Fo Research Advances

OPTICAL FIBERS
RESEARCH ADVANCES

OPTICAL FIBERS
RESEARCH ADVANCES

JRGEN C. SCHLESINGER
EDITOR

Nova Science Publishers, Inc.
New York

Copyright 2007 by Nova Science Publishers, Inc.

All rights reserved. No part of this book may be reproduced, stored in a retrieval system or
transmitted in any form or by any means: electronic, electrostatic, magnetic, tape, mechanical
photocopying, recording or otherwise without the written permission of the Publisher.

For permission to use material from this book please contact us:
Telephone 631-231-7269; Fax 631-231-8175
Web Site: http://www.novapublishers.com

NOTICE TO THE READER
The Publisher has taken reasonable care in the preparation of this book, but makes no expressed or
implied warranty of any kind and assumes no responsibility for any errors or omissions. No
liability is assumed for incidental or consequential damages in connection with or arising out of
information contained in this book. The Publisher shall not be liable for any special,
consequential, or exemplary damages resulting, in whole or in part, from the readers use of, or
reliance upon, this material. Any parts of this book based on government reports are so indicated
and copyright is claimed for those parts to the extent applicable to compilations of such works.

Independent verification should be sought for any data, advice or recommendations contained in
this book. In addition, no responsibility is assumed by the publisher for any injury and/or damage
to persons or property arising from any methods, products, instructions, ideas or otherwise
contained in this publication.

This publication is designed to provide accurate and authoritative information with regard to the
subject matter covered herein. It is sold with the clear understanding that the Publisher is not
engaged in rendering legal or any other professional services. If legal or any other expert
assistance is required, the services of a competent person should be sought. FROM A
DECLARATION OF PARTICIPANTS JOINTLY ADOPTED BY A COMMITTEE OF THE
AMERICAN BAR ASSOCIATION AND A COMMITTEE OF PUBLISHERS.

LIBRARY OF CONGRESS CATALOGING-IN-PUBLICATION DATA
Schlesinger, Jrgen C.
Optical fibers research advances / Jrgen C. Schlesinger, Editor.
p. cm.
Includes index.
ISBN-13: 978-1-60692-607-9

1. Optical communications. 2. Fiber optics. 3. Optical fibers. I. Title.
TK5103.59.S35 2008
621.36'92--dc22 2007031168

Published by Nova Science Publishers, Inc. New York

CONTENTS

Preface vii

Short Communication 1

Ignition with Optical Fiber Coupled Laser Diode 3
Shi-biao Xiang, Xu Xiang , Wei-huan Ji and Chang-gen Feng

Research and Review Studies 13

Chapter 1 Evanescent Field Tapered Fiber Optic Biosensors (TFOBS):
Fabrication, Antibody Immobilization and Detection
15
Angela Leung, P. Mohana Shankar and Raj Mutharasan

Chapter 2 New Challenges in Raman Amplification
for Fiber Communication Systems
51
P.S. Andr, A.N. Pinto, A.L.J. Teixeira, B. Neto, S. Stevan Jr.,
Donato Sperti, F. da Rocha, Micaela Bernardo, J.L. Pinto,
Meire Fugihara, Ana Rocha and M. Faco

Chapter 3 Fiber Bragg Gratings in High Birefringence Optical Fibers 83
Rogrio N. Nogueira, Ilda Abe and Hypolito J. Kalinowski

Chapter 4 Applications of Hollow Optical Fibers in Atom Optics 119
Heung-Ryoul Noh and Wonho Jhe

Chapter 5 Advances in Physical Modeling of Ring Lasers 161
Vittorio M.N. Passaro and Francesco De Leonardis

Chapter 6 Investigation of Optical Power Budget
of Erbium-Doped Fiber
187
Hideaki Hayashi, Setsuhisa Tanabe and Naoki Sugimoto

Chapter 7 Recent Developments in All-Fibre Devices
for Optical Networks
205
Nawfel Azami and Suzanne Lacroix

Contents vi
Chapter 8 Advances in Optical Differential Phase Shift Keying and Proposal
for an Alternative Receiving Scheme for Optical Differential
Octal Phase Shift Keying
231
M. Sathish Kumar, Hosung Yoon and Namkyoo Park

Chapter 9 A New Generation of Polymer Optical Fibers 257
Rong-Jin Yu and Xiang-Jun Chen

Chapter 10 Dissipative Solitons in Optical Fiber Systems 279
Mrio F.S. Ferreira and Sofia C.V. Latas

Chapter 11 Bright - Dark and Double - Humped Pulses in Averaged,
Dispersion Managed Optical Fiber Systems
301
K.W. Chow and K. Nakkeeran

Chapter 12 Dynamics and Interactions of Gap Solitons in Hollow Core
Photonic Crystal Fibers
315
Javid Atai and D. Royston Neill

Chapter 13 Multiwavelength Optical Fiber Lasers
and Semiconductor Optical Amplifier
Ring Lasers
335
Byoungho Lee and Ilyong Yoon

Chapter 14 Aging and Reliability of Single-Mode Silica Optical Fibers 355
M. Poulain, R. El Abdi and I. Severin

Index 369

PREFACE

An optical fiber is a glass or plastic fiber designed to guide light along its length by
confining as much light as possible in a propagating form. In fibers with large core diameter,
the confinement is based on total internal reflection. In smaller diameter core fibers, (widely
used for most communication links longer than 200 meters) the confinement relies on
establishing a waveguide. Fiber optics is the overlap of applied science and engineering
concerned with such optical fibers. Optical fibers are widely used in fiber-optic
communication, which permits transmission over longer distances and at higher data rates
than other forms of wired and wireless communications. They are also used to form sensors,
and in a variety of other applications.
The term optical fiber covers a range of different designs including graded-index optical
fibers, step-index optical fibers, birefringent polarization-maintaining fibers and more
recently photonic crystal fibers, with the design and the wavelength of the light propagating
in the fiber dictating whether or not it will be multi-mode optical fiber or single-mode optical
fiber. Because of the mechanical properties of the more common glass optical fibers, special
methods of splicing fibers and of connecting them to other equipment are needed.
Manufacture of optical fibers is based on partially melting a chemically doped preform and
pulling the flowing material on a draw tower. Fibers are built into different kinds of cables
depending on how they will be used.
This new book presents the latest research in the field.
Optical fibers, an important and promising material, have attracted more and more
attention and extended their applications to various scientific and practical aspects. In the
short communication, the key role of fibers, as the carriers of information and energy in our
times, was briefly summarized. Afterwards, the configuration of fiber coupled laser diode
ignition system was elucidated as well as the advantages, developments and applications of
this technology. Furthermore, the energy-transmitting characteristics of single-mode fibers
and multi-mode ones and the key points of fiber-coupled technology were analyzed. In a
practical case, the effect of the diameters of core on laser ignition, from both theory and
experiments, was studied. The findings suggest that the smaller the diameters of core, the
lower the ignition threshold under the same laser power. That is to say, the ignition becomes
easier while using fibers with smaller core. Finally, the issue on selection of core was
clarified based on the consideration of both laser power density and the endurance of fibers.
Tapered Fiber Optic Biosensors (TFOBS) are sensors that operate based on fluctuations
in the evanescent field in the tapered region. In the laboratory, TFOBS are made by heat
Jrgen C. Schlesinger viii
pulling commercially-available single mode optical fibers. They have been investigated for
various applications, including measurement of physical characteristics (refractive index,
temperature, pressure, etc.), chemical concentrations, and biomolecule detection. In this
chapter, an up-to-date review of TFOBS research is provided, with emphasis on applications
in biosensing such as pathogen, proteins, and DNA detection. The physics of sensing and
optical behavior based on taper geometry is discussed. Methods of fabrication, antibody
immobilization, sample preparation, and detection from our laboratory are described. This
chapter presents results on the non-specific response, simulation, and detection of E.coli
O157:H7 and BSA. Chapter 1 will conclude with an analysis of the future direction of the
Tapered Fiber Optic Biosensors.
Raman fiber amplifiers (RFA) are among the most promising technologies in lightwave
systems. In recent years, Raman optical fiber ampliers have been widely investigated for
their advantageous features, namely the transmission fiber can be itself used as the gain media
reducing the overall noise figure and creating a lossless transmission media. The introduction
of RFA based on low cost technology will allow the consolidation of this amplification
technique and its use in future optical networks.
Chapter 2 reviews the challenges, achievements, and perspectives of Raman
amplification in optical communication systems. In Raman amplified systems, the signal
amplification is based on stimulated Raman scattering, thus the peak of the gain is shifted by
approximately 13.2 THz with respect to the pump signal frequency. The possibility of
combining many pumps centered on different wavelengths brings a flat gain in an ultra wide
bandwidth.
An initial physical description of the phenomenon is presented as well as the
mathematical formalism used to simulate the effect on optical fibers.
The review follows with one section describing the challenging developments in this
topic, such as using low cost pump lasers, in-fiber lasing, recurring to fiber Bragg grating
cavities or broadband incoherent pump sources and Raman amplification applied to coarse
wavelength multiplexed networks. Also, one of the major issues on Raman amplifier design,
which is the determination of pump powers in order to realize a specific gain will be
discussed. In terms of optimization, several solutions have been published recently, however,
some of them request extremely large computation time for every interaction, what precludes
it from finding an optimum solution or solve the semi-analytical rate equation under strong
simplifying assumptions, which results in substantial errors. An exhaustive study of the
optimization techniques will be presented.
This paper allows the reader to travel from the description of the phenomenon to the
results (experimental and numerical) that emphasize the potential applications of this
technology.
Fiber Bragg gratings (FBG) are a key element in optical communication devices and in
fiber sensors. This is mainly due to its intrinsic characteristics, which include low insertion
loss, passive operation and immunity to electromagnetic interferences. Basically a FBG is a
periodic modulation of the core refractive index formed by exposure of a photosensitive fiber
to a spatial pattern of ultraviolet light in the region of 244248 nm. The lengths of FBGs are
normally within the region of 120 mm. Usually a FBG operates as a narrow reflection filter,
where the central wavelength is directly proportional to the periodicity of the spatial
modulation and to the effective refractive index of the fiber. The production technology of
these devices is now in a mature state, which enables the design of gratings with custom-
Preface ix
made transfer functions, crucial for all-optical processing. Recently, some work has been
done in the application of FBG written in highly birefringent fibers (HiBi). Due to the
birefringence, the effective refractive index of the fiber will be different for the two
transversal modes of propagation. Therefore, the reflection spectrum of a FBG will be
different for each polarization. This unique property can be used for advanced optical
processing or advanced fiber sensing.
Chapter 3 will describe in detail this unique device. The chapter will also analyze the
device and demonstrate different applications that take advantage of its properties, like
multiparameter sensors, devices for optical communications or in the optimization of certain
architectures in optics communications systems.
A hollow optical fiber (HOF) has a lot of interesting applications in atom optics
experiments such as atom guiding and the generation of hollow laser beam (HLB). In this
article the authors present theoretical and experimental works on the use of hollow optical
fibers in atom optics. Chapter 4 is divided into two parts: One is devoted to the atom guide
using HOFs and the other describes the atom optics researches that utilizes laser lights
emanated from the HOF. First, the authors describe the electromagnetic fields inside the HOF
and characterize the electromagnetic modes diffracted from the HOF. Then they describe two
guiding schemes using red and blue detuned laser lights. Finally, they describe the various
relevant experiments using LP
01
or LP
11
modes such as the generation of HLB from the HOF,
funneling atoms using the diffracted fields, diffraction-limited dark laser spot, and a dipole
trap using LP
01
mode of the diffracted field from the HOF.
In Chapter 5, an overview on fiber ring lasers and III/V semiconductor integrated ring
lasers is presented. In particular, some aspects of mathematical modelling of these devices are
reviewed. In the first part of the chapter, the authors have focused our attention on the more
recent theoretical and experimental studies concerning fiber ring laser architectures. Then, a
complete quantum-mechanical model for integrated ring lasers is presented, including the
evaluation of all the involved physical parameters, such as self and cross saturation and
backscattering. Finally, the influence of sidewall roughness on either unidirectional or
bidirectional regime in multi-quantum-well III/V semiconductor ring lasers is demonstrated.
In Chapter 6, the authors investigated optical power budget of an erbium-doped fiber
(EDF). In addition to the output signal and amplified spontaneous emission (ASE) powers
from the fiber end, lateral spontaneous emissions and scattering laser powers in the EDF were
measured quantitatively by using an integrating sphere. Compared with the signal and ASE
powers, it was found that considerable powers were consumed by the laterally emitting lights.
As an optically undetected loss which limits power conversion efficiency (PCE) of the fiber
amplifier, the effect of nonradiative decay from the termination level of pump excited state
absorption (pump ESA) was estimated from decay rate analyses of the relevant levels. The
nonradiative loss was comparable to amplified signal power in the EDF when pumped with a
980 nm LD. Nonradiative decay following cooperative upconversion (CUP) process is also
discussed using rate equations analysis.
All-fibre components are essential components of optical networks systems.
Development of such devices is of great importance to allow network functions to be
performed in the glass of the optical fibre itself. Among of all fabrication techniques, the
Fused Fibre Biconical Taper (FBT) technique allows optical devices with high performances.
Although fibre devices are mainly based on the passive directional coupler basic structure,
research is made to design components that perform complex functionalities in today optical
Jrgen C. Schlesinger x
networks systems. Recent developments on all-fibre devices in network systems are
presented. Research is mainly focused on enhanced fabrication and stability of FBT
fabrication technique, passive thermal compensation for stable interferometer optical
structure, broadband spectral operation for multi-wavelength operations and new
interferometer designs. An overview of recent fused fibre devices for optical
telecommunications is presented to understand the main functionalities of these fibre devices.
The limiting factors are explained in Chapter 7, to understand challenges on fibre devices
development.
Optical Differential Phase Shift Keying (oDPSK) with delay interferometer based direct
detection receiver was proposed as an alternative for the conventional On-Off Keying (OOK)
modulation schemes. Compared to OOK, oDPSK was predicted to have a 3dB improvement
in performance due to its balanced detection receiver structure. It was also predicted that due
to the optical signal occupying all the symbol slots, unlike in OOK, symbol pattern dependent
fiber nonlinear effects will make less of an impact on long haul optical transmission schemes
based on oDPSK. Subsequent successful demonstrations of these positive attributes of
oDPSK resulted in active investigations into multilevel formats of oDPSK namely, optical
Differential Quadrature Phase Shift Keying (oDQPSK) and optical Differential Octal Phase
Shift Keying (oDOPSK). Significant developments in theoretical models of optically
amplified lightwave communication systems based on the Karhunen-Loeve Series Expansion
(KLSE) method assisted such investigations. In Chapter 8, the authors discuss some of the
recent advances in oDPSK and its multilevel formats that have been achieved such as
proposals for receiver schematics, theoretical analysis of receiver schematics, electronic
techniques to counter polarization mode dispersion induced penalties, and application of
coded modulation techniques. The chapter also proposes an alternative receiver schematic for
oDOPSK which can separately detect the three constituent bits from an oDOPSK symbol.
Chapter 9 describes the background to the development of Polymer Optical fibers
(POFs), discusses the optical and temperature resistant properties of polymers while
emphasizing the intrinsic high attenuation of them. The first generation of POFs which
consists of a solid-core surrounded by cladding and transmits light by total internal reflection,
is puzzled by the difficulty of high attenuation. Then, the method of using a specific structure
(i.e. hollow-core Bragg fiber) to solve the problem is presented. A new generation of POFs
based on the hollow-core Bragg fibers with cobweb-structured cladding can guide light with
low transmission loss and high bandwidth in the wavelength range of visible to terahertz
(THz ) radiation. Efficient hollow-core guiding for delivery of power laser radiation and solar
radiation can be achieved by replacing the traditional polymethylmethacrylate (PMMA) with
heat-resistant polymers. Lastly, this chapter concludes with a discussion of applications in
diverse areas.
Chapter 10 introduces the concept of dissipative solitons, which emerge as a result of a
double balance: between nonlinearity and dispersion and also between gain and loss. Such
dissipative solitons have many unique properties which differ from those of their conservative
counterparts and which make them similar to living things. The authors focus our discussion
on dissipative solitons in optical fiber systems, which can be described by the cubic-quintic
complex Ginzburg-Landau equation (CGLE). The conditions to have stable solutions of the
CGLE are discussed using the perturbation theory. Several exact analytical solutions, namely
in the form of fixed-amplitude and arbitrary-amplitude solitons, are presented. The numerical
solutions of the quintic CGLE include plain pulses, flat-top pulses, and composite pulses,
Preface xi
among others. The interaction between plain and composite pulses is analyzed using a two-
dimensional phase space. Stable bound states of both plain and composite pulses are found
when the phase difference between them is 2 / . The possibility of constructing
multisoliton solutions is also demonstrated.
As explained in Chapter 11, the envelope of the axial electric field in a dispersion
managed (DM) fiber system is governed by a nonlinear Schrdinger model. The group
velocity dispersion (GVD) varies periodically and thus realizes both the anomalous and
normal dispersion regimes. Kerr nonlinearity is assumed and a loss / gain mechanism will be
incorporated. Due to the big changes in the GVD parameter, the correspondingly large
variation in the quadratic phase chirp of the DM soliton will be identified. An averaging
procedure is applied. In many DM systems, an amplifier at the end of the dispersion map will
compensate for the energy dissipated in that map. Here the case of gain not exactly
compensating the loss is considered, in other words, a small residual amplification /
attenuation is permitted. The present model differs from other similar ones on variable
coefficient NLS, as the inhomogeneous features involve both time and the spatial coordinate.
The goal here is to extend the model further, by permitting coupled modes or additional
degree of freedom. More precisely, the coupling of fiber loss and initial chirping is considered
for a birefringent fiber. The corresponding dynamics is governed by variable coefficient,
coupled NLS equations for the components of the orthogonal polarization of the pulse
envelopes. When the self phase and cross phase modulation coefficients are identical for
special angles, several new classes of wave patterns can be found. New stationary wave
patterns which possess multiple peaks within each period are found, similar to those found for
the classical Manakov model. For situations where the self- and cross-phase modulation
coefficients are different, symbiotic solitary pulses are studied. A pair of bright-dark pulses
exists, where either or both pulse(s) cannot propagate in that waveguide if coupling is absent.
The existence and stability of gap solitons in a model of hollow core fiber in the zero
dispersion regime are analyzed in Chapter 12. The model is based on a recently introduced
model where the coupling between the dispersionless core mode and nonlinear surface mode
(in the presence of the third order dispersion) results in a bandgap. It is found that similar to
the anomalous and normal dispersion regimes, the family of solitons fills up the entire
bandgap. The family of gap solitons is found to be formally unstable but in a part of family
the instability is very weak. Consequently, gap solitons belonging to that part of the family
are virtually stable objects. The interactions and collisions of in-phase and the -out-of-
phase quiescent solitons and moving solitons in different dispersion regimes are investigated
and compared.
Chapter 13 reviews various schemes for multiwavelength fiber lasers and semiconductor
optical amplifier (SOA) ring lasers. Multiwavelength fiber lasers have applications in
wavelength division multiplexing (WDM) optical communication systems, optical fiber
sensors and optical spectroscopy. Erbium-doped fiber amplifiers (EDFAs), Raman amplifiers
and SOAs are mainly used as gain media for multiwavelength fiber lasers.
Because EDFAs are homogeneously broadened gain media, various methods have been
researched to enable the multiwavelength generation. Due to the introduction of liquid
nitrogen cooling, four-wave mixing, frequency shifted feedback, and so on, multiwavelength
erbium-doped fiber lasers could become realized.
Jrgen C. Schlesinger xii
On the other hand, because SOA and Raman amplifiers are gain media with
inhomogeneous broadening, multiwavelength generation is relatively easy. The useful
features of the multiwavelength lasers are mainly dependent on a comb filter. One of the most
important features of multiwavelength lasers is tunability. The tunability of wavelengths and
channel spacing is required for WDM optical communication systems. Much research has
been conducted to enable implementation of tunable multiwavelength fiber lasers. Various
comb filters such as Fabry-Perot filters, fiber Bragg gratings, and polarization-maintaining
fiber loop mirrors can be used for multiwavelength fiber lasers. The authors review several
schemes for multiwavelength SOA-fiber and Raman fiber lasers in this chapter.
The optical fiber reliability in telecommunication networks has been still an issue, thats
why the question of how long an optical fibers might been used without a significant
probability of failure isnt out of interest. Much work was developed around this issue, but the
optical fiber fatigue and aging process has not been yet fully understood.
The reliability of the optical fibers depends on various parameters that have been
identified: time, temperature, applied stress, initial fiber strength and environmental
corrosion. The major and usually unique corrosion reagent is water, either in the liquid state
or as atmospheric moisture. Glass surface contains numerous defects, either intrinsic, the so-
called Griffiths flaws and extrinsic, in relation to fabrication process. Under permanent or
transient stress, microcracks grow from these defects, and growth kinetics depend on
temperature and humidity. Although polymeric coating efficiently protects glass surface from
scratches, it does not prevent water to reach glass fiber.
The work carried out during the last years made possible to apprehend in a more coherent
way the problems of failure and rupture of fibers subjected to severe aging conditions.
In Chapter 14, some informations on the used characterization methodology for the silica
optical fibers are given. In addition, Optical fibers analysis advantages, expected percussions
and theoretical background are given to enlighten the potential concerned persons. The
principal optical fiber test benches are described and some results are commented. Finally,
final remarks are noted.

SHORT COMMUNICATION
In: Optical Fibers Research Advances ISBN: 1-60021-866-0
Editor: Jurgen C. Schlesinger, pp. 3-11 2007 Nova Science Publishers, Inc.

IGNITION WITH OPTICAL FIBER
COUPLED LASER DIODE

Shi-biao Xiang
1,2*
, Xu Xiang
3
, Wei-huan Ji
2

and Chang-gen Feng
4

1
Department of Technical Physics, Zhengzhou Institute of Light Industry,
No.5 Dongfeng Road, Zhengzhou 450002, P.R. China
2
Key Laboratory of Informationalized Electric Apparatus of Henan Province,
Zhengzhou 450002, P.R. China
3
State Key Laboratory of Chemical Resource Engineering, Beijing University of
Chemical Technology, P.O. BOX 98, Beijing 100029, P.R. China
4
School of Mechanics and Engineering, Beijing Institute of Technology,
Beijing 100081, P.R. China
Abstract
Optical fibers, an important and promising material, have attracted more and more
attention and extended their applications to various scientific and practical aspects. In this
article, the key role of fibers, as the carriers of information and energy in our times, was
briefly summarized. Afterwards, the configuration of fiber coupled laser diode ignition system
was elucidated as well as the advantages, developments and applications of this technology.
Furthermore, the energy-transmitting characteristics of single-mode fibers and multi-mode
ones and the key points of fiber-coupled technology were analyzed. In a practical case, the
effect of the diameters of core on laser ignition, from both theory and experiments, was
studied. The findings suggest that the smaller the diameters of core, the lower the ignition
threshold under the same laser power. That is to say, the ignition becomes easier while using
fibers with smaller core. Finally, the issue on selection of core was clarified based on the
consideration of both laser power density and the endurance of fibers.

*
E-mail address: shibiaoxiang@zzuli.edu.cn. Tel: 86-371-63557226 (Corresponding author: S. B. Xiang)
Shi-biao Xiang, Xu Xiang, Wei-huan Ji et al. 4
1. Introduction
Optical fibers as carrier of information and energy have intrigued intensive interest worldwide
due to its scientific and technological significance in various practical fields. For instance, in
optical communications, fibers have received tremendous attention from both experimental
and theoretical aspects not only on the type of fiber materials but also on various
communicating techniques [1-4], in which the most primary function of fibers is to transmit
information like voice, images and videos from one place to another. A wide variety of
optical fiber devices have been designed and exploited in the field of fiber-based
communications, such as fiber optical amplifiers, frequency or phase modulators, planar
waveguides and fiber polarizers.
Furthermore, the developments of microstructured optical fibers (MOFs) and photonic
crystal fibers [5-9] enable a number of potential functionalities including tunability and
enhanced nonlinearity, and extend novel fiber device applications to fiber Bragg gratings,
tunable resonant filters, variable optical attenuators and nonlinear optics devices owing to
their unique characteristics [10-15].
More interestingly, chemical sensors based on optical fibers have been widely explored
in the past few years [16-18]. For example, sensors for gases or vapors [19-20], humidity [21-
22], metallic ions, specific chemical compounds [23], viscosity [24], intensity [25] and
miniature pressure [26] have been delicately designed and rapidly developed. Also,
biosensors [27, 28] for enzymes, antibodies or antigens, DNA [29] and bacteria are becoming
a prevailing research topic on the basis of fiber materials. They have been exhibiting
promising applications in a variety of fields such as chemical analysis, biological monitoring
and environmental detection. In this article, the emphasis has been highlighted on the
fundamental principles and the important practice of fiber-coupled laser diode ignition.
2. Fiber Coupled Laser Diode Ignition
2.1. Brief Review on Laser Diode (LD) Ignition
Laser ignition is a kind of ignition technique, which refers to detonation or ignition of
energetic materials such as solids or fluids [30-33] by laser beam. At early stage of laser
ignition technique, the types of laser used for the experimental and application research are
mostly Nd:YAG, Nd: GSGG, Nd: glass laser and CO
2
laser [34-40]. These lasers possess the
characteristics of high output power or energy, small radiation angle of light, long life-span
and low price. However, the obvious disadvantages of this kind of laser are low energy
conversion efficiency, in which the ratio of output light energy and input electric energy is
usually lower than 3%, as well as large volume and heavy weight. With the born of LD and
the naissance of LD ignition, the research and evolution of laser ignition technique come into
a new era. The experimental studies for laser diode ignition began in the middle of 1980s.
Ewick, Kunz, Kramer, Jungst, Merson, Glass and Roman et al have made great devotion to
the field of LD ignition, of which Ewick [41] and Kunz [42] published their literatures firstly.
LD belongs to a kind of semiconductor laser stimulated by current. In LD ignition, LD is
utilized as energy source, and the energy is transmitted to powders by using optical fiber,
which detonates or ignites the energetic materials. This ignition configuration has the
characteristics of safety, reliability, and strong capability of anti-interference of
electromagnetism. In addition, the following advantages are also realized.

(1) It is easy for LD ignition system to realize miniaturization of apparatus due to its
small volume and light weight.
(2) LD ignition system has excellent adaptability to the ambient environment because of
the input of low voltage and electric energy.
(3) LD ignition system can output multi-channel laser signals by using LD arrays and
consequently control multi-point ignition through the selection to time and order of
signals.

As a result, LD ignition has received extensive attention, and exhibits promising
application especially in the field of aviation and aerospace.
Fig. 1 illustrates the schematic diagram of ignition system induced by laser diode. Laser
diode is employed as light source, and energy is transmitted to powders by optical fibers. The
powders are ignited and subsequently exploded while enough energy is provided.

powder
fiber
fiber coupler
connecting
laser
aperture
lock
device

Figure 1. Schematic illustration of ignition system induced by laser diode.
2.2. Optical Fiber and Fiber Coupled Technology
As a carrier to transmit laser, optical fiber plays a crucial role in LD ignition. The materials of
optical fiber should possess the favorable characteristics of optical and mechanical properties
as well as the characteristic of temperature. The widely used fibers are made of silica glass or
plastic. The fibers can be classified into two types, one is step-index fiber and the other is
grade-index one according to the distribution of refractive-index of fiber core. The refractive
index of core is a constant for step-index fiber, schematically shown in Fig. 2. However, for
the grade-index fiber, the refractive index of core gradually decreases outwards along the
radial direction. Due to the self-focusing characteristic of the grade-index fiber, the output
beam has higher energy density close to the axis of fiber. As a consequence, the laser power
density can be enhanced by using the grade-index fiber.

n
y
n
2
n
1
Cladding
Core
z
y
r
Fiber axis

Figure 2. The schematic diagram of step-index fibers.
Both theoretical analysis and experimental results indicate that the increase of power
density is considerably favorable to LD ignition. That is to say, the combination of thin
diameter, low attenuation, small numerical aperture and grade-index fiber is advantageous to
LD ignition. Ewick and coworkers found that the threshold of ignition using grade-index fiber
was decreased by around 30% than that using step-index fiber in the ignition experiments of
Ti/KClO
4
and CP/carbon black.
Generally, optical fibers can be classified into single-mode fibers (SMFs) and multi-mode
fibers (MMFs) according to the transmission modes. SMFs exhibit excellent capability in
optical communications. And the light energy transmitted by SMFs presents to be Gauss
distributions, which means the more centralized energy can be obtained, and is thus favorable
to LD ignition. Nevertheless, the diameter of core in SMFs is confined to a large extent.
The fiber waveguide parameters can be expressed as
2 / 1 2
2
2
1
) ( n n kr V = , where
1
n and
2
n are the refractive indices of the core and the cladding, respectively, and r is the core radius.
And
0
/ 2 = k is the wave number, where
0
represents the wavelength in vacuum. Single
mode operation is obtained for V <2.405, and it can be observed for the wavelength longer
than
c
( 405 . 2
0
=
=
c
V

). For example, in order to obtain V =2.405 at =10.6 m, a silver
halide fiber can have core diameter 2r = 8 m while the normalized difference between the
refractive indices of the core and the cladding [
1 2 1 1
/ ) ( / n n n n n = = ] should be 0.1.
Alternatively, V =2.405 will be obtained for 2r = 80 m and = 0.001. In both cases one can
observe a SMF operation. In typical silica SMFs, the value is of the order of 0.002. Shalem et
al [43] selected to design silver halide SMFs which have 2r = 60 m and = 0.002, for
which the estimated waveguide parameter was at least 10% lower than 2.405 (at = 10.6 m).
To make the laser power transmitted by fiber enough high in the transmitting process, it
is necessary to ensure that the fiber has enough thick core, otherwise the fiber will be
inevitably damaged by laser. Based on the findings of Schmidt-Uhlig and his colleagues [44],
the feasibility of transmitting 20 mW, 5 ns laser pulse from a frequency doubled Nd:YAG
laser through a standard 1500 m multi-mode optical fiber has been demonstrated.
Furthermore, the experiments on the delivery of more than 20000 pulses with mean energy of
110 mJ with no damage to fiber have been performed. Consequently, multi-mode optical
fibers are prevailing in LD ignition.
Commonly, the specification of optical fiber used in LD ignition is as follows, the
diameter of core (2r) of 100-200 m, numerical aperture (NA) <0.3, attenuation per kilometer
dB
km
<3dB and output power of fiber P 1W.
Additionally, the end-face quality of fiber is an important aspect, which includes the
perpendicularity of end-face to axes, degree of levelness and cleanliness of end-face, etc.
These factors have a significant effect on the transmission of laser. Surface defects and
contamination not only debase the transmission efficiency but also result in the strong electric
field and large thermal stress in local area under higher power density, and eventually damage
fiber itself. Accordingly, the cleansing and polishing treatment to the end-face of fiber is a
serious issue. Ewick adopted tailor-made polishing apparatus for fiber to polish the end-face
and directly observed it by a microscope with 400 magnifications in order to determine the
quality of polishing.
Besides transmitting energy, optical fiber can also be applied in coupling and splitting.
The fiber-coupled technology was firstly used in the coupling between fiber and LD. LD
emits the elliptic radiation, and thus the proper convergent lens is required to effectively
couple laser into output fiber. The difficulty of coupling is greatly increased because of the
small fiber core used in LD ignition, and the loss of coupling can even be as high as 5dB.
Apparently, it is crucial to enhance the coupling efficiency at coupling sites. The second
problem of this technology is the coupling between fibers, which includes two types of
coupling: one is the coupling between single and single fiber (one in and one out), and the
other is single and multi fiber (one in and several out). The latter is indispensable for multi-
point ignition of LD [45]. The coupling between fibers is realized by linkers and commutators.
To reduce the loss of coupling, it is necessary to improve the techniques of collimation, tight-
fittings, and fixed-airproof. Roman and coworkers [46] used STC linkers in LD ignition to
link two fibers for light in and light out, one of which has a core diameter of 100 m and the
other has an outer diameter of 140 m. The linkers, having an attenuation of as low as 0.56dB
and jack diameter of 144 m, can operate in the temperature range of 40 to +80
o
C. This is a
kind of linker with low attenuation, easy to manipulate, and good performance.
Further, it is also a noticeable issue on the coupling technique between fibers and ignited
powders. Kramer et al [47] designed and developed two kinds of components i.e., the fiber
foot and the optical window, to resolve the coupling between fibers and ignited powders.
These components make LD ignition system more convenient and practical in operation.
They are required not only to meet the demand of mechanical strength but also to reduce the
energy loss as could as possible. The fiber foot with high mechanical strength is prepared
through the following steps, firstly envelop a short fiber into glass preform within a metal
shell under high temperatures, and secondly polish two ends of the shell. The advantages of
the component are small cross-sections of fiber, high mechanical strength, and meanwhile
fiber itself plays a role as waveguide with the characteristic of high transmission quality.
However, the disadvantage is the energy attenuation caused by surface reflection and non-
collimation.
The optical window is a kind of transparent solid made of glass material, which is fixed
between the fiber and ignited powders. There is a little probability resulting from non-
collimation. However, the material for windows can absorb laser and disperse radiation of
beam, thus leading to the decrease of power density. By selecting a proper material and
appropriate thickness of windows combined with the convergence method, the above-
mentioned loss can evidently be decreased. Also, they compared sapphire glass with
phosphor glass as window material. When the thickness of window is the same 0.4 mm, the
ignition energy for carbon black-doped powders is 3.4 mJ by using sapphire glass as window
material, while the ignition energy is 2.3 mJ using phosphor glass as window, and
correspondingly the ignition energy is 1.6 mJ with no window. Compared with sapphire glass
window, phosphor glass window has a better performance due to its lower thermal
conductivity and lower refractive index.
2.3. Experimental Study
The dependence of ignition threshold on diameter of core (2r) can be demonstrated by
experimental studies and numerical calculations. The experimental setup, schematically
illustrated in Fig. 3, is divided into two main parts: A for ignition part and B for testing one.
T-type fiber junction separates the emitted laser into two ways, one of which directly delivers
laser into photoelectric detector and then to oscillograph. The other transmits laser to ignitor,
where energetic powders are ignited and shone. The time difference of two ways of light
reaching the oscillograph is defined as ignition delay time t
i
, measured by a photoelectric
detector. The threshold energy of ignition E
i
can be calculated in the following formula:

E
i
= P
i
t
i
(1)

where P
i
refers to laser power imported ignitor.

Figure 3. Schematic diagram of experimental setup of laser ignition.
The maximal output power of LD used in the experiments is 1W and the wavelength is
808 nm. The laser is continuously output and is power-adjustable. The diameter of the
coupled fibers is 100 m, 200 m and 400 m, respectively. And the powder is Zr/KClO
4

with a mass ratio of 1:1. The ignition experiments were carried out at room temperature. Fig.
4 shows the relationship of the ignition threshold and laser power with regard to three types
of fibers with different diameters of core.
0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2
6.0
6.5
7.0
7.5
8.0
8.5
9.0
9.5
10.0
10.5
11.0
2r =400m
2r =200m
2r =100m
E
i

/
m
J
P/W

Figure 4. The relationship of ignition threshold and laser power.
The three plots in Fig. 4 from bottom to top correspond to the diameter of core of 100 m,
200 m and 400 m, respectively. According to the three plots, one can conclude that the
ignition threshold decreases with the increasing laser power when the diameter of core has a
fixed value. And the threshold increases as the diameter of core becomes thicker under the
same laser power. This suggests that the ignition threshold decreases as the increasing power
density under certain conditions.
3. Conclusion
Both theoretical and experimental results indicate that two issues need to be considered for
the selection of diameters of core in fiber coupled laser ignition system. One is to enhance
power density as could as possible i.e., to select the core with thinner diameters, which can
decrease the ignition threshold. And the other is to take into account the endurance of fiber
itself i.e., the fiber with too thin diameter of core is not suitable. As a result, the commonly
used fibers are multi-mode fibers with diameters of core of 100-200 m.
References
[1] Litchinitser, N. M.; Sumetsky, M.; Westbrook, P. S. J. Opt. Fiber. Commun. Rep. 2007,
4, 41-85.
[2] Li, X. Y.; Voss, P. L.; Chen. J.; Sharping, J. E.; Kumar, P. Opt. Lett. 2005, 30, 1201-
1202.
[3] Hutchinson, D. P.; Richards, R. K. J. Opt. Fiber. Commun. Rep. 2007, 4, 1-11.
[4] Wong, K. K.Y.; Lu, G. W.; Chen, L. K. Opt. Commun. 2007, 270, 429-432.
[5] Monro, T. M.; Ebendorff-Heidepriem, H. Annu. Rev. Mater. Res. 2006. 36, 467-495.
[6] Cregan, R. F.; Mangan, B. J.; Knight, J. C.; Birks, T. A.; Russell, P. St. J; Roberts, P. J.;
Allan, D. C. Science 1999, 285, 1537-1539.
[7] Knight, J. C.; Broeng, J.; Birks, T. A.; Russell, P. St. J. Science 1998, 282, 1476
[8] Benabid, F.; Couny, F.; Knight, J. C.; Birks, T. A.; Russell, P. St J. Nature 2005, 434,
488-491.
[9] Talukder, A. I.; Totsuka, K.; Tomita. M. Appl. Phys. Lett. 2006, 89, 054103.
[10] Eggleton, B. J.; Kerbage. C.; Westbrook, P. S.; Windeler, R. S.; Hale, A. Opt. Express
2001, 9, 698-713.
[11] Villatoro, J.; Minkovich, V. P.; Pruneri, V.; Badenes, G. Opt. Express 2007, 15, 41491-
1496.
[12] Hu, J.; Marks, B. S.; Menyuk, C. R.; Kim, J.; Carruthers, T. F.; Wright, B. M.; Taunay,
T. F.; Friebele, E. J. Opt. Express 2006, 14, 4026-4036.
[13] Bian, S. P.; Zhang, W. J.; Kuzyk, M. G. Opt. Lett. 2003, 28, 929-931.
[14] Akimov, D. A.; Schmitt, M.; Maksimenka, R.; Dukelskii, K. V.; Kondratev, Y. N.;
Khokhlov, A. V.; Shevandin, V. S.; Kiefer, W.; Zheltikov, A. M. Appl. Phys. B 2003, 77,
299-305.
[15] Skryabin, D. V.; Luan, F.; Knight, J. C.; Russell, P. St. J. Science 2003, 301, 1705-1708.
[16] Wolfbeis, O. S. Anal. Chem. 2004, 76, 3269-3284.
[17] Lee S. T.; George, N. A.; Sureshkumar, P.; Radhakrishnan, P.; Vallabhan, C. P. G.;
Nampoori, V. P. N. Opt. Lett. 2001, 26, 1541-1543.
[18] Hassani, A.; Skorobogatiy, M. Opt. Express 2006, 14, 11616-11621.
[19] Minkovich, V. P.; Monzn-Hernndez, D.; Villatoro, J.; Badenes, G. Opt. Express 2006,
14, 8413- 8418.
[20] Pickrell, G.; Peng, W.; Wang, A. Opt. Express 2004, 29, 1476-1478.
[21] Xu, L.; Fanguy, J. C.; Soni, K.; Tao, S. Q. Opt. Express 2004, 29, 1191-1193.
[22] Kronenberg, P.; Rastogi, P. K.; Giaccari, P.; Limberger, H. G. Opt. Lett. 2002, 27, 1385-
1387.
[23] Santoyo, A. T.; Shlyagin, M. G.; Jimenez, F. J. M.; Oyarzabal, L. N. R. Opt. Commun.
2007, 271, 386-390.
[24] Haidekker, M.A.; Akers, W.J.; Fischer, D.; Theodorakis, E.A. Opt. Lett. 2006, 31, 2529-
2531.
[25] Baptista, J. M.; Marques, J. M.; Frazao, O.; Santos, S.; Santos, J. L.; Marques, M. B. Opt.
Commun. 2007, 271, 224-227.
[26] Wang, X. W.; Xu, J. C.; Zhu, Y. Z.; Cooper, K. L.; Wang, A. B. Opt. Lett. 2006, 31,
885-887.
[27] Emiliyanov, G.; Jensen, J. B.; Bang, O.; Hoiby, P. E.; Pedersen, L. H.; Kjr, E. M. Opt.
Lett. 2007, 32, 460-462.
[28] Ye, J. Y.; Myaing, M. T.; Norris, T. B.; Thomas, T.; Baker, Jr. J. Opt. Lett. 2002, 27,
1412-1414.
[29] Wang, X. W.; Cooper, K. L.; Wang, A. B.; Xu, J. C.; Wang, Z.; Zhang, Y.; Tu, T. Z.
Appl. Phys. Lett. 2006, 89, 163901.
[30] Starik, A. M.; Lukhovitskii, B. I. Kinet. Katal. 2006, 47, 333-340.
[31] Krauns, Ch.; Barelko, V.; Fabre, G.; Tredicce, J.; Krinsky, V. Catal. Lett. 2001, 72(3-4),
161-165.
[32] El-Rabii, H.; Gaborel, G. Appl. Phys. B 2007, DOI: 10.1007/s00340-006-2563-9.
[33] Korepanov, V. I.; Lisitsyn, V. M.; Oleshko, V. I.; Tsipilev1, V. P. Combust. Expl. Shock
Waves, 2006, 42, 94-106.
[34] Ostrowski, P. P.; Grant, J. F.; Petrow, E. D.; Downs. D. S. Proceedings of the 8th
International Pyrotechnics Seminar, 1982, 526.
[35] Hill, R. A. Appl. Opt. 1981, 20, 2239-2242.
[36] Holy, J. A.; Girmann, T. C. Proceedings of the 13th international pyrotechnic seminar,
Grand Junction, CO, 11-15 July 1988, 449-469.
[37] Carleton, F. B.; Krallis, K.; Weinberg, F. J. AD-A212 342, 1989.
[38] Barrows, A. W.; Forch, B. E.; Beyer, R. A.; Cohen, A.; Newberry, J. E. AD-A261 049.
1993.
[39] Phuoc, T. X.; Mathur, M. P.; Ekmann, J. M.; Durbetaki, P. High-Energy Flame, 1993,
94, 349-362.
[40] Ostrowski, P. P.; Petrick, J. T.; Petrow, E. D.; Smith, T. C. Proceedings of the 8th
International Pyrotechnics Seminar, 1982, 543.
[41] Ewick, D. W.; Dosser, L. R.; Mcconb, S. R.; Brodsky, L. P. Proceedings of the
thirteenth international pyrotechnic seminar, 263-277, Grand Junktion,CO, 1988
[42] Kunz, S. C.; Salas, F. J. DE88008771/HDM, 1988.
[43] Shalem, S.; Tsun, A.; Rave, E.; Millo, A.; Nagli, L.; Katzir, A. Appl. Phys. Lett. 2005,
87, 091103.
[44] Schmidt-Uhlig, T.; Karlitschek, P.; Marowsky, G.; Sano, Y. Appl. Phys. B 2001, 72,
183-186.
[45] Boucher, C.; Schulze, N. 31AIAA95-2982, 1995.
[46] Oestmark, H.; Roman, N. J. Appl. Phys. 1993, 73, 1993-2003.
[47] Kramer, D. P.; Beckman, T. M.; Ewick, D. W. Proceedings of the 15th international
pyrotechnic seminar, 1990, 569-579.

RESEARCH AND REVIEW STUDIES

Chapter 1

EVANESCENT FIELD TAPERED FIBER OPTIC
BIOSENSORS (TFOBS): FABRICATION, ANTIBODY
IMMOBILIZATION AND DETECTION

Angela Leung
1
, P. Mohana Shankar
2
and Raj Mutharasan
1*

1
Department of Chemical and Biological Engineering
2
Department of Electrical and Computer Engineering, Drexel University,
Philadelphia, PA 19104
Abstract
Tapered Fiber Optic Biosensors (TFOBS) are sensors that operate based on fluctuations
in the evanescent field in the tapered region. In our laboratory, TFOBS are made by heat
pulling commercially-available single mode optical fibers. They have been investigated for
various applications, including measurement of physical characteristics (refractive index,
temperature, pressure, etc.), chemical concentrations, and biomolecule detection. In this
chapter, an up-to-date review of TFOBS research is provided, with emphasis on applications
in biosensing such as pathogen, proteins, and DNA detection. The physics of sensing and
optical behavior based on taper geometry is discussed. Methods of fabrication, antibody
immobilization, sample preparation, and detection from our laboratory are described. We
present results on the non-specific response, simulation, and detection of E.coli O157:H7 and
BSA. This chapter will conclude with an analysis of the future direction of the Tapered Fiber
Optic Biosensors.
1.0. Introduction
Tapered fiber optic biosensors (TFOBS) are made from optical fibers, and, are capable of
detecting specific analytes using optical responses. They have been used for the measurement
of physical and chemical properties [4-8], [9-14] of biological molecules [2, 15-19] and have
several applications including environmental monitoring, drug screening, clinical diagnostics,
and defense. TFOBS offer many advantages including flexibility, ease of use, affordability,

*
E-mail address: mutharr@drexel.edu. Tel.: (215) 895-2236. Fax: (215) 895-5837. (Corresponding author)
Angela Leung, P. Mohana Shankar and Raj Mutharasan 16
and ability to perform sensing using a small amount of sample. These sensors are based on
the evanescent field associated with fiber, and, often are also referred to as Evanescent Field
Tapered Fiber Sensors. In this chapter, the basics of TFOBS are discussed, along with an up-
to-date literature review of TFOBS. Experimental methods and recent results from our
laboratory are also presented.
2.0. Physics of Evanescent Field Sensing in Tapered Fibers
Optical fibers are cylindrical waveguides, and, are made of a silica core surrounded by a silica
cladding. The core refractive index is higher than the cladding refractive index (RI) because it
is doped with Ge. Light propagates through the core by total internal reflection (TIR). Besides
the light propagating in the core, there is a small component of light, known as the evanescent
field, which decays into the cladding.
Evanescent light penetration is described by its penetration depth (d
p
), which is the
position away from the core/cladding interface at which the light decays to 1/e of its value at
the core-cladding interface, and is given by:

2 2 2
2 sin
p
co cl
d
n n
(1)

In eqn. (1), is the operating wavelength of light, n
co
the index of the core and n
cl
the
index of the cladding. The angle of incidence at the core cladding interface is . The
evanescent field in a uniform diameter fiber does not interact with the outside environment
because it decays to a negligible value as it reaches beyond the cladding. This is due to the
fact that in typical fibers the cladding thickness is several times that of the core. However, if
the cladding is removed or the fiber is tapered down to a diameter less than the original core
diameter, evanescent field can interact with the external medium affecting the transmission
through the fiber.
The penetration depth in a tapered fiber depends on the local diameter of the tapered
fiber, the RI of the core, and RI of the external medium. Since there is a continuous change in
the diameter along the fiber in the tapered region (except in the waist), coupling of light
among the modes can occur [20]. Coupling in the tapered region causes the transmission
properties of the fiber to change. Presence of analytes in the tapered region can lead to RI
changes in the taper. This results in changes in the coupling characteristics and causes
changes in the optical throughput.
Physical characteristics of the fiber such as RI of the core and cladding, core diameter,
and operating wavelength determine the number and type of modes that propagate through
the fiber. The lowest order mode has the tightest confinement of the field, and hence the
weakest evanescent field. As the mode order goes up, the associated evanescent field also
increases. In a tapered or de-cladded fiber, the optical characteristics of the surrounding
medium such as its index, absorption, etc. can affect the optical throughput.
Evanescent Field Tapered Fiber Optic Biosensors (TFOBS) 17
2.1. Wave Propagation in Absorption Sensors
The shape of the optical field in a fiber is determined by the number of modes present. Figure
1 shows a tapered fiber with a short region of constant thickness (waist) and contracting and
expanding regions. The number of modes that can be supported in a fiber is determined by the
V-number,

2 2 0
2
co cl
a
V n n
= (2)

where a
0
is the radius of the core. When V<2.405, only the lowest order mode is supported,
and as V increases, number of modes increase. Although in a single-mode (SM) fiber only the
lowest order is supported, in the tapered region higher order modes can potentially be
supported because of the larger difference in refractive index between the core and the sample
(~0.12) compared to a regular fiber (~0.01). Reduction of the fiber radius increases the
evanescent field strength, and enhances the interaction of the evanescent field with the analyte
leading to variations in optical throughput (transmitted light).

Figure 1. Photograph of a TFOBS. The region of interest in a tapered fiber is identified by the region
where V
core
<1.
2.2. Wave Propagation in Continuous Bi-conical Tapered Fibers
In our laboratory, tapered fibers were made by heat-pulling an optical fiber without removing
the cladding. Unlike uniform fibers, the V-number changes along the length of a tapered
fiber. When V-number becomes less than unity, the core is too small to contain the light and
light guidance is determined by the original cladding which acts as the core and the external
medium of RI n
ext
which serves as the cladding. The new V-number is called V
clad
where the
parameter, a
0
in Eq. (2) is replaced by the radius of the overall fiber, b(z), and is given by

2 2
2 ( )
( )
clad cl ext
b z
V z n n
= (3)

In tapered fibers, the V-value is generally referred to as V
clad
to distinguish it from V
core
,
given in Eq. (2). Note that the diameter b in eqn. (3) is a function of the location (z),
indicating the existence of tapering.
2.3. Numerical Simulation of Light Transmission in a Tapered Fiber
Numerical simulation of light transmission through a tapered fiber can provide useful insight
into its properties. To simplify the analyses, the simplified mode theory based on linearly
polarized modes (LP) can be used to determine the transmission behavior [21]. Assuming that
light enters into the fiber parallel to the axis, the only modes that are excited are the LP
0m

modes. The transverse components of the electrical field inside the fiber are:

( )
( )
0 0
0 0
0 0
0 0
( ), 1
( )
( ), 1
m
X
m
J U R R
E r A J U
K W R R
K W

=

>

(3)

where R is the normalized radial coordinate, r/a
0,
whereas U and W are constants They
depend on the wavelength and RI of the core and cladding,

0
2
2 2 2 2
0 0
2
m co m
U a n

=

(4a)

0
2
2 2 2 2
0 0
2
m m cl
W a n

=

(4b)

where c is the speed of light and is the propagation constant. When m=1, only the
fundamental mode exists. In eqn. (3) J
0
(.) and K
0
(.) are the Bessel and modified Bessel
functions of zero
th
order, respectively. A is a constant determined from orthogonality
principle [21]. The subscript m represents the various circularly symmetric LP
0m
modes that
may be present in the fiber.
When V
core
<1 and V
clad
>2.405, many modes are supported since the index difference
between the cladding and the external medium (n
ext
) is large. As mentioned previously,
tapering leads to coupling among LP
0m
modes [20, 22] . A simple means to visualize the taper
is to model taper geometry by approximating the slopes of the taper by stepwise linear
approximation. At each step i, the parameters U
0m
and W
0m
are analogous

to the constants U
and W in a uniform diameter fiber. They can be expressed using the local radius
i
as

0 0
2
2 2 2 2
2
m m
i i i
cl
U n

=

(5a)


0 0
2
2 2 2 2
2
m m
i i i
ext
W n

=

(5b)

The V-number for each step is given by:

1/ 2
2 2
2
i i
cl ext
V n n
=

(6)

The values of U, W and are calculated using the LP mode approximation [21]. The
relationship between the modal amplitudes of the LP
0m
modes of the i
th
and (i+1)
th
step is:

1 1
1 1
1 1
( ) ( )
i i
i i
q
m
j z
j z i i i i
m m q q
m q
A E r e B E r e

+ +
+ +
= =
=

(7)

where E(r) is the electric field,
m
is the propagation constant on the left and
q
is the
propagation constant on the right. It has been assumed that E(r) are orthonormal [21]. A
m
is
the amplitude of the modes on the left and B
q
is the amplitude of the modes on the right. That
is,

2
2
0 0
( ) 1 E r rdrd
=

(8)

The amplitude on the right is obtained by applying the orthogonality principle:

1 1
1
;
+ +
+
=
i i
i i
q
m
j z
j z i i
q m nm pq
n m
B e A e C

(9)

1
;
0
2 ( ) ( )
+
=

i i
nm pq q m
C E r E r rdr (10)

In the tapered region, light is coupled among the various LP
0m
modes. When V
core
=1,
power in the LP
01
cladding mode is transferred to LP
01
core mode and appears at the output
end of the fiber. The light remaining in other modes stays in the cladding and is lost. A
MATLAB
program was used to estimate the amplitude and output power. The taper
geometry, wavelength and number of steps were varied to determine the resulting changes in
power.
Sample simulation results, illustrated in Figure 2, show changes in transmission vs. waist
diameters for two taper geometries. In Panel A, the taper geometry resembles a symmetric
taper made by the fusion splicer, while in Panel B, the simulation is for a long taper similar to
a heat-drawn taper. The transmission is normalized with respect to air, so that a value of 1.2
indicates a transmission increase of 20% in water compared to air. Figure 2 show that as the
waist radius increases, the difference in transmission between water and air decreases.
However, at intermediate values the ratio may be higher or lower than unity, particularly at
smaller waist diameters. For certain values of the radius, the transmission through water is
higher than in air. At a longer wavelength (550 nm, for example) and for the same waist
diameter values, the difference in transmission between air and water differ by less than 10%.
To explore further, simulation was undertaken by varying the diameter of the waist in much
smaller steps of 0.001 m. These results are shown in Figure 3 for two starting diameters, 5
m and 6.25 m. The transmission characteristics change significantly for small changes in
diameter. For example, a 5.54 m diameter taper exhibits 30% higher transmission in water

0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5
1.6
5 9 13 17
Waist diameter (m)
N
o
r
m
a
l
i
z
e
d

T
r
a
n
s
m
i
s
s
i
o
n
0.7
0.75
0.8
0.85
0.9
0.95
1
1.05
1.1
5 9 13 17
Waist diameter (m)
N
o
r
m
a
l
i
z
e
d

t
r
a
n
s
m
i
s
s
i
o
n
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5
1.6
5 9 13 17
Waist diameter (m)
N
o
r
m
a
l
i
z
e
d

T
r
a
n
s
m
i
s
s
i
o
n
0.7
0.75
0.8
0.85
0.9
0.95
1
1.05
1.1
5 9 13 17
Waist diameter (m)
N
o
r
m
a
l
i
z
e
d

t
r
a
n
s
m
i
s
s
i
o
n

Figure 2. Transmission in water normalized with respect to air as waist diameter is altered. Top: A short
symmetric taper: a= 0.425 mm, b=0.325 mm, c=0.500 mm. Operating wavelength = 470 nm. Bottom:
A longer asymmetric taper: a=2.25 mm, b=0.245 mm, c=4.5 mm. Operating wavelength = 550 nm.
Adapted from [3].
while a 5.58 m waist diameter taper transmits 20% less transmission. The differences,
however, become smaller for larger diameters. The example of 6.25 m in Figure 3 shows
that the changes in transmission were less than 10 %.

0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
0 0.02 0.04 0.06 0.08 0.1
Change in waist diameter (mm)
N
o
r
m
a
l
i
z
e
d

T
r
a
n
s
m
i
s
s
i
o
n
at 6.25 mm
at 5.5 mm

Figure 3. Transmission characteristics of a taper (a= 2.25 mm, b=0.25 mm, c=4.5 mm) in water at 470
nm as a function of change in waist diameter. Transmission is normalized with respect to transmission
in air at the corresponding geometric values. Smaller starting diameter tapers show large changes in
transmission for small (0.01 m) changes is waist diameter. Adapted from [3].
These simulation results can serve as a guide to the analysis and interpretation of the
experimental data on the tapered fibers. It is important to recognize that in the simulation we
considered only the effect of refractive index in the waist region. In actual sensing
experiments, the cells absorb at the operating wavelength, and the resulting sensor response is
a complex interplay of these two phenomena. Furthermore, cells do not have homogeneous
RI because the cells constitute particulate matter. Finally, the cell attachment onto the taper
surface is often not uniform as we showed in our earlier report [2].
3.0. Literature Review
In this section, the applications of TFOBS for pathogen detection, toxins measurements,
clinical measurements, and DNA detection are presented. In tables 1 to 3, we summarize the
analytes detected, matrices in which they were detected, detection principle, basis of sensors,
and detection limits.

Table 1. TFOBS For Pathogen and Toxin Measurement
Target Analyte LOD Matrix
Taper
Geometry
Fiber Type Detection Principle References
Bacillus anthracis
3.2E5
spores/mL
buffer BT Polystyrene MM Fluorescent sandwich assay [32]
Bacillus subtilis var. niger
8 x 10(4)
spores/mL
buffer NA (chip) NA (chip) Leaky wave (SPR) [59]
LacZ DNA in Escherichia coli 25 pM buffer Uniform MM Fluorescent intercalating agents [55]
Staphylococcus aureus Protein A 1 ng/mL ND ND MM plastic Fluorescent sandwich assay [35]
Escherichia coli O157:H7
0.016 dB/h/N
o
,
Initial number
(N
o
): 10-800 *
buffer BT MM Absorption [23]
Escherichia coli O157:H7 70 cells/mL Buffer BT SM Intensity [2]
Escherichia coli O157:H7 1 CFU/ml
ground beef
samples
Uniform MM polystyrene Fluorescent sandwich assay [25, 26]
Salmonella 50 CFU/g
irrigation
water used in
the sprouting
of seeds
RAPTOR
uniform
Waveguide Fluorescent sandwich assay [27]
Salmonella 10(4) CFU/ml
Hotdog
samples
RAPTOR
uniform
Salmonella 10(4) CFU/mL Nutrient broth TT MM Fluorescent sandwich assay [28]

Table 1. Continued
Target Analyte LOD Matrix
Taper
Geometry
Fiber Type Detection Principle References
Staphylococcal Enterotoxin B
Min: 0.5 ng/ml
(buffer)
Buffer, human
serum, urine,
and aqueous
extract of ham
CTT ND Fluorescent sandwich assay [33]
C. Botulinum toxin A, Pseudexin
Toxin
Min: 30 pM (C.
Botulinum
toxin A), 60
pM (Pseudexin)
ND CTT MM Fluorescent sandwich assay [61]
Clostridium-Botulinum Toxin-A 5 ng/mL buffer TT MM Fluorescent sandwich assay [36]
E. coli lipopolysaccharide
endotoxin
Min: 10 ng/ml
Buffer and
plasma
CTT ND Fluorescent sandwich assay [38]
Ricin Concentration
Min: 100 pg/ml
(buffer)
Max: 1 ng/mL
(river water)
Buffer, river
water
CTT
MM (plastic clad
silica)
Fluorescent sandwich assay [37]
Listeria monocytogenes
5 x 10(5)
CFU/ml
frankfurter
sample
RAPTOR
uniform
5.4 x 10(7)
CFU/ml
Hotdog
samples
RAPTOR
uniform
4.3x10(3)
CFU/ml
Buffer Uniform MM polystyrene Fluorescent sandwich assay [29]
Abbreviations: BT = Biconical Taper, TT = Tapered Tip, CTT = Combination Taper Tip, SM = Single Mode, MM = Multimode, ND = Not Described, * = the change
in dB per hour per number of cells at inoculation, NA = Not Applicable.

Table 2. TFOBS For Biochemical Measurements
Target Analyte LOD Matrix Taper Geometry Fiber Type Detection Principle References
NADH, NADPH
Concentration
Min: 0.2 M
(NADH), 0.5
M (NADPH)
buffer BT SM Absorption [18]
Chinese Hamster Ovary
Cell Concentration
Min: 10
5

cells/ml
buffer BT SM Absorption [18]
Paraoxon Sub ppm buffer TT MM Chemiluminescence [62]
STAT3 ND Buffer Uniform MM
Fluorescent sandwich
assay
[63]
Abbreviations: BT = Biconical Taper, TT = Tapered Tip, CTT = Combination Taper Tip, SM = Single Mode, MM = Multimode, ND = Not Described.
Table 3. TFOBS For Clinical Measurements
Protein A 1 g/mL ND NA (chip) NA Leaky wave (SPR) [64]
BSA 10 fg/mL Buffer BT SM Intensity [1]
BSA 7.4 ng/mL buffer Chip (NA) Chip (NA) SPR [65]
BSA 2.5 g/ml Buffer BT
ND (plastic
clad silica)
Dye-protein complex
absorption
[43]
Ovalbumin 2.5 g/ml Buffer BT
ND (plastic
clad silica)
Dye-protein complex
absorption
[43]

Table 3. Continued
Hemoglobin 2.5 g/ml Buffer BT
ND (plastic
clad silica)
Dye-protein complex
absorption
[43]
IgG 20 fM Buffer TT MM
Fluorescent
competitive assay
[44]
IgG 75 pg/mL
Serum and jejunal
fluids diluted with
buffer
BBT SM
assay
[45]
Protein C 0.1 g/mL Buffer TT ND
assay
[16]
Protein C 0.5 g/mL Plasma TT MM
assay
[46]
Protein C 0.5 g/mL Plasma Uniform MM
assay
[66]
Protein S 0.5 g/mL Plasma Uniform MM
assay
[66]
Antithrombin III (ATIII) 30 g/mL Plasma Uniform MM
assay
[66]
Plasminogen (PLG) 30 g/mL Plasma Uniform MM
assay
[66]
B-type natriuretic peptide (BNP) 0.1 ng/mL Plasma Uniform MM
assay
[66]
cardiac troponin I (cTnI) 1 ng/mL Plasma Uniform MM
assay
[66]
C-reactive protein (CRP) 1 g/mL
Plasma

Uniform MM
assay
[66]

Table 3. Continued
Myoglobin (MG) 75 ng/mL Plasma Uniform MM
assay
[66]
L. donovani Antibody Concentration
Min: 0.244
ng/ml
Serum CTT
MM
(plastic clad
silica)
assay
[17]
Progesterone ng/mL Buffer ND ND
assay
[42]
Adriamycin 0.01 g/mL blood Straight core tip MM
Fluorescence
quenching
[49]
Cytochrome c ND Cell TT MM
assay
[50]
Cytochrome c 2.5 g/ml buffer BT
ND (plastic
clad silica)
Dye-protein complex
absorption
[43]
Yersinia pestis fraction 1 50 ng/mL
Buffer, serum,
plasma, and whole
blood
BT MM
assay
[15]
cTnI 31 pM plasma TT MM quartz
Nano gold particle
enhanced fluorescence
[67]
BNP 26 pM plasma TT MM quartz
Nano gold particle
enhanced fluorescence
[67]
Intracellular Benzopyrene Tetrol 6.4 pM cell TT ND Autofluorescence [51]
Benzo\c\phenanthridinium alkaoids ND buffer Chip NA SPR [68]
Fumonisin B
1
10 ng/ml
methanol/water-
extracted corn
TT
MM
(plastic clad
silica)
assay
[69]
Myoglobin 2.9 ng/mL buffer tip MM SPR [39]

Table 3. Continued
Myoglobin 5 nmol/L buffer Uniform probe MM
Fluorescent Energy
Transfer
[40]
Thrombin 1 nM Buffer Spheres (NA) NA
Fluorescent
competitive assay
[48]
Thrombin 1nM Buffer Uniform MM
Coagulation of
fluorescently labeled
fibrinogen to
unlabelled fibrinogen
bound to the surface of
the fibre optic
[47]
RNA pM Buffer TT SM Fluorescence [56]
DNA 70 fM Buffer Uniform MM Fluorescence [52]
interleukin-1 (IL-1), interleukin-6
(IL-6), and tumor necrosis factor-a.
(TNF-alpha)
1 ng/mL
Buffer and spiked
cell culture medium
(CCM)
ND MM
Fiber-optic surface
plasmon resonance
(SPR)
[41]
DNA 5 nM buffer ND MM Fluorescence [53]
Abbreviations: BT = Biconical Taper, TT = Tapered Tip, CTT = Combination Taper Tip, SM = Single Mode, MM = Multimode, ND = Not Described, NA = Not
Applicable.

3.1. Pathogen Detection
Escherichia coli O157:H7 [2, 23-26], Salmonella typhimurium [27, 28], Listeria
monocytogenes [29-31], and Bacillus anthracis [32] are some of the pathogens which have
been detected using TFOBS. Most pathogen detection studies done to date used fluorescence
TFOBS [25-32], but a few of them used intensity-based TFOBS [2, 23, 24].
Ferreira et al. developed an intensity-based evanescent sensor, to be used with a 840 nm
light source, to detect Escherichia coli O157:H7 growth [23]. This evanescent sensor was
fabricated by chemically etching. Transmission is reduced due to light absorption by the
bacteria, and the power loss is proportional to the intrinsic bulk absorption and scattering,
which depends on the concentration of the bacteria. The sensitivity of this sensor was 0.016
dB / hour-N
o
, where N
o
is initial cell concentration and ranges from 10 to 800. Similarly,
Maraldo et al. used TFOBS to detect Escherichia coli JM 101 growth on poly-L-lysine [24].
E.coli JM 101 expressing green fluorescent protein was immobilized on the poly-L-lysine
coated fibers, and growth was monitored by light transmission at 480 nm. The transmission
decreased exponentially with cell growth on the tapered surface. In a follow up study by Rijal
et al., Escherichia coli O157:H7 (EC) was covalently bonded to the surface of a TFOBS via
an antibody, and concentrations as low as 70 cells/mL was detected by changes in intensity at
470 nm [2]. Detection of EC in real samples is of great interest and was investigated by
DeMarco et al. [25]. EC in seeded ground beef samples was prepared and detected by a
sandwich immunoassay using cyanine 5 dye-labeled polyclonal anti-E. coli O157:H7. Light
was launched at 635 nm and the fluorescence was emitted at 670 to 710 nm. Responses were
obtained within 20 minutes, and E. coli O157:H7 at 3 to 30 CFU/mL were detected. A similar
study was recently conducted by Geng et al., where a sandwich immunoassay was used with
FOBS to detect EC in ground beef [26]. Light was launched at 635 nm and the fluorescence
was emitted at 670 to 710 nm. The sensor detected 10(3) CFU/ml of pure cultured EC grown
in culture broth. Artificially inoculated EC at concentration of 1 CFU/ml in ground beef was
detected after 4 hours of enrichment.
Kramer et al. [27] studied the detection of Salmonella typhimurium in sprout rinse water
using RAPTOR, an evanescent fluorescence sensor developed by Research International,
Monroe, Washington.. Alfalfa seeds contaminated with various concentrations of Salmonella
typhimurium were sprouted, and the sprout water was measured by the instrument.
Salmonella typhimurium was identified for seeds that were contaminated with 50 CFU/g.
Zhou et al. [28] also used a sandwich immunoassay to detect Salmonella. Light was launched
at 650 nm and the fluorescence was emitted at 680 nm. Tapered fiber tips with various
geometries and treatments were studied and optimized, and Salmonella was detected at 10(4)
CFU/mL.
An antibody-based sandwich fluorescence FOBS was developed by Geng et al. to detect
Listeria monocytogenes [29]. Light was launched at 635 nm and the fluorescence emission
was in the range of 670 to 710 nm. The sensor was specific, as shown by the significantly
lower signals caused by other Listeria species or microorganisms. The LOD was 4.3x10(3)
CFU/ml for a pure culture of L. monocytogenes. In less than 24 h, L. monocytogenes in hot
dog or bologna was detected at 10 to 1,000 CFU/g after enrichment. Recently, Kim et al. also
detected L. monocytogenes using the RAPTOR sensor [30]. This method achieved a LOD
of 5.4 x 10(7) CFU/ml. L. monocytogenes was detected in phosphate buffered saline (PBS) by
Nanduri et al. using RAPTOR to evaluate the effect of flow on antibody immobilization
[31]. Light was launched at 635 nm and the fluorescence was emitted at around 670 nm. It
was found that both the static and the flow through mode method had a LOD of 1 x 10(3)
CFU/ml. However, the effective disassociation constant and the binding valences for static
modes were higher than for flow through method of antibody immobilization. The flow
through mode was chosen to test real samples, and the LOD was 5 x 10(5) CFU/ml.
Bacillus anthracis, is a serious threat to national security. Tims et al. addressed the need
to detect Bacillus anthracis, and achieved detection at a concentration of 3.2 x 10(5)
spores/mg in spiked powders in less than 1 hour [32]. The method used was based on
fluorescent sandwich assay and a polystyrene tapered fiber. The excitation wavelength was
635 nm.
3.2. Toxin Measurement
TFOBS have been used to detect toxins such as enterotoxins [33-36], ricin [37], and
endotoxins [38]. Fluorescence was used for all the toxins measurements which are discussed
here.
Staphylococcal enterotoxins are a major cause of food poisoning. Tempelman et al.
quantified Staphyloccoccal enterotoxin B (SEB) in a fluorescent sandwich immunoassay on a
fiber optic biosensor [33]. A 635 nm diode laser was used to excite the labeled antibody. The
fluorescence level was measured and gave a detection limit of 0.5 ng/mL. Shriver-Lake et al.
used an array biosensor to detect SEB at a LOD of 0.5 ng/mL in buffer and six different types
of food samples [34]. Staphylococcus aureus is the only species which produces protein A
and was detected by Chang et al. using a fluorescent sandwich FOBS at a LOD of 1 ng/mL
[35]. Excitation of this sensor was at 488 nm. Similar to SEB, Clostridium botulinum toxin A
was detected by a fluorescent sandwich FOBS at 5 ng/mL [36]. A light source at 514 nm was
used in this case.
Narang et al. reported a sandwich fluorescent TFOBS ricin detection in buffer and in
river water [37]. The light source was 635 nm. Antibody to ricin was immobilized onto
tapered fiber surface using silanization and avidin-biotin linkage. The avidin-biotin method
had a higher sensitivity and wider linear dynamic range. The response of the avidin-biotin
sensor was linear in the range of 100 pg/mL to 250 ng/mL. The LOD for ricin in buffer
solution was 100 pg/mL, and in river water it is 1 ng/ml. At concentrations greater than 50
ng/ml, there was a strong interaction between ricin and avidin due to the lectin activity of
ricin. This interaction was reduced for fibers coated with neutravidin or with the addition of
galactose.
James et al. developed a method to detect lipopolysaccharide (LPS) endotoxin, which is
the most powerful immune stimulant and causes sepsis [38]. LPS from E. coli was detected at
a LOD of 10 ng/mL using fluorescent FOBS based on the competitive assay. Polymyxin B
was used as a recognition molecule and was covalently immobilized onto the surface of the
probe. Fluorescent labeled LPS was introduced to the fiber and attached to the Polymyxin B.
Unlabeled LPS was then introduced and competed with the labeled LPS for the binding sites
on the Polymyxin B. As LPS concentration increases, fluorescence decreases.
3.3. Clinical Measurements
Most clinical measurements done with TFOBS used proteins as analytes. Notable examples
include cardiac markers [39, 40], cytokines [41], and hormones [42]. Investigators have
detected model proteins using TFOBS in order to characterize TFOBS potential. Preejith et
al. detected model proteins using fiber optic evanescent wave spectroscopy [43]. They
immobilized Comassie Blue on a multimode fiber surface using a porous glass coating.
Comassie Blue normally absorbs at 467 nm, but it forms a dye-protein complex with the
protein when exposed to an acidic environment, and such a complex absorbs at 590 nm. The
protein concentration is inversely proportional to the output power at 590 nm, because
increase in protein concentrations causes the evanescent absorption to increase. Calibration
curves were obtained for BSA, hemoglobin, ovalbumin, and cytochrome c in the range of 0 to
20 g/mL. In our laboratory, BSA was recently detected at 10 fg/mL in stagnant condition
using intensity-based TFOBS [1]. Tromberg et al. detected antibody to IgG at 20 fM on a
fluorescent FOBS tip using a competitive assay [44]. Light was launched at 488 nm and the
fluorescence was emitted at 520 nm. Rabbit IgG was immobilized on the fiber tip, and
exposed to fluorescein isothiocyanate (FITC) labeled and unlabeled anti-IgG. The response
was inversely proportional to the amount of unlabeled anti-IgG, because the unlabeled anti-
IgG displaced the labeled one. Hale et al. developed a fluorescent optical fiber loop sensor to
detect antibody to IgG [45]. The sensor was used with a two-step sandwich assay. IgG was
labeled with the fluorescent dyes fluorescein isothiocyanate or tetramethyl rhodamine.
Antibody to IgG was detected at 75 pg/mL with this method.
Deficiency in Protein C (PC), if left untreated, may result in thrombotic complications,
and, thus presents an important clinical challenge. Spiker et al. detected PC at 0.1 g/mL in
buffer using a sandwich fluorescent fiber optic sensor [16]. Real-time detection of PC in
plasma is an important challenge in the clinical setting. Convective flow plays a vital role in
the transport of PC in a viscous medium such as plasma. Tang et al. who examined PC
detection in plasma with fluorescent sandwich FOBS and obtained a detection limit of 0.5
g/mL [46].
Cardiac markers myoglobin (MG) and cardiac tropinin I (cTnI) can be measured to
predict the occurrence of myocardial infarction, because they are released from cardiac
muscles when they are damaged. A fiber-optic SPR sensor was developed by Masson et al. to
detect MG and cTnI at 3 ng /mL [39]. A direct fluorescence FOBS was also used to detect
myoglobin at 5nM [40]. An excitation wavelength of 425 nm was used to excite the Cascade
Blue-labeled antibody, which was entrapped in the sensing element and fluoresces at 425 nm.
Fluorescence quenching occurred when myoglobin attaches to the labeled antibody. Recently,
Tang et al. developed a fiber-optic multi-analyte system which simultaneously quantifies two
groups of multi-biomarkers related to cardiovascular diseases (CVD): anticoagulants (protein
C, protein S, antithrombin III, and plasminogen) for deficiency diagnosis; and cardiac
markers (B-type natriuretic peptide, cardiac troponin I, myoglobin, and C-reactive protein) for
coronary heart disease diagnosis.
Garden et al. detected thrombin at 1 nM using fluorescent FOBS [47]. Excitation was at
495 nm and emission was at 520 nm. Unlabeled fibrinogen was first attached to the FOBS
surface. Then, coagulation of solution phase fluorescently labeled fibrinogen to unlabelled
fibrinogen bound to the surface was observed. Lee et al. detected thrombin at 1 nM using a
fluorescent FOBS immobilized with an antithrombin DNA aptamer receptor [48]. The
aptamer was immobilized on the surface of silica microspheres, which were distributed in
microwells on the distal tip of an imaging fiber that was coupled to a modified
epifluorescence microscope system. Another set of microspheres was prepared with a
different oligonucleotide to measure the non specific binding. The distal end of the imaging
fiber was incubated with fluorescein-labeled thrombin (F-thrombin), and the non-labeled
thrombin was detected using the competitive method.
Progesterone was found to have evidence of carcinogenicity based on animal studies.
Progesterone can be found in various surface waters commonly used for drinking water. In a
study by Tschmelak et al., a fluorescence FOBS was immobilized with a labeled-antibody
and used successfully to detect progesterone at concentrations lower than ng/L [42].
A fluorescent tip FOBS was used to measure adriamycin (ADM) at 10 ng/mL in vivo in a
blood vessel [49]. A polymeric fluorescent D-70 membrane with pore sizes of 1-2 m was
immobilized on the fiber tip. Fluorescence was quenched by ADM present in the blood and
the fluorescence signal was measured by a photomultiplier tube (PMT) at a wavelength of
530 nm.
The protein cytochrome c is involved in apoptosis and was detected by a sandwich
fluorescent nanobiosensor fabricated by Song et al. [50]. -Aminolevulinic acid (5-ALA), a
photodynamic therapy (PDT) drug, was activated by a He-Ne laser at 632.8 nm to induce
apoptosis in MCF-7 human breast carcinoma cells. When mitochondria are damaged by PDT,
cytochrome c is released into the cytoplasm; therefore cytochrome c concentration is an
indication of apoptosis. Results indicate that 5-ALA PDT-treated cells had a much higher
fluorescence signal, pointing to high cytochrome c concentrations in the treated cells.
Yersinia pestis is an etiologic agent of plague. A sandwich fluorescent FOBS devised by
Cao et al. was used to detect Yersinia pestis Fraction 1 antigen at a limit of 5 ng/mL [15]. The
light source was a 514 nm argon ion laser. This system detected 50 400 ng/mL of protein in
serum, and the results were in excellent agreement with ELISA results.
Nath et al. developed a fluorescent FOBS to detect L. donovani specific antibodies [17].
The sensor was made by de-cladding an optical fiber so that the evanescent wave propagated
outside the tapered region. The sensor was used with a 488 nm light source. Cell surface
protein of L. donovani was immobilized covalently on the sensing region. Then, the sensor
was incubated with patient serum for 10 minutes, followed by incubation with goat anti-
human IgG tagged with FTIC, which excites at 525 nm. The amount of L. donovani specific
antibodies in the patient serum was proportional to the fluorescence. There were no false
positive results from leprosy, tuberculosis, typhoid, and malaria serum.
Cullum et al. detected benzo[a]pyrene tetrol (BPT) at 6.4 1.7 E pM in mammary
carcinoma cells using a sandwich fluorescent fiber-optic nanosensor tip [51]. BPT is a
metabolite of benzo[a]pyrene. Using a 325 nm light source, the authors were able to calibrate
the sensor and obtain an unknown concentration by observing the level of fluorescence. This
technique is useful for cancer screening since carcinogens bind to DNA and form substances
such as BPT.
Three cytokines related to chronic wound healing are interleukin-1 (IL-1), interleukin-6
(IL-6), and tumor necrosis factor- (TNF-alpha) [41]. A fiber-optic SPR sensor was modified
with antibodies at the surface, and detected these proteins with LOD of 1 ng/mL in buffered
saline solution and spiked cell culture medium (CCM).
3.4. DNA Hybridization
Kleinjung et al. detected DNA hybridization at 3.2 attomoles (70 fM) using a fluorescent
multimode FOBS with 13-mer probe attached to the de-cladded core [52]. The
complementary strands were labeled and detected when introduced to the sensor. This sensor
was able to distinguish between matching sequences, single nucleotide mismatch, and
mismatch caused by additional deviations.
Zeng et al. examined the interfacial hybridization kinetics of oligonucleotides
immobilized onto silica using a fluorescent FOBS that was excited at 632 nm [53]. A dT20
DNA probe was used as recognition molecules, while target fluorescein-labeled non-
complementary DNA (ncDNA) dT20 and fluorescein-labeled dA20 were detected. The target
DNA concentrations were 5 nM to 0.1 M. The response of the sensor fit the second order
Langmuir model.
Molecular beacons (MB) are oligonucelotide probes that fluoresces upon hybridization
with target DNA or RNA molecules [54]. Liu et al. immobilized MB on a fluorescent FOBS
and determined the effects of ionic strength and target DNA concentration on hybridization
kinetics. Using an excitation wavelength of 514 nm, they found the LOD was 1.1 nM of
DNA. The sensor showed selectivity by distinguishing between 100 nM of ncDNA, 100 nM
of one-base mistmatch, and 100 nM of cDNA [54].
3.4.1. Pathogen Detection via DNA
A fluorescent FOBS was developed by Almadidy et al. to detect short sequences of
oligonucleotides that identify E. coli microbial contamination [55]. DNA probes were first
immobilized to silica surface via a silane reagent. Then, stepwise synthesis of
oligonucleotides by the -cyanoethyl-phosphoramidite protocol took place on the surface.
The sensor was exposed to both complementary (cDNA) and non-complementary (ncDNA)
20-mers, as well as genomic DNA from E.coli. The cDNA and ncDNA were introduced at a
concentration of about 1.7 nM, whereas genomic DNA was introduced at 1.7 pM to 170 pM.
Fluorescent intercalating dye was used to detect hybridization. Quantities as low as 100 fM
were detected using this method.
Pilevar et al. detected Helicobacter pylori total RNA using a fluorescent FOBS that had
probes immobilized on its surface [56]. IRD-41 is a near-infrared fluorophore which is
excited by 785 nm light. Real-time hybridization measurement of IRD 41-labeled
oligonucleotide at various concentrations to the surface bound probes was performed.
Complementary DNA at lower than nM concentration was detected. Sandwich assays were
performed with Helicobacter pylori total RNA, and results showed that this sensor could
detect H. pylori RNA in a sandwich assay at 25 pM.
4.0. Methods
4.1. Fabrication
Corguide fibers (Corning Glass Works, NY, attenuation at 1300 and 1500 nm of 0.36 and
0.26 dB km1, respectively) with a core diameter of 8 m and total diameter of 125 m were
used in all the fabrication methods described here. The fabrication methods commonly used
in our laboratory are chemical etching, heat pulling by flame, and heat pulling by fusion
splicer.
4.1.1. Chemical Etching
Chemical etching using hydrofluoric acid (HF) is one of the simplest ways to create tapers
with a step change in radius. Acrylic (Plexiglas) was used to construct the etching reactor
because HF does not attack most plastic materials. In order to monitor the etching, a
spectrofluorometer was used to detect the transmission through the fiber as etching took
place. This instrument has a compact 75 W Xenon arc lamp (Ushio Inc., Japan) coupled to a
monochromator and a PMT (model R1527P in housing 710, PTI Inc.) coupled to a
monochromator.
The plastic sheathing of the fiber was removed by immersing the fiber in acetone (Fisher
Scientific) for 1520 min followed by mechanical removal with a fiber optic stripper (NO-
NIK). A fiber optic cleaver (NO-NIK) was used to make a clean-cut fiber tip so as to enhance
the efficiency of light collection into the fiber.
HF (Fisher Scientific, Philadelphia) at a concentration of 49.5 wt.% was used. Two
hundred microliter of HF was introduced into the reaction chamber. Once HF was injected,
the transmission was monitored at 350 nm. When the diameter of the fiber was etched to a
certain fraction of the initial diameter, the etching process was stopped by first removing the
HF and then washing the chamber twice with 5 N NaOH as rapidly as possible. The fiber was
then immersed in a 200 mL of 0.1 N NaOH bath for 60 min to stabilize the fiber. If this step
was not carried out, any remaining HF would have continued to etch the fiber until it
dissolved completely. It was found that the length of the etching time needed at room
temperature was about 40 min.
4.1.2. Heat Pulling Using a Manual Propane Torch
Heat pulling using a micro-propane torch is another method of obtaining tapers. The
apparatus for this method is illustrated in Figure 4. The polymeric sheathing of a 30-35 cm
long fiber was removed similarly as in the chemical etching method. The fiber was then
mounted on the apparatus with two paper clips of identical weights (2.8 g) on either ends to
provide tension to the fiber. A micro-flame (Model 6000, Microflame, Inc., MN) was
positioned such that the fiber was approximately one third distance from the top end of the
visible end of the flame, and the flame was removed as soon as the paper clip touched the
stop. The ends of the tapered fiber were cleaved using a fiber-optic cleaver (NO-NIK) to give
clean cut ends. The fiber was placed in an optical fiber holder to be used in the experiments.
The dimensions of a fiber were measured after taking micrographs of the taper using an IMT-
2 optical microscope (Olympus, Japan) equipped with a video camera (Cohu Corp., Japan)
linked to a computer. The dimensions were measured in the Scion Image software (Scion
Corp., MD) after a calibration was performed according to the microscope objective used.

Fiber
Heat
4 mm
2 mm
Magnification
Microflame torch
Weight
Fiber Tapering stand
Fiber
Heat
4 mm
2 mm
Magnification
Microflame torch
Weight
Fiber Tapering stand

Figure 4. Some of the fibers discussed in this paper were fabricated using a simple fiber tapering
device. Fiber without sheathing was mounted on a stand and two pieces of weights were attached to the
ends to provide tension required for tapering. The fiber was then heated with a flame while carefully
monitoring the diameter of the taper. When the desired diameters were reached, the two ends of the
fibers were clipped and the tapered fiber were placed on an optical fiber holder to be used in the
experiments. Adapted from [2].
4.1.3. Heat Pulling Using a Fusion Splicer
The polymeric sheathing was removed over a distance of 5 cm at the center and both ends of
the fiber. The fiber was cleaned with isopropanol and the ends were cut clean using a fiber
cleaver (Ericsson EFC 11-4). The fiber was inserted into the programmable fusion splicer
(Ericsson FSU975), where electric current was applied via a pair of electrodes for up to 60
seconds while the taper was pulled automatically. Various current levels (3-13 mA) and pull
times (2-30 s) were used to produce fibers of varying taper diameters and lengths. A
micrograph of the fiber was taken via a camera inside the fusion splicer, and the dimensions
were measured in the Scion Image software (Scion Corp., MD).
4.2. Optical Characterization of Tapers
4.2.1. Preliminary Characterization Using Water
The preliminary characterization method used in our laboratory for determining the
evanescent field strength is the comparison of the transmission in water to that in air. The
reason for the choice of these two media was that they provide the most difference in
refractive index that is expected to be present in the waist region for biological samples. If a
taper exhibited little or no transmission change going from air to water, its transmission was
not expected to change significantly at the presence of dilute analyte solutions.
4.2.2. Characterization in the Visible Range Using E.coli JM101
Characterization in the visible range was performed on fusion spliced and torch heat-drawn
tapers using E.coli JM101 (ECJ) as the analyte. Tapers that showed little or no transmission
change in water compared to air, also showed no or low response to the presence of ECJ
suspensions. Both symmetric and asymmetric tapers of small and large waist diameters had
this behavior.
Several tapers exhibited a significant transmission difference (~50%) in water compared
to air. These tapers also showed little or no change in the presence of the ECJ. Small RI
changes due to the presence of ECJ suspension were not sufficient to produce the
transmission changes, resulting in poor sensitivity. Any impact on the light through the fiber
was already saturated from change due to water itself, such that the presence of ECJ
suspension had little further impact. Other tapers that had this characteristic property, but
were of smaller waist diameter showed weak sensitivity. Most of such tapers were symmetric
tapers. On the other hand, the asymmetric tapers showed lower relative transmission through
water, but allowed further modulation in transmission from the presence of ECJ.
We compared relative transmission at 470 nm for fusion splicer tapered fibers. In general,
there were two types of responses. In the first type, the transmission increased or decreased
monotonically, as ECJ concentration increased. In the second type, an initial increase for low
cell concentration is followed by a decrease at higher cell concentration. There were tapers
which also showed an increase in transmission at low ECJ concentrations, followed by
decrease at intermediate concentrations and then an increase at 7 million cells/mL.
Heat drawn tapers have typically a much longer convergent, waist and divergent sections,
each on the order of millimeters. Similar to the fusion spliced tapers, HD tapers showed two
basic characteristics. In one case, tapers showed a decrease in transmission as concentration
increased. In the other case, tapers showed a slight increase and then a decrease in
transmission for higher concentrations. At low concentration, the RI of cellular suspension
influenced transmission response, and caused the increase in transmission. At higher
concentrations, the evanescent light absorption by the cells dominated the response. These
results suggest that torch-drawn tapers have excellent potential as biosensors.
4.2.3. Characterization in the RI Range Using Glucose Solutions
In order to characterize the tapered fibers in the IR region, we measured transmission
properties under various RI fluids in the tapered region, using an experimental setup similar
to Figure 5 but in flow condition. The transmissions at 1310 and 1550 nm were monitored
and recorded simultaneously using a spectrum analyzer and LabView program. Once the
transmission stabilized in air, de-ionized (DI) water was flowed in at 0.5 mL/min. Glucose
solutions of various concentrations were then flowed past the taper, with de-ionize (DI) water
flowed in to rinse out the taper in between glucose solutions.
Ando AQ-6310B Spect rum Analyzer
Anrit su GB5A016
1550nm Laser
sensor
Thorlabs
TED200
Temperat ure
cont roller
FC-FC
adapt er
Thorlabs
LDC202
Laser diode
cont roller
Const ant t emperat ure
wat er bat h or incubat or
Reservoir
Ando AQ-6310B Spect rum Analyzer
Anrit su GB5A016
1550nm Laser
sensor
Thorlabs
TED200
Temperat ure
cont roller
FC-FC
adapt er
Thorlabs
LDC202
Laser diode
cont roller
Const ant t emperat ure
wat er bat h or incubat or
Reservoir

Figure 5. Experimental setup at 1550 nm in stagnant condition.
4.3. Antibody Immobilization
The most commonly used antibody immobilization method in our laboratory was adapted
from Hermanson [57] with modification for the fiber surface and geometry. Prior to
immobilization, the taper was cleaned with 1 M hydrochloric acid for 30 minutes, sulfuric
acid for 10 minutes, and 1 M sodium hydroxide for 10 minutes. The sample holder and taper
were rinsed several times with de-ionized water between cleaning steps. The cleaning
procedure produced reactive hydroxyl groups on tapered surface. The surface was then
silanylated with 3-aminopropyl-triethoxysilane (APTES; Sigma-Aldrich) in de-ionized water
for 2-24 hours. The fiber was then dried overnight in a vacuum oven at 40
o
C, or in a regular
oven at 75
o
C. The APTES reaction creates amine groups at the surface, which can further
react with carboxylic groups in the antibody to form a peptide bond. The polyclonal antibody
to BSA (anti-BSA; Sigma Catalog # B1520) contains carboxyl groups which were activated
using 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide (EDC; Sigma-Aldrich) and stabilized
by sulfo-N-hydroxysuccinimide (Sigma-Aldrich). EDC converts carboxylic groups into
reactive unstable intermediates which are susceptible to hydrolysis. However, Sulfo-NHS
replaces the EDC, resulting in a more stable reactive intermediate which catalyzes reaction
with amine groups. To prepare the antibody, 0.4 mg of EDC and 1.1 mg of sulfo-NHS was
added to each mL of antibody solution and the reaction was left on for 30 minutes at room
temperature. Then, 1.4 L of 2-mercaptoethanol was added to quench the EDC. This
intermediate was added to the silanylated tapered fiber surface and covalent bonding was
carried out at room temperature for 2 hours, in stagnant condition. At the end of antibody
immobilization, Hydroxylamine was added to regenerate the carboxylic groups of the
antibody. Transmission through the fiber was recorded during antibody immobilization and is
shown in Figure 6.

0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 20 40 60 80
time (min)
C
h
a
n
g
e

i
n

T
r
a
n
s
m
i
s
s
i
o
n

(
d
B
)
0
5
10
15
20
25
30
35
40
T
e
m
p
e
r
a
t
u
r
e

(
C
) Temperature
Transmission

Figure 6. Transmission change vs. time for antibody immobilization at 1550 nm. Temperature was held
constant at 30
o
C 0.5
o
C as indicated. Adapted from [1].
Alternatively, the antibody can be activated via carbohydrate groups. For this protocol, 1
mg/ml of antibody was dissolved in PBS and protected from light. Then, 100 L of 0.1 M
NaIO4 solution was added to antibody and allowed to react for 30 minutes. The silanized
fibers were exposed to the solution for 2 hours. Then, 10 L of NaCNBH3 was added for 30
minutes to reduce the Schiff Base to a second amine.
Another possible method of functionalization which is currently under investigation is the
use of Protein G with gold. In this method, the taper was first coated with a 1:1000/v:v
Polyurethane/Toluene mixture and dried overnight. The taper was then coated with 10 to 100
nm of gold using Denton Vacuum Desk IV system. After gold coating, the taper was
enclosed in the fiber holder by epoxy. During the first step of immobilization, Protein G was
flowed into the sample chamber and left there in stagnant condition for 90 minutes. The
sample chamber was then rinsed thoroughly with PBS, and antibody was flowed in and left
there for 90 minutes. Then, the chamber was rinsed thoroughly prior to using it in a detection
experiment.
4.4. Sample Preparation
All biological samples were prepared as per the instructions of the manufacturer using
solutions of 0.1% Sodium Azide in PBS as the solvent. Usually a bulk solution of antibody is
made and then aliquots of 4 mL are dispensed into sterilized scintillation vials. The vials are
then stored at -30 C freezer until use. The antibody vials were for single use and were
disposed at the end of the experiment. As for the analytes such as BSA and E.coli, they were
prepared in bulk in sterile centrifuge tubes each holding a maximum of 50 mL. Each tube
contained one concentration of analyte, and they were all stored at 4
o
C. The tubes were
placed back refrigerated at 4
o
C after each use.
4.5. Detection
4.5.1. E.coli O157:H7 in Stagnant Condition
E.coli O157:H7 (EC) was detected using a wavelength of 470 nm in stagnant conditions.
Tapers were fabricated using heat pulling by torch or fusion splicer. The surfaces of the tapers
were functionalized with antibody to E.coli O157:H7 using APTES and carboxylic linkage.
The taper was exposed to various concentrations of pathogen, and showed transmission
changes as the antigen attached.

0.97
0.99
1.01
1.03
1.05
1.07
0 5 10 15 20 25
Time, min
N
o
r
m
a
l
i
z
e
d

T
r
a
n
s
m
i
s
s
i
o
n
E coli 0157:H7 at 1e06/mL

0.81
0.82
0.83
0.84
0.85
0.86
0.87
0.88
27 29 31 33 35 37 39
Time, min
N
o
r
m
a
l
i
z
e
d

T
r
a
n
s
m
i
s
s
i
o
n
Release of E coli 0157:H7

Figure 7. Detection and release of 1 million cells/mL of EC on antibody immobilized tapered fiber. EC
detection and release experiment were performed on a 8.8 m diameter TFOBS. After attachment (top
panel), release buffer (glycine-HCl/ethylene glycol buffer, pH 1.7) was injected into the chamber to
release EC (bottom panel).
An EC stock solution (7x10
9
cells/mL) was prepared as per the vendors (KPL)
rehydration protocol in 10 mM PBS at pH 7.4. Lower concentrations (7x10
7
cells/mL, 7x10
5

cells/mL, 7x10
3
cells/mL, and 70 cells/mL) were prepared in PBS (pH 7.4) by serial dilution.
150 L of each sample was injected into the sample chamber. After EC attachment, the
sample was removed and the chamber was loaded with either HCl/PBS buffer at pH of 2.3 or
Glycine-HCl/ethylene glycol (1:1 v/v) buffer at pH 1.7.
The response due to attachment and release of EC cells are shown in Figure 6.
Immediately upon addition of the 1E6 cells/mL of EC sample, there was a rapid increase in
transmission due to the RI change of the medium. Subsequently, a gradual and exponential
decrease in transmission occurred due to EC attachment. Cells change the RI surrounding the
fiber and absorb light from the evanescent field. When the attachment reached equilibrium,
no further light is absorbed and the transmission remained constant.
The antigen attached to the sensor may be released by altering the pH as the antibody-
antigen binding is pH-dependent. The response due to release was equal in magnitude and
opposite in direction, as shown in Figure 7. This change occurred because cells released into
the bulk are too far away from the taper surface to influence light transmission. When cells
released reached equilibrium, the transmission reached a constant value.

1
1.05
1.1
1.15
1.2
1.25
1.3
1.35
1.4
1.45
0 10 20 30 40
Time, minutes
N
o
r
m
a
l
i
z
e
d

T
r
a
n
s
m
i
s
s
i
o
n

a
t

4
7
0

n
m
PBS
7000 #/mL
70 #/mL

Figure 8. The response at 470 nm due to different concentrations of EC cells. Adapted from [2].
Intuitively, one would imagine that the transmission change would be directly
proportional to the concentration. However, results show that the magnitude of the change is
inversely proportional to the pathogen concentration, as shown in Figure 8. In addition, the
response for this experiment was an increase in transmission, contrary to the experiment
shown in Figure 7. The cause of this is not entirely clear, but we believe that it is due to the
combined effects of evanescent absorption and scattering of the evanescent light. As cells
cover the taper surface, the evanescent light is absorbed by the cells in proportion to the
surface coverage. On the other hand, the RI is increased due to cell attachment. If the sample
were homogeneous, increase in refractive index tends to increase transmission through the
core due to reduction in the evanescent field. Hence, cell attachment result in transmission
increase or decrease.

0.92
0.94
0.96
0.98
1.00
1.02
1.04
1.06
0 1000 2000 3000 4000
Time, S
N
o
r
m
a
l
i
z
e
d

T
r
a
n
s
m
i
s
s
i
o
n
50% Wild
70% Pathogen
30% Wild
100% Wild
50% Pathogen
50% Wild

Figure 9. Effect of pathogenic and non-pathogenic EC mixture. Experiments were performed on a 9.5
m diameter TFOBS. When 0% pathogen (100% wild strain JM101) was injected around the taper,
there was no significant transmission change through the taper. When a solution containing an EC and
the wild strain is added to the solution, EC bind to the antibody thus resulting in a decrease in
transmission through the fiber. As the concentration of EC is increased to 50% and 70%, there is a
greater binding of pathogen to the antibody on the surface and thus greater change in transmission
occurs. Adapted from [2].
In order to evaluate specificity, the response to non-pathogenic E. coli was measured.
Stock solution containing EC was mixed with a wild strain of E. coli (JM101) in volumetric
proportions of 0%, 50% and 70%. The total bacterial count was 7x10
7
cells/mL. The
detection experiments were carried out in the same manner as with pure EC. The sensor
showed good selectivity to the pathogenic antigen as shown in Figure 9.
It is useful to obtain the kinetics EC attachment on antibody-immobilized surfaces. The
immobilization and detection responses show exponential behavior, similar to the adsorption
process often referred to as Langmuir kinetics. The Langmuir kinetics model can be
expressed as [58]:

1
o b s
k t
e

= (11)

where ( ) 1 0 is the fractional coverage of the reactive sites at time t. The
parameter,
obs
k , is the observed binding rate constant, which depends on the bulk
concentration of the reactant. We hypothesize that the transmission is indicative of
attachment, and express the Langmuir model as follows:

( ) ( )
( )
1
b
kC t
I I e
=
(12)

where ( ) I is the transmission change at time, t , ( )
I is the steady state transmission

change, and C
b
is the bulk concentration. Taking the natural log on both sides of Eq. (12) we
obtain:

( ) ( )
( )
ln
b
I I
kC t
I

(13)

-4
-3
-2
-1
0
0.E+00 1.E+08 2.E+08 3.E+08 4.E+08
Cb t, # bacteria-min/mL
l
n
[
(
I
*

I
)
/

I
*
)
]

Slope = 9.2 E-7 min
-1
(#/mL)
-1

Figure 10. Calculation of rate of attachment (slope k) for EC.
The above suggests that the characteristic rate constant k during initial time can be
determined from a plot of the left hand side versus C
b
*t in Eq. (13). Figure 10 is an example
of a graph displaying Eq. (13). The kinetic constant (k) was found to be in the range of 4x10
-9

min
-1
(pathogen/mL)
-1
to 7x10
-9
min
-1
(pathogen/mL)
-1
.
4.5.2. BSA in Stagnant Condition
Although we were able to detect E.coli O157:H7 at 470 nm, the sensitivity of the sensors was
limited due to the diameter in relation to the wavelength. Because the fibers are very fragile,
we are unable to fabricated tapers that are less than 5 m in diameter. However, at 470 nm the
penetration of the evanescent field is limited. Also, cells are relatively large compared to the
evanescent field generated at 470 nm. On the other hand, according to Eq. (1), there are
reasons to believe that the evanescent field would be larger at a longer wavelength. Therefore,
detection of BSA was detected similarly to EC but performed mostly using near-IR
wavelengths. We first reported the use of a 1550 nm laser with TFOBS to monitor the real-
time attachment of BSA to the antibody-immobilized surface [1]. While cuvette
measurements established that BSA was non-absorbing at 1550 nm, antibody-immobilized
TFOBS showed transmission changes at bulk concentrations of 10 fg/mL of BSA. The
experimental setup for near-IR detection is shown in Figure 5.
Solutions of BSA from 10 fg/mL to 1mg/mL were prepared. After antibody was
immobilized, it was rinsed with PBS, and 200 L of BSA was injected into the sample
chamber. Only one concentration was used in each experiment of attachment and release.
After attachment, the BSA was removed and the sensor was rinsed with PBS. Then, PBS
adjusted to a pH of 2 by H
2
SO
4
was added to release the BSA. The acidic PBS weakens the
binding of BSA to the antibody because it changes the conformation of the protein. After this
the fiber surface was regenerated with the cleaning sequence, followed by modification by
APTES.

-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0 5 10 15 20 25 30 35
Time (minutes)
C
h
a
n
g
e

i
n

T
r
a
n
s
m
i
s
s
i
o
n

(
d
B
)
Attachment
Release

Figure 11. BSA attachment and release of 10 pg/mL sample. The attachment response was obtained
when the tapered region was first exposed to 10 pg/mL of BSA. Data was collected for 30 minutes,
rinsed with PBS, and then the tapered region was exposed to pH2 PBS for BSA release. The
transmission changes due to attachment and release are in opposite direction and have approximately
the same magnitude. Adapted from [1].
When BSA was injected into the sample holder, transmission decreased due to change in
surface refractive index caused by the presence of BSA. When BSA was replaced by low pH
PBS, transmission increased back almost to the starting value. Similar experiments were
performed with many tapers using different concentrations, and we conclude that
experimental results are reproducible with the different fibers. The results of the attachment
and release of 10 pg/mL of BSA are shown in Figure 11.
Multiple step attachment experiments was performed on three TFOBS with up to five
different solutions of BSA, with concentration ranging from 100 fg/mL to 10 ng/mL. The
experiments were initially performed with a starting concentration of 100 fg/mL. In the last
experiment, shown in Figure 12, the initial concentration was set at 10 fg/mL. The BSA
solutions were added in order from the lowest to the highest concentration, with removal of
each sample after collection of data for up to 40 minutes. The transmission decreased as a
function of time as BSA attached to the antibody. The steady state transmission for each
concentration also decreased.

