several ytterbium doped air-clad MOFs have been fabricated. We have experimentally shown that
the inner cladding NA can be as high as 0.5, while the inner cladding diameter is reduced to be as
small as 28µm. Using these fibres, we have shown extended tunability and 980nm operation of
ytterbium doped cladding pumped fibre lasers and that were only possible due to the short
(sub-metre) device lengths enabled by the combination of the small inner cladding dimensions and
high NA.
A photosensitive GeO2-B2O3 air-clad MOF has been fabricated, and the effect of the multiple
air-silica interfaces on UV written grating fabrication has been studied. It was found that it is
possible to inscribe a reasonably strong grating (~5dB) without hydrogen loading in this fibre.
Hence the air-silica interfaces do not degrade the grating writing process at least when the number
of the air silica interfaces within the fibre is small. Further study is required to investigate the effect
of increasing the number of air-silica interfaces, on the scale of the air holes within cladding, and
how the maximum strength of the grating changes with respect to total UV dose. This research is
on-going at the ORC, and an examination of the temperature and environmental stability of the
gratings written in this novel photosensitive air-clad fibre will soon be conducted.
Chapter.6
Large mode area microstructured optical fibres
6.1. Introduction
Large mode area (LMA) optical fibres are becoming increasingly important as a high power
transmission medium. One major driving force results from the constant increase in the output
powers from laser diodes that allows for higher output powers from diode-pumped solid-state
lasers. As a result, higher demands are being placed on the power handling capability of optical
fibres. For instance, the recently developed high average power ultrashort pulse sources has
effectively utilised a LMA fibre as a nonlinear fibre to compress the pulse duration below
50fs[257], a pulse duration that is difficult to obtain directly from high average power oscillators.
Moreover, for medical and industrial use of high power lasers such as excimer, Nd:YAG, Er:YAG,
and CO2 lasers, optical fibres are expected to provide a convenient and flexible means of handling
the light without the difficulties of alignment. Traditionally, hollow fibres with appropriate inner
coatings or hollow tapers have been used[258-260]. However, given the recent advances in diode-
pumped solid state lasers with significantly improved beam quality, in conjunction with the
nonlinear wavelength conversion technique[261], single mode optical fibres with low nonlinearity
such as LMA fibres need to be developed for a range of wavelengths.
LMA fibres are also becoming increasingly important as an active medium due to the recent
interest in the possibility of very high power fibre lasers. Single mode output powers of more than
100W are now routinely reported[212-214] and 1kW output has been reported from a multimode
fibre laser[215]. For many applications, a diffraction limited quality of the output beam is one of
the key features that the active optical fibres can provide, since it allows for the maximum
brightness achievable for a given fluence or average power. It is thus important to increase the
mode size of the fundamental mode, in order to avoid the limiting factors of both optical damage
and nonlinear thresholds.
Chapter.6 Large mode area microstructured optical fibres
125
So far, conventional LMA optical fibres have been fabricated using the MCVD technique. These
fibres are characterised by a very low numerical aperture (NA<0.1) and an additional ring profile
(see Fig.7.2.1). In order to control the desired refractive index profile, a precise flow control of the
chemical vapour is required during the deposition process. Unless fabrication systems are fully
optimised, it is challenging to reliably fabricate such low contrast refractive index profiles,
particularly at short wavelengths. Despite this, an optimised MCVD process, combined with the
solution doping technique, pioneered here in Southampton[128], has allowed for an increase in the
effective area by a factor of four using LMA fibres at 1550nm[262].
High power single mode operation is now routinely achieved using single mode excitation of step
index multimode fibres[263,264]. As a result, millijoule pulse energies can now be obtained from a
Q-switched fibre laser[265], an ultrashort pulse amplifier[266], and a MOPA configuration[267]. In
particular, nearly diffraction limited beam quality has been obtained with the aid of preferential
gain[268] in the case of active fibres, where the modal overlap with the doped region is designed
such that only the fundamental mode is efficiently amplified even when several modes are
supported within the fibre.
However, it has been shown that the single mode propagation distance in MCVD based multimode
fibres is proportional to Ls~2x1015xD6λ4/(ρ10n6), where D and ρ are the fibre and core radii, λ and n
are the wavelength and the refractive index of the core[263]. It is readily recognised that D must be
very large (>>125µm) in order to achieve Ls>10m at 1.06µm, and that it is increasingly difficult to
achieve single mode operation at short wavelength due to the λ4 dependence. In ref. [262], a tight
bend was imposed on the fibre to prevent high order modes. For passive applications, these
restrictions can severely limit the usefulness of MM fibres. Furthermore, for active fibres, large
cladding dimensions suggest reduced absorption and thus long device lengths are required. This
requirement can impose other limitations such as Brillouin scattering[41]. (Further issues
associated with use of the long active device lengths are also discussed in Chapter 5)
MOF technology provides an alternate route for low-NA LMA fibres, and possesses an advantage
that perfectly homogeneous pure silica core fibres can be readily made. This implies that MOFs are
inherently suitable for pulsed and/or short wavelength operations owing to the wide electron band
gap of pure silica and to the fact that the refractive index contrast is not limited by doping levels.
Although large mode area microstructured fibres (LMA-MOFs) have been already
demonstrated[7], practical issues such as macroscopic bend loss and power handling of such fibres
were not well understood prior to this study. Furthermore, no active devices based on LMA-MOF
had been reported.
Chapter.6 Large mode area microstructured optical fibres
126
This chapter describes the fabrication and characterisation of LMA-MOFs, aiming at forming a
foundation for the work on active LMA-MOFs that is presented in Chapter 7. The author begins by
discussing the fabrication issues related to LMA-MOFs in Section 6.2. The basic optical properties
of LMA-MOFs are experimentally characterised including effective mode area in Section 6.3 and
bend losses in Section 6.4, in both of which a comparison with conventional LMA fibres is carried
out. Furthermore, the transmission losses of LMA-MOFs are presented in Section 6.5. Finally, the
conclusion is given in Section 6.6.
6.2. Fabrication of large mode area microstructured optical fibres
In this section, the refractive indices of the materials that can be used for silica based LMA-MOFs
are reviewed first, as they can greatly influence the performance of LMA-MOFs. In addition, a
doped LMA-MOF incorporates a core material with different refractive indices, as shown in
Chapter 7. As we learnt more about the sensitive characteristics of LMA-MOFs, a number of
modifications of the fabrication process became necessary. In the near future, further modifications
are likely to be required. Therefore, in Section 6.2.2, an overview of the evolution of the fabrication
process so far is presented.
6.2.1. Refractive indices of silica based materials
As is in any low NA fibre, any index difference within the transverse index profile of the LMA-
MOF can strongly influence the modal properties of the fibres. This is because the index contrast
created by the air holes is typically very small and may be comparable to any slight index
difference introduced by the use of different silica based materials within the fibre cross section. It
is thus naturally important to know the refractive indices of the start materials that can be used
within such fibres. The measurement of the refractive indices of various silica glasses is presented
below.
The author made a trial preform, which consisted of several different silica glasses by collapsing
and jacketing different tubes on the lathe. Then, the refractive index profile of this preform at
633nm was measured using a preform index analyser (P-104, York technology). We also made an
index measurement on bulk samples using a refractometer, in which the sodium D-line (589nm)
was used as a light source. Although the relative index differences between some of the glasses
were near to the resolution limit (∆n~10-4) of these two measuring techniques, the index values and
differences with respect to the high quality silica are summarised in Table.6.2.1. The absolute
offset of the respective measurements was eliminated for the purpose of comparison.
Chapter.6 Large mode area microstructured optical fibres
127
Suprasil® F300 and F320 are synthesised silica glasses with an extremely low OH content
(<0.2ppm), and the difference between them are due to the agents used during their dehydration
process. Consequently, F300 contains a substantial amount of chlorine (~1500ppm) whilst F320
contains an amount of fluorine (~3500ppm). Suprasil® F100 is also a synthesised silica made
without dehydration, containing <1000ppm of OH ions. However, the transmission window of
F100 extends below 200nm, thereby it is often used for UV optics and fibres. HLQ-210 and
Vycor® are fused silica, and they contain significant amounts of impurities as well as high water
content. The latter contains ~3wt.% of boron and ~4wt.% of sodium, leading to a significantly
reduced softening temperature, compared with the others glasses discussed here.
Table. 6.2.1 Measured refractive indices for different raw materials. ∆ is the index difference with respect to the values measured for Suprasil® F300. (*unknown)
Both chlorine and sodium increase the refractive index of silica whilst fluorine and boron reduce it,
as long as the overall silica content is high enough. Note that the degree of index modification can
be dependent on both the dopant concentrations and the species. The measured values can be well
explained by taking into account these dopants. The relative refractive index differences measured
from the preform analyser are smaller than those measured from the refractometer. This can be
attributed to the diffusion of impurities that took place when the glass was heated to prepare the
sample used for the preform analyser, and which led to a reduction in the refractive index contrasts.
Note that when different materials are used in the same MOF preform, heat is also applied to the
glasses although the effect is not as significant as that during the collapsing process on the lathe.
Throughout both of the measurements, it was found that the refractive index of HLQ-210 is not as
spatially uniform as those of the synthesised glasses. This resulted in a noisy index profile,
particularly in the low quality silica layer, in the preform analyser. The reading was also blurred in
the refractometer for HLQ-210. This may be attributed to some inhomogeneously localised
impurities.
