02whole PDF

A Study of Tellurite Glasses for
Electro-optic Optical Fibre Devices
by
Sean Manning
Supervisors:
Prof. Tanya M. Monro
Prof. Jesper Munch
A thesis submitted in fulfilment of the

degree of Doctor of Philosophy
in the
Faculty of Science
School of Chemistry & Physics
November 2011
Declaration of Authorship
I, Sean Manning, declare that this thesis titled, ‘A Study of Tellurite Glasses for Electro-
optic Optical Fibre Devices’ and the work presented in it are my own. I confirm that:
This work contains no material which has been accepted for the award of any other
degree or diploma in any university or other tertiary institution to Sean Manning
and, to the best of my knowledge and belief, contains no material previously
published or written by another person, except where due reference has been made
in the text.
I give consent to this copy of my thesis, when deposited in the University Library,
being made available for loan and photocopying, subject to the provisions of the
Copyright Act 1968.
The author acknowledges that copyright of published works contained within this
thesis (as listed below) resides with the copyright holder(s) of those works.
I also give permission for the digital version of my thesis to be made available on
the web, via the University’s digital research repository, the Library catalogue,
the Australasian Digital Theses Program (ADTP) and also through web search
engines, unless permission has been granted by the University to restrict access
for a period of time.
Signed:
Date:
iii
iv
List of Publications
1. Manning, Sean; Ebendorff-Heidepriem, Heike; Heike Monro, Tanya Mary.

Sodium Zinc Tellurite Glass: a Candidate Material for Core/Clad Fi-
bres for Electro-optic Devices.
Proceedings of the 9th Pacific Rim Conference on Ceramic and Glass Technology
(PACRIM9), held in Cairns, Queensland, Australia July 10-14 2011.
2. Manning, Sean; Monro, Tanya Mary; Munch, Jesper; Ottaway, David John.
Improved maker fringes data analysis using genetic algorithms.
Proceedings of the Australasian Conference on Optics, Lasers and Spectroscopy
and Australian Conference on Optical Fibre Technology in association with the
International Workshop on Dissipative Solitons 2009.
3. Manning, Sean; Ebendorff-Heidepriem, Heike; Monro, Tanya Mary; Munch, Jes-

per.
Tellurite glasses for photonic devices with enhanced nonlinearity.
(PACRIM8), held in Vancouver, British Columbia, Canada May 31- June 5 2009.
On the application of genetic algorithms to maker fringes analysis.
18th Australian Institute of Physics (AIP) Congress Conference, held in Adelaide,
South Australia Nov 30-Dec 5 2008..
“It doesn’t matter how beautiful your theory is, it doesn’t matter how smart you are. If
it doesn’t agree with experiment, it’s wrong.”
Richard P. Feynman
THE UNIVERSITY OF ADELAIDE
Abstract
Faculty of Science
School of Chemistry & Physics
Doctor of Philosophy
by Sean Manning
Optical fibre devices that can control light via the application of electric fields are of
enormous technological interest. These so called electro-optic devices have potential
applications in many varied places, such as data systems, pulsed lasers and sensing
technologies.
We have identified tellurium dioxide (tellurite) based glasses as being especially suit-
able for electro-optic fibre devices owing to their large nonlinear coefficients and high
crystallisation stabilities. Furthermore, tellurite glass is compatible with the extrusion
technique for producing optical fibre preforms, this being a fabrication strength of our
research group.
We developed tellurite glasses based on the general formula 10Na2 O.xMO.(90 − x)TeO2
with M=Magnesium, Zinc and Barium and x = 5, 10, 15 and 20. Raman spectroscopy
was utilised to determine the structure of the glasses under study, from which definite
compositional trends were observed. Further, we measured physical, thermal and optical
properties of these glasses that are critical for the design of electro-optical optical fibres.
Certain of these properties displayed compositional trends that were correlated with the
structural data, thus indicating physical origins for the properties. This information can
thus be used to guide future glass composition design.
We investigated thermal poling as a potential post processing technique for inducing

second order nonlinearities thereby enhancing the efficiency of the electro-optic effects.
The Maker fringes technique was applied to measuring the induced second order nonlin-
earities. We have made refinements to the standard way in which these measurements
are made, both in terms of the experimental technique as well as the analysis of the
data.
We developed computational models of optical fibres with internal electrodes for deter-
mining the properties, such as optical attenuation resulting from the presence of internal
electrodes. The results of these computations in combination with the measurements of
the glass properties are used to guide the design of prototype electro-optic fibres. Finally,
we developed various techniques for the fabrication of electro-optic fibre devices, such
as optical fibre preform extrusion, fibre drawing techniques and electrode insertion.
Acknowledgements
First and foremost I would like to thank my supervisors Tanya Monro and Jesper
Munch for their guidance, support and advice. Both have contributed heavily to form-
ing the physicist I have become. Additionally, much thanks and appreciation must go
to Heike Ebendorff-Heidepriem and David Ottaway who provided me with invaluable
co-supervision, if however unofficially, their experience and guidance was greatly appre-
ciated.
During my PhD I had the pleasure of working with so many great people, all of whom
helped me at some stage or other, whether directly or in spirit. People such as: Adrian
Selby, Aidan Brooks, Alastair Dowler, Blair Middlemiss, Bob Nation, David Hosken,
Eric Schartner, Herbert Fu, Ka Wu, Keiron Boyd, Kevin Kuan, Kristopher Rowland,
Matt Heintze, Matt Henderson, Michael Oermann, Mifta Ganja, Murray Hamilton,
Nikita Simakov, Peter Veitch, Shahraam Afshar, Trevor Waterhouse, Neville Wild,
Roger Moore and Tilanka Munasinghe.
I also had the fortune to travel far and wide to meet some fantastic collaborators such
as: Walter Margulis and Oleksandr Tarasenko at ACREO, Steve Madden, Barry Luther-
Davies and Khu Vu at ANU and Kathleen Richardson at Clemson University.
Finally, I would like to acknowledge my Wife and kids for their love and support. They
are the stabilising force in my life and I’d be lost without them.
Thank you.
viii
Contents
Declaration of Authorship iii
Abstract vi
Acknowledgements viii
List of Figures xiii
List of Tables xix
1 Introduction 1
1.1 Background and Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Review of Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.1 Tellurite Glass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.2 Electro-optics in Optical Fibres . . . . . . . . . . . . . . . . . . . . 7
1.3 Thesis Aims and Methodology . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.5 Statement of Original Work and Author Contribution . . . . . . . . . . . 13
1.6 List of Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2 Nonlinear Optical Theory 15

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 Nonlinear Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3 Second Order Nonlinearities . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.1 Propagation of the Fields . . . . . . . . . . . . . . . . . . . . . . . 21
2.3.2 Derivation of the Maker Fringes Expression . . . . . . . . . . . . . 22
2.3.3 Second Order Nonlinear Effects in Thermally Poled Materials . . . 27
2.4 Third Order Nonlinearities . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.4.1 Origin of the Nonlinear Refractive Index . . . . . . . . . . . . . . . 35
2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3 Glass Theory, Design & Fabrication 39

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2 What is a Glass? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.3 General Properties of Glasses . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.3.1 Structural Properties . . . . . . . . . . . . . . . . . . . . . . . . . . 42
ix
Contents x
3.3.2 Relationships Between the Structure and Properties . . . . . . . . 44

3.4 The Tellurite Glass System . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.5 Designing the Glass Compositions . . . . . . . . . . . . . . . . . . . . . . 47
3.6 Procedure for Glass Production . . . . . . . . . . . . . . . . . . . . . . . . 49
3.7 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4 Analysis of the Microscopic Structure of Glass 55

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2 Raman Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.2.1 Raman Spectroscopy of Tellurite . . . . . . . . . . . . . . . . . . . 58
4.2.2 Experimental Details . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.2.3 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.2.4 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5 Measurements of Physical and Thermal Properties 75

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.2 Density, Molecular Mass and Molar Volume . . . . . . . . . . . . . . . . . 75
5.3 Thermal Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.3.1 Differential Scanning Calorimetry . . . . . . . . . . . . . . . . . . . 81
5.3.1.1 Experimental Details . . . . . . . . . . . . . . . . . . . . 84
5.3.1.2 Results and Discussion . . . . . . . . . . . . . . . . . . . 84
5.3.2 Measurement of the Thermal Expansion Coefficient . . . . . . . . 88
5.3.2.1 Experimental Details . . . . . . . . . . . . . . . . . . . . 90
5.3.2.2 Results and Discussion . . . . . . . . . . . . . . . . . . . 90
5.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6 Measurements of the Optical Properties 97

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.2 UV-VIS Absorption Spectra . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6.3 Fourier Transform Infrared Absorption Spectroscopy . . . . . . . . . . . . 103
6.4 Refractive Index Measurements . . . . . . . . . . . . . . . . . . . . . . . . 108
6.5 Nonlinear Refractive Index Measurements . . . . . . . . . . . . . . . . . . 113
6.5.2 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
Contents xi
7 Thermal Poling 127

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
7.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
7.3 Experimental Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
7.3.1 Experimental Plan . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
7.3.2 Thermal Poling Apparatus and Configurations . . . . . . . . . . . 132
7.4 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
7.4.1 Charge Migration Thermal Poling . . . . . . . . . . . . . . . . . . 134
7.4.2 Charge Migration Thermal Poling Using a Blocking Electrode . . . 138
7.4.3 Charge Injection Thermal Poling . . . . . . . . . . . . . . . . . . . 141
7.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
8 Measurements of Second Order Nonlinearities 147

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
8.2 Maker Fringes Analysis: Background . . . . . . . . . . . . . . . . . . . . . 148
8.3 Experimental Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
8.3.1 Experimental Plan . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
8.3.2 Description of Maker Fringes Experiment . . . . . . . . . . . . . . 151
8.3.3 Data Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
8.3.4 Calibration and Alignment . . . . . . . . . . . . . . . . . . . . . . 155
8.4 Data Fitting Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
8.4.1 Root Mean Square Error Minimisation Fitting Procedure . . . . . 161
8.4.2 Genetic Algorithm Fitting Procedure . . . . . . . . . . . . . . . . . 161
8.5 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
8.5.1 Quartz Reference Measurements . . . . . . . . . . . . . . . . . . . 163
8.5.2 Thermally Poled Infrasil . . . . . . . . . . . . . . . . . . . . . . . . 165
8.5.3 Thermally Poled Tellurite . . . . . . . . . . . . . . . . . . . . . . . 167
8.6 Measurement Techniques for the Thickness of the Nonlinear Region . . . 169
8.6.1 SHG Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
8.6.2 Differential Etching . . . . . . . . . . . . . . . . . . . . . . . . . . 172
8.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
9 Fibre Preliminaries 177

9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
9.2 Computational Modelling of Electro-optic Optical Fibres . . . . . . . . . . 178
9.2.1 Electric Fields Between Internal Electrodes . . . . . . . . . . . . . 179
9.2.2 Electrode Induced Optical Attenuation . . . . . . . . . . . . . . . . 186
9.2.3 Discussion of Results . . . . . . . . . . . . . . . . . . . . . . . . . . 191
9.3 Insertion of Electrodes into Optical Fibres . . . . . . . . . . . . . . . . . . 191
9.3.1 The Physics of Capillary Filling . . . . . . . . . . . . . . . . . . . . 192
9.3.1.1 Contact Angle Measurements . . . . . . . . . . . . . . . . 194
9.3.2 Design and Operation of the Fibre Filling Apparatus . . . . . . . . 195
9.3.2.1 Selective Filling of Optical Fibres . . . . . . . . . . . . . 197
9.4 Optical Fibre Preform Fabrication . . . . . . . . . . . . . . . . . . . . . . 201
9.4.1 Fabrication of the Electrode Jacket . . . . . . . . . . . . . . . . . . 202
9.4.2 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
Contents xii
9.5 Preliminary Investigations of Optical Fibre Fabrication . . . . . . . . . . . 206

9.5.1 Tellurite Step Index Fibre Fabrication Experiments . . . . . . . . . 207
9.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
10 Concluding Remarks 213

10.1 Conclusion of Thesis Findings and Results . . . . . . . . . . . . . . . . . . 213
10.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
List of Figures
1.1 An assortment of microstructured optical fibres . . . . . . . . . . . . . . . 2

1.2 Optical transmission of fused silica . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Voltage dependent wave plate as an example of an electro-optic device . . 4
1.4 Number of papers published relating to tellurite glass from 1952 to 2011 . 6
1.5 Thesis structure flow diagram . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1 Second harmonic generation energy level diagram . . . . . . . . . . . . . . 18

2.2 Diagram for second harmonic generation in a dispersive medium . . . . . 22
2.3 electro-optical response of glass before and after poling . . . . . . . . . . . 29
2.4 Illustration of the symmetry possessed by the electrical potential in ther-
mally poled glass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.5 Configuration for SHG in the pp-configuration . . . . . . . . . . . . . . . 32
2.6 Third harmonic generation energy level diagram . . . . . . . . . . . . . . 34
3.1 Atomic arrangements in the crystalline and glassy state . . . . . . . . . . 40

3.2 Potential cooling curves for glass and crystal formation. . . . . . . . . . . 41
3.3 The effect of network modifying ions on the atomic arrangement in a glass 43
3.4 Structural representation of Te-O subunits . . . . . . . . . . . . . . . . . . 46
3.5 Brass moulds for glass casting and the resulting billets . . . . . . . . . . . 51
4.1 Conversion between Raman shift and Frequency (blue) and Wavelength
shift (red) for input light at 514 nm . . . . . . . . . . . . . . . . . . . . . 57
4.2 Example of deconvoluted raman spectrum . . . . . . . . . . . . . . . . . . 58
4.3 Ball and stick representation of the tellurite lattice vibrations that con-
tribute to the Raman A mode . . . . . . . . . . . . . . . . . . . . . . . . . 59
tribute to the Raman B mode . . . . . . . . . . . . . . . . . . . . . . . . . 59
tribute to the Raman C mode . . . . . . . . . . . . . . . . . . . . . . . . . 60
tribute to the Raman D mode . . . . . . . . . . . . . . . . . . . . . . . . . 60
tribute to the Raman E mode . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.8 Raman spectra for the TMN glass series . . . . . . . . . . . . . . . . . . . 64
4.9 Raman spectra for the TZN glass series . . . . . . . . . . . . . . . . . . . 64
4.10 Raman spectra for the TBN glass series . . . . . . . . . . . . . . . . . . . 65
4.11 Effect of modifier content on the relative intensities of the Raman A band
for the TMN, TZN and TBN glasses . . . . . . . . . . . . . . . . . . . . . 67
4.12 Effect of modifier content on the position of the Raman A band for the
TMN, TZN and TBN glasses . . . . . . . . . . . . . . . . . . . . . . . . . 67
xiii
List of Figures xiv
4.13 Effect of modifier content on the relative intensities of the Raman B band
4.14 Effect of modifier content on the position of the Raman B band for the
4.15 Effect of modifier content on the relative intensities of the Raman C band
4.16 Effect of modifier content on the position of the Raman C band for the
4.17 Effect of modifier content on the relative intensities of the Raman D band
4.18 Effect of modifier content on the position of the Raman D band for the
4.19 Effect of modifier content on the relative intensities of the Raman E band
4.20 Effect of modifier content on the position of the Raman E band for the
5.1 Experimental configuration for Archimedes Density Measurement . . . . . 77

5.2 Densities of TMN, TZN and TBN glass series . . . . . . . . . . . . . . . . 79
5.3 Molar masses of TMN, TZN and TBN glass series . . . . . . . . . . . . . 80
5.4 Molar volume of TMN, TZN and TBN glass series . . . . . . . . . . . . . 80
5.5 Schematic representation of a DSC apparatus . . . . . . . . . . . . . . . . 82
5.6 Example of DSC trace with key temperature features indicated . . . . . . 83
5.7 DSC traces for TMN, TZN and TBN glass series . . . . . . . . . . . . . . 85
5.8 Plot of ΔT=Tg −Tx for TMN, TZN and TBN glass series . . . . . . . . . 86
5.9 Plot of enthalpy of crystallisation for TMN, TZN and TBN glass series . 87
5.10 Schematic of apparatus for thermal expansion coefficient measurement . . 90
5.11 Example of thermal expansion measurement raw data for IG5 reference
glass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
5.12 Example of thermal expansion measurement raw data for TMN glass series 92
5.13 Example of thermal expansion measurement raw data for TZN glass series 92
5.14 Example of thermal expansion measurement raw data for TBN glass series 93
5.15 Thermal expansion coefficients for TMN, TZN and TBN glass series . . . 94
6.1 Proposed energy band diagram for tellurite glass . . . . . . . . . . . . . . 101

6.2 Examples of indirect band gap fitting procedure for the TZN series . . . . 102
6.3 Compositional dependence of indirect band gap for TMN, TZN and TBN
glass series. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
6.4 Infrared absorption spectrum of TMN1 and theoretical multiphonon edge 105
6.5 Compositional dependence of multiphonon onset wavelength for TMN,
TZN and TBN glass series . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
6.6 Compositional dependence of attenuation due to OH− contamination . . . 107
6.7 Ray diagram for an optical fibre . . . . . . . . . . . . . . . . . . . . . . . 108
6.8 Prism coupling configuration . . . . . . . . . . . . . . . . . . . . . . . . . 109
6.9 Example trace from a prism coupler . . . . . . . . . . . . . . . . . . . . . 111
6.10 Refractive index of tellurite glasses measured at 1064 nm . . . . . . . . . 112
6.11 Simple Z scan configuration . . . . . . . . . . . . . . . . . . . . . . . . . . 114
6.12 Examples of Z scan signals for self focusing and self defocusing . . . . . . 115
List of Figures xv
6.13 Schematic of Z-scan experimental setup . . . . . . . . . . . . . . . . . . . 116

6.14 Overlap of Gaussian with truncated Bessel function . . . . . . . . . . . . . 117
6.15 Representative Z scan data set . . . . . . . . . . . . . . . . . . . . . . . . 118
6.16 Z Scan measurements at various incident intensities . . . . . . . . . . . . 119
6.17 Nonlinear refractive indices of Tellurite glasses . . . . . . . . . . . . . . . 121
6.18 Millers law plot with various representative glass types indicated along
with TMN, TZN and TBN glasses . . . . . . . . . . . . . . . . . . . . . . 124
7.1 Schematic of the key steps in the charge migration thermal poling process 128
7.2 Schematic of the key steps in the charge injection thermal poling process . 129
7.3 Thermal poling apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
7.4 Ohms law plot for TZN2 at 250◦ C . . . . . . . . . . . . . . . . . . . . . . 136
7.5 Time dependence of the current through TZN2 at 250◦ C . . . . . . . . . . 136
7.6 Optical micrograph of the cathodic surface of TZN2 after thermal poling . 137
7.7 Reproduction of an optical micrograph of the chemically reduced region
at the cathodic surface of a Bismuth borate glass. . . . . . . . . . . . . . . 138
7.8 Two possible configurations for the blocking electrode thermal poling
technique. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
7.9 Time dependence of the current for TZN2 in the two blocking electrode
configurations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
7.10 Charge injection thermal poling configuration . . . . . . . . . . . . . . . . 142
7.11 Time dependence of the current for TZN2 in the charge injection config-
urations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
7.12 Schematic illustration of the electrode configuration for thermal poling of
optical fibres . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
7.13 Simulated equipotential maps and ion distributions for fibres with two
anodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
8.1 Experimental configuration for Maker et al. original Maker fringes exper-
iment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
8.2 First recorded Maker fringes . . . . . . . . . . . . . . . . . . . . . . . . . . 149
8.3 Simulated Maker fringes with the inclusion of higher order interference
and without . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
8.4 Maker fringes measurement apparatus . . . . . . . . . . . . . . . . . . . . 152
8.5 Circuit diagram for the Maker fringe data conditioning and acquisition
system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
8.6 Timing diagram for the Maker fringes experiment data acquisition system 156
8.7 Photomultiplier signal linearity plot . . . . . . . . . . . . . . . . . . . . . 157
8.8 Relative error in the Maker fringes data vs number of pulses averaged over158
8.9 Plot of SH power vs lens to sample separation . . . . . . . . . . . . . . . . 159
8.10 Optimised fit to y-cut quartz Maker fringes data . . . . . . . . . . . . . . 165
8.11 Convergence of Maker fringes fitting parameters for thermally poled Infrasil166
8.12 Optimised fit to thermally poled Infrasil Maker fringes data . . . . . . . . 168
8.13 Measured angular dependence of second harmonic power for thermally
poled TZN3 sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
8.14 Schematic of SHG microscopy sample, optical micrograph of the region
imaged in SHG micrograph and SHG micrograph . . . . . . . . . . . . . . 171
List of Figures xvi
8.15 Mean line scan of the SHG channel taken from an SHG micrograph of
thermally poled Infrasil . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
8.16 SEM of the etched depletion region of thermally poled Infrasil . . . . . . . 173
8.17 SEM of the anodic face of thermally poled infrasil . . . . . . . . . . . . . 173
8.18 Overlay of SON profile with SEM of depletion region . . . . . . . . . . . . 174
8.19 Second order nonlinear susceptibility as a function of depth under the
anodic surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
9.1 Comparison between experiment and simulation for a poled SIF and the
modelled electric field distribution . . . . . . . . . . . . . . . . . . . . . . 181
9.2 Modelled electric field distribution for a step index fibre with internal
electrodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
9.3 Modelled electric field distribution for a hexagonal three ring MOF with
internal electrodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
9.4 Modelled electric field distribution for a wagon wheel MOF with internal
electrodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
9.5 Core electric field strengths for three representative fibre types with in-
ternal electrodes calculated over a range of electrode separations . . . . . 186
9.6 Optical fibre with internal electrodes . . . . . . . . . . . . . . . . . . . . . 187
9.7 Convergence of electrode induced loss . . . . . . . . . . . . . . . . . . . . 188
9.8 Electrode induced loss vs. electrode to core separation . . . . . . . . . . . 189
9.9 Electrode induced loss for a wagon wheel MOF calculated for various
mesh element sizes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
9.10 Schematic representation of pressure assisted filling of capillaries for wet-
ting and nonwetting liquids . . . . . . . . . . . . . . . . . . . . . . . . . . 192
9.11 Illustration of the contact angles between a liquid on a solid substrate . . 193
9.12 Photographs of BiSn contacting silica, bismuth and tellurite substrates . . 195
9.13 Photograph of fibre filling apparatus . . . . . . . . . . . . . . . . . . . . . 196
9.14 Schematic of filling apparatus showing fibre positioning and sealing tech-
nique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
9.15 Illustration of manual fibre hole blocking . . . . . . . . . . . . . . . . . . . 198
9.16 Photograph of an optical fibre with a manually blocked hole . . . . . . . . 199
9.17 Photograph of an optical fibre with pressure assisted filling of UV glue . . 199
9.18 Illustration of UV glue pressure filling set up . . . . . . . . . . . . . . . . 200
9.19 Photograph of an optical fibre prepared for selective filling demonstrating
the ideal cleave position . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
9.20 Photograph of the cross section of an optical fibre with one hole blocked
for selective filling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
9.21 Photograph of an optical fibre with a selectively filled electrode hole . . . 201
9.22 Schematic of the optical fibre preform extrusion apparatus . . . . . . . . . 202
9.23 Illustration of the extruded jacket and core for creating the preform for a
step index optical fibre with internal electrodes . . . . . . . . . . . . . . . 203
9.24 Die for electrode jacket preform . . . . . . . . . . . . . . . . . . . . . . . . 203
9.25 Electrode jacket preform extrusion trials . . . . . . . . . . . . . . . . . . . 204
9.26 Linear fit to TZN refractive index data . . . . . . . . . . . . . . . . . . . . 208
9.27 Cut back measurement data showing optical attenuation of a core glass
bare fibre . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
9.28 Microscopy images of tellurite step index fibre . . . . . . . . . . . . . . . . 210
List of Figures xvii
9.29 Cut back measurement data showing optical attenuation of a tellurite

step index fibre . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
List of Tables
3.1 Deitzel’s Field Strength Parameters for Glass Formation . . . . . . . . . . 45

3.2 Modifier properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.3 Purity and supplier for glass raw material . . . . . . . . . . . . . . . . . . 50
3.4 Table of glass compositions investigated. . . . . . . . . . . . . . . . . . . . 52
4.1 Comparison of certain properties of Quartz and Fused Silica . . . . . . . . 56

4.2 Raman band assignment to structural subunits and nomenclature . . . . . 61
4.3 Table of Raman band relative intensities. . . . . . . . . . . . . . . . . . . 66
4.4 Table of Raman band centre positions. . . . . . . . . . . . . . . . . . . . . 66
5.1 Molar masses, Densities and molar volumes of tellurite glasses. . . . . . . 78

5.2 Thermal data for TMN, TZN and TBN glass series . . . . . . . . . . . . . 86
5.3 Measured thermal expansion coefficients for the TMN, TZN and TBN
glass series. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.1 Optical energy gaps calculated from UV-Vis spectra and corresponding
wavelength cut off for optical transmission. . . . . . . . . . . . . . . . . . 103
6.2 Multiphonon onset wavelength for TMN, TZN and TBN glass series. . . . 106
6.3 Measured refractive indices . . . . . . . . . . . . . . . . . . . . . . . . . . 111
6.4 Measured nonlinear refractive indices and calculated third order suscep-
tibilities for TMN, TZN and TBN glasses . . . . . . . . . . . . . . . . . . 120
7.1 Brief summary of literature reported tellurite thermal poling . . . . . . . 131
8.1 Optical parameters for quartz reference sample . . . . . . . . . . . . . . . 164

8.2 Optimised fitting parameters for y-cut quartz Maker fringes . . . . . . . . 164
8.3 Optimised fitting parameters for thermally poled Infrasil Maker fringes . . 167
9.1 Input parameters for SIF with internal electrodes electric field model . . . 180
9.2 Input parameters for hexagonal three ring MOF with internal electrodes
electric field model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
9.3 Input parameters for wagon wheel MOF with internal electrodes electric
field model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
9.4 Input parameters for electrode induced loss model . . . . . . . . . . . . . 187
9.5 Contact angle between BiSn solder and some optical glasses . . . . . . . . 194
xix
For my wife, who deserves a PhD in patience.
xxi
Chapter 1
Introduction
P
hotonics research has, over the last several decades, been characterised by a pro-
gressive shift from free space optical components towards fibre optical equivalents.
This shift has been driven by the desirable features of optical fibres, namely; compact-
ness, ruggedness, environmental insensitivity and, of particular importance, low cost. In
the spirit of this trend the primary focus of the work reported herein is to develop a new
class of fibre optical component to replace existing free space electro-optical devices.
This will further facilitate the advancement of optical systems towards true all-fibre
equivalents.
In the following sections we outline our motivation for undertaking this work this is given
context with a discussion of the background concepts (Section 1.1). We then provide
a review of the relevant literature pertaining to tellurite glass and electro-optical fibres
(Section 1.2). For clarity, our aims and research methodology are explained in Section
1.3 followed by an outline of the structure of this thesis in Section 1.4. This Chapter
concludes with a statement outlining those aspects of the work contained within this
thesis that were produced by the Author (Section 1.5).
1.1 Background and Motivation
A
revolution in optical fibre technology occurred in 1996 with the work of Birks et al
with the development of a new generation of optical fibres known as microstruc-
tured optical fibres1 [1, 2]. Microstructured optical fibres contain an array of small holes
that run the length of the fibre and reside in the region where the light is guided. These
holes enable the fibre to guide light via effective index guidance or photonic band gap
effects [3–5]. Additionally, size, shape and distribution of these internal holes offer fine
1
Sometimes referred to as photonic crystal fibres (PCF). We restrict the use of this name to those
fibres which guide light via photonic band gap effects and instead refer generally to any fibre with
internal structure as a MOF.
1
2 Introduction
control over the guidance properties, thus providing a versatile platform for the design
of devices [6]. Figure 1.1 shows some examples of some in-house microstructured optical
fibres, in particular, these images give an indication of the diversity in micorstructures
mirroring the diversity in guiding properties2 .
Figure 1.1: An assortment scanning electron microscope images of various microstruc-

tured optical fibres made at The University of Adelaide. (A) Air core photonic band
gap fibre. (B) Hexagonal array effective index guiding fibre. (C) Dispersion engineered
fibre. (D) Large suspended core fibre. (E) Large mode area fibre. (F) Air core photonic
band gap fibre. (G) Large suspended core fibre. (H) Exposed core wagon wheel fibre.
(I) Small core wagon wheel fibre.
To date, the vast majority of optical fibres found in commercial devices have been made
from silica (SiO2 ) [7]. The ubiquity of silica fibres is a result of the many desirable prop-
erties possessed by silica. For example, silica is transparent between, approximately,
200 nm and 2000 nm as shown in Figure 1.2. It should be noted that in Figure 1.2 the
reduction of the transmission which results from Fresnel reflection at the interfaces has
2
Acknoledgments for these fibres is as follows: (A) Kristopher Rowland, Heike Ebendorff-Heidepriem,
Roger Moore. (B) Roger Moore and Heike Ebendorff-Heidepriem and Asahi Glass Co. (for sup-
plying the bismuth glass) (C) Wenqi Zhang, Heike Ebendorff-Heidepriem, Roger Moore. (D) Heike
Ebendorff-Heidepriem, Roger Moore. (E) Heike Ebendorff-Heidepriem, Roger Moore. (F) Kristopher
Rowland, Heike Ebendorff-Heidepriem, Roger Moore. (G) Heike Ebendorff-Heidepriem, Roger Moore.
(H) Stephen Warren-Smith, Heike Ebendorff-Heidepriem, Roger Moore. (I) Heike Ebendorff-Heidepriem,
Roger Moore.
Chapter 1 3
not been removed. This transparency window makes silica an ideal transmitter of many
existing commercial laser sources. Mechanically, silica is very strong and fibres fabri-
cated from silica show exceptional resilience against tensile loads and bending. Another
factor contributing to the near ubiquity of silica fibres is the maturity of the processing
technologies surrounding their manufacture. The enormous quantities of money spent
by the telecommunications industry towards the end of the 20th century have produced
extremely well refined silica glass with optical absorptions below 0.2 dB.km−1 at 1550 nm
which is near to the theoretical limit [8, 9]. Additionally, the processing techniques for
producing the preforms, and ultimately the optical fibres themselves, have been opti-
mised.
NOTE:
This figure is included on page 3 of the print copy of
the thesis held in the University of Adelaide Library.
Figure 1.2: Optical transmission of fused silica. Note: Fresenel losses are included in
the data. Reproduced from [10]
Despite its ensemble of excellent physical and optical properties, silica has limitations.
Most notably, in the context of this work, the nonlinearity of silica is inherently low
thus making it a less than ideal material for nonlinear applications such as electro-
optics. Additionally, fused silica has strong absorption features beyond approximately
2 μm (see Figure 1.2) that preclude its use for mid IR transmission. There is, therefore,
a need to develop alternative materials to circumvent the shortcomings of silica based
optical fibres.
A plethora of potential glass compositions exist that could, in principle, pave the way
for efficient, low cost electro-optic fibre devices. The search for such materials is well
documented in the literature. One of the most promising candidate materials is the
tellurium dioxide (TiO2 ) based family of glasses, collectively known as tellurite glasses
[11]. Tellurite glasses are amongst the most nonlinear oxide materials known, as well as
possessing optical transmission up to the mid IR region of the electromagnetic spectrum,
as high as 5 μm.
4 Introduction
The class of optical device we are investigating are the so called electro-optical devices.
electro-optics is the control of the optical properties of a material via the use of external
electric fields. Typically, this involves a change in the birefringence of the material which
in turn produces a polarisation dependent phase change in the the light passing through
the material. Figure 1.3 illustrates this with a specific example of a voltage dependent
wave plate. Light linearly polarised at 45◦ to the vertical passes through a material with
a voltage applied across it. When the correct voltage is applied a relative phase difference
of π/4 is produced between the two orthogonal polarisation states, thereby generating
circularly polarised light. Such a devices are used in optical isolators, Q-switches and a
myriad other applications.
Figure 1.3: The voltage dependent wave plate as an example of an electro-optic device.
Light linearly polarised at 45◦ to the vertical passing through a material with electrodes
connected to opposing faces. The strength of the applied field is sufficient to produce a
relative phase difference of π/4 between the vertical and horizontal components of the
lights polarisation. The light exiting from the material is now circularly polarised.
The two electro-optic effects of practical importance are the Kerr effect and the Pockels
effect. The Kerr effect involves a change in refractive index of a material that is pro-
portional to the square of an applied electric field. This is sometimes referred to as the
quadratic electro-optic effect3 . The Pockels effect, or linear electro-optic effect provides
change in refractive index is linearly dependent on the applied electric field.
The Kerr effect is related to the third order nonlinear susceptibility, χ(3) , of the material.
All materials have non-zero χ(3) and therefore display the Kerr effect. On the other hand,
the Pockel’s effect is determined by the second order nonlinear susceptibility, χ(2) , which
3
To be specific we are referring to the effects in which the applied electric field is external. As opposed
to the situation when the field responsible for inducing the modifications to the refractive index is the
optical field of the light propagating through the material
Chapter 1 5
is only possessed by certain materials. For glass χ(2) = 0. It is shown in Chapter 2 that
any material with inversion symmetry (in particular glass) has this property.
It is always the case that χ(2) effects in a material will be of a larger magnitude, and
therefore more efficient, than χ(3) effects. It is for this reason that χ(2) materials are
desirable for device manufacture.
The quadratic electro-optic effect can be exploited in optical fibres, indeed, this has
been reported in the literature. To do this, however, large electric fields (in the order
of 108 V.m−1 ) are required to produce even modest phase changes. Devices utilising
electric fields of this magnitude are susceptible to failure via dielectric break down and
consume large amounts of electrical power.
A much better solution resides in the development of various processing techniques that
in some way or other destroy the (macroscopic) inversion symmetry of the glass. In
this thesis we focus solely on thermal poling as a means for achieving this. During the
thermal poling process the glass is heated to enhance the mobility of various ionic species
within, then under the application of strong electric fields these ions are encouraged to
migrate in one preferred direction. With the field still applied the sample is cooled to
room temperature thus leaving a nonequilibrium charge distribution which produces a
n electrostatic field within the glass. Through rectification of the intrinsic third order
nonlinearity by the electrostatic field an effective second order nonlinearity is generated.
In this situation even a modest external electric field can produce phase shifts sufficient
for any practical purpose.
1.2 Review of Literature
I
n the following review of literature we examine the key developments in the two
areas of research of direct relevance to the overall aims of this thesis. These areas are
tellurite glass and electroopitcal optical fibre devices. This review is based on information
obtained largely from the Scopus citation database [12].
1.2.1 Tellurite Glass
The first recorded investigation of the glass forming ability of TeO2 was made in 1834
by Berzelius [13]. Berzelius mainly looked at binary glasses of TeO2 and BaO but
not in the context of optical materials. A proper, systematic study of tellurite glasses
was not begun until the work of Stanworth in 1952 [14]. Stanworth considered the
glass forming ability of TeO2 in light of recent advances in the understanding of the
6 Introduction
factors that determine the glass forming ability of oxides. He reasoned that tellurium
oxide should behave analogously to phosphorous oxide as tellurium and phosphorus have
very similar electronegativities. This paper is significant as it was the first time that
the technological significance of tellurite glasses was noted. Stanworth discovered that
tellurites possess, relative to other glasses known at the time, large thermal expansion
coefficients, large refractive indices and low softening points. The latter two properties
are of particular relevance to the work in this thesis. These initial discoveries prompted
others to investigate this intriguing material in a technological context.
Following the work of Stanworth the research in tellurite glasses has been an extremely
active area. Figure 1.4 demonstrates this by showing the number of published papers
from 1952 to the present (mid 2011). This data was compiled using the Scopus citation
database with “tellurite glass” as the search term and searching within article title,
abstract and key words.
150
125
Number of Published Papers
100
75
50
25
0
1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015
Year
Figure 1.4: Chart of the number of papers published relating to tellurite glass from
1952 to 2011 [12].
An important result was established in 1985 when it was confirmed that many of the
glasses reported in the literature up to then were contaminated with material from the
crucible they had been melted in, usually SiO2 (silica) or Al2 O3 (alumina) [15]. These
impurities were found in concentrations as high as 6 mole% which has a large effect on
the properties of the resulting glasses. It is now common place to use gold or platinum
crucibles as these materials are much less soluble in the molten solutions.
Tellurite was first recognised as a potential material for optical fibres in 1976 in the
context of a host for lasing ions [16]. Tellurite was noted to be favourable relative to
silicate glasses for its spectroscopic parameters such as the florescence life times. In 1985
Chapter 1 7
El-Zaidia et al studied the infrared transmission of certain tellurite glasses and noted
that the low absorption in spectral region beyond 2.5 μm would make these glasses
desirable for low loss mid infrared optical fibres [17].
To date the most comprehensive review of literature is the Tellurite Glasses Handbook
Physical Properties and Data by El-Mallawany [18]. This volume contains a vast amount
of data on many tellurite glass systems as well as details relating to the measurement of
these properties. Much of the fundamental understanding of the physical mechanisms
for the origins of these properties is absent from the book and therefore the interested
reader needs to pursue the referenced papers. Since the publication of El-Mallawany’s
handbook (2002) there has been almost 1000 publications on tellurite glass, essentially
twice as many as existed at the time of publication. Since 1984 there has been over 350
patents filed relating to technologies based on tellurite glass.
1.2.2 Electro-optics in Optical Fibres
The history of electro-optical fibres begins in 1986 with the first report of the generation
of a permanent second order nonlinearity χ(2) , in glass by Österberg and Margulis [19].
At the time it was conjectured that this could be a result of the forced migration of
charged species in the glass and/or the reorientation of dipoles within the glass. Follow-
ing this discovery Myers et al deliberately caused ions to migrate under the influence
of an applied electric field using heat to increase the mobility of the ions [20]. This
processing technique was later called ‘thermal poling’ (often just ‘poling’) due to the
requirement of heat and the resulting polarity produced by the separation of charged
species. This is not to be confused with periodic poling which is a technique for quasi-
phase matching in nonlinear crystals to increase the efficiency of the interaction4 .
It is now understood that the mechanisms for the poling produced by optical fields alone
(often referred to as ‘optical poling’) and those for thermal poling are subtly different
[21–23]. Regardless, the hypothesis that it was resulting from charge migration and
the subsequent experimental work lead to the discovery that permanent second order
nonlinearities can be generated in this way.
Having a technique for creating a second order nonlinearity in glass is a significant

technological achievement as it paves the way for the fabrication of low cost electro-
optical components for telecommunications, industry and science. Armed with this new
information researchers began to investigate two separate directions. One focused on
investigating the effects of this process on glasses other than silica. The other sought
4
In some cases fibres have been thermally poled in a periodic manner, thus combining both techniques.
8 Introduction
to generate second order nonlinearities the region in which the light is guided within an
optical fibre.
In terms of the non-silica glass poling the breadth of the work is very large. In general,
it has been demonstrated that there is typically several ionic species that undergo mi-
gration as well as the impregnation of ionic species from the air into the glass. Further,
the conductive nature of the glass plays an integral role. That is, whether or not the
glass is a pure ionic conductor, pure electronic conductor or some combination of both
[24].
The work towards thermal poling of optical fibres has to this point been done exclusively
in silica based fibres. This is primarily due to the maturity of the fabrication technologies
surrounding silica fibres.
In 2001 a European consortium involving nine partners was established to develop and
fabricate electro-optic devices in the optical fibre and planar wave guide platforms [25].
The consortium known as GLAMOROUS (GLAss based MOdulators, ROUters and
Switches) investigated a large variety of materials and processes over the four year
duration of the project. Several techniques and devices that were not originally con-
templated by the GLAMORUS consortium were developed, such as X-ray poling and
in-fibre lithography, thus opening up many new research avenues [25].
Significantly the GLAMOROUS work resulted in the fabrication of an optical fibre

electro-optic switch. This device was constructed from silica fibre with internal electrodes
which were used first to pole the fibre then provide the modulating electric field for the
electro-optic interaction. Their work with non-silica glasses was not translated into an
optical fibre platform despite the fact that several of these glasses exhibited χ(2) values
at least an order of magnitude larger than silica. Again, it should be noted that this
is primarily due to the relative infancy of the fabrication processes required to produce
optical fibres from these materials. To date there have been no reports of such a device
being constructed from glasses other than silica.
At the conclusion of the GLAMORUS project it was noted that the fundamental limi-
tation to the efficiency of the devices was the low values of the induced χ(2) [25]. This
result provided us with the impetus for the work in this thesis, where our primary fo-
cus was to develop high efficiency electro-optical fibre devices fabricated from highly
nonlinear glasses.
Chapter 1 9
1.3 Thesis Aims and Methodology
A
t the outset of this work our primary objective was to develop an electro-optic fibre
device, such as a phase modulator, from a glass with a high intrinsic nonlinearity.
In Chapter 10 we present a post mortem in which the reasons for not meeting this
primary objective are discussed. Despite not producing a device there were many sub-
goals that had to be met in order to achieve the primary objective and much progress
was made in these areas. Listed below are the sub-goals that comprise this work:
1. To develop a range of tellurite glasses and characterise their properties so that in-
formed decisions can be made regarding the choice of glass. Adding to the collective
understanding of this glass system and the origin of its properties this knowledge
can be applied to the design of devices and new glasses.
2. To identify suitable tellurite glasses for nonlinear optical fibre devices.
3. To generate permanent second order nonlinearities in some of these glasses via

thermal poling, thus making possible efficient electro-optical devices for controlling
the passage of light in optical systems.
4. To design and fabricate prototype electro-optical devices from in house fabricated

optical fibres.
The way we approached the problem was to begin with a well known ternary tellurite
glass: 5Na2 O.20ZnO.75TeO2 (TZN), which has been demonstrated to be suitable for
the fabrication of optical fibres [26], and explore compositional variations of this system.
We then sought to measure as many of the material properties as possible, given the
availability of equipment. Next, we looked for trends and correlations in the properties
that could shed insight into the underlying mechanisms for such properties. Further to
this, we use the observed data and trends to identify candidate compositions for optical
fibre fabrication.
Some of the more promising glasses were investigated for their suitability for post pro-
cessing. We explored the possibility of inducing permanent second order nonlinearities
via thermal poling.
In parallel to the glass development section of this work we investigated the process-
ing techniques required to produce the fibre optic devices. These include methods for
providing the fibres with internal hole for later insertion of electrodes. Techniques for
inserting the electrodes were learnt during a visit with collaborators in ACREO Swe-
den. These were generalised sufficiently to be suitable for the application of providing
microstructured optical fibres.
10 Introduction
To compliment the practical side of the fibre development we produced some compu-
tational models to guide the design and fabrication of the fibres for future optimised
devices.
1.4 Thesis Outline
T
o aid the reader the following is an outline of the structure of this thesis. To aid the
discussion we refer to the flow chart in Figure 1.5 which illustrates the structuring
of the Chapters and the way information flows from certain Chapters into others.
Chapter 2
The focus of our glass making efforts is to produce suitable materials for nonlinear
fibre optic devices. We therefore discuss some important aspects of nonlinear
optical theory. We cover the general theory of second harmonic generation and then
apply it to the specific case of thermally poled glass. The Chapter concludes with
a derivation of the Maker fringes equation which used to analyse the magnitude
of the induced nonlinearities.
Chapter 3
In this Chapter we present a detailed discussion of the main concepts underlying
the glassy state: What is a glass? How are they made and what general properties
they possess? This is followed by a discussion of family of glasses under study in
this thesis, namely tellurite glass.
Following this general discussion we then document the details concerning the
design and fabrication of the glasses that were used in this work. We describe
the process for the design of the compositions, including; the choice of chemical
components and the proportions thereof. The procedure for glass fabrication is
presented followed by the results of our trials.
Chapter 4
In Chapter 4 we present our investigations of the structure of the glasses under
study. Specifically, we have determined the arrangement of chemical species into
molecular groupings, the type of bonding within and between these groupings
and how this varies for differing compositions is investigated. The trends that we
discovered are continuously referred to in subsequent sections as it is the structure
of the glass and the components from which it is made that largely determine these
properties.
Chapter 1 11
Figure 1.5: Flow diagram for the structure of this thesis. As indicated by the legend in
the bottom left, the chapters are split into three main sections: Introductory material,
Experimental work and Concluding material. The chapter order is indicated by thick
arrows. Information flowing from one chapter to another is indicated by the arrows.
12 Introduction
Chapter 5
We have measured various physical and thermal properties of the glasses which we
then correlate to the structure via Chapter 4. The measured properties include:
Density, molar volume, the glass transition and crystallisation temperatures and
the coefficient of thermal expansion. The measured properties give us useful infor-
mation about the suitability of these glasses for the use as an optical fibre material.
Chapter 6
Measurements of the optical properties of the glasses are presented in Chapter 6.
We have determined the transmission and absorption spectra as well as the linear
and nonlinear refractive indices. There are correlations between these properties
and the trends displayed by the structure of the glasses and the components from
which they are made. We provide a detailed discussion of the significance and
interpretation of these correlations.
Chapter 7
In Chapter 7 we present a review of thermal poling techniques and present our
experiments to induce second order nonlinear properties in the glass samples.
Chapter 8
The analysis of the second order nonlinearity produced in the glasses that were
thermally poled is preformed in this Chapter. The details of the experimental
apparatus that was constructed to measure these properties are set out with par-
ticular attentions to the factors that guided the design and calibration of the ap-
paratus. Data analysis of the measurements is considered in great detail, with two
methodologies presented and the pros and cons of each discussed. Also presented
are some complementary experimental techniques for determining the properties
of the thermally poled materials. The Chapter is concluded with a discussion of
the results.
Chapter 9
This Chapter is an eclectic collection of: simulations, fabrication techniques and
design stages for future nonlinear optical fibre devices. We have performed simula-
tions in order to gain a qualitative understanding to guide the design of potential
nonlinear optical fibre devices. Several of the critical fabrication steps necessary
to produce devices have been explored and tested. The design of these steps and
preliminary results are discussed. We conclude with a summary and make some
suggestions for future work.
Chapter 10
Finally, we make concluding remarks concerning the totality of the work presented
in this thesis and suggest the future directions.
Chapter 1 13
1.5 Statement of Original Work and Author Contribution
T
he following is a declaration of the aspects of this work that are, to the best of
our knowledge, original and the contributions to this work by the Author. This
discussion is broken down into Chapters:
Chapter 2
The Author derived the Maker Fringes expression and reduced the nonlinear tensor
elements for C∞ ν symmetry.
Chapter 3
All of the composition design and glass fabrication was performed by the Author.
Chapter 4
Analysis of the Raman spectral data and interpretation thereof was undertaken
by the Author.
Chapter 5
All measurements and the subsequent data analysis were made by the Author
with the exception of the coefficient of thermal expansion. These measurement
were made by a student under direct supervision of the Author. The design of the
equipment and the experimental technique was performed by the Author as too
was the data analysis.
Chapter 6
All measurements and the subsequent data analysis were made by the Author.
Chapter 7
Equipment for thermal poling was designed and fabricated by the Author.
Chapter 8
The Maker fringes experiment was constructed by the Author with the exception
of the electronics for the data acquisition, which was designed in consultation with
Mr. Neville Wild and Mr. Robert Nation, who also built the equipment. Mr.
Neville Wild also provided input into the construction of the LabView control
software.
Chapter 9
The computational models for calculating the DC fields in the electrode containing
fibres was written in Matlab using the COMSOL software package by the Author.
Kristopher Rowland wrote the code upon which the electrode loss simulations were
based with additional modifications made by the Author.
14 Introduction
The design of the fibre filling apparatus was designed by the Author with close
consultation with colaborator Walter Margulis at ACREO in Stockholm. It’s fab-
rication was performed by the Author.
Fibre filling experiments were performed by the Author and an Honours student,
Hashan Tilanka Munasinghe, under his tutelage. So too was the preform fabrica-
tion.
Design of the optical fibres was performed by the Author. The author assisted in
the fabrication of these fibres.
1.6 List of Publications
1. Manning, Sean; Ebendorff-Heidepriem, Heike; Heike Monro, Tanya Mary.