-34.3
-34.1
-33.9
-33.7
-33.5
-33.3
-33.1
-32.9
-32.7
50 100 150 200 250 300
Time (minutes)
T
r
a
n
s
m
i
s
s
i
o
n

(
d
B
)
19
21
23
25
27
29
31
T
e
m
p
e
r
a
t
u
r
e

(
o
C
)
Temperature
10 fg/mL
100 fg/mL
1 pg/mL
Transmission

Figure 12. Semi-batch staircase experiment showing attachment of BSA from10 fg/mL to 10 pg/mL.
Temperature was maintained at 30
o
C 0.5
o
C using an incubator. The BSA solutions were added
sequentially from lowest to highest upon removal of the previous solution. The purple line with peaks
represents transmission through the fiber. The peaks correspond to time instants when the samples were
introduced. The dotted line at the bottom represents the trend exhibited by the steady state transmission
with respect to time. Adapted from [1].
Like the EC results, transmission change is not linearly proportional to concentration. We
believe that the reason for this is that at low concentrations, the surface of the fiber is not
saturated with the antigen BSA. An estimate of the antibody/antigen surface coverage can be
made with a few simplifying assumptions, and it was suggested that the concentration
required for saturation is less than 4 ng/mL.
Transmission changes are caused by the evanescent field interaction with the surface
layer of antigen. Once the concentration approaches ng/mL levels, the surface is saturated
with BSA. Additional BSA molecules would attach on top of the surface layer. However, the
evanescent field magnitude decays away from the surface. Therefore the effect of BSA on top
of the first layer results in much smaller changes. In addition, the condition for
immobilization varies from one experiment to another. It is possible that nonlinearity was
observed because at the lowest concentration, the bulk refractive index is approximately the
same as that of PBS, and the BSA molecules on the fiber surface act as isolated points of high
refractive index. When the concentration increases to saturation point, the fiber surface is
covered with a layer of BSA which has a higher refractive index than PBS.
5.0. Conclusion
It is seen that TFOBS have several advantages in in terms of detection, including sensitivity,
selectivity, ease of use, affordability, ability for remote sensing, and small sample volumes.
They have been used for many applications such as pathogen detection, medical diagnostics
based on protein or cell concentration, and detection of DNA hybridization.
As far as the sensor physics is concerned, intensity-based sensors have been used to a
limited extent in cell detection. On the other hand, fluorescence based TFOBS are widely
used for protein and DNA detection because amplification is a convenient tool, and often
necessary to achieve low LODs. In addition, SPR is commonly used for protein
characterization and has also been used for the detection of DNA hybridization. The ng/mL
LOD of SPR makes it suitable for many medical applications. While fluorescence is very
selective, its LOD is higher than SPRs. In addition, fluorescence requires multiple steps for
the preparation of the sensor or the sample.
In terms of target analytes, one possible area of growth is the use of SPR or intensity-
based TFOBS protein and DNA detection. Another application may be drug screening using
TFOBS. Because of recent concerns of homeland security, there will likely be a significant
push for research in bio-threat detection. Pathogen detection also remains important in
maintaining a safe environment and food supply. Clinical applications of TFOBS will likely
be important as medical professionals seek convenient methods to diagnose diseases.
TFOBS have been used as intensity-based sensors in our laboratory. We have used three
methods of fabrication: step-etching using hydrofluoric acid, heat pulling by flame, and heat
pulling by fusion splicer. The sensing ability of TFOBS was characterized by measuring the
transmission in water, E.coli JM101 solutions, and glucose solutions. TFOBS were
functionalized with antibodies using covalent bonding or surface coating with gold and
Protein G.
TFOBS was used in our laboratory to measure E.coli O157:H7 in stagnant condition.
One surprising finding was that concentration had an inverse effect on the transmission.
TFOBS was shown to be selective to the pathogens. BSA was detected at 10 fg/mL in
stagnant condition at 1550 nm. Transmission data was fitted to the Langmuir absorption
model to determine the attachment rate.
As TFOBS evolves, new efforts will be focused on enhancing the sensitivity and
selectivity. Improved surface chemical modification and stability of the recognition molecule
can increase the sensitivity and robustness of TFOBS, especially for intensity-based TFOBS
because it is the most sensitive when molecules are bound to its surface. As was shown in this
chapter, there is a solid foundation of work to support the use of TFOBS and a wide variety of
applications. Given its promising advantages, it is likely that TFOBS will remain a popular
choice for detection in the future.
Acknowledgements
This work was supported through the National Science Foundation Grant # CBET-0329793,
Ultra Sensitive Continuous Tapered Fiber Biosensors for Pathogens and Bioterrorism Agents.
References
[1] Leung, A., Shankar, P.M., Mutharasan, R. (2006) Real-time monitoring of bovine serum
albumin at femtogram/mL levels on antibodyimmobilized tapered fibers Sens. Actuators
B: Chem. in press, doi:10.1016/j.snb.2007.03.010 .
[2] Rijal, K., Leung, A., Shankar, P.M., Mutharasan, R. (2005) Detection of vathoizen
Escherichia coli O157 : H7 AT 70 cells/mL using antibody-immobilized biconical
tapered fiber sensors. Biosensors & Bioelectronics 21: 871-880.
[3] Leung, A., Rijal, K.S., Shankar, P.M., Mutharasan, R. (2006) Effects of geometry on
transmission and sensing potential of tapered fiber sensors. Biosensors & Bioelectronics
21: 2202-2209.
[4] Arregui, F.J., Matias, I.R., Lopez-Amo, M. (2000) Optical fiber strain gauge based on a
tapered single-mode fiber. Sensors and Actuators a-Physical 79: 90-96.
[5] Datta, P., Matias, I., Aramburu, C., Bakas, A., LopezAmo, M., Oton, J.M. (1996)
Tapered optical-fiber temperature sensor. Microwave and Optical Technology Letters
11: 93-95.
[6] Bariain, C., Matias, I.R., Arregui, F.J., Lopez-Amo, M. (2000) Optical fiber humidity
sensor based on a tapered fiber coated with agarose gel. Sensors and Actuators B-
Chemical 69: 127-131.
[7] Diaz-Herrera, N., Navarrete, M.C., Esteban, O., Gonzalez-Cano, A. (2004) A fibre-optic
temperature sensor based on the deposition of a thermochromic material on an adiabatic
taper. Measurement Science & Technology 15: 353-358.
[8] Bariain, C., Matias, I.R., Arregui, F.J., Lopez-Amo, M. (2000) Tapered optical-fiber-
based pressure sensor. Optical Engineering 39: 2241-2247.
[9] Villatoro, J., Luna-Moreno, D., Monzon-Hernandez, D. (2005) Optical fiber hydrogen
sensor for concentrations below the lower explosive limit. Sensors and Actuators B-
Chemical 110: 23-27.
[10] Villatoro, J., Diez, A., Cruz, J.L., Andres, M.V. (2003) In-line highly sensitive hydrogen
sensor based on palladium-coated single-mode tapered fibers. Ieee Sensors Journal 3:
533-537.
[11] Bariain, C., Matias, I.R., Romeo, I., Garrido, J., Laguna, M. (2000) Detection of volatile
organic compound vapors by using a vapochromic material on a tapered optical fiber.
Applied Physics Letters 77: 2274-2276.
[12] Kim, Y.C., Peng, W., Banerji, S., Booksh, K.S. (2005) Tapered fiber optic surface
plasmon resonance sensor for analyses of vapor and liquid phases. Optics Letters 30:
2218-2220.
[13] McCue, R.P., Walsh, J.E., Walsh, F., Regan, F. (2006) Modular fibre optic sensor for the
detection of hydrocarbons in water. Sensors and Actuators B-Chemical 114: 438-444.
[14] Zalvidea, D., Diez, A., Cruz, J.L., Andres, M.V. (2006) Hydrogen sensor based on a
palladium-coated fibre-taper with improved time-response. Sensors and Actuators B-
Chemical 114: 268-274.
[15] Cao, L.K., Anderson, G.P., Ligler, F.S., Ezzell, J. (1995) Detection of Yersinia-Pestis
Fraction-1 Antigen with a Fiber Optic Biosensor. Journal of Clinical Microbiology 33:
336-341.
[16] Spiker, J.O., Kang, K.A. (1999) Preliminary study of real-time fiber optic based protein
C biosensor. Biotechnology and Bioengineering 66: 158-163.
[17] Nath, N., Jain, S.R., Anand, S. (1997) Evanescent wave fibre optic sensor for detection
of L-donovani specific antibodies in sera of kala azar patients. Biosensors &
Bioelectronics 12: 491-498.
[18] Haddock, H.S., Shankar, P.M., Mutharasan, R. (2003) Evanescent sensing of
biomolecules and cells. Sensors and Actuators B-Chemical 88: 67-74.
[19] Ko, S.H., Grant, S.A. (2006) A novel FRET-based optical fiber biosensor for rapid
detection of Salmonella typhimurium. Biosensors & Bioelectronics 21: 1283-1290.
[20] Bobb, L.C., Shankar, P.M., Krumboltz, H.D. (1990) Bending Effects in Biconically
Tapered Single-Mode Fibers. Journal of Lightwave Technology 8: 1084-1090.
[21] Okamoto, K. (2000): In Fundamentals of Optical Waveguides, Academic Press, pp.
323-330.
[22] Moar, P.N., Huntington, S.T., Katsifolis, J., Cahill, L.W., Roberts, A., Nugent, K.A.
(1999) Fabrication, modeling, and direct evanescent field measurement of tapered
optical fiber sensors. Journal of Applied Physics 85: 3395-3398.
[23] Ferreira, A.P., Werneck, M.M., Ribeiro, R.M. (2001) Development of an evanescent-
field fibre optic sensor for Escherichia coli O157 : H7. Biosensors & Bioelectronics 16:
399-408.
[24] Maraldo, D., Shankar, P.M., Mutharasan, R. (2006) Measuring bacterial growth by
tapered fiber and changes in evanescent field. Biosensors & Bioelectronics 21:
1339-1344.
[25] DeMarco, D.R., Saaski, E.W., McCrae, D.A., Lim, D.V. (1999) Rapid detection of
Escherichia coli O157 : H7 in ground beef using a fiber-optic biosensor. Journal of Food
Protection 62: 711-716.
[26] Geng, T., Uknalis, J., Tu, S.I., Bhunia, A.K. (2006) Fiber-optic biosensor employing
Alexa-Fluor conjugated antibody for detection of Escherichia coli O157 : H7 from
ground beef in four hours. Sensors 6: 796-807.
[27] Kramer, M.F., Lim, D.V. (2004) A rapid and automated fiber optic-based biosensor
assay for the detection of salmonella in spent irrigation water used in the sprouting of
sprout seeds. Journal of Food Protection 67: 46-52.
[28] Zhou, C.H., Pivarnik, P., Auger, S., Rand, A., Letcher, S. (1997) A compact fiber-optic
immunosensor for Salmonella based on evanescent wave excitation. Sensors and
Actuators B-Chemical 42: 169-175.
[29] Geng, T., Morgan, M.T., Bhunia, A.K. (2004) Detection of low levels of Listeria
monocytogenes cells by using a fiber-optic immunosensor. Applied and Environmental
Microbiology 70: 6138-6146.
[30] Kim, G., Morgan, M.T., Ess, D., Hahm, B.K., Kothapalli, A., Valadez, A., Bhunia, A.
(2006) Detection of Listeria monocytogenes using an automated fiber-optic biosensor:
RAPTOR. Advanced Nondestructuve Evaluation I, Pts 1 and 2, Proceedings 321-323:
1168-1171.
[31] Nanduri, V., Kim, G., Morgan, M.T., Ess, D., Hahm, B.K., Kothapalli, A., Valadez, A.,
Geng, T., Bhunia, A.K. (2006) Antibody immobilization on waveguides using a flow-
through system shows improved Listeria monocytogenes detection in an automated fiber
optic biosensor: RAPTOR (TM). Sensors 6: 808-822.
[32] Tims, T.B., Lim, D.V. (2004) Rapid detection of Bacillus anthracis spores directly from
powders with an evanescent wave fiber-optic biosensor. Journal of Microbiological
Methods 59: 127-130.
[33] Tempelman, L.A., King, K.D., Anderson, G.P., Ligler, F.S. (1996) Quantitating
staphylococcal enterotoxin B in diverse media using a portable fiber-optic biosensor.
Analytical Biochemistry 233: 50-57.
[34] Shriver-Lake, L.C., Shubin, Y.S., Ligler, F.S. (2003) Detection of staphylococcal
enterotoxin B in spiked food samples. Journal of Food Protection 66: 1851-1856.
[35] Chang, Y.H., Chang, T.C., Kao, E.F., Chou, C. (1996) Detection of protein a produced
by Staphylococcus aureus with a fiber-optic-based biosensor. Bioscience Biotechnology
and Biochemistry 60: 1571-1574.
[36] Ogert, R.A., Brown, J.E., Singh, B.R., Shriverlake, L.C., Ligler, F.S. (1992) Detection of
Clostridium-Botulinum Toxin-a Using a Fiber Optic-Based Biosensor. Analytical
Biochemistry 205: 306-312.
[37] Narang, U., Anderson, G.P., Ligler, F.S., Burans, J. (1997) Fiber optic-based biosensor
for ricin. Biosensors & Bioelectronics 12: 937-945.
[38] James, E.A., Schmeltzer, K., Ligler, F.S. (1996) Detection of endotoxin using an
evanescent wave fiber-optic biosensor. Applied Biochemistry and Biotechnology 60:
189-202.
[39] Masson, J.F., Obando, L., Beaudoin, S., Booksh, K. (2004) Sensitive and real-time fiber-
optic-based surface plasmon resonance sensors for myoglobin and cardiac troponin I.
Talanta 62: 865-870.
[40] Hanbury, C.M., Miller, W.G., Harris, R.B. (1997) Fiber-optic immunosensor for
measurement of myoglobin. Clinical Chemistry 43: 2128-2136.
[41] Battaglia, T.M., Masson, J.F., Sierks, M.R., Beaudoin, S.P., Rogers, J., Foster, K.N.,
Holloway, G.A., Booksh, K.S. (2005) Quantification of cytokines involved in wound
healing using surface plasmon resonance. Analytical Chemistry 77: 7016-7023.
[42] Tschmelak, J., Proll, G., Gauglitz, G. (2004) Sub-nanogram per litre detection of the
emerging contaminant progesterone with a fully automated immunosensor based on
evanescent field techniques. Analytica Chimica Acta 519: 143-146.
[43] Preejith, P.V., Lim, C.S., Kishen, A., John, M.S., Asundi, A. (2003) Total protein
measurement using a fiber-optic evanescent wave-based biosensor. Biotechnology
Letters 25: 105-110.
[44] Tromberg, B.J., Sepaniak, M.J., Vo-Dinh, T., Griffin, G.D. (1997) Fiber-optic chemical
sensors for competitive binding fluoroimmunoassay. Anal. Chem. 59: 1226-1230.
[45] Hale, Z.M., Payne, F.P., Marks, R.S., Lowe, C.R., Levine, M.M. (1996) The single
mode tapered optical fibre loop immunosensor. Biosensors & Bioelectronics 11:
137-148.
[46] Tang, L., Kwon, H.J., Kang, K.A. (2004) Theoretical and experimental analysis of
analyte transport in a fiber-optic, Protein C immuno-biosensor. Biotechnology and
Bioengineering 88: 869-879.
[47] Garden, S.R., Doellgast, G.J., Killham, K.S., Strachan, N.J.C. (2004) A fluorescent
coagulation assay for thrombin using a fibre optic evanescent wave sensor (vol 19, pg
737, 2004). Biosensors & Bioelectronics 19: 1775-1775.
[48] Lee, M., Walt, D.R. (2000) A fiber-optic microarray biosensor using aptamers as
receptors. Analytical Biochemistry 282: 142-146.
[49] Wen-xu, L., Jian, C. (2003) Continuous monitoring of adriamycin in vivo using fiber
optic-based fluorescence chemical sensor. Analytical Chemistry 75: 1458-1462.
[50] Song, J.M., Kasili, P.M., Griffin, G.D., Vo-Dinh, T. (2004) Detection of cytochrome c in
a single cell using an optical nanobiosensor. Analytical Chemistry 76: 2591-2594.
[51] Cullum, B.M., Griffin, G.D., Miller, G.H., Vo-Dinh, T. (2000) Intracellular
measurements in mammary carcinoma cells using fiber-optic nanosensors. Analytical
Biochemistry 277: 25-32.
[52] Kleinjung, F., Bier, F.F., Warsinke, A., Scheller, F.W. (1997) Fibre-optic genosensor for
specific determination of femtomolar DNA oligomers. Analytica Chimica Acta 350:
51-58.
[53] Zeng, J., Almadidy, A., Watterson, J., Krull, U.K. (2003) Interfacial hybridization
kinetics of oligonucleotides immobilized onto fused silica surfaces. Sensors and
Actuators B-Chemical 90: 68-75.
[54] Liu, X., Tan, W. (1999) A fiber-optic evanescent wave DNA biosensor based on novel
molecular beacons. Anal. Chem. 71: 5054-5059.
[55] Almadidy, A., Watterson, J., Piunno, P.A.E., Foulds, I.V., Horgen, P.A., Krull, U.
(2003) A fibre-optic biosensor for detection of microbial contamination. Canadian
Journal of Chemistry-Revue Canadienne De Chimie 81: 339-349.
[56] Pilevar, S., Davis, C.C., Portugal, F. (1998) Tapered optical fiber sensor using near-
infrared fluorophores to assay hybridization. Analytical Chemistry 70: 2031-2037.
[57] Hermanson, G.T.: Bioconjugate Teshniques. Elsevier, San Diego 1996.
[58] Yan, D., Saunders, J.A., Jennings, K. (2002) Kinetics of formation for n-alkanethiolates
self-assembled monolayers onto gold in aqueous micellar solutions of C12E6 and
C12E7. Langmuir 18: 10202.
[59] Zourob, M., Mohr, S., Brown, B.J.T., Fielden, P.R., McDonnell, M.B., Goddard, N.J.
(2005) An integrated metal clad leaky waveguide sensor for detection of bacteria.
Analytical Chemistry 77: 232-242.
[60] Morgan, M.T., Kim, G., Ess, D., Kothapalli, A., Hahm, B.K., Bhunia, A. (2006) Binding
inhibition assay using fiber-optic based biosensor for the detection of foodborne
pathogens. Advanced Nondestructuve Evaluation I, Pts 1 and 2, Proceedings 321-323:
1145-1150.
[61] Shriverlake, L.C., Ogert, R.A., Ligler, F.S. (1993) A Fiberoptic Evanescent-Wave
Immunosensor for Large Molecules. Sensors and Actuators B-Chemical 11: 239-243.
[62] Chen, Z., Kaplan, D.L., Gao, H., Kumar, J., Marx, K.A. (1996) Molecular assembly of
multilayer enzyme: toward the development of a chemiluminescent-based fiber optic
biosensor. Materials Science and Engineering C4: 155-159.
[63] Kapoor, R., Kaur, N., Nishanth, E.T., Halvorsen, S.W., Bergey, E.J., Prasad, P.N. (2004)
Detection of trophic factor activated signaling molecules in cells by a compact fiber-
optic sensor. Biosensors & Bioelectronics 20: 345-349.
[64] Zourob, M., Goddard, N.J. (2005) Metal clad leaky waveguides for chemical and
biosensing applications. Biosensors & Bioelectronics 20: 1718-1727.
[65] Wu, S.Y., Ho, H.P., Law, W.C., Lin, C.L., Kong, S.K. (2004) Highly sensitive
differential phase-sensitive surface plasmon resonance biosensor based on the Mach-
Zehnder configuration. Optics Letters 29: 2378-2380.
[66] Tang, L., Ren, Y.J., Hong, B., Kang, K.A. (2006) Fluorophore-mediated, fiber-optic,
multi-analyte, immunosensing system for rapid diagnosis and prognosis of
cardiovascular diseases. Journal of Biomedical Optics 11: 21011-21011 to 21011-21010.
[67] Hong, B., Kang, K.A. (2006) Biocompatible, nanogold-particle fluorescence enhancer
for fluorophore mediated, optical immunosensor. Biosensors & Bioelectronics 21:
1333-1338.
[68] Minunni, M., Tombelli, S., Mascini, M., Bilia, A., Bergonzi, M.C., Vincieri, F.F. (2005)
An optical DNA-based biosensor for the analysis of bioactive constituents with
application in drug and herbal drug screening. Talanta 65: 578-585.
[69] Thompson, V.S., Maragos, C.M. (1996) Fiber-optic immunosensor for the detection of
fumonisin B-1. Journal of Agricultural and Food Chemistry 44: 1041-1046.


Chapter 2

NEW CHALLENGES IN RAMAN AMPLIFICATION
FOR FIBER COMMUNICATION SYSTEMS

P.S. Andr
1,2
, A.N. Pinto
1,3
, A.L.J. Teixeira
1,3
, B. Neto
1,2
,
S. Stevan Jr.
1,3
, Donato Sperti
1,3,4
, F. da Rocha
1,3
,
Micaela Bernardo
2,5
, J.L. Pinto
1,2
, Meire Fugihara
1,3
,
Ana Rocha
1,2
and M. Faco
2

1
Instituto de Telecomunicaes, Aveiro Portugal
2
Departamento de Fsica, Universidade de Aveiro, Aveiro, Portugal
3
Departamento de Electrnica, Telecomunicaes e Informtica,
Universidade de Aveiro, Aveiro, Portugal
4
Universit Degli Studi di Parma, Parma, Italy
5
Portugal Telecom Inovao SA, Aveiro, Portugal
Abstract
Raman fiber amplifiers (RFA) are among the most promising technologies in lightwave
systems. In recent years, Raman optical fiber ampliers have been widely investigated for
their advantageous features, namely the transmission fiber can be itself used as the gain media
reducing the overall noise figure and creating a lossless transmission media. The introduction
of RFA based on low cost technology will allow the consolidation of this amplification
technique and its use in future optical networks.
This paper reviews the challenges, achievements, and perspectives of Raman
amplification in optical communication systems. In Raman amplified systems, the signal
amplification is based on stimulated Raman scattering, thus the peak of the gain is shifted by
approximately 13.2 THz with respect to the pump signal frequency. The possibility of
combining many pumps centered on different wavelengths brings a flat gain in an ultra wide
bandwidth.
An initial physical description of the phenomenon is presented as well as the
mathematical formalism used to simulate the effect on optical fibers.
The review follows with one section describing the challenging developments in this
topic, such as using low cost pump lasers, in-fiber lasing, recurring to fiber Bragg grating
cavities or broadband incoherent pump sources and Raman amplification applied to coarse
wavelength multiplexed networks. Also, one of the major issues on Raman amplifier design,
P.S. Andr, A.N. Pinto, A.L.J. Teixeira et al. 52
which is the determination of pump powers in order to realize a specific gain will be
discussed. In terms of optimization, several solutions have been published recently, however,
some of them request extremely large computation time for every interaction, what precludes
it from finding an optimum solution or solve the semi-analytical rate equation under strong
simplifying assumptions, which results in substantial errors. An exhaustive study of the
optimization techniques will be presented.
This paper allows the reader to travel from the description of the phenomenon to the
results (experimental and numerical) that emphasize the potential applications of this
technology.
1. Introduction
The deployment of optical communication systems through long haul networks required the
development of transparent optical amplifiers, for replacement of the expensive and limitative
optoelectronic regeneration. The increasing distance between amplification sites saves
amplification huts reducing by this way the investment and operational cost in the network
management.
The first choice for transparent optical amplification pointed out to the Erbium Doped
Fiber amplifiers (EDFA), which was a mature technology by the beginning of the last decade
of the XX century. However, the growing demand in terms of transmission capacity has been
increasing dramatically, fulfilling the entire spectral band of the EDFA, and wideband
amplifiers are now required. Raman fiber amplifiers (RFA) have emerged as a key technology
for the optical networks.
In lumped amplified systems (using for example EDFAs) the amplification modules are
placed every 40~50 km of span. This module amplifies back to the initial power level, the
transmission signal attenuated during propagation. The distance between amplifiers is
determined by the span loss, by the limit imposed from the maximum admissible power
allowed in the fiber without inducing nonlinear effects and by the minimum acceptable power
that avois a degradation of the signal-to-noise-ratio.
The use of Raman amplification allows the confinement of the signal inside the limits
imposed by the nonlinearities and of the signal-to-noise-ratio degradation resulting from
higher span distances. This advantage of the distributed (Raman) over lumped amplification
is illustrated in figure 1.

Figure 1. Distributed and lumped amplification signal evolution.
New Challenges in Raman Amplification for Fiber Communication Systems 53
The distributed amplification scheme can be used to cover very long span links or to
increase the distance of ultra-long haul systems.
Raman fiber amplifiers are based on the power transfer from pump(s) signal(s) to
information carrying signals (usually described as probes) due to stimulated Raman scattering
(SRS) which occurs when there is sufficient pump power within the fiber. Since the gain peak
of this amplification is obtained for signals downshifted approximately by 13.2 THz (for
Silica), relative to the pump frequency, to achieve gain at any wavelength we need to select a
pump whose frequency complies with this relation. In this way, it is possible to optimize the
number of pumps to obtain a wide and flat gain [1-3]. However, it is necessary to bear in
mind, that due to the pump-to-pump interaction, the shorter wavelength pumps demand more
power to be effective [4,5].
From a telecommunications point of view, the pump wavelengths must be placed around
1450 nm because the signal wavelengths used on the so called 3
rd
transmission window are
centered around 1550 nm and the maximum gain occurs for a Stokes frequency shift of
13.2 THz.
The RFA had become attractive just after the development and commercialization at a
reasonable cost of a key component: the high power pump laser [6]. Typically a high power
laser for Raman amplification, provides an optical power of 300 mW, launched over an
optical fiber, which for a standard single mode fiber (SMF) is equivalent to a power density
of 3.75 GW/m
2
. This high power injected into the fiber, especially when multipump lasers are
utilized, imposes new concerns in terms of safety.
Therefore, the use of RFAs requires the utilization of automatic power reduction or
automatic laser shut down systems to prevent the hazard of high power leakage from the
optical cables or service cabinets. Also, as the optical power rises, the nonlinear effects, such
as the fiber fuse effect, start to become relevant. This effect has threshold intensity of 10~30
GW/m
2
and it is responsible by a catastrophic destruction of the fiber core. This destruction
once started propagates in direction to the optical source, resulting also in the destruction of
the pumping laser [7]. For operating wavelengths of 1550 nm, the fuse effect power threshold
is ~1.5 W for SMF fibers, while for dispersion shift fibers (DSF) this power is reduced to
~1.2 W [8]. This effect is also responsible by the damage of the optical connectors
interface [8].
In terms of implemented systems, several architectures have been proposed, based in all
Raman or hybrid Raman/EDFA amplification [10]. The use of bidirectional Raman
amplification has also been reported for long reach access networks. Experimental results
have shown the feasibility of systems with symmetric up-and-downstream signals with
bitrates up to 10 Gb/s, supported by distributed Raman amplification over 80 km of fiber [11].
Field transmission experiments have been reported with 8 170 Gb/s over 210 km of single
mode standard fiber, achieving spectral efficiency of 0.53 bit/s/Hz [12].
As the traffic increases, wavelength division multiplexing (WDM) arises to enlarge the
transmission capacity. This, in turn, requires flexible and broadband architectures which
reinforces the interest in Raman amplification. Nowadays, WDM exists in two formats:
Dense WDM (DWDM) working at C and L spectral windows, allocating a maximum of 150
channels spaced by 0.8 nm [13], and Coarse WDM (CWDM) working at O, E, S, C, and L
spectral windows, allocating a maximum of 18 channels spaced by 20 nm [14]. The DWDM
solution is extensively used in long haul systems, sending as much information as possible.
CWDM is a good solution whenever less information is transmitted over short distances in a
less expensive way than DWDM. As CWDM works with far apart signals, it can make use of
uncooled distributed feedback (DFB) lasers [15,16] needing multiplexing components with
flexible tolerances. However, as the channels in CWDM systems are far apart, optical
amplification is still a matter of concern. Traditional EDFA bandwidth (20~40 nm) cannot
support the full band of CWDM channels [17].
Other technical solution to amplification of signals is the semiconductor optical amplifier,
which presents a low saturation power (around 13 dBm) when compared with other fiber
based amplifiers, but with a signal-to-noise ratio degradation quite considerable. A good
solution for the amplification of both DWDM and CWDM relies on Raman amplifiers. A
wide and flat spectral gain profile is achievable thanks to the combination of several pumping
lasers operating at specific powers and wavelengths. The composite amplification is
determined from the mutual interactions among the pump and signal wavelengths. Gain
spectra as large as 100 nm were obtained using multiple pumps. Emori et al. have presented
an experimental Raman amplifier with a 100 nm bandwidth using a WDM laser diode unit
with 12 wavelengths ranging from 1405 to 1510 nm, whose maximum total power was equal
to 2.2 W [4, 18]. Therefore, a gain equal to 2 dB is obtained over a 25 km SMF link and a 6.5
dB gain using a 25 km DSF link, both with 0.5 dB of maximum ripple. Kidorf et al. provided
a mathematical model to implement a 100 nm Raman amplifier using low power pumps with
maximum power of each pump equal to 130 mW [14]. They used 8 pumps from 1416 nm to
1502 nm along 45 km of SMF, obtaining a gain around 4 dB with a maximum ripple equal to
1.1 dB.
The growing maturity of high pump module technologies is providing competitive
solutions based on Raman amplification and currently many alternative techniques are being
developed to overcome the ordinary one pump and dual pumping methods [19, 20]. In
particular, we report here two major techniques. First, the use of low power pumping lasers
provides gain comparable to the ordinary one pump Raman amplification. This technique is
especially interesting for combining commercial and low cost lasers [21]. The second
particular technique corresponds to an evolution of the cascaded Raman amplification.
Actually, a sixth order cascade Raman amplifier was recently proposed [22]. In the cascade
Raman amplification, the pump power is downshifted in frequency by using a pair of fiber
Bragg gratings (FBG) placed in spectral positions multiples of 13 THz, from the pump
frequency. In a particular case, the generation of the fiber pump laser is obtained by using
only one passive reflector element and distributed reflectors over the long optical fiber,
established by a nonlinear fiber intrinsic effect called Rayleigh backscattering.
The enlargement of the bandwidth of Raman amplifiers is also achieved using incoherent
pumping instead of multi-pump schemes [23-27]. Vakhshoori et al. proposed a high-power
incoherent semiconductor pump prototype that uses a low-power seed optical signal, coupled
into a long-cavity semiconductor amplifier. It was achieved 400mW of optical power over a
35nm spectral window [27]. A 50 nm bandwidth amplifier was obtained with an on/off gain
equal to 7 dB. It was also demonstrated that the use of six coherent pumps is less efficient, in
terms of flatness, than the use of two incoherent pumps [24]. The signal wavelengths were
comprised between 1530 nm and 1605 nm and the transmission occurs over 100 km of optical
fiber. Another advantage of using incoherent pumping is the reduction of nonlinear effects,
such as Brillouin scattering, four wave mixing of pump-pump, pump-signal and pump-noise
[28].
RFAs have become a crucial component for the implementation of fiber optic
communication systems [9]. An exponential increase on the product distance capacity of
the transmission experiments on optical communication systems was observed in the last
decade. The majority of these experiments, especially since the year 2000, have employed
RFA as amplification technology [9]. This survey attempts to cover the most recent aspects in
the field of Raman amplification for fiber communication systems.
2. Theoretical Description of Raman Scattering
In 1928 Raman scattering was discovered independently and almost simultaneously by two
research groups, one working in India and lead by Sir C. V. Raman [29], and the other by G.
S. Landsberg and L. I. Mandelstam working in Russia [30]. In 1930, the Nobel committee
distinguished Sir C. V. Raman for his discovery of the molecular scattering of light and since
then this effect has been known as the Raman effect.
Raman effect is a scattering effect of light. Light scattering occurs as a consequence of
fluctuations in the optical properties of a medium. In optical fibers three types of scattering
effects are relevant: Rayleigh, Brillouin and Raman scattering.
Rayleigh scattering is an elastic process, i.e., the incident and the scattered photon have
the same energy, therefore the same frequency. Rayleigh scattering in fibers couples light
from guided modes to unguided ones leading to optical attenuation. Indeed, in modern fibers
operating in the near infrared, Rayleigh scattering is the major source of attenuation, as
absorption is practically negligible. In fact, Silica lattice and electronic resonances are in the
mid infrared and in the ultra-violet, respectively. Therefore in the near infrared, fibers
operate, essentially in an off-resonance regime, apart from impurities, which in nowadays
fibers are reduced to an extremely low level [31]. However, besides the off-resonant
interaction with bound electrons, optical waves also interact with molecules inside Silica
fibers, through scattering.
Raman and Brillouin scattering are both inelastic processes, i.e., the incident and
scattering photons have different energies. The energy lost by the incident field is stored into
the medium in the form of vibrational energy, named phonons. Indeed, the origin of both
Raman and Brillouin scattering effects resides in the interaction of light with these vibrational
states (phonons). In the Brillouin scattering low frequency vibrational states are involved,
usually referred as acoustic phonons. In the Raman process high frequency vibrational states
are presented, named as optical phonons.
Raman scattering can occur in two distinguished forms: Spontaneous Raman Scattering,
and Stimulated Raman Scattering (SRS).
In the spontaneous form, Raman scattering occurs when the incident field interacts with
vibrational modes, mainly excited by thermal effects, of the molecules constituting the
medium. From this interaction, it can result another optical phonon, with frequency , and a
down shifted optical photon with frequency
0
=
S
, or a up shifted photon of
frequency
0
= +
A
and in this case an optical phonon is annihilated,
0
is the frequency
of the incident signal. As the frequency is related to the normal vibrational modes of the
molecules constituents of the medium, by analyzing the scattered light, information about the
medium can be retrieved. This is the main idea behind Raman spectroscopy, a widely used
technique for materials characterization. In amorphous materials, like Silica, can assume a
value belonging to a broad spectral range, starting from zero and going up to 40 THz.
Experimentally both down shifted and up shifted frequencies waves have been observed and
have been named as Stokes and anti-Stokes, respectively.
Stimulated Raman Scattering was discovered by E. J. Woodbury and W. K. Ng, almost
accidentally in 1962, when working with a Ruby laser [32]. They observed a strong spectral
line not coincident with any spectral line of the fluorescence spectrum of Ruby. To
understand this process let us assume that an incident photon is scattered by an optical
phonon in the medium, and in this process a down shifted photon and an optical phonon are
created. We can see that we have two ways of creating phonons, the scattering process and
the thermal mechanism. If the intensity of the incident light is small, the rate of phonons
created by scattering is low and due to thermal equilibrium the density of phonons in the
medium is unchanged, and therefore the medium maintains the same optical properties. If the
intensity of light is increased above a certain threshold, the optical properties of the medium
can be changed in a way that the scattering process is enhanced [33]. In this situation, the
incident light stimulates the scattering process and we are in the presence of Stimulated
Raman Scattering. Through this positive feedback the scattering process can be enhanced by
several orders of magnitude. Due to the Bosonic nature of the photons, this process can
indeed provide gain. The photon emission process by a scattering center, it can be stimulated
by the presence of another photon, and this stimulated emission is the origin of the gain. The
term emission is used in this context in a quite abusive way because there is no absorption to
a real state, but this process can be treated considering that the scattering photon is initially
absorbed to a virtual state and after re-emitted.
If we consider that the decay from the virtual states only occurs spontaneously, the
Stokes power grows linearly with the pump power. In the other way, if we consider that the
decays from the virtual states must be triggered by another photon, the Stokes power grows
exponentially with the pump power. Off course, in reality both spontaneous and stimulated
emission occurs. If the photon that triggers the stimulated emission is part of a signal we are
in the presence of optical gain, which can be beneficial for optical communication systems
[34]. If this photon was initially generated by spontaneous emission we are in the presence of
amplified spontaneous emission noise which is usually considered as harmful, at least for
telecommunications purposes. The spontaneous emission process always leads to an excess of
noise in the system.
The optical gain provided by the Raman process can be completely characterized by the
Raman-gain coefficient ( )
R
g , which is related with the imaginary part of the third-order
nonlinear susceptibility. The characterization of the amplified spontaneous emission process
requires, besides the Raman-gain coefficient
( )
R
g , another coefficient named noise
spontaneous emission factor
( )
sp
n . However, it turns out that another source of noise must
be also considered to characterize the noise in Raman amplifiers. This source of noise arises
from Rayleigh scattering. Most of the Rayleigh scattered photons are lost through non-guided
modes, but some of them are coupled to the counter-propagating mode. Those photons can be
amplified and through another Rayleigh scattering process can appear as extra-noise at the
amplifier output. This effect is usually named as double Rayleigh scattering and will be
described in more detail in section 4.3.
3. Modeling of Raman Amplifiers
The implementation of RFA, using an optical fiber as gain medium, requires that the pump
and information signals must be injected into the same fiber. A basic scheme for a RFA
architecture is displayed in figure 2. The signal and pump waves are launched into the optical
fiber (the gain medium) by a coupler, so, that stimulated Raman scattering can occur. Since
the SRS effect occurs uniformly for all the orientations between pumps and signals, Raman
amplifiers can work both in forward and/or backward pumping configuration.

Figure 2. General scheme for a distributed Raman amplifier. For simplicity the optical isolators used to
protect the pumps and signals sources, were omitted.
The model for power evolution in Raman amplifiers assuming a multipump multisignal
configuration is often based on an unified treatment of channels, pumps and spectral
components of the amplified spontaneous emission (ASE). The major interactions can be
reasonably drawn by considering the pump-to-pump, signal-to-signal and pump-to-signal
power transfer, attenuation, Rayleigh back scattering, spontaneous Raman emission and its
temperature dependence. Other effects, such as noise generation due to spontaneous anti-
Stokes scattering, polarization and nonlinear index are neglected, but they can reach
considerable importance in certain regimes of transmission. It must be noted that signal
channels and Raman pumps are treated as fields at single frequencies, so ignoring the
interactions due to the spectral shape of signals and pumps.
In a general approach, the power evolution of pumps, signals and ASE (in forward and
backward directions), with time along the fiber distance is given by the following set of
coupled differential equation [35]. For N
p
pumps, N
s
probe signals and N
ASE
spectral
components for ASE, the system is formed by N
p
+N
s
+2N
ASE
equations.

( ) ( )
( ) ( ) [ ] ( ) ( ) [ ] ( ) ( )
( ) ( ) [ ]( )

+ + +
+
+ + + +
=
+ = + =
+ +

ji j j
i
j
ji i
i i i
m
i j
ij ij i
m
i j
j j ij
j
i
j j
i
j
ji i
i
i
i
t z P t z P g h
t z P t z P g h t z P t z P g t z P t z P g
t
t z P
V z
t z P
1 , ,
) , ( , 1 2 , , , ,
, 1 ,
1
1
1 1
1
1
m
m
(1)

V
i
is the frequency dependent group velocity. The signs stand for the forward or
backward propagating waves, being
i
and
i
the coefficients of attenuation and Rayleigh of
the i
th
wave at frequency
i
. h and k
B
are the Planck and Boltzmann constants, respectively,
and T is the fiber absolute temperature. The phonon occupancy factor is given by:

( )
1
1 exp

=
T k
h
B
j i
ij

(2)

The frequencies
i
are numbered by their decreasing value (the lower order corresponds
to the higher frequency). Thus, the terms in the summation in expression 1, from j=1 to j=i-1
cause amplification since the wave i is receiving power from the lower order waves (with
higher frequency). For the same reason, the terms in the summation from j=i+1 to j=m
originate depletion. For mathematical convenience the gain spectrum was divided into slices
of width , spanning the range over which ASE spectral components are significant.
The terms that contain a product of powers describe the coupling via stimulated Raman
Scattering, being its strength determined by the Raman gain coefficient of the fiber, g
ij

obtained by equation 3.

( )
eff
j i R
ij
A
g
g

=

(3)

where A
eff
is the effective area of the fiber and the factor is a dimensionless quantity
comprised between 1 and 2 that takes into account the polarization random effects. The
achieved gain, as well as the slope of the gain spectrum, depends on the transmission fiber
[36, 37]. In figure 3, two Raman gain coefficient spectra are displayed, showing the different
strengths of the Raman coupling of a SMF fiber and a dispersion compensating fiber (DCF).

-30 -20 -10 0 10 20 30
-4
-3
-2
-1
0
1
2
3
4
R
a
m
a
n

g
a
i
n

c
o
e
f
f
i
c
i
e
n
t

(
W
-
1
k
m
-
1
)
pump-signal frequency difference (THz)
SMF
DCF

Figure 3. Raman gain coefficient spectra for two germanosilicate fibers: Single mode fiber (SMF) and
dispersion compensating fiber (DCF), for a pump wavelength of 1450 nm.
As a matter of fact, the small effective area of the DCF (15 m
2
) is determinant for its higher
Raman gain coefficient when compared to the SMF (80 m
2
) or when compared with DSF
fibers (50 m
2
). Those spectra also show peaks that are broader than those presented by
crystalline materials, since the amorphous nature of Silica allows a continuum of molecular
vibrational frequencies.
To obtain a steady-state power distribution, the time derivative in equation 1 is settled
equal to zero, and the set of equation takes the form of expression 4.

[ ] [ ] ( )
[ ]( )

+ + + +
+
+ + + +
=
+
+ = + =
+ +

ji j j
i
j
ji i i i
i
m
i j
ij ij i
m
i j
j j ij
j
i
j j
i
j
ji i
i
P P g h P
P g h P P g P P g
dz
dP
1
1 2
1
1
1 1
1
1
m
(4)

In spite of the simplification, the modeling is still computationally intensive, especially
for the situation of backward or bidirectional pumping. In those situations, the mathematical
problem that describes the power evolution of pumps and signals along the fiber is a
boundary value problem (BVP) which is more difficult to solve than the initial value problem
(IVP) in the forward pumping scheme. An immediate approach to the numerical solution of
such problem is the shooting method [38]. There are other allowable numerical methods, such
as relaxation methods, or collocation methods [39]. Generally, shooting methods are faster
than relaxation ones. In shooting methods, we choose values for all the dependent variables at
one boundary, solve the system of ordinary differential equation (ODE) as an IVP and verify
if the obtained values on the other boundary are consistent with the stipulated values
(boundary conditions) [40]. Then, the parameters are repeatedly changed using some
correction scheme until this goal is attained. The selection of the correction scheme is crucial
for stability and efficiency of the resulting algorithm. An other variant of the shooting
method, we can guess boundary values at both ends of the domain, integrate the equation to a
common midpoint and repeatedly adjust the guessed boundary values so that the solution
tends to the same value at the middle point. This adjustment task is usually performed by the
Newton-Raphson method.
Recently, some shooting algorithms with different correction schemes for the design of
Raman fiber amplifiers have been proposed in order to improve convergence of the solutions
even for larger fiber lengths [41]. This scheme is obtained by modifying the numerical
method used to perform the IVP integration (fourth order Runge-Kutta, Runge-Kutta-
Felhberg, etc). Other approaches to solve the equation 4 propose a shooting method to a
fitting point using a correction scheme based on a modified Newton approach. Therefore, by
introducing the Broydens rank-one method into the modified Newton method, the algorithm
becomes more efficient and stable. This happens because the intensive numerical calculations
of the Jacobi matrix are substituted by simpler algebraic calculations [41].
The use of projection methods such as collocation, gives a continuous approximation of
the solution as a function of the fiber length. The basic idea is to approximate the BVP
solution by a simpler function that represents an approximation.
Nevertheless, the Raman equations (equation 4) are also solvable through semi-analytical
methods, using the average power analysis (APA) presented by Min et al. [42]. The amplifier
is split into n small segments, in order to avoid the position dependency of the powers of
equation 4. The equations are then solved analytically in each segment, considering as input
conditions the outputs provided by the solution on the previous segment. Equations 5 to 8
show how the powers are iteratively computed. The output pump/signal power at each section
end is given by:

) , ( z G P P
in out

=
(5)

being G(z,) the section gain,

[ ] z B A z G + = )) ( ) ( ) ( ( exp ) , ( (6)

The constants, A() and B() are obtained through:

j
m
i j
ji
j
i
j
ij
P g B
P g A
+ =
=
=
=
1
1
1
) (
) (
(7)

The optical power term in each section can be substituted by its length averaged values
given by:

)) ( ln(
1 ) (
G
G
P P
in
=

(8)

For a RFA, the net gain is usually defined as the ratio between the signal powers at the
end and at the beginning of the fiber link, as defined in equation 9:

) 0 (
) (
=
=
=
z P
L z P
G
signals
signals
net
(9)

The so-called on/off gain is another useful quantity that measures the increase in signal
powers at the amplifier output when the pumps are turned on, as follows:

off pumps with ) (
on pumps with ) (
/
L z P
L z P
G
signals
signals
off on
=
=
=
(10)

The numerical issues due to the backward pumping can be surpassed by assuming that
the pump inputs are located at the same fiber end that the signal inputs. Therefore, the pump
equations are integrated reversely as if they were backward by multiplying them by (1). A
guessed initial input is necessary to perform the integration, but the algorithm is able to adjust
it using an optimization routine that adjust the initial input until the output at the fiber end
reaches the real backward pump input. The use of the APA approach has shown a reduction
of two orders of magnitude in the computation time, being the obtained results in agreement
with the ones resulting from traditional numerical methods.
To demonstrate the numerical resolution of the steady-state Raman propagation
equations, we assume the scheme in figure 4, where three bidirectional pumps (two backward
and one forward) and four probe signals are considered. The counterpropagated pumps have
power levels set equal to 0.1W, working at 1450 nm and 1460 nm, respectively. The
copropagated pump is working at 1470 nm with an output power also equal to 0.1W. The
forward pumping signal are then injected into 40 km of SMF fiber and combined with
41000 GHz spaced C band probe signals with an initial optical power equal to 1W. The
spatial evolution of pumps and probe signals are displayed in figure 4.

0 10 20 30 40
0.6
0.7
0.8
0.9
1.0
1.1
1.2
P
r
o
b
e

P
o
w
e
r

(
W
)
Fiber length (km)
1530 nm
1538 nm
1546 nm
1554 nm

0 10 20 30 40
0.00
0.02
0.04
0.06
0.08
0.10
1450 nm
1460 nm
1470 nm
P
u
m
p

P
o
w
e
r

(
W
)
Fiber length (km)

Figure 4. Spatial evolution of two counterpropagated pumps, one copropagate pump and four probe
signal along a 40 km SMF fiber span amplifier. Probe signals evolution (top) and pump signals
evolution (bottom).
The implementation of equation 4 also allows the calculation of the total noise for each
signal (forward and backward ASE) within the amplifier, whose spatial evolution for this
system can be followed in figure 5.

0 10 20 30 40
1
2
3
4
5
6
7
8
9
T
o
t
a
l

N
o
i
s
e

P
o
w
e
r

(
n
W
)
Fiber length (km)
1530 nm
1538 nm
1548 nm
1554 nm

Figure 5. Spatial evolution of the total noise (forward and backward ASE) power along 40 km SMF
fiber span.
The noise figure of an optical amplifier amounts the degradation of the signal to noise
ratio (SNR) when the signals are amplified. The most important source of noise in optical
amplifiers is ASE, which, for Raman amplifiers is due to spontaneous scattering. Assuming
that the signals are initially as noiseless as possible, and that their degradation is due to
signals spontaneous beat noise produced by ASE, the noise figure, in linear units, is given by
equation 11 [36]:

net
ASE
G h
L z P 1
1
) ( 2
NF
+

(11)

where h is the photon energy and P
ASE
+
is the forward ASE measured over the reference
bandwidth . The first term corresponds to the noise from the signal spontaneous beating
and the second one to shot noise.
Another quantity, named effective noise figure, accounts the noise that a discrete
amplifier placed at the end of an unpumped fiber link would need to have the same noise
performance that a distributed Raman amplifier. In decibel units, the effective noise figure is
computed using:

( )
dB dB dB
eff
L NF NF =
(12)

Typically, WDM systems allocate a large number of channels spaced over wide
bandwidths. Considering the previous system but doubling the pump powers and using 64
probe signals (instead of 4), we obtained the spectra of the gain and noise figure which are
plotted in figure 6.

1530 1540 1550 1560 1570 1580
0
2
4
6
8
10
Signal Wavelengths (nm)
N
e
t

G
a
i
n

(
d
B
)
-2.0
-1.5
-1.0
-0.5
0.0
E
f
f
e
c
t
i
v
e

N
o
i
s
e

F
i
g
u
r
e

(
d
B
)

Figure 6. Net gain and effective noise figure spectra for a system with two counterpropagated pumps,
one copropagate pump and 64100 GHz probe signals along 40 km SMF fiber span.
As depicted in this section and despite some remaining numerical issues, the modeling of
a multipump Raman amplifier anticipates many valuable applications for WDM systems,
namely the broadband gain. It is important to notice that gain spectra as wide as 100 nm are
achievable and that the gain value can be kept quite constant by an appropriate tailoring of the
amplifier architecture. This procedure involves solely the proper dimensioning of the pump
power levels and operating wavelengths, as discussed more extensively in section 4.
Another interesting feature of RFA is the noise performance. The ASE noise in RFA is
intrinsically low (as suggested by the negative effective noise figure presented above). The
reason relies in the fast relaxation of the optical phonons, the absorption of signal photon to
the upper virtual state is extremely small. The inversion of population is almost complete.
4. Challenges in Raman Amplification
4.1. Gain Profile Optimization
One of the most impressive features of Raman fiber amplifiers is assuredly the possibility to
achieve gain at any wavelength, by selecting the appropriate pump wavelength. Therefore, it
is possible to operate in spectral regions outside the Erbium doped fiber amplifiers bands over
a wide bandwidth (encompassing the S, C and the L spectral transmission bands).
Nevertheless, some studies have been reporting that despite the Raman gain dependence is
essentially due to the pump-signal frequency difference; there is also some weaker
dependence on the pump absolute frequency [43]. However, since a deeper study of this topic
is beyond the scope of this work, we will not consider it in the gain optimization.

1400 1420 1440 1460 1480 1500 1520 1540 1560 1580 1600
0.20
0.22
0.24
0.26
0.28
0.30
0.32
0.34
Net gain (around 5 dB)
Gain contribution from each individual pump
pumps powers and wavelengths
P
u
m
p

p
o
w
e
r

(
W
)
Wavelength (nm)
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
N
o
r
m
a
l
i
z
e
d

g
a
i
n

c
o
e
f
f
c
i
e
n
t

Figure 7. Numerical simulation of broadband Raman amplifier gain. Bars show backward input pump
powers and wavelengths. Ticker line show 14400 GHz probe signals optimized net gain and thin lines
the gain contribution of each individual pump. The simulation was carried out through 25 km of SMF
fiber.
A flat spectral gain profile is achievable with the combination of several pumping lasers
operating at specific powers and wavelengths. The Raman gain created by pumps at different
frequencies is slightly shifted from each other to partly overlap and form a composite gain.
When the pump powers and frequencies are properly chosen, this wide gain can also be
considerably flat. Another important feature to take into account when designing a flat gain
scheme, is the strong Raman interaction between the pumps, since the higher frequency pump
is responsible for the amplification of the lower frequency signals, more pumping power is
needed, as some will also be transferred to the lower frequency pumps. This interaction
between pumps also affects the noise properties of the amplifier. However, some novel
pumping schemes have been recently proposed in order to prevent those unwanted effects:
copumping, time dependent Raman pumping, higher order pumping and broad-band pumping
[44].
Typically, laser diodes with output powers in the 100-200 mW range can be used in a
multipump scheme. This scheme is normally composed of a set of laser diodes operating in
the 14XX nm region, whose spectral width is narrowed and stabilized by a FBG. Optical
couplers are used to combine and depolarized them, in order to suppress the polarization
dependent gain. The multipumping allows bandwidth upgradeability by the addition of new
laser diodes. Theoretically, the larger the number of pumps the better the gain ripples.
Nevertheless, there are economic issues that prohibit the use of an arbitrary number of pumps.
For this reason, we have to find a balance between the system performance and the cost of
amplification.
Optimization of the gain spectrum has been widely performed making use of several
global search methods, such as neural networks [45], simulated annealing [46] and genetic
algorithm (GA) [47]. During the search process, the pump powers and frequencies are
directly substituted into the system of propagation equations to calculate de gain profile.
Depending on the speed of the numerical method used to integrate the system of equations,
the amount of numerical computations involved can be considerably large and the
optimization inevitably time consuming. Those solutions are not suitable for practical
applications where the real optimal solution must be provided in a short time.
However, some alternatives can be found by replacing the usual intensive numerical
integrations with simpler algebraic calculations using the APA method while integrating the
Raman propagation equations.
Another simple but important issue when using a global optimization method relies in a
proper dimensioning of the search domain. Using the APA method, all the inputs are located
at the same fiber end, even for the counter pump situations. Therefore, the pump power inputs
are chosen by presuming a typical propagation profile. By this way, it is advisable to try
lower power values for the higher frequency pumps and higher power values for the lower
frequency pumps (the opposite happens at other fiber end). Regarding to the optimization of
the pump frequencies, it is advisable to divide our spectral range into the number of pumps
and then shift those values by 13 THz.
A second approach to speed up the search of the optimal pump configuration uses the
genetic algorithm (GA) method only to search the pump frequencies and a quadratic
programming method to solve the power integral [48]. The search domain of the GA method
is by this way reduced to a half, enabling faster convergence.
Another approach combines GA with the Nelder-Mead search. This so called hybrid GA
can be useful in certain situations for the purpose of saving some function evaluations and
consequently to perform the optimization in the least time possible [49]. The hybrid GA
follows the routine depicted in Figure 8. Firstly, the initial population, as well as the other GA
operators are dimensioned: selection, crossover and mutation. The selection, together with the
crossover, is responsible for the bulk of GA processing power. The mutation is an operator
that plays a secondary role in the GA. Since, the genetic operators can be performed by
different methodologies, it is important to choose the ones that are more adequate to the
problem we are dealing with, in order to improve the GA search procedure [50]. It must be
noted that if the search space is not large, it can be searched exhaustively and the best
possible solution will be probably found. The maximum number of allowed generations is
also an important feature because, when carefully chosen, it can save a large number of
function evaluations. The Nelder-Mead method uses a simplex in a n-dimensional space,
characterized by the n+1 distinct vectors that are its vertices. At each step of the search, a
new point in or near the current simplex is generated. The function value at the new point is
compared with the function values at the vertices of the simplex and one of the vertices is
replaced by the new point, giving a new simplex. This step is repeated until the diameter of
the simplex is less than the specified tolerance

Figure 8. Scheme of the hybrid GA implementation.
0 5 10 15 20 25 30 35 40 45
0.5
0.6
0.7
0.8
0.9
1.0
B
A
Number of Generations

Figure 9. Total simulation time for a hybrid GA against the number of generations for a population size
of 50 individuals. The line is a visual guide.
By determining properly the right moment to switch from one method to another, it is
possible to reduce the simulation time to a half when compared to simple GA. This result can
be observed in figure 9. The normalized total simulation time () against the number of
generations is plotted. Here, the reference is the slowest simulation, the one that use 40
generations, identified by the letter B. The best situation (tagged as A) tooks a simulation time
equal to about a half of the time of the worse situation and it was attained with 17
generations. An heuristic explanation relies on the intrinsic nature of the GA. We verify that
for small number of generations (bellow 15) the GA time is small but the system reaches a
worse fittest solution. Thus, the Nelder-Mead method needs more time to reach a desirable
solution. When the number of generations increases the GA reaches a best solution but the
needed computation time increase accordingly.

0 5 10 15 20
0
20
40
60
80
100
120
140
160
1410.0 nm
1424.2 nm
1437.9 nm
1452.1 nm
1465.5 nm
1494.8 nm
1502.4 nm
P
u
m
p

P
o
w
e
r

(
m
W
)
Fiber Length (km)

Figure 10. Power evolution of optimized pumps along 20 km of SMF (lines). The geometric shapes
stand for the used experimental values.
1520 1530 1540 1550 1560 1570 1580 1590 1600 1610
7.5
7.6
7.7
7.8
7.9
8.0
O
n
/
O
f
f

G
a
i
n

(
d
B
)
Channel Wavelength (nm)

Figure 11. Experimental (arrows) and simulation (line) on/off spectral gain for the 20 probe signals and
7 counter propagated pumps, over 20 km of SMF fiber.
In order to enlighten the conclusions provided by the hybrid GA algorithm, a laboratorial
implementation was carried out to test the optimization results. A Raman amplified system
with 20 km of SMF fiber, 20 probe signals and 7 backward pumps was implemented. Since
the pump wavelengths are already settled, only the optimization of the power levels is needed.
The simulation used the stochastic uniform method for selection, the scattered crossover
method and the uniform mutation. A population of 50 individuals and a number of
generations equal to 35 were considered. The spatial evolution of the pumps signals
optimized values are displayed in figure 10 jointly with the pump signal experimental values.
In figure 11, the optimized and experimental on/off gain spectra are presented.
This is a good agreement between the optimization modeling and the experiment. The
maximum ripple attained by the optimization is 0.41 dB being the experimental maximum
ripple equal to 0.23 dB. The mean square deviation between simulation and experimental
results is equal to 0.0036. Indeed, a flat gain over a wide bandwidth (~80 nm) was attained,
using seven pumps with a total input power equal to 453 mW.
4.2. Raman Amplification Using Multiple Low Power Lasers
One of the main issues in Raman amplification is related to the stability of the high power
lasers, the costs and the need for efficient cooling. To go around these problems, the usual
solution is the use of several pump signals, what results in added advantages, like high, flat
and wide-gain bandwidth [51-53].
The technology evolution allowed that high power pumps are nowadays commercially
available, although some problems still limited [54]. The pressure on optical components
prices, lead to the creation of CWDM standards [55]. This is reflected specially on price
dropping of uncooled lasers with relatively high powers (>10mW). The price to pay is
wavelength wondering, however, neither for CWDM nor for Raman, wavelength stability is
not a stringent requirement, allowing simple control. With this technology the possibility of
achieving Raman gain by combining multiple of these low power lasers was successfully
implemented [21].
Teixeira et el proposed the use of an array of low cost lasers to achieve wideband Raman
amplification, providing both experimental and simulation results [21]. In this work a
counterpropagating topology was implemented, using 40 C band lasers with 20 mW output
power spaced by 0.8 nm (1533 nm 1557 nm). These lasers were combined using a
multiplexer, bringing up a total power of more than 200mW (23 dBm). This power is enough
to generate SRS. Several fibers were tested: True Wave and dispersion compensating. To
characterize the gain profile, an array of 40 L-band 0.8nm spaced probe lasers (1565 nm -
1605 nm) with a total optical power of 1 mW was used. Figure 12 (a) shows the implemented
setup. Figure 12 (b) presents the simulation results for the implemented system to four
different pumping configurations. The first curve corresponds to the traditional approach,
where one high power pump (23.6 dBm) at a single wavelength (located at 1530 nm) is used.
In the second case, three lasers spaced by 0.8 nm starting at 1530 nm having total power of
23.6 dBm were multiplexed. The results for the two above pumping configurations are
approximately equal, having only a wavelength shift of 0.8 nm as expected due to the average
pumps wavelength difference. Similar simulation was experienced considering 40 lasers, each
with 7.6 dBm (after the multiplexer), resulting in a total power of 23.6 dBm. In this case, the
gain curve appears even smothered and the peak shifted by ~16 nm; the peak gain is similar,
however a small enhancement on the 3 dB gain bandwidth was obtained and the gain profile
smothered. In order to explore the advantages of the methodology (gain flatness), the power
distribution for the pumps was optimized to reach an equalized gain, while maintaining the
same total pump comb. The peak gain was decreased at the expense of an increased flattened
profile.

a)
1600 1620 1640 1660 1680
-2
-1
0
1
2
3
4
5
G
a
i
n

(
d
B
)
Wavelength (nm)
1 laser
3 laser
40 laser
40 laser eq.

b)
Figure 12. a) Implement setup for the simulation and experimental systems, b) simulated Raman gain
profiles for several sets of pumping configurations, with 23.6 dBm of total power. PM demotes an
optical power meter, OSA denotes an optical spectrum analyzer and MUX is an optical multiplexer.
Due to limitations on available probe and pump signals, the experimental implemented
system only can scope part of the spectral bands used in simulation. The gain is only
measurable when the pump powers go above 10 dBm. A maximum of 3 dB net gain was
achieved in the L band for full pump power, 23.6 dBm, as displayed in figure 13 b). Also, in
the same figure, a minimum of 2dB gain over more than 30nm, with 1dB ripple, was achieved
without any power distribution optimization. The results have demonstrated the effectiveness
of the technique to achieve Raman amplification.

1530 1535 1540 1545 1550 1555 1560
0
5
10
15
20
25
1575 1580 1585 1590 1595 1600 1605 1610
0
1
2
3
4

R
e
l
a
t
i
v
e

P
u
m
p

P
o
w
e
r

(
d
B
m
)
Wavelength (nm)

G
a
i
n

(
d
B
)
Wavelength (nm)
0dBm
10dBm
23.6dBm

a) b)
Figure 13. a) Pump to pump Raman effect; b) experimental Raman gain achieved for several values of
the pump power, with probes at 0 dBm.
In figure 13 (a) it is illustrated the pump to pump effect which is commonly occurring in
dense WDM systems. This effect starts to be noticeable above 10 dBm of total power and is
evident for 7.6 dBm per channel. This phenomenon can be harmful due to uneven distribution
of power during transmission, however, if correctly considered can be used to obtain
beneficial extra gain in the system, if a pre equalization is also implemented .
4.3. Raman Amplification Using Rayleigh Backscattering
Raman amplification pumping can also be achieved by recurring to the traditional methods of
shifted gain [19, 20]. In these methods, several FBG reflector pairs are used to generate
resonant cavities in the maximum of the Raman gain spectrum. Thus, with a Ytterbium laser
operating in the vicinity of 1090 nm, where it exhibits its maximum efficiency, it is possible
to generate pumps in the E band, as demonstrated by Papernyi et al, where a set of 6 FBG
reflector pairs were used to generate pumping in the E-band [22]. The latter amplifies the C
band, where the probe signals transmission usually occurs.
The main penalties of traditional Raman amplification are associated with intrinsic
nonlinear phenomena such as nonlinear refraction and Rayleigh backscattering, since it is
required to use high powers and long fiber spans. This last effect occurs when a fraction of
scattered light is backreflected towards the launch end of the optical waveguide. This
reflection is called single Rayleigh backscattering (SRB). Part of this scattered light is also
backreflected in the forward direction and it is called double Rayleigh backscattering (DRB),
as shown in figure 14 [56]. SRB and DRB can be controlled by actuating properly on the fiber
drawing process or by a correct power design [57]. The Rayleigh backscattering has been
studied, modeled and characterized by many authors [56-59]. It is known that the process
results from multiple reflections of light inside the fiber and therefore spontaneous and
unstable lasing can occur [60]. However, this phenomenon has been observed as an
impairment to signal transmission [61, 62].


Figure 14. Simple Rayleigh backscattering (SRB) and double Rayleigh backscattering (DRB)
representations over an infinitesimal length of fiber.
Recently, a method that, up to some extent, allows the control of this phenomenon was
reported [56,63]. With the possibility of controlling the SRB and DBR effect, novel
applications can be drafted. One suggestion is the use of this effect to generate distributed
resonant cavities, which will degenerate in lasing if enough gain is achieved. These are
achieved with the help of only one end FBG set [63]. This is advantageous when compared to
the previously described methods to obtain cascaded Raman amplification, since it needs only
one FBG set, minimizing the need for identical FBG to be used and tuned at different sites
which can be not colocated.
In order to demonstrate the application of this technique to control SRB and DRB, the
experimental system reported in figure 15 was implemented. A Raman pump in the E-band, at
1428 nm, was coupled to the transmission fiber, with controllable power up to 1.5 W. A
circulator was used to protect the laser from back reflections and, simultaneously, to allow the
measurements of the back reflected power spectrum. Two different scenarios were observed:
the FBGs are absent between the fiber and the coupler; and the setup was complete as
described in figure 15. These two scenarios target to show the controlling effect achieved by
the FBGs.
A set of three FBG with wavelengths centered at: 1520 nm, 1531.6 nm, 1535.6 nm all
having 95% reflectivity, were placed after the pump and act as reflective elements.
In a first setup, a 14 km DCF fiber with dispersion parameter equal to -1393 ps/nm and
Raman coefficient of 3.05 x 10
-3
m
-1
W
-1
was used as transmission medium.

Figure 15. Experimental setup for the double shifted Raman experiments; WDM denotes a band coupler
and Att denotes an optical attenuator.
Considering the first scenario, where no FBGs were present, the common Raman effect
in the C band was observed, figure 16 a) for a pump power of 350 mW. When the power of
the pump was increased to 600 mW, Rayleigh backscattering spontaneous lasing effect is
observed, as displayed in figure 16 b). This effect presents random behavior, both in
wavelength and power, being the spectrum time dependent.
In a second scenario the FBGs were present, control of the random process generated by
the SRB and DRB was achieved and the lasing was stabilized in the FBGs wavelengths. In
this situation a virtual cavity was established, formed by the FBG and the Rayleigh
backscattered light. To generate more than one laser in the C-band a set of cascaded
wavelength mismatched FBGs were used. These gratings are responsible for a multipeak
frequency dependent reflection back into the fiber of the amplified spontaneous emission and
DRB light from the fiber. This, in conjunction with the FBG, create resonant cavities, which
generate stable wavelength constant lasing actions, from now on called as FBG-DRB lasing.
Due to the different reflectivities of the FBGs and the Raman gain profile, different lasing
powers for each configuration occur in the C-band. Whenever the power of the generated
lasers in the C-band is high, cascaded Raman effects will occur that generate gain in the far L
and U-band. The FBG-DRB lasing and consequent stabilization process with the
simultaneous L-U band spontaneous emission is reported in figure 16 c), where a pump
power of 1.2 W was used [63].

1400 1450 1500 1550 1600 1650 1400 1450 1500 1550 1600 1650
-90
-80
-70
-60
-50
-40
-30
-20
-10
1400 1450 1500 1550 1600 1650

U-band ASE
C-band
FBG-DRS lasing

(c) (a)
C-band ASE
P
o
w
e
r

(
d
B
m
)
Pump
(b)
C-band
Spontaneous
Lasing
Pump
Pump
Wavelength (nm)

Figure 16. Transmission spectra for 14 km DCF fiber: a) Spontaneous ASE for a pump of 300mW; b)
spontaneous lasing for a pump of 600mW; c) C-Band FBG-DRB lasing and far L and U-band Raman
generated ASE for a 1.2 W pump.
The results show a 38 nm flattened ASE bandwidth in the U-band, generated by the FBG-
DRB. By introducing a copropagating probe at 1625 nm, a gain of 10 dB was measured for an
E-band pump power of 1 W.
In a second setup, different optical fibers were tested in order to compare the pump power
laser threshold. A 14 km long DCF fiber, a 50 km long DSF fiber and a 50 km long non zero
dispersion shift fiber (NZDSF) were used [60]. Figure 17 shows the different lasing
thresholds and curve shapes resulting from the intrinsic differences between the optical fibers.
From figure 17, it can be observed that this process is more efficient in the DCF fibers, where
the threshold power is 350 mW, while for the NZDSF fiber is 650mW and 1W for the DSF
fiber.

0,2 0,4 0,6 0,8 1,0 1,2
-50
-40
-30
-20
-10
0
10
0,2 0,4 0,6 0,8 1,0 1,2 0,2 0,4 0,6 0,8 1,0 1,2

P
e
a
k

P
o
w
e
r

(
d
B
m
)
1520.0nm
1531.6nm
1535.4nm
1428.0nm

14 km DCF Fiber
Selected Input Power (W)
50 km NZD Fiber 50 km DSF Fiber

Figure 17. Depletion of the E-band pump and peak power of the C band lasers as a function of the
pump power for several fiber types; from left to right: DSF, NZD and DCF.
Usually, the simulation of Raman amplification as convergence and stability problems,
especially for high pump powers, has reported in previous sections. The simulation of the
lasing effect with high pump power has similar difficulties. To avoid such problems the
solving method for the differential equation system is simplified to an analytical method
based on the transfer matrix (APA) as proposed in section 4.1. Inside these fiber slices, the
parameters are considered to have small variations and the solution of the equation system is
obtained by stabilization after multiple passes along the length of the fiber [64]. The initial
solution uses an analytical approach that was based in the undepleted case. The approach to
the pump depletion is included in the attenuation of the pump.
In each fiber slice, the Rayleigh backscattering is calculated at the boundary and this
backscattering power is added to the signals in the same direction and wavelength, that also
suffer amplification and depletion.

0.2 0.4 0.6 0.8 1.0
-50
-40
-30
-20
-10
0
10

Simulated
Experimental
O
u
t
p
u
t

P
e
a
k

P
o
w
e
r

(
d
B
m
)
Input Power (W)

Figure 18. (a) Optical power density spectra for 14 km DCF fiber from E-band to U-band; (b) Output
power evolution of the lasing effect of the FBG-DRB at 1520 nm.
Figure 18 (a) presents the simulation results of such algorithm for 29.03 dBm of pump
power. The evolution of the power densities from E-band to U-band spectra is shown for a
long fiber span. The E-band pump signal suffers depletion in the long propagation fiber. This
pump works as a seed of the C-Band FBG-DRB lasing, which generate the L-U-Band Raman
gain.
Since the process of lasing is not stable, the simulation process presents a slow
stabilization, but, the boundary powers over the FBG are quickly stabilized. Figure 18 (b)
presents a comparison of the threshold laser power obtained by experimental and semi-
analytical methods. The output peak power of the FBG-DRB signal at 1520 nm is related
with the input pump power.
As observed for the DCF fiber, stable multiple laser actions were achieved for moderate
pump powers (350 mW) for both simulation and experiment.
4.4. Amplification with Incoherent Pumps
A technique to increase the bandwidth and decrease the spectral ripple of RFA is available
with incoherent pump lasers. A Raman amplifier with incoherent pumps can be modeled as a
multipump Raman amplifier. In such case, the spectrum of the incoherent pump is well
approximated by a large number of pumps of infinitesimal spectral width and whose power
sum equals the integral power of the incoherent pump. Therefore, the theoretical model used
for incoherent pump schemes is based on the model, previously presented, for coherent
multipump configurations.

1470 1480 1490 1500 1510 1520
0.0
0.2
0.4
0.6
0.8
1.0
Wavelenght (nm)

O
p
t
i
c
a
l

p
o
w
e
r

(
a
u
)

Figure 19. Pump spectrum for the incoherent pump.
An incoherent pump spectrum, as displayed in figure 19, with 10 nm FWHM, can be
approximated by 100 pumps of infinitesimal spectral width, having an aggregate power equal
to the integral power of the incoherent pump. The incoherent pump here considered was
obtained from a high power FBG (Fiber Bragg Grating) laser, from which the stabilization
grating was removed [65].
To evaluate the advantages of this technique, the Raman on/off gain and the noise figure
were measured for coherent and incoherent pumping over 40 km of SMF fiber. The probe
signal combo consists of 13 channels, with 1 mW power, spaced by 100 GHz over the 1546-
1556 nm spectral region. Both co-propagating and counter-propagating architectures were
considered. The coherent pumping source was a high power FBG laser with a wavelength of
1490 nm. In both cases the pump power was 290 mW.
The results of Raman on/off gain and effective noise figure are shown in figure 20. The
relatively low on/off gain is due to the fact that the pump wavelengths have not been
optimized for this signal band.

1546 1548 1550 1552 1554 1556
0
2
4
6
8
10
12
14
incoherent co puming incoherent counter pumping
coherent co pumping coherente counter punping
E
f
f
e
c
t
i
v
e

N
o
i
s
e

f
i
g
u
r
e

(
d
B
)
Wavelenght (nm)
1546 1548 1550 1552 1554 1556
1.0
1.2
1.4
1.6
1.8
2.0
2.2
2.4
2.6
2.8
3.0

Incoherent co pumping Coherent co pumping
Incoherent counter pumping Coherent counter pumping
G
a
i
n

(
d
B
)
Wavelenght (nm)

Figure 20. Raman gain and Effective Noise Figure. Lines are simulated results and points represents to
experimental data.
The incoherent pumping gain slopes are 0.0150.008 dB/nm and 0.0170.004 dB/nm for
co and counter propagation configurations, respectively. For coherent pumping, the gain
slopes are 0.0420.01 (co-propagation) and 0.0520.005 dB/nm (counter-propagation). Such
results show that the incoherent pumping configuration presents a flatter gain.
The noise figure is approximately the same for coherent and incoherent pumping in the
counter-propagating configuration. However, in the co-propagating case, the noise figure is
considerably lower for coherent pumping.
In agreement with previous works [23-26], these results indicate that the incoherent
pumping technique can be used to decrease the spectral ripple of the Raman gain.
4.5. Raman in CWDM Systems
Another important challenge is the deployment of RFA for access networks, namely for
CWDM networks. Since CWDM systems require large bandwidths to guarantee the
transmission of a reasonable number of channels, spaced by 20 nm, wide band Raman
amplifiers are well suited for this purpose.
The Raman amplifier bandwidth can be enlarged by using multiple pumps. Optimization
of the number of pumps and their wavelengths enables the large needed gain spectra and that
could be placed in any range of wavelengths used in optical communications.
The design of an amplifier that fits more than two CWDM channels can be achieved,
with the following procedure. The number of channels to be transmitted is determined in
order to define the required bandwidth. The optical fiber characteristics impose a minimum to
the required gain, and finally the number of pumps as well as their characteristics are decided.
The scheme of figure 21 illustrates the important issues to be considered to design a multi-
pumped Raman amplifier for a CWDM system.
Since the number of CWDM channels and the length of the link as well as its losses are
defined, the minimum required gain to compensate the transmission losses and the minimum
bandwidth to transmit all the required channels may be determined using the rectangle shown
in figure 21. The gain has to be high enough to compensate the losses caused by the optical
fiber and the bandwidth should be large enough to support all the transmitted channels. The
purpose is to obtain a spectrum that encloses this rectangle.

Bandwidth
Ripple
Gain
Bandwidth
Ripple
Gain

Figure 21. Design concerns for a multi-pumped Raman amplifier for CWDM systems.
Another point is to guarantee that the maximum deviation between the values of the
designed and the needed gain as smallest as possible. The curved line in figure 21 represents
the obtained spectrum after optimization of pumps characteristics. The ripple represents the
maximum deviation cited above.
Another concern in designing the spectrum is to make it flat, with all the channels at the
same level, in order to avoid reception constrains.
The multipumped Raman amplifier can be designed using a set of coupled nonlinear
equations as equation 4. Solving the coupled equations for one signal and one pump, may be
simplified when pump depletion is ignored. This approximation is valid because the pump
power is higher than the signal power, P
p
P
s
[66]. However, whenever multiple pumps are
used this simplification cannot be used due to the interaction between pumps which enhances
the effect of depletion due to the higher powers involved.
1350 1400 1450 1500 1550 1600 1650 1700 1750
0
5
10
15
20

G
a
i
n

(
d
B
)
Wavelength (nm)

Figure 22. Example of a multi-pumped Raman amplifier applied to CWDM systems.
Figure 22 shows an example of an optimized Raman amplifier spectrum applied to
CWDM systems. It was designed to transmit five probe channels at 1490 nm, 1510 nm, 1530
nm, 1550 nm and 1570 nm. The transmission link is based on 80 km SMF fiber, with 0.23
dB/km losses, which implies a 18.4 dB gain with a minimum spectral bandwidth of 80 nm.
The graph in figure 22 is the result of a forward pumping configuration. The number of
pumps used in this example was six. The continuous line represents the gain spectrum
obtained with the six pumps the arrows represent the transmitted probe channels.
The bandwidth is 100 nm, 20 nm larger than the minimum required bandwidth, in order
to guarantee that all the signals are amplified. The gain is around 18.4 dB with a maximum
deviation between the designed and needed gain equal to 1 dB, and a maximum gain
deviation for each channel being 0.9 dB.
The six pumps used are centered at 1380 nm, 1393 nm, 1405 nm 1428 nm, 1444 nm, and
1468 nm with powers of 450 mW, 200 mW, 330 mW, 160 mW, 45 mW, and 55 mW,
respectively. The pump of lower wavelength needs the highest power due to the interactions
between pumps: The lower wavelength pump loses energy to the higher wavelengths, causing
its depletion.
This optimization scheme was verified experimentally, with 3 probe channels CDWM
system, pumped with 3 pump signals at 1470 nm, 1490 nm and 1510 nm. For the
optimization the hybrid GA algorithm, previously presented, was used. This pump allocation
problem is less exigent, in terms of ripple, than for a DWDM system, since the probe signal
are far apart.
The implemented scenario consists of a 40 km SMF fiber, with a counterpropagating
pump scheme and a 7 dB gain target. The optimized pump powers were 128.1 mW, 65.0 mW
and 146.9 mW, respectively. The maximum gain excursion was 0.002 dB and 0.12 dB for the
simulation and experimental systems, respectively.
Experimentally, we can observe that the Raman amplification improves the eye opening
penalty of a signal transmitted along a fiber link allowing a good reception at the end of the
transmission path. To illustrate this behavior, the eye diagram of a signal after a 40 km link is
shown in figure 23.

Figure 23. Comparison between eye diagrams with and without Raman amplification.
Figure 23 illustrates a real case where there is a signal centered at 1567 nm and two
pumps centered at 1508.8 nm, one in forward configuration and another in backward
configuration. The powers of the pumps are chosen to be 100 mW each.
The eye openings are given in Volt and the gain was obtained using the on/off definition
(equation 10). Using the relation between power and voltage, P=V
2
/R, the on/off gain
becomes G
Voltage
= 10log
10
(V
with pump
/V
without pump
).
It is notorious that an increase of the eye opening obtained when both pumps are turned
on. The scale is the same to all the graphs in figure 23 to allow comparisons of the eye
opening amplitude. The eye opening to the bidirectional configuration is higher due to the
higher pump power, while the forward and backward systems have 100 mW, the bidirectional
system uses 200 mW. The respective gains of the eye openings are 1.94 dB, 2.35 dB, and
2.98 dB to the forward, backward and bidirectional systems, respectively. The results show a
higher gain for the counter propagating situation.
5. Conclusion
Raman fiber amplifiers are a technological key component that fulfill the challenging strict
requirements of the beginning of this century, enabling applications not feasible with
conventional EDFAs.
In this contribution, we have discussed the origin of Raman scattering and the critical
properties for system design, such as pumping allocation, cascade pump and broadband
amplification for multiple CDWM networks. It was also presented solutions that provide that
gain, such as the use of low power pumps or incoherent pumps.
These issues are, in the authors point of view, the relevant questions and challenges
associated with Raman amplification on communication systems.
Acknowledgments
This work was supported by the POSC program, financed by the European Union FEDER
fund and by the Portuguese scientific program. The authors also greatly acknowledge the
ARPA (POSI/EEA-CPS/55781/ 2004) and TECLAR (POCI/A072/2005) projects and to FCT
and ALBAN scholarship program.
References
[1] Hu, J.; Marks, B. S.; Menyuk, C. R. J Lightwave Technol. 2004, 22, 1519-1522.
[2] Giltrelli, M.; Santagiustina, M. IEEE Photon Technol Lett. 2004, 16, 2454-2456.
[3] Cui, S.; Liu, J.; Ma, X. IEEE Photon Technol Lett. 2004, 16, 2451-2453.
[4] Kidorf, H.; Rottwitt, K.; Nissov, M.; Ma, M.; Rabarijaona, R. IEEE Photon Technol
Lett. 1999, 11, 530-532.
[5] Bromage, J. J Lightwave Technol. 2004, 22, 79-93
[6] Namiki, S.; Seo, K.; Tsukiji, N.; Shikii, S. Proceedings IEEE 2006, 94, 1024-1035.
[7] Shuto, Y.; Yanagi, S.; Asakawa, S.; Kobayashi, M.; Nagase, R. IEEE Journal Quantum
Electronics 2004, 40, 1113-1121.
[8] Seo, K.; Nishimura, N.; Shiino, M.; Yuguchi, R.; Sasaki, H. Furukawa Review 2003,
24, 17-22.
[9] Bromage, J. J Lightwave Technol. 2004, 22, 79-93.
[10] Chen D. Z., Wellbrock G., Peterson D. L., Park S. Y., Thoen E., Burton C., Zyskind J.,
Penticost S. J., Mamyshev P., ECOC 2006, Cannes.
[11] Monroy I. T., Kjaer R., Palsdottir B., Koonen A: M. J., Jeppesen P., ECOC 2006,
Cannes.
[12] Scheiders, M.; Vorbeck, S.; Leppla, R.; Lach, E.; Schimidt, M.; Papernyi, S.; Sanapi, K,
Optical fiber conference 2005, Anaheim, Post-deadline paper.
[13] ITU -T G.694.1, 2002.
[14] ITU -T G.694.2, 2002.
[15] Eichenbaum, B. R.; Das S. K. National Fiber Optic Engineers Conference 2001,
Baltimore, USA, 1444-1448.
[16] Wang, D.; He, C.; Li, Y. FibreSystems Europe/Lightwave Europe. 2005, 13-15
[17] Desurvire E. Erbium-Doped Fiber Amplifiers - Principles and Applications; John Wiley
& Sons, New York, NY: 1994.
[18] Emori, Y.; Tanaka, K.; Namiki, S. Electron Lett. 1999, 35, 1355-1356.
[19] D. I. Chang, H. K. Lee and K. H. Kim, Electron. Lett. 1999, 35, 1951-1952.
[20] S.B.Papernyi,V.I.Karpov,andW.R.L.Clements, Optical Fiber Communication (OFC)
2002, Anaheim, USA, PostdeadlinePaperFB4-1.
[21] Teixeira, A.; Andre, P.; Stevan, S.; Silveira, T.; Tzanakaki, A.; Tomkos, I. International
Conference on Internet and Web Applications and Services/Advanced International
Conference AICT-ICIW 2006, on 19-25 Feb. 2006 Page(s):85 - 85 ,Digital Object
Identifier 10.1109/AICT-ICIW.2006.158
[22] Papernyi, S.B., Ivanov, V.B. ,Koyano, Y., Yamamoto, H., Sixth order cascaded Raman
Amplification .- Ivanov, Optical Fiber Communication Conference, 2005. Technical
Digest. OFC/NFOEC
[23] Han, B.; Zhang, X. P.; Zhang, G. D.; Lu, Z. G.; Yang, G. X. Opt Express. 2005, 13,
6023-6032.
[24] Zhang, T.; Zhang, X.; Zhang, G. IEEE Photon Technol Lett. 2005, 17, 1175-1177.
[25] Wen, S. F. Opt Express. 2006, 14, 3752-3762.
[26] Wen, S. F. Opt Express. 2006, 15, 45-55.
[27] Vakhshoori, D.; Azimi, M.; Chen, P.; Han, B.; Jiang, M.; Knopp, K. J.; Lu, C. C.; Shen,
Y.; Vander Rhodes, G.; Vote, S.; Wang, P. D.; Zhu, X. OFC. 2003, 3, PD47-P1-3.
[28] X. Zhou, M. Birk and S. Woodward, IEEE Photon. Technol. Lett 2002, vol 14, 1686-
1688.
[29] C. V. Raman, K. S. Krishnan, "A New Type of Secondary Radiation", Nature, Vol. 121,
pp. 501-502, March, 1928.
[30] G. S. Landsberg, L. I. Mandelstam, "Eine neue Erscheinung bei der Lichtzerstreuung in
Krystallen", Naturwissenschaften, Vol. 16, pp. 557-558, July, 1928.
[31] P. S. Andr, A. N. Pinto, Chromatic Dispersion Fluctuations in Optical Fibers Due to
Temperature and Its Effects in High-Speed Optical Communication Systems, Optics
Communications, Vol. 246, Issues 4-6, 15 February 2005, pp. 303-311, 2005;
[32] E. J. Woodbury, W. K. Ng, "Ruby Laser Operation in the Near IR", Procedings of IRE
(correspondence), Vol. 50, No. 11, pp. 2367, November, 1962.
[33] R. W. Boyd, Nonlinear Optics, second edition, San Diego, Academic Press, 2003;
[34] C. Headley, G. P. Agrawal, Raman Amplification in Fiber Optical Communication
Systems, San Diego, Academic Press, 2005.
[35] Karsek M., Kanka J., Honztko P., Peterka P., Int. J. Numerical Modelling: Electronic
Network, Devices and Fields , 2004, vol. 17, n2, 165-176.
[36] Bromage J., J. Lightwave Technol, 2004, vol. 22, n1, 79-93.
[37] Andr, P., Teixeira, A., Kalinowsky, H., Pinto, J. L. Optica Aplicatta 2003, 33
559 573.
[38] Liu X., Lee B., Opt. Express, 2003, vol. 11, n.12, 1452-1461.
[39] Neto B.; Stevan S; Teixeira A. T; Andr P. S.; ICT 2006, Funchal.
[40] Han Q., Ning J., Chen Z., Shang L., Fan G., J. Opt A Pure Appl. Opt., 2005, 7, 386-390.
[41] Han Q., Ning J., Zhang H., Chen Z., J. Lightwave Technol, 2006, 24, 1946-1952.
[42] Min B., Lee W. J., Park N., IEEE Photon. Technol Lett., 2002, 12, 1486-1488.
[43] Newbury N. R., , J. Lightwave Technol, 2003, 21, 3364-3373.
[44] Mollenauer L. F., Grant A. R., Mamyshev P. V., Opt. Lett., 2001, 592-594.
[45] Xiao P. C., Zeng Q. J., Huang J., Liu J. M, IEEE Photon. Technol Lett. 2003, 15,
206-208.
[46] Yan M., Chen J., Jiang W, Li J., Chen J., Li X., IEEE Photon. Technol Lett. 2001, 13,
948-950.
[47] Zhou X., Lu C., Shum P., Cheng T. H., IEEE Photon. Technol Lett. 2001, 13, 945-947.
[48] Cui S, Liu L, Ma X, IEEE Photon. Technol Lett. 2004, 16, 2451-2453.
[49] Neto, B., Junior, S., Teixeira, A., Andr, P. European Conf. on Networks and Optical
Communications - NOC 2006, Berlin, Germany.
[50] Goldberg D. E., Genetic Algorithms in Search, Optimisation and Machine Learning,
Massachusetts: Addison-Wesley co, 1989, pp 28-56
[51] V. E. Perlin and H. G. Winful J. Lightwave Technol. 2002, 20, 250254.
[52] X. Liu, J. Chen, C. Lu, and X. Zhou, Opt. Express 2004, 12, 6053-6066.
[53] X. M. Liu, et al., J. Lightwave Technol. 2003, 21, 3446-3455.
[54] M. Islam, IEEE Journal of Selected Topics in Quantum Electronics 2002, 8.
[55] T. Miyamoto, R. Lindsay, Lightwave Magazine 2003, 1.
[56] Kobyakov, A., Gray S. and Vasilyev M. Electronics Letters 2003, 39, 732 733.
[57] Tsujkawa K., Tajima K., Ohashi, M. J. Lightwave Technol. 2000, 18, 1528-1532.
[58] Essambre R.; Winzer P.; Bromage J.; Kim C. H., Photonics Tech. Letters 2002, 14,
914 916.
[59] Park, J., Kim N. Y., Choi W.; Lee, H., Park N. Photonics Tech. Letters 2004, 16, 1459
1461.
[60] Teixeira A., Stevan Jr., S. Silveira T.; Nogueira R.; Tosi Beleffi G. M., Forin D., Curti
F. European Conf. on Networks and Optical Communications - NOC 2005,London, UK.
[61] P. B. Hansen, L. Eskildsen, A. J. Stentz, T. A. Strasser, J. Judkins, J. J. DeMarco, R.
Pedrazzani, and D. J. DiGiovanni, IEEE Photon. Technol. Lett. 1998, 10, 159-161.
[62] Faralli, S. Di Pasquale, F., IEEE Photonics Technology Letters 2003, 15, 804- 806.
[63] S. Stevan Jr., A. Teixeira, T. Silveira, P. Andr, G. M. Tosi Beleffi, A. Reale and A.
Pohl, Double shifted Raman amplification by means of spontaneous Rayleigh
Backsattering lasing control, ITC 2006.
[64] S. Stevan Jr.

, A. Teixeira, P. Andr, G. M. Tosi Beleffi, A. Pohl, simulation of Raman
amplification and Rayleigh Scattering laser using the transference matrix method, MTPT
2006, Leiria, Portugal.
[65] Andr, P., A. N. Pinto, Teixeira, A.T., Neto, B., Junior, S., Spertti, D., Rocha, F.,
Bernardo, M., Fujiwara, M., Rocha, A., Faco, M. ICTON 2007, Rome, Italy.
[66] Agrawal, G. P. Nonlinear Fiber Optics, 3rd ed. San Diego: Academic Press, 2001.


Chapter 3

FIBER BRAGG GRATINGS IN HIGH
BIREFRINGENCE OPTICAL FIBERS

Rogrio N. Nogueira, Ilda Abe
and Hypolito J. Kalinowski
Instituto de Telecomunicacoes, polo de Aveiro,
Aveiro, Portugal
Abstract
Fiber Bragg gratings (FBG) are a key element in optical communication devices and in
fiber sensors. This is mainly due to its intrinsic characteristics, which include low insertion
loss, passive operation and immunity to electromagnetic interferences. Basically a FBG is a
periodic modulation of the core refractive index formed by exposure of a photosensitive fiber
to a spatial pattern of ultraviolet light in the region of 244248 nm. The lengths of FBGs are
normally within the region of 120 mm. Usually a FBG operates as a narrow reflection filter,
where the central wavelength is directly proportional to the periodicity of the spatial
modulation and to the effective refractive index of the fiber. The production technology of
these devices is now in a mature state, which enables the design of gratings with custom-made
transfer functions, crucial for all-optical processing. Recently, some work has been done in the
application of FBG written in highly birefringent fibers (HiBi). Due to the birefringence, the
effective refractive index of the fiber will be different for the two transversal modes of
propagation. Therefore, the reflection spectrum of a FBG will be different for each
polarization. This unique property can be used for advanced optical processing or advanced
fiber sensing.
The chapter will describe in detail this unique device. The chapter will also analyze the
device and demonstrate different applications that take advantage of its properties, like
multiparameter sensors, devices for optical communications or in the optimization of certain
architectures in optics communications systems.
1. Introduction
The development of the fiber optical technology was an important step in the revolution of
global communications and in information technology. One of these developments happened
Rogrio N. Nogueira, Ilda Abe and Hypolito J. Kalinowski 84
in the 70s with the first optical fibers with low attenuation [1], a feature that enabled long-
distance communication with high bandwidth. The intrinsic optical bandwidth of the optical
fibers has also allowed the propagation of different simultaneous channels, allowing the
transmission of data at Tbit/s rates [2]. In these systems, in addition to transmission and
amplification, it is often necessary to do all-optical processing to the signal. This is due to the
inherent advantages of the optical processing, relative to the optic-electric-optic processing,
like the higher flexibility to operate at different bit rates and modulation formats and also at
the higher bandwidth. The evolution of the fiber optical technology has also enabled the
development of devices for all optical processing. In this way, the insertion loss is reduced
and the processing quality improved. One of the factors contributing to all-fiber optical
processing devices was the discovery of the photosensitivity in optical fibers. It was
documented for the first time in 1978 by Hill et al. [3] and led to the development of fiber
Bragg gratings (FBG).
A FBG is, generally speaking, a periodic perturbation, along the longitudinal axis, of the
refractive index in the fiber core. The production of the refractive index perturbation is done
optically in a photosensitive fiber. With the current techniques, it is possible to produce fiber
Bragg gratings with different optical properties, which can be designed according to the
desired optical processing. In addition to the high flexibility in the production of gratings with
custom amplitude and phase responses, the compatibility with common transmission fiber
also reduces the insertion loss and decreases the production costs.
The application in optical sensors is also a large potential market for FBG. Their intrinsic
low immunity to electromagnetic interference, high dynamic range, passive operation,
resistance to corrosion and the possibility of multiplexing hundreds of sensors have made
FBG a quite interesting sensor for different applications including medicine, civil, aeronautics
or biomechanics. Their properties enable the measurement of temperature and also
deformation with extremely high resolution. Nevertheless, it can also be used to measure
other parameters using indirect measurements [4-7]. The high potential of these devices has
also induced the creation of several companies dedicated to the production and installation of
fiber sensors.
There are already good references for the study of FBGs [8,9]. The purpose of this
chapter is not to study in detail these devices, but to describe a special case when a fiber
Bragg grating is written in high birefringence fibers (HiBi FBG). These special gratings have
unique polarization properties that give them exclusive capabilities for optical
communications. This is due to the possibility of applying a different optical processing for
different polarization components of the signal being transmitted.
HiBi FBGs are also quite interesting for multiparameter sensors, due to their response to
temperature variations and deformation. Sensors capable of measuring simultaneously several
physical parameters have increased in importance in todays technological world. In
particular, there are various applications of such sensors in civil, mechanical, biomedical or
aeronautical engineering, where measurements of different parameters are required

[10].
Engineering structures are an example of an application area for the multiparameters sensors,
where strain sensing can lead to better understanding about their lifetime and failure. Such
knowledge can be critical for some applications like smart skins for airplanes and
aeronautical vehicles.
Fiber Bragg Gratings in High Birefringence Optical Fibers 85
2. Fiber Bragg Gratings
A FBG is an optical device produced within the core of a standard optical fiber (figure 1).
Basically, it is a periodic modulation of the core refractive index formed by exposure of a
photosensitive fiber to a spatial pattern of ultraviolet light. The length of a FBG is dependent
on its application, but it generally varies between a few millimeters to a few centimeters.

Fiber Bragg grating

Fiber

Figure 1. Scheme of a Fiber Bragg grating written in an optical fiber.
The periodic modulation of the refraction index generates a resonant condition at the
Braggs wavelength (
B
) which is given by the Braggs condition:

2
B eff
n =
(1)

where n
eff
is the effective refraction index of the fiber and is the modulation period.
Therefore, when a FBG is illuminated by a broadband source, a spectral band centered at
B

will be reflected back. The reflection function can be determined using the coupled mode
theory [11-14], since it is difficult to determine analytically. The exception is the uniform
FBG, where it is possible to calculate the reflectivity in an analytical way. Considering a
uniform periodic modulation of the refractive index, with amplitude n, the reflection
coefficient of the grating can be given by

( )
( )
( ) ( )
sinh
sinh cosh
L
L i L

=
+
(2)

where L is the length of the FBG, the propagation constant mismatch, , is given by

2
eff
n

, (3)

2 2
= , and is the coupling constant given by


n

= (4)

where is the overlap integral and can be approximated as 1 for single mode fibers with
step index.
The reflectivity is given by

( )
( )
2
2
2
2
2
sinh
cosh
L
R
L
= =
(5)

and the phase by

( )
( )
Im
arctan
Re
R

=

(6)

Figure 2 shows the calculated reflectivity and the phase of a uniform FBG with L =5 mm
and n = 2x10
-4
as given by the above equations.

-1 -0.5 0 0.5 1
0
0.2
0.4
0.6
0.8
1
-1 -0.5 0 0.5 1
-4
-2
0
2
4
P
h
a
s
e

(
r
a
d
)
[nm]
[nm]
R
e
f
l
e
c
t
i
v
i
t
y

Figure 2. Reflectivity and phase of a uniform FBG. Parameters: L=mm and n= 2x10
-4
.
If the period changes linearly with the length of the grating, the FBG is said to have a
linear chirp. Figure 3 shows the simulation of the reflectivity and group delay of a linear
chirped FBG. The simulation method is based on the coupled mode theory.

z

1546 1547 1548 1549 1550 1551 1552 1553 1554
-20
-18
-16
-14
-12
-10
-8
-6
-4
-2
0

Wavelength [nm]
R
e
f
l
e
c
t
i
v
i
t
y

(
d
B
)
-50
0
50
100
150
200
250
300
G
r
o
u
p

d
e
l
a
y

[
p
s
]

Figure 3. Reflectivity and group delay of a linear chirped FBG.
3. High Birefringence Fibers
In an ideal monomode fiber, with a perfect cylindrical core, and with uniform diameter, the
fundamental propagation mode is a degenerated combination of two orthogonal propagation
modes. However, in real fibers, that degeneration does not exist. In fact, small variations of
diameter in the fibers core generate a birefringence in the optical fiber. The birefringence can
also be a result of an anisotropic stress in the fiber. The local birefringence, B, in each
position of the fiber, is defined as

( )
x y f x y
B n n C = =
(7)

where
x
n and
y
n are the mean refractive index of the orthogonal polarization modes,
x
and
y
are the main stress on the polarization axes and C
f
is the photoelastic constant of the fiber.
In monomode silica fibers C
f
is around 3.08 x10
-6
mm
2
/N for wavelengths near 1500 nm,
while B is typically B 10
-7
. Due to this small birefringence value, the two polarization
components of the light propagating in the fiber have a propagation velocity very similar.
Therefore, small environmental perturbations will lead to an energy coupling between one
polarization to another. As a result, a linearly polarized light will rapidly evolve to a random
polarization. This situation can be avoided with high birefringence fibers. In these fibers, the
core has an anisotropic stress, which is generated due to the geometric properties of the fiber.
Due to the photoelastic effect, the stress induces a birefringence in the core. Typical values
are B 10
-4
[15]. Due to the high birefringence, the propagation constant is different for the
two orthogonal propagation modes, which means that the coupling between both transversal
propagation modes is far lower as compared to standard fibers. Therefore, the higher the
birefringence, the easier will be for a linearly polarized light, propagating in one of the
orthogonal modes, to maintain its state of polarization. Due to this feature, HiBi fibers are
also known as polarization maintaining fibers. Figure 4 shows the main structure of the most
common HiBi fibers.

Fast axis (Y)
Slow axis (X)
air
PANDA
IEC Bow-Tie
Elliptical Core
Side-Hole
D-Shaped
Elliptical Core

Figure 4. Schematic of the transversal section of some of the most known HiBi fibers.
The PANDA (Polarization-maintaining AND Attenuation-reducing), IEC (Internal
Elliptical Cladding) and Bow tie fibers have anisotropic glass structures around the core, with
a Poisson coefficient different from the rest of the fiber. These structures create the
anisotropic stress in the core, which produces the birefringence. The Side-Hole, the Elliptical
Core and the D-Shaped Elliptical Core fibers have an elliptical core to generate the
birefringence, aided by two air structures, in the case of the Side-Hole or by the shape of the
cladding, in the case of the D-Shaped elliptical core. The main axes of the HiBi fibers are
designated as fast axis (Y) for the lower refraction index and slow axis (X) for the higher
refraction index.
Coherence Length
If a linearly polarized light propagates in a monomode fiber, with a polarization angle of 45,
relatively to the main axes of the fiber, both orthogonal polarization modes will be excited
with equal power. If the fiber has a constant birefringence, the mismatch,
HB
(z), between the
orthogonal polarization components will change as a function of the propagation distance on
the fiber, z, and its given by

( ) ( )
HB x y
z z = (8)

where
x
and
y
are the propagation constants in the X and Y axes respectively. The mismatch
will change periodically with the fiber, leading to a change in the state of polarization from
linear to elliptical and back again to linear (figure 5).

? =0
? =/2
? =
? =3/2 ? =2

Figure 5. Evolution of the state of polarization in a birefringence fiber.
The spatial periodicity of the evolution of the state of polarization is designated as
coherence length (L
B
). It is determined by the birefringence of the fiber and can be expressed
as

/
B
L B = (9)

where is the operating wavelength. Typical coherence lengths for HiBi fibers are in the
millimeter scale [16].
4. Fiber Bragg Gratings Written in HiBi Fibers
HiBi fibers can have two linear polarization modes with refractive indexes n
x
and n
y
for the
slow and fast modes respectively. When a FBG is written in one of these fibers, the periodic
modulation will be the same for the two orthogonal polarization modes; however since the
effective refraction index is different for the two polarizations, the Bragg wavelength will also
be different for each mode. Consequently, expression (1) can be rewritten for the two
orthogonal modes:

2 , X,Y
i i
n i = = (10)

where
i
are the Bragg wavelengths for each polarization mode.
The wavelength difference between the two reflection peaks,
HB
, can be calculated by

2 2
HB x y
x y
n n
=
=
(11)

The reflectivity of a HiBi FBG will be given by the linear sum of the reflectivity of the
two polarization components, i.e. R()=R
x
()+R
y
(). R
x
and R
y
are the reflectivity for each
polarization given by

( )
( )
2
2
2
2
2
sinh
cosh
i
i
i
i
L
R
L
= =
, i=x,y (12)

where

2
i
i
n
, i=x,y (13)

and

2 2
i i
= , i=x,y (14)

Figure 6 shows a simulation, using the previous model, for the reflectivity of a HiBi FBG
with birefringence of B = 3.2 10
-4
.