In fabricating MOFs, there are three possible combinations of materials as follows:
Chapter.6 Large mode area microstructured optical fibres
128
A) Single material
B) High index material core and low index material cladding
C) Low index material core with high index material cladding
Most of MOFs reported to date use type A). Type C) has been reported using F320 in the core with
F300 in the cladding, where an antiguide was formed at short wavelengths since the index
difference created by the air holes is compensated by the index difference between the core and
cladding materials[269]. Type B) is also a typical case when doped fibres are made, as described
Chapter 4 and 7. However, the index difference has to be sufficiently small to still receive the
benefits of the MOFs since their NAs are increased by the doped sections. Amongst the materials
investigated here, the highest index difference is between those of F300 and F320 (~8.6x10-4). The
corresponding NA is ~0.05. To achieve rigorously single mode operation at 1µm using
conventional step index fibre designs, the core diameter has to be less than 15µm for this material
combination. When air holes are introduced in such a step index fibre, it is likely to become
multimoded due to the extra contribution of the air holes to the total NA. Thus, particularly in
LMA fibres, it is not so trivial to incorporate air holes together with a high index section created by
the use of different silica based materials. In the following sections, we demonstrate types A) and
B), and examine the difference in their optical properties.
6.2.2. Evolution of the LMA-MOFs fabrication process
Using the materials described in the previous section, a range of LMA-MOF s have been fabricated.
The fabricated fibres are summarised below.
A) The first generation
At the very early stage of LMA-MO F development, the preform consisted of a F300 core, HLQ-
210 capillaries, and a Vycor® jacket since it was anticipated to be the cheapest combination (from a
cost perspective) whilst achieving an acceptable loss level. The capillaries were prepared by
drilling a 13mm diameter air hole within a 25mm O.D. HLQ-210 rod using an ultrasonic drilling
machine and then by drawing it into ~2mm O.D. capillaries after a fire-polishing step since drilling
results in surface roughness of ~1µm. Due to the thick walls of the tube, it was difficult to fire-
polish the inner wall on the lathe. This may be improved by flowing helium inside the tube during
the fire-polishing stage due to the increased thermal conductivit y of such gases. The reason why
this approach was taken is because it is the most straightforward way of obtaining a uniform thick
walled tube (I.D./O.D. < 0.5) with a large O.D.
Chapter.6 Large mode area microstructured optical fibres
129
It was difficult to stack these 2mm thick walled capillaries to form the preform, since they do not
deform significantly unlike thin walled capillaries. Combined with the poor qualit y of the
unprocessed Vycor® jacket tube, the resultant preform assembly was hopelessly disordered. Note
that it is now possible to improve the circularity of Vycor® tubes on the lathe. The preform and the
draw parameters are summarised in Table.6.2.2.
Table. 6.2.2 The fibre draw parameters, the preform dimensions, and the fibre dimensions of LMA- 0001.
Draw parameters:
vf [mm/min.] vd [m/min.] Temp [°C]
1.216 10~30 ∆75~85
Preform dimensions:
Core (F300) Capillaries (HLQ-210) Jacket (Vycor®)
O.D. [mm] I.D. [mm] O.D. [mm] I.D. [mm] O.D. [mm]
~2.0 ~0.5 ~2.0 ~18 ~20
Final fibre dimensions:
O.D. [µm] Λ [µm] d/Λ
125~200 8.5~15 0.1~0.2
A range of different dimensions of fibre were fabricated by varying the draw temperature and the
draw speed, and an example of this is shown in Fig.6.2.1 (LMAHF-0001). Although the structure
was approximately retained along the entire length of the fibre (~500m), unexpected gaps were
formed between the jacket tube and the inner capillary bundle in addition to the interstitial air holes
between the capillaries.
Fig. 6.2.1 SEM photographs of LMAHF-0001. (a) the entire cross section and (b) the core region of the fibre. The fibre O.D. was varied from 125µm to 200µm by varying the drawing conditions.
(a) (b)
Chapter.6 Large mode area microstructured optical fibres
130
As discussed in Chapter.2, the idea of the collapse ratio can be applied to the individual preform
components. By dividing the collapse ratio of the capillaries by that of the jacket tube, a ratio of
~0.07 is obtained assuming the same viscosity and the surface tension. This means that the
capillaries collapse far more than the jacket, which is the cause of the gaps between the stack and
the jacket tube and interstitials. Since the jacket almost fitted on to the capillary stack (i.e. nearly
the same collapse ratio), the viscosity difference between F300 and Vycor® is estimated to be
nearly a factor of 10 at these temperatures. The gaps between the capillary stack and the jacket tube
are mainly attributable to the low filling factor of the initial preform due to the poor quality of the
stack. This could be eliminated by using a thicker jacket so that the collapse ratio of the jacket can
be slightly increased.
B) The second generation
A second generation of LMA-MOFs was developed with the aim of realising a single material fibre,
the development of which is described in Chapter 2. The difference from the previous generation
was that it became possible to collapse a sufficiently long tube on the lathe, so that higher quality
capillaries can be obtained. The difficulty involved in this process is to obtain good control of the
inner diameter without cutting the tube, since the O.D. of the tube does not change as much as the
I.D. during the collapsing process. However, recently developed laser diameter gauges allow us to
measure both I.D. and O.D. of the tubes simultaneously using the beam deflection technique. Thus,
this issue should soon be overcome by utilising such a device.
Another technique introduced in this generation was a screening procedure. By using short lengths
(~10cm) of capillaries selected in terms of their dimensions, it becomes possible to stack the
capillaries neatly. The author initially concentrated on fabricating fibres with a very low fraction of
air (d/Λ<0.1). However, it turned out to be difficult to guide light within these fibres because their
extremely low NA leads to significant bend losses (and possibly also because of the confinement
losses) (see Fig.6.2.2 (a)).
These losses can be significantly reduced by increasing the number of rings of air holes in the
cladding[150]. Therefore, the number of the rings was increased up to seven (>150 elements).
However, this approach was not successful in helping the light guidance.
The other option to cope with this issue could be to use a double clad structure as shown in
Fig.6.2.2 (c). By this means, it was anticipated that the modal leakage could be reduced owing to
the stronger modal confinement in the outer cladding without significantly influencing the mode.
However, the mode was strongly confined within the inner cladding rather than in the core. As a
Chapter.6 Large mode area microstructured optical fibres
131
result, it was impossible to determine the guidance properties of the core mode due to the bend
losses that couples its power into the cladding modes within the inner cladding.
Fig. 6.2.2 Impractical fibres because the air holes are too small. The single clad fibre: (a) and (b). The double clad fibre: (c) and (d).
By comparing with LMA-0001, which possesses similar structural parameters in terms of d and Λ,
it is understood how significantly the wave guidance in LMA-0001 was assisted by the slight index
difference between F300 and HLQ-210.
It was found that it is necessary to increase d/Λ to >0.2 in order to realise practically usable single
material fibres. Note that the practical choice of d/Λ can be dependent on Λ. Seven rings of air
holes were used in the examples shown in Fig.6.2.3 (LMAHF-0100 and LMAHF-0101). The
typical draw parameters used for these fibres are summarised in Table.6.2.3. Due to the large
number of capillaries, it became difficult to align the core centrally with respect to the jacket tube.
This was found to be the cause of the orientation dependent bend losses, described in Section
6.4[270].
Finally, it should be noted that when using this fibre, some of the author’s colleagues found that it
was difficult to obtain a reasonable agreement between the theoretical predictions (localised
function method[18] and multi-pole method[25,26]) and the experimental measurements using a
(a) (b)
(c) (d)
Chapter.6 Large mode area microstructured optical fibres
132
mode field diameter (MFD) measurement[44] and a scanning near field microscope (SNOM)[271].
This problem was solved using the fibre that was made of all F300, in the third generation. The
refractive index inhomogeneity within the HLQ-210 rod may be the reason for this discrepancy, as
discussed in the refractive index measurement section.
Table. 6.2.3 The draw parameters used for LMAHF-00 and 01.
Fig. 6.2.3 LMAHF-0100: (a) the entire cross section and (b) the core region, and LMAHF-0101 : (c) the entire cross section and (d) the air-clad region.
Chapter.6 Large mode area microstructured optical fibres
133
C) The third generation
The third generation was developed using only Suprasil® F300 glass as a raw material. Capillaries
are drawn from the tube that was prepared by slightly etching, using hydrofluoric acid (<5µm), and
then fire-polishing in the furnace by purging with argon. The jacket tube was prepared in the same
fashion. The reasons why the preparation was modified are explained below. When gas phase
etching is applied using SF6, a thin fluorine doped silica layer is formed, as is found in many
refractive index profiles of MCVD preforms (see Fig.7.2.1). This low index fluorine doped silica
layer remains during the fabrication process, and may help trap the cladding modes when such a
tube is used as a jacket.
By using a wet etching process, the formation of this fluorine layer can be prevented. In addition,
the OH ions in the final fibre can also be reduced as described in Chapter 3. The only drawback of
this method is that it is difficult to judge the completion of the fire-polishing of the etched tube as
the furnace is enclosed. By observing the scattered light when capillaries are drawn, the
temperature was adjusted so as to reduce the amount of scattered light.