Sodium Zinc Tellurite Glass: a Candidate Material for Core/Clad Fi-
bres for Electro-optic Devices.
(PACRIM9), held in Cairns, Queensland, Australia July 10-14 2011.
Improved maker fringes data analysis using genetic algorithms.
Proceedings of the Australasian Conference on Optics, Lasers and Spectroscopy
and Australian Conference on Optical Fibre Technology in association with the
International Workshop on Dissipative Solitons 2009.
3. Manning, Sean; Ebendorff-Heidepriem, Heike; Monro, Tanya Mary; Munch, Jes-

per.
Tellurite glasses for photonic devices with enhanced nonlinearity.
(PACRIM8), held in Vancouver, British Columbia, Canada May 31- June 5 2009.
On the application of genetic algorithms to maker fringes analysis.
18th Australian Institute of Physics (AIP) Congress Conference, held in Adelaide,
South Australia Nov 30-Dec 5 2008..
Chapter 2
Nonlinear Optical Theory
2.1 Introduction
A
s the focus of this thesis is on the the elaboration of experimental tellurite glasses
and their suitability for base materials for nonlinear optical fibres, it is pertinent
to discuss the physics of nonlinear optics. Accordingly, in the following chapter there is a
review of some key elements of the theory of nonlinear optics with particular focus on the
phenomena of nonlinear refractive index and second harmonic generation. In addition to
the review there are some derivations and calculations that infrequently appear in texts
and literature and which were performed by the author in order to properly understand
the origin of key formulae and results. Such as the specific form of the effective second
order nonlinear susceptibility matrix for thermally poled glass and the origin of the
Maker fringes expression.The information in this chapter is of direct relevance to the
material in Chapters 6 and 8.
2.2 Nonlinear Optics
T
he optics of our everyday experience, reflection, refraction and scattering, are linear
optical phenomena. This is because the strength of the electric field for sun light,
light bulbs and flames are weak with respect to the binding strength of electrons and
nuclei, which is ≈ 1010 V.m−1 . This was, in general, the case until the invention of the
laser by Theodore Maiman [27]. Pulsed lasers can produce very strong electric fields in
the order of 107 V.m−1 , in a collimated beam1 . Focusing this, already intense radiation
down to a spot of approximately λ in diameter will increase the electric field strength
to around ≈ 1010 V.m−1 . If this radiation is incident on a material the electric field of
1
e.g a fairly typical Nd:YAG laser can produce a 1 mm diameter beam with 1 mJ Q-switched pulses
of 1 ns duration. The peak power will be 1 MW and the peak electric field strength will be 107 V.m−1
15
16 Nonlinear Optical Theory
the light is now sufficiently strong to perturb the electrons from their host nuclei and
produce new and interesting optical effects.
Unsurprisingly, shortly after the invention of the laser Peter Franken et al [28] observed
the first nonlinear optical effects and today nonlinear optics is arguably one of the most
active areas of optical research.
To understand why such high field strengths are required and specifically what is occur-
ring in materials when they are being influenced by these fields we will need to consider
the polarisation P̃(t), of the material in the presence of the optical field Ẽ(t). In the
linear regime we can describe the relationship between polarisation and optical field as:
P̃(t) = 0 χ(1) Ẽ(t), (2.1)
where 0 is the free space permittivity and χ(1) is the linear susceptibility which, for
transparent materials, is related to the refractive index n0 , by
χ(1) = n20 − 1 (2.2)
and therefore to the relative permitivity r via
r = 1 + χ(1) (2.3)
Therefore χ(1) determines much of the optics we experience on a day-to-day basis. We

can generalise the polarisation of the material by expressing it as a power series in Ẽ(t),
as

P̃(t) = 0 χ(1) Ẽ(t) + χ(2) Ẽ2 (t) + χ(3) Ẽ3 (t) + . . .
(2.4)
≡ P̃(1) (t) + P̃(2) (t) + P̃(3) (t) + . . . .
We now have the additional terms χ(2) , χ(3) etc which are the second, third and in
general nt h order nonlinear optical susceptibilities, respectively. The strength of second
order processes is determined by the size of χ(2) just as the strength of third order
processes is determined by χ(3) and so on for higher order nonlinear processes. Typical
values for χ(2) are in the order of 10−12 m.V−1 and χ(3) is around 10−21 m2 .V−2 [29].
This is why such high intensity optical fields are required for these effects to become
appreciable.
For a full and accurate description of nonlinear optical phenomena we need to use vector
quantities for the electric field and polarisation and tensors for the susceptibilities. The
Chapter 2 17
optical electric field can be represented as a discrete sum of frequency components by

Ẽ(r, t) = E(ωn )e−iωn t (2.5)
n
where we are summing over positive and negative frequencies (i.e. n = ±1, 2, 3 . . . ) and
E(ωn ) satisfies the reality condition, i.e.
E(−ωn ) = E(ωn )∗ . (2.6)
Using a similar notation as in Equation 2.5 we can express the nonlinear polarisation as

P̃(r, t) = P(ωn )e−iωn t (2.7)
n
We are now in a position to analyse specific nonlinear processes.
2.3 Second Order Nonlinearities
F
or the second order nonlinear processes we need only consider the terms in P̃(r, t)
(2)
which relate to the second order nonlinear susceptibility tensor χijk (ωn +ωm , ωn , ωm ).
The components of the second order susceptibility tensor are the constants of propor-
tionality which relate the amplitude of the nonlinear polarisation to the product of the
optical field amplitudes, as in:
(2)
Pi (ωn + ωm ) = 0 χijk (ωn + ωm , ωn , ωm )Ej (ωn )Ek (ωm ) (2.8)
jk (nm)
In Equation 2.8 ijk are the Cartesian components of the fields and the notation (nm)
indicates that, despite ωn and ωm being able to vary, the sum ωn + ωm is held constant.
This is necessary to conserve energy. Taking the product of the n and m electric field
components:
E(ωn )e−iωn t E(ωm )e−iωm t = E(ωn )E(ωm )e−iωn t e−iωm t

= E(ωn )E(ωm )e−iωn t e−iωm t
= E(ωn )E(ωm )e−i(ωn +ωm )t ,
we see that E(ωn )E(ωm ) is associated with a time dependent product which oscillates
at ω = ωn + ωm . As this is always the case we can rewrite the argument of the second
order susceptibility tensor as χ(2) (ω3 ; ω2 , ω1 ) where it is understood that ω3 = ω2 + ω1 .
Thus
(2)
(2)
Pi (ω3 ) = 0 χijk (ω3 ; ω2 , ω1 )Ej (ω1 )Ek (ω2 ). (2.9)
jk
becomes the expression for the second order nonlinear polarisation.
Consider Equation 2.9 for the situation where there is a single optical field of frequency
ω1 polarised along the x direction and that it is significantly intense for nonlinear inter-
actions to take place. Thus, we have a single optical field at frequency ω1 producing an
nonlinear polarisation at ω1 + ω1 = 2ω1 i.e.
(2) (2)
Pi (2ω1 ) = 0 χixx (ω3 ; ω1 , ω1 )Ex (ω1 )2 . (2.10)
This process, as illustrated in Figure 2.1, is known as second harmonic generation (SHG).
SHG involves the conversion of two input photons of frequency ω into a new photon of
frequency 2ω.
Figure 2.1: Energy level diagram describing second harmonic generation. Second
harmonic generation involves the conversion two of input photons at frequency ω into
a single output photon of frequency 2ω, i.e. ω + ω = 2ω. Note that the process is never
100% efficient consequently there is always unconverted input radiation.
(2)
We now turn our attention to χijk which is a rank 3 tensor with 3 × 3 × 3 = 27 Cartesian
elements. For three interacting waves with frequencies ω1 , ω2 and ω3 = ω1 + ω2 there
will be a nonlinear polarisation influencing each one, each with it’s own second order
nonlinear susceptibility tensor,
(2) (2)
χijk (ω3 , ω1 , ω2 ), χijk (ω3 , ω2 , ω1 ),
(2) (2)
χijk (ω2 , ω3 , −ω1 ), χijk (ω2 , −ω1 , ω3 ),
(2) (2)
χijk (ω1 , ω3 , −ω2 ), χijk (ω1 , −ω2 , ω3 )
There are another 6 tensors for the negative frequencies and thus there are 12×27 = 324
separate elements to consider. This can be reduced to a matrix of 3 × 6 = 18 elements
Chapter 2 19
by considering certain symmetries. These include the reality of the fields, intrinsic
permutation symmetry, lossless media symmetry and Kleinman symmetry [30].
After the application of these symmetries the second order nonlinear susceptibility tensor
can be written in the contracted notation as
⎡ ⎤
d11 d12 d13 d14 d15 d16
⎢ ⎥
dil = ⎢
⎣d21 d22 d23 d24 d25 d26 ⎦
⎥ (2.11)
d31 d32 d33 d34 d35 d36
where the indices for dil come from the following prescription and
jk: 11 22 33 23,32 31,13 12,21

l: 1 2 3 4 5 6
1 (2)
dijk = χijk . (2.12)
2
For glassy materials there is an immediate implication for the second order nonlinear
susceptibility tensor. Amorphous materials possess what is called inversion symmetry,
that is any crystal axis of the material can have its sign changed (e.g. X → −X,
Y → −Y and Z → −Z) and the material remains unchanged. Otherwise known as
centrosymmetry, this has the effect of making all of the elements of χ(2) vanish. For
example, assume that the nonlinear polarisation is given by
P̃(2) (t) = 0 χ(2) Ẽ2 (t) (2.13)
where the optical field is give by
Ẽ(t) = E0 cos ωt (2.14)
If the sign of Ẽ(t) is changed then we necessarily must change the sign of the P̃(2) (t), as
we have assumed that the material has inversion symmetry. Thus we obtain
− P̃(2) (t) = 0 χ(2) |−Ẽ(t)|2 . (2.15)
Thus
− P̃(2) (t) = 0 χ(2) Ẽ2 (t) (2.16)
which implies that

P̃(2) (t) = −P̃(2) (t) (2.17)
which can only be true if and only if
χ(2) = 0. (2.18)
Thus all materials with centrosymmetry have zero second order nonlinear coefficients.
In fact, this applies for all even ordered nonlinear coefficients as
|−Ẽ(t)|n = Ẽ(t) ∀ n even. (2.19)
Conversely, for third ordered nonlinearities we have
−P̃(3) (t) = 0 χ(3) |−Ẽ(t)|2 (−Ẽ)

−P̃(3) (t) = −0 χ(3) Ẽ3 (t)
Implying that
P̃(3) (t) = P̃(3) (t), (2.20)
which provides no contradiction and thus centrosymmetric structures do possess third

order nonlinearities. Furthermore, this result is easily generalised to account for all odd
ordered nonlinearities.
Equation 2.11 allows us to express the nonlinear polarisation as a matrix equation which,
for the specific example of SHG, has the form:
⎡ ⎤
Ex (ω)2
⎢ ⎥
⎡ ⎤ ⎡ ⎤ ⎢ Ey (ω)2 ⎥
⎢ ⎥
Px (2ω) d11 d12 d13 d14 d15 d16 ⎢ ⎥
⎢ ⎥ ⎢ ⎥ ⎢ Ez (ω)2 ⎥
⎢Py (2ω)⎥ = 20 ⎢d21 d22 d23 d24 d25 ⎥ ⎢
d26 ⎦ ⎢ ⎥ (2.21)
⎣ ⎦ ⎣ ⎥
⎢2Ey (ω)Ez (ω)⎥
Pz (2ω) d31 d32 d33 d34 d35 d36 ⎢ ⎢2Ex (ω)Ez (ω)⎥
⎥
⎣ ⎦
2Ex (ω)Ey (ω)
The precise form of the second order nonlinear susceptibility tensor will depend on the
material in question. Firstly on the symmetry properties of the material, which will
reduce the number of non-zero elements in the tensor and secondly the actual values for
these non-zero elements will be determined by the constituents of the material.
There are additional second order effects that will not be discussed in detail, however,
we can briefly summarise all of the second order processes as:
Sum frequency generation (SFG): ω1 + ω2 = ω3

Second harmonic generation (SHG): ω + ω = 2ω
Difference frequency generation (DFG): ω1 − ω2 = ω3
Chapter 2 21
The common aspect to all of these processes is that two photons are converted into a
single photon of a new frequency, with overall conservation of energy.
2.3.1 Propagation of the Fields
To understand how the fields propagate during nonlinear processes we begin with Maxwell’s
equations in a medium with no free charges or currents:
∂ B̃
∇ × Ẽ = − (2.22a)
∂t
∇.D̃ = 0 (2.22b)
∂ D̃
∇ × H̃ = (2.22c)
∂t
∇.B̃ = 0 (2.22d)
which relate the electric and magnetic fields (Ẽ and H̃) to the electric displacement and
the magnetic flux density (D̃ and B̃). These quantities can be further related by the
constitutive relations give by
D̃ = 0 Ẽ + P̃ (2.23)
B̃ = μ0 H̃ (2.24)
where P̃ is given by Equation 2.4. Therefore we can write Equation 2.24 as
D̃ = D̃(1) + P̃N L (2.25)
where we have explicitly separated the linear and nonlinear parts of D̃ such that
D̃(1) = 0 Ẽ + P̃(1) (2.26)
Introducing the dielectric tensor (1) we can thus relate D̃(1) and Ẽ by
D̃(1) = (1) · Ẽ (2.27)
By manipulating Maxwell’s equations and using the fore mentioned definitions we obtain
the inhomogeneous wave equation for electromagnetic radiation
n2 ∂ 2 Ẽ 1 ∂ 2 P̃NL
− ∇2 Ẽ + = − (2.28)
c2 ∂t2 0 c2 ∂t2
This is the inhomogeneous wave equation, or nonlinear wave equation, in which the
nonlinear response of the material acts as a source term for additional frequencies. In
the following section we will utilise this equation to derive the Maker fringe expression.
2.3.2 Derivation of the Maker Fringes Expression
We can now use Equation 2.28 to analyse the case of SHG from a dispersive, nonlinear
material. First, however, consider the situation in Figure 2.2.
Figure 2.2: Diagram for second harmonic generation in a dispersive medium. Ra-
diation with frequency ω (fundamental) is incident at an angle θi to the surface of
a material with non-zero χ(2) and thickness L. The fundamental radiation generates
a second harmonic at 2ω. Refraction and dispersion cause the fundamental and sec-
ond harmonic radiation to be transmitted at angels θtω and θt2ω respectively. Both
frequencies leave the material at the angle θi .
Monochromatic light of frequency ω is incident onto a material with non-zero second

order nonlinearity. The light is incident at an angle θi and has sufficient intensity to
generate optical harmonics. The input light therefore induces a nonlinear polarisation in
the material which generates second harmonic radiation at frequency 2ω. For the fields
to be continuous across the boundary there must be reflected and transmitted radiation
at 2ω. Due to the dispersion of the refractive index, the fundamental and SH propagate
Chapter 2 23
at angles θtω and θt2ω , respectively as determined via Snell’s Law

−1 sin θi
θtω = sin (2.29)
nω

sin θi
θt2ω = sin−1 . (2.30)
n2ω
The wave vectors for the fundamental and SH are kω and k2ω and are directed along z
and z . Therefore, we can write the electric fields for the fundamental and SH as
1 ω
Ẽω (z, t) = A (z)e−i(kω z−ωt) + c.c. (2.31)
2
1 2ω −i(k2ω z −2ωt)
Ẽ2ω (z , t) = A (z )e + c.c. (2.32)
2
where Aω\2ω is a slowly varying amplitude for the respective fields and
nω ω
kω = (2.33)
c
n2ω 2ω
k2ω = . (2.34)
c
With these expressions for the fields and 2.9 we can therefore write the nonlinear polar-
isation as
(2)
(z, t) = 0 χijk (2ω, −ω)A2ω ∗ω −i(k2ω z −kω z) −iωt
N L(ω)
P̃i j (z )Ak (z)e e (2.35)
and
1 (2) −i(k2ω )z −i2ωt
(z , t) = 0 χijk (ω, ω)Aωj (z)A∗ω
N L(2ω)
P̃i k (z)e e (2.36)
2
We can calculate the derivative terms in Equation 2.28 as
∂2 ω
Ẽ (z, t) = −ω 2 Ẽωi (z, t) (2.37)
∂t2 i
∂ 2 N L(ω)
(z, t) = −ω 2 P̃i
N L(ω)
2
P̃i (z, t) (2.38)
∂t
and
∂ 2 Ẽωi (z, t) dAωi (z)
≈ − 2ikω + kω Ai (z) e−i(kω z−ωt) + c.c.
2 ω
(2.39)
∂z 2 dz
With equivalent expressions for the field at 2ω. In order to simplify Equation 2.39
we have utilised the slowly varying envelope approximation which assumes that the
fractional variation of in Ẽ on the scale of the wavelength associated with the field must
be smaller than unity, i.e.
d2 Ẽ dẼ

2 k ω (2.40)
dz dz
Substituting Equations 2.37, 2.38 and 2.39 into the NL wave Equation 2.28 and dropping
the time dependence, we obtain expressions for the field amplitudes Aωi and A2ω
i , given
by
dAωi (z) ω2 NL(ω)

= −i 2
Pi (z)e−i(kω z) (2.41)
dz kω c 0
dA2ω
i (z ) 2ω 2 NL(2ω) i(k2ω z )

= −i 2
Pi (z )e (2.42)
dz k2ω c 0
Introducing the expressions for the polarisations at ω and 2ω, given by Equations 2.35
and 2.36 and assuming that the susceptibilities are independent of frequency (i.e. Klein-
man Symmetry) we obtain:
dAωi (z) ω (2) 2ω ω∗

= −i χ A (z )Ak (z)e−i(Δk·Z) (2.43)
dz 2cnω ijk j
dA2ω
i (z ) ω (2)

= −i χijk Aωj (z)Aω∗
k (z)e
i(Δk·Z)
(2.44)
dz 2cn2ω
where nω and n2ω are the refractive indices of the nonlinear material at frequencies ω
and 2ω respectively and Δk = k2ω − kω represents the difference in the propagation
constants of the two frequencies. We have also applied a change of variable from z and
z to Z. This involves a projection of the propagation directions for the frequencies
at ω and 2ω onto the Z axis of the material (see Figure 2.2). This occurs under the
transformations
Z
z→ (2.45)
cos θtω
Z
z → (2.46)
cos θt2ω
The justification for performing this transformation comes from the fact that the planes
of equal phase mismatch between the fundamental and second harmonic are all perpen-
dicular to the material axis Z. Accordingly, Δk · Z has the form

k2ω kω
Δk · Z = − Z (2.47)
cos θt2ω cos θtω
In nonlinear processes where the conversion efficiency is low it is customary to make the
undepleted pump approximation, which can be expressed as
dAωi (z)
≈0 (2.48)
dz
that is the amplitude of the fundamental wave is essential unchanged under the con-
version of ω → 2ω. Further, if we introduce the effective nonlinear coefficient which
deals with the projection of the electric fields of the two waves onto the crystal axis (see
Chapter 2 25
Section 2.3.3 and specifically Equation 2.76) and complete the change of variables from
z → Z can rewrite Equation 2.44 as
dA2ω (Z) ω [Aω (Z)]2 i(Δk·Z)

= −i deff e (2.49)
dZ cn2ω cos θtω
Having developed the expression for the spatial variation of the amplitude of the wave
at frequency 2ω in terms of one variable we can integrate Equation 2.49 with respect to
Z over the thickness of the nonlinear region (see Figure 2.2) to obtain an expression for
the amplitude of the second harmonic wave
L
ω
2ω
A (Z) = −i deff [Aω (Z)]2 ei(Δk·Z) dZ. (2.50)
cn2ω cos θtω
0
By taking the product A2ω (Z)A2ω ∗ (Z) and using the general fact that intensity is
related to the filed amplitude via
0 cn 2
I= |A| (2.51)
2
we can find an expression for the intensity of the generated SH I2ω , as a function of
incidence angle θi , give by:
L 2

ω 2 Iω2
I2ω (θi ) = 3 deff e i(Δk·Z)
dZ. (2.52)
c 0 n2ω n2ω cos2 θtω

0
It must be noted that Equation 2.53 assumes no depletion of the fundamental, i.e.
Iω =constant. This is a reasonable assumption for the low values of χ(2) observed in
most material. The dependence of Equation 2.53 is implicit and can be found explicitly
through Snell’s law.
We will assume that the effective nonlinear coefficient has no dependence on the spatial
coordinate Z and thus Equation 2.53 can be written as:
L 2

ω 2 Iω2 2

I2ω (θi ) = 3 2 2
deff e i(Δk·Z)
dZ. (2.53)
c 0 n2ω nω cos θtω
0
Which enables us to complete the integration, yielding:

2
ω 2 Iω2 2πL sin2 Ψ
I2ω (θi ) = 3 d2eff (2.54)
c 0 n2ω n2ω cos2 θtω λ Ψ2
We have simplified the appearance result by making the replacement of

2Lπ
Ψ= (nω cos θtω − n2ω cos θt2ω ) (2.55)
λ
The expression in Equation 2.54 gives the angular dependence of the intensity of the
second harmonic, however, it falls short in predicting what one would observe as the
nonlinear sample is rotated with respect to the beam because it fails to take into account
the losses due Fresnel reflection at the interfaces.
The form of the Fresnel transmission coefficients depend on the polarisations of the
fundamental and second harmonic. For light incident onto an interface between materials
1 and 2 at an angle θ1 and refracted via Snell’s law to an angle θ2 , for the polarisations
states s and p they are give by:
2 cos θ1 sin θ2
t1→2
p = (2.56)
sin(θ1 + θ2 ) cos(θ1 − θ2 )
2 cos θ1 sin θ2
t1→2
s = (2.57)
sin(θ1 + θ2 )
If the materials have refractive indices n1 and n2 for materials 1 and 2 respectively then
we can define the amplitude transmission factor as
n2 cos θ2
T 1→2 = t1→2 (2.58)
n1 cos θ1
Given the situation illustrated in Figure 2.2 the amplitude transmission factor for the
fundamental will be given by:
nω cos θtω
Tωa→g = (ta→g ) (2.59)
cos θi
Further, the SH generated in the material crosses the glass/air interface as it exits the
glass at an incident angle θt2ω and thus its amplitude transmission coefficient is
g→a cos θi
T2ω = (tg→a ) . (2.60)
n2ω cos θt2ω
Equation 2.54 is proportional to the intensity in the second harmonic and the square of
the intensity of the fundamental, it is therefore clear that by multiplying Equation 2.54
g→a 2
by T2ω and (Tωa→g ) we will obtain an expression for the angular dependence of the
SH intensity.
2 2
(Tωa→g ) T2ω g→a 2 2
ω Iω 2πL sin2 Ψ
I2ω (θi ) = d2eff (2.61)
c3 0 n2ω n2ω cos2 θtω λ Ψ2
Finally, using the definition for intensity of the fundamental as the power Pω divided
Chapter 2 27
by the area of the beam A (and assuming that the fundamental and second harmonic
beams have the same area), we obtain the so called Maker fringes expression of which
more discussion will be made in Section 8.1.
2 2
(Tωa→g ) T2ωg→a 2
ω 2πL sin2 Ψ
P2ω (θi ) = 3 2 2
P2 d2eff (2.62)
Ac 0 n2ω nω cos θtω ω λ Ψ2
2.3.3 Second Order Nonlinear Effects in Thermally Poled Materials
It is known experimentally that thermally poled glasses produce second harmonics when
exposed to intense radiation. To see how this occurs consider the first term in the
nonlinear polarisation for a centrosymmetric material (i.e. χ(2) = 0):
PN L = 0 χ(3) E3 . (2.63)
When we have generated the field EDC via poling this contributes with the optical field
to the nonlinear polarisation and the total field can be written as a superposition of the
DC field and a monochromatic optical field at frequency ω, i.e. ET = E0 cos ωt + EDC .
Substituting this into Equation 2.63 we obtain:
PN L =0 χ(3) E3T

=0 χ(3) (E0 cos ωt + EDC )

=0 χ(3) E0 cos3 ωt + 3EDC E20 cos2 ωt + 3E2DC E0 cos ωt + E3DC (2.64)
1 3
Using the trigonometric replacements: cos3 ωt = 4 cos 3ωt + 4 cos ωt and cos2 ωt =
1
2 + 12 cos 2ωt
Equation 2.64 takes the from:

1 3 3 3 3
PN L = 0 χ(3) E cos 3ωt + E30 cos ωt + EDC E20 + EDC E20 cos 2ωt + . . .
4 0 4 2 2

2 3
3EDC E0 cos ωt + EDC (2.65)
The fourth term in Equation 2.65 indicates that there will be radiation at frequency
2ω i.e. a second harmonic. We can thus write an effective second order nonlinear
polarisation as:
(2) 3
P = 0 χ(3) EDC E20 (2.66)
eff 2
or
(2) (2)
P = χ E20 (2.67)
eff eff
(2)
where we have defined the effective second order nonlinear susceptibility χ = 32 0 χ(3) EDC .
eff
The presence of a linear electro-optic effect in thermally poled glasses can be understood
by considering Equation 2.3 for the relative permittivity and the constitutive relation
D = 0 E + P (Equation 2.24) Dividing by 0 E we obtain
D P
=1+
0 E 0 E
Combining this result with Equation 2.4 we get
D
= 1 + χ(1) + χ(2) E2 + χ(3) E3 + . . . (2.68)
0 E
Which in combination with Equation 2.3 can be expressed as
D
=r + χ(2) E2 + χ(3) E3 + . . .
0 E
=r + Δ
= (2.69)
where is the altered permittivity of the material under the influence of strong electric
fields. Using the relationship between permittivity and refractive index
= n2 (2.70)
and considering a small change in the refractive index of the material Δn, corresponding
to the change in permittivity produced by the electric field E we obtain
=(n + Δn)2
≈n20 + 2n0 Δn. (2.71)
For glass, where χ(2) = 0 and ignoring all terms of E higher than order 2 it follows that:
− n20 − r Δ
Δn = = =
2n0 2n0 2n0
1 (3) 2
= χ E (2.72)
2n0
Consider now the case of thermally poled glass which has associated with it a frozen
in DC field EDC onto which we superpose an external electric field Eext . Then E =
EDC + Eext , which with Equation 2.72 gives:
1 (3) 2
Δn = χ EDC + 2EDC Eext + E2ext (2.73)
2n0
The second term in 2.73 implies that for a fixed value of EDC under the influence of the
Chapter 2 29
external electric field Eext the change in refractive index (and therefore phase of light
propagating through the glass) is linear in Eext . This is the linear electro-optic effect.
The third term can be neglected because Eext EDC [31]. If the glass has not been
poled and therefore EDC = 0 then the change in refractive index will be proportional to
E2ext i.e. a quadratic electro-optic effect .
A convenient way of visualising these two situations is shown in Figure 2.3: To begin with
the glass has an intrinsic third order nonlinearity which as we have shown can produce
a phase change that is quadratic in the applied field (red curve in Figure 2.3). After
poling this parabolic curve has shifted by an amount equal to the value of the frozen in
field EDC , such that now the phase shift is approximately linear with the applied field
(blue curve in Figure 2.3). Therefore producing a pseudo-linear electro-optic response.
Figure 2.3: electro-optical response of glass before (red parabola) and after poling
(blue parabola).
Now that we have established that thermal poling does indeed produce an effective
second order nonliearity we can apply Equation 2.21 to analysing SHG from in such a
sample. But first we must first apply the symmetries of the material to the second order
nonlinear susceptibility tensor. In doing so we will be able to show that many of the
elements are equal to zero, thus simplifying the equations for the nonlinear polarisation.
Figure 2.4 is a schematic illustration of a thermally poled glass sample with the electric
field shown. By considering which transformations will leave the field in the figure un-
changed we are able to apply these to the elements of the effective nonlinear susceptibility
matrix (Equation 2.11) and determine which are the non-zero elements.
We are using the conventional capital letters notation (X, Y, Z) for crystal axis, the
laboratory frame will have lower case letters (z, y, z). However, we must stress that
the symmetries are not related to any crystal structure, instead it is the symmetry of
the electric field which has been created during poling. The field possesses rotation
Z
Ep
-X
-Y Y
X
-Z
Figure 2.4: Illustration of the symmetry possessed by the electrical potential in ther-
mally poled glass.
symmetry about the Z axis. We can rotate the figure about Z by any angle and the
field will remain unchanged. This type of symmetry is called C∞ν . Consider a clockwise
rotation by 90◦ , under this operation the coordinates transform like
X → −Y
Y → X
Z → Z .
This can be represented by the following matrix equation

⎡ ⎤⎡ ⎤ ⎡ ⎤
0 −1 0 X −Y
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢1 0 0⎥ ⎢Y ⎥ = ⎢ X ⎥ . (2.74)
⎣ ⎦⎣ ⎦ ⎣ ⎦
0 0 1 Z Z
Then, using the full index notation for the dijk elements and using Equation 2.74 as a
prescription for changing the ijk’s we obtain as series of relationships between elements
of dijk . For example
d122 = −d211
d211 = d122
Chapter 2 31
This produces a contradiction which can only be resolved if and only if
d122 = d211 = 0
Many similar relationships can be established via this symmetry and others which in-
clude:
Rotation about Z by 180◦ ,
X → −X
Y → −Y
Z → Z .
Reflection symmetry in the XZ plane,.
X → −X
Y →Y
Z → Z .
and in the Y Z plane
X → X
Y → −Y
Z → Z .
Inversion symmetry in the X and Y axes which is equivalent to a reflection.
The final result is a nonlinear susceptibility matrix with the following form:
⎡ ⎤
0 0 0 0 d31 0
⎢ ⎥
dil = ⎢
⎣ 0 0 0 d31 0 0⎥⎦. (2.75)
d31 d31 d33 0 0 0
Now, if we select a specific experimental configuration, in terms of sample orientation

and input and output polarisation, we can derive an expression for the effective nonlinear
polarisation def f . The effective nonlinear polarisation is defined as [29]
deff = ê2ω · dil : êω êω (2.76)
where êω is the electric field polarisation vector at frequency ω. If we consider the
following experimental configuration (Figure 2.5):
X Y
Figure 2.5: Configuration for SHG in the pp-configuration. Coordinate axes (X, Y, Z)
belong to the glass and (x, y, z) are the laboratory frame.
We select the lab frame axis to coincide with those of the crystal frame for convenience.
The polarisation of the wave at frequency ω, which we will call the fundamental, lies in
the xz plane of the laboratory frame at some angle θω to the z axis. Upon entering the
glass with refractive index nω , Snell’s law,
sin θω = nω sin θω . (2.77)
tells us that it will be refracted to a new angle θω , where the prime indicates an angle
within the glass. Thus, in the glass the fundamental is propagating with an electric field
polarisation vector given by:
êω = (cos θω , 0, sin θω ). (2.78)
We now assume that the fundamental wave has sufficient intensity to produce some
second harmonic (SH). Further, if we use a polariser to select the SH that has its
Chapter 2 33
polarisation in the xz plane, the polarisation vector for the SH will have the form:

ê2ω = (cos θ2ω , 0, sin θ2ω ). (2.79)
The angle θ2ω is determined by the incidence angle of the fundamental and the refractive
index of the glass at the second harmonic n2ω , by

sin θω = n2ω sin θ2ω . (2.80)
Equation 2.76 then becomes

⎡ ⎤
cos2 θω
⎢ ⎥
⎡ ⎤ ⎡ ⎤⎢ 0 ⎥
⎢ ⎥
cos θ2ω 0 0 0 0 d31 0 ⎢ 2 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ sin θω ⎥

deff = ⎢
⎣
⎥
0 ⎦·⎣ 0⎢ 0 0 d31 0 ⎥ ⎢
0⎦ ⎢ ⎥
⎥ (2.81)
⎢ 0 ⎥

sin θ2ω d31 d31 d33 0 0 0 ⎢⎢ ⎥
0 ⎥
⎣ ⎦
2 sin 2θω
which expands to give
deff = 2d31 sin θω cos θω cos θ2ω

+ (d31 cos2 θω + d33 sin2 θω ) sin θ2ω

. (2.82)
The configuration we have considered, fundamental polarisation and SH polarisation,

is commonly called the pp-configuration. Here pp is short hand for p-polarised int-
put (fundamental) and p-polarised output (second harmonic). There is an alternative
configuration for SHG in poled glass which is the sp-configuration. Here the input is
s-polarised and the output is p-polarised. In the sp-configuration the effective nonlinear
coefficient is given by

deff = d31 sin θ2ω . (2.83)
In general these expressions can be used to determine the angle required to maximise
the effective nonlinear coefficient for a particular experimental configuration, although
first we must square def f to obtain a value proportional to intensity. For the pp- and
sp-configuration d2ef f is maximum for θω = 90◦ .
We can use this specific forms of the effective second order nonlinear coefficients given in
Equation 2.82 and Equation 2.83 with then general form of the Maker fringes expression
(Equation 2.62), to derive the Makers fringe expression for thermally poled glass.
2.4 Third Order Nonlinearities
T
he third order order contribution to the nonliear polarisation as given by
P̃(3) (t) = 0 χ(3) Ẽ(t)3 , (2.84)
determines the third order nonliearities of a material. If we consider a monochromatic

wave of the form
Ẽ(t) = E0 cos ωt, (2.85)
substituting into Equation 2.85 into Equation 2.84 we get
P̃(3) (t) = 0 χ(3) E30 cos3 ωt. (2.86)
1 3
By appliing the trigonometric identity cos3 ωt = 4 cos 3ωt + 4 cos ωt, the nonlinear po-
larisation can be expressed by

(3) 1 (3) 3 3 (3) 3
P̃ (t) = 0 χ E0 cos 3ωt + χ E0 cos ωt . (2.87)
4 4
The first term in Equation 2.87 predicts third hamonic generation (THG), a process
analogous to SHG, where three input photons with frequency ω, are converted into a
single photon of frequency 3ω i.e. ω + ω + ω = 3ω (see Figure 2.6).
Figure 2.6: Energy level diagram describing third harmonic generation. Third har-
monic generation involves the conversion three of input photons at frequency ω into a
single output photon of frequency 3ω, i.e. ω + ω + ω = 3ω. Note that the process is
never 100% efficient consequently there is always unconverted input radiation.
There are a multitude of additional third order effects (44 in fact), far too many to
describe in detail. It suffices to say that they all involve the interaction of four photons
Chapter 2 35
and sums and differences of their frequencies. The only process of direct interest to this
thesis is that of the intensity dependent refractive index, otherwise referred to as the
nonlinear refractive index. This process is discussed in detail in the following Section.
2.4.1 Origin of the Nonlinear Refractive Index
The second term in Equation 2.87 predicts a nonlinear contribution to the polarisation at
the frequency of the incident field. This term causes the material to change its refractive
index in response to the intensity of the input beam in the form
n2
n = n0 + I (2.88)
2
where n0 is the refractive index of the material and n2 is the nonlinear refractive index
which corresponds to the change in n observed with an input intensity I. The factor
of 1/2 appears in Equation 2.88 to maintain consistency with the majority of literature
reported values of n2 . An alternative expression exists which originates as a result of a
different definition of the electric field and therefore intensity. This alternative differs
by a factor of 2 from the more standard definition. It is therefore not uncommon to
find reported values of the nonlinear refractive index to differ a factor of 2, thus causing
some initial confusion.
The change in refractive index described by Equation 2.88 is often called the optical
Kerr effect. This can be restated in terms of the electric field as
n = n0 + γ|E(ω)|2 (2.89)
where the intensity and the electric field are related via:
I = 2n0 o c|E(ω)|2 (2.90)
To understand how the optical Kerr effect arises from the nonlinear polarisation we
must consider the term in the nonlinear polarisation which influences the propagation
of a beam of frequency ω;
P(3) (ω) = 30 χ(3) (ω = ω + ω − ω)|E(ω)|2 E(ω) (2.91)
here we have three photons of frequency ω combining to produce a photon at frequency

ω. To avoid the complicated tensor nature of the third order susceptibility we are
assuming the light is linearly polarised. The total polarisation, up to third order, of the
material is thus described by

P(ω) = 0 χ(1) E(ω) + 3χ(3) |E(ω)|2 E(ω) (2.92)
Note that we have assumed this material to have no second order response. We can
introduce an effective susceptibility
χeff = χ(1) + 3χ(3) |E(ω)|2 (2.93)
which allows us to write Equation 2.92 as
P(ω) = 0 χeff E(ω). (2.94)
It is generally true that

n2 = 1 + χeff (2.95)
thus by substituting Equation 2.90 on the left and Equation 2.93 on the right we obtain
[n0 + γ|E(ω)|2 ]2 = 1 + χ(1) + χ(3) |E(ω)|2 (2.96)
and expanding gives
n20 + 4n0 γ|E(ω)|2 = (1 + χ(1) ) + (χ(3) |E(ω)|2 )). (2.97)
We can equate the relevant terms to get relation ships for n0 and γ:
n0 = (1 + χ(1) )1/2 (2.98)

3χ(3)
γ= (2.99)
4n0
Finally, we note that the total refractive index must be the same regardless of description.
Thus from Equations 2.88 and 2.90 we see that
n2
γ|E(ω)|2 = I (2.100)
2
and therefore the intensity dependent refractive index is related to the third order non-
linear susceptibility via:
3χ(3)
n2 = . (2.101)
4n20 0 c
Chapter 2 37
2.5 Conclusion
T
he essential physics describing the nonlinear processes that are of relevance to
this thesis has been outlined. Furthermore, calculations and derivations of key
relationships have been made. Notably, the form of the effective second order nonlinear
susceptibility matrix for thermally poled glass and the Maker fringes expression. These
expressions are utilised in the analysis of second harmonic generation from samples in
Chapter 8. As well as showing the relationship between the nonlinear refractive index
and the third order susceptibility. This result is employed in Chapter 6 to calculate the
third order susceptibilities from the measured nonlinear refractive indices.
Chapter 3
Glass Theory, Design &

Fabrication
3.1 Introduction
I
n this chapter I describe the process that was used to design the tellurite glass compo-
sitions that were the focus of my PhD research. The procedure for fabricating theses
glasses is also outlined. Finally, I summarise the results, indicating which compositions
are shown to be good glass formers and therefore become the focus of all subsequent
investigations within this thesis.
3.2 What is a Glass?
T
he glassy, or vitreous state, as it is sometimes called, refers to the underlying
structural arrangement of atoms in these solids. Glasses are solid materials that
are typified by a random arrangement of atoms and therefore lack long-range atomic
order (figure 3.1 right) thus resembling liquids more than crystalline solids in terms of
their atomic structure. In contrast to crystalline material, glasses do not have a precisely
defined melting point but rather soften gradually over a range of temperatures. This
range is in the order of 10◦ C.
Many materials can be induced into the glassy state provided certain requirements are
met. Glass formation is really a process of rapid solidification where the rapidity is
sufficient to suppress the tendency for crystallisation. For a crystal to form, firstly it
must be nucleated and then it must have time to grow. This second step requires the
migration of atoms into the vicinity of the crystal nucleus. If either or both of theses
processes can be avoided, then a glass can be formed. Therefore, by cooling the molten
39
40 Glass Theory, Design & Fabrication
Figure 3.1: Left: Regular atomic arrangement in a crystal. Right: Irregular atomic
arrangement in a glass.[32]
solution very rapidly, the mechanisms for crystallisation do not have time to occur and
a glass may be formed.
With reference to Figure 3.2, we begin with a liquid at the point ‘a’. If we cool the liquid
slowly the volume decreases at a constant rate along the line ‘ab’, until we reach the
melting temperature, Tm . For T=Tm the liquid and solid state are in equilibrium, that
is they have the same Gibbs energy1 . For T Tm a rapid and discontinuous change in
the materials volume is encountered along line ‘bc’ and a crystal is formed. Subsequent
cooling will result in a further constant decrease in volume, along ‘cd’, as the thermal
oscillations of the nuclei reduce. However for any given temperature, T < Tm , the
material has a minimum volume for that particular temperature2 .
Now consider what will happen if we begin at ‘a’ but cool rapidly. The material will first
move along ‘ab’ but at T=Tm the processes required for crystallisation (nucleation and
growth) do not have time to occur. Further cooling below Tm will result in a continuous
reduction in volume and the material will move through the ‘super cooled liquid’ region
and into the ‘glass transformation range’ whereupon dV /dT decreases and the glassy
state is obtained. Further cooling serves to reduce the volume of the now formed glass
by decreasing thermal oscillations just as in the crystalline state along a line of constant
slope ‘f e’, which is called the glass cooling line.
1
The Gibbs energy is a measure of the capacity of the system to do non-mechanical work. As such
it is analogous to a potential energy.
2
There are some notable exceptions to this rule in an analogous way to which the volume of water
increases upon freezing. These exceptions are curious in their own right but irrelevant to this work and
we do not lose generality by disregarding them
Chapter 3 41
Figure 3.2: Potential cooling curves for glass and crystal formation. Glass with fictive
temperature Tf , obtained with fastest cooling rate. Glass’ with fictive temperature T’f ,
obtained with slower cooling rate. The approximate positions of the glass relaxation
temperatures Tg are indicated with respect to the corresponding fictive temperatures.
Crystal with melting temperature Tm , is obtained with slowest cooling rate
The intersection of the glass cooling line and the liquid cooling line is defined to be the
fictive temperature, Tf . Figure 3.2 indicates that the final volume (for a given temper-
ature) of the glass and hence Tf depends on the cooling rate. Faster cooling produces
a glass with a larger volume (lower density) and hence higher fictive temperature than
a more slowly cooled melt. The fictive temperature is closely related to the glass relax-
ation temperature3 Tg . The glass transition temperature is the temperature at which
upon heating the solid glass takes on the character of an extremely viscous liquid.
It has been demonstrated experimentally that Tg is typically slightly larger than Tf ,

however there is no simple relation between the two [33]. It is reasoned that this is
essentially due to the way in which each is determined. Tf is determined during cooling
of the melt, while Tg is determined via heating. Both Tg and Tf are indicators of the
transition from super cooled liquid to solid glass (or vice versa), however, Tg is much
easier to measure and therefore more commonly used.
3
Alternatively refereed to as the glass transition temperature or glass transformation temperature.
3.3 General Properties of Glasses
T
he physical and optical properties of glasses will typically differ from the chemically
equivalent crystalline structures due to the amorphous nature of glasses. Further-
more, we cannot use constructs such as the unit cell and periodic boundary conditions,
as we would for a crystal, to calculate or derive the properties because of the lack of
structural periodicity. This fact makes glasses fundamentally difficult systems to model.
The following are some general properties that typify all glassy systems:
Glasses do not cleave in any preferred direction owing to the absence of crystalline
planes.
They will have zero-valued even-ordered nonlinear coefficients.
The property that best typifies glass is an ill-defined melting point. Therefore a
commonly adopted convention is to define the melting point as the temperature
at which the viscosity of the glass reaches 100 dPa.s.
The properties of a glass, as with all materials, are determined by the properties of its
constituents. Because a glass is a frozen mixture of materials, stoichiometry need not be
satisfied and so there is an continuous range of possible combinations of materials that
can produce a glass.
3.3.1 Structural Properties
It is useful to think of glasses as being comprised of two different elements:
The first is the network. This is the structure of the glass and makes up the vast
majority of the material and is the principal determinant of the properties of the
glass. The right side of Figure 3.1 represents a glass network.
Second, there are the network modifiers which perturb the network and pro-
duce small changes to the properties of the glass. Network modifiers are present
in the form of ionic species and will alter physical and optical properties of the
glass by disrupting the network. They will reside in locations which facilitate the
neutralisation of excess charge produced by the breaking of bonds (figure 3.3).
As to which elements will be network formers, and which will be modifiers, there are
several models for predicting this behaviour. The model we will consider is Dietzel’s
Chapter 3 43
NOTE:
Figure 3.3: The effect of network modifying ions on the atomic arrangement in silica
glass. Reproduced from [34]
field strength criterion [35] which is well suited for glasses in which the bonds between
elements are highly polar, such as silica and tellurite. Dietzel’s model is, however, poorly
suited for highly covalent glasses like the chalcogenides where the bonding between atoms
is very weak.
Dietzel’s model for glass formation looks at the process in terms of the strength of
the oxygen−cation bonds and the cation−cation repulsion in the molten state and the
resulting battle for oxygen between two or more components. Dietzel’s Field Strength is
defined as:
Zc Zc
FD = 2
= 2. (3.1)
(rc + ro− ) a
Where Zc is the valency of the cation and re and ro− are the ionic radii of the cation
and the oxygen ion respectively and so a = rc + ro− is the ionic bond length between the
cation and oxygen. Dietzel’s Field Strength has units of C.m−2 , however FD is typically
used as a figure of merit to identify the roles various chemical components will play
during glass formation. It is therefore confusing to refer to it as a field strength. As
such we refer to FD as Dietzle’s figure of merit for glass formation, or simply Dietzle’s
figure of merit.
Dietzel defines network formers to be those cations with figures of merit with values of
1.3 FD 2. Network modifiers will be those cations with a figure of merit in the
range 0.1 FD 0.4. The intermediates will have figures of merit of 0.5 FD 1.1.
This is summarised in Table 3.1
To illustrate how this model predicts glass forming behaviour, consider the following
situation: Oxides of the components A and B are present in a melt. Let component
A have a figure of merit > 1.3 and B < 0.4 strength. As we cool the melt, A repels
other cations and preferentially bonds to form oxygenated complexes. Now B will have
lost the battle for oxygen and therefore do the best it can to neutralise its charge by
coordinating it’s self with oxygens already bonded to A. This situation looks very much
like Figure 3.3.
3.3.2 Relationships Between the Structure and Properties
Because there are no stoichiometric requirements for glass compositions we can produce
a continuum of compositions and we can expect that as we continuously vary the pro-
portions of the components the physical properties will also vary continuously. Many
researchers attempt to develop empirical models to predict the continuous variation in
properties of glasses, the simplest possible are the additivity relationships which predict
linear relationships between the concentration of the components and the associated
property. Additivity relations are only valid under the assumption that there are no
nonlinear interactions between the different components of the glass.
For example, a glass with N components could have an additivity relation for refractive
index (refractive index often obeys additivity relationships), in the form