1547.5 1548.0 1548.5 1549.0
0.0
0.2
0.4
0.6
0.8
1.0
Polarization x

R
e
f
l
e
c
t
i
v
i
t
y
Wavelength [nm]
Polarization y

Figure 6. Reflectivity of a simulated HiBi FBG. Simulation parameters: B=3.2 10
-4
, =535 nm,
L=10 mm.
If the HiBi FBG is illuminated with light having the two orthogonal components, the
reflection spectrum will have those two peaks at orthogonal polarizations. This feature can be
very important in some applications, namely in optical communications, as it will be
confirmed further in this chapter.
The production of HiBi FBGs uses the same techniques as the ones used in regular FBGs. The
only difference will be in the utilization of photosensitive HiBi fiber. Generally it is used a
hydrogenated HiBi fiber.
Table 1 shows the dimensions of the anisotropic glass structures around the core of some
HiBi fibers obtained through the photographs of the transverse section. The table also
displays the main characteristics of HiBi fibers obtained from the manufacturers data sheet.
Table 1. Characteristics of different HiBi fibers. The structures of the HiBi fibers were
obtained by microphotography.
Fiber type
Commercial
provider
Wavelength
(nm)
Core
diameter
(m)
Cladding
diameter
(m)
Intrinsic stress-applying region
IEC (FS-
PM-6621)
3M 1300 8 125
Ellipse: Major axis: 75 m; Minor
axis: 30 m
Bow tie (F-
SPPC-15)
Newport 1550 8 125
From core center to extremity of
bow tie lobe: 18.4 m
Bow tie (HB-
1500G)
Fibercore 1550 8 80
From core center to extremity of
bow tie lobe: 16.5 m
PANDA
(SM-13-P-7)
Fujikura 1300 8 125
From core center to opposite
extremity of side cylinder: 41 m;
Diameter of side cylinder: 32 m

The reflection spectra for gratings written in the above fibers are shown on figure 7,
where the plots of the best-fitted bands are also presented [19]. All the gratings were
produced with the phase mask technique. The estimated length of the grating is 10 mm.
From these spectra it can be seen the effect of the intrinsic birefringence of the HiBi
fibers. The IEC fiber has the higher birefringence, corresponding to larger spectral splitting
between both polarizations bands, while the bow tie fiber presents the lowest birefringence.

1546.0 1546.5 1547.0 1547.5 1548.0
0.0
0.2
0.4
0.6
0.8
1.0
I

N
O
R
M
A
L
I
Z
E
D

I
N
T
E
N
S
I
T
Y
WAVELENGTH (nm)

Figure 7. Continued on next page.

1548.0 1548.5 1549.0 1549.5
0.0
0.2
0.4
0.6
0.8
1.0
I

N
O
R
M
A
L
I
Z
E
D

I
N
T
E
N
S
I
T
Y
WAVELENGTH (nm)

1548.0 1548.5 1549.0 1549.5
0.0
0.2
0.4
0.6
0.8
1.0
I

N
O
R
M
A
L
I
Z
E
D

I
N
T
E
N
S
I
T
Y
WAVELENGTH (nm)

Figure 7. Reflection spectra of Bragg gratings written in different HiBi fibers: IEC ( ), Panda ( )
and bow tie ( ). The continuous line represents the simulated best fit.
Table 2 shows the best-fit parameters obtained in the simulation process. From the fit it is
also possible to obtain the values of birefringence of the HiBi fibers.
Table 2. Parameters of FBGs written in HiBi fibers obtained for the best fit for the
experimental data.
HiBi Fiber Bands (nm) n
eff
kL (nm) B
IEC
(FS-PM-6621)
X
1546.57
1547.29
1.44539
1.44606
1.7212
1.7196
535 6.7 x 10
-4
@ 1550 nm
PANDA
(15P8)
X

1548.39
1548.82
1.44709
1.44750
1.7172
1.7162
535 4.1 x 10
-4
@ 1550 nm
Bow tie
(SPPC-15)
X

1548.61
1548.95
1.44730
1.44762
1.7167
1.7159
535 3.2 x 10
-4
@ 1550 nm

The Bragg wavelength peaks of the optical spectrum for both polarizations can change
with temperature and strain. Therefore, considering a HiBi FBG under a temperature variation
of T and under a strain aligned with the main axes of the fiber
X
,
Y
and
Z
, the
resultant wavelength shift,
x
and
y
of both wavelength peaks,
x
and
y
, can be
expressed as

( )
T
n
T n
)] ( p p [
2
n
X
Y Z 12 X 11
2
X
Z
X
X

+ + + + =

(15)

( )
T
n
T n
)] ( p p [
2
n
Y
X Z 12 Y 11
2
Y
Z
Y
Y

+ + + + =

(16)

where p
11
and p
12
are the components of the photoelastic tensor and is the thermal
expansion coefficient of the fiber, = 0.5510
-6
K
-1
[17]. For a fiber based on germanium
and silica, p
11
=0.113, p
12
=0.252 and the thermo-optic coefficient is
( )
n
T n
=8.610
-6
[18].
Figure 8 shows schematically the effect on the reflection spectrum of a HiBi FBG when it
is under temperature variations, under transversal strain or longitudinal strain. The effect of
temperature variations or longitudinal strain in the reflection spectrum is equivalent to a
translation in the wavelength. On the other hand, when under a transversal strain, the peak
separation will change. This difference can be used in multiparameter sensors as it will be
discussed further in this chapter.

Wavelength
P
o
w
e
r
Wavelength
P
o
w
e
r
Increasing longitudinal stress and/or temperature
Increasing transversal strain

Figure 8. Evolution of the reflection spectrum of a HiBi FBG when under a longitudinal stress,
temperature variation or transversal strain.
4.1. Characterization of Bragg Gratins Written in High-Birefringence Fiber
Optics
4.1.1. Transverse Strain
The sensitivity of HiBi FBG to transversal strain can be characterized using a mechanical set-
up, like the one shown in figure 9. The transversal load is applied using a micro scratch
mechanical system. The system uses an arm to apply a load with a precision of 0.1 N. A
grating written in HiBi fiber was placed between two plates having a length of 13 mm. The
apparatus arm applies the load to the upper plate. The transverse loads were made for several
orientations of the birefringence axis with respect to the direction of the applied load through
two fiber rotators. Figure 9 also shows the optical system used to analyze the FBG reflection
spectrum. Optical spectra were recorded using an amplified spontaneous emission (ASE) of
an erbium doped fiber amplifier as light source, an optical circulator and conventional optical
spectrum analyzer (OSA).

Circulator

Figure 9. Set-up for the characterization of HiBi FBGs under a transversal load. The detail shows the
transverse section of a HiBi fiber oriented along the angle .when subjected to applied force F. ASE:
Broadband optical source (amplified spontaneous emission); OSA: optical spectrum amplifier.
Figure 10 (a), (b) and (c) shows an example of the reflection spectra of a FBG written in
a IEC HiBi fiber as a function of an applied load of 0, 45 and 90, respectively.
The results show that, if a load is applied to one of the main axes, fast or slow, it leads to a
change in the wavelength of the spectral band associated with the orthogonal axis, while the band
associated with the correspondent axis will show a smaller variation. For the applied load angle of
45 both polarization bands present similar evolution. The figure also shows, for different
applied load angles (), the evolution of the peak wavelength of each reflection band with the
transverse strain applied to the sample. The strain calibration points in the spectra deformed
areas were obtained by identifying and measuring local maximum, minimum and inflexion
points. The band split that occurs in some of the spectra is due to a phase shift induced by the
applied load. The resulting complex structure is known to be responsible for spectral changes
of FBG subject to mechanical stress [9].

F
X
Y
1546.0 1546.5 1547.0 1547.5 1548.0
0.0
0.2
0.4
0.6
0.8
1.0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
WAVELENGTH (nm)
L
O
A
D

(
N
/m
m
)
N
O
R
M
A
L
I
Z
E
D

I
N
T
E
N
S
IT
Y
-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
1546.4
1546.6
1546.8
1547.0
1547.2
1547.4
1547.6
1547.8
Y
X

W
A
V
E
L
E
N
G
T
H

(
n
m
)
LOAD (N/mm)
(a)
-0.1 0.0 0. 1 0.2 0. 3 0.4 0.5 0.6 0.7
1546.4
1546.6
1546.8
1547.0
1547.2
1547.4
1547.6
1547.8

Y
X
W
A
V
E
L
E
N
G
T
H

(
n
m
)
LOAD (N/mm)
1546.0 1546.5 1547.0 1547.5 1548.0
0.0
0.2
0.4
0.6
0.8
1.0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
L
O
A
D

(
N
/
m
m
)

WAVELENGTH (nm)
N
O
R
M
A
L
I
Z
E
D

I
N
T
E
N
S
I
T
Y
F
Y
X
(b)
-0.1 0. 0 0.1 0.2 0.3 0. 4 0.5 0.6 0.7
1546.4
1546.6
1546.8
1547.0
1547.2
1547.4
1547.6
1547.8

Y
X
W
A
V
E
L
E
N
G
T
H

(
n
m
)
LOAD (N/ mm)
1546.0 1546.5 1547.0 1547.5 1548.0
0.0
0.2
0.4
0.6
0.8
1.0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
WAVELENGTH (nm)
L
O
A
D

(
N
/
m
m
)
N
O
R
M
A
L
IZ
E
D
I
N
T
E
N
S
I
T
Y

F
Y
X
(c)

Figure 10. Left: Changes in the spectral response of a FBG written in an IEC HiBi fiber when subjected
to an applied load oriented along the angle : (a) 0 (X-axis); (b) 45 and (c) 90 (Y-axis). Right: Peak
position of each band as a function of the applied load. The lines represent the linear best fit for the
experimental data.

-90 -60 -30 0 30 60 90
-0.2
0.0
0.2
0.4
0.6

Y

X

W
A
V
E
L
E
N
G
T
H

S
E
N
S
I
T
I
V
I
T
Y

(
n
m
/
N
/
m
m
)
(degrees)

Figure 11. Curves of peak sensitivities of the FBG in IEC HiBi fiber as a function of the applied load
angle.
Figure 11 shows the wavelength sensitivity curves obtained for both polarization bands.
The graph also displays the periodic evolution of the bands as a function of the applied load
angle.
Identifying and measuring the reflection peaks as a function of the applied load can be
used to obtain the calibration line for each polarization band. The respective slopes can be
evaluated and, from them, the dependence of the Bragg wavelength position with the strain
can be obtained. Table 3 shows some measurements obtained with FBGs written in IEC
and PANDA fibers.
Table 3. Slopes and strain sensitivities of FBGs written in IEC and PANDA HiBi fibers
as a function of the direction of applied load (module values). Both fibers have a
diameter of 125 m.
X- polarization band Y- polarization band
HiBi Fiber
Angle of
applied
load
Slope
(nm/N/mm)
Strain sensitivity
(pm/)
Slope
(nm/N/mm)
Strain sensitivity
(pm/)
= 90 0.51 7.02 0.07 1.02 IEC

= 0 0.02 0.29 0.11 1.55
= 90 0.46 3.78 0.02 0.24 PANDA

= 0 0.01 0.11 0.13 2.80

4.1.2. Longitudinal Strain
The Bragg wavelength dependence with the longitudinal strain can be measured by gluing
one extremity of the fiber in a holder, while the other is glued to a translation stage, which
applies a known deformation using a calibrated micrometer.
Figure 12 shows the reflection spectra of a FBG and peak position of each band, written
in an IEC fiber as a function of longitudinal strain.

1545.0 1545.5 1546.0 1546.5 1547.0
0.0
0.2
0.4
0.6
0.8
1.0
0
83.7
167.5
251.2
334.9
418.7

L
O
N
G
I
T
U
D
I
N
A
L

S
T
R
A
I
N

(
)

WAVELENGTH (nm)
N
O
R
M
A
L
I
Z
E
D

I
N
T
E
N
S
I
T
Y
0 100 200 300 400
1545.2
1545.6
1546.0
1546.4
1546.8
Y

X
.
.
.

W
A
V
E
L
E
N
G
T
H

(
n
m
)
LONGITUDINAL STRAIN ()

Figure 12. Left: Changes in the spectral response of a FBG written in IEC HiBi fiber when subjected to
a longitudinal strain. Right: Peak position of each band as a function of the longitudinal strain. The
lines represent the linear best fit for the experimental data.

Both bands show the same behavior, which is an increase of peak wavelengths as the
strain increases. The slopes and the Bragg wavelength sensitivity to longitudinal strain are
given in table 4. The obtained ratios between strain and applied load were 758 /N (X-axis)
and 755 /N (Y-axis).
Table 4. Slopes and longitudinal strain sensitivity of a FBG written in an IEC HiBi fiber.
Bands Slope (nm/N) Longitudinal strain sensitivity (pm/)
X polarization 1.44 1.9
Y - polarization 1.51 2.0

4.1.3. Temperature
The temperature dependence of the reflection bands of HiBi FBGs can be characterized using
a cooling/heating system. Figure 13 shows the evolution of the reflection bands and peak
position of each band of a Bragg grating written in an IEC fiber as a function of temperature.

1546.0 1546.5 1547.0 1547.5 1548.0
0.0
0.2
0.4
0.6
0.8
1.0
15
20
25
30
35
40
45
50
55

T
E
M
P
E
R
A
T
U
R
E
(
C
)

WAVELENGTH (nm)
N
O
R
M
A
L
I
Z
E
D
I
N
T
E
N
S
I
T
Y

10 20 30 40 50 60
1546.5
1546.8
1547.1
1547.4

Y

X

W
A
V
E
L
E
N
G
T
H

(
n
m
)
TEMPERATURE (C)

Figure 13. Left: Changes in the spectral response of a FBG written in an IEC HiBi fiber when subjected
to different temperatures. Right: Peak position of each polarization band as a function of the
temperature. The lines represent the linear best fit for the experimental data.
Table 5 shows the temperature sensitivity values for a FBG written in IEC, PANDA and
bow tie HiBi fibers.
Table 5. Slopes of temperature for FBGs written in HiBi fibers.
Slope (pm/C)
HiBi fiber
X polarization band Y polarization band
IEC 125 m
6.76 6.71
PANDA 125 m
3.28 3.40
Bow Tie 125 m
10.93 11.12
Bow Tie 80 m
8.02 8.46

The results show that there are quite large variations between the sensitiveness to
temperature for different HiBi fibers. The values changed between ~ 3 pm/C for the
PANDA fiber to ~ 11 pm/C for the IEC fiber. There are also differences in the
coefficients between polarization bands of the same fiber. For example, in the PANDA
fiber this difference is 0.24 pm/C. These results can be used for simultaneous
measurements of temperature and longitudinal strain with only one FBG in a HiBi fiber.
This approach, along with others, will be described in the next section.
5. Application in Multiparameter Sensors
FBG sensors are generally based on a unique grating written in a standard fiber optic. The
wavelength shift in the reflection spectrum may be used to measure a single component of
strain or temperature variation, but not both simultaneously. An adequate measurement of
both temperature and strain requires a suitable sensor with a differential sensitivity between
parameters. HiBi FBGs can be used as sensors to simultaneously measure one component of
transverse strain, temperature and/or longitudinal strain. As it was shown previously in this
chapter there are differences in the calibration coefficients of both polarization bands, which
can be used to simultaneous measure the temperature and longitudinal strain with only one
HiBi FBG. Since the variations of temperature or longitudinal strain causes both bands to
shift, and the variation of strain causes asymmetric spectral response in the polarization bands
depending of the direction of the applied load, allows the FBG in the HiBi fiber to measure
simultaneously transverse strain and temperature or transverse strain and longitudinal strain.
Several types of optical sensors using FBG written in HiBi fibers, which simultaneously
measure longitudinal strain and temperature have been proposed and demonstrated [20-24].
Some of the methods include the recording by a CCD camera of the LP
01
and LP
11
spatial
modes [22], using a HiBi FBG partially exposed to chemical etching [20] or by using a quasi-
rectangular HiBi fiber to increase the birefringence [21]. In those works, only the longitudinal
strain component was measured in simultaneous with temperature. But, there are many
applications where it is desirable to determine the transverse strain components in addition to
longitudinal strain. Several techniques based in HiBi FBG have already been reported for
transverse strain sensing [19, 25-29]. However, when a transverse strain is applied to a
HiBi FBG, depending of the fiber orientation relatively to the applied load, the
separation of the two Bragg wavelengths can be quite low, so it becomes impossible to
resolve the two peaks. To overcome this problem, it can be used an interrogation system
capable of detecting independently and simultaneously the two orthogonally polarized signals
reflected from the HiBi FBG [26].
There are many applications where it is necessary an ultra small sensor to measure
simultaneously components of transverse strain, longitudinal strain and temperature. The use
of two superimposed Bragg gratings in HiBi fiber have been described in the literature like
potential sensors for monitoring four parameters, two components of transverse strain,
longitudinal strain and temperature. [30-34].
5.1. Simultaneous Measurement of Transverse Strain and Temperature
The change in the Bragg wavelength of a HiBi FBG, for each polarization, due to a
temperature change T and a transversal strain , is given by


X X
X
T
T

= +

(17)

Y Y
Y
T
T

= +

(18)

were
X
/T and
Y
/T are the temperature coefficients and
X
/ and
Y
/ are the
transverse deformation coefficients.
Expressions (17) and (18) can be rearranged and written in matrix form in order to
calculate the transverse strain and temperature, given the measured wavelength shifts for each
polarization band:

1 X
Y
T

=

(19)

where K is a matrix given by

Y Y
X X
,
T
,
T
K (20)

A simultaneous measurement of transverse strain and temperature can be obtained by
determining the coefficients of K, which are determined with previous characterization.
Two examples of simultaneous measurement of these parameters are shown in the table 6
for an IEC fiber and table 7 for a PANDA fiber. The results were obtained using the values of
the Bragg wavelength changes for both polarizations bands.
Table 6. Simultaneous measurements of temperature and transverse strain using a FBG
written in a IEC HiBi fiber. The set values were determined by the experimental system
equipment. Direction of applied load: 0. [19].
Set values 12 C 23 C 31 C 46 C
11.8 C 24.6 C 31.5 C 45.5 C
61

65 71 70 75
12.0 C 25.4 C 32.6 C 46.3 C
76

69 79 73 81
12.7 C 26.8 C 34.0 C 48.0 C
91

76 83 78 79
Table 7. Simultaneous measurements of temperature and transverse strain using a FBG
in written in a PANDA HiBi fiber. The set values were determined by the experimental
system equipment. Direction of applied load: 90.
Set values 7 C 21 C 22 C 40 C 53 C
7 C 20 C 23 C 39 C 51 C
11

12 12 9 11 10
7 C 19 C 22 C 38 C 51 C
21

21 20 17 16 18
6 C 18 C 22 C 38 C 53 C
31

32 31 31 22 24
5 C 18 C 21 C 37 C 55 C
41

39 44 44 37 39
5 C 18 C 21 C 37 C 55 C
51

43 42 48 41 43
5.2. Simultaneous Measurement of Transverse Strain and Longitudinal
Strain
For the measurement of the longitudinal (
Z
) and transverse (
X
or
Y
) strain, the
equations can also be written in matrix form, given the measured wavelength shifts for each
polarization band:

1
,
Z X
X Y Y

=

(21)

where K is now given by:

,
,
,
,
X X
Z X Y
Y Y
Z X Y
K

=

(22)

Table 8 shows an example of simultaneous measurements of longitudinal and transversal
strain obtained using the wavelength changes of both polarizations bands.
Table 8. Simultaneous measurements of longitudinal and transverse strain using an
FBG written in a IEC HiBi fiber. The set values were determined by the experimental
system equipment. Direction of applied load: 90.
Set values 0 9 14
1 8 13 83
64 66 68
1 8 13 167

160 161 154
0 7 12 251
240 246 244
1 8 13
335
320 316 313
5.3. Simultaneous Measurement of Transverse Strain, Longitudinal Strain
and Temperature
Two superimposed Bragg gratings can be written in high birefringence fiber optics to
measure simultaneously temperature, transverse and longitudinal strain.
This section demonstrates the use of a pair of Bragg gratings written in high birefringence
fiber optics to measure, simultaneously, three physical parameters [31]. The Bragg gratings
are superimposed in the same position of the fiber optic, in order to behave as a single sensor
with reduced dimension.

5.3.1. Superimposed Bragg Gratings
1534 1536 1546 1548
0
8
16
24
32
X
2
X
1
Y
2
Y
1

I
N
T
E
N
S
I
T
Y

(
n
W
)
WAVELENGTH (nm)

Figure 14. Optical reflection spectrum of two superimposed Bragg gratings written in HiBi IEC fiber
[31].
Figure 14 shows an optical reflection spectrum of two gratings recorded at the same fiber
position. The two FBG were written with different periods in an IEC HiBi optical fiber with
125 m diameter. The figure shows the polarization bands (Y-polarization and X-
polarization) of each pair. Their relative intensity is not the same as the optical source was not
flat along the full wavelength range.
The superimposed HiBi FBGs were characterized by longitudinal, transversal
strain and temperature. The measurements of transversal load were made with the fiber
oriented with the fast or slow birefringence axis in the direction of the applied load.
Figure 15 shows the dependence of the peak position of each reflection band against the
transversal strain applied to the sample (load applied along the Y-axis direction). The best-
fitted lines are not parallel; their slopes are different depending on the polarization band. This
asymmetric behavior can be used to distinguish the effects of longitudinal and transversal
strain acting upon the grating pair.

0 20 40 60 80
1534
1535
1536
1546
1547
1548

W
A
V
E
L
E
N
G
T
H

(
n
m
)
TRANSVERSE STRAIN ()

Figure 15. Dependence of the peak wavelength on transverse strain for the reflection bands [X (V) and
Y (U)] of the two superimposed FBGs written in an IEC HiBi fiber. Direction of applied load: 90. The
lines represent the linear best fit to the experimental data [31].
The behavior of the reflection bands, when the sensor is under longitudinal strain, is the
same for both gratings. The temperature dependence of the reflection bands of the both FBGs
has also approximately the same behavior, which is an increase in the wavelength with an
increase of temperature.
Table 9. Slopes of temperature, longitudinal and transverse strain for the two superimposed
FBGs in an IEC HiBi fiber. Direction of applied transverse load: Y-axis [31].
Polarization bands
Slopes
Y
1
X
1
Y
2
X
2

/T (pm/C) 8.4 7.8 7.8 7.5
/
Y
(pm/) 0.08 4.02 0.19 4.11
/
Z
(pm/) 1.3 1.39 1.39 1.36
The corresponding slopes of temperature, longitudinal and transversal strain for both
polarization bands, for the best-fitted lines of superposing FBGs in IEC HiBi fiber, are given
in table 9.

5.3.2. Simultaneous Measurements
The change in the Bragg wavelength of the reflection spectrum of the both FBGs, due to a
temperature change T, a transverse strain (
X
or
Y
) and longitudinal strain
Z
, for each
polarization, is given by

1 1 1
1 ,
,
X X X
X X Y Z
X Y Z
T
T

= + +

(23)

1 1 1
1 ,
,
Y Y Y
Y X Y Z
X Y Z
T
T

= + +

(24)

2 2 2
2 ,
,
X X X
X X Y Z
X Y Z
T
T

= + +

(25)

2 2 2
2 ,
,
Y Y Y
Y X Y Z
X Y Z
T
T

= + +

(26)

where
X1
/T,
X2
/T,
Y1
/T and
Y2
/T are the temperature coefficients,
X1
/
X,Y
,
X2
/
X,Y
,
Y1
/
X,Y
and
Y2
/
X,Y
are the transversal deformation coefficients, and
X1
/
Z
,
X2
/
Z
,
Y1
/
Z
and
Y2
/
Z
are the longitudinal deformation coefficients.

Equations (23) to (26) can be rearranged and written in matrix form, in order to calculate
the transverse, longitudinal strain and temperature, given the measured wavelength shifts for
each polarization band. In this way, the calculation of the three parameters being measured
can be made using the following (the choice of reflection bands was arbitrary):

1
, 1
2
Y
X Y X
Z Y
T
K

=

1
(27)

where K is assembled from the several sensitivities for temperature and deformation:

1 1 1
,
1 1 1
,
2 2 2
,
K
Y Y Y
X Y Z
X X X
X Y Z
Y Y Y
X Y Z
T
T
T

=

(28)

After a previous characterization, in order to obtain K, the temperature, longitudinal and
transversal strain components can be simultaneously measured. Some of the obtained results
with the grating pair described above are given in table 10.
Table 10. Simultaneous measurements of temperature, transverse and longitudinal
strain using two superimposed FBGs in IEC HiBi fiber. The set values were determined
by the experimental system equipment [31].
167 251
Set values
15 C 45 C 15 C 45 C
12 C 42 C 12 C 37 C
13 10 16 10 12
117 99 228 177
16 C 33 C 16 C 43 C
18 16 16 24 22
141 132 252 187
16 C 36 C 18 C 43 C
32 29 21 32 32
116 139 236 178
5.4. Bragg Gratings in Reduced Diameter High Birefringence Fiber Optics
Bragg gratings written in reduced diameter high birefringence fiber optics can also be used
for multiparameter sensing. Changes in the stress profile of HiBi fibers due to reduced
diameter can modify the response of a FBG sensor system to strain or temperature optimizing
the simultaneous measurement of those parameters. Chemical etching can be a good tool to
reduce the fiber diameter. The changes in the birefringence properties of HiBi fibers as a
function of fiber diameter can be analyzed using fiber samples chemically etched in
hydrofluoric acid (HF), while the optical spectra of pre-recorded gratings are measured [34].

0 10 20 30 40 50
70
80
90
100
110
120

D
I
A
M
E
T
E
R

(
m
)
EXPOSURE TIME (min)

Figure 16. Diameter of an IEC HiBi fiber as a function of the exposure time. HF concentration: 20%
[34].
The diameter of the fibers during the etching can be measured by having several samples of
the fiber in the acid. The samples are removed successively from the acid, rinsed in distilled
water, dried, and then measured under a microscope with a calibrated scale.
The evolution of the diameter, as a result of etching, for an IEC fiber is presented in
figure 16. HF acid was diluted to 20 % (parts per volume) in order to reduce the velocity of
chemical etching and to increase the sampling points along the process. Figure 17 shows the
changes in the transversal section of the IEC fiber, with 125 m of diameter (left) and after
etching (right), with 86 m of diameter. The internal elliptical cladding can be observed in
these photographs. The major axis of the ellipse has approximately 75 m. The etched IEC
fiber shows a higher asymmetry on the borders close to the axes along the major axis of the
internal elliptical cladding.

Figure 17. Microphotographs of the transverse section of an IEC HiBi fiber. Left: standard HiBi fiber
with 125 m of diameter. Right: etched HiBi fiber with 86 m of diameter [34].
1545.2 1545.6 1546.0 1546.4
0
50
100
150
200
4
12
20
28
36
44
P
O
W
E
R

(
p
W
)
E
X
P
O
S
U
R
E

T
I
M
E

(
m
i
n
)

WAVELENGTH (nm)

70 80 90 100 110 120 130
1545.2
1545.4
1545.6
1545.8
1546.0
1546.2

W
A
V
E
L
E
N
G
T
H

(
n
m
)
DIAMETER (m)

Figure 18. Left: evolution of the reflection bands of a FBG written in an IEC HiBi fiber as a
function of the etching time. Right: peak position of the polarized bands (Y-polarized () and X-
polarized ())as a function of the fiber diameter. The lines represent the linear best fit for the
experimental data. HF concentration: 20 % [34].
Figure 18 (left) illustrates the optical reflection spectra of the FBG in IEC fiber, obtained
as a function of HF exposure time. After 36 minutes of exposition time, the optical spectrum
had a single band, which means that, the fiber birefringence was almost zero. That is a
consequence of the stress release due to the etching.
Figure 18 (right) shows the changes in the peak position of the reflected polarized bands
as a function of the IEC fiber diameter. The different slopes for the X and Y polarized bands
can be related to asymmetric changes of the internal stress applied by the internal elliptical
cladding.
The evolution of the birefringence, as a function of the diameter, can be seen in figure 19.

70 80 90 100 110 120 130
0.0
1.0x10
-4
2.0x10
-4
3.0x10
-4
4.0x10
-4
5.0x10
-4
6.0x10
-4

B
I
R
R
E
F
R
I
N
G
E
N
C
E
DIAMETER (m)

Figure 19. Calculated birefringence of the IEC HiBi fiber as a function of diameter. HF concentration:
40 % () and 20 % () [34].
1545.6 1545.9 1546.2 1546.5
0.0
2.0x10
-5
4.0x10
-5
6.0x10
-5
8.0x10
-5
90
110
130
150
170
183
187
E
X
P
O
S
U
R
E

T
I
M
E

(
m
i
n
)

WAVELENGTH (nm)
P
O
W
E
R

(
m
W
)

40 50 60 70 80 90 100110120130
1545.6
1545.8
1546.0
1546.2
1546.4

W
A
V
E
L
E
N
G
T
H

(
n
m
)
DIAMETER (m)

Figure 20. Left: evolution of the polarized bands of a FBG written in a bow tie HiBi fiber as a function
of the etching time. Right: peak position of polarized bands (Y-polarized () and X-polarized ()) as a
function of the fiber diameter. The lines represent the linear best fit for the experimental data. HF
concentration: 20 % [34].
A similar characterization can be made to other types of HiBi fibers. For example,
figure 20 (left) shows the effect of chemical etching in the optical spectrum of a Bragg grating
written in a bow tie fiber. The etching rate is lower and it is possible to observe that the two
polarization bands collapse. Initially both bands show a trend to longer wavelengths on their
peak position, as the diameter changes from 100 m to 65 m (figure 20 (left)). Further
etching now causes the X polarized band to shift sharply to shorter wavelengths, until both
bands collapse when the diameter reaches approximately 40 m. This value agrees with the
intrinsic stress-applying region dimensions, where the distance between the boundaries of the
two internal side-lobes is approximately 37 m.
Figure 21 shows the birefringence for a bow tie fiber as a function of the diameter. The
results show that IEC and bow tie fibers have vanishing birefringence for diameters that are
close to the value of the maximum dimension of the stress-applying region.

B
I
R
E
F
R
I
N
G
E
N
C
E
DIAMETER (m)

Figure 21. Measured birefringence of bow tie HiBi fiber as a function of diameter. HF concentration:
20 % [34].

5.4.1. Reduced Diameter for the Simultaneous Measure of Transverse Strain and
Temperature
A FBG in an etched HiBi fiber can be applied as a sensor to simultaneously measure the
transverse strain and temperature. Once again, a previous calibration of the different
sensitivities must be made. The temperature and transverse strain coefficients for an etched
IEC fiber is shown in table 11. It also displays the coefficients for a non-etched bow tie fiber
with a similar diameter.
Table 11. Slopes of temperature and transverse strain of a FBG written in etched IEC
and non-etched bow tie HiBi fibers [34].
Temperature Transversal strain
Fiber
(diameter)
x
/T (pm/C)
y
/T (pm/C)
x
/ (pm/)
y
/ (pm/)
Etched IEC
(82 m)
7.00 6.90 0.7 (X -axis)
3.4 (Y -axis)
2.23(X -axis)
0.1 (Y -axis)
Bow tie
(80 m)
8.02 8.46 0.02 (X -axis)
1.2 (Y -axis)
1.16 (X -axis)
0.3 (Y -axis)

The results of simultaneous transversal strain and temperature measurements obtained
with matrix K and the values of
X
and
Y
of the reflection spectra are displayed in
table 12.
Table 12. Simultaneous measurements of temperature and transverse strain using
etched FBGs in IEC HiBi fiber (diameter of 82 m). The set values are determined by
the experimental system equipment. Direction of applied load: 90 [34].
Set values 16 C 26 C 36 C 46 C 56 C
15 C 28 C 33 C 44 C 56 C
33
37 37 39 42 42
17 C 29 C 36 C 47 C 56 C
48
54 57 56 54 59
17 C 28 C 36 C 46 C 57 C
64
49 48 48 54 53
17 C 29 C 36 C 48 C 58 C
79
66 80 71 65 73
17 C 29 C 36 C 48 C 58 C
94
80 100 94

91 85

The errors obtained using a FBG in normal and reduced diameter HiBi fibers as a sensor
are of comparable magnitude, but the dynamic range for strain measurements with the later
ones is almost doubled as compared to the former sensors. This fact is important for
technological applications where FBG can be tailored to attend a specific measurement range.
6. Applications to Optical Communications
All optical processing devices are becoming a key element in the next generation of optical
communication systems, since they play a critical role in pulse formatting, spectral shaping
and optimized all-optical routing and switching. These devices dont have the typical
bottleneck associated to the optical-electrical-optical conversion and the majority is
transparent to modulation format and bit-rate. FBGs are quite interesting for these
applications, due to their low insertion loss and due to the avoidance of the decoupling of the
signal outside the fiber. Moreover, the production technology is now in a mature state, which
enables the design of gratings with custom made transfer functions, crucial for all-optical
processing. Some advanced processing can be made if the transfer function is different for the
two transversal modes of propagation in the fiber. This can be achieved by a HiBi FBG. One
of the devices that take full advantage of the optical processing capabilities of the HiBi FBG
is the orthogonal pumps source [35-37], which can be used in all optical wavelength
converters [38, 39]. A tunable PMD compensator can also be developed based on the
polarization processing properties of these special gratings [40, 41]. Also, a tunable
microwave-photonic notch filter that makes use of a time delay element based on tunable
HiBi chirped FBG has been demonstrated [42, 43] In addition, the interference due to laser
coherence, typical in those micro-wave photonic filters was also reduced due to the
polarization properties of the HiBi FBGs.
The following sections describe some example application of HiBi FBG in optics
communications.
6.1. Optical Delay Line for PMD Compensation
In a linearly chirped grating, written in a HiBi fiber, each position of the grating will reflect
two wavelengths at orthogonal polarizations (figure 22). This means that the group delay of
these gratings is a combination of two linear functions, one for each polarization, with the
same slope (D
FBG
) and shifted by
HB
:

( )
( )
( )
y FBG
x FBG HB
D b
D b

= +
= +
(29)

where b in (29) is a constant.
Therefore, the relative group delay induced by a linearly chirped FBG written in a HiBi
fiber (=
x
-
y
) is calculated using the following expression

2
FBG
FBG
D
D B
=

(30)

Expression (30) shows that the dynamic tuning of the induced PMD can be made by
adjusting the birefringence of the fiber, which can be done by applying a transversal stress in
the fiber, as shown before in this chapter.

1546 1548 1550 1552 1554
-50
-40
-30
-20
-10
0
y polarization
x polarization
Wavelength [nm]
R
e
f
l
e
c
t
i
v
i
t
y

[
d
B
]
0
50
100
150
200
250
300
HB
G
r
o
u
p

d
e
l
a
y

[
p
s
]

Figure 22. Reflectivity and group delay of a linearly chirped HiBi FBG for both transversal propagation
modes [47].
6.1.1. Compensation Using a Linear Chirp
As can also be observed in expression (30), it is also possible to tune the PMD by adjusting
the dispersion of the grating. That can be done using different methods [44-46]. One of them
is by using thermal gradients to induce a linear chirp to a uniform FBG. Let us consider a
uniform HiBi FBG put in a thermal contact with metal substrate. By applying different
temperatures to the substrate, different linear temperature gradients will be generated. This
gradient will induce a linear chirp to the FBG, due to thermo-optic and photoelastic effects.
By changing the temperature gradient, the dispersion will also change, inducing a tunable
differential delay line [47]. Figure 23 shows the experimental results of the evolution of as
a function of the applied temperature gradient to a 24 mm uniform HiBi FBG.

10 20 30 40 50
-50
0
50
100

[
p
s
]
T[C]

Figure 23. Relative group delay as a function of the applied linear gradient to a uniform HiBi FBG with
24 mm length.
Therefore, the presented device can be included in a PMD compensator as a tunable
optical relative group delay line.

6.1.2. Compensation Using a Nonlinear Chirp
Let us now suppose that we have a HiBi FBG with a quadratic chirp. The group delay is now
composed by two parabolic functions (one for each polarization) shifted by
HB
. If the
grating is tuned by temperature or longitudinal stress, the relative induced delay between the
orthogonal polarizations, for a specific wavelength will change [40]. Figure 24 shows a
simulation of a quadratic chirped FBG, with a length of 25 mm, written in a HiBi fiber with
birefringence B = 5x10
-4
. For a tuning of 4.5 nm in the central wavelength, the relative group
delay at 1550 nm changed from 41.6 ps to 12.1 ps. In this way, with this method, it is possible
to do small corrections in the relative group delay.
The advantages of this method are its tuning simplicity and the flexibility in the operation
range. However, the technique needs a FBG with a nonlinear chirp, which is quite complex to
produce. It is generally produced with a custom made phase mask with a nonlinear chirp.

1544 1546 1548 1550 1552 1554 1556
0
50
100
150
200
250
= 12.1 ps
G
r
o
u
p

d
e
l
a
y

[
p
s
]
Wavelength [nm]
= 41.6 ps
Tuning

Figure 24. Simulation of the group delay of a HiBi FBG with quadratic chirp. Line: Y polarization;
dots: X polarization [47].
6.2. Tunable Multiwavelength Linear Polarized Fiber Lasers
Fiber lasers have different applications in sensors and telecommunications due to their
reduced linewidth, power and spectral profile. Like other lasers, fiber lasers need two
components: a gain medium and a resonant cavity. For a fiber laser operating around
1550 nm, it is generally based on an optical pump with 980 or 1480 nm of wavelength, an
erbium-doped fiber and an optical filter. The gain is obtained from the amplified spontaneous
emission due to the optical pump.
Generally, fiber optical lasers based on an optical ring with erbium-doped fiber dont
enable the generation of more than one laser line [48,49]. This is a consequence of the fact
that erbium is a medium with homogeneous gain at room temperature, resulting in strong
mode competition, which induces laser instability. A method was proposed to reduce the
homogeneity of the fiber by cooling the fiber to 77 K [50, 51]. However, by obvious reasons,
it is not very practical. Other methods used special fibers like the elliptical core fibers [52] or
the twincore fibers [53].

PC
Circulator
Two Tunable
HiBi FBG
Output
Optical Pump
(980 nm)
EDF
WDMc

Figure 25. Diagram of a multiwavelength fiber laser based on HiBi FBGs. EDF: Erbium doped fiber;
PC: Polarization controller; WDMc: WDM optical coupler;
Another way to reduce the homogeneity of the fiber is to use different laser lines
operating at different longitudinal modes. For this kind of implementation, HiBi FBGs can
have an important role, since they will reflect two wavelengths at orthogonal polarizations.
An implementation method for a tunable laser with up to four laser lines is depicted in
figure 25.

-60
-50
-40
-30
-20
-10
0
-60
-50
-40
-30
-20
-10
0
1530 1535 1540 1545 1550 1530 1535 1540 1545 1550
-60
-50
-40
-30
-20
-10
0
x
x
yy
x
x
x
x

x

y
y
y
P
o
w
e
r

[
d
B
m
]

Wavelength [nm]

Figure 26. Optical spectra at the output of the fiber laser with different operation modes. The
operating laser lines are at a linear polarization (x or y).
The two tunable HiBi FBGs enable the selection of 4 different wavelengths. By tuning
the polarization controller (PC) inside the optical cavity, it is possible to select the appropriate
laser lines. Figure 26 shows some of the possibilities that can be achieved with just two HiBi
FBGs.
One of the advantages of this technique is its ability to generate two laser lines at
orthogonal polarizations (see last spectrum of figure 26). Therefore, it can be used as two
orthogonal pumps in a polarization insensitive wavelength converter [38].
6.3. Optical Networks Architectures Using HiBi FBG for Performance
Improvement
6.3.1. Optical Code Division Multiple Access
Metro optical code division multiple access (OCDMA) networks can benefit from the
polarization multiplexing, since two users using codes in the same time-wavelength chip can
be given orthogonal polarizations to operate, therefore reducing interference. One of the
implementation techniques is the polarization assisted OCDMA with HiBi FBG [54]. The
technique uses the polarization properties of the HiBi FBG along with a special code
generation scheme to improve the performance of OCDMA based networks. The coders are
based on HiBi FBGs. To implement the suggested polarization assisted OCDMA, each HiBi
FBG will reflect a pair of wavelengths
i
j
, which are consecutive and cross polarized. In the
case of the proposed method, a set of three of these HiBi FBGs, spaced by the fiber length
needed for achieving the corresponding time chip spacing, results in two subsequent codes.
Here,
i
corresponds to a X polarized reflection and
j
to a Y polarized one. This allows two
consecutive user spreading sequences to share the same encoder. An implementation example
is depicted in figure 27 .

User A
PBC
Circ
HiBi FBG

24, 25

3,4

11, 12

X
User B
Y
Encoder
HiBi
fiber
X
Y
A
B

Figure 27. Schematic of the proposed encoder implementation showing the use of the polarization to
encode simultaneously two users with different wavelengths at orthogonal polarizations. Legend: PBC:
polarization beam combiner; Circ: optical circulator [54] ( 2006 IEEE).

X
Y
11,12

3,4

HiBi
fiber
Circ
PBS
Decoder
HiBi FBG

24,25

Figure 28. Schematic of the proposed implementation for the decoder based on HiBi FBG. Legend:
PBS: polarization beam splitter; Circ: optical circulator [54] ( 2006 IEEE).
Each bit of information from users A and B is a wavelength comb which includes at least
the wavelengths of the correspondent code (or a standard modulated broadband source can be
used). Both bit sequence signals are multiplexed using a polarization beam combiner, thereby
ensuring that they enter the encoder at the correct orthogonal polarizations. The encoder is
based on three HiBi FBG reflecting the wavelengths
3
,
11
and
24
for X polarization and
4
,
12
and
25
for Y polarization. To achieve networking operation many such encoders need to
operate simultaneously, and due to the properties of the technique, the number of encoders
needed reduces to almost half.
The decoder can be imprinted with standard FBG, since each user has its own code.
However, for reduced user interference and to reduce the number of decoders needed, it can
be based on HiBi FBG like the one exemplified in figure 28.
The decoding process is similar to the encoding, where the HiBi FBG correlates two
codes simultaneously. Afterwards a polarization beam splitter is used to separate both users.
If no polarization maintaining fiber is used in the transmission link between the encoder and
the decoder, the former must be preceded by a polarization rotator to ensure correct
polarization coupling to the receiver. The polarization rotator can be automatically controlled
by the receiver of one of the users using simple electronics. Even if no alignment of the
polarization is made between non adjacent users, on average, only half the power will induce
interference since the decoder will process only one of the two available polarizations. In the
same way, the sensitivity to heterodyne crosstalk is also reduced since the power of the
adjacent user, generated by the same encoder, is orthogonally polarized. In opposition to other
coding/decoding techniques, like the ones based on arrayed waveguide gratings and optical
delay lines, the proposed coder/decoders are quite compact, simple to use and have low
insertion losses. On the other hand, since the gratings can have a length down to 1-2 mm and
still have a high reflectivity, the time slots can be as low as a few picoseconds which can be
considered enough for the majority of applications.

6.3.2. Radio over Fiber
In radio over fiber systems (RoF), using the same central station to transmit to different local
stations, one can use frequency interleaving to improve the bandwidth efficiency, exploiting
the unused band between the carrier and data when high modulation frequencies are used with
single side band (SSB) format. However, frequency interleaving also increases the bit error
rate (BER), due to the interference of adjacent carriers. This drawback can be minimized if
polarization multiplexing is used, i.e., the carriers and data are at orthogonal polarizations
(figure 29).

Interference: same polarization
(40 GHz spacing)
Traditional Implementation
Implementation with HiBi FBG
Data of interest
Interference: both
polarizations
(20 GHz spacing)

o o o o o o o o
X Y X Y X Y X Y
o o o o o o o o
x y x y x y x y

Figure 29. Diagram of the concept of interleaving using polarization multiplexing between carriers and
data.
The implementation of this concept can be made using a HiBi FBG filter at the
transmission which creates the SSB format and, at the same time, selects one polarization for
the carrier and the orthogonal one for the data. At the local station another HiBi FBG removes
the selected channel with reduced interference, since the interference will only be made by
the data of the adjacent channels, which are at higher wavelength spacing and with lower
power, relatively to the adjacent carriers. This technique has an impact on the overall
performance of the system since the bandwidth efficiency can be improved without increasing
the BER [55].
7. Conclusion
This chapter described some of the characteristics and functionalities associated with HiBi
FBG. Their anisotropic behavior, relative to stress and/or strain, make them well suited for
multiparameter sensors, including temperature, transversal strain and longitudinal strain.
Moreover, their polarization processing capabilities also give them an interesting potential for
different applications in optics communications. These applications include the development
of new devices, like multiwavelength fiber lasers or in the optimization of certain
architectures, like OCDMA. Some of the applications for sensing and optical communications
were described but many more are yet to come.
References
[1] Kapron, F. P.; Keck, D. B. Appl. Phys. Lett. 1970, vol.17, 423-425.
[2] Yamada, Y.; Nakagawa, S. I.; Kurosawa, Y.; Kawazawa, T.; Taga, H.; Goto, K. Elect.
Lett. 2002, vol. 38, 328-330.
[3] Hill, K. O.; Fufii Y.; Johnson D. C.; Kawasaki, B. S. Appl. Phys. Lett. 1978, vol. 32,
647-649.
[4] Mora, J.; Dez, A.; Cruz, J. L.; Andrs, M. V. IEEE Photon. Technol. Lett. 2000, vol. 12,
1680-1682.
[5] Guan, B.; Tam, H.; Liu, S. IEEE Photon. Technol. Lett. 2004, vol. 16, 224-226.
[6] Tjin, S. C.; Suresh, R.; Ngo, N. Q. J. Lightwave Technol. 2004, vol. 22, 17281733.
[7] Liu, Y.; Chiang, K. S.; Chu, P. L. IEEE Photon. Technol. Lett. 2005, vol. 17, 450-452.
[8] Kashyap, R. Fiber Bragg gratings, Publisher: Academic Press, San Diego, CA, 1999.
[9] Othonos, A.; Kalli, K. Fiber Bragg gratings, Publisher: Artech House, Norwood, NA,
1999.
[10] Higuera, J. M. L. Handbook of Optical Fiber Sensing Technology; Publisher: John
Wiley & Sons, New York, NY, 2002.
[11] Erdogan, T. J. Lightwave. Technol. 1997, vol. 15, 1277-1294.
[12] Lam, D. K. W.; Garside, B. K. Appl. Optics. 1981, vol. 20, 440-456.
[13] Yariv, A. J. Quantum Electron. 1973, vol. 9, 919-933.
[14] Yariv, A.; Nakamura, M. J. Quantum Electron. 1977, vol. 13, 233-253.
[15] Dyott, R. B. Elliptical Fiber Waveguides; Publisher: Artech House, Boston, MA, 1995.
[16] Rashleigh, S. C. J. Lightwave. Technol. 1983, vol. LT-1, 312-330.
[17] Kersey, A. D.; Davis, M. A.; Patrick, H. J.; LeBlanc, M.; Koo, K. P.; Askins, C. G.;
Putnam, M. A.; Friebele, E. J. J. Lightwave Technol. 1997, vol. 15, 1442-1463.
[18] Othonos, A. Rev. Sci. Instrum. 1997., vol. 68, 4309-4341,
[19] Abe, I.; Kalinowski, H. J.; Nogueira, R. N.; Pinto, J. L.; Frazo, O. IEE P-Circ. Dev.
Syst. 2003, vol. 12, 495-500.
[20] Frazo, O.; Pereira, D. A.; Santos, J. L.; Arajo, F. M.; Ferreira, L. A. Proc. SPIE 2005,
vol.5855, 765-758.
[21] Chen, G; Liu, L; Jia, H; Yu, J; Xu, L.; Wang, W. IEEE Photon. Technol. Lett. 2004, vol.
16, 221-223.
[22] Urbanczyk, W.; Chmielewska, E.; Bock, W. Meas. Sci. Technol. 2001, vol. 12, 800-804.
[23] Ferreira, L. A.; Arajo, F. M.; Santos, J. L.; Farahi, F. Opt. Eng. 2000, vol. 39, 2226-
2234.
[24] Sudo, M.; Nakai, M.; Himeno, K.; Suzaki, S.; Wada, A.; Yamauchi, R. Proc. 12 th Int.
Conf. Optical Fiber Sensors 1997, vol. 16, 170-173.
[25] Chehura, E.; Ye, C. C.; Staines, E. S.; James, S. W.; Tatam, R. P. Smart Mater. Struct.
2004, vol.13, 888-895.
[26] Ye, C.C.; Staines, S. E.; James, S. W.; Tatam, R. P. Meas. Sci. technol. 2002, vol. 13,
1446-1449.
[27] Bosia, F.; Giaccari, P.; Facchini, M.; Botsis, J.; Limberger, H.; Salath, R. Proc. SPIE
2002, vol. 4694, 175-186.
[28] Lawrence, C. M.; Nelson, D. V.; Udd, E.; Bennett, T. Exp. Mech. 1999, vol. 39,202-209.
[29] Lawrence, C. M.; Nelson, D. V.; Udd, E. Proc. SPIE 1997, vol. 3042, 218-228.
[30] Caucheteur, C.; Ottevaere, H.; Nasilowski, T.; Chah, K.; Statkiewicz, G.; Urbanczyk,
W.; Berghmans, F.; Thienpont, H.; Mgret, P. Proc. SPIE 2005, vol. 5952, 1-10.
[31] Abe, I.; Kalinowski, H. J.; Frazo, O.; Santos, J. L.; Nogueira, R. N.; Pinto, J. L. Meas.
Sci. Technol. 2004, vol. 15, 1453-1457.
[32] Udd, E.; Schulz, W. L.; Seim, J. Proc. SPIE 1999, vol. 3538, 206-214.
[33] Udd, E.; Nelson, D., Lawrence, C.; Ferguson, B. Proc. SPIE 1996, vol. 2718,104-107.
[34] Abe I.; Frazo O.; Schiller M. W.; Nogueira R. N.; Kalinowski H. J.; Pinto, J. L. Meas.
Sci. Technol. 2006, vol.17, 14771484.
[35] Nogueira, R. N.; Teixeira, A. L. J.; Andr, P. S.; Rocha, J. F.; Pinto, J. L. Proc. Conf.
Lasers Electro-Optics 2003, 545.
[36] Andr, P. S.; Nogueira, R. N.; Teixeira, A. L.; Lima, M. J. N.; Rocha, R. F.; Pinto, J. L.
Laser Physics Lett. 2004, vol. 1, 1 - 4.
[37] Zhao, C. L.; Yang, X.; Ng, J. H.; Dong, X.; Guo, X.; Wang, X.; Zhou, X.; Lu, C.
Microw. Opt. Technol. Lett. 2004, vol. 41, 73 75.
[38] Nogueira, R. N.; Teixeira, A. L. J.; Andr, P. S; Rocha, J. F.; Pinto, J. L. Opt. Commun.
2006, vol. 262/1, 38-40.
[39] Nogueira, R. N.; Teixeira, A. L. J.; Pinto, J. L.; Rocha, R. F. IEE Elect. Lett. 2004, vol.
40, 616617.
[40] Lee, S.; Khosravant, R.; Peng, J.; Grubsky, V.; Starodubov, D. S.; Willner, A. E.;
Feinberg, J., IEEE Photon. Technol. Lett. 1999, vol. 11, 12771279.
[41] Willner, A. E.; Feng, K.-M.; Cai, J.; Lee, S.; Peng, J.; Sun, H. IEEE J. Select. Topics
Quantum Electron. 1999, vol. 5, 1298-1311.
[42] Yi, X.; Lu, C.; Yang, X.; Zhong, W.-D.; Wei, F.; Ding, L.; Wang, Y. IEEE Photon.
Technol. Lett. 2003, vol. 15, 754-756.
[43] Zhang, W.; Williams, J. A. R.; Bennion, I. IEEE Photon. Technol. Lett. 2001, vol. 13,
523-525.
[44] Eggleton, B. J.; Mikkelsen, B.; Raybon, G.; Ahuja, Rogers, A.; J. A.; Westbrook, P. S.;
Nielsen, T. N.; Stulz, S.; Dreyer, K. IEEE Photon. Technol. Lett. 2000, vol. 12, 1022-
1024.
[45] Mora, J.; Ortega, B.; Andrs, M. V.; Capmany, J.; Cruz, J. L.; Pastor, D.; Sales, S. IEEE
Photon. Technol. Lett. 2003, vol. 15, 951-953.
[46] Lauzon, J.; Thibault, S.; Martin J.; Ouellettet, F. Opt. Lett. 1994, vol. 19, 2027-2029.
[47] Nogueira, R. N.; Pinto, J. L.; Rocha, J. F. Microw. Opt. Tech. Lett. 2006, vol. 48 , 2357-
2359.
[48] Gloag, A. J.; Langford, N.; Bennion I.; Zhang, L. Opt. Commun. 1996, vol. 123,
553-557.
[49] Inaba, H.; Akimoto, Y.; Tamura, K.; Yoshida, E.; Komukai T.; Nakazawa, M. Opt.
Commun. 2000, vol. 180, 121125.
[50] Yamashita S.; Hotate, K. Electron. Lett. 1996, vol. 32, 1298-1299.
[51] Wei, D.; Li, T.; Zhao Y.; Jian, S. Opt. Lett. 2000, vol. 25, 1150-1152.
[52] Das G.; Lit, J. W. Y. IEEE Photon. Technol. Lett. 2002, vol. 14, 606-608.
[53] Graydon, O.; Loh, W. H.; Laming, R. I.; Dong, L. IEEE Photon. Technol. Lett. 1996,
vol. 8, 63-65.
[54] Nogueira, R. N.; Teixeira, A. L. J.; Pinto, J. L.; Rocha, J. F. IEEE Photon. Technol. Lett.
2006, vol. 18, 841 843.
[55] Teixeira, A. T.; Nogueira, R. N.; Andr, P. S.; Lima, M. J. N.; Rocha, J. F., Electron.
Lett. 2005, vol. 41, 30 - 32.

In: Optical Fibers Research Advances
Editor: Jurgen C. Schlesinger, pp. 119-159
ISBN 1-60021-866-0
c 2007 Nova Science Publishers, Inc.
Chapter 4
APPLICATIONS OF HOLLOW OPTICAL FIBERS
IN ATOM OPTICS
Heung-Ryoul Noh
1
and Wonho Jhe
2
1
Department of Physics and Institute of Opto-Electronic
Science and Technology, Chonnam National University,
Gwangju 500-757, Korea
2
School of Physics and Astronomy,
Seoul National University, Seoul 151-742, Korea
Abstract
A hollow optical ber (HOF) has a lot of interesting applications in atom optics
experiments such as atom guiding and the generation of hollow laser beam (HLB).
In this article we present theoretical and experimental works on the use of hollow
optical bers in atom optics. This article is divided into two parts: One is devoted
to the atom guide using HOFs and the other describes the atom optics researches that
utilizes laser lights emanated from the HOF. Firstly, we describe the electromagnetic
elds inside the HOF and characterize the electromagnetic modes diffracted from the
HOF. Then we describe two guiding schemes using red and blue detuned laser lights.
Finally, we describe the various relevant experiments using LP
01
or LP
11
modes such
as the generation of HLB from the HOF, funneling atoms using the diffracted elds,
diffraction-limited dark laser spot, and a dipole trap using LP
01
mode of the diffracted
eld from the HOF.
PACS: 32.80.Pj, 42.50.Vk, 39.25.+k, 32.80.-t, 03.75.Be
1. Introduction
For the last two decades, there has been much progress on atom optics that manipulates
atoms by using laser lights [1, 2, 3, 4, 5, 6, 7, 8, 9]. This eld includes the studies such
as focusing, reecting, diffracting, and guiding atoms. In particular, it is an atom guide
that provides a high spatial resolution of atom manipulation. So far several types of atom
E-mail address: whjhe@snu.ac.kr. (Corresponding author)

120 Heung-Ryoul Noh and Wonho Jhe
guides using the optical, magnetic means, and using hollow optical bers (HOFs) have been
demonstrated. Of these, guidance of atoms by using HOFs is widely investigated, since it
enables atoms to be controlled very accurately and guided over a long distance at high
spatial accuracy without much atomic loss. A hollow optical ber has many interesting
applications in sensor [10] and harmonic generation [11], and optical communications [12,
13, 14, 15]. In this article, we describe the application of the HOF in atom optics.
The basic principle of an optical atom guide is that atoms in laser beams are either
attracted to or repelled from the regions of high intensity laser light, depending on the sign
of the laser frequency detuning with respect to the atomic resonance. For two level atoms,
the optical dipole potential due to the guiding laser is described by [16]
U(r) =

2
ln
1 +
I(r)/I
s
1 + 4
2
/
2
, (1)
where =
L

0
is the laser detuning from the atomic resonance, I
s
is the saturation
intensity, I(r) is the guiding laser intensity, and is the spontaneous decay rate of the
upper level. When the laser frequency is tuned slightly below the atomic resonance, or
red-detuned, atoms are attracted to the high-intensity regions. One the other hand, atoms
are repelled from the high-intensity regions when the light is tuned above the resonance, or
blue-detuned.
One way to guide atoms is to launch the laser light inside the hollowregion of the hollow
ber and tune the laser frequency to the red side of the atomic resonance as proposed by
Olshanni et al. [17]. This guiding scheme was successfully realized by Renn et al. at JILA
[18]. In that study, a red-detuned laser was coupled to the lowest-order grazing incidence
mode inside the glass capillary. The second method is to introduce the laser light into the
glass core of the hollow ber and tune the laser frequency to the blue side of the atomic
resonance. This method was rst proposed by Savage et al. [19, 20] and later by Jhe et
al. [21] for different waveguide congurations. This guiding scheme was demonstrated for
the capillary ber by Renn et al. [22] and then subsequently demonstrated for the micron-
sized hollow optical ber by Ito et al. [23]. A similar experiment was later performed by
Workurka et al. [24] and also by Dall et al. [25, 26].
In addition to the atom guidance, the HOF has another application in atom optics, which
is the generation of laser beams diffracted from the facet of the HOF. The rst use of the
HOF for the generation of a hollow laser beam (HLB) by imaging the eld distribution of
LP
01
mode was demonstrated by Yin et al. [27]. Such hollow laser beams have been used
for atom guidance [28], atom fountains [29], and atom traps [30]. The characterization
of the output-eld distribution of the hollow optical ber was described in detail [31, 32].
The hollow laser beams made by a combination of two orthogonal LP
11
modes without an
imaging lens can be used for funneling and guiding atoms [32]. They can also be used for
generation of the diffraction-limited dark laser spot [33]. Furthermore the diffracted output
of the LP
01
mode has a bright focused spot, which can be used for a tight optical dipole trap
when a red-detuned laser is used [34].
This article is organized as follows: In the next section, we characterize the electromag-
netic elds inside the HOF and the eld distributions diffracted from it. In Secs. 3. and 4.,
the atom guidance by using the red and blue-detuned laser lights are presented, respectively.
The experimental works with diffracted LP
11
modes from the HOF are described in Sec.
Applications of Hollow Optical Fibers in Atom Optics 121
5., and the discussions on the applications of diffracted LP
10
modes such as atom guiding,
atom fountain, crossed HLB trap, and single optical dipole trap follow in Sec. 6. The nal
section presents the summary of the work.
2. Characteristics of Electromagnetic Field for a Hollow Optical
Fiber
2.1. Electromagnetic Field Modes Inside the Hollow Optical Fibers
We describe the electromagnetic eld modes for both inside and outside the hollow optical
bers. The electromagnetic elds inside HOF are given in the rst subsection, and then the
discussion on the electromagnetic elds diffracted fromHOF follows in the next subsection.
The schematic diagram of the HOF with the hollow diameter of 2a and the core thickness
d b a is shown in Fig. 1 [20, 35]. Since the difference of the refractive indices of
core and hollow region is not small, the weakly-guiding approximation seems to be not
applicable. However, this approximation proved to be well applicable for the HOF [36, 37].
Therefore, instead of discussing the cumbersome vectorial approach, we will use the scalar
theory for the analysis of electromagnetic elds modes. A capillary ber composed of a
hole of radius a and outer glass part is also used in the guiding experiment. The discussions
on the guiding modes for the capillary ber with red-detuned laser beam will be given in
Sec. 4.1..
Figure 1. The schematic diagram of the hollow optical ber. The diameter of the hollow
region and the thickness of the cylindrical core are 2a and d, respectively. The refractive in-
dices of the hollow, the core, and the cladding are 1, n
1
, and n
2
, respectively. The thickness
of cladding can be taken to be innite.
In the cylindrical coordinate (r, , z), the longitudinal component of the electric eld
E
z
(r, ) with E
z
(r, , z, t) = E
z
(r, ) exp[i(tz)] satises the Maxwell equation given
by
2
E
z
r
2
+
1
r
E
z
r
+
1
r
2
2
E
z
2
+
k
2
n
2
E
z
= 0 , (2)
where , , k and n are the angular frequency, the propagation constant, the wave number,
and the refractive index, respectively. The solution of Eq. (2) is given by
E
z
(r, ) =
AI
(vr) sin( +) (r < a),

(BJ
(ur) +CN
) sin( +) (a r b),
DK
(wr) sin( +) (r > b),

(3)
where J
and N
(I
and K
) are the (modied) Bessel functions of the rst and the sec-
ond kind of order , respectively, n
1
(n
2
) is the refractive index of the core (cladding),
and is a phase constant. The constants A, B, C, and D can be determined by the con-
tinuity conditions of E
z
at r = a and b. The quantity u =
k
2
n
2
1
2
is the transverse
propagation constant, whereas v =
2
k
2
and w =
2
k
2
n
2
2
are the transverse
attenuation constants. The magnetic longitudinal components, H
z
, can be obtained by re-
placing sin( + ) in Eq. (3) with cos( + ). The transverse components (E
r
, E
,
H
r
, and H
) can be derived from these longitudinal components [38, 39]. The solutions of
Eq. (3) are called HE
(EH
) modes, when E
z
and H
z
have the same (different) signs
[40]. The dispersion equations describing the light propagation modes of the HOF can then
be derived from the secular equation obtained by applying the boundary conditions to the
tangential components E
z
, E
, H
z
, and H
[39]. The explicit expressions of the dispersion

relation are presented in Eqs. (5)-(7) of Ref. [35].
Let us assume the silica-glass HOF with 2a = 7m, d = 3.8m, and n
2
= 1.45,
in which the core is germanium doped with the relative refractive index difference n =
n
2
1
n
2
2
/2n
2
1
= 0.0018. We also consider the guiding of rubidium atoms with the wave-
length = 780 nm of D
2
line. From the dispersion equation, one can nd that six propa-
gation modes, TE
01
, TM
01
, HE
11
, HE
21
, HE
31
, and EH
11
can be excited at the wavelength
of 780 nm (Fig. 2). The lowest mode is the HE
11
mode, whereas the TE
01
, TM
01
, and
HE
21
modes exhibit almost the same dispersion curves, so that they form the second group
of propagation modes. In the same way, EH
11
and HE
11
modes consist of the third group.
When the refractive indices of the core and the cladding are nearly the same in a step-
index solid ber, the weakly guiding approximation is generally used [39, 41]. In this
case, since one of the tangential eld components is far larger than the other orthogonal
transverse or longitudinal components, the guided mode can be approximately described
only by the dominant transverse component, for which the linearly polarized LP
lm
-mode
description is usually employed [38, 39, 41] where l is the azimuthal mode number and m
is the radial mode number. In case of an HOF, despite the large difference of the refractive
index between the core and the hollow region, one can still use the LP
lm
modes due to
the relatively small intensity in the hollow region. The transverse component for an LP
lm
mode is given in the same form as the longitudinal one [Eq. (3)] and thus we use it as a
guided mode in the following calculations due to its simplicity over the traditional mode
description method.
From the continuity conditions at the boundaries r = a and r = a + d, one can derive
the following simple dispersion equation describing the LP modes as [42]
J
m
(ua)
I
m
(va)

u
v
J
m
(ua)
I
m
(va)

N
m
(ub)
K
m
(wb)

u
w
N
m
(ub)
K
m
(wb)
N
m
(ua)
I
m
(va)

u
v
N
m
(ua)
I
m
(va)

J
m
(ub)
K
m
(wb)

u
w
J
m
(ub)
K
m
(wb)
. (4)
The dispersion equation in Eq. (4) yields the LP
m1
modes (m = 0, 1, 2, ) for the
hollow bers considered here. In fact, from the numerical analysis of Eq. (4), one observes
that three LP modes can be excited in the 7-m hollow core at the wavelength of 780
nm. Figure 2 shows the dispersion curves with respect to several lower modes. The solid
circles indicate three LP modes: LP
01
, LP
11
, and LP
21
mode. Comparing with the the exact
numerical results described above, one can nd that: (i) LP
01
mode is approximately equal
to HE
11
mode, (ii) LP
11
mode is made up of TE
01
, TM
01
, and HE
21
modes, and (iii) LP
21
mode consists of EH
11
and HE
31
modes. It should be noted that in Fig. 2(a), the points
where each dispersion curve intersects the horizontal axis represent the cutoff frequencies.
According to Eq. (4), the 7- and 2-m hollow-core bers become multi-moded at 780 nm,
whereas the 1.4- and 0.3-m hollow bers become single-moded.
Figure 2. (a) Dispersion curves of the propagating modes at the wavelength of 780 nm in
the 7-m hollow optical ber with the core thickness of 3.8 m and the relative refractive
index difference of 0.18%. The cross-sectional intensity proles for (b) LP
01
mode, (c)
LP
11
mode, and (d) LP
21
mode (Figure from Ref. [35]).
Figures 2(b-d) showthe cross-sectional mode-patterns of the 7-mhollowber; (b), (c),
and (d) present the LP
01
, LP
11
, and LP
21
modes, respectively. Note that one can selectively
excite one of these LP modes by careful alignment of the incident angle of the laser beam.
These CCD camera images show that the LP
01
mode is suitable for guiding atoms: LP
01
mode has no nodes around the cylindrical inner wall of the hollow core, as shown in Fig.
2(b).
0 2 4 6 8 10 12 14
0.00
0.05
0.10
0.15
0.20
0.25
Cladding Core Hollow
I
n
t
e
n
s
i
t
y
(
a
r
b
.
u
n
i
t
)
r (mm)
LP
01
LP
11
Figure 3. The radial intensity distributions of LP
01
(solid curve) and LP
11
(dotted curve)
modes inside the HOF.
The intensity distributions of LP
01
and LP
11
modes along the radial direction are shown
in Fig. 3 as solid and dotted curves, respectively. In Fig. 3, the both the diameter of the hole
and the core thickness are typically assumed to be 4.5 m. While the intensity distribution
of LP
01
mode is azimuthally symmetric as shown in Fig. 2(b), that of LP
11
mode has
an angular dependence sin
2
as shown in Fig. 2(c). Therefore the real two-dimensional
intensity distribution of LP
11
mode should be obtained by multiplying the angular factor
sin
2
to the function in Fig. 3. The guided LP
01
mode produces the blue-detuned optical
evanescent elds on the core-vacuum interface, which then generates the optical potential
barrier so that atoms can be guided in the dark hollow region of the core.
2.2. Characterization of Diffracted Fields from a Hollow Optical Fiber
In the previous subsection, we have described the electromagnetic eld modes inside the
hollow optical bers. The various applications using the inside modes will be described in
Secs. 3. and 4. In this subsection, we will provide a theory of the diffracted laser lights from
an HOF. As was mentioned in the previous section, we will adopt a simple scalar approach
for the calculation.
Since we know the electric eld, Eq. (3), on the facet of the HOF (z = 0), we can
calculate the diffraction pattern at (x, y, z) using the Huygens-Fresnel integral [43]
E(x, y, z) =
z
2

E
0
(x
0
, y
0
)
+ ik
e
ik
2
dx
0
dy
0
, (5)
where (x
0
, y
0
) is the coordinate of a source point, =

z
2
+ (x x
0
)
2
+ (y y
0
)
2
1/2
,
and E
0
(x
0
, y
0
) is the electric eld at the ber facet. In the near-eld region where z
(kb
2
/2) 200 m, using the Rayleigh-Sommerfeld theory [44, 45], one can calculate the
diffraction pattern without any approximation on as in the Fresnel diffraction calculation.
For a given LP
lm
mode represented by E
0
lm
at z = 0, we then obtain
E
lm
(x, y, z) = E
0
lm
(x
0
, y
0
) h(x, y, z) (6)
where
h(x, y, z) =
_
exp
_
ikz
_
1 + (x/z)
2
+ (y/z)
2
__
_
iz
_
1 + (x/z)
2
+ (y/z)
2
__
1
(7)
and denotes a two-dimensional convolution integral.
Eq. (6) becomes a normal product in the transformed space,
U
lm
(, ; z) = U
0
lm
(, ) H(, ; z) (8)
where U
lm
(U
0
lm
) is the Fourier transformation of E
lm
(E
0
lm
), and
H(, ; z) = exp
_
ikz
_
1
2
(
2
+
2
)
_
. (9)
The main task is to calculate the diffracted eld in the transformed space and then
convert the results to real space by an inverse transformaton. In the cylindrical coordinates
(r,) and (,) in each space, the source eld in the transformed space is given as:
U
0
lm
(, ) = 2i
l
sin(l +
0
)
_
+
0
r
0
E
0
lm
(r
0
)J
l
(2r
0
)dr
0
U
0
lm
() sin(l +
0
), (10)
where E
0
lm
(r
0
,
0
) E
0
lm
(r
0
) sin(l
0
+
0
). Substituting U
0
lm
into Eq. (8), we inverse-
transform U
lm
to obtain
E
lm
(r, , z) = 2(i)
l
sin (l +
0
)
_

0
U
0
lm
() exp[ikz
_
1
2
2
]J
l
(2r)d. (11)
Here, U
0
lm
is given analytically by
U
0
lm
() = 2i
l
[+
aA
_
J
1+l
(a
)I
l
(av) + vI
1+l
(av)J
l
(a
)
_
v
2
+

aB
_
J
1+l
(a
)J
l
(au) uJ
1+l
(au)J
l
(a
)
_
u
2
2
+
bB
_
J
1+l
(b
)J
l
(bu) uJ
1+l
(bu)J
l
(b
)
_
u
2
aC
_
J
1+l
(a
)N
l
(au) uN
1+l
(au)J
l
(a
)
_
u
2
2
+
bC
J
1+l
(b
)N
l
(bu) uN
1+l
(bu)J
l
(b
u
2
bD
J
1+l
(b
)K
l
(bw) wK
1+l
(bw)J
l
(b
w
2
+

2
] (12)
where

2. Using these results, one can calculate the general prole of the diffracted
beam at any position z. The numerical results of the radial intensity distributions for small
zs are presented in steps of 5 m in Fig. 4. Figure 4(a) explains how the LP
01
mode
diffracts in free space near the HOF: the two peaks (which represents a cross-section of
ring-shaped mode) at z = 0 diminish away while an additional central peak grows up. In
Fig. 4(b), one can see that the peaks of LP
11
also diminish whereas another pair of peaks
grow. Nevertheless, there still does exist a dark column along the central axis.
-10 -5 0 5 10
LP01
distance(um)
0 10 r(mm)
0
70
z
(mm)
(a) LP
-10 -5 0 5 10
LP01
distance(um)
-10 -5 0 5 10
LP01
distance(um)
0 10 r(mm)
0
70
z
(mm)
(a) LP
-10 -5 0 5 10
LP11
distance(um)
0 10 r(mm)
0
70
z
(mm)
(b)LP
-10 -5 0 5 10
LP11
distance(um)
-10 -5 0 5 10
LP11
distance(um)
0 10 r(mm)
0
70
z
(mm)
(b)LP
Figure 4. Development of the radial intensity distributions due to the diffraction of (a) LP
01
and (b) LP
11
mode near the facet of HOF (z = 0 at the facet).
We now describe the experimental verication of theoretical calculated results for the
developments of LP
10
and LP
11
modes using a simple imaging technique. In wave optics,
the light propagating through a lens from a source plane to a screen can be described by
equation (4.3-18) of reference [38], which shows that the eld distribution at the screen is
identical to the original distribution within a magnication constant m d
o
/d
i
when
(d
i
1
+ d
o
1
f
1
) = 0, where d
i
is the distance between the lens and the source eld,
d
o
, the distance between the lens and the screen, and f, the focal length of the lens. This
condition is easily realized in experiments when d
i
= f d
o
. Therefore, one can observe
the intensity distribution of z = 0 at the screen far enough from the lens by locating the lens
at a distance of the focal length from the tip of the HOF. In addition, one can observe the
intensity distribution (diffraction pattern) of an arbitrary z by moving the lens by z toward
the screen.
The experimental results are shown in Fig. 5. We measured the intensity distribution
of z = 0 at the screen 100 cm away from the lens with a focal length of 4.3 mm, and
repeated the measurement at several values of z by moving the lens. As shown in Fig.
5, the experimental results show good agreement with the theoretical results. The output
intensity distribution of LP
01
mode at z = 100 m has a maximum at the center with the
10 m m
(a)LP (z=0 m)
01
m (b)LP (z=150 m)
01
m
(c)LP (z=0 m)
11
m (d)LP (z=150 m)
11
m
-20 -10 0 10 20
LP
01
(z=150 mm)
I
n
t
e
n
s
i
t
y
(
a
r
b
.
u
n
i
t
)
Radial Distance (mm)
(e)
Figure 5. Output intensity distributions of the LP
01
mode ((a) at z = 0; (b) at z = 150 m)
and the LP
11
mode ((c) at z = 0; (d) at z = 100 m), and (e) beam prole of the output of
the LP
01
mode at z = 150 m in a radial direction. The solid curve in the graph represents
the tting curve obtained from the calculation.
full width at half maximum(FWHM) of 7.6 m although it is in the shape of a ring inside
the HOF. In case of LP
11
mode, on the other hand, both the dark center and the angular
node line are preserved continuously even after propagation into free space.
3. Atom Guidance by Hollow Optical Fiber with Red-Detuned
Laser
In this section, we describe the atom guidance experiment by a hollow optical ber with the
red-detuned Gaussian laser lights. This scheme was suggested by OlShanii et al. [17], and
experimentally realized by Renn et al. for the hollow glass capillary using the red-detuned
Gaussian laser beam [18, 46]. We rst describe the experiment performed by Renn et al.,
and then the parametric excitation experiment will be briey discussed [47].
Let us briey summarize the theory of atom guidance by the red-detuned laser in the
hollow capillary ber which is composed of a hole of radius a and a glass with the refractive
index of n. The thickness of the surrounding glass is assumed to innite compared to radius
of the hole. In the cylindrical coordinate (r, , z), the electric eld of the lowest guided
EH
11
mode is given by
E(r, z, t) = eE
0
(r)e
i(tz)
, (13)
where e is a unit transverse vector, and E
0
(r) is given by
E
0
(r) =
AK
0
(a)J
0
(r) (r < a),
AJ
0
(a)K
0
(r) (r > a),
(14)
where J
(K
) is the rst (second) kind of the (modied) Bessel functions of order , A is

a normalization constant,
2
= k
2
2
,
2
=
2
n
2
k
2
, and k = 2/ .
The value can be determined by the following characteristic equation,
J
0
(a)K
1
(a) + J
1
(a)K
0
(a) = 0 , (15)
When ka 1, Eq. (15) can be simplied as
J
0
(a) =
i
2k
n
2
+ 1
n
2
1
J
1
(a) . (16)
a is 2.405 + 0.022i for a 40-m hollow-core diameter capillary at = 780 nm. The
imaginary part of is the attenuation coefcient of the mode amplitude and is given by
Im() =
a
2
2
2a
3
n
2
+ 1
n
2
1
, (17)
for ka 1. For the 40-m-diameter capillary ber, the attenuation length, [Im()]
1
, was
6.2 cm. Figure 6 shows the intensity prole of EH
11
mode. As can be seen from Fig. 6 and
Eq. (14), the intensity prole of the laser beam undergone by the atoms inside the hollow
region is approximately given by I(r) = I
0
J
2
0
(r), where I
0
is the peak laser intensity.
When the guiding laser is red-detuned to the atomic resonance frequency, the atoms can
be guided along the capillary hole owing to the attractive dipole force exerted by the laser
light.
-40 -30 -20 -10 0 10 20 30 40
0.0
0.2
0.4
0.6
0.8
1.0
Glass Glass Hole
I
n
t
e
n
s
i
t
y
(
a
r
b
.
u
n
i
t
)
r (mm)
Figure 6. Radial intensity prole of EH
11
mode inside the capillary ber with the inner
diameter of 40 m.
The experiment of Renn et al. [18] consisted of two separate vacuum chambers con-
nected by a 3.1-cm-long capillary ber with an outer diameter of 144 m and capillary-core
diameter of 40 m (Fig. 7). The guiding light from a Ti-Sapphire laser was coupled into
the EH
11
mode. They observed a 50% coupling efciency and an attenuation loss of 7%
per cm of ber length for the EH
11
mode. The guided atoms leaving the capillary ber are
surface ionized on a heated Pt or Re wire in the second chamber. Then the resulting ions
are detected with a channeltron electron multiplier and recorded with a pulse counter.
Figure 7. Experimental apparatus.
Figure 8. Guided atom signal versus laser detuning. The intensities of the guiding laser are
I
0
= 3.6 10
3
W/cm
2
(upper curve) and I
0
= 1.3 10
3
W/cm
2
(lower curve) (Figure
from Ref. [18]).
Figure 8 shows the detuning dependence of the number of guided atoms for two laser
intensities. The zero point of the detuning () was taken to be the average transition fre-
quency of the 5S
1/2
5P
3/2
multiplet of
85
Rb and
87
Rb. The guiding signal rose to half
maximum at 2 GHz and full maximum at 3 GHz. For the laser detunings
larger than a few GHz, the guided signal decreased approximately as
1
. With increasing
intensity the signal increased and the position of maximum ux shifts to larger negative
detunings.
Figure 9. Guided atomux vs laser detuning fromresonance at several laser intensities. The
ber lengths are 3.1 and 6.2 cm for (a) and (b)(d), respectively, and the inner diameter is
40 m (Figure from Ref. [46]).
In a subsequent paper, Renn et al. reported on the improved results for the dependence
of the guided atom signals on the laser detuning for several laser intensities [46]. Figure 9
shows the typical dependence of guided atom ux on the laser detunings. At an intensity
of 0.6 MW/m
2
[Fig. 9(a)], the curve follows the similar trend as those shown in Fig. 8:
The ux reaches a maximum near a detuning of about -1 GHz and then falls off rapidly for
larger detunings. As the intensity increased, as shown in Figs. 9(b)9(d), they observed
a substantial ux at signicantly higher detunings and found a dip formed in the prole
at intermediate detunings. As a function of increasing intensity the dip grows deeper and
broader [Figs. 9(c)9(d)]. They found that the formation of the hole was attributed to
viscous dipole forces which heat the atoms to larger transverse energies than can be guided
[46].
Recently Hayashi et al. reported on the parametric excitation of laser guided Cs atoms
in an HOF [47]. Periodic modulation of the optical potential achieved by the intensity mod-
ulation of the red-detuned guiding laser resonantly enhances the amplitude of the atomic
motion. Eventually the atoms hit the inner wall of the HOF and absorbed on the surface.
Thus the parametric resonance causes a reduction in the number of atoms coming out of the
HOF.
The experimental apparatus consists of two separate vacuum chambers connected by a
3-cm long borosilicate glass capillary having an outer diameter of 2 mm and a hollow-core
diameter of 86 m. One chamber serves as a Cs oven, which contains a Cs ampoule and
is heated to 310 K to produce Cs vapor at 1023 Pa. The other chamber is a high vacuum
detection chamber having a background pressure of 1026 Pa. Guided atoms leaving the
capillary outlet are surface-ionized on a heated Pt wire. The generated ions are detected by
a Channeltron electron multiplier and counted by a pulse counter. A Ti:sapphire laser with
a wavelength of 852 nm was used to guide Cs atoms. For parametric excitation, the laser
intensity is modulated by an acousto-optical modulator (AOM).
Figure 10. Dependence of the count rate of Cs atoms on the modulation frequency of the
laser intensity (Figure from Ref. [47]).
Figure 10 shows the count rate of Cs atoms as a function of the modulation frequency
of the laser intensity. The modulation depth is set at 0%, 10%, 40%, and 60%. The guiding
laser is -15 GHz red-detuned from D
2
resonance line. As the modulation depth is increased,
the guided atom ux lowers appreciably. For 10% modulation, a broad resonant structure
appears in the parametric excitation spectrum. As the modulation frequency is increased,
the number of guided Cs atoms starts to decrease around 100 Hz, takes its minimum at 5
kHz, and then grows for higher frequencies. Above 100 kHz, the ux recovers to its original
level. For 40% modulation, parametric excitation occurs over a much wider frequency
range. Still, the dip of the spectrum is centered around 5 kHz. When the modulation depth
is increased to 60%, severe guiding loss takes place over the whole frequency range, and
the resonant structure disappears.
4. Atom Guidance by Hollow Optical Fiber with Blue-Detuned
Evanescent Light
4.1. Atomic Guide by Glass Capillary Fiber
In this section, the atom guidance experiments by the blue-detuned laser beams through
capillary bers or micron-sized hollow optical bers are presented. First, in this subsection,
we will describe the guiding experiments by means of hollow capillary bers, performed
by two groups Renn et al. [22] and Baldwin et al. [25] for the Rb and metastable He atoms,
respectively.
Renn et al. [22] at JILA reported the rst atom guidance with the blue-detuned evanes-
cent waves. They used a hollow glass capillary whose hollow-core diameter (2a) was 10
m and outer diameter was 77 m. In the absence of the laser eld, atoms are supposed to
be attracted to the glass surface due to the long-range van der Walls forces. At low inten-
sities, the optical repulsive potential is weaker than the van der Waals potential, which is
given by [48, 49]
U
vdW
=
3
4f
n
2
1
n
2
+ 1
(kr)
3
, (18)
where f is the oscillator strength, k is the wave number, n is the refractive index, so that the
total potential is always attractive. Therefore, one can dene the threshold intensity above
which atoms feel the net repulsive potential, which is the position of 3/(2k) apart from the
wall.
The experimental setup consisted of two vacuum chambers that were connected by the
hollow capillary ber like the red-detuned case. A blue-detuned high-power laser with 500
mW was focused on the annular region of the ber end and a weaker, red-detuned escort
laser with 10 mW was focused in the hollow region where it was coupled to the EH
11
grazing incidence mode. The escort laser facilitates atom loading into the capillary guide,
which might be hampered by the scattered light near the core facet. Figure 11 shows the
enhancement of the atomic ux due to optical guidance by a 6-cm-long glass capillary with
the 20-m core diameter. With the red-detuned escort laser alone, 200 atoms/s were guided
in the ber. When the evanescent light was excited on the glass interface, on the other hand,
the ux increased at least by a factor of 3 at the optimum guide-laser detuning of +3 GHz
and for the escort laser detuning of -1.6 GHz. When the escort laser detuning was decreased
to -9.4 GHz, the number of atoms injected into the evanescent guide was also reduced. On
the other hand, tuning the escort laser to the blue side of the resonance inhibited injection
of atoms into the guide and thus completely suppressed the guided atom ux.
They also measured the intensity dependence of the atomic guidance signal. Unlike the
red-detuned guidance discussed in the previous subsection, there seems to exist threshold
Figure 11. The dependence of the guided atomic ux on the guide-laser detuning in the
presence of the red-detuned escort laser. The detuning of the escort laser is -1.6 GHz (-9.4
GHz) for the upper (lower) curve (Figure from Ref. [22]).
behaviour near the intensity of about 6 MW/m
2
, and the ux increases roughly linearly
above this value. However, it was not straightforward to account for this behaviour in terms
of the cavity potential because the hollow diameter was rather large, unlike the micron-sized
HOFs where the van der Waals interaction near the core wall becomes signicant (refer to
next section for quantitative discussions of the cavity quantum electrodynamic effects).
The JILA group also demonstrated evanescent-light guidance of laser cooled
87
Rb
atoms in the hollow-core bers [50] using the similar apparatus. The ux of an atomic
beam generated from the source chamber by the low velocity intense source (LVIS) method
[51] was approximately 10
9
atoms/s and the brightness was 10
13
atoms/(srs). The
transverse velocity of LVIS was in the range v
t
= 8.01.5 cm/s and its longitudinal veloc-
ity was measured to be v
r
= 10.0 2.0 m/s. The bers used in their experiment were glass
capillary tubes with 100-m inner and 160-m outer diameter. They have guided atoms
by using several bers of different lengths varying from 17 to 30.5 cm. In the 30.5-cm-
long ber, in particular, the atomic ux of 7 10
4
atoms/s was measured and 5.9 10
5
atoms/s was the highest ux obtained in the 23.5-cm-long ber. The transverse velocity of
the guided atoms was found to be 9.4 1.7 cm/s.
Dall et al. demonstrated guidance of metastable helium atoms by blue-detuned evanes-
cent waves in hollow capillary bers [25]. A bright helium atomic beam in the metastable
2
3
S
1
state is generated by the nitrogen-cooled discharge source. The atomic beam is
initially collimated in transverse directions by using a diode laser operating at 1083 nm
500
400
300
200
100
0
-4 -2 0 2 4 6 8
Guide Laser Detuning
(GHz)
F
l
u
x

(
H
z
)
Escort Laser Detuning
-1.6 GHz
-9.4 GHz
(2
3
S
1
2
3
P
2
). Then the atoms are cooled longitudinally by a second laser using a Zee-
man slower, where the longitudinal velocity is decreased from 900 m/s down to 100
m/s. Finally, the atoms are compressed in two-dimensional magneto-optical trap by a sepa-
rate laser, which results in the beam ux of up to 10
10
atoms/s over 1 cm
2
area. The guided
atoms are detected by the channeltron detector. To couple the light into the capillary ber,
one end of the capillary is optically polished at 45
angle so that the focused guiding laser

can be coupled from the side as discussed in Section 4.2. They have used three types of cap-
illaries: square-section capillary (350 m wide, 49 m hole), round-section capillary (110
m wide, 40 m hole), and further round-section capillary (150 m wide, 10 m hole).
With the input light power of 23 mW, the coupling efciency exceeds 70% for the square
capillary, whereas it is as low as 10% for the 110/40 m thin-walled capillary.
Figure 12 shows the guided atom signals with respect to the copropagating guide-laser
detuning for the 350/49 m capillary with the guide-laser power of 15 mW (the saturated
absorption signal is also recorded for frequency reference). As can be seen, the guiding
signal increases rapidly and decreases gradually as the detuning is increased. Figure 12(b)
shows the effect of reducing the laser intensity to 1/3 with respect to that shown in
Fig. 12(a). The reduced width of the transmitted signal is consistent with the decrease
of the evanescent-wave potential. The signal in Fig. 12(b) can be well tted with the
2
dependent function rather than
1
, as expected fromthe expression of the dipole potential.
Note that the rapid decline of the transmitted signal with respect to in Fig. 12(b) was not
due to capillary curvature. Rather, it was attributed to the rapid change of the area of the
bright regions associated with the speckle pattern (formed by the multimode guide light) in
the capillary. They also performed the atomic guidance experiment in the 110/40 m and
150/10 m round capillaries and obtained qualitatively similar results with respect to the
case of the square 350/49 m capillary [25].
4.2. Atomic Guide in Micron-Sized Hollow Optical Fiber
In a series of works, the Japan-Korea collaboration has reported on optical guidance of
thermal rubidium atoms by the blue-detuned evanescent waves induced in the micron-sized
HOFs. Figure 13 shows the schematic diagram of the experimental setup [23]. In the
vacuum chamber, a rubidium atomic beam from the hot oven, well collimated by several
apertures, is introduced into the HOF that is coaxially placed behind a holed mirror. The
ber has the hollow diameter of 7 m (2 m), core thickness of 3.8 m (4 m), and length
of 3 cm. The typical incident atomic ux was of the order of 10
6
atom/s.
The transmitted Rb atoms through the hollow ber were detected by a channel electron
multiplier, via two-step photoionization detection with two overlapping lasers; a diode laser
tuned to the Rb D
2
line and a high-power Ar-ion laser at the wavelength of 476.5 nm (the
ionization energy is 4.177 eVabove the 5S
1/2
ground state). The condition for efcient two-
step photoionization of the ground-state atom is given by P
i
P
0
(
0
/
i
) where P
i
is the
light intensity required for ionizing atoms in the excited state and
i
is the ionization cross-
section from the excited state to the ionization level [52]. For the Ar-ion laser intensity of
0.5 GW/m
2
, assuming the resonant excitation cross-section
0
= (3/2)
2
= 310
9
cm
2
for the 5S
1/2
5P
3/2
transition [53] and
i
= 2.5 10
17
cm
2
[54], the photoionization
efciency of about 30% was estimated.
Figure 12. (a) The transmitted atom signal (dots) and saturated absorption signal (solid
curve) as functions of the laser detuning for the square 350/49 m capillary in the copropa-
gating conguration. The inset shows a detailed view at small detunings. (b) Same as(a) but
with laser power reduced by 1/3. The dashed curve shows the
2
t for the far-detuned
blue wing (Figure from Ref. [25]).
Figure 13. Schematic of the experimental setup.
Figure 14 shows the typical photoionization spectrum of the
85
Rb atoms guided by
the 7-m HOF over the distance of 3 cm as a function of the frequency detuning of
the guiding laser with respect to the 5S
1/2
, F = 3 upper ground state. As expected, the
guided atomic ux is greatly enhanced in the blue-detuning region. The foot level of the
photoionization signal extends to the detuning over +20 GHz. The broken line in Fig. 14
shows the background transmission level that was obtained without the guiding laser, which
came from those atoms ballistically ying through the hollow ber [23]. By comparison
of the maximum guided atomic ux with the background transmission, the atomic-guide
enhancement factor was found to be about 20. From a similar experiment with the 2-m
hollow ber, in which the axis of the optical ber was slightly tilted against that of the Rb
atomic beam, a higher enhancement factor of 80 was obtained [55].
Figure 15 shows the novel characteristics of atomic isotope separation achieved in the
7-m hollow ber. The upper curve of Fig. 15 shows the case where the guide laser is
blue-detuned for both isotopes. In this case, both isotopes can be guided in the hollow
ber. On the other hand, the lower curve of Fig 15 shows the case where the guide laser is
blue-detuned for
87
Rb atoms but nearly red-detuned for
85
Rb atoms. As is clear, the
87
Rb
atoms are guided by the hollow ber, while the transmission of the
85
Rb atoms is greatly
suppressed, which represents the interesting feature of an in-line atomic-state lter.
In a subsequent experiment Ito et al. investigated a novel atomic guiding scheme in
which the guide light beam is coupled to the hollow core sideways at a 45
angle via total

internal reection near the edge [56]. There are several advantages for this method: First,
the scheme can easily remove unwanted light scattering such as the propagating modes
inside the hollow region. Second, this scheme prevents the leak of the incident light near
the entrance of the hollow ber, so that heating or optical pumping of atoms near the ber
entrance can be minimized. Moreover, this scheme also enables one to couple the guide
light from the rear direction. Recently, Fatemi et al. reported a side-coupling method
for the mutitimode HOF using embedded microprism [57]. Microprisms embedded into a
multimode, double-clad hollowber, allowlaser light to be coupled into the ber at multiple
locations along the length of the ber.
In the atom guidance through HOF, in addition to the repulsive optical dipole interac-
tion, the attractive cavity quantum electrodynamic (QED) interaction is also induced be-
Figure 14. Two-step photoionization spectrum of the
85
Rb atoms in the 5S
1/2
, F = 3 state
guided by the 7-m hollow ber over a distance of 3 cm (solid line). The broken line shows
the background transmission level without the guide laser (Figure from Ref. [23]).
Figure 15. In-line spatial separation of
85
Rb and
87
Rb in the 7-m hollow optical ber. The
upper curve shows the case where a guide laser is blue detuned for both isotopes while the
lower curve shows the case where a guide laser is blue-detuned for
87
Rb but red-detuned
for
85
Rb (Figure from Ref. [23]).
tween the dielectric wall and the nearby atoms [48, 58, 59]. Moreover, when the cavity
potential exceeds the optical potential near the surface, atoms will be attracted to the inner
wall and will be lost. Therefore, one can expect to observe the threshold behaviour of the
atomic transmission in the cylindrical dielectric cavity [49, 60, 61, 62, 63].
Figure 16. Two-step photoionization signal of
87
Rb atoms guided by the 1.4-m hollow-
core optical ber as a function of the guide-laser power (Figure from Ref. [65].
Figure 17. Atomic transmission signal of the guided
87
Rb atoms in the 0.3-m hollow ber
near the low-intensity threshold region (Figure from Ref. [64]).
Ito et al. have measured the threshold guide-laser intensities where the atomic trans-
mission starts increased [55, 64]. The hollow-core diameter is only of the order of or less
than the resonant wavelength of 780 nm for the
87
Rb D
2
transition, and consequently the
induced cavity effects are much signicant so that the small threshold intensities are now
easily measurable even in the 300-nm hollow ber. Figure 16 shows experimental data of
photoionization signal due to the atoms guided in the 1.4 m ber near the threshold [65].
As can be seen, the threshold behaviour in the atomic transmission is clearly observed at the
small power of 125 W. The laser power can be equivalently expressed in terms of the pure
optical potential U
op
(r = a) normalized to the mean transverse kinetic energy K
av
= mv
2
tr
of the guided
87
Rb atoms (m is the atomic mass and v
tr
is the transverse root-mean-square
velocity of
87
Rb atoms). Then the threshold intensity for the case of 1.4 m ber corre-
sponded to the ratio of U
op
(r = a)/K
av
= 2.5. When the total potential including the
cavity potential is considered, on the other hand, the ratio of the total potential to the trans-
verse kinetic energy becomes approximately one, as observed in the case of plane surface
[66]. The same experiment is also done in the slightly larger HOF with the core diame-
ter of 2.0 m. The results show very similar threshold behaviours and the corresponding
threshold laser-power is 40 W, which is only 30% of the value for the 1.4-m case. The
atomic guidance experiment was also performed in the much smaller 0.3-m hollow ber.
From the atomic transmission data near the threshold region presented in Fig. 17, one can
obtain the threshold guide-laser power of about 2.6 mW. This laser power is much larger
than those for the other two cases of larger HOFs. Note that the larger values of the thresh-
old intensities for the smaller hollow-core bers can provide the direct manifestation of the
cavity QED effects in the cylindrical cavity.
Figure 18. Spatial distribution of Rb atoms guided by the 7-m hollow ber (Figure from
Ref. [55]).
Ito et al. reported on the possibility of fabricating an arbitrary pattern by using the
atomic waveguide with the hollow optical ber, which may lead to a novel lithographic
technique of optically controlled atom deposition [55]. The experimental setup is similar
to that in Fig. 13. Figure 18 shows the surface-ionization signal of Rb atoms guided by
the 7-m hollow optical ber over the distance of 3 cm. The spatial distribution of the
Figure 19. Two-step photoionization spectrum of the
87
Rb atoms in the 5S
1/2
, F = 2 state
guided through the 1.4-m hollow ber at a low oven temperature (Figure from Ref. [55]).
guided atomic ux is obtained by the cross-sectional scan of the hot-wire at the distance
of 12 mm downstream from the exit facet of the hollow ber. The guide-laser frequency is
blue-detuned at the optimal value of +3 GHz with respect to the
85
Rb, 5S
1/2
, F = 3 upper
ground state. The FWHM of the spatial distribution shown in Fig. 19 is 20 m. Considering
the quantum efciency of 0.9 of the channeltron and the cross section of the hot wire, the
guided Rb ux is measured to be 10
5
atom/s above the background level.
5. Experiments with Diffracted LP
11
Modes
5.1. Generation of a Hollow Laser Beam Diffracted from a Hollow Optical
Fiber
In this subsection, we describe the method of generation of hollow laser beams (HLBs)
by means of diffracted LP
11
modes from an HOF. Although a doughnut-shaped divergent
beam generated directly from an HOF is needed, for the simultaneous realization of an
atomic funnel and an HOF atomic guiding, however, the calculations and experimental
results in section 2.2. reveal that such an HLB cannot be produced in a simple method.
These problems can be solved by using one of the TE
01
, TM
01
, and HE
21
modes which
result in LP
11
modes. All of these modes are, like the LP
01
mode, ring-shaped inside the
HOF, and, in addition, each output beam forms a doughnut-shaped hollow beam since they
are only the superposition of the output-eld pair of the LP
11
modes (For example, for TE
01
mode, the pair is shown in the upper row of Fig. 20(a)). Consequently we have superposed
two LP
11
modes instead of exciting the desired mode directly. In this combination, one
LP
11
mode has a polarization and a node line orthogonal to those of the other mode.
TE
01
+HE
21
TE
01
-HE
21
TM
01
+HE
21
TM
01
-HE
21
APP
hal f-wave
PBS1
M1
M2
PBS2
4X
HOF
40X
CCD
SCREEN
pl ate
camera
Screen
Figure 20. Generation of an HLB by a proper superposition of LP
11
modes: (a) diagram for
four degenerate congurations of LP
11
modes and (b) a sketch of the experimental setup.
LD, APP, M, and PBS, in (b) stand for laser diode, anamorphic prism pair, mirror, and
polarizing beam splitter, respectively.
The experimental setup is presented in Fig. 20(b). If one wants a specic mode to be
excited dominantly in a multimode ber, the transverse distribution of the incident light
should resemble that of the mode as much as possible and, in particular, its polarization
should be also matched [67]. A half-wave plate just after the laser makes it possible to
balance a relative intensity ratio of one mode to the other, which allows generation of a
more symmetric mode in an azimuthal direction. The resulting combined beam in front
of the ber may look like a single linearly-polarized beam with its plane of polarization
rotated 45
with respect to the horizontal plane, but one should note that each beam can be
adjusted separately, which was important in exciting modes different from each other.
The rst two pictures in Fig. 21, which were obtained by blocking one of the two optical
paths, represent the patterns of perpendicular modes at z = 0 before they are merged, and
5 m m
(a) (b) (c)
Figure 21. Superposition of two orthogonal LP
11
modes. Transverse intensity proles
at z = 0 are shown: (a) and (b) before, and (c) after superposition. Polarization and
angular variation of the corresponding electric eld can be described by, for example, (a)
xsin( + ) and (b) y cos( + ).
the last one shows their combined pattern, which is similar to that of LP
01
mode. Figure 22
shows its output intensity distribution at z = 250 m. The peak-to-peak distance is about
17 m and the dark spot size is about 8.2 m. We have checked its azimuthal isotropy
by measuring the beam proles along eight different radial axes and they showed good
uniformity within the maximum error of about 7%.
-20.0 -10.0 0.0 10.0 20.0
10 m m
Radialdistance( m) m
I
n
t
e
n
s
i
t
y
(
a
r
b
.
u
n
i
t
)
(a) (b)
Figure 22. HLB made of the diffracted output of the superposed mode. (a) CCD images of
the intensity distribution (z = 250 m) and (b) its prole in a radial direction.
Let us now discuss briey on the application of this beam to an atomic funnel. Figure
23 shows a possible conguration of our overall atom guiding system. It consists of three
main parts: a pyramidal or axicon mirror trap [68, 69], a HLB atomic funnel, and an HOF
evanescent-wave atomic guiding.
85
Rb atoms trapped in a pyramidal or an axicon mirror
trap are pushed through a small hole by the power imbalance of the laser light along the
mirror axis, and guided inside the HLB. With the HLB converging into the hollow region
of the HOF, they are nally guided through the HOF. The guiding laser light is coupled
into the core of the HOF from the side by the total reection at the glass-vacuum interface
which is ground and polished at an angle of 45
as employed in reference [56]. The cold