In addition, in order to reduce the orientation dependence of the bend losses, as described in the
previous section, the boundary between the cladding and the jacket was hexagonally shaped by
inserting rods in the vicinity of the jacket tube. By using sealed capillaries with dc/Λc~0.4, high
temperatures were used to collapse the air holes in order to obtain d/Λ=~0.3. However, the collapse
of the air holes begins at the outer-most ring as the temperature is higher here due to the presence
of the temperature gradient within the preform, as discussed in Chapter 2. This degrades the
transverse uniformity of the structure, leading to the partial collapse of the air holes near the jacket
tube due to the lower viscosity. Furthermore, the drawing becomes somewhat unstable due to the
reduced tension at high temperatures. These two factors limited the highest temperature that could
be used. The examples of the draw parameters and the fabricated fibres (LMA_0200) are shown in
Table.6.2.4 and Fig.6.2.4, respectively.
At low temperatures, interstitial air holes appeared within the jacket region, where the glass rods
are stacked in order to shape the cladding.
Thus, the useful temperature range, where no interstitials appear and the partial collapse is achieved,
is at most within ±10°C in this example (d/Λ=0.25~0.35). The temperature range is generally
narrower when the difference in dimensions between the capillary and the jacket is larger (i.e.
when a substantial number of silica rods were inserted to shape the cladding boundary, which
enlarges the absolute difference in dimensions between the jacket and the capillaries). Such a
preform contains a greater difference in the collapse ratio in comparison with LMAHF-0100,
Chapter.6 Large mode area microstructured optical fibres
134
making interstitials appear more easily at low temperatures. In addition, the effect of the thermal
gradient across the preform is more pronounced at high temperatures.
Table. 6.2.4 The preform and the fibre draw parameters used for LMAHF-0200.
Draw prameters:
vf [mm/min.] vd [m/min.] Temp [°C]
2.605 15 ∆0~60
Preform dimensions: Capillaries (sealed) Jacket
I.D. [mm] O.D. [mm] I.D. [mm] O.D. [mm]
~0.4 ~0.94 ~16 ~20
Final fibre dimensions:
O.D. [µm] Λ [µm] d/Λ
230~250 11~13 0.25~0.5
Fig. 6.2.4 SEM photographs of LMAHF-0200 (Λ=11.3µm, d/Λ=0.35).
The optical properties of these fibres are currently under investigation.
6.3. Effective mode area
Here, experimental measurements of the effective mode areas of LMA-MOFs are presented.
Initially, fibre nonlinearity measurements were used to characterise the effective mode area.
However, the measurement technique has been subsequently shifted to mode field diameter (MFD)
measurement using a knife-edge technique, as is commonly used for standard telecommunication
fibres. These two different measurement techniques are compared and the measured effective mode
areas are discussed with respect to their structural parameters.
(a) (b)
Chapter.6 Large mode area microstructured optical fibres
135
6.3.1. Measurement via nonlinearity
The effective mode areas of the fibres drawn from LMA-0001 were characterised by measuring the
four wave mixing (FWM) of intense two colour pulses propagating through the fibres[272]. The
experimental set-up is shown in Fig.6.3.1.
Fig. 6.3.1 Experimental setup for the nonlinearity (effective mode area) measurement. DFB: distributed feed-back laser, EESL: grating stabilised edge-emitting-semiconductor laser, EDFAs: erbium doped amplifier modules, AOM: acousto-optic modulator, FI: Faraday isolator, LM-EDFA: Large mode area EDFA, P.M.: power meter, and OSA: optical spectrum analyser.
A master oscillator power amplifier (MOPA) configuration was used. The CW semiconductor laser
outputs were combined and then gated by using an electric-optical modulator (EOM) to generate
5ns pulses, which were then seeded into an 3-stage erbium doped fibre amplifier (EDFA)
chain[262]. At the output we obtained micro-joule level pulses with a variable repetition rate, and
thus a variable pulse energy. An acousto-optic modulator (AOM) was inserted prior to the LMA-
EDFA to suppress the amplified spontaneous emission (ASE).
By launching the two colour pulses into a sample fibre, one could observe the generation of side
bands via FWM. The ratio between the main peak Imain and side-band peak Iside is related to the
nonlinear phase shift φ of the pulses by the relation
� � � �� � � �����
����� ����
��
�� ��
���
� � ������ � � �
��� , (6.1)
where Jn is the nth Bessel function of the first kind. Thus, using this equation, the launched power
versus the nonlinear phase shift can be obtained. The nonlinear phase shift is written as follows.
Chapter.6 Large mode area microstructured optical fibres
136 � �� �� �� ���� ����� �������� � ���� , (6.2)
where P is the launched pump power and
�� � ��
is the effective mode area of the fibre, respectively.
λ, α and L are the pump wavelength, absorption coefficient, and the length of the fibre,
respectively. Therefore, by assuming the nonlinear coefficient (n2~2.16x10-20 [m2/W]) for pure
silica[45], the effective mode area of the sample fibre can be estimated[272].
The accuracy of this measurement technique is limited by the inherent low nonlinearity of the
LMA-MOFs, which was comparable to the background nonlinear phase shift accumulated by the
propagation of the incident pulses through the amplifier chain. Though the nonlinear phase shift
within the LMA-MOFs can be increased by reducing the repetition rate (and thus increasing the
pulse energy), this is limited by the distortion in the pulse shape due to the saturation of the
amplifier. In addition, the background phase shift is also increased, leading to high order side band
generation.
The measurement was performed so that the dynamic range was as large as possible for a given
length of the fibre (~3m), whilst the pulse shape was retained. However, the bend loss present in
some of the fibres with the largest
�� � ��
prevented us from accurately characterising them.
Nevertheless,
�� � ��
as large as 436µm2 could be measured using this approach for fibres drawn
from LMA-0001. A typical plot of the nonlinear phase versus internal peak power is shown in
Fig.6.3.2. The measurement error is mainly due to the power fluctuation of the Ti:sapphire laser,
which pumped the final amplifier. The estimated
�� � ��
for these fibres with a range of different
dimensions are summarised in Table.6.3.1.
Fig. 6.3.2 Measured nonlinear phase shift with respect to the launched peak power of LMA-
MOF. The data was taken at the repetition rate of 15kHz and the calculated
�� � ��
is 436µm2.
� � !! � "! � #! � $! � "
� ! � � " � � % � � # � � & � �
' ( ) *+ , - . */ 0 (1324 5 6 7 4 83936 5 6 6 4 : ; < =?> 2@6 5 A < 8 B ;132C6 5 A ; A 7 A3936 5 6 6 4 : B B 4 =D> 2C6 5 A B 4 7 ;132C6 5 A B 6 8 4E936 5 6 6 4 : : < : =D> 2C6 5 A A 7 ; F
GH IJKL MINPOQ
RTSVU@W XY Z\[
Chapter.6 Large mode area microstructured optical fibres
137
Note that because of the irregularities present in the transverse structure of the fibre, it is difficult to
explicitly quantify d and Λ for these fibres. The limit of the measurement can clearly be observed
from the fibres E and J, where the core area A is more than ~450µm2. �� � ��
values are far smaller
than those of D and I, which have smaller core areas than E and J, respectively. Although the
author set the fibre such that the bend curvature was very loose (>50cm) and almost homogeneous
along the length, these results indicate that the bend losses were present during the measurement
for these fibres, leading to underestimates of �� � ��
(see Eq.(6.2)).
The results obtained from the other fibres qualitatively follow the anticipated tendency that the
value of �� � ��
increases with increased Λ or decreased d/Λ. For instance, the fibres C (Λ=11.36µm,
d/Λ=0.189) and D (Λ=12.57µm, d/Λ=0.187) have a similar d/Λ, and the measured values of �� � ��
are 268µm and 311µm, respectively. By comparing the fibres C and I (Λ=11.01µm, d/Λ=0.105),
the measured �� � ��
values are 268 and 294µm2, respectively. This confirms that the use of small
d/Λ drastically increases the effective mode area. These examples imply that �� � ��
is either a strong
function of d/Λ or Λ, or both, depending on these parameters.
Table. 6.3.1 Summary of the fibre characterisation. O.D.: outer diameter, Λ: hole spacing (average), d: hole diameter (average), A: calculated geometric core area from the SEM photographs, Aeff
N: effective mode area via nonlinearity measurement, and Aeff
M: effective mode area via MFD measurement.
O.D. Λ D d/Λ A AeffN Aeff
M
[µm] [µm] [µm] [µm2] [µm2] [µm2]
A H2 108 8.67 1.786 0.206 203 123 126 B H6 145 10.38 2.19 0.211 282 190 195 C 163 11.36 2.147 0.189 336 268 - D H7 173 12.57 2.357 0.187 401 311 300 E 198 14.71 2.088 0.142 577 220 - F 193 14.59 1.967 0.135 587 - - G H4 170 14.09 2.237 0.159 533 282 378 H 158 11.46 1.546 0.135 353 243 - I H5 168 11.01 1.159 0.105 350 294 305 J H3 168 12.58 1.334 0.106 447 232 - K 188 13.76 1.391 0.101 508 436 - L H1 195 12.97 1.246 0.096 536 - 680
Indeed, our theoretical model has revealed such characteristics[45]: when d/Λ is >~0.12, Aeff is
relatively insensitive to d/Λ, whilst it becomes a strong function of both Λ and d/Λ when d/Λ
becomes smaller, as shown in Fig.6.3.3. This, in turn, suggests that it is difficult to precisely
control Aeff in the region of d/Λ <~0.1 in practice, because of the increased sensitivity to both d and
Λ. Note that as long as collapse is utilised, Λ changes with d. However, d is more directly affected
Chapter.6 Large mode area microstructured optical fibres
138
by the fluctuation of the collapse than Λ. Hence, it is expected that the control over Aeff can be
improved by increasing the value of d/Λ.