N
n0 = ni ai , (i = 1, 2, 3, . . . , N ). (3.2)
i
Where ni are the coefficients for the refractive index n0 and ai is the proportion of the
ith component in the glass. Refractive index and density are examples of properties
that often obey additivity relations (see Sections 5.2 and 6.4) Additivity relationships,
however, do not apply in general as it is very common for the components to inter-
act in a nonlinear fashion. Thus more complicated empirical models often need to be
constructed.
3.4 The Tellurite Glass System
P
ure TeO2 is not a good glass former [36], however, with the addition of network
modifying cations, TeO2 can be readily induced into the glassy state. Tellurium
oxide is therefore referred to as a conditional glass former. The structural elements
Table 3.1: Deitzel’s Field Strength Parameters for Glass Formation of Various Oxides. Reproduced from [35].
Chapter 3
NOTE:
This table is included on page 45
of the print copy of the thesis held in
the University of Adelaide Library.
45
present in the glassy state are known to be distorted variants of the TeO4 trigonal bi-
pyramidal (tbp), TeO3+1 polyhedron and TeO3 trigonal pyramidal (tp) sub units present
in crystalline TeO2 as illustrated in Figure 3.4.
Figure 3.4: Structural representation of Te-O subunits. Left: TeO4 trigonal bi-
pyramidal (tbp). Middle: TeO3+1 distorted trigonal bi-pyramidal. Right: TeO3 trigo-
nal pyramidal (tp). Bond lengths in nm.
In tellurite glass, Te-O bond lengths for the TeO4 tbp structural subunit range from
0.185 to 0.195nm for the equatorial oxygens, and 0.205 to 0.215nm for the axial oxygens.
The elongated axial Te-O bond belonging to the TeO3+1 subunit ranges from 0.220 to
0.260nm, while the Te-O bonds in the TeO3 tp subunits lie in the range 0.185 to 0.20nm
[37]
We can calculate the glass-forming nature of tellurium oxide via Equation 3.1. Using
the mean bond length from Figure 3.4 (2 Å) and the most common valency of tellurium
(Zc = 4), we obtain a figure of merit of ≈ 1. This implies that TeO2 is an intermediate
i.e. it can behave as a network former or modifier depending on the glass forming nature
of other species which may be present in the melt. For example when TeO2 is melted in
the presence of another oxide with a figure of merit much less than 1 (such as Na2 O),
TeO2 readily forms a glass. It is common to refer to modifiers that induce intermediates
to become network formers as ’stabilising ions’. In isolation, however, TeO2 is prone to
crystallisation.
Some important properties of tellurite glass are:

Chapter 3 47
TeO2 is highly polarisable (αi = 6.95/10−30 m3 i.e. 2.5×SiO2 [38]) and therefore
possesses large linear and nonlinear refractive indices. This makes it an ideal
candidate for nonlinear optical devices.
Tellurite is optically transparent between ≈ 400 nm and 5 μm, which makes tel-
lurite an ideal candidate for visible and mid-IR applications [11].
It has a low glass transition temperature which makes it a good glass for optical
fibre preform extrusion [11].
It is not hygroscopic and thus does not suffer from environmental degradation
unlike the flouride glasses which are often considered for mid-IR applications [11].
3.5 Designing the Glass Compositions
T
he problem we have is this: How can we understand the properties and the variation
of properties observed in tellurite glasses so that predictions can be made about
which composition is most suitable for a particular application? To address this question
we fabricated a variety of tellurite glasses with systematically varied compositions to
make possible general conclusions about the way the properties of tellurite glass depend
on the additional components within them. It can be reasonably assumed that the
physical properties of a glass vary continuously with variation in the proportion of the
components within the glass. Knowing how said properties vary with the different
chemical components enables us to predict which combination of components will provide
an optimised composition for a specific purpose.
In terms of the main objective of this thesis, which is to establish the suitability of
various tellurite glasses for use as a base material for the fabrication of nonlinear optical
fibre devices, we need to consider what properties are critical to the production and
subsequent operation of such devices. Thus we make the following considerations:
1. A nonlinear device requires the use of materials that possess large nonlinear co-
efficients. Optical nonlinearities are related to the electronic polarisability (and
hyperpolarisability etc.) of the material and in particular its constituents. For
this reason it is necessary to select chemical components that will imbue the glass
with high nonlinear coefficients.
2. Any optical device must have excellent optical transmission. In particular for a
nonlinear fibre device there are two main contributing factors to the overall loss of
the device. Material absorption presents the fundamental lower limit to the devices
loss and can be minimised by selecting a material with a transmission range that
suits the desired operating wavelength or wavelengths. The purity of this material
should then be made as high as possible by starting from high purity raw materials
and using a very clean and controlled fabrication environment.
Scattering loss produced by crystals can be a major source of loss in optical fibres
and thus a glass with high resistance to crystallisation should be used. Crystalli-
sation of glass is a common problem encountered when trying to form glasses from
new or little studied glass compositions
3. We seek glasses that can be used for the fabrication of optical fibres. As such,
volume scalability is also crucial so that we may produce practical size billets for
preform production. We fabricate our optical fibres via the extrusion technique
(see Section 9.4) which requires the use of glass billets with volumes of 20 cm3 .
For a glass to be produced in volumes such as these it is critical for the glass to
possess a high degree of crystallisation stability. This is because as the volume
of the glass increases its surface area to volume ratio decreases and therefore the
ability of the cooling melt to shed heat, which it must do so rapidly to form a
glass, is reduced.
Much of the glass research reported in the literature focuses on relatively small
volumes of glass. As such the reports of what are often referred to as ’promising
candidates’ for optical fibre materials a frequently unsuitable as they cannot be
easily scaled to practical billets sizes.
To summarise our requirements; we require glasses that are highly transparent, highly
nonlinear and resistant to crystallisation to allow crystal free fibres to be fabricated.
We address the first requirement by selecting modifiers that have a variety of electronic
polarisabilities (see Table 3.2). This provides an opportunity to probe the contribution
of the modifier to the optical nonlinearity. The second requirement is met by including
10 mole% of sodium oxide (Na2 O) in all the glass samples as a stabilizing ion [39].
Finally the third requirement, which is also related to the glass forming stability of the
material, is met by only studying glass compositions which have shown their ability
to be produced in large volumes. This should ensure that we can fabricate optical
fibre preforms with sufficient size so as to provide practical lengths of optical fibre.
Furthermore, glasses which can be fabricated in large volumes by default have a good
resistance to crystallisation and so will have a good chance of tolerating the fabrication
process for optical fibres.
By fabricating a series of glasses with systemically varied compositions we can determine

the compositional dependence of various properties, whilst ensuring that if a sample
Chapter 3 49
Electronic Ionic Most Common

Polarisability Electronegativity Radius Coordination
Modifier [10−23 cm3 ] [Pauling Scale] [Å] Number
Na+ 0.41 0.93 0.95 6
Zn2+ 0.8 1.65 0.83 6
Ba2+ 2.5 0.89 1.43 8
Mg2+ 0.09 1.31 0.78 6
Sr2+ 1.6 0.95 1.27 8
Pb2+ 4.9 2.33 1.32 8
Tellurite
Te4+ - 2.10 1.87 3,4
TeO4 6.71 - - -
TeO3 5.30 - - -
Table 3.2: Potential modifiers and their electronic polarisabilities [40], ionic radii
[18, 40], electronegativities, molecular masses and coordination numbers [41]. Where
possible the properties of the tellurium ion [18] and the TeO4 and TeO3 sub units [42]
are listed.
shows particular promise for a nonlinear device it should have the necessary stability to
survive the MOF manufacturing process.
We can express the glass compositions that were investigated by the general formula:
10Na2 O.xMO.(90 − x)TeO2 (3.3)
Where MO is a metal oxide comprised from one of the divalent cations listed in Table
3.2.
3.6 Procedure for Glass Production
A
ll glasses were produced using high purity raw materials as listed in Table 3.3. For
Na2 O and BaO we started with the carbonates Na2 CO3 and BaCO3 which readily
decompose to the oxides under the glass melting conditions. This provides higher purity
at lower cost.
The following list outlines the general procedure for producing the glass samples. The
specific details of which were developed by trial and error with some reference to pub-
lished techniques:
1. Raw materials are combined in an inert atmosphere (N2 ) in the appropriate pro-
portions to an accuracy of ± 0.05 mg. Combining (batching) the raw materials in
Component Purity [%] Supplier

TeO2 99.995 CERAC & Alfa Aesar
Na2 CO3 99.99 Alfa Aesar
ZnO 99.99 CERAC
BaCO3 99.99 CERAC
PbO 99.99 CERAC
SrO 99.99 CERAC
Table 3.3: Purity and supplier for glass raw material.
a controlled environment such as this minimises the likelihood of contamination

by dust particles and moisture from the air.
2. The combined materials are mixed by shaking in a sealed plastic container.
3. The powders are transferred to a gold crucible of either 100 mL or 300 mL volume
depending on the size of the desired billet. A lid made from gold is used in
combination with the crucible to minimise the evaporation loss while melting.
Thus we can consider the melting environment to be, approximately, a closed
system.
4. We place the crucible into a preheated furnace in open air at 500◦ C and ramp the
temperature at 10◦ C.min−1 . This gives the volatile carbonate components time to
decompose to their oxides before the mixture has melted.
5. We hold the mixtures at a melting temperature of 900◦ C for >1 h.
6. Periodically the melted solution is swirled during the melting period to ensure
homogeneity, taking care to capture any undissolved material on the inner sides
of the crucible.
7. We then cast the melt into brass moulds which were pre-heated to ≈ 250◦ C. This
temperature was arrived at empirically through trial and error. If the temperature
of the mould is too low then the formed glass will possess too much internal stress
and will crack and fracture. On the other hand, for mould temperatures that are
too high the melt is wetting4 on the metal surface and consequently sticks.
8. The glass is annealed around 280◦ C with the temperature decreased to ambient
at 0.3 − 0.1◦ C.min−1 (depending on melt volume).
9. Thermal characteristics of the glass are determined via Differential Scanning Calorime-
try (DSC) (Section 5.3.1).
4
For a general discussion of wetting behaviour turn to Section 9.3.2.
Chapter 3 51
10. Finally, to eliminate any residual stresses in the glass, the samples are re-annealed
at just below Tg for > 1 h and cooled to ambient temperature at 0.3−0.1◦ C.min−1
(depending on melt volume).
Typically glass melts of three different volumes were attempted for confirmation of the
scalability of the compositions; 4 cm3 , 20 cm3 and 40 cm3 which, given the densities of
tellurite glasses (Section 5.2), corresponds to ≈20 g, ≈100 g and ≈200 g respectively.
Figure 3.5 shows the brass moulds used and examples of the resulting glass billets.
Figure 3.5: Brass moulds for glass casting and the resulting billets. Approximate
volumes and masses from left to right: 4 cm3 /20 g, 20 cm3 /100 g and 40 cm3 /200 g
3.7 Results
T
able 3.4 details the attempted compositions and the maximum achieved melt vol-
ume.
We stress that, although some of the attempted compositions did not form glasses, we
can not conclude that the composition is not a glass former. All that can be concluded is
that under our fabricating conditions the composition did not form a glass. This result
indicates that the failed compositions are not suitable for our optical fibre manufactur-
ing process as it stands, however after further optimisation of the melting and casting
conditions they may prove suitable. As such these compositions have been abandoned
for the purposes of this thesis.
The successful glass forming compositions all formed optically clear samples with a slight
green colouration, which is due to the position of the absorption band edge for these
materials which is around 450 nm (see Section 6.2 for more details on the band edge
and optical transmission in general).
Vol.
Sample TeO2 Na2 O MgO ZnO BaO PbO SrO [cm3 ]
TMN1 85 10 5 - - - - 4
TMN2 80 10 10 - - - - 42
TMN3 75 10 15 - - - - 4
TZN1 85 10 - 5 - - - 42
TZN2 80 10 - 10 - - - 42
TZN3 75 10 - 15 - - - 42
TZN4 70 10 - 20 - - - 42∗
TZN5 65 10 - 25 - - - N/A
TZN6 60 10 - 30 - - - N/A
TBN1 85 10 - - 5 - - 22
TBN2 80 10 - - 10 - - 22
TBN3 75 10 - - 15 - - N/A
TBN4 70 10 - - 20 - - N/A
TPN1 85 10 - - - 5 - 4
TPN2 80 10 - - - 10 - N/A
TPN3 70 10 - - - 20 - N/A
TSN1 85 10 - - - - 5 N/A
Table 3.4: Table of glass compositions investigated. Amounts of components are

in molar % and the volumes indicated are the current maximum volumes. Where a
stable glass was not achieved the volume is labelled as N/A. ∗ Some small crystals were
observed in the 42 cm3 billet only, thought to be undissolved ZnO.
We label the compositions of the tellurite glasses in the following way:
The first letter in the label is ’T’, indicating tellurium oxide.
The second letter is taken from the first letter in the chemical symbol of metal
species in the modifying oxide, e.g. zinc oxide → Z, barium oxide → B, etc.
The third letter, which is common to all glasses is ’N’ and indicates the stabilising
compound, sodium oxide.
We changed the molar percentages of the modifying oxides in increments of 5 molar

% for all of the attempted compositions. Accordingly we are able to represent the
modifier concentration with a single number at the end of the label. Where that
Chapter 3 53
number is the multiple of 5 mole % of the modifying oxide, i.e. 1 → 5 mol.%,

2 → 10 mol.%, 3 → 15 mol.%, etc.
3.8 Conclusion
I
n the preceding sections I have discussed the material requirements for fabrication of
a nonlinear fibre device. Specifically we require the glasses to have large nonlinear
coefficients, low absorption and high crystallisation stability. To meet these require-
ments we must select an appropriate material, however, in order to do so a thorough
understanding of the chosen glass system (tellurite) is required. Therefore, to address
this need we have designed a range of compositionally varied potential glasses, so that
the correlation between composition and property can be established. After experimen-
tal determination of the glass forming ability of these proposed compositions we have
arrived at three tellurite glass families; TMN, TZN and TBN, appropriately summarised
by the following formulae:
⎫
TMN: 10Na2 O.xMgO.(90 − x)TeO2 x = 5, 10, 15 ⎪
⎪
⎪
⎬
TZN: 10Na2 O.xZnO.(90 − x)TeO2 x = 5, 10, 15, 20 (3.4)
⎪
⎪
⎪
⎭
TBN: 10Na2 O.xBaO.(90 − x)TeO2 x = 5, 10
They possess different modifying species (Mg, Zn and Ba) with a spectrum of electronic
polarisabilities (Table 3.2). Additionally, within each family of glass, the concentration
of the modifier ranges over at least 5 mol.% (Formulas 3.4). Importantly, all of these
glasses have been successfully fabricated into ≈40 cm3 (200 g) billets which provides a
preliminary indication that these glasses have a high crystallisation stability and could
then be suitable for fibre optic devices.
Chapter 4
Analysis of the Microscopic

Structure of Glass
4.1 Introduction
T
he properties of a material are manifest from the collective properties of the con-
stituents of that material. To understand the origins of a material’s properties
there are two things that need to be considered: what elements is the material made
from and how are they arranged?
In this chapter we seek to discover the arrangement of elements within the glasses under
study. knowing what structural elements are present will enable us to better understand
the origins of trends and changes in the physical and optical properties presented in
Chapters 4, 5 and 6.
We begin this chapter with an example that illustrates the effect of different structural
arrangements of the same elements on the properties of a material. Consider silicon
dioxide (SiO2 ), the elemental components of which are silicon and oxygen in the ratio 1:2.
If these elements are arranged in a regular lattice we obtain the well known crystalline
material quartz. In quartz there are well defined bond lengths between the silicon and
oxygen atoms. However, if the silicon and oxygen are arranged haphazardly the resulting
material is known as fused silica, which is a common glass. As discussed in Chapter 3
the glassy state of a material is typified by chemical bonds that are distorted from the
ideal lengths and angles. Now, these two materials are chemically identical but, owing
to their structural differences, have differing properties.
For example, in Table 4.1 we list side by side some representative properties of crystalline
quartz and fused silica.
55
56 Analysis of the Microscopic Structure of Glass
Table 4.1: Comparison of certain properties of Quartz and Fused Silica. ∗ Values of
the ordinary (ne ) and extraordinary (ne ) refractive indices respectively. + Values for β
tridymite and β cristobalite respectively.
Property Quartz Fused Silica

Refractive Index [@ 627.8 nm] 1.542819 / 1.551880∗ 1.45716
Density [g.cm3 ] 2.65 2.203
Melting / Softening Point [◦ C] 1670 / 1713+ 1665
Owing to the strong dependence of properties on the structural arrangements of atoms

it is therefore critical to determine this structure to gain a proper understanding of
a material. To this end we have undertaken to determine the microscopic structural
arrangement of the tellurite glasses under study. The tool with which we achieved this is
Raman spectroscopy which, as will be seen in the following sections, can provide excellent
information regarding the arrangement of the chemical elements in these materials.
In the next sections we describe the basic principles behind Raman spectroscopy and how
they relate to the molecular structure of tellurite glass. Then we outline the experimental
technique for recording these spectra and discuss the data analysis procedure. Finally, we
present the data for the samples, accompanied by a detailed discussion of the results. In
later chapters, we present data for various properties of the tellutite glasses. We will be
referring to the structural information herein so that property-to-structure relationships
can be established.
4.2 Raman Spectroscopy
R
aman spectroscopy is a measurement of the molecular rotational and vibrational
modes of a material. In solid state systems molecular rotations are impossible. We
can therefore only study the vibrational transitions. The technique relies on inelastic
scattering of monochromatic light from phonons, which essentially involve the transfer
of energy between the incident light and the vibrational modes of the material.
An incoming photon excites a vibrational mode into a virtual energy state. The state
then decays into a real vibrational state. The difference in energy between the initial
state and the final state corresponds to a change of energy, or Raman shift, of the
emitted photon. If the final state is more energetic than the initial state the emitted
photon will have a lower frequency than the incoming photon. This is called the Stokes
shift. Alternatively, if the final state is lower in energy than the initial state then the
emitted photon will have acquired some energy and therefore an increase in frequency
will be observed. This shift to higher frequency is called the Anti-Stokes shift. The
Chapter 4 57
intensity of the Anti-Stokes shifted light is much less than the Stokes shifted light and
therefore Raman spectrometers are configured to detect the Stokes shifted photons.
It is typical to measure Raman intensity with respect to wavenumber (ν̄ [cm−1 ]) and
not frequency. Therefore, to maintain consistency with the literature we adopt this con-
vention. Wavenumber and frequency are proportional to one another and, importantly,
proportional to energy. In SI units the relationships between frequency, energy and
wavenumber are
E
ν= = ν̄c × 102 . (4.1)
h
The measurement of the Raman shift represents a change in frequency Δν, of the input
light determined by
Δν = Δν̄c × 102 (4.2)
Alternatively, in terms of the change in the wavelength of the input light
Δλ = λ2 Δν̄ × 102 . (4.3)
Where λ is the wavelength of the input light. Figure 4.1 is a conversion plot for the
frequency and wavelength shifts for a give Raman shift. The vales of the Raman shifts
have been chosen based on what is typically encountered in tellurite glass and Δλ’s are
calculated at 514 nm, which is the wavelength used in the measurements presented in
Section 4.2.2.
4
x 10
3 35
2.5 30
25
2
Δν [GHz]
Δλ [nm]
20
1.5
15
1
10
0.5
5
0
100 200 300 400 500 600 700 800 900 1000
−1
Raman Shift [cm ]
Figure 4.1: Conversion between Raman shift and Frequency and Wavelength shift.
Δλ calculated using λ = 514 nm (To match experiments performed within this thesis).
Values of the Raman shift are taken between 100 and 1000 cm−1 which correspond to
typical Raman shifts produced in tellurite glass.
In the following analysis we will use the terms wavenumber and frequency interchange-
ably.
4.2.1 Raman Spectroscopy of Tellurite
Raman spectra of tellurite glasses, where tellurium is the only network former, have
five characteristic peaks that correspond to the inhomogeneously broadened vibrational
features found in TeO2 crystals [37, 43]. Figure 4.2 shows a typical Raman spectrum
the details of which will be discussed subsequently. In general all tellurite glasses of
this type vary in the relative intensities of these bands for different compositions of
glass [36, 37, 44, 45].Figure 4.2 shows a representative Raman spectra which has been
deconvolved into its Gaussian modes. The method for performing this deconvolution is
explained later in Section 4.2.3.
Figure 4.2: Example of a deconvoluted raman spectrum (sample: TZN2 see Section
4.2.2 for experimental details). Black: Measured spectrum. Green: Gaussian vibra-
tional modes. Red: Reconstructed spectrum achieved by addition of gaussians. Note:
The fit is poor in the region ν̄ < 400 cm−1 . This is due to the edge filter that is used
to remove the Rayleigh scattered light therefore corrupting the spectra in this region.
The following is a summary of the assignment of the Raman bands taken from Sekiya et
al [37]. Refer to Section 3.4 for details relating to the TeO4 , TeO3+1 and TeO3 polyhedra.
Chapter 4 59
The A band
Located at ≈ 450 cm−1 , results from symmetrical bending and stretching modes of Te-
O-Te bonds found at corner sharing sites of TeO4 , TeO3+1 and TeO3 polyhedra (see
Figure 4.3).
Figure 4.3: Ball and stick representation of the tellurite lattice vibrations that con-
tribute to the Raman A mode. Tellurium = black circles. Oxygen = white circles. From
left to right: Corner sharing subunits: TeO3 with a NBO of the type Te=O, TeO4 and
TeO3+1 with a NBO of the type Te-O− . Red arrows indicate the symmetrical bending
of Te-O-Te bonds. Blue arrows indicate the symmetrical stretching of Te-O-Te bonds.
The B band
Located at ≈ 611 cm−1 , is attributed to anti-symmetric stretching of continuous net-
works of TeO4 subunits (see Figure 4.4). Accordingly the A and B bands are a measure
of the networking of the glass structure.
tribute to the Raman B mode. Tellurium = black circles. Oxygen = white circles. Red
arrows indicate the anti-symmetric stretching in a continuous chain of TeO4 subunits.
The C band
Located at ≈ 659 cm−1 , is caused by anti-symmetric vibrations of Te-ax O-eq constructed
by two unequivalent Te-O bonds (see Figure 4.5). It is often reported that this band
is related to TeO4 content, however, there are many contributing vibrations to this
band from the TeO3+1 and TeO3 subunits where the shared oxygen is bonded to a
neighbouring tellurium with unequivalent bonds, e.g. Figure 4.5 shows a TeO3 sharing
an oxygen with an axial TeO4 oxygen.
tribute to the Raman C mode. Tellurium = black circles. Oxygen = white circles. Red
arrows indicate the anti-symmetric vibrations of two unequivalent Te-O bonds. In this
example one oxygen from a TeO3 subunit with a NBO of the type Te=O is bonded to
an axial TeO4 oxygen.
The D band
Located at ≈ 716 cm−1 , is due to stretching modes of Te-O− and Te=O bonds containing
non-bridging oxygens (NBO) which are formed by TeO3+1 and TeO3 polyhedra (see
Figure 4.6). As such the D band is the most unambiguous indicator of the presence of
TeO3+1 and TeO3 sub units.
tribute to the Raman D mode. Tellurium = black circles. Oxygen = white circles. Red
arrows indicate the stretching of the NBO’s both Te=O and Te-O− .
Chapter 4 61
The E band
Located at ≈ 773 cm−1 , band is ascribed to stretching modes of non-bridging oxygen
atoms of the type Te-O− in TeO3+1 and TeO3 subunits (top of Figure 4.7).The Raman
E peak is a good measure of one particular type of NBO (Te-O− ), which we shall denote
NBO− . There are also contributions to this band from the symmetric stretching of
continuous network of TeO4 subunits (bottom of Figure 4.7).
tribute to the Raman E mode. Tellurium = black circles. Oxygen = white circles. Top:
Red arrow indicates the stretching of Te-O− type NBO. Bottom: Red arrows indicate
the symmetric stretching of a continuous chain of TeO4 subunits.
The intensities of the Raman bands are used as a measure of the glass structure as
follows in Table 4.2:
Raman Band Characteristic Structural Elements Nomenclature

A Te-O-Te bonds Network connectivity
B Connectivity of TeO4 chains TeO4 connectivity
C Te-ax O-eq bonds TeO4 concentration
D TeO3 and TeO3+1 subunits TeO3 concentration
E Non bonding oxygens of the type Te-O− NBO− content
Table 4.2: Raman band assignment to structural subunits and nomenclature.
In addition to the five band listed above, all glasses possess a spectral feature known
as the Boson peak. This is the lowest frequency (ν̄ < 100 cm−1 ) scattering feature and
can account for a large portion of the total Raman intensity. The origin of this feature
is still poorly understood and remains somewhat controversial. It is, however, believed
to be related to medium range order within the glass [46–48]. We do not attempt to
analyse the Boson peak data in our spectra because of the lack of accepted knowledge
relating to its origin. Also, during the measurement the Raman shifted light was filtered
from the Rayleigh scattered pump light with an edge filter consequently the Boson peak
was not detected for the TZN and TBN glasses and only partially detected for the TMN
glass.
4.2.2 Experimental Details
The Raman spectra presented were kindly recorded by Dr. Elizabeth Carter at the
University of Sydney1 . We prepared spectroscopic samples by cutting ≈ 2 mm thick
slices from the 4 cm3 rectangular billets. These slices were then polished to optical
quality on both faces. The following experimental details were provided by Dr. Carter:
Raman spectra were recorded using a Renishaw Raman InVia Reflex Microscope. The
excitation wavelength of 514 nm was provided by an argon ion laser operating at a max-
imum power of 5.2 mW. This was attenuated using neutral density (ND) filters from 0%
to 99% depending on sample. The light is collected in the 180 ◦ configuration using a
Leica DMLM microscope equipped with 50x, 20x and 5x objectives with NA=0.75, 0.40
and 0.12 respectively. Raman signals were detected with an air cooled charge coupled
device (CCD) camera equipped with a holographic notch filter with 2400 lines.mm−1 for
removal of the Rayleigh scattered pump light. The spectrometer is regularly calibrated
using the 520.50 cm−1 Raman band of silicon to an accuracy of ±0.10 cm−1 . We ob-
tained triplicate spectra for each sample from 100 cm−1 to 1000 cm−1 with a resolution
of 0.25 cm−1 .
4.2.3 Data Analysis
Upon review of the literature it was noticed that there are several methods for analysing
the spectroscopic data. There appears to be a concerning lack of consistency in the
techniques and despite the frequent qualitative agreement there is a need to determine a
single and well justified approach. For example, it is common for authors to discuss the
relative intensities of the peaks in the Raman spectra. However, this is frequently done
in terms of the amplitudes of the peaks. This is an insufficient definition for intensity in
particular for the broadened peaks observed in glassy materials. There are contributions
to the peaks that are frequency shifted as a result of the distorted bond lengths in the
amorphous system. It is conceivable that changes in composition, for example, could
1
As a part of the NCRIS scheme
Chapter 4 63
distort the structural element responsible for the spectral feature without changing the
number of these elements. Simply measuring the peak hight would not be sensitive to
such a change. The only way to guarantee that all of the contributions to the particular
vibrational mode are accounted for is to take the area under the peak as representing the
intensity. Another questionable technique that is used is to normalise the amplitudes
of each peak to that of the Raman C band. As we have already stated, this band is
frequently ascribed to the presence of the TeO4 subunit which is a mistake that has
propagated through the literature. We have instead chosen to normalise the intensities
of the individual Raman bands to the total intensity. That is we take the ratio of the
area under the peak to that of the total area under all peaks.
The Raman spectra within this thesis were analysed via the following procedure.
1. Spectra for each sample were recorded three times and averaged to reduce the
random error.
2. Background removal was performed by subtracting the minimum value from the
spectra (i.e. where the materials have no Raman response.)
3. Using a commercial software package (OriginPro 8) we de-convolve the spectra

into five Gaussians centred at approximately: 450, 611, 659, 716 and 773 cm−1
[37]. No restrictions were placed on the width, amplitude or position of the peak.
4. Raman band intensities are calculated by integrating the de-convolved Gaussians.
5. We normalise to the total Raman intensity which is the total area under the
spectrum. This normalisation procedure provides relative Raman intensities which
can then be compared between samples.
6. The intensity error was estimated from the maximum standard error for the am-
plitudes.
4.2.4 Results and Discussion
Shown in Figures 4.8 to 4.10 are the recorded Raman spectral data for the three glass
series. The low frequency cut off of the TMN glasses is at approximately 100 cm−1 as
opposed to the cut off of 200 cm−1 in the case of the TZN and TBN spectra. This is
due to the inadvertent use of a different filter during the measurement. As a result the
tail of the Boson peak is observed in the spectra for the TMN glasses.
The immediate observation that can be made about all of the glass series is that as
the concentration of the modifier is increased the spectra take on more high frequency
content, around 800 cm−1 , to the detriment of the lower frequency components of the
spectra near 450 cm−1 . To assess the changes in the spectra more precisely, an analysis
of the deconvolved spectra is necessary.
Figure 4.8: Raman spectra for the TMN glass series. The arrows indicate general
trends observed as the Mg2+ concentration is increased. The spectral content is trun-
cated below ≈100 cm−1 due to the use of filter in the spectrometer.
Figure 4.9: Raman spectra for the TMN glass series. The arrows indicate general
trends observed as the Zn2+ concentration is increased. The spectral content is trun-
cated below ≈200 cm−1 due to the use of filter in the spectrometer.
The deconvoultion of the spectra was performed as per the steps outlined in Section
4.2.3. We assess the quality of the de-convolution via the correlation coefficient, R2 , for
the fitted Gaussians where 0 R2 1 and 1 indicates a perfect fit. In all cases for the
data obtained here R2 > 0.997.
Chapter 4 65
Figure 4.10: Raman spectra for the TMN glass series. The arrows indicate gen-
eral trends observed as the Ba2+ concentration is increased. The spectral content is
truncated below ≈200 cm−1 due to the use of filter in the spectrometer.
Table 4.3 contains the relative Raman band intensities. We have indicated the position
of each band relative to the analogous band for the crystalline phase, however, we note
that in the glass phase these centre locations are shifted. Centre positions for the bands
as found in the glasses are tabulated in Table 4.4. In general, an increase in the frequency
of a Raman band indicates an increase in bond strength which is produced by shortening
of the bond length. Conversely, for a decrease in the Raman band frequency there is an
implied increase in bond length and thus an decrease in bond strength. If we refer to the
Szigeti relation [49], which relates the vibrational frequency of the stretching vibration
to the inter-atomic force F and the reduced mass, μ = m1 m2 /(m1 + m2 ), via:

1 F
ν= . (4.4)
2π μ
The force F, is Coulombic which is ∝ 1/r2 (where r is the bond length). Therefore F
will increase if the bond length decreases and decrease if the bond length increases.
Table 4.3: Relative Raman band intensities for glasses under study. Values have been
multiplied by 100 to give a %. The intensities have a maximum error of ±0.5
Raman Band
A B C D E
Sample (450 cm−1 ) (611 cm−1 ) (659 cm−1 ) (716 cm−1 ) (773 cm−1 )
TMN1 33.0 6.8 28.9 2.3 29.1
TMN2 31.6 5.5 28.0 15.8 19.2
TMN3 29.7 4.1 23.9 39.8 2.5
TZN1 34.4 5.9 33.2 6.8 19.7
TZN2 33.4 4.9 31.7 8.2 21.8
TZN3 33.0 3.9 30.8 9.7 22.6
TZN4 31.6 3.2 30.5 10.5 24.2
TBN1 30.8 4.8 37.2 12.5 14.7
TBN2 26.5 5.0 26.4 37.7 4.5
Table 4.4: Table of Raman band centre positions obtained from the Gaussian decon-
volution. Bands are labelled with unperturbed centre positions for reference.
Raman Band
A B C D E
Sample (450 cm−1 ) (611 cm−1 ) (659 cm−1 ) (716 cm−1 ) (773 cm−1 )
TMN1 454 597 661 718 752
TMN2 451 595 660 725 785
TMN3 448 592 659 753 817
TZN1 451 596 663 721 766
TZN2 447 594 663 724 770
TZN3 443 592 664 728 774
TZN4 438 591 667 733 778
TBN1 456 591 663 726 778
TBN2 457 588 654 735 792
We summarise the observed trends in the Raman bands of the studied glass samples as:
Raman A Band: The relative intensity of the Raman A band decreases with
increasing modifier content for the TMN, TZN and TBN glasses (Figure 4.11).
This is interpreted as representing a progressive decrease in the network connec-
tivity of the glasses by the conversion of Te−O−Te bonds into Te−O−M−O−Te
chains, where M represents Mg, Ba or Zn. The effect is largest for the Ba2+
containing glasses, because Ba2+ possesses the smallest electronegativity and the
highest coordination number of the three modifiers (see Table 3.2). Thus Ba2+
demands more oxygens and does so with greater strength than Mg2+ and Zn2+ .
Consequently, disruption of the Te−O−Te bonds is produced, in order of degree,
by Ba2+ , Mg2+ and Zn2+ .
Chapter 4 67
36
34 TZN1
Network Connectivity [arb. units]

TZN2
TMN1 TZN3
32
TMN2 TZN4
TBN1
30
TMN3
28
TBN2
26
24
5 10 15 20
Modifier Concentration [mol. %]
Figure 4.11: Network connectivity (Raman A band) variation with modifier content
(Mg2+ , Zn2+ and Ba2+ ) for the TMN, TZN and TBN glasses.
460
TBN2
TBN1
455 TMN1
Raman A Band Centre [cm ]
-1
TZN1 TMN2
450
TZN2 TMN3
445
TZN3
440
TZN4
435
5 10 15 20
Figure 4.12: Position of the Raman A band variation with modifier content (Mg2+ ,
Zn2+ for the TMN, TZN and TBN glasses.
The frequency of the Raman A band decreases with increasing modifier content
for the TMN and TZN glasses and increases for the TBN glasses (see Figure
4.12). We note that this implies an increase in the Te−O−Te bond lengths which
is associated with the modifiers (Mg2+ and Zn2+ ) distorting and breaking up
the network connectivity). The Ba2+ modified TBN glasses, despite displaying a
decrease in network connectivity (Figure 4.11), display an increase in Te−O−Te
bond strength, implying that the bond lengths are shortening.
Raman B Band: Figure 4.13 shows that for increasing modifier content the rel-
ative intensities of the Raman B bands decrease for the TZN and TMN glasses.
This indicates that, as Zn2+ and Mg2+ is added the connectivity of the TeO4 sub-
units is destroyed. Conversely, for increasing Ba2+ content the TeO4 connectivity
increases. This can be understood by considering the shift to higher frequencies
displayed by the Raman A peak in the TBN glasses. We interpret this to imply
that the TeO4 subunits are clustering into chains so that, whilst the overall con-
nectivity of the network decreases for increased Ba2+ (Figure 4.11), there is local
formation of TeO4 chains with less perturbed and thus shorter Te−O−Te bond
lengths.
7.5
7.0
TMN1
6.5
TeO4 Connectivity [arb. units]
6.0 TZN1
5.5 TMN2
5.0 TBN2
TBN1 TZN2
4.5
TMN3
4.0 TZN3
3.5
TZN4
3.0
2.5
5 10 15 20
Figure 4.13: TeO4 connectivity (Raman B band) variation with modifier content
The Raman B band frequency decreased for increasing modifier content for TMN,
TZN and TBN glasses as shown in Figure 4.14. This behaviour implies that the
bond lengths for the TeO4 chains increase which is likely to be as a result of
Chapter 4 69
598
TMN1
596 TZN1
Raman B Band Centre [cm ]

-1
TMN2
594 TZN2
592 TZN3 TMN3
TBN1
TZN4
590
588
TBN2
5 10 15 20
Figure 4.14: Position of the Raman B band variation with modifier content (Mg2+ ,
the increase in the axial Te−O bond (0.21 nm→ 0.24 nm) associated with the
conversion of TeO4 to TeO3+1 (Figure 3.4) which is common to all of the glasses
under study as shown in Figure 4.17.
Raman C Band: The Raman C band is the most ambiguous band as its presence
is indicative of the anti-symmetric vibrations of unequivalent Te-O bonds in a Te-
O-Te linkage that can occur between TeO4 , TeO3+1 and TeO3 subunits. It is
frequently claimed that this band is representative of the presence of the TeO4
subunit alone as this feature is observed in pure TeO2 crystals which are known
to be comprised exclusively of the TeO4 . This is an erroneous assertion because
as has been shown the inhomogeneously broadened feature observed in glassy
tellurite systems has multiple contributing structural elements [36, 37, 44, 45].
We can, however, utilise the information from this band to determine the degree
to which the connectivity of the glass is effected by the changes in the modifier
concentration.
Figure 4.15 shows that the number of Te-ax O-eq Te bonds is reduced as the con-
centration of the modifier is increased. The cause for this lays in the breaking up
of the network by the modifying species via the creation of TeO3 subunits and
NBOs.
Plotted in Figure 4.16 are the centre positions of the Raman C band for each
glass sample. It is noted that the position of the C band for the TZN series
40
TBN1
Raman C Band Relative Intensity 36
TZN1
TZN2
32 TZN3 TZN4
TMN1
TMN2
28
TBN2
TMN3
24
20
5 10 15 20
Figure 4.15: Relative intensity of the Raman C band variation with modifier content
progressively moves to higher frequencies as the amount of Zn2+ is increased.

This can be understood by considering that as TeO4 →TeO3 both the axial and
equatorial bonds decrease in length thus increasing in vibrational energy. The
change to higher frequencies of this band therefore indicates an overall shortening
of the Te-ax O-eq bonds which occurs as TeO4 is converted into TeO3 .
Conversely, the position of the C band for the TMN and TBN series is shifting
towards lower frequencies as additional modifier is introduced. In the case of TMN
glasses it is possible that this is resulting from the presence of the intermediate
TeO3+1 subunit with its elongated axially bonded oxygen. However, this is difficult
to determine. As for the TBN glasses a strong shift towards the lower frequencies is
observed if we also consider the data for the B band we see that there is evidence to
suggest that the TeO4 subunits are taking part in the connectivity of the network,
more so that the other species.
Raman D Band: There is an increase in the relative intensity for the Raman D
band with increasing modifier content for the TMN, TZN and TBN glasses. As
this peak is associated with the presence of TeO3+1 and TeO3 subunits we conclude
that the addition of Mg2+ , Zn2+ and Ba2+ promotes the conversion of TeO4 into
TeO3 via the TeO3+1 intermediate. This behaviour is commonly seen in tellurite
glass in which an increase in modifier distorts the TeO4 subunits to produce TeO3
[44].
Chapter 4 71
670
668
666 TZN4
Raman C Band Centre [cm ]

-1 664 TBN1
TZN3
662 TZN1 TZN2
660 TMN1
TMN2
658 TMN3
656
654 TBN2
652
650
5 10 15 20
Figure 4.16: Position of the Raman C band variation with modifier content (Mg2+ ,
45
40 TBN2 TMN3
35
TeO3 Concentration [arb. units]
30
25
20
15 TMN2
TBN1
10
TZN3 TZN4
TZN2
5 TZN1
TMN1
0
5 10 15 20
Figure 4.17: TeO3 concentration (Raman D band) variation with modifier content
755
TMN3
750
745
Raman D Band Centre [cm ]
-1
740
TBN2
735
TZN4
730
TBN1 TZN3
725 TMN2
TZN2
720
TZN1
TMN1
715
5 10 15 20
Figure 4.18: Position of the Raman D band variation with modifier content (Mg2+ ,
The Raman D band frequency increases for increasing modifier content for TMN,
TZN and TBN glasses (see Figure 4.18). This peak is indicative of the conversion of
TeO3+1 into TeO3 which is accompanied by a shortening of the equitorial oxygen
bonds from 0.19 nm→ 0.185 nm (Figure 3.4) and the axial oxygen bond from
0.21 nm→ 0.20 nm.
Chapter 4 73
Raman E Band: For increasing modifier content we observe a decrease in the

intensity of the Raman E band for the TMN and TBN glasses and an increase
for the TZN glasses. This indicates that both Mg2+ and Ba2+ are inhibiting the
formation of non-bonding oxygens of the type NBO− , whereas Zn2+ is promot-
ing it. It is possible that this behaviour is somehow related to the Zn-O bonds
being the least ionic of all the modifier-oxygen bonds, as Zn2+ has the highest
electronegativity of all the modifiers (Table 3.2). We are, however, unsure of the
precise mechanism and thus leave it for future study.
30
TMN1
25 TZN4
TZN3
TZN2
NBO Content [arb. units]
TZN1
20
TMN2
15
TBN1
10
-
5
TBN2
TMN3
0
5 10 15 20
Figure 4.19: NBO− concentration (Raman E band) variation with modifier content
As shown in Figure 4.20, the Raman E band frequency increases for increasing
modifier content for TMN, TZN and TBN glasses. This band is also associated
with the conversion of TeO3+1 into TeO3 and thus by the same reasoning for the
Raman D band we expect to see an increase in the vibrational frequency.
820
TMN3
810
800
Raman E Band Centre [cm ]
-1
TBN2
790
780 TBN1 TMN2 TZN4

TZN3
TZN2
770 TZN1
760
750 TMN1
740
5 10 15 20
Figure 4.20: Position of the Raman E band variation with modifier content (Mg2+ ,
4.3 Conclusion
T
hrough the measurements of the Raman spectra and via comparison to the pub-
lished data on analogous crystalline phases of TeO2 and similar tellurite glasses
we have been able to determine the structural units that are present in the glasses under
study. The compositionally driven trends in the relative magnitudes of the Raman bands
will be used in later Chapters to aid in the identification of the physical mechanisms
leading to other material properties, such as the linear and nonlinear refractive indices.
The results obtained are consistent with similar studies reported in the literature and
indicate that, in general, the addition of a modifying specie will promote the conversion
of TeO4 subunits into TeO3 subunits via the intermediate TeO3+1 state. Additionally,
there is a decrease in the network connectivity associated with this process which is more
pronounced for modifiers with high coordination numbers and low electronegativities.
Chapter 5
Measurements of Physical and

Thermal Properties
5.1 Introduction
I
n this Chapter we present measurements of a number of the physical and thermal
properties of the glasses that are important for their application in determining the
suitability of glasses for use in nonlinear devices. The measured properties include:
Density, Section 5.2
Molecular mass, Section 5.2
Molar volume, Section 5.2
Glass transition temperature and enthalpy, Section 5.3.1
Crystallisation temperature and enthalpy, Section 5.3.1
Thermal expansion coefficient, Section 5.3.2
Each measurement is presented with an explanation of the relevant underlying physics

followed by a detailed description of the experimental technique. We conclude each
section by presenting the data followed by a discussion of the results and correlating the
trends in these properties to the structural changes occurring in the glasses (see Chapter
3) as the relative proportions of the components are varied.
5.2 Density, Molecular Mass and Molar Volume
T
he density of a glass is principally determined by the chemical constituents of the
glass and how they are arranged in space. As discussed in Section 3.2 the density
75
76 Measurement of Physical and Thermal Properties
of a glass can vary depending on the rate of cooling used to form the glass from the
melt. For this reason care was taken to anneal the samples as close to Tg as possible
and cool them slowly to maximise the density (see Section 3.6 and Figure 3.2).
Molar volume is defined as the volume occupied by one mole of a substance. Thus, in
order to calculate the molar volume one requires the molecular mass and the volume,
which can in turn be obtained from the density. As glasses can be considered to be solu-
tions, the molar volume is generally considered to be a more useful parameter than the
density because it provides less ambiguous structural information. We can demonstrate
the molar volume’s ability to reveal structural information with the following generalised
example. Consider the following situation: A particular glass has some proportion of a
monovalent modifier A+ in it. We then produce an identical glass, except, we substitute
a different monovalent modifier B + that has, for example, a larger molecular mass than
A+ . In this situation the B + containing glass will have a higher density, however, if
there is no difference in the way B + in incorporated into the glass matrix then there
will be no change in the molar volume.
This example clearly demonstrates the superiority of the molar volume over the density
as a measure of glass structure. In the following Sections the details of the measurements
and the interpretation of the data are discussed.
Densities of the glasses were measured by using the Archimedes method [50]. Shown
in Figure 5.1 is a schematic of the density measurement experiment. The force due to
gravity on the sample with mass ms when suspended in the air is expressed as
Fs = m s g (5.1)
When the sample is immersed in the liquid it displaces a volume equal to its own. This
produces a buoyancy force Fb , in the opposite direction to gravity. The net force on the
immersed sample Fi is
F i = F s − Fb (5.2)
Therefore
Fb = Fs − Fi (5.3)
Taking the ratio of Fs /Fb gives

Fs Fs
= (5.4)
Fb Fs − F i
Chapter 5 77
Archimedes principle tells us that the displaced volume is equal to the volume of the
sample thus Equation 5.4 can be written as
ρs V s ρs Fs
= = Sρ = (5.5)
ρl V s ρl F s − Fi
Where we define Sρ to be the specific gravity. We now make the replacement of Fs →

Wout and Fi → Win to denote the weight of the sample out of the liquid and in the liquid
respectively, finally yielding:
Wout
Sρ = (5.6)
Wout − Win
The following is a description of the procedure for measuring the density.
With reference to Figure 5.1 we measure the mass of the suspended sample which
has been cleaned thoroughly with an organic solvent so as to remove any dirt and
oils from previous handling. Let this mass be denoted mout .
Glass sample Reference liquid
Fb
Fg
Fg
Figure 5.1: Experimental configuration for Archimedes Density Measurement. Left:

Sample is weighed suspended above a beaker the force of gravity Fg acts on the sample
and is recorded as a weight on a set of scales. Right: Sample is immersed in the water
where the buoyancy of the sample produces an opposing force, Fb , to Fg hence reducing
the measured weight.
The sample is then immersed in a liquid with known density and temperature.
We take care to ensure that there is no air trapped under the sample as this will
reduce the measured density.
The mass of the immersed sample is then measured. Let this mass be denoted
min .
Using Equations 5.5 and 5.6 with the value for waters density ρwater = 1.00 g.cm−1
(at 25 ◦ C) we obtain the density of the sample.
We repeat this process several times and take the mean value obtained for the
density. In this way we minimise the errors such as fluctuations in the measured
submerged weight, which are primarily due to vibrations of the equipment.
It should be noted that we made no corrections for the fact that the density of water
changes with temperature. All measurements were preformed at temperatures conser-
vatively estimated to be between 20◦ C and 30◦ . Base on literature reported values of
the density of pure water we calculate the error associated with this procedure is in the
order of 0.26% [51].
The results of the density measurements are tabulated in Table 5.1 and plotted against
modifier content in Figure 5.2. The measurement error, determined from the standard
error accumulated over five measurements per sample, is found to be no larger than the
error associated with not correcting for the temperature dependence of the density of
water, i.e. ≈ ± 0.006 g.cm−3 .
Molar Mass Density Molar Volume

Sample [g.mole−1 ] [g.cm−3 ] [cm3 .mole−1 ]
TMN1 143.87 5.18 27.77
TMN2 137.91 5.06 27.25
TMN3 131.94 4.95 26.66
TZN1 145.93 5.23 27.90
TZN2 142.02 5.18 27.42
TZN3 138.11 5.15 26.82
TZN4 134.20 5.13 26.15
TBN1 149.52 5.22 28.64
TBN2 149.21 5.25 28.42
Table 5.1: Calculated molecular masses, Densities measured via the Archimedes
method in pure water. Maximum measurement error of ± 0.006 g.cm−3 . Molar volumes
calculated from measured densities with a maximum error of ± 0.01 cm3 .mole−1 .
We can calculate the molar mass of the glass from the following equation:

N
M mglass = n i M mi (5.7)
i=1
Chapter 5 79
5.3
TZN1 TBN2
5.2 TBN1
TZN2
TMN1 TZN3
Density [g.cm ] TZN4
-3
5.1
TMN2
5.0
TMN3
4.9
5 10 15 20
Figure 5.2: Densities of the TMN, TZN and TBN glass series plotted against modifier
content.
Where ni is the molar fraction of the ith component so that

N
ni = 1
i=1
and M mi is the molar mass of the ith component. The molar volume is then calculated
from the density, ρ, and molar mass of the glass via:
M mglass
VM = . (5.8)
ρ
The volume occupied by one of the components of the glass is called the partial molar
volume. The partial molar volume of the ith component Vi is defined as:

∂VM
Vi = (5.9)
∂ni
We have tabulated the molar masses for each glass and the corresponding molar volumes
in Table 5.1. Figures 5.3 and 5.4, respectively, are plots of the dependence of molar mass
and molar volume on the modifier content for each glass family.
In Figure 5.2 we see a clear, decreasing trend for the TZN and TMN series with increasing
modifier content. The TBN series displays an increase in density with increasing modifier
content. These results can be interpreted in the following way: As we increase the
155
150
TBN1 TBN2
TZN1
Molecular Mass [g.mol ]
-3
145
TMN1 TZN2
140
TZN3
TMN2
135 TZN4
130
TMN3
125
5 10 15 20
Figure 5.3: Calculated molar masses (Equation 5.7) for TMN, TZN and TBN glass
series plotted against modifier content.
29.0
28.5 TBN1
TBN2
Molar Volime [cm .mol ]
-1
28.0 TZN1
3
TMN1
27.5 TZN2
TMN2
27.0
TZN3
TMN3
26.5
TZN4
26.0
5 10 15 20
Figure 5.4: Calculated molar volumes (Equation 5.8) of TMN, TZN and TBN glass
series plotted against modifier content.
Chapter 5 81
modifier content, we are in turn decreasing the Te content. The masses for the various
elements are ordered as: Mg<Zn<Te<Ba. Therefore if we ignore volume changes due
to structural rearrangements and consider only the masses of the components, we can
see that substitution of Te with Mg, for example, will decrease the mass per unit volume
of the system and therefore decrease the density. Alternatively, when Te is substituted
by Ba the converse is true.
The variation of molar volume with modifier content is essentially linear for all considered
modifying species. Accordingly, it is straightforward to estimate and compare the partial
molar volumes simply as the gradients of the interpolated lines in Figure 5.4. In doing so
we see that the TMN and TZN series have very similar slopes and thus Mn2+ and Zn2+
must be incorporating them selves into the matrix in approximately the same manner.
On the other hand, the TBN series has a smaller gradient which leads us to conclude
that Ba2+ is producing a more tightly packed matrix. We see support for this in the
Raman spectroscopy data for the TBN glasses (Section 4.2.4) where the trend in the
Raman B peak suggests an increase in the network connectivity for increasing amounts
of barium.
5.3 Thermal Properties
O
ur overall aim is produce glasses which could be used for fabricating optical fibres.
We therefore require glasses with excellent thermal stability (Section 3.5). We
have determined the thermal stability for each of the glass compositions via two methods.
Firstly by observation; when we attempt to form a glass from the melt we can determine
qualitatively whether or not the composition has a good glass forming stability simply
by noting if it forms a glass. Secondly, and more quantitatively, we can measure the
temperatures at which the glass undergoes phase transitions and use this data as a
measure of glass forming stability.
5.3.1 Differential Scanning Calorimetry
Differential Scanning Calorimetry or DSC, is a measurement technique that provides us

with the necessary phase information in order to determine the glass forming stability.
Figure 5.5 is a schematic illustration of a DSC, showing all of the essential elements of
such a device.
The basic operating principle is the following: The sample under study is heated in
a sample pan simultaneously with a reference sample (empty sample pan) inside of a
Figure 5.5: Schematic representation of a DSC apparatus. The temperature of sample

and reference pans are increased at a constant rate. The control PC adjusts the heat
flow into the pans in order to achieve this.
thermally insulated container. The rate of temperature increase is kept constant for
both sample and reference by a computer controller. In order to achieve this, during
phase transitions of the sample, either more or less heat (depending on the nature of
the transition, i.e. whether it is exothermic of endothermic) will need to be put into
the sample. Thus a plot of heat flow against temperature (Figure 5.6) will reveal these
phase transitions and the temperatures at which they occur.
With reference to Figure 5.6, the important features of the graph are the:
Beginning at the left of Figure 5.6 there is a horizontal line of heat flow q W.s−1 ,
vs. temperature T . From this section of the data the heat capacity of the glass
can be determined by:
q/t q
Cp = = (5.10)
ΔT /t ΔT
where ΔT /t ◦ C.s−1 is the heating rate.
The next feature of interest is the glass relaxation peak otherwise known at the
glass transition temperature. We adopt the conventional definition of the onset
temperature Tg , of the glass relaxation as the intersection of the tangents to the
region of constant heat flow prior to relaxation and the region of constant rate of
change of heat flow during the relaxation. These tangents are shown in Figure 5.6.
Chapter 5 83
The glass relaxation temperature1 represents the temperature at which the atoms
and structural sub units become mobile and thus the glass takes on the nature of
an extremely viscous liquid.
The glass relaxation peak has also associated with it an enthalpy of relaxation
ΔHg . This corresponds to the amount of energy absorbed during the transition
from glass to super cooled liquid. ΔHg is proportional to the area under the glass
relaxation peak and normalised to the sample mass to obtain units of J.g−1 .
Following the relaxation peak there is another region of approximately constant

heat flow vs. temperature. This region can be used to determine the heat capacity
of the glass in the super cooled liquid state using Equation 5.10. The heat capacity
in this region is always greater than the heat capacity in the solid region.
The crystallisation peak has an onset temperature Tx , which is defined as the

intersection of the two tangents, indicated in Figure 5.6. In addition to the onset
temperature we also define the peak crystallisation temperature Tc .
The enthalpy of crystallisation ΔHc is determined via integration of the crystalli-

sation peak and represents the amount of energy released during crystallisation.
Figure 5.6: Example of DSC trace, red line, (sample TZN1) with key temperature fea-
tures indicated. The determination of enthalpies requires the introduction of baselines
(black) to bound the peaks for integration.
1
In fact, this transition occurs over a range of temperatures due to the amorphous nature of the glass.
Despite this, the convention is to specify one temperature which indicates an onset of this transition.
5.3.1.1 Experimental Details
The particular DSC instrument used was a TA Instruments DSC 29202 . Calibration of
the device was performed using the following procedure:
1. We used gold sample pans and an air environment. Our reasoning is that these
conditions match exactly the melting conditions we used for producing the glass
(Section 3.6).
2. A baseline is obtained by heating two empty sample pans over a temperature range
of 100 ◦ C to 550 ◦ C with an isochronal heating rate of 10 ◦ C.min−1 .
3. The DSC’s cell constant was calibrated by recording a DSC trace for zinc which
displays a sharp, well defined melting peak at 419.53 ◦ C. We chose zinc as a
calibration standard because its melting point falls within our desired measurement
range.
4. Next we calibrated for the absolute temperature scale by performing a DSC mea-
surement of the melting temperature of tin, which melts at 231.93 ◦ C.
We then performed DSC on all tellurite glass samples the results of which are listed in
Table 5.2. The temperature range for the measurements was 250 ◦ C to 500 ◦ C with an
isochronal heating rate of 10 ◦ C.min−1 . This heating rate is commonly used for DSC
studies of glasses [52].
Initially, we used ≈ 20 mg samples which were ground to a fine powder in an agar

mortar and pestle. This produced DSC traces with very noisy crystallisation peaks. We
interpreted this to be due to individual particles of glass crystallising, but due to their
small size the crystal growth halts quickly. We then turned to using single ≈ 20 mg
chips of glass produced by smashing the edges off of large glass samples with a mortar
and pestle. We collected the fragments which had approximately the right mass for
use in the DSC. The DSC traces acquired in this way did not present the same noise
in the crystallisation peaks as once the crystal growth has been seeded it can progress
unencumbered through out the entire sample. The DSC traces are shown in Figure 5.7.
5.3.1.2 Results and Discussion
Figure 5.7 shows the DSC traces for all glass samples and the temperatures and en-
thalpies measured are listed in Table 5.2. It is interesting to note that the position of
2
Kindly provided for our use by Milena Ginic-Markovic and Rachel Pillar at Flinders University
Chapter 5 85
Tg does not change significantly for increasing modifier content in the TZN and TBN
series, however, for the TMN series there is a near linear increase in Tg with increasing
Mg content of ≈ 2.2 ◦ C.mol−1 . With the exception of TZN3, all of the crystallisation
peaks were easily determined.
Figure 5.7: DSC traces for TMN (magenta), TZN (red) and TBN (blue) glass series.
The vertical scale is in W.g−1 with an arbitrary vertical offset introduced to separate
the traces. The trends in Tg have been indicated by the dashed lines.
For all glass series, the amount of modifier has a significant effect on the position and
enthalpy of crystallisation which has ramifications for the glass forming stability [53].
We can estimate the glass crystallisation stability via [18]:
ΔT = Tg − Tx . (5.11)
Where values of ΔT > 100 ◦ C are generally considered to represent glasses with high
crystallisation stability [53]. Glasses with ΔT > 100 ◦ C will require a reasonably large
amount of heat energy to activate the crystallisation processes (i.e. 150 J.g−1 [18]) and
are therefore considered stable. A ΔT ≈ 100 ◦ C also provides us with a large working
range for processes such as optical fibre preform extrusion, where we are required to
heat the glass beyond Tg and force it through a die (see Section 9.4). Glasses with
a ΔT > 100 ◦ C can comfortably be heated above Tg and softened without inducing
crystallisation. Note in Table 5.2 that all glasses under study have ΔT > 100 ◦ C which
is to be expected as we selected these glasses from the investigated compositions for
their ability to be produced in large volumes (Section 3.7).
Tg Tx Tc ΔT ΔHg ΔHc
Sample [◦ C] [◦ C] [◦ C] [◦ C] [J.g−1 ] [J.g−1 ]
TMN1 295 436 446 141 7.49 73.7
TMN2 307 435 497 128 8.24 69.6
TMN3 317 453 478 136 6.72 105
TZN1 291 405 423 114 7.63 37.0
TZN2 292 421 455 129 7.49 25.2
TZN3 293 464 484 171 8.44 ≈1.2
TZN4 293 465 482 172 8.60 17.6
TBN1 287 454 475 167 3.72 15.8
TBN2 288 410 436 122 5.78 17.2
Table 5.2: Thermal data for TMN, TZN and TBN glass series. Temperatures are
accurate to ±2 ◦ C and enthalpies are accurate to approximately 1%.
180
TZN3 TZN4
170
TBN1
160
150
ΔT [ C]
140
TMN1
o
TMN3
130 TZN2
TMN2
120 TBN2
TZN1
110
100
5 10 15 20
Figure 5.8: Plotted values of ΔT=Tg −Tx for TMN, TZN and TBN glass series. This
is a measure of crystallisation stability where ΔT > 100 indicates a stable glass.
Chapter 5 87
120
100 TMN3
80
TMN1
ΔHc [J.g ]
TMN2
-1
60
40 TZN1
TZN2
20 TZN4
TBN1 TBN2
TZN3
0
5 10 15 20
Figure 5.9: Plot of enthalpy of crystallisation ΔHc for TMN, TZN and TBN glass
series.
Our criterion for a stable glass is ΔT > 100◦ C. Both TZN3 and TZN4 have the largest
value for ΔT and therefore represents the most stable glasses. However, during the glass
production stage it was noted that the 42 cm3 TZN4 billet showed a small amount of
crystallisation. We believe this to be an issue of reaching the solubility limit of ZnO
in TeO2 and not glass forming stability. The ΔT > 100◦ C criterion has its limitations
and, in fact, is much more related to the workability of the glass at temperatures above
Tg and less so to the formation of the glass from the melt. There is another parameter
suggested by Hruby and Stourac [52], the so called Hruby parameter, which is calculated
as
Tc − Tg
Kg = (5.12)
Tm − Tc
where Tm is the melting temperature. The only sample for which melting was observed
was TMN1, where a distinct endothermic peak was observed at ≈ 480◦ C (see Figure
5.7).
Tellurite glasses have been shown to obey the classical two-third rule which is an empir-
ical relationship between the transition temperature and the melting temperature, i.e.
Tg /Tm = 2/3 (in absolute temperature units) [54]. Using the two-third rule we esti-
mate the melting temperatures of the glasses to be in the range 570◦ C→610◦ C which is
outside of the range of the DSC that was used and therefore helps to explain why they
were not observed, except in the case of TMN1. The TMN1 glass only weekly obeys the
two-third rule, as the ratio of Tg to Tm is 0.75.
Some authors use the two-thirds rule to calculate the melting temperature from incom-
plete DSC data and then use this value to calculate the Hruby parameter. This is a
reasonable method provided the glass is known to obey the two-third rule, however,
based on TMN1 not showing close agreement with the rule we do not adopt this prac-
tise and instead simply use the crystallisation stability criterion expressed by Equation
5.11.
In addition to the temperature data obtained from the DSC, the crystallisation enthalpy
can also be used as a measure of crystallisation stability [55]. At every temperature,
the glass attempts to minimise its internal energy through the motion of its atomic con-
stituents. A low crystallisation enthalpy implies that the glass is, structurally speaking,
already in a reasonably stable state. Thus a low value for crystallisation enthalpy im-
plies a stable glassy state and therefore that particular glass will have a low tendency
to crystallise.
Based on this criterion for crystallisation stability we can see in Figure 5.9 that TZN3 has
the lowest enthalpy of crystallisation and therefore the highest crystallisation stability.
5.3.2 Measurement of the Thermal Expansion Coefficient
Of particular importance for the fabrication of core/clad optical fibres is good knowledge
of the thermal expansion coefficients for the core and cladding materials. To achieve the
refractive index contrast for optical guidance we must necessarily use different materials
for the core and the cladding. There will inevitably be a difference between the thermal
expansion coefficients for two materials which will result in axial mechanical stresses
being produced when the fibre is drawn [56]. The stress in the fibre will effect both
its optical and mechanical properties, often detrimentally. In particular, if the internal
stresses are comparable to the failure strength of the material, which is approximately
equal to the Young’s modulus of the material (50 to 70 GPa for glass) the manufacture
of a mechanically stable fibre is not possible.
It is possible to calculate the axial stress σ, via [57]:
(α1 − α2 )(T − Tg1 )E (α1∗ − α2 )(Tg2 − Tg1 )

σ= + . (5.13)
(R/r)2 (1 − μ) (1 − (R/r)2 )/3K1∗ − 1/3K2
The first term in Equation 5.13 applies to the situation where the core and cladding
are both solid. The subscripts 1 and 2 represent the core and cladding respectively
Chapter 5 89
and α is the coefficient of thermal expansion, T is the room temperature, Tg the glass
transition temperature, E is Young’s modulus, R the radius of the cladding, r the core
radius, μ is Poisson’s ratio. The second term provides the hydrostatic stress i.e. when
the core is liquid and the cladding solid. In this term a superscript of ∗ denotes values
of parameters just above the transition temperature and K is the bulk modulus.
It is clear that to minimise σ we should find materials with α1 ≈ α2 and Tg2 ≈ Tg1 . We
note that the TZN glasses satisfy this second condition (refer to Section 5.3.1.2). Further,
it has been noted in the literature that acceptable Δα = α1 − α2 is approximately
50 × 10−7 ◦ C−1 [58].
We have measured the thermal expansion coefficients for all of the glasses under study
so that, through careful choice of composition, we can minimise the stress in an optical
fibre produced from them.
The thermal expansion of a material is simply a change in length for a given change in
temperature. For most materials this relationship is linear in the region of interest and
can be expressed as
l2 = l1 (1 + αT ) (5.14)
where l2 is the length at temperature T ◦ C, l1 is the length at 0◦ C and α is the ther-

mal expansion coefficient. Thus to measure the thermal expansion coefficient we simply
measure the length of the sample over a range of temperatures and use the linear rela-
tionship to find α. Note that the amount by which the material increases in length is
in direct proportion to the value l1 . Accordingly, the changes become easier to measure
the larger the sample.
The key material property that determines the thermal expansion coefficient is bond
strength. That is, as we put energy in the form of heat into the material the oscillation
amplitude of the atoms increases. This has the effect of increasing the mean separation
of the atoms and therefore the size of the material. A bond with a high strength requires
more energy to produce an equivalent oscillation amplitude than does a weak bond.
It is therefore reasonable to assume that compositional trends in the thermal expansion

coefficient will be explainable in terms of the reported compositionally driven structural
changes (see Section 4.2.4).
In the following sections we describe a simple experimental technique for determining

the thermal expansion coefficients and present the data for the glasses under study.
Attempts to correlate the observed trends in the thermal expansion coefficients with the
structural information presented in Chapter 3 are made
5.3.2.1 Experimental Details
In general tellurite glass has a large thermal expansion coefficient, typically of the order
10−5 ◦ C−1 [18]. Compare this with silica which is 5.5×10−7 ◦ C−1 . It is therefore possible
to measure the expansion of a relatively small sample of telluite which, conveniently,
corresponds to the size of the spectroscopic samples described in Section 6.2.1. These
samples measure about 15 mm across their longest dimension, which is a distance that
can be determined mechanically using a micrometer.
The measurement apparatus consisted of the following (see Figure 5.10): A hot plate
with a controllable temperature range of ambient→300◦ C had mounted on it a specifi-
cally designed metal sample holder with a clamp both machined from aluminium. The
sample holder was designed to maximise the thermal contact between it and the glass,
leaving ≈ 0.2 mm of over hanging sample with which to contact the micrometer to in
order to make a measurement of the sample length3 .
Figure 5.10: Schematic of apparatus for thermal expansion coefficient measurement.

The glass sample is held in an aluminium clamp to ensure even heating. Measurement
of the sample length is made as shown.
5.3.2.2 Results and Discussion
To establish the accuracy of the measurements we used a sample of commercial glass

with a thermal expansion coefficient of similar size to that expected for tellurite. We
chose a chalcogenide glass produced by Schott Glass Co. known as IG5 with a reported
thermal expansion coefficient of α = 14 × 10−6 ◦ C−1 [59].
3
All measurements were performed by an undergraduate student, Matthew Mielczarek, who was
under my supervision
Chapter 5 91
We recorded the length of the sample for temperatures between ambient and ≈ 180◦ C.
Equation 5.14 can be rearranged to give
l2 − L 1
= αT (5.15)
l1
l2 −L1
The acquired data is plotted as l1 vs T in Figure 5.11 along with a linear fit. The
slope of the linear fit is equal to α as per Equation 5.15. Via this analysis we obtain
a value for α = 14 ± 0.12 × 10−6 ◦ C−1 , in good agreement with the reported value.
From this test we can conclude that this technique for measuring the thermal expansion
coefficient is accurate to about 1%.
0.0020
0.0015
(l2-l1)/l1
0.0010
0.0005
0.0000
25 50 75 100 125 150

o
Temperature [ C]
l2 −L1
Figure 5.11: Plot of l1 vs T for IG5. The slope of line of best fit equal to α.
Using the same technique, we measured the thermal expansion of the TMN, TZN and
l2 −L1
TBN. Figures 5.12, 5.13 and 5.14 show the data plotted as l1 vs T . From the slopes
of the lines of best fit we obtain the thermal expansion coefficients for each glass, the
results of which are tabulated in Table 5.3 and plotted in Figure 5.15.
It can be seen in Figure 5.15 that the addition of Zn to the TZN glasses causes the
value of the thermal expansion coefficient to decrease. Conversely, the addition of Mg
and Ba into the TMN and TBN glasses, produces an increase of the thermal expansion
coefficient.
0.0030
TMN1
TMN2
0.0025 TMN3
0.0020
(l2-l1)/l1
0.0015
0.0010
0.0005
0.0000
0 50 100 150 200 250

o
Temperature [ C]
l2 −L1
Figure 5.12: Plot of l1vs T for the TMN glass series. The slopes of the lines of
best fit are equal to α for each glass.
0.004
TZN1
TZN2
TZN3
0.003 TZN4
(l2-l1)/l1
0.002
0.001
0.000
0 50 100 150 200 250

o
Temperature [ C]
l2 −L1
Figure 5.13: Plot of l1 vs T for the TZN glass series. The slopes of the lines of
Chapter 5 93
0.0018
0.0016
TBN1
0.0014 TBN2
0.0012
0.0010
(l2-l1)/l1
0.0008
0.0006
0.0004
0.0002
0.0000
-0.0002
0 50 100 150 200 250
o
Temperature [ C]
l2 −L1
Figure 5.14: Plot of l1 vs T for the TBN glass series. The slopes of the lines of
α
Sample [10−6 ◦ C]
TMN1 7.57
TMN2 1.98
TMN3 1.88
TZN1 19.8
TZN2 19.8
TZN3 18.8
TZN4 17.4
TBN1 3.21
TBN2 8.79
IG5 14.1
Table 5.3: Measured values of the thermal expansion coefficient for TMN, TZN and
TBN glass series. Accuracy of measurements ≈ 1% determined from a reference stan-
dard chalcogenide glass (IG5 Schott).
To understand the trends observed in Figure 5.15, we consider the electronegativities of

the modifying cations, which are listed in Table 3.2. The relative electronegativities of
elements determines the strength of the chemical bonds between them. As such, when
discussing a substitution of one element for another in a glass matrix the electronega-
tivities of the elements being exchanged are directly related to the bond strengths. Of
the modifying species barium has the lowest electronegativity and thus the highest bond
20
15
αx10 [ C ]
-1
o
-6
10
0
5 10 15 20
Figure 5.15: Thermal expansion coefficients α, for TMN, TZN and TBN glass series.
strength. Magnesium has the next lowest electronegativity and then zinc. Accordingly
the values for the thermal expansion coefficient for the TBN glasses are the lowest.
Followed closely by the TMN series and then the TZN series.
Within the individual glass series the trends are explained in terms of the conversion
of TeO3 →TeO3+1 →TeO3 . With reference to Figure 3.4 we can see that during this
conversion the equatorial bond lengths are decreasing from 0.19 nm to 0.185 nm. Mean-
while one of the axial bonds has its length increased from 0.21 nm to 0.24 nm. Now,
the shortening of a bond is associated with an increase in bond strength and vice versa.
From the Raman spectroscopy data presented in Chapter 3, and in particular Figures
4.17 and 4.19 it can been seen that, for the TMN and TBN glass series, there is a rapid
conversion of TeO4 to TeO3 with increasing modifier content as indicated by the slopes
of the graphs. This conversion is associated with the intermediate TeO3+1 state which
possesses a elongated axial bond and further, because the elongation of this bond is sig-
nificant with respect to the shortening of the equatorial bond lengths (i.e. 10× greater),
the increase in thermal expansion coefficient can be understood in terms of the decrease
in bond energy associated with the creation of TeO3+1 subunits.
Conversely, the TZN series displays a very low tendency for the conversion of TeO4 to
TeO3 (Figure 4.17) and so the change in thermal expansion coefficient across the series
Chapter 5 95
is small. Further, we are substituting the Te4+ ion with an electronegativity of 2.10
for Zn2+ ion which has an electronegativity of 1.65 (Table 3.2). This substitution has
the effect of increasing the average bond strength and therefore decreasing the thermal
expansion coefficient. The effect of substitution appears to not dominate for the TMN
and TBN glasses.
In terms of suitable materials for an optical fibre, it is generally recognised that for a
core/clad fibre if the thermal expansion coefficient cannot be matched exactly then the
next best scenario is for the core to have a slightly higher thermal expansion coefficient
that the cladding to ensure that the core is under compression [60]. Additionally, the
Δα should be no greater than 50 × 10−7 ◦ C−1 [58].
If we consider the data in Figure 5.15 we can see that the TZN series for Zn content
greater than approximately 10 mol.% satisfies this first criterion. Furthermore, the total
variation of α over the entire composition range is ≈ 25 × 10−7 ◦ C−1 , thus satisfying the
second criterion. Later, in Chapter 7, we will revisit this data in terms of designing a
core clad fibre.
5.4 Conclusion
W
e measured the densities of the glasses, from which molar volumes were calculated.
The compositional dependence of the molar volumes are indicative of how the
modifying ions are incorporated into the glass network. Zinc and magnesium display
similar rates of change in molar volumes with modifier concentration (i.e. partial molar
volumes). This we attribute to the similarity of the coordination numbers for the Zn2+
and Mg2+ ions ( coordination number of 6, see Table 3.2). Barium, on the other hand
possesses a larger coordination number and therefore perturbs the glass network to a
higher degree. This is observed as a larger partial molar volume.
We have determined the thermal properties of the tellurite glasses including the transi-
tion and crystallisation temperatures and enthalpies (Table 5.2) as well as the coefficient
of thermal expansion (Table 5.3). Using the criterion for crystallisation stability we have
determined the most stable compositions within each family and overall. The most sta-
ble compositions are listed below:
TMN1: Within the TMN series TMN1 has the highest ΔT and very close to the
lowest ΔHc (Table 5.2). Thus making it the most stable TMN composition.
TZN3: The TZN series is in general a much more stable series than the TMN
series. The optimal composition is TZN3 which one of the highest ΔT values and
by far the lowest ΔHc (Table 5.2).
TBN1: For both TBN compositions the ΔHc varies very little. However, TBN1
has the highest ΔT and therefore represents the most stable glass in this series
(Table 5.2).
In addition to the thermal stability of the glasses studied we measured the thermal
expansion coefficients (Table 5.2). We find that the size of the thermal expansion coeffi-
cient is determined by composition as well as the microscopic structure of the glass. For
glasses with a strong tendency to convert into a matrix of TeO3 subunits, with the addi-
tion of modifier, the changes in the thermal expansion coefficient are largely determined
by the elongation of the axial Te-O bond typified by the intermediate TeO3+1 state.
Glasses that display a low tendency for this conversion, such as the TZN glasses, have
thermal expansion coefficients that appear to be determined by the electronegativities
of the components.
Of all the studied glass compositions TZN3 has the highest crystallisation stability.
Additionally, the TZN glass series displays the lowest variation of α. We therefore
conclude that glasses in the compositional vicinity of TZN3, that is glasses with zinc
concentrations around 15 ± 2 mol.%, are ideal for fabricating core/clad optical fibres.
This is further evidenced by the fact that other researchers have also chosen this glass
from which to make optical fibres [61, 62]. As such much of the subsequent work on
thermal poling (Chapter 7) and optical fibre design and fabrication (Chapter 9) will
focus on this composition range.
Chapter 6
Measurements of the Optical

Properties
6.1 Introduction
I
t is essential to measure the optical properties of the tellurite glasses so that an
optimal choice for a particular nonlinear optical device can be made. Furthermore,
it is also desirable to build an understanding of the physics determining these properties
so that, through control of the composition of the glasses, the optical properties can be
optimised.
The following sections describe the various characterisation techniques that were used
to analyse the glass samples. They include:
Ultra violet and visible light spectroscopy, Section 6.2.
Infrared spectroscopy, Section 6.3.
Refractive index measurement, Section 6.4.
Nonlinear refractive index and third order susceptibility, Section 6.5.
We begin each section by describing the information we intend to obtain from the mea-
surements and how that information fits into the context of this thesis. Next, the physics
required to understand and interpret the measurement are discussed. Once we are armed
with the requisite understanding the specific experimental details are outlined along with
the data analysis techniques employed. Finally, we present the data accompanied by a
detailed discussion of the results, including correlations between the measured properties
and the composition and structure of the glasses.
97
98 Measurements of the Optical Properties
6.2 UV-VIS Absorption Spectra
F
rom the measurement of a materials absorption spectra in the ultra violet and visible
regions of the spectrum we can determine several things about that material. Most
fundamentally, we determine the wavelengths over which the material is transparent.
This is an obvious necessity for determining the suitability of a material for an optical
device. At the short wavelength end of the transmission window of the glass is the UV
edge or band edge. This feature of the transmission spectrum originates from electronic
transitions in the material, which are determined by the constituents of the material and
their spatial arrangement. Thus certain correlations can be made between the structure
of the glass and the position of the band edge, in particular, how the position of the
edge depends on the proportions of the elements in the glass.
Glasses fall into the family of materials known as dielectrics. Typical to dielectric ma-
terials is the existence of forbidden electron energies which result in the well known
electronic band structure of these materials. The phenomenon of the band structure
of crystalline dielectrics is well understood in terms of resonant scattering of electrons
from the periodic potential presented by the regular arrangement of nuclei in the crys-
tal, so called Bragg scattering. The consequence of this is that photons with energies
which correspond to energies less than the band gap cannot be absorbed by electrons
in the material and thus the material is said to be transparent in that range of photon
wavelengths.
As for amorphous materials, such as glass, the regular ordering of the atoms is not
present and the band theory based on periodic potentials breaks down. Despite the
lack of ordering glasses do possess electronic bands. These are much more complicated
than those exhibited by crystalline systems. Indeed, it is common instead to explain the
optical absorption of glasses in terms of molecular orbital theory [41]. In this picture the
electronic structure of the structural subunits present in the glass determine the band
gap. The presences of modifying ions and polymerised chains of these subunits serve
to perturb the energies hence leading to a shift of the band edge [63]. For example; ab
initio calculations of the molecular orbitals of the subunits found in tellurite glass have
determined the energy gaps between highest occupied molecular orbital (HOMO) and
the lowest unoccupied molecular orbital (LUMO) to be 4.4 eV for the TeO4 subunit
and 5.5 eV for the TeO3 [42]. Indicating that pure tellurite glass should have an optical
absorption edge in the vicinity of 280 nm. In reality, pure tellurite glass has its absorption
edge at ≈ 330 nm and other multi-component tellurite glasses have absorption edges in
the range of to 350 - 600 nm because of distortions of the subunits and the influence of
modifying species and polymerised chains of subunits on the electronic structure of the
structural subunits [18, 63].
Chapter 6 99
We prepared the spectroscopic samples by cutting 10×10×2 mm thick slices from the
4 cm3 blocks. Then they were ground and polished to optical quality. Absorption spectra
were recorded with a Cary UV-Vis-NIR spectrophotometer1 between 200−3300 nm with
a 1 nm resolution.
The acquired data is in the form of absorbance A, which is defined in terms of the input
and output intensities, I0 and I, as:

I
A = − log . (6.1)
I0
We can combine this with the Beer-Lambert law for optical absorption over a distance
z in a material,
I = I0 e−αz (6.2)
to obtain the absorption coefficient α, by2

A = − log e−αz
1
= αz.
ln(10)
Therefore,
ln(10)
α(ω) = A(ω) (6.3)
z
where z is the sample thickness (in cm) and we have made the frequency dependence of
the optical density and optical absorption coefficient explicit. The absorption coefficient
consequently has units of cm−1 .
Losses due to Fresnel reflection and scattering from surface imperfections are removed
by subtracting the minimum absorption value from the entire spectrum. In doing so
we are assuming that at some wavelength there is zero loss. This assumption is not
strictly correct, however, it does provide a means of eliminating the effects of reflection
and surface scatter to allow relative comparisons between the samples. To perform this
analysis without this assumption would require the measurement of the refractive indices
of each glass at every wavelength within the spectral region of interest, or at least enough
1
Kindly provided by the Defence Science Technology Organisation (DSTO), Edinburgh, SA
2 logb N
Using the relation loga N =
logb a
wavelengths such that an interpolating function (e.g. Cauchy or Sellmeire equation [64])
could be fitted to the data . Both of these options are outside of our present capabilities.
In Chapter 9 measurement of the attenuation in an optical fibre made from TZN glass
with the composition 10Na2 O.12ZnO.73TeO2 is presented (Figure 9.27). The attenua-
tion measured at 1550 nm is 2.8±0.09 dB.m−1 or 6.4±0.2 × 10−3 cm−1 . Assuming that
the absorption of the glasses under study are roughly equal in their transparency regions,
the absolute value of the absorption is, in principle, in error of at least this magnitude.
All glass samples showed excellent transmission from 370 nm, below which there is
strong optical absorption due to electronic transitions, and 2800 nm whereupon ab-
sorption due to the presence of water begins to dominate.
We have analysed the band edge using the theory of Davis and Mott [65] to obtain
values for the optical energy gaps. Davis and Mott found empirically that Equation 6.4
provides an accurate representation for the shape of the band edge [65].
(ω − Eopt )r
α(ω) = A . (6.4)
ω
Here α(ω) is the frequency dependent optical absorption coefficient, A is a fitting con-
stant, is the reduced Plank constant, Eopt is the optical energy gap and r is an
empirically derived index which can take the values of 2, 3, 1/2 and 3/2 depending on
the nature of the electronic transition. It has been established [66–70] that for glass
and amorphous materials in general, r = 2 best represents the experimental data, this
further implies that the band edge is dominated by indirect transitions. We illustrate
this situation with the hypothetical band map, Figure 6.1, which shows how an indirect
transition Eind , can have a lower energy than a direct transition Edir . For the direct
transition, electrons in the valence band acquire energy from an incoming photon with
energy Edir , and move into the conduction band (blue arrow). The indirect transition
involves a valence electron acquiring energy from an incoming photon Eind (red arrow)
and momentum from a phonon (yellow arrow). The net result is that the valence elec-
trons are elevated into the conduction band and therefore radiation with energy equal to
Eind is absorbed. In this scheme Eind <Edir and therefore Eind marks the fundamental
absorption edge of the material .
Using r = 2 in Equation 6.4 we can obtain the following relation:
√
(αω)1/2 = A(ω − Eopt ). (6.5)
Chapter 6 101
Figure 6.1: Proposed energy band diagram for tellurite glass (Electron energy (E)
vs. wavevector (k)). For the direct transition, electrons in the valence band acquire
energy from an incoming photon with energy Edir , and move into the conduction band
(blue arrow). The indirect transition involves a valence electron acquiring energy from
an incoming photon Eind (red arrow) and momentum from a phonon (yellow arrow).
As implied by Equation 6.5, a plot of (αω)1/2 vs. ω will produce a straight line in the
vicinity of the band edge. Furthermore, the x−intercept of this line will be the value of
Eopt .
Figure 6.2 shows the absorption spectra for the TZN series plotted as (αω)1/2 vs. ω.
One can clearly see the linear regions in the high photon energy region of the figure. We
have fitted straight lines to the data for all glass series in this region and obtained the
values for Eopt which are tabulated in Table 6.1 and plotted against modifier content in
Figure 6.3. The data were analysed to determine if the indirect gap is in fact dominant.
By setting r = 1/2 and following an analogous procedure we determined that all of the
direct transitions occurred at photon energies approximately 0.15 eV higher than the
indirect transitions. Thus confirming that the indirect transition determine the band
edge for these glasses.
It can be seen in Figure 6.3 that the band edge moves towards shorter wavelengths (blue
shifted) for increasing Zn2+ and Mg2+ content and shifts towards longer wavelengths
(red shifted) for increasing Ba2+ content. The trends observed for the TZN and TBN
glasses are consistent with results for similar tellurite glasses reported in the literature
[71, 72].
Figure 6.2: (αω)−1/2 as a function of photon energy for the TZN series. The inter-
section of the linear fits with the horizontal axis provide the band gap energies, Eopt
(see Equation 6.5.
3.400
TMN3
3.375
TZN4
3.350 TMN2
TZN3
Eopt [eV]
3.325
TMN1
TBN1
3.300
TZN2
3.275
TBN2
TZN1
3.250
5 10 15 20
Figure 6.3: Compositional dependence of indirect band gap for TMN, TZN and TBN
glass series.
Chapter 6 103
Eopt λcutoff
Sample [eV] [nm]
TMN1 3.310 375
TMN2 3.335 372
TMN3 3.376 368
TZN1 3.267 380
TZN2 3.293 377
TZN3 3.332 373
TZN4 3.359 370
TBN1 3.299 376
TBN2 3.253 382
Table 6.1: Optical energy gaps calculated from UV-Vis spectra, maximum error of
±0.005 eV, and corresponding wavelength cut off for optical transmission ±1 nm.
It is generally noted in the literature that tellurite glasses experience a blue shift of
the band edge for decreasing amounts of TeO2 [18]. As exemplified by the TZN and
TMN glasses. One possible explanation for this is the conversion of TeO4 → TeO3 as
the modifier concentration is increased. As previously mentioned, the energy gaps for
the individual TeO4 and TeO3 subunits have been calculated to be 4.4 eV and 5.5 eV
respectively [42]. These energy gaps corresponds to wavelengths of 283 nm and 226 nm
and so as the relative proportions of TeO4 and TeO3 shift to favour the TeO3 species
the band edge should shift to shorter wavelengths.
In specific cases, such as with the TBN glasses, additional mechanisms take precedence.
We hypothesise that the disordering of the glass network structure induced by Ba2+ has
produced defect states in the band gap thus decreasing the energy required to bridge
this gap. We can see from Figure 4.11 that the TBN glasses posses the least network
connectivity and so there is tentative evidence that disordering of the network structure
could be producing the observed narrowing of the band gap.
Additional analysis of the band edge trends will be made in Section 6.4, where compar-
isons with compositional trends in the refractive index can be analysed.
6.3 Fourier Transform Infrared Absorption Spectroscopy
F
ourier Transform Infrared (FTIR) Absorption Spectroscopy is a complementary
spectroscopic technique to Raman spectroscopy. Both techniques probe the low
frequency vibrations of a solid, however, in contrast to Raman, the interaction of the
light with the material is elastic. Pure TeO2 crystalline Tellurite is known to have
characteristic IR absorption bands at 780 cm−1 , 714 cm−1 , 675 cm−1 and 635 cm−1
[73].
Furthermore, there are some reported IR spectra for various crystalline binary tellurite
oxides which include MgTe2 O5 , ZnTeO3 , Zn2 Te3 O8 and BaTeO3 . These crystalline
compounds have IR spectral features that fall in the range 620 cm−1 → 790 cm−1
[73]. Based on the crystalline vibrational features glasses comprised from these elements
will have resonances at similar frequencies, however, these will be inhomogeneously
broadened due to the amorphous structure.
The samples we used for this measurement are much too thick (1 mm) to resolve these
features. As such we are only able to determine the onset of the multiphonon edge
which results from harmonics and overtones of the fundamental IR absorption bands.
We do, however, make inferences as to the behaviour of the fundamental IR absorption
bands by noting the changes in the position of the multiphonon onset wavelength and
by comparison to the Raman spectra.
We recorded inferred absorption spectra using a Bruker Vertex 70 FTIR spectrometer3 ,

for wavelengths from 2 μm to 10 μm with a resolution of 1.55 nm. These spectra
were recorded using the same glass samples that were described in Section 6.2.1 in
transmission. We have concentrated on determining the location of the multiphonon
edge and the vibrational spectra of impurities, in particular hydroxide groups.
The raw data from the spectrometer is measured as absorbance K, which is related to
the transmission through the material via:

1
K = log . (6.6)
T
This is converted to absorption units of dB.m−1 via Equation 6.7, which is a convenient
unit for assessing materials for their potential use in optical fibres. Just as in Section
6.2.2, just as for the UV-VIS data we eliminate the effect of Fresnel reflection and surface
scattering by subtracting the minimum loss value from the entire spectrum.
ln(10)
αdB.m−1 = K (6.7)
L
3
Kindly provided by the Defence Science Technology Organisation (DSTO), Edinburgh, SA
Chapter 6 105
For the samples described here, we melt the tellurite glass in an open atmosphere furnace
and thus in the presence of atmospheric water. Accordingly there is a strong presence
of water in the glasses which can be easily seen in the infrared absorption spectra (see
Figure 6.4) by the presence of characteristic vibrations of the OH− groups at ≈ 3.5 μm
and ≈ 4.5 μm.
Figure 6.4: Infrared absorption spectrum of TMN1 (Red) and theoretical multi-
phonon edge (Blue).
Figure 6.4 also shows a sharp increase in absorption at the mulitphonon edge where
combinations and overtones of the fundamental lattice vibrations occur. We can estimate
the position of the multiphonon edge by fitting an exponential of the form:
B
L = Ae−( λ ) . (6.8)
Where L is the loss in dB.m−1 and A and B are fitting constants with units of dB.m−1
and μm respectively. Equation 6.8 predicts the location of the multiphonon edge for the
tellurite glasses without the contribution from water. Therefore this is a theoretical fea-
ture. We define the multiphonon onset wavelength as the point at which the absorption
exceeds 1 dB.m−1 . Listed in Table 6.2 are the multiphonon onset wavelengths for the
tellurite glass samples with the associated fitting parameters A and B.
Note that all of the glasses show similar multiphonon onset wavelengths. The multi-
phonon onset wavelengths are plotted for each glass sample in Figure 6.5, which shows a
general trend of increasing multiphonon edge wavelength for increasing modifier content.
This result is indicative of an increase in the mean bond length possibly related to the
Multiphonon onset A B
Sample wavelength [μm] [dB.m−1 ] [μm]
TMN1 4.27 1.5×1012 120
TMN2 4.29 2.7×1012 123
TMN3 4.39 1.4×1013 133
TZN1 4.56 9.0×1013 146
TZN2 4.54 6.7×1013 145
TZN3 4.57 1.4×1014 149
TZN4 4.66 7.6×1014 160
TBN1 4.52 7.9×1013 145
TBN2 4.76 1.1×1016 176
Table 6.2: Multiphonon onset wavelength for TMN, TZN and TBN glass series..
4.8
TBN2
4.7
TZN4
Multiphonon Onset [μm]
4.6 TZN3
TZN1
TZN2
4.5 TBN1
4.4
TMN3
4.3
TMN2
TMN1
4.2
5 10 15 20
Figure 6.5: Compositional dependence of multiphonon onset wavelength for TMN,

TZN and TBN glass series.
general loss of network connectivity displayed by the glasses as the amount of modifier
is increased.
Perhaps the most significant result from the FTIR data is the high absorption from
≈ 2 μm onwards, due to the presence of hydroxyl (OH− ) groups in the glass. In Figure
6.6 we have plotted the attenuation in dB.m−1 at the peak of the OH− absorption which
Chapter 6 107
is at ≈ 3.4 μm. The magnitude of the absorption is a direct measure of the concentration
of OH− in the glass and therefore Figure 6.6 can also be read as the dependence of the
OH− concentration on modifier concentration [74]. What we observe is that the amount
of OH− contamination decreases with increasing modifier content. One exception to that
conclusion is the data point for TZN3. It is suspected that this is an outlier, possibly
due to surface contamination of atmospheric H2 O, alternatively, contamination could
have occurred during fabrication of the glass. There is strong evidence to suggest that
TZN3 is indeed an outlier based on the consistent linear trend displayed by the TMN
series and the other TZN glasses.
2200
2000
TBN1
Absorption @ 3.4 μm [dB.m ]
-1
TBN2
TZN3
1800
TMN1
1600 TMN2
TZN1 TMN3
TZN2
1400
TZN4
1200
5 10 15 20
Figure 6.6: Compositional dependence of attenuation due to OH− contamination.
The observed reduction in OH− content can be explained by considering the way in which
the modifier species incorporate themselves into the glass structure. These charged
entities preferentially locate themselves in the regions of NBOs to fill in voids in the
structure. These are precisely the same locations that the hydroxyl groups would form.
As such, the increase of modifier concentration displaces the OH− from the glass.
Absorption due to hydroxyl groups is deleterious to the operation of optical fibre devices
that are intended to operate in the near to mid infrared region of the spectrum, i.e.
≈ 2 → 5 μm. For these glasses to be considered for use in the mid infrared it will
be necessary to reduce the water content (dehydroxylation) to a point where the loss
is below 1 dB.m−1 . The reduction achieved via increase in modifier concentration are
inadequate for this purpose. It is possible to reduce the water induce loss by fabricating
in a dry atmosphere of N2 . Alternatively, the addition of fluorinated compounds such as
NaF can reduce the OH− contamination via the reaction: OH− +F− →HF+O2− . These
dehydroxylation methods are left for future work.
6.4 Refractive Index Measurements
O
ne of the most fundamental properties of an optical material is the refractive index.
The refractive index determines the speed of light in that particular medium v,
as per the relation:
c
v= (6.9)
n
where c is the speed of light in vacuum and n is the refractive index. Besides simply
determining the speed of light in the material, the refractive index also dictates the
angle of refraction as light moves from one material to another via Snell’s law. This is of
consequence to optical fibres as it determines whether or not light can be guided within
a fibre. In the simplest picture of an optical fibre we have a cylindrically symmetric
interface between two material of differing refractive index as per Figure 6.7. The central
region, the core, has a slightly higher refractive index than the outer region or cladding.
The ray picture of light in combination with Snell’s law predicts that certain rays incident
onto the core/cladding interface will be totally internally reflected at every point along
the fibre and thus confined to the core.
Figure 6.7: Ray diagram for an optical fibre. Ray 1 is incident onto the core/cladding
interface at an angle less than the critical angle for TIR and it transmitted out of the
fibre. Ray 2 is incident onto the core/cladding interface at an angle greater than the
critical angle and is thus transmitted along the fibre.
For the design of an optical fibre it is essential to know the refractive indices of the
material or materials being used as these will determine the operation of the fibre.
Chapter 6 109
Further, typical index contrasts between core and cladding are in the order of Δn ≈ 0.01
therefore necessitating accurate measurements of the indices of refraction, at least as
accurate as 1/10th of the desired index contrast.
We measured the refractive indices using a prism coupler. This method is superior to
most alternative techniques such as the method of minimum deviation [75] as it requires
relatively little sample preparation. The measurements were made at two wavelengths,
1064 nm and 532 nm. These wavelengths were chosen to facilitate the Maker fringes
analysis in Chapter 8.
A prism coupler works by measuring the angle at which total internal reflection (TIR)
ceases at the interface between a prism of know refractive index and the sample under
study. From Figure 6.8 it is clear that for certain incidence angles θ at the prism/sample
interface, θ will be larger than the critical angle for total internal reflection θc . In this
case a signal will be recorded at the photo diode. However, as θ is decreased the point
at which θ < θc TIR no longer occurs. This will be recorded at the photo diode as a
sharp decrease recorded signal at θ = θc (see Figure 6.9).
Figure 6.8: Schematic diagram of the internal configuration of the Metricon prism
coupler. Laser light is directed via two mirrors, M1 and M2, onto the interface between
the sample and a reference prism. The polariser P, selects the polarisation state of
the light (this is only relevant for birefringent materials). The light reflected from
the interface is detected with a photodiode PD1. The sample, prism and detector are
mounted on a rotation stage. Mirror M2 is partially transmitting, allowing the reflected
light from the prism to pass through the aperture A, and onto the photodiode PD2.
This reflected beam is used to calibrate the angular position of the rotation stage.
We used a prism coupler (Metricon4 ) which has been retrofitted to include an additional
laser wavelength. The schematic representation of this equipment is shown in Figure
6.8. The measurements were made at 1046 nm and 532 nm using a rutile (TiO2 ) prism.
The absolute accuracy of the measurement was determined by first measuring the index
of a commercial glass SF57 (SCHOTT Glass Co.) It was found that at 532 nm the
measured index was in agreement with the value reported by SCHOTT to 0.08 % and
0.05 % at 1064 nm. The main source of this error is likely to stem from uncertainty
in the refractive index of the reference prism which is quoted to be ±1.5 × 10−1 [76].
Clearly, from the results at hand, this is a conservative estimate.
Shown in Figure 6.9 is a representative data set from the prism coupler in which the
position of the knee is indicated, and a line intersection method is used to find the
critical angle from which the refractive index is calculated. Using Snells law for a prism
refractive index np , and sample index ns and setting the transmitted angle to 90◦ we
obtain the following equation
ns = np sin θc . (6.10)
Software which accompanies the prism coupler performs this data analysis automatically
provided a distinct knee is recorded. The absence of a distinct knee results when the
contact between the prism and sample is not conformal. In these cases the sample was
removed and, using adhesive tape, dust was removed from the sample and prism surfaces
that make contact.
Table 6.3 contains the measured refractive index values and plotted in Figure 6.10 are
the refractive index values measured at 1064 nm for each composition. We note that,
for increasing modifier content, the TMN, TZN and TBN glasses display a decrease in
refractive index.
These compositional trends in refractive index can be explained by considering the rel-
ative contributions to the electronic polarisability by the individual components of the
glasses. The linear susceptibility can be related to the electronic polarisability of an
ensemble of n identical entities through:
n
χ(1) = α (6.11)
V
4
Prism Coupler was kindly provided by S. Madden and B. L. Davies at The Australian National
University (ANU)
Chapter 6 111
Figure 6.9: Example trace from a prism coupler.
Sample n532 n1064