Pyramid
Mirror
Trap
AtomicFunnel
madeby
DarkHollowBeam
Hollow-core
Optical
Fiber
Evanascent
Field
Blue-detuned
LaserBeam
TrappingBeam
Channeltron
Figure 23. Set-up of a novel atomic guiding system. The guiding beam is launched into the
ber from the side, and channeltron is used to detect the guided atoms.
atom funnel by employing similar apparatus with the red-detuned Gaussian funneling beam
rather than a blue-detuned HLB was recently realized [70, 71].
5.2. Diffraction-Limited Dark Laser Spot Produced by a Hollow Optical
Fiber
Shin et al. reported on the generation of the diffraction-limited HLB having submicron-
sized dark spot by using the diffracted LP
11
mode from the HOF [33]. From the practical
point of view, leakage of light fromthe cladding region on the ber facet is the main obstacle
to obtaining the good optical quality of diffracted HLB since a short ber is generally used
for atom optical experiments, where contamination due to the leaked light is inevitable. To
produce an ideal HLB with HOF, it is required to block the cladding-guided light so that
unwanted scattering can be avoided. For this purpose, a microsphere is employed as an
evaporation mask for the core of HOF.
Figure 24(a) shows the image of the eld distribution of the LP
01
mode on the ber facet
observed by the imaging method [27]. The scattered light on the cladding surface is clearly
observed. They blocked the cladding-mode light by selectively metal-coating the cladding
on the bers facet with a microsphere used as an evaporation mask. They rst attach the
Figure 24. Intensity distribution of the LP
01
mode on the HOF-end facet imaged on the
screen for (a) the normal HOF and (b) the metal-coated HOF. (c) scanning electron micro-
scope (SEM) image of the microsphere attached to the HOF center after metal evaporation.
(d) Enlarged SEM picture near the hollow-core region with the sphere removed.
microsphere on the ber center masking the hollow core, evaporate thin lm of aluminum,
and then remove the microsphere afterwards. A 20-m-diameter microsphere is used to
effectively mask the hollow core having the diameter of 2a + 2d 12.3m. In Fig. 24(c),
the scanning electron microscope (SEM) image shows that the microsphere is positioned
well on the ber center. Once it is attached to the ber, the van der Waals force holds the
sphere tightly in its position. The more-detailed SEM picture of the facet after removing
the microsphere, i.e. the metal-coated HOF facet, is presented in Fig. 24(d). Fig. 24(b)
shows that the cladding-mode light is completely blocked by the described procedures.
They measured the beam proles and the dark hollow size of the diffracted LP
11
mode
as shown in Fig. 25. In Fig. 25(a), one can observe the beam-propagation characteristics
of the freely-diffracting HLB: whereas the bright ring on the ber-end facet is diminished,
the new peaks are developed from the center at around 30 m, maintaining the dark region
along the axis (the inner peaks diverge with a diffraction angle of 40 mrad). In particular,
the dark spot is preserved along the central axis even when HLB is spontaneously focused
due to diffraction. The experimental results are in good agreement with the numerical
simulations obtained by the Rayleigh-Sommerfeld theory, as also shown in Fig. 25(a).
Fig. 25(b) shows the radial intensity distributions at z = 15, 20, 25, 30 m, respectively.
Note that the central dark region near z = 30 m is slightly contaminated, which may be
associated with the intensity imbalance of two peaks, a slight excitation of LP
01
mode, or
the resolution limit of imaging system.
To estimate the size of the dark hollow region, they tted the measured prole for a
given z with two independent Gaussian curves and dene the radius of dark spot R
max
as
a half the distance between the central maxima of each curve. As shown in Fig. 26, the
Figure 25. Characteristic dimensions of the dark hollow region, measured in terms of the
dark-spot radius R
max
and the half-width w, which are in good agreement with the numer-
ical simulation.
0 20 40 60 80 100
0
1
2
3
4
5
hole radius of
hollowfiber
(
m
m
)
z (mm)
R
max
(simulation)
R
max
(experiment)
w
Figure 26. (a) Experimental and numerical results of the radial-intensity proles of the
diffracted LP
11
mode measured in steps of 5 m from the HOF-end facet (z = 0). (b)
Intensity proles measured near the focus at z = 15, 20, 25, 30 m.
minimum radius of dark spot is about 2 m which is similar to the hollow radius a itself.
As an alternative denition characterizing the dark spot, on the other hand, we also have
tted the dark region prole between the two curves with an inverted Gaussian curve, and
estimate the half width of dark spot w as the half width at half maximum of the single
inverted Gaussian. In this way, as in Fig. 26, we obtain that the smallest value of w is less
than 1 m (about 0.8 m around z = 35 m).
6. Experiments with Diffracted LP
10
Modes
In this section, we describe the various experiments performed by using the hollow laser
beams produced by a hollow optical ber. First, we describe the micro-imaging method
to generate HLBs and discuss the experiments of atom guiding, atom fountain guided by
an HLB, and crossed atom trap. Finally, as an application of Diffracted LP
10
modes, the
discussions on the optical dipole trap is presented.
6.1. Micro-Imaging Method for Hollow Fiber Modes
Yin et al. [27] obtained an HLB by using a micro-collimation technique for the output
beam of a micron-sized hollow optical ber. The principle of this method is very simple:
for a ber waveguide consisting of a hollow cylindrical core, some low-order modes can be
guided in the hollow core, such as the LP
01
, LP
11
, LP
21
, and LP
31
mode [35]. Therefore,
when one uses a microscope objective with a short focal length to image the output intensity
distribution at the facet of a hollow ber, a simple HLB can be obtained.
The inner and outer diameter of the hollow-core of the ber was 7 m and 14.6 m,
respectively and the outer diameter of the cladding of the ber was 123.4 m. The relative
refractive index difference, n = (n
2
1
n
2
2
)/(2n
2
1
) = 0.0018 and n
2
= 1.45, where n
1
and n
2
are the refractive index of the core and the cladding, respectively. The numerical
aperture is about 0.124. Figure 27 shows the relationship between the dark spot size (DSS)
and the propagation distance Z of the dark HLB. It can be observed that (i) the DSS of
the dark hollow beam collimated by a M-20 objective is about 50 m at Z = 100 mm
and about 100 m at Z = 500 mm, and (ii) the relative divergent angle in the near eld of
HOF is about 6.5 10
5
, whereas the divergent angle in the far eld is 4.0 10
4
. If one
uses an HOF having a slightly larger hollow-core, an HLB with a smaller DSS and better
propagation invariance may be obtained.
The HLB generated by the micro-imaging method was used to couple into the HOF
to increase the coupling efciency by Takamizawa et al. [72]. In order to eliminate the
undesirable preinteraction before atoms enter the hollow region, they used an annular beam
whose shape and diameter are approximately the same as those of the core. The use of
annular beam also makes it feasible to excite the LP
01
mode with a relatively low power
of 10 mW to reect atoms. The evanescent wave produced with the LP
01
mode has
a cylindrical shape around the inner-wall surface without nodes, and consequently it is
suitable for the atom guidance.
A Gaussian light beam coupled to the HOF excites LP
01
mode. The output light beam
from the HOF, which has an annular shape just after exiting the HOF, is collimated and
recoupled to another HOF with two convex lenses. Then, annular light beam with the same
Figure 27. The relationship between the DSS and the propagation distance Z of the HLB
measured with (a) M-20 and (b) M-40 objective lens.
intensity prole as that of LP
11
mode is reproduced at the focal point behind the second
lens. The typical values of the focal length and the distance of the two lenses are 25.2 mm
and 500 mm, respectively. The coupling efciency was increased to 75% when the annular-
shaped beam was used compared to 59%, which was the maximum coupling efciency
when a Gaussian beam was coupled to the HOF [72]. The hole diameter, core thickness,
and the length of the used HOF was 7 m, 3.8 m, and 3 m, respectively.
6.2. Atom Guiding with Hollow Laser Beams
Xu et al. performed optical guiding of trapped cold atoms by a hollow laser beam produced
by micro-collimation and micro-imaging technique as discussed in the previous subsection
[28]. The atomic guiding direction was downward along the gravity (+z direction), whereas
the HLB propagated along the z direction (counterpropagating scheme) or along the +z
direction (copropagating scheme). A Ti:sapphire laser was used as the guiding laser source
with a maximum output power of 1.8 W. It was coupled to the core of HOF with a coupling
efciency of about 30%. The typical HLB power used for guiding atoms was 250 mW.
They used a micron-sized HOF that has a hollow diameter of 4 m, core thickness of 2
m, and length of 25 cm. In both guiding schemes, they obtained the identical radius of the
maximum-intensity ring
m
(z) that varies linearly with the distance z [
m
(z) =
m
(0)
z, where
m
(0) = 1.4 mm is the value at the trap center (z=0) and = 1.27(4) 10
3
].
They used a standard vapor-cell magneto-optical trap (MOT) of
85
Rb atoms. The num-
ber of trapped atoms was typically 2 10
7
and the trap diameter was about 1 mm so that
the loading efciency of the trapped atoms into HLB was 98%. By time-of-ight measure-
ment, the temperature of atoms in the MOT was found to be about 140 K, which was
further cooled down to 16 K by the polarization-gradient cooling. After the sub-Doppler
cooling, the cooling and repumping lasers were blocked by mechanical shutters, and the
HLB was simultaneously introduced to the atoms to guide their gravitational falling. The
number and the temperature of guided atoms were detected by observing the probe-induced
uorescence with a photomultiplier. The probe laser beam was placed horizontally at 105
mm below the trap center.
0.00 0.05 0.10 0.15
0.0
0.5
1.0
1.5
2.0
2.5
Free
16 GHz
10 GHz
6 GHz
2 GHz
1 GHz
G
u
i
d
e
d
A
t
o
m
F
l
u
x
(
a
r
b
.
u
n
i
t
s
)
Time of Flight (s)
0.10 0.12 0.14 0.16 0.18 0.20 0.22 0.24
0.0
0.5
1.0
1.5
2.0
2.5
Free
16 GHz
10 GHz
6 GHz
G
u
i
d
e
d
A
t
o
m
F
l
u
x
(
a
r
b
.
u
n
i
t
s
)
Time of Flight (s)
Figure 28. Typical TOF signals of atoms guided by a single HLB. (a) In the copropagating
scheme, the laser detuning
2
is 1, 2, 6, 10, and 16 GHz, respectively. (b) In the counter-
propagating scheme,
2
is 6, 10, and 16 GHz, respectively.
Figure 28 shows time-of-ight signals of guided cold atoms in both guiding schemes
at various laser detunings with respect to the 5S
1/2
, F = 2 5P
3/2
transition line. For
comparison, the detected signal without the HLB for the freely falling atoms is also shown.
In particular, it is observed that the number of atoms guided by the copropagating HLB is
about 20-fold enhanced with respect to that without the HLB at 2 GHz detuning. In this
case, the guided atoms also become accelerated along the +z direction due to the increased
radiation pressure at small detunings [Fig. 28(a)]. In the counterpropagating case, on the
other hand, the guided atoms are decelerated as the detuning is decreased [Fig. 28(b)].
-4 -2 0 2 4 6 8 10 12 14 16
0
10
20
30
40
50
Copropagating
Counterpropagating
(b)
(a)
G
u
i
d
i
n
g
E
f
f
i
c
i
e
n
c
y
h
(
%
)
Detuning d (GHz)
Figure 29. Guiding efciency as a function of the detuning
2
in the copropagating (a) as
well as the counterpropagating (b) scheme. The solid curves represent numerical simulation
results.
Figure 29 presents experimental and numerical guiding efciencies versus detuning in
the copropagating (a) as well as in the counterpropagating (b) scheme. Note that in numer-
ical simulation, the HLB was assumed, to a good approximation, as the lowest Laguerre-
Gaussian (LG
1
0
) mode given by
I(, z) =
4P
w
2
2
w
2
exp
2
2
w
2
, (19)
where P is the laser power, w = w
0
1 + (z/z
R
)
2
is the beam waist at distance z, w
0
is the beam waist at z = 0, and z
R
= w
2
0
/ is the Rayleigh length. It can be observed
that at small detuning, atoms are most efciently guided in the copropagating scheme (for
example, the maximum guiding efciency is about 50% at the detuning of 2 GHz). On
the other hand, the counterpropagating guiding is generally less efcient as found in Fig.
29. However, for large detunings, both schemes provide similar guiding efciencies and
the maximum efciency of 23% is obtained at 10 GHz detuning in the counterpropagating
scheme.
6.3. Atom Fountain with Hollow Laser Beams
The development of an atomic fountain based on laser-cooled atoms [73, 74] has created
prospects for an improved accuracy and stability of frequency standards. In such a clock,
one approach to solve the line shift due to cold collisions is to use laser light for guiding the
upward launched atoms [75]. This is because the guiding can enhance the number of atoms
which come back into the microcavity without increasing the atomic densities.
Optical guiding of an atomic fountain by using a cylindrical HLB was demonstrated by
Kim et al. [29]. The generated HLB by using micro-imaging method was collimated by
the objective lens and propagated downwards toward the center of the Rb MOT. The power
of the guiding laser was 250 mW and the beam waist was 3 mm. The HLB was nearly
collimated in order to remove the dipole force of the guiding direction, which can cause
broadening of the spatial distribution of guided atoms. With an intensity of 3 mW/cm
2
in
each beam, the typical diameter of an atomic cloud in the MOT was about 1 mm, and the
number of trapped atoms was typically 2 10
7
. Cold atoms were then launched upwards
in a rather simple way by varying rapidly the vertical magnetic eld resulting in the atomic
Zeeman shift. After 1-ms acceleration, the detuning of the laser beams was changed from
2.5 to 70 , lowering the atomic temperature to 33.7 Kin the frame moving upwards.
A typical launching velocity of ascending atoms was 1.4 m/s and the atoms were launched
up to 10 cm.
The number of guided atoms was detected by observing the uorescence with a photo-
multiplier tube, which was induced by a horizontally placed probe laser at 10.5 cm below
the center of the MOT. They observed that 0.5% of the launched atoms were detected with-
out the HLB. On the other hand, a tenfold enhancement of the HLB-guided atomic fountain
was clearly obtained without appreciable heating. In Fig. 30, the line (a) is the time-of-
ight (TOF) signal of atoms that are launched without the HLB, while the other line (b)
is the TOF signal with the HLB at a detuning of 19 GHz. From this TOF signal, one can
deduce the guiding efciency of atoms and the temperature. Without the HLB, the temper-
ature was about 33.7 (2.1) K and about 34.4 (1.7) K with the HLB.
0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45
0.0
0.2
0.4
0.6
0.8
1.0
P=250 mW
d
2
=19 GHz
v
launch
=1.4 m/s
HLBoff (a)
HLBon (b)
N
u
m
b
e
r
o
f
A
t
o
m
s
(
a
r
b
.
u
n
i
t
s
)
Time of Flight (s)
Figure 30. TOF signals in the HLB-guided atomic fountain experiment. The line (a) is for
the case without the HLB whereas the line (b) is for the case with the HLB.
To characterize the enhancement due to the guiding HLB, they introduced the enhance-
ment factor, dened as the ratio of the number of atoms guided with the HLB to that without
HLB. In Fig. 31, the line with lled squares shows the relationship between the enhance-
ment factor and the detuning measured with respect to the 5S
1/2
, F = 2 5P
3/2
transition
line. The inset shows the enhancement of the guiding efciency for larger detuning. The
line with empty circles shows how the number of scatterings, or the heating, is changed with
0 5 10 15 20
0
10
20
30
40
0
500
1000
1500
2000
0 50 100
0
10
20
30
40
N
u
m
b
e
r
o
f
S
c
a
t
t
e
r
i
n
g
s E
n
h
a
n
c
i
n
g
F
a
c
t
o
r
d
2
[GHz]
Enhancing Factor
Number of Scattering
Figure 31. Dependence of the enhancement factor () and the number of scatterings () on
the detuning . The inset shows the enhancement factor for larger detuning.
the detuning. They observe that for small detuning, the enhancement factor is more than 35,
but there is serious heating. As the detuning increases, the enhancement factor as well as
the heating decrease. Note that the number of scatterings decreases more rapidly (
2
)
than the enhancement factor. At a detuning of 19 GHz, the enhancement factor is over 10
and an atom experiences spontaneous emissions about 40 times during the launching and
falling processes. According to the calculation, however, they found that the heating due to
spontaneous emissions was not so serious.
In order to reduce a loss of atomic coherence, an HLB with a large detuning may be
used. For example, if a 15-W Ar-ion laser is used for a tenfold enhancement of guiding
efciency, the average rate of spontaneously scattered photons is calculated to be 10
3
Hz.
While the number of atoms being guided in the fountain is increased, the HLB introduces
inhomogeneous energy shifts of the ground-state hyperne levels. In a trap based on a sheet
of blue-detuned light supporting against gravity, a Stark shift of 270 mHz is obtained for
4-s trapping time, which is larger than the line shift due to cold atom collisions [76]. One
possibility for reducing the light shift in the HLB is to use a much higher-order Bessel beam
for the HLB or to use the evanescent waves of a hollow optical ber. Since the evolution
of atoms in an HLB depends on the shape of the HLB, it is suggested that the ensemble-
averaged heating and the light shift will be changed with the shape of the HLB [77]. In
principle, if the potential of the HLB is square, then atoms in the HLB may not feel any
scattering or light shift.
6.4. Crossed-HLB Trap of Rb
Xu et al. constructed a blue-detuned optical dipole trap by intersecting two horizontal,
cylindrical HLBs at a right angle in the center of a Rb MOT [30]. The polarizations of the
beams were chosen to be orthogonal in the crossed region in order to suppress standing-
wave effects. The detuning of HLBs was 20 GHz from 5S
1/2
, F = 3 5P
3/2
, F
= 4
transition line of
85
Rb and the trap depths were about 10 K in the x-direction and 90 K
in the z-direction, respectively. They estimated that 60% of the atoms in the MOT were
initially loaded in the HLB trap.
0 20 40 60 80 100
0
20
40
60
80
100
Experimental Data
Simulation Results
T
r
a
p
p
i
n
g
E
f
f
i
c
i
e
n
c
y
(
%
)
Trapping Time (ms)
Figure 32. The trapping efciency of a crossed-HLB trap as a function of the trapping time.
The two horizontal HLBs have a power of 200 and 400 mW, the detuning of each HLB is
20 GHz, and the initial temperature of atoms is 16 K.
The number and the temperature of the trapped atoms could be deduced by the TOF
measurements. They found that the temperature of the trapped atoms was about 7 K and
this value was close to the minimum height of potential barrier of 10 K, where about 10
5
atoms stayed inside the trap for 100 ms and the lifetime of trapped atoms was about 20
ms as shown in Fig. 32. Figure 32 also shows the trapping efciency as a function of the
trapping time. For comparison, the simulation results are also shown as the solid line in
Fig. 32 [78]. Since the detuning was much larger than the splitting between the hyperne-
structure levels of the excited state, the three-level interaction mode was quite good for the
simulation.
6.5. Optical Dipole Trap
Shin et al. proposed a three-dimensional, microscopic, and diffraction-limited far-off-
resonance optical dipole trap (DFORT) for neutral atoms operating in the Lamb-Dicke
regime, which is produced by employing the diffracted LP
01
mode of an HOF [34]. DFORT
provides, in particular, a large trap volume so that it can be loaded with a large number of
cold atoms. Moreover, the 3D DFORT can be also operated as an elongated 1D optical
trap. Such a microscopic and tight optical trap can be realized with moderate laser power
and a simple experimental setup. As the LP
01
mode propagates in free space, its initial
annular intensity distribution, which is represented by the two peaks on the HOF exit facet
diminishes away while the central bright peak starts to appear. The resultant generation
of a tightly focused bright spot near the ber facet can be qualitatively understood as the
diffraction of the plane wave (i.e., the uniform LP
01
mode) by a ring-shaped aperture and
the subsequent constructive interference near the central axis. Note that the intensity along
the axial direction rst increases to a maximum and then gradually decreases down to zero
[Fig. 33(b)], and this axial asymmetry of the intensity gradient may be useful for efcient
loading of the nearby precooled atoms into DFORT. Note also that the axial intensity gradi-
ent is larger than that of a typical focused Gaussian beam and the spot is even more tightly
focused in the transverse radial direction [Fig. 33(b)].
r/l
z/l
r/l
-10 0 10
0.0
0.5
1.0
r
/
l
z/l
z/l
0 200 400
Figure 33. (a) Typical diffraction prole of the LP
01
mode of HOF when a = 4 and
d = 3.5. (b) The contour plots of the produced optical potential. A, the radial distribution
of DFORT at z = z
0
= 50.9, B, axial distribution from z = 0 to z = 600.
As a specic application, they consider the
87
Rb atoms and the laser power P = 100
mW at the wavelength = 800 nm, which is far detuned from the D
1
and D
2
resonance
lines. Then they obtain a cigar-shaped DFORT having the following basic parameters: the
maximum trap depth U
0
= 7.9 mK, the axial trap frequency f
z
= 4.9 kHz, the radial trap
frequency f
r
= 125 kHz, the trap volume V
trap
10
8
cm
3
, and the scattering rate at
the trap center
sc
= 2 310 Hz. In this case, the Lamb-Dicke parameters in the axial
and the radial direction, dened by the ratio of the corresponding ground-state size a
0i
to
the laser wavelength,
i
= 2a
0i
/ (i = z, r), are given by
z
= 0.86 and
r
= 0.18,
respectively. For
133
Cs atoms at = 900 nm, on the other hand, the trap depth is U
0
= 7.0
mK, the oscillation frequencies are f
z
= 3.3 kHz and f
r
= 842 kHz, the Lamb-Dicke
parameters are
z
= 0.75 and
r
= 0.15, and the scattering rate is
sc
= 2 317 Hz.
The DFORT has the unique feature of a tightly focused trap with a large trap volume and
convenient loading and cooling of the precooled atoms, and may be also operated as an
elongated one-dimensional optical trap.
7. Conclusion
We have reviewed various atom optical experiments using the hollow optical bers. The
detailed study on characteristics of electromagnetic elds inside and outside the HOF is
presented. We review the neutral atom guidance by means of red- and blue-detuned laser
beams, and the applications of the laser beams emanating from the HOF. In particular,
an atomic guide in small hollow-core bers has interesting applications. For example, by
using submicron-sized hollow bers, it is possible to carry cold atoms at arbitrary points on
a substrate for precise control of atomic deposition. By using bent hollow bers, an opto-
gravitational trap [79] and an atom-laser cavity [80] can be also constructed. Furthermore,
the nonlinear dynamics has been theoretically studied for the guided atoms inside HOF with
the potential depth periodically modulated [81]. Moreover, cold atoms can be manipulated
by a sharpened, nanometric optical ber tip or trapped by an evanescent eld induced near
the tip [82], just like an optical tweezer is used for manipulating nanometric particles [83].
In particular, this may lead to the possibility of single atom manipulation so that atomic-
scale crystal growth can be realized.
Cold atom guidance by hollow ber can be also applicable to crystal growth of silicon
on the atomic scale with a near-eld optical device if a laser of 252 nm wavelength is
available. The cavity QED effect in the near-eld region is also an interesting subject:
hollow ber or sharpened ber with an induced optical near-eld can be a unique tool
for experimental study of the cavity QED effects by measuring the atomic deection and
comparing it with the dipole forces [64]. Recently, the scope of the atomic guidance through
the HOF is extended: As well as cold atoms, the micro- or nano-sized particles were guided
through the HOF [84]. Furthermore, in stead of hollow optical bers, a photonic band-gap
ber (PBG) is also used for guiding atoms [85] or Bose-Einstein condensed atomic sample
[86].
In the second part of this article, we have investigated the characteristics of the output
intensity distribution of each LP
lm
mode in HOF using the diffraction theory, and compared
them with experiments. We have observed that the LP
01
or LP
11
mode itself cannot satisfy
the basic requirement for an atomic funnel since the former does not support a dark column
along the central axis while the latter causes loss of atoms due to the line of nodes. To
overcome these limitations, two LP
11
modes were excited separately with their node lines
and polarizations orthogonal to each other and then combined at an even fraction. The
resultant mode has an annular intensity distribution nearly similar to that of LP
01
and its
output forms a divergent doughnut-shaped beam with the minimum dark spot of a few m,
and this HLB can provide a deep optical potential for an atomic funnel which focuses atoms
from their source onto a micron-sized hollow region of an HOF with a high efciency,
and the evanescent eld associated with the corresponding mode can allow atoms guided
through an HOF. Moreover, the microscopic dark spot may be useful for the increase of the
atomic density in the optical funnel trap suggested in reference [87, 88]. They can also be
used for generation of the diffraction-limited dark laser spot [33]. Small focused dark spot
of HLB may be useful in atom optical experiments such as atomic lens, atom trap, and atom
switch.
The hollow laser beam generated by imaging the guided laser output of the hollow ber
has been used for atom guidance [28], atom fountains [29], and atom traps [30]. Further-
more, the diffracted output of the LP
01
mode has a bright focused spot, which can be used
for a tight optical dipole trap when a red-detuned laser is used [34].
References
[1] Adams, C. S; Siegel, M.; Mlynek, J. Phys. Rep. 1994, 240, 143210.
[2] Balykin, V. I.; Letokhov, V. S. Atom Optics with Laser Light; Harwood Academic
Publishers: Chur, 1995.
[3] Dowling, J. P.; Geo-Banacloche, J. Adv. At. Mol. Opt. Phys. 1996, 34, 194.
[4] Berman, P. R. Atom Interferometry; Academic Press: New York, 1997.
[5] Dholakia, K. Contemp. Phys. 1998, 39, 351369.
[6] Balykin, V. I. Adv. At. Mol. Opt. Phys. 1999, 41, 181260.
[7] Hinds, E. A.; Hughes, I. G. J. Phys. D 1999, 32, R119R146.
[8] Friedman, N.; Kaplan, A.; Davidson, N. Adv. At. Mol. Opt. Phys. 2002, 48, 99151.
[9] Folman, R.; Kr uger, P.; Schmiedmayer, J.; Denschlag, J.; Henkel, C. Adv. At. Mol.
Opt. Phys. 2002, 48, 263356.
[10] Sudo, S.; Yokoyama, I.; Yaska, H.; Sakai, Y.; Ikegami, T. IEEE Photon. Tech. Lett.
1990, 2, 128131.
[11] Christov, I.; Kapteyn, H.; Murnane, M. Optics Exp. 1998, 3, 360365.
[12] Choi, S.; Eom, T. J.; Yu, J. W.; Lee, B. H.; Oh, K. IEEE Photon. Rechnol. Lett. 2002,
14, 17011703.
[13] Choi, S.; Oh, K.; Shin, W.; Park, C. S.; Paek, U. C.; Park, K. J.; Chung, Y. C.; Kim,
G. Y.; Lee, Y. G. IEEE Photon. Technol. Lett. 2002, 14, 248250.
[14] Choi, S.; Oh, K. Opt. Commun. 2003, 221, 307312.
[15] Oh, K.; Choi, S.; Jung, Y.; Lee, J. W. J. Lightw. Technol. 2005, 23, 524532.
[16] Ashkin, A. Phys. Rev. Lett. 1978, 40, 729732.
[17] Olshanii, M. A.; Ovchinnikov, Yu. B.; Letokhov,V. S. Opt. Commun. 1993, 98,
7779.
[18] Renn, M. J.; Montgomery, D.; Vdovin, O.; Anderson, D. Z.; Wieman, C. E.; Cornell,
E. A. Phys. Rev. Lett. 1995, 75, 32533256.
[19] Savage, C. M.; Marksteiner, S.; Zoller, P. In Fundamentals of Quantum Optics III;
Ehlotzky F.; Ed.; Springer-Verlag: Berlin, 1993; pp 6074 .
[20] Marksteiner, S.; Savage, C. M.; Zoller, P.; Rolston, S. Phys. Rev. A 1994, 50,
26802690.
[21] Jhe, W.; Ohtsu, M.; Hori, H.; Friberg, S. R. Jpn. J. Appl. Phys. 1994, 33,
L1680L1682.
[22] Renn, M. J.; Donley, E. A.; Cornell, E. A.; Wieman, C. E.; Anderson, D. Z. Phys. Rev.
A 1996, 53, R648R651.
[23] Ito, H.; Nakata, T.; Sakaki, K.; Ohtsu, M.; Lee, K. I.; Jhe, W. Phys. Rev. Lett. 1996,
76, 45004503.
[24] Workurka, G.; Keupp, J.; Sengstock, K.; Ertmer, W. IQEC 98, Technical Digest 1998,
235.
[25] Dall, R. G.; Hoogerland, M. D.; Baldwin, K. G. H.; Buckman, S. J. J. Opt. B 1999, 1,
396401.
[26] Dall, R. G.; Hoogerland, M. D.; Tierney, D.; Baldwin, K. G. H.;Buckman, S.J. Appl.
Phys. B 2002, 74, 1118.
[27] Yin, J. P.; Noh, H. R.; Lee, K.; Kim, K.; Wang, Y.; Jhe, W. Optics Comm. 1997, 138,
287292.
[28] Xu, X.; Kim, K.; Jhe, W.; Kwon, N. Phys. Rev. A 2001, 63, 063401.
[29] Kim, K.; Noh, H. R.; Yeon, Y. H.; Xu, X.; Jhe, W.; Kwon, N. J. Korean Phys. Soc.
2001, 39, 877880.
[30] Xu, X.; Kim, K.; Jhe, W. J. Korean Phys. Soc. 2000, 37, 661664.
[31] Won, C.; Yoo, S. H.; Oh, K.; Paek, U. C.; Jhe, W. Opt. Commun. 1999, 161, 2530.
[32] Yoo, S. H.; Won, C. Y.; Kim, J. A.; Kim, K.; Oh, K.; Paek, U. C.; Jhe, W. J. Opt. B
1999, 1, 364370.
[33] Shin, Y. I.; Kim, K.; Kim, J. A.; Noh, H. R.; Jhe, W. Opt. Lett. 2001, 26, 119121.
[34] Shin, Y. I.; Heo, M.; Kim, J. W.; Shim, W.; Noh, H. R.; Jhe, W. J. Opt. Soc. Am. B
2003, 20, 937941.
[35] Ito, H.; Sakaki, K.; Nakata, T.; Jhe, W.; Ohtsu, M. Opt. Commun. 1995, 115, 5764.
[36] Chaudhuri, P. R.; Lu, C.; Xiaoyan, W. Opt. Commun. 2003, 228, 285293.
[37] Hautakorpi, M.; Kaivola, M. J. Opt. Soc. Am. A 2005, 22, 11631169.
[38] Marcuse, D. Light Transmission Optics; Van Nostrand Reinhold: New York, 1982.
[39] Okoshi, T. Optical Fibers; Academic Press: New York, 1989.
[40] Snitzer, E. J. Opt. Soc. Am. 1981, 51, 491498.
[41] Gloge, D. Appl. Opt. 1971, 10, 22522258.
[42] Ito, H.; Sakaki, K.; Nakata, T.; Jhe, W.; Ohtsu, M. Ultramicroscopy 1995, 61, 9197.
[43] Akhmanov, S. A.; Nikitin, S. Y. Physical Optics; Oxford University Press: New York,
1997.
[44] Gaskill, J. D. Linear Systems, Fourier Transform, and Optics; John Wiley & Sons:
New York, 1978.
[45] Sheppard, C. J. R.; Gan, X. S.; Fatemi, H. Electron. Lett. 1996, 32, 11281130.
[46] Renn, M. J.; Zozulya, A. A.; Donley, E. A.; Cornell, E. A.; Anderson, D. Z. Phys.
Rev. A 1997, 55, 36843696.
[47] Hayashi, S.; Ishimizu, A.; Tohei, T.; Tachikawa, M. Phys. Rev. A 2003, 68, 053408.
[48] Chevrollier, M.; Fichet, M.; Oria, M.; Rahmat, G.; Bloch, D.; Ducloy, M. J. Phys. II
France 1992, 2, 631657.
[49] Courtois, J. Y.; Courty, J. M.; Mertz, J. C. Phys. Rev. A 1996, 53, 18621878.
[50] M uller, D.; Cornell, E. A.; Anderson, D. Z.; Abraham, E. R. I. Phys. Rev. A 2000, 61,
033411.
[51] Lu, Z. T.; Corwin, K. L.; Renn, M. J.; Anderson, M. H.; Cornell, E. A.; Wieman, C.
E. Phys. Rev. Lett. 1996, 77, 33313334.
[52] Letokhov, V. S. Laser Photoionization Spectroscopy; Academic Press: San Diego,
1987.
[53] Drake, G.W. Atomic, Molecular, and Optical Physics Handbook; AIP Press: New
York, 1996.
[54] Dinneen, T. P.; Wallace, C. D.; Tan, K.-Y. N.; Gould, P. L. Opt. Lett. 1992 17,
17061708.
[55] Ito, H.; Sakaki, K.; Jhe, W.; Ohtsu, M. Appl. Phys. Lett. 1997, 70, 24962498.
[56] Ito, H.; Sakaki, K.; Jhe, W.; Ohtsu, M. Opt. Commun. 1997, 141, 4347.
[57] Fatemi, F. K.,; Bashkansky, M.; Moore, S. Opt. Express 2005, 13, 48904895.
[58] Casimir, H. B. G.; Polder, D. Phys. Rev. 1948, 73, 360372.
[59] Berman, P. R. Cavity Quantum Electrodynamics; Academic Press: San Diego, 1994.
[60] Jhe, W.; Kim, J. W. Phys. Rev. A 1995, 51, 11501153.
[61] Jhe, W.; Kim, J. W. Phys. Lett. A 1995, 197, 192196.
[62] Wylie, J. M.; Sipe, J. E. Phys. Rev. A 1984, 30, 11851193.
[63] Nha, H.; Jhe, W. Phys. Rev. A 1996, 54, 35053513.
[64] Ohtsu, M. Near-Field Nano/Atom Optics and Technology; Springer-Verlag: Tokyo,
1998.
[65] Jhe, W.; Noh, H. R.; Ito, H.; Ohtsu, M. unpublished.
[66] Landragin, A.; Courtois, J. Y.; Labeyrie, G.; Vansteenkiste, N.; Westbrook, C. I.;
Aspect, A. Phys. Rev. Lett. 1996, 77, 14641467.
[67] Saleh, B. E. A.; Teich, M. C. Fundamentals of Photonics; Wiley Interscience Publica-
tion: New York, 1991.
[68] Lee, K. I.; Kim, J. A.; Noh, H. R.; Jhe, W. Opt. Lett. 1996, 21, 11771179.
[69] Kim, J. A.; Lee, K. I.; Noh, H. R.; Jhe, W.; Ohtsu, M. Opt. Lett. 1997, 22, 117119.
[70] Takamizawa, A.; Ito, H.; Ohtsu, M. Appl. Phys. Lett. 2004, 85, 17901792.
[71] Takamizawa, A.; Ito, H.; Yamada, S.; Ohtsu, M. Jpn. J. Appl. Phys. Pt. 1 2002, 41,
62156218.
[72] Takamizawa, A.; Ito, H.; Ohtsu, M. Jpn. J. Appl. Phys. Pt. 1 2000, 39, 67376742.
[73] Kasevich, M.; Riis, E.; DeVoe, R. G.; Chu, S. Phys. Rev. Lett. 1989, 63, 612615.
[74] Clairon, A.; Salomon, C.; Guellati, S.; Phillips, W. D. Europhys. Lett. 1991, 16,
165170.
[75] Szymaniec, K.; Davies, H. J.; Adams, C. S. Europhys. Lett. 1999, 45, 450455.
[76] Davidson, N.; Lee, H. J.; Adams, C. S.; Chu, S. Phys. Rev. Lett. 1995, 74, 13111314.
[77] Song, Y.; D. Milam, D.; Hill III, W. T. Opt. Lett. 1999, 24, 18051807.
[78] Xu, X.; Minogin, V. G.; Lee, K.; Wang, Y. Z.; Jhe, W. Phys. Rev. A 1999, 60,
47964804.
[79] Harris, D. J.; Savage, C. M. Phys. Rev. A 1995, 51, 39673971.
[80] Moy, G. M.; Hope, J. J.; Savage, C. M. Phys. Rev. A 1997, 55, 36313638.
[81] Liu, X. M.; Milburn, G. J. Phys. Rev. A 2000, 61, 053401.
[82] Kobayashi, K., Sangu, S.; Ito, H.; Ohtsu, M. Phys. Rev. A 2001, 63, 013806.
[83] Novotny, L.; Bian, R. X.; Xie, X. S. Phys. Rev. Lett. 1997, 79, 645648.
[84] Renn, M. J.; Pastel, R.; Lewandowski, H. J. Phys. Rev. Lett. 1999, 82, 15741577.
[85] Bigelow, M.; Takekoshi, T.; Maes, C.; Knize, R. Abstract of DAMOP Meeting
2005, 42.
[86] Will, S.; Christensen, C.; Jo, G.-B.; Pasquini, T. A.; Saba, M.; Shin, Y.; Ketterle, W.;
Pritchard, D. E. Abstract of ICAP Meeting 2006, 506.
[87] Morsch, O.; Meacher, D. R. Opt. Comm. 1998, 148, 4953.
[88] Yin, J. P.; Zhu, Y. Optics Comm. 1998, 152, 421428.
Chapter 5
ADVANCES IN PHYSICAL MODELING
OF RING LASERS
Vittorio M.N. Passaro
1*
and Francesco De Leonardis
2**
1
Photonics Research Group, Dipartimento di Elettrotecnica ed Elettronica,
Politecnico di Bari, via Edoardo Orabona n. 4, 70125 Bari, Italy
2
Photonics Research Group, Dipartimento di Ingegneria dellAmbiente e per lo Sviluppo
Sostenibile, Politecnico di Bari, viale del Turismo n. 8, 74100 Taranto, Italy
Abstract
In this chapter, an overview on fiber ring lasers and III/V semiconductor integrated ring
lasers is presented. In particular, some aspects of mathematical modelling of these devices are
reviewed. In the first part of the chapter, we have focused our attention on the more recent
theoretical and experimental studies concerning fiber ring laser architectures. Then, a
complete quantum-mechanical model for integrated ring lasers is presented, including the
evaluation of all the involved physical parameters, such as self and cross saturation and
backscattering. Finally, the influence of sidewall roughness on either unidirectional or
bidirectional regime in multi-quantum-well III/V semiconductor ring lasers is demonstrated.
Keywords: Semiconductor Ring lasers, Fiber Ring Lasers, Multi quantum well, Modeling
Introduction
Nowadays, fiber ring lasers are typical fiber devices obtained by placing a fiber amplifier
inside a cavity to provide optical feedback, so involving a number of linear and non linear
physical effects. Moreover, semiconductor ring lasers are of great interest for monolithic and
integrated optoelectronic applications. Since these lasers do not require the use of cleaved
facets or difficult Bragg gratings for optical feedback, they can be conveniently integrated so
reducing the occupation area. Moreover, the semiconductor ring lasers offer new capabilities

*
E-mail address: passaro@deemail.poliba.it
**
E-mail address: f.deleonardis@poliba.it
Vittorio M.N. Passaro and Francesco De Leonardis 162
in the travelling-wave unidirectional oscillation. Unidirectional operation is desired because it
offers the advantages of enhanced mode purity (high side-mode suppression ratio), reduced
sensitivity to feedback and higher single beam power. Unidirectional ring lasers are used for
telecommunications systems, feedback laser diodes, multi-wavelength and all-optical flip flop
operation.
In this work we present both fiber and semiconductor ring lasers, with some recent
advances on their physical models. The chapter is substantially divided in two parts. The first
part is essentially a review of fiber ring laser architectures, their relevant physical effects,
technologies and main applications, including erbium doped devices, and continuous wave
and mode-locking operations. In the second part, we focus on integrated architectures of
semiconductor ring lasers and their physical models. In particular we investigate the influence
of some technological parameters (ring sidewall roughness, ring radius) over the
backscattering coefficient influencing the operation regime and performance of any
semiconductor ring laser. The theory of the physical model based on quantum mechanical
approach is briefly summarized and some numerical results and simulations applied to multi-
quantum-well III/V semiconductor ring lasers are presented. Here how either unidirectional or
bidirectional regime is related to the ratio between the injection current and the backscattering
coefficient value is shown. Finally simulation results relevant to an architecture of the ring
laser with asymmetrical output coupler to obtain a stable unidirectional regime are presented.
In general, our model is based on four differential equations in total, two coupled
equations for the counter-propagating modes, one rate equation for the carriers injected in the
laser active region and one equation describing the phase difference dynamics between the
two modes. Both the self- and cross- saturation effects and the backscattering effect over both
microscopic and macroscopic scale have been taken into account, where with microscopic
scale we mean the effect of the backscattered wave on the gain medium, while as
macroscopic scale the effect of the backscattering on the Maxwell's equations is considered.
By our model we can evaluate all the physical coefficients of the rate equations by means of a
full quantum mechanical analysis. In fact, starting by the energy band information of the
MQW structure and assuming typical values for the time constants of the electron scattering,
our model is able to evaluate the linear and non linear gain coefficients by using a compact
density matrix formalism. Then, these parameters are put inside the rate equations to calculate
the dynamic evolution of the oscillating modes in the ring resonator. By the study of this
dynamics, the investigation of any ring laser operating regime, bidirectional or unidirectional,
should be possible.
Fiber Ring Lasers: An Overview
Fiber ring lasers have attracted tremendous interest because of their many important
applications in fiber-optic test and measurement [1], fiber communications [2], fiber sensor
systems [3], and high-resolution spectroscopy. This is mainly due to their features, such as
wide tunable range, narrow linewidth, and tuning at high speed allowing fast component
characterization. However, despite of these advantages, the fiber ring lasers present a number
of drawbacks essentially related to the possibility of multimode operation, which induces the
beat-noise generated as a result of beating effect between the lasing longitudinal modes and
could severely limit the lasers applications.
Advances in Physical Modeling of Ring Lasers 163
Differents methods to realise a fiber laser have been proposed in recent years. All the
experimental architectures can be classified in two main groups: fiber lasers based on
nonlinear effects, and fiber lasers based on doped fiber amplifiers. Basically, both ring and
Fabry-Perot geometries have been used to realise the resonant cavity in which a piece of
nonlinear or doped fiber as active medium must be inserted. Nonlinear effects play an
important role in high power fiber lasers. They are induced by the large amount of power
density in the small area fiber core. Among the nonlinear effects, the Stimulated Raman
Scattering (SRS) and the Stimulated Brillouin Scattering (SBS) have shown significant
potentials.
The fiber-Raman lasers not only have a lower threshold compared with the single pass
SRS, but also they can be tuned over a wide frequency range (10 THz). Fig. 1 schematically
shows a typical fiber Raman laser [4].
Fiber
Pump
Pump
M
1
M
2
Prism
Lens Lens
Figure 1. Schematic architecture of a tunable fiber Raman laser.
A piece of single-mode fiber is placed inside a Fabry-Perot cavity formed by the partially
reflecting mirrors M1 and M2. The cavity provides a resonant, wavelength-selective feedback
for Stokes light generated inside the fiber by means of the SRS effect. The intracavity prism
is needed to tune the laser wavelength by spatially dispersing various Stokes wavelengths
which can be selected by rotating the mirror M2. It is evident that the laser threshold
corresponds to the pump power at which the Stokes amplification during a round trip is large
enough to balance the cavity losses. As mentioned in [4], a threshold level of about 1 W can
be obtained using a fiber length of about 10 m. By adding separate mirrors for each order of
Stokes waves, the fiber-Raman laser can be operated at several wavelengths simultaneously,
each of which can be independently tuned by tuning the mirrors [5]. Several experiments
have demonstrated that it is relatively easy to achieve synchronization in fiber-Raman lasers.
The reason is that the laser can select a particular wavelength satisfying the synchronous-
pumping requirement among the wide range of possible wavelengths near the peak of the
Raman-gain spectrum. Moreover, the laser wavelength can be tuned by simply changing the
cavity length. This technique is referred to as time-dispersion tuning to distinguish it from
prism tuning based on spatial dispersion provided by the prism. Thus, synchronously pumped
fiber Raman lasers have attracted attention for generating ultrashort optical pulses. In
addition, if the Raman pulse falls in the anomalous group velocity dispersion (GVD) regime
of the fiber, the soliton effects can create pulses with widths of about 100 fs or less.
Similar to the SRS case, the Brillouin gain in optical fibers can be used to make fiber-
Brillouin lasers by placing the fiber inside a cavity. To this aim, both ring cavity and Fabry-
Perot geometries have been used. Fiber-Brillouin can lead to obtain a threshold input power
of 0.56 mW by using an all-fiber ring resonator, as shown schematically in Fig. 2 [4]
Fiber-Brillouin lasers consisting of a Fabry-Perot cavity exhibit features which are
qualitatively different from those of a ring cavity configuration. The difference arises from
the simultaneous presence of forward and backward propagating components associated with
the pump and Stokes waves. The simultaneous presence of many equi-spaced spectral lines in
the output of a fiber-Brillouin laser indicates the possibility of obtaining ultrashort optical
pulses if the laser can be mode-locked. Thus, an intra-cavity modulator could be useful to
realise the mode-locking process [4].
However, Brillouin backscattering can be the limitation of the output power for narrow-
band signals; and Raman scattering can generate a frequency shift which decreases the pump
power and signal power. For these reasons, a great attention has been reserved to fiber lasers
based on doped fiber amplifier.
Pump
Laser
Spectrum Analyzer
Lens
Directional
Coupler
Fiber ring
Polarization
controller
Figure 2. Schematic architecture of a fiber-Brillouin ring laser.
The erbium-doped fiber amplifier (EDFA) was invented by D.N. Payne and co-workers
in 1987 [6]. The EDFA commercialization made long haul optical communications
inexpensive and reliable and optical fiber became the standard for long-haul
telecommunication systems. However, in situations where high-power lasers are needed, an
EDFA does not work very well because the high power density damages the fiber. Nd-doped
fiber lasers (NDFL) and Yb-doped fiber lasers (YDFL) were developed for scaling output
power of fiber lasers. Neodymium can be pumped at 808 nm to get good absorption, while
ytterbium can be pumped at 975 nm. Both of these elements can emit light at around 1060 nm
with slightly different energy transition mechanisms. Initially, YDFLs attracted less attention
with respect to NDFLs essentially because Nd
3+
has the advantage of a four-level pumping
scheme, while Yb
3+
works with a three or quasi four-level scheme. A four level laser system
tends to easier lasing because it has a lower threshold. However, ytterbium offers higher
power conversion efficiencies and larger output powers. In fact, even if its energy level
structure and the re-absorption effect make the threshold pump power relatively high,
ytterbium does not have self-quenching effects [7]-[8] as neodymium, and it can have a
higher ion concentration. Moreover, ytterbium can be more efficient due to the small quantum
defects.
Several theoretical works have been presented in literature to model and design fiber ring
lasers. Generally speaking, these models are mainly focused on the study of the gain medium
(in case of doped fiber lasers) and on nonlinear effects that origin in the ring cavity. In the
former case, the mathematical modelling is constituted from a system of rate equations
describing the time evolution of the carrier population and photon densities. These two kinds
of rate equations contain complete information about the dynamics of the laser system. Since
they are non linear coupled equations, a straightforward analytical solution is not feasible. A
numerical solution of these set of equations is presented in [9]. They can be analyzed under
certain approximations corresponding to real physical situations and an analytical solution is
possible in some cases [9].
A generalization of the mathematical model to the case of multi-frequency erbium-doped
fiber ring lasers employing a periodic filter and a frequency shifter has been also presented
[10]. In this work, the rate equation systems describing the laser dynamics are coupled with
three spatial differential equations related to the pump propagation inside the fiber ring and
both forward and backward propagating amplified spontaneous emission (ASE) powers. An
iterative solution of the rate equations and propagation equations for two counter-propagating
ASE powers and pump has been implemented using a fourth-order RungeKutta routine in
which appropriate boundary conditions were been imposed at beginning and end of the active
fiber, according with the experimental setup.
More recently, a steady-state model has been proposed to analyse and design a quasi-
continuous wave injection-seeded ring laser [11]. As it is known, frequency stabilized laser
diodes can achieve a high degree of wavelength stability and are well-suited as master
oscillators for spectroscopic lidar systems. However, diode lasers alone do not have sufficient
power to provide for a high signal-to-noise ratio in these systems. One method to efficiently
scale the system power is to use the stabilized laser as a seeder in an injection-locked fiber
ring laser.
Two factors can limit the performance of seeded ring lasers. First, the spectral overlap
between the seed laser and a ring laser cavity mode. The use of a low-finesse ring cavity has
been shown to relax this spectral requirement making the seeding process relatively easy. The
second issue is that the laser self oscillation can dominate the seeding process. Self-seeding
can be overcome through the incorporation of intracavity filters or using sufficient seed
power. The latter condition has been assumed in a model proposed in literature [11].
However, it is worth noting that the above cited mathematical models do not require
heavy computational efforts, essentially because they do not include optical nonlinear effects,
that can originate in the ring fiber in high power regime. Generally, the nonlinear effects in
fiber ring lasers can be modelled by means of the nonlinear Schroedinger equation (NLS) [4].
In particular, the NLS equation represents a powerful mathematical tool not only to design the
fiber-Raman or fiber-Brilluoin lasers [4], but also to analyse the wave behaviour of an optical
fiber exhibiting a weak Kerr nonlinearity. The NLS has been extensively investigated in this
context, with particular emphasis given to the robust and stable soliton solutions that result
from a fundamental balance between linear dispersion and cubic nonlinearity [4]. Thus,
soliton pulses are ideal carriers for transmitting optical data. For applications for which
polarization effects are important, one must consider a system of coupled NLSs, that is
generally not integrable. Despite the increased numerical complexity, in order to manage
soliton like solutions it is not only useful to optimise the design of long-haul communications
systems [12], but also to develop efficient optical fiber ring lasers [13]-[16]. For the ring laser
configuration, Kerr nonlinearity of the birefringent optical fiber generates a nonlinear rotation
of the polarization state that depends on the pulse intensity. Then, the insertion of a passive
polarizer provides an effective intensity filter that stabilizes, or mode locks, a propagating
pulse by periodically attenuating all components of the pulse that are not aligned with the
polarizer. Simple devices such as these have been shown experimentally to generate stable
and robust soliton-like pulse trains, that can be used for a wide variety of telecommunications
purposes [13]-[17].
Recently, an advanced mathematical model has been proposed in literature [18] for fiber
ring lasers. This model includes in the standard NLS equations two key modeling elements
that describe the mode-locking dynamics: the nonlinear polarization rotation induced by
cross-phase modulation, and the polarization control through the passive polarizer. In
particular, the theoretical model consists of the coupled NLSs with periodic perturbations that
are due to the polarizer. In deriving this model, the contributions from continuum radiation
have been neglected. This is due to the filtering function of the passive polarizer in the mode-
locking process. Although a residual amount of radiation is expected from the periodicity of
the cavity, it remains negligible in comparison with the energy in the localized mode-locked
pulse. In addition, a source of amplitude fluctuations arising from the interplay between
nonlinearity and polarization control has been considered. In conclusion, the mathematical
model proposed in [18] represents a useful theoretical tool to examine the underlying mode-
locking mechanism of the fiber ring and to describe the systematic and predictable amplitude
fluctuations that result from this interaction.
As mentioned above, the main drawback in the fiber ring laser is related to the multimode
longitudinal behaviour. This disadvantage has induced many research groups to investigate
experimental solutions and search different configurations to optimise the fiber ring laser
performance. A number of studies have been presented in literature to reduce the beat noise,
and to realise a fiber ring laser with single-longitudinal-mode (SLM) operation. For example,
a compound-ring cavity to reduce the beat-noise has been proposed [19] where, due to the
short ring-length, the dual-coupler fiber ring acts as a small free spectral range (FSR) etalon
filter and, combined with the tunable optical bandpass filter, selects one longitudinal mode.
However, this proposed solution needs PZT to accurately control the length of the cavity.
Recently, the research on SLM operation using saturable absorbers has been reported to
overcome the limited spectral width of the mode selection filters [19]. Additionally, the
longitudinal lasing mode become unstable when the fiber ring cavity and the filter are highly
sensitive to the temperature drift and other external disturbances. To prevent the unstable
operation, the fiber lasers of various schemes have been reported so that length of tunable
filter and fiber ring cavity were stabilized [20]-[21], or a saturable absorber was used to
suppress multimode operation [22]-[24]. As it is known, a large free spectrum range FSR is
also required to facilitate the SLM operation. In this sense, an architecture based on S-band
ring lasers has been recently proposed [25]. Fig. 3 shows the proposed configuration and
experimental setup of an S-band erbium fiber laser with a triple-ring cavity resonator.
The architecture is constituted by the main ring (Ring-1), coupled with two rings having
different lengths (Ring-2, and Ring-3) by means of two 50:50 optical couplers (C). The Ring-
1 is composed of an S-band EDFA module, a 90:10 optical coupler (C
1
) to couple out the
optical beam, a fiber Fabry-Perot tunable filter (FFP-TF), and two polarization controllers
(PCs). It is worth noting that the presence of two PCs lead to align the state of polarization of
the Ring-1 cavity to guarantee a SLM oscillation. In addition, the FFP-TF is necessary not
only to determine a lasing wavelength, but also it serves as a mode-restricting component to
provide the first restriction on the possible laser modes.
C
980 nm
Pump LD
W W
C
C
C
1
PC PC
FFP-TF
S-Band EDFA Module
Output
Ring 3
Ring 2
Ring 1
EDF
Isolator
Figure 3. Experimental setup of an S-band erbium fiber laser with a multiple-ring cavity structure.
Finally, the S-band amplifier, constituted by three isolators, a pump laser at 980 nm, and
two erbium doped fibers (EDF) with a saturated output power of 16.1 dBm at 1498 nm, has a
depressed-cladding design and a power-sharing 980 nm pump laser to generate EDF gain
extension effect [26].
The behaviour of this architecture is based on Vernier effect. In fact, indicating with
FSRm and FSRs

the free spectral range of the main and sub-ring cavities, respectively, the
value of effective FSR becomes the least common multiple number of both FSRm and FSRs.
As a result, the mode suppression can be achieved and governed by the length of the main-
ring and sub-ring cavities. Thus, the laser mode oscillates only at a frequency that
simultaneously satisfies the resonant conditions of main cavity and all the sub-ring cavities.
Due to the combination of a FFP-TF with a triple-ring cavity, a SLM selection in this fiber
laser is successfully achieved. The polarization state of proposed laser should be maintained
by adjusting the PCs as that of the Ring-1.
The experimental results performed by means of an optical spectrum analyzer (OSA)
with a 0.05 nm resolution [25], indicate that an output power larger than 0 dBm, a power
stability less than 0.05 dB, a wavelength variation less than 0.02 nm and a side-mode
suppression ratio (SMSR) larger than 54.3 dB / 0.05 nm can be obtained over an operating
range of 1481 to 1513 nm.
As mentioned before, one of the main problems considered in the fiber ring laser is the
beat-noise. Recently, a novel method to suppress the beat-noise fiber ring laser using a Fabry
Prot laser diode (FP-LD) has been proposed in [27]. As reported there, the beat-noise of
fiber ring lasers is primarily in the low-frequency region of about 10 MHz due to the long
ring cavity length, which is typically of the order of several tens of meters. Thus, basically an
injected FP-LD can act as a high-pass filter to suppress the low-frequency beat-noise of fiber
ring lasers [28]. This is mainly due to the fact that the FP-LD has a fast carrier recovery rate
(1 ns) and experiments the gain saturation effect. Fig. 4 shows the experimental set-up used
[27].
Pump
Fiber Fabry-Perot
Filter
FP-LD
Circulator 1
Circulator 2
Output
3 dB Coupler
Angled Splices
Fusion Splice
Bi-EDF
A
B
1
2
3
3
2
1
Figure 4. Configurations of the highly polarized, low beat-noise, tunable fiber ring laser.
The architecture is constituted by an Bismuth oxide-based EDF (Bi-EDF) [29] pumped
by one 1480-nm semiconductor laser diode via Port 1 of an optical circulator (circulator 1),
which fairly exhibits a flat pass-band in the wavelength range from 1460 to 1630 nm. The
length of the Bi-EDF was of 51.4 cm. In the experiments, the refractive index of the Bi-EDF
core and cladding were 2.03 and 2.02, while the diameters of core and cladding were 3.9 and
124.7 m, respectively. The erbium concentration in the Bi-EDF was 6500 wt ppm and Boron
and Lanthanum are co-doped in the Bi
2
O
3
-based fiber to increase the pump efficiency. The
Bi-EDF peak absorption measured at 1480 and 1530 nm were 141 and 219 dB/m,
respectively. Both ends of Bi-EDF were first angle spliced to high numerical aperture fiber
(Nufern 980-HP fiber) before splicing to Port 2 (SMF-28 fiber) of Circulator 1 and to Port 3
(SMF-28 fiber) of Circulator 2, providing better mode field diameter matching. The splicing
loss attained was less than 0.2 dB for the angled splices. The angled splices reduce the
reflection in the laser cavity to less than 60 dB.
The set-up presents also a large FSR fiber FabryProt (FFP) filter employed to tune the
laser wavelength and a double-channel planar-buried hetero-structure FP-LD with cavity
mode spacing and threshold current of 1.12 nm and 10.9 mA, respectively. The optical
bandwidth and FSR of FFP filter were about 50 pm ( 6.25 GHz) and 110 nm, respectively.
Thus, since the gain bandwidth of Bi-EDF is less than 110 nm [14], only one wavelength in
the fiber ring laser cavity is excited. The FP-LD was biased at 12 mA, slightly above the
threshold current, to realize the low beat-noise laser output. Finally, a 3-dB fused fiber taper
was included in the laser cavity to provide the laser output.
The experimental results [27] show that the configuration described in Fig. 4 leads to
obtain high performance, including a fiber ring laser wavelength tuning from 1536.82 to
1570.72 nm in 1.12 nm steps, with a maximum output power of about +3 dBm. The
polarization degree and extinction ratio of the laser output are about 99% and 60 dB,
respectively. Finally, the beat-noise, with and without FP-LD, was dramatically reduced by
50 dB.
In parallel to the amount of research focused to realise SLM fiber ring laser, a number of
studies has been proposed about multi-wavelength erbium-doped fiber lasers (MW-EDFLs).
They have attracted considerable interest for potential applications in optical test and
measurement, and optical wavelength-division-multiplexing communication and sensing
systems [30]. Compared with compact semiconductor-based lasers, EDFLs are competitive
because of their all-fiber structure, as well as their capacity to provide high power and
ultrashort pulse width [31]. They can be used in applications that require multiple wavelength
sources, with small equal-wavelength spacing, a large number of peaks within a broad band,
and high output uniformity across the channels. These requirements pose a very challenging
task for building a cost-effective multi-wavelength EDFL for continuous-wave or pulsed
operation. Previously, due to the homogeneously broadened gain property, many MW-EDFLs
were developed with wavelength spacing larger than homogeneous linewidth of about 3.5nm,
to overcome gain competition. Many different approaches have also been explored for
developing MW-EDFLs, including use of polarization or spatial hole burning, use of
independent gain media, frequency shifting, and phase modulation [32]-[36]. However,
recently a novel room-temperature-operated MW-EDFL with wavelength spacing less than
the homogeneous broadening linewidth, based on interchannel four-wave mixing (FWM), has
been proposed [37]. The EDF gain-clamping effect is compensated by the parametric four
wave mixing (FWM) between multi-wavelength channels in a highly nonlinear fiber section
that is inserted into the fiber ring cavity and it is based on highly nonlinear photonic crystal
fiber (HNL-PCF). Finally, sampled-fiber Bragg grating (SFBG) is used in input to one port of
the circulator (see Fig. 5).
EDFA
Sampled FBG
Circulator
10% Output
Polarization controller
HNL-PCF
Output coupler
Figure 5. Schematic diagram of the multi-wavelength erbium-doped fiber laser.
The multiple wavelength operation has been initiated, by adjusting the polarization
controller and a specific sampled fiber Bragg grating (SFBG), with 0.8nm wavelength
spacing. The architecture uses the inter-channel four-wave mixing-induced dynamic gain-
flattening mechanism to stabilize the output. Thus, the FWM processes created by means of
the HNL-PCF suppress the EDFL homogeneous line broadening and stabilize the multiple
wavelength oscillation. By tuning the intracavity polarization controller and then the FWM
efficiency, the number of concurrent lasing wavelengths can be changed from two to five, and
the peak power differences for the main oscillation wavelengths are less than 2.0 dB.
Channel spacing of 0.5nm operation of 10GHz dual-wavelength mode locking with the
help of fiber birefringence, has been obtained with cavity structure with 60 m HNL-PCF [37].
In addition, a supermode suppression ratio higher than 60dB, and a time-bandwidth products
ranging from 0.39 to 0.41, have been measured.
Semiconductor Ring Lasers
Nowadays, multi-quantum-well (MQW) semiconductor ring lasers are of great interest for
monolithic and integrated optoelectronic applications [38]. Since these lasers do not require
the use of cleaved facets or difficult Bragg gratings for optical feedback, they can be
conveniently integrated so reducing the occupation area. Moreover, the semiconductor ring
lasers offer new capabilities in the travelling-wave unidirectional oscillation. Unidirectional
operation is desired because it offers the advantages of enhanced mode purity (high side-
mode suppression ratio), reduced sensitivity to feedback and higher single beam power.
Unidirectional ring lasers have been used for feedback laser diodes [39]. Other important
applications of semiconductor ring lasers include multi-wavelength [40] and all-optical flip
flop operation [41-42]. A number of experimental studies have been also carried out to
analyse the operation regimes of the semiconductor ring lasers [43-45].
The main drawback of the semiconductor ring lasers could be represented by the instable
regime of behaviour (switching between unidirectional and bidirectional). Thus, this physical
situation needs an accurate physical model to individuate the design criteria in order to realise
stable unidirectional regime.
In this sense, recently it has been experimentally demonstrated that GaAs-AlGaAs ring
lasers show either bidirectional or unidirectional regimes depending on the injection current
[45]. In particular, bidirectional operation reveals that just above threshold the ring laser
operate in a regime where the two counter-propagating modes are continuous waves. As the
injected current is increased, a new regime appears where the intensities of the counter-
propagating modes undergo alternate sinusoidal oscillations. Finally, for injection currents
larger than a critical value, the unidirectional regime appears in a stable way. Therefore, the
control of the unidirectional operation is of fundamental importance in order to use the
semiconductor ring laser as an integrated source. In this chapter, we propose a very accurate
physical model to analyse the operating regimes of a MQW semiconductor ring laser.
Similarly, a physical model has been recently presented to analyse the operating regimes of
any MQW semiconductor ring laser [46], as a strong generalization of the models previously
proposed in literature. In fact, it can be applied to a generic MQW semiconductor structure,
not only to a two level gas system [47][48] (for gas laser), and it considers the macroscopic
effect of the backscattering non included in the model proposed in [47][48]. In addition, the
model in [46], differently by [45], leads to calculate all the physical parameters without any
semi-empirical approximation, taking into account the backscattering effect on the gain
medium. It is based on four differential equations in total, two coupled equations for the
counter-propagating modes, one rate equation for the carriers injected in the laser active
region and one equation describing the phase difference dynamics between the two modes.
Anyway, in the following section the theory of the physical model proposed in [46] is
briefly summarized with the aim to introduce some recent advances of this numerical model.
In particular we have investigated the influence of some technological parameters (ring
sidewall roughness, ring radius) over the backscattering coefficient influencing the operation
regime and performance of any MQW semiconductor ring laser.. Then, we present some
numerical results and simulations applied to a GaAs-AlGaAs MQW semiconductor ring laser.
In that section we will show as either unidirectional or bidirectional regime is related to the
ratio between the injection current and the backscattering coefficient value, i.e. technological
considerations have been summarized. Finally, simulation results relevant to an architecture
of the ring laser with asymmetrical output coupler to obtain a stable unidirectional regime are
presented.
Theory
The theory is based on the semi-classical interaction between radiation and matter. Then, the
atomic systems are modelled as quantum phenomena while the electromagnetic (e.m.) field is
classically described by the Maxwells equations. In particular, the electric dipole e
r
operator
relates the system quantum-mechanical description with the polarization P of the medium
used as a source of the e.m. field.
Assuming a predominant single transverse mode as electric field inside the ring laser, we
can write:
( ) ( )
( ) ( )
( ) E , F . .
n n
j t t
n
n
t E t e c c
+
= +
r r (1)
where c.c. indicates the conjugate complex terms, ( )
n
E t is the electric field amplitude,
n
is
the angular pulsation of the optical mode inside the MQW ring cavity and ( )
n
t is the time-
dependent phase of the electric field. In general, the field function ( ) F r can be written as
( ) ( , ) ( ) F r G x y U z = , where ( , ) G x y and ( ) U z indicate the transverse and longitudinal profile
of the electric field, z representing the propagation direction (curvilinear coordinate). The
subscript n takes into account all possible longitudinal modes in the ring cavity. With this
representation for the field, the polarization P induced by the gain medium is given by:
( ) ( )
( ) ( )
( ) P , F .
n n
j t t
n
n
t P t e cc
+
= +
r r (2)
where ( )
n
P t is the complex, slowly-varying component of the polarization of the n-th
longitudinal mode. The wave equation for the time evolution of the electric field is given by:
2 2
2
0 0
2 2 2
1
t v t t

+ + =

J E P
E (3)
where v is the velocity of the electric wave in the ring resonator. The second term is included
to take into account the cavity losses. In particular the current density is expressed in terms of
the fictional conductivity = J E, where the conductivity is assumed as the sum of two
contributions,
1
and
2
. The former includes all kind of optical losses (propagation loss,
radiation loss, leakage loss, etc..) of the n-th longitudinal mode. It is related to the quality
factor
n
Q of the cold cavity by means of the relationship ( )
1 n n
Q = , where is the
permittivity of the cavity. The latter contribution is included to take into account the effect of
the backscattering induced by the cavity sidewall roughness. According to [49], the
backscattered wave induces a contribution to the fictional conductivity given by:
2
b = ,
where b is the backscattering rate. This rate coefficient is related to the amplitude
reflectivity R due to the backscattering as ( )
eff eff
b cR n R = , being c the free-space light
velocity,
eff
n the effective index of the optical wave in the ring cavity and
eff
R the ring
effective radius.
With the aim to study the dynamics of the two counter-propagating modes, we can
particularise Eqs. (1)-(2) to only two modes, one clockwise (CW, as mode 1) and the other
counter-clockwise (CCW, as mode 2). Under slowly-varying amplitude and phase
approximations, extensively used in the laser dynamics modelling, the wave equation (3)
produces the following set of equations:
( ) { }
1 2 1
1 1 2 2 1
1 1
1 1
cos Im
2 2 2
E E bE P
Q

= + +
(4)
( ) { }
2 1 2
2 2 1 1 2
2 2
1 1
cos Im
2 2 2
E E bE P
Q

= + +
(5)
{ } ( )
1 2 2
1 1 2 1 1
1 1 1
1 1
Re sin
2 2
E
P b
E E

= + +
(6)
{ } ( )
2 1 1
2 2 1 2 1
2 2 2
1 1
Re sin
2 2
E
P b
E E

= + +
(7)
where
2 1
= and
n
are the eigen-frequencies of the cold cavity eigen-modes. Eqs.(4)-
(7) are the Lamb's self consistency equations and take into account the effects related to the
gain medium and the macroscopic effect of the backscattering. The system (4)-(7) can be
solved when the polarization vector is known.
This model leads to evaluate the linear and nonlinear terms of P by performing quantum
calculations applied to the MQW semiconductor structure. By this way, it is possible to
calculate all physical parameters, involved in the ring laser dynamics, starting from the
physical description of the MQW energy band without any semi-empirical approximation. To
evaluate the P vector we use the density matrix formalism [50] to write:
( )
,
Tr( )
ba ab ab ba
b a
P n M n M M = = +
(8)
where Tr is the transposte of matrix and is the density matrix operator, given by:
aa ab
ba bb

=

being
aa
the probability to have an electron in a state of a BV (valence band) sub-band,
bb
the probability to have an electron in b state of a BC (conduction band) sub-band,
ab
and
ba
the probabilities to have a transition between a and b levels or b and a levels,
respectively (it holds
*
ba ab
= ). M is the dipole moment operator (formed by the electron-
hole pair relevant to two sublevels of the same order, one in BC and the other in BV) in the
form of a 2x2 matrix as:
0
0
ab
ba
M
M
M

=

ab
M ,
ba
M have been calculated as in [36]. In Eq. (8), n denotes the electron total density
included in the conduction and valence sub-bands which must verify the following
relationship for MQW semiconductor lasers:
( ) ( ) ( )
(0) (0)
,
g
bb aa cv ba c ba v ba ba
b a
W
n g W f W f W dW

=

(9)
being ( )
cv ba
g W the state density, W W W
ba b a
= the energy difference between level b and
a , f
c
( f
v
) the Fermi-Dirac distribution function at level b and a , respectively.
By solving the continuity equation as proposed in our previous work [46], we derive the
elements of the density matrix and, then, the polarization vector as a sum of one linear and
one non linear terms. By substituting the components of the polarization vector obtained by
the quantum-mechanical analysis in Eq. (4)-(5), considering only two modes, one clockwise
(CW, 1) and the other counter-clockwise (CCW, 2), rearranging the equations in terms of the
mode intensities
2
1 1
I E =

and
2
2 2
I E = , and assuming
1 2
= , we obtain the rate equations
for any semiconductor MQW ring laser:
( )
( ) ( )
1
1 1 1 1 12 2 1 2 1 1 12 2 1 2
2 2 cos
tot
dI
I I I I I I I b I I
dt
= + + +

(10)

( ) ( ) ( )
2
2 2 2 1 21 1 1 2 2 2 21 1 1 2
2 2 cos
tot
dI
I I I I I I I b I I
dt
= + +

(11)
with ( )
2 1
1
2
tot
= + , ( )
1 2
1
2
= + and:
4
2
i
i iiii iiii
M
= N I

4
2
i
ij iijj iijj ijji ijji ijij ijij
M
= + +

N I N I N I

4
2
i
i iiij iiij iiji iiji ijii ijii
M
= + +

N I N I N I

4
2
i
ij ijjj ijjj
M
= N I
1, 2 i = and j i
where ( ) ( ) ( )
{ }
Im
nqkm
g
j
nqkm cv ba c ba v ba qkm ba
W
g W f W f W je dW

=

I C ,
* *
( )
I
n q k m
V
nqkm
n
F r F F F dr
=

N
N
is a real quantity (field overlapping integral,

*
( ) ( )
n n n
F r F r dr =
N ), which is not zero just in the active region volume, and

2
n
n n
n
g
Q
=
is the net linear gain of MQW structure, being
n
g the gain of the MQW structure. The
parameters
nqkm
and
qkm
C are defined in [31].
The introduction of the variable in Eq. (10)-(11) avoids the explicit dependence on the
individual scattering phases
1
and
2
, and, then, all physical quantities will depend only on
the average value . The equation system (10)-(11) has a stable solution as
tot
= 0 when
= and
1 2
= . The coefficients included in the model, i.e.
i
,
ij
,
i
and
ij
, represent
the self-saturation coefficient, the cross-saturation coefficient, the self interference coefficient
and the cross interference coefficient, respectively. They are responsible of the mode
competition phenomenon [46]. Then, by manipulating Eq. (6)-(7), we obtain:
( ) ( ) ( ) ( )
( ) ( )
( ) ( )
2 1 2 1 2 1 2 2 1 1
1 2
21 1 12 2 1 2 21 12 21 1 12 2
2 1
2 1
1 2
sin sin
2
tot
tot tot
d
I I
dt
I I
I I I I I I
I I
I I b
I I

= +

+ + +

+ +

(12)
with
2
i
i iiii iiii
R
= N
2
i
ij iijj iijj ijji ijji ijij ijij
R R R
= + +

N N N

2
i
ij iiij iiij iiji iiji ijii ijii
R R R
= + +

N N N

2
i
ij ijjj ijjj
R
= N
where
1
and
2
are the cold cavity frequencies for CW and CCW mode, respectively,
1
,
2
are the laser beam frequencies,
1
,
2
are the pulling effect coefficients,
1
,
2
are the
self-pushing effect coefficients,
12
and
21
are the cross-pushing effect coefficients. The
terms
12
,
21
,
12
, and
21
are nonlinear coefficients induced by the nonlinear
contribution in the polarization vector. Moreover, it holds:
( ) ( ) ( )
{ }
4
Re
nqkm
g
j
nqkm cv ba c ba v ba qkm ba
W
R M g W f W f W je dW

=

C
and
( ) ( ) ( )
( )
( )
2
2
2
2
( )
2
I
g
V i ba i
i cv ba c ba v ba ba
m W
i ba
in
F r dr
W W
M g W f W f W dW
h
W W
=

/
+

N
being
in
the electron average relaxation time.
In order to complete the ring laser model, we have introduced the classical rate equation
for the injected carriers, in the form:
( )
( )
( )
( )
2
0 eff 2 3
1 1 1 1 12 2 1 2 1 1 12 2
1
2
0 eff
2 2 2 2 21 1 1 2 2 2 21 1
2
2 n
dN J
= -AN-BN -CN - I g - I - I - I I I + I
dt ed h
2 n
- I g - I - I - I I I + I
h

(13)
where
o
is the vacuum dielectric permittivity, d is the active region thickness, J is the laser
current density, n is the refractive index, A, B, C are the leakage recombination coefficient,
the bimolecular recombination coefficient and the Auger coefficient, respectively. Eqs. (10)-
(13) are the coupled differential equations for the two counter-propagating modes inside any
MQW ring laser.
It is worth noting that the self and cross interference coefficients are equal to zero only in
absence of any backscattering (i.e. R = 0) inside the laser cavity. In this ideal case Eqs (10)-
(11) include only the self- and cross-saturation coefficients and, then, the only stable regime
is the unidirectional one. In the following section we will show that the backscattering effect
can be also responsible of a stable bidirectional regime, depending on R values.
Since the backscattering effect depend on technological effects like the ring sidewall
roughness, the goal of our model is also to evaluate the operating regime of the MQW ring
laser related to the statistical information about this sidewall roughness. The sidewall
roughness or boundary imperfections can have two significant effects: 1) energy scattering
towards the radiation field, reducing the total quality factor of the optical mode; 2) power
redistribution between the two counter-propagating waves influencing the operational regime
of the ring laser. Differently from [45], where the backscattering effects are estimated in an
empirical way, we calculate the R coefficient and the influence of the scattering losses by
means of an analytic approach [52]. Thus, the sidewall imperfections can be described by a
random function having a Gaussian distribution for its self-correlation function as:
( )
2
' 2 ' 2
( ) exp /
c c
corr s s s s L

=

(14)
where s is the curvilinear coordinate,
c
L is the correlation length and
c
is the standard
deviation of the roughness correlation function.
We have analyzed the scattering loss due to sidewall imperfections by using the volume
current method [52]. This method consists of the calculation of the current density associated
to the sidewall roughness profile and, thus, solving the Maxwells equations in presence of
this current source. Since the wave electric field has only components along the r and
directions in the plane of the ring (for TE polarization), we concentrate our attention on the
dominant component E
, because it has a peak at the ring edges. This component is

influenced by the ring cross-section and radius. We have used a mixed numerical technique
based on both effective index, conformal mapping and Wentzel-Kramers-Brillouin (WKB)
methods to take into account these influences [52]. Thus, the electric field of the wave
travelling along the ring excites an additional contribution to the current density in the regions
as perturbed by the presence of roughness, and this contribution will induce the field
radiation, calculated by the azimuth component of the vector potential
A . The determination
of vector potential has been executed by evaluating the free-space Greens function. Then, we
have determined the power density radiated by means of the radially directed Poynting vector.
Both types of radiations have been considered, i.e. tunneling radiation and phase-matched
radiation. However, for large resonator radii (>50 m), we have found the phased-matched
radiation as the only significant contribution to the scattering loss. Now, we have evaluated
the scattering-related quality factor of the ring resonator,
scatt
Q . It is defined as the ratio
between the stored energy and the power lost by scattering per each round trip, in the form:
2 exp( / 2)
1 exp( )
eff eff scatt eff
stored
scatt
lost scatt eff
n R R
P
Q
P R

= =

(15)
by which the scattering coefficient
scatt
has been calculated as:
2
1
2
sinh
eff eff
scatt
eff scatt
n R
R Q

=

(16)
and the total quality factor of the ring cavity has been estimated as:
4
2 1 exp( / 2)
1 1 exp( )
eff eff total eff
n
total eff
n R R
Q
R

=

(17)
where the parameter is the coupling efficiency between the ring laser and an output bus
waveguide and the total optical losses in the ring laser is given by
total scatt prop bend leak
= + + + , being
scatt
the sidewall roughness scattering loss,
leak
the
leakage loss to the substrate,
prop
the propagation loss and
bend
the curvature-induced
bending loss. The leakage loss to the substrate is negligible by introducing in the ring laser
structure a buffer layer. The bending loss coefficient can be considered negligible due to the
strong confinement in the ring resonator and to its large radius (>50 m). Thus the optical
losses are mainly dominated by
prop
and
scatt
.
Finally, the current induced by the sidewall roughness is a field source inducing a power
transfer between the two counter-propagating waves with a backscattering amplitude
reflectivity R given as [52]:
( ) ( )
2 2
2
2 2
0 0
2 / 4 exp
eff rib c eff c
R R n t F k n L

=

(18)
where
2
n is the ring-air relative permittivity change,
rib
t is the ridge overall height of the
ring cavity, F
is the azimuth component of the normalized electric field travelling inside the
ring cavity. Therefore, the backscattering coefficient is related by our model to the ring laser
technological parameters, i.e.
2
n ,
c
L ,
c
,
rib
t ,
eff
R .
Numerical Results
The numerical simulations have been performed by considering a standard GaAs-AlGaAs
MQW ring laser structure. The active region is constituted from one (or more) GaAs wells
sandwiched between two Al
0.2
Ga
0.8
As waveguide regions, each 100 nm thick. The p-type and
n-type Al
0.4
Ga
0.6
As cladding layers are 1.0 and 1.5 m thick, respectively. A top GaAs cap
has been also included. A total optical loss of 25 cm
-1
has been taken into account. We have
assumed a ring radius of 200 m (small ring), a rib width of 2 m and considered the
backscattering coefficient as a parameter. By this way we can show as the physical operation
of the MQW ring laser depends on the relationship between the injection current and
backscattering coefficient. Anyway, it is possible to calculate the backscattering coefficient
starting from the statistical information of the ring sidewall roughness taken from
experimental data, as explained by Eq. (18).
Analysing the stationary solution of Eqs. (10)-(11), the relationship
1 2
0, + = must be
satisfied. In particular
1 2
0 + = gives rise to a condition of minimum stimulated energy,
whereas
1 2
+ = gives rise to an instable condition of maximum stimulated energy and
can be discarded.
We have investigated the impact of the output coupler configuration [46] on the operating
characteristics of the semiconductor MQW ring laser. We have assumed an evanescent field
coupler, a very common element in integrated optics. However, one of the main problems
with these couplers is the sensitivity of the coupling efficiency to the changes of coupler
dimensions, particularly the coupling gap. Such variations make it difficult to accurately
obtain a given coupling efficiency, a high reproducibility and good stabilization of the ring
laser operating regime. Fig. 6 shows the stationary regimes of
1
I and
2
I versus the coupling
efficiency.
Figure 6. Intensities of both beams and phase difference versus the coupling efficiency.
In this simulation we have assumed
c
L =0.07 m,
c
= 0.012 m and an injection current
of I =100mA. The plot shows that the operating regime of the ring laser becomes
bidirectional by increasing the coupling efficiency, starting from an unidirectional condition.
In fact, for a coupling efficiency ranging from 5% to 16%, the quality factor of the ring
resonator (see Eq. 17) assumes relatively large values, so inducing the current I to be well
greater than the threshold,
th
I . A dominant beam in the ring cavity grows due to the mode
competition effect, as induced by the self- and cross-saturation coefficients
i
,
ij
. For a
coupling efficiency larger than 16%, a significant part of the optical power leaves the ring
resonator and this induces the threshold current to be close to I = 100mA. In this case, it is
not possible for only one of the counter-propagating beams to be completely extinguished. In
fact, the backscattering effect between the beams always induces radiation travelling in the
opposite direction.
Fig. 7 shows the stationary regimes of
1
I and
2
I versus the correlation length
c
L for
different values of the standard deviation
c
, by assuming an injection current value of
I =100mA (one well) and
1 2
0, + = . As usual, mode 1 designates the CW solution and
mode 2 the CCW one, respectively.
Figure 7. Intensities of both beams versus the correlation length for various standard deviations of ring
sidewall roughness function.
The plot shows that for a value of
c
smaller than a critical value, depending of the
injection current (
, c th
=0.0047 m in this case), the MQW ring laser works in an
unidirectional regime without any dependence on
c
L . This means that for each value of
c
L
the backscattering coefficient is too small to compensate the mode competition effect induced
by the self- and cross- saturation coefficients
i
,
ij
(see Eqs. (10)-(11)). Therefore, a

dominant beam in the ring cavity grows while the backscattering effect will always maintain a
very weak wave travelling in the opposite direction, being orders of magnitude below the
dominant beam. The dominant beam can be randomly either the CW or CCW beam,
depending mathematically on the initial conditions or, physically, on the local optical losses
inside the ring cavity. For values of roughness standard deviation
c
>
, c th
, there exists a
range for
c
L where the MQW ring laser shows a bidirectional operating regime. In particular,
this range increases with increasing
c
.
Since the self-correlation function of the sidewall roughness is described as a Gaussian
function (see Eq. (14)), this means that exists a range of
c
L , close to the peak of the Gaussian
shape, where the backscattering coefficient compensates the mode competition effect. In the
range of
c
L where the regime is bidirectional, it is possible to observe a maximum in the plot.
This maximum occurs where the peak of the Gaussian self-correlation function is situated.
The previous discussion is also confirmed in Fig. 8, which shows the intensities of CW
and CCW beams versus the injection current in the stationary condition, for different values
of the statistical parameters of the sidewall roughness.
Figure 8. Intensities of both beams versus the laser current for various roughness correlation lengths.
The plot again shows that the operation regime of the MQW semiconductor ring laser
depends on the relationship between (
c
L ,
c
) and the injection current. If the current is close
to the threshold, i.e.
th
I I , the beams hold the same intensity (bidirectional condition). For
currents well larger than
th
I , a dominant beam in the ring cavity grows inducing the
unidirectional regime. In Fig.8 we have assumed
c
=12 nm and used different values of
c
L ,
larger than the position of the Gaussian peak. We can observe that the range of injection
currents where the MQW ring laser works in a bidirectional regime increases by increasing
the
c
L value. Figs. 7 and 8 are particularly important because they lead to an estimation of the
MQW ring laser operation regime related to the etching step of the fabrication process. In
fact, by performing a number of measurements of the etching profile of ring sidewalls
obtained on different samples, it is possible to extract the statistical (Gaussian) information on
the sidewall roughness function and, then, give theoretical predictions by our model on the
laser working regime.
Fig. 9 shows the intensities of CW and CCW beams versus the effective ring radius for
different values of injection current in case of
c
L =0.07 m and
c
= 0.012 m. It is possible
to observe that the operating regime of the MQW ring laser is influenced by the ring cavity
sizes. In fact, for each value of the injection current the graph shows that the operational
regime converts from unidirectional to bidirectional by increasing the ring radius. This
behaviour depends on the circumstance that the density current decreases by increasing the
ring radius and, therefore, the intensity of the dominant beam is reduced. In this condition the
backscattering effect is not negligible and the beams hold the same intensity. The curves
stopped for an appropriate value of the ring radius, depending on the injection current. For
ring radii larger than this value, the MQW ring laser remains under the threshold.
Figure 9. Intensities of both beams versus the effective ring radius for various laser currents.
On the basis of our results, one additional component has to be included in the
architecture of the MQW ring laser to favour only one circulating direction over the other, i.e.
to achieve an unidirectional regime also for injection current values where the ring laser
should be bidirectional. The solution consists of an output coupler including a grating [46].
Then, we have investigated the influence of the grating reflectance
g
R and the output coupler
reflection and transmission coefficients,
c
R and
c
T respectively, over the laser behaviour.
It is clear that the presence of the grating strengthens the CW solution with respect to the
CCW one. A number of parametric simulations by varying
g
R and
c
R show that, at constant
injection current I =100 mA, correlation length
c
L =0.07 m and standard deviation
c
= 12
nm (the value for bidirectional regime of the MQW ring laser without output coupler), it is
possible to realise a purely unidirectional condition for
g
R > 0.9 and
c
R =10%. It is also
possible to obtain an unidirectional condition with a value of
g
R < 0.9, but it still needs to
increase
c
R .
Conclusion
In this chapter a short review on ring fiber lasers and recent advances of a highly detailed
physical model of MQW semiconductor ring lasers is presented. In particular, solutions to
multimode operation in fiber ring lasers and high performance in multi-wavelength erbium-
doped fiber lasers have been described. Moreover, the operation regimes of GaAs-based
semiconductor ring lasers have been demonstrated and related to the physical coefficients
included in the model. Different behaviour of the ring laser can be observed as a function of
injection current, ring radius and statistical information of the ring sidewall roughness as
determining the backscattering coefficient. Results relevant to output coupling by a grating
are discussed to guarantee a unidirectional regime in any condition. Thus, this chapter puts
into evidence as integrated MQW ring lasers are promising candidates to realise sources with
enhanced mode purity, reduced sensitivity to feedback and higher single beam power.
However, in particular applications such as optical test and measurements, optical
wavelength-division-multiplexing communications systems and sensing, the fiber ring lasers
lead to obtain high performance essentially due to their wide tunable range and very narrow
linewidth.
References
[1] Massicott, J. F., Armitage, J. R., Wyatt, R., Ainslie, R. J., and Craig-Ryan, S. P., High gain
broadband 1.6 m Er doped silica fiber amplifier, Electronics Letters, 1990, 26, 16451646.
[2] Haber, T., Hsu, K., Miller, C. and Bao, Y., Tunable erbium-doped fiber ring laser precisely
locked to the 50-GHz ITU frequency grid, IEEE Photonics Technology Letters, 2000, 12,
14561458.
[3] Yu, Y., Tam, H. Y., Lui, L. F. and Chung, W. H., Fiber-laser-based wavelength division
multiplexed fiber Bragg grating sensor system, IEEE Photonics Technology Letters, 2001,
13, 702704.
[4] Agrawal, G. P., Nonlinear Fiber Optics, Academic Press, 2001.
[5] Jain, R. K., Lin, C., Stolen, R. H. and Ashkin, A., A tunable multiple Stokes cw fiber Raman
oscillator, Applied Physics Letters, 1977, 31, 89-90.
[6] Mears, R. J., Reekie, L., Juancey, I. M., Payne, D. N., Low-noise Erbium-doped fiber
amplifier operating at 1.54 m, Electronics Letters, 1987, 23, 1026-1028.
[7] Pask, H. M., Carman, R. J., Hanna, D. C., Tropper, A. C., Mackechnie, C. J., Barber, P. R.,
Dawes, J. M., Ytterbium-doped silica fiber lasers: versatile sources for the 1-1.2 m region,
IEEE J. Selected Topics in Quantum Electronics, 1995, 1, 2-13.
[8] Oron, R., Bunimovich, D., Nagli, L., Katzir, A., Hardy, A. A., Design considerations for rare
earth doped silver halide fiber amplifiers, Optical Engineering, 2001, 8, 1516-1520.
[9] Sun, Y., Zyskind, J. L. and Srivastava, A. K., Average inversion level, modeling, and physics
of erbium-doped fiber amplifiers, IEEE J. Selected Topics in Quantum Electronics, 1997, 3,
991-1007.
[10] Bellemare, A., Karsek, M., Riviere, C., Babin, F., He, G., Roy, V. and Schinn, G.W., A
broadly tunable erbium-doped fiber ring laser: Experimentation and modeling, IEEE J.
Selected Topics in Quantum Electronics, 2001, 7, 2229.
[11] Dragic, P. D., Analytical Model for Injection-Seeded Erbium-Doped Fiber Ring Lasers,
IEEE Photonics Technology Letters, 2005, 17, 1629-1631.
[12] Mollenauer, L. F., Mamyshev, P. V., Gripp, J., Neubelt, M. J., Mamysheva, N., Gruner-
Nielsen, L. and Veng, T., Demonstration of massive wavelength-division multiplexing over
transoceanic distances by use of dispersion-managed solitons, Optics Letters, 2000, 25,
704706.
[13] Tamura, K., Haus, H. A. and Ippen, E. P., Self-starting additive pulse mode-locked erbium
fiber ring laser, Electronics Letters, 1992, 28, 22262227.
[14] Haus, H. A., Ippen, E. P. and Tamura, K., Additive-pulse mode-locking in fiber lasers, IEEE
J. Quantum Electronics, 1994, 30, 200208.
[15] Fermann, M., Andrejco, M. J., Silverberg, Y. and Stock, M. L., Passive mode locking by
using nonlinear polarization evolution in a polarizing-maintaining erbium-doped fiber laser,
Optics Letters, 1993, 18, 894896.
[16] Hofer, M., Fermann, M. E., Haberl, F., Ober, M. H. and Schmidt, A. J., Mode locking with
cross-phase and self-phase modulation, Optics Letters, 1991, 16, 502504.
[17] Collings, B. C., Cundiff, S. T., Akhmediev, N. N., Soto-Crespo, T. J., Bergman, K. and
Knox, W. H., Polarization-locked temporal vector solitons in a fiber lasers: experiment, J.
Optical Society of America B, 2000, 17, 354365.
[18] Spaulding, K. M., Yong, D. H., Kim, A. D., Kutz, J. N., Nonlinear dynamics of mode-
locking optical fiber ring lasers, J. Optical Society America B, 2002, 19, 1045-1054.
[19] Zhang, J. L., Yue, C. Y., Schinn, G. W., Clements, W. R. L. and Lit, J. W. L., Stable single-
mode compound-ring erbium-doped fiber laser, J. Lightwave Technology, 1996, 14,104109.
[20] Horowitz, M., Daisy, R., Fisher, B. and Zyskind, J., Narrow-linewidth, singlemode erbium-
doped fiber laser with intracavity wave mixing in saturable absorber, Electronics Letters,
1994, 30, 648649.
[21] Kim, S. K., Stewart, G., Johnstone, W. and Culshaw, B., Mode-hop-free single-longitudinal-
mode erbium doped fiber laser frequency scanned with a fiber ring resonator, Applied Optics,
1999, 38, 51545157.
[22] Cheng, Y., Kringlebotn, J. T., Loh, W. H., Laming, R. I. and Payne, D. N., Stable single-
frequency travelling wave fiber loop laser with integral saturable-absorber-based tracking
narrow-band filter, Optics Letters, 1995, 20, 875877.
[23] Song, Y. W., Havstad, S. A., Starodubopv, D., Xie, Y., Willner, A. E. and Feinberg, J., 40-
nm-wide tunable fiber ring laser with single-mode operation using a highly stretchable FBG,
IEEE Photonics Technology Letters, 2001, 13, 11671169.
[24] Libatique, N. J. C., Wang, L. and Jain, R. K., Single-longitudinal-mode tunable WDM-
channel-selectable fiber laser, Optics Express, 2002, 10, 15031507.
[25] Yeh, C.-H., Huang, T.-T., Chien, H.-C., Ko, C.-H. and Chi, S., Tunable S-band erbium-
doped triple-ring laser with single-longitudinal-mode operation, Optics Express, 2007, 15,
382387.
[26] Yeh, C. H., Lee, C. C. and Chi, S., A Tunable S-Band Erbium-Doped Fiber Ring Laser,
IEEE Photonics Technology Letters, 2003, 15, 10531054.
[27] Liu, H. L., Tam, H. Y., Chung, W. H., Wai, P. K. A. and Sugimoto, N., Low Beat-Noise
Polarized Tunable Fiber Ring Laser, IEEE Photonics Technology Letters, 2006, 18, 706-708.
[28] Xu, L., Glesk, I., Rand, D., Baby, V. and Prucnal, P. R., Suppression of beating noise of
narrow-linewidth erbium-doped fiber ring lasers by use of a semiconductor optical amplifier,
Optics Letters, 2003, 28, 780782.
[29] Liu, H. L., Tam, H. Y., Chung, W. H., Wai, P. K. A. and Sugimito, N., La-codoped bismuth-
based erbium-doped fiber ring laser with 106-nm tuning range, IEEE Photonics Technology
Letters, 2005, 17, 297299.
[30] Wei, D., Li, T., Zhao, Y. and Jian, S., Multi-wavelength erbium-doped fiber ring lasers with
overlap-written fiber Bragg gratings, Optics Letters, 2000, 25, 11501152.
[31] Bellemare, A., Karasek, M., Rochette, M., LaRochelle, S. and Tetu, M., Room temperature
multifrequency erbium- doped fiber lasers anchored on the ITU frequency grid, J. Lightwave
Technology, 2000, 18, 825831.
[32] Graydon, O., Loh, W. H., Laming, R. I. and Dong, L., Triple-frequency operation of and Er-
doped twin core fiber loop laser, IEEE Photonics Technology Letters, 1996, 8, 6365.
[33] Das G. and Lit, J. W. Y., L-band multiwavelength fiber laser using an elliptical fiber, IEEE
Photonics Technology Letters, 2002, 14, 606608.
[34] Zhao, C.-L., Yang, X., Lu, C., Ng, J. H., Guo, X., Chaudhuri, P. R. and Dong, X., Switchable
multi-wavelength erbium-doped fiber lasers by using cascaded fiber Bragg gratings written
in high birefringence fiber, Optics Communications, 2004, 230, 313317.
[35] Poustie, A. J., Finlayson, N. and Harper, P., Multiwavelength fiber laser using a spatial mode
beating filter, Optics Letters, 1994, 19, 716718.
[36] Mao, Q. and Lit, J. W. Y., Switchable multiwavelength erbium-doped fiber laser with
cascaded fiber grating cavities, IEEE Photonics Technology Letters, 2002, 14, 612614.
[37] Shum, P., Tang, M., Gong, Y., Dong, X., Fu, S., Dong, H. and Yang, X., Nonlinearity
enhances operation of fiber ring laser, SPIE, 2007, DOI: 10.1117/2.1200701.0562.
[38] Liu, Y., Hill, M. T., Calabretta, N., de Waardt, H., Khoe, G. D. and Dorren, H. J. S., Three-
state all-optical memory based on coupled ring lasers, IEEE Photonics Technology Letters,
2003, 15, 1461-1463.
[39] Mahnkopf, S., Kamp, M., Marz, R. and Forchel, A., Unidirectional ring laser diode with
gain-coupled distributed feeback, Electron. Letters, 2003, 39, 1055-1056.
[40] Bente, E. A. J., Barbarin, Y., den Besten, J. H., Smit, M. K., Binsma, J. J. M., Wavelength
selection in an integrated multiwavelength ring laser, IEEE J. Quantum Electronics, 2004,
40, 1208-1216.
[41] Zhang, S., Liu, Y., Lenstra, D., Hill, M. T., Ju, H., Khoe, G.-D., and Dorren, H. J. S., Ring
laser optical flip-flop memory with single active element, IEEE J. Selected Topics in
Quantum Electronics, 2004, 10, 1093-1100.
[42] Tangdiongga, E., Yang, X., Li, Z., Liu, Y., Lenstra, D., Khoe, G.-D., Dorren, H. J. S.,
Optical flip-flop based on two coupled mode-locked ring lasers, IEEE Photonics Technology
Letters, 2005, 17, 208-210.
[43] Sorel, M., Laybourn, P. J. R., Giuliani, G. and Donati, S., Unidirectional bistability in
semiconductor waveguide ring lasers, Applied Physics Letters, 2002, 80, 3051-3053.
[44] Sorel, M., Laybourn, P. J. R., Scir, A., Balle, S., Giuliani, G., Miglierina, R., Donati, S.,
Alternate oscillations in semiconductor ring lasers, Optics Letters, 2002, 27, 1992-1994.
[45] Sorel, M., Giuliani, G., Scir, A., Miglierina, R., Donati, S., Laybourn, P. J. R., Operating
regimes of GaAs-AlGaAs semiconductor ring lasers: experiment and model, IEEE J.
Quantum Electronics, 2003, QE-39, 1187-1195.
[46] De Leonardis, F. and Passaro, V. M. N., Accurate Physical Modeling of Multi Quantum Well
Ring Lasers, Laser Physics Letters, 2005, 2, 59-70.
[47] Shimosako, Y., and Numai, T., Semiclassical approach in analysis of ring laser. 1: Derivation
of rate equations including backscattering and interference, Japanese J. Applied Physics Part
1, 2000, 39, 3983-3990.
[48] Shimosako, Y., and Numai, T., Semiclassical approach in analysis of ring laser. 2: Mode
coupling due to backscattering and interference, Japanese J. Applied Physics Part 1, 2000,
39, 3991-3996.
[49] Haus, H. A., Statz, H., Smith, I. W., Frequency locking of modes in a ring laser, IEEE J.
Quantum Electronics, 1985, QE-21, 78-85.
[50] Yamada, M. and Suematsu, Y., Analysis of gain suppression in undoped injection lasers, J.
Applied Physics, 1981, 52, 2653-2664.
[51] Asada, M., Kameyama, A., Suematsu, Y., Gain and intervalence band absorption in
quantum-well lasers, IEEE J. Quantum Electronics, 1984, QE-20, 745-753.
[52] Armenise, M. N., Passaro, V. M. N., De Leonardis, F., Armenise, M., Modeling and Design
of a Novel Miniaturized Integrated Optical Sensor for Gyroscope Applications, J. Lightwave
Technology, 2001, 19, 1476-1494.

Chapter 6

INVESTIGATION OF OPTICAL POWER BUDGET
OF ERBIUM-DOPED FIBER

Hideaki Hayashi
a, b
, Setsuhisa Tanabe
a
and Naoki Sugimoto
b

a
Graduate School of Human and Environmental Studies, Kyoto University, Sakyo-ku,
Kyoto 606-8501, Japan
b
Research Center, Asahi Glass Co.,Ltd., Kanagawa-ku, Yokohama 221-8755, Japan
Abstract
We investigated optical power budget of an erbium-doped fiber (EDF). In addition to the
output signal and amplified spontaneous emission (ASE) powers from the fiber end, lateral
spontaneous emissions and scattering laser powers in the EDF were measured quantitatively
by using an integrating sphere. Compared with the signal and ASE powers, it was found that
considerable powers were consumed by the laterally emitting lights. As an optically
undetected loss which limits power conversion efficiency (PCE) of the fiber amplifier, the
effect of nonradiative decay from the termination level of pump excited state absorption
(pump ESA) was estimated from decay rate analyses of the relevant levels. The nonradiative
loss was comparable to amplified signal power in the EDF when pumped with a 980 nm LD.
Nonradiative decay following cooperative upconversion (CUP) process is also discussed using
rate equations analysis.
1. Introduction
With the spreading and popularization of Internet and broadband communication, larger data
traffic and higher processing speed are required in optical telecommunication systems. To
meet these demands, optical fiber communication networks have developed rapidly. By
using transmission fibers, metropolitan area networks (MANs) in inner-city has been
established for several years as well as long-haul networks [1]. There are two method of
increasing the information capacity in a single fiber; one is time division multiplexing
(TDM), and the other is wavelength division multiplexing (WDM) [2]. The TDM is a
technology of increasing the bit rate. On the other hand, the WDM is a technology of
coupling optical signals of different wavelengths in the same fiber. The transposable
Hideaki Hayashi, Setsuhisa Tanabe and Naoki Sugimoto 188
bandwidths are represented as the TDM speed times the number of wavelengths in the WDM
system.
In the optical fiber networks, optical amplifiers are one of the key components. The
optical signals decay due to the background losses of the transmission fibers (typically 0.2
dB/km at around 1550 nm). The insertion losses of optical add-drop multiplexer (OADM)
components also decrease the signal intensity, thus the amplifications of the signals are
necessary at every few tens kilometers. For the optical amplifiers in the WDM systems, it is
required that as many optical signals with different wavelengths as possible are amplified at
the same time. As a practical amplification medium, erbium doped fibers (EDFs) have been
extensively studied due to their excellent gain operation around 1.5 m in the loss minimum
window of transmission silica fiber [3, 4]. Figure 1 shows loss spectrum of a transmission
silica based fiber and amplification bandwidths of EDFs. Since the development of an
efficient silica-based erbium doped fiber amplifier (EDFA) in 1987 by a research group of
University of Southampton [5], considerable research efforts have been made to improve the
efficiency and broaden the bandwidths. To extend the bandwidths from conventional C-band
(1530-1565 nm) to L-band (1570-1610 nm) in the WDM system, several glass hosts for
EDFA such as fluoride, tellurite, Bi
2
O
3
-based, and multi component antimony silicate (MCS)
have been proposed since the later half of the 1990s [6-9].

1400 1500 1600
Wavelength (nm)
T
r
a
n
s
m
i
s
s
i
o
n

l
o
s
s

(
d
B
/
k
m
)
0.2
0.3
0.4
0.5
C-band L-band
Silica EDF for
Bi
2
O
3
-based EDF for
C-band: 1530-1565
L-band: 1580-1605
3 dB down bandwidth
C-band: 1530-1565
C+L-band: 1535-1610
Extended L-band:
1545-1620

Figure 1. Loss spectrum of a transmission silica based fiber and the bandwidths of Silica EDFs and
Bi
2
O
3
-based EDF. Here bandwidth means 3 dB down bandwidth.
Silica based EDFs have been installed in the actual optical network system and
practically played a critical role. However, their power conversion efficiency (PCE) is
limited to 50-55% when pumped with 980 nm LD at present (i.e. ER-1090 amplifier by
Sumitomo Electric Industries, Ltd. or HP980 amplifier by OFS). Since pump LD cost
represents a significant proportion of the total amplifier cost, increasing the PCE is a concern
for the amplifier development. Although the PCE is one of the most important factors for the
amplifier design, it is not perfectly understood what limits the PCE in the EDFs. Except for
Investigation of Optical Power Budget of Erbium-Doped Fiber 189
emissions or loss origins that can be evaluated at the output end of the fiber (i.e. amplified
signal, amplified spontaneous emission (ASE) or splice loss), considerable optical power
budget of the EDFs is not clear. For example, lateral spontaneous emission of 1550 nm band
has not ever been evaluated quantitatively although there is a report that the lateral emission
spectra have been measured to calculate the cross section [10]. In order to optimize the PCE
and amplifier performance, understanding of overall optical power budget of the EDFs is
essential.
In this study, we constructed a novel evaluation system for measuring lateral emissions
from the EDF by using an integrating sphere. We used a Bi
2
O
3
-based EDF (BIEDF) for the
evaluation due to its potential for high performance amplifier [8, 11-13]. The lateral
emissions such as spontaneous emission, upconversion emission, scattering light of laser
diode (LD), and the scattering light of signal or ASE were measured quantitatively as well as
the in-situ data results of the gain properties as a fiber amplifier. The variations of the lateral
emissions with signal wavelength, signal power, or pump power were investigated. In
addition, we estimated the effect of other nonradiative decay processes that follow pump
excited state absorption (pump ESA) or cooperative upconversion (CUP). To investigate the
nonradiative decay from the termination level of the pump ESA, the luminescence decay of
the 550 nm band was measured. The effect of the CUP is then discussed theoretically using
rate equations and optical propagation equations. Finally, we present the optical power
budget of the BIEDF and clarify what decreases the PCE in the amplifier.
2. Background
The configuration and principle of an EDFA is shown in Fig. 2. An EDFA is composed of
pump lasers, WDM couplers that couple input signals with pumping lights, isolators that
prevent the reflection of output signals, and an EDF as an amplification medium. The EDF
can be operated as a laser for the signal wavelength ranging in the band around 1550 nm, by
utilizing a pump beam of a LD at the wavelength of 980 nm or 1480 nm [14, 15].
The 4f energy diagram of Er
3+
ion and the main transitions involved in the three level
laser operation are shown in Fig. 3. The ground state absorption (GSA) cross-section of the
Er
3+
ion exhibits a peak at 980 nm, and the Er
3+
ions are excited from the ground
4
I
15/2
level to
the
4
I
11/2
level. They decay to the metastable
4
I
13/2
level immediately, and the stimulated
emission from the
4
I
13/2
level to the
4
I
15/2
level takes place. In addition to these transitions, the
following transitions are accounted in this study: the quantum noise due to the ASE; the CUP
via two photons in the first excited level of
4
I
13/2
. Energy transfers from one Er
3+
ion to other,
and then the remaining excited Er
3+
ion rapidly decay back to the
4
I
13/2
level; 1550 nm-band
spontaneous emission (1550 nm-SE); the pump ESA from the
4
I
11/2
level to the
4
F
7/2
level; the
upconversion emission around 550 nm-band (550 nm-SE); Nonradiative transitions (NRs)
between the
4
F
7/2
level and the
4
I
13/2
level.
The PCE of optical amplifiers is calculated using following expression:

PCE(%) = ( (P
sOUT
P
sIN
) / P
pIN
) 100, (1)

where P
sOUT
, P
sIN
, and P
pIN
are output signal power, input signal power, and launched pump
power, respectively (Unit: W).
980nm
1480nm
Pump LD Pump LD
Er-doped fiber
Signal
h
WDM
coupler
Isolator
E
n
e
r
g
y
Er
3+
ions at ground state
Excited ions in active-center

Figure 2. Configuration and principle of Er-doped fiber amplifier.