Fig. 6.3.3 Predicted effective mode area as a function of structural parameters d and Λ at 1.55µm wavelength. Contour level unit is µm2. Labels H1-H6 corresponds to the measured values for some LMAHF-0001 with different dimensions. (Courtesy J.C.Baggett)
In fibres drawn from LMAHF-0001, the preform parameter dc/Λc~0.5 was used and the control
over the fibre structure heavily relied on the collapse of the air holes during the fibre drawing
process. Thus, d and Λ could not be controlled independently and are also sensitive to the draw
temperature. Note that the temperature range used for these fibres was only within ±5°C. If the
degree of the collapse was smaller, the sensitivity of Λ to the draw temperatures would be reduced
and thus the change in Λ would be negligible compared with that of d. Therefore, the preform
parameter dc/Λc should ideally be similar to d/Λ in the final fibre.
6.3.2. Measurement via mode field diameter (MFD)
The mode field diameter of the fibre can be characterised by measuring the divergence of the
output beam versus the distance from the fibre end assuming Gaussian beam propagation in free
space. A Gaussian beam, which has a beam radius of w0 and radius of curvature ��� at z=0,
expands as follows. ��� � � �� � ����� � , (6.3)
where �� ��������� . By differentiating by z, we obtain
� � ������ ������ �
�!! "
#$%& (z>>zR). (6.4)
Chapter.6 Large mode area microstructured optical fibres
139
Thus, by measuring dw/dz, w0 is obtained, from which the effective mode area and the mode field
diameter are immediately calculated using ������
� � � � and MFD=2w0, respectively.
Fig. 6.3.4 Experimental setup for the MFD measurement.
The experimental setup is shown in Fig.6.3.4. By detecting the beam through a beam chopper
blade, which rotates with a well defined frequency, the signal rise follows the integrated Gaussian
function, in which the half width positions correspond to a height of 16 to 84%. Here, we assume
that the chopper diameter is sufficiently large, compared to the beam diameter. Therefore, by
measuring the rise time of the leading edge of the signal at different distances from the fibre end, it
is possible to measure FWHM of the beam diameter with the known blade speed. The advantages
that this method includes are as follows.
Its adaptability to any light sources as long as their linewidth is narrow enough.
The fibre length is not a severe parameter as long as the cladding modes are stripped off.
The measurement is independent of bend losses.
Disadvantages can be the fact that a relatively small beam divergence is required to ensure
adequate beam capture by the lens behind the chopper blade, and that accurate measurement of
both the distance in the z direction and the small beam diameter with respect to the chopper blade is
required. An additional approximation results from the non-circular modal shapes in MOFs.
Furthermore, Gaussian beam propagation of the modes is assumed. However, the error caused by
these facts is theoretically estimated to be 10% at most for LMA-MOFs[44]. Therefore, it can be
concluded that this technique is particularly suitable for the characterisation of LMA fibres.
For some of the LMAHF-0001 fibres and conventional LMA fibres, we examined the wavelength
dependence of the �� �
measurement as shown in Fig.6.3.5 using various lasers of different
2w
84%
16% Power transmission through the chopper
Fibre Time
Chopper blade
Light source
Detector Collecting lens
Chapter.6 Large mode area microstructured optical fibres
140
wavelengths. For comparison, MOFs, which had similar characteristics at 1550nm to the
conventional LMA, were chosen. Note that �
� � ��
as large as 680µm2 was measured at 1550nm.
Fig. 6.3.5 Wavelength dependence of the effective mode area measured by MFD measurement. H1-4: LMA-MOF (see Table.6.3.1), S1: conventional LMA (NA=0.06,
a=9µm, �
� � ��
=405µm2 at 1550nm), and S2: SMF (NA=0.11, a=4µm, �
� � ��
=126µm2 at
1550nm). (courtesy J.C.Baggett)
The conventional fibres are limited below 1.3µm due to the presence of the modal cut-off. Several
tens of modes are present in the visible wavelengths in these fibres, making it difficult even to
selectively excite the fundamental mode. Although the set of fibres LMA-0001 contain an index
difference between the core and the cladding materials and thus display cut-off, it was confirmed to
be only few moded at 488nm. Thus, it was still possible to measure the effective mode areas at
short wavelengths in the LMA-MOFs.
By referring to Table.6.3.1, the following trends can be understood. If d/Λ is very small, as it is for
H1, �
� � ��
drastically increases at longer wavelengths, while the fibres with d/Λ ~0.2 give a
relatively flattened wavelength dependence of �
� � ��
over a wide spectral range[44]. This is because
the modal penetration into the cladding at longer wavelengths can be suppressed with these
structural parameters. Generally, smaller d/Λ also allows for deeper modal penetration into the
cladding. On the other hand, when d/Λ is relatively larger, modal confinement at shorter
wavelengths is significant. Thus, there is a certain value for which both of these characteristics are
less pronounced. Note also that the strong wavelength dependent confinement of light is
particularly significant for smaller Λ. Therefore, Aeff can be flattened over a wide range of the
spectrum by controlling the value of d/Λ for a given core diameter (or Λ), especially for relatively
large core fibres. This may be useful for maximising nonlinear optical parametric processes within
Chapter.6 Large mode area microstructured optical fibres
141
the fibre[273], where two or more wavelength components interact with each other, and
upconversion type lasers, where the pump and signal wavelengths are significantly different.
6.4. Bend losses
Bend losses are an important factor in LMA-MOFs as they practically limit the largest mode areas
that can be used. The bend losses are customarily categorised into two components: micro-bend
loss and macro-bend losses, depending on the physical scale of the bend, with respect to the core
diameter. The former is primarily related to the issues arising from cabling and installation, where
the propagating modes within the fibre continuously suffer from modal distortion, whereas the
latter results from the bends that are significantly greater in scale than the wavelength of the light
guided by the fibre and represents the factors that limit the practical handling.
Macro-bend losses can be categorised into two components: transition loss and pure bend loss[155].
The former is induced when the mode enters or exits a bend due to the abrupt change in curvature,
which leads to a change in modal shape. The latter happens continuously as the mode propagates
along the bend, and this is known to be the main contributor to macro bend losses for practical
situations[270]. This mechanism can alternatively be explained as follows. On a straight fibre, the
modal field at every point in the cross-section propagates parallel to the fibre axis with a constant
phase velocity, thus the phase front is orthogonal to the propagation axis. Under the presence of the
bend, the phase front has to rotate about the centre of curvature of the bend. This means that the
phase velocity parallel to the fibre axis increases with distance from the centre of curvature. Once
the phase velocity matches that of the cladding modes, a large loss occurs due to the coupling to the
cladding and radiation modes.
In Ref.[14], it has been predicted that the bend losses in MOFs are qualitatively different from the
conventional step index fibres, as they exhibit a short wavelength edge in the bend loss spectrum.
In conventional fibres, the bend loss is dominant at long wavelengths due to the fact that the
refractive index difference or NA of the core becomes small. This leads to low V values, in which
the guided mode is prone to be affected by environmental perturbations. At the same time, the
transverse wave vector components of the guided mode become small, allowing for deep modal
penetration into the cladding. This increases the modal overlap with the radiation modes. As a
result, the bend loss is also increased due to the enhanced coupling.
However, in single material MOFs, the NA decreases with the wavelength. This fact leads to an
additional bend loss edge at short wavelengths, whilst the bend loss edge due to the deep modal
penetration into the cladding at the long wavelengths also exists. The bend losses at short
wavelengths in MOFs can also be understood by considering the hole spacing Λ with respect to the
Chapter.6 Large mode area microstructured optical fibres
142
mode field diameter MFD. If MFD<<Λ, the mode can leak out from the core through the silica
bridges in the holey cladding. Note that the MFD is small at short wavelengths despite the weak
guidance since it is determined by the transverse wave vector components of the mode, which is
inversely proportional to the NA in endlessly single mode MOF since the V value is constant at the
short wavelength limit. Indeed, one physical interpretation to the large transverse wave vector
components is that the light on the short wavelength edge can ‘resolve’ the holey cladding and thus
escape from the core.
Although quantitative predictions for the bend loss calculation have been controversial even for
conventional fibres, at least qualitative agreements have been obtained by several authors[274,275].
By applying one of the theoretical models to the MOFs, the presence of both the short and the long
wavelength bend loss edges has been predicted, indicating that the bend loss minimum is located
around ~Λ/2[276]. Given the Λ scale in LMA-MOFs (~10µm), the transmission window of the
silica fibres lies on the short wavelength edge. Thus, it appears that the bend loss characteristics of
MOFs are likely to be worse than conventional fibres.
However, the refractive index profiles of the depressed-cladding fibres (or W-fibres)[277] have
been engineered such that desired dispersion characteristics at a certain wavelength range are
obtained while retaining low bend loss. Therefore, it is anticipated that LMA-MOFs can also be
engineered to obtain a large Aeff while the bend loss is reduced. To start such engineering, it is
important to quantify how the bend losses vary with the structure. Below, the bend loss of LMA-
MOFs are first characterised at a fixed wavelength (1550nm) and then compared with their
conventional counterparts. Then, the measurement of the bend loss spectra for different types of
LMA-MOFs is presented.