TMN1 2.09568 2.02124
TMN2 2.06038 1.99058
TMN3 2.02314 1.95850
TZN1 2.10540 2.02636
TZN2 2.07768 2.00218
TZN3 2.05680 1.98418
TZN4 2.02854 1.96012
TBN1 2.09494 2.01668
TBN2 2.06268 1.98926
SF57 1.85996 1.81272
Table 6.3: Measured refractive indices at 532 nm and 1064 nm. Absolute measurement
error 0.08 % at 532 nm and 0.05 % at 1064 nm
where V is volume of material and n/V , the atomic number density can be expressed
as N = n/v. Thus by Equation 2.98 we can relate the electronic polarisability to the
linear refractive index by
n0 = (1 + 4πN α)1/2 . (6.12)
To relate the changes in polarisabilty and hence refractive index to the composition of
the glasses we consider the individual electronic polarisabilities of the modifiers (Mg2+ ,
Zn2+ and Ba2+ ) and the TeO4 and TeO3 subunits (Table 3.2). Sodium need not be
considered as it is in equal portions in all glass samples. The behaviour is similar for all
glass families (TMN, TZN and TBN) and so we can generalise the discussion by referring
to a general modifier M2+ (where M2+ = Mg2+ , Zn2+ or Ba2+ ). As we increase the
2.04
2.03 TZN1
2.02 TMN1
TBN1
2.01
TZN2
n (1064nm)
2.00
1.99 TMN2
TBN2 TZN3
1.98
1.97
TZN4
1.96
TMN3
1.95
5 10 15 20
Figure 6.10: Refractive index of tellurite glasses measured at 1064 nm.
amount of M2+ we are decreasing the amount of Te4+ as a result there are fewer TeO4
and TeO3 subunits present in the glass. We are therefore substituting components with
electronic polarisabilities of ≈ 6 × 10−23 cm3 (approximate electronic polarisibility of
TeO2 ) with components with electronic polarisabilities of 2.5 × 10−23 cm3 (Ba2+ has
an electronic polarisability of 2.5×10−23 cm3 which is the highest of the three modifiers)
and so we should expect a decrease in polarisability. Any decrease in polarisability will
be accompanied by a corresponding decrease in refractive index as predicted by by
Equation 6.12.
Furthermore, the Raman spectra in Section 4.2.4 indicate that as we increase the amount
of M2+ in the glass there is a tendency for the TeO4 subunits to convert into TeO3
subunits. As the electronic polarisability of the TeO4 subunit is ≈ 20% greater than the
TeO3 subunit [42] the progression to lower refractive index with increasing M2+ is also
mediated by the conversion of TeO4 to TeO3 .
The progression to lower polarisability with increasing modifier content, as indicated by

the trends in refractive index, leads us to an interesting apparent contradiction. With
reference to the data for the band edges of the glasses in Section 6.2.2 the band edge
is blue shifted for the TMN and TZN glass series as we increase the modifier content.
Typically, a decrease in polarisability will be associated with a blue shift in the band
Chapter 6 113
edge. This behaviour is exhibited by the TMN and TZN glasses. However, the TBN glass
displays a red shifted band edge with increasing Ba2+ content despite the poarisability
decreasing which we have hypothesised to be resultant from progressive and extreme
disordering of the TeO2 network by Ba2+ (Section 6.2.2). This disordering leads to
longer band tailing and, subsequently, a red shifted band edge.
6.5 Nonlinear Refractive Index Measurements
T
here are several experimental techniques for measuring χ(3) including degenerate
four wave mixing [77], electric field induced second harmonic generation [78] and
third harmonic maker fringes [79] to name a few of the more common. The technique we
used is the Z-Scan technique which used the phenomenon of Kerr focusing to determines
the value of the intensity dependent refractive index n2 as given in Equation 2.88.
In the Chapter 2.4.1, for simplicity, we did not consider the imaginary parts of the
nonlinear susceptibilities. The imaginary part of χ(3) mediates an effect known as two
photon absorption and can be thought of as a nonlinear loss mechanism. For two photon
absorption to occur the energy of the conduction band of the material must be less than
twice the energy of the photons. This can be negated by choosing a probe wavelength
for which the second harmonic is below the UV cut off [80]. The following sections
describe the theory and experimental details, followed by the results for the studied
glass samples.
As shown in Section 2.4, there is a component of the refractive index that scales with
the intensity of the transmitted light. If we have light with a plane phase front and
a Gaussian transverse intensity distribution incident onto a nonlinear medium then by
Equation 2.88 the modified refractive index of the material in the vicinity of the beam
can be written as
n2 −2r2
n = n0 + I0 exp . (6.13)
2 w2
In Equation 6.13 I0 is the on axis intensity, r is the transverse spatial coordinate and w
is the radius of the incident beam. In the pariaxial approximation we can expand the
exponential to lowest order and obtain:

n2 I 0 r 2
n ≈ n0 1− . (6.14)
n0 w 2
It is straight forward to show that when a material has a parabolic index profile as
described by Equation 6.14 it will cause plane waves to be focused. Provided the material
with thickness L is thin and the paraxial approximation holds the focal length of the
induced lens is given by

n0 w 2
f= . (6.15)
2n2 I0 L
Now if the input wave does not have a flat phase front, then the expressions become
more complex. However, we can gain a qualitative understanding of the Z scan by
keeping these approximations. Consider then experimental configuration as shown in
Figure 6.11.
Figure 6.11: A Simple Z scan configuration. Top: Sample positioned before the un-
perturbed focal point causes less light to reach the detector. Bottom: Sample positioned
after the unperturbed focal point causes more light to reach the detector.
A collimated beam with a Gaussian intensity profile is focused to a point some distance
form an aperture. We then translate a thin sample along the propagation direction.
Assume that the beam has sufficient intensity to generate a change in the refractive
index of the sample of the form in Equation 6.14. The diagram at the top of Figure 6.11
shows the additional focusing from the sample causing the actual focal point to move
away from the aperture. The spreading out of the beam results in less light reaching
the detector. At the bottom of Figure 6.11 the sample is now closer to the aperture and
therefore the focusing of the beam results in more light passing through the aperture
and this being detected. An example of a Z scan signal is shown in Figure 6.12
This simplified version of the Z scan illustrates that as the sample is translated along
the beam and through the waist the signal detected after the aperture first decreases
then increases. The analysis for a curved wave front, which is in fact the case due to the
lens, is more complicated and is expressed, not in terms of a focal length but, instead,
in terms of the on axis phase change caused by the material. The analysis performed
Chapter 6 115
NOTE:
Figure 6.12: Examples of Z scans. Red: Self focusing produced by a positive phase
shift for a material with n2 > 0. Black: Self defocusing produced by a negative phase
shift for a material with n2 < 0. Reproduced from [81]
by Shiek-Bahae et al [82] derives a useful approximation for the on axis phase shift,
provided that the material is thin (i.e. L << z0 ) and the beam has a Gaussian intensity
profile. Under these conditions the on axis phase shift is given by
ΔTpv
|ΔΦ0 | = (6.16)
0.406(1 − S)0.27
where ΔTpv is the difference between the peak and the valley of the normalised trans-
mittance and S is the fraction of the beam transmitted through the aperture (before
the self focusing). The nonlinear refractive index is related to ΔΦ0 via

λ ΔΦ0
n2 = . (6.17)
2π I0 Leff
Combining Equations 6.16 and 6.17 we obtain

λ ΔTpv 1
n2 = (6.18)
2π I0 Leff 0.406(1 − S)0.27
The experimental Z-scan setup for measuring the nonlinear refractive indices is illus-
trated in Figure 6.13.

"#
$ "
# $
#
"
$

%" &

%" $

!
Figure 6.13: Schematic of Z-scan experimental setup. Optical paths red dotted lines,
electrical paths grey lines. M1-3; Mirrors, F; IR filter, LP; Linear Polariser, A1-3;
Apertures, ND1-2; Neutral density filter wheels, BS1-2; Beam splitters, L1-2; Lenses,
LS; Linear translation stage, PD1-3; Photo diodes
The mode locked output from a Clark-MXR CPA-2001 Ti-Sapphire laser is injected
into a Quantronix TOPAS OPA. Via nonlinear processes the 788 μJ 775 nm pulses are
converted into 120 μJ 1500 nm pulses. As per the UV-VIS spectroscopy results in
Section 6.2 the band edge is at ≈ 3.3 eV for all glasses under study. For this reason
we chose a probe wavelength of 1500 nm, for which the two photon energy is ≈ 1.7 eV.
Therefore making two photon absorption negligible.
The duration of the frequency converted pulses is approximately 150 fs (full width half
maximum), thus providing a peak power of ≈752 MW. Several wavelengths are emitted
from the TOPAS; there is the 1500 nm signal accompanied by the 1603.4 nm idler.
Additionally there are doubled signal and idler at around 750 nm and 800 nm respec-
tively and there is also some sum frequency radiation resulting from the addition of the
pump and idler which is at 525 nm. All of the spurious wavelengths are filtered by an
appropriate filter (F).
The intensity incident on the sample is controlled by the ND1 and ND2 filter wheels
by between 3 % and 0.3 % of maximum. The beam is passed through an aperture
Chapter 6 117
(A1) to produce diffraction such that at aperture A2 there is an airy disc. By carefully
adjusting the size of A2 we are able to select the central peak of the airy disc and
generate a near Gaussian beam with an approximate diameter of 2 mm. Figure 6.14
illustrates the validity of the Gaussian approximation for a truncated Airy disc. We
define the percentage overlap between the two functions as:
⎛ a ⎞
∞

⎜ B(x) dx −
2
G(x)dx ⎟
⎜ ⎟
⎜ ⎟
⎜ −a −∞ ⎟
%Overlap = ⎜1 − ⎟ × 100 (6.19)
⎜ ∞ ⎟
⎜ ⎟
⎝ G(x)dx ⎠
−∞
The functions B(x) and G(x) are Bessel and Gaussian functions, respectively, which
provide a simple 2D method for evaluating the overlap. The integration limit a, is the
position of the first root of the Bessel function which corresponds to the position of the
aperture. Under these conditions the degree of overlap is ≈ 96% which therefore shows
that the Gaussian approximation is valid.
Figure 6.14: Illustration of overlap between a truncated airy disc (red) and a Gaussian
(green).
A small fraction of the beam is directed into a photo diode (PD1) by a glass microscope
slide (BS1). The signal from PD1 is used for normalising the intensity fluctuations in the
laser beam. Next the beam is focused by lens L1 which has a focal length of 17 cm. This
produces a beam diameter at the focal plane of 20 μm and a Reighlegh range of ≈7 mm
and so the 2-3 mm thick samples we are using meet the criterion for being thin i.e.
L < z0 . LS, the computer controlled linear translation stage has a full range of 90 mm
and is used to move the sample along the beam. At BS 2, which is also a microscope
slide, the signal is split into the open aperture channel recorded at PD 3 and the closed
aperture channel recorded at PD 2. The aperture in the closed aperture channel, A3,
has its diameter set to ≈3 mm. This diameter is adjusted iteratively to give the best
Z-scan signal by viewing the closed aperture signal on an oscilloscope as the sample is
manually translated through the beam focus. We have an optimised aperture diameter
when the signal on the oscilloscope is symmetric about the focal point.
6.5.2 Data Analysis
The raw data from the Z scan analysis is analysed in the following way:
1. For each glass sample we have Z scan traces for a variety of laser powers. The
signal from PD1 is averaged over the entire scan. This value is proportional to
the mean intensity I0 , at the sample over the duration of the experiment which is
used later to calculate the relative nonlinear refractive index (see Equations 6.18
and 6.21).
2. The data from the closed aperture channel is divided by the open aperture channel.
This serves two functions. Firstly, it normalises the power fluctuations in the
signal. Secondly, if there is any two photon absorption the power loss in the closed
aperture channel that results from this loss is recovered.
Figure 6.15: Representative Z scan data set: Closed aperture signal, red, open aper-
ture signal, blue and ΔTpv indicated. This particular Z scan is for TZN3.
3. We determine ΔTpv and the mean value of the open aperture channel at each laser
power and plot this against the averaged PD1 signal. Figure 6.16 shows the data
Chapter 6 119
for the TZN3 glass. As expected, the signals are proportional to the intensity. The
slope of the ΔTpv vs intensity is obtained for each glass sample and is equal to
ΔTpv /I0 .
0.40 3.2
Mean Open Channel Sgnal [V]

0.36 2.8
0.32 2.4
ΔTpv
0.28 2.0
0.24 1.6
1.4 1.5 1.6 1.7 1.8 1.9
Mean Laser Power [arb. units]
Figure 6.16: Z Scan measurements at various incident intensities. Red: Peak to

trough value of the normalised transmission at various incident intensities. Blue: Mean
open aperture signal at various incident intensities.
4. We also perform the Z scan on a reference sample with a well known nonlinear
refractive index. A commercial lead silicate glass (SF57, SCHOTT) was used.
Silica is a commonly used reference material [77], however, it is desirable to use
a reference that possesses a nonlinear refractive index more closely matched to
the sample under study as this eliminates the problem of detector dynamic range.
Silica, for example, requires approximately 10× the incident intensity to achieve
a phase shift comparable to tellurite. SF57, on the other hand, has a nonlinear
refractive index within a factor of 2 of tellurite. Thus we can use the dynamic
range of the detector more efficiently without the need for beam attenuation as
would be required is silica was used as the reference.
5. If we apply Equation 6.18 to the reference sample:

r
λ ΔTpv 1
nr2 = . (6.20)
2π I0 Lef f 0.406(1 − S)0.27
r r
and the sample under study (Equation 6.18) and take the ratio,(assuming that S
is held the same, i.e. the aperture is not adjusted in between measurements.), we
obtain an expression for the relative nonlinear refractive index:
r r
1 ΔTpv I0 Lef f
n2 = . (6.21)
nr2 ΔTpv
r I L
0 ef f
where a superscript of r indicates a reference quantity. The ratios ΔTpv /I0 and
r /I r have been calculated from the Z scan data (as described above), the
ΔTpv 0
sample lengths (Lef f and Lref f are known as is the reference nonlinear refractive
index nr2 . We can therefore calculate the value of the nonlinear refractive index of
the samples under study, these are presented in Section 6.5.3.
The results of the Z-scan measurements are displayed in Table 6.4 where we also quote
the reported nonlinear refractive index for SF57. Further, we have used Equation 2.101
to estimate the values of χ(3) for each sample, also listed in Table 6.4. The validity
of this estimation is based on the assumption that the value of the refractive index
at 1064 nm is approximately equal to that at 1500 nm, i.e. the wavelength at which
the n2 measurements were made. The refractive index at these two wavelenght was
kindly measured for the TZN3 sample by Jeremy VanDerslice at J.A. Woolam on an
spectroscopic ellipsometer. The difference in the values at 1064 nm and 1500 nm is
determined to be approximately 6% giving an error of approximately 12% for χ(3) , as
χ(3) ∝ n20 (Equation 2.101).
n2 χ(3)
Sample [10−19 m2 .W−1 ] [10−21 m2 .V−2 ]
TMN1 5.89 8.52
TMN2 4.49 6.30
TMN3 3.49 4.74
TZN1 5.76 8.37
TZN2 5.55 7.88
TZN3 4.86 6.77
TZN4 4.82 6.56
TBN1 5.41 7.79
TBN2 5.08 7.12
SF57 4.10 5.02
Table 6.4: Measured nonlinear refractive indices and calculated third order suscepti-
bilities (Equation 2.101) for TMN, TZN and TBN glasses at 1500 nm. Reference value
of SF57 is quoted [83]
Chapter 6 121
The values of the nonlinear indices are plotted in Figure 6.17 vs the concentration of
the various modifier.
6.0
TMN1
TZN1
5.5 TZN2
TBN1
n2 [10 m .W ]
5.0 TBN2
-1
TZN3 TZN4
2
-19
4.5 TMN2
4.0
3.5 TMN3
5 10 15 20
Figure 6.17: Nonlinear refractive indices of Tellurite glasses.
At the incident power levels used we observed no two photon absorption, this is largely
because of our choice of probe wavelength and the lack of material absorption at the
two photon energy. Figure 6.15 shows the Z scan data for TZN3 with the highest
input power, ≈ 25 MW peak. The open aperture signal (blue) is essentially constant
with position. Two photon absorption would appear as a sharp downwards peak in the
centre of the data no such features were observed in any of the data.
Observe that for all glass series the nonlinear refractive index is decreasing with increas-
ing modifier. An understanding of why this occurs is necessary if we wish to discover
alternative glass compositions with larger nonlinear susceptibilities. Larger nonlinear
susceptibilities will enable more efficient nonlinear devices and are therefore of techno-
logical interest. To a first approximation, we can understand the progressive decrease in
terms of the substitution of the highly polarisable tellurium ions for the less polarisable
modifying ions (see Table 3.2). If this were the only mechanism we would expect the
slopes in Figure 6.17 to have be ranked in the order TMN<TZN<TBN. This is because
the polarisabilities of the ions Mn2+ , Zn2+ and Ba2+ follow that ordering. What is ob-
served is that the slope of the TMN glasses is the lowest but the TBN and TZN glasses
have approximately equal slopes. As the polarisability of the Ba2+ ion is greater than
2× that of the Zn2+ ion this requires further analysis.
We therefore draw attention to the Raman spectroscopy data in Chapter 3. In particular

the trend in the Raman D bands (as shown in Figure 4.17) which can be used as a
measure of the concentration of the TeO3 subunits. Notice that the increase in the
TeO3 concentration with increasing modifier content is more pronounced for the TBN
series than for the TZN series. It is known from quantum chemical ab initio models that
the polarisability of the TeO3 subunit is ≈ 80% of the TeO4 subunit [42]. Therefore, as
the amount of TeO3 increases the nonlinear refractive index is expected to decrease. As
this trend is much more pronounced for TBN glasses than for TZN glasses the tendency
towards lower nonlinear refractive index, based on TeO3 content alone is less for the
TZN glass series.
With these two mechanisms in action we can summarise the trends in nonlinear refractive
index as follows:
TMN: The Mg2+ ion has a polarisability that is approximately 2 orders of mag-
nitude lower than tellurium (Table 3.2). This substitution thus decreases the net
polarisability. Additionally, there is a pronounced conversion of TeO4 →TeO3 as
the concentration of Mg2+ is increased (Figure 4.17). Together these two trends
contribute to produce the steepest slope in the n2 vs. modifier content graph in
Figure 6.17.
TZN: Zn2+ ions have a polarisability that is ≈ 10× less than tellurium (Table
3.2) therefore when tellurium us substituted for zinc there is a decrease in the
polarisability. However, the progression of TeO4 →TeO3 is relatively slow (Figure
4.17) and so the dominant effect for the lowering of the nonlinear refractive index
is simply related to the substitution of lower polarisabilty ions.
TBN: The polarisability of the Ba2+ ion is of the same order as tellurium, however,
slightly lower (Table 3.2). Consequently, as we substitute tellurium for barium the
change in polarisabilty is small. In particular with respect to the changes observed
in the TMN and TZN glasses. On the other hand, the increase in the concentration
of TeO3 subunits as Ba2+ is added is large (Figure 4.17). Therefore the decrease
in nonlinear refractive index observed in the TBN glass series is dominated by the
conversion of TeO4 →TeO3 .
Chapter 6 123
In 1964 Miller analysed the relationship between the linear and nonlinear susceptibility
for a large range of crystals [84]. An empirical relationship was observed, later named
Miller’s rule, that showed that the ratio
χ(2) (ω1 + ω2 , ω1 , ω2 )
(6.22)
χ(1) (ω1 + ω2 )χ(1) (ω1 )χ(1) (ω2 )
was essentially constant for all materials with non-zero second order nonlinear suscepti-
bilities. It has since been shown that this ratio is related through
χ(2) (ω1 + ω2 , ω1 , ω2 ) ma20

= (6.23)
χ(1) (ω1 + ω2 )χ(1) (ω1 )χ(1) (ω2 ) N 2 e3
where m is the mass of the electron, e is the electrons electronic charge, N is the atomic
number density and a is a constant that is related to the strength of the nonlinearity
[30]. N is of the order 1022 cm−3 for nearly all solid materials. Furthermore, a can be
shown to be roughly constant for all materials [30]. It is therefore clear that the ratio
in Equation 6.22 should be approximately constant for most materials.
This relationship can be further generalised to show that there exists a similar scaling
law for the higher order susceptibilities as was done in the work of Wang [85]. Wang
derived an empirical relationship for the first and third order susceptibilities given by:
g (1) 2
χ(3) = χ (6.24)
N f ω0
where f is the oscillator strength, ω0 is an average transition frequency and g is a

dimensionless parameter ≈ 1 that is assumed to be constant for all materials [30]. This
generalisation of Miller’s law has been shown to be accurate for many materials including
glass.
With reference to the data for nonlinear refractive index and the measurements of linear
refractive index in Section 6.4 we notice when the refractive index decreases the nonlinear
refractive index also decreases. This trend is consistent with the scaling law of Wang,
however, despite the fact that Wang developed the relation between χ(3) and χ(1) (and
therefore n2 and n) this scaling law is referred to as Miller’s law in the literature.
Figure 6.18 (adapted from [6]) is a plot of Millers law with various glass families indicated
(circles) along with the glasses under study in this thesis (squares). As can be seen the
results correlate well with Millers prediction.
We note that the nonlinear refractive index of silica, the current material of choice for
electro-optic fibres is 2.7 × 10−20 m2 .W−1 . This is some 20× smaller that the values
NOTE:
Figure 6.18: Millers law plot with various representative glass types indicated along
with TMN, TZN and TBN glasses. Adapted from [6]
measured for the studied tellurite glasses. Indicating the superiority of tellurite as a
material for electro-optical fibres.
6.6 Conclusion
I
n conclusion, the tellurite glasses under study have optical transmission from ≈370 nm
to ≈4500 μm. While the edges of this transmission window are largely determined
by the properties of TeO2 , there are some small changes in the positions that are subtly
effected by the modifiers. The infrared transmission of the glasses is dominated by
absorption due to the presence of water (OH− ) in the glass in the spectral region ≈
2 → 5 μm. For progress in the area of mid IR glasses this will have to be reduced or
eliminated via dehydroxylation processes.
Refractive indices display a clear linear decrease for increasing modifier which results
from an overall decrease in the net polarisability as the proportion of the most polarisable
component, TeO2 , is decreased. The addition of modifier forces the conversion from a
more polarisable subunit (TeO4 ) into a less polarisable subunit (TeO3 ). These two
processes in combination produce the measured decrease in index. Furthermore, the
nonlinear refractive index displays close correlation to the compositional trends observed
for the refractive index. We claim that these trends are also related to the substitution
Chapter 6 125
of lower polarisability modifiers in place of the highly polarisable tellurite sub groups, as
well as the progressive conversion of the more polarisable TeO4 into the less polarisable
TeO3 subunits.
In terms of identifying suitable glasses from which to fabricate electro-optic fibres devices
we note the following:
All of the studied glasses possess approximately the same transmission windows
from 350 nm to 4.5 μm. However, the presence of hydroxyl ions limits the usable
portion of this this transmission window up to approximately 2 μm after which
the absorption is prohibitively high (i.e. > 10 dB.m−1 ).
We observe near linear variation in the refractive index with modifier concentration
for each of the glass series. This offers the possibility of tuning the index with subtle
alterations in composition so as to achieve the necessary refractive index contrasts
for step index fibres.
The measured nonlinear refractive indices and calculated third order susceptibili-
ties are in excess of one order of magnitude larger than those of silica. As such, all
of the studied glasses offer increased electro-optic efficiency as compared to silica.
In Chapter 5 the TZN glasses with zinc concentrations in the vicinity of 15 ± 2 mol.%
were identified as being the most stable. Taking this into consideration we note that
the refractive index variation in the TZN system over this range of compositions is
Δn ≈ 0.02. This is sufficient for producing a large range of refractive index contrasts
necessary for step index optical fibres.
Chapter 7
Thermal Poling
7.1 Introduction
T
here are several processing techniques for glass that can generate a non-zero χ(2)
by producing an anisotropy in the electrostatic field. Generally speaking they all
involve the displacement of charged species to create a DC electric field inside of the
glass. In this thesis we have studied thermal poling as a possible method for creating
nonzero χ(2) glasses.
The work presented in this Chapter documents efforts to create permanent second order
nonlinearities in the glasses under study. The results obtained elucidate some subtleties
relating to the thermal poling of tellurite glass. This information will be invaluable for
future research in this area.
7.2 Background
W
e have identified that there are two, essentially different, glass thermal poling
techniques that are often not sufficiently differentiated from one another in the
literature. Both involve the forced migration of charged species under the application of
heat and high electric fields and both have been applied to the thermal poling of tellurite
glass.
We shall distinguish the two thermal poling techniques in the following way:
Charge migration thermal poling

This, the most common style of thermal poling, relies on the temperature activated
mobility of charged species within the glass being poled. The mobile species are
usually sodium ions due to their presence as impurities in many raw materials and
the high mobility of such small ions. The poling configuration for this technique
127
128 Thermal Poling
employs electrically conductive electrodes, usually stainless steel or doped silicon

wafers, which are put in physical contact with the glass to be poled (Figure 7.1
(A))
Figure 7.1: Schematic of the key steps in the charge migration thermal poling process.
(A) The glass to be poled in sandwiched between two electrodes. Initially there is
a homogeneous distribution of positive and negative charge. (B) Heat and a strong
electric field is applied. This causes the mobile positive charges to migrate away from
the positive electrode, producing a region depleted of positive charge. (C) The glass
cooled and the field removed. Positive charges from the air are attracted to the highly
negative depletion region. (D) The negative charges at the surface have been neutralised
by the positive charges from the air. Leaving a thin buried layer of negative charge in
the glass.
Under the application of heat and a strong electric field, cations migrate away
from the anode producing a region depleted of positive charge (Figure 7.1 (B)).
After the sample is cooled and the field is removed, there is a high negative surface
charge on the anodic side of the glass. Positive ions from the air, namely H+ and
H3 O+ , move in to neutralise the charge (Figure 7.1 (C)), however, as they have
limited mobility in the glass at room temperature they are only able to neutralise
a thin layer at the surface. This leaves behind a thin region of negative charge
buried under the anodic surface of the glass (Figure 7.1 (D)).
Silica is poled in this way as are many other glasses such as borosilicates [86],
borophosphates [87, 88] lead borates [89] and chalcogenides [90]. Tellurite glass
has also been poled in this manner, however, a borosilicate glass microscope cover
Chapter 7 129
slip is often used at the cathode as a blocking electrode to prevent electrons from
flowing into the glass and reducing the tellurite [91].
Charge injection thermal poling

This technique involves supplying the glass to be poled with a source of mobile
ions. A typical configuration for this type of poling consists of sandwiching the
glass between the electrodes with a commercial borosilicate microscope cover slip
on the anodic side of the glass (Figure 7.2 (A)).
Figure 7.2: Schematic of the key steps in the charge injection thermal poling process.
(A) The glass to be poled in sandwiched between two electrodes with an additional
glass with a high cation content at the anodic side. Initially there is a homogeneous
distribution of positive and negative charge in both glasses. (B) Heat and a strong
electric field is applied. This causes the mobile positive charges in the anodic glass
to migrate away from the positive electrode and jump the gap into the glass sample,
producing a region of excess positive charge. (C) The glass cooled and the field removed.
Negative charges from the air are attracted to the highly positive region. (D) The
positive charges at the surface have been neutralised by the negative charges from the
air. Leaving a thin buried layer of positive charge in the glass.
The cover slips are a rich source of sodium ions which are mobile above ≈ 150 C◦ .
Provided there is conformal contact between the glass to be poled and the cover
slips, which can be ensured by polishing the glass surface and providing some addi-
tional pressure from the electrodes, the ions can jump from one glass to the other.
At the anodic side of the glass sodium ions move from the cover slip and impreg-
nate them selves into the glass forming a positively charged layer ≈ 10 μm thick
(Figure 7.2 (B)). After the sample has been cooled and the field removed, negative
130 Thermal Poling
ions, most likely O− and OH− from the environment will neutralise the high sur-
face charge (Figure 7.2 (C)). However, because the mobility of these negative ions
is small, due to the temperature being low, they cannot completely neutralise the
entire charged layer. Thus, a thin buried positive layer exists in the glass (Figure
7.2 (D)). It is the interaction between the DC electric field and the third order
nonlinear coefficient that then produces an effective second order nonlinearity (see
Section 2.3.3).
Table 7.1 summarises reported tellurite thermal poling results for a wide variety of
tellurite compositions. Where possible the significant poling conditions have been noted.
In particular the poling temperature Tp , poling voltage Vp , poling time tp and the
measured χ(2) a presented. Importantly, we indicate the method that was used to pole
the glasses, either charge migration CM, or charge injection CI. Further, if the poling
was performed under vacuum, we indicate this with a -V in the method column. We
note that this list is not exhaustive, but does provide a picture for the range of glass
compositions and conditions that have been successfully used to produce second order
nonlinearities in the tellurite glass system.
Also in Table 7.1 are some reported poling conditions and measured χ(2) values for
Infrasil 301 (Heraeus Holding GmbH) [92]. This commercial glass has been studied
extensively for thermal poling and was used by us to test and calibrate our apparatus,
measurements and data analysis techniques.
7.3 Experimental Details
7.3.1 Experimental Plan
Prior to beginning the experimental poling work an experimental plan was developed.
This plan, described below, was intended to provide us with samples poled under a range
of conditions such that optimal conditions could be determined.
1. First, a sample of commercial silica glass is poled under conditions that will pro-
duce a small second order nonlinear coefficient. We used Infrasil 301 (Schott Glass
Co.), which is a well studied material in the area of thermal poling and thus pro-
vides predictable outcomes.
This step is preformed to provide us with a means of: 1) Testing the performance
of the poling apparatus. 2) Determining the limitations of the characterisation
experiment as described in Chapter 8.
Chapter 7 131
Tp Vp tp χ(2)
Glass [◦ C] [kV] [min] [pm.V−1 ] Method Ref.
Li2 O.Nb2 O5 .TeO2 250 4.0 30 * CM [91]
30ZnO.70TeO2 250 4.0 30 0.44 CI [93]
10MgO.20ZnO.70TeO2 260-300 4.0 20 0.24 CI [93]
15MgO.15ZnO.70TeO2 260-300 4.0 20 0.26 CI [93]
20B2 O3 .80TeO2 260-300 4.0 20 0.22 CI [93]
25B2 O3 .75TeO2 260-300 4.0 20 0.02 CI [93]
30NaO1/2 .70TeO2 225 3.0 20 0.16 CI [94]
10NaO1/2 .20ZnO.70TeO2 260 3.0 20 0.46 CI [94]
30ZnO.70TeO2 280 3.0 20 0.9 CI [94]
15Nb2 O5 .85TeO2 * 3.0 15 * CI [95]
15Nb2 O5 .85TeO2 + Na+ * 3.0 15 * CI [95]
5Sb2 O3 .25Pb(PO3 )2 .70TeO2 200 4.0 60 0.4 CM-V [96]
” 225 4.0 60 0.58 CM-V [96]
” 250 4.0 60 0.33 CM-V [96]
70TeO2 .15Bi2 O3 .15ZnO 280 5.0 60 1.2 CI-V [97]
” 300 5.0 60 0.26 CM-V [97]
” 300 3.0 60 0.22 CM-V [97]
90TeO2 .5Bi2 O3 .5ZnO 300 3.0 60 0.36 CM-V [97]
Infrasil 301 290 4.0 90 0.20 CM [98]
” 300 4.0 60 0.01 CM [87]
” 325 4.0 60 0.08 CM [87]
” 350 4.0 60 0.06 CM [87]
Table 7.1: A summary of some reported thermal poling experiments performed on

various tellurite glass systems. When possible the poling conditions such as poling
temperature Tp , voltage Vp and time tp as well as the measured χ(2) are stated. The
poling method has been indicated where: CM, charge migration. CI, charge injection
and -V indicates that the poling was performed under vacuum. An * indicates that the
value was not explicitly reported in the cited reference.
2. Using the results in Chapter 5 regarding the crystallistion stability of the glasses
we narrowed our focus onto the TZN2 and TZN3 compositions. These glasses rep-
resented the most promising candidates for optical fibre materials, as they possess
the largest crystallisation stabilities.
3. Results in the literature (see Table 7.1 indicate that most tellurite glasses have
optimal poling temperatures in the region of Tg − 50◦ C. This temperature is thus
used as a starting point for our investigations. The resistivity of the samples is
measured at this temperature by recording the current vs. voltage. During these
measurements we note the voltage at which the sample undergoes dielectric break-
down. This then represents the maximum voltage permissible, at this temperature.
We are able to discuss the dielectric breakdown in terms of voltage rather than
electric field because all of the samples under study were prepared with the same
132 Thermal Poling
thickness (1 mm). For the poling experiments the poling voltage will be chosen to
be ≈ 500 V lower than the breakdown voltage.
4. The poling program for each glass type involves first poling for a duration sufficient
to reach a sate wherein the current is zero. This situation is understood to represent
the case where the resistivity of the poled region is high enough to negate the
applied voltage, thus indicating the cessation of poling.
5. Following this, several samples will be poled for times less than the steady state
time.
6. Steps 4 and 5 are then repeated for several temperatures either side of the initial
Tg − 50◦ C.
Following the initial poling investigations it became apparent that the proposed exper-
imental plan was not adequate. Specifically, after each experiment the samples were
destroyed. Further to this, initial characterisation experiments revealed that no measur-
able second harmonic generation was observed in the sample fragments. This resulted
in a series of experiments designed to determine the issues identified with the initial
approach.
7.3.2 Thermal Poling Apparatus and Configurations
We constructed an apparatus for thermally poling glass. Shown in Figure 7.3 is the
apparatus which is placed in an oven during the poling process. Stainless steel electrodes
are held in ceramic (Macor, Corning Inc.) blocks to electrically insulate them from the
support frame and oven. The cathode is mounted on a spring, as indicated schematically
in Figure 7.3. The anode is fixed to a plate, the vertical position of which can be adjusted.
The electrical connections are made with glass fibre insulated high voltage cables with a
stainless steel braid over-jacket. These cables are capable of supporting >10 kV at over
500◦ C.
The final design was achieved through trial and error. It was discovered that it is critical
to have no sharp edges on the poling electrodes as this often results in dielectric break
down due to the high electric fields that form at these edges. We also found that spring
loading one of the electrodes enables one to ensure good physical contact between the
sample and the electrodes, the force of which is kept fairly constant during the periods
of thermal expansion and contraction during heating and cooling of the sample as well
as from experiment to experiment.
Chapter 7 133
Figure 7.3: Thermal poling apparatus with the electrical configuration indicated
schematically. Stainless steel electrodes contained within ceramic insulating blocks are
fixed to metal plates, one of which has vertical adjustment. The lower electrode, chosen
to be the ground is spring loaded to ensure good physical contact between the electrode
and the sample. The entire assembly can be placed in an oven.
134 Thermal Poling
7.4 Results and Discussion
B
ased on the crystallisation stability results in Section 5.3.1, where it was determined
that TZN2 and TZN3 represented promising base glasses for optical fibres due to
the large ΔT values and, in the case of TZN3, very small enthalpy of crystallisation,
we decided to focus the efforts to create a second order nonlinearity in these glasses.
Furthermore, we had many more TZN2 glass samples than any other composition and
as the process of thermal poling, as will be discussed, is often destructive most of the
poling results are for TZN2.
In the following sections the two poling techniques are discussed separately and the
results for each are presented in each respective section. The successful generation of a
second order nonlineariry is determined by exposing the poled sample to the output of a
Q-Switched Nd:YAG laser operating at 1064 nm, and 10 Hz. The SH signal is detected
with a photomultiplier tube and an oscilloscope. This measurement provides a yes/no
answer to weather or not a second order response has been generated. Subsequent to
this measurement, a precise determination of the vale of χ(2) is made via the Maker
fringes experiment (Section 8.1).
7.4.1 Charge Migration Thermal Poling
Thermal Poling of Infrasil

As a test of the performance of the thermal poling apparatus and the Makers fringes
experiment, that was constructed for determining the value of the induced nonlinear
coefficient (see Section 8.1), we commenced this work by thermally poling a commercial
fused silica glass, Infrasil 301. There are many reports of the successfully poling of
this material [98–102], additionally, many analytical techniques have been developed
for studying the depletion region that is formed [98, 101–104] and finally, an excellent
theoretical model that describes the charge migration dynamics has been established
[105]. This being such a well studied material it makes an ideal test material for our
experiments and procedures.
The poling conditions for the Infrasil sample were chosen intentionally to produce a
sample with a low χ(2) , as a means of testing the limitations of the characterisation
experiment described in Chapter 8. From the work of Triques et al. it is known that
the observed second harmonic signal is smallest for samples poled for long times, i.e.
greater than 90 min at 280◦ C [99]. Therefore we used the following poling conditions:
A 1 mm thick Infrasil disc was thermally poled between two stainless steel electrodes
at 290◦ C for 90 min with an applied potential of 4 kV. No attempt was made to record
Chapter 7 135
the current during the poling, however, it was noted to be approximately 18 μA at the
commencement of the poling and ≈ 0 at its completion.
We recorded a non zero second order response from this sample using the Maker fringes
experiment. In Section 8.1 there is a detailed discussion relating to the measurement of
the induced nonlinearity.
Thermal Poling of Tellurite

Using the TZN2 glass we assessed the possibility of charge migration thermal poling using
doped silicon electrodes in physical contact with the glass sample. The temperature was
set to 250◦ C and the applied voltage was increased from 0→2.5 kV at a rate of 100 V
every 10 s. This enabled the current to stabilise between each voltage step also helping
to avoid thermal run away and dielectric breakdown [106]. It was determined that, at
the elevated temperature of 250◦ , for voltages above ≈2.5 kV the sample underwent
dielectric breakdown. The current through the sample was recorded at each voltage,
this relationship is plotted in Figure 7.4. The axis of Figure 7.4 have been chosen such
that a linear fit to the data provides us with the resistance via Ohm’s Law, V = IR.
We find the resistance of the TZN2 glass at 250◦ C to be 6×107 Ω. With a sample
thickness of ≈ 1 mm and an electrode surface area of ≈ 1.27 cm2 the corresponding
resistivity ρ is equal to 7.3×108 Ω.cm. For comparison the resistivity of fused silica is
1011 Ω.cm [107] at 250◦ C.
The next step in the process is to hold the voltage constant for some period of time to
allow the migration of charge and the formation of a depletion region. The formation of
a depletion region will be observable as an increase in resistivity and therefore a decrease
in the poling current. The explanation of this assertion is based on the fact that as we
form the depletion region the number of charge carriers decreases which means that the
conductivity of that thin region of glass decreases. Thus, the resistivity increases and
the voltage drop across the depletion region increases via Ohm’s Law. This effect has
been reported for glasses that pole via charge migration [106].
In Figure 7.5 are the data taken over 1 h of constant voltage for the TZN2 sample. Note
that the current in fact increases over this time period, indicating that no depletion
region was formed. A possible explanation for this observation is that at this temperature
and in this poling configuration the sample is conducting electrons from the cathode to
the anode. These electrons can then in turn alter the oxidation state of the cations within
the glass and thus change the conductivity. In particular if the redox state of Te4+ or
were to change to produce elemental metallic Te then we would naturally expect the
conductivity to increase. We see evidence for this in Figure 7.6 which shows an optical
micrograph of the cathodic surface of the glass sample in question. There is evidence
136 Thermal Poling
3500
DATA
3000 Linear Fit
2500
2000
Voltage [V]
1500
1000
500
0
-5 -5 -5 -5 -5
0.0 1.0x10 2.0x10 3.0x10 4.0x10 5.0x10
Current [A]
Figure 7.4: Ohms law plot for TZN2 at 250◦ C. Blue squares: Data for TZN2 and
linear fit in red.
Figure 7.5: Time dependence of the current through TZN2 at 250◦ C.

Chapter 7 137
of chemical reduction of Te4+ →Te in the region of contact between the cathode and
the glass at the left of Figure 7.6. The dark spots are thought to be evidence of this
reduction. The part of the sample that was in the vicinity of the edge of the silicon
electrode has a higher concentration of these dark regions, which is consistent with the
higher field strengths encountered at the edges of a conductor.
Figure 7.6: Optical micrograph of the cathodic surface of TZN2 after thermal poling.
The region to the left was in contact with the Si electrode.
The phenomenon of chemical reduction at the cathode has been reported for certain
tellurites [91] as well as other non-silica glasses [24]. In particular Deparis et al published
a photograph of the cathodic surface of a bismuth borate glass that experienced chemical
reduction (reproduced in Figure 7.7) the appearance of which closely matches what is
observed in the TZN2 glass. Deparis et al analysed the sample with x-ray photoelectron
spectroscopy (XPS) and found that the proportion of oxygenated bismuth is significantly
reduced in the dark features. Thus indicating that the bismuth was indeed reduced into
metallic bismuth. We observe similar features in the thermally poled TZN2 sample.
Another possible cause for the increase in conductivity would be the establishment of
microscopic cracks in the sample. Such a crack can conduct electrons on the surface of
the material instead of through the volume. Surface conductivities are, in general, higher
than volume conductivities which would match with the observed change in current. The
microscopic inspection of the sample did not show unequivocal evidence for micro cracks,
however, higher spatial resolution may be required.
Following the period of constant voltage and temperature, we decreased the temperature
to ambient in two steps: First, from 250→200◦ C at 5◦ C.min−1 then from 200→30◦ C at
a cooling rate of 3◦ C.min−1 . We used such a slow cooling rate to avoid breaking the
138 Thermal Poling
Figure 7.7: Reproduction of the chemically reduced region at the cathodic surface of
a Bismuth borate glass. Inset scanning electron micrographs of the individual particles.
(Reproduced from [24]).
tellurite, as it was observed to be an extremely fragile material that would often break
upon cooling.
Notice, that in Figure 7.5, after the initiation of the cooling period the current initially
decreases from ≈ 7.5 μA to ≈ 6.5 μA and then remains constant. We stopped recording
when the sample had reached 150◦ because it was noticed that the sample had cracked
and was discharging through this crack.
The sample did produce light at 532 nm when exposed to the output of a Q-switched
Nd:YAG laser. However, further investigations revealed that this was, in fact, due to
ablation of the surface caused by the strong optical absorption of the black particles
on the cathodic face. The ablation produced white light the 532 nm component of the
white light made it to the PMT and was incorrectly recorded as a SH event. Therefore,
we cannot state definitively whether or not a second order response was induced.
7.4.2 Charge Migration Thermal Poling Using a Blocking Electrode
In an effort to pole the tellurite glass via charge migration while avoiding chemical
reduction at the cathode we trailed two blocking electrode configurations as shown in
Figure 7.8.
Chapter 7 139
Figure 7.8: Two possible configurations for the blocking electrode thermal poling
technique (anodic sides towards the top). (A) A borosilicate blocking electrode is in
physical contact with the glass to be poled. The poling electrodes are n-type silicon
for the anode and stainless steel (s.s.) at the cathode. (B) The borosilicate blocking
electrode is separated from the glass to be poled by a p-type silicon wafer. The poling
electrodes are the same as for (A).
Tellurite Sample A
The configuration shown in Figure 7.8 (A) has a doped silicon anode and a stainless steel
cathode, of which the anode is in physical contact with the tellurite (TZN2), blocking
the flow of electrons from the cathode is a 0.15 mm thick borosilicate microscope cover
slip. Borosilicate is a good choice for a blocking electrode as it has a relatively high
ionic conductivity at elevated temperatures as well as a very low electron conductivity.
It is thus able to prevent the flow of electrons from the stainless steel cathode with out
reducing the field across the sample.
We equilibrated the temperature of the assembly to 250 ◦ C then increased the voltage
across the electrodes from 0→3.0 kV at 100 V every 10 s. The voltage was then held at
3.0 kV for 1 h.
Plotted in Figure 7.9 (blue squares) are the data for this experiment. Note that once
the voltage is held constant that the current steadily decays then, during the cooling
phase the current further decreases to zero.
A current profile such as this is consistent with the formation of a depletion region. As
the decrease in current is associated with an increase in the resistivity of the ion depleted
layer. After the sample was removed from the oven and inspected, three things were
observed. First, the tellurite sample was stuck to the borosilicate cover slip. Second,
the tellurite was severely cracked and third there was a white precipitate on the cathode
From these observations we conclude the following: Charge migration occurred in the
borosilicate glass whereupon Na2+ left the glass and deposited itself on the cathode. The
Na2+ combined with atmospheric oxygen to produce oxides of sodium. The migration
140 Thermal Poling
Figure 7.9: Time dependence of the current for TZN2 in the two blocking electrode
configurations as shown in Figure 7.8. Blue squares: Data for configuration (A). Red
circles: Data for configuration (B).
of positive charges away from the anodic side of the borosilicate produced a strong elec-
trostatic attraction between the cover slip and the tellurite. Upon cooling, the difference
in thermal expansion coefficients between tellurite and borosilicate induced stress in the
tellurite which ultimately cracked under the compressive forces. This produced irregular
glass fragments with dimensions of ≈ 2 × 2 mm.
We were unable to determine if a second order nonliearity was induced in the sample,
due to the difficulties in mounting such small, irregular samples and directing the probe
laser beam through them at angles sufficient to observe second harmonic generation
(θ ≈ 60◦ , for details of this experiment see Chapter 8).
Tellurite Sample B
Following this we repeated the experiment, instead using the configuration (B) in Figure
7.8. The same temperature, voltage and times as were used for the configuration A were
applied here. As shown in Figure 7.9 (red circles) the dynamics of the process are
subtly different. The key differences are that the maximum achieved current is slightly
smaller. Secondly, the decay of the current during the constant voltage and temperature
phase of the experiment essentially halts after ≈ 15 min then remains constant until the
temperature is decreased. The first observation can be explained by the added resistance
Chapter 7 141
introduced by the p-type silicon, which has quoted resistivity range of 1-20 Ω.cm, as well
as the resistance of the interfaces, the value of which is difficult to estimate. As for the
constant current during the period of constant voltage, we believe that a surface current
was established during this period thus corrupting the data. Supporting this claim is the
visual observation of arcing across the surface immediately prior to the cooling period.
After the sample had cooled it was inspected and the following observations were made.
The tellurite sample had a dark area where the p-type silicon had been in contact with
it just as did the sample poled between two silicon electrodes (see Section 7.4.1). As
the flow of electrons should, in principal, be blocked by the borosilicate then the p-type
silicon must have acted as a source of electrons which caused chemical reduction of the
tellurium ions. It was also noted that there was a white precipitate on the cathode, most
likely Na2 O. Therefore, there was a reasonably high electric field across the borosilicate
and so there must have been conduction of charge through or across the tellurite.
Sample B displayed the same tendency to ablate at the cathodic surface as did the TZN2
sample poled in between two silicon electrodes with no blocking electrode.
7.4.3 Charge Injection Thermal Poling
The charge injection method is by far the most common approach towards poling tellurite
glass. Of particular relevance is the work done by Narazaki et al where sodium zinc
tellurites were studied [94]. They report a maximum d33 of 0.23 pm.V−1 for sodium
containing zinc tellurites with an optimised poling temperature that depends on the Tg
with the following linear relationship:
Tp = 0.72 × Tg + 45 (7.1)
Using Equation 7.1 to predict an optimal poling temperature for TZN2 and TZN3 we
obtain Tp ≈ 250◦ C.
Using this poling temperature we attempted to thermally pole TZN2 and TZN3 glass
samples of 1 mm thickness sandwiched between two commercial borosilicate microscope
cover slips1 . The poling configuration is illustrated schematically in Figure 7.10,
where the electrodes are made from stainless steel and the blocking electrode is a borosil-
icate microscope cover slip. Following is the descriptions of the TZN2 and TZN3 exper-
iments covered separately:
1
Frequently in the cited references these are referred to as ’microscope slides’. This is an unfortunate
misuse of the language. Microscope slides are ≈ 1 mm thick, the electric fields that result when this
thickness of glass is used are insufficient for poling.
142 Thermal Poling
Figure 7.10: Charge injection thermal poling configuration (anodic sides towards the
top). The glass to be poled is sandwiched in between two commercial borosilicate
microscope coverslips which are sandwiched in between two stainless steel electrodes.
TZN2 Charge Injection Thermal Poling