0
5
10
15
20
E
n
e
r
g
y

(
1
0
3

c
m
-
1
)
Er
3+
4
I
13/2
4
I
15/2
4
I
11/2
4
I
9/2
4
S
3/2
4
F
9/2
2
H
11/2
4
F
7/2
980 nm
Amp.Sig.
NR
CUP
1550nm-
SE
550 nm-SE
Signal
ASE
ESA
GSA
0
5
10
15
20
E
n
e
r
g
y

(
1
0
3

c
m
-
1
)
Er
3+
4
I
13/2
4
I
15/2
4
I
11/2
4
I
9/2
4
S
3/2
4
F
9/2
2
H
11/2
4
F
7/2
980 nm
Amp.Sig.
NR
CUP
1550nm-
SE
550 nm-SE
Signal
ASE
ESA
GSA

Figure 3. 4f energy diagram of Er
3+
ion and the relevant transitions.
3. Preparation and Gain Characteristics of BIEDF
The glass preform containing Bi
2
O
3
and SiO
2
as main constituents was prepared using a
conventional melting method. For the fiber core composition, 0.5 mol% of Er
2
O
3
was added
to the glass batch. Single mode EDF (cladding diameter of 125 m) with plastic coatings was
then fabricated. The core diameter of the BIEDF was 3.9 m. The refractive index of the
core and the numerical aperture (NA) of the fiber at 1550 nm were 2.03 and 0.20,
respectively. A BIEDF of 16 cm length was fusion-spliced to high NA fibers (Nufern
980HP) using a commercial fusion-splicer. The insertion loss of the spliced BIEDF at 1310
nm was 0.61 dB. By using a cutback method, the propagation loss of the BIEDF at 1310 nm
was estimated to be 0.77 dB/m. Accordingly, the average splice loss per point was estimated
to be 0.24 dB. Angled cleaving and splicing were applied to suppress the reflection due to the
large difference of refractive induces between the BIEDF and the silica fibers [12]. It was
confirmed that pig-tailed BIEDFs passed Bellcore (Telcordia) GR-1221 CORE qualification
test [16]. Gain and noise figure profiles of the 16 cm BIEDF are shown in Fig. 4. The BIEDF
was pumped with 140 mW by forward direction at 980 nm. The gain of the BIEDF reached
18.8 dB at 1535 nm in case the input signal power was 10 dBm.

1520 1540 1560 1580 1600 1620
0
5
10
15
20
25
Wavelength (nm)
G
a
i
n

a
n
d

N
o
i
s
e

F
i
g
u
r
e

(
d
B
)
Gain
NF

Figure 4. Gain and NF profiles of the 16 cm BIEDF. Launched pump: 140 mW forward at 980 nm;
Input signal: 10 dBm at 1535 nm.
4. Lateral Emission Properties of BIEDF
4.1. Lateral Emission Measurement
Experimental setup for evaluating the lateral and fiber-propagating emission powers is shown
in Fig. 5. The BIEDF of 16 cm length was coiled with 6 cm diameter and set in an
integrating sphere (10inch: Model LMS-100s, Labsphere Inc.). The input and output end of
high NA silica fibers were connected with instruments through small hole (5mm diameter) of
the integrating sphere. The splice points were set just outside of the sphere. It was then
pumped with a LD (FITEL) by forward direction at the wavelength of 980 nm. The pump
power and temperature of the LD were controlled with a LD-driver (Model 525, Newport
Corp.) and a temperature-controller (Model 325, Newport Corp.), respectively. A tunable
laser (Model TLS210, Santec Corp.) was used for a single-channel signal source, and then the
pump light and the signal light was coupled using a WDM coupler/Isolator (WDM/ISO). The
spontaneous emissions and scattering lights laterally emitted from the BIEDF were detected
with two kinds of fiber multi-channel CCDs with Si and InGaAs detectors. Each CCD
coupled proprietary spectrometer. The maximum wavelength range of the visible
spectrometer (Model USB-2000, Ocean Optics Inc.) and the near-infrared spectrometer
(Model NIR-512, Ocean Optics Inc.) were 350-1000 nm and 900-1700 nm, respectively. A
premium-grade fiber with 1mm core (Model QP1000-2-VIS/NIR, Ocean Optics Inc.) was
used to link the CCD and the output port of the sphere. For the spectral calibration, a
standard halogen lamp (Model SCL-600, Labsphere Inc.) was used. The lamp was set at the
center of the sphere and driven at 2.60A with a current-regulated DC stabilized power supply
(Model PAN-5A, Kikusui Electronics Corp.). The absolute powers of total radiant flux of
lateral emissions were then calculated. At the same time, the output spectra of fiber-
propagating signal and ASE were detected with an optical spectrum analyzer (OSA: Model
MS9780A, Anritsu Corp.) with 1 nm resolution.
First, we measured the spectral power distribution of various emissions. The pump
power, the input signal power, and the signal wavelength dependences of the emissions were
then investigated.

Integrating sphere
CCD/
Spectrometer
PC
BIEDF
OSA
WDM
/ISO
Pump LD
Device Under
Test
Splice point
Signal
source

Figure 5. Experimental setup for evaluating the lateral and fiber-propagating emission powers of the
BIEDF. Basically the splice points were set outside of the integrating sphere.
4.2. Spectral Power Distribution
First, we show absolute power spectrum of lateral emissions and output emissions from the
fiber end (Fig. 6). The ordinate represents spectral power distribution of radiant flux. The
upconversion emission around 520 nm (
2
H
11/2

4
I
15/2
) and 550 nm (
4
S
3/2

4
I
15/2
), scattering
light of LD around 980 nm, spontaneous emission of 1550 nm band, and ASE were detected
by the two kind of multi-channel CCD which was connected with integrating sphere. We can
also see weak emission around 660 nm that is related to the pump ESA process [17, 18]. The
spectral shapes of the upconversion emission and 1550 nm band spontaneous emission were
approximately identical with those in bulk glass [8]. When 100 mW of pump power and 0
dBm of signal power at 1530 nm were input, the optical powers of the upconversion
emission, the LD scattering, and the 1550 nm band spontaneous emission, were 0.2 mW, 0.2
mW, and 3.1 mW, respectively. Here the splice points were set outside of the integrating
sphere. In the case that the splice points were set inside of the sphere, the scattering of pump
LD was increased to 4.3 mW, and 1.8 mW of the scattering light of the amplified signal was
detected by the CCD. The optical powers of amplified signal at 1530 nm and ASE band that
detected by the OSA were 11.9 mW and 0.2 mW, respectively.
When the splice points were set outside of the integrating sphere, the sum of emission
powers detected by the OSA and the CCDs were 12.1 mW and 3.5 mW, respectively. On the
other hand, when the splice points were set inside of the sphere, the optical powers of the LD
and signal scattering lights increased. The differences should represent the scatterings at the
splice points. That is, the LD and signal scattering lights at the splice points result in the
power losses of 4.1 mW and 1.8 mW, respectively.

500 1000 1500
10
-2
10
-1
10
0
10
1
10
2
10
3
10
4
Wavelength (nm)
S
p
e
c
t
r
a
l

p
o
w
e
r

d
i
s
t
r
i
b
u
t
i
o
n

(
W
/
n
m
)
550nm
-SE
980nm
-Scat.
ASE
1550nm
-SE
Amp.Si g.

Figure 6. Spectral power distribution of various lateral emissions and amplified signal from the BIEDF.
Launched pump and input signal power were 100 mW and 0 dBm, respectively. The splice points of
the BIEDF were set outside of the integrating sphere. 550 nm-SE = 550 nm band spontaneous
emission; 980 nm-Scat. = Scattering light of the LD at 980 nm; 1550 nm-SE = 1550 nm band
spontaneous emission; Amp.Sig. = Amplified signal at 1530 nm; ASE = Amplified spontaneous
emission.
4.3. Signal Wavelength Dependence
Figure 7 shows the signal wavelength dependences of the optical powers of the lateral
emissions and the fiber propagating emissions. The launched power of the pump LD at 980
nm was fixed to 100 mW. Input signal power was set to 0 dBm. The right axis in the figure
shows the gain of the output signal (square plots, unit: dB). In the wavelength range from
1530 nm to 1560 nm that corresponds to the C-band, we can see that the signal gains more
than 10 dB were obtained with the BIEDF of only 16 cm length. The optical power of the
1550 nm band spontaneous emission was larger than that of the ASE in the entire C-band
region. As for the ASE, the spontaneous emission of 1550 nm band, and the scattering light
of the 980 nm LD, the optical powers of their emissions showed negative correlations with
that of the amplified signal at measured wavelengths. The correlation of the ASE was the
strongest among these emissions. These results indicates that more powers are consumed for
the output signal power in the C-band by lowering the ASE and lateral powers in vain, which
is desirable as a fiber amplifier.
On the other hand, the optical power of the upconversion emission around 550 nm
showed weak positive correlation. In other words, the upconversion emission power was
large at the wavelength that the output signal power was large. This suggests that the
upconversion emission is populated by the signal photons. In addition to the pump ESA,
signal ESA using the signal photons can also occur. The initial level of upconversion
emission,
4
S
3/2
, would be populated by the signal photons through pump ESA. It can be said
from the above correlation that the effect of the input or output signal wavelength on the
signal ESA process is smaller than that of the output signal power.

1500 1520 1540 1560 1580
10
2
10
3
10
4
Wavelength (nm)
O
p
t
i
c
a
l

p
o
w
e
r

(
W
)
Amp. Sig.
ASE
0. 55u-SE
0. 98u-Scat.
1. 55u-SE
S
i
g
n
a
l

g
a
i
n

(
d
B
)
10
0
-10

Figure 7. Signal wavelength dependence of optical powers of various emissions in the BIEDF.
Launched pump and input signal power were100 mW and 0 dBm, respectively. The splice points of the
BIEDF were set outside of the integrating sphere. 550 nm-SE = 550 nm band spontaneous emission;
980 nm-Scat. = Scattering light of the LD at 980 nm; 1550 nm-SE = 1550 nm band spontaneous
emission; Amp.Sig. = Amplified signal; ASE = Amplified spontaneous emission.
4.4. Signal Power Dependence
The signal power dependences of the optical powers of various emissions are shown in Fig. 8.
The launched pump power was set to 100 mW, and input signal wavelength was fixed to
1530 nm. The ASE, the spontaneous emission of 1550 nm band, and the scattering light of
the 980 nm LD decreased with increasing the input signal power. Even in the small signal
region, the lateral emission power was larger than the ASE at the same 1550 nm band. The
lateral 1550 nm spontaneous emission was larger than the amplified signal when the input
signal power was smaller than -20 dBm. On the other hand, the upconversion emission
around 550 nm increased with the input signal power. This positive correlation also suggests
the existence of the signal ESA process using the input and amplified signal photons, because
the output signal power of a fiber amplifier increases with increasing the input signal power.

-30 -20 -10 0
10
2
10
3
10
4
Input signal power (dBm)
O
p
t
i
c
a
l

p
o
w
e
r

(
W
)
Amp. Sig.
ASE
0. 55u-SE
0. 98u-Scat.
1. 55u-SE

Figure 8. Input signal power dependence of optical powers of various emissions in the BIEDF.
Launched pump power were 100 mW. The splice points of the BIEDF were set outside of the
integrating sphere. 550 nm-SE = 550 nm band spontaneous emission; 980 nm-Scat. = Scattering light
of the LD at 980 nm; 1550 nm-SE = 1550 nm band spontaneous emission; Amp.Sig. = Amplified
signal; ASE = Amplified spontaneous emission.
4.5. Pump Power Dependence

10
1
10
2
10
0
10
1
10
2
10
3
10
4
Exci tati on power (mW)
O
p
t
i
c
a
l

p
o
w
e
r

(
W
)
Amp.Si g.
ASE
0.55u-SE
0.98u-Scat.
1.55u-SE

Figure 9. Pump power dependence of optical powers of various emissions in the BIEDF. Input signal
power was 0 dBm. The splice points of the BIEDF were set outside of the integrating sphere. 550 nm-
SE = 550 nm band spontaneous emission; 980 nm-Scat. = Scattering light of the LD at 980 nm; 1550
nm-SE = 1550 nm band spontaneous emission; Amp.Sig. = Amplified signal; ASE = Amplified
spontaneous emission.
Figure 9 shows the pump power dependence of the optical powers of various emissions. All
the emission species increased with the pump power, and the dependence of the upconversion
emission around 550 nm was nearly 2 and obeying quadratic law. This means that the
emission occurs as a result of the pump ESA or the CUP, each of which is due to a two-
photon process. The pump power dependence of the lateral 1550 nm spontaneous emission
was small and almost saturated under the pump power of larger than 60 mW. This indicates
that there exist sufficient photons at the
4
I
13/2
level even when the excitation power is very
small.
5. Nonradiative Loss
Other than lateral emissions described above, various nonradiative decay processes can be
considered; deactivation by hydroxyl group in glass, nonradiative decay which is related to
the pump ESA, the decay which is related to the CUP, and the multiphonon relaxation from
the
4
I
11/2
level. Among these origins, the effect of hydroxyl groups was neglected here
because this BIEDF was sufficiently dehydrated during the fabrication [19, 20].
5.1. Pump ESA Process
5.1.1. Lifetime Measurement of Er
3+
:
4
S
3/2
Level
To analyze the effect of the nonradiative decay from the termination level of the pump ESA,
luminescence decay of 550 nm band was measured, and then the quantum efficiency of the
Er
3+
:
4
S
3/2
level was calculated from the measured lifetime [21]. Second harmonic of Nd:
YVO
4
laser at 532 nm (Model J80-H10-532QW, Spectra Physics) was used as a pump source.
The pump power was adjusted to 1 W, and the pump light that was modulated into pulses
(Repetition: 15000 Hz; Pulse width: 13 ns) was incident on the optically polished Er-doped
Bi
2
O
3
-based glass sample (18153.5 mm in size). The luminescence of 550 nm band of the

0 1 2 3 4
[10
-5
]
I
n
t
e
n
s
i
t
y
(
a
r
b
.
u
n
i
t
)
Ti me(s)
Excitation: 532 nm
Power: 1 W
Monit ering: 550 nm
Sl it: 8 mm
f
=2.7 s

Figure 10. Luminescence decay curve of the Er
3+
:
4
S
3/2
level in the Bi
2
O
3
-based glass. Circle plots
represent measured data, and solid line represents single exponential fitting of these data.
glass sample was monochromated (Model 1681B, Spex) and detected with a photomultiplier
(Model 1424M, Spex) that 0.8 kV of voltage was applied. The signal was collected using a
sampling oscilloscope (500 MHz; Model TDS520, Tectronix Corp.), and the lifetime was
determined by least square fitting of the obtained decay curve with exponential functions.
Measured luminescence decay curve of 550 nm band (
4
S
3/2

4
I
15/2
) is shown in Fig.
10. Sharp peak that can be observed near zero of the decay time must be the scattering of
the pump LD, because the monitoring wavelength is relatively close to the pumping one.
After excluding the effect of the LD scattering, the lifetime of the
4
S
3/2
level was
determined to be 2.7 s by using single exponential function.

5.1.2. Nonradiative Losses Following Pump ESA
By using the obtained lifetime value, we discuss decay from the
4
S
3/2
level as a result of the
pump ESA process. For simplicity, we assumed that all the photons that were excited to the
4
F
7/2
levels relax nonradiatively to the
4
S
3/2
level. This assumption will be valid because the
energy gap between the
4
F
7/2
and the
4
S
3/2
is narrow (750 cm
-1
) [22].
Generally quantum efficiency of an emission, , is written as follows:

= A
f
= A / (A + W), (2)

where A is spontaneous emission probability, W is nonradiative transition probability, and
f

is fluorescence lifetime. We calculated the A coefficient from the Judd-Ofelt analysis (3100
s
-1
) [23-25]. Accordingly, the quantum efficiency was estimated to be 0.8%.
The nonradiative energy loss from the
4
S
3/2
level to the
4
I
11/2
level (unit: W), P
NR

(
4
S
3/2
4
I
11/2
), can be then expressed as follows:

NR
P (
4
S
3/2
4
I
11/2
) = {
R
P (
4
S
3/2
4
I
15/2
) / } {E (
4
S
3/2
4
I
11/2
) / E (
4
S
3/2
4
I
15/2
)}, (3)

where P
R
(
4
S
3/2
4
I
15/2
) is upconversion emission power, E is energy gap between two 4f
levels. The nonradiative energy loss from the
4
I
11/2
level to the
4
I
13/2
level, P
NR
(
4
I
11/2
4
I
13/2
),
can be considered separately.

NR
P (
4
I
11/2
4
I
13/2
) = [P
L
{
R
P (
4
S
3/2
4
I
15/2
) +
NR
P (
4
S
3/2
4
I
11/2
) }P
R
(
4
I
11/2
4
I
15/2
)]
{ E (
4
I
11/2
4
I
13/2
) / E (
4
I
11/2
4
I
15/2
)}. (4)

Here P
L
is launched pump power, P
R
(
4
I
11/2
4
I
15/2
) is the optical power of 1000 nm
emission band.
By using these expressions described above, P
NR
(
4
S
3/2
4
I
11/2
) and P
NR
(
4
I
11/2
4
I
13/2
)
were calculated to be 13 mW and 31 mW, respectively. Although the visible upconversion
luminescence power was only 0.2 mW, we have to count the nonradiative decay from the
4
S
3/2
level due to low .
5.2. Coorpelative Upconversion Process
5.2.1. Calculation of Cup Process
We can estimate the effect of the CUP process using the rate equations analysis. Figure 11
shows the 4f energy level diagram of Er
3+
ions and transitions used for the analysis. When a
BIEDF is pumped with a 980 nm LD by forward direction, the time dependence of
populations can be expressed as follows [26-28]:

( ) ( )
4 41 3 31
2
2 1 12 13 2 21 21 21 1
/ N A N R CN N R R N W R A dt dN + + + + + + = , (5)

2
2 3 32 1 12 2 21 21 21 2
2 ) ( / CN N W N R N W R A dt dN + + + + = , (6)

( )
2
2 4 43 3 31 34 32 1 13 3
/ CN N W N R R W N R dt dN + + + + = , (7)

3 34 4 43 41 4
) ( / N R N W A dt dN + + = , (8)

where N
1
, N
2
, N
3
, and N
4
represent the population of the
4
I
15/2
,
4
I
13/2
,
4
I
11/2
, and
4
F
7/2
levels,
respectively. For simplicity, we neglected the intermediate levels between the
4
I
11/2
and the
4
F
7/2
levels, and assumed that all the photons pumped at the
4
F
7/2
level via the pump ESA
transit nonradiatively to the
4
S
3/2
level. Total Er
3+
ion number density for the calculation was
set to 1.54 10
26
m
-3
, which corresponded to 0.5 mol% of Er
2
O
3
. R
21
, R
12
, R
31
, R
13
, and R
34

are radiation transition rate between these levels that are calculated from absorption and
emission cross sections (
s
e
,
s
a
,
p
e
,
p
a
, and
ESA
, respectively). A
21
and A
41
represent
spontaneous emission probabilities that are calculated by the Judd-Ofelt analysis [25].
Nonradiative transition probability, W
43
, is calculated in the way described in Section 5.1.2.
W
21
and W
32
can be also estimated from the lifetime measurements of the bulk glasses in the
same way as described in Section 5.1.1. The fiber length and the numerical aperture were set
to 16 cm and 0.20, respectively. C represents cooperative upconversion coefficient. Here we
assumed homogeneous distribution of Er
3+
ions in the glass and homogeneous upconversion
process [29-31]. Mode field diameter at 980 nm and at 1530 nm were set to 4.2m and 6.3
m, respectively.
The signal and pump lightwaves propagating along the fiber (I
s
and I
p
) are expressed as
the following set of ordinary differential equations [26-28].

dI
s
/ dz = (
s
e
N
2

s
a
N
1
)
s
I
s

s
I
s
(9)

dI
p
/ dz = (
p
a
N
1

p
e
N
3
+
ESA
N
3
)
p
I
p

p
I
p
(10)

s
and
p
are overlap factor at the signal wavelength and pumping wavelength, respectively.
s
and
p
are parameters that represent the intrinsic fiber background loss at the signal and
pumping wavelength, respectively. Here we assumed that the
s
were identical with
p
, and
treated them as fitting parameters. We applied the Quimbys assumption that
ESA
is equal to
2
p
a
[32]. Although spontaneous decay was accounted for, the ASE was neglected since the
input signal power was sufficiently large (0 dBm) and the fiber length was sufficiently short.
Splice loss from the BIEDF to high-NA silica fiber was set to 0.24 dB/point.
Assuming a steady state condition (the time derivatives to be zero), the set of differential
equations were numerically integrated using the fourth order Runge-Kutta method with an
initial condition at the input end of the fiber (z=0 m) [27]. The parameters used for numerical
calculations are shown in Table 1. Input signal and launched pump powers were set to 1
mW (0 dBm) and 100 mW, respectively. By using these calculations, we obtained the
relationship between the output signal power and the CUP coefficient.
Table 1. Parameters used for numerical calculations.

Parameter Symbol Value Unit
Spontaneous emission rate A
21
250 s
-1
A
41
3100 s
-1
Nonradiative decay rate W
32
69 s
-1
W
32
3.3010
4
s
-1
W
43
3.7010
5
s
-1
Signal emission cross section at 1530 nm
s
e
7.41
10
-25
m
2
Signal absorption cross section at 1530 nm
s
a
8.19
10
-25
m
2
Pump emission cross section at 980 nm
p
e
3.06
10
-25
m
2
Pump absorption cross section at 980 nm
p
a
2.36
10
-25
m
2
Overlap factor at 980 nm
s
0.82
Overlap factor at 1530 nm
p
0.52
Er
3+
ion density 1.54
10
26
m
-3
Cooperative upconversion coefficient C Fitting parameter m
3
/s
Background loss Fitting parameter

N
1
0
5
10
15
20
E
n
e
r
g
y

(
1
0
3

c
m
-
1
)
Er
3+
4
I
13/2
4
I
15/2
4
I
11/2
4
I
9/2
4
S
3/2
4
F
9/2
4
F
7/2
R
13
R
34
A
41
Amp.Sig.
R
21
W
21
W
32
W
43
C
N
2
N
3
N
4
A
21
R
12
R
31

Figure 11. 4f energy diagram of Er
3+
ion and the transitions used for the rate equations analysis.
5.2.2. Effect of Cooperative Upconversion
Here we estimate the effect of the cooperative upconversion (CUP) using rate equations
analysis as described in the former section. Figure 12 shows the variation of calculated
output signal with the CUP coefficients. The difference between the output power at a given
CUP coefficient and the output at zero of the coefficient (value at y-intercept) represents
energy loss via the CUP process. The calculations were performed for three values of . For
any , the output power decreased exponentially with increasing the CUP coefficient.
Snoeks et al. reported that the value of the CUP coefficient was 3.2 10-
24
m
3
/s in a soda
lime silicate glass that was doped with 1.410
-26
m
-3
of Er
3+
ions [31]. When we assume that
the CUP coefficient of the BIEDF (1.5410
-26
m
-3
of Er
3+
ion number density) is same as that
of the soda lime silicate, the curve of = 4 seems reasonable. In this case, the effect of the
CUP process results in approximately 10 mW. If we decrease the Er concentration in glass,
the CUP will be reduced because the CUP coefficient is a function of the Er
3+
ion density
[33].

0 0.5 1 1.5 2
[10
-23
]
0
5
10
15
20
25
30
35
40
O
u
t
p
u
t

p
o
w
e
r

(
m
W
)
CUP coeffi cient (m
3
/s)
=0
=4
=8
Soda li me
silicate

Figure 12. Variation of signal output power with the CUP coefficient in the BIEDF doped with 1.54
10
26
m
-3
of Er
3+
ions. Plots represent calculation data, and solid lines are exponential fitting of these
data. Dashed line represents literature value for a soda lime silicate glass doped with 1.4 10
26
m
-3
of
Er
3+
ions (C = 3.2 10
-24
m
3
/s) [31].
6. Energy Budget of BIEDF
The optical power budget of the BIEDF that has been clarified in this study is shown in Table
2. Here the launched pump power and the input signal power were 100 mW and 1 mW (0
dBm), respectively. The insertion loss of 0.61 dB corresponds to 13.1 mW. The output
signal power at 1530 nm and the sum of lateral emissions and scattering powers were 11.9
mW and 9.4 mW, respectively. It can be said that considerable powers were consumed by the
lateral emissions and scatterings in the BIEDF.
Taking into account the output signal, the ASE, the lateral emissions, and the insertion
loss, 65% of total power (65 mW) was not detected either by the CCDs or the OSA. The
power of the nonradiative decay from the termination level of the pump ESA to the
4
I
11/2
level
was estimated to be 13 mW. That from the
4
I
11/2
level to the
4
I
13/2
level was 31 mW.
Approximately 10 mW can be attributed to the nonradiative decay following the CUP. We
can say that nonradiative decays above also affect the decrease of the PCE in the BIEDF.
Even counting all sources of loss described above, however, we could not identify
approximately 11% of total launched power. A possible reason is that we underestimate
nonradiative losses at present. For precise estimation of the pump ESA effect, high
measurement accuracy of the very weak upconversion luminescence is necessary. For the
CUP effect, we will have to consider the clustering of the Er
3+
ions and resulting pair induced
quenching [15, 34, 35].
Table 2. Energy budget of the BIEDF when pumped with 100 mW of launched power.

Emission species and source of loss mW
Amplified signal 12
Insertion loss (splice loss+background loss) 13
980 nm LD scattering (at splice point) 4.1
980 nm LD scattering (w/o splice point) 0.2
Signal scattering (at splice point) 1.8
Amplified spontaneous emission 0.2
1550 nm band spontaneous emission 3.1
550 nm band upconversion emission 0.2
Nonradiative decay from the
4
S
3/2
to the
4
I
11/2
13
Nonradiative decay from the
4
I
11/2
to the
4
I
13/2
31
Nonradiative decay following CUP aprx.10
Unidentified aprx.11
Launched pump power: 100 mW
Input signal power: 1 mW

7. Conclusion
We have analyzed optical power budget of an erbium-doped amplifier (EDF). Lateral
spontaneous emissions and scattering laser powers in a Bi
2
O
3
-based EDF (BIEDF) were
evaluated quantitatively by using an integrating sphere. Comparing with amplified signal, it
was clarified that considerable power was consumed by the laterally emitting lights. While
the LD scattering, the signal scattering, and the 1550 nm band emission powers decreased
with increasing input signal power, the lateral 550 nm emission power increased. In the same
way, among the lateral emissions, only 550 nm band showed positive correlation with the
spectrum of the output signal. These results suggested that the upconversion emission was
promoted by the signal ESA.
As a result of decay rate analysis, it was revealed that the nonradiative power loss related
to the pump excited state absorption (pump ESA) was comparable with the output signal
power because the quantum efficiency of the initial level of the upconversion emission was
only 0.8%. In addition, as a result of rate equations analysis, it was suggested that the effect
of nonradiative decay following the cooperative upconversion (CUP) was not negligible when
Er
3+
ion density was an order of 10
-26
m
-3
.
These analyses performed in this study can be applicable for not only a BIEDF but also
commercial silica based EDF that the power conversion efficiency (PCE) is usually limited to
50-55%, other rare earth-doped amplifiers or lasers. The measurement system using an
integrating sphere is also useful to analyze the lateral emissions from waveguide amplifiers in
which precise control of their structures is necessary.
References
[1] Kartalopoulos, S. V. Introduction to DWDM Technology; IEEE Press: NJ, 1999, pp.
209-222.
[2] Keiser, G. E. Opt. Fiber Technol. 1999, vol. 5, 3-39.
[3] Sudo, S. Optical Fiber Amplifiers; Artech House: MA, 1997, pp. 1-53.
[4] Tanabe, S.; Hanada, T. J. Non-Cryst. Solids 1999, vol. 196, 101-105.
[5] Mears, R.J.; Reekie, L.; Jauncey, I. M.; Payne, D. N. Electron. Lett. 1987, vol. 23, 1026-
1028.
[6] Yamada, M.; Ono, H.; Kanamori, T.; Sudo, S.; Ohishi, Y. Electron. Lett. 1997, vol. 33,
710-711.
[7] Oishi, Y.; Mori, A.; Yamada, M.; Ono, H.; Nishida, Y.; Oikawa, K. Opt. Lett. 1998, vol.
23, 274-276.
[8] Tanabe, S.; Sugimoto, N.; Ito, S.; Hanada, T. J. Lumin. 2000, vol. 87-89, 670-672.
[9] Goforth, D. E.; Minelly, J. D.; Ellison, A. J. G.; Trentelman, J. P.; Samson, B. N.
Technical Digest of Optical Amplifiers and their Applications 2000, Nara, 2000, pp.
OTuA4-1.
[10] Zech, H. IEEE Photon. Technol. Lett. 1995, vol. 7, 986-988.
[11] Sugimoto, N. J. Am. Ceram. Soc. 2002, vol. 85, 1083-88.
[12] Ohara, S.; Sugimoto, N.; Ochiai, K.; Hayashi, H.; Fukasawa, Y.; Hirose, T.; Nagasima,
T.; Reyes, M. Opt. Fiber Technol. 2004, vol. 10, 283-295.
[13] Hayashi, H.; Sugimoto, N.; Tanabe, S. Opt. Fiber Technol. 2006, vol. 12, 282-287.
[14] Barnes, W. L.; Laming, R. I.; Tarbox, E. J.; Morkel, P. R. IEEE J. Quantum Electron.
1991, vol. 27, 1004-1010.
[15] Prudenzano, F. J. Lightwave Technol. 2005, vol. 23, 330-340.
[16] Bell Communications Research, Generic requirements GR-1221-CORE 1994, Issue 1.
[17] Auzel, F. Chem. Rev. 2004, vol. 104, 139-173.
[18] Sun, H.; Xu, S.; Dai, S.; Wen, L.; Zhang, J.; Hu, L.; Jiang, Z. J. Non-Cryst. Solids 2005,
vol. 351, 288-292.
[19] Hayashi, H.; Sugimoto, N.; Tanabe, S.; Ohara, S. J. Appl. Phys. 2006, vol. 99, 093105.
[20] Hayashi, H.; Sugimoto, N.; Ochiai, K.; Ohara, S.; Fukasawa, Y.; Tanabe, S. Extended
Abstract of International Congress on Glass XX, Kyoto, 2004, pp. O-14-028.
[21] Tanabe, S.; Hayashi, H.; Hanada, T.; Onodera, N. Opt. Mat. 2002, vol. 19, 343-349.
[22] Miyakawa T.; Dexter, D.L. Phys. Rev. B 1970, vol. 1, 2961-2969.
[23] Judd, B.R. Phys. Rev. 1962, vol. 127, 750-761.
[24] Ofelt, J.S. J. Chem. Phys. 1962, vol. 37, 511-520.
[25] Tanabe, S.; Photonics Based on Wavelength Integration and Manipulation; IPAP
Books: Tokyo, 2005, vol. 2, pp. 101-112.
[26] Becker, P. C.; Olsson, N. A.; Simpson, J. R. Erbium-Doped Fiber Amplifiers; Academic
Press: NY, 1999, pp 153-195.
[27] Komukai, T.; Yamamoto, T.; Sugawa, T.; Miyajima, Y. IEEE J. Quantum Electron.
1995, vol. 31, 1880-89.
[28] Khoptyar, D.; Jaskorzynska, B. J. Opt. Soc. Am. B 2005, vol. 22. 2091-2098.
[29] Myslinski, P.; Nguyen, D.; Chrostowski, J. J. Lightwave Technol. 1997, vol. 15,112-120.
[30] Bilxt, P. IEEE Photon. Technol. Lett. 1991, vol. 3, 996-998.
[31] Snoeks, E.; van den Hoven, G. N.; Polman, A.; Hendriksen, B.; Diemeer, M. B. J.;
Priolo, F. J. Opt. Soc. Am. B 1995, vol. 12, 1468-74.
[32] Quimby, R.S. Appl. Opt. 1991, vol. 30, 2546-52.
[33] Gapontsev, V.P.; Platonov, N.S. Materials Science Forum 1989, vol. 50, 165-222.
[34] Nilsson, J.; Bilxt, P.; Jaskorzynska, B.; Babonas, J. J. Lightwave Technol. 1995, vol. 13,
341-349.
[35] Masuda, H.; Takada, A.; Aida, K. J. Lightwave Technol. 1992, vol. 10, 1789-99.

Chapter 7
RECENT DEVELOPMENTS IN ALL-FIBRE DEVICES
FOR OPTICAL NETWORKS
Nawfel Azami
Institut National des Postes et Tlcommunications, Madinat Al Irfane,
Rabat-Instituts, Rabat, Morocco.
Suzanne Lacroix
Ecole Polytechnique de Montral, Laboratoire des fibres optiques,
Montral, Qubec, Canada
Abstract
All-fibre components are essential components of optical networks systems.
Development of such devices is of great importance to allow network functions to be
performed in the glass of the optical fibre itself. Among of all fabrication techniques, the
Fused Fibre Biconical Taper (FBT) technique allows optical devices with high performances.
Although fibre devices are mainly based on the passive directional coupler basic structure,
research is made to design components that perform complex functionalities in today optical
networks systems. Recent developments on all-fibre devices in network systems are
presented. Research is mainly focused on enhanced fabrication and stability of FBT
fabrication technique, passive thermal compensation for stable interferometer optical
structure, broadband spectral operation for multi-wavelength operations and new
interferometer designs. An overview of recent fused fibre devices for optical
telecommunications is presented to understand the main functionalities of these fibre devices.
The limiting factors are explained to understand challenges on fibre devices development.
Introduction
The fibre is not only the choice transmitting medium for high speed long-haul
telecommunication. It is also currently used in sensing networks applications and more
recently in quantum information systems. Components are key elements of such networks.
All-fibre devices and their full compatibility with the transmission medium make them
particularly attractive to perform operations such as multiplexing, routing, or filtering with
Nawfel Azami and Suzanne Lacroix 206
low insertion loss, low polarization mode dispersion, and ease of interconnection. In this
chapter, new developments of all-fibre components for optical networks systems are
presented.
A number of techniques have been developed to fabricate all-fibre components. Among
them, the fusion-tapering or, in short, the FBT technique is extensively used especially for the
fabrication of 2x2 couplers. Because of their intrinsic low loss, they offer the possibility of
high power handling (such as in all-fibre lasers), as well as individual photon manipulation
(such as quantum information processing in quantum key distribution and quantum
computing.) Most devices and components considered herein are however firstly designed
for use in standard telecommunication networks. In the first part of this chapter, Fused
Biconical Taper (FBT) fabrication technique is described as well as the basic designs
structures of all-fibre devices.
A basic optical fibre communication system is presented in Fig. 1. An electrical signal
from the data source is fed into the optical transmitter, which is contains a laser or an LED.
The modulated light from the transmitter is launched into the fibre and transmitted to the
receiver via a demodulator. The receiver consists of a light detector with appropriate
amplification and noise filtering. In a digital system a decision gate is also included. Optical
fibres prove economic when good use can be made of the bandwidth that they offer. Optical
Wavelength Division Multiplexing (WDM) and Dense WDM (DWDM) systems have been
developed to perform multi channels propagation in a single optical fibre. Development of
stable multiplexers/demultiplexers is of great importance to combine wavelength channels in
the optical fibre. These types of multiplexers can also be used as demodulators when
Differential Phase Shift keying modulation is used. Designs of all-fibre wavelength
multiplexers/demuliplexers are usually complex since they require techniques for thermal
compensation of the wavelength channel drift. Moreover the sinusoidal spectral response of
basic structures such as tapered fibre couplers or Mach-Zehnder interferometers is not
appropriate. A flattened spectral response is more appropriate since it allows minimizing
insertion loss even when the carrier wavelength drifts. It also reduces crosstalk between
adjacent channels. The second part of this chapter is dedicated to new developments on stable
WDM/DWDM. In particular, passively temperature-independent all-fibre devices techniques
and new design of flat top multiplexers are presented.
During the last twenty years, interest in communicating by sending signals along optical
fibres has grown enormously. This interest lies in the very high capacity of transmission in
optical fibres, the very low attenuation of the signal during the propagation, as well as the
high performances of Erbium doped fibre amplifiers (EDFA) and Raman amplifiers.
Development of these amplifiers allows achievement of multi-channel lightwave systems
with high bit rates performances. For silica fibres, the attenuation is quite small, particularly
in the C-band, between 1525 nm and 1570 nm. Erbium doped optical fibres demonstrated
high performances on amplification of signals with low noise. However, multi-channel
systems need additive components when compared to single-channel communicating systems.
As an example, Erbium gain non-uniformity causes power divergence of WDM channels,
limiting the system performances. Gain flattening filters (GFF) and Dynamic gain equalizers
modules are requested to flatten the amplifiers gain. Development of such devices using FBT
fabrication technique is presented in the third part.
Raman amplifiers have also proved their high capacity of achieving high gain with low
noise. Distributed Raman amplification is very attractive since it allows amplification of
Recent Developments in All-Fibre Devices for Optical Networks 207
signals in the transmission fibre. Because of the high pump power needed in Raman
amplification, all-fibre components are of more benefit thanks to their high optical power
handling. Multi-channel signal systems need multi-wavelength pump lasers when Raman
amplification is used. For this reason, and for system reconfiguration agility, wideband
devices are of particular interest for Raman amplifiers. In the last part, new developments on
large bandwidth all-fibre devices for Raman amplification are presented.
Figure 1. Schematic optical fibre network configuration.
1. Fused Biconical Taper
Fibre optic couplers either split optical signals into multiple paths or combine multiple signals
on one path. The number of input and output ports, expressed as an NxM configuration,
characterizes a coupler, N representing the number of input fibres, and M the number of
output fibres. Fused couplers can be made in any configuration, but the simplest is the 2x2
symmetric directional coupler, which is the equivalent in guided optics of a beam splitter in
bulk optics. Although the most frequent components are 2x2 couplers, tapered single fibres
are also a basic component of interest in themselves and, as such, are studied in the following.
1.1. Manufacturing
The fusion-tapering manufacturing technique consists in fusing laterally two (or more) fibres
together using, as an heat source, a micro-torch, an oven or a CO
2
laser. Depending on the
fusion duration, one obtains a cross section with a degree of fusion ranging theoretically from
zero (for unfused fibres) to 1 (corresponding to a circular cross section theoretically obtained
after an infinite duration). From a practical point of view, the degrees of fusion usually range
between 0.5 and 0.7. Example cross sections are shown in Fig. 2.
0.005 0.25 0.5 0.75 1
Figure 2. Cross sections of 2x2 symmetric fused fibre couplers. The degrees of fusion and indicated
below each cross section. Note the deformation of the cores as the fusion increases.
As shown schematically in Fig. 3, the fused structure is then stretched so as to create a
biconical structure until the desired profile or the desired response is obtained. Tools and
rules for the elaboration of recipes to design specific components are given in Ref. [1,2]. Each
tapering recipe includes several tapering segments, usually no more than 3 or 4, except for
complex concatenated structures, such as Mach-Zehnder interferometers. Apart from the fibre
local temperature, each segment is characterized by four main parameters, namely, the pulling
speed, the elongation, the flame position, and the effective width the flame or length of the
hot zone during the process. Temperatures range is typically between 1450 50 C but may
reach 1700 C. The ends of the tapered structure are usually pulled apart at equal and opposite
speeds relative to the centre of the heat source. As a result, the tapered structure is symmetric
so that the slopes of the down-taper transition and of the up-taper transition regions are
identical. Pulling speeds, typically of the order of millimetres per minute, are usually constant
for a given segment and depend on the fibre temperature. For a given temperature, it is
adjusted so that the fibre neither breaks nor sags during the process. The final elongation of
the component for a given segment determines the end of this particular segment. During the
tapering process, diagnostics are made: the shape of the device is controlled through a
binocular microscope; its optical transmissions (in both arms) are recorded at a given
wavelength as a function of elongation or for a whole range of wavelengths using a
broadband light source and an Optical Spectrum Analyzer (OSA) as a detector.
Broadband
light
source
Diagnostics
(OSA)
Heat source
(flame, CO
2
laser, oven)
Streching motors
Figure 3. Manufacturing of a 2x2 coupler using the FBT technique.
1.2. Adiabaticity Concept
The slopes of the longitudinal structure largely determine the behaviour of the component.
The propagation along a tapered fibre is said to be adiabatic whenever the fibre transmission
is not affected by the taper slope [2]. This is only possible for gentle slopes. In contrast, when
the slopes are abrupt, transfer of power to higher mode may occur. This is, from a general
point of view, undesirable for couplers as this causes power leakage. An adiabaticity criterion
is derived for every particular structure, whether a single fibre, or a coupler made of two or
more fused fibres. The fused fibres may be identical to create a symmetric structure or not, in
the more general case of asymmetric couplers. The adiabaticity criterion provides the upper
limit normalized slope that a structure may have for an adiabatic behaviour. Details of the
calculation and graphical representation of adiabaticity criteria are given in Ref. [1,2]. For
most structures made of standard 125 m diameter fibres, the limit slope is of the order of 10
-
3
m
-1
. While adiabaticity is usually required for couplers, non-adiabaticity of tapered single
fibres can be used to design a variety of all-fibre spectral filters (see Section 3.1). Their
principle of operation is overviewed below.
1.3. Non-adiabatically Tapered Fibres
When one tapers a fibre with steep slopes, e.g. heating over a zone of the order of a few
millimetres, one observes oscillations in the transmitted power as a function of elongation, at
a given wavelength. For a given elongation, these oscillations are also present as a function of
wavelength. As explained in more details below, this is the result of the alternation of local
modes coupling and beating effects along the tapered structure.
In the downtaper region, as the fibre diameter decreases, the fundamental LP
01
core mode
expands in the cladding. When the diameter is reduced by a factor of 2 or more, the
fundamental mode becomes guided by the cladding-air interface. The mode is said to be cut
off as a core mode: it becomes a cladding mode. If the slope is steep, some power is
transferred to other cladding modes (LP
02
, LP
03,
...) by coupling effects.
In the central region, where the slopes are small, the adiabaticity criterion is again obeyed
and the excited LP
0m
modes, all of them being cladding modes, accumulate phase differences
through the beating effect.
While arriving on the uptaper region, mode coupling again occurs before the power is
finally recovered in the core. Depending on the relative phase of the excited LP
0m
modes,
(therefore on the wavelength and on the elongation) power may be partially or totally
recovered in the core. All the power, which is not recovered in the core, is in the cladding
modes and possibly trapped by the protective jacket of the fibre, thus lost.
This process of coupling-beating-coupling thus confers to a tapered fibre an oscillatory
behaviour according to the various parameters affecting the modal phase differences
accumulated mostly in the beating region. The LP
01
and LP
02
modes are responsible for the
main oscillation. Higher order modes (LP
03
, ...) possibly superimpose to it smaller amplitude
and larger frequency oscillations. For a pair of modes, e.g. LP
01
and LP
02
, the wavelength
response is essentially sinusoidal and it is exploited to design spectral filters, such as those to
flatten the Erbium doped fibre gain described in Section 3.1.
1.4. Transfer Matrices of 2x2 FBT Symmetric Couplers
In the following, the principle of operation of adiabatic 2x2 FBT symmetric couplers is
overviewed. For a coupler made of individual guides in close proximity, the power transfer
from branch to branch is usually analyzed in terms of coupling between the modes of the
individual guides. However, in the case of FBT couplers, it is necessary to call for
supermodes. For a 2x2 symmetric coupler (the only coupler considered herein for the sake of
simplicity), these are referred to as SLP
01
and SLP
11
, respectively. They are the fundamental
and the first asymmetric modes of the superstructure, i.e., the fused structure. Note that the
adiabaticity criterion, which is supposed to be obeyed, refers to these supermodes. This
concept of supermodes is essential for the following reasons.
For the power transfer to occur, the fused structure is tapered down to a diameter such
that the cores no more play their guiding roles.
As the cores are reduced, the fields spread out of the cores and the guiding process is
ensured by the cladding-air interface.
As a result, individual guides can no more be identified in the fused and tapered region,
where the power transfer occurs.
The transfer of power is then described in terms of a beating phenomenon between the
fist two supermodes SLP
01
and SLP
11
, which are equally excited at the entrance of the coupler
when light is launched only one of the entrance branch. Due to their different propagation
constants, they accumulate a phase difference along the structure. Whenever they are in
phase, the power is retrieved in the main branch, while, whenever they are out of phase, the
power is retrieved in the secondary branch. An intermediate phase difference value
corresponds to a branching ratio between 0 and 1. More quantitatively, a coupler is
characterised by its transfer matrix
M
= e
i
cos i sin
i sin cos

(1)
where

is an average common propagation phase, which is, for this reason often omitted. It
is defined as
2 =
0
L
(
01
+
11
)dz
(2)
and 2 is the accumulated phase difference between both supermodes along the length of the
coupler
2 =
0
L
(
01
11
)dz
(3)
01
and
11
being, in these formulas, the propagation constants of the supermodes SLP
01
and
SLP
11
, respectively. Note that these propagation constants are wavelength dependent, which
confer the coupler a spectral dependence. The transfer matrix relates the amplitudes in the
two exit branches to those in the entrance branches. For example, an excitation in a single
branch corresponds to the entrance vector
1
0

and thus to an exit vector
e
i
cos i sin
i sin cos

1
0

= e
i
cos
i sin

Note the i factor, corresponding to a /2 phase factor between both branches, which is
unusual referring to the analogy between a fibre coupler and a beam splitter.
The corresponding intensity transmissions in main and secondary branches, respectively
labelled 1 and 2, are
T
1
= cos
2
=
1+ cos2
2
T
2
= sin
2
=
1cos2
2
(4)
A 50%/50% splitter, also referred to as 3 dB coupler, is thus a coupler having
=/4+p/2, with p integer (usually null, to ensure small spectral dependence).
Due to the lack of circular symmetry of the guiding structure, the couplers are inherently
polarisation dependent. The cross section has two symmetry axes x and y, which define the
principal polarisation axes. The coupler transmissions must more generally be written as a
superposition of transmissions in each polarisation
T
1
= T
1x
+ T
1y
=cos
2
x
+ (1)cos
2
y
T
2
= T
2x
+ T
2y
=sin
2
x
+ (1)sin
2
y
(5)
where and (1-) are the proportions of power launched in the x and y polarisations,
respectively. However, strongly fused couplers are virtually polarisation insensitive inasmuch
their waist is not too small. This is the case of most 3dB and other standard beam splitters, the
polarisation dependence of which is ignored.
The transfer matrices are very useful tools to predict the responses of more complex
structures made of concatenation of several couplers. The simplest one is the Mach-Zehnder
(MZ) interferometer made of two concatenated couplers, which may be different, thus
characterised by . Such an interferometric structure is sketched in Fig. 4.
Figure 4. All-fibre Mach-Zehnder structure. The phase difference between the arms =
1
L
1
2
L
2
is
realised through a length difference L
1
L
2
and/or a propagation constant difference
2
=2(n
1
n
2
)/c.
For a phase difference between the two MZ arms, the transmission column vector
(containing individual guide amplitudes) may be calculated by using matrix products as
follows
0
1
cos sin
sin cos
.
0
0
.
' cos ' sin
' sin ' cos
2
2
'

i
i
e
e
e
i
i
e
i
i
i
i
(6)
The intensity transmissions in each branch are then easily derived to be
T
1
=
1
2
(1+ cos2 cos2' sin2 sin2' cos)
T
2
=
1
2
(1 cos2 cos2' +sin2 sin2' cos)
(7)
The different parameters , , and give a flexibility to design a variety of different
components with specific functionalities. As an example, the DWDM components are
examined in Section 2. In this case, one has the couplers parameters ==/4, which are
almost wavelength independent over the range of interest.
2. Stable Wavelength Division Multiplexer All-Fibre Devices
WDM Mach-Zehnder Interferometers (MZI) are extensively used as multiplexer,
demultiplexer, add-drop modules, and in many other applications. For most of these
applications, a control of the thermal dependence of the refractive index is required. In order
to simplify the description of the MZI transmission as a function of the optogeometrical
parameters of the two fibres (Fig. 4), let us suppose an ideal MZI with no loss and an infinite
isolation by using 3dB couplers. Using eq. 7, the transmittivity from port 1 to port 2 can be
written as:
= ) (
2
cos ) (
2 2 1 1
2
L n L n
c
T

(8)
where is the signal frequency, c is the light velocity, L
1
and L
2
are the lengths of fibres 1
and 2 in the central zone respectively, and n
1
and n
2
their effective indices. The Free Spectral
Range () and the p
th
transmission peak frequencies
p
are given by:
( ) ( )
2 2 1 1 2 2 1 1
2
.
2

L n L n
c
p
L n L n
c
p
=
(9)
Inter-channel spectral distance is then induced by fibres with different refractive index
profiles or/and different lengths. MZIs are known for their narrow band capabilities. For this
purpose, they must be stable over a range of environmental conditions, such as temperature,
within a defined range in case of temperature variations. However, the refractive indices or
the optical path lengths of the two connecting fibres of the device between the two couplers
usually vary with temperature. If the thermooptic coefficients (i.e. the temperature
dependence of their refraction indices) of the two fibres are not equal or if the optical paths of
the two fibres are not equal, the temperature variations cause variations in the differential
phase shift. Consequently, the channel spacing of the device, defined as the wavelength
separation between the transmission peaks of two adjacent channels, as well as the peak
wavelengths and the pass-band, become unstable. This would cause significant problems for
WDM applications, due to the small separation between channels. In the next section, results
of temperature-independent all-fibre MZI manufactured with the FBT technology are
presented.
The thermal dependence of an optical fibre can be expressed with the aid of the thermo-
optic coefficient, which describes the change of the index of refraction with the change of
temperature dn/dT. If both arms in the central zone are equal (L
1
=L
2
=L), the thermal shift of
the MZI transmission peaks is given by:
d
p
dT

p
1
n
dn
1
dT
dn
2
dT
1
L
dL
dT
+
(10)
where n= n
1
n
2
and L is the length of the central zone of the MZI. The thermal expansion
coefficient for silica (L
-1
.dL/dT) is about 5.10
-7
C
-1
. The contribution of the thermal
expansion of silica fibre to the thermal shift of the MZI transmission peaks is nearly
0.75 pm/C for a transmission peak at 1.55 m. Thermal expansion of silica is usually
neglected in Mach-Zehnder interferometer structures and is not an issue for thermal
compensation. However, thermal expansion of silica is of great importance in tapered
couplers design because of the impact on supermodes propagating index.
2.1. Passive Thermal Compensation Using UV Treatment
It is well known that photosensitive fibre hydrogenation may produce large refractive index
changes if the fibre is exposed to UV radiation [3,4]. This process has been extensively used
for fabrication of Bragg gratings and balanced MZI. More recently, it has been shown that
hydrogenation of an optical fibre followed by UV exposure can control the thermal
dependence of the refractive index. This may be used in a device, such as an all-fibre MZI. In
the following section, the process applied to one fibre-arm of the MZI is presented. This
process, applied before the fabrication of the MZI, consists of hydrogenation and exposition
to UV radiation. The optical fibre is put in a pressure chamber, filling the chamber with
hydrogen at a suitable pressure (about 1800 psi) and left there for a period of time suitable to
achieve the desired photosensitivity (about 12 hours). This process produces an increase in
the index of refraction of the fibre, which becomes n+dn [3,5,6]. Thereafter, the
photosensitive fibre is exposed to UV radiation. As is mentioned in ref. [6], such an exposure
can lead to a further increase of the fibre refractive index. It has been recently found that one
can control or adjust the thermal dependence of the optical fibre by controlling the UV
exposure time of the photosensitive fibre [7]. Moreover, it has been discovered that the
change of the thermal dependence provided by this method remains constant even though the
index of refraction is further changed, for example by exposing the fibre to heat. Thus, by
heating the fibre to a temperature greater than 800 C, one can bring down the fibre index of
refraction back to the value n, without affecting the adjusted thermal dependence.
As a demonstration, all-fibre MZIs were fabricated using the treated fibre (fibre 1) and a
dissimilar fibre (fibre 2) with different refractive indices (Fig. 5). Thermal dependences of the
fibres are also different. Fig. 6 illustrates the change of the thermal dependence of the all fibre
MZI as a function of the UV exposure time of the hydrogenated fibre (Corning SMF-28).
The thermal dependence of the MZI does not change during the first 5 minutes of exposure to
UV radiation. The change in thermal dependence then starts to occur gradually and continues
more steeply as shown in Fig. 6. Between 10 and 25 minutes of exposure, the thermal
dependence change is essentially linear. As is shown in Fig. 6, the reproducibility of the
thermal dependence is good. The small variability of the thermal dependence may be due to
variations in the fabrication process of the MZI (for example small variations of the
temperature from device to device) and also to variations in the final free spectral range of the
MZI that have been tested (20nm 1nm).
3 dB 3 dB
Fibre 1
Fibre 2
Figure 5. All-fibre Mach-Zehnder interferometer with different fibres.
-20
-10
0
10
20
30
0 5 10 15 20 25 30 35
Time of exposure to UV (min)
T
h
e
r
m
a
l

d
e
p
e
n
d
e
n
c
e

(
p
m
/
C
)
Figure 6. Experimental thermal dependence of all fibre MZI as a function the UV exposure time of one
of its fibre arms.
2.2. Passive Thermal Compensation Using Specific Dopants in Fibres
The cores of silica fibres are usually doped with Germanium to increase the refractive index
with respect to the undoped cladding. However, many other dopants can be used to control
the refractive index, such as Fluor or Phosphorus. Concentrations of such dopants in the fibre
core or cladding have direct impact on the effective index of the fibre and on the temperature
dependence of this effective index. The adjustment of the composition with dopants can take
place in the core of the fibre or in the cladding or both. It has been demonstrated that the type
of dopant used and its concentration can be selected to control the thermal wavelength drift of
a MZI to about 1-2 pm/C accuracy within a desired temperature range which is generally
between about -35C and +85C [8].
2.3. Flat-top WDM Devices
WDM optical systems allow multi-channels communication in a single optical fibre. A
channel is spectrally characterised by a wavelength and a width. Channel spacing in WDM
systems is constantly decreasing and can be as low as 25 GHz. In many cases, sinusoidal
spectral response of multiplexers/demultiplexers is not appropriate, especially in long haul
optical networks where tight specifications on insertion loss, crosstalk and differential group
delay are required. Moreover, fluctuations of the signal laser wavelength may induce loss
fluctuations in Dense Wavelength Division Multiplexing (DWDM) systems when a
sinusoidal transmission device is used. Flat-top spectral responses are preferred because they
allow minimizing the crosstalk between adjacent channels, the fluctuations of channel loss,
and the differential group delay. Thin films DWDM devices can be easily designed to meet
Figure 7. Typical spectral response of flat top interleaver with three cascaded couplers.
flat top channel specifications. However, all-fibre DWDM devices are still attractive because
of their low polarization mode dispersion. In their usual two couplers Mach-Zehnder
interferometers configuration, only sinusoidal spectrum can be achieved. Flat-top channel
spacing multiplexer has been implemented by three cascaded couplers of different coupling
ratios linked by two differential delays [9,10,11]. A non sinusoidal spectrum is obtained by
using three couplers instead of the usual two, while the second differential delay is exactly
twice the first one. Cascading wavelength insensitive couplers allow constant isolation and
insertion loss over more than 100 nm. As an example, Fig. 7 shows typical optical spectra of
a 100 GHz spacing DWDM interleaver consisting of three cascaded couplers.
2.4. All Fibre Optical Add Drop Module
Optical Add-Drop Modules (OADMs) are key devices for optical networks. OADMs are the
access points to the optical network and allow adding or droping wavelengths at different
sites along the network. The most usual all-fibre design used a balanced MZI with two
identical Fibre Bragg Gratings (FBGs) embedded in the two MZI arms [12]. Optical signals
are launched into port 1 (Fig. 8). The 3dB coupler splits the input power evenly into the two
MZ arms. Only those signals carried at the Bragg wavelength get reflected by the FBGs and
return back into the first 3dB coupler. Whenever the optical paths of both reflected waves are
balanced, all the wavelengths over the bandwidth of interest are phase-matched and all the
optical energy is transferred into port 4 with little energy returning back to the bar path (see
eq. 7 with =/4 and =0). The port 4 becomes the drop-port, at which signals at the Bragg
wavelength of the FBGs get filtered out from other channels. Signals carried at wavelengths
other than the Bragg wavelength transmit through the FBGs and merge into the second 3dB
coupler. Similar to the reflected one, all the transmitted waves over the wavelength span of
interest are phase-matched under a balanced MZ structure and most of the energy is carried
into port 3. Port 3 then becomes the pass-port, through which signals outside of the FBG
reflection band are transmitted. Port 2 can then be used as the add-port, into which other
signals carried at the Bragg wavelength are launched. Those additional signals get reflected
by the FBGs, carried through the cross path arm of the second 3dB coupler, and join port 3
without interfering with each other
The most common fabrication method approach is that a MZI is made first and the FBG
pair is then written on the established interferometer [13-15]. Another approach for which
available FBGs are integrated into a MZ interferometer has also been demonstrated [16].
3 dB 3 dB
1
4

g
g
-dropp
g
-add
IN
OUT
2
3
Figure 8. Balanced all-fibre Optical Add Drop Module.
2.5. All Fibre Differential Phase Shift Keying Demodulator
The differential phase shifted keying (DPSK) modulation has been attracted great attention
for its application for dense wavelength division multiplexing (DWDM) transmission, since
DPSK with optical Mach-Zehnder interferometer (MZI) demodulation provides several
advantages over the conventional intensity modulation detection [17]. Optical differential-
phase shift keying (DPSK) is a modulation format that offers high receiver sensitivity, high
tolerance to major nonlinear effects in high-speed transmissions, and high tolerance to
coherent crosstalk [18,19]. In DPSK, data information is carried by the optical phase
difference between successive symbols. As an example, a Conventional DPSK (CDPSK) uses
phase difference in the set (0,) [20].
Figure 9. DPSK demodulator. (a) successive symbols with -phase difference. (b) successive symbols
in phase.
For direct detection of DPSK signal (by conventional intensity detectors), a DPSK
demodulator is used to convert the phase-coded signal into an intensity-coded signal. Fig. 9
illustrates demodulation of a DPSK optical signal using 1-bit-time-unbalanced Mach-Zehnder
interferometer designed with 3 dB couplers (also called delay line interferometer). The
incoming differential phase-shift keying optical signal is first split into two equal-intensity
beams in two arms of a Mach Zehnder, in which one beam is delayed by an optical path
difference corresponding to 1-bit time delay. After recombination, the two beams interfere
with each other constructively or destructively, depending on the optical phase difference
between adjacent bits. Using eq. 7 with ==/4 (3 dB couplers) and the phase difference
corresponding to 1 bit time, one can easily show that the resultant interference intensity of
two adjacent bits in phase is directed in port 3 (output port), while the resultant interference
intensity of two adjacent bits having -phase difference is directed in port 2. The resultant
interference intensity is the intensity-keyed signal in output port 3. The all-fibre Mach-
Zehnder interferometer design demonstrated low insertion loss, low polarization dependent
loss and low polarization dependent isolation over a wide spectral band by using 3dB
wavelength insensitive couplers [21]. Tunable all fibre DPSK demodulators are also very
attractive since they allow re-configuration of wavelength channels. In an all-fibre structure,
phase tuning is typically achieved by applying an electrical voltage to one arm of the MZ that
has been metallized to this aim. The refractive index of the metallized arm change because of
the thermo-optic effect, which allows tuning of the phase difference .
3. Developments on All-Fibre Devices for Erbium Amplifiers
Erbium-based optical amplifiers have been developed during the 1980s to replace the
expensive and complex electronic repeaters. The advantage of Erbium-doped fibre amplifiers
(EDFAs) lies in the practical issues related to coupling losses, polarization insensitivity, high
gain, low noise, and capability to regenerate several channels simultaneously. However,
EDFAs need components for their integration in optical networks. As an example, EDFAs
usually incorporate a gain equalizer filter to flatten the gain spectrum. Because of their high
performance Erbium amplifiers are also used in a two-stage configuration. The mid-stage
allows incorporating many devices to optimize the network performance, such as a chromatic
dispersion compensation fibre, a Polarization Mode Dispersion (PMD) compensation module,
a dynamic gain compensator, a gain flattening filter, and add-drop modulesFig. 10
illustrates a basic configuration of a two-stage EDFAs.
Pump/Signal combiner
EDFA
Isolator
EDFA
Mid-stage
GFF
DGC
pump pump
Figure 10. Two-stage EDFA basic configuration.
3.1. Fibre Gain flattening Filters
All-fibre amplifiers are commonly used in telecommunication networks to amplify signals on
a wide bandwidth. Filters are required to flatten the non-uniform gain of EDFAs or Raman
amplifiers. Fixed gain flattening filters (GFFs) flatten the gain profile of optical amplifiers by
selectively removing excess power. These filters are often fabricated using short- or long-
period fibre gratings. However, efficient gain flattening filters can also be achieved with a
cascade of tapered fibres [22]. As discussed in Section 1.3, abruptly tapered fibres allow the
coupling between the fundamental mode and several cladding modes. The controlled taper
profile is used for tailoring the filter loss spectrum. Tapered fibres show a sinusoidal spectral
response. They can thus be combined to create any spectral filter as in a Fourier series
(Fig. 11).
GFF are static filters used to flatten the amplifier spectrum gain for a given gain shape.
Hence, GFFs are designed for given operating conditions, such as total signal power, pump
power, number of channels, and temperature conditions. When these parameters vary, the
gain shape of the amplifier also varies, and the GFF is no longer able to flatten the amplifier
spectrum.
Figure 11. Spectral response of a gain flattening filter for erbium-doped amplifiers made by
concatenation of four tapered fibre filters.
3.2. Dynamic Gain Tilt Compensation
In this section, we focus on recent development of the EDFA gain control using all-fibre
devices. An optical amplifier may not always operate at the gain value for which the gain
flatness is optimized. Many factors contribute to this sub optimal operating condition: span-
loss variation, input channel count change, and spectral tilt due to stimulated Raman
scattering. As a result, the amplifier gain is tilted, and such tilt can have significant impact on
the system performance. Generally, spectral gain flatness of an EDFA due to change of
operating conditions is characterized by the Dynamic Gain Tilt parameter (DGT). DGT
(dB/dB) is defined as the gain variation at wavelength ,when the gain variation at a reference
wavelength
0
is 1 dB.
) (
) (
) (
0
G
G
DGT
=
(11)
The DGT is a characteristic function of erbium ions and do not depend on fabrication
techniques or opto-geometrical parameters of the fibre. It is then an efficient parameter to
characterize the variation of the gain in EDFAs. It can be easily shown that the DGT is a
function of the absorption and emission Giles parameters:
) ( ) (
) ( ) (
) (
0 0

s
s
g
g
DGT
+
+
=
(12)
In the case of C-band EDFAs, the dynamic gain-tilt (DGT) is the main factor of the gain
flatness deterioration, especially in the case of 980 nm pumping. Another important amplifier
control function is to maintain the output signal level per channel. The dynamic control of the
per-channel output power in the EDFA is important to avoid SNR degradation. Many designs
have been proposed for dynamic gain tilt compensation. Variable optical attenuator (VOA) is
the most common device used for gain tilt compensation [23]. However, the large insertion
loss of the VOA deteriorates the signal to noise ratio and/or the power conversion efficiency
of the amplifier. Automatic power control (APC) scheme and a variable attenuation slope
compensator (VASC) demonstrate better performances than the VOA [24]. The APC is
employed in the first EDFA stage and the VASC in the mid-stage does not change its
insertion loss in spite of the attenuation slope change. In reference [25], an all-fibre Mach-
Zehnder interferometer with appropriate couplers is presented. The design allows dynamic
gain tilt compensation by only changing the isolation of the interferometer while the centre
wavelength remains unchanged. The all-fibre structure allows high optical performance
including low insertion loss, low polarization dependent loss and low polarization mode
dispersion. The gain slope tuning is made using the thermo-optic effect while the device still
passively insensitive to external temperature variations. For illustration, this all-fibre device is
presented in the following.
Figure 12. Mach-Zehnder interferometer with metallised fibre in one arm.
A MZ can be used in a linear spectral region to compensate the gain tilt of an EDFA. The
MZ couplers are identical and wavelength dependent such that 0 dB insertion loss is realized
at
0
(1520 nm) and 3dB is realized at
1
(1580 nm). The MZ is then characterized by a
minimum insertion loss at
0
and a maximum insertion loss at
1
. One of the two branches of
the MZ is metallised to allow phase tuning between the two arms of the interferometer
(Fig. 12). Applying an electric voltage allows to increase the temperature, and thus change
the refractive index of the fibre. As a result, phase changes occur between the two arms via
thermo-optic effect, allowing the change of the slope in the 1530-1570 nm spectral band. The
phase change has only an impact on the isolation at
1
, while central wavelengths stay
unchanged. Figure 13 shows the spectral response of the MZs for different phase differences
induced between the two branches. Initial configuration is such that the isolation of the
components is 0 dB. A flat transmission near 0 dB over the C-band is realised for no applied
voltage. We focus on the 1540-1565nm spectral band where the EDFA has a linear DGT. The
maximum insertion loss at 1540nm is 1 dB for 5 dB gain tilt. The deviation from linearity is
less than 0.25 dB. The polarization dependent loss (PDL) is less than 0.3dB. A maximum of
3 Volts allow 5 dB tilt between 1540 nm and 1565 nm. The response time, defined by the
characteristic time allowing a change of the attenuation slope from 5 dB to 5/e dB was 210
ms.
-25
-20
-15
-10
-5
0
1500 1550 1600
Wavelength (nm)
d
B
-6
-4
-2
0
1540 1550 1560
Wavelength (nm)
d
B
-25
-20
-15
-10
-5
0
1500 1550 1600
Wavelength (nm)
d
B
-6
-4
-2
0
1540 1550 1560
Wavelength (nm)
d
B
Figure 13. (a) Transmission of MZI for different applied voltage. V=0, 1, 2.25, 3 and 3.6 Volts. (b)
Zoom of the 1540-1565 nm range.
Two concatenated MZIs are required to allow positive and negative slope compensation
as well as a constant average insertion loss for any slope. The first MZI has optical
characteristics presented in the previous section. The second MZI has a complementary
transmission (couplers with 0 dB at
1
and 3 dB at
0
). ). A great advantage of this all-fibre
DSC is that a very low electrical power is needed to compensate the gain tilt. A maximum
total electrical power of 250 mW is needed due to the low thermal conductivity of silica.
4. Developments on All-Fibre Devices for Raman Amplifiers
Raman Fibre Amplifiers (RFAs) are of great interest for the development of long distance,
high capacity WDM systems. Their main advantages are their low noise, wide amplification
bandwidth and saturation characteristics. RFAs have also the advantage that the optical
amplification occurs in the transmission fibre transmission itself. RFAs differ in principle
from EDFAs as they utilize the stimulated Raman scattering effect to create optical gain.
However, RFA suffer from polarization dependent gain (PDG). A solution to reduce PDG is
the use of pump laser with low degree of polarization (DOP). One can scramble the state of
polarization of the pump with the aid of a depolarizer. Experimental and theoretical
investigations have been reported on the statistical properties of PDG [26-30]. These reports
show that the PDG is linked to the PMD of the fibre. Fig. 14 illustrates the basic design of an
optical fibre system using Raman amplification. In a RFA, a strong pump laser provides gain
to signals at longer wavelengths through stimulated Raman scattering. One of the major
attractions of Raman amplification is that it can be used over a very wide wavelength range
by multiplexing together different pumps wavelengths.
Figure 14. Raman amplification configuration.
Polarization beam combiners (PBC) are key component of diode-pumped Raman
amplifiers. They allow combining the output of two pumps operating at the same wavelength
but in different polarization modes into a single fibre, thereby doubling the effective power
output. PBCs can also be used in EDFAs, typically to combine 1,480 nm pumps in the second
stage. The combination of the two linear-orthogonal polarizations having the same power,
allows in addition a complete depolarization of the output pump. Moreover, polarization
combiners allow protecting the system from the failure of any single laser. RFAs may also be
subject to dynamic reconfiguration of the pumps lasers wavelengths. Broadband polarization
combiners can be necessary to ensure system reconfiguration. Optical depolarizers play also a
key role by scrambling the state of polarization of wave pumps that are not doubled.
4.1. Broadband All-Fibre Polarization Combiners
The fibre optic coupler made by the FBT technology has been one of the most widely used
devices in optical fibre systems. Other than the most common function of optical power
splitting, such couplers may have other applications. In particular, they can be designed to
operate as optical polarization beam splitter/combiners [31,32]. For example, optical
networks use optical polarization beam splitters in their PMD compensator modules, and
pump depolarizers in Raman amplifiers. A large bandwidth is usually specified in case of
multi-channel lightwave systems. Ideally, a bandwidth as large as the complete optical band
(S+C+L) is suitable. However, all-fibre polarization beam splitters suffer from a narrow
bandwidth, which limits the spectral operating wavelength range to a few nanometers. A
spectral width of 17 nm for 15 dB extinction ratio has been reported [33]. A wider spectral
width of 38 nm for 15 dB extinction ratio has been demonstrated using a weakly fused
coupler design [34]. Recently, an all-fibre polarization beam splitter/combiner has been
reported on a very wide band of more than 200 nm for 15 dB extinction ratio [35]. In the
following section a brief description of this device is presented.
An all-fibre MZI design is used for which specific couplers are designed. The first
coupler is an all-fibre polarization beam combiner (PBC) and the second coupler is an all-
fibre WDM coupler. The structure is schematized on Fig. 15. Input fibres (fibres 1 and 4) are
polarization-maintaining fibres (PMF). The central zone of the MZI (fibres A and B) and
outputs (fibres 2 and 3) are standard single mode fibres (SMF). In regard to Fig. 15, the x-
polarization is coupled into port 4 and the y-polarization is coupled into port 1. Experimental
transmissions from input ports 1 and 4 to output ports A and B are illustrated in Fig. 16 for
both polarizations. By using transfer matrix formalism, the transmitivity of the MZI is given
by
) ( ). ( ). (
,
,

Y X
PBC MZ WDM Y X
M M M M =
(13)
where the transfer matrix of the WDM and the PBC are given in section 1.4.
Let us note
x,y
() and
w
() the phase difference between the symmetric and anti-
symmetric super-modes of the fused fibre PBC for x- and y-polarizations, and the WDM
coupler respectively. These parameters are defined in section 1.4 (eq. 2). The central zone of
the interferometer structure is characterized by a -phase shift between the two branches of
the MZI (fibres A and B). It has been shown that, If
w
()=
x
() and = then the device
transfer matrix for x- and y-polarizations becomes wavelength independent. The spectral
dependence of the PBC can be counterbalanced by a -phase between the two MZI arms
while using a WDM coupler having the same spectral dependence transmitivity as the PBC.
Figure 15. Wideband polarization combiner design.
Fig. 17 shows typical experimental spectral transmissions of the interferometer structure.
The non-perfect extinction ratio at the input ports (PM fibres 1 and 4) induces polarization
beating and spectral ripples on the device transmission. The extinction ratio at the PM fibres
inputs is estimated to be 30 dB. For such an extinction ratio, the ripples are only observed on
the isolation ports. The insertion loss is lower than 0.2 dB in the 1440 nm-1550 nm-
wavelength range. Insertion loss increases near 1440 nm because of the water absorption
peak, which induces excess loss, and increases above 1550 nm because the spectral
transmissions shift between the PBC and the WDM coupler degrades the isolation. More than
200 nm spectral width for 15 dB extinction ratio is achieved.
Figure 16. Experimental transmissions of PBC. 1 and 4: input ports; A and B: output ports.
-30
-25
-20
-15
-10
-5
0
1400 1450 1500 1550 1600
Wavelength (nm)
d
B
X(4-3)
Y(1-3)
X(4-2), Y(1-2)
-30
-25
-20
-15
-10
-5
0
1400 1450 1500 1550 1600
Wavelength (nm)
d
B
X(4-3)
Y(1-3)
X(4-2), Y(1-2)
1 and 4: input ports; 2 and 3: output ports.
Figure 17. Experimental spectral transmissions of the interferometer structure.
4.2. Stable All-Fibre Depolarizers
Passive depolarizers are used to scramble the State Of Polarization (SOP) of an incoming
light source, reducing the mutual coherence between the orthogonal polarization components
of the light source. Highly birefringent (Hi-Bi) fibres are usually used to depolarize wide-
band sources but are not suitable for narrow-band sources because of the long length required.
Recently described passive devices are based on incoherent fibre ring structures, using a
cascaded fibre ring design [36] or a dual fibre ring design [37]. These designs allow
depolarizing any SOP by using directional couplers in a fibre ring scheme and by adjusting
the ring birefringence. Recently a new technique for the single-mode fibre depolarizer based
on polarization combiners using a linear design (as opposed to ring design) has been
proposed. The use of a linear design allows a one-way propagation of the two orthogonal
polarization components, which make the device very stable. This new design provides low
loss, low polarization dependent loss and high depolarization of light for any input SOP. The
assembly of such a device is made using power light detection that makes the integration
device easier than the degree of polarization optimization technique.
Figure 18. All-State of Polarization all-fibre depolarizer design.
The all-SOP all-fibre depolariser linear design presented in ref [38] is a combination of
two polarisation combiners (PC) and a 2x2 directional coupler (Fig. 18). An optical phase
delay (delay1) is induced between the waves propagating in the two branches A and B. A
polarisation rotator device is used to realise half- rotation of the light wave SOP propagating
in fibre B. Interference occurs at the coupler since the SOPs at the inputs of the coupler are
parallel. The average intensities at the outputs of the 2x2 coupler are equal if the delay
induced is much greater than the coherence length of the light source. A second phase delay
(delay2) is induced between the waves propagating in A and B fibres. A half- rotator device
is used such that the SOP of the wavelength propagating in A and B fibres are orthogonal and
aligned with the eigen axis of PC
2
. To increase the polarisation scrambling at the output of the
depolariser, the condition on equal average intensities I
X
and I
Y
has to be satisfied. If the
optical delay induced by delay1 is much greater than the coherence length of the source and if
a 3 dB transmission coupler is used then the average intensities for x and y-polarisations at the
input of PC
2
are equals for any SOP at the input. Also, the two orthogonal components x and
y are completely uncorrelated when the delay lengths (L
1
and L
2
) as well as the difference
length (L
1
-L
2
) are greater than the coherence length of the laser
DOP less than 4.5% is obtained for coherence length less than 1 m. The linear design
eliminates the re-circulations in PC or coupler fibre occurring in a fibre ring configuration.
The loss is then given by the sum of the losses of each subcomponents. Low loss (1.5 dB),
and low polarization dependent loss (0.5 dB) over 14001500 nm spectral range, for all
SOPs, and for a 070C temperature range, has been demonstrated with this type of design.
By using a wide band PC, the depolariser presented is made achromatic over a 100 nm
spectral band (Fig. 19). The one passage light propagation in symmetrical branches (identical
fibres) makes it very stable. Although the PM fibre is often temperature sensitive the small
length of half wave PM fibre used as a rotator device (1.8 mm) keeps the SOP thermally
stable. The DOP variation obtained is +/-2% and loss variation is 0.05 dB for a 0C to 70C
temperature range. In addition, the all silica-fibre structure allows the depolarisation of any
laser with a coherence length lower than the loop length and permits high power handling. By
design, this depolariser allows 2 inputs and 2 outputs, for each input corresponds an output.
Figure 19. Maximum degree of polarization (for any input SOP) versus wavelength.
Conclusion
Fused Biconical Taper fibre devices have shown great integration in today optical networks.
From basic directional coupler to cascaded Mach-Zehnder interferometers designs, all fibre
components are used to perform many functionalities in all parts of the optical network. In the
transmission part, temperature-independent WDM and DWDM interleavers with flat-top
spectrum are attractive because of their very low chromatic dispersion, differential group
delay, and polarization dependent loss. In the amplification part, all-fibre devices based on
fused biconical taper fabrication technique demonstrated high potential to multiplex pumps
and signals. Cascaded tapered fibres and cascaded couplers demonstrated their capability to
correct the amplifier gain non-uniformity being respectively used as Gain Flattening Filters
and dynamic gain compensators. For Raman amplification, broadband polarization combiners
and depolarizers have been developed to limit gain and channels power fluctuations while
high power handling and wide spectral band are ensured, which allow broadband
multichannel amplification and amplifiers re-configuration. Access to the network can be
performed anywhere along the fibre by using all-fibre optical add-drop modules while
polarization mode dispersion is minimized compared to circulators based design. Many other
functionalities can be performed by all-fibre devices.
References
[1] S. Lacroix, N. Godbout, and X. Daxhelet, Fused Biconical Components, Chapter 6 of
Optical Fiber Components: Design and Applications Editor Habib Hamam, Research
Signpost (2006).
[2] W. J. Stewart, and J. D. Love, Design limitation on tapers and couplers in single-mode
fibres, in 5th Int. Conf. Integrated Opt. &Opt. Fibre Commun., 11
th
European Conf.
Opt. Commun., IOOC/ECOC'85, ed. Instituto Internazionale delle Comunicazioni,
Venice, Italy, 1985, pp 559-562.
[3] G. Meltz, W. W. Morey, and W. H. Glenn, Formation of Bragg gratings in optical
fibers by a transverse holographic method, Opt. Lett., 1989, vol. 14, pp 823825.
[4] K. O. Hill, Y. Jujii, D. C. Johnson, and B. S. Kawasaki, Photosensitivity in optical fiber
waveguides: application to reflection filter fabrication, Appl. Phys. Lett., 1978, vol. 32,
pp 647 649.
[5] M. Douay, W. X. Xie, T. Taunay, P. Bernage, P. Niay, P. Cordier, B. Poumellec, L.
Dong, J. F. Bayon, H. Poignant, and E. Delevaque, Densification involved in the UV-
based photosensitivity of silica glasses and optical fibers, J. Lightwave Technol., 1997,
vol. 15, pp 13291342.
[6] P. J. Lemaire, R. M. Atkins, V. Mizrahi, and W. A. Reed, High pressure H2 loading as
a technique for achieving ultrahigh UV photosensitivity and thermal sensitivity in GeO2
doped optical fibers, IEE Electron. Lett., 1993, vol. 29, pp 11911193.
[7] N. Azami, A. Villeneuve, F. Gonthier, Method of adjusting thermal dependence of an
optical fiber, Optics Communications, 2005, Vol. 251, Issues 1-3, pp 6-9.
[8] F. Gonthier, F. Seguin; Francois, N. Godbout, A. Villeneuve, Passive thermal
compensaton of all fibre Mach Zehnder interferometer US Patent 6,850,654, 2005.
[9] C. Chon, H. Luo, C. H. Huang, R. Huang, J. Chen, and J. R. Bautista, Expandable 50-
GHz and 100-GHz dense wavelength division multiplexers based on unbalanced and
cascaded-fiber Mach-Zehnder architectures, in Tech. Dig. National Fiber Optic
Engineers Conference, NFOEC99, 1999.
[10] C. H. Huang, Y. Li, J.Chen, E. Sidick, J. Chon, K. G. Sullivan, and J. Bautista, Low-
loss flat-top 50-GHz DWDM and add/drop modules using all-fiber Fourier filters, in
Tech. Dig. National Fiber Optic Engineers Conference, NFOEC2000, 2000, pp
311-316.
[11] J. C. Chon, C. H. Huang, and J. R. Bautista, Ultra small dispersion, low loss, flat-top,
and all-fiber DWDM and NWDM devices for high speed optical network applications,
European Conference on Optical Communication 2000, paper 11.3.1, 2000.
[12] D.C Johnson, K.O. Hill, F. Bilodeau,,S. Faucher,. New design concept for a
narrowband wavelength selective optical tap and combiner, IEE Electronics Letters,
Vol.23, No.13, pp.668-669, 1987.
[13] F. Bilodeau, D.C, Johnson, S. Theriault, B. Malo, J. Albert, K.O. Hill, An all-fiber
dense wavelength-division multiplexer/demultiplexer using photoimprinted Bragg
Gratings, IEEE Photonics Technology Letters, Vol.7, No.4, pp.388-390, 1995.
[14] S.J. Madden, P.G. Jacob, J. Harradence, C. Robilliard, M.C. Ellas, S.J. Hewlett, K.
Sugden, P.A. Otten, W.T. Holloway, M.D. Wallace, R.K. Selwood, High performance
modular DWDM implementation based on fiber Bragg Grating Mach Zehnder
Interferometers, in Tech. Dig. National Fiber Optic Engineers Conference, NFOEC98,
1998, Vol.1, pp.453-463.
[15] F. Bakhti, X. Daxhelet, P. Sansonetti, S. Lacroix, Influence of Bragg grating location in
fused 100% coupler for add and drop multiplexer realization, OFC98 Technical
Digest, 1998, pp.333-334.
[16] L.Y. Liu, J. Chen, A. Luo, J. Chon, K.Sullivin, J. Bautista, Ultra low loss all fiber
based FBG , in Tech. Dig. National Fiber Optic Engineers Conference, NFOEC99,
1999.
[17] A.H. Gnauck, "40-Gb/s RZ-differential phase shift keyed transmission", in Proc.
Optical Fiber Communication Conf. OFC 2003, 2003, Paper ThE1, pp 450-451.
[18] 18 X. Liu, Nonlinear effects in phase shift keyed transmission, in Proc. Optical Fiber
Communication Conf. OFC 2004, 2004, Paper ThM4, pp 851-852.
[19] X. Liu, C. Xu and X. Wei, Performance analysis of time/polarization multiplexed40-
Gb/s RZ-DPSK DWDM transmission, 2004, IEEE Photon. Technol. Lett., Vol. 16, pp
302-304.
[20] J. Wang, and J.M. Kahn, Conventional DPSK Versus Symmetrical DPSK:Comparison
of Dispersion Tolerances, IEEE Photon. Technol. Lett., 2004, Vol. 16, pp 1585-1587.
[21] F. Seguin and F. Gonthier, Tuneable all-fiber, delay-line interferometer for DPSK
demodulation, 2005, OFC 2005, Vol. 5, pp 248 250.
[22] X. Daxhelet and F. Gonthier, Tapered fiber filters: Theory and applications, in Proc.
SPIE, Photonic East Boston, MA, vol. 4216A, no. 5, 2000, pp. 343-354.
[23] M. Suzuki and S. Shikii, The gain equalizing method of Erbium-doped fiber amplifiers
for C-band and L-band, IEEE/LEOS Annual Meeting, 1999, ThJ2.
[24] H. Nakaji, M. Kakui, H. Hatayama, C. Hirose, H. Kurata, and M. Nishimura, Superior
noise performance and wide dynamic range Erbium-doped fiber amplifiers employing
Variable Attenuation Slope Compensator, IEICE Trans. Electron, 2001, Vol. E84-C,
pp 598-604.
[25] N. Azami, All-Fiber Dynamic Gain Slope Compensator, Optics Communications,
2003, Vol. 230/4-6 pp 325-329.
[26] 26 P. Ebrahimi, M.C. Hauer, Q. Yu, R. Khosravani, D. Gurkan, D.W. Kim, D.W. Lee,
and A.E. Willner, Statistics of Polarization Dependent Gain in Raman Fiber Amplifiers
due to PMD, in Conference on Laser and Electro-Optics, 2001, Vol. 56 of OSA trends
in Optics and Photonics Series, pp 143-144.
[27] S. Popov, E. Vanin, And G. Jacobsen, Influence of polarization mode dispersion value
in dispersion-compensating fibers on the polarization dependence of Raman gain,
Optics Letters, 2001, Vol. 27, No 10, pp 848-850.
[28] Q. Lin and G. P. Agrawal, Statistics of Polarization-dependant gain in fiber-based
Raman amplifiers, Optics Letters, 2003, Vol. 28, No 4, pp 227-229.
[29] E. Son, J. Lee, and Y. Chung, Gain variation of Raman amplifier in birefringent fiber,
Optical Fiber Conference (OFC), 2003, Paper TuC5.
[30] N. Azami, Characterization of Polarization Dependent Gain in Raman fiber
amplifiers, Optics Communications, 2003, Vol. 230/1-3, pp 181-184.
[31] M.S. Yataki, D.N. Payne, and M.P. Varnham, All-fiber polarizing beamsplitter,
Electron. Lett., 1985, Vol. 21, pp 249-251.
[32] T. Bricheno and V. Baker, All-fiber polarization splitter/combiner, IEE Electron. Lett,
1985, Vol. 21, pp 251-252.
[33] M. Eisenhamm, E. Weidel, Single mode fused biconical coupler optimized for
polarization beam splitting, J. of Lightwave Technol., 1991, Vol. 9, No 7, pp 853-858.
[34] C.W. Wu, T.L. Wu, H.C. Chang, A novel fabrication method for all-fiber, weakly
fused, polarization beamsplitter, IEEE Photonics Technol. Lett., 1995, Vol. 7, No. 7, pp
786-788.
[35] N. Azami, A. Villeneuve, F. Gonthier, All-Fiber Wide-Band Polarization Beam
Combiner, IEE Electron. Letters, 2003, Vol. 40, No 17, pp 1043-1044.
[36] P. Shen, and J.C. Palais, Passive single-mode fiber depolarizer, Appl. Opt., 1999, Vol.
38, pp 16861691.
[37] M. Martinelli, and J.C. Palais, Dual fiber-ring depolarizer, J. Lightwave. Technol.,
2001, Vol. 19, pp 899905.
[38] N. Azami, E. Villeneuve, A. Villeneuve, F. Gonthier, All SOP all fiber depolarizer
design, IEE Electron. Letters, 2003, Vol. 39, pp. 1573-1575.
Chapter 8
ADVANCES IN OPTICAL DIFFERENTIAL PHASE SHIFT
KEYING AND PROPOSAL FOR AN ALTERNATIVE
RECEIVING SCHEME FOR OPTICAL DIFFERENTIAL
OCTAL PHASE SHIFT KEYING
M. Sathish Kumar
1a
, Hosung Yoon
2b
and Namkyoo Park
1c
,
1
Optical Communication Systems Lab, School of EECS, Seoul National University,
Seoul, Korea, 151-742,
2
Network Infra Laboratory, Korea Telecom, Daejeon, Korea, 305-811,
Abstract
Optical Differential Phase Shift Keying (oDPSK) with delay interferometer based direct
detection receiver was proposed as an alternative for the conventional On-Off Keying (OOK)
modulation schemes. Compared to OOK, oDPSK was predicted to have a 3dB improvement
in performance due to its balanced detection receiver structure. It was also predicted that due
to the optical signal occupying all the symbol slots, unlike in OOK, symbol pattern dependent
fiber nonlinear effects will make less of an impact on long haul optical transmission schemes
based on oDPSK. Subsequent successful demonstrations of these positive attributes of
oDPSK resulted in active investigations into multilevel formats of oDPSK namely, optical
Differential Quadrature Phase Shift Keying (oDQPSK) and optical Differential Octal Phase
Shift Keying (oDOPSK). Significant developments in theoretical models of optically
amplified lightwave communication systems based on the Karhunen-Loeve Series Expansion
(KLSE) method assisted such investigations. In this chapter, we discuss some of the recent
advances in oDPSK and its multilevel formats that have been achieved such as proposals for
receiver schematics, theoretical analysis of receiver schematics, electronic techniques to
counter polarization mode dispersion induced penalties, and application of coded modulation
techniques. The chapter also proposes an alternative receiver schematic for oDOPSK which
can separately detect the three constituent bits from an oDOPSK symbol.

a
E-mail address: mskuin@yahoo.com
b
E-mail address: hsyoon@ieee.org
c
E-mail address: nkpark@snu.ac.kr
M. Sathish Kumar, Hosung Yoon and Namkyoo Park 232
1. Introduction
Single mode optical fibers with their enormous bandwidth of around 15THz and extremely
low attenuation of 0.2dB/Km in the 1550nm window offer immense promise as a viable
medium to realize high bit rate long distance data transmission systems. Over the years, there
has been a remarkable growth in the data transmission rates achieved with one of the most
recent experiments claiming 14Tbps over a distance of 160Km [1]. As the transmission
distance and signaling rates increase, certain problems inherent to the optical fiber medium
such as attenuation, chromatic dispersion, nonlinear effects, and Polarization Mode
Dispersion (PMD) start to crop up which inhibits increments in the link distance and data
transmission rates.
Developments in the area of optical fiber amplifiers have resulted in mature technologies
such as doped fiber amplifiers and fiber Raman amplifiers to overcome the attenuation limits.
Also, developments in optical fiber device technology such as fiber gratings, both long period
and Bragg, and high degree of control over refractive index profiles of core and cladding have
helped in identifying feasible and effective methodologies to counter chromatic dispersion.
Combating the ill effects of fiber nonlinearities and PMD still continue to be challenging
problems primarily due to their statistical nature.
More recently, alternate modulation techniques such as optical Differential Phase Shift
Keying (oDPSK) [2] a bi-level version of optical differential phase modulation and optical
duobinary signaling have been proposed and actively investigated upon. Optical duobinary
schemes are based on the principle of introducing controlled inter symbol interference so that
compared to On-Off Keying (OOK), for a given data transmission rate, the bandwidth of the
optical signal propagating through the fiber is reduced. This obviously has an advantage over
OOK in that spectral width dependent signal distorting mechanisms such as chromatic
dispersion and PMD will have less of an impact. A tutorial discussion on duobinary signaling
schemes could be found in [3]. In optical differential phase modulation, irrespective of
whether it is bi-level or multilevel, the phase of the optical field during the current signaling
interval is modulated relative to its phase in the previous signaling interval. The detection of
the data at the receiver side at any particular signaling interval is hence dependent on the
phase of the received optical signal in the previous interval. It is worth to note that optical
differential phase modulation was under investigation during the late eighties and early
nineties while coherent optical communication systems were aggressively explored [4].
However, the idea of optical differential phase modulation as most of the recent publications
concentrate on is based mainly on an interferometric delay line based direct detection
technique and will be the one discussed in this chapter.
In comparison to OOK, oDPSK provides a 3dB performance improvement [2]. The 3dB
improvement offered by oDPSK can easily be translated to an increase of approximately
15Km in the transmission length or a reduction in signal intensity dependent nonlinear effects
such as stimulated Raman scattering, four wave mixing, cross phase modulation, etc.
Moreover, since oDPSK has all the bit slots occupied by optical intensity, unlike in OOK, bit
pattern dependent undesirable impacts of fiber nonlinear effects also get alleviated.
While oDPSK transmits one bit per signaling interval, its multilevel versions, namely
oDQPSK and oDOPSK, transmit two and three bits respectively per signaling interval.
Obviously, for a given bit rate and pulse format (Return to Zero (RZ) or Non Return to Zero
Advances in Optical Differential Phase Shift Keying 233
(NRZ)), the oDQPSK and oDOPSK schemes can provide an increment in the spectral
efficiency by factors of two and three respectively over OOK and oDPSK. This is significant
in that as with optical duobinary transmissions, oDQPSK and oDOPSK due to their reduced
bandwidth requirements will be more immune to spectral width dependent signal distorting
mechanisms. Over and above these advantages, oDQPSK and oDOPSK carry with them more
or less the same other advantages that oDPSK has over OOK. However, the price that needs
to be paid for this improved spectral efficiency of the multilevel versions of oDPSK is an
inferior error rate performance.
This chapter aims at providing a review of some of the advances that have been achieved
in the domain of optical differential phase modulation schemes such as proposals for receiver
schematics, theoretical analysis of receiver schematics, electronic techniques such as
equalizers to counter PMD induced penalties, and application of coded modulation techniques
such as Trellis Coded Modulation (TCM) [5]. The chapter also proposes an alternative
receiver schematic for oDOPSK which can separately detect the three constituent bits from an
oDOPSK symbol.
2. Transmitter and Receiver Schematics for oDPSK and oDQPSK
Figure (1) shows the transmitter and receiver schematics for oDPSK along with the resultant
one dimensional signal space diagram and constellation in the inset. The coherent optical field
emitted by the laser is phase modulated by a suitable optical phase modulator which is driven
by differentially encoded NRZ data. The phase modulated output is passed through a pulse
carver to obtain RZ optical pulses which are subsequently transmitted through the fiber. A
pulse carver is essentially a Mach Zehnder Modulator (MZM) complimentarily driven by a
sinusoidal clock [2] [6]. The phase modulator can be a simple optical phase modulator or a
MZM [2].
At the receiver side, the optically amplified signal is first passed through an optical band
pass filter and then through a delay line interferometer with one arm of interferometer
introducing a time delay of T where T is the signaling interval. The constructive and
destructive port outputs of the delay line interferometer are used to illuminate a pair of
identical photo detectors to facilitate optoelectronic conversion. The difference of the two
photo detector outputs is then passed through an electrical post detection filter whose output
is sampled at appropriate time instants once every signaling interval to obtain y as shown in
figure (1). The obtained sample y is compared with a threshold of zero to obtain the estimates
of the transmitted binary data. It may be noted that the constructive port output (the top output
in figure (1)) effectively feeds in duobinary modulated optical signal to the photo detector
while the destructive port feeds in alternate mark inversion modulated optical signal [2]
[7][8]. As mentioned in [2], it is this balanced detection using a pair of photo detectors that
effectively gives a 3dB advantage for oDPSK over OOK.
The concepts of multilevel optical differential phase modulation schemes such as
oDQPSK and oDOPSK are in effect a two dimensional extension of oDPSK. In these
multilevel versions of oDPSK, inphase and quadrature components of the optical carrier are
phase modulated independently, combined and then transmitted. The inherent orthogonality
between the inphase and quadrature components of the optical carrier enables unambiguous
identification of the modulating data at the receiver side.
Figure 1. Transmitter and receiver schematics for oDPSK. The inset shows one dimensional signal
space diagram.
When it comes to receiver schematic for oDQPSK, there is no much of an option and
more or less the same technique as the one used in electrical communication systems [3] is
employed. However, oDOPSK opens up a range of options for detection. As such, we
dedicate two separate sections later on for discussing the receiver schematics for oDOPSK.
The rest of this section will discuss the transmitter and receiver schematic for oDQPSK.
Figure (2) shows the transmitter and receiver schematic for oDQPSK. Comparing this
schematic with that of oDPSK as given in figure (1), it can be readily observed that the
transmitter schematic is in effect a parallel concatenation of two oDPSK transmitters. The
incoming laser field is split into two equal parts in terms of power and passed through parallel
phase modulators ensuring that the split optical fields have a relative phase shift of /2
between them. This is to separate out the inphase and quadrature components of optical
carrier. The inphase and quadrature phase modulated optical fields are then combined and
guided through an optical fiber towards the oDQPSK receiver. At the receiver side, the
received signal is split into two equal parts and passed through parallel concatenated delay
interferometers. Compared to the delay interferometer setup depicted in figure (1) with
regards to oDPSK, the difference here is that the arms of the delay interferometers introduce
phase shifts that have to be such that the absolute value of the phase difference is /2. More
specifically
2 /
2 1
= (1)
Figure 2. Transmitter and receiver schematics for oDQPSK. The inset shows the signal space diagram
with
1
and
2
as p/4 and q/4, such that p and q are odd integers satisfying the condition given in
equation (1).
The post detection electronic processing to extract the data bits can be simplified
considerably by selecting
1
and
2
as p/4 and q/4, such that p and q are odd integers
satisfying the condition given in equation (1), and using an electronic precoding circuit at the
transmitter side as reported in [9]. The precoder is designed to satisfy the following Boolean
expressions (for p=1, q=-1).
) ( ) ( ) ( . ) (
1 1 1 1 1 2 1 1
+ =
k k k k k k k k k
I b I Q I b I Q I (2.a)
) )( ( ) ( . ) (
1 2 1 1 1 1 1 1
+ =
k k k k k k k k k
I b I Q I b I Q Q (2.b)
Here I
k
and Q
k
are the NRZ data driving the top and bottom phase modulators respectively
in figure (2) during the kth signaling interval and b
1k
and b
2k
are the two binary data bits
which constitute the kth transmitted oDQPSK symbol and which are subsequently detected
directly from the outputs y
1
and y
2
respectively.
With the parameters selected as given in the above paragraph and with the appropriate
precoder identified through equations (2.a) and (2.b), it is possible to obtain the information
bit sequence carried by the inphase and quadrature components of the optical carrier from a
bi-level detection of the outputs y
1
and y
2
respectively [9]. The resultant signal space diagram
and constellation when
1
and
2
is p/4 and q/4 with p and q as discussed above is also
shown in figure (2) in the inset. It may be noted that when one of the phase shifts of the delay
interferometers of the receiver is set as |/2| and the other as 0, the orientation of the signal
vectors get rotated by /4 from what it was when
1
and
2
were p/4 and q/4 with p and q
as discussed above. This has an obvious disadvantage in that the outputs y
1
and y
2
will now
be three valued as opposed to binary valued when
1
and
2
were p/4 and q/4 such that p
and q are odd integers satisfying the condition given in equation (1). It can be inferred from
the signal space diagram and the discussion above that the receiver schematic for oDQPSK
with
1
and
2
as p/4 and q/4 in effect treats the inphase and quadrature components of the
optical carrier as two separate independent oDPSK channels.
3. Transmitter and Receiver Schematics for oDOPSK
Figure (3) depicts a possible transmitter schematic for oDOPSK [10]. The idea is based on
the fact that an oDOPSK signal comprises essentially of two oDQPSK signals having a
relative phase offset of /4 between them [3]. The differentially encoded data b
1
and b
2
drive
two parallel phase modulators as it was in the case of oDQPSK transmitters. However, the
differentially encoded data b
3
brings about a phase shift of 0 or /4 in the signal propagating
through the last phase modulator. This effectively rotates the signal constellation by 0 or /4
radians.
Unlike in oDQPSK, wherein a direct mapping of the ideas from electrical communication
systems was followed to arrive at possible receiver schematics, for oDOPSK, the major
driving factors in identifying receivers have been optimized performance as well as ability to
extract the three constituent bits directly through a bi-level detection of samples. This has led
to suggestions of receiver schematics for oDOPSK which employ more than two delay
interferometers. Before venturing into a review of such receiver schematics, it is worth to note
that in principle it is possible to extract the eight distinct symbols that comprise the oDOPSK
symbol set using a receiver schematic like the one used for oDQPSK. This should be quite
elementary to understand since the signal space diagram of oDOPSK is two dimensional and
to uniquely represent a point in a two dimensional space, only two coordinates are required.
Those two coordinates could be obtained readily from a schematic exactly like the oDQPSK
receiver by treating the outputs y
1
and y
2
as multilevel [11].
Figure 3. Transmitter schematic for oDOPSK.
Figure (4) depicts two receiver schematics proposed in [10] and [11] for oDOPSK. In
case of figure (4.a), each interferometer output, after passing through the electrical post
detection filter, is sampled once every signaling interval and treated as a bi-level sample.
(a) (b)
Figure 4. Receiver schematics for oDOPSK; (a) is after [11] and (b) after [10].
To be more specific, the absolute value of the samples obtained (y
1
to y
4
) are disregarded
and only their numerical signs are taken into consideration. From the fact that in oDOPSK the
phase difference between successive symbols can take on only values which are integer
multiples of /4 mod 2, table (1) can readily be formulated from the receiver schematic
shown in figure (4.a). Table (1) clearly shows that for each value of phase difference there is
a unique combination of the numerical signs of y
1
to y
4
which easily enables the identification
of the transmitted symbol and subsequent decoding of the symbol to its respective binary
triplet. However, it should also be noted that out of 2
4
possible combinations, only 2
3
are
made use of and the inevitable redundancy that exists throws up the option of error correction
using maximum likelihood estimation techniques [11] [12].
Table 1. Relation between polarity of detected samples and phase of oDOPSK symbols
for receiver shown in figure (4.a)
Phase difference y
1
y
2
y
3
y
4
0 + + - -
/4 + - - -
/2 - - - -
3/4
- - - +
- - + +
5/4
- + + +
3/2 + + + +
7/4 + + + -
Coming to the receiver schematic given in figure (4.b), the samples obtained from the
delay interferometers outputs (y
1
, y
2
) and the XOR logic block output (y
3
) are treated as bi-
level as was in the case of figure (4.a). We defer discussions on this receiver schematic for a
later section wherein we discuss an alternative receiver schematic for oDOPSK.
4. Error Rate Performance Evaluation
As is well known, optical amplifiers have become essential components of current state-of-
the-art long haul fiber optic communication systems. Consequently, current fiber optic
communication systems are essentially Amplified Spontaneous Emission (ASE) noise limited.
As such, in performance evaluations, usually the sole noise source taken into account is the
ASE. An accurate method to arrive at the Characteristic Function (CF) of the detected
optically preamplified signal using the Karhunen Loeve Series Expansion (KLSE) method and
subsequent evaluation of probability of error through saddle point integration was reported in
[13] and modified later in [14] for computational efficiency. Though the KLSE method
reported in [13] and [14] was implied for OOK systems, the method could easily be modified
to use it for oDPSK as well as its multilevel versions as suggested in [2][15][16]. The major
advantage of the KLSE method compared to others such as those developed in [17] is that the
KLSE method could be used to evaluate the error rate performance for arbitrary pulse shapes
and can account for the pre and post detection filter transfer functions along with other linear
impairments of the fiber medium such as chromatic dispersion and PMD [14].
In the following subsections we first discuss the essence of KLSE method for error rate
evaluation of optically amplified lightwave communication systems and then the error rate
performance.
4.1. Essence of the KLSE Method for Error Rate Evaluation
Consider the general setup of a direct detection fiber optic communication receiver as shown
in figure (5). Let the received optically amplified signal be
) ( ) ( ) ( t w t s t e + = (3)
where s(t) and w(t) stand respectively for the desired signal and ASE noise sample function.
The ASE noise is a complex Additive White Gaussian Noise (AWGN) process with a two
sided Power Spectral Density of N
O
W/Hz. As the first step in deriving the KLSE method, it
can be shown that [13]

= ' ) ) ' ( 2 exp( ) ( ) , ' ( ) ' (
2
1
*
dfdf t f f j f E f f K f E y
k k
(4)
where y
k
is the detected sample as shown in figure(5), E(f) is the Fourier transform of the
received signal e(t), t
k
is the time instant at which the post detection filter output is sampled
and K(f,f) is as given below
Figure 5. General setup of a direct detection fiber optic communication receiver.
) ' ( ) ' ( ) ( ) , ' (
*
f H f f H f H f f K
o e o
= (5)
with H
o
(f) and H
e
(f) standing respectively for the transfer function of the optical and electrical
filters.
Rewriting equation (4) above as

=

b
a
b
a
k k
b
a k
df df ft j f E f f K t f j f E Lt y ' ) 2 exp( ) ( ) , ' ( ) ' 2 exp( ) ' (
2
1
*
(6)
and by treating a and b as finite but sufficiently large in magnitude so as to cover several
times the filter bandwidths [13], the inner integral in equation (6) becomes effectively the
right hand side of a Fredholm integral equation of second kind which can be represented
formally as [18]
=
b
a
m m m
df f f f K f ) ( ) , ' ( ) ' ( (7)
with
m
(f) and
m
being its eigenfunction and eigenvalue respectively and K(f,f) acting as the
Hermitian kernel.
Based on the fact that the eigenfunctions form a complete set of orthonormal basis
functions in the interval a<f<b, ) 2 exp( ) (
k
ft j f E can be expressed as a series expansion of
the basis functions as
+ =
m
m m m k
f N S ft j f E ) ( ) ( ) 2 exp( ) ( (8)
In equation (8) above the coefficients S
m
correspond to the desired signal and N
m
to the
ASE noise. It can be shown that N
m
will have both its real and imaginary parts as Gaussian
random variables with zero mean and variance N
o
/2 W/Hz each [13][14]. Substituting
equation (8) into (6) and making use of the definition of the Fredholm integral equation as
given in equation (7), it can readily be shown that
+
=
m
m
m m
k
N S
y
2
2
1
(9)
Since the real and imaginary parts of N
m
are mutually uncorrelated and Gaussian
distributed [13] [14], the CF of y
k
will be products of CF of non central chi square
distributions each with single degree of freedom. This CF of y
k
will be as given below
=
m o
m
m
m
o
N
j
S j
N
j
2
2
exp
2
1
1
) (
2
(10)
The probability density function (pdf) of y
k
can be obtained by inverting the CF. From the
pdf so obtained, the probability of error can be evaluated. Since the CF given above cannot be
inverted in a straightforward manner, saddle point integration method is used [14].
The alteration of the above discussed KLSE method so as to make it suitable for
evaluating the error rate performance of oDPSK and its multilevel versions is rather
straightforward in that all it needs to be done is to incorporate appropriately the transfer
functions of the delay line interferometers into the Hermitian kernel. Thus for oDPSK and its
multilevel versions, each of the detected outputs (y
k
) when expressed in the form of equation
(6) will have its Hermitian kernel as
) ( ) ' (
2
)) 2 ( exp( )) ' 2 ( exp(
) ' ( ) , ' (
*
f H f H
fT j T f j
f f H f f K
o o e

+ + +
=

(11)
where is the phase shift introduced by the corresponding delay interferometer and T is as
given earlier.
Treating the signal s(t), i.e. the data sequence transmitted by the transmitter, as a periodic
signal of sufficiently large period facilitates the usage of the FFT/IFFT algorithms which
makes the computational process more efficient [14]. Also, by treating the data sequence as
periodic and by representing it using appropriate pseudorandom sequence such as the de-
Bruijn sequence [16], it is possible to estimate the impacts of fiber induced inter symbol
interference caused due to chromatic dispersion and PMD. To incorporate these into the
KLSE model, the fiber transfer function has to be included in the Hermitian kernel. The
transfer function of the fiber which incorporates the chromatic dispersion is well known [14]
and the principal states model of Poole [19] facilitates rather easy incorporation of the PMD
effects into the fiber transfer function [16].
4.2. Error Rate Performance
Figure (6) shows the Bit Error Rate (BER) performance for oDPSK, oDQPSK and oDOPSK
as a function of the Optical Signal to Noise Ratio (OSNR) in a back-to-back condition. The
receiver schematics assumed in these evaluations are the ones discussed through figures (2)
and (3) for oDPSK and oDQPSK respectively and the one discussed through figure (4.a) for
oDOPSK. The optical filter was modeled as a first order Gaussian filter with its 3dB
bandwidth as three times the baud rate and the electrical post detection filter was modeled as
a fifth order Bessel filter with 3dB bandwidth as equal to the symbol rate. From here on, we
refer to the 3dB bandwidth as just the bandwidth. The OSNR was evaluated with unpolarized
ASE noise within a reference spectral width of 0.1nm. The bit rate was taken as 40Gbps for
all three cases. The binary data sequence was modeled as pseudo random sequences of length
31, 63 and 511 respectively for oDPSK, oDQPSK and oDOPSK. These results as well as the
others to be presented later in this chapter are for RZ pulse formats. We restrict ourselves to
the RZ formats because in general as per the results in [11], [15] and [16], the RZ format gives
a better error rate performance compared to the NRZ formats.
Though the above results in back-to-back condition give a very good picture of the
relative error rate performance of oDPSK, oDQPSK and oDOPSK schemes, it has to be noted
that in actual long haul systems, signal distorting mechanisms like PMD will have to be dealt
with. Electronic post detection equalization techniques such as Linear Equalizers (LE) [3]
have been proposed as viable technologies to deal with PMD induced pulse broadening [20].
In the following we discuss the performance of the above discussed schemes in the
presence of first order PMD with and without the use of LE. We begin with a very brief
review of LE schemes. More details could be found in standard digital communication text
books such as [3].
Figure 6. Bit Error Rate performance for oDPSK, oDQPSK and oDOPSK as a function of the OSNR in
back-to-back condition.
Figure 7. A five tap Linear Equalizer.
The LE is essentially a transversal filter as shown in figure (7). The tap delay elements
shown as square boxes in figure (7) introduce a delay of T
s
which can be equal to the
signaling interval or a fraction of that. The tap weights c
k
with which each delay element
output is multiplied is constantly updated using a tap updater algorithm called as the LMS
algorithm. Details of the LMS algorithm are beyond the scope of this chapter and could be
found in [3]. The sequence {v
k
} stands for the received distorted symbol sequence. The tap
updater (the LMS algorithm) is in turn fed by an error signal e
k
as shown in figure (7). This
error signal is derived from the difference between the equalized output
k
I
and its nearest

information symbol
k
I as estimated by the decision circuit. It should be easily
understandable from this figure that the tap updater algorithm works towards minimizing the
magnitude of this error signal which drives it.
It may be noted that to estimate the kth information symbol, the LE discussed above has
to depend not only on the kth received symbol v
k
but also on two before that and two after
that.
To incorporate the LE into the KLSE method of error rate evaluation so as to theoretically
evaluate the performance improvement achievable with the use of LE in the presence of PMD
induced pulse broadening, the LE is modeled as an electrical filter with transfer function [15]
=
=
2
2
) 2 exp( ) (
k
s k LE
fkT j c f H (12)
where, it is assumed that the LE has five taps as in figure (7).