6.4.1. Bend losses at 1550nm
The bend losses were measured for one loop of the fibre on a flat bench, while maintaining as small
a fibre tension as possible. The fibre loop was positioned using pins on a well defined circle that
determined the bend radius. The bend loss characteristics of a selection of LMA-0001 fibres and
the conventional fibres are shown in Fig.6.4.1. The measurement was performed at 1550nm. The
bend losses increase with the mode area, as expected, and it is understood that LMA-MOFs (H2: �
� � ��
~126µm2, H4: �
� � ��
~378µm2) possess almost comparable bend loss characteristics to their
conventional counterparts (S2: �
� � ��
~126µm2, S1: �
� � ��
~405µm2) with the similar values of Aeff.
Chapter.6 Large mode area microstructured optical fibres
143
Fig. 6.4.1 Normalised transmission intensity versus the bend radius. (Courtesy J.C.Baggett)
The curve may be not smooth since there is a portion of power that can be coupled back from the
cladding or the radiation modes. The coupling can be rather strong for some angular orientations
due to partial reflections at the air-silica interface. For this reason, the bend losses of LMAHF-0001
(see Fig.6.2.1) displayed clear orientation dependence. However, this effect can, in turn,
immediately be used for improving the bend losses of MOFs. Hoping to confirm the improvement
of the bend loss characteristics, LMAHF-0102 was fabricated as shown in Fig.6.4.2. However, it
was found to be difficult to strip off the cladding modes in this fibre, despite the deliberately thick
bridges in the outer cladding. Note that there is another index difference between the holey
cladding and the jacket of LMAHF-0001 although this is typically small (NA~0.04) since the
fabrication of this fibre was based upon the second generation, as described in 6.2.2.
The orientation dependent variations of the bend losses were also observed for LMAHF-0100 (see
Fig.6.2.3). Furthermore, it has been found that there is a correlation between the extent of the
cladding (the distance between the core and the outer most air hole) and the bend losses[270]. This
suggests that it is important to increase the large cladding dimensions further.
Fig. 6.4.2 The SEM photograph of LMAHF-0102.
Chapter.6 Large mode area microstructured optical fibres
144
Fig.6.4.3 shows the relation between �� � �� and the critical bend radii Rcr, which the author here
defines as the radius at which the power reduces by half the initial power observed from a straight
fibre, for different types of LMA-MOFs. Although the critical bend radii are almost the same for
all the fibres with relatively small �� � ��, LMAHF-0100 exhibits a sharp increase with �� � ��
whilst
those of LMAHF-0001 and the conventional fibres increase almost linearly. These different trends
imply the difference in the bend loss mechanisms involved in different fibres. Recalling that
LMAHF-0001 is a hybrid fibre, in which there is a small index difference between the core and
cladding (∆n~2x10-4), the bend losses at this wavelength may be dominated by the long wavelength
edge, resulting in the similar trend to the conventional fibres. On the other hand, LMAHF-0100 is a
single material fibre that is likely to suffer from the short wavelength bend loss edge. Thus, it is
understood that the small index difference between the core and the cladding helps to prevent the
onset of the short wavelength edge at this wavelength.
Fig. 6.4.3 Relation between the Aeff and Rcr for different fibres (at 1550nm).
6.4.2. Wavelength dependence of the bend losses
The measurement was performed using a white light source, where the dynamic range of the
measurement varies depending on the wavelength. Because of the lack of source power at short
wavelengths, the obtainable intensity contrast dropped (almost exponentially) from ~20dB at
800nm to just 2dB at 400nm. The bend was applied by using several bobbins with different
diameters. The author only discusses qualitative characteristics, which are always observable for
any orientation of the fibre, by neglecting the orientation dependence.
Examples of the bend loss spectra for LMAHF-0001 are shown in Fig.6.4.4. The peaks at 700nm
(a) and at 550nm (b) correspond to the cut-off of the higher order modes. This demonstrates that
Chapter.6 Large mode area microstructured optical fibres
145
these fibres are no longer endlessly single moded due to the small index difference between the
core and the cladding although the dimensions of the small air holes within the structures are small.
Notice that the tolerable bend diameters are very different for (a) ~3.3cm and (b) ~15cm due to
their different mode areas. For a relatively small core fibre (H2, �� � ��~126µm2), significant losses
(>10dB) occurred when the bend radius was 1.65cm and the losses suddenly increased toward
shorter wavelengths. The relative transmission loss around 1550nm was less than 2dB, clearly
indicating the onset of the short wavelength bend loss edge. The many peaks between 800 and
1200nm were possibly induced by the irregular cladding structure since the bend losses are
sensitively dependent on the positions and the dimensions of the air holes. In fact, these peaks
showed orientation dependence.
Fig. 6.4.4 Bend loss spectrum of LMA-0001 for the different bend diameters (units in cm). (a) H2, and (b) Η5. The peak losses between 500~800nm occur due to the leakage of the higher order modes.
On the other hand, for relatively large core fibres (H5, �� � ��~305µm2) in Fig.6.4.4 (b), the long
wavelength spectral components clearly show larger losses even without a tight bend. The spectral
interference in the bend loss spectrum is also similar to that observed in conventional fibres[275].
The short wavelength bend loss edge is completely negligible in this case. This can be understood
by considering the relative contribution to the NA of this hybrid fibre, as described below.
The NA between F300 and HLQ-210 is ~0.02, which may be regarded as approximately constant
over the wavelength range of interest. By assuming Λ to be the core diameter (the F300 region),
NA~0.08 is calculated from the cut-off wavelength of ~900nm for H2, and is a factor of four higher
than that of the NA due to the material index difference. This implies that the dominant guiding
mechanism is provided by the air holes. On the other hand, for H5, the NA is calculated to be
~0.047 at 700nm. Thus, the NA created by the air holes should be comparable to the NA due to the
material index difference. Note that the relative contribution of the air holes to the total NA can be
By considering the relative contribution to the NA term in eq.(7.1), the following condition can be
used to define the regime where the holey cladding dominates the guidance resulting from the high
index doped core: @ A
BC D C EGFEGDBC D C EGF�HHHH
IJJ KLMN . (7.2)
Therefore, in addition to the single mode condition (V<2.405), the above equation should be
satisfied. In the region beneath the dashed line in Fig.7.3.2, the single mode operation is achieved,
whilst structures that satisfy eq.(7.2) lie under the solid line. Here, Λ=10µm and λ=1.06µm are
assumed. As d/Λ is increased, the possible range of values for ∆n of the doped section increases
Chapter.7 A ytterbium-doped all-glass double-clad large mode area microstructured fibre 160
due to the strong effect of the holey cladding however the fibre tends towards multi-mode
operation. The fabricated fibre corresponds to the diamond symbol in Fig.7.3.2. Given the fact that
the actual doped section is commonly ~30% of the hexagonal region defined by the inner most air
holes, both curves are underestimated. Therefore, both of these two conditions are well satisfied in
Yb-LMAHF-00.
Fig. 7.3.2 Optimum parameter regime for the doped LMA-MOF (Λ=10µm, λ=1.06µm). The impact of the doped core is relatively small compared to that of the holey cladding for points under the solid line eq.(7.2) whilst robust single mode operation is achieved under the dashed line eq.(7.1).
B) Effective mode area
The effective mode area of Yb-LMAHF-00 was characterised by using the mode field diameter
(MFD) measurement described in Chapter 6. A Nd:YLF laser operating at 1047nm was used.
Fig.7.3.3 shows the experimental plot, from which MFD=12.3µm ( �� � ��=119µm2) was estimated.
The measurement was repeated by rotating the fibre at the output facet and repeating the
measurement >10 times. The error bars correspond to the data range (e.g. maximum and the
minimum values). The large error suggests that the beam is slightly elliptical, as observed in the
following section, although it is to be appreciated that any light associated with excited cladding
modes would also result in an underestimate of the mode-area due to the increased divergence
associated with these modes.
It has been shown that the extent of the cladding impacts the macroscopic bend losses[270].
Therefore, it is fair to consider whether the effective mode area might also be influenced by the
outer cladding structure. Here, we show the measured �� � �� values for LMA-HF-0100 and 0101
(see Fig.6.2.3) with different structural dimensions as given in Table.7.3.1, in order to compare the
effective mode area of undoped LMA-MOFs and undoped double clad LMA-MOFs.
It can be seen that at Λ~7.6µm, � ! !" of the double clad fibre is comparable to that of the single
clad fibre. However, at Λ~10µm, #$ % %" of the single clad fibre drastically increases to a value of
more than 200µm2, whilst that of the double clad fibre is still ~140µm2. Thus, an increase of #$ % %"
is effectively suppressed by the presence of the outer cladding. The measured #$ % %" value for
Yb-LMAHF-00 thus seems reasonable by taking into account the combined action between the
high index doped section and the holey outer cladding.
C) Inner cladding NA
The inner cladding NA can also be roughly estimated by the effective index model to be & ' & '(((*),+),+),+ -.-./ " 0102 33 44 , where nocl is the effective index of the outer cladding as a
function of its hole diameter do and spacing Λo. Note that in reality, the extent of both inner and
outer cladding can affect the NA, but these factors are omitted here as a first order approximation.
Due to the fact that the effective index of the inner cladding is lower than that of silica, the inner
cladding NA of this fibre type is slightly more restricted relative to the case of the air-clad MOFs
Chapter.7 A ytterbium-doped all-glass double-clad large mode area microstructured fibre 162
that are described in Chapter.5 (i.e. the lower the fraction of air within the inner cladding the higher
the inner cladding NA that can be achieved).