We sandwiched a 1 mm thick sample of TZN2 in between two borosilicate microscope
cover slips as per Figure 7.10. This in turn was placed between two stainless steel
electrodes and placed in an oven. Once the temperature of the assembly had equilibrated
at 250◦ C for approximately 1 h the voltage across the electrodes was steadily increased
in 100 V steps up to 3.0 kV. We then held this voltage and temperature for 60 min
before decreasing the temperature to ambient with the voltage held constant.
Figure 7.11 shows the time dependence of the current through the poling circuit. As
can be seen, the increase in current is linear during the initial stage of the experiment
while the voltage was being increased linearly with time. Subsequent to this there is an
observed decay in the current during the phase of constant voltage. This is produced by
the migration of charge and the associated increase in resistivity due to the depletion of
charge carriers.
After the sample had cooled, it was dismantled and inspected. The TZN2 sample cracked
into small pieces, and physically stuck to the borosilicate cover slip at the cathodic
side. As with the experiment in Section 7.4.2 we believe that these two observations
can be explained by the electrostatic attraction between the TZN2 sample and the ion
depleted anodic surface of the cover slip. This sticking produces a stress on the glasses
during thermal contraction. We made visual inspections of the sample at several stages
during the poling process and this cracking only began to appear during the temperature
decrease phase of the experiments.
The size of the fragments was approximately 2 × 2 mm, consequently we were unable to
measure the induced nonlinearity.
TZN3 Charge Injection Thermal Poling

The experiment above was repeated using a sample of TZN3 and a slightly lower poling
Chapter 7 143
Figure 7.11: Time dependence of the current for TZN2 in the charge injection con-
figurations.
temperature of 230◦ C. We used this new poling temperature to see if there was any
effect on the sample destruction that always accompanied the use of a borosilicate cover
slip as a blocking electrode. With reference to the data presented in Section 5.3.2, the
coefficient of thermal expansion of TZN3 is similar to TZN2.
The time dependence of the current was not recorded in for this experiment, however,
it was noted that the current did follow approximately the same profile as for the TZN2
charge injection thermal poling experiment above. i.e. there was an initial linear in-
crease of current during the voltage ramping phase, an exponential-like decay during
the constant temperature and constant voltage phase, followed by a more rapid decay
during the temperature decrease phase.
The observations made after the sample had cooled were consistent with the TZN2
experiment. The average size of the fragments was ≈ 4× the size of the TZN2 fragments.
By chance one fragment was just large enough to use for a SHG measurement, which is
described in the following Section 8.1.
144 Thermal Poling
SHG was detected coming from the TZN3 sample. Further measurements were made of
its angular dependence and are shown in Section 8.1.
7.5 Conclusion
T
hermal poling experiments were performed on the TZN2, TZN3 and Infrasil 301
samples. Both poling configurations were explored for the tellurite glasses.
In the charge migration configuration we observed, relative to Infrasil, high conductiv-

ities as well as probable chemical reduction of the tellurium ions. We were unable to
determine if a second order nonlinearity had been produced because the formation of
black particles on the surface which ablated when exposed to the pulsed laser radiation.
The high conductivities can be attributed to the tellurite glass possessing a relatively
high proportion of sodium ions. For comparison, Infrasil has approximately 1 ppm of
sodium where as the tellurite glass has approximately 100,000 ppm. We note that there
was not a 5 order of magnitude difference between the observed conductivities of the
Infrasil and tellurite and thus there are necessarily other factors to consider. Factors
such as; different mobilities of the sodium ions in the two glasses and the reduction of
the tellurium could also contribute.
Tellurite glass poled via the charge injection method developed a white precipitate,
most likely an oxide of sodium. Additionally, the structural integrity of the glass in
compromised when poled in this manner. Thus resulting in significant cracking of the
sample. Second harmonic generation was observed from charge injection thermally poled
tellurite glass.
In terms of future work, the development of thermally poling tellurite based optical fibres
is likely to be challenging. The charge injection method, which was shown to produce
a non zero χ(2) , is not feasible in the fibre format. The electrodes must be positioned
near the core of the fibre (see Figure 7.12) and thus are internal to the fibre and, given
current knowledge, it is not possible to have a ion source in between the anode and the
tellurite glass. The charge migration method, may or may not produce a non zero χ(2) ,
however it is now known that significant chemical changes occur in the region of contact
between the glass and the electrode. In a fibre this will compromise the strength of the
fibre as well as its optical absorption.
The problem of chemical reduction at the cathode could also be detrimental for a ther-
mally poled fibre, however, a very encouraging result from Margulis et al shows that a
fibre may be poled in the charge migration configuration with both internal electrodes
acting as the anode [108]. In this case the cathode, or ground potential is provided by
Chapter 7 145
Figure 7.12: Schematic illustration of the electrode configuration for thermal poling
of optical fibres. This cross section of the fibre shows the core positioned bewteen the
two internal electrodes which are provided via the technique in Section 9.3.
ionisation of the air surrounding the fibre and so the exterior surface behaves as the
ground. After charges begin to migrate from near the anodes towards the surface of
the fibre the resistivity of the glass immediately next to the electrodes increases causing
the potential at the fibre core to decrease. As this process progresses, the potential
difference between electrodes and core increases, pushing the depletion region towards
the core and thus poling the fibre. Fibres poled via this method have shown a factor of
2 increase in the achieved χ(2) [108].
NOTE:
Figure 7.13: Simulated equipotential maps prior to (a) and during (b) poling with
two anodes. (c) Initially Na+ ions are evenly distributed the voltage drop is mainly
between the electrodes and the surface of the fibre. (d) After charges have migrated
the resistivity of the glass around the electrodes increases and the voltage drop across
the core increases. distributions for fibres with two anodes. Reproduced from [108].
This result offers a promising solution to the problem of chemical reduction at the cath-
ode for thermally poled tellurite fibres, as it takes the reducing electrode (the cathode)
146 Thermal Poling
out of the vicinity of the fibre core. It is also probable that the enhanced χ(2) observed
in silica fibres will also be seen in tellurite fibres.
It is our conclusion that the only potentially feasible way to thermally pole an optical
fibre made from the tellurite glasses under study is the dual anode poling as reported
by Margulis et al [108]. As this will avoid the precipitation of metallic tellurium in the
vicinity of the core.
Chapter 8
Measurements of Second Order

Nonlinearities
8.1 Introduction
A
t the outset of this work our goal was to thermally pole the glass compositions
deemed most suitable for optical fibre fabrication under various conditions of tem-
perature, voltage and poling time. These samples would then have their second order
nonlinear properties measured so as to determine optimal compositions and poling con-
ditions which could then be investigated in the fibre format eventually leading to χ(2)
fibre devices. The complexities of thermally poling these glasses was not anticipated,
primarily due to the lack of definitive information in the literature. That being so, the
anticipated requirement of characterising the thermally poled glasses necessitated the
development of an experimental apparatus for measuring χ(2) as well as various compli-
mentary characterisation techniques whereby the spatial extent of the poled region in
the glasses could be investigated. This work progressed in parallel to the thermal poling
program during which much progress was made in the understanding of the experimental
techniques and data analysis.
There are several techniques for measuring the second order nonlinear properties of
materials such as parametric florescence, difference frequency generation and second
harmonic generation techniques such as Maker fringes [109]. Based on a review of the
literature it was determined that the Maker fringes technique was the most suited for
measuring the nonlinear properties of thermally poled glass samples. The nonlinear
region in these samples is typically very thin, ≈ 10 μm, thereby ruling out parametric
florescence and difference frequency generation which require long sample lengths in
order to increase signal to noise levels [109].
The following sections describe the Maker fringes technique, including a qualitative
description of the origin of the fringes (a derivation of the Maker fringes can be found
147
148 Measurements of Second Order Nonlinearities
in Section 2.3.2), the design and characterisation of the experimental apparatus as well
as the data analysis techniques. Measurements of the second order nonlinear properties
of a quartz reference sample, thermally poled Infrasil and tellurite glass are presented
in Sections 8.5.1, 8.5.2 and 8.5.3 respectively.
We explored two methods for determining independently the spatial extent of the nonlin-
ear region. The impetus for this stems from the reliance on the accuracy of the measured
χ(2) values on the accurate knowledge of the size of the thermally poled region. The
Maker fringes analysis does provide a measure of this region, however, in order to be
certain of the results we obtained independent measurements. Two techniques were iden-
tified in the literature as being suitable, namely: Second harmonic microscopy (Section
8.6.1) and differential etching (Section 8.6.2) [86, 101, 110].
8.2 Maker Fringes Analysis: Background
T
he predominate technique for determining the strength of the second order nonlin-
ear susceptibilities of a material is the Maker fringes analysis. Maker fringes were
first discovered by Maker et al [111] in 1962, shortly after the the invention of the laser
and the advent of nonlinear optical research. Maker’s experiment (Figure 8.1) involved
exposing a quartz plate to the output of a ruby laser, the transmitted light was filtered
to remove the red light and the blue SHG light was detected. The quartz sample was
rotated about its crystalline Z axis and the change in power of the blue light signal was
recorded as a function of incidence angle (Figure 8.2).
Blue Filter
T
Ti:Sapphire Detector
Quartz plate
Figure 8.1: Experimental configuration for Maker et al. original Maker fringes exper-
iment. The high intensity light from a Ti:Sapphire laser in passed through a sample of
quartz. The quartz is rotated with respect to the incident beam by an angle θ. A filter
removes the red light from the beam and the generated blue SH is detected.
A theoretical description of the Maker fringes was first put forth in a paper by Jer-
phagnon and Kurtz (JK) [112] in 1970. In Section 2.3.2, we have derived a version of
the Maker fringes expression that closely resembles that of Jerphangnon and Kurtz.
Jerphagnon and Kurtz made certain assumptions when deriving the expression, such as
no reflection of the SH at the material/air interface, no wavelength dispersion in the
Chapter 8 149
NOTE:
Figure 8.2: Maker fringes for quartz first measured by Maker et al. Reproduced from
[111].
refractive index (i.e. nω = n2ω ) and no absorption of the fundamental or SH waves. In

1995 Herman and Hayden (HH) published an extended theory of MF that accommodated
for these additional effects. The HH Maker fringes expression is applicable for materials
that have high refractive index, thus making the reflection at the material/air interface
appreciable, high dispersion, or materials that are absorbing at either of the frequencies.
To demonstrate the difference between the JK Maker fringe and those of HH we have
plotted the theoretical fringes for a y-cut quartz with a thickness of 2 mm as shown in
Figure 8.3.
With reference to the two sets of fringes in Figure 8.3 the observed detail can be explained
by braking it down into four separate contributions:
Maker fringes are characterised by a small scale modulation seen on the order
of 10’s of degrees (Figure 8.2). This occurs because the effective thickness of the
material changes with angle. The effective thickness increases as t cos θω , where t is
the actual thickness of the material and θω is the incidence angle. As this thickness
increases the optical path difference between the fundamental, bound SH and free
SH waves changes. Thus the fringes in the pattern are in fact interference fringes
produced by the interaction between the three propagating waves.
On a very small angular scale (Δθ 1◦ ) there is a high frequency interference

pattern that results from SH interfering with reflected SH in the nonlinear material.
JK’s analysis did not predict this because they assumed that the reflection of the
Figure 8.3: Simulated Maker fringes with the inclusion of higher order interference
(red line) and without (blue line).
SH is negligible. They are predicted by the analysis of Herman and Hayden [113],
however, researchers rarely use an angular resolution sufficiently small to resolve
them. The red line in Figure 8.3 shows the simulated Maker fringes for y-cut quartz
with the higher order interference included. The HH Maker fringes expression is
derived assuming that the light is composed of infinite plane waves. Therefore, this
high order interference is observed for all incidence angle. In reality a laser is used
to probe the Maker fringes. Laser beams have finite widths thus at larger angles
the reflected beam does not overlap well with the transmitted beam. Therefore
this interference will only be observed at incidence angles near to normal incidence.
The envelope of the fringes is determined by the effective nonlinear coefficient deff
(Equation 2.76) which, depending on the material, generally has a period of the
order of 45 ◦ θ 90◦ , and the Fresnel reflection and transmission coefficients at
the fundamental and SH.
Subsequent to their initial Maker fringes paper, Jerphagnon and Kurtz applied the
measurement of Maker fringes to characterising relative nonlinear susceptibilities for
several materials [114]. This was the beginning of Maker fringes analysis and today it
is the most common technique for determining second order nonlinear coefficients.
Chapter 8 151
8.3 Experimental Details
8.3.1 Experimental Plan
The experimental plan for the determination of the second order nonlinear coefficients
for the thermally poled samples was developed under the following considerations:
There is no way to know in advance the magnitude of the nonlinearities that could
be induced in the samples. Therefore we were guided by results in the literature to
give an indication of what to expect. As per the information collated in Table 7.1
tellurite glasses have been observed to possess χ(2) in the order of 1 to 0.1 pm.V−1 .
Accordingly, it was deemed necessary for the Maker fringes apparatus to be capable
of measuring nonlinearities at least one order of magnitude lower than the lowest
value reported in the literature, i.e. 0.01 pm.V−1 .
It is always possible to measure a low nonlinear coefficient by using higher intensity

laser pulses to produce a measurable amount of SHG. Provided, that is, that
the laser pulse is not sufficiently intense as to ablate or damage the material
under study. The problem we face is that tellurite glasses have low laser damage
thresholds. Literature reported results indicate that tellurites have surface damage
thresholds in the order of 10 GW.cm−2 which, for comparison, are three orders
of magnitude lower than silica (7 × 103 GW.cm−2 ) [60]. We therefore sought to
design a Maker fringes experiment that had high SHG detector sensitivity and
high signal to noise rather than high laser intensity.
These considerations and the overall design of the Maker fringe experiment are addressed
in the following Section 8.3.2.
8.3.2 Description of Maker Fringes Experiment
We constructed a Maker fringes measurement system (see Figure 8.4) described as fol-
lows:
The fundamental excitation source was provided by a modified Spectra Physics Quan-
taRay GCR-16 that produces p-polarised, 10 ns, Q-switched pulses at 1064 nm with
a pulse repetition rate of 10 Hz. We introduced an intra-cavity aperture of ≈ 3 mm
diameter to force the laser to run in a Gaussian-like transverse mode.
First the output of the laser is passed through an edge filter with > 99.9% attenuation
below 550 nm. This element was critical as it eliminated the flash lamp light from
PD SP
PC Oscilloscope
DM DM
ND P2
PMT2
F532
L3
PMT1
DAQ
Sample F532
P3
O/2 L1 T L2 ND SP
M
50:50 DM
Power
P1 Reference
O/2
(Quartz)
Nd:YAG BD
1064nm W
LP
Figure 8.4: Configuration for the Maker fringes experiment. Red dotted lines: Op-
tical path for the fundamental beam. Green dotted line: Optical path for the second
harmonic beam. LP: Long pass (> 550 nm) filter. W: Glass wedge. λ/2: Halfwave
plate. P1-3: Polariser. M: Mirror. L1-3: Lens. ND: Neutral density filter. DM:
Dichroic mirror. SP: Short pass (< 600 nm) filter. F532: 532 nm laser line filter. PD:
Photodiode. PMT1-2: Photomultiplier tube. Grey lines: Electrical connections.
the beam path which contained a significant amount of light at the SH (532 nm). To
maintain intensities sufficiently low to ensure that the samples are not damaged we
attenuated the output via reflection from a glass wedge (W). This provided a nominal
pulse energy of 200 μJ as determined by the calibrated photodiode (PD).
The light passed through the wedge is directed into a beam dump (BD). Additional
control of the pulse energy was achieved with a half wave plate (λ/2) and polariser (P1)
combination.
The beam is then directed through a second λ/2 plate to allow control of the polarisation.
The beam is focused with a 20 cm focal length lens (L1) through a 50:50 non-polarising
beam splitter, provides us with two focused fundamental beams with pulse energies of
100 μJ each. We calculate the intensity at the focus to be in the order of 10 MW.cm−2 ,
i.e. three orders of magnitude below the reported damage threshold of tellurite glass
[60].
One of the beams was directed onto the sample under study, which is mounted on a
computer controlled rotation stage. The SHG that is produced in the sample, as it
Chapter 8 153
is rotated about the input beam, is collected by the lens L3. We attenuate the SH
signal with a calibrated neutral density filter (ND) the strength of which depends on
the sample being used and is chosen to reduce the signal to coincide optimally with the
dynamic range of the detector electronics. Residual fundamental radiation is filtered
from the signal by two dichroic mirrors (DM), with approximately 70% transmission at
1064 nm, a short pass filter (SP) and a 532 nm laser line band pass filter (F532) which
is mounted directly onto a photomultiplier tube (PMT1). This combination of filters
provides approximately 83 dB suppression of the fundamental wavelength.
Prior to detection a polariser (P2) is used to select a specific polarisation state the nature
of which depends on the material under study and the form of its nonlinear susceptibility
tensor.
The other focused fundamental beam is passed through a reference sample of y-cut
quartz which is held at a fixed angle to the beam. The signal is conditioned in the
same was as for the sample and detected with a photomultiplier tube (PMT2). The
second harmonic radiation generated from this sample was used as a reference channel
to normalise the power fluctuations in the sample under study. Normalising to the SHG
signal is preferable to normalising to the laser output power as the SHG is sensitive
to the square of the laser power and thus provides a more sensitive measure of the
fluctuations in that power. In order to minimise stray light signals from the room the
entire apparatus is contained within a light tight box.
8.3.3 Data Acquisition
We designed a built the data collection system for the MF experiment. The system is
comprised of a pulse conditioning and sampling system in combination with a 16 bit
dynamic range National Instruments data acquisition (DAQ) module (USB-6210) and a
PC running LabView1 .The electronic circuit diagram for this system is shown in Figure
8.5.
The following list details the sequence of events that take place during the data acqui-
sition:
1. The flash lamp synchronised trigger signal is sent to the system which initiates the
sample and hold (S/H) timing.
1
The Author specified the requirements of the system which was then designed and built by Mr.
Neville Wild and Mr. Bob Nation.
154
CHARGE SENSITIVE AMPLIFIER SHAPING CIRCUIT GAIN BUFFER
*
100pF
+12
1N4148 220pF 220pF
500k
FROM PMT ANODE 2
2 1 3
U2A MONITOR
220R 1 3 1 3
1N4148 U1A U3A
3 120k 120k 2 1
100nF U4A
LF412 1M 2 2k2
NOTE 3
LF412 LF412 NOTE 4
-12 PART B 220pF LF412
5
100pF
7
6 * *
R 120k
LF412
AI 0
100k USB-6210
THIS CIRCUITRY REPEATED FOR SECOND CHANNEL - USE "B" PARTS
AI 8
SAMPLE AND HOLD 100k
AI GND
+5 +5
IN
2
+5 8
U5/U6 OUT AI 1/AI 2
10k _
1nF 1nF
* * 3 S/H
GND AI 9/AI 10
47k 200k 68k 7
1
2
15
14
AD781
D GND
TRIGGER IN
C
C
R/C
R/C
(FROM LASER)
4 6 12 10
A Q A Q AI SAMPLE CLOCK (PFI 0)
1k U7A U7B _
5 7 11 9 S/H
B Q B Q
50R 2N3646
+5 +5 V POWER
3 13 U8B
+5 CLR +5 CLR U8A 10uF
_
74HC4538 74HC4538 S/H 1 2 3 4
SAMPLE (PFI 4)
+5
DELAY: 50 us to 150 us HOLD TIME: 50 us 74HC04 74HC04 CONVERT (PFI 5)
1k
SPARE INVERTERS
COUNTER OUTPUT EVENT (PFI 6)
U8C U8D
USB ON
DC ON 5 6 9 8
TRIGGER OUT
74HC04 74HC04 (TO LASER)
U8F 39R
3k3
DC-DC CONVERTER 13 12 CONVERT
+12 VDC IN TEN 3-1223 SUPPLY CONNECTIONS: NOTES: 39R
23 14 +15 V +12 V
+Vin +Vout 74HC04
47uH 22 78L12 LF412 PIN 4 = -12 V 1: ALL IC SUPPLY PINS SAMPLE
+Vin 10uF PIN 8 = +12 V BYPASSED WITH 100 nF U8E 39R
CERAMIC CAPACITORS
16 AD781 PIN 5 = -12 V 11 10 S/H MONITOR
COM
9 PIN 1 = +12 V
other power is supplied by the USB-6210, internal circuitry not shown.

COM
2: * DENOTES MF RESISTOR 39R
PIN 3 = GND OR COG CAPACITOR
74HC04
3 10uF
-Vin
2 11 79L12 74HC4538 PIN 8 = GND 3: PROTECTION COMPONENTS
-Vin -Vout
-15 V -12 V PIN 16 = +5 V (OMIT FOR LESS NOISE)
THE UNIVERSITY OF ADELAIDE - SCHOOL OF CHEMISTRY & PHYSICS
cated in the device for the additional input channel. The pulse conditioning circuitry
is powered from an external DC supply (shown bottom right of Figure). The pulse
system. Pulse conditioning circuitry for one input channel shown on left. This is dupli-
sampling circuitry (shown in the centre of the Figure) is utilised for both inputs. All
Figure 8.5: Circuit diagram for the Maker fringe data conditioning and acquisition
74HC04 PIN 7 = GND 4: OPTIONAL COMPONENTS SEAN'S PMT PULSE MEASUREMENTS
PIN 14 = +5 V
DAQ ELECTRONICS
21-Apr-2009 DAQ ELECTRONICS.SCH
Measurements of Second Order Nonlinearities
Chapter 8 155
2. Shortly after, a pulse from each of the PMTs enters a charge sensitive amplifier.
This integrates the voltage pulse producing a signal the peak of which is propor-
tional to the total charge in the pulse, which is in turn proportional to the power
in the optical pulse.
3. We shape the short, asymmetric output from the charge sensitive amplifier by
using low pass filters to produce a ≈ 100 μs long symmetric pulse with a peak
voltage which is still proportional to the optical power.
4. We then use a sample and hold circuit to sample the peak of the shaped pulse.
5. The output of the sample and hold is acquired by the DAQ and sent to the PC.
The LabView program that controls this process is also responsible for controlling the
angular position of the sample rotation stage and dividing the sample signal by the
reference signal.
8.3.4 Calibration and Alignment
The linearity of the detectors was investigated to identify the optimal operating condi-
tions. For PMT bias voltages of 1500 V and 2000 V we used a series of calibrated neutral
density filters to attenuate the signal generated from a y-cut quartz sample which was
oriented at an angle, ≈ 10◦ , to the incident beam that maximised the SHG signal (see
Figure 8.10). The plots in Figure 8.7 show that the detector has a linear response up
to approximately 4.5 V independent of PMT bias voltage. As such, we ensured that all
acquired signals were appropriately attenuated to ensure that the signal fell below this
value.
We determined the number of pulses needed to ensure a measurement error below 1%

buy collecting data on the signal (CH1), reference (CH2) and the normalised signal
(CH1/CH2) channels. Using the LabView program we collected the standard error,
represented as a percentage and plotted this data as a function of the number of pulses.
Figure 8.8 shows this data. When the number of pulses exceeds ≈ 150 the relative
error is < 1% in the normalised signal thus ensuring an accuracy in calculated nonlinear
coefficients of approximately the same order.
It is important that the focal point of the beam coincide with the rotation axis of the
sample otherwise, as the sample rotates, the sample will either move away from or
towards the focal point thus altering the intensity at the sample.This makes the Maker
fringe appear asymmetric thus increasing the error in the nonlinear coefficient. For
the alignment transverse to the beam a 0.1 mm diameter aperture was mounted in the
Figure 8.6: Timing diagram for the Maker fringes experiment data acquisition system.
sample holder which was translated on a micrometer driven stage orthogonal to the
beam to maximise the optical throughput.
The position of the sample relative to the focus of the beam is also critical for an
accurate measurement of the Maker fringes. It was shown by Jerphagnon and Kurtz
that the minima of the fringes are influenced by the divergence of the beam [112] which
must be minimised. The optimal position for the sample is therefore at the focal point.
Also, the Rayleigh range should be made to be large compared to the sample thickness
to minimise the divergence at the sample.
Chapter 8 157
4
PMT Signal [V]
y = 424.35x + 0.3287
R² = 0.9607 y = 25.045x + 0.08
3 R² = 0.9966
1500V
2 2000V
Linear (1500V)
1 Linear (2000V)
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Transmission
Figure 8.7: Photomultiplier tube output signal vs. fractional transmitted intensity
from the y-cut quartz reference oriented at 10◦ to the beam to ensure maximal SHG
(see Figure 8.10) Red triangles: Data for 1500 V bias with linear fit. Blue squares:
Data for 2000 V bias with linear fit.
To ensure that the sample is correctly positioned longitudinally, with respect to the
focus, we mounted the focusing lens on a translation stage and recorded the SH signal
generated from a thermally poled Infrasil sample, which has its nonlinear region at the
surface, at various lens-to-sample separations. We have plotted the second harmonic
power P2ω , vs the micrometer reading on the lens translating stage in Figure 8.9 (Red
circles). It is clear from the data that the micrometer reading of 10.5 mm corresponds
to the maximum SHG signal and thus the position at which the sample is at the focal
point. This is because the second harmonic power is proportional to the intensity of the
fundamental to the second power and thus P2ω ∝ 1/w4 where w is the beam radius.
−1/4
Interestingly, if we plot P2ω vs position we have a parameter that is proportional to
the beams radius and its dependence on position. From the theory of Gaussian beams
[115] the expression for the beam radius as a function of the propagation coordinate is
2 1/2
z
w(z) = w0 1 + (8.1)
z0
where w0 is the beam radius at the focus and z0 is the Rayleigh range given by
πw02
z0 = (8.2)
λ
8
CH1
7 CH2
CH1/CH2
6
Relative Error [%]
0
0 100 200 300 400 500 600 700 800
Number of Pulses
Figure 8.8: Relative error in the Maker fringes data vs number of pulses averaged
over.
By first fitting a function of the form

2 !1/2
(z − c)λ
w(z) = a b 1 + (8.3)
bπ
we are able to find the constants a, b and c where a is the scaling constant to turn the
arbitrary units from the measurement into beam radius in mm. The constant c is the
horizontal shift required to put the waist at z = 0 and b = w02 . In so doing we find that
a = 7.72, w0 = 0.096 mm and c = 10.5 mm. Using this information and Equation 8.2
we calculate the Rayleigh range to be z0 = 27 mm.
We stress that this measurement has limited accuracy due the measurements being made
close to the waist and, in particular, no far field measurements being made. Neverthe-
less, this measurement indicates that, to a reasonable approximation, the beam can be
considered Gaussian and therefore the standard expressions for the Maker fringes can
be applied. Furthermore, the Rayleigh range is > 10× the thickness of the reference
quartz sample and so the divergence problem is avoided.
Chapter 8 159
4.0 0.86
P2ω
-1/4
P2ω 0.84
Gaussian Fit
3.5
0.82
[arb. units]
P2ω [arb. units]
0.80
3.0
0.78
-1/4
P 2ω
0.76
2.5
0.74
2.0 0.72
-5 0 5 10 15 20 25
Position [mm]
−1/4
Figure 8.9: Red: Plot of SH power vs lens to sample separation. Blue: P2ω vs lens
to sample separation and Gaussian beam fit to data.
8.4 Data Fitting Techniques
I
n general, determining the value of the second order nonlinear coefficient for a material
involves; first recording the Maker fringe for a material with a known second order
nonlinear coefficient, fitting the theoretical Maker fringes expression (Equation 2.62) to
the data then under identical conditions (i.e. laser power, beam spot size, etc) recording
the Maker fringes for the sample under study and fitting Equation 2.62 to the data.
The fitted equations can then be compared to provide a relative measure for the second
order coefficient of the sample.
The fitting and subsequent calculation of the nonlinear coefficient is what we refer to as
‘Analysis’. The following is a discussion of the steps required to preform this analysis.
The expression for the Maker fringes is given by Equation 2.62, which, for convenience,
is reproduced below:
2 2
(Tωa→g ) T2ωg→a 2
ω 2πL sin2 Ψ
P2ω (θi ) = 3 2 2
P2 d2eff
Ac 0 n2ω nω cos θtω ω λ Ψ2
The Makers fringes expression can be written simply as a function of incidence angle,
θi by applying Snell’s law. We make the replacement θi → θ and collect the constants
into a single constant, α, in order to simplify the expressions. We can therefore write
Equation 2.62 as:
P2ω = αL2 T (θ)d(θ)2 Ψ(θ) (8.4)
where we have summarised the contributions to the expression in terms of the inter-
ference term Ψ(θ) = sin2 Ψ/Ψ2 , the transmission term T (θ), and the angular part of
the effective nonlinear coefficient d(θ) which depends on the material in question, its
orientation and the input and output polarisation. Note that we are making a relative
measurement and will thus be interested in a ratio of the constant α for the two mate-
rials. We can therefore neglect the common constants such as c and 0 . Furthermore,
the power measurements are normalised and so Pω is equal to unity. By making these
considerations we obtain
α = d2 (8.5)
where d2 = d2ij which is simply the square of the particular element of the second order
nonlinear susceptibility tensor under study.
If we acquire the Maker fringes data for a reference sample and fit Equation 8.4 to the
data we obtain a reference value of α = αr . We then record the Maker fringes for the
sample under study, making sure that the same experimental conditions are used, i.e.
laser power and beam area. After fitting Equation 8.4 to this data we have the sample
value of α = αs . The ratio of αs /αr can be used to calculate the value of the nonlinear
coefficient for the sample by:
2
αs ds
= (8.6)
αr dr
Thus "
αs
ds = dr (8.7)
αr
It is occasionally the case that the SH signals exceed the dynamic range of the PMTs, in
particular for the reference material. In this case it is necessary to attenuate the signals.
Equation 8.7 is modified to yield
"
αs (NDs −NDr )/2
ds = dr .10 (8.8)
αr
which accounts for attenuation of the sample and reference signals with neutral density
filters with values of NDs/r , respectively. χ(2) is then calculated via the relationship in
Equation 2.12, i.e. χ(2) = 2ds .
In the following sections we describe two techniques for fitting Equation 2.62 to exper-
imental data. We quote the nonlinear coefficients in terms of elements of the reduced
Chapter 8 161
nonlinear susceptibility tensor suing the contracted notation, i.e. the dil coefficients
(See Section 2.3), as these are what is measured. In this context χ(2) is a less useful
parameter, but can be easily calculated from d if necessary.
8.4.1 Root Mean Square Error Minimisation Fitting Procedure
This data fitting technique considers the Maker fringes expression to be a function of
two parameters, α and L. Using the condensed version of the Maker fringes Equation
8.4 we can write it as:
P2ω (θ : α, L) = αL2 T (θ)d(θ)2 Ψ(θ) (8.9)
Which indicates that Equation 8.9 is a function of the angle θ and the fitting parameters
α and L.
Maker fringes data will inevitably have some small angular offset which arises from the
difficulty in perfectly aligning the sample with respect to the beam. We denote this
offset as Δθ. Maker fringes data is always symmetric about the zero degrees, or normal
incidence angle and so in order to remove the angular offset from the data we need to
find the centre of the data and the difference between it and the zero angle position. We
do this prior to the following data fitting procedure.
The fitting procedure begins with the generation of an array of pairs of α and L, where
the parameters are allowed to vary over some physically meaningful range. This range
is determined by the sample at hand. From each of these pairs of α and L a theoretical
Maker fringes pattern is generated. We then find the root mean square error (RMSE)
between the function and the data. The minimum of this array and the parameters
corresponding to that minimum provides us with the best fitting parameters.
8.4.2 Genetic Algorithm Fitting Procedure
The RMSE minimisation technique described above has one inadequacy which needs
to be addressed. The angular offset Δθ is obtained by finding the point of symmetry
in the data and then displacing that point so as to coincide with zero degrees. We
found no precise method for achieving this, other than finding the mid point between
corresponding features in the data which are subject to experimental error. Instead,
we developed an alternative data fitting algorithm that finds the value of Δθ as well
as α and L. This alternative data fitting procedure uses a genetic algorithm to breed
an optimal set of input parameters for a nonlinear least squares (NLS) fit. While this
is not the first time genetic algorithms have been applied to data fitting [116–118] the
approach is quite uncommon in literature despite its effectiveness.
The details of the algorithm are summarised below:
1. The algorithm begins by generating an initial population, subject to upper and

lower bounds, of potential fits in the form of a vector of the form:
Fn = [αn , Ln , Δθn ] (8.10)
where Fn is the nth member of the population with guess values for the scale
factor αn , nonlinear thickness Ln and an additional parameter Δθn which removes
any angular misalignment from the measurement by shifting the data set about
until it is centred on 0◦ (the value of Δθ is typically less than about ±2◦ ). We
typically use a population of 500 members and the upper and lower bounds used
will depend on the sample under study. For example: The quartz reference we use
can have its thickness measured mechanically with a micrometer to an uncertainty
of ≈ ±2.5μm.
2. Next, the algorithm assesses the fitness of each member of the population by using
its parameters as the input for a NLS fit. The input for the fit requires upper U ,
and lower bounds L, in which to search for a good fit and these are determined
from Equation 8.10 in the following way:
Un = [αn + δαn , Ln + δLn ]

Ln = [αn − δαn , Ln − δLn ]
where 0 < δ < 1, typically ≈ 10−5 . We keep δ small to prevent the NLS from
wandering too far from the initial guess values. Note that the angular offset Δθ is
not an input into the NLS fit. This is because we add Δθ to the angular position
of the measurement before performing the NLS fit, hence simplifying the solution
space for the NLS fitting algorithm.
3. We compile the RMSE for each potential fit. This is used as the so called fitness
function that the GA requires to guide the next stages of the algorithm.
4. The next step in the algorithm is called crossover and is akin to sexual reproduction
in animals. This step is performed by pairing members of the initial population
and exchanging their parameters. For example: Member a and b are paired each
with parameters Fa = [αa , La , Δθa ] and Fb = [αb , Lb , Δθb ] which after crossover
Chapter 8 163
could look like:
Fab1 = [αa , Lb , Δθb ]

Fab2 = [αb , La , Δθa ] .
We have indicated the ‘children’ of Fa and Fb as Fab1 and Fab2 . Members of the
population are paired preferentially based on fitness.
5. After crossover we also apply some mutation to the new ‘children’. This has the
effect of preventing the GA form converging onto local minimum by introducing
some freak members of the population that may (or may not) accidentally acquire
superior qualities. 20% of the population is subject to mutation. The mutation is
performed by randomly selecting members and adding a random number to one
or all of the parameters. This random number is subject to the upper and lower
bounds imposed on the population from the outset.
6. Once the new population has been produced we return to Step 2. This process
is repeated until we satisfy the halting criterion, which is reached if there is no
improvement in the fitness for 5 consecutive generations.
8.5 Results and Discussion
8.5.1 Quartz Reference Measurements
The chosen reference material was a sample of y-cut quartz, cut to 2 mm thickness
and optically polished. This material was chosen for a number of reasons: It possesses
a second order nonlinearity. Quartz is a very stable material. It is not hygroscopic
in contrast to some nonlinear materials. Quartz has a relatively high laser damage
threshold ensuring that the samples will not be damaged when exposed to the pulsed
laser source. Quartz possesses a relatively low nonlinear coefficient and thus will not
require excessive attenuation to avoid saturation in the detection system. Finally, it is a
well studied material with reliable refractive index and second order nonlinear coefficients
data available in the literature. For future reference, the optical parameters of quartz
that are used in this work are given in Table 8.1.
We measured the Maker fringes for the quartz reference sample with p-polarised laser
excitation and p-polarised detection. The SH power was recorded for incidence angles
between ±45◦ with an angular resolution of 0.2◦ . A neutral density filter with N D = 2.0
was used to decrease the SH power to a level sufficient for the detection system to not
be saturated.
Parameter Value Units

n1064 1.53413 -
n532 1.54702 -
d11 0.49 pm.V−1
Table 8.1: Optical parameters for quartz reference sample. Refractive index values
taken from [119]. Nonlinear coefficient taken from [114]
Figure 8.10 shows the recorded data (blue dots). The central region of the data contains
the interference fringes resulting from the interaction between the generated SH and the
SH reflected at the material/air interface. The Maker fringes model we are using does
not account for this interference. The data was first treated by removing the central
±10◦ to omit the SH interference, thereby reducing the fitting error. We then analysed
this data using the two methods outlined in Section 8.4. We summarise the results for
the two methods below:
Root Mean Square Minimisation

From the raw data an estimate of the angular offset Δθ was obtained by taking equivalent
points in the fringes and finding the difference in their angular position. By applying this
method to several points an estimate of Δθ ≈ −1.5◦ was obtained. We then computed
the RMSE for a range of fitting parameters 0.1 ≤ α ≤ 1.5 with increments of 0.0005 and
1.8 mm≤ L ≤ 2.2 mm with increments of 0.0005 × 10−3 mm. The parameters which
resulted in the lowest RMSE and therefore best fit are tabulated in Table 8.2.
Genetic Algorithm
The MF expression was also fitted to the data using the Genetic algorithm fitting pro-
cedure. The upper and lower bounds for the GA were:
0.5 α 2
1.99 × 10−3 L 2.10 × 10−3
and −2 Δθ 2.
The optimised fitting parameters are listed in Table 8.2. We note that the two methods
Method α L [mm] Δθ
RMSE Minimisation 1.113 2.035 -1.5
Genetic Algorithm 1.116 2.036 -1.500
Table 8.2: Optimised fitting parameters for y-cut quartz Maker fringes as determined
by the RMSE minimisation and GA fitting algorithms.
provide values for the coefficients that are in good agreement, within 0.3% for α and
0.05% for L. Further, the GA has arrived at the same value for Δθ as did we, i.e. −1.5◦ .
Chapter 8 165
2.5
2
[arb. units]
1.5
2ω
P
0.5
0
−40 −30 −20 −10 0 10 20 30 40
Incidence Angle θ [°]
i
Figure 8.10: Optimised fit to y-cut quartz Maker fringes data. Blue dots: Recorded
Maker fringes data. Red line: Best fit as determined by GA fitting procedure.
Using the fitting parameters from the GA method we plot the Maker fringes (red line
in Figure 8.10) fitted to the experimental data (blue dots in Figure 8.10). Note, that
in the central ±10◦ there is a vertical spread in the data about the fitted curve. This
is due to the interference between the generated SH and the SH that is reflected at the
material/air interface. The interference is only observed within the ±10◦ region due to
the finite width of the beam.
The thickness of the quartz sample was measured with a micrometer which yielded
2.039±0.0025 mm, in close agreement with the value determined from the Maker fringes
analysis. Thus validating the Maker fringes measurement.
8.5.2 Thermally Poled Infrasil
The Maker fringes of the thermally poled Infrasil were recorded using p-polarised input
light and p-polarised detection for incidence angles between ±80◦ and at a resolution of
0.5◦ . A neutral density filter of OD = 0.6 was used to decrease the SH power to avoid
saturation in the detection system. The acquired data was analysed using the two data
fitting techniques as described below.
Root Mean Square Minimisation

We first established an estimate for the angular offset in the data of Δθ = 0.33 by
averaging the difference in the angular position of three pairs of corresponding points.
The algorithm then computed the RMSE for fits to the experimental data with input
parameters ranging over 0.1 ≤ α ≤ 1 with increments of 0.001 and 1.0 mm≤ L ≤ 20 μm
with increments of 0.001μm. The optimal fitting parameters obtained from this method
are tabulated in Table 8.3.
Genetic Algorithm
When applied to the data for thermally poled glass a problem arises. Despite the ability
of the GA to avoid getting stuck in a local minimum in the RMSE landscape, the
algorithm if run multiple times returns very different parameters for fits with a range
of RMSEs. A scheme was developed to overcome this difficulty, the details of which are
summarised below:
1. We choose an educated guess value for the nonlinear thickness, say 1μm then we
tighten the upper and lower bounds on the nonlinear length. We typically use
L ± 0.5μm.
2. The algorithm is then run for this guess value for the nonlinear length.
3. We obtain the optimised fit parameters and the RMSE.
4. Next, we increment the guess value for the nonlinear thickness and re-run the
algorithm.
5. This process is repeated over a range of valued of L. We obtain a plot of RMSE

vs L and by using value of αr and Equation 8.8 a plot of RMSE vs d33 as shown
in Figure 8.11.
Figure 8.11: Convergence of Maker fringes fitting parameters for thermally poled
Infrasil. Left: Thickness of the nonlinear region, L. Right: Nonlinear coefficient, d33 .
Chapter 8 167
It is clear from Figure 8.11 that there is indeed an optimal set of parameters despite the
inability of the GA to find these without being made to scan the parameter space. The
optimised parameters are tabulated in Table 8.3.
Method α d33 [pm.V−1 ] L [μm] Δθ

RMSE Minimisation 0.196 0.035 11.70 0.33
Genetic Algorithm 0.192 0.033 11.79 0.589
Table 8.3: Optimised fitting parameters for thermally poled Infrasil Maker fringes as
determined by the RMSE minimisation and GA fitting algorithms.
Comparing the fitting results from the two methods (Table 8.3) we see that the agreement
is good. In particular the percentage differences are 2% for α, 5.7% for d33 and 0.8%
for the thickness of the nonlinear region L. The discrepancies can be attributed to the
difference in the values of Δθ used in each method. In our assessment, the value for
Δθ as determined by the GA technique is more accurate because this particular value
optimises the fit, whereas, the value use in the RMSE minimisation method was simply
an estimate based on a smaller set of experimental data points. As such, we conclude
that the GA data fitting algorithm has increased the accuracy of the measurement of
d33 by almost 5.7%. Although, it must be stated that one could conceive of a more
complicated scheme for obtaining Δθ from the experimental data and then feed this
into the RMSE minimisation method. In this case the expectation would be for the
parameters from each technique to converge. The GA method simply provides a single
technique for determining all of the fitting parameters.
Plotted in Figure 8.12 are the measured data (blue dots) and the fitted curve (red line)
using the optimised fitting parameters obtained from the GA method. In Section 8.6
we present independent measurements of the thickness of the nonlinear region for the
thermally poled Infrasil.
Given that we used a N D = 0.6 neutral density filter in the detection system and the
dynamic range of the DAQ is 16 bits, we estimate that the minimum detectable second
order nonlinear coefficient is in the order of 1 × 10−4 pm.V−1 .
8.5.3 Thermally Poled Tellurite
The angular dependence of the SH was recorded for the thermally poled TZN3 sample
(discussed in Chapter 7). The experimental conditions were identical to conditions used
to measure the the thermally poled infrasil (refer to Section 8.5.2). Figure 8.13 shows the
experimental data. There are several points to make regarding this data. Specifically,
P2ω does not go to zero at 0◦ as would be expected. Also, the fringes are not perfectly
1.5
1
[arb. units]
2ω
P
0.5
0
−80 −60 −40 −20 0 20 40 60 80
i
Figure 8.12: Optimised fit to thermally poled Infrasil Maker fringes data. Blue dots:
Recorded Maker fringes data. Red line: Best fit to data provided by GA fitting routine.
symmetric about 0◦ . There are several possible causes for this. These include: The
creation of an inhomogeneous nonlinear region caused by imperfect contact between the
TZN3 sample and the borosilicate cover slip. Ablation of surface contaminants by the
laser. One potential source of surface contamination are oxides of sodium (or other
alkalis) that precipitated after the poling treatment.
0.7
0.6
0.5
[arb. units]
0.4
0.3
2ω
P
0.2
0.1
0
−80 −60 −40 −20 0 20 40 60 80
i
Figure 8.13: Measured angular dependence of second harmonic power for thermally
poled (CI) TZN3 at 230◦ C and 3 kV for 60 min.
Both of the afore mentioned fitting techniques were applied to the data for the TZN3
sample. Neither technique provided a fit with reliable parameters. In particular the
Chapter 8 169
RMSE minimisation technique predicts a value for d33 in excess of 100 pm.V−1 . The
validity of this results is highly dubious, considering this is at least two orders of mag-
nitude higher than any previously reported second order nonlinear coefficient observed
for tellurite glass. As for the GA method, we find no converging solution. From this we
can only conclude that fitting a Maker fringes expression to this data is unjustified, due
to its poor quality. Accordingly, we have no estimate for the strength of the induced
nonlinearity for this sample.
8.6 Measurement Techniques for the Thickness of the Non-

linear Region
T
o verify the effectiveness of the Maker fringes fitting algorithms an independent
verification of the width of the nonlinear thickness, otherwise referred to as the de-
pletion region, is necessary. We employed two complimentary techniques for determining
the nonlinear thickness second harmonic microscopy (SHG microscopy) and differential
etching.
8.6.1 SHG Microscopy
SHG microscopy is becoming a common imaging technique in the physical and biological
sciences. The technique looks at a materials second order nonlinear response on the scale
of ≈ 500 nm and upward. This is therefore a useful technique for analysing thermally
poled glasses as the region of nonlinearity is typically very thin, in the order of several
microns.
A sample was prepared from the thermally poled Infrasil for SHG microscopy in the
following way2 :
1. The thermally poled glass disc was cleaved in two along a diameter by scoring with
a diamond scribe and bending of the sample about this scribe mark.
2. One half of the disc was then sandwiched between two microscope slides and fixed
with glue.
3. A thin slice was cut from this assembly and this was fixed with glue to another
microscope slide.
2
This sample preparation was performed by a technician at the University of Sydney’s Electron
Microscopy Group
4. The exposed surface of the poled glass was polished to an optical finish for viewing
in the SHG microscope.
SHG microscopy was performed on an inverted Leica DMIRBE microscope equipped

with a Leica TCS2MP confocal system and coherent Verdi-Mira tunable pulsed titanium
sapphire laser3 . This laser provided the fundamental excitation at 830 nm, with pulses
of approximately 150 fs in duration in addition to a continuous wave optical channel, at
543 nm, for sample viewing. The weak SH signal was separated from the optical channel
with a dichroic mirror and recorded by photomultipliers.
Figure 8.14 is composed of a schematic illustration of the sample configuration (Figure