Figure 8. OSNR penalty of equalized and unequalized oDQPSK and oDOPSK systems.
The improvement in performance in oDQPSK and oDOPSK systems by the use of a five
tap LE in the presence of first order PMD effects is depicted in figure (8). The OSNR penalty
to attain a reference BER of 10
-12
is plotted as a function of Differential Group Delay (DGD).
The OSNR penalty for a particular DGD value was computed as the difference between the
OSNR sensitivity in back-to-back condition (DGD=0ps) and the OSNR sensitivity at the
DGD value of interest. The bit rate for all the considered systems was taken as 40Gbps and
the tap delay T
s
of the LE was set as the symbol duration. The receiver optical and electrical
filter bandwidths were optimized to provide maximum instantaneous DGD tolerances. It has
been reported in [15] that the maximum instantaneous DGD tolerance was provided when the
receiver optical and electrical filter bandwidths were optimized for a reference DGD of 25ps
which turns out to be 1/2 and 1/3 times respectively of the symbol duration for oDQPSK and
oDOPSK considered herein. What we mean by optimization of filter bandwidths for
maximum DGD tolerance is as follows. When the filter bandwidths were optimized for
maximum OSNR sensitivity at a DGD value of 25ps, the OSNR penalty of the system to
instantaneous DGD values, especially around and beyond DGD values of 20ps, was lesser
than what it was when the filter bandwidths were optimized for maximum OSNR sensitivity
in a back-to-back condition [15]. The optimum filter bandwidths for a reference DGD of 25ps
was found to be Bo = 3.5, Be = 0.8 times the baud rate for oDQPSK without equalizers and
Bo = 3, Be = 0.6 times the baud rate for oDQPSK with LE where Bo and Be are the optical
and electrical filter bandwidths respectively. For oDOPSK, Bo and Be were found to be 3.4
and 0.9 times the baud rate and 3.7 and 0.7 times the baud rate respectively for unequalized
and equalized systems.
Two important observations can be made from figure (8). The use of LE results in lower
OSNR penalty for both the systems and the oDOPSK system has lower OSNR penalty
compared to oDQPSK for a given bit rate. Reason for the latter is the fact that the oDOPSK
symbols, at the same bit rate as supported by oDQPSK, are broader in time domain.
Before moving on to the next section, we would like to mention that Decision Feedback
Equalizers (DFE) could as well be used in place of the LE to electronically compensate for
PMD induced pulse distortions. Results of OSNR penalty reduction in the presence of first
order PMD by the usage of DFE is presented in [15]. A four tap DFE with two feed forward
filter taps and one feed back filter tap was considered in [15]. The results in [15] show that
improvements over LE occur only at relatively larger values of DGD that are above 50ps for
the oDOPSK systems whereas for oDQPSK systems, there was practically no difference in
the OSNR penalty between the LE and DFE systems.
5. Application of Coded Modulation Techniques
As the demands on data transmission rate and distance increase, one of the prime candidates
that can assist meeting these requirements is coded modulation technique. Coded modulation
technique is the name given to such error control techniques which combine coding and
modulation into a single operation unlike in conventional error control coding schemes
wherein coding and modulation are treated as two different operations. One of the coded
modulation techniques called as TCM has been successfully employed in electrical
communication systems and will be discussed here from an application view point to
oDQPSK systems.
(a)
(c)
(b)
(d)
Figure 9. The concept of TCM; (a) is a rate convolutional encoder with one of the input bits driving it
and the other connected directly to the output, (b) is the resultant trellis diagram, (c) is the partitioning
of oDOPSK constellation and (d) is the trellis with the state transitions marked with the transmitted
oDOPSK symbol selected based on the partitioning shown in (c).
One of the first reports of TCM applied to optical differential phase modulation systems
was [21] wherein TCM was applied to oDPSK systems by incorporating a rate 1/2, constraint
length 3 convolutional encoder into the modulation process. Later, in [22] the idea was
extended to application of TCM to oDQPSK systems. TCM was essentially an invention of
Ungerboeck [5] in the eighties and ever since has grown considerably and has found useful
applications in electrical communication systems. For the benefit of readers who are not
familiar with TCM techniques, we provide a brief overview.
TCM is best explained with an example. Consider the setup depicted in figure (9.a). We
have a rate 1/2 convolutional encoder [3] [5] whose input is driven by one of two parallel
binary data streams. The other binary data stream appears as it is at the output of the system.
At the final output of the system, there are three parallel binary data streams (two from the
rate 1/2 convolutional encoder and one from the direct output). Thus, in a way, the combined
setup can be viewed as a rate 2/3 convolutional encoder with its trellis structure as given in
figure (9.b). During each signaling interval, depending on the combination of the three
parallel binary data streams, one of 2
3
possible signal points from an octal signal set such as
that of the oDOPSK is selected for transmission. The key to the success and effectiveness of
TCM is the way the signal point to be transmitted is selected. A heuristic approach towards
this, as explained in [5] [23], is to partition the constellation of the chosen modulation scheme
progressively so that the minimum Euclidean distance after each level of partitioning is more
than what it was before. The partitioning of the oDOPSK constellation based on this heuristic
guideline is as shown in figure (9.c). The two outputs from the convolutional encoder elect
that partitioned subset that has two signal points in it and the output that is directly connected
to the input selects the final signal point to be transmitted. It should be obvious from the
above discussion as to how the coding and modulation steps are interconnected and are
inseparable in a TCM system. Figure (9.d) shows the trellis diagram of the TCM scheme
discussed above with the transmitted signal points labeled on each transition.
To complete this example on TCM, we explain briefly the demodulation/decoding
operations at the receiver. Having received a noisy sequence of oDOPSK symbols of a
particular length from a TCM based transmitter, the receiver computes the Euclidean distance
between the received symbol sequence and all other possible symbol sequences of length the
same as that of the received sequence. Due to the usage of a convoutional encoder to select
the transmitted symbols, only certain particular symbol sequences would only be valid
transmitted sequences. For example in the present system, with the initial state of the
convolutional encoder as 00, a look at the trellis structure shown in figure (9.d) readily tells
that the sequence a-g-b cannot be a valid transmitted sequence of length three. It is this fact
that arms the TCM system with its error correction capability. The sequence that is nearest in
terms of Euclidean distance with the received noisy sequence is estimated as the actual
transmitted sequence. This method of calculating the Euclidean distance and taking decisions
based on the Euclidean distance is derived from the principle of maximum likelihood
sequence estimation and from the assumption that the noise corrupting the transmitted
symbols are statistically independent and Gaussian in distribution [3]. However, in the case of
non Gaussian noise as well, if the noise corrupting the symbols is statistically independent
and the statistical distribution of the noise is nearly Gaussian, Euclidean distance based
estimation is justified.
The maximum likelihood sequence estimation and the resultant Euclidean distance based
estimation can be carried out efficiently using the Viterbi algorithm [3] [5]. An error in the
decoding stage occurs when the Viterbi algorithm selects a wrong path in the trellis. This can
happen if the Euclidean distance between the received sequence and a valid sequence other
than the actual transmitted sequence is less than the Euclidean distance between the received
sequence and the actual transmitted sequence. Due to the manner in which the oDOPSK
symbol sequence to be transmitted is selected, namely the partitioning of the constellation and
the involvement of a convolutional code, the minimum Euclidean distance between any two
given sequences will be equal to the maximum Euclidean distance (between diametrically
opposite symbols) in the oDOPSK constellation [5]. Now since the dominant error event in
the decoding stages of conventional error control coding scheme is free distance or minimum
Hamming distance respectively for convolutional codes and block codes, it is understood that
the dominant error event in the decoding of TCM will be the minimum Euclidean distance
error which, in the present example, will be the event of erroneous estimation between
diametrically opposite symbols in the oDOPSK constellation [3] [5]. Under high OSNR
scenarios such as the ones often encountered in present day systems due to the usage of
optical amplification technologies, the probability of such an error event will thus very well
approximate and provide a lower bound on the Symbol Error Rate (SER) of TCM encoded
systems. It is worth to note that though the basic modulation scheme involved in this example
is oDOPSK, from the data transmission viewpoint the system is an oDQPSK system since
there are only two information bits carried per signaling interval. In further discussions in this
chapter, we refer to the above discussed TCM system as oDQPSK-TCM system.
An important point that needs to be mentioned here with regards to the receiver for
oDQPSK-TCM is that out of the two possible receiver schematics discussed in section 3 for
oDOPSK, due to the fact that the multiple outputs from the receiver are not statistically
uncorrelated and hence statistically dependent, they are not the ideal ones for TCM
demodulation [24]. Due to this reason, we use a receiver scheme as was depicted in figure (2)
with
1
= /4 and
2
= 3/4. The procedure for SER evaluation with such a receiver schematic
under the assumption of the minimum Euclidean distance error event dominating significantly
the error events in the decoding process is reported in [22] and references therein and is quite
straightforward.
Figure (10) depicts SER performance of oDPSK, oDQPSK and oDQPSK-TCM system as
a function of OSNR. The OSNR was calculated as earlier with unpolarized ASE noise within a
reference spectral width of 0.1nm. The optical and electrical filter parameters have been
retained the same as those for the results depicted in figure (6) of the previous section.
However, in these results, the signaling rate is taken as 20Gbaud for all the three cases
whereby the oDPSK system supports a bit rate of 20Gbps and the oDQPSK and oDQPSK-
TCM systems support a bit rate of 40Gbps. Also, a symbol error is said to have occurred in
oDQPSK if any one or both the bits that constitute a symbol is received erroneously
As can be seen, the application of TCM improves the performance of an uncoded
oDQPSK system to that of the binary oDPSK system at identical OSNR values. More
importantly, for a symbol rate of 20Gbaud, the oDQPSK-TCM supports an information bit
rate of 40Gbps at the same performance as a binary oDPSK system that supports 20Gbps
while consuming the same bandwidth. Thus the application of TCM effectively doubles the
bandwidth efficiency without having to tradeoff SER.
Figure 10. SER performance of oDPSK, oDQPSK and oDQPSK-TCM system as a function of OSNR.

Figure 11. OSNR penalty for increasing values of DGD for oDQPSK-TCM, uncoded oDQPSK without
equalizer and uncoded oDQPSK which uses LE.
Figure (11) shows the OSNR penalty of uncoded oDQPSK with and without LE discussed
in the previous section and that of the unequalized oDQPSK-TCM for increasing values of
instantaneous Differential Group Delay (DGD). The results are referenced to the OSNR
sensitivity of unequalized oDQPSK optimized in back-to-back condition for a SER of 10
-12
[15]. The equalized uncoded oDQPSK as well as the unequalized oDQPSK- TCM were
optimized for a reference DGD of 25ps as was done in the last section with optical and
electrical filter bandwidths as 3.5 and 0.8 times the baud rate respectively for unequalized
oDQPSK and 3 and 0.8 times the baud rate respectively for oDQPSK-TCM. These results
bring out the robustness of oDQPSK employing TCM in the presence of first order PMD. It
can be observed that the coding gain due to TCM over uncoded, unequalized oDQPSK is
retained throughout the range of DGD values considered in this figure.
6. Alternative Receiver Schematic for oDOPSK
In this section, we discuss an alternative receiver schematic for oDOPSK which needs only
two delay interferometers in contrast to the four required by the schematics depicted in figure
(4). The receiver discussed in this section is very closely associated with an earlier one
reported in [25].
The maximum likelihood detection principle [3] as applicable to oDOPSK can be readily
explained with the help of figure (12) wherein we depict the receiver schematic discussed
earlier through figure (2) with
1
= /8 and
2
= 5/8 along with the resultant oDOPSK
constellation and signal space diagram. The term
k
in this diagram stands for the phase
difference between the received oDOPSK symbols in the present and the previous intervals.
This phase difference can take on discrete values ranging from 0 to 7/4 in steps of /4.
Now having received y
1
and y
2
during a particular signaling interval, classical detection
theory works on the principle of hypothesis testing as given below
Figure 12. Receiver schematic with
1
= /8 and
2
= 5/8 and associated oDOPSK constellation and
signal space diagram.
) / ( ) / ( y k P
s
k
y s P
>
<
(13)
where P(a/b) is the conditional probability of event a given event b and s , k , and y are
respectively the vectors [s
1
s
2
], [k
1
k
2
] and [y
1
y
2
] with the first entry within the parenthesis
standing for the inphase component and the second entry for the quadrature component. The
vectors s and k are two different signal points in the signal space diagram. It may be noted
that what the above equation does in principle is it compares the probability of the transmitted
symbol being s or k given that the vector y was received. This hypothesis test is conducted
over all symbols and the transmitted symbol is estimated as the one that has the highest
probability of being transmitted given that the vector y is received.
Now if the noise that corrupts the transmitted signals is AWGN, and the two components
of the received vector y namely y
1
and y
2
are statistically independent (which they will be if
they are uncorrelated and have Gaussian distribution), the hypothesis test given in the last
equation can be rewritten as
2
2 2
2
1 1
2
2 2
2
1 1
) ( ) ( ) ( ) ( y k y k
s
k
y s y s +
<
>
+ (14)
which is in fact a Euclidean distance comparison in the signal space between the received
vector y and the two signal points s and k . Since all the eight signal points in oDOPSK
constellation are equidistant from the origin, equation (14) can be rewritten as
0 ) ( ) (
2 2 2 1 1 1
s
k
k s y k s y
>
<
+ (15)
The numerical sign of y
1
and y
2
can be readily made use of to identify the quadrant in
which the transmitted symbol is most likely to be. Thus, while assigning binary triplets to the
eight different symbols of the oDOPSK constellation, if two of the bits in these triplets are
made to tally with the quadrant in the signal space diagram where the symbol lies, those two
bits could be readily detected by detecting the numerical sign or in other words a bi-level
detection of y
1
and y
2
. As per the signal space diagram and constellation shown in figure (12),
starting from the point
k
=7/4 in the first quadrant, the triplets could be assigned in an
anticlockwise manner in the order (111), (110), (010), (011), (001), (000), (100) and (101). It
may be noted that this assignment is in agreement with gray encoding rules according to
which closest points in the signal constellation have to differ by only one bit.
To identify the third bit which separates the two signal points within a quadrant, we
revert back to equation (15) and analyze as follows. Within the same quadrant, (s
1
-k
1
) and (s
2
-
k
2
) would be of same magnitude and the only difference if any would be in the numerical
signs. Thus, while conducting the test as given in equation (15) above to estimate the most
probable transmitted symbol among two symbols from the same quadrant, what matters is not
the magnitude of the differences (s
1
-k
1
) and (s
2
-k
2
) but their numerical signs. Thus equation
(15) can be rewritten for symbols within the same quadrant as
0
2 2 1 1
s
k
y y
>
<
+ (16)
with
1
and
2
being the numerical signs of (s
1
-k
1
) and (s
2
-k
2
) respectively.
With reference to the signal space diagram shown in figure (12) and the concepts
presented above, the following with regards to estimating the third bit can readily be arrived
at
Ist quadrant
1
= +,
2
= -, if s = 7/4 and k = 3/2; therefore, 0
1
0
2 1
>
<
y y and
2 1
y y + = +
IInd quadrant
1
= -,
2
= -, if s = and k = 5/4; therefore, 0
0
1
2 1
>
<
+ y y and
2 1
y y = -
IIIrd quadrant
1
= -,
2
= +, if s = 3/4 and k = /2; therefore, 0
0
1
2 1
>
<
y y and
2 1
y y + = -
IVth quadrant
1
= +,
2
= +, if s = 0 and k = /4; therefore, 0
1
0
2 1
>
<
+ y y and
2 1
y y = +
From the above we note that either (y
1
-y
2
) or (y
1
+y
2
) is the deciding factor and when one
is the deciding factor the other has a constant numerical sign. From this observation, the four
different decision rules given above can be combined into a single decision rule for estimating
the third bit as given below.
0
1
0
) )( (
2 1 2 1
>
<
+ y y y y (17)
The above equation suggests that a binary decision on the sum and difference of y
1
and y
2
followed by an XNOR operation on those decisions can readily provide an estimate of the
third bit. In fact, this is the receiver schematic suggested in [25] with a minor variation in that
the XNOR is replaced by the XOR apparently due to the swap in positions of 0s and 1s in the
third bit of the triplet as compared to what it is herein.
The receiver schematic depicted in figure (4.b) also works as per the same principle as
discussed above. The two inputs to the XOR are effectively (y
1
-y
2
) and (y
1
+y
2
) [25]. Also,
with an appropriate precoding of the binary data as given in [10], it is possible to directly
obtain the three constituent data bits from the detected binary levels of y
1
, y
2
and the product
(y
1
-y
2
)(y
1
+y
2
).
Further, if equation (17) is rewritten as
0
1
0
) (
2
2
2
1
>
<
y y (18)
it becomes obvious that the decisions can be taken depending solely on the difference of the
absolute values of y
1
and y
2
. The conversion of the detected samples to their absolute values
can be achieved in effect by considering the fact that the detected analog samples y
1
and y
2
are in fact dealt with in the receiver electronics in the digital domain through an analog to
digital converter. More the resolution of the analog to digital converter better will be the
resultant digital representation of the detected analog voltage. This is the methodology used in
almost all the electrical soft decision decoding receivers [3]. In an analog to digital converter,
it is possible in principle to identify the numerical sign of the digitally converted sample and
as such it is possible to alter that numerical sign. Thus, if the detected sample y
1
or y
2
is
negative, its numerical sign can be altered and the following decision rule can be applied to
detect the third bit.
0
1
0
) (
2 1
>
<
y y (19)
The advantage of this receiving scheme is that the dependence of the decision making
variable on the sum as well as difference of y
1
and y
2
is removed and is now dependent only
on the difference between y
1
and y
2
. This is of course at the cost of an additional electronic
operation of changing the numerical sign of the detected samples. It may also be noted that a
mere change in numerical sign does not alter the pdf of the detected samples. The complete
schematic representation of this receiver is as given in figure (13).
Figure 13. Schematic representation of an oDOPSK receiver which employs only two delay
interferometers.
The BER or probability of error for this receiver schematic can be readily arrived at as
BER = ( P(y
1
>0/ b
1
= 0)+P(y
1
<0/b
1
= 1)+ P(y
2
>0/b
2
=0)+P(y
2
<0/b
2
=1)+P((y
1
-y
2
)>0/b
3
=0)+ P((y
1
-y
2
)<0/b
3
=1))/6 (20)
where b
1
, b
2
and b
3
stand for the constituents of the binary triplets assigned to the eight
symbols of the oDOPSK constellation and are assumed to take on logic levels 1s and 0s with
equal probability. It has to be specially taken note of the fact that y
1
-y
2
as it appears in the
above equation is after the numerical signs of both y
1
and y
2
are converted to positive.
The above BER can be computed as discussed earlier in section (4) using the KLSE
method. However, while computing the last two probabilities in equation (20) that involves
the difference of y
1
and y
2
, the CF of y
1
-y
2
is required. This can be obtained by replacing the
Hermitian kernel in the procedure outlined in section (4) by the difference of the Hermitian
kernels of the two arms of the delay interferometer after appropriately accounting for the
power splitting at the front end of the receiver after the optical filter.

(a) (b)
Figure 14. BER performance for the oDOPSK receiver structure depicted in figure (13).
Figure (14.a) and (14.b) shows the BER as a function of OSNR. Figure (14.b) zooms on to
a portion of the BER curve. These figures show the individual BER for b
1
, b
2
and b
3
as well as
the average BER. The optical and electrical filter bandwidths while evaluating these results
were taken as 3 times and 1 time the baud rate respectively. The bit rate was set as 40Gbps. It
can be noted that while the BER curves for b
1
and b
2
overlap as they should due to identical
variances and same average means for y
1
and y
2
, the BER of b
3
is slightly more than that of b
1
and b
2
. This is obviously due to the fact that the variance of y
1
-y
2
is twice the variance of y
1
and y
2
.
Before we conclude this chapter, it needs to be mentioned that we have not included a
performance evaluation of the oDOPSK receiver schematic shown in figure (4.b). This is due
to the fact that the BER evaluation of that receiver schematic using the accurate KLSE method
becomes a cumbersome problem due to the fact that the XOR output is a function of both
y
1
+y
2
as well as y
1
-y
2
. This renders the BER evaluation an exercise of double integration of
the joint pdf of y
1
+y
2
and y
1
-y
2
. The BER performance of that receiver using other techniques
of evaluation has appeared in [10] and its inclusion here is not suitable for a fair comparison
with the results reported in this chapter since all the results in this chapter have been
evaluated using the KLSE method.
7. Conclusion
This chapter has presented a detailed review of optical differential phase modulation schemes
touching upon oDPSK, oDQPSK and, oDOPSK. Various proposed receiver schematics for
these modulation schemes were presented along with their error rate performance. An
introduction to the KLSE method for error rate evaluation of the discussed systems was also
presented. Use of electronic equalization methods in the form of LE was discussed and the
resultant improvement in OSNR penalty to achieve a target BER of 10
-12
was discussed.
Coded modulation techniques in the form of TCM which can improve error rate performance
of optical differential phase modulation schemes were introduced and a comparison was made
between uncoded oDQPSK system and oDQPSK systems that use TCM. An alternative
receiver schematic for oDOPSK which needs only two delay interferometers was presented
along with its error rate performance.
While oDOPSK, oDQPSK and, oDOPSK with electronic equalizers and/or TCM schemes
hold a lot of promise for realizing future long haul optical transmission systems, the main
stumbling block that can arise towards the successful application of post detection electronic
processing based performance enhancers such as equalizers and TCM is the speed of the
electronic hardware. In this regards, some developments in electronic hardware as reported in
[26] and [27] holds some promise.
References
[1] Sano, A.; Masuda, H.; Kisaka, Y.;Asawa, S.;Yoshida, E.;Miyamoto, Y.;Koga, M.;
Hagimoto, K.;Yamada,; Furuta, T.;Fukuyama, H. ECOC 2006. Cannes, (Post Deadline
Paper)
[2] Gnauck, A. H.; Winzer, P. J. J. Lightwave Tech. 2005, Vol. 23, 115-129
[3] Proakis, J. Digital Communication. McGraw Hill, 2001
[4] Betti, S.; Marchis, G. D.; Iannone, E. Coherent Optical Communication Systems. Wiley,
New York, 1995.
[5] Ungerboeck, G. IEEE Commun. Magazine. 1987, Vol.25, 5-11
[6] Winzer, P.J.; Dorrer, C.; Essiambre, R. J.; Kang, I. IEEE Photon.Technol.Lett. 2004,
Vol 16, 1379-1381
[7] Penninckx, D.; Bissessur, H.; Brindle, P.; Gohin, E.; Bhakhti, F. Proc. ECOC 2001.
Amsterdam, The Netherlands, 2001, 456-457
[8] Wei, X.; Liu, X.; Chandrashekar, S.; Gnauck, A. H.; Raybon, G.; Leuthold, J.; Winzer,
P. J. Proc ECOC 2002. Copenhagen, Denmark, 2002, paper 9.6.3.
[9] Griffin, R.A.; Carter, A.C. OFC 2002. 367-368.
[10] Kim, C.; Li, G. Optics Express. 2004, Vol.12, 3415-3421.
[11] Yoon, H.; Lee, D.; Park, N. Optics Express. Vol.13, 371-376
[12] Yoon, H.; Lee, D.; Park, N. Ninth Optoelectronics and Communications Conference and
Third Conference on Optical Internet (IEEE Communication Society). 2004. paper
14C3-4.
[13] Lee, J. S.; Shim, C.S. J. Lightwave. Tech. 1994, Vol.12, 1224-1228.
[14] Forestieri, E. J. Lightwave Tech. 2000, Vol. 18, 1493-1503
[15] Yoon, H.; Kim, N.Y.; Park, N. IEEE Photons. Tech. Letters. 2005, Vol. 17, 2577-2579
[16] Wang, J.; Kahn, J.M. J. Lightwave Tech. 2004, Vol. 22, 362-371
[17] Marcuse, D. J. Lightwave Tech. 1990, Vol. 8, 1816-1823
[18] Cohen, H. Mathematics for Scientists and Engineers, Prentice Hall, 1992.
[19] Poole, C.D.; Wagner, R.E. Electron. Lett. 1986, Vol. 22, 1029-1030
[20] Haunstein, H. F.; Wolfgang, S. G.; Dittrich, A.; Sticht, K.; Urbansky, R. J. Lightwave.
Tech. 2004, Vol. 22, 1169-1182
[21] Buelow, H.; Thielecke, G.; Buchali, F. Optical Fiber Communication (OFC) , Los
Angeles, CA 2004, Paper WM5
[22] Kumar, M. S.; Yoon, H.; Park, N. IEEE Photons. Tech. Letters. 2007, Vol. 19,
1245-1247
[23] Ungerboeck, G. IEEE Commun. Magazine. 1987, Vol.25, 12-21
[24] Wang, Jin-Der.; Chung, H. Y. IEEE Trans. Commun. 1990, Vol. 38, 1549-1556
[25] Han, Y.; Kim, C.; Li, G. Electron. Lett, 2004, Vol. 40, 1372-1373
[26] Mizuochi, T.; Ouchi, K.; Kobayashi, T.; Miyata, Y.; Kuno, K.; Tagami, H.; Kubo, K.;
Yoshida, H.; Akita, M.; Motoshima, K. OFC 2003, PD 21, Atlanta, 2003
[27] Dawid, H.; Fettweis, G.; Meyr, H. IEEE Trans. VLSI Systems. 1996, Vol. 4, 17-31

Chapter 9

A NEW GENERATION OF POLYMER OPTICAL FIBERS

Rong-Jin Yu and Xiang-Jun Chen
School of Information Science and Engineering, Yanshan University,
Qinhuangdao 066004, China
Abstract
This chapter describes the background to the development of Polymer Optical fibers
(POFs), discusses the optical and temperature resistant properties of polymers while
emphasizing the intrinsic high attenuation of them. The first generation of POFs which
consists of a solid-core surrounded by cladding and transmits light by total internal reflection,
is puzzled by the difficulty of high attenuation. Then, the method of using a specific structure
(i.e. hollow-core Bragg fiber) to solve the problem is presented. A new generation of POFs
based on the hollow-core Bragg fibers with cobweb-structured cladding can guide light with
low transmission loss and high bandwidth in the wavelength range of visible to terahertz
(THz ) radiation. Efficient hollow-core guiding for delivery of power laser radiation and solar
radiation can be achieved by replacing the traditional polymethylmethacrylate (PMMA) with
heat-resistant polymers. Lastly, this chapter concludes with a discussion of applications in
diverse areas.
1. Introduction
The optical fiber is a circular dielectric waveguide that guides optical information and energy.
The first theoretical study of wave propagation on circular cylindrical dielectric structures
was made by Hondros and Debye [1]. It was not until 1966 that the proposed to use circular
glass fibers as optical transmission lines was made [2]. Since then, the attenuation of more
than / 1000dB km for silica fiber is reduced to about 0.2 / dB km by advanced purifying and
manufacturing processes. The optical fiber is generally made from glasses, polymers or other
materials. Compared to glass materials, polymeric materials have the shortcoming of intrinsic
high attenuation while having the advantage of great flexibility. The elastic limits of
polymeric materials are high. It is this high elasticity which allows the fabrication of tough
and flexible POF with diameters of one millimeter and above. However, inorganic glasses
with low elastic limits dictate that remarkable flexibility may only be observed in the optical
Rong-Jin Yu and Xiang-Jun Chen 258
fibers of very small diameter, typically 125 m . The inherent brittleness of glass requires that
a more elastic polymer coating be applied to the very fragile core-cladding structure to protect
its surface and prevent the growth of Griffiths cracks and consequent fracture.
For a long time in the past, the optical fibers consist of a solid-core surrounded by
cladding, and transmit light by total internal reflection, with the addition of a few hollow
waveguides with inner metallic or dielectric coatings to provide high reflectivity. The
transmission losses of a solid-core fiber, including solid-core photonic crystal fiber, are larger
than (at least equal to) the losses of the fiber-core material itself. Thus, first generation POFs
which consist of a solid-core surrounded by cladding are facing the difficulty of high
attenuation.
One feasible way of solving the problem is to use a hollow-core fiber instead of the solid-
core fiber. But the hollow-core fibers do not get any special attention and exploitation due to
the very large success of low-loss high-index core (solid-core) silica fibers by total internal
reflection and their extensive use in various fibers for transmitting light. Until a few years ago,
the conventional wisdom in most books related to the fiber had been that confined and
lossless propagation in fibers is accomplished by total reflection from the dielectric interface
between the core and the cladding, or must make use of the concept of total internal reflection
to save light inside the core of the optical fiber. In fact, very early in the developmental
history of optical fibers, the idea of using Bragg reflection in a cylindrical fiber to obtain
lossless confined propagation in a core with a refractive index lower than that of the cladding
medium was proposed in the year 1978 [3]. Up to the late 1990s, the hollow-core Bragg
fibers [4] and hollow-core photonic band gap fibers [5] were demonstrated. Both
experimental and theoretical investigations confirm that transmission losses by using hollow-
core Bragg fibers are dramatically lower than the losses of its constituent materials. It seems
that hollow-core Bragg fibers are a suitable structure that can very effectively confine the
transverse leakage of guided wave.
Once adopting the hollow-core Bragg structures, surmounting the shortcoming of high
attenuation and taking the advantage of great flexibility of POFs, a new generation of POFs
can guide light with low transmission loss and high bandwidth in the wavelength range of
visible to THz radiation. Efficient hollow-core guiding for delivery of power laser radiation
and solar radiation can be achieved by replacing the traditional PMMA with heat-resistant
polymers. We expect that a new generation of POFs will find many applications in diverse
areas, and is irreplaceable for some applications.
2. Development History of POFs
The first report of POF was in 1960s, which was almost the same as the invention of silica-
base optical fiber. In 1966, Du Pont invented the first POF named Crofon that was of step-
index (SI) type composed of PMMA core surrounded by a partially-fluorinated polymer
cladding. In 1975, Mitsubishi Rayon commercialized the first SI POF whose trade name was
Eska and Asahi Chemical and Toray soon followed in 1970s. The POF market was
originally dominated by these three major companies who have been manufacturing SI type
POFs composed of PMMA core [6]. In the more than forty years history of POFs, efforts
were made to improve the performance of attenuation, bandwidth and temperature resistance
of POFs. In the mean time, the work of Kaino (NTTs Ibaraki Laboratories), Koike (Keio
A New Generation of Polymer Optical Fibers 259
University) and their co-workers is very important in the advancement of technology of POFs
in the area of the reduction of attenuation and the realization of high bandwidth.
PMMA, which has been the most typical core material for POF, has large attenuation due
to the intrinsic absorption loss of carbon-hydrogen stretching vibration. Theoretical
attenuation limit of PMMA based POF is around 100 / dB km at 0.65 m wavelength, while
the attenuation abruptly increases from near infrared to infrared region. The intrinsic
absorption loss can dramatically decreased by using a polymer which has no carbon-hydrogen
bonding in it. The work of Kaino and co-workers has brought about the reduction of the
losses of PMMA-D8-core fibers to 20 / dB km at 680nm [7]. By substituting all hydrogen
bonding in polymer molecules with fluorine, remarkable low attenuation of 10 / dB m was
achieved by the perfluorinated (PF) polymer based graded-index (GI) POF. The first PF
polymer based GI-POF was reported in 1994. In 2000, PF polymer based GI-POF named
Lucina was commercialized from Asahi Glass Co., using a PF polymer named CYTOP [8].
However, the fibers reduce attenuation at the expense of increased cost. The deployment of
both deuterated and fluorinated POFs is limited by the high cost of the fiber materials.
Considering the transmission bandwidth of POFs, commercially available POFs have
been of the SI type with a numerical aperture ( NA) of about 0.5, whose bandwidth of
transmission is limited to about 5MHz Km . For a SI-POF, quite a degree of improvement in
its bandwidth can be achieved by reducing NA of the POF. For the low NA SI-POF ( NA=
0.31), its bandwidth is 160MHz at 100m ( 200MHz > at 50m ), which is currently
commercialized. It meets the requirement of standardization of 156 / Mb s , transmission 50m
approved by the ATM forum in May 1997. It is a common knowledge that the main limitation
on the bandwidth of multimode optical fibers is modal dispersion, which means that different
optical modes propagate at different velocities and the dispersion grows linearly with length.
One way to overcome the modal dispersion is to use single mode (SM) POF. The first
SMPOF was reported in 1991, which was successfully prepared by the interfacial-gel
polymerization technique [9]. In the fiber, the core diameter was 3 15 m and the
attenuation of the transmission was about 200 / dB Km at 652nm wavelength. Another way to
solve the problem for POFs with large cores is to use multi-layer step-index (ML-SI) POF
[10], multi-core step-index (MC-SI) POF [11], or GI-POF [12]. In ML-SI POFs, the core
region is composed of several layers with different refractive index. This concentric multi-
layer structure decreases modal dispersion compared to conventional SI type POF and a data
rate as high as 500 / Mb s for 50m transmission is achieved experimentally. MC-SI POF has a
low numerical aperture (0.25) and a core region composed of 19 cores of small-core. By
reducing the core diameter, not only modal dispersion but also bending loss is decreased. A
data transmission at 500 / Mb s for 50m is also achieved by the MC-SI POF. For GI POF, the
refractive index of the fiber core is graded parabola-like from a high index at the fiber core
center to a low index in the outer core region. For the GI POF produced by the interfacial-gel
polymerization method, its bandwidth measured is 3GHz for a fiber length of 100m. A low-
loss PF polymer based GI POF has been developed and PF polymer based GI POF is able to
transmit a data rate of 10 / Gb s or higher because of its material dispersion property [13].
For the temperature resistance of POFs, the high-temperature performance of a polymer
is limited by its glass transition temperature (
g
T ). For PMMA,
g
T is about 105. Maximum
operating temperature for PMMA-core SI-POFs is 80. Ziemann et al. [14] had carried out a
accelerated aging test for the fibers from the three leading POF manufacturers. The results of
the test show that for all fibers and wavelengths ( 650nm, 590nm, 525nm and 520nm) the
estimated possible operating temperature for 20 years use is over 70. Some applications,
such as in automobiles, aerospace environments and transmitted power demand performance
at temperature in excess of 80. The
g
T of polycarbonate (PC) is around 170. The use of
PC and partially fluorinated PC as core material enables temperatures of up to 115 and
145, respectively. The
g
T of polyethersulfone (PES) is about 225, maximum operating
temperature of PES is 197. Polyimide material has even more high operating temperature
(316). The attenuation of these high temperature resistant polymers is generally larger than
that of PMMA, therefore making the polymers useless in fabricating the fibers, but using the
polymers (such as polyimide) as the coating of high temperature resistant silica fiber.
In a word, in the forty years development of POFs, there is no better position in both
performance (especially attenuation) and cost comparing with silica glass fibers. Thus, first
generation POFs have limited their penetration in important market-segments, and are only
suited to ornament, illumination, sensors and short-distance data transmission applications.
3. Hollow-Core Fibers
Hollow-core fibers reported to date in the literature can generally be classified into four types:

(1) those in which the refractive index of the cladding is greater than that of the core,
(2) those in which inner wall coating has high reflectivity,
(3) hollow-core photonic bandgap fiber, and
(4) hollow-core Bragg fiber.
3.1. Those in Which the Refractive Index of the Cladding Is Greater Than
That of the Core
As is known to all, waveguiding is achieved in conventional solid-core fibers due to the total
internal reflection from the interface between the core with the refractive index
core
n and the
cladding with the refractive index ( > )
clad core clad
n n n . For the hollow-core fiber in which the
refractive index of the core is lower than that of the cladding, the propagation of light is
achieved by the regime of grazing incidence and is accompanied by radiation losses (leaky
guide). In fact, this hollow-core fiber is a capillary tube, as shown in Fig.1. The coefficient of
optical losses in the hollow fibers scales as
2 3
/a , where is the radiation wavelength and
a is the core radius of the fiber. Thus, most of applications are performed by using the
hollow-core fibers with large inner radii and short length. For example, a 10cm-long and
150 m -diameter hollow-core fiber filled with argon gas is used on extreme ultraviolet (EUV)
light generated through the process of high-harmonic up-conversion of femtosecond laser
[15].


Figure 1. Cross section of the hollow-core fiber with
clad core
n n > (leaky guide).
3.2. Those in Which Inner Wall Coating Has High Reflectivity

Figure 2. Cross section of two hollow-core fibers with inner wall coatings.
For hollow-core fiber in which inner wall coating has high reflectivity as shown in Fig.2,
the fibers are metallic, glass or polymer tubes with inner metallic or (and) dielectric coatings
provided with high reflectivity in the wavelength range of interest. Among them, the hollow-
core fiber whose inner wall material has a refractive index less than one is referred to as an
attenuated total reflectance (ATR) guide. In general, the 1 n < or ATR guides are made of
sapphire,
2
GeO or some special 1 n < oxide glass. The idea of an 1 n < structure originated
from Hidaka, et al. in 1981 [16]. To be useful for laser transmission, the ATR guides must
have the region of anomalous dispersion, where n is less than 1, fall within some useful laser
wavelength range. The first 1 n < guides studied by Hidaka, et al. focused on glass tubes
made from lead and germanium doped silicates. By adding heavy ions to silica glass, he was
able to shift the infrared edge to longer wavelengths so that the 1 n < region of anomalous
dispersion occurred within the
2
CO laser wavelength band. Gregory and Harrington [17]
pointed out that sapphire or
2 3
Al O has 1 n < from 10 to 16.7 m and, in addition, it has a
very small k value of 0.05 at 10.6 m . This means that the theoretical loss is very low (less
than 0.1 / dB m for a 1000 m -bore tube) for this material. But sapphire has a high modulus.
Therefore, it cannot be bent to small diameters. These hollow-core fibers are an attractive
alternative to solid-core infrared fibers. These fibers have losses as low as 0.1 / dB m at
10.6 m and may be bent to radii less than 5cm . For applications in high-power laser
delivery, the fibers have been shown to be capable of transmitting up to 2.6kW of
2
CO laser
power [18]. They also have usage in both temperature and chemical fiber sensor applications.
Recently, hollow polycarbonate tubing with inner Cu coating is used on broadband THz
hollow-core
cladding(glass)
hollow-core
n<1 material
structural tube
hollow-core
structural tube
metallic film
dielcetric film

transmission, and the lowest loss of 3.9 / dB m was obtained from a 3mm -bore fiber at
158.51 m [19].
3.3. Hollow-Core Photonic Bandgap Fiber
In 1999, the hollow-core photonic bandgap fiber have been experimentally demonstrated for
the first time [5]. For hollow-core photonic bandgap fibers as shown in Fig.3, they are optical
fibers with cladding made of fused silica or polymer incorporating arrays of air holes. The

Figure 3. Cross section of hollow-core photonic bandgap fibers.
core is formed by omitting several unit cells of material from the cladding. The holey
cladding has a two-dimensional photonic bandgap that can confine light to the core for
wavelengths around a minimum-loss wavelength. For example, the transmission losses of
hollow-core plastic (PMMA) photonic bandgap fibers can be decreased by an order of
magnitude with nine rings of air holes in comparison with conventional plastic fibers
according to our numerical analysis by using multipole method [20]. Xu, et al. [21] analyzed
the loss of an air-core silica glass photonic bandgap fiber and demonstrated that it is possible
reduce the transmission loss to a level below / 0.01dB km , with eight rings of air holes at
1.53 m . Experimentally, the lowest loss reported in hollow-core photonic bandgap fibers
with fused silica is 1.2 / dB km at 1620nm[22]. The ultimate limit to the attenuation of such
fibers is determined by surface roughness due to frozen-in capillary waves. The attenuation of
1.2 / dB km at 1620nm already appears to be dominated by this mechanism. On the other hand,
under consideration of fiber design, Roberts, et al. [22] pointed out that the reduction in
reported attenuation from 13 / dB km to 1.7 / dB km (and now 1.2 / dB km ) was partly from
enlarging the core from 7 to 19 unit cells, reducing F by a factor of
3/ 2
(19 7) / 4.5 . F could
be further reduced by enlarging the core to 37 cells for example. However, this would be
accompanied by propagation of more higher-order core modes, increased bending loss and
closer spectral packing of surface modes. Since these are already apparent in our 19-cell
hollow-core photonic crystal fiber (HC-PCF), we expect 37-cell HC-PCFs to be of very
limited practical value. In view of this, it is very difficult for hollow-core photonic bandgap
fibers to further reduce the attenuation and achieve single mode at the same time. The hollow-
core fiber needs a cladding structure that can confine the transverse leakage of guided wave
more effectively. Until now, it seems that hollow-core Bragg fiber is a suitable selection.
3.4. Hollow-Core Bragg Fiber
So far, there are three classes of hollow-core Bragg fibers:

(a) OmniGuide fibers with very large cladding indices contrast [23],
(b) ring-structured hollow-core fibers with a single material [24, 25],
(c) cobweb-structured hollow-core fibers with a single material and a certain number of
supporting strips [26, 27].

(a) OmniGuide fiber (b) Ring-structured fiber (c) Cobweb-structured fiber
Figure 4. Cross section of hollow-core Bragg-fibers.
In Bragg fibers, the hollow-core is surrounded by a 1-dimensional Bragg reflector
consisting of alternating layers of high- and low-index materials. The cladding of
OmniGuide hollow-core fiber consists of two (solid) materials with different indices as
shown in Fig.4(a), that of ring-structured hollow-core fiber consists of a single material in
which rings of holes are used to define the low-index layers as shown in Fig.4(b), and that of
cobweb-structured hollow-core fiber consists of a single material in which air layers are used
as the low-index layers as shown in Fig.4(c). These Bragg fibers made a breakthrough in
fibers transmission losses less than the absorption losses of the material. The attenuation,
which is lower than the material loss, has been observed in OmniGuide fibers and ring-
structured fibers. The transmission losses of OmniGuide hollow-core fibers are
dramatically lower than the losses of its constituent materials [28]: Recent data show the
losses of 0.65 / dB m for such fiber, when made of a material with losses of 30000 / dB m .
Thus, its structure suppresses the losses of constituent materials by a factor of more than
45000. But, it is difficult to find two materials they have larger indices contrast, similar
thermal and mechanical properties, as well as compatible processing technique in realizing
the structure. Therefore, only two material combinations have been demonstrated: Te ( n =4.6)
in combination with a polymer ( n =1.59) [4], and
2 3
As Se ( n ~2.8) in combination with PES
( n ~1.55) [29]. For ring-structured hollow-core fibers, the attenuation, which is lower than
the material (PMMA) loss, has been observed in the infrared ( 1120nm > ), specifically at
1390nm wavelength, at which the transmission loss is only 40 / dB m compared with the
420 / dB m material loss [30].
We proposed a modified cladding structure, i.e. a hollow-core Bragg fiber with cobweb-
structured cladding [26]. The structure uses a single dielectric material and may solve the
problem of structural support by using a certain number of supporting strips. The supporting
strips are always symmetric in the cross-section and use the same dielectric material as
alternating layers. Our research shows that the field profiles are slightly deformed due to the
introduction of supporting structure. Although a small fraction of power is leaked out as a
result of the introduction of supporting structure, properly selected parameters of supporting
structure will keep the loss at a low level, neglecting the presence of the supporting strips.
The number and width of supporting strips should be as small as possible, generally,
m (number of supporting strips) = 6~12 and
s
w (width of supporting strips) = 3 /3 ~ 0 / ,
where is the operating wavelength of the fibers.
In comparison to OmniGuide fibers, the feasibility of cobweb-structured fibers is
greatly improved. For ring-structured fibers, the refractive index of low-index layers in the
cladding is between high-index (host material) and 1 (air). As a result, the cladding indices
contrast of ring-structured fibers is smaller than that of cobweb-structured fibers. As far as the
ability to confine the transverse leakage of guided wave is concerned, the ring-structured
fibers are smaller compared to the cobweb-structured fibers. In order to compare the
confinement losses of hollow-core ring-structured Bragg fiber with hollow-core cobweb-
structured Bragg fiber, we make the design of analogous structure. Argyros et al. [25] have
presented the design that supports a single-polarization, circularly symmetric nondegenerate
mode in an air-core ring-structured Bragg fiber. The design presented has
0.403
i
m =
,
0.578
e
m =
,
0.355 d m =
and core radius(
o
r ) =2.89 m , giving / 0.83
i
d = . The host
material was assumed to be lossless with a refractive index of 1.49 (corresponding to PMMA
material). When N (number of rings in cladding) = 9, the confinement losses of the
01
TE
mode (lowest-loss mode) and
02
TE mode (second-lowest-loss mode) are about 0.83 / dB m
and 57.14 / dB m, respectively. The ratio of the loss of the
02
TE mode to the loss of the
01
TE
mode reaches approximately 70. In our design, the same parameters:
2
n (PMMA) = 1.49,
co
r (core radius) = 2.89 m ,
2
d (thickness of high-index layers) = 0.243 m ,
1
d (thickness of
low-index layers) = 0.335 m and N (number of alternating layers in cladding) = 9, as well
as
1
1 n = are used. The host material was also assumed to be lossless. The calculated results
show that the least-loss wavelength of the
01
TE mode is located at 0.72 m . The confinement
losses of the
01
TE mode and
02
TE mode at 0.72 m wavelength are
5
5.32 10 / dB m
and
3
2.97 10 / dB m
, respectively. The ratio of the loss of the

02
01
TE
mode reaches approximately 56. Thus it can be seen that the confinement loss of the
01
TE
mode in the hollow-core cobweb-structured Bragg fiber is reduced by 15600 times in
comparison to that of the air-core ring-structured Bragg fiber.
These hollow-core Bragg fibers not only can reduce unwanted material properties, such
as absorption, scattering, dispersion and nonlinearity to a large extent, but also can act as a
modal filter [3]. Sterke et al. [31] found that such Bragg fibers can be guaranteed to be
effectively single-moded. Johnson et al. [23] presented their work of how the lowest-loss
01
TE mode can propagate in a single-mode fashion through even large-core fibers, with other
modes eliminated asymptotically by their higher losses and poor coupling, analogous to
hollow metallic microwave waveguides. The single-mode operation of the Bragg fibers is
achieved through asymptotic way during the transmission of guided waves, i.e. the number of
modes in large-core Bragg fibers causes the change as follows, at the beginning, the
transmission with multimode is followed by a few modes, and then the transmission becomes
single moded at last. Thus the single mode is achieved in a certain length range of the fiber.
Moreover, Bassett and Argyros [32] presented a method for calculating the single-mode
length range: The individual modes are characterized by two lengths,
1%
L at which the
transmitted power in that mode is reduced to 1%, and
0.01% 1%
2 L L = , at which the power is
reduced to 0.01%. We characterize each fiber as a whole by two lengths,
max 1%
L L = for the
best guided mode, and
0.01% sm
L L = for the second best guided mode. We consider the
usefully single moded for lengths between
sm
L and
max
L .
4. Hollow-Core Bragg Fiber with Cobweb-Structured Claadding
The refractive index profiles of hollow-core Bragg fiber with cobweb-structured cladding,
together with those of ring-structured and OmniGuide hollow-core Bragg fibers are shown
in Fig.5 for comparison. The parameters of the fiber with cobweb-structured cladding are
co
r (hollow-core radius),
1
n (=1, air),
2
n (high-index),
1
d (thickness of air layers),
2
d (thickness of high-index layers),
2 1
( / ) d d = ,
1 2
( ) d d = + , N , m and
s
w , where N is
the number of alternating layers in cladding, m and
s
w are the number and the width of the
supporting strips, respectively.

(a) cobweb-structured fiber (b) ring-structured fiber (c) OmniGuide fiber
Figure 5. Profiles of refractive index for hollow-core Bragg fibers.
In cylindrical waveguides, modes can be labeled by their angular momentum integer m ;
the ( , , ) z t dependence of the modes is given by
( ) j z t mf
e
+
. In the hollow-core fiber with
cobweb-structured cladding the modes will be affected by the supporting strip. Because the
supporting strip is periodic in , the modes can be written as
( ) 2 / j z t m j n
n
e e
+
, where
n is integer, is the periodicity of supporting strip in direction. The effective wavevector
/ k m r
= in the direction goes to zero for r . So the bandgap of this structure is the
same as OmniGuide Bragg fiber in Ref. [23] and purely depends on
r
k and
as long as
s
w is small enough.
For designing hollow-core Bragg fiber with cobweb-structured cladding, some important
structural parameters related to the permitted normalized frequency range of the
01
TE mode,
and their varying rule were analyzed by using a plane wave expansion method [27]. The
lowest-loss mode in Bragg fiber is
01
TE mode. The simulated results for hollow-core Bragg
fiber with cobweb-structured cladding show that leakage losses of
02
TE mode for the fibers
with
0.05 =
,
2
0.25 d m = ,
2
1.49 n = , 4 N = and different core radii ( 10
co
r m = and
50 m ) at 0.65 m = are
6
5.4 10 and
3
8.7 10 times larger than those of
01
TE mode,
respectively. Thus, the permitted frequency range of
01
TE mode is of especially interest. The
most commonly used material in POF is PMMA, its refractive index is 1.49. Using PMMA as
the high-index material of the Bragg reflection layers, the first two TE modes in the Bragg
reflection layers are calculated with the plane wave expansion method [33]. Figure 6 shows
the mode index of the first two TE modes in the Bragg reflection layers.
01
TE mode is the
fundamental mode in the hollow core. Its mode index must be below 1 and approach to 1. The
frequency range formed by two intersecting points ( P and Q) of the two TE mode curves
and the air line ( 1
eff
n = ) is approximately the permitted frequency range of
01
TE mode in
hollow core. For 0.01 = and 0.05,
1
n (air),
2
1.49 n = (PMMA), we can see from Fig.6 that
this kind of structure can guide light in the hollow core over a wide frequency range.
Different have a strong effect on the permitted normalized frequency range of the
01
TE
mode. For 0.01 = , normalized frequency can achieve the range from 2.91 to 45.76, while
for 0.05 = , normalized frequency is within the range 1.34 to 9.5. The permitted normalized
frequency range of
01
TE mode shrinks more than 5 times as changes from 0.01 to 0.05. In
order to figure out the influence of the structural parameter of Bragg reflection layers on
the permitted normalized frequency range of
01
TE mode, the permitted normalized frequency
range of
01
TE mode with different at a fixed
2
n (1.49) was calculated. The results are
listed in Table 1.

Figure 6. Permitted normalized frequency range of TE
01
mode for Bragg fiber with =0.01, 0.05. The
two curves indicate the first two TE modes in the Bragg reflection layers, and the solid line is the air
line ( n =1) [27].
Table 1. The permitted normalized frequency range of
01
TE mode vs different at a
fixed
2
n (1.49) [27]
0.5 0.4 0.3 0.2 0.1 0.08 0.06 0.05 0.04 0.02 0.01
Q point value
1.36 1.59 1.96 2.72 4.98 6.11 8.00 9.5 11.76 23.07 45.76
P point value
0.57 0.60 0.65 0.75 0.99 1.09 1.24 1.34 1.49 2.07 2.91
/ range
( ~ ) Q P

0.79 0.99 1.31 1.97 3.99 5.02 6.76 8.16 10.27 21.00 42.85

0 0.1 0.2 0.3 0.4 0.5
0
0.1
0.2
0.3
0.4
0.5
n
2
=1.49
upper limit of d
2
lower limit of d
2

Figure 7. Range of normalized high index layer thickness (
2 2
d d = ) vs. [27]

01
mode vs.
1
d [27]
In regard to the range of allowed values of
2
d , we define
2 2
/ d d = as the normalized
high-index-layer thickness, where is the operating wavelength of the fiber. The upper and
lower limit of
2
d can be obtained by means of the Q and P point values for each in
Table 1. Take 0.05 = as an example, the upper limit ( Q point value) of the permitted
normalized frequency range of
01
TE mode is 9.5, which means
1 2
( ) / 9.5 d d + = .
Substituting
1 2
/ 0.05 d d = into it, we can obtain
2 2
/ 0.4524 d d = = . The relationship
between
2
d and is shown in Fig.7. From Fig.7, we can see that the values of
2
d for the
upper line are approximately 0.45, indicating that the maximum thickness of
2
d cannot go
beyond 0.45 . In general,
2
d takes 0.4 ~ 0.3 . The minimum thickness of
2
d decreases
when decreasing .

01
mode as a function of
2
n for = 0.05 [27].
In regard to the relationship between the
1
d and permitted normalized frequency range of
01
TE mode, the permitted normalized frequency range of
01
TE mode with
2
0.25 d m = and
2
1.49 n = at different
1
d is illustrated in Fig.8. One obvious feature of Fig.8 is that the
permitted normalized frequency ranges of
01
TE mode and the corresponding thickness
1
d of
air layer are approximately a linear relationship. Thus, so long as the thickness
1
d of air layer
increases at a fixed
2
d , the normalized frequency range broadens.
In regard to the relationship between
2
n and permitted normalized frequency range of
01
TE mode, a series of
2
n ranging from 1.45 to 5.8 at a fixed (0.05) are calculated, as
shown in Fig.9. The permitted normalized frequency range of
01
TE mode increases when
2
n
decreases. Most of polymers have the refractive index smaller than 1.8. Therefore, they are
advantageous as the materials of the fiber with a large transmission frequency range.
In regard to the tolerance of the parameters, we take a dielectric material PMMA as an
example. The design objective is a hollow-core fiber to use as optical fiber communication in
the wavelength range from 0.65 m to 1.65 m . The design parameters are 0.05 = ,
2
0.25 d m = and
2
1.49 n = . Its normalized frequency / is in the range from
5.25/0.65=8.1 to 5.25/1.65=3.2, all within the permitted normalized frequency range of
01
TE
mode (9.5-1.34) as shown in Fig.6(b). If
2
d has a error of
2
20% d in the production
process, this corresponds to
2
0.2 d m = and 0.3 m . For
2
0.2 d m = , the normalized
frequency range is from 5.2/0.65=8 to 5.2/1.65=3.15. This is within the permitted normalized
frequency range of
01
TE mode (11.76-1.49) as shown in Table 1 for
0.04 =
. For
2
0.3 d m = , the normalized frequency range is from 5.3/0.65=8.15 to 5.3/1.65=3.21. This is
almost within the permitted normalized frequency range of
01
TE mode (8.00-1.24) as shown
in Table 1 for 0.06 = . If
1
d has a error of
1
25% d , this corresponds to
1
3.75 d m = and
6.25 m . For
1
3.75 d m = , the normalized frequency range is from 4/0.65=6.15 to
4/1.65=2.42. This is within the permitted normalized frequency range of
01
TE mode (7.21-
1.18) for 0.067 = . For
1
6.25 d m = , the normalized frequency range is from 6.5/0.65=10
to 6.5/1.65=3.94. This is within the permitted normalized frequency range of
01
TE mode
(11.76-1.49) for
0.04 =
. Finally, polymers are considered to have different refractive
indices for the same material, due to different molecular weight or polymerization condition.
If the index of PMMA has a variation of
2
0.02 n , which corresponds to
2
1.47 n = and 1.51,
then the permitted normalized frequency range of
01
TE mode are (9.74-1.38) and (9.27-1.31),
respectively. They are essentially consistent with the normalized frequency range (9.5-1.34)
as originally designed for
2
1.49 n = and
0.05 =
.
The confinement loss and transmission loss for hollow-core Bragg fiber with cobweb-
structured cladding were modelled by using an asymptotic formalism [34]. Many results show
that the fibers with only 3-4 alternating layers in cladding can achieve the low confinement
loss and transmission loss, and the confinement and transmission losses decrease with
increasing the hollow-core radius (
co
r ). In order to achieve both low loss and wide
wavelength range, fiber design should adopt smaller
2
d value and lager
1
d value, besides
increasing
co
r and N .
5. Functional Exploiting of Hollow-Core Bragg Fiber with
Cobweb - Structured Cladding
With the appealing properties described above, the possibility of using hollow-core Bragg
fiber with cobweb-structured cladding for transmitting the information and delivering the
laser energy was analyzed.
5.1. Fibers for Use in Optical Communications from Visible to near Infrared
Region
Today, the capacity of optical fiber communications has expanded gigabits per second into
terabits per second, enough to meet the current traffic demand due to the explosive growth of
data transfer and internet services. Large-capacity and long-distance optical fiber
communication trunk line has been installed in many countries. The next big step will be
extending the network from fiber-to-the-curb into every building and home.
In the area of fiber to the home (FTTH) or fiber to the premises (FTTP) application,
passive optical networks (PON), especially ethernet passive optical networks (EPON) and
gigabit ethernet passive optical networks (GEPON), are generally preferred for home fiber
connections. Usually, the transmission bandwidth and transmission distance required for the
networks are 100MHz - 10GHz and 100 -10 m km, respectively. Therefore, the fibers with
lower loss, higher bandwidth and cheaper cost are in demand. People have been trying to find
materials and methods to meet those requirements and POF is one of the major approaches
being explored in addition to silica glass single-mode fiber and multi-mode hard plastic clad
fiber (HPCF).
The simulated results for hollow-core Bragg fibers with cobweb-structured cladding had
proved that depending on the modal-filtering effect, they may realize the transmission of
01
TE single-mode or a few modes, thus achieving the transmission of higher bandwidth
( GHz ) [35].

Figure 10. Absorption loss spectrum of PMMA [36].
A fiber design for use in optical communication from visible to near infrared region is
presented. The fiber parameters are
2
1.49 n = (PMMA),
1
1 n = ,
2
0.25 d m = ,
1
5 d m = ,
75
co
r m = and 3 N = . According to absorption loss spectrum of PMMA [36] as shown in
Fig.10, we calculate the transmission losses of the fiber. The absorption losses of PMMA at
the wavelengths of 0.65 , 0.85 , 1.3 and 1.55 m are about 100 / dB km ,
3
2.5 10 / dB km ,
4
2.5 10 / dB km and
4
7.8 10 / dB km , respectively. The transmission losses of
01
TE mode at
these wavelengths are
4
3.9 10 / dB km
,
3
4.3 10 / dB km
, 0.13 / dB km and 0.80 / dB km ,

respectively. The results show that after inevitable factors (material purity, imperfection and
nonuniformity of fiber structure and existence of supporting strips) being considered, the
transmission losses of the fiber are still very low.
Thus, by using an inexpensive material (PMMA), it allows the fibers to meet the needs of
the transmission distance and bandwidth for EPON and GEPON, and to realize the
wavelength division multiplexing (WDM).
5.2. Fibers for Use in THz Waveguiding
The THz radiation, whose frequency range is about 0.1 10THz , has important applications
in spectroscopy, imaging, space science and information transmission. To date, progress in
THz wave generation and detection techniques has been enormous. However, most of the
present THz systems rely on free space propagation due to the absence of low loss
waveguides and transparent materials in the THz region. The waveguides constructed with
some metals suitable for microwave guides or some dielectrics (such as silica) suitable to
optical waveguiding have very high losses for THz wave. Even if for high-resistivity silicon,
the most common material for the passive devices in the THz technology, its absorption
coefficient is of the order of
1
0.04cm
. In recent years, THz waveguides have been

fabricated from some dielectrics (such as sapphire, plastics) except from metals such as Cu,
brass, and stainless steel. The loss coefficients of high-index core (solid-core) photonic
crystal fibers using high-density polyethylene (HDPE) [37] and polytetrafluoroethylene
(Teflon) [38] are less than
1
0.5cm
( 0.1 3THz ) and approximately

1
0.12cm
, respectively.
Hollow polycarbonate waveguides with inner Cu coatings for broadband THz transmission
have been reported [39]. The lowest loss 3.9 /m dB (
1
0.00898cm
) was obtained from a 3mm

core diameter fiber at 158.51 m wavelength. Recently, a simple subwavelength-diameter
( 200 m ) plastic (polyethylene) wire, similar to an optical fiber for guiding a THz wave has
been reported as well [40]. Its attenuation constant is reduced to less than
1
0.01cm
in the
frequency range near 0.3THz .
A fiber design for use in THz waveguiding is presented. The structural parameters of
fibers (A, B, C) are as follows: 9
co
r mm = ,
2
1.52 n = ,
1
1 n = ,
2
25 d m = ,
1
500 d m = and
3 N = (fiber A); 12
co
r mm = ,
2
1.52 n = ,
1
1 n = ,
2
70 d m = ,
1
1050 d m = and 3 N = (fiber
B); 16
co
r mm = ,
2
1.52 n = ,
1
1 n = ,
2
150 d m = ,
1
2250 d m = and 3 N = (fiber C). The host
material was assumed to be lossless with a refractive index of 1.52 (corresponding to HDPE
material). The confinement loss as a function of wavelength for
01
TE ,
02
TE ,
01
TM and
02
TM modes is shown in Fig.11. The lowest-loss mode is
01
TE mode. The confinement loss
of the
01
TE mode at the least-loss wavelength is
8
1.13 10 / dB km
at 83.5 m (fiber A),

7
2.15 10 / dB km
at 233 m (fiber B), and

7
5.05 10 / dB km
at 500 m (fiber C).

Figure 11. Confinement loss as a function of wavelength for
01 02 01
, , , TE TE TM and
02
TM modes.

Figure 12. Transmission loss as a function of wavelength for
01 02 01
, , , TE TE TM and
02
TM modes.
Then, we attempt to take the calculation of losses further by including the material
absorption. Based on the absorption spectra of HDPE in wavelength range
50 1200 m m [41], the transmission losses of three hollow-core fibers (A, B, C) with
cobweb cladding are calculated. The calculated results are shown in Fig.12. The data in
Fig.12 show that the transmission losses of
01
TE mode for fiber A in the wavelength range of
2 6 0 5 0 m m are below 5.5 / dB km. The lowest loss is 0.63 / dB km (corresponding to loss
coefficient
6 1
1.45 10 cm

) at 90 m . The transmission losses of
01
TE mode for fiber B in
the wavelength range of 2 5 0 4 0 0 m m are below 5.0 / dB km . The lowest loss is
2.0 / dB km at 270 m . The transmission losses of
01
TE mode for fiber C in the wavelength
range of 1 4 0 20 00 m m are below 5.6 / dB km. The lowest loss is 2.09 / dB km at 560 m .
The above transmission losses were taken into account only the absorption spectra of the
material (HDPE). In fact, certain spectral region in the THz waves may not be available for
signal transmission due to the strong absorption of water present in the constituent materials
and air-core for the polymer fibers [42]. Therefore, while using hollow-core polymer Bragg
fiber with cobweb-structured cladding in transmitting light through air-core, it is very
important to eliminate the water from the constituent material and avoid moist air in the
environment during fabrication and storage.
5.3. Fibers for Infrared (IR) Applications
IR optical fibers may be defined as fiber optics transmitting wavelengths greater than
approximately 2 m . IR fibers can be useful for the medical, industrial, civil, and military
arenas. For example, they are used in surgical applications by transmitting
2
CO laser
radiation(10.6 ) m and : Er YAG laser radiation (2.94 ) m . When used as fiber sensors, IR
fibers are generally used either to transmit blackbody radiation for temperature measurements
or as an active or passive link for chemical sensing, achieving non-contact temperature
monitoring and remote spectroscopic chemical sensing. The application in the industrial arena
includes welding and cutting. Scanning near-field microscopy by using high-quality single-
mode and multimode IR fiber-tapered tips can obtain 20-nm topographic resolution and
about 200-nm optical resolution for a variety of samples. IR fibers are also used for military
applications including anti-aircraft missile defense. The development of infrared fiber optics
began in the year 1960. The first IR fibers were fabricated in the mid-1960s using arsenic-
sulphur glasses [43]. So far, there are four classes of infrared fibers:

(i) fluoride, germanate, tellurite or chalcogenide glass based solid-core fibers;
(ii) crystalline silver halide solid-core fibers;
(iii) hollow-core fibers in which inner wall coatings have high reflectivity; and
(iv) solid-core photonic crystal fibers and hollow-core photonic bandgap fibers.

The optical-loss values of the sulfide based chalcogenide glass fibers at the Naval
Research Laboratory have been reduced to only 0.1 to 0.2 / dB m in fiber lengths of about
500m by using improved chemical purification and better fiber fabrication techniques [44].
The optical losses of crystalline silver halide solid-core fibers by an extrusion process have
been reduced to lower than 50 / dB km in a broad IR region from 9 to 14 m and lower than
1 /m dB in the region from 3 to 20 m [45]. The losses of rectangular hollow waveguides
with 1- m -long and 1 1 mm mm cross-section by first depositing thin-film coatings of
2
PbF
on phosphor bronze strips and then soldering four of these phosphor bronze metal strips
together are as low as 0.1 / dB m at 10.6 m [46]. Photonic crystal fibers for the middle
infrared were fabricated by multiple extrusions of silver halide crystalline materials [47].
These fibers are composed of two solid materials: the core consists of pure AgBr (n=2.16)
and the cladding includes AgCl (n=1.98) fiberoptic elements arranged in two concentric
hexagonal rings around the core. IR transmissive As-S glass and As-Se glass triangular
photonic band gap fiber structures were theoretically modeled [48]. From numerical
simulations, Pottage et al. [49] discovered a new type of air-line bandgap that is of
considerable importance in the design of practical hollow-core photonic bandgap fibers made
from high-index glass (n2.0) for guidance in the mid/far-IR. A silica based hollow-core
photonic bandgap fiber in which fiber-core diameter is 40 m (nineteen capillaries were
omitted from the centre of the stack to form the core), the overall outside diameter is 150 m
and the nearest-neighbor hole spacing is around 7 m , has been fabricated [50]. The peak of
the bandgap is at 3.14 m with a typical attenuation of 2.6 / dB m . By further optimization
of the structure, modeling suggests that a loss below 1 / dB m should be achievable.
The design is a hollow-core Bragg fiber with cobweb-structured cladding for the mid-IR
region. In the wavelength region between 100 m and 1 m , many longitudinal and rotational
resonances of molecules are present in almost all substances, especially the long-chain
polymers [2]. Polymers such as teflon and polyethylene show relatively strong absorption at
1
1000cm
(10 m ). The absorption coefficient at 10 m wavelength is about

1
100cm
for
teflon and about
1
50cm
for polyethylene [16]. A fiber design for use in infrared is presented.

The structural parameters of fibers (A, B) are as follows: 1500
co
r m = ,
2
1.4 d m = ,
1
30 d m = ,
1
1 n = ,
2
1.37 n = (teflon) and 3 N = (fiber A); 1200
co
r m = ,
2
2.8 d m = ,
1
28 d m = ,
1
1 n = ,
2
1.55 n = (PES) and 3 N = (fiber B). The absorption coefficient of the
host material (teflon) is
1
100cm
(corresponding to absorption loss

7
4.343 10 / dB km ). The
calculated results are shown in Fig.13(a). The data in Fig.13(a) show that the transmission
loss of
01
TE mode for fiber A in the wavelength range of 2.8 m to 10.6 m are below
39.5 / dB km . The lowest loss is 1.47 / dB km at 3.9 m . The absorption loss of the host
material (PES) is
7
3 10 / dB km . The calculated results are shown in Fig.13(b). The data in
Fig. 13(b) show that the transmission loss of
01
TE mode for fiber B in the wavelength range
of 8 m to 13 m are below 30.9 / dB km. The loss for the 10.6 m wavelength of
2
CO laser
is 18.9 / dB km.

Figure 13. Transmission losses of the mid-IR region for fibers (A, B).
The numerical results show that despite the strong absorption of the polymers in the mid-
IR region, the transmission losses of the fibers are lower by comparison with those of other
IR fibers reported in the literature. And the polymer fibers have an advantage over other
fibers in flexibility.
5.4. Circular-Polarization-Maintaining Single-Mode Fibers
Standard single-mode fibers support two degenerate, orthogonally polarized modes (
11
HE
mode). Random imperfections in the fiber structure and external forces on the fibers can
create asymmetries that break the polarization degeneracy, resulting in polarization mode
dispersion and polarization fading in interferometers. Conventional polarization-maintaining
fibers (highly birefringent fibers) and some single-polarization single-mode photonic crystal
fibers supported a linear polarization mode. The fibers require accurate alignment of the
birefringence axes of the two fibers when coupling, splicing and some sensing applications
are considered. Therefore, in the year 1980, Jeunhomme and Monerie [51] have suggested the
design of a circular-polarization-maintaining single-mode fiber cable . Recently, Argyros et al.
[25] have presented the design that supports a single-polarization, circularly symmetric
nondegenerate mode in an air-core ring-structured Bragg fiber.
We presented the design that supports a circular-polarization-maintaining single mode in
a hollow-core and cobweb-structured cladding Bragg fiber. The structural parameters of the
fiber are 10
co
r m = ,
1
1 n = ,
2
1.585 n = (PC),
2
0.21 d m = ,
1
2.1 d m = and 3 N = . The
intrinsic losses of the host material (PC) are 166 / dB km at 650 656nm and 224 / dB km at
764nm[52]. The calculated results show that the transmission losses of
01
TE mode (lowest
loss mode) are 0.226 / dB km at 650 656nm and 0.170 / dB km at 764nm , those of
02
TE
mode (second-lowest loss mode) are 3.513 / dB km at 650 656nm and 1.848 / dB km at
764nm . The ratio of the loss of the
02
01
TE mode is 15.54
( 650 656nm ) and 10.87 ( 764nm). In accordance with the research reported in Ref.32, the
fiber is single moded for lengths between 11.4km and 88.5km( 650 656nm ), and 21.7km
and 117.6km ( 764nm).
We expect that this type of hollow-core Bragg fibers with circular-polarization-
maintaining single-mode and low-losses will find many applications, such as gyroscopes,
current sensors and coherent communication systems.
6. Applications of Hollow-Core Bragg Fiber with Cobweb-
Structured Cladding
A new generation of POFs has the advantages of both low-cost and high-performance in
terms of attenuation, bandwidth and flexibility. It will find many applications in diverse areas
and increases market acceptance.
As respects information transmissions, the new generation of POFs can guide the light of
visible to terahertz radiation, and can be applied to optical fiber communications and optical
fiber sensing, such as LANs, specially FTTH, THz wave fiber communications. It can also
be used as an active or passive links for chemical sensing and remote spectroscopic chemical
sensing, a variety of physical quantity sensing as well as medical diagnostics including
noninvasive blood glucose monitoring and detection of tumors. As respects delivery of power
laser radiation and solar radiation, hollow-core Bragg fibers with cobweb-structured cladding
can deliver solar radiation into darkroom, be used for indoor illumination, replacing former
guided light tube or solid-core polymer fiber. Efficient hollow-core guiding for delivery of
power laser radiation (
10.6 m

2
CO laser,
2.94 m
: Er YAG laser, etc) can be achieved by
replacing the traditional PMMA with heat-resistant polymers, and can be used for medical
therapy and processing including micro-processing and material processing.
By using gas-filled hollow-core Bragg fibers with cobweb-structured cladding, it is
possible to obtain the EUV light generated through the process of high-harmonic up-
conversion of femtosecond laser and ultrahigh efficiency laser wavelength conversion by pure
stimulated rotational Raman scattering, as well as to use laser light to levitate and guide
particles through the hollow-core fiber, etc.
Circular-polarization-maintaining single-mode low-loss fibers and high-strength,
flexibility and resistance to shock fibers will provide the possibilities for some new
applications. These fibers will stimulate further progress, both in fiber and allied systems
technologies.
The new generation of POFs based on hollow-core Bragg fiber with cobweb-structured
cladding will find many applications and is irreplaceable for some applications such as THz
wave low-loss transmission.
References
[1] D. Hondros and P. Debye. Ann. Physik. 1910, vol.32, 465-476
[2] K.C. Kao and G.A. Hockham. Proc. IEE 1966, vol.113(7), 1151-1158
[3] P. Yeh, A. Yariv, and E. Marom. J. Opt. Soc. Am. 1978, vol.68, 1196-1201
[4] Y. Fink, D.J. Ripin, S. Fan, C. Chen, J.D. Joannopoulos, and E.L. Thoms. J. Lightwave
Technol. 1999, vol.17, 2039-2041
[5] R.F. Cregan, B.J. Mangan, J.C. Knight, T.A. Birks, P.st.J. Russell, P.J. Roberts, and D.C.
Allan. Science 1999, vol.285, 1537-1539
[6] Y. Koike. Status of POF for fiber to the display. 15th Inter. Conf. on POF(POF 2006),
17-22, Seoul, Korea
[7] T. Kaino, K. Jinguji, and S. Nara. Appl. Phys. Lett. 1983, vol.42, 567-569
[8] http://www.lucina.jp/eg_lucina/productseng f2.htm
[9] Y. Koike and E. Nihei. Proc. 17th Euro. Conf. Opt. Comm.(ECOC91), pp.41-44, Paris,
France, 1991
[10] K. Nakamura, J. Okumura, K. Irie, and K. Shimada. Proc. Int1 POF Conf.(POF95),
pp.114-116, Boston, USA, 1995
[11] H.Munekuni, S. Katsuta, and S. Teshima. Proc. Int1 POF Conf.(POF94), pp.148-151,
Yokohama, Japan, 1994
[12] Y. Koike, T. Ishigure, and E. Nihei. J. Lightwave Technol. 1995, vol.13(7), 1475-1489
[13] T. Ishigure, Y. Koike, and J.W. Fleming. J. Lightwave Technol. 2000, vol.18(2),178-184
[14] O. Ziemann, W. Daum, A. Bruer, J. Schlick, and W. Frank. Conf. Proc. POF 2000,
173-177, Cambridge, USA
[15] R.A. Bartels, A. Paul, H. Green, H.C. Kapteyn, M.M. Murnane, S. Backus, I.P. Christov,
Y. Liu, D. Attwood, and C. Jacobsen. Science 2002, vol.297, 376-378
[16] T. Hidaka, T. Morikawa, and J. Shimada. J. Appl. Phys. 1981, vol.52, 4467-4471
[17] C.C. Gregory and J.A. Harrington. Appl. Opt. 1993, vol. 32, 5302-5309
[18] A. Hongo, K. Morosawa, K. Matsumoto, T. Shiota, and T. Hashimoto. Appl. Opt. 1992,
vol 31, 5114-5120
[19] J.A. Harrington, R. George, P. Pedersen, and E. Mueller. Opt. Express 2004, vol.12,
5263-5268
[20] T.S. Guan, M.Y. Chen, Z.L. Zhang, and R.J. Yu. Chin. Opt. Lett. 2005, vol.3, 313-315
[21] Y. Xu and A. Yariv. Opt. Lett. 2003, vol.28(20), 1885-1887
[22] P.J. Roberts, F. Couny, H. Sabert, B.J. Mangan, D.P. Williams, L. Farr, M.W. Mason, A.
Tomlinson, T.A. Birks, J.C. Knight, and P.st.J. Russell. Opt. Express 2005, vol.13, 236-
244
[23] S.G. Johnson, M. Ibanescu, M. Skorobogatiy, O. Weisberg, T.D. Engeness, M. Soljai,
S.A. Jacobs, J.D. Joannopoulos, and Y.Fink. Opt. Express 2001, vol.9 , 748-779
[24] A. Argyros, I.M. Bassett, M.A. van Eijkelenborg, M.C.J. Large, J. Zagari, N.A.P.
Nicorovici, R.C. McPhedran, and C.M. de Sterke. Opt. Express 2001, vol.9, 813-820
[25] A. Argyros, N. Issa, I. Bassett, and M.A. van Eijkelenborg. Opt. Lett. 2004, vol.29,
20-22
[26] M.Y. Chen, R.J. Yu, Z.G. Tian, and X.Z. Bai. Chin. Opt. Lett. 2006, vol.4, 63-65
[27] L. Huo, R.J. Yu, B. Zhang, M.Y. Chen, and B.X. Li. Chin. Phys. Lett. 2006, vol.23,
2121-2124
[28] G. Dellemann, T.D. Engeness, M. Skorobogatiy, and U. Kolodny. Photonics Spectra
June 2003, vol.37, 60-64
[29] B. Temelkuran, S.D. Hart, G. Benoit, J.D. Joannopoulos, and Y. Fink. Nature 2002,
vol.420, 650-653.
[30] A. Argyros, M.A. van Eijkelenborg, M.C.J. Large, and I.M. Bassett. Opt .Lett. 2006,
vol.31, 172-174
[31] C.M. de Sterke, I.M. Bassett, and A.G. Street. J. Appl. Phys. 1994, vol.76, 680-688
[32] I.M. Bassett and A. Argyros. Opt. Express 2002, vol.10, 1342-1346
[33] S.G. Johnson and J.D. Joannopoulos. Opt. Express 2001, vol.8, 173-190
[34] Y. Xu, A. Yariv, J.G. Fleming, and S.-Y. Lin. Opt. Express 2003, vol.11, 1039-1049
[35] R.J. Yu, B. Zhang, M.Y. Chen, L. Huo, Z.G. Tian, and X.Z. Bai. Opt. Commun. 2006,
vol.266, 536-540
[36] T. Kaino, M. Fujiki, and K. Jinguji. Rev. Elec. Commun. Lab. 1984, vol.32, 478-488
[37] H. Han, H. Park, M. Cho, and J.Kim. Appl. Phys. Lett. 2002, vol.80, 2634-2636
[38] M. Goto, A. Quema, H. Takahashi, S. Ono, and N. Sarukura. Jpn. J. Appl. Phys. 2004,
vol.43(No. 2B), L317-L319
[39] J.A. Harrington, R. George, P. Pedersen, and E. Mueller. Opt. Express 2004, vol.12,
5263-5268
[40] L.-J. Chen, H.-W. Chen, T.-F. Kao, J.-Y. Lu, and C.-K. Sun. Opt. Lett. 2006, vol.31,
308-310
[41] G.W. Chantry, J.W. Fleming, P.M. Smith, M. Cudby, and H.A. Willis. Chem. Phys. Lett.
1971, vol.10, 473-477
[42] R.-J. Yu, B. Zhang, Y.-Q. Zhang. C.-Q Wu, Z.-G. Tian, and X.-Z. Bai. IEEE Photon.
Technol. Lett. 2007, vol.19, 910-912
[43] N.S. Kapany and R.J. Simms. Infrared Phys. 1965, vol.5, 69-80
[44] J.S. Sanghera, I.D. Aggarwal, L.E. Busse, P.C. Pureza, V.Q. Nguyen, F.H. Kung, L.B.
Shaw, and F. Chenard. Laser Focus World April 2005, vol.41, 83-87
[45] L.N. Butvina, E.M. Dianov, N.V. Lichkova, V.N. Zavgorodnev, an L. Kuepper. Proc.
SPIE 2000, vol.4083, 238-253
[46] H. Machida, Y. Matsuura, H. Ishikawa, and M. Miyagi. Appl. Opt. 1992, vol.31, 7617-
7622
[47] E. Rave, P. Ephrat, M. Goldberg, E. Kedmi, and A. Katzir. Appl. Opt. 2004, vol.43(11),
2236-2241
[48] L.B. Shaw, J.S. Sanghera, I.D. Aggarwal, and F.H. Kung. Opt. Express 2003, vol.11,
3455-3460
[49] J.M. Pottage, D.M. Bird, T.D. Hedley, T.A. Birks, J.C. Knight, P.st.J. Russell, and P.J.
Roberts. Opt. Express 2003, vol.11(22), 2854-2861
[50] J.D. Shephard, W.N. MacPherson, R.R.J. Maier, J.D.C. Jones, D.P. Hand, M. Mohebbi,
A.K. George, P.J. Roberts, and J.C. Knight. Opt. Express 2005, vol.13(18), 7139-7144
[51] L. Jeunhomme and M. Monerie. Electron. Lett. 1980, vol.16, 921-922
[52] T. Yamashita and K. Kamada. Jpn. J. Appl. Phys. 1993, vol.32, 2681-2686

Chapter 10
DISSIPATIVE SOLITONS IN OPTICAL FIBER SYSTEMS
Mrio F.S. Ferreira and Sofia C.V. Latas
Department of Physics, University of Aveiro, 3800-193 Aveiro, Portugal
Abstract
We introduce the concept of dissipative solitons, which emerge as a result of a double
balance: between nonlinearity and dispersion and also between gain and loss. Such dissipative
solitons have many unique properties which differ from those of their conservative
counterparts and which make them similar to living things. We focus our discussion on
dissipative solitons in optical fiber systems, which can be described by the cubic-quintic
complex Ginzburg-Landau equation (CGLE). The conditions to have stable solutions of the
CGLE are discussed using the perturbation theory. Several exact analytical solutions, namely
in the form of fixed-amplitude and arbitrary-amplitude solitons, are presented. The numerical
solutions of the quintic CGLE include plain pulses, flat-top pulses, and composite pulses,
among others. The interaction between plain and composite pulses is analyzed using a two-
dimensional phase space. Stable bound states of both plain and composite pulses are found
when the phase difference between them is 2 / . The possibility of constructing
multisoliton solutions is also demonstrated.
1. Introduction
Solitary waves have been the subject of intense theoretical and experimental studies in many
different fields, including hydrodynamics, nonlinear optics, plasma physics, and biology [1]-
[5]. In fact, the history of solitons dates back to 1834, the year in which James Scott Russell
observed that a heap of water in a canal propagated undistorted over several kilometres [6].
However, the term soliton was coined only in 1965, to reflect the particle-like nature of
solitary waves that remain intact even after mutual collisions [7]. Such waves correspond to
localized solutions of integrable equations such as the Korteveg de Vries and nonlinear
Schrdinger equations. In these circumstances, solitons were usually attributed only to
integrable systems. However, the concept of soliton was subsequently broaden to include also
the localized solutions of non-integrable systems.
Mrio F.S. Ferreira and Sofia C.V. Latas 280
Concerning the field of nonlinear optics, one can distinguish between temporal and
spatial solitons [8]. Spatial optical solitons are beams of light in which nonlinearity
counteracts diffraction, leading to a robust structure which propagates without change of
form. Such structures will play a major role in the future in the field of all-optical processing
and logic. Temporal solitons, on the other hand, represent shape invariant (or breathing)
pulses, formed by a balance between nonlinearity and dispersion. It is believed that temporal
solitons will play a major role in future all-optical high-capacity transmission systems [9]
[10].
Until now, the main emphasis has been given to the well-known conservative soliton
systems, where only the diffraction or dispersion needs to be balanced by the nonlinearity.
However, a new field has emerged in the last few years concerning the formation of solitons
in systems far from equilibrium [11]. These solitons are termed dissipative solitons or auto-
solitons and they emerge as a result of a double balance: between nonlinearity and dispersion
and also between gain and loss. Such dissipative solitons have many unique properties which
differ from those of their conservative counterparts. For example, except for very few cases
[5], they form zero-parameter families and their properties are completely determined by the
external parameters of the optical system. They can exist indefinitely in time, as long as these
parameters stay constant. However, they cease to exist when the source of energy or matter is
switched off, or if the parameters of the system move outside the range which provides their
existence.
Even if it is a stationary object, a dissipative soliton shows non-trivial energy flows with
the environment and between different parts of the pulse. Hence the dissipative soliton is an
object which is far from equilibrium and which presents characteristics similar to a living
thing. In fact, we can consider animal species in nature as elaborate forms of dissipative
solitons. An animal is a localized and persistent structure which has material and energy
inputs and outputs and complicated internal dynamics. Moreover, it exists only for a certain
range of parameters (pressure, temperature, humidity, etc.) and dies if the supply of energy is
switched off. The same analogy can be applied to individual organs within an animal, since
each maintains its shape and function over time.
Many non-equilibrium phenomena, such as convection instabilities, binary fluid
convection and phase transitions, can be described by the complex Ginzburg-Landau equation
(CGLE) [12]-[14]. In the field of nonlinear optics, the CGLE can describe various systems,
namely optical parametric oscillators, free-electron laser oscillators, spatial and temporal
soliton lasers, and all-optical transmission lines [9][15]-[27]. In these systems there are
dispersive elements, linear and nonlinear gain, as well as losses. In some cases, the CGLE
admits a multiplicity of solutions for the same range of system parameters. This reality again
resembles the world of biology, where the number of species existing in the same
environment is trully impressive.
In this chapter we will discuss the cubic-quintic CGLE and the characteristics of some of
its solutions. In Section 2 we present the CGLE and in Section 3 the perturbation approach to
solve this equation is discussed. Some analytical and numerical solutions of the CGLE are
presented in Sections 4 and 5, respectively. Finally, Section 6 summarizes the main
conclusions.
Dissipative Solitons in Optical Fiber Systems 281
2. The Complex Ginzburg-Landau Equation
In one of the forms used in nonlinear optics, the cubic-quintic complex Ginzburg-Landau
equation (CGLE) can be written as [5][19]-[27]:
q q q q i q q q i
T
q
i q i q q
T
q D
Z
q
i
4 4 2
2
2
2
2
2
2

+ + + = + + (1)
where Z is the propagation distance or the normalized number of round trips, T is the retarded
time, q is the normalized envelope of the electric field, stands for spectral filtering ( >0),
is the linear gain or loss coefficient, accounts for nonlinear gain-absorption processes
(for example, two-photon absorption), represents a higher order correction to the nonlinear
gain-absorption, and is a higher order correction term to the nonlinear refractive index.
The parameter D is the group velocity dispersion coefficient, with 1 = D , depending on
whether the group velocity dispersion (GVD) is anomalous or normal, respectively.
The CGLE is rather general, as it includes dispersive and nonlinear effects, in both
conservative and dissipative forms. It is known in many branches of physics, including fluid
dynamics, nonlinear optics and laser physics.
Equation (1) becomes the standard nonlinear Schrdinger equation (NLSE) when the
right-hand side is set to zero. When this does not happen, Eq. (1) is non-integrable, and only
particular exact solutions can be obtained. In the case of the cubic CGLE ( 0 = = ), exact
solutions can be obtained using a special ansatz [28], Horota bilinear method [29] or
reduction to systems of linear PDEs [30]. Concerning the quintic CGLE, the existence of
soliton-like solutions in the case 0 > has been demonstrated both analytically and
numerically [5][20][26][31]. Exact solutions of the quintic CGLE, including solitons, sinks,
fronts and sources, were obtained in [32], using Painlev analysis and symbolic computations.
It must be noted that Eq. (1) can not be used as it stands to describe the behaviour of
femtosecond optical pulses. For such ultrashort pulses, some higher-order nonlinear and
dispersive effects must be taken into account, which results in additional terms to be added to
the right-hand side of Eq. (1) [33]-[38].
3. Results from the Soliton Perturbation Theory
Assuming that D=+1 and that all the other coefficients in the right-hand side of Eq. (1) are
small, we can use the adiabatic soliton perturbation theory [9][34][39][40] to evaluate the
dynamical evolution of the soliton parameters the amplitude and the frequency , with
which the one soliton solution is given by:
[ ] { } [ ]

+ + = i Z Z Z
i
T Z i Z T Z h Z Z T q
2 2
) ( ) (
2
) ( exp ) ( ) ( sec ) ( ) , (
(2)
Applying the perturbation procedure, we get the following set of ordinary-differential
equations:
5 3 2 2
15
16
3
4
3
1
2 2
+ +
+ =
dZ
d
(3)

2
3
4
=
dZ
d
(4)
As can be seen from Eq. (4), the soliton frequency approaches asymptotically to 0 =
(stable fixed point) if 0 . The stable fixed points for the soliton amplitude, on the other
hand, are given by minimums of the potential function defined by:

d
d
dZ
d
= (5)
Considering the Eq. (3), we have the following expression for the potential function:
( )
6 4 2
45
8
2
6
1
) ( + = (6)
For the zero-amplitude state to be stable, the potential function must have a minimum at
0 = , in addition to a minimum at 0 =
s
. These objectives can be achieved if the
following conditions are verified [20]:
0 < , 0 < , 2 / > ,
4
8 15
s
> (7)
We can verify from the above conditions that the inclusion of the quintic term in Eq. (1)
is necessary to have the double minimum potential.
The stationary value for the soliton amplitude can be obtained from Eq. (6) and is given
by:
8
5 / 24 ) ( 5 ) ( 5
2
2

=
s s
s
(8)
where 2 / =
s
for small values of . However, the result given by Eq. (8) can be
generalized for arbitrary values of using
s
given by [20][26][41]:
2
2
9 2
1 4 1 3
2

+
+
=
s
(9)
From Eq. (8) it can be seen that a stationary amplitude 1 =
s
occurs when the
coefficients satisfy the relation:
0 8 ) 2 ( 5 15 = + + (10)
The discriminant in Eq. (8) must be greater than or equal to zero for the solution to exist.
For given values of , , and , the allowed values of to guarantee a stable pulse
propagation must satisfy the condition 0
min
, where
( )
24
5
2
min
s
= (11)
When 0 = , the peak amplitude is found to achieve a maximum value:
( )
=
4
5
max
(12)
For 0 = and
s
= the peak amplitude becomes arbitrary.
On the other hand, for given values of , , and , the minimum value of allowed
becomes
5 / 24
min
+ =
s
(13)
Considering the last condition in Eq. (7) or, alternatively, from Eq.s (8) and (13) we find
that there is a minimum value for the peak amplitude, given by:
4
min
8
15
= (14)
Fig. 1 shows the potential function given by Eq. (6) when the relation (10) is satisfied for
3 . 0 = , 5 . 0 = , 25 . 0 = (curve a), 34375 . 0 = (curve b) and 5 . 0 = (curve
c). Curves a and b present a minimum at 1 = and 0 = since they satisfy the conditions
(7), corresponding to negative values of the linear gain ( 05 . 0 = and 1 . 0 = ,
respectively). However, curve c has no minimum at 0 = , since the corresponding value of
is positive ( 033 . 0 = ).
Figure 1. Potential versus soliton amplitude when the relation (10) is satisfied for 3 . 0 = ,
5 . 0 = , 0 = , 5 . 0 = (curve a), 34375 . 0 = (curve b) and 25 . 0 = (curve c).
Figure 2. Phase portrait of Eq.s (3) and (4) corresponding (A) to curve c and (B) to curve b of Figure 1.
Fig. 2 illustrates the stability characteristics of the stationary solutions using the phase-
plane formalism. Fig. 2a corresponds to curve c in Fig. 1, and we observe that, in this case,
soliton propagation can be affected by background instability due to the amplification of
small-amplitude waves. The steady-state solution shows a limited basin of attraction. For
example, initial conditions with 7 . 0 =
i
and 1 =
i
evolve toward the trivial solution
0 =
s
of Eq.s (3) and (4). For these initial conditions, the nonlinearity is not sufficiently
strong to balance dispersion, and the pulse disperses away. The dashed curves in Fig. 2a give
approximate limits between different basins of attraction. From a perturbation analysis of Eq.s
(3) and (4) around 0 = , one can show that these curves cross the 0 = axis at
33 . 0 =
c
. Thus, waves weak initial amplitudes grow up to 1 =
s
if 33 . 0 <
i
. In this
case, soliton propagation can be severely affected by the background instability. Fig. 2b
corresponds to curve b in Fig. 1, and we can see that, in this case, the background instability
is avoided, since the small-amplitude waves are attenuated, irrespective of their frequency .
Besides the stable stationary point at 1 =
s
, we note, in this case, the existence of another
stationary point at 5 . 0
s
, which is unstable.
This simple approach shows that, in general, the CGLE has stationary soliton-like
solutions, and that, for the same set of equation parameters, there may be two of them
simultaneously (one stable and one unstable). Moreover, this approach shows that soliton
parameters are fixed.
4. Exact Analytical Solutions
Several types of exact analytical solutions of the CGLE have been obtained considering a
particular ansatz [5][26]. However, due to restrictions imposed by the ansatz, these solutions
do not cover the whole range of parameters. In the following, we will assume a stationary
solution of Eq. (1) in the form:
[ ] { } Z i T a id T a Z T q = ) ( ln exp ) ( ) , ( (15)
where a(T) is a real function and d, are real constants.
4.1. Solutions of the Cubic CGLE
The cubic CGLE is given by Eq. (1) with 0 = = . Inserting Eq. (15) in this equation we
obtain the following solution for a(T):
a T A h BT ( ) sec ( ) = (16)
where
2
2 2
3
2
) 2 (
B d
d B
A +
= (17)