Fig.7.3.4 shows the effective inner cladding NA for the case Λ=10µm and d/Λ=0.3 for the inner
cladding at 915nm. The outer cladding hole pitch is also assumed to be Λo=10µm as an ideal case.
For do/Λo=0.8, NAcl is only ~0.19. Given the deformation involved in Yb-LMAHF-00, it is hard to
predict the actual value since the NA is sensitive to both Λo and do/Λo. However, it is understood
that high d/Λ (>0.9) is essential to be comparable to the conventional cladding pumped fibre with a
polymer coating. One possible approach to improve the value of NAcl is to use smaller capillaries in
the outer cladding than in the inner cladding since the effective index of the outer cladding can be
reduced.
The effect of the index difference between Suprasil® F300 and Vycor® was found to be so small
within the parameter range that there is no actual difference compared with the case in which
Suprasil® F300 is used for both inner and outer cladding. Thus, the use of Suprasil® F300
capillaries should result in higher NA since the deformation of the outer cladding can be prevented.
Fig. 7.3.4 Possible inner cladding NA with respect to the outer cladding structural parameter do/Λo. (Λ=Λo=10µm, d/Λ=0.3, and the wavelength of 915nm is assumed.)
In Chapter 5, it is shown that do/Λo~0.9 is readily achievable in practice within the air clad MOFs.
Therefore, even when a holey cladding is used for both the outer and inner cladding, it is expected
that an inner cladding NA of more than 0.3 can be obtained, although there will be a trade-off with
respect to a corresponding decrease in the effective mode area. This trade-off can be virtually
eliminated by increasing the inner cladding dimensions. However, since the doped section within
this fibre type is relatively small compared with the conventional fibres, the device length is
elongated, which is not a preferred option for this kind of cladding structure since the effective
inner cladding cladding NA is length dependent[253]. Therefore, improving the core structure is a
key issue for the future so that a large doped section can be incorporated.
Chapter.7 A ytterbium-doped all-glass double-clad large mode area microstructured fibre 163
7.3.2. Absorption and background losses
The absorption of Yb-LMAHF-00 was measured by using the cut-back technique with a white light
source launched into the fibre inner cladding. The ytterbium absorption peak was measured to be
~2.0dB/m at 976nm and ~1.0dB/m at 910nm, respectively, using ~5m of the fibre on an 8cm radius
bobbin. Compared with the absorption cross section data[296] in germanosilicate, the 976nm peak
height is lower, indicating that a fraction of power is not interacting with the core. By slightly
decreasing the bend radius to 7.5cm, the absorption was slightly enhanced to ~3dB/m at 976m and
~1.25dB/m at 915nm.
The area ratio of the inner cladding to the core is readily estimated to be ~3n(n+1)(1-(d/Λ)2)=82,
where n is the number of the rings. Therefore, the dopant concentration can be estimated to be
1900~2900ppm (by weight) for 976nm and 3500~4500ppm for 915nm, respectively, by assuming
the cross section values. Given that the conventional fibres, made using the same ytterbium
concentration in the solution, exhibited ~3000ppm these values seem reasonable. The slightly
higher value at 915nm is presumably due to the increased modal number (~V2/2)[252] and to the
fact that the majority of the extra modes at 915nm overlap well with the core. This in turn implies
that some cladding modes rarely interact with the core.
In fact, the pump absorption measurement along the length using a low brightness laser diode
source (at 915nm) can be best fitted by an exponential curve with a positive offset. This indicates
that there are still some non-absorbed modal components present in the fibre even at 915nm. Thus,
despite an irregular cladding boundary and an offset core, the pump interaction with the core in this
MOF is not as good as anticipated. However, given that the total absorption observed during the
white light measurement at 976nm was ~10dB on the 8cm bobbin with a length of ~8m, the
absorption characteristics resulting from the holey inner cladding seem acceptable from a device
perspective.
Although the author tried to characterise the background losses, it was practically difficult to
measure using the available length of 30m. The accuracy in detection of the white light
measurement is approximately 0.5dB, which restricted the measurement accuracy to ~15dB/km.
The measured loss around 1100nm was ~30dB/km. This is sufficiently low for the given device
lengths of ~10m. On the other hand, the author measured ~30ppm of OH absorption at 1380nm,
which ultimately limits the high power operation due to the energy transfer from the ytterbium ions
because of the spectral overlap at 940nm. The main contribution to the background loss was
primarily due to this OH contribution. It can be seen that there is an overtone peak at 1130nm
which ‘plugs’ the transmission minimum around this wavelength. The long wavelength (>1550nm)
losses are presumably due to the leakage at the outer cladding interface. Thus, by introducing the
Chapter.7 A ytterbium-doped all-glass double-clad large mode area microstructured fibre 164
dehydration process (see Chapter 3), and by increasing the number of rings in the outer cladding, a
further reduction in the pump propagation losses can be expected.
Fig. 7.3.5 Absorption spectra of Yb-LMAHF-00 for different bend radii (left) and the water absorption (right).
7.4. CW laser characteristics
7.4.1. Core pumping at 976nm
The author performed laser experiments using a single-mode Ti:sapphire laser at 976nm as the
pump laser. However, it was found to be difficult to efficiently launch the pump light into just the
fundamental mode of the core. As a result, the absorption measurement of the fibre showed an
absorption coefficient of 3.1dB/m with negligible bends and this is only a slightly higher value than
the value given by the white light measurement. This suggests that a large portion of the pump
power is used to excite the cladding modes. The value was relatively insensitive to the focusing
element (f=10mm~20mm). In fact, the diameter of the incident beam was ~5mm. Therefore, the
focusing beam NA was still ~0.1 when using an f=20mm lens. On the other hand, the effective NA
is estimated to be ~0.06 from Fig.7.3.1. Thus, the poor coupling efficiency to the fundamental
mode can be understood. The total coupling efficiency of the pump was measured using a 5cm
piece of the fibre and was 70%.
The laser cavity was formed by a high reflector and 4% Fresnel reflection from the cleaved end of
the fibre, to which the pump laser was coupled. The output power was extracted as a reflection
from a dichroic mirror (HR:1030nm, HT:980nm). Using a 4m length of the fibre, we recorded 82%
slope efficiency with respect to the absorbed pump power, as shown in Fig.7.4.1. Despite the
relatively long length of the fibre, the spectral peaks of the free running laser were around 1040nm,
indicating relatively weak re-absorption. The laser threshold pump power was ~360mW and a
maximum power of ~580mW was obtained.
Chapter.7 A ytterbium-doped all-glass double-clad large mode area microstructured fibre 165
7.4.2. Cladding pumping at 915nm
A pure cladding pumped laser was demonstrated using a low brightness pump laser diode operating
at 915nm and that had a fibre-pigtailed output with ~100µm spot and 0.22 NA. The pump beam
was imaged on to the fibre end using a 1x telescope. The coupling efficiency was measured to be
65% again using a 5cm piece of the fibre. The slight degradation of the launch efficiency can be
attributed to the higher NA of the pump beam and possibly to the uniform illumination of the pump
beam over the air holes, and which can cause random scattering at the launch interface. The cavity
arrangement was otherwise exactly the same as the one employed for the core pumping
experiments.
The bend was carefully imposed so that no significant leakage of the fundamental mode occurred
while enhancing the pump absorption. It was possible to reduce the bend radius to a value as small
as 7.5cm without significant modal leakage, while enhancing the pump absorption.
The highest slope efficiency was obtained using an 8.5m length of the fibre, where an output power
well in excess of 1W was obtained. Since the absorption cross section is approximately 2.5 times
smaller at 915nm than 976nm, the optimum absorption length should be 2.5 times longer. The
shorter than expected length can be explained by the enhancement of the pump absorption due to
the bend. The slope efficiency and the pump power threshold were calculated to be 70% and
200mW with respect to the absorbed pump power as shown in Fig.7.4.1. The reduction of the slope
efficiency compared with 976nm pumping is primarily because of the different quantum limit
efficiencies (~85% for 915nm pump and 1080nm lasing, ~94% for 976nm pump and 1040nm
lasing). This, in turn, resulted in a smaller pump threshold using the 915nm pump since the laser
operation here is more similar to a four level system than when using the 976nm pump.
The modal profile of the output was taken using an IR camera, and single mode guidance of the
fibre was confirmed, as shown in the inset of Fig.7.4.1. The mode is slightly elliptical although the
hexagonal shape can be deduced from the tail of the modal profile. This is consistent with the MFD
measurement, in which a relatively large variation in angular orientation was observed. The reason
behind this is possibly due to the deformation of the doped section through the fabrication, which
can be induced by the thin silica layer that remains on the surface of the etched preform. The
difference in silica layer thickness can lead to a variation in a compressive force imposed on the
doped section during the drawing. To verify this argument, a SNOM measurement of the near field
modal profile[271] would be necessary.
Chapter.7 A ytterbium-doped all-glass double-clad large mode area microstructured fibre 166
Fig. 7.4.1 CW laser output power characteristics of Yb-LMAHF-00. The circular dots: core pumping at 976nm using 4m of the fibre. The square dots: cladding pumping at 915nm using 8.5m of the fibre. Inset: the measured output modal profile in the cladding pumped configuration.