8.14 (a)), an optical micrograph of the sample assembly (Figure 8.14 (b)) and an overlay
of the optical channel and the SHG channel from the SHG micrograph (Figure 8.14 (c)).
The glue that fixed it to the microscope slides is observed as a strip of dark markings on
the SHG micrograph. From the optical channel we were able to calibrate the position of
the SHG signal with respect to the edge of the glass (i.e. the edge of the glued region).
Using MATLAB we took a line scan of the SHG signal which was averaged over the entire
image. One can naively deduce that by taking the square root of the SHG intensity signal
a spatial profile for the second order nonlinear (SON) susceptibility can be determined.
This is because P2ω ∝ (d33 )2 EDC . Figure 8.15 shows the square root of the SHG
intensity as a function of scan position, where the anode side of the glass is to the right
of the SON profile plot.
The plot in Figure 8.15 displays the characteristic shape displayed by thermally poled
silica [86, 110]. There is a sharp increase in the SON profile near the anodic surface
of the glass which rises to a maximum a few micron into the glass. The signal then
decreases at a slower rate further into the glass.
It appears that the width of the SON profile is ≈ 16 μm (i.e. from a scan position of
93 μm to 109 μm), however, the actual spatial extent of the SON profile is most likely
shorter than this. The reason for this statement is as follows: As we have exposed a
part of the glass that contains excess charge, this charge produces fringing fields. Thus
we have disturbed the electric field that existed inside of the glass and consequently the
recorded SON distribution is not exactly representative of the distribution that exists
inside of the glass.
3
SHG microscopy was kindly performed by Mr. Guy Cox at the University of Sydney’s Electron
Microscopy Unit.
Chapter 8 171
Figure 8.14: Top left: A schematic representation of the sample assembly used for
SHG microscopy. The poled Infrasil is sandwiched in between two microscope slides
and glued in place. Right middle. The sample is orientated with the anodic face of
the poled Infrasil towards the bottom of the page: An optical micrograph of the region
of interest for the SHG microscopy. The highly polished Infrasil can been seen at the
top and the microscope slide at the bottom. In between the two glass layers there is a
layer of glue ≈ 40 μm thick. Left bottom: An overlay of the optical (green) and SHG
(false colour orange) channels taken from the SHG microscope. The darker region to
the bottom of the SHG is the glue layer.
Figure 8.15: Graph: Mean line scan of the SHG channel taken from an SHG micro-
graph of thermally poled Infrasil with the anodic side of the glass to the right of the
peak. Inset: False colour image from the SHG channel. Note: the red dot to the right
of the inset figure is an SHG response from a contaminant and has been removed from
the line scan.
8.6.2 Differential Etching
It has been established that the depletion region formed in a thermally poled silica glass
displays a different etching rate in HF acid than the unpoled glass [104, 120]. It was
shown that the depletion of ionic species from a region in fused silica reduces the rate at
which HF acid etches that region by approximately 1/2. This effect has been exploited
in order to reveal the spatial extent of the depletion region by several authors [86, 101].
There are various experimental techniques that have been developed to glean information
from the HF etching of poled silica. In particular an interferometric technique was
developed by Margulis and Laurell [120] which simultaneously measures the amount of
material being removed and the SHG signal being generated from a pump laser. This
technique allows accurate measurement of the spatial profile of the nonlinear region
and represents the ultimate characterisation experiment for poled glasses. We, however,
have opted for a much simplified version of this technique. The poled glass was simply
etched and then the depletion region was imaged with an electron microscope so that
measurements of the width of the depletion region could be made.
As with the SHG sample we sandwiched the cleaved glass disc between two microscope
slides, fixed in place with a contact adhesive, and polished the assembly to provide a
flat surface, perpendicular to the poling direction. The microscope slides were removed
from the sample by immersion in dichloromethane for several hours. We then etched
the polished face of the sample in a 45 w.t.% solution of hydrofluoric acid for 60 s. The
sample was then mounted for viewing on an electron microscope. The etched surface
was imaged in the secondary electron emission configuration.
The SEM micrographs of the HF etched thermally poled Infrasil are shown in Figures
8.16 and 8.17. An etch rate of ≈ 1μm.min−1 for 45 w.t.% HF acid on fused silica is
reported in [121] and, as stated in [104], the etching rate of the ion depleted region
should be ≈ 1/2 the rate of the undepleted glass. Despite this we see clear evidence that
the etch rate of the anodic surface was indeed higher than this. With reference to Figure
8.17 we see that the anodic surface of the poled glass has developed large (≈ 2 μm) open
holes in a 2 μm thick layer below the anodic surface. The implication is that the etch
rate was closer to 2 μm.min−1 for the anodic layer. It is likely that this result stems
from the documented phenomenon of H+ injection that is associated with long poling
times [102]. Furthermore, it has been observed by Triques et al. that the initial ≈2 μm
of glass does indeed etch faster than the undepleted glass [99].
Chapter 8 173
The edges of the depletion region are indicated by the dashed lines in Figure 8.16. The
top line indicates the position of the glass surface prior to etching, the position of which
was calculated via the argument above. The lower dashed line coincides with a ridge
resulting from the different etch rates of the ion depleted region and the pristine glass.
We estimate the thickness of the depletion region to be approximately 12 μm (see Figure
8.16).
Figure 8.16: SEM of the etched depletion region of thermally poled Infrasil. Assuming
an etch rate of ≈ 2 μm.min−1 we have indicated the approximate position of the glass
surface prior to etching.
Figure 8.17: SEM of the anodic face of thermally poled infrasil. A region of higher
etch rate is clearly visible at the top of the image.
We have overlaid the trace obtained from the SHG microscopy with the SEM of the
depletion region as shown in Figure 8.18. There was some freedom with respect to
the precise positioning of the SHG profile so it was positioned by eye to provide best
overlap with the proposed thickness of the depletion region. When this is done we can
see that the peak of the SHG response is located in the region of maximal etching that
we hypothesise is rich in H+ due to charge injection during poling.
Figure 8.18: Overlay of SON profile with SEM of depletion region. SON profile (in
green) is overlaid with the SEM of the depletion region. The thickness of 12 μm is
indicated to show the approximate position of the surface of the unetched glass.
It has previously been reported that there is indeed a peak in the SHG response within
the first couple of microns of a thermally poled Infrasil sample for long poling durations
[102]. Figure 8.19, which is reproduced from [102] shows (white circles) that for an
Infrasil sample that has been poled for 100 min at 250◦ and 4.0 kV there is an initial
peak in the value of the second order nonlinear susceptibility, χ(2) . As the amount of
SHG is determined by χ(2) it is therefore clear that a peak in the SHG response should
also be observed.
Additionally, the SON profile in Figure 8.18 has reasonably distinct knees at either
side of the peak which closely coincide with the edges of the depletion region. It is
possible that the discontinuities in electric field at the edges of the charged depletion
are responsible for the knees in the SON profile. However, to verify this claim more
investigations are required.
Chapter 8 175
NOTE:
Figure 8.19: Second order nonlinear susceptibility as a function of depth under the
anodic surface. White circles: This data shows that there is an initial peak in the SON
response in Infrasil in the long poling regime. Reproduced from [102]
8.7 Conclusion
A
Maker fringes measurement experiment was constructed that was capable of deter-
mining the angular dependence of the second harmonic power generated in a poled
glass sample. The apparatus provided measurements of SH power with a maximum er-
ror of 1%. Based on a measurement of a prepared sample of thermally poled Infrasil,
the apparatus can measure second order nonlinear coefficients of at least 0.03 pm.V−1 .
Further, we estimate that second order nonlinear coefficients as low as 1 × 10−4 pm.V−1
could be measured with this system.
We developed two methods for fitting the Maker fringes expression to the measured
data for determining the strength of the induced nonlinearity and the spatial extent of
the thermally poled region. Both methods were in close agreement with one another
when the sample was very think as compared to the coherence length. For poled glass
the absence of fringe detail in the Maker fringes data made it difficult to obtain an
accurate estimate of the angular offset. The GA method showed superiority to the
RMSE minimisation method in that a more accurate measure of the angular offset was
obtained leading to more accurate values for the thickness of the nonlinear region and
the second order nonlinear coefficient. Measurement of the thermally poled tellurite
sample yielded data from which neither algorithm was able to yield valid results. This is
primarily due to the poor quality of the data that resulted from surface contamination
during the thermal poling procedure.
Independent measurements of the thickness of the nonlinear region of the thermally

poled Infrasil sample were made. Both the SHG microscopy and differential etching
results are in good agreement with the thickness determined via the data analysis.
Chapter 9
Fibre Preliminaries
9.1 Introduction
T
his Chapter comprises some preliminary work that was undertaken as a first step
towards fabricating optical fibre devices. This was conducted in parallel with
the production and characterisation of the tellurite glasses reported in the previous
Chapters. Several computational models were developed to give qualitative indications
of the performance of a range of potential optical fibre designs for electro-optic fibre
devices as well as rules of thumb for the design of fibres with internal electrodes. The
details and results from these computational models are found in Section 9.2.
In addition to modelling the properties of optical fibres with internal electrodes, some ex-
perimental work in establishing fibre processing capabilities, necessary for the fabrication
of optical fibres with internal electrodes, was performed. Techniques such as fabrication
of extruded optical fibre preforms with holes with which to accommodate electrodes.
As well as electrode insertion techniques were developed. The work on electrode in-
sertion was greatly facilitated by a collaboration with Walter Margulis at ACREO in
Stockhlom. During a two week visit by the Author to ACREO’s research facility, fibre
electrode insertion techniques were learnt and developed further. The techniques were
adapted so as to be more generally applicable to optical fibres with microstructuring.
The details of this work are discussed in Section 9.3.2.1.
Also discussed in this Chapter are the preliminary fibre fabrication trials dedicated
towards developing the capability to produce core/clad optical fibres made from tellurite
glass. In the future such core/clad fibres could be fabricated with internal electrodes,
which would be a significant milestone on the path to creating practical tellurite glass
based electro-optic fibre devices.
177
178 Fibre Preliminaries
9.2 Computational Modelling of Electro-optic Optical Fi-

bres
E
lectro-optic fibre devices require internal electrodes positioned close to the core
of the fibres in order to manipulate the light propagating through them. The
degree to which the refractive index is modified is related to the strength of the electric
field across the core, a relation that varies for the two specific cases of χ(3) and χ(2)
devices. The relationship that describes the change in index for a χ(3) device is given in
Equation 2.73. Where Eext is the strength of the applied electric field (and EDC = 0).
Typical values for χ(3) are in the order of 10−20 m2 .V−2 , thus Eext needs to be in the
order of 108 V.m−1 to obtain changes in the refractive index of around 1 part in 1000.
Note that the relationship is quadratic in the applied electric field. Typical electrode
separations would be in the order of 10 μm and thus voltages in the range of several kV
would be necessary.
For fibres that have been thermally poled a permanent DC electric field EDC , exists
across the core. The expression for the index change when an additional external field
Eext , is applied is given by the relationship given in Equation 2.73 (in this case EDC is
non zero). Accordingly, the change in refractive index is linear with respect to the applied
external electric field which, by virtue of the size of the frozen-in-field typically being in
the order of 108 V.m−1 , can be as low as 106 V.m−1 to achieve similar refractive index
variations as in the case of χ(3) devices. [105]. For electrodes separated by ≈ 10 μm this
equates to tens of volts of potential. It is for this reason that thermally poled electro-
optic devices are desirable over χ(3) devices as the required strength of the applied field
is so much lower.
Much of this Chapter addresses the issue of providing optical fibres with internal elec-
trodes and the effects that electrodes would have on the optical properties of the fibre.
There are several methods for introducing the electrodes, each with its advantages and
disadvantages. Below, we have listed some of the more common techniques with discus-
sion regarding the advantages and disadvantages of each:
Wire insertion: An optical fibre is fabricated with approximately 30 - 50 μm

diameter holes running along the length of the fibre parallel to the core as per
Figure (make the figure). Thin tungsten wire with a diameter as similar to the
holes in the fibre as piratical are then manually inserted into the fibre [122]. This
technique is limited as it can only produce optical fibres with internal electrodes
of up to approximately 10 cm and there is always an air gap between the electrode
and the surrounding glass.
Chapter 9 179
Simultaneous fibre+wire drawing: In this process the internal electrodes are

introduced into the fibre during the drawing process. In the fibre drawing tower
the fibre preform has a spool of thin tungsten wire above it and as it is drawn
down into the fibre the wire is drawn with it [123]. Fibres with internal electrodes
in excess of 200 m have been produced in this manner and there is no gap between
the electrode and the glass. However, the draw back with this technique is that
stress in introduced into the fibre due to the difference in the thermal expansion
coefficients of the cooling glass/metal combination.
Liquid metal injected electrodes: Electrodes are introduced by first producing

an optical fibre with 20-30 μm diameter holes running the length of the fibre
parallel to the core. Then one end of the fibre is sealed in a pressure vessel
immersed in a molten solution of metal and placed into a furnace. Typical alloys
used are BiSn and AuSn solders with melting temperatures of 170 ◦ C and 280 ◦ C
respectively. Under the application of high pressure (10-30 bar) the liquid metal
is pushed into the holes of the fibre [124, 125]. The fibre+metal combination is
allowed to cool slowly by gradually removing the fibre from the furnace from the
end furthest from the pressure vessel. In this way the liquid metal is allowed to
contract longitudinally and therefore a minimum of stress in introduced into the
fibre. Fibres with internal electrodes of ≈100 m have been produced in this way
and indeed this process could be scaled further. Additionally, this technique allows
multiple fibres to be filled simultaneously.
In the subsequent sections we analyse the properties of such a fibre. Specifically, the
DC electric field that can be produced between the electrodes and how the structure
of the fibre may effect the strength of the field across the core region. We have also
calculated the optical attenuation that inevitably results from having metal placed near
a propagating optical mode. This attenuation is then analysed for different optical fibre
types, i.e. core/clad and MOF as well as the influence of the distance from the optical
field and the electrodes.
9.2.1 Electric Fields Between Internal Electrodes
The strength of the electric field between the internal electrodes of the optical fibre is
an important factor for determining the performance of any device fabricated from such
a fibre. We have modelled the electric field that would exist in the core region of several
representative fibre types. The fibres we considered were a core/clad step index fibre
(SIF), a hexagonal three ring MOF (HexMOF) and a wagon wheel MOF (WWMOF).
In principle we can use the results for optimising of the positions of the electrodes for
electro-optic modulators and poling fibres and to better understand how poling might
take place in optical fibres with microstructured regions.
Using the generalised electromagnetic toolbox in COMSOL, a finite element software

package, we computed the electric field distributions over a range of electrode separa-
tions. The results for each fibre type are presented separately below:
Step Index Fibre (SIF): The SIF we modelled is based closely on a particular
fibre reported by Myrén et al [126]. This fibre represents the state of the art in
electro-optic fibres, thus providing an excellent point of comparison. Table 9.1
details the parameters used in the model.
Parameter Value/s
Fibre diameter 125 μm
Core diameter 4.3 μm
Electrode diameter 29.5 μm
Anode potential 4.0 kV
Cathode potential 0V
Electrode separation 6.2 de 32.3 μm
Boundary conditions:
Fibre surface Ground
Core Continuity
Anode Electric potential (4.0 kV)
Cathode Ground
Table 9.1: Input parameters for SIF electric field model. Optical fibre parameters
taken from [126]
The most difficult aspect of setting up this model was that of boundary condition
selection. We chose, for the fibre surface, a grounded boundary condition. This
may seem like an odd condition to use for a non-conductive material, however, by
comparing modelled potential fields with imaged depletion regions for real poled
fibres [126] we see in Figure 9.1 that this boundary condition gives the best agree-
ment with experiment. Note in Figure 9.1, left; the depletion region has extended
towards the fibre surface from the top of the anode. It is only possible for this to
occur if there is a potential difference between the anode and the fibre surface. The
simulation of the electric filed in the left of Figure 9.1 where a grounded boundary
condition was used for the fibre surface displays, qualitatively, the same shape as
the depletion region.
Further to this, a reported experiment in which two anodes were used proved
conclusively that the surface of the fibre is indeed grounded [108]. In this work
it was established that there is a mechanism for charge neutralisation at the fibre
Chapter 9 181
Figure 9.1: Comparison between experiment and simulation for a poled SIF and the
modelled electric field distribution. Left: The chemically etched depletion region of a
thermally poled fibre viewed with a phase contrast microscope [126]. Right: Simulated
electric field distribution for a similar fibre using a grounded boundary condition at the
fibre surface.
surface. The neutralisation is most likely to be produced by ionised species from

the air being attracted to the surface and negating the surface charge.
Figure 9.2 shows the electric field distribution for one particular electrode separa-
tion. The blown up core region (Right of Figure 9.2) shows that the field is fairly
uniform across the core region of the fibre. The electric field in the centre of the
core is plotted as a function of electrode separation in Figure 9.5 (Blue line). The
electric field falls away with electrode separation slightly faster than a typical 1/r2
relation. This is because the grounded boundary condition on the fibre surface
serves to bend the field lines towards it and as the electrode separation increases
this effect becomes more pronounced.
Hexagonal Array Microstructured Optical Fibre (HexMOF): A represen-

tative geometry was chosen for analysis of the electric field produced in the core
region of a microstructured optical fibre with a hexagonal array of holes (Hex-
MOF). We chose a design that is both possible to fabricate given our current
fabrication capabilities and, importantly, has close to the best confinement of the
optical mode. This confinement is determined by the ratio of the hole diameter
to their pitch, d/Λ = 0.5 for this fibre, as well as the number of rings of holes, in
this case 3. The need for high optical confinement will become apparent when we
analyse the optical attenuation of these fibres (see Section 9.2.2). The parameters
for the HexMOF model are listed in Table 9.2.
Figure 9.2: Modelled electric field distribution for a step index fibre with internal
electrodes. Left: The entire fibre cross section. Right: Magnified core region.
Parameter Value/s
Pitch (Λ) 2.0 μm
Hole diameter (d) 1.0 μm
d/Λ 0.5
Electrode separation 12 de 28 μm
Internal holes Continuity
Cathode Ground
Table 9.2: Input parameters for HexMOF electric field model.
The boundary condition for the air hole/glass interfaces in the microstrucured re-
gion of the fibre was chosen to be continuity. It is possible that, like the fibre
surface, these boundaries will in fact be grounded, however, we have no evidence
that this is the case and therefore we chose the simplest possible boundary condi-
tion that makes the fewest assumptions. The spatial extent of the microstructured
region prohibits the electrodes from being positioned close to the core. This is why
we have considered a smaller range of electrode positions than for the SIF.
It can be seen in Figure 9.3 that the electric field in the air holes of the fibre is
≈ 3 × 108 V.m−1 . This exceeded the break down field strength of air by two orders
Chapter 9 183
Figure 9.3: Modelled electric field distribution for a hexagonal three ring MOF with
internal electrodes. Left: The entire fibre cross section. Right: Magnified core region.
of magnitude1 . The existence of such high electric fields in the air holes can be
understood in terms of the continuity condition for the tangential components of
the electric field at the interface can be written as
Et1 = Et2 (9.1)
where Et1 is the component of the electric field that lays tangential to the interface
in region 1 and Et2 is the tangential component of the electric field in region 2.
However, for the normal components the boundary condition is written as
1 En1 = 2 En2 (9.2)
The change from one material to another at the interface implies that there will
be a change in the value of at the interface. Thus there is a discontinuity in the
normal component of the electric field En , across the boundary. The ratio of 1 /2
for the glass/air boundary is ≈ 2 which explains the observed difference between
the electric field in the glass and that seen in the air holes.
This result is concerning, as it would appear to limit the maximum possible field
strength to a value insufficient for thermal poling and Kerr modulation. However,
this result is predicated on the validity of the continuity boundary condition at the
air/glass interface of these holes, an assumption which may prove to be invalid. As
has been shown by Margulis et al, there is a mechanism for charge neutralisation
1
Dielectric strength of air is approximately 3 × 106 V.m−1 [127]
at the surface of the glass [108]. However, there is as yet no report of this occurring
within the microstructured region of and optical fibre.
Further to this, the question of how the depletion region would form in such a
fibre is an interesting one. For a first approximation, we would assume that during
poling the migrating species follow the field lines. The is no doubt that this would
be the initial situation. However, after the charges have migrated some distance
there would be a change to the net electric field, due to the non-equilibrium charge
distribution. To assess this more thoroughly a time dependent model would need
to be constructed. We leave this for future work.
This assumption is tested in Figure 9.1 and is shown to be approximately correct.

Thus, the depletion region which would form in this fibre would be very compli-
cated and it is difficult to predict how well it would overlap with the core. This
being so, there is at least one report of a fibre with a similarly complex holey
region being successfully thermally poled [128].
The dependence of the electric field in the centre of the core on the separation
of the electrodes is plotted in Figure 9.5 (red line). It is observed that the field
strength is slightly smaller than for the equivalent electrode separation in the
SIF. Coulomb’s law predicts an inverse square dependence of electric field on the
separation of the charges. However, this is only valid for charges in a medium of
constant permittivity. The differing permittivities of the air and glass produce the
observed difference between the SIF and HexMOF cases.
Wagon Wheel Microstructured Optical Fibre (WWMOF): The wagon

wheel optical fibre is extreme in terms of the amount of air surrounding the core of
the fibre. As such it is an interesting fibre to analyse in this way. There are many
benefits to such a fibre including low confinement loss, high mode confinement and
thus large fibre nonlinearity as well as large NA. The fibre considered in the model
is a straight forward design for fabricate and can be applied to many materials.
As with the HexMOF, we employed the continuity boundary condition at the

air/glass boundary of the internal holes surrounding the core. Furthermore, due
to the spatial extent of the air holes we are limited in terms of the proximity of the
electrodes to the core region. For the purposes of this simulation we had to choose
an orientation of the structure within the fibre. It was determined that the most
logical orientation is with one of the three struts aligned along the imaginary line
joining the two electrodes. This is because the air holes serve to shield the core
from the electric field and if we want charge migration to occur in the core region
we need to minimise this shielding.
Chapter 9 185
Parameter Value/s
Effective core diameter 1.0 μm
Electrode separation 13 de 37 μm
Internal holes Continuity
Cathode Ground
Table 9.3: Input parameters for wagon wheel structured electric field model. Note,
Effective core diameter is defined as the diameter of the largest circle that fits inside of
the triangular core region.
Figure 9.4: Modelled electric field distribution for a wagon wheel MOF with internal
electrodes. Left: The entire fibre cross section. Right: Magnified core region.
Again, as in the HexMOF case, the field in the air holes is prohibitively high with
respect the the dielectric break down of air. In terms of the possible evolution
of a depletion region the wagon wheel fibre looks promising. There exists a large
electric field in the core of the fibre and if we assume the ions will migrate along
the field lines then one would expect the depletion region to spread out to occupy
the entire core region (see Figure 9.4).
Figure 9.5 shows the electric field in the core for various electrode separations and
reveals that the wagon wheel provides the lowest field strength of the three fibres.
This is determined by the amount of air surrounding the core.
8
x 10
7
6 Wagon wheel
Step index
Hexagonal
Electric Filed [V.m ]
5
−1
0
5 10 15 20 25 30 35 40
Electrode Separation [μm]
Figure 9.5: Core electric field strengths for three representative fibre types with
internal electrodes calculated over a range of electrode separations. Blue: SIF. Red:
Hexagonal three ring MOF. Green: Wagon wheel MOF.
9.2.2 Electrode Induced Optical Attenuation
The presence of internal metal electrodes in an optical fibre will result in attenuation of
the optical field. This additional loss, despite being unavoidable, should be minimised.
To gain insight into how this might be achieved we modelled the loss induced by the
electrodes over a range of electrode to core separations using the finite element method.
This is, to the best of our knowledge, the first time this attenuation has been analysed
in this manner. We note that Myr’en reports a computation of electrode induced atten-
uation in his Thesis [31]. Their method involved using commercial beam propagation
software (BeamProp) and specifying the absorption of the metal (in their case Bismuth)
with a complex refractive index. Our investigations with finite element methods showed
that this approach leads to non-physical outcomes, with electric filed lines not meet-
ing the conductive surfaces at 90◦. Instead, we specify, within the software, that the
electrodes are metallic with finite conductivity.
Figure 9.6 shows an optical fibre with internal electrodes in cross section. We have
highlighted the upper right plane of the diagram to indicate that we performed a quarter
plane analysis [129].
When using a quarter plane analysis, the boundary conditions for the edges along the
x and y axis need to be chosen correctly to produce the correct optical modes and
polarisations. For this analysis we have only considered the fundamental LP01 mode of
the fibre, furthermore, the internal electrodes produce birefringence and therefore there
Chapter 9 187
are two non-degenerate orthogonally polarised LP01 -like modes. One has its electric
x . The other has
field Ex , aligned with the x axis of Figure 9.6, let it be denoted LP01
y
its electric field Ey , aligned with the y axis, denoted LP01 . To produce these two modes
we use the perfect electric conductor E0 , and perfect magnetic conductor H0 , boundary
x mode is obtained with the E condition on the x axis and
conditions such that the LP01 0
x mode is obtained with E on
the H0 condition on the y axis. Conversely for the LP01 0
the y axis and H0 on the x axis.
Figure 9.6: Optical fibre with internal electrodes used for modelling of electrode
induced loss. Left: Optical fibre cross section with the electrode-core spacing indicated
d. The 1/4 plane used for the calculations is highlighted with the annulus of high mesh
density in the electrode shown. Centre: Regions of important mesh density have been
indicated. Right: The boundary conditions used to obtain the two non-degenerate
orthogonally polarised LP01 type modes.
Parameter Value/s
Cladding index 1.438
Core diameter 4.3 μm
Core index 1.453
Electrode surface resistivity 0.1×10−6 Ω.m
Electrode conductivity 6.67×106 S.m−1
Core to electrode spacing 1 d 14 μm
Table 9.4: Input parameters for electrode induced loss model.
In the electrode region the optical field decays extremely rapidly. The depth over which
the current density decays to the 1/e value is the so called skin depth and can be
estimated via: " "

1 ρ ρ
δ=√ ≈ 503 . (9.3)
πμ0 μr ν μr ν
Where ρ is the resistivity and μr is the relative permeability of the metal and ν is the
frequency of the electromagnetic wave. Taking the values for BiSn as ρ = 1.29 μΩ.m2
and μr ≈ 1 and considering an incident wave at 1550 nm we obtain a skin depth of
≈ 41 nm. To obtain accurate calculations the region of the electrode over which the
optical field decays requires mesh elements with a maximum size less than the skin depth.
We performed a convergence calculation whereby the imaginary part of the propagation
constant (converted into optical loss in dB.m−1 ) for the two polarisations of the LP01
mode at 1550 nm was determined over a range of decreasing mesh element size. The
results of these calculations are presented in Figure 9.7. It can been seen that the value
for the loss is converging for decreasing mesh size. Convergence to 5 significant figures
was obtained.
1.17 11.6
1.16 11.5
1.15 11.4
Ey Loss [dB.m ]
Ex Loss [dB.m ]
-1
-1
1.14 11.3
1.13 11.2
1.12 11.1
1.11 11.0
1.10 10.9
-8 -7 -7 -7
0.0 5.0x10 1.0x10 1.5x10 2.0x10
Maximum Element Size [m]
Figure 9.7: Convergence of electrode induced loss as mesh element size is decreased.
x y
Blue: Loss of the LP01 mode. Red: Loss of the LP01 mode.
We next computed the loss for the two polarisations of the LP01 mode at 1064 nm and
1550 nm over a range of electrode to core separations. The results of these calculations
are presented in Figure 9.8 where we have changed the horizontal axis such that the 0
point conincides with the position of the mode field radius. This better illustrates the
dependence of the loss on the distance from the optical field. We calculated the mode
field radius by applying the Gaussian approximation for the fundamental mode of a SIF
2
We are assuming that the resistivity of BiSn is approximately the same as that of pure Bismuth
Chapter 9 189
with core radius a given by [130]

1.619 2.87
w0 ≈ a 0.65 + 3/2 + 6 . (9.4)
V V
The induced loss is sensitive to the distance between core and electrode. The computed
loss ranges over 4 orders of magnitude for a change in the position of the electrode
of only several micron. For a practical device the losses should be as low as possible.
Acceptable losses for the entire device are around 3 dB. Allowing for losses as the light
enters and exits the device (i.e. Fresnel losses) ,which could be as high as 1 dB, the loss
due to the electrodes can be no higher than 2 dB.
It is noteworthy to mention that the loss experienced by the y polarised mode is roughly
10× that of the x polarised mode. This can be understood if we consider that the electric
field lines must meet the metal surface at 90◦ thus the y polarised mode is coupled into
the metal much more effectively that the x polarised mode.
6
Ey 1064 nm
4 Ex 1064 nm
Ey 1550 nm
2 Ex 1550 nm
0
Log(dB.m )
-1
-2
-4
-6
-8
0.0 2.5 5.0 7.5 10.0 12.5 15.0
Mode field radius - Electrode Separation [μm]
Figure 9.8: Base 10 log of the electrode induced loss against mode field radius to
y x
electrode separation. Blue (): LP01 mode at 1064 nm. Blue (): LP01 mode at
y x
1064 nm. Red (): LP01 mode at 1550 nm. Red (): LP01 mode at 1550 nm.
Attempts were made to calculate the electrode induced loss for a wagon wheel MOF and
a HexMOF. In the case of the wagon wheel quarter plane analysis is not possible because
the wagon wheel structure has three fold symmetry. Thus, we needed to construct a
model for the entire fibre cross section. Just as for the SIF case we first attempted to find
a converged solution. Figure 9.9 shows the computed loss value for decreasing electrode
mesh element size calculated for a wagon wheel fibre with the same parameters as in
Section 9.2.2 and at a wavelength of 1550 nm. For this convergence test the electrode
was positioned as close to the core as possible (≈ 6 μm) to provide the largest possible
loss value.
4.00E-008
3.00E-008 Ex
Ey
2.00E-008
1.00E-008
Loss [dB.m ]
-1
0.00E+000
-1.00E-008
-2.00E-008
-3.00E-008
-4.00E-008
-5.00E-008
-8 -8 -8 -8 -8 -8 -8 -8
2.0x10 2.5x10 3.0x10 3.5x10 4.0x10 4.5x10 5.0x10 5.5x10
Maximum Element Size [m]
Figure 9.9: Electrode induced loss for a wagon wheel MOF calculated for various
x y
mesh element sizes. Blue: LP01 mode. Red: LP01 mode.
As indicated in Figure 9.9 a converged solution for this model was not obtained. We
can however state that when the electrodes are positioned as close to the wagon wheel
structure as possible, that is where the loss should be maximum, the model was unable to
resolve their influence on the light. Therefore we assert that the attenuation produced by
the electrodes is approximately zero. This indicates that the optical mode is, effectively,
completely shielded from the electrode. For example an equivalent electrode to core
separation for the SIF example produces a loss of ≈ 17 dB.m−1 . Such high optical
confinement is expected given the high refractive index contrast between the core and
the cladding (i.e. the surrounding air).
A convergent solution was obtained for the HexMOF model. At approximately 6 μm

from the core the loss is 112 dB.m−1 . This value is > 10× the equivalent electrode to
core separation for the SIF, a result which is consistent with the lower confinement of
optical modes in such fibres. The loss in such a fibre can, in principle be reduced by
increasing the size of the air holds and/or increasing the number of rings of holes. In
each case the refractive index distribution is becoming more like that of the step index
Chapter 9 191
fibre. However, as previously stated, this fibre geometry represents the best case for
optical confinement achievable with current fabrication capabilities.
9.2.3 Discussion of Results
The computational models indicates that the DC electric field strength in the vicin-
ity of the core varies approximately as 1/r2 where r is the electrode to core distance.
Fields of sufficient strength to induce the necessary electro-optical effects are possible,
in particular in the case of a thermally poled fibre were the required fields are much
lower.
We have shown that the loss induced by the internal electrodes varies exponentially with
core to electrode separation. This effect can be mitigated by confining the light more
strongly, i.e. higher refractive index contrast. The wagon wheel MOF and HexMOF
represent two extreme cases for electrode induced loss. The wagon wheel MOF, with
its very good modal confinement, experiences virtually no loss at all. On the other
hand, the HexMOF design considered here, has relatively poor mode confinement and
therefore displays high electrode induced loss.
It is thus clear that in order to minimise the attenuation produced by the internal
electrodes, the optical mode should be well confined to the core. This can be achieved,
in the case of a SIF, by using high refractive index contrasts. For MOFs this requirement
is met by incorporating large proportions of air in between the electrodes and the core.
9.3 Insertion of Electrodes into Optical Fibres
W
e have determined the liquid metal injection method (Section9.2) to be the supe-
rior method for providing optical fibres with internal electrode. This assessment
is based on the following observations drawn from the available literature: The entire
cross section of the electrode hole is filled with metal, unlike wire insertion wherein there
will be an air gap between the electrode and the fibre. The air gap leads to non-uniform
electric fields along the fibre. Furthermore, as we are able keep the majority of the
thermal contraction longitudinal to the fibre by slowly drawing it from the furnace the
stresses can be minimised. This is not possible with the simultaneous fibre+wire draw-
ing technique. As such we designed and built an apparatus for filling optical fibres with
liquid metal3 .
3
This was greatly facilitated by the advice and guidance of Walter Margulis at ACREO, Stockholm,
Sweden
The following sections describe the theory and design of the filling apparatus and the
subsequent testing and filling results.
9.3.1 The Physics of Capillary Filling
The holes in an optical fibre for the insertion of electrodes are essentially capillaries and
therefore, to understand how they will fill with a liquid metal, we need to refer to the
theory of capillary filling as described by the Hagen-Poiseuille equation as given by:
|ΔP |πrc4
Φ= . (9.5)
8ηLc
Where Φ is the volumetric flow rate, rc is the radius of the capillary, η is the dynamic
viscosity of the liquid, Lc is the length of the capillary and ΔP is the pressure difference
between the ends of the capillary. We can calculate the time taken for a capillary of
length Lc to fill by dividing the cross sectional area of the capillary by Φ to obtain
8ηL2c
t (Lc ) = (9.6)
|ΔP |rc2
Figure 9.10 illustrates the specific example of pressure assisted filling of a capillary.
Figure 9.10: Schematic representation of pressure assisted filling of capillaries for

wetting and nonwetting liquids. Left: Pressure assisted filling a capillary with a wetting
liquid. Capillary pressure Pc , is positive and thus acts in the same direction as the
applied pressure Papp . Right: Filling a capillary with a nonwetting liquid. In this case
Pc is negative and thus works against Papp .
Chapter 9 193
In this case we can write ΔP as
ΔP = Papp − Pext − Pc (9.7)
Where Papp is the pressure applied to the pressure vessel, Pext is the external atmo-
spheric pressure and Pc is the capillary pressure which can be calculated from
2γ cos θc
Pc = (9.8)
rc
In Equation 9.8 γ is the surface tension of the liquid (Bismuth tin solder has a surface
tension of γ = 0.0319 N.m−1 ) and θc is the contact angle that the liquid makes with
the surface of the capillary.
The contact angle θc , for a liquid on a solid substrate (as illustrated in Figure 9.11) is
determined by the thermodynamic equilibrium which is established by the three phases.
The liquid phase of the droplet, the solid phase of the substrate and the gas phase
which is a mixture of the atmosphere and the equilibrium concentration of the droplets
vapour. This equilibrium is established by balancing the solid-vapour interfacial energy,
the solid-liquid interfacial energy and the liquid-vapour energy (i.e. surface tension).
Figure 9.11: Illustration of the contact angles between a liquid on a solid substrate.
Left: Wetting implies that the contact angle is less than 90◦ . Right: For contact angles
> 90◦ the liquid is Nonwetting.
Contact angles fall into one of two distinct groups: Contact angles measuring less than
90◦ indicate that the liquid is wetting (left of Figure 9.11) the surface of the substrate.
The capillary pressure in this instance will be greater than zero as cos θc > 0 for 0
θc < 90. If the contact angle is greater than 90◦ then the liquid is non-wetting for that
particular substrate (right of Figure 9.11). For 90 < θc 180 the sign of the capillary
pressure is negative which serves to expel the liquid from the capillary.
The effect of the applied pressure is to push the liquid into the capillary, against gravity
for a wetting liquid and against the negative capillary force for a non-wetting liquid.
9.3.1.1 Contact Angle Measurements
In order to be able to calculate approximate filling times and the required pressures we
measured the contact angle for the BiSn alloy on a number of glass substrates including
silica4 , bismuth borate5 and tellurite6 . This was achieved in the following way:
1. A polished, clean and flat substrate was placed onto a electric hot plate.
2. We placed a small sample of the BiSn onto the substrate.
3. The temperature of the hot plate was gradually increased until the BiSn was
completely melted and formed a into a bead.
4. We then cooled the substrate and BiSn to room temperature.
5. The bead of BiSn was removed from the substrate and glued onto a flat piece of
aluminium.
6. A saw was used to cut the BiSn+aluminium in half then the exposed edge was
polished smooth.
7. Under a microscope we imaged the region of the BiSn bead which had been in con-
tact with the glass substrate and from this image we used image analysis software
to determine the contact angle (Figure 9.12).
Images of the BiSn beads and the contact angle are shown in Figure 9.12 and the results
of these measurements are summarised in Table 9.5.
Substrate Contact Angle [◦ ]

Silica 153
Bismuth 166
Tellurite 176
Table 9.5: Measured contact angles for BiSn solder on silica, bismuth and tellurite
glass.
It is observed that the contact angle is increasing for increasing heavy metal content i.e.
silica¡bismuth¡tellurite. This is consistent with the general observation that the surface
tension of a glass decreases with increasing heavy metal content [131].
We can now apply the contact angle information with some reasonable approximations
for capillary dimensions to design a fibre filling apparatus.
4
Commercial fused silica.
5
Ahahi Glass Co. Japan.
6
Produced in house.
Chapter 9 195
Figure 9.12: Photographs of BiSn contacting silica, bismuth and tellurite substrates.
Contact angle measurements performed with graphics editing software (GIMP).
Electrode containing capillaries are typically in the order of 10 − 30 μm in diameter.

Furthermore, all measured contact angles range between ≈ 150 − 175◦ which, translates
into capillary pressures in the range −0.4 → −1.3 bar. Accordingly we require pressures
larger that this to fill the fibre. With reference to Equation 9.6 we can determine
the applied pressure required to provide reasonable filling times. If we consider a 1 m
length of fibre a reasonable filling time would be in the order of 10 minutes, under these
requirements we calculate a required applied pressure of ≈ 1bar.
9.3.2 Design and Operation of the Fibre Filling Apparatus
An apparatus for filling optical fibres was constructed, the components of which can
been seen in Figure 9.13. The apparatus consists of a brass high pressure cell with a
removable crucible for holding the molten BiSn (or any other electrode material). A
teflon gasket was fabricated to act as a high temperature seal between the pressure cell
and its lid and the fibre was sealed using a teflon plug (Figure 9.14).
In the following list we outline the procedure for loading the fibres into the filling appa-
ratus.
1. Approximately 10−15 lengths of fibres are cut at about 1 m length.

Figure 9.13: Photograph of fibre filling apparatus.
2. The lower 2 cm of the fibres had the acrylite coating stripped using a mechanical
stripping tool.
3. We then cleave the fibres leaving ≈ 3 cm of stripped fibre which was cleaned with
an organic solvent. This step is important to ensure that the end of the fibre is
free from dirt or glass to avoid obstructing the holes during the filling.
4. All of the fibres are brought together into a bundle which is then threaded through
the pressure cell lid the fibre seal and fibre seal clamp from the un-cleaved end
first so as to avoid contamination.
5. Next, we carefully tighten the fibre seal clamp until it is no longer possible to move
the fibres.
6. The lid+fibre assembly is fitted to the pressure cell which has been preheated such
that the metal for filling is molten. It is important to ensure that the fibre ends
penetrate the surface of the molten metal.
7. The entire apparatus is placed into an oven and connected to the high pressure
gas line.
Chapter 9 197
Figure 9.14: Schematic of filling apparatus showing fibre positioning and sealing
technique. Red arrows indicate the direction of forces as the fibre seal clamp is screwed
down to deform the teflon to produce a seal.
8. After approximately 1 h the apparatus should be at the same temperature as the

oven and importantly the teflon seals and remaining acrylite will now have softened
slightly. We then tighten the pressure cell lid bolts as much as possible. A small
pressure is applied to the cell such that escaping air can be heard from the fibre
seal. The fibre seal clamp is then tightened until this noise can no longer be heard.
We repeat this step gradually increasing the applied pressure until we have reached
the predetermined filling pressure.
9. The oven is then closed for the required duration for filling to occur.
10. Once the fibres have been filled the entire apparatus is cooled inside of the oven
with the pressure still applied.
9.3.2.1 Selective Filling of Optical Fibres
If we consider the additional difficulty of filling holes for internal electrodes in a MOF
it becomes immediately obvious that filling of the microstructured holes which provide
the guidance must be avoided. We therefore require a selective hole filling technique.
At present we do not have microstructured optical fibres with the necessary holes for
internal electrodes. Accordingly, to test the selective filling techniques we instead used a
twin hole silica fibre7 . We were able to develop a simple technique for selectively filling
electrode holes which should be easily applicable to a microstructured optical fibre with
internal electrodes. The following is a step by step outline of the procedure:
1. We begin by sealing the holes we wish to fill with metal. This counter intuitive step
is performed because the electrode holes are always towards the outside of the fibre
away from any microstructure. The fibre is viewed in a stereoscopic microscope.
The large working distance of these microscopes is ideal for manipulating fibres in
this way. A small amount of UV curable glue on the end of a sharp object (such as
a pin or syringe needle) is gently brought into contact with the desired hole. The
best results were achieved by gently dragging the glue down the edge of the fibre
and allowing the wetting of the glue and the capillary forces do most of the work.
Figure 9.15: Illustration of manual fibre hole blocking. Left: With the aid of a
microscope and a steadied hand we draw a glue tipped needle past the hole we wish to
block. Right: The wetting action of the glue and capillary forces server to block the
hole if performed correctly.
Capillary forces draw the glue a small distance into the hole or holes. The glue
is then cured with exposure to UV radiation. Figure 9.16 shows a twin hole fibre
that has had one hole capped in this way. The other hole is unblocked. We were
able to reliably perform this step by hand to an accuracy of ≈ ±5 μm.
2. We next connect the uncapped end into a negative pressure cell. This was con-
strued from a small glass beaker and a thick rubber membrane. A large gauge
syringe needle is pushed trough the rubber membrane then the fibre is fed through
7
Provided by Walter Margulis at ACREO, Sweden
Chapter 9 199
Figure 9.16: Photograph of an optical fibre with a manually blocked hole.
the needle until it is inside of the pressure cell. The syringe needle is then removed
from the seal leaving the seal around the fibre. A second syringe needle connected
to a syringe is pushed through the seal to provide a means of reducing the pressure
in the cell.
Figure 9.17: Photograph of an optical fibre with pressure assisted filling of UV glue.
The capped end of the fibre is immersed in a small quantity of UV glue after which
we apply suction to the pressure cell. This causes UV glue to be drawn into the
holes of the fibre that we wish block so that they will not fill with metal during the
electrode insertion process. The configuration for this filling set up is illustrated
schematically in Figure 9.18.
It is important to fill this hole significantly further than the capped hole i.e. ≈
50 mm. With the negative pressure still applied a UV lamp is used to cure the
glue. It is important to keep the negative pressure applied to avoid air bubbles
from forming in the capillary. The result of this step can be seen in Figure 9.17
where the image shows one hole that is filled only a short distance (This is the
capped hole from step 1) and the other is filled several times further.
Figure 9.18: Illustration of UV glue pressure filling set up.
3. The next step is to cleave the fibre at a point between the capped hole and the
pressure filled hole as illustrated in Figure 9.19.
Figure 9.19: Photograph of an optical fibre prepared for selective filling demonstrating
the ideal cleave position.
Figure 9.20 shows the fibre end after cleaving. The cleave is very bad due to the
presence of the glue in the electrode hole.
4. Now we are left with an open hole which we wish to fill with metal and a very well
blocked hole that we wish to keep free of metal. We filled the blocked twin hole
fibre with BiSn and Figure 9.21 shows the result after cleaving the glued end from
the fibre to leave only metal filled fibre.
Therefore, we can conclude that under the conditions of high temperature (≈