+
=
d d
B
2
(18)
and d is given by
) 2 ( 2
) 2 ( 8 ) 2 1 ( 9 ) 2 1 ( 3
2 2

+ + +
= d (19)
On the other hand, we have
( )
( )
2
2
2
4 1
d d
d d

+
+
= (20)
The solution (16)-(18) is known as the solution of Pereira and Stenflo [28]. Although the
amplitude profile of the solution (16)-(18) is an hyperbolic secant as in the case of the NLSE
solitons, two important differences exist between the CGLE and the NLSE solitons. First, for
CGLE pulses the amplitude and width are independently fixed by the parameters of (1),
whereas for NLSE solitons A=B. The second difference is that the CGLE solitons are chirped.
The solution given by Eq.s (16)-(18) has a singularity at d d + =
2
0 , which takes
place on the line ) (
s
in the plane ( , ) defined by Eq. (9). For a given value of , the
denominator in the expression for B in Eq. (18) is positive for
s
< and negative for
s
> . Hence, for solution (16)-(18) to exist, the excess linear gain must be positive for
s
< and negative for
s
> . In the last case, both numerical simulations and the soliton
perturbation theory show that the soliton is unstable relatively to any small amplitude
fluctuations [20][26]. On the other hand, for 0 > and
s
< the solution (16)-(18) is
stable, since after any small perturbation it approaches the stationary state. However, the
background state is unstable in this case, since the positive excess gain also amplifies the
linear waves coexistent with the soliton trains. The general conclusion is that either the
soliton itself or the background state is unstable at any point in the plane ) , ( , which
means that the total solution is always unstable.
The stationary value of the pulse width 1/B can be significantly reduced by a convenient
choice of the system parameters [42]. In fact, it can be verified from Eq.s (11) and (12) that,
for a given value of the filter strength , as the nonlinear gain coefficient approaches the
value
s
given by Eq. (9), the amplitude A increases to infinity and its width 1/B tends to
zero. This singularity can be used in soliton lasers to vary the pulse parameters by a small
variation of the material parameters.
If and satisfy the Eq. (9) and = 0, a solution of the cubic CGLE with arbitrary
amplitude exists, given by [5][26]:
) ( sec ) ( DT h C T a = (21)
where C is an arbitrary positive parameter and C/D is given by:
( ) ( )
( ) 1 4 1 3 2
1 4 1 4 1 9 2
2 2
2 2 2
+
+ + +
=

D
C
(22)
We have also
d =
+ 1 4 1
2
2
(23)
2
2
2
4 1
D d
+
=
Figure 3. Simultaneous propagation of four arbitrary-amplitude solitons with with amplitudes 2, 1.5, 1,
and 0.5, for 0 = , 2 . 0 = 0 = = and
s
= .
It can be verified that the cubic CGLE becomes invariant under the scale transformation
Dq q , DT T , Z D Z
2
when = 0. This is the reason for the existence of the
arbitrary-amplitude solitons. On the ther hand, we can see that the limiting value of the
amplitude-width product A/B for the fixed-amplitude solitons coincide with the value C/D on
the line (22) [20]. This shows that arbitrary amplitude solitons can be considered as a limiting
case of fixed amplitude solitons when 0. However, the arbitrary amplitude solitons have
stability properties different from those for fixed amplitude solitons. In fact, arbitrary
amplitude solitons are stable pulses, which propagate in a stable background because = 0.
This feature is illustrated in Fig. 3, which shows the simultaneous propagation of four stable
solitons with amplitudes 2, 1.5, 1, and 0.5, for 0 = , 2 . 0 = and
s
= .
4.2. Solutions of the Quintic CGLE
Considering the quintic CGLE and inserting Eq. (15) in Eq. (1), the following general
solution can be obtained for
2
a f = [5][26]:
( ) T f f f f f f
f f
T f
2 1 2 1 2 1
2 1
2 cosh ) ( ) (
2
) (
+
= (24)
where
2
2 3 d d

= (25)
and d is given by Eq. (19). The parameters
1
f and
1
f are the roots of the equation:
0
) 4 1 ( 3
) 2 ( 2
3 8
2
2 2
2
2
=
+
+

+
+ d d
f
d
f
d d

(26)
and the coefficients are connected by the relation:
0 1
2
3 16 2
2
2
2 4 12
2 2
=
+
d d

(27)
One of the roots of Eq. (26) must be positive for the solution (24) to exist, while the other
can have either sign.
When the two roots are both positive, the general solution given by Eq. (24) becomes
wider and flatter as they approach each other. These flat-top solitons correspond to stable
pulses, whereas the solution (24) is generally unstable for arbitray choice of parameters. If
2 1
f f = , the width of the flat-top soliton tends to infinity and the soliton splits into two
fronts. The formation and stable propagation of a flat-top soliton will be demonstrated
numerically in Section 5.
If and satisfy the Eq. (9) and = 0, a solution of the quintic CGLE with arbitrary
amplitude exists, given by:
[ ]
( )
( ) ( ) T P S
P d
T a T f
2 cosh 2
4 1 3
) ( ) (
2
2
+
+
= =

(28)
where P is an arbitrary positive parameter and
( )
( )
P
d d
d
S
2
2
2 2
2
2 3
4 1 9
2

+
+ = (29)
2
1 4 1
2
+
= d
(30)
P d
2
4 1
2
+
=
When 0 , the solution (28) transforms to the arbitrary-amplitude solution of the
cubic CGLE, given by Eq. (21)-(22).
5. Numerical Solutions
Due to restrictions imposed by the ansatz, the analytic solutions of the quintic CGLE
presented above do not cover the whole range of parameters and almost all of them are
unstable. To find stable solutions in other regions of the parameters, different approximate
methods [41], a variational approach [43]-[45], or numerical techniques must be used.
As shown by the perturbative analysis presented in Section 3, the parameter space where
stable solitons exist has certain limitations. We must have 0 > in order to stabilize the
soliton in frequency domain. The linear gain coefficient must be zero or negative in order
to avoid the background instability. The parameter must be negative in order to stabilize
the soliton against collapse. Concerning the parameter , it can be positive or negative.
Stable solitons can be found numerically from the propagation equation (1) taking as the
initial condition a pulse of somewhat arbitrary profile. In fact, such profile appears to be of
little importance. For example, Fig. 4 illustrates the formation of a fixed amplitude soliton of
the cubic CGLE starting from an initial pulse with a rectangular profile. It must be noted that,
in this case, the linear gain is positive but relatively small ( 003 . 0 = ) and the soliton
propagation remains stable within the displayed distance.
In general, if the result of the numerical calculation converges to a stationary solution, it
can be considered as a stable one, and the chosen set of parameters can be deemed to belong
to the class of those which permit the existence of solitons. In the following we show some
examples of stable soliton solutions found with this method.
Figure 4. Formation of a fixed-amplitude soliton solution of the cubic CGLE starting from an initial
pulse with a rectangular profile of amplitude Ao = 0.7 (a) and Ao = 1.0 (b), when 003 . 0 = ,
2 . 0 = , and 09 . 0 = .
Figure 5. (a) Evolution of the peak amplitude and (b) the final pulse profile when 01 . 0 = ,
15 0. = , 2 0. = , 0 = , 1375 0. = (dashed curves) or 4 0. = , 3875 0. = (full
curves), considering an input pulse ) T ( h sec ) T , ( q = 0 .
Fig. 5 shows (a) the evolution of the peak amplitude and (b) the final pulse profile
obtained numerically from Eq. (1), assuming an input pulse with a sech profile and
considering the following parameter values: 01 . 0 = , 15 0. = , 0 = , 2 0. = ,
1375 0. = (dashed curves) or 4 0. = , 3875 0. = (full curves). When inserted in
Eq. (8), these values provide a stationary amplitude 1 =
s
. This prediction of the
perturbation theory, as well as the stability of the stationary solution are confirmed by the
numerical results of Fig. 5.
For small values of the parameters in the right-hand side of Eq. (1) the stable soliton
solutions of the CGLE have a sech profile, similar to the soliton solutions of the NLSE, and
correspond to the so-called plain pulses (PPs). However, rather different pulse profiles can be
obtained for non small values of those parameters. As an example, Fig. 6 illustrates the
formation and stable propagation of a flat-top soliton, starting from an initial pulse with a
sech profile. The following parameter values were considered: 1 . 0 = , 5 . 0 = ,
66 . 0 = , 01 . 0 = = .
Figure 6. Formation and evolution of a flat-top soliton, considering an input pulse
) T ( h sec ) T , ( q = 0 , for 1 . 0 = , 5 . 0 = , 66 . 0 = , 01 . 0 = = .
Fig. 7 shows (a) the amplitude profiles and (b) the spectra of a plain pulse, as well as of
two composite pulses (CPs). The following parameter values were considered: 01 . 0 = ,
5 . 0 = , 03 . 0 = , 0 = , 5 . 1 = (plain pulse), 0 2. = (narrow composite pulse)
and 5 2. = (wide composite pulse). Fig. 7c illustrates the formation and propagation of the
wide composite pulse starting from the plain pulse solution represented in a) and b). A
composite pulse exhibits a dual-frequency but symmetric spectrum (Fig. 6b) and can be
considered as a bound state of a plain pulse and two fronts attached to it from both sides [5].
The hill between the two fronts should be counted as a source, because it follows from the
phase profile that energy flows from the centre to the CP wings.
If one of the fronts of a CP is missing one has a moving soliton (MS) [5]. The MS always
moves with a velocity smaller than the velocity of the front for the same set of parameters. In
fact, the front tends to move with its own velocity but the soliton tends to be stationary, due to
the spectral filtering. The resulting velocity of the MS is determined by competition between
these two processes.
Increasing slightly the nonlinear gain coefficient and keeping the values of the other
parameters equal to those used in Fig. 7 the stationary wide composite pulse shown in Fig. 7c
is lost and a non stationary expanding structure appear, as illustrated in Fig. 8.
Figure 7. (a) Amplitude profiles and (b) spectra of a plain pulse and of two composite pulses when
01 . 0 = , 5 . 0 = , 03 . 0 = , 0 = , 5 . 1 = (plain pulse), 0 2. = (narrow composite
pulse) and 5 2. = (wide composite pulse). Figure 7c illustrates the formation and propagation of the
wide composite pulse, starting from the plain pulse solution.
Figure 8. Nonstationary expanding structure obtained from an initial plain pulse when 01 . 0 = ,
5 . 0 = , 03 . 0 = , 0 = and 183 . 2 = .
Pulsating and exploding soliton solutions of the CGLE were also observed recently [46].
Pulsating solitons correspond to fixed solutions in the same way as the stationary pulses and
can be found when the parameters of the CGLE are far enough from the NLSE limit. On the
other hand, exploding solitons appear for a wide range of parameters of the CGLE and
originate from soliton solutions which remain stationary only for a limited period of time.
Following the explosion, there is a cooling period, after which the solution becomes
stationary again. This is a periodic phenomenon, like other phenomena occurring in the
nature.
It can be verified that different stable stationary solutions of the quintic CGLE can exist
simultaneously for the same set of parameters [5][19]. This can be understood considering
that solitons, fronts and sources are elementary building units which can be combined to form
more complicated structures. In more complex systems, the number of solutions may be very
high. This reality again resembles the world of biology, where the number of species is trully
impressive.
6. Soliton Bound States
After finding the conditions for the existence of stable solitary-pulse solutions of the CGLE
equation, the next natural step is to consider their interactions and, in particular, the
possibility of the existence of bound states of these pulses [19][25][47]-[52]. In fact, the
problem of soliton interaction is crucial for the transmission of information. In the case of
Hamiltonian systems, the interaction between the pulses is inelastic. Energy exchange
between the pulses is one of the mechanisms that makes the two-soliton solutions of these
systems unstable, even when such stationary solutions do exist. The situation is rather
different for dissipative systems. In this case, all solutions are a result of a double balance:
between nonlinearity and dispersion and also between gain and loss. Moreover, the properties
of dissipative solitons are completely determined by the external parameters of the optical
system.
For given values of the CGLE parameters, the amplitude and width of its soliton
solutions are fixed. As a consequence, during the interaction of two solitons, basically only
two parameters may change: their separation r and the phase difference, , between them.
These two parameters provide a two-dimensional plane in which we may analyze of pulse
interaction, namely the formation of bound states, their stability and their global dynamics
[19][25][51][52]. This reduction in the number of degrees of freedom is a unique feature of
systems with gain and loss. In the case of Hamiltonian systems, the amplitudes of the solitons
can also change, which can affect the stability of the possible bound states.
In order to analyze numerically the soliton interaction in the 2-D space provided by the
separation, r, and phase difference, , between the two solitons, Eq. (1) can be solved with
an initial condition
) exp( ) 2 / ( ) 2 / ( ) (
0 0
i r T q r T q T q + + = (31)
where
0
q is the stationary solution obtained numerically from Eq. (1) when the values of its
parameters are specified. Initial condition (31) with arbitrary values for r and will result in
a trajectory on the interaction plane. Bound states will be singular points of this plane.
Figure 9. Trajectories on the interaction plane showing the evolution of two plain pulses for
01 . 0 = , 5 . 0 = , 5 . 1 = , 0 = , and 03 . 0 = . We have X ) cos( r = and
Y ) sin( r = .
Fig. 9 shows an example of a numerical simulation of an interaction between the two
solitons on the interaction plane, considering the following parameter values: 01 . 0 = ,
5 . 0 = , 5 . 1 = , 0 = , 03 . 0 = . This figure indicates that, for the given set of
parameters, there are at least four singular points. The points
3
P and
4
P are saddles and
correspond to unstable bound states. In these states, the phase difference between the solitons
is zero or . In addition, there are two symmetrically located stable foci (points
1
P and
2
P ),
which correspond to stable bound states of two solitons with a phase difference 2 / =
between them. The stationary pulse separation in these bound states is 62 . 1 r .
As a consequence of its asymmetric phase profile, the two-soliton solution corresponding
to the stable bound states
1
P and
2
P in Fig. 9 moves with a constant velocity. The direction
of motion depends on the sign of . An example of stable propagation of a two-soliton
bound state with a phase difference of 2 / between the pulses is given in Fig. 10.
Stable bound states of two CPs, with a phase difference 2 / = between them, can
also be observed. This is illustrated in Fig. 11, which shows the stable propagation of a bound
state of two composites pulses with a phase difference 2 / . The following parameter values
were assumed: 01 . 0 = , 5 . 0 = , 0 . 2 = , 0 = , 3 . 0 = . In contrast with the
behaviour of the plain pulse bound state shown in Fig. 10, the CP bound state moves at the
group velocity.
Figure 10. Propagation of a bound state of two plain pulses with a phase difference of 2 / between
them.
Figure 11. Propagation of a bound state of two composite pulses with a phase difference of 2 /
between them.
The two-soliton solution can be assumed as the building block to construct various multi-
soliton solutions. An example is given in Fig. 12, corresponding to a four-plain pulse
solution, with a phase difference of 2 / between adjacent pulses. As observed in the case of
the two-PP solution, multisoliton solutions formed by plain pulses move with a constant
velocity along the T axis.
Figure 12. Four-plain pulse solution with a phase difference of 2 / between adjacent pulses. The
dash-dotted (full) lines in (b) correspond to the initial (final) phase profiles.
Figure 13. Five-plain pulse solution and the correspondent phase profiles. The dash-dotted (full) lines in
(b) correspond to the initial (final) phase profiles of the pulses in (a).
Multisoliton solutions formed by plain pulses with zero velocity can be obtained by
choosing appropriately its phase profile. Fig. 13a illustrates the evolution of a five-soliton
solution whose initial phase profile is given by the dash-dotted line in Fig. 13b. This phase
profile evolves during the propagation, and achieves a final profile given by the full curve in
Fig. 13b. In spite of some oscillations, this multisoliton bound state remains relatively stable
and propagates with zero velocity. The final phase profile shown in Fig. 13b corresponds
indeed to a stationary stable solution. From Fig. 13 we can infer that a zero velocity multi-
soliton solution formed by plain pulses must present a symmetric and concave phase profile,
such that the temporal displacement of half of the structure is balanced by the opposite
displacement of the other half. These solutions can be the basic building blocks for more
complicated structures.
7. Conclusion
The concept of dissipative solitons was explained in this chapter. In fact, this concept is wide-
ranging and provides a new paradigm for the investigation of phenomena involving stable
structures in nonlinear systems far from equilibrium. Here, we have considered the particular
case of nonlinear optical fiber systems with gain and loss, which can be described by the
cubic-quintic complex Ginzburg-Landau equation (CGLE). These include spatial and
temporal soliton lasers, parametric amplifiers and optical transmission lines. However, the
model can also be applied in other fields of physics.
The conditions to have stable solutions of the CGLE were discussed using the
perturbation theory. Several exact analytical solutions, namely in the form of fixed-amplitude
and arbitrary-amplitude solitons, were presented. The numerical solutions of the quintic
CGLE include plain pulses, flat-top pulses, and composite pulses, among others. We used the
two-dimensional phase space (distance-phase difference) to analyze the dynamics of the two
soliton system. We have found stable bound states of both plain pulses and composite pulses
when the phase difference between them is 2 / . Two-composite pulses bound states have
zero velocity, which is in contrast with the behaviour of the bound states formed by plane
pulses. As a consequence of the existence of two-soliton bound states, three-soliton and other
multisoliton bound states also exist. In particular, we have shown the possibility of
constructing stable bound states of multiple plain pulses with zero velocity by choosing
appropriately the phase profile of the whole solution.
References
[1] Ablowitz, M. J.; Clarkson, P. A. (1991). Solitons, Nonlinear Evolution Equations, and
Inverse Scattering, Cambridge University Press, New York.
[2] Taylor, J. T., Editor (1992). Optical Solitons Theory and Experiment, Cambridge
University Press, New York.
[3] Drazin, P. G. (1993). Solitons: An Introduction, Cambridge University Press, New York.
[4] Gu, C. H. (1995). Soliton Theory and its Applications, Springer, New York.
[5] Akhmediev, N. N., and Ankiewicz, A. A. (1997). Solitons: Nonlinear Pulses and
Beams, Chapman and Hall, London.
[6] Russell, J. S. (1844). Report of 14
th
Meeting of the British Association for Advancement
of Science, York, September, 311-390.
[7] Zabusky, N. J., and Kruskal, M. D. (1965). Interactions of solitons in a collisionless
plasma and the recurrence of initial statea, Phys. Rev. Lett., 15, 240-243.
[8] Kivshar, Y. A., and Agrawal, G. P. (2003). Optical Solitons. From Fibers to Photonic
Crystals, Chapman and Hall, London.
[9] Hasegawa, A., and Kodama, Y. (1995). Solitons in Optical Communications, Clarendon
Press, Oxford.
[10] Mollenauer, L. F., and Gordon, J. P. (2006). Solitons in Optical Fibers. Fundamentals
and Applications. Elsevier Academic Press, San Diego.
[11] Akhmediev, N. N., and Ankiewicz, A. A. (2005). Dissipative Solitons, Springer, Berlin.
[12] Normand, C., and Pomeau, Y. (1977). Convective instability: a physicists approach.
Rev. Mod. Phys., 49, 581-623.
[13] Kolodner, P. (1991). Collisions between pulses of travelling-wave convection, Phys.
Rev. A, 44, 6466-6479.
[14] Graham, R. (1975). Fluctuations, Instabilities and Phase Transictions, Springer, Berlin.
[15] Staliunas, K. (1993). Laser Ginzburg-Landau equation and laser hydrodynamics, Phys.
Rev. A, 48, 1573-1581.
[16] Jian, P.-S., Torruellas, W. E., Haelterman, M., Trillo, S., Peschel, U., and Lederer, F.
(1999). Solitons of singly resonant optical parametric oscillators, Optics Letters, 24,
400-402.
[17] Dunlop, A. M., Wright, E. M., and Firth, W. J. (1998). Spatial soliton laser, Optics
Commun., 147, 393-401
[18] Ng, C. S., and Bhattacharjee, A. (1999). Ginzburg-Landau model and single-mode
operation of a free-electron laser oscillator, Phys. Rev. Lett., 82, 2665-2668.
[19] Akhmediev, N. N., Rodrigues, A., and Townes, G. (2001). Interaction of dual-frequency
pulses in passively mode-locked lasers, Opt. Commun. 187, 419-426.
[20] Ferreira, M. F., Faco, M. V., and Latas, S. V. (2000). Stable soliton propagation in a
system with spectral filtering and nonlinear gain, Fiber Integrated Opt. 19, 31-41.
[21] Ferreira, M. F., Faco, M. V., Latas, S. V., and Sousa, M. H. (2005). Optical solitons in
fibers for Communication systems, Fiber Integrated Opt., 24, 287-314.
[22] Ferreira, M. F., Faco, M. V., and Latas, S. V. (1999). Solitonlike pulses in a system
with nonlinear gain, Photonics and Optoelect. 5, 147-153.
[23] Matsumoto, M., Ikeda, H., Uda, T., and Hasegawa, A. (1995). Stable soliton
transmission in the system with nonlinear gain, J. Lightwave Technol., 13, 658-665.
[24] Latas, S. V., and Ferreira, M. F. (1999). Soliton stability and compression in a system
with nonlinear gain, SPIE Proc., 3899, 396-405.
[25] Akhmediev, N. N., Ankiewicz, A. A., and Soto-Crespo, J. M. (1998). Stable soliton
pairs in optical transmission lines and fiber lasers, J. Opt. Soc. Am. B 15, 515-523.
[26] Akhmediev, N. N., Afanasjev, V. V., and Soto-Crespo, J. M. (1996). Singularities and
special soliton solutions of the cubic-quintic complex Ginzburg-Landau equation, Phys.
Rev. E 53, 1190-1201.
[27] Soto-Crespo, J. M., Akhmediev, N. N., and Afanasjev, V. V. (1996). Stability of the
pulselike solutions of the quintic complex Ginzburg-Landau equation, J. Opt. Soc. Am. B
13, 1439-1448.
[28] Pereira, N. R., and Stenflo, L. (1977). Nonlinear Schrdinger equation including growth
and damping, Phys. Fluids, 20, 1733-1734.
[29] Nozaki, K., and Bekki, N. (1984). Exact solutions of the generalized Ginzburg-Landau
equation, Phys. Soc. Japan, 53, 1581-1582.
[30] Conte, R., and Musette, M. (1993). Linearity inside nonlinearity: exact solutions to the
complex Ginzburg-Landau equation, Physica, D, 69, 1-17.
[31] Thual, O. and Fauvre, S. (1988). Localized structures generated by subcritical
instabilities, J. Phys, 49, 1829-1833.
[32] Marcq, P., Chat, H., and Conte, R. (1994). Exact solutions of the one-dimensional
quintic complex Ginzburg-Landau equation, Physica D, 73, 305-317.
[33] Agrawal, G. P. (1989). Nonlinear Fiber Optics, Academic Press, San Diego.
[34] Ferreira, M. F. (1994). Analysis of femtosecond optical soliton amplification in fiber
amplifiers, Opt. Commun., 107, 365-368.
[35] Latas, S. V., and Ferreira, M. F. (2005). Soliton propagation in the presence of
intrapulse Raman scattering and nonlinear gain, Opt. Commun., 251, 415-422.
[36] Faco, M. V. and Ferreira, M. F. (2001). Analysis of the timing jitter for ultrashort
soliton communication systems using perturbation methods, J. Nonlinear Math. Phys., 8,
112-117
[37] Latas, S. V., and Ferreira, M. F. (2007). Stable soliton propagation with self-frequency
shift, J. Mathematics and Computers in Simulation, 74, 370-387
[38] Ferreira, M. F. (1997). Ultrashort soliton stability in distributed fiber amplifiers with
different pumping configurations. In Applications of Photonic Technology. Ed.s G.
Lampropoulos and R. Lessard, Plenum Press, New York, 2, 249-254.
[39] Karpman, V. I., and Maslov, E. M. (1977). Perturbation theory for solitons. Zh. Eksp.
Teor. Fiz. 73, 537-559 (1977. Sov. Phys. JETP. 46, 281-291).
[40] Essiambre, R. J., and Agrawal, G. P. (1997). Soliton communication systems. In
Progress in Optics XXXVII. Ed. E. Wolf. Amsterdam: North Holland Physics and
Elsevier Science.
[41] Soto-Crespo, J. M., and Pesquera, L. (1977). Analytical approximation of the soliton
solutions of the quintic complex Ginzburg-Landau equation, Phys. Rev. E 56 , 7288-
7295.
[42] Ferreira, M. F., and Latas, S. V. (2002). Soliton stability and compression in a system
with nonlinear gain, Optical Eng. 41, 1696-1703.
[43] Anderson, D, (1983). Variational approach to nonlinear pulse propagation in optical
fibers, Phys. Rev. A, 27, 3135-3145.
[44] Kaup, D. J., and Malomed, B. A. (1995). The variational principle for nonlinear waves
in dissipative systems, Physica D, 87, 155-159.
[45] Ankiewicz, A. A., Akhmediev, N. N., and Devine, N. (2007). Dissipative solitons with a
Lagrangian approach, Optical Fiber Technol., 13, 91-97.
[46] Akhmediev, N., Soto-Crespo, J. M., and Town, G. (2001). Pulsating solitons, chaotic
solitons, period doubling, and pulse coexistence in mode-locked lasers: complex
Ginzburg-Landau equation, Phys. Rev. E 63, 056602.
[47] Malomed, B. A. (1991). Bound solitons in the nonlinear Schrdinger-Ginzburg-Landau
equation, Phys. Rev. A 44, 6954-6957.
[48] Malomed, B. A. (1993). Bound states of envelope solitons, Phys. Rev. E 47, 2874-2880.
[49] Afanasjev, V. V., and Akhmediev, N. N. (1996). Soliton interaction in nonequilibrium
dynamical systems, Phys. Rev. E 53, 6471-6475.
[50] Ferreira, M. F., and Latas, S. V. (2001). Interaction and bound states of pulses in the
Gnizburg-Landau equation, SPIE Proc. 4271, 268-279.
[51] Ferreira, M. F., and S. V. Latas. 2002. Bound states of plain and composite pulses in
optical transmission lines and fiber lasers. In Applications of Photonic Technology. Ed.s
R. Lessard, G. Lampropoulos, and G. Schinn. SPIE. 4833:845-854.
[52] Latas, S. V. , Ferreira, M. F., and Rodrigues, A. (2005). Bound states of plain and
composite pulses: multi-soliton solutions, Optical Fiber Technol., 11, 292-305
ISBN 1-60021-866-0
Chapter 11
BRIGHT - DARK AND DOUBLE - HUMPED PULSES
IN AVERAGED, DISPERSION MANAGED OPTICAL
FIBER SYSTEMS
K.W. Chow
and K. Nakkeeran
Department of Mechanical Engineering

University of Hong Kong, Pokfulam, Hong Kong
School of Engineering, Fraser Noble Building, Kings college

University of Aberdeen, Aberdeen AB24 3UE, UK
Abstract
The envelope of the axial electric eld in a dispersion managed (DM) ber sys-
tem is governed by a nonlinear Schr odinger model. The group velocity dispersion
(GVD) varies periodically and thus realizes both the anomalous and normal dispersion
regimes. Kerr nonlinearity is assumed and a loss / gain mechanism will be incorpo-
rated. Due to the big changes in the GVD parameter, the correspondingly large vari-
ation in the quadratic phase chirp of the DM soliton will be identied. An averaging
procedure is applied. In many DM systems, an amplier at the end of the dispersion
map will compensate for the energy dissipated in that map. Here the case of gain not
exactly compensating the loss is considered, in other words, a small residual ampli-
cation / attenuation is permitted. The present model differs from other similar ones
on variable coefcient NLS, as the inhomogeneous features involve both time and the
spatial coordinate. The goal here is to extend the model further, by permitting coupled
modes or additional degree of freedom. More precisely, the coupling of ber loss and
initial chirping is considered for a birefringent ber. The corresponding dynamics is
governed by variable coefcient, coupled NLS equations for the components of the
orthogonal polarization of the pulse envelopes. When the self phase and cross phase
modulation coefcients are identical for special angles, several new classes of wave
patterns can be found. New stationary wave patterns which possess multiple peaks
within each period are found, similar to those found for the classical Manakov model.
For situations where the self- and cross-phase modulation coefcients are different,
symbiotic solitary pulses are studied. A pair of bright-dark pulses exists, where either
or both pulse(s) cannot propagate in that waveguide if coupling is absent.
302 K.W. Chow and K. Nakkeeran
1. Introduction
Transmission of information (voice, video, and data) over distances (short, moderate, long,
and ultra-long) is a common requirement in the past, present and future. Carrier communi-
cation of information using the electromagnetic waves is the best technology for high-speed
transmission. Out of different frequency bands in the electro-magnetic wave spectrum, op-
tical regime has various advantages. Optical bers are commonly used in optical commu-
nication for channelling the light pulses for digital transmission. Both linear and nonlinear
optical effects in bers play vital roles in determining the dynamics of pulse propagation.
The eld of nonlinear optics has blossomed and is undergoing a new revolution in recent
years. The nonlinear optical response is now a key element for new emerging technologies.
This is particularly true for soliton and other types of nonlinear pulse transmission in opti-
cal bers/nonlinear materials, since this form of light propagation can be used to realize the
long-held dream of very high capacity dispersion-free communications. In the recent past,
it has been proved beyond doubt that solitons do exist not only in optics but also in many
other areas of science. Solitons that exist in optics called optical solitons have been draw-
ing a greater attention among the scientic community, as they seem to be right candidates
for transferring information across the world through optical bers.
Nonlinear pulse propagation in a long-distance, high speed optical ber transmission
system can be described by the (perturbed) nonlinear Schr odinger equation (NLSE). NLSE
includes linear dynamics due to group velocity dispersion of the pulse, and nonlinear mech-
anism due to the Kerr effect [1]. Much research efforts on the development of such a system
have been made with the intention to overcome or control these effects [2, 3]. In this di-
rection, recent numerical studies [46] and experiments [7] have shown that a periodic
dispersion compensation seems to be the most effective way for improving the optical trans-
mission system. The main purpose of dispersion management is to reduce several effects,
such as radiation due to lumped ampliers compensating the ber loss [8, 9], resonant four-
wave mixing [10, 11], modulational instability [12], jitters caused by the collisions between
signals [13], and the Gordon-Haus effect resulting from the interaction with noise [14], also
to decide a desired average value for the dispersion [12].
Basically, dispersion-management technique utilizes a transmission line with a periodic
dispersion map, such that each period consists of two types of ber, generally with different
lengths and opposite group-velocity dispersion (GVD) [4]. Lakoba has proved the non-
integrability of the system equation governing the pulse propagation in dispersion-managed
(DM) bers [15]. As analytical solution for DM solitons is not available, researchers have
so far utilized the Lagrangian method to study the dynamics of DM solitons [4]. Very
recently we have developed a complete collective variable theory for DM solitons which
effectively includes the residual eld due to soliton dressing and radiation [16]. Many
works have reported on tting a Hermite-Gaussian ansatz function for the oscillating tails
of the numerical stationary solution (xed point) of the DM solitons [4, 1719]. From
numerical studies [5, 6] of DM ber line, the pulse is deformed from the ideal soliton, has
a chirp and requires an enhanced power for the average dispersion. Meanwhile Kumar and
Hasegawa [20] have obtained a new nonlinear pulse (quasi-soliton) by programming the
dispersion prole such that the wave equation has a combination of the usual quadratic
potential and the linear potential.
Bright - Dark and Double - Humped Pulses... 303
The envelope of the axial electric eld in a DM ber system is governed by a NLS
model. The GVD varies periodically and thus realizes both the anomalous and normal
dispersion regimes. Kerr nonlinearity is assumed and a loss / gain mechanism will be
incorporated. Due to the big changes in the GVD parameter, the correspondingly large
variation in the quadratic phase chirp of the DM soliton will be identied. An averaging
procedure is applied [21].
In many DM systems, an amplier at the end of the dispersion map will compensate for
the energy dissipated in that map. Here the case of gain not exactly compensating the loss
is considered, in other words, a small residual amplication / attenuation is permitted.
The present model differs from other similar ones on variable coefcient NLS [22], as
the inhomogeneous features involve both time and the spatial coordinate. The goal here is
to extend the model further, by permitting coupled modes or additional degree of freedom.
More precisely, the coupling of ber loss and initial chirping is considered for a birefrin-
gent ber [23]. The corresponding dynamics is governed by variable coefcient, coupled
NLS equations for the components of the orthogonal polarization of the pulse envelopes.
When the self phase and cross phase modulation coefcients are identical for special angles,
several new classes of wave patterns can be found.
The rst result will be a stationary wave pattern which possesses multiple peaks within
each period, similar to those found for the classical Manakov model [24].
Another new result is the family of symbiotic solitary pulses, and this novel nding is
applicable to conguration where the self phase and cross phase modulation coefcients are
different. Indeed the constraints imposed on these coefcients extend or generalize results
obtained earlier in the literature. As a simple example, a pair of bright - dark pulses exists
where each individual wave guide separately will only admit bright solitons. This coupling
nonlinearity is truly remarkable. As a second example, bright or dark solitons are allowed
to propagate in waveguides which would otherwise prohibit their existence.
2. Double-Hump Bright - Dark Periodic and Solitary Pulses
We consider the averaged, dispersion management system for coupled waveguides:
i
A
z
+

2
A
t
2
+ (AA
+ BB
)A + iA +
2
t
2
A = 0, (1)
i
B
z
+

2
B
t
2
+ (AA
+ BB
)B + iB +
2
t
2
B = 0. (2)
A, B are the complex envelopes of the axial electric elds, z is the distance and t is the
retarded time. The quantity measures the quadratic phase chirp, and the residual gain or
loss is specically selected to match this parameter. The parameter represents cross phase
modulation coefcient arising from the coupling. The derivation of system (1, 2) from the
rst principle of averaging over a dispersion map can be found in our earlier work [21].
System (1, 2) can be solved exactly by several techniques, but we shall focus on the Hi-
rota bilinear method. As the description has been given in our earlier work, the presentation
here will be brief. The quadratic phase factor or chirp is rst isolated as
A = exp
it
2
2
, B = exp
it
2
2
. (3)
The reduced governing equations for the auxiliary variables and will be free of
quadratic terms in time. To search for the special modes of optical pulses, we further
constrain the wave pattern to be expressed as
=
g exp(i
1
)
f
, =
Gexp(i
2
)
f
. (4)
G, g and f are dependent variables for the bilinear operation with the restriction that f is
real. Typically they are combinations of exponential functions for solitary pulses but elliptic
functions for periodic patterns. The phase factors
1
,
2
are functions of the distance (z)
only. They will have their derivatives determined in the bilinear equations, and hence they
themselves are readily recovered by quadrature.
The resulting bilinear equations are then
f
D
2
t
+ 2itD
t
+

1
z
+ 2i
g f
+ g(D
2
t
f f + gg
+ GG
) = 0, (5)
f
D
2
t
+ 2itD
t
+

2
z
+ 2i
G f
+ G(D
2
t
f f + gg
+ GG
) = 0. (6)
They are solved by using rather straightforward differentiation formulas developed from
rst principles. D is the bilinear operator, with its denition and properties described more
fully in Appendix A.. For periodic wave patterns, Hirota derivatives of theta functions can
be simplied by identities involving products of theta functions (Appendix B.).
As illustrative examples, the simplest periodic wave pattern will be given by the choice,
g = A
0
1
(t[h
1
(z)]), G = B
0
3
(t[h
1
(z)]), f =
4
(t[h
1
(z)]). (7)
The theta functions are Fourier series with exponentially decaying coefcients and the
classical Jacobi functions can be expressed as ratios of theta functions. The amplitude
parameters, A
0
, B
0
, isolated here for convenience will also be functions of z. The distance
dependent wave number function h
1
(z) will render the period of the pattern to change with
location, and the precise form is determined by forcing the odd Hirota derivatives to vanish.
The loss / gain factor is not arbitrary as it has to match the precise forms of the functions
A
0
, B
0
. Theta functions will be convenient in the intermediate calculations. However, the
Jacobi elliptic functions are preferred in the nal expressions, as they can be easily handled
by most established routines in computer algebra.
A summary on existing results will be instructive:
1. When the cross phase modulation coefcient, , is arbitrary, the wave system will
permit periodic patterns in terms of a single elliptic function. The long wave limit
will, not surprisingly, yield solitary bright or dark pulses.
2. When is constrained to be unity, there are other varieties of solutions. In particular,
one class of wave patterns can be expressed in terms of products of elliptic functions.
The physical implication is that the intensity will display two, or perhaps more, peaks
per period.
The goal of this section is to derive still further new wave patterns by choosing products
of elliptic functions as the starting point of these calculations, while still assuming the cross
phase modulation coefcient, , to be one. The motivation comes from the choice of wave
patterns for the case of coupled nonlinear Schr odinger models with constant coefcients.
Proceeding along the lines of reasoning just described will yield
A =
6r
1 k
2
1 2c
1 k
2
c
dn
2
(rte
2z
)
1 k
2
exp
it
2
2
2z i
1
, (8)
B =
6 rk sn(rte
2z
)dn(rte
2z
)
1 2c
1 k
2
exp
it
2
2
2z i
2
. (9)
1
=
r
2
exp(4z)
4
6c
2
(1 k
2
)
1 2c
1 k
2
+
2
1 k
2
c
, (10)
2
=
r
2
exp(4z)
4
6c
2
(1 k
2
)
1 2c
1 k
2
2(5 4k
2
)
, (11)
r is a free parameter and represents the wave number at the initial location (z = 0). The
quantity c will be one of the roots of the quadratic equation
3c
2
2c
1 k
2
+
1
1 k
2
+ 1 = 0, (12)
k is the modulus of the elliptic function. Waveguide B will generally exhibit two peaks per
period. Waveguide A will degenerate to a dark solitary pulse with multiple peaks in the
long wave period. Figures 1a, 1b illustrate this pattern.
3. A Generalized Model with Different Dispersion Coefcients
In many applications involving wave propagation along different channels or waveguides,
the optical pulses will experience different measures of group velocity dispersion. Hence
the coefcients of the second derivative terms of the coupled NLS equations will generally
be different. Remarkably a special model system will still permit analytical progress, and
we shall consider pairs of bright - dark solitons in this model. Generally bright (dark) soli-
tons occur for the conventional NLS model in the anomalous (normal) dispersion regimes
respectively. However, due to the special nonlinear effects in coupled NLS systems, these
bright / dark solitons can occur in the appropriate waveguide which are otherwise prohib-
ited in the single mode NLS. They have sometimes been termed symbiotic solitons in the
literature.
In optical physics, such waves have indeed been studied for congurations associated
with conventional NLS with Kerr nonlinearity [25], intra-pulse stimulated Raman scattering
-10
-5
0
5
10
1.0
0.5
0.0
0
0.5
1
1.5
2
t [N
o
rm
. U
n
it]
z

[
N
o
r
m
.

U
n
i
t
]
|
A
|
2

[
N
o
r
m
.

U
n
i
t
]
(a)
-5
0
5
1.0
0.5
0.0
0
0.02
0.04
0.06
0.08
0.1
t [N
o
rm
. U
n
it]
z

[
N
o
r
m
.

U
n
i
t
]
|
B
|
2

[
N
o
r
m
.

U
n
i
t
]
(b)
Figure 1. Evolution of the periodic solution (8) and (9) for the parameters = 0.05, r = 1,
k = 0.9 and c = 1.61.
[26], quasi-phase matched parametric oscillator [27], second harmonic generation [28], and
three-wave solitons [29]. In other systems, symbiotic solitons also occur in phenomena
connected with Bose - Einstein condensates [30], discrete systems [31], multi-dimensional
NLS by separation of variables [32], and quadratic, nonlinear media with loss and gain [33].
More precisely, we shall consider
i
A
z
+
2
A
t
2
+ (AA
+ BB
)A + iA +

2
t
2
A
= 0, (13)
i
B
z
+

2
B
t
2
+ (AA
+ BB
)B + iB +
2
t
2
B = 0. (14)
Here A and B are again complex envelopes but the rst waveguide is permitted to have
a dispersion coefcient (positive or negative) relative to waveguide B. The chirp factors,
however, must be modied to
A = exp
it
2
2
, B = exp
it
2
2
.
A periodic pattern is obtained earlier in the literature as
A = rkQ
1
sn(rte
2z
) exp
2z +
it
2
2

ir
2
e
4z
4
(Q
2
1
+(1 k
2
))
, (15)
B = rQ
2
dn(rte
2z
) exp
2z +
it
2
2

ir
2
e
4z
4
(Q
2
2
k
2
)
, (16)
Q
1
=
2( )
2
1
, Q
2
=
2( 1)
2
1
. (17)
The restrictions are either
> if > 1, (18)
or
< if < 1, (19)
The long wave limit is
A = rQ
1
tanh(rte
2z
) exp
2z +
it
2
2

ir
2
e
4z
Q
2
1
4
, (20)
B = rQ
2
sech(rte
2z
) exp
2z +
it
2
2

ir
2
e
4z
(Q
2
2
1)
4
. (21)
(20) represents a dark soliton, and propagates in the anomalous regime if is positive, while
(21) remains a bright soliton in the anomalous dispersion regime.
The contribution in the present work is to document another set of periodic / solitary
wave pattern which relieves or compensates the constraints imposed by (18) and (19). Fur-
thermore, for some parameter regimes, one can achieve a pair of symbiotic solitons with
bright (dark) solitons propagating in the normal (anomalous) dispersion regimes respec-
tively. Following reasoning similar to the previous sections, we can attain another set of
wave proles by exchanging the choice of elliptic functions in (15) and (16):
A = rR
1
dn[rt exp(2z)] exp
2z +
it
2
2
i
1
, (22)
B = rkR
2
sn[rt exp(2z)] exp
2z +
it
2
2
i
2
, (23)
where the parameters R
1
, R
2
are
R
1
=
2( )
2
1
, R
2
=
2(1 )
2
1
, (24)
and the requirement of real square roots dictates that
>
1
if < 1, (25)
or
<
1
if > 1, (26)
(25) and (26) are different from (18) and (19).
1
,
2
are angular frequency parameters
given by
1
z
= r
2
exp(4z)
(2 k
2
) +
2(1 )
2
1
, (27)
2
z
= r
2
exp(4z)
1 k
2
+
2(1 )
2
1
. (28)
The long wave limit is even more instructive. On taking the limit k 1, where in general
(sn, cn, dn) become (tanh, sech, sech) respectively, one now has
A = rR
1
sech[rt exp(2z)] exp
2z +
it
2
2
i
10
, (29)
B = rR
2
tanh[rt exp(2z)] exp
2z +
it
2
2
i
20
. (30)
where
10
,
10
are the long wave (k 1) limits of
1
and
2
respectively.
For < 1, both waveguides are in the anomalous dispersion regimes (as > 1). For
> 1, can be either positive or negative. In particular, negative values of here will imply
that waveguide A is in normal dispersion regime. Remarkably, a bright (dark) soliton now
propagates in the normal (anomalous) dispersion regime respectively. These phenomena
are quite contrary to the well known results.
4. Conclusions
A class of periodic and solitary waves has been studied for a system of coupled envelope
equations. This systemcan model averaged, dispersion managed systems where the residual
gain / loss in each cycle of the dispersion map has been carefully chosen. Waves with mul-
tiple peaks per period or symbiotic pairs of solitary pulses are obtained analytically. They
will enhance our capability in modeling such systems and strengthen our understanding in
this and similar optical systems.
Several aspects still allow rooms for future work and expansions. In particular, cong-
urations where both waveguides are in the normal dispersion regime have not been worked
out in details yet, although the same physics is expected to hold true qualitatively.
One issue which has not been addressed is the stability of these wave patterns. Numer-
ical simulations of perturbed wave trains must be pursued. Recent works and experience
have indicated that stability will probably still prevail in some parameter regimes. The
precise elucidation will await further efforts.
Acknowledgement
Partial nancial support has been provided by the Research Grants Council through the con-
tract HKU7123/05E. KWCand KNwish to thank The Royal Society for their support in the
form of an International Joint Project Grant. KWC and KN are very grateful to Prof. John
Watson for his valuable support for this research collaboration. KN also wishes to thank the
Nufeld Foundation for nancial support through the Newly Appointed Lecturer Award.
A. Hirota Bilinear Operator
The Hirota operator for any two functions f and g is dened as [34, 35]
D
m
x
D
n
t
g f =

x

x
m

t

t
n
g(x, t)f(x
, t
x=x
,t=t
, (31)
and the properties in association with differentiation of exponential functions are espe-
cially striking (m, n are constants):
D
x
[exp(imx)g exp(inx)f] = [D
x
g f + i(m n)gf] exp[i(m + n)x], (32)
D
2
x
[exp(imx)g exp(inx)f] = [D
2
x
g f + 2i(m n)D
x
g.f (m n)
2
gf]
exp[i(m + n)x]. (33)
Most existing works on the Hirota operator focus on the case of constant wave number or
frequencies. The important point in this work is to extend Hirota derivatives to the case of
time or space dependent wavenumbers.
The bilinear identities for Hirota derivatives, even for the case of variable wave number,
can be obtained from simple, straightforward differentiation. Examples are:
D
z
exp[t
1
(z) +
2
(z)] exp[t
1
(z) +
2
(z)]
= {t[
1
(z)
1
(z)] +
2
(z)
2
(z)} exp{t[
1
(z) +
1
(z)] +
2
(z) +
2
(z)}, (34)
D
z
exp(a) m(z) exp(b) = m
D
z
exp(a) exp(b)
1
m
m
z
exp(a + b)
. (35)
B. Theta Functions
The theta functions
n
(x), n = 1, 2, 3, 4 in terms of the parameter q (the nome) are dened
by [3638]:
1
(x) = 2
n=0
(1)
n
q
(n+1/2)
2
sin[(2n + 1)x] , (36)
2
(x) = 2
n=0
q
(n+1/2)
2
cos [(2n + 1)x] , (37)
3
(x) = 1 + 2
n=1
q
n
2
cos (2nx) , (38)
4
(x) = 1 + 2
n=1
(1)
n
q
n
2
cos (2nx) , 0 < q < 1. (39)
Basically they are Fourier series with exponentially decaying coefcients. Relationships
between the theta and elliptic functions are:
sn(u) =

3
(0)
1
(z)
2
(0)
4
(z)
, cn(u) =

4
(0)
2
(z)
2
(0)
4
(z)
, dn(u) =

4
(0)
3
(z)
3
(0)
4
(z)
, (40)
z =
u
2
3
(0)
, k =

2
2
(0)
2
3
(0)
. (41)
Arguments of the theta and elliptic functions are related by a scale factor. The modulus of
the elliptic functions, k, is related to the theta constants by (41).
Theta functions possess a huge number of identities involving addition and subtraction
of arguments:
3
(x + y)
3
(x y)
2
2
(0) =
2
4
(x)
2
1
(y) +
2
3
(x)
2
2
(y), (42)
4
(x + y)
4
(x y)
2
2
(0) =
2
4
(x)
2
2
(y) +
2
3
(x)
2
1
(y), (43)
Such identities can be proven by re-arranging terms of the multiple sums [37]. By con-
sidering the leading and quadratic terms in the Taylor series of y in identities of the form
(42-43), one obtains
4
(0)
4
(0)

3
(0)
3
(0)
=
4
2
(0),

4
(0)
4
(0)

2
(0)
2
(0)
=
4
3
(0),

3
(0)
3
(0)

2
(0)
2
(0)
=
4
4
(0). (44)
D
2
x
3
(x)
3
(x) =
2
2
(0)
2
3
(x)
2
(0)
+ 2
2
3
(0)
2
4
(0)
2
4
(x), (45)
D
2
x
4
(x)
4
(x) = 2
2
3
(0)
2
4
(0)
2
3
(x) +
2
2
(0)
2
4
(x)
2
(0)
. (46)
Hence formulas for D
x
m

n
, D
2
x
m

n
can be developed for m, n integers using this
line of reasoning.
References
[1] A. Hasegawa and F. Tappert, Transmission of stationary nonlinear optical pulses
in dispersive dielectric bers. I. Anomalous dispersion, Appl. Phys. Lett., 23, 142
(1973).
[2] A. Hasegawa and Y. Kodama, Solitons in Optical Communication, (Oxford University
Press, New York, 1995).
[3] G. P. Agrawal, Nonlinear Fiber Optics, (Academic Press, San Diego, 1989).
[4] V. E. Zakharov and S. Wabnitz, Optical Solitons: Theoretical Challenges and Indus-
trial Perspectives, (Springer-Verlag, Heidelberg, 1998).
[5] N. J. Smith, F. M. Knox, N. J. Doran, K. J. Blow and I. Bennion, Enhanced power
solitons in optical bres with periodic dispersion management, Electron. Lett., 32, 54
(1996).
[6] T. Georges and B. Charbonnier, Reduction of the dispersive wave in periodically
amplied links with initially chirped solitons, IEEE Photon. Technol. Lett., 9, 127
(1997).
[7] M. Suzuki, I. Morita, N. Edagawa, S. Yamamoto, H. Toga and S. Akiba, Reduction
of Gordon-Haus timing jitter by periodic dispersion compensation in soliton transmis-
sion, Electron. Lett., 31, 2027 (1995).
[8] W. Forysiak, F. M. Knox and N. J. Doran, Stepwise dispersion proling of periodi-
cally amplied soliton systems, J. Lightwave Technol., 12, 1330 (1994).
[9] S. Kumar, A. Hasegawa and Y. Kodama, Adiabatic soliton transmission in bers with
lumped amplier: Analysis, J. Opt. Soc. Am. B, 14, 888 (1997).
[10] C. Kurtzke, Suppression of ber nonlinearities by appropriate dispersion manage-
ment, Photon. Technol. Lett., 5, 1250 (1993).
[11] P. V. Mamyshev and L. F. Mollenauer, Pseudo-phase-matched four-wave mixing in
soliton wavelength-division multiplexing transmission, Opt. Lett., 21, 396 (1996).
[12] N. J. Doran, N. J. Smith, W. Forysiak and F. M. Knox, in Physics and Applications of
Optical Solitons in Fibers 95, (Kluwer Academic Press, 1996).
[13] A. Hasegawa, S. Kumar and Y. Kodama, Reduction of collision-induced time jitters
in dispersion-managed soliton transmission systems, Opt. Lett., 21, 39 (1996).
[14] W. Forysiak, K. J. Blow and N. J. Doran, Reduction of Gordon-Haus jitter by post-
transmission dispersion compensation, Electron. Lett., 29, 1225 (1993).
[15] T. I. Lakoba, Non-integrability of equations governing pulse propagation in
dispersion-managed optical bers, Phys. Lett. A, 260, 68 (1999).
[16] P. Tchofo Dinda, A. B. Moubissi and K. Nakkeeran, A collective variable approach
for dispersion-managed solitons, J. Phys. A, 34, L103 (2001).
[17] T. I. Lakoba and D. J. Kaup, Hermite-Gaussian expansion for pulse propagation in
strongly dispersion managed bers, Phys. Rev. E, 58, 1998 (1998).
[18] S. K. Turitsyn, T. Sch afer, K. H. Spatschek and V. K. Mezentsev, Path-averaged
chirped optical soliton in dispersion-managed ber communication lines, Opt. Com-
mun., 163, 122 (1999).
[19] P. Tchofo Dinda, K. Nakkeeran and A. B. Moubissi, Optimized Hermite-gaussian
ansatz functions for dispersion-managed solitons, Opt. Commun., 187, 427 (2001).
[20] S. Kumar and A. Hasegawa, Quasi-soliton propagation in dispersion-managed opti-
cal bers, Opt. Lett., 22, 372 (1997).
[21] C. C. Mak, K. W. Chow and K. Nakkeeran, Soliton Pulse Propagation in Averaged
Dispersion-managed Optical Fiber System, J. Phys. Soc. Japan, 74, 1449 (2005).
[22] V. N. Serkin and A. Hasegawa, Novel Soliton Solutions of the Nonlinear Schr odinger
Equation Model, Phys. Rev. Lett., 85, 4502 (2000).
[23] R. Ganapathy, V. C. Kuriakose and K. Porsezian, Soliton propagation in a birefrin-
gent optical ber with ber loss and frequency chirping, Opt. Commun., 194, 299
(2001).
[24] K. W. Chow and D. W. C. Lai, Periodic solutions for systems of coupled nonlinear
Schrdinger equations with ve and six components, Phys. Rev. E, 65, 026613 (2002).
[25] M. Lisak, A. H o ok and D. Anderson, Symbiotic solitary-wave pairs sustained by
cross-phase modulation in optical bers, J. Opt. Soc. Am. B, 7, 810 (1990).
[26] K. Hayata and M. Koshiba, Bright-kink symbions resulting from the combined effect
of self-trapping and intra-pulse stimulated Raman-scattering, J. Opt. Soc. Am. B, 11,
61 (1994).
[27] A. Picozzi and M. Haelterman, Spontaneous formation of symbiotic solitary waves
in a backward quasi-phase-matched parametric oscillator, Opt. Lett., 23, 1808 (1998).
[28] S. Trillo, Bright and dark simultons in second-harmonic generation, Opt. Lett., 21,
1111 (1996).
[29] C. Durniak, C. Montes and M. Taki, Temporal walk-off for self-structuration of
three-wave solitons in CW-pumped backward optical parametric oscillators, J. Opt.
B: Quantum and Semi-Classical Optics, 6, S241 (2004).
[30] V. M. Perez-Garcia and J. B. Beitia, Symbiotic solitons in heteronuclear multicom-
ponent Bose-Einstein condensates, Phys. Rev. A, 72, 033620 (2005).
[31] E. P. Fitrakis, P. G. Kevrekidis, B. A. Malomed and D. J. Frantzeskakis, Discrete
vector solitons in one-dimensional lattices in photorefractive media, Phys. Rev. E,
74, 026605 (2006).
[32] K. Hayata and M. Koshiba, Bright-dark solitary-wave solutions of a multidimen-
sional nonlinear Schrdinger equation, Phys. Rev. E, 48, 2312 (1993).
[33] S. Darmanyan, L. Crasovan and F. Lederer, Double-hump solitary waves in quadrat-
ically nonlinear media with loss and gain, Phys. Rev. E, 61, 3267 (2000).
[34] M. J. Ablowitz and H. Segur, Solitons and the Inverse Scattering Transform, (SIAM,
Philadelphia, 1981).
[35] Y. Matsuno, The Bilinear Transformation Method, (Academic Press, New York,
1984).
[36] M. Abramowitz and I. Stegun, Handbook of Mathematical Functions, (Dover, New
York, 1964).
[37] D. F. Lawden, Elliptic Functions and Applications, (Springer, New York, 1989).
[38] K. W. Chow, A class of doubly periodic waves for nonlinear evolution equations,
Wave Motion, 35, 7190.
ISBN 1-60021-866-0
Chapter 12
DYNAMICS AND INTERACTIONS OF GAP
SOLITONS IN HOLLOW CORE PHOTONIC
CRYSTAL FIBERS
Javid Atai and D. Royston Neill
School of Electrical and Information Engineering
The University of Sydney, NSW 2006 Australia
Abstract
The existence and stability of gap solitons in a model of hollow core ber in the
zero dispersion regime are analyzed. The model is based on a recently introduced
model where the coupling between the dispersionless core mode and nonlinear surface
mode (in the presence of the third order dispersion) results in a bandgap. It is found
that similar to the anomalous and normal dispersion regimes, the family of solitons lls
up the entire bandgap. The family of gap solitons is found to be formally unstable but
in a part of family the instability is very weak. Consequently, gap solitons belonging
to that part of the family are virtually stable objects. The interactions and collisions
of in-phase and the-out-of-phase quiescent solitons and moving solitons in different
dispersion regimes are investigated and compared.
1. Introduction
Gap solitons (GSs) were originally introduced in Ref. [1]. Recent years have witnessed an
upsurge of research activity on gap solitons in various areas of physics such as nonlinear
optics and Bose-Einstein condensation (BEC). In optics, a nonlinear dispersive medium
whose spectrum contains one or more forbidden bands can support gap solitons. An ex-
ample of such a system is a ber Bragg grating (FBG). The periodic variation of linear
dielectric constant in an FBG leads to a photonic band structure. The linear cross coupling
between the counter-propagating waves results in a large effective dispersion (5 to 6 orders
of magnitude larger than the dispersion of standard optical ber) [2, 3]. For sufciently
high light intensities, the large Bragg grating induced dispersion may be counterbalanced
by Kerr nonlinearity resulting in a gap soliton.
Signicant theoretical [36] and experimental [710] efforts have been directed to-
ward understanding and characterizing GSs in periodic media. In particular, it has been
316 Javid Atai and D. Royston Neill
shown that GSs in an FBG form a two-parameter family of solutions [4]. It has also been
shown that approximately half of the soliton family is stable against oscillatory pertur-
bations [1113]. Experimental activities in this area have focused on generating zero-
velocity (quiescent) GSs due to their potential applications in optical buffers and storage
elements. To date, GSs with a velocity of 0.23 of the speed of light in the ber have been
observed [14]. GSs have been studied in more sophisticated systems such as in the pres-
ence of higher order dispersion [15], quadratic nonlinearity [16], cubic-quintic nonlinear-
ity [18], dual core bers with FBGs [17] and in waveguide arrays [19] and photonic crystal
bers [20].
Since their demonstration in 1996 [21], photonic crystal bers (PCFs) have been the
subject of extensive research due their interesting and peculiar properties. PCFs are spe-
cially designed optical bers with many microstructured air holes running along the bers
length. They can be divided into two main categories depending on the mechanism of light
guidance, namely the solid core and hollow core PCFs. Solid core PCFs are similar to
conventional optical bers in that they guide light through total internal reection. On the
other hand, in hollow core PCFs (HC-PCFs) the microstructured cladding surrounding the
hollow core creates a photonic bandgap that guides the light [2224].
Introduction of atomic or molecular gases into the core of HC-PCF results in efcient
nonlinear optical interactions due to strong connement of light in the core region. Some
recent results include demonstrations of generation of stimulated Raman scattering (SRS)
in hydrogen [25], and electromagnetically-induced transparency (EIT) [26, 27]. They have
also been utilized in delivery of high energy pulses [2830] and in soliton lasers [31].
In Ref. [34] a model for pulse propagation in HC-PCF based on experimental [32] and
numerical [33] results was proposed. The model took into account the coupling of a linear
dispersionless mode propagating in a gas-lled core with a nonlinear dispersive surface
mode propagating in silica. In Ref. [35] a simpler model was considered where the second
and third order group velocity dispersion terms were absent. The model contained a linear
loss term which accounted for the power leakage from the core to the cladding. In both
models a bandgap opens in the systems spectrum. The models in Ref. [34, 35] belong to
a general class of models that give rise to wavenumber bandgap [36, 37]. A wavenumber
bandgap arises as a result of the coupling between the co-propagating waves (in this case
the core and surface modes). On the other hand, a frequency bandgap (e.g. the above-
mentioned bandgap structure in a FBG) is due to the coupling of counterpropgating waves.
The stability of GSs in the model of Ref. [34] has been investigated in both anoma-
lous [38] and normal [39] dispersion regimes. It is shown that, strictly speaking, GSs in
both anomalous and normal dispersion regimes are unstable. However, due to the fact that
instability is weak in a part of the soliton family, the GSs belonging to that part of the family
are virtually stable objects. In addition, an important nding reported in Ref. [39] is that
GSs in the normal dispersion are far more stable than their counterparts in the anomalous
dispersion.
In this article, we will rst investigate the existence and stability of gap solitons in a
HC-PCF in the special case when the second order group velocity dispersion is negligible.
The model, which is based on the model of Ref. [34], and the characteristics of the bandgap
and soliton solutions will be discussed in Section 2.. In Section 3. stability of quiescent and
moving GSs will be presented and their stability will be compared with that of GSs in the
Dynamics and Interactions of Gap Solitons... 317
anomalous and normal dispersion regimes. In Section 4. we will investigate and compare
the interactions of quiescent GSs and collision dynamics of moving solitons in the normal,
anomalous and zero dispersions. In particular we will analyze the effect of initial phase
difference and separation on the outcome of collisions and interactions. The results are
summarized in Section 5..
2. The Model and Gap Soliton Solutions
The system of equations governing the propagation of the above-mentioned surface and
core modes in the zero GVD are based on the model introduced in Ref. [34]. In the normal-
ized form it reads:
iu
z
icu
+ iu
+|u|
2
u + v = 0, (1)
iv
z
+ icv
+ u = 0, (2)
where u and v are the amplitudes of the surface and core waves, respectively, z and are
the propagation distance and reduced time and c represents the group velocity mismatch
between the modes. The coefcients of Kerr nonlinearity and the linear coupling between
the modes have been scaled to unity. Therefore, there are two free parameters in the model
namely and c.
It should be noted that when = 0 Eqs. (1) and (2) reduce to the model of Ref. [35].
However, as was pointed out in Refs. [34,38,39], due to the small temporal width of solitons
and that the carrier wavelength may be close or exactly equal to zero GVD point, the third
order dispersion needs to be present. Undoing the rescalings and using a typical value of
|
3
| = 0.2 ps
3
/km the solitons width is found to be in the range of 100-300 fs. This value
of
3
corresponds to a normalized value of 0.3. Also, z = 1 and = 1 correspond
to ranges 1-10 cm and 30-100 fs in physical units.
In order to determine the linear spectrum of the system, we substitute (u, v)
exp(ikz i) into the linearized Eqs. (1) and (2). This results in the following dispersion
relation:
2k
=
3

(
3
+ 2c)
2
+ 4. (3)
By denition we set c > 0 in which case the wavenumber bandgap exists for > 0.
Straightforward analysis of Eq. (3) shows that the bandgap is 1 < k < +1 provided that
c
2
1
4
. The solid curves in Fig. 1 represent the branches of the dispersion relation (3) for
c = 1 and = 0.3 with the bandgap being 1 < k < 1.
In the gap, soliton solutions to (1) and (2) were sought in the form of
{u(z, ), v(z, )} = {U(), V ()} exp(ikz). Substituting this ansatz into (1) and (2)
results in a set of equations for complex functions U() and V (). These equations can be
solved numerically using the relaxation method. It is found that, similar to the anomalous
and normal dispersion regimes, the family of gap solitons completely ll the bandgap, and
5 2.5 0 2.5 5
10
5
0
5
10
k
Figure 1. Dispersion diagramcorresponding to c = 1 and = 0.3 for quiescent gap solitons
(solid lines) and moving ones with = 0.7 (dashed lines). The bandgap for quiescent
solitons is 1 < k < 1. The bandgap for moving solitons is 0.77 < k < 0.78.
|U()| and |V ()| are always single-humped. As is shown in Fig. 2, the real and imaginary
parts of U() and V () are even and odd functions of , respectively.
The GSs in the model of Eqs. (1) and (2), similar to their counterparts in the anomalous
and normal dispersion regimes [34, 39], satisfy Vakhitov-Kolokolov criterion [40]. This
criterion states that a necessary condition for the stability of solitons against nonoscillatory
perturbations with purely real growth rates is
dE
dk
> 0 where E is the energy of the soliton
family and is given by:
E(k) =
|U(; k|
2
+|V (; k|
2
d. (4)
However, the soliton family or part thereof may be unstable against oscillatory pertur-
bations. A nding of Ref. [39] was that the energy of the soliton family in the normal
dispersion regime is considerably lower than that of the anomalous dispersion regime. As
is shown in Fig. 3, the energy of solitons in (1) and (2) is less than that of the anomalous
case and greater than the normal dispersion case. Based on the results of Ref. [39] one
may conjecture that the solitons in the zero dispersion regime are more stable than their
counterparts in the anomalous dispersion and less so compared with the ones in the normal
dispersion regime. This issue will be considered in the next section.
Moving solitons can be obtained by rewriting Eqs. (1) and (2) in the boosted reference
frame through the coordinate transform (z, ) (z, z) where is the velocity shift.
The dispersion relation of the transformed system of equations is given by:
20 10 0 10 20
2
1
0
1
2
Im(U)
Re(U)
(a)
20 10 0 10 20
2
1
0
1
2
3
4
Im(V)
Re(V)
(b)
Figure 2. The real and imaginary parts of the U() and V () for a quiescent gap soliton
with c = 1 and = 0.3 and k = 0.
2k
= (
3
+ 2)
(
3
+ 2c)
2
+ 4. (5)
The bandgap dened by Eq. (5) varies with . The dashed curves in Fig. 1 display
the branches of Eq. (5) for = 0.7. The bandgap for moving solitons exists in the range
0.9 0.6 0.3 0 0.3 0.6 0.9
k
0
80
160
240
320
E
n
e
r
g
y
Anomalous
Normal
Zero
Figure 3. The total energy of gap solitons with c = 1 and = 0.3 in the anomalous
dispersion (see Ref. [34, 38]), the normal dispersion (see Ref. [39]) and the zero dispersion
regime (Eqs. (1) and (2)).
min
< < c where
min
is negative and can be obtained numerically. It is also found that
the bandgap dened by (5) is completely lled with soliton solutions all of which satisfy
VK criterion. In addition, similar to the case of quiescent solitons, the energy of moving
solitons in this model is found to be greater than that in the normal dispersion and less than
that of moving GSs in the anomalous dispersion regime.
3. Stability of Solitons
In this section we investigate the stability of GSs in this model by means of direct numer-
ical simulations and linear stability analysis. Evolution of GSs were simulated by numeri-
cally solving Eqs. (1) and (2) using the symmetrized split-step Fourier method. Absorbing
boundary conditions were implemented in order to attenuate any radiation that reaches the
boundaries of computational window. To seed any inherent instability in the system, the
GSs found by the above-mentioned relaxation algorithm were initially perturbed asymmet-
rically and then propagated. It is found that, the GSs in the model of (1) and (2), like their
counterparts in the anomalous and normal dispersion regimes, are unstable against oscilla-
tory perturbations. But, in a part of the GS family the instability is weak and as a result
solitons may propagate for long distances before the instability is manifested. As a conse-
quence, the GSs belonging to this part of the family can be considered as being practically
stable
A key result of Ref. [39] was that GSs in the normal dispersion regime are signicantly
more stable than their counterparts in the anomalous dispersion. Moreover, it was con-
jectured that the higher degree of stability of GSs in the normal dispersion regime was, at
least in part, due to the fact that their total energy was considerably smaller than GSs in the
anomalous dispersion. Based on this conjecture and since the total energy of GSs in Eqs.
(1) and (2) is greater (smaller) than those in the normal (anomalous) dispersion (see Fig. 3),
one expects the GSs in the zero dispersion to be more stable than those in the anomalous and
less stable compared to the GSs in the normal dispersion regime. Our simulations corrobo-
rate this prediction. A comparison between the propagation of GSs in different dispersion
regime is provided in Fig. 4.
To quantify the degree of instability of GSs in this model, we have utilized a linear sta-
bility analysis to calculate the instability growth rates for small perturbations. Substituting
the following perturbed soliton solution
{u(z, ) , v (z, )} = {U
() + f() e
z
, V
() + g () e
z
} e
kz
(6)
into the boosted equations (see Section 2.) and linearizing, we arrive at the following eigen-
value problem:
Ay = y (7)
where U
() and V
() are the soliton solutions corresponding to velocity and f () and

g () are the eigenmodes of the small perturbations and is the corresponding complex
eigenvalue. y = [f, f
, g, g
]
T
, and
A =
ik + 2i |U
|
2
+ D
f
iU
2
i 0
iU
2
ik 2i |U
|
2
+ D
f
0 i
i 0 ik + D
g
0
0 i 0 ik + D
g
.
with D
f
= D
f
= (c + )
d
d

d
3
d
3
, and D
g
= D
g
= (c )
d
d
. In the above
expressions, asterisk represents complex conjugate.
The eigenvalue problem posed by Eq. (7) can be solved using standard numerical tech-
niques. The results of the stability analysis are summarized in Fig. 5 as graphs of Re() vs.
k for c = 1, = 0.3 and = 0, 0.25 and 0.5. Since the instability growth rate for all the
cases is positive the GSs are formally unstable. However, one observes that the growth rates
for a part of the family, particularly toward the lower edge of the bandgap, are very small. In
this case, the instability will only be observable after a long propagation distance. Solitons
exhibiting this character are therefore virtually stable objects. We have adopted the de-
nition of Ref. [39] for virtual stability and quasi-stability. That is, for a soliton to be virtually
stable it must remain stable for at least 300Z
nonlin
(where Z
nonlin

1
|U
|
2
) and for it to be
quasi-stable it must remain stable for propagation distances 50Z
nonlin
< z < 300Z
nonlin
.
There are a number of noteworthy features in Fig. 5. Firstly, we note that the growth
rates in the zero dispersion regime are greater than those in the normal dispersion region (c.f.
Fig. 5 in Ref. [39]) and smaller than those in the anomalous dispersion regime (c.f. Figs. 3
and 4 in Ref. [38]). This is consistent with the results of the direct numerical simulations
shown in Fig. 4. Secondly, increasing gives rise to larger growth rates, particularly for
solitons near the upper edge of the bandgap. Nevertheless, varying does not have an
appreciable effect on the border of stable and quasi-stable regions. The weak dependence
of boundary of stable and unstable regions on the velocity of solitons has also been reported
for GSs in a FBG [11, 13].
60 40 20 0 20 40 60
48
(a)
z
0
60 40 20 0 20 40 60
(b)
z
0
250
60 40 20 0 20 40 60
1600
0
(c)
z
Figure 4. Examples of propagation of asymmetrically perturbed quiescent gap soliton corre-
sponding to k = 0.4, c = 1, and = 0.3 in (a) anomalous dispersion, (b) zero dispersion
and (c) normal dispersion. In (c) the initial perturbation causes the soliton to acquire a small
velocity. In this gure and all others below, only the u-component is shown.
0.8 0.4 0 0.4 0.8
k
0
0.1
0.2
0.3
0.4
R
e
(
)
Stable Quasi-Stable
(a)
0.8 0.4 0 0.4 0.8
k
0
0.1
0.2
0.3
0.4
0.5
R
e
(
)
Stable Quasi-Stable
(b)
0.8 0.4 0 0.4 0.8
k
0
0.2
0.4
0.6
0.8
1
R
e
(
)
Stable Quasi-Stable
U
n
s
t
a
b
l
e
(c)
Figure 5. Instability growth rate of GSs in the model of Eqs. (1) and (2) with c = 1 and
= 0.3 versus k for (a) quiescent gap solitons, (b) moving gap solitons with = 0.25 and
(c) moving gap solitons with = 0.5. In the Stable region, the solitons propagate for
long distances i.e. z 300Z
nonlin
without any conspicuous instability development. In the
Quasi-Stable region, instability occurs in the range 50Z
nonlin
< z < 300Z
nonlin
.
4. Interactions and Collisions of Solitons
In view of nonintegrability of the model, the collision dynamics and interactions between
the solitons may be quite complex. In Refs. [38, 39], the collisions between in-phase GSs
in the anomalous and normal dispersion regimes were considered. Moreover, in [39] the
interaction of in-phase and -out-of-phase quiescent GSs in the normal dispersion regime
was investigated and it was shown that in the case of -out-of-phase solitons the outcome
of interaction depends on k and the initial separation of solitons.
In this section we will investigate the interaction of quiescent solitons in the anomalous
and zero dispersion regimes. In addition, the collisions of in-phase and -out-of-phase
moving GSs in anomalous, normal and zero dispersion regimes will also be studied.
4.1. Interactions of Quiescent Solitons in Anomalous and Zero Dispersions
The interaction of GSs in zero and anomalous dispersion regimes was simulated by propa-
gating two identical quiescent solitons belonging to the Stable regions with a time sepa-
ration of and a phase difference of .
It is found that, irrespective of dispersion regime, when = 0 (i.e. GSs are in-phase),
the solitons always repel each other. This behavior was reported for the GSs in the normal
dispersion regime (see [39]) .
In the case of = , the interaction of solitons becomes dependent on and
k.These interactions can be divided into three types, denoted here as Types A, B, and C. In
the Type A interactions the pulses initially attract each other and collide without merging
and then bounce back. An example of this type of interaction in the zero dispersion regime is
shown in Fig. 6(a). In the Type B interactions the pulses attract and temporarily merge and
form a lump which subsequently disintegrates into two separating solitons with different
amplitudes and velocities. Fig. 6(b) displays an example this type of interaction. In the Type
C interactions the pulses repel each other resulting in two separating pulses with different
amplitudes and velocities (Fig. 6(c)).
It should be noted that the velocity and amplitude of resulting solitons in the Type
C interaction as well as the interaction between in-phase solitons depend on the degree of
initial overlap of solitons. If the solitons are initially weakly overlapping then the difference
between the velocities and amplitudes of the eventual moving solitons will be small (see for
example Fig. 7 and Fig 8(b) in Ref. [39]). On the other hand, increasing the initial overlap
between solitons (i.e. reducing ) leads to generation of solitons whose velocities and
amplitudes differ considerably.
Fig. 7 displays the regions in the plane of (,k) where the types A, B and C in-
teractions occur in zero and anomalous dispersion regimes. A noteworthy feature in Fig.
7(a) is that in the zero dispersion the boundary between the types C and A is very weakly
dependent on the initial separation of solitons.
4.2. Collisions of Moving Solitons
In Refs. [38, 39] it was shown that in the anomalous and normal dispersion regimes the
collisions between in-phase counterpropagating solitons belonging to the Stable region
60 40 20 0 20 40 60
50
0
(a)
z
60 40 20 0 20 40 60
50
0
(b)
z
60 40 20 0 20 40 60
50
0
(c)
z
Figure 6. Examples of interaction of quiescent gap solitons in the zero dispersion regime
with c = 1, = 0.3, and = . (a) k = 0.64, = 10; (b) k = 0.64, = 8; (c)
k = 0.9, = 8.
with = 0.5 are always elastic. In particular, It was also found the relative collision
induced loss of energy is 0.1%.
In this section the effect of phase and velocity shift on the collisions will be considered.
First, we consider the collisions between GSs with initial velocities 0.25. As shown in
Fig. 8, in-phase GSs in different dispersion regimes with = 0.25 bounce off each other
elastically. On the other hand, as is displayed in Fig. 9, the -out-of-phase solitons with
= 0.25 collide and merge temporarily and form a lump which then quickly breaks
8 9 10 11 12
0.9
0.8
0.7
0.6
0.5
k
A
A
B
C
(a)
8 9 10 11 12
0.9
0.8
0.7
0.6
k
A
B
C
(b)
Figure 7. Regions of different types of interaction in the plane of (, k) for (a) the zero
dispersion and (b) the anomalous dispersion.
up into two separating solitons. In addition, the collisions do not generate any noticeable
radiation.
Figs. 10 and 11 show that the collisions of in-phase and -out-of-phase solitons with
= 0.5 in different dispersion regimes . In this case, regardless of the initial phase
60 40 20 0 20 40 60
50
0
(a)
z
60 40 20 0 20 40 60
50
0
(b)
z
60 40 20 0 20 40 60
50
0
(c)
z
Figure 8. Examples of collisions of moving gap solitons with k = 0.6, c = 1, =
0.3, = 0.25 and = 0. (a) Anomalous dispersion; (b) zero dispersion; (c) normal
dispersion.
difference, the solitons collide and form a lump which breaks up into two solitons which
travel at almost the same velocity as the initial solitons. The effect of the initial phase
difference is that in the case of = the emerging solitons have different velocity shifts
compared to those with = 0.
5. Conclusion
In this article, we have characterized the gap soliton solutions in a recently introduced model
in the absence of the second order dispersion. Similar to the anomalous and normal disper-
sion regimes, the family of GSs in this case is found to be formally unstable but in a part of
the family the instability is very weak and the solitons belonging to that part of the family
are therefore virtually stable. Interactions of quiescent solitons and collisions of moving
60 40 20 0 20 40 60
50
0
(a)
z
60 40 20 0 20 40 60
50
0
(b)
z
60 40 20 0 20 40 60
50
0
(c)
z
Figure 9. Examples of collisions of moving gap solitons with k = 0.6, c = 1, = 0.3,
= 0.25 and = . (a) Anomalous dispersion; (b) zero dispersion; (c) normal
dispersion.
60 40 20 0 20 40 60
50
0
(a)
z
60 40 20 0 20 40 60
50
0
(b)
z
60 40 20 0 20 40 60
50
0
(c)
z
0.3, = 0.5 and = 0. (a) Anomalous dispersion; (b) zero dispersion; (c) normal
dispersion.
solitons in zero, anomalous and normal dispersion regimes are analyzed. Depending on the
initial separation and the wavenumber, the solitons may either attract and bounce, attract
and merge temporarily and break up into separating solitons, or repel each other. We also
nd that the outcome of the collisions of moving solitons depends on the initial phase and
the velocity shift. In all dispersion regimes, when = 0.25, the in-phase solitons collide
and bounce off each other elastically whereas the -out-of-phase solitons collide and form
a lump which subsequently disintegrates into two separating solitons. On the other hand,
60 40 20 0 20 40 60
50
0
(a)
z
60 40 20 0 20 40 60
50
0
(b)
z
60 40 20 0 20 40 60
50
0
(c)
z
0.3, = 0.5 and = . (a) Anomalous dispersion; (b) zero dispersion; (c) normal
dispersion.
when = 0.5, the solitons always collide elastically and two separating solitons emerge. In
this case, the velocity shifts of the emerging solitons depend on the initial phase difference.
References
[1] W. Chen, D.L. Mills, Gap solitons and the nonlinear optical response of superlat-
tices, Phys. Rev. Lett. 58, 160-163 (1987).
[2] P.St.J. Russel, Bloch wave analysis of dispersion and pulse propagation in pure dis-
tributed feedback structures, J. Mod. Opt. 38, 1599-1619 (1991).
[3] C.M. de Sterke, J.E. Sipe, Gap solitons, Prog. Opt. 33, 203-260 (1994).
[4] A.B. Aceves, S. Wabnitz, Self-induced transparency solitons in nonlinear refractive
periodic media, Phys. Lett. A 141, 37-42 (1989).
[5] J.E. Sipe, H.G. Winful, Nonlinear Schr
1
2
inger soliton in a periodic structure, Opt.
Lett. 13, 132-133 (1988).
[6] D.N. Christadoulides, R.I. Joseph, Slow Bragg Solitons in Nonlinear Periodic Struc-
tures, Phys. Rev. Lett. 62, 1746-1749(1989).
[7] B.J. Eggleton, R.E. Slusher, C.M. de Sterke, P.A. Krug, J.E. Sipe, Bragg Grating
Solitons, Phys. Rev. Lett. 76, 1627-1630 (1996).
[8] B.J. Eggleton, C.M. de Sterke, R.E. Slusher, Nonlinear pulse propagation in Bragg
gratings, J. Opt. Soc. Am. B 14, 2980-2993 (1997).
[9] C.M. de Sterke, B.J. Eggleton, P.A. Krug, High-intensity pulse propagation in uni-
form gratings and grating superstructures, J. Lightwave Technol. 15, 1494-1502
(1997).
[10] D. Taverner, N.G.R. Broderick, D.T. Richardson, R.I. Laming, M. Ibsen, Nonlinear
self-switching and multiple gap-soliton formation in a ber Bragg grating, Opt. Lett.
23, 328-330 (1998).
[11] B.A. Malomed, R.S. Tasgal, Vibration modes of a gap soliton in a nonlinear optical
medium, Phys. Rev. E 49, 5787-5796 (1994).
[12] A. De Rossi, C. Conti, S. Trillo, Stability, Multistabiity, and Wobbling of Optical
Gap Solitons, Phys. Rev. Lett. 81, 85-88 (1998).
[13] I.V. Barashenkov, D.E. Pelinovski, E.V. Zemlyanaya, Vibrations and oscillatory in-
stabilities of gap solitons, Phys. Rev. Lett. 80, 5117-5120 (1998).
[14] J.T. Mok, C.M. de Sterke, I.C.M. Littler, B.J. Eggleton, Dispersionless slow light
using gap solitons, Nature Phys. 2, 775-780 (2006).
[15] A.R. Champneys, B. A. Malomed and M.J. Friedman, Thirring Solitons in the Pres-
ence of Dispersion, Phys. Rev. Lett. 80, 4169-4172(1998).
[16] W.C.K. Mak, B.A. Malomed and P.L. Chu, Three-wave gap solitons in waveguides
with quadratic nonlinearity, Phys. Rev. E 58, 6708-6722(1998).
[17] W.C.K. Mak, P.L. Chu and B.A. Malomed, Solitary waves in coupled nonlinear
waveguides with Bragg gratings, J. Opt. Soc. Am. B 15, 1685-1692 (1998); J. Atai
and B. A. Malomed, Bragg-grating solitons in a semilinear dual-core system, Phys.
Rev. E 62, 8713-8718 (2000); J. Atai and B.A. Malomed, Solitary waves in systems
with separated Bragg grating and nonlinearity , Phys. Rev. E 64, 066617 [5 pages]
(2001).
[18] J. Atai and B.A. Malomed, Families of Bragg-grating solitons in a cubic-quintic
medium, Phys. Lett. A 284, 247-252 (2001); J. Atai, Interaction of Bragg grating
solitons in a cubic-quintic medium, J. Opt. B: Quant & Semiclass. Opt. 6, S177-
S181 (2004) .
[19] D. Mandelik, H.S. Eisenberg, Y. Silberberg, R. Morandotti and J.S. Aitchison, Band-
Gap Structure of Waveguide Arrays and Excitation of Floquet-Bloch Solitons, Phys.
Rev. Lett. 90, 053902 (2003); D. Mandelik, H.S. Eisenberg, Y. Silberberg, R. Moran-
dotti and J.S. Aitchison, Gap Solitons in Waveguide Arrays, Phys. Rev. Lett. 92,
093904 (2003).
[20] A. Ferrando, M. Zacares, P. Fernandez de Cordoba, D. Binosi and J. Monsoriu, Spa-
tial soliton formation in photonic crystal bers, Opt. Express 11, 452-459 (2003); D.
Neshev, A.A. Sukhorukov, B. Hanna, W. Krolokowski and Y.S. Kisvshar, Controlled
Generation and Steering of Spatial Gap Solitons, Phys. Rev. Lett. 92, 093904 (2004).
[21] J.C. Knight, T.A. Birks, P.St.J. Russell, D.M. Atkin, All-silica single-mode ber with
photonic crystal cladding, Opt. Lett. 21, 1547-1549 (1996).
[22] T.A. Birks, P.J. Roberts, P.St. J. Russell, D.M. Atkin, T.J. Shepherd, Full 2-D pho-
tonic bandgaps in silica/air structures, Electron. Lett. 31, 1941-1943 (1995).
[23] R.F. Cregan, B.J. Mangan, J.C. Knight, T.A. Birks, P.St. J. Russell, P.J. Roberts, D.C.
Allan, Single mode photonic band gap guidance of light in air, Science 285, 1537-
1539 (1999).
[24] P.St. J. Russell, Photonic crystal bers, Science 299, 358-362.
[25] F. Benabid, G. Bouwmans, J.C. Knight, P.St.J. Russell, F. Couny, Ultra-high ef-
ciency laser wavelength conversion in gas-lled hollow core photonic crystal ber by
pure stimulated Raman scattering in molecular hydrogen, Phys. Rev. Lett. 93, 123903
(2004).
[26] S. Ghosh, J. Sharping, D.G. Ouzounov, A.L. Gaeta, Resonant optical interactions
with molecules conned in photonic bandgap bers, Phys. Rev. Lett. 94, 093902
(2005).
[27] F. Benabid, P.S. Light, F. Couny, P.St.J. Russell, Electromagnetically-induced trans-
parency grid in acetylene-lled hollow-core PCF, Opt. Express 13, 5694 (2005).
[28] D.G. Ouzounov, F.R. Ahmad, D. Mller, N. Venkataraman, M.T. Gallagher, M.G.
Thomas, J. Silcox, K.W. Koch and A.L. Gaeta, Generation of Megawatt optical soli-
tons in hollow-core photonic band-gap bers, Science 301, 1702-1704 (2003);
[29] F. Luan, J.C. Knight, P.St.J. Russell, S. Campbell, D. Xiao, D.T. Ried, B.J. Mangan,
D.P. Williams, P.J. Robert, Femtosecond soliton pulse delivery at 800 nm wavelength
in hollow-core photonic bandgap bers Opt. Express 12, 835-840 (2004).
[30] W. Gobel, A. Nimmerjahn, and F. Helmchen, Distortion free delivery of nanojoule
femtosecond pulses from Ti:sapphire laser through a hollow-core photonic crystal
ber Opt. Lett. 29, 1285-1287 (2004).
[31] H. Lim and F.W. Wise, Control of dispersion in a femtosecond ytterbium laser by use
of hollow-core photonic bandgap ber, Opt. Express 12, 2231-2235 (2004).
[32] C.M. Smith, N. Venkataraman, M.T. Gallagher, D. Mller, J.A. West, N.F. Borrelli,
D.C. Allan and K.W. Koch, Low-loss hollow-core silica/air photonic bandgap bre,
Nature 424, 657-659 (2003).
[33] K. Saitoh, N. A. Mortensen, M. Koshiba, Air-core photonic band-gap bers: the
impact of surface modes, Opt. Express 12 (2004) 394.
[34] I.M. Merhasin and B.A. Malomed, Gap solitons in a model of a hollow optical ber,
Opt. Lett. 30, 1105-1107 (2005).
[35] D.V. Skryabin, Coupled core-surface solitons in photonic crystal bers, Opt. Ex-
press 12, 4841-4846 (2004).
[36] S. Wabnitz, Forward mode coupling in periodic nonlinear-optical bers: modal dis-
persion cancellation and resonance solitons, Opt. Lett. 14, 1071-1073 (1989).
[37] G. van Simaeys, S. Coen, M. Haelterman and S. Trillo, Observation of Resonance
Soliton Trapping due to a Photoinduced Gap in Wave Number, Phys. Rev. Lett. 92,
223902 (2004).
[38] J. Atai, B.A. Malomed, I.M. Merhasin, Stability and collisions of gap solitons in a
model of a hollow optical ber, Opt. Comm. 265, 342-348 (2006).
[39] D.R. Neill, J. Atai, Gap solitons in a hollow optical ber in the normal dispersion
regime, Phys. Lett. A (in press).
[40] M.G. Vakhitov, A.A. Kolokolov, Stability of stationary solutions of nonlinear wave
equations, Radiophys. Quantum Electron. 16, 783 (1973).
Chapter 13
MULTIWAVELENGTH OPTICAL FIBER LASERS
AND SEMICONDUCTOR OPTICAL AMPLIFIER
RING LASERS
Byoungho Lee
*
and Ilyong Yoon
School of Electrical Engineering, Seoul National University
Gwanak-Gu Sinlim-Dong, Seoul 151-744, Korea
Abstract
We review various schemes for multiwavelength fiber lasers and semiconductor optical
amplifier (SOA) ring lasers. Multiwavelength fiber lasers have applications in wavelength
division multiplexing (WDM) optical communication systems, optical fiber sensors and
optical spectroscopy. Erbium-doped fiber amplifiers (EDFAs), Raman amplifiers and SOAs
are mainly used as gain media for multiwavelength fiber lasers.
Because EDFAs are homogeneously broadened gain media, various methods have been
researched to enable the multiwavelength generation. Due to the introduction of liquid
nitrogen cooling, four-wave mixing, frequency shifted feedback, and so on, multiwavelength
erbium-doped fiber lasers could become realized.
On the other hand, because SOA and Raman amplifiers are gain media with
inhomogeneous broadening, multiwavelength generation is relatively easy. The useful
features of the multiwavelength lasers are mainly dependent on a comb filter. One of the most
important features of multiwavelength lasers is tunability. The tunability of wavelengths and
channel spacing is required for WDM optical communication systems. Much research has
been conducted to enable implementation of tunable multiwavelength fiber lasers. Various
comb filters such as Fabry-Perot filters, fiber Bragg gratings, and polarization-maintaining
fiber loop mirrors can be used for multiwavelength fiber lasers. We review several schemes
for multiwavelength SOA-fiber and Raman fiber lasers in this chapter.