7.5. Pulsed laser characteristics
7.5.1. Q-switching
The development of conventional LMA fibres[262] has resulted in major advances in the
performance of high energy pulsed fibre lasers due to the high energy storage capacities and
relatively large saturation energy within these fibres. Using multi-mode ytterbium doped
conventional LMA fibres, nearly single transverse mode, millijoule pulses have been
reported[265,297,298]. The author constructed a Q-switched laser using Yb-LMAHF-00 since its
performance under Q-switched operation provides a direct measure of the energy storage capability
of the fibre. Here, it is shown that despite the relatively small doped area within the current fibre it
is possible to obtain pulses with ~50µJ pulse energies with a peak power of ~1kW.
The extractable pulse energy in a Q-switched fibre laser is roughly determined by the difference
between the stored energy and the energy at which the fibre gain becomes positive to overcome the
signal absorption due to the propagation along the fibre (e.g. a bleaching level)[297]. Note that both
of these quantities are proportional to the doped area. The bleaching level is also proportional to the
saturation energy and the fibre length.
Decreasing the bleaching level in principle provides one way of increasing the extractable energy
as long as either the excitation density is low or the pump is injected within a period shorter than
Fig. 7.5.5 The tuning curve obtained from the mode-locked laser using Yb-LMAHF-00.
An average power in excess of 500mW was obtained from 1030nm to 1090nm with pulse at the
fundamental repetition rate. (Note that the orientation of the wave-plate had to be changed to
recover the mode-locking operation when the wavelength was changed substantially because of the
wavelength dependence of the retardation through the waveplate). Note that the corresponding
pulse energy is more than 40nJ. By employing a more appropriate mode-locking for short pulse
generation and incorporation of suitable dispersion compensation elements within the cavity it
should be possible to obtain reasonably short pulse trains with pulse energies comparable to those
achievable using bulk mode-locked lasers [311].
7.6. Discussion
It has been shown that the novel cladding pumped fibre, that the author has presented herein,
displays comparable CW laser performance to its conventional counterpart, and demonstrates the
high quality of the fabricated MCVD preform and the quality of our MOF fabrication process. This
ensures that the doped core of the preform retains its performance. However, it will be necessary to
improve the structural design (in particular the inner cladding NA) if this technological approach is
to find successful applications.
In terms of modal properties, the limitation of obtaining larger effective mode area arises from both
the high index doped section and the impact of the outer-cladding. The latter may be solved by
Chapter.7 A ytterbium-doped all-glass double-clad large mode area microstructured fibre 173
increasing the inner cladding dimensions, although this will reduce the net pump absorption
because of the reduced core clad area ratio and the corresponding increase in bend sensitivity.
Furthermore, in Q-switching experiments, the major obstacle for the scaling of the output pulse
energy was primarily identified as the limited doped area in the current fibre.
In order to enlarge the doped section within LMA-MOFs, the index boundary should be placed
outside the MOF core, for instance, between the first and second rings of air holes (see Fig.7.6.1).
By doing this, the effect of the finite refractive index difference between the core and cladding
glasses is different to the current case (dotted line). The impact of the outer cladding on the
effective area will also be different and may slightly modify some of the details in terms of trade
off of mode area versus bend loss, and parameters for which more robust single mode guidance is
achieved. In order to fabricate such structures a fabrication technique that allows us to make large
scale rods of doped glasses is necessary. Thus, the key issues relate to the fabrication of large scale
pieces of doped glasses with an accurately controlled refractive index.
Fig. 7.6.1 The possible area of the doped section with respect to the hexagonal arrangement of air holes (the dotted line corresponds to the one in Yb-LMA-00.)
In terms of controlling the refractive index, it is clear that it will be difficult in practice to reduce
the aluminium content since the dopant concentration of the cladding pumped fibre has to be
reasonably high to reduce the device length. Therefore, incorporation of another index decreasing
element such as boron or fluorine is likely to be required. Fortunately, boron is trivalent and thus
has good characteristics for the incorporation of ytterbium ions[312,313] although fluorine
incorporation through MCVD is difficult since freon can no longer be used. Thus, optimising the
balance between the aluminium and boron, the refractive index of the doped core can be matched
to that of Suprasil® F300 while a sufficiently high ytterbium concentration for cladding pumped
fibres is preserved. Although the boron doped preforms are stressed in general, the boron required
to compensate for the positive contribution from both aluminium and ytterbium should be so small
that the core extraction by polishing is unlikely to be an issue (see Chapter 4 and 5). The author
therefore concludes that there is a possibility for obtaining matched doped glasses without
compromising the dopant concentration.
Chapter.7 A ytterbium-doped all-glass double-clad large mode area microstructured fibre 174
In terms of the dimensions of the doped glasses, MCVD preforms suffer from the disadvantages of
the solution doping technique that relies upon the deposition of a thin soot layer. Historically, it is
well known that a heat source must be separated from the substrate to deposit thick soot as within
OVD, VAD, and PCVD. However, the author restricts this discussion to the possibilities of using
MCVD systems below.
There are two possible approaches to overcome this issue. One is to increase the thickness of the
silica frit that is to be solution doped. However, it is difficult to obtain a uniform and thick silica
frit because of its poor thermal conductivity. When the deposition temperature is too high, the frit
near the substrate tube becomes insufficiently porous to incorporate the rare-earth ions from the
solution. On the other hand, when the temperature is too low, the inner frit becomes too porous to
attach it to the surface, preventing further deposition. Therefore, thermally conductive gas (such as
helium) may be used during the soot deposition process. Alternatively, water/air cooling of the
deposition tube[314] can help since a large temperature gradient can be created in front of the
deposition burner increasing the thermophoretic force[315]. Another approach is to perform
solution doping many times by repeating the process. However, given that a single step of the
doping process typically takes ~3hours, it is therefore impractical to perform multiple solution
doping stages since the perform core dimensions realised by a single step are typically just
2~3mm2.
Given these difficulties in the MCVD method, it seems that the synthesis of the bulk doped silica
glasses is ultimately needed for the future improvement of the doped LMA-MOFs although the
refractive index control will become more challenging than in the MCVD preform fabrication as a
result. One possible approach is to incorporate organometallic materials as raw materials[316],
which allows for continuous deposition of aluminosilicate and rare-earth incorporation via aerosol
formation[317]. Therefore, it is anticipated that routes forward will be found with MCVD
technique.
7.7. Conclusions
The author has fabricated an efficient all-glass double-clad ytterbium doped LMA-MOF fibre using
the core extracted from a low NA MCVD preform. It was demonstrated that the use of a low NA
doped section is essential, and that LMA-MOFs offer unique opportunities within double clad
structures.
It has been shown that, using the effective index model, it is possible to roughly predict the
optimum structural parameters to preserve single mode guidance while taking advantage of the
holey cladding, and that the V value of the doped LMA-MOF is flattened, allowing for a robust
Chapter.7 A ytterbium-doped all-glass double-clad large mode area microstructured fibre 175
single mode operation over a wide spectral range. This implies interesting opportunities for core
pumped variants using suitable glass materials.
In laser experiments, a slope efficiency of 80% with respect to the absorbed pump power at 976nm
was obtained for continuous wave laser operation. A corresponding value of 70% was obtained
using a low brightness laser diode at 915nm. The modal profile of the laser output was also
confirmed to be single mode.
Furthermore, both Q-switching and mode-locking operations were demonstrated using the doped
LMA-MOF. In Q-switching experiments, it has been shown that the peak output power obtained of
~1kW was limited by the lasing of ASE components due to the relatively small doped area. In
mode-locking experiments a broad tunability of over 70nm was demonstrated within the
fundamental mode-locking regime.
Finally, the future directions and possibilities for doped LMA-MOF fabrication were discussed.
When the existing fabrication issues are solved, it will open up a host of new opportunities for the
use of doped LMA-MOFs.
Chapter.8
Conclusions and future directions
This thesis has described the fabrication, characterisation, and applications of a broad range of
microstructured optical fibres (MOFs) with a particular emphasis on incorporating rare-earth ions.
Chapter 2: The fabrication related issues were studied. First, a mathematical model of the capillary
drawing process was described, which forms a basic building block for MOF fabrication. Its
experimental verification was carried out and good agreement was obtained.
Second, general guidelines for preform preparation were provided. It was shown that there are two
types of preforms: a single material preform with sealed capillaries or a combination of silica
capillaries and a Vycor® jacket tube since different collapse ratios imposed on the capillaries and
the jacket tube must be compensated either by pressure or viscosity. Practical issues such as
cleaning of the elements and the preform dimensions are also discussed.
Third, caning and fibre drawing processes were discussed. It was found that the capillary drawing
model also provides useful insights for all these drawing processes. General criteria as to how the
drawing parameters can be allotted for the case of the two step drawing approach were discussed. It
was found that reducing the tension of the fibre is a key for the successful drawing of small scale
structures. Precise internal pressure control of the preforms was found to be important for
controlling the final structures in the fibres.
Chapter 3: Investigations into the loss properties of highly nonlinear MOFs (HNL-MOFs) were
presented. The first part focused upon the possible loss mechanisms involved in these fibres.
Forward, backward, and homogeneous scattering mechanisms were individually discussed and the
relevant perturbation scales were identified.
The second part described experimental measurements of losses. Using the cut-back method, the
homogeneous scattering losses were identified as a main contributor to the total losses of these
Chapter.8 Conclusions and future directions 177
fibres since the wavelength dependence of the losses shows λ-2 dependence at long wavelength
range (λ~1.5µm), which agrees with a simple model where the scattering elements are assumed to
be homogeneously distributed within the cladding. Furthermore, significant dependence of the
losses on the structural dimensions of the fibres indicates that the effective Rayleigh scattering
coefficient of holey cladding is more than 1000 times greater than that of conventional fibres. This
indicates that surface treatment during the fabrication process must be improved for these fibre
designs. A comparison with attenuation and backscattering factors within the OTDR trace implies
that there are also substantial radiation losses. To confirm this conclusion, , further details, in
particular for fibres with different d/Λ, need to be studied by improving the backscattering
measurement.