*
E-mail address: byoungho@snu.ac.kr. Tel: +82-2-880-7245, Fax: +82-2-873-9953
Byoungho Lee and Ilyong Yoon 336
1. Introduction
The realization of the laser has made many applications possible. Among those applications, a
light source for optical communication systems is one of the most important applications. As
wavelength-division-multiplexing (WDM) optical communication systems have become
more developed, multiwavelength light sources have also been widely researched. In the first
stage, multiwavelength lasers could be made as a simple structure consisting of the array of
lasers and a multiplexer [1, 2]. However, there have been difficulties with these lasers such as
large insertion loss and bulky size. Therefore, multiwavelength fiber lasers using a single gain
medium are desired. There are many possible gain media for optical communication such as
erbium-doped fiber (EDF), semiconductor optical amplifier (SOA), and stimulated Raman
scattering (SRS). In this chapter we review a wide variety of multiwavelength fiber lasers
employing a single gain medium.
Because EDF is a homogeneously broadened gain medium, a laser using EDF normally
lases at a single wavelength. Various methods have been researched to enable the multiple
wavelength generation, such as the introduction of liquid nitrogen cooling, four-wave mixing,
frequency shifted feedback, and so on. For the multiwavelength EDF laser (EDFL), schemes
to suppress mode competition are a main subject.
On the other hand, Raman amplifiers and SOAs are inhomogeneously broadened gain
media. Therefore, multiwavelength generation is relatively simple compared with an EDFL.
Many methods have also been proposed to implement multiwavelength lasers using these
technologies. One of the useful characteristics of multiwavelength fiber lasers is tuning
capability. The tunability of lasing wavelengths and channel spacing is required for WDM
optical communication systems. Therefore, much research has been conducted for the
implementation of tunable multiwavelength fiber lasers. For multiwavelength SOA-fiber and
Raman fiber lasers, the schemes for tuning and these lasers characteristics are main subjects.
We classify and review many schemes for the design of tunable fiber lasers in this chapter.
2. Multiwavelength Fiber Lasers Using EDFA
The most challenging difficulty of an EDF amplifier (EDFA) for a multiwavelength laser is
that the EDFA is a homogeneously broadened medium. In a homogeneously broadened
medium, all atoms in the excited state have the same gain spectrum. Therefore, when a laser
employs a homogeneously broadened gain medium, only the wavelength which has the
largest net gain (gain minus cavity loss) can survive. The other wavelengths decay due to
loss. When a number of wavelengths are in a cavity, each channel experiences mode
competition. However, because Er
3+
ions are surrounded by a glass host, the interaction with
the silica and other dopants leads to some degree of inhomogeneous broadening contribution.
Therefore, the schemes for the multiwavelength EDFL involve increasing of the
competitiveness of weak wavelengths or decreasing of the homogeneously broadened
linewidth of the EDFA.
Multiwavelength Optical Fiber Lasers and Semiconductor Optical Amplifier 337
2.1. Cavity Loss Balancing
The first multiwavelength operation of an EDFL was demonstrated in 1992 [3]. The cavity
losses of the lasing wavelengths are carefully controlled to suppress single channel lasing as
shown in Fig. 1. The cavities of lasing wavelengths are separated and the losses of cavities are
controlled independently so that many wavelengths can lase. This is equivalent to the net gain
flattening. There is no dominant wavelength due to the flattened net gain. However, this
method requires a careful control of cavity losses. Thus it can be easily expected that lasers
employing this scheme are relatively unstable and sensitive to environmental conditions.
Gain
medium
W
D
M
PC
FLM
FLM
1
,
2
, ,
8
8
(a)
W
D
M
8
W
D
M
Gain
medium
Polarizer
Isolator
1
,
2
, ,
8
(b)
Figure 1. (a) An eight-channel laser configuration based on a linear cavity. (b) An eight-channel laser
configuration based on a ring cavity (PC: polarization controller, FLM: fiber loop mirror, WDM:
wavelength division multiplexer) [3].
2.2. Liquid Nitrogen Cooling
When an EDFA is cooled, the homogeneous linewidth of the EDFA is narrowed. Spectral
hole burning and homogeneous linewidth were measured as a function of temperature in Ref.
[4]. The homogeneous linewidth was measured as 1.3 nm at 61 K. It was shown that the
homogeneous linewidth exceeded 11.5 nm at room temperature.
For a multiwavelength application of an EDFA, there was other research to make an
inhomogeneously broadened EDFA by liquid nitrogen cooling [5]. In Ref. [5], the main idea
was the suppression of dynamic crosstalk between adjacent channels. The liquid nitrogen
cooling made an 11 dB suppression of crosstalk. The channel spacing was 4 nm and the
homogeneous linewidth was measured as ~1 nm.
Comb filter
Isolator
Doped fiber
77K
75:25 coupler
Output
WDM coupler
Pump
Figure 2. A multiwavelength EDF ring laser configuration using a comb filter in the cavity [6].
The first multiwavelength EDFL by liquid nitrogen cooling (77 K) was presented in 1996
[6]. Figure 2 shows the configuration. 11 stable laser lines were demonstrated with 0.65 nm
channel spacing around 1535 nm. Two types of comb filters were used in the experiment.
Those were a chirped fiber Bragg grating (CFBG) Fabry-Perot filter and a sampled grating.
2.3. Four-Wave Mixing
The self-stabilizing effect of four-wave mixing (FWM) can be used for a multiwavelength
EDFL. The powers of lasing wavelengths are automatically balanced by several degenerated
FWMs,
1 2 3
2 = + , or nondegenerated FWMs,
1 2 3 4
+ = + . For a phase
matching condition, dispersion-shifted fiber (DSF) or photonic crystal fiber (PCF) are
required. The self-stabilizing effect may be described as a photonic Robin Hood. This
means that FWM takes the energy of a rich wavelength and gives it to a poor wavelength.
Because power is transferred between wavelengths by the nonlinear process, this scheme can
be thought of as automatic net gain equalization.
PC
1
PC
2
VOA
1
VOA
3
VOA
4
VOA
2
FBG
1
FBG
3
FBG
2
FBG
4
3-dB coupler
3
-
d
B

c
o
u
p
l
e
r
HN-PCF PC
3
Output
980nm Pump
laser diode
WDM coupler
EDF
Output
F
u
s
e
d

c
o
u
p
l
e
r
1
F
u
s
e
d

c
o
u
p
l
e
r
2
Figure 3. An experimental setup for four-wavelength EDFL (HN-PCF: highly nonlinear photonic
crystal fiber, VOA: variable optical attenuator, FBG: fiber Bragg grating, PC: polarization controller)
[7].
Liu and Lu demonstrated a four-wavelength EDFL using a highly nonlinear PCF to
suppress the mode competition at room temperature as shown in Fig. 3 [7]. Experimental
results showed lasing wavelengths of 1540.28, 1543.58, 1546.79 and 1550.08 nm.
EDFA
Output
10 90
AWG(1xn)
PC
FBG
1 FBG
n
1

2

n
Isolator
DSF
PC
Figure 4. A schematic of the multiwavelength EDFL based on degenerate four-wave mixing in the DSF
(EDFA: erbium-doped fiber amplifier, FBG: fiber Bragg grating, AWG: arrayed waveguide grating,
PC: polarization controller) [8].
Han et al. also presented similar multiwavelength EDFLs by using a DSF [8, 9]. Figure 4
shows a schematic diagram of the multiwavelength EDFL employing multiple fiber Bragg
gratings (FBGs) and a 1 km DSF for 10 channels lasing with 0.8 nm channel spacing. In
addition, they showed channel spacing tunability through the elimination of the effects of
several FBGs.
Output
10 90
Isolator
DSF
PC
1
980nm pump
laser diode
EDF
PC
3
PC
4
PC
2
PMF
1
L
1
PMF
2
L
2
Lyot-Sagnac filter
Figure 5. Schematics of the multiwavelength EDFL using DSF and Lyot-Sagnac filter (PMF:
polarization-maintaining fiber, EDF: erbium-doped fiber, PC: polarization controller) [9].
A tunable multiwavelength EDFL was demonstrated in 2005 [9]. The tunability
originated from a tunable Lyot-Sagnac filter as shown in Fig. 5. The wavelength spacing of
the two-segment Lyot-Sagnac filter was
[ ]
2
1 2
/ ( ) n L L = , where n was the
effective birefringence between two orthogonal polarization modes and
1 2
, L L were the
lengths of the two polarization-maintaining fibers (PMFs) shown in Fig. 5. Thus, the channel
spacing was switchable by polarization control. In the Lyot-Sagnac filter, clockwise and
counterclockwise lights experienced optical path difference due to PMF segments. Therefore,
the optical path difference between two lights led to comb-like filter characteristics. The
birefringence of the two PMF segments may be summed or subtracted depending on the state
of the polarization controllers (PCs).
Experimental results showed 11 laser lines with 1 nm spacing and 17 laser lines with 0.8
nm spacing. Stable lasing characteristics and tuning capability were obtained due to FWM.
2.4. Frequency Shifting Technique
Another scheme for a multiwavelength EDFL is a frequency shifted feedback scheme [10].
Figure 6 is a schematic diagram of the multiwavelength EDFL. In the scheme, an acousto-
optic modulator (frequency shifter) shifts the frequency of light by 100 MHz for each round
trip. This prevents single frequency lasing. Experimental results showed stable ~13 laser lines
with 0.8 nm spacing. A Fabry-Perot etalon with a CFBG or sampled grating was used for the
periodic filter. The experimental results showed good agreement with the simulation results.
EDFA
1
EDFA
2
3-dB coupler
Isolator
Output
Frequency
shifter
Periodic
filter
Figure 6. A schematic diagram of the multiwavelength EDFL employing a frequency shifted feedback
scheme (EDFA: erbium-doped fiber amplifier) [10].
Table 1. Multiwavelength EDFLs
Year
First author
[Reference]
Comments
Channel
number
Channel
spacing
(nm)
Scheme
1992 N. Park [3] The first multiwavelength EDFL 6 4.8
Cavity loss
balancing
1996 J. Chow [6] 11 0.65
Liquid nitrogen
cooling
2000 A. Bellemare [10] ~13 0.8 Frequency shifting
2002 R. Slavik [11] High uniformity 18 0.8 Frequency shifting
2005 X. Liu [7] 4 3.3 Four-wave mixing
2005 Y.-G. Han [9] Channel spacing switching 17, 11 0.8, 1 Four-wave mixing
2006 Y.-G. Han [8] Channel spacing switching 10 0.8, 1.6, 2.4 Four-wave mixing
For the frequency shifting technique, a more in-depth study was published in 2002 [11].
These researchers improved the uniformity of the lasing wavelengths in the EDFL. Uniform
18 laser lines with 0.8 nm channel spacing were obtained in the experiments.
Similar to the frequency shifted feedback scheme, the phase-modulation feedback
scheme was also presented [12, 13]. A LiNbO
3
phase modulator was used for phase
modulation. In Ref. [12], the sawtoothed and sinusoidal phase modulation of a few tens of
kHz generated a multiwavelength operation. In Ref. [13], the authors reported that sinusoidal,
sawtoothed, triangular and square waveforms are all suitable for multiwavelength lasing.
They also indicated that the phase modulation of 500 Hz to a few tens of kHz is good.
Important features of the above multiwavelength EDFLs are shown in Table 1 as a
summary.
3. Multiwavelength Fiber-SOA and Fiber-Raman Lasers
In a SOA, the gain medium is a semiconductor and not a single atom or ion. The
recombination of electron-hole pairs makes spontaneous or stimulated emission. The SOA is
electrically pumped. More electrons in the conduction band and more holes in the valance
band lead to higher gain. The gain spectrum of the SOA depends on materials and structure.
The intrinsic inhomogeneous broadening is an advantage of the SOA in its application to
multiwavelength lasers. High gain per unit length and compact size are other advantages. On
the other hand, the rectangular structure leads to a coupling loss for optical fiber and
polarization-dependent gain. The fast carrier lifetime (~200 ps) leads to cross saturation and
stronger nonlinear processes.
SRS is an interaction between photon energy and molecular vibrational energy (optical
phonon). The amplification is performed by the energy transfer from a pump beam to the
signal beam (or light to lase). Unlike that of the EDFA and the SOA, Raman scattering does
not require a population inversion for amplification. Very broad gain bandwidth is the main
characteristic of the Raman amplification process. In the SRS, specific resonant frequency
does not exist in contrast to the EDFA and the SOA. The wavelength of the pump beam
determines the location of the gain spectrum which has a peak at 13.2 THz off the pump
wavelength. It is a main advantage of a Raman amplifier that a specific gain medium is not
required, i.e., amplification occurs in a common optical fiber. Therefore, lumped or
distributed schemes are all possible. If several pump wavelengths are used properly, a flat
gain over a wide bandwidth can be obtained [14]. Because Raman scattering is a weak effect,
the SRS requires very high pump power (typically a few Watts) and a long length of fiber.
Intrinsic inhomogeneous linewidth broadening is very attractive for a multiwavelength laser.
In the SOA and Raman fiber lasers, multiwavelength generation is relatively easy
because of their inhomogeneous broadening. Therefore, tunable capability has been a main
subject of research involving these multiwavelength SOA and Raman fiber lasers. Tunability
was even considered in the demonstration of the first multiwavelength Raman fiber laser [15,
16]. Thus, in this section, we focus on the tunable capability of SOAs and Raman fiber lasers.
To avoid confusion, we use two different terminologies: switchability and tunability.
Switchability and tunability denote discrete tuning and continuous tuning, respectively. Thus
a wavelength switchable laser means a laser which can shift the spectral position of lasing
wavelengths by some discrete steps.
3.1. Wavelength or Channel Spacing Switchability
By using a sampled high-birefringence (Hi-Bi) fiber grating as a switchable comb filter, the
wavelength switchable laser was demonstrated by Yu et al. as shown in Fig. 7 [17]. We can
think of this as if two different sampled FBGs (SFBGs) are used due to the difference of the
refractive indexes in the fast and slow axes of the Hi-Bi fiber. The control of a rotatable
polarizer is equivalent to the selection of one of two SFBGs. In a SFBG, the center Bragg
wavelength is 2
B eff
n = and the wavelength separation is
2
/ 2
B eff
n p = , where
eff
n
is the effective refractive index of the fiber core, the individual grating pitch, and p is the
sampling period. Therefore, while is maintained at nearly the same value, it is possible
to move only the center Bragg wavelength. Because the birefringence n is of the order of
10
-4
, the channel spacing is hardly influenced by the choice of polarization axis. However, if
we control the polarization of light incident on the sampled grating by using the rotatable
polarizer, transmission peaks can be shifted by 2 n . The experimental result showed an
interleaving characteristic. The laser output was shifted by 0.4 nm with the 0.8 nm channel
spacing fixed. It had a disadvantage in that the number of switchable wavelength set was
intrinsically limited to two. The amount of switchable wavelengths was determined by the
choice of PMF. Thus, the maximum switchable range was limited by the birefringence of the
PMF.
PC
SOA
Isolator
Isolator
Polarizer
Output
Variable
coupler
SMF
Sampled Hi-Bi
fiber grating
Figure 7. A schematic diagram of a wavelength switchable SOA-fiber ring laser employing sampled Hi-
Bi FBG (SMF: single mode fiber) [17]
Lee et al. presented wavelength switchable SOA fiber lasers employing two SFBGs [18]
and a reflection type interleaver [19]. The former used two SFBGs connected to a polarization
beam splitter (PBS) for waveband switching as shown in Fig. 8. A rotatable linear polarizer
selected one of the two SFBGs. Contrary to the work by Yu et al. [17], the amount of
waveband switching depended on the design of the SFBGs. In other words, the amount of
wavelength shift was not limited by the choice of a PMF. The experimental result in Fig. 9
showed that 5 laser lines with 0.8 nm spacing could be switched by a spectral displacement of
10 nm.
SOA
Isolator
PC
75: 25
coupler
Output
Rotatable liner
polarizer
Light
absorber
Sampled fiber
Bragg gratting
1
Sampled fiber
Bragg gratting
2
PBS
Figure 8. A schematic diagram of the waveband-switchable SOA-fiber laser using two SFBGs (PBS:
polarization beam splitter, PC: polarization controller) [18].
1535 1540 1545 1550 1555 1560 1565
-40
-30
-20
-10
0
Wavelength [nm]
O
p
t
i
c
a
l

p
o
w
e
r

[
d
B
m
]
Figure 9. Output optical spectra showing the waveband switching operation.
SOA
Isolator
Isolator
75: 25
coupler
Output
50:50
coupler
PC
1
PC
2
PMF
1
2
Figure 10. A schematic diagram of a SOA-fiber laser employing a reflective type interleaver (PMF:
polarization maintaining fiber, PC: polarization controller) [19].
L1 L2
50:50 coupler
/4 /4
/2
/2 /2
/2
a
b
c
Tunable filter
WDM
2
WDM
1
Raman fiber
Pump laser
Output
Figure 11. An experimental setup for a tunable Raman fiber ring laser (WDM: wavelength division
multiplexer) [21].
Another scheme using a reflective interleaver is shown in Fig. 10 [19]. The interleaver is
composed of a PBS and a PMF loop mirror. A PC in the PMF loop mirror consists of two
quarter-wave plates. The control of waveplates leads to an interleaving characteristic. The
feature of this filter is that the transmission and reflection characteristics show interleaved
sets of multiple wavelength peaks. Theoretically, for transmission, the filter has infinite
channel isolation and 3 dB insertion loss. On the other hand, for reflection, the filter shows 3
dB channel isolation and 0 insertion loss. 17 wavelengths were generated with 0.8 nm channel
spacing. The laser lines could be shifted by 0.4 nm with channel spacing fixed.
The PMF Lyot-Sagnac filter was also used for a multiwavelength SOA-fiber laser [20].
With the PMF Lyot-Sagnac filter, the SOA-fiber laser could have channel spacing
switchability. In addition, the rotation of a quarter-wave plate made the lasing wavelength
shift. Channel spacing switchability from 0.8 nm to 4.1 nm was demonstrated. Laser lines
from 5 to 20 were observed. Continuous wavelength tuning was also shown.
There have also been intensive research efforts for tunable multiwavelength Raman fiber
lasers. Kim et al. demonstrated a multiwavelength Raman fiber ring laser with switchable
channel spacing and a tunable lasing wavelength [21]. The multiwavelength source was
composed of a Raman fiber and a Lyot-Sagnac filter as shown in Fig. 11. The experimental
results showed a multiwavelength generation of up to 20 laser lines with 0.43 nm spacing.
PMF
2
PMF
1
Lyot-Sagnac
filter
PC
2
(/2)
PC
1
(/2)
Fiber
grating
97%
Raman gain fiber
Fiber
grating
90%
Pump combiner
Output
WDM
coupler
Pump
Pump laser
Figure 12. An experimental setup for a tunable multiwavelength Raman laser based on an FBG cavity
incorporating PMF Lyot-Sagnac filter (PC: polarization controller, WDM: wavelength division
multiplexer) [22].
Han et al. demonstrated a multiwavelength Raman laser with a similar filter as shown in
Fig. 12 [22]. Although a similar PMF Lyot-Sagnac filter was used, there were two
differences: a linear cavity structure employing FBGs and the use of PMFs with different
birefringences. In the experimental result, the multiwavelength laser generated 7 channels
with 0.6 nm spacing and 5 channels with 0.8 nm spacing.
A phase modulator loop mirror filter (PM-LMF) could be used for wavelength
switchability [23]. The PM-LMF is a sort of PMF loop mirror where a phase modulator is
inserted. Because DC bias, RF power, or modulation frequency changes the birefringence of
the phase modulator slightly, the spectral comb position can be shifted while the channel
spacing is fixed. Experimentally, 21 laser lines with a 0.8 nm channel spacing were obtained.
Optical amplifier
10%
Output
Coupler
Programmable
Hi-Bi FLM
Output Input
Residual pump power
DCF
6.9 km
Pump
laser
WSC
Isolator
In
Out
Combiner
3 dB
coupler
PC
1
PC
2
PC
n
Hi-Bi, L
1
Hi-Bi, L
2
Hi-Bi, L
n
22 switch
Figure 13. A schematic of a multiwavelength laser and a programmable Hi-Bi fiber loop mirror (FLM)
(DCF: dispersion compensation fiber, WSC: wavelength selection coupler) [24].
Chen demonstrated channel spacing switchable fiber lasers by using a programmable Hi-
Bi fiber loop mirror as shown in Fig. 13 [24]. He employed a SOA or a Raman amplifier as
an optical amplifier. Although many switchable sections are theoretically possible in Fig. 13,
the two sections of the PMF were demonstrated. The use of 22 switches changes the
combination of the PMF section more flexibly. The channel spacing expression is essentially
equivalent to that of the PMF Lyot-Sagnac filter except that it has a more flexible
birefringence combination. The experimental result showed 3.2 nm and 1.6 nm channel
spacing switching.
A tunable Raman fiber laser employing an electro-optical tuning scheme was presented in
2004 [25]. The comb filter uses an electro-optic polarization controller (EOPC) inserted in the
PMF Sagnac loop filter. The PMF Sagnac loop filter is sometimes called a Lyot-Sagnac filter.
Due to the PMF Sagnac loop filter, the channel spacing was switchable. When a driving
voltage was applied to the EOPC, the additional birefringence was induced. This effect is
equivalent to changing the length of the PMF slightly. Therefore, lasing wavelengths can be
shifted as channel spacing is nearly fixed. This Raman laser had both channel spacing
switchability and wavelength tunability. Experimental results indicated channel spacing
switchability between 0.95 and 2.95 nm. Interleaved switching operation of 11 laser lines was
also demonstrated with 0.88 nm channel spacing.
3.2. Wavelength Tunability
In a SOA fiber laser, wavelength tuning possibility was presented in 2001 [26]. In fact, the
work in Ref. [26] was about a multiplexed sensor. The sensor was a SOA ring laser using a
transmission-type filter consisting of a circulator and multiple FBGs. Eight laser lines were
observed when 8 FBGs were used. The center wavelengths of the FBGs were 1534.44,
1543.68, 1546.32, 1549.38, 1552.5, 1554.06, 1556.28 and 1558.92 nm. The strain on each
FBG changed each lasing wavelength. Therefore, the sensor was indeed a sort of a tunable
laser although it was not clarified in the reference.
Isolator
Isolator
PC
GC-
SOA
PBS
HWP2
HWP1 QWP
PMF
Port 2
Port 1
Output
75:25
coupler
CW
CCW
Figure 14. The schematic diagram of a multiwavelength SOA-fiber ring laser employing a PDLC (GC-
SOA: gain-clamped semiconductor optical amplifier, HWP: half-wave plate, QWP: quarter-wave plate,
PMF: polarization-maintaining fiber, PBS: polarization beam splitter, PC: polarization controller, CW:
clockwise, CCW: counterclockwise) [27].
1545 1550 1555 1560 1565 1570 1575
-50
-40
-30
-20
-10
O
p
t
i
c
a
l

p
o
w
e
r

[
d
B
m
]
Wavelength [nm]
1557 1558 1559 1560 1561 1562 1563
-50
-30
-10

O
p
t
i
c
a
l

p
o
w
e
r

[
d
B
m
]
Wavelength [nm]
0
T
h
e

a
n
g
l
e

o
f

H
W
P

1
-50
-30
-10

2
0
-50
-30
-10

4
0
-50
-30
-10
6
0
-50
-30
-10
8
0
(a) (b)
Figure 15. (a) The output spectrum of a multiwavelength laser (b) The tuning characteristic.
Yoon et al. published their research on the wavelength tunable SOA fiber ring laser [27].
They employed a polarization-diversity loop configuration (PDLC) comb filter as shown in
Fig. 14. The PDLC comb filter consists of two half-wave plates, a quarter-wave plate, a
polarization beam splitter, and a PMF. The PDLC filter can tune lasing wavelengths with the
control of a half-wave plate alone. A single polarization which comes out of the PBS enters
the PMF after passing through wave plates. The light experiences polarization change
according to its wavelength by the PMF. When the light meets the PBS again, the
transmittivity is determined by its polarization. The role of the wave plates is to make the
polarization-change by the rotation of the first half-wave plate reproduce the polarization-
change by the PMF. When we represent polarization change in the Poincare sphere, the
trajectory of polarization change becomes a circle. Because the rotation of the first half-wave
plate makes the trajectory rotate, the rotation of the first half-wave plate can shift the position
of the filter comb. In the PDLC filter, clockwise and counterclockwise lights exist. Two
counter propagating lights experience the same transmittivity and reflectivity. A 90 rotation
of the angle of the first HWP corresponds to the sweep of the entire channel spacing. The
channel spacing is determined by the length and birefringence of the PMF. In the experiment,
18 laser lines were observed with 0.8 nm channel spacing. Figs. 15 (a) and (b) show the
output spectrum and tuning characteristics. The rotation of a half-wave plate can shift the
position of lasing wavelengths linearly.
Few-mode Bragg grating
PC
Raman fiber
(SMF 50 km)
Tunable chirped
FBG
Output
Pump combiner
1425nm14351455 1465
Pump laser
Pump
(a)
d
Flexible metal plate
Compression
(Negative bending)
Tension
(Positive bending)
(b)
Figure 16. (a) An experimental setup for a multiwavelength Raman fiber laser based on few-mode
FBGs (b) A tuning method based on the symmetrical bending of a flexible metal plate (FBG: fiber
Bragg grating, PC: polarization controller) [28].
For a wavelength-tunable Raman fiber laser, Han et al. demonstrated the few-mode
FBG scheme as shown in Fig. 16 [28]. Because the few-mode FBG has multiple
resonant wavelengths, a multiwavelength Raman fiber laser could be obtained
without additional multichannel filters. The CFBG is used to form a linear cavity because
of the broad reflection spectrum. Tuning of the CFBG is needed to match the
reflection spectrum to that of a few-mode FBG. The lasing wavelength shift ( )
can be defined as (1 )
p
= , where
p
is the lasing wavelength of the Raman
fiber laser, is the photo-elastic coefficient, and is the strain induced by the bending
of the fiber. The experimental results showed 3 laser lines with 3.5 nm spacing and
the wavelength tuning characteristics. The number of laser lines was limited by the
few-mode FBG.
Han et al. also demonstrated temperature tuning of the lasing wavelength of a
multiwavelength Raman laser using the few-mode FBG [29]. Three laser lines
were also obtained and the temperature sensitivity was measured as 10.5 pm/C. They
also applied a similar structure to a temperature and strain sensor using a
multiwavelength Raman laser with a phase-shifted FBG [30]. The experimental result
showed that two lasing wavelengths could be shifted by strain and temperature with a fixed
spacing.
3.3. Channel Spacing Tunability
Dong et al. presented a multiwavelength SOA-fiber laser and Raman fiber laser
employing a fiber Fabry-Perot filter based on a superimposed CFBG [31, 32]. Two
super imposed CFBGs form the Fabry-Perot filter when the writing positions of the
two CFBGs are slightly different. The tuning of the chirp rate changes the channel
spacing. The SOA-fiber laser and Raman fiber laser could be implemented by using the
same filter. The SOA-fiber laser generated 10~13 laser lines with 0.3~0.6 nm channel
spacing and the Raman fiber laser generated 2~10 laser lines with 0.3~0.6 nm channel
spacing.
3.4. Both Wavelength and Channel Spacing Tunability
Roh et al. demonstrated a SOA-fiber laser with both wavelength and channel
spacing tunability [33]. They employed a PDLC comb filter with a differential delay
line (DDL) as shown in Fig. 17. The use of the DDL instead of a PMF could lead to
spacing tunability as well as wavelength tunability. Both the channel spacing and
lasing wavelength are continuously tunable. Channel spacing is tuned electrically and
wavelength is tuned by the rotation of a half-wave plate. Because this laser adopted a
PDLC comb filter, the wavelength tuning characteristics are totally equivalent to Ref. [27].
Experimental results showed channel spacing tunability of 0.4 ~ 1.6 nm with up to 23
laser lines.
PC
Isolator
SOA
Isolator
Output
CW
DDL
QWP HWP1
HWP2
CCW
1
2
75:25
coupler
S
F
h1
S
F
q
S
F
D
S
F
h2
PBS
Figure 17. The schematic diagram of the channel spacing and wavelength tunable SOA-fiber ring laser
(DDL: differential delay line, QWP: quarter-wave plate, HWP: half-wave plate, PBS: polarization beam
splitter, PC: polarization controller, CW: clockwise, CCW: counterclockwise) [33]
The features of the demonstrated SOA-fiber lasers and Raman fiber lasers are
summarized in Tables 2 and 3.
Table 2. Multiwavelength SOA-fiber lasers
Year
First author
[Reference]
Advantage
Channel
number
Channel
spacing
(nm)
Scheme
2001 S. Kim [26] 8
Fiber Bragg
grating
2003 B.-A. Yu [17] Wavelength switching 4 0.8
Sampled fiber
Bragg grating
2004 L. R. Chen [24] Spacing tuning 11, 6 1.6, 3.2
Programmable Hi-
Bi FLM
2004 Y. W. Lee [19] Waveband tuning 17 0.8 Hi-Bi FLM
2005 M. P. Fok [23] Wavelength tuning 21 0.8 PM-LMF
2005 Y.-G. Han [20] Spacing and waveband switching 5~20 0.8~4.1
PMF Lyot-Sagnac
filter
2005 X. Dong [31] Spacing tuning 13 0.4 Fabry-Perot
2006 I. Yoon [27] Wavelength tuning 18 0.8 PDLC
2006 S. Roh [33]
Wavelength and channel spacing
tuning
23 0.8 PDLC with DDL
Table 3. Multiwavelength Raman fiber lasers
Year
First author
[Reference]
Advantage
Channel
number
Channel
spacing
(nm)
Scheme
2001 F. Koch [15] Potential angle tuning 24 0.8 Fabry-Perot
2001
C. J. S. de Matos
[16]
Potential individual tuning 4 ~4.5 FBG
2003 C.-S. Kim [21]
Channel spacing switching and
wavelength tuning
20 0.4~3 Sagnac
2004 C.-S. Kim [25]
Channel spacing switching and
wavelength tuning
11
0.95, 0.88,
2.95
PMF Sagnac loop filter
with electro-optic
polarization controller
2004 Y.-G. Han [22] Channel spacing switching 7, 5 0.6, 0.8 PMF Lyot-Sagnac
2005 Y.-G. Han [28] Wavelength tuning 3 3.5 Few-mode fiber
2005 Y.-G. Han [30] Sensing (wavelength tuning) 2 1.4 Phase-shifted fiber
2006 X. Dong [32] Channel spacing tuning 2~10 0.3~0.6
Sample fiber Bragg
grating
4. Conclusion
As multiwavelength light sources become more important in WDM optical communication,
there is an increasing amount of research on the multiwavelength fiber lasers. In this chapter
we reviewed various schemes for a multiwavelength fiber laser to date. Feasible gain media
are the EDFA, the SOA and the SRS. Each of these schemes has different challenging
difficulties for multiwavelength generation.
For the EDFA, the most difficult problem is its homogeneous broadening. The unstable
lasing characteristic of the EDFA results from homogeneous broadening. As possible
schemes to overcome homogeneous broadening, we reviewed the techniques of cavity loss
balancing among wavelengths, self stabilization from FWM, liquid nitrogen cooling and
frequency shifted feedback.
On the other hand, inhomogeneous broadening of a SOA and Raman amplifier makes
multiwavelength generation easy. The tunability in lasing wavelength and channel spacing
becomes more important as optical communication systems become more flexible and
efficient. We focused on the challenging issues of tuning multiwavelength SOAs and Raman
fiber lasers.
Practically, tunable channel spacing and tunable lasing wavelengths are required features
in WDM optical communication systems. Tuning capabilities can be classified into
switchability and tunability. Todays tunable SOA-fiber and Raman fiber lasers were
classified and reviewed. Because the tunable characteristic originates from characteristics of a
filter, various filters for tunable lasers were introduced.
References
[1] Hamakawa. A.; Kato, T.; Sasaki, G.; Shigehara, M. Conf. Opt. Fiber Comm. 1997,
297-298.
[2] Takahashi, H.; Toba, H.; Inoue, Y. Electron. Lett. 1994, 30, 44-45.
[3] Park, N.; Dawson, J. W.; Vahala, K. J. IEEE Photon. Technol. Lett. 1992, 4, 540-541.
[4] Desurvire, E.; Zyskind, J. L.; Simpson, J. R. IEEE Photon. Technol. Lett. 1990, 2,
246 248.
[5] Goldstein, E. L.; Eskildsen, L.; da Silva, K.; Andrejco, M.; Silberberg, Y. IEEE Photon.
Technol. Lett. 1993, 5, 937 939.
[6] Chow, J.; Town, G.; Eggleton, B.; Ibsen, M.; Sugden, K.; Bennion, I. IEEE Photon.
Technol. Lett. 1996, 8, 60-62.
[7] Liu, X.; Lu, C. IEEE Photon. Technol. Lett. 2005, 17, 2541-2543.
[8] Han, Y.-G.; Tran, T. V. A.; Lee, S. B. Opt. Lett. 2006, 31, 697-699.
[9] Han, Y.-G.; Lee, S. B. Opt. Express 2005, 13, 10134-10139.
[10] Bellemare, A.; Karasek, M.; Rochette, M.; LaRochelle, S.; Ttu, M. J. Lightwave
Technol. 2000, 18, 825-831.
[11] Slavik, R.; LaRochelle, S.; Karasek, M. Opt. Commun. 2002, 206, 365-371.
[12] Zhou, K.; Zhou, D.; Fengzhong, F.; Ngo, Q. N. Opt. Lett. 2003, 28, 893-895.
[13] Ahmed, F.; Kishi, N.; Miki, T. IEEE Photon. Technol. Lett. 2005, 17, 753-755.
[14] Liu, X.; Lee, B. IEEE Photon. Technol. Lett. 2004, 16, 428-430.
[15] Koch, F.; Reeves-Hall, P. C.; Chernikov, S. V.; Taylor, J. R. Conf. Opt. Fiber Comm.
2001, 54, WDD7/1-WDD7/3.
[16] De Matos, C. J. S.; Chestnut, D. A.; Reeves-Hall, P. C.; Koch, F.; Taylor, J. R. Electron.
Lett. 2001, 37, 825-826.
[17] Yu, B.-A.; Kwon, J.; Chung, S.; Seo, S.-W.; Lee, B. Electron. Lett. 2003, 39, 649-650.
[18] Lee, Y. W.; Yu, B.-A.; Lee, B.; Jung, J. Opt. Eng., 2003, 42, 2786-2787.
[19] Lee, Y. W.; Jung, J.; Lee, B. IEEE Photon. Technol. Lett. 2004, 16, 54-56.
[20] Han, Y.-G.; Kim, G.; Lee, J. H.; Kim, S. H.; Lee, S. B. IEEE Photon. Technol. Lett.
2005, 17, 989-991.
[21] Kim, C.-S.; Sova, R. M.; Kang, J. U. Opt. Commun. 2003, 218, 291-295.
[22] Han, Y.-G.; Lee, J. H.; Kim, S. H.; Lee, S. B. Electron. Lett. 2004, 40, 1475-1476.
[23] Fok, M. P.; Lee, K. L.; Shu, C. IEEE Photon. Technol. Lett. 2005, 17, 1393-1395.
[24] Chen, L. R. IEEE Photon. Technol. Lett. 2004, 16, 410-412.
[25] Kim, C.-S.; Kang, J. U. Appl. Opt. 2004, 43, 3151-3157.
[26] Kim, S.; Kwon, J.; Kim, S.; Lee, B. IEEE Photon. Technol. Lett. 2001, 13, 350-351.
[27] Yoon, I.; Lee, Y. W.; Jung, J.; Lee, B. J. Lightwave Technl. 2006, 24, 1805-1811.
[28] Han, Y.-G.; Moon, D. S.; Chung, Y.; Lee, S. B. Opt. Express 2005, 13, 6330-6335.
[29] Han, Y.-G.; Lee, S. B.; Moon, D. S.; Chung, Y. Opt. Lett. 2005, 30, 2200-2202.
[30] Han, Y.-G.; Tran, T. V. A.; Kim, S.-H.; Lee, S. B. Opt. Lett. 2005, 30, 1114-1116.
[31] Dong, X.; Shum, P.; Xu, Z.; Lu, C. IEEE LEOS Ann. Meeting 2005, 814 815.
[32] Dong, X.; Shum, P.; Ngo, N. Q.; Chan, C. C. Opt. Express 2006, 14, 3288-3293.
[33] Roh, S.; Chung, S.; Lee, Y. W.; Yoon, I.; Lee, B. IEEE Photon. Technol. Lett. 2006, 18,
2302-2304.
Chapter 14
AGING AND RELIABILITY OF SINGLE-MODE SILICA
OPTICAL FIBERS
M. Poulain
1
, R. El Abdi
2
and I. Severin
3
1
UMR 6226, Universit de Rennes1, F-35042 Rennes, France
2
LARMAUR, Fre-Cnrs 2717, Universit de Rennes1, F-35042 Rennes, France
3
Universita Politechnica, Splaiul Independentei, IMST, 06042 Bucarest, Romania
Abstract
The optical fiber reliability in telecommunication networks has been still an issue, thats
why the question of how long an optical fibers might been used without a significant
probability of failure isnt out of interest. Much work was developed around this issue, but
the optical fiber fatigue and aging process has not been yet fully understood.
The reliability of the optical fibers depends on various parameters that have been
identified: time, temperature, applied stress, initial fiber strength and environmental
corrosion. The major and usually unique corrosion reagent is water, either in the liquid state
or as atmospheric moisture. Glass surface contains numerous defects, either intrinsic, the so-
called Griffiths flaws and extrinsic, in relation to fabrication process. Under permanent or
transient stress, microcracks grow from these defects, and growth kinetics depend on
temperature and humidity. Although polymeric coating efficiently protects glass surface from
scratches, it does not prevent water to reach glass fiber.
way the problems of failure and rupture of fibers subjected to severe aging conditions.
In the proposed chapter, some informations on the used characterization methodology for
the silica optical fibers are given. In addition, Optical fibers analysis advantages, expected
percussions and theoretical background are given to enlighten the potential concerned
persons. The principal optical fiber test benches are described and some results are
commented. Finally, final remarks are noted.
M. Poulain, R. El Abdi and I. Severin 356
1. Introduction
Terrestrial and submarine telecommunication networks depend critically on optical fibers.
While main emphasis is put on transmission and signal characteristics [1], more basic features
such as reliability and expected lifetime has appeared also as major concerns [2, 3]. However,
these concerns become less popular with the deep crisis that occurred in the telecom market
and the emergence of more advanced fibers: one may think that new fibers should replace the
existing ones earlier than expected. In addition, until now, operators did not face serious
problems in relation to fiber failure.
Nevertheless the reliability issue remains more than ever a topical question for several
reasons. Firstly, the impressive increase of the bit rate is accompanied by a power increase
which is supported by the fiber core and can generate catastrophic failure phenomena and
generate damage of the fiber ends or losses in the connectors. Secondly, current models
include humidity, applied stress and temperature as major aging factors, but their accuracy for
lifetime prediction is questionable.
Aging of silica fibers is now rather well understood as numerous studies have been
implemented in this area [2-17]. One must separate the case of the fatigue static behavior
where fibers are subjected to a permanent strain, e.g. bended fibers, and the dynamic fatigue
corresponding to an unexpected tensile stress arising from environmental changes. Failure
mechanism involves surface phenomena, which raise fundamental questions.
Surface defects, initiator for cracks grow have not yet been identified neither by Scanning
Electron Microscopy (SEM) nor by Atomic Force Microscopy (AFM). The current random
network model used actually to describe glass structure gives no explanation for the so-called
Griffiths flaws [18] and does not account for density fluctuations and inhomogeneities in
glass. While other models, such as the vacancy model [14, 19], may provide a physical
picture of these defects, they are still in an emerging state that limits their application.
Water is also critical in fiber failure: fiber strength may increase by 100 % if water is
missing, for example under vacuum, in a dry box or at liquid nitrogen temperature [3, 4, 16].
It is assumed that water molecules break the Si-O-Si chemical bonds of the vitreous network.
This simple and logical model may be incomplete and ignore some aspects of the whole
phenomenon.
Polymeric coatings are largely used to inhibit surface flaws and proved to be efficient.
However the reinforcement mechanism is not well understood.
The general use of Weibulls statistics in data processing may be inappropriate in some
cases: it is widely observed that fiber strength value calculated from Weibull plots decreases
as sample length increases, but Weibull formula is precisely expressed to be independent on
fiber length.
These questions, and others, have not only a fundamental interest, but still could have
notable economic implications.
The technology evolution and the research for low cost solutions lead to use new fibers
and new components. Thus, polymeric fibers are being considered for the local distribution,
while Bragg grating fiber components are now largely used in optical amplifiers. However,
the reliability of these new components has still to be evaluated.
A traditional stake is referred to the future fibers for the local distribution networks (Fiber
To The Home, FTTH). These fibers will be submitted to notable permanent stresses, for
Aging and Reliability of Single-Mode Silica Optical Fibers 357
example at door corners, thus exposure to temperature, humidity and sudden stress may be
larger than in classical cables.
The better understanding of the factors ruling aging and reliability of optical fibers should
lead not only to scientific advances, but also to economical spin-offs. The telecom market will
require light cables for local area networks, and the design of mini cables can be optimized on
this basis. In addition various markets are likely to open in other fields where optical fiber
components are key elements. This concerns optical fiber sensors, laser power delivery, fiber
lasers, monitoring and control, remote spectroscopy, in line imaging, etc Of particular
interest is automotive industry in those case reliability and cost make essential points.
On the fundamental level, the principal goal is to collect new information elements that
could help answering recurrent questions.
2. Background
Failure of fibers is rather well understood as it might be considered as a particular case of
fragile material fracture [2-18]. While such materials exhibit a large resistance to compressive
stress, they are much more sensitive to traction. Failure originates from surface flaws that
may be described as microcracks. Their cracks growth under tensile stress as effective stress
is amplified at the bottom of the crack. This growth is enhanced by water activity in most
materials, including oxide glasses. It is generally assumed that water acts by breaking the
chemical bonds between oxygen and silicium or other cations. For this reason the intrinsic
strength K
1C
of the material can be observed only in extremely dry conditions, e.g. vacuum or
liquid nitrogen.
In practice, optical fiber aging depends on various factors that may decrease effective
fiber strength: residual applied stress, temperature and water. It is assumed that surface flaws
are enlarged, consequently crack growth promotes. Maximum water activity is in aqueous
solutions and it is expressed by the relative humidity (RH) in current atmosphere.
Figure 1. Fracture morphology of silica optical fibre (see silica core typical fragile surface fracture
surrounded by the two layer epoxi-acrylate polymer coating).
As fiber surface has determined fracture to a large extent, external coating appears
critical. This coating is polymeric in most cases, and modern optical fibers are coated by two
different layers, a soft coating at glass surface and a hard coating at external surface (Fig. 1).
The coating first makes a protection against scratches that occur in normal handling; it also
fills the surface flaws gluing in some a way the two sides of the micro cracks and finally, it
reduces water activity at glass surface. Ideally, coating should prevent any water molecule to
reach glass surface. Unfortunately, polymeric coatings, including hydrophobic coatings, are
very permeable to water. Only inorganic hermetic coating could make an efficient barrier
against water. Polymeric coatings (e.g. epoxyacrylates) are preferred in practice because they
are more efficient to inhibit surface defects [13, 20-24].
Various theoretical models are applied for mechanical characterization of optical fibers
[25- 26], but the most common one is based on Weibull's statistics.
The Weibull law expresses the failure probability F of a fiber with a length L subjected to
an applied stress :
[ ] ) ( ) ( ]
1
1
[
1
o
Ln Ln m
F
Ln
L
Ln =
(1)
where m is a size parameter and
is a scale parameter.
The evolution of
]
1
1
[
1
F
Ln
L
Ln in function of Ln () is known as the Weibull
plot.
The values of m and
are calculated from the slope of the curve and the intersection
with the stress axis. The m parameter characterizes the defect size dispersion [26]. A high m
value indicates that the distribution of the defect size is homogeneous while a low m value
means that surface defects are varying in size. When the curve appears as a broken line with
two distinct slopes one small for low stress and the second one large, respectively one has
assumed two different families of defects, the first one corresponding to large extrinsic
defects, and the second one relating to intrinsic flaws. Other plots encompass several straight
lines relating to different groups of defects. The failure probability F is calculated from the
relation:
N
i
F
i
5 . 0
= (2)
where i represents the rank of the measurement and N the total number of values. The
parameter represents the stress corresponding to the fiber cumulative fracture probability F of
is 50%.
In the static fatigue measurements, the fiber is subject to a constant stress and one
measures the time to failure. This time t
f
is ruled by the following relation:
n
a
n
i
f
S
B t
2
= (3)
where B is a constant that depends on environment typically water, S the initial inert
strength of the fiber and n the stress corrosion parameter, and
a
the failure stress. The failure
probability of F can be written as:
F t L L
t
B S
f
f a
n
o
n
m
n
( , ) exp
.
.
=

1
2
2
(4)
Or, in the logarithmic form:
( )
( )
Ln t nLn LnB Ln S
f a i
n
= + +

2
(5)
The plot Ln t
f
( ) as a function of Ln
a
( ) describes the static fatigue behavior and gives
access to the fatigue parameters n and BS
i
n2
.
3. Experimental: Mechanical Measurements
The mechanical strength of the fibers may be measured in different ways corresponding to
static fatigue and dynamic fatigue [27, 28]. In the static fatigue tests, fibers are subject to a
permanent stress, and the time to failure is recorded for a set a fibers. Then a statistical
analysis gives values for the mean failure strength and the mean lifetime in the testing
conditions: type of fiber, temperature, applied stress and water activity. The dynamic fatigue
test consists in applying an increasing tensile strength until fiber breaks. From a convenient
data processing one finds the mean fiber strength. Special equipments are used for these
measurements, presented as follows.
3.1. Vertical Bench
The static fatigue under axial tensile loading consists in subjecting a fiber sample to a uniform
load as a suspending weight of known value. The two fiber ends are rolled up on a pulley
provided with a system allowing to block the fiber sample ends and to avoid any slip (Fig.
2a). The higher pulley (noted 1) is fixed on a support, while the lower pulley (noted 2) is
mobile and interdependent of a plate on which a chosen mass is applied.
This set up leads to carry out static tensile tests on high length fiber samples (usually 4 m
in length) for applied loads ranging between 5 and 50 N. A number of forty samples can
simultaneously be tested (Fig. 2b). The measurement of the fiber fracture time for different
loads allows determining the static stress corrosion parameter. Thus, the fibers are submitted
to different aging conditions and subsequently to mechanical tensile testing.
Silica fiber
Mass
Beam of light
Optical
sensor
Pulley2
Pulley1
Plate
(a) (b)
Figure 2. Vertical static tensile-test bench: (a) diagram of sample fiber under mass loading, (b) general
view.
3.2. Static Fatigue under Permanent Curvature
Another testing bench can be used for static testing [29]. Optical fibers, one meter in length,
are subjected to bending stresses by winding around alumina mandrel with calibrated
diameter sizes (Fig. 3a). The constant level of applied stress is adjusted by the proper choice
of the mandrel size. The time to failure is measured, and this corresponds to the time required
for the fiber strength to degrade until it equals the stress applied through winding round the
mandrel. The time to failure is measured by optical detection when the ceramic mandrel
moves out of the special holder. When fiber breaks, the mandrel rocks from its vertical static
position and the time to failure is directly recorded with an accuracy of 1 s. The testing setup
consists of a large number of vats containing 16 holders each (Fig. 3b).
The applied stress on the fiber depends on the mandrel diameter according to the
Mallinder and Proctor relation [30] as follows:
+ =
2
1
'
0

E
;
4
3
'
= ;
d
d
fiber
glass
+
=
(6)
where is the applied stress (in GPa), E
0
is Young modulus (equal to 72 GPa for the
silica), is the relative fiber deformation, is the constant of the elastic nonlinearity (equal to
6), is the mandrel diameter (in m), d
glass
is the glass fiber diameter and d
fiber
is the fiber
diameter including the polymer coating. For example, for standard silica optical fibers used
for telecommunication networks, d
glass
is equal to 125 m and d
fiber
is equal to 250 m; this
leads to the corresponding stress of 3.92, 3.76, 3.34 and 3.22 GPa for the calibrated diameter
mandrel of 2.3, 2.4, 2.7 and 2.8 mm respectively. The testing environmental conditions
during static fatigue measurements (temperature and relative humidity) should be also taken
into account.
E R
Clamping rings
Wound fibre on
calibrated
mandrel
Light beam
(a) (b)
Figure 3. Static bending test.
3.3. Vertical Dynamic Tensile Test
Movable
pulley
Dynamometric
cell
Engine
Device
speed
control
Fixed
pulley
5
0
0

m
m
Fiber
Higher
plate
(a) (b)
Figure 4. Schematic description of the dynamic tensile-test bench.
During a dynamic tensile test, the fiber is subjected to a deformation under a constant speed
until the rupture. The two fiber ends are rolled up on pulleys, having 65 mm in diameter and
covered with a powerful adhesive so as to prevent any fiber slip during the test (Fig. 4). The
lower pulley is fixed while the higher pulley is mobile and its displacement velocity (v,
mm/min) corresponds to the chosen deformation speed to carry out the test. Typical fiber
length is 500 mm.
During the test, the deformation and the tensile load are measured using a dynamometric
cell while the fiber deformation is deduced from the displacement between the fixed lower
pulley and the mobile higher plate.
The test velocity has an important influence for the failure stresses as this might be seen
in Fig. 5. High speeds lead to failure cracks with the same geometry (not curve slope variation
for v=500 mm/min), while the low speeds lead to various crack forms.
-4
-3
-2
-1
0
1
2
0,60 0,80 1,00 1,20 1,40
Failure stress (GPa)
L
n

(
-

l
n
(
1
-
F
)
)
v- 50mm/min
v- 150mm/min
v-300mm/min
v - 500mm/min
(F represents the cumulative fracture probability)
Figure 5. Evolution of failure stresses for different tensile test velocities v (mm/min).
3.4. Long Length Dynamic Tensile Bench
This mechanical bench (Fig. 6) allows to carry out tensile tests on fibers with high lengths
(from 0.5 m to 18 m) with broad speeds (ranging between 30 mm/min to 30 m/min with an
accuracy of less than 2 per 1000) and under very diverse environmental conditions
(temperature, aqueous solution...). Using the set up, one can obtain information on the defect
size dispersion onto the fiber surface and can determine the dynamic stress corrosion
parameter n. Indeed, this parameter is related to the velocity by the following relation:
K
A
n
I
V = (7)
where A is a parameter environment dependent, K
I
is the stress intensity factor and n is a
parameter characterizing the material capacity to resist to a stress.
3.5. Two Point Bending Bench
Fibers can also be characterized by using a two point bending testing device (Fig. 7). Samples
of 10 cm in length are bent and placed between the grooved faceplates of the testing
apparatus, in order to avoid the fiber slipping during the faceplate displacements and to
maintain the fiber ends in the same vertical plane.
Figure 6. Thirty meters long dynamic tensile bench.
Generally, a series of 30 samples are tested for different faceplate velocities (for example,
100, 200, 400 and 800 m/s, respectively). The failure stress is calculated from the distance
separating the faceplates, using the Proctor and Mallinder relation, improved by Griffioen [8].
Subsequently failure stress is obtained for each tested sample and tracing the classical
Weibull plots one might calculate the statistical parameters. Due to a very short fiber sample
part subjected under stress, this testing method is preferentially used to study the intrinsic
defects or selected flaws.
Fiber
Fiber
Faceplates
Control
Block
Stepper
motor
Faceplates
Computer
Piezo Electric
Sensor
Fiber

Figure 7. Dynamic two point bending bench.
4. Results
way the failure of the fibers subjected to severe aging conditions. An empirical relation
defining the fiber lifetime t
f
according to the temperature, residual stress and water content
was established. For this purpose, silica optical fibers were either immersed in hot water or
heated in wet atmosphere with a controlled relative humidity (RH).
The two relations derived from this set of measurements are the followings [14]:
( )

+
=
RT
T E
A t
a
f

0
0
exp (8)
( )

+
=

RT
T E
Z A t
a
f

0
0
exp (9)
The first relation (8) applies to fibers aged in liquid water, while the second one (9)
concerns fibers exposed to humid atmosphere with variable RH (RH = Z), temperature T and
applied stress . It is worth noting that the factor corresponds to some kind of relaxation
that decreases the effect of the applied stress. Its magnitude grows with temperature and
applied stress.
In a second set of measurements, fibers were immersed in hot water at 65C or at 85C
for a long time (up to 24 and 27 months). Then they were characterized in static and dynamic
fatigue. As one could expect, fiber strength in dynamic fatigue decreases as aging time
increases. More surprisingly, the time to failure of aged fibers subjected to static fatigue
increased enormously by comparison to non-aged fibers [31]. However this unexpected effect
did not follow a regular evolution versus time, but a rather cyclic one (Fig. 8).
(In legend calibrated mandrel diameter, in mm, in the case of the static fatigue testing set-up Fig.3)
Figure 8. Evolution of the fiber failure time in function of different aging conditions.
The explanation for this lifetime increase can be found in the structural relaxation
identified in the previous relations. The failure mechanism of the aged fibres involves surface
phenomena, in relation to water activity. A layer of hydrated silica is likely to be formed at
fibre surface [15]. This vitreous hydrated phase may relax under stress at room temperature,
which partly compensates the external applied stress in static fatigue. The change of the glass
surface was exemplified by the indentation behaviour that is different from that of normal
silica [32].
New experiments are carried out as well on standard silica fibers as on new fibers [33].
Fibers with a hermetic coating, fibers before and after photo-printing, fibers of polymer, on
average have diameters between 85 m and 125 m. The influence of temperature, water and
various corrosive agents on the mechanical fiber strength is determined. The coating aging is
also taken into account. Characterizations are also carried out on fibers belonging to different
vitreous systems (fluorides, oxides, sulphides) to detect and analyze less visible phenomena
when silica fibers are studied.
For several silica fibers subjected to vertical static tensile testing (see Fig. 2) under
various loadings, one can notice that more the suspended mass value is high; more the time of
rupture is large (Fig. 9). For weak loads (15 N), two families of cracks exist (a slope break
indicates the dispersion of the microcrack shapes).
-5
-4
-3
-2
-1
0
1
2
0 2 4 6 8
ln (time to rupture) (h)
l
n
(
-
l
n
(
1
-
F
)
)
20N
15N
25N
30N
Figure 9. Time to rupture evolution for different loadings (F represents the cumulative fracture
probability).
5. Final Remarks
The huge development of the telecommunication networks has been made possible by the
availability of low cost and high quality silica optical fibers. As industrial production reaches
millions of km, research rather focuses on networks and advanced components. Fiber
reliability is not a critical issue at this time because few problems were encountered, most of
them being accidental. However future fiber local loops will put fibers under large and
permanent stress. In addition, lighter and less expansive cables could be manufactured if
transient or permanent stresses have no significant influence on fiber lifetime.
Fiber aging has been the subject of numerous studies leading to theoretical models for
lifetime assessment. While ground observations do not contradict these predictions, the
accuracy of the models is questionable due to the complexity of the aging mechanism. In this
respect, experiments implemented on a long time scale are likely to bring new information.
There are some questions underlying the reliability studies. Aging parameters encompass
time, temperature, applied stress and water activity. While the critical part of water in failure
mechanism is well known, its real impact varies according to the physical state - liquid or
vapor - and partial vapor pressure. The stress applied to the fiber may be temporary, for
example during the proof test or network installation, or permanent when the fiber is bent in
cable and connecting areas. The aging mechanism is assumed to enlarge or to extend the
"Griffith flaws" which are spread at the fiber surface. These defects may be described as
micro-cracks which grow under applied stress in wet environment. Although this mechanism
is believed to be irreversible, water may also induce some curing effect which could
correspond to the geometrical smoothing of the crack tip [34, 35].
There is a practical interest in collecting quantitative information on the aging of the
commercial optical fibers over a long period of time. Such observations should allow a more
accurate comparison between experimental and calculated strengths and make lifetime
assessments more realistic as testing periods (> 2 years) become closer to the lifetime
required by network users that is at least 20 years.
Acknowledgments
Authors express their gratitude to France Telecom for technical assistance and equipment
supply and to Region Bretagne for financial support.
References
[1] Pal B. P., Fundamentals of fiber optics in telecommunication and sensor systems. (Wiley
Eastern ltd, Delhi, 1992).
[2] Olshansky R. and Maurer R. D., (1976). Tensile strength and fatigue of optical fibers. J.
Appl. Phys. 47, 4497-4499.
[3] Sakaguchi S., Kimura T., (1981). Influence of temperature and humidity on dynamic
fatigue of optical fibers. J. Amer. Ceram. Soc. 64 [5], 259-262.
[4] Duncan W. J., France P. W. and Craig S. P., The effect of environment on the strength
of optical fiber. Pp. 309-328 in Strength of Inorganic Glass, Edited by C.R. Kurkjian,
Plenum press, New York, 1985.
[5] Matthewson M. J. and Kurkjian C. R., (1988). Environmental effects of the static fatigue
of silica optical fiber. J. Amer. Ceram. Soc. 71 [3], 177-183.
[6] Kurkjian C. R., Krause J. T. and Mathewson M. J., (1989). Strength and fatigue of silica
optical fibers. J. ligthwave Tech. 7, 1360-1370.
[7] Michalske T., Smith W., Bunker B., (1991). Fatigue mechanisms in high-strength silica-
glass fibers. J. Am. Ceram. Soc. 74, [8], 1993-1996.
[8] Griffioen, W. Optical fiber reliability. Thesis edited by Royal PTT, The Netherlands
NV, PPT Research, Leidschendam, 1994.
[9] Glaseman G. S., (1994). Assessing the long term reliability of optical fibers. Proc.
National Fiber Optics Engineers Conference, 297.
[10] Muraoka M., Ebata K., Abe H., (1993). Effect of humidity on small-crack growth in
silica optical fibers. J. Am. Ceram. Soc. 76, [6], 1545-1550.
[11] Volotinen T. T. Water tests on optical fibers Proc. SPIE 3848, 134-143, (1999).
[12] Semjonov S. L., Kurkjian C. R., (2001). Strength of silica optical fibres with micron size
flaws. J. Non-Cryst. Solids, 283, 220-224.
[13] Armstrong J. M. and Matthewson M. J., (2000). Humidity dependence of fatigue of
high-strength fused silica optical fibers. J. Am. Ceram. Soc. 83, [12], 3100-3108.
[14] Poulain M., Evanno N., Gouronnec A. Static fatigue of silica fibers Optical fiber and
fiber component mechanical reliability and testing II, M. J. Matthewson, C. R. Kurkjian,
Editors, Proc. SPIE 4639, 64-74, (2002).
[15] Berger S., Tomozawa M., (2003). Water diffusion into silica optical fiber. J. Non-Cryst.
Solids, 324, 256-263.
[16] Gougeon N., El Abdi R and Poulain M., (2004). Evolution of strength of silica fibers
under various moisture conditions. Optical Materials, 27, 75-79.
[17] Severin I., El Abdi R. and Poulain M., (2007). Strength measurements of silica optical
fibers under severe environment. Optics & Laser Techn.. 39, [2], 435-441.
[18] Griffith A. A., Phil. Trans. 221A, 163 (1920).
[19] Poulain M., Vacancy model of ionic glasses. Proc Int. symp. Non Oxide Glasses, Part B,
pp 22-26, Corning USA and What is glass?(briton langage) KIANT, 1, 13-26, (1996).
[20] Wei T., Skutnik J., (1988). Effect of coating on fatigue behavior of optical fiber. J. Non-
Cryst. Solids, 102, 100-105.
[21] Kurkjian C. R., Simpkins P. G., Inniss D., (1993). Strength, degradation and coating of
silica lightguides. J. Am. Ceram. Soc. 76, [5], 1106-1112.
[22] Shiue S. T., Ouyang H., (2001). Effect of polymeric coating on the static fatigue of
double-coated optical fibers. J. App. Phy. 90, [11], 5759-5762.
[23] Mrotek J. L., Matthewson M. J., Kurkjian C. R., (2001). Diffusion of Moisture through
optical fiber coatings. Journal Light-wave Technol. 19, [7], 988-993.
[24] Mrotek J. L., Matthewson M. J., Kurkjian C. R., (2003). Diffusion of Moisture through
fatigue and aging-resistant polymer coatings on lightguide fibers. Journal Light-wave
Technol. 21, [8], 1775-1778.
[25] Schmitz G. K. and Metcalfe A. G., (1967). Testing of fibers. Mat. Res. Stand., 7 [4],
862-865.
[26] Matthewson M. J., (1994). Optical fiber reliability models. Proc. SPIE, Critical Reviews,
CR 50, 3-31.
[27] Matthewson M. J., (1994). Optical fiber mechanical testing techniques. Proc. SPIE,
Critical Reviews, CR 50, 31-59.
[28] Severin I., El Abdi R., Poulain M. and Amza G., (2005). Fatigue testing of silica optical
fibres. Journal of Optoelectronics and Advanced Materials, 7 [3], 1581-1588 .
[29] International standard IEC 793-1-3, First Edition 1995-10 .
[30] Mallinder, F.P., Proctor, B.A., (1964) Phys. Chem. Glass, 5, 91.
[31] Gougeon N., El Abdi R. and Poulain M. (2003). Mechanical reliability of silica optical
fibers. J. Non-Cryst. Solids, 316, 125-130.
[32] Gougeon N., Sangleboeuf J. C., El Abdi R., Poulain M. and Borda C. T., (2005).
Indentation Behavior of Silica Optical Fibers Aged in Hot Water. Fiber and Integrated
Optics. 24, [5], 491-500.
[33] Severin I., Poulain M., ElAbdi R. (2005). Phenomena associated to aging of silica
optical fibers. Photonic Applications in Devices & Communication Systems, P.
Mascher, A. P. Knights, eds., Proc. SPIE 5970.
[34] Hirao K., Tomozawa M. (1987). Kinetics of crack tip blunting of glasses. J. Am. Ceram.
Soc. 70, [1], 43-48.
[35] Hirao K., Tomozawa M., (1987). Dymanic fatigue of treated high-silica glass:
Explanation by crack tip blunting. J. Am. Ceram. Soc. 70 [6], 377-382.

INDEX

A
absorption spectra, 272
access, 53, 75, 112, 216, 359
accounting, 254
accuracy, 120, 149, 201, 215, 356, 360, 362, 366
acetone, 33
acetylene, 332
achievement, 206
acid, 31, 33, 36, 44, 104, 105
acrylate, 357
adaptability, 5
adjustment, 59, 215
adriamycin, 31, 48
adsorption, 40
aerospace, 5, 260
AFM, 356
agent, 31
aging, xii, 260, 355, 356, 357, 360, 364, 365, 366,
367, 368
aging process, xii, 355
albumin, 45
algorithm, 59, 60, 65, 68, 74, 77, 243, 246, 247, 320
alternative(s), x, 54, 65, 146, 231, 233, 238, 249,
255, 261
aluminum, 144
amplitude, x, 19, 78, 84, 85, 128, 131, 166, 171, 172,
177, 209, 279, 281, 282, 283, 284, 285, 286, 287,
288, 289, 290, 291, 293, 297, 304, 324
AN, 175, 231
annealing, 65
antibody, viii, 15, 28, 29, 30, 31, 36, 37, 38, 39, 40,
42, 43, 45, 46
antigen, 31, 38, 39, 40, 43
antimony, 188
APC, 220
apoptosis, 31
argon, 31, 260
arsenic, 273
assessment, 366
assignment, 251
assumptions, viii, 43, 52
asymmetry, 105, 152
atoms, ix, 119, 120, 122, 124, 128, 130, 131, 132,
133, 134, 136, 137, 138, 139, 140, 142, 143, 146,
147, 148, 149, 150, 151, 152, 153, 154, 336
attachment, 21, 38, 39, 40, 41, 42, 43, 44
attention, vii, ix, 3, 4, 5, 161, 163, 164, 176, 217,
258, 302
attenuated total reflectance, 261
Australia, 315
automobiles, 260
availability, 365
averaging, xi, 301, 303
avoidance, 108
B
Bacillus, 22, 28, 29, 47
Bacillus subtilis, 22
backscattering, ix, 54, 70, 71, 72, 73, 161, 162, 164,
170, 171, 172, 176, 177, 179, 180, 181, 182, 185
bacteria, 4, 28, 41, 48
bandgap, xi, 258, 260, 262, 265, 273, 315, 316, 317,
318, 319, 320, 321, 322, 332, 333
bandwidth, viii, x, 51, 54, 62, 63, 65, 68, 69, 72, 74,
76, 77, 84, 114, 115, 168, 170, 188, 206, 207, 216,
218, 221, 222, 232, 233, 241, 247, 257, 258, 259,
269, 270, 275, 342
beams, 120, 132, 140, 146, 150, 151, 154, 178, 179,
180, 181, 217, 280
beef, 22, 28, 46
behavior, viii, 15, 18, 35, 40, 72, 78, 97, 102, 115,
324, 356, 359, 367
Beijing, 3
bending, 177, 259, 262, 349, 350, 360, 361, 363
Index 370
bias, 346
binary decision, 252
binding, 29, 31, 39, 40, 41, 42, 47
biomarkers, 30
biomechanics, 84
biomolecule(s), viii, 15, 46
biosensors, 4, 15, 35
biotin, 29
birefringence, ix, 83, 84, 87, 88, 89, 90, 91, 92, 94,
98, 101, 102, 104, 106, 107, 109, 110, 170, 184,
225, 274, 340, 343, 346, 347, 349
bismuth, 184
blackbody radiation, 272
blocks, 297
blood, 26, 31, 275
BN, 175
BNP, 25, 26
Boltzmann constant, 58
bonding, 36, 44, 259
bonds, 356, 357
boundary value problem, 59
Bragg grating, viii, xii, 4, 51, 54, 83, 84, 85, 92, 97,
98, 101, 104, 106, 115, 161, 169, 170, 182, 184,
213, 227, 228, 315, 331, 332, 335, 338, 339, 349,
351, 356
branching, 210
brass, 271
breast carcinoma, 31
breathing, 280
broadband, viii, x, 51, 53, 63, 64, 79, 85, 113, 182,
187, 205, 208, 227, 261, 271
buffer, 22, 23, 24, 25, 26, 27, 29, 30, 38, 39, 177
building blocks, 297
burning, 169, 338
C
cabinets, 53
cables, vii, 53, 357, 366
calibration, 33, 94, 96, 98, 107, 192
Canada, 205
cancer screening, 31
candidates, 182, 244, 302
capillary, 120, 121, 127, 128, 131, 132, 133, 134,
135, 260, 262
carbohydrate, 37
carbon, 6, 8, 259
carboxylic groups, 36
carcinogenicity, 31
carcinogens, 31
carcinoma, 31, 48
cardiovascular disease, 30, 49
carrier, 4, 5, 114, 115, 165, 167, 206, 233, 234, 236,
317, 342
cDNA, 32
cell, 21, 26, 27, 28, 31, 35, 40, 44, 48, 148, 262, 361,
362
cell culture, 27, 31
cell growth, 28
ceramic, 360
CFBG, 338, 341, 350
CGLE, x, 279, 280, 281, 285, 286, 287, 288, 289,
290, 291, 293, 297
channels, 53, 54, 57, 63, 75, 76, 77, 84, 115, 169,
206, 213, 215, 216, 218, 219, 227, 236, 305, 338,
346
chemical bonds, 356, 357
chemical etching, 33, 98, 105, 106
chemical properties, 15
China, 3, 257
Chinese, 24
cladding, x, 6, 16, 17, 18, 19, 31, 88, 105, 106, 121,
122, 143, 144, 146, 167, 168, 177, 190, 209, 210,
215, 218, 232, 257, 258, 260, 261, 262, 263, 264,
265, 266, 269, 270, 272, 273, 274, 275, 316, 332
cladding layer, 177
classes, xi, 263, 273, 301, 303
cleaning, 36, 42
clustering, 201
CO2, 4, 207, 208, 261, 272, 274, 275
coagulation, 30, 48
coatings, 190, 258, 261, 271, 273, 356, 358, 367
codes, 112, 113, 114, 247
coding, 114, 244, 246, 247, 249
coherence, 89, 109, 151, 224, 225, 226
collaboration, 134, 309
collisions, xi, 149, 151, 279, 302, 315, 317, 324, 325,
326, 327, 328, 329, 330, 333
combined effect, 39, 312
communication, vii, viii, x, xi, xii, 51, 52, 55, 56, 79,
83, 84, 108, 169, 187, 206, 215, 231, 232, 234,
236, 238, 239, 242, 244, 246, 268, 269, 270, 275,
299, 302, 312, 335, 336, 352
communication systems, viii, x, xi, xii, 51, 52, 55,
56, 79, 108, 231, 232, 234, 238, 239, 244, 246,
275, 299, 335, 336, 352
community, 302
compatibility, 84, 205
compensation, x, 205, 206, 213, 218, 220, 221, 302,
311, 347
competition, 111, 169, 174, 179, 180, 291, 336, 339
competitiveness, 336
complementary DNA, 32
complexity, 165, 366
complications, 30
Index 371
components, ix, xi, 7, 18, 54, 57, 58, 68, 84, 87, 89,
90, 91, 93, 98, 104, 111, 122, 164, 166, 173, 176,
188, 205, 206, 207, 208, 212, 218, 220, 224, 225,
226, 233, 234, 236, 238, 250, 301, 303, 312, 356,
357, 365
composites, 295
composition, 190, 215
compounds, 4
computation, viii, 52, 61, 67
computing, 206, 253
concentration, 28, 29, 30, 31, 32, 33, 35, 37, 39, 40,
41, 42, 43, 44, 104, 105, 106, 107, 164, 168, 200,
215
condensation, 315
conduction, 173, 342
conductivity, 8, 172, 221
configuration, vii, 3, 4, 49, 57, 65, 72, 75, 77, 78,
135, 142, 164, 165, 166, 169, 178, 189, 207, 216,
218, 220, 222, 226, 227, 303, 337, 338, 349
confinement, vii, 16, 52, 177, 264, 269, 271, 316
confusion, 342
Congress, 202
conjecture, 318, 321
consolidation, viii, 51
constituent materials, 258, 263, 272
constraints, 303, 307
contaminant, 47
contamination, 7, 32, 48, 143
continuity, 122, 173
control, 5, 68, 71, 72, 81, 154, 166, 170, 202, 212,
213, 215, 219, 220, 232, 244, 247, 302, 337, 340,
343, 345, 349, 357, 361
convergence, 8, 59, 65, 73
conversion, ix, 4, 108, 164, 187, 188, 202, 220, 233,
252, 260, 275, 332
cooling, xi, 68, 97, 111, 148, 154, 293, 335, 336,
338, 341, 352
Copenhagen, 255
corn, 26
coronary heart disease, 30
correlation(s), 176, 179, 180, 182, 193, 194, 201
correlation function, 176, 180
corrosion, xii, 84, 355, 359, 360, 362
costs, 68, 84
couples, 55
coupling, xi, 7, 16, 18, 58, 85, 87, 114, 128, 134,
136, 146, 147, 177, 178, 179, 182, 185, 187, 209,
216, 218, 264, 274, 301, 303, 315, 316, 317, 333,
342
covalent bonding, 36, 44
coverage, 40, 41, 43
crack, 357, 362, 366, 367, 368
C-reactive protein, 25, 30
critical value, 170, 179
cross-phase modulation, xi, 166, 301, 312
CRP, 25
crystal growth, 154
crystalline, 59, 273
culture, 27, 28, 31
curing, 366
CVD, 30
cytochrome, 30, 31, 48
cytokines, 30, 31, 47
cytoplasm, 31
D
damping, 298
data processing, 356, 359
data transfer, 269
decay, ix, 56, 120, 187, 188, 189, 196, 197, 199, 201,
202, 336
decibel, 62
decision making, 253
decisions, 246, 252
decoding, 114, 238, 246, 247, 252
decoupling, 108
defects, xii, 7, 164, 355, 356, 358, 363, 366
defense, 15, 272
deficiency, 30
definition, 78, 146, 240, 304, 317, 321
deformation, 84, 96, 99, 103, 207, 361, 362
degenerate, 71, 141, 274, 305, 339
degradation, 52, 54, 62, 220, 367
delivery, x, 6, 257, 258, 261, 275, 316, 333, 357
demand, 7, 52, 53, 260, 269, 270
Denmark, 255
density, vii, 3, 5, 6, 7, 9, 53, 56, 73, 154, 162, 163,
164, 172, 173, 175, 176, 181, 198, 199, 200, 202,
240, 271, 356
density fluctuations, 356
dependent variable, 59
depolarization, 222, 225
deposition, 45, 139, 154
derivatives, 199, 304, 309
destruction, 53
detection, viii, x, 4, 15, 21, 28, 29, 30, 37, 38, 40, 42,
44, 45, 46, 47, 48, 49, 131, 134, 217, 225, 231,
232, 233, 234, 235, 236, 237, 238, 239, 241, 249,
250, 255, 271, 275, 360
detection techniques, 271
detonation, 4
deviation, 68, 76, 77, 176, 179, 180, 182, 221
dielectric constant, 315
dielectric permittivity, 175
dielectrics, 271
Index 372
differential equations, 162, 165, 171, 175, 198
differentiation, 304, 309
diffraction, ix, 119, 120, 124, 125, 126, 143, 144,
152, 153, 154, 280
diffusion, 367
digital communication, 242
diode laser, 29, 133, 134, 165
diodes, 64, 65, 162, 165, 170
dipole, ix, 119, 120, 121, 128, 130, 134, 136, 146,
150, 151, 152, 154, 155, 171, 173
dipole moment, 173
dispersion, x, xi, 53, 58, 68, 71, 72, 110, 122, 123,
163, 165, 183, 206, 216, 218, 220, 226, 227, 228,
231, 232, 238, 241, 259, 261, 264, 274, 279, 280,
281, 285, 293, 301, 302, 303, 305, 307, 308, 309,
311, 312, 315, 316, 317, 318, 320, 321, 322, 324,
325, 326, 327, 328, 329, 330, 331, 333, 338, 347,
358, 362, 365
displacement, 297, 343, 362
distortions, 244
distribution, 5, 59, 69, 70, 120, 124, 126, 139, 140,
141, 142, 143, 144, 146, 150, 152, 153, 154, 173,
176, 192, 193, 198, 206, 246, 250, 356, 358
distribution function, 173
divergence, 206
diversity, 349
division, xi, 53, 112, 169, 182, 183, 187, 217, 227,
228, 270, 311, 335, 336, 337, 345, 346
DNA, viii, 4, 15, 21, 22, 27, 31, 32, 44, 48, 49
DOP, 221, 226
dopants, 215, 336
Doppler, 148
dream, 302
drinking water, 31
DRS, 72
DSC, 221
duration, 207, 244
dyes, 30
dynamic control, 220
dynamical systems, 299
E
E. coli, viii, 15,23, 28, 29, 32, 35, 37, 38, 40, 41, 44
earth, 183, 202
EEA, 79
eigenvalue, 240, 321
Einstein, Albert, 154, 306, 312, 315
elaboration, 208
elasticity, 257
electric current, 34
electric energy, 4, 5
electric field, xi, 19, 121, 124, 127, 142, 171, 176,
177, 281, 301, 303
electrical power, 221
electrodes, 34
electromagnetic, viii, ix, 83, 84, 119, 120, 121, 124,
154, 171, 302
electromagnetic fields, ix, 120, 121, 154
electromagnetic waves, 302
electromagnetism, 5
electron(s), 55, 128, 131, 134, 144, 162, 173, 175,
280, 298, 342
ELISA, 31
elongation, 208, 209
emission, ix, 28, 30, 56, 57, 72, 94, 111, 165, 187,
189, 191, 192, 193, 194, 195, 196, 197, 198, 199,
201, 202, 219, 342
encoding, 114, 251
endotoxins, 29
endurance, vii, 3, 9
energetic materials, 4
energy, vii, xi, 3, 4, 5, 6, 7, 8, 55, 62, 77, 87, 134,
139, 151, 162, 164, 166, 173, 176, 178, 189, 190,
197, 198, 199, 200, 216, 257, 269, 280, 291, 301,
303, 316, 318, 320, 321, 325, 338, 342
energy density, 5
energy transfer, 342
enlargement, 54
environment, 5, 16, 30, 44, 272, 280, 359, 362, 366,
367
environmental change, 356
environmental conditions, 212, 337, 361, 362
enzyme(s), 4, 48
epoxy, 37
equilibrium, 39, 56, 280, 297
equipment, vii, 99, 100, 101, 104, 108, 366
erbium, ix, xi, 94, 111, 162, 164, 165, 166, 167, 168,
169, 182, 183, 184, 187, 188, 201, 219, 335, 336,
339, 340, 341
Escherichia coli, 22, 28, 45, 46
estimating, 251, 252
etching, 28, 33, 44, 98, 104, 105, 106, 107, 181
ethylene glycol, 38, 39
Euro, 276
European Union, 79
evanescent waves, 132, 133, 134, 151
evaporation, 143, 144
evidence, 31, 182
evolution, 4, 52, 54, 57, 59, 61, 62, 67, 68, 73, 74,
84, 89, 94, 96, 97, 105, 106, 110, 151, 162, 165,
171, 183, 281, 290, 291, 294, 296, 313, 356, 358,
364, 365
excitation, 29, 30, 32, 46, 127, 130, 131, 132, 134,
144, 196, 210
Index 373
exercise, 254
exploitation, 258
exponential functions, 197, 309
exposure, viii, 83, 85, 104, 106, 213, 214, 357
extinction, 169, 222, 223, 224
extrusion, 273
F
fabrication, viii, ix, x, xii, 15, 33, 44, 181, 196, 205,
206, 213, 214, 216, 219, 226, 227, 229, 257, 272,
273, 355
Fabry-Perot filters, xii, 335
failure, xii, 84, 222, 355, 356, 358, 359, 360, 362,
363, 364, 365, 366
family, xi, 303, 315, 316, 317, 318, 320, 321, 328
fatigue, xii, 355, 356, 358, 359, 361, 364, 365, 366,
367, 368
feedback, xi, 54, 56, 161, 162, 163, 170, 182, 331,
335, 336, 341, 342, 352
FFT, 241
fiber aging, 357
fiber optics, 101, 104, 272, 366
fibers, vii, viii, ix, x, xii, 3, 4, 5, 6, 7, 8, 9, 15, 16, 17,
18, 21, 28, 29, 32, 34, 35, 37, 41, 42, 45, 51, 53,
55, 58, 59, 68, 72, 83, 84, 86, 87, 88, 89, 91, 92,
96, 97, 98, 104, 105, 106, 107, 108, 111, 119, 120,
121, 123, 124, 132, 133, 139, 154, 163, 167, 187,
188, 190, 191, 227, 228, 232, 257, 258, 259, 260,
261, 262, 263, 264, 265, 266, 269, 270, 271, 272,
273, 274, 275, 298, 299, 302, 311, 312, 316, 332,
333, 340, 355, 356, 357, 358, 359, 360, 361, 362,
364, 365, 366, 367, 368
fibre laser, 206
fibrinogen, 27, 30
film(s), 144, 215, 261, 273
filters, xii, 4, 109, 165, 166, 206, 209, 218, 219, 227,
228, 239, 335, 338, 350, 352
financial support, 309, 366
first generation, x, 257, 258
flame, 33, 34, 44, 208
flatness, 54, 69, 219, 220
flexibility, 15, 84, 110, 212, 257, 258, 274, 275
flight, 148
fluctuations, vii, 15, 55, 166, 215, 227, 286, 356
fluid, 280, 281
fluorescence, 26, 28, 29, 30, 31, 44, 48, 49, 56, 148,
150, 197
fluorine, 259
fluorophores, 48
focusing, 5, 119
food, 29, 44, 47
food poisoning, 29
Fourier, 125, 157, 219, 227, 239, 304, 310, 320
Fourier transformation, 125
four-wave mixing, xi, 169, 170, 311, 335, 336, 338,
339
France, 157, 161, 162, 164, 166, 168, 170, 172, 174,
176, 178, 180, 182, 184, 276, 355, 366
freedom, xi, 240, 293, 301, 303
FTTH, 269, 275, 356
function values, 65
functionalization, 37
fusion, 19, 33, 34, 35, 38, 44, 190, 206, 207
G
gases, 4, 316
Gaussian, 127, 143, 144, 146, 147, 149, 153, 176,
180, 181, 239, 240, 241, 246, 250, 302, 312
gel, 45, 259
generalization, 165, 170
generation, ix, x, xi, xii, 54, 57, 108, 111, 113, 119,
120, 140, 141, 143, 152, 154, 257, 258, 260, 271,
275, 306, 312, 316, 324, 335, 336, 342, 346, 352
geometrical parameters, 219
germanium, 93, 122, 261
Germany, 81
Ginzburg-Landau equation, x, 279, 280, 297, 298,
299
Ginzburg-Landau equation (CGLE), x, 279, 297
glass(es), vii, ix, xii, 4, 5, 7, 8, 30, 88, 91, 120, 121,
122, 127, 131, 132, 133, 142, 188, 190, 192, 196,
197, 198, 200, 205, 227, 257, 258, 259, 260, 261,
262, 270, 273, 336, 355, 356, 357, 358, 360, 361,
365, 366, 367, 368
glass transition temperature, 259
global communications, 83
glucose, 35, 44, 275
glycine, 38
glycol, 38, 39
gold, 26, 37, 44, 48
graph, 41, 77, 96, 127, 181
gratings, viii, xii, 4, 54, 72, 83, 84, 91, 92, 98, 101,
102, 104, 108, 109, 114, 115, 161, 170, 184, 213,
218, 227, 232, 331, 332, 335, 340
gravity, 147, 151
grazing, 120, 132, 260
Greens function, 176
Griffiths flaws, xii, 355, 356
groups, 30, 36, 37, 55, 132, 163, 166, 196, 358
growth, xii, 28, 44, 46, 154, 232, 258, 269, 298, 318,
321, 323, 355, 357, 367
growth rate, 318, 321, 323
guidance, 17, 120, 127, 132, 133, 134, 136, 139, 146,
154, 155, 273, 316, 332
Index 374
H
halogen, 192
Hamiltonian, 293
HD, 35
HDPE, 271, 272
HE, 122
healing, 31, 47
heart disease, 30
heat(ing), vii, x, 15, 17, 19, 33, 35, 38, 44, 97, 130,
136, 150, 151, 207, 208, 209, 213, 214, 257, 258,
275
height, 152, 177
helium, 133
hemoglobin, 30
high power density, 164
hip, 24, 26, 113
homeland security, 44
homogeneity, 111, 112
Hong Kong, 301
host, 264, 271, 273, 274, 336
house(ing), 33, 115, 202
humidity, xii, 4, 45, 280, 355, 356, 357, 361, 364,
366, 367
hybrid, 53, 65, 66, 68, 77
hybridization, 32, 44, 48
hydrocarbons, 45
hydrochloric acid, 36
hydrofluoric acid, 33, 44, 104
hydrogen, 45, 213, 259, 316, 332
hydrogenation, 213
hydrolysis, 36
hydroxide, 36
hydroxyl groups, 36, 196
hypothesis test, 249, 250
I
identification, 233, 238
ignition energy, 8
IL-6, 27, 31
illumination, 260, 275
images, 4, 124, 142
imaging, 31, 120, 126, 143, 144, 146, 147, 150, 155,
271, 357
immobilization, viii, 15, 28, 29, 36, 37, 40, 43, 47
immunity, viii, 83, 84
impairments, 238
implementation, xii, 55, 57, 62, 66, 68, 112, 113,
115, 228, 335, 336
impurities, 55
in situ, 164
in vivo, 31, 48
incidence, 16, 120, 132, 260
inclusion, 254, 282
India, 55
indication, 31
indices, 6, 121, 122, 212, 213, 214, 263, 264, 269
indirect measure, 84
industrial production, 365
industry, 357
inelastic, 55, 293
infarction, 30
infinite, 121, 127, 207, 212, 345
information processing, 206
information technology, 83
inhibition, 48
initial state, 246, 297
inoculation, 23
insertion, viii, 83, 84, 108, 114, 165, 188, 190, 200,
201, 206, 215, 216, 218, 220, 221, 223, 336, 346
insight, 18
instability, xi, 111, 285, 289, 298, 302, 315, 316,
320, 321, 323, 328
instruments, 191
integrated optics, 178
integration, 59, 60, 218, 225, 226, 238, 240, 254
intensity, 4, 28, 30, 44, 53, 56, 102, 120, 122, 123,
124, 126, 127, 128, 130, 131, 132, 133, 134, 138,
139, 141, 142, 144, 145, 146, 147, 150, 152, 153,
154, 165, 180, 181, 188, 211, 212, 217, 232, 305,
331, 362
interaction(s), viii, xi, 17, 29, 43, 52, 53, 54, 55, 57,
64, 76, 77, 133, 136, 152, 166, 171, 279, 293, 294,
298, 299, 302, 315, 316, 317, 324, 325, 326, 332,
336, 342
interface, 16, 53, 124, 132, 142, 209, 210, 258, 260
interference, 5, 84, 108, 113, 114, 115, 152, 174,
176, 185, 217, 232, 241
internet, 80, 187, 255, 269
interpretation, 21
interval, 232, 233, 236, 237, 240, 243, 246, 247, 249
inversion, 63, 183, 233, 342
investment, 52
ionization, 134, 139
ions, 4, 128, 131, 189, 190, 198, 200, 201, 219, 261,
336
IR, 35, 42, 80, 272, 273, 274
Islam, 81
isolation, 212, 216, 218, 220, 223, 224, 346
isotope(s), 136, 137
Italy, 51, 81, 161, 227
iterative solution, 165
Index 375
J
Japan, 33, 134, 187, 276, 298, 312
K
Karhunen-Loeve Series Expansion (KLSE), x, 231,
238, 239, 240, 241, 243, 253, 254
kernel, 240, 241, 253
kinetics, xii, 32, 40, 48, 355
Korea, 119, 134, 231, 276, 335
L
laser(s), vii, viii, ix, x, xi, xii, 3, 4, 5, 6, 7, 8, 9, 29,
31, 42, 51, 53, 54, 56, 64, 65, 68, 69, 70, 71, 72,
73, 74, 75, 81, 108, 111, 112, 115, 119, 120, 121,
123, 124, 127, 128, 129, 130, 131, 132, 133, 134,
135, 136, 137, 138, 139, 140, 141, 142, 146, 147,
148, 149, 150, 151, 152, 153, 154, 155, 161, 162,
163, 164, 165, 166, 167, 168, 169, 170, 171, 172,
173, 175, 176, 177, 178, 179, 180, 181, 182, 183,
184, 185, 187, 189, 191, 196, 201, 202, 206, 207,
208, 215, 221, 222, 226, 233, 234, 257, 258, 260,
261, 269, 272, 274, 275, 280, 281, 287, 297, 298,
299, 300, 316, 332, 333, 335, 336, 337, 338, 339,
340, 341, 342, 343, 344, 345, 346, 347, 348, 349,
350, 351, 352, 357
laser radiation, x, 257, 258, 272, 275
lasing effect, 72, 73
lattices, 313
leakage, 53, 143, 172, 175, 177, 208, 258, 262, 264,
266, 316
lectin, 29
LED, 206
lens, 7, 120, 126, 147, 150, 155
leprosy, 31
lifetime, 84, 152, 196, 197, 198, 342, 356, 359, 364,
365, 366
light beam, 136, 146
light scattering, 136
light transmission, 18, 28, 39
likelihood, 238, 246, 249
limitation, 164, 227, 259
linear function, 109
linkage, 29, 38
links, vii, 53, 275, 311
liquid nitrogen, 336, 338, 352, 356, 357
liquid phase, 45
Listeria monocytogenes, 23, 28, 46, 47
literature, 16, 98, 164, 165, 166, 170, 200, 260, 274,
303, 305, 307
local area networks, 357
location, 18, 228, 304, 305, 342
long distance, 120, 221, 232, 320, 323
LPS, 29
luminescence, 189, 196, 197, 201
lysine, 28
M
magnetic field, 150
malaria, 31
management, 52, 302, 303, 311
Manakov model, xi, 301, 303
manipulation, 119, 154, 206
manufacturer, 37
manufacturing, 207, 257, 258
mapping, 176, 236
market(s), 84, 258, 260, 275, 356, 357
Marx, 48
masking, 144
Massachusetts, 81
matrix, 59, 73, 81, 99, 100, 103, 108, 162, 173, 210,
211, 223
Maxwell's equations, 162
measurement, viii, 15, 32, 46, 47, 84, 98, 99, 100,
104, 108, 126, 148, 162, 169, 201, 202, 358, 359
measures, 60, 303, 305, 358
mechanical properties, vii, 5, 263
mechanical stress, 94
mechanical testing, 367
media, viii, xi, xii, 35, 47, 51, 169, 306, 313, 315,
331, 335, 336, 352
medical diagnostics, 44, 275
medicine, 84
melting, vii, 190
memory, 184
metals, 271
methanol, 26
microarray, 48
microcavity, 149
micrometer, 96
microorganisms, 28
microscope, 7, 31, 33, 105, 144, 146, 208
microscopy, 272
microspheres, 31
microwave, 108, 264, 271
military, 272
miniaturization, 5
mitochondria, 31
mixing, xi, 54, 169, 170, 183, 232, 302, 311, 335,
336, 338, 339, 341
model system, 305
modeling, 46, 59, 63, 68, 166, 183, 273, 308
Index 376
models, x, 162, 164, 165, 170, 231, 305, 316, 356,
358, 366, 367
modules, 52, 206, 212, 218, 222, 227
modulus, 261, 305, 310, 361
moisture, xii, 355, 367
molecular weight, 269
molecules, 15, 32, 43, 44, 49, 55, 259, 273, 332, 356
moon, 353
Morocco, 205
morphology, 357
motion, 131, 294
motivation, 305
multidimensional, 313
multiples, 54, 237
multiplicity, 280
multiplier, 128, 131, 134
muscles, 30
mutation, 65, 68
myocardial infarction, 30
myoglobin, 30, 47
N
NADH, 24
nanometers, 222
National Science Foundation, 45
national security, 29
Nd, 4, 6, 164, 196
necrosis, 27, 31
neodymium, 164
Netherlands, 255, 367
network(ing), ix, x, 52, 114, 188, 205, 207, 216, 218,
226, 227, 269, 356, 366
neural networks, 65
New York, 79, 115, 155, 156, 157, 158, 255, 297,
299, 311, 313, 366
next generation, 108
NIR, 192
nitrogen, xi, 133, 335, 336, 338, 341, 352, 356, 357
nodes, 124, 146, 154
noise, viii, 51, 52, 54, 56, 57, 62, 63, 64, 75, 162,
165, 166, 167, 168, 169, 182, 184, 189, 191, 206,
218, 220, 221, 228, 238, 239, 240, 241, 246, 247,
250, 302
nonequilibrium, 299
nonlinear dynamics, 154
nonlinear optical response, 302, 331
nonlinear optics, 4, 279, 280, 281, 302
nonlinear Schrdinger model, xi
nonlinear systems, 297
normalization constant, 127
numerical analysis, 123, 262
numerical aperture, 6, 7, 168, 190, 198, 259
numerical computations, 65
O
observations, 244, 366
OFS, 188
oligomers, 48
one dimension, 233, 234
On-Off Keying (OOK), x, 231, 232, 233, 238
operator, 65, 171, 173, 304, 309
optical communications, ix, 4, 6, 76, 83, 91, 115,
120, 164
Optical Differential Phase Shift Keying (oDPSK), x,
231, 232, 233, 234, 235, 236, 237, 238, 239, 240,
241, 242, 243, 245, 246, 247, 248, 249, 251, 253,
254, 255
optical fiber, vii, viii, ix, x, xi, xii, 4, 5, 6, 7, 15, 17,
30, 31, 33, 34, 45, 46, 48, 51, 53, 54, 55, 57, 72,
76, 84, 85, 87, 102, 119, 120, 121, 123, 124, 127,
132, 136, 137, 138, 139, 146, 151, 154, 163, 164,
165, 183, 187, 188, 227, 232, 234, 257, 258, 259,
268, 269, 271, 272, 275, 279, 297, 302, 311, 312,
315, 316, 333, 335, 342, 355, 356, 357, 358, 361,
364, 365, 366, 367, 368
optical gain, 56, 221
optical polarization, 222
optical properties, 55, 56, 84
optical pulses, 163, 233, 281, 305, 311
optical solitons, 280, 302, 332
optical systems, 215, 308
optics, vii, ix, 4, 83, 101, 104, 109, 115, 119, 120,
126, 178, 207, 272, 279, 280, 281, 302, 315, 366
optimization, viii, ix, 52, 60, 64, 65, 68, 69, 76, 77,
83, 115, 225, 244, 273
optimization method, 65
ordinary differential equations, 198
orientation, 98, 236
orthogonality, 18, 19, 233
OSA, 69, 94, 167, 192, 193, 201, 208, 228
oscillation, 153, 162, 165, 166, 170, 209
oscillograph, 8
oxides, 365
oxygen, 357
P
palladium, 45, 46
PAN, 192
parameter, xi, 6, 17, 41, 71, 177, 199, 219, 266, 280,
281, 287, 289, 290, 291, 294, 295, 301, 303, 305,
307, 309, 310, 316, 358, 359, 360, 362
Paris, 276
Index 377
particles, 154, 275
particulate matter, 21
partition, 246
passive, viii, ix, x, 54, 83, 84, 165, 166, 205, 224,
269, 271, 272, 275
pathogens, 28, 44, 48
PBC, 113, 222, 223, 224
PCF, 169, 170, 262, 316, 332, 338, 339
PDEs, 281
penalties, x, 70, 231, 233
performance, x, 7, 8, 62, 63, 65, 113, 115, 162, 165,
166, 169, 171, 182, 189, 218, 219, 220, 228, 231,
232, 233, 236, 238, 239, 240, 241, 242, 243, 244,
247, 248, 254, 255, 258, 259, 260, 275
periodicity, viii, 83, 89, 166, 265
permit, 290, 304, 305
permittivity, 172, 175, 177
pH, 38, 39, 42
phase shifts, 234, 236
phase transitions, 280
phonons, 55, 56, 63
phosphate, 28
photoelastic effect, 87, 110
photographs, 91, 105
photonic crystal fiber, vii, 258, 262, 273, 316, 332,
333, 338
photonic crystal fiber (PCF), 338
photons, 55, 56, 151, 189, 194, 196, 197, 198
photosensitivity, 84, 213, 227
physics, viii, 15, 44, 183, 279, 281, 297, 305, 309,
315
pitch, 343
plane of polarization, 141
plasma, 23, 26, 30, 279, 297
plasminogen, 30
plastics, 271
PM, 69, 92, 223, 226, 346, 351
PMMA, x, 257, 258, 259, 260, 262, 263, 264, 266,
268, 269, 270, 275
POFs, x, 257, 258, 259, 260, 275
polarity, 238
polarization, vii, ix, x, xi, xii, 57, 58, 64, 83, 84, 87,
88, 89, 90, 94, 96, 97, 98, 99, 100, 102, 103, 106,
108, 109, 110, 111, 112, 113, 114, 115, 140, 141,
148, 165, 166, 167, 169, 170, 171, 172, 173, 175,
176, 183, 206, 216, 218, 220, 221, 222, 223, 224,
225, 226, 227, 228, 229, 231, 264, 274, 275, 301,
303, 335, 337, 339, 340, 342, 343, 344, 345, 346,
347, 348, 349, 351, 352
polarized light, 87, 88
polycarbonate, 260, 261, 271
polyethylene, 271, 273
polyimide, 260
polymer(s), x, 257, 258, 259, 260, 261, 262, 263,
268, 269, 272, 273, 274, 275, 357, 361, 365, 367
polymer molecule, 259
polymeric materials, 257
polymerization, 259, 269
polymethylmethacrylate, x, 257
polystyrene, 22, 23, 29
poor, 35, 264, 338
population, 63, 65, 66, 68, 165, 198, 342
population size, 66
ports, 207, 223, 224
Portugal, 48, 51, 81, 83, 279
positive correlation, 194, 201
positive feedback, 56
power, vii, ix, x, 3, 4, 5, 6, 7, 8, 9, 19, 28, 30, 52, 53,
54, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 68, 69,
70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 88, 111, 114,
115, 132, 134, 135, 138, 139, 142, 146, 147, 149,
150, 152, 153, 162, 163, 164, 165, 167, 169, 170,
176, 177, 179, 182, 187, 188, 189, 191, 192, 193,
194, 195, 196, 197, 199, 200, 201, 202, 206, 207,
208, 209, 210, 211, 216, 218, 219, 220, 221, 222,
225, 226, 227, 234, 254, 257, 258, 260, 261, 263,
265, 275, 302, 311, 316, 338, 342, 344, 346, 347,
348, 356, 357
prediction, 290, 321, 356
pressure, viii, 4, 15, 45, 68, 131, 149, 213, 227, 280,
366
prices, 68
probability, xii, 7, 173, 197, 198, 238, 240, 247, 250,
253, 355, 358, 359, 362, 365
probability density function, 240
probe, 27, 29, 32, 57, 61, 63, 64, 67, 68, 69, 70, 72,
75, 77, 148, 150
production, viii, 83, 84, 91, 108, 268, 365
production costs, 84
production technology, viii, 83, 108
progesterone, 31, 47
prognosis, 49
program(ming), 19, 35, 65, 79, 302
propagation, ix, 18, 19, 52, 61, 65, 74, 75, 83, 84, 85,
87, 89, 108, 109, 122, 127, 144, 146, 147, 165,
171, 172, 177, 189, 191, 206, 208, 210, 211, 225,
226, 257, 258, 260, 262, 271, 281, 283, 285, 287,
288, 289, 291, 292, 294, 295, 296, 298, 299, 302,
305, 311, 312, 316, 317, 321, 322, 331
propane, 33
protein(s), viii, 15, 24, 25, 26, 28, 29, 30, 31, 42, 44,
46, 47
protocol, 32, 37, 39
prototype, 54
PTT, 367
Index 378
pulse(s), x, xi, 6, 108, 128, 131, 163, 164, 165, 166,
169, 183, 196, 232, 233, 238, 241, 243, 244, 279,
280, 281, 283, 285, 286, 287, 288, 289, 290, 291,
292, 293, 294, 295, 296, 297, 298, 299, 300, 301,
302, 303, 304, 305, 308, 311, 312, 316, 324, 331,
333
pumps, viii, 51, 53, 54, 57, 59, 60, 61, 63, 64, 65, 67,
68, 69, 70, 74, 76, 77, 78, 79, 108, 112, 221, 222,
226
purification, 273
pyrene, 31
Q
QED, 136, 139, 154
quantum phenomena, 171
quantum well, 161
quartz, 26
R
radial distribution, 153
radiation, x, 4, 7, 80, 149, 166, 171, 172, 176, 179,
198, 213, 214, 257, 258, 260, 271, 272, 275, 302,
320, 326
radio, 114
radius, 6, 17, 18, 19, 33, 121, 127, 144, 145, 146,
147, 162, 171, 172, 176, 177, 181, 182, 260, 264,
265, 269
Raman, viii, xi, xii, 51, 52, 53, 54, 55, 56, 57, 58, 59,
60, 61, 62, 63, 64, 65, 67, 68, 69, 70, 71, 72, 73,
74, 75, 76, 77, 78, 79, 80, 81, 163, 164, 165, 182,
206, 207, 218, 219, 221, 222, 227, 228, 229, 232,
275, 299, 305, 312, 316, 332, 335, 336, 342, 345,
346, 347, 349, 350, 351, 352
Raman and Brillouin scattering, 55
Raman spectroscopy, 55
Raman-scattering, 312
range, vii, x, 7, 28, 29, 30, 35, 41, 56, 58, 64, 65, 76,
84, 102, 108, 110, 132, 133, 162, 163, 166, 167,
168, 180, 182, 184, 191, 193, 207, 208, 212, 214,
215, 221, 222, 224, 226, 228, 234, 249, 257, 258,
261, 265, 266, 267, 268, 269, 271, 272, 274, 280,
285, 289, 293, 317, 319, 323, 343
reactant, 41
reactive sites, 41
reality, 56, 280, 293
reasoning, 305, 307, 310
reception, 76, 77
receptors, 48
recognition, 29, 32, 44
recombination, 175, 217, 342
recovery, 167
recurrence, 297
redistribution, 176
reduction, 40, 53, 54, 61, 131, 232, 244, 259, 262,
281, 293
redundancy, 238
reflection, vii, viii, ix, x, 7, 16, 70, 72, 83, 85, 90, 91,
93, 94, 96, 97, 98, 101, 102, 103, 105, 106, 108,
113, 136, 142, 168, 182, 189, 191, 216, 227, 257,
258, 260, 266, 316, 343, 345, 346, 350
reflectivity, 71, 85, 86, 90, 114, 172, 177, 258, 260,
261, 273, 349
refraction index, 85, 88, 89
refractive index(ices), viii, ix, 5, 6, 8, 15, 16, 17, 21,
35, 40, 42, 43, 44, 83, 84, 85, 87, 89, 121, 122,
132, 146, 168, 175, 190, 212, 213, 214, 215, 218,
220, 232, 258, 259, 260, 261, 264, 265, 266, 268,
271, 281, 343
regenerate, 36, 218
regeneration, 52
rehydration, 39
reinforcement, 356
relationship, 8, 9, 19, 146, 147, 150, 172, 173, 177,
178, 180, 199, 267, 268
relaxation, 59, 63, 175, 196, 317, 320, 364, 365
reliability, xii, 5, 355, 356, 357, 365, 366, 367
remote sensing, 44
reparation, viii, 15
resistance, 84, 258, 259, 275, 357
resolution, 61, 84, 119, 144, 162, 167, 192, 252, 272
resonator, 162, 163, 166, 172, 176, 177, 178, 179,
183
response time, 221
RF, 346
rings, viii, 51, 166, 262, 263, 264, 273, 361
RNA, 27, 32
robustness, 44, 249
Romania, 355
room temperature, 8, 36, 111, 338, 339, 365
root-mean-square, 139
roughness, ix, 161, 162, 171, 172, 176, 177, 178,
179, 180, 181, 182, 262
routines, 304
routing, 108, 205
Royal Society, 309
rubidium, 122, 134
Russia, 55
S
SA, 43, 44, 51
safety, 5, 53
Salmonella, 22, 28, 46
Index 379
sample(ing), viii, 15, 16, 17, 23, 36, 37, 39, 40, 42,
43, 44, 94, 102, 105, 154, 196, 197, 233, 237, 239,
252, 343, 356, 359, 360, 363
sapphire, 8, 131, 147, 261, 271, 333
saturation, ix, 43, 44, 54, 120, 161, 162, 168, 174,
176, 179, 221, 342
scaling, 164
scattered light, 55, 70, 132, 143
scattering, viii, ix, 28, 39, 51, 53, 54, 55, 56, 57, 62,
79, 136, 143, 151, 153, 162, 164, 174, 176, 177,
187, 189, 191, 192, 193, 194, 197, 200, 201, 219,
221, 232, 264, 275, 299, 305, 312, 316, 332, 336,
342
schema, x, 231, 233, 234, 235, 236, 237, 241, 247,
249, 254
scholarship, 79
Schrdinger equation, 279, 281, 298
science, vii, 271, 302
scientific community, 302
search, 65, 166
security, 29, 44
seed(ing), 54, 74, 165, 320
selecting, 7, 63, 235
selectivity, 32, 40, 44
self-phase modulation, 183
semiconductor, ix, xi, 4, 54, 161, 162, 168, 169, 170,
171, 172, 173, 178, 180, 182, 184, 185, 335, 336,
342, 348
semiconductor lasers, 173
sensing, viii, ix, 15, 16, 21, 30, 31, 44, 45, 46, 83, 84,
98, 104, 115, 169, 182, 205, 272, 274, 275
sensitivity, 28, 29, 35, 41, 44, 94, 96, 97, 98, 114,
162, 170, 178, 182, 217, 227, 244, 249, 350
sensors, vii, viii, ix, xi, 4, 15, 16, 21, 41, 44, 45, 46,
47, 83, 84, 93, 98, 108, 111, 115, 260, 272, 275,
335, 357
separation, 93, 98, 136, 137, 213, 293, 294, 306, 317,
324, 329, 343
sepsis, 29
series, 134, 219, 240, 268, 304, 310, 363
serum, 23, 26, 31, 45
shape(ing), 17, 57, 88, 108, 127, 146, 151, 180, 208,
219, 280
sharing, 167
shock, 275
sign(s), 57, 120, 122, 237, 238, 250, 251, 252, 253,
288, 294
signaling, 49, 232, 233, 236, 237, 243, 246, 247, 249
signals, 5, 28, 53, 54, 57, 59, 61, 62, 63, 64, 67, 68,
69, 70, 73, 77, 98, 114, 130, 134, 148, 150, 164,
187, 188, 189, 206, 207, 216, 218, 221, 226, 236,
250, 302
signal-to-noise ratio, 54, 165
silane, 32
silica, xii, 5, 6, 16, 23, 24, 25, 26, 31, 32, 48, 87, 93,
122, 182, 183, 188, 191, 199, 202, 206, 213, 215,
221, 226, 227, 257, 258, 260, 261, 262, 270, 271,
273, 316, 332, 333, 336, 355, 356, 357, 361, 364,
365, 366, 367, 368
silicon, 154, 271
silver, 6, 183, 273
simulation, viii, 15, 18, 19, 20, 21, 64, 66, 67, 68, 69,
73, 74, 77, 81, 86, 90, 92, 110, 145, 149, 152, 162,
171, 178, 294, 341
SiO2, 190
sites, 7, 29, 41, 52, 71, 216
smoothing, 366
sodium, 36
sodium hydroxide, 36
software, 33, 34
solitons, x, xi, 183, 279, 280, 281, 286, 287, 288,
289, 290, 293, 294, 297, 298, 299, 302, 303, 305,
306, 307, 311, 312, 313, 315, 316, 317, 318, 319,
320, 322, 323, 324, 325, 326, 327, 328, 329, 330,
331, 332, 333
solvent, 37
species, 28, 29, 196, 201, 280, 293
specificity, 40
spectral component, 58
spectroscopy, xi, 30, 55, 162, 271, 335, 357
spectrum, ix, 35, 56, 58, 65, 69, 70, 71, 72, 74, 76,
77, 83, 91, 92, 93, 94, 98, 101, 102, 103, 106, 112,
131, 132, 136, 137, 140, 163, 166, 167, 188, 192,
201, 216, 218, 219, 226, 270, 291, 302, 315, 316,
317, 336, 342, 348, 349, 350
speed, 18, 65, 162, 187, 188, 205, 208, 217, 227,
255, 302, 316, 361, 362
speed of light, 18, 316
spin, 357
sprouting, 22, 46
stability, x, xi, 44, 59, 68, 73, 149, 165, 167, 205,
285, 288, 291, 293, 298, 299, 309, 315, 316, 318,
320, 321
stabilization, 72, 73, 74, 75, 178, 352
stages, 247
standard deviation, 179, 180, 182
standardization, 259
standards, 68, 149
staphylococcal, 47
Staphylococcus, 22, 29, 47
statistics, 356, 358
steel, 271
sterile, 37
stimulant, 29
stock, 39
storage, 272, 316
Index 380
strain, 40, 45, 84, 92, 93, 94, 96, 97, 98, 99, 100,
101, 102, 103, 104, 107, 108, 115, 348, 350, 356
strength, xii, 7, 17, 32, 34, 58, 132, 275, 286, 355,
356, 357, 359, 360, 364, 365, 366, 367
stress, xii, 7, 87, 88, 91, 93, 94, 104, 106, 107, 109,
110, 115, 355, 356, 357, 358, 359, 360, 361, 362,
363, 364, 365, 366
stress intensity factor, 362
stretching, 259
strong interaction, 29
structural relaxation, 365
subtraction, 310
sulphur, 273
sun, 116, 183, 202, 277
supply, 44, 192, 280, 366
suppression, 162, 167, 170, 185, 338
surface layer, 43
susceptibility, 56
suspensions, 35
switching, 108, 170, 331, 341, 343, 344, 347, 351,
352
symbols, 217, 236, 237, 238, 244, 246, 247, 249,
250, 251, 253
symmetry, 211
synchronization, 163
synthesis, 32
systems, viii, ix, x, xi, xii, 51, 52, 53, 54, 55, 56, 63,
69, 70, 75, 76, 77, 78, 79, 83, 84, 108, 114, 162,
164, 165, 169, 171, 182, 187, 188, 205, 206, 207,
215, 221, 222, 231, 232, 234, 236, 238, 239, 241,
243, 244, 246, 247, 254, 255, 271, 275, 279, 280,
281, 293, 297, 298, 299, 301, 303, 305, 306, 308,
311, 312, 316, 332, 335, 336, 352, 365, 366
T
technical assistance, 366
technology, vii, viii, 3, 7, 51, 52, 55, 68, 83, 84, 108,
187, 213, 222, 232, 259, 271, 302, 356
teflon, 273
telecommunication networks, xii, 206, 218, 355, 356,
361, 365
telecommunications, x, 53, 56, 111, 162, 166, 205
temperature, viii, x, xii, 5, 7, 8, 15, 33, 36, 45, 57,
58, 84, 92, 93, 97, 98, 99, 100, 101, 102, 103, 104,
107, 108, 110, 111, 115, 140, 148, 150, 152, 166,
169, 184, 191, 206, 208, 212, 213, 214, 215, 219,
220, 226, 257, 258, 259, 260, 261, 272, 280, 338,
339, 350, 355, 356, 357, 359, 361, 362, 364, 365,
366
temperature dependence, 57, 97, 102
temperature gradient, 110
tensile strength, 359
tensile stress, 356, 357
tension, 33, 34
theory, vii, x, 3, 18, 85, 86, 121, 124, 125, 127, 144,
154, 162, 171, 249, 279, 281, 286, 291, 297, 299,
302
therapy, 31, 275
thermal expansion, 213
threat, 29, 44
threshold(s), vii, 3, 6, 8, 9, 53, 56, 72, 74, 132, 138,
139, 163, 164, 168, 170, 178, 179, 180, 181, 233
threshold level, 163
thrombin, 30, 31, 48
tics, x, 231, 233, 234, 235, 236, 237, 241, 247, 249,
254
time(ing), viii, xi, xii, 5, 8, 30, 32, 33, 37, 41, 42, 43,
45, 46, 47, 52, 57, 59, 61, 64, 65, 66, 67, 72, 84,
104, 105, 106, 108, 112, 113, 114, 115, 148, 150,
151, 152, 162, 163, 165, 170, 171, 175, 187, 188,
192, 197, 198, 199, 213, 214, 217, 221, 228, 233,
239, 244, 254, 258, 262, 280, 281, 293, 299, 301,
303, 309, 311, 317, 324, 355, 358, 359, 360, 364,
365, 366
tin, 175
TIR, 16
TNF-alpha, 27, 31
Tokyo, 158, 203
topology, 68
total energy, 320, 321
total internal reflection, vii, x, 16, 257, 258, 316
toxin, 23, 29
tracking, 183
trade, 258
traffic, 53, 187, 269
trajectory, 294, 349
transference, 81
transformation, 125, 288
transition(s), 130, 134, 138, 148, 150, 151, 164, 173,
189, 190, 197, 198, 199, 208, 245, 246, 259, 280
transition rate, 198
transition temperature, 259
translation, 93, 96
transmission, vii, viii, x, 6, 7, 16, 18, 19, 20, 21, 28,
33, 34, 35, 38, 39, 40, 41, 42, 43, 44, 45, 51, 52,
53, 54, 55, 57, 58, 63, 70, 71, 75, 76, 77, 78, 84,
114, 115, 136, 137, 138, 139, 182, 187, 188, 205,
206, 207, 208, 211, 212, 213, 215, 217, 220, 221,
223, 226, 228, 231, 232, 244, 246, 247, 255, 257,
258, 259, 260, 261, 262, 263, 264, 268, 269, 270,
271, 272, 273, 274, 275, 280, 293, 297, 298, 300,
302, 311, 343, 345, 347, 356
transmission path, 78
transmits, x, 8, 21, 232, 257
transparency, 316, 331, 332
Index 381
transport, 30, 48
transverse section, 91, 94, 105
trend, 43, 106, 130
triggers, 56
tuberculosis, 31
tumor necrosis factor, 27, 31
tumors, 275
tunneling, 176
typhoid, 31
U
UK, 81, 301
ultraviolet light, viii, 83, 85
uniform, 16, 17, 18, 21, 22, 23, 68, 85, 86, 87, 110,
152, 218, 331, 359
urine, 23
users, 112, 113, 114, 366
UV, 213, 214, 227
UV radiation, 213, 214
V
vacuum, 6, 36, 124, 128, 131, 132, 134, 142, 175,
356, 357
Vakhitov-Kolokolov criterion, 318
valence, 173
values, 19, 20, 21, 59, 60, 65, 67, 68, 70, 76, 87, 92,
96, 97, 99, 100, 101, 104, 108, 126, 139, 147, 162,
176, 178, 179, 180, 181, 200, 237, 244, 247, 248,
249, 252, 267, 273, 282, 283, 290, 291, 293, 294,
295, 308, 358, 359
vapor, 45, 131, 148, 366
variability, 214
variable(s), xi, 4, 59, 174, 220, 240, 253, 301, 302,
303, 306, 309, 312, 339, 364
variance, 240, 254
variation, xi, 92, 93, 94, 98, 142, 167, 200, 219, 226,
229, 252, 269, 287, 301, 303, 315, 362
vector, 127, 172, 173, 175, 176, 183, 210, 211, 250,
313
vehicles, 84
velocity, xi, 57, 87, 105, 133, 134, 139, 150, 163,
172, 212, 281, 291, 294, 295, 296, 297, 301, 302,
305, 316, 317, 318, 321, 322, 324, 325, 327, 329,
330, 362
vibration, 259
viscosity, 4
voice, 4, 302
W
water absorption, 224
wave number, 6, 122, 132, 304, 305, 309
wave propagation, 257, 305
wavelengths, viii, xii, 42, 51, 53, 54, 63, 64, 68, 71,
72, 75, 76, 77, 87, 89, 97, 98, 106, 107, 109, 112,
113, 114, 163, 170, 188, 193, 208, 213, 216, 220,
221, 222, 260, 261, 262, 270, 272, 335, 336, 337,
338, 339, 342, 343, 346, 347, 348, 349, 350, 352
welding, 272
wells, 177
windows, 7, 8, 53
workers, 164, 259
wound healing, 31
writing, 350
X
X-axis, 95, 97
Y
Y-axis, 95, 97, 102
yield, 304, 305
ytterbium, 164, 333

Fo Research Advances

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Fo Research Advances

Hochgeladen von

Copyright:

Verfügbare Formate

OPTICAL FIBERS

I is the steady state transmission

E-mail address: whjhe@snu.ac.kr. (Corresponding author)

(vr) sin( +) (r < a),

(wr) sin( +) (r > b),

[39]. The explicit expressions of the dispersion

) is the rst (second) kind of the (modied) Bessel functions of order , A is

angle so that the focused guiding laser

angle via total

as employed in reference [56]. The cold

is a real quantity (field overlapping integral,

N ), which is not zero just in the active region volume, and

, because it has a peak at the ring edges. This component is

(see Eqs. (10)-(11)). Therefore, a

and its nearest

, respectively. The ratio of the loss of the

, 0.13 / dB km and 0.80 / dB km ,

. In recent years, THz waveguides have been

( 0.1 3THz ) and approximately

) was obtained from a 3mm

at 83.5 m (fiber A),

at 233 m (fiber B), and

at 500 m (fiber C).

(10 m ). The absorption coefficient at 10 m wavelength is about

for polyethylene [16]. A fiber design for use in infrared is presented.

(corresponding to absorption loss

Department of Mechanical Engineering

School of Engineering, Fraser Noble Building, Kings college

() are the soliton solutions corresponding to velocity and f () and

Das könnte Ihnen auch gefallen