Finally, continuous efforts to reduce OH incorporation in HNL-MOFs were described. Technique
for dehydrating the capillary stack preforms was examined and the concentration of OH ions was
reduced to a few ppm level. Although it was found to be effective to etch off the surface layer that
contains OH ions, there is a trade-off with the increased background losses due to the degradation
of the silica surface. By minimising the exposure to the atmosphere, combined with the developed
dehydration process, an OH concentration as low as 1ppm can be anticipated in HNL-MOFs, even
when the two-step drawing approach is used. For a further reduction of the OH incorporation the
exposure to the atmosphere must be prevented during the capillary drawing and caning.
Chapter 4: rare-earth doped MOFs with small scale cores, thus with high nonlinearity, were
developed and three device demonstrations were carried out using them.
First, the mode-locked operation of an ytterbium doped MOF was presented. By employing the
frequency shift feedback technique, a mode-locking threshold of 17mW and tunability over 20nm
was obtained. The mode-locking threshold was higher than anticipated. This was attributed to end
facet reflection due to the extremely high NA design of the fibre. In order to fullly take advantage
of anomalous dispersion of the fibre in an oscillator, a lower NA design is required but maintains
the anomalous dispersion.
Second, the ytterbium doped MOF was incorporated into an amplifier configuration using a fibre
based ultrashort pulse source as a seed. Taking advantage of the broad anomalous dispersion
regime of the fibre, mono-colour tunable Raman soliton generation was demonstrated covering the
1.06µm~1.33 µm range. Further complex regimes including multi-colour soliton generation up to
1.58µm and even supercontinuum generation were observed by increasing the seed/pump power.
The fibre lengths used for these experiments were less than 10m and seed pulse energy was less
than 10 picojoules. These facts demonstrate the possibility of realising truly compact, widely
tunable ultrashort pulse fibre sources using a highly nonlinear ytterbium doped MOF. For further
Chapter.8 Conclusions and future directions 178
improvements, a more detailed understanding of the system is required. In addition, optimisation of
fibre design in terms of effective mode area and dispersion characteristics as well as the fibre
fabrication (i.e. low OH content fibres) are necessary. By overcoming these issues, an all-fibre
femtosecond source, which is continuously tunable from 1µm to 2µm, could be realised.
Finally, the continuous wave operation of an erbium doped MOF laser was presented, resulting in a
pump power threshold of 0.55mW and a slope efficiency of 57.3%. This demonstrates that it now
is possible to fabricate rare-earth doped MOFs with improved performance relative to the fibres
drawn from the original MCVD preform. It was found to be difficult to excite the single
polarisation axis within this fibre despite the high birefringence of these fibre types. These
polarisation issues need to be addressed further for use within a variety of devices.
Chapter 5: Development of the novel cladding pumped fibres; air clad MOFs, which incorporate a
high fraction of air within the outer cladding whilst possessing the conventional core and inner
cladding, were described. This allows one to realise a high NA and small inner cladding at the
same time, which had been difficult to realise using the existing cladding pumped fibres.
Improved laser performance was demonstrated by presenting two examples. A broad tunability
over 110nm was achieved using only 1.7m of the fabricated ytterbium doped air clad MOF, in
cladding pumped configuration with pump feedback. By selecting the wavelength via a bulk
grating, efficient pure three level operation of cladding pumped ytterbium doped fibre at 980nm
was also achieved. The output exceeded 3.5W by employing high quality laser diodes. Furthermore,
a photosensitive air-clad MOF was developed and was used to investigate the impact of air-silica
interfaces during the grating fabrication process. It was found that it is possible to inscribe gratings
within the simplest form of the air clad structure, which contains a single layer of air holes. A
further study is necessary for understanding the effects of the air-silica interface for UV writing. By
overcoming these issues, truly integrated CPFLs with extended performance can be realised.
Chapter 6: Passive large-mode-area MOFs (LMA-MOFs) were fabricated and their optical
properties were studied in terms of effective mode area and bend losses.
Two different approaches to characterise the effective mode areas were examined, and effective
mode areas as large as 680µm2 was measured using the knife-edge method.
It was found that when different silica materials are used in the core and the cladding, the bend loss
characteristics are qualitatively modified and this was qualitatively understood by considering the
relative contribution to the total NA, that can be estimated from the observed cut-off wavelengths.
It was shown that single material fibres suffer from a short wavelength bend loss edge and this
Chapter.8 Conclusions and future directions 179
results in worse bend loss characteristics for fibres with d/Λ~0.25 at 1550nm as Aeff is increased.
Comparable bend loss performance can be obtained by using slightly higher index material within
the core. However, this is accompanied by the onset of the multimode regime in visible
wavelengths. By exploring different regimes (near the multimode regime, i.e. d/Λ~0.4), the bend
loss performance is anticipated to be improved. The flexibility of the cladding design of
LMA-MOFs offers further research opportunities for better bend loss performance with greater
effective mode area (see ref.[318]).
Finally, the propagation losses of LMA-MOFs with different structures have been characterised.
The loss at 1550nm varied from 1~10dB/km for fibres with d/Λ>0.4. From the loss spectra, it was
found that the longitudinal uniformities of these fibres are very important because low NA fibres
are prone to suffer from radiation losses when any perturbations are imposed on the fibres.
Chapter 7: A novel cladding pumped ytterbium doped LMA-MOF, which uses different sizes of air
holes to define the inner and the outer cladding, was developed and a range of laser operations were
demonstrated.
By using the low NA MCVD preform, it was possible to retain the single mode output, which was
confirmed during the laser experiments. In continuous wave operation, slope efficiencies of
70~80% were obtained, depending on the pump wavelength, which were comparable to the
conventional counterpart. In cladding pumped operation, an output power in excess of 1W was
obtained. In Q-switching operation, a peak power of ~1kW and a pulse energy of ~50µJ was
obtained. In mode-locking operation, a wide tunability of 70nm was obtained with a fundamental
repetition rate with an average power of ~500mW corresponding to a pulse energy of >40nJ.
Finally, possibilities of improving energy storage capability of these fibres were discussed.
An output power up to 280W has been reported using this fibre type by improving the fibre
design[319]. Thus, it is anticipated that further improvement and refinement of the design will lead
to 1kW output from this fibre type near future.
Appendix.A The effective index model The effective index model is often used to analyse the fibres in this thesis. However, as explained,
this model is not accurate as a modal model when the modal index of the fundamental space filling
mode is calculated[12,13,19,23]. For instance, group velocity dispersion cannot be calculated using
this approach. The reasons for this are:
1. one must first calculate the effective cladding index neff, which may be incorrect, and
2. one must assume a value of the effective core radius aeff, which may be wavelength dependent
for a given structure.
Although the effective index model is not an accurate modal model, it can conveniently be used for
adapting the models for conventional fibres, as shown in Chapter 3. In other words, the effective
index model has a convenient interface to the other models, that can be used for analysing
transmission losses, for instance. Therefore, it is important to improve the accuracy of the effective
index model so that the physical quantities are consistent with the other accurate numerical
methods[14-31].
Below, it is shown that using the propagation constant β and the effective mode area Aeff, both of
which are calculated from the other numerical method, it is possible to determine modal parameters
within the effective index model without any ambiguities. Therefore, it is possible to retain the
consistency between the numerical model and the effective index model.
The dispersion relation of a perfectly circular step index fibre is given by[155]
� �
� � ��
��
����
�
�
� � �� � �
��� � ��� � �
� � ���� � ��� � �
�����
� ����
�� ����
��
��
����
�
� , (A.1)
where k is the wave vector, aeff the effective core radius, β the propagation constant. Here, unknown
variables are aeff and neff. Therefore, these values are dependent and their relationship can readily be
calculated.
Appendix.A Effective index model 181
The effective mode area is defined as[41]
������� �� � ��� �� � �� ����� � �����
� �������� � � � ���
�, (A.2)
where E is the electric field. Since the field distribution of the perfectly circular fibre is known, Aeff
can be calculated using a tentative combination of aeff and neff and then be compared with the value
obtained from the other modal method. Let the author define a minimisation function as
� ��� �� � �� � �� � �� � � �� �! "# , (A.3)
where Aeffo is the effective area obtained from the other model. By minimising f, it is possible to
determine a combination of aeff and neff, with which both β and Aeff are retained to be consistent with
the other modal model.
An example is shown below. The wavelength of 1.06µm and the structure with Λ=2.0µm and
d/Λ=0.4 are used, for which β~8.514852460143605x106 m-1 and $% & &� ~7.61325µm2 is calculated
using the localised function method[19]. In order to evaluate Aeff, the exact solution to the fields is
used and the numerical integration was performed using the adaptive recursive trapezoid
algorithm[316]. The integration range was varied by monitoring the convergence.
Fig.A.1 shows the relation between the core radius aeff and the effective cladding index neff, which
Fig.A. 1 Relation between aeff and neff that satisfies the dispersion relation (A.1) that gives the same propagation constant as that calculated from the orthogonal function method.
Appendix.A Effective index model 182
Using this relation, the function f is plotted with respect to aeff in Fig.A.2. A local minimum is
clearly seen, which corresponds to the consistent aeff (and thus neff).