180 ◦ C) and high pressure (≈ 12 bar) the UV glue is sufficient to block the hole.
Chapter 9 201
Figure 9.20: Photograph of the cross section of an optical fibre with one hole blocked
for selective filling.
Figure 9.21: Photograph of an optical fibre with a selectively filled electrode hole.
9.4 Optical Fibre Preform Fabrication
A
first step towards producing optical fibres with internal electrodes is to produce
the optical fibre preforms. The key requirements for this preform are a central
region to act as the waveguide. This can be either a high index region, such as would
be found in a step index fibre. Alternatively, the central region can be formed from
MOF type structures (see Figure 1.1). Critical to electro-optical devices is the presence
of internal electrodes, thus requiring large holes in the preform, for later introduction of
metal.
Some methods for producing preforms with internal features, such as the ones described,
are: Direct casting, wherein a mould with the necessary shape is fabricated and the glass
is poured directly into it [132]. Glass billets can be ultrasonically drilled to produce the
necessary holes [133]. We currently use the extrusion method for creating optical fibre
preforms. This involves heating a glass billet above its transition temperature (Tg ) and
forcing it through a metal die [134], which imprints the desired structure, as shown in
Figure 9.22
NOTE:
Figure 9.22: Schematic of the optical fibre preform extrusion apparatus. This cross
section shows the essential elements of preform extrusion. A glass billet is held in a
body, this is heated above Tg and pushed through a die (arrow indicating the direction
of the force). The preform that exits the die has imprinted in it the desired structure.
Modified from [135]
The process for fabricating the fibre then involves inserting the core (high index rod
or MOF preform) into the jacket (Figure 9.23) and drawing into fibre the assembled
preform on a fibre drawing tower.
In the following Section 9.4.1 we describe experiments that were carried out to develop
the capabilities required to produce the optical fibre preforms for electro-optic fibres.
9.4.1 Fabrication of the Electrode Jacket
Our initial investigations focused on the fabrication of the electrode jacket alone, as this
is required regardless of fibre type. Our goal was to produce a preform with a small
central hole, Diameter≈ 1 mm, and two larger outer holes for > 2× the central hole (see
left of Figure 9.23).
The photographs in Figure 9.24 show an example of an electrode jacket die, both the die
insert, which produces the features in the preform and the die outer, which is responsible
Chapter 9 203
Figure 9.23: Illustration of the extruded jacket and core for creating the preform for
a step index optical fibre with internal electrodes. The electrode jacket has three holes
in order to accommodate the fibre core and the two internal electrodes. A core made
from a higher index glass is extruded and caned on a fibre drawing tower so that it fits
snugly into the core hole of the electrode jacket. This entire assembly is then drawn
into a fibre.
for determining the exterior shape of the preform. The key features are the pins that
create the holes for the core and electrodes and the many feed holes that allow the glass
to pass from the extrusion body and through the die.
Figure 9.24: Die for electrode jacket preform. Left: The components of the electrode
jacket die. The die insert produces the features internal to the preform and the die
outer produces the exterior shape. Right: Shown is the assembled die. The region
shown is where the extruded glass exits the die and the features are imprinted onto the
preform.The array of many small holes in the die insert allow glass to flow from the
extrusion body and into the die (see Figure 9.22). Pins are used to create the circular
holes for the core and electrodes.
To begin with, we chose an non-optimised design for the electrode jacket preform. These
trials were performed with the express purpose of solving fabrication related problems
inherent in the extrusion of a low rotational symmetry preform such as the electrode
jacket. Accordingly, the die design has the essential elements of the electrode jacket
without including carefully optimised dimensions. Once the fabrication issues have been
solved the optimal preform can then, in principle, be fabricated.
We experimented with various die designs in an effort to produce an electrode jacket

free from distortions. Shown in Figure 9.25 is the evolution of jacket extrusions as
modifications were made to the feed holes and die pins.
Figure 9.25: Electrode jacket preform extrusion trials. Top row: Photographs of the
extruded preforms for each trial. Second row: Feed hole configurations, where a black
circle indicates a hole blocked by a pin and red circles are blocked holes to mediate the
flow rate. Third row: The pin dimensions and distances. All dimensions in mm.
The following list details the conditions for each extrusion trial as well as a discussion of
the results from each trial. All extrusion trials were performed using a commercial lead
silicate glass (F2, Schott Glass Co.) to minimise cost. Glass billets of 30 mm diameter
and 30 mm height were used. The extrusion temperature for F2 is approximately 580◦
(Tg = 436◦ ) and the extrusions were carried out with extrusion speeds of 0.2 mm.min−1 .
Trial 1
In trial 1 we used a uniform distribution of equal size feed holes, with the exception
Chapter 9 205
of the holes around the electrode pins. These had to be smaller to account for the
diameter of these pins.
The resulting preform, shown in Figure 9.25 under Trial 1, is very distorted. There
is swelling of the regions to the left and right of the central hole. Also, the central
hole is itself distorted into an elliptical shape.
Deviations from circularity of the central hole present a significant issue. It is
critical for the central hole to remain circular as it must accept the cane. On the
other hand, distortion of the electrode holes is less important. This is because the
electrodes are introduced into the fibre as a liquid which will conform to the shape
of the holes. Additionally, during the fibre draw it is likely that the surface tension
of the molten glass will cause these holes to become circular.
Trial 2
To reduce the swelling of the left and right regions and distortion of the central
hole we decreased the number of feed holes in the left and right regions, thereby
reducing the amount of glass flowing into these regions. This is indicated by the
red holes in Figure 9.25 under Trial 2. In addition to reducing flow, we increased
the length of the central pin. We based this decision on the assumption that the
flow of the glass becomes predominantly longitudinal the further it is from the exit
of the die.
We can draw the following conclusions from preform resulting from Trial 2: The
swelling in the left and right portions of the preform appears to be unchanged by
the reduction in glass flow caused by the blocked holes. Further, the lengthened
central pin produced a circular feature.
Trial 3
Trial 3 was designed to assess the effect of the distance between central and elec-
trode pins on the preform. The central pin was kept at the same length as the
electrode pins for direct comparison with Trial 1.
The preform produced in Trial 3 (shown under Trial 3 in Figure 9.25) shows that
the distance between the central and electrode pins has been reduced, however, the
preform also displays significant swelling of the left and right regions, a distorted
central hole and electrode holes that are much smaller than the pins from which
they were formed.
9.4.2 Conclusion
It can be seen from the results of our extrusion trials that more work is required in order
to produce an electrode jacket preform free from distortions. The distortions that were
observed are due to the two-fold symmetry of the jacket. As glass flows through the
die the flow nearest the walls of the die and the pins is slowest due to friction. Thus in
regions with the fewest obstructions the flow is maximal. This results in these regions
swelling and consequently distorting the holes in the jacket. If these distortions are
translated into the optical fibre, the performance of a device based on this fibre would
be compromised.
The problem of guaranteeing a circular central hole is overcome by increasing the length
of the central pin. We were unable to mitigate the swelling the the regions to the left and
right of the central hole (see Figure 9.25). These distortions may be an inevitable conse-
quence of the two fold symmetry possessed by the die. As such, alternative fabrication
techniques, such as preform drilling etc., may need to be applied to this problem.
9.5 Preliminary Investigations of Optical Fibre Fabrica-

tion
T
o assess any potential increase in electro-optic performance offered by tellurite glass
over silica a direct comparison should be made. We therefore elected to fabricate
a fibre from tellurite glass that is otherwise as similar as possible to the state-of-the-art
silica electro-optic fibres.
At present the most efficient electro-optic fibre has been reported by Margulis et al [108].
They claim an induced χ(2) = 0.25 pm.V−1 in a silica fibre, produced via the dual anode
technique. To make a direct comparison requires fabricating an identical fibre made
from a glass with superior nonlinear properties, such as the tellurite glasses under study
in this thesis. This in turn requires us to develop fabrication techniques necessary to
fabricate a step index fibre from tellurite glass, a capability which we did not possess.
Considering our current fabrication facilities and experience we determined that an

extrusion based rod in tube method would be a logical first step. One that would require
minimal development and/or modifications of our facilities. The extrusion based rod in
tube method involves the extrusion of a tube, that will comprise the fibre cladding.
Dimension for this tube will be in the range of ≈ 10 mm outer diameter and ≈ 1 mm
inner diameter.
Following this a rod of glass with higher refractive index is extruded. This rod is extruded
with an outer diameter of approximately 10 mm. We then reduce the diameter of this
rod such that it fits into the extruded tube. This is achieved using a fibre drawing
tower, using the so called caning procedure. The cane is inserted into the tube and the
assembly which is then drawn into fibre using the fibre drawing tower.
Chapter 9 207
In order to close the small gap in between the cane and tube to produce a good interface
a vacuum is applied to the preform during the fibre drawing process.
In the following Section 9.5.1 we describe initial experiments related to the fabrication
of a tellurite step index fibre.
9.5.1 Tellurite Step Index Fibre Fabrication Experiments
For these preliminary experiments no attempt was made to produce an optimal geometry,
as in core size etc. We were simply concerned with determining the effectiveness of an
extrusion based rod in tube fabrication method using tellurite glass. From the results on
crystallisation stability (Section 5.3.1), thermal expansion (Section 5.3.2) and refractive
indices (Section 6.4) the TZN glass series was determined to be the best candidate for
this work.
To obtain the desired difference in refractive indices between core and cladding Δn, we
refer to the refractive index data for the TZN glass series (Table 6.3). If we plot the
refractive index value as a function of Zn content and perform a linear fit we obtain
an interpolation expression that predicts the refractive index for molar concentrations
of zinc ions such that 5 [Zn] 20 mol%, assuming a linear dependence. The fit in
Figure 9.26 we can see that a linear approximation is valid, the R2 value for the fit is
0.997. Using the fitted linear function:
n = −0.0043[Zn] + 2.0474, 5 [Zn] 20. (9.9)
With reference to the crystallisation stability data in Section 5.3.1 where it was noted
that the most stable composition under study was the TZN3 composition. As such, we
choose for the core and cladding glass the following two compositions respectively:
Core: 10Na2 O.12ZnO.77TeO2

Cladding: 10Na2 O.15ZnO.75TeO2
Both of these glass compositions will have excellent crystallisation stability and therefore
make ideal choices for optical fibre materials. Furthermore, the coefficients of thermal
expansion of these glasses have been determined to be decreasing, very slightly, with
increasing Zinc content (see Section 5.3.2). As such the two glasses will have closely
matched coefficients of thermal expansion whit the core having a slightly higher coeffi-
cient than the cladding. This has been deemed to be preferable as it means the cladding
is under compression resulting in a more robust fibre [60].
Figure 9.26: Linear fit to TZN refractive index data.
It was our goal to fabricate a fibre with a core of approximately 10 μm diameter core
and a 160 μm diameter cladding. We can estimate the mechanical stress that would be
present in this fibre by applying Equation 5.13. We first note that for the TZN glass
series the transition temperature Tg varies by approximately 2◦ C. Specifically, within the
compositional vicinity of [Zn]= 15 ± 2 mol.% there is essential no variation in Tg above
the uncertainty of the measurements (see Table 5.2). Thus Tg2 −Tg1 ≈ 0. Accordingly,
the second term in Equation 5.13 can be neglected. The values of the parameters used in
the calculation are: Tg = 293◦ C for the core glass, α1 = 19.3 × 10−6 ◦ C−1 (interpolated
from data in Table 5.3), α2 = 18.8 × 10−6 ◦ C−1 (measured). We assumed values for
Young’s modulus to be equal to 54.5 GPa and Poisson’s ratio to be equal to 0.253 [60].
The calculated stress is σ = −38 kPa, which is well below the tensile fracture strength
of tellurite fibres which is reported to be approximately 119 MPa [60]. Furthermore, the
negative stress is indicitive of a cladding under compression, which is considered ideal
for mechanical stability [60]. This result indicates that a mechanically stable fibre can
be fabricated.
We first extruded a tube from the cladding glass with a 10 mm outer diameter and
1 mm inner diameter. The length of the tube was cut down to 180 mm. Next, a rod
of the cladding glass was extruded which measured 10 mm in diameter. Following this
the rod was reduced in diameter in the fibre drawing tower to have a diameter equal to
that if the tube inner diameter. This process is commonly refered to as ‘caning’. We
Chapter 9 209
obtained the correct size cane by periodically selecting section of the cane during the
caning process and checking the fit with the tube. Once an appropriately sized cane was
produced, we drew > 50 m of bare fibre with a diameter of 160 μm.
We measured the optical attenuation of the fibre with the cut back method. This
measurement was made such that the lower limit on the attenuation of the eventual step
index fibre could be established. Provided we make the assumption that any additional
attenuation observed in the step index fibre is as a result of the interface between the
core and the cladding .
The measurements were made with a diode laser operating at 1550 nm with an initial
fibre length of 3 m. Plotted in Figure 9.27 are the data for the measurement. The
attenuation of the bare fibre was measured as 2.8±0.09 dB.m−1 at 1550 nm.
8
Data
Linear fit
6
Attenuation [dB]
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Total Cutback Length [m]
Figure 9.27: Cut back measurement data showing optical attenuation of the core
glass bare fibre.
We next assembled the rod in tube preform, which was then placed into the fibre drawing
tower and drawn down to a diameter of 160 μm. Initially during the draw there was a
large gap between the core and cladding. We subsequently applied a vacuum to the tube
with the pressure difference between ambient and internal ranging between 0.1 mBar and
10 mBar. This had the desired effect of closing the interface gap. However, small holes
were observed at the interface the presence and size of these holes was independent on the
strength of the applied vacuum beyond the minimum. Figure 9.28 A (top illumination)
and B (bottom illumination) show optical microscope images of the fibre where the core
is clearly visible. The core of the fibre was measured to be ≈ 11 μm in diameter.
As can be seen in Figure 9.28 B the fibre is guiding the light from the microscope and
there is a dark ring around the core. Further investigations showed that this dark ring
results from microscopic holes at the interface, as seen in the electron microscope images
in Figure 9.28 C and D. These holes are present along the entire length of fibre and are
randomly distributed. It is unclear to us what the small structure observed within the
interface hole is (shown in Figure 9.28).
Figure 9.28: Microscopy images of tellurite step index fibre. A) Illumination from
above showing fibre core as a dark circle. B) Illumination from beneath, core shown as
bright circle. C) Electron microscope image of the fibre core showing interface holes.
D) Electron microscope image of interface hole.
6
Data
Linear fit
5
4
Attenuation [dB]
0.0 0.1 0.2 0.3 0.4 0.5
Total Cutback Length [m]
Figure 9.29: Cut back measurement data showing optical attenuation of the tellurite
step index fibre
Chapter 9 211
Using the cutback method we measured the optical attenuation of the step index fibre,
see Figure 9.29. We determine the loss to be 11.3±2 dB.m−1 at 1550 nm.
Subtracting the loss of the core material alone we obtain 8.5 dB.m−1 , which we attribute
to the interface imperfections. This compares unfavourably with other reports of core
clad fibres fabricated via extrusion based rod in tube methods. For example Lousteau
et al claim 6.8 dB.m−1 at 1550 nm [136]. However, the optical attenuation will depend
heavily on core size and refractive index contrast (i.e. mode confinement) and therefore
it is difficult to make direct comparisons. Future work will involve experiments designed
to minimise the interface imperfections and hence the loss.
9.6 Conclusion
T
he qualitative analysis provided by the computational models for electrode induced
loss and the DC electric field across the core show us that for a efficient device
we require a well confined optical mode, as far as possible from the electrode in order
to minimise the loss. The strength of the DC field reduces linearly with separation and
can therefore be maintained at a high value (i.e. 108 V.m−1 ) while minimising the loss
which decreases exponentially over the distance of the wavelength of the propagating
light.
We have made good progress towards the ability to fabricate optical fibres with internal
electrodes. The technical challenge of selectively filling the fibres with metal electrodes
has been overcome and much progress towards fabricating a suitable fibre preform has
been made.
Finally, based on the material characterisations made in the previous chapters, we have
identified a pair of tellurite glass compositions that can be combined to produce a pro-
totype fibre. Steps toward the fabrication of a step index fibre made from these glasses
have been made and the key issue of interface imperfections was identified.
Chapter 10
Concluding Remarks
10.1 Conclusion of Thesis Findings and Results
W
e conclude this thesis by addressing the research goals stated in Chapter 1. Future
research in this area will benefit from an analysis of the factors, if any, that
impeded progress and/or the realisation of the stated goals.
The overall goal for this work was the fabrication of a prototype electro-optical fibre
modulator. This goal was not achieved. The contributing factors that prevented the
realisation of this goal can be separated into two types of experimental difficulties:
those that were unavoidable and those resulting from flawed initial assumptions. The
first category includes things such as: unreliable supply of raw materials that slowed
progress in the glass fabrication and malfunctioning equipment.
The second category of experimental difficulties represent serious issues that will need to
be overcome if work in this area is to develop further. The most fundamental error that
was made was not recognising that certain glasses can not be thermally poled via the
charge migration method (see Section 7.2). Our initial assumption that a tellurite glass
containing sodium ions should be able to be thermally poled by the charge migration
method was based on the evidence from silica poling wherein the sodium enables the
creation of a permanent χ(2) . It is now evident that the situation if more complicated
than this.
A further difficulty that eventuated due to an incorrect assumption was the inability to
extrude an optical fibre preform with the necessary features for the subsequent intro-
duction of electrodes into the fibre. We encountered large distortions in the obtained
preforms which appear to be endemic to the low rotational symmetry of the geometry.
Indeed, in a private communication with collaborator Prof. Darren Crowdy at Imperial
College London, mathematical models of the glass flow have indicated that this type of
213
214 Concluding Remarks
geometry may always suffer from these distortions [137]. Presently, more investigations
are being undertaken to better understand the process.
Despite not achieving the final goal set at the outset of this work much progress was
made toward this goal. In particular, the sub-goals outlined in Chapter 1 are reiterated
below with comments made as to the various achievements for each:
1. To develop a large range of tellurite glasses and characterise fully their properties
so that informed decisions can be made regarding the choice of glass. Adding to the
collective understanding of this glass system and the origin of its properties this
knowledge can be applied to the design of devices and new glasses.
We produced several families of tellurite glass within each compositional variations
were explored. Of the glass families that were studied (TMN, TZN and TBN)
the TZN glass family stands out as being the most suitable candidate for optical
fibre fabrication. Its high crystallisation stability makes it a suitable materiel for
the processing required for optical fibre fabrication. The small variation in its
coefficient of thermal expansion over the range of compositions makes it an ideal
candidate for core/clad optical fibres. Furthermore, the liner variation in refractive
index with composition enables a wide range of index contrasts. Significantly, for
electro-optical applications, the nonlinear refractive indices of these glasses are in
excess of > 20× that of silica, the current material of choice for electro-optical
fibre devices.
2. To identify suitable tellurite glasses for nonlinear optical fibre devices.

The suitability of the TZN glasses was established via the fabrication of prototyp-
ical optical fibres.
3. To generate permanent second order nonlinearities in some of these glasses via

thermal poling, thus making possible efficient electro-optical devices for controlling
the passage of light in optical systems.
Thermal poling experiments were conducted on selected compositions. The com-
plexity of the process was underestimated and progress was limited. However, fu-
ture investigations into this area will be benefited by the knowledge gained herein.
Significant progress was made in the area of characterisation of the second order
nonlinear properties of thermally poled glass. The Maker fringes experiment that
was constructed had very high sensitivity due, mainly, to the novel data acquisition
system developed. Furthermore, the analysis techniques for the experimental data
were analysed in detail and improvements were made. In particular in terms of
the application of genetic algorithms to finding optimal data fits and therefore
measured parameters.
Chapter 10 215
4. To design and fabricate prototype electro-optical devices from in house fabricated

optical fibres.
Progress was made towards the design and fabrication of electro-optical devices.
That being so, we have determined useful information regarding influence of fibre
structure on the electric fields produced between the internal electrodes. These
results will compliment future work wherein the thermal poling of microstructured
optical fibres is explored.
Additionally, the attenuation produced by internal electrodes was studied via com-
puter models. Results indicate the the confinement of the optical mode has a large
influence on the strength of the attenuation. Certain mircostructured fibres such
as the wagon when fibre offer extreme confinement and thus, potentially, very low
loss alternatives to current step index designs.
Fabrication techniques such as selective electrode insertion and electrode jacket

preform extrusion were also explored. A sound technique for selectively filling
electrode holes within a microstructured fibre was developed. Thereby providing
one of the key elements required for fabricating electro-optical fibre devices.
10.2 Future Work
F
uture work towards the fabrication of electro-optical fibre devices will focus on the
refining the processing of the TZN glasses that were identified as the most suitable
candidates for this work. In addition to this, It will be useful to explore tellurite glass
compositions that do not contain sodium as a constituent. Following from the results of
the thermal poling experiments, glasses that have only trace amounts of mobile species
may indeed be preferable for poling. In this way the glasses are expected to behave
somewhat more like silica during poling.
Our fabrication trials for the preform jacket all resulted in significant deviations from
the target preform shape. Further extrusion trials will be necessary to determine if this
can be overcome, or indeed, if an alternative preform fabrication method is required.
Finally, our preliminary step index tellurite fibre experiments, although successful in
creating a fibre that guides light, resulted imperfect interfaces which produced unaccept-
ably high optical losses. More work is required to refine this process and/or determine
whether or not the extruded rod in tube method has any merit.
Bibliography
[1] J. C. Knight, T. A. Birks, P. St. J. Russell, and D. M. Atkin. All-silica single-

mode optical fiber with photonic crystal cladding. Opt. Lett., 21(19):1547–1549,
Oct 1996. doi: 10.1364/OL.21.001547. URL http://ol.osa.org/abstract.cfm?
URI=ol-21-19-1547.
[2] J. C. Knight, T. A. Birks, P. St.J. Russell, and D. M. Atkin. All-silica single-mode

optical fiber with photonic crystal cladding: errata. Opt. Lett., 22(7):484–485, Apr
1997. doi: 10.1364/OL.22.000484. URL http://ol.osa.org/abstract.cfm?URI=
ol-22-7-484.
[3] T. A. Birks, J. C. Knight, and P. St.J. Russell. Endlessly single-mode photonic

crystal fiber. Opt. Lett., 22(13):961–963, Jul 1997. doi: 10.1364/OL.22.000961.
URL http://ol.osa.org/abstract.cfm?URI=ol-22-13-961.
[4] P. Russell. Photonic crystal fibers. Science, 299(5605):358–362, 2003. doi: 10.
1126/science.1079280. URL http://www.sciencemag.org/content/299/5605/
358.abstract.
[5] J. C. Knight. Photonic crystal fibres. Nature, 424:847–851, 2003.
[6] T. M. Monro and H. Ebendorff-Heidepriem. Progress in microstructured optical

fibers. Annual Review of Materials Research, 36(1):467–495, 2006. doi: 10.1146/
annurev.matsci.36.111904.135316. URL http://www.annualreviews.org/doi/
abs/10.1146/annurev.matsci.36.111904.135316.
[7] J P Goure, I Verrier, and J P Meunier. Linear and nonlinear optical fibre devices.
Journal of Physics D: Applied Physics, 22(12):1791, 1989. URL http://stacks.
iop.org/0022-3727/22/i=12/a=001.
[8] D.A. Pinnow, T.C. Rich, Jr. F.W. Ostermayer, and Jr. M. DiDomenico. Funda-
mental optical attenuation limits in the liquid and glassy state with application to
fiber optical waveguide materials. Applied Physics Letters, 22(10):527–529, 1973.
doi: 10.1063/1.1654495. URL http://link.aip.org/link/?APL/22/527/1.
[9] URL http://www.corning.com/WorkArea/showcontent.aspx?id=27659.
217
218 Bibliography
[10] HEBO Spezialglas. Hebo spezialglas datasheet. URL http://www.hebo-glass.

com/en/quarz.
[11] J. S. Wang, E. M. Vogel, and E. Snitzer. Tellurite glass: a new candidate for fiber
devices. Optical Materials, 3(3):187 – 203, 1994. ISSN 0925-3467. doi: DOI:10.
1016/0925-3467(94)90004-3. URL http://www.sciencedirect.com/science/
article/B6TXP-46JYHGM-2V/2/4d9156322a0462f8ae3d1f4232c52eef.
[12] URL http://www.scopus.com.
[13] J. T. Berzelius. Ann. Phys. Chem., 32:577, 1834.
[14] J. E. Stanworth. Tellurite glasses. Nature, 169:581–582, 1952.
[15] E. F. Lambson, G. A. Saunders, B. Bridge, and R. A. El-Mallawany. The elas-

tic behaviour of TeO2 glass under uniaxial and hydrostatic pressure. Journal of
Non-Crystalline Solids, 69(1):117 – 133, 1984. ISSN 0022-3093. doi: DOI:10.
article/B6TXM-48KDH5W-12F/2/6ae144b503340c1adfa4bb1a00bc295b.
[16] Van Uitert L.G. Grodkiewicz-W.H. Singh, S. Laser spectroscopic properties

of Nd3+ -doped tellurite glasses. Optics Communications, 17(3):315–319, 1976.
ISSN 00304018. URL http://www.scopus.com/inward/record.url?eid=2-s2.
0-0016967875&partnerID=40&md5=5f1a48c92b0613d46d9648454a08f6eb. cited
By (since 1996) 2.
[17] Ammar A.A. El-Mallwany R.A. El-Zaidia, M.M. Infra-red spectra, electron
spin resonance spectra, and density of (TeO2 )100−x -(WO3 )x and (TeO2 )100−x -
(ZnCl2 )x glasses. Physica Status Solidi (A) Applied Research, 91(2):637–
642, 1985. URL http://www.scopus.com/inward/record.url?eid=2-s2.
0-0022145564&partnerID=40&md5=901dd1599e87f0422ae6c0986b423e92. cited
By (since 1996) 6.
[18] Raouf A. H. El-Mallawany. Tellurite Glasses Handbook Physical Properites and

Data. CRC PRess LLC, 2002.
[19] U. Österberg and W. Margulis. Dye laser pumped by nd:yag laser pulses frequency
doubled in a glass optical fiber. Opt. Lett., 11(8):516–518, Aug 1986. doi: 10.1364/
OL.11.000516. URL http://ol.osa.org/abstract.cfm?URI=ol-11-8-516.
[20] R. A. Myers, N. Mukherjee, and S. R. J. Brueck. Large second-order nonlinearity

in poled fused silica. Optics Letters, 16(22):1732–1734, November 1991.
Bibliography 219
[21] Evgenii M Dianov, P G Kazansky, and D Yu Stepanov. Problem of the photoin-

duced second harmonic generation in optical fibers. Soviet Journal of Quantum
Electronics, 19(5):575, 1989. URL http://stacks.iop.org/0049-1748/19/i=5/
a=L01.
[22] Evgenii M. Dianov, P. G. Kazansky, and D. Yu Stepanov. Mechanism of effi-

cient photoinduced second harmonic generation in fiber waveguides. Soviet Jour-
nal of Quantum Electronics, 20(7):849, 1990. URL http://stacks.iop.org/
0049-1748/20/i=7/a=A39.
[23] P.G. Kazansky, P.St.J. Russell, and H. Takebe. Glass fiber poling and applications.
Lightwave Technology, Journal of, 15(8):1484 –1493, aug 1997. ISSN 0733-8724.
doi: 10.1109/50.618381.
[24] O. Deparis, F.P. Mezzapesa, C. Corbari, P.G. Kazansky, and K. Sakaguchi.

Origin and enhancement of the second-order non-linear optical susceptibil-
ity induced in bismuth borate glasses by thermal poling. Journal of Non-
Crystalline Solids, 351(27-29):2166 – 2177, 2005. ISSN 0022-3093. doi: DOI:10.
1016/j.jnoncrysol.2005.06.004. URL http://www.sciencedirect.com/science/
article/B6TXM-4GJVBHJ-3/2/fde3074f52e6d60d1b10b6e7d095bf8e.
[25] W. Margulis, M. Ferraris, M. Kristensen, P. Kazansky, B. Ortega, D. Pastor,

V. Pruneri, B. Poumellec, and G. Martinelli. GLAMOROUS final report. Technical
report, GLAMOROUS Consortium, 2005.
[26] A. Belwalkar, W. Z. Misiolek, and J. Toulouse. Viscosity study of the optical tellu-
rite glass: 75TeO2 -20ZnO-5Na2 O. Journal of Non-Crystalline Solids, 356(25-27):
1354 – 1358, 2010. ISSN 0022-3093. doi: DOI:10.1016/j.jnoncrysol.2010.03.011.
URL http://www.sciencedirect.com/science/article/B6TXM-503HRR2-1/2/
12b99a487ded87c372b40d3d699fdab6.
[27] T. H. Maiman. Stimulated optical radiation in ruby. Nature, 187:493–494, 1960.

doi: 10.1038/187493a0.
[28] P. A. Franken, A. E. Hill, C. W. Peters, and G. Weinreich. Generation of optical

harmonics. Phys. Rev. Lett., 7(4):118–119, Aug 1961. doi: 10.1103/PhysRevLett.
7.118.
[29] Richard L. Sutherland. Handbook Of Nonlinear Optics. Marcel Dekker, inc., 1996.
[30] Robert W. Boyd. Nonlinear Optics. Academic Press Inc., 3 edition, 2008.
[31] N. Myrén. Poled Fiber Devices. Phd, Royal Institute of Technology, 2005.
220 Bibliography
[32] W. H. Zachariasen. The atomic arrangement in glass. Journal of the American

Chemical Society, 54:42–43, 1932.
[33] P. Badrinarayanan, W. Zheng, Q. Li, and S. L. Simon. The glass transition temper-
ature versus the fictive temperature. Journal of Non-Crystalline Solids, 353(26):
2603 – 2612, 2007. ISSN 0022-3093. doi: DOI:10.1016/j.jnoncrysol.2007.04.025.
URL http://www.sciencedirect.com/science/article/B6TXM-4NY4WMN-1/2/
bd041ece53c058656ca42ff89f140e0d.
[34] M. Yamane and Y. Asahara. Glasses for photonics. Cambridge University Press,
2000.
[35] A. Dietzel. Z. Electrochem., 48:9 – 23, 1942.
[36] H. Burger, K. Kneipp, H Hobert, W. Vogel, V. Kozhukharov, and S. Neov. Glass

formation, properties and structure of glasses in the TeO2 -ZnO system. Journal
of Non-Crystalline Solids, 151:134–142, 1992.
[37] T. Sekiya, N. Mochida, A. Ohtsuka, and M. Tonokawa. Normal vibrations of two

polymorphic forms of TeO2 crystals and assignment of raman peaks of pure TeO2
glass. Nippon Seramikkusu Kyokai Gakujutsu Ronbunshi, 97(12):1435–1440, 1989.
[38] S. Fujino and K. Morinaga. Material dispersion and its compositional pa-
rameter of oxide glasses. Journal of Non-Crystalline Solids, 222:316 –
320, 1997. ISSN 0022-3093. doi: DOI:10.1016/S0022-3093(97)90130-7.
URL http://www.sciencedirect.com/science/article/B6TXM-49V0B2G-1G/
2/2868b536f8a8dead577791a2ab2e938a. Practical Implications of Glass Struc-
ture.
[39] J. C. McLaughlin, S. L. Tagg, J. W. Zwanziger, D. R. Haeffner, and S. D.

Shastri. The structure of tellurite glass: a combined NMR, neutron diffrac-
tion, and X-ray diffraction study. Journal of Non-Crystalline Solids, 274(1-
3):1 – 8, 2000. ISSN 0022-3093. doi: DOI:10.1016/S0022-3093(00)00199-X.
URL http://www.sciencedirect.com/science/article/B6TXM-4117SWB-1/2/
ad49bc4f5e8fdfeec0d413e36c15dbcc. Physics of Non-Crystalline Solids 9.
[40] Tessman, Jack R., Kahn, A. H., and William Shockley. Electronic polarizabilities
of ions in crystals. Phys. Rev., 92(4):890–895, Nov 1953. doi: 10.1103/PhysRev.
92.890.
[41] Arun K. Varshneya. Fundamentals of Inorganic Glasses. Academic Press Inc.,

1994.
Bibliography 221
[42] E. Fargin, A. Berthereau, T. Cardinal, G. Le Flem, L. Ducasse, L. Canioni,

P. Segonds, L. Sarger, and A. Ducasse. Optical non-linearity in oxide glasses. Jour-
nal of Non-Crystalline Solids, 203:96 – 101, 1996. ISSN 0022-3093. doi: DOI:10.
article/B6TXM-3VSCHWM-13/2/67ee63ff2b449d6df232c0f73d27d828. Optical
and Electrical Propertias of Glasses.
[43] A. S. Pine and G. Dresselhaus. Raman scattering in paratellurite, TeO2 . Phys.

Rev. B, 5(10):4087–4093, May 1972. doi: 10.1103/PhysRevB.5.4087.
[44] Y. Himei, A. Osaka, T. Nanba, and Y. Miura. Coordination change of

Te atoms in binary tellurite glasses. Journal of Non-Crystalline Solids, 177:
164 – 169, 1994. ISSN 0022-3093. doi: DOI:10.1016/0022-3093(94)90526-6.
URL http://www.sciencedirect.com/science/article/B6TXM-48CX6DS-PK/
2/6445c004500013d018c6c283f3c1697a. First PAC RIM Meeting on Glass and
Optical Materials.
[45] T. Komatsu, H. G. Kim, and H. Mohri. Raman scattering study on local structures
of Te4+ and Nb5+ in LiNbO3 -TeO2 glasses. Journal of Materials Science Let-
ters, 15:2026–2029, 1996. ISSN 0261-8028. URL http://dx.doi.org/10.1007/
BF00278611. 10.1007/BF00278611.
[46] V. K. Malinovsky and A. P. Sokolov. The nature of boson peak in

Raman scattering in glasses. Solid State Communications, 57(9):757 –
761, 1986. ISSN 0038-1098. doi: DOI:10.1016/0038-1098(86)90854-9.
URL http://www.sciencedirect.com/science/article/B6TVW-46TYWHV-2KN/
2/5866f2b21ac630f04281036441329059.
[47] C. Duverger, F. Romain, S. Khatir, M. Bouazaoui, and S. Turrell. Low-frequency

raman spectra and structural correlations in metal oxide glasses. Journal of
Molecular Structure, 410-411:285 – 289, 1997. ISSN 0022-2860. doi: DOI:10.
1016/S0022-2860(96)09574-9. URL http://www.sciencedirect.com/science/
article/B6TGS-49YDV9C-2B/2/5d03e79140ead4cd929d80cb30a2cc18. Molecu-
lar Spectroscopy and Molecular Structure 1996.
[48] E Stavrou, C Tsiantos, R D Tsopouridou, S Kripotou, A G Kontos, C Raptis,

B Capoen, M Bouazaoui, S Turrell, and S Khatir. Raman scattering boson peak
and differential scanning calorimetry studies of the glass transition in tellurium-
zinc oxide glasses. Journal of Physics: Condensed Matter, 22(19):195103, 2010.
URL http://stacks.iop.org/0953-8984/22/i=19/a=195103.
222 Bibliography
[49] B. Szigeti. Compressibility and absorption frequency of ionic crystals. Pro-

ceedings of the Royal Society of London. Series A. Mathematical and Physi-
cal Sciences, 204(1076):51–62, 1950. doi: 10.1098/rspa.1950.0161. URL http:
//rspa.royalsocietypublishing.org/content/204/1076/51.abstract.
[50] Archimedes. The Works of Archimedes with the Method of Archimedes. Dover
Publications, 2002.
[51] CRC Press. CRC Handbook of Chemistry and Physics. CRC Press, 70 edition,
1990.
[52] A. Hrub. Evaluation of glass-forming tendency by means of DTA. Czechoslovak

Journal of Physics, 22(11):1187–1193, 1972. doi: 10.1007/BF01690134. URL
http://www.springerlink.com/content/gg8532r863217151.
[53] El-Mallawany. Devitrification and vitrification of tellurite glasses. J. Mater. Sci.,

6:1–3, 1995. doi: 10.1007/BF00208125.
[54] S. Sakka and J. D. Mackenzie. Relation between apparent glass transition

temperature and liquids temperature for inorganic glasses. Journal of Non-
Crystalline Solids, 6(2):145 – 162, 1971. ISSN 0022-3093. doi: DOI:10.
article/B6TXM-48CXSBC-2C3/2/9e02b70c2d6af782cca016fe202bb2b1.
[55] Deepika and Narendra S. Saxena. Thermodynamics of glass/crystal transformation

in Se58 Ge42−x Pbx (9 ≤ x ≤ 20) glasses. The Journal of Physical Chemistry B, 114
(1):28–35, January 2010. ISSN 1520-6106. URL http://dx.doi.org/10.1021/
jp9066845.
[56] U. C. Paek and C. R. Kurkjian. Calculation of cooling rate and induced stresses in
drawing of optical fibers. Journal of the American Ceramic Society, 58(7-8):330–
335, 1975. ISSN 1551-2916. URL http://dx.doi.org/10.1111/j.1151-2916.
1975.tb11490.x.
[57] D. A. Krohn and A. R. Cooper. Strengthening of glass fibers: I, cladding. Jour-

nal of the American Ceramic Society, 52(12):661–664, 1969. ISSN 1551-2916.
doi: 10.1111/j.1151-2916.1969.tb16072.x. URL http://dx.doi.org/10.1111/j.
1151-2916.1969.tb16072.x.
[58] J. Wang, J. R. Lincoln, W. S. Brocklesby, R. S. Deol, C. J. Mackechnie, A. Pearson,

A. C. Tropper, D. C. Hanna, and D. N. Payne. Fabrication and optical properties of
lead-germanate glasses and a new class of optical fibers doped with tm3+ . Journal
of Applied Physics, 73(12):8066–8075, 1993. doi: 10.1063/1.353922. URL http:
//link.aip.org/link/?JAP/73/8066/1.
Bibliography 223
[59] VITRON Spezialwerkstoffe GmbH. Infrared chalcogenide glass ig5, December

2009.
[60] M. D. O’Donnell, K. Richardson, R. Stolen, A. B. Seddon, D. Furniss, V. K.

Tikhomirov, C. Rivero, M. Ramme, R. Stegeman, G. Stegeman, M. Couzi, and
T. Cardinal. Tellurite and fluorotellurite glasses for fiberoptic raman amplifiers:
Glass characterization, optical properties, raman gain, preliminary fiberization,
and fiber characterization. Journal of the American Ceramic Society, 90(5):1448–
1457, 2007. URL http://dx.doi.org/10.1111/j.1551-2916.2007.01574.x.
[61] H. Li1, J. Lousteau, W. N. MacPherson, X. Jiang, H. T. Bookey, J. S. Barton1,

A. Jha, and A. K. Kar. Thermal sensitivity of tellurite and germanate optical
fibers. Optics Express, 15:8857–8863, 2007.
[62] A. Lin, A. Zhang, E. J. Bushong, and J. Toulouse. Solid-core tellurite glass fiber
for infrared and nonlinear applications. Opt. Express, 17(19):16716–16721, 2009.
URL http://www.opticsexpress.org/abstract.cfm?URI=oe-17-19-16716.
[63] A. P. Mirgorodsky, M. Soulis, P. Thomas, T. Merle-Mjean, and M. Smirnov. Ab

initio study of the nonlinear optical susceptibility of TeO2 -based glasses. PHYSI-
CAL REVIEW B, 73:1–13, 2006. doi: 10.1103/PhysRevB.73.134206.
[64] W. Sellmeier. Zur erklarung der abnormen fabenfolge im spectrum einiger sub-
stanzen. Annalen der Physik und Chemie, 219:272–282, 1871.
[65] E. A. Davis and N. F. Mott. Conduction in non-crystalline systems v. conductivity,

optical absorption and photoconductivity in amorphous semiconductors. Phil.
Mag., 22:903–922, 1970.
[66] S. K. J. Al-Ani, C. A. Hogarth, and R. A. El-Malawany. A study of optical

absorption in tellurite and tungsten-tellurite glasses. Journal of Material Science,
20:661–667, 1985.
[67] D. Stentz, H. B. George, S. E. Feller, and M. Affatigato. Comparison of the optical

cutoffs of bismuth borate and bismuth germanate glasses. Phys. Chem. Glasses,
41:406–408, 2000.
[68] G. Vijaya Prakash, D. Narayana Rao, and A. K.Bhatnagar. Linear optical

properites of niobium-based tellurite glasses. Solid State Communications, 119:
39–44, 2001.
[69] G. S. Murugan and Y. Ohishi. TeO2 -BaO-SrO-Nb2 O5 : glasses a new glass system
for waveguide devices applications. Journal of Non-Crystalline Solids, 341:86–92,
2004.
224 Bibliography
[70] Halimah M.K., Daud W.M., Sidek H.A. A., Zainal A.T., Zainul H., and Jumiah
Hassan. Optical properties of borotellurite glasses. American Journal of Applied
Sciences, (Special Issue):63–66, 2005.
[71] J. Ozdanova, H. Ticha, and L. Tichy. Remark on the optical gap in ZnOBi2 O3 TeO2
glasses. Journal of Non-Crystalline Solids, 353:2799–2802, 2007.
[72] M. A. HASSAN, W. I. KHLEIF, and C. A. HOGARTH. A comparison of the

optical properties of glass and of evaporated amorphous thin films of Bao-TeO2 .
JOURNAL OF MATERIALS SCIENCE, 24:1607–1611, 1989.
[73] M. Arnaudov, V. Dimitrov, Y. Dimitriev, and L. Markova. Infrared-

spectral investigation of tellurites. Materials Research Bulletin, 17(9):1121
– 1129, 1982. ISSN 0025-5408. doi: DOI:10.1016/0025-5408(82)90059-9.
URL http://www.sciencedirect.com/science/article/B6TXC-48FM86J-7F/
2/0c7a5b469c77a0297a8d6146a17d1523.
[74] X. Feng, S. Tanabe, and T. Hanada. Hydroxyl groups in erbium-doped

germanotellurite glasses. Journal of Non-Crystalline Solids, 281(1-3):48 –
54, 2001. ISSN 0022-3093. doi: DOI:10.1016/S0022-3093(00)00429-4.
URL http://www.sciencedirect.com/science/article/B6TXM-42JRD0W-6/2/
10ce01e32c90a5eea4028446b114db52.
[75] H. Onodera, I. Awai, and J. Ikenoue. Refractive-index measurement of bulk ma-

terials: prism coupling method. Applied Optics, 22(8):1194–1197, April 1983.
[76] Metricon Data Sheet.
[77] M. Samoc, A. Samoc, B. L. Davies, Z. Bao, L. Yu, B. Hsieh, and U. Scherf.

Femtosecond Z-scan and degenerate four-wave mixing measurements of real and
imaginary parts of the third-order nonlinearity of soluble conjugated polymers. J.
Opt. Soc. Am. B, 15(2):817–825, 1998. URL http://josab.osa.org/abstract.
cfm?URI=josab-15-2-817.
[78] H. G. Chatellus and E. Freysz. Measurement of the third-order susceptibility of

glasses by EFISH of femtosecond pulses. Opt. Express, 9(11):586–591, 2001. URL
http://www.opticsexpress.org/abstract.cfm?URI=oe-9-11-586.
[79] F. D’Amore, A. Zappettini, G. Facchini, S. M. Pietralunga, M. Martinelli,

C. Dell’Erba, C. Cuniberti, D. Comoretto, and G. Dellepiane. Third order op-
tical characterisation of a π-conjugated polydiacetylene by Maker fringes tech-
nique. Synthetic Metals, 127(1-3):143 – 146, 2002. ISSN 0379-6779. doi:
DOI:10.1016/S0379-6779(01)00605-1. URL http://www.sciencedirect.com/
science/article/B6TY7-44JD5KT-B/2/93c3d9db8438f3ae63a88fee7df36cb6.
Bibliography 225
[80] M. Sheik-Bahae, D.C. Hutchings, D.J. Hagan, and E.W. Van Stryland. Disper-
sion of bound electron nonlinear refraction in solids. Quantum Electronics, IEEE
Journal of, 27(6):1296 –1309, June 1991. ISSN 0018-9197. doi: 10.1109/3.89946.
[81] APPLICATION NOTE Z-Scan for the Characterization of Transparent Optical

Materials. Newport Corporation, Worldwide Headquarters 1791 Deere Avenue
Irvine, CA 92606, 2007.
[82] M. Sheik-Bahae, A. A. Said, T. H. Wei, D. J. Hagan, and E. W. Van Stryland.

Sensitive measurement of optical nonlinearities using a single beam. IEEE Journal
of Quantum Electronics, 26(4):760–769, April 1990.
[83] A single-mode, high index-contrast, lead silicate glass fibre with high nonlinearity,
broadband near-zero dispersion at telecommunication wavelengths, sept. 2010. doi:
10.1109/ECOC.2010.5621359.
[84] R. C. Miller. Optical second harmonic generation in piezoelectric crystals. Applied

Physics Letters, 5(1):17–19, 1964. doi: 10.1063/1.1754022. URL http://link.
aip.org/link/?APL/5/17/1.
[85] Charles C. Wang. Empirical relation between the linear and the third-order non-
linear optical susceptibilities. Phys. Rev. B, 2:2045–2048, Sep 1970. doi: 10.1103/
PhysRevB.2.2045. URL http://link.aps.org/doi/10.1103/PhysRevB.2.2045.
[86] H. An and S. Fleming. Second-order optical nonlinearity in thermally

poled pyrex borosilicate glass. Optics Communications, 281(5):1263 –
1267, 2008. ISSN 0030-4018. doi: DOI:10.1016/j.optcom.2007.10.098.
URL http://www.sciencedirect.com/science/article/B6TVF-4R5P5S3-1/2/
693738dc3c36e07a2864664766b3d1b4.
[87] B. Ferreira, E. Fargin, J.P. Manaud, G. Le Flem, V. Rodriguez, and T. Buffeteau.

Second harmonic generation induced by poling in borophosphate bulk and thin
film glasses. Journal of non-crystalline solids, 343:121–130, 2004.
[88] M. Dussauze, . Fargin, M. Lahaye, V. Rodriguez, and F. Adamietz. Large second-

harmonic generation of thermally poled sodium borophosphate glasses. Opt. Ex-
press, 13(11):4064–4069, May 2005.
[89] Y. Xia, Z. Xua, Z. Houa, L. Liua, L. Xua, W. Wanga, M. Affatigatob, and

S. Fellerb. Second-order optical nonlinearity in bulk PbO/B2 O3 glass. Optics
Communications, 210:367–373, 2002.
[90] M. Guignarda, V. Nazabala, F. Smektalaa H. Zeghlacheb, A. Kudlinskib,

Y. Quiquempoisb, and G. Martinellib. High second-order nonlinear susceptibility
226 Bibliography
induced in chalcogenide glasses by thermal poling. Opt. Express, 14:1524–1532,

2006.
[91] K. Tanaka, K. Kashima, K. Hirao, N. Soga, A. Mito, and H. Hasu. Second har-
monic generation in electrically poled Li2 O-Nb2 O5 -TeO2 glasses. Journal of Non-
Crystalline Solids, 185, 1995.
[92] Infrasil 301, 302 and 303 data sheet. URL www.optosolutions.com/doc/
Infrasil.pdf.
[93] K. Tanaka, A. Narazaki, K. Hirao, and N. Soga. Optical second harmonic

generation in poled MgO-ZnO-TeO2 and B2 O3 -TeO2 glasses. Journal of
Non-Crystalline Solids, 203:49 – 54, 1996. ISSN 0022-3093. doi: DOI:10.
1016/0022-3093(96)00333-X. URL http://www.sciencedirect.com/science/
article/B6TXM-3VSCHWM-V/2/16a8761fb2b8e93fbb2df4c7fbfff9cb. Optical
and Electrical Propertias of Glasses.
[94] A. Narazaki, K. Tanaka, K. Hirao, and N. Soga. Effect of poling temperature

on optical second harmonic intensity of sodium zinc tellurite glasses. Journal
of Applied Physics, 83(8):3986–3990, 1998. doi: 10.1063/1.367154. URL http:
//link.aip.org/link/?JAP/83/3986/1.
[95] Y. Yonesaki, K. Tanaka, A. Narazaki, J. Si, and K. Hirao. Relaxation phenomena

in second-order nonlinearity of thermally and optically poled Nb2 O5 -TeO2 glasses.
Journal of Physics D: Applied Physics, 35(16):2026, 2002. URL http://stacks.
iop.org/0022-3727/35/i=16/a=318.
[96] B. Ferreira, E. Fargin, B. Guillaume, G. Le Flem, V. Rodriguez, M. Couzi, T. Buf-

feteau, L. Canioni, L. Sarger, G. Martinelli, Y. Quiquempois, H. Zeghlache, and
L. Carpentier. Second harmonic generation in poled tellurite glass. Journal of
Non-Crystalline Solids, 332:207–218, 2003.
[97] G. Senthil Murugan, E. Fargin, V. Rodriguez, F. Adamietz, M. Couzi,

T. Buffeteau, and P. Le Coustumer. Temperature-assisted electrical poling of
TeO2 Bi2 O3 ZnO glasses for non-linear optical applications. Journal of Non-
Crystalline Solids, 344:158–166, 2004.
[98] Y. Quiquempois, G. Martinelli, P. Dutherage, P. Bernage, P. Niay, and M. Douay.

Localisation of the induced second-order non-linearity within infrasil and suprasil
thermally poled glasses. Optics Communications, 176:479–487, 2000.
[99] A. L. C. Triques, I. C. S. Carvalho, M. F. Moreira, H. R. Carvalho, R. Fischer,

B. Lesche, and W. Margulis. Time evolution of depletion region in poled silica.
Bibliography 227
Applied Physics Letters, 82(18):2948–2950, 2003. doi: 10.1063/1.1572472. URL

http://link.aip.org/link/?APL/82/2948/1.
[100] Y. Quiquempois, A. Kudlinski, and G. Martinelli. Zero-potential condition in

thermally poled silica samples: evidence of a negative electric field outside the
depletion layer. J. Opt. Soc. Am. B, 22(3):598–604, 2005. URL http://josab.
osa.org/abstract.cfm?URI=josab-22-3-598.
[101] C. S. Franco, G. A. Quintero, N. Myrén, A. Kudlinski, H. Zeghlache, H. R.

Carvalho, A. L. C. Triques, D. M. González, P. M. P. Gouvêa, G. Martinelli,
Y. Quiquempois, B. Lesche, W. Margulis, and I. C. S. Carvalho. Measurement of
depletion region width in poled silica. Appl. Opt., 44(27):5793–5796, 2005. URL
http://ao.osa.org/abstract.cfm?URI=ao-44-27-5793.
[102] Y. Quiquempois, A. Kudlinski, G. Martinelli, G. A. Quintero, P. M. Gouvea, I. C.

Carvalho, and Walter Margulis. Time evolution of the second-order nonlinear dis-
tribution of poled infrasil samples during annealing experiments. Opt. Express, 14
(26):12984–12993, 2006. URL http://www.opticsexpress.org/abstract.cfm?
URI=oe-14-26-12984.
[103] D. E. Carlson. Ion depletion of glass at a blocking anode: I, theory and ex-
perimental results for alkali silicate glasses. Journal of the American Ceramic
Society, 57(7):291–294, 1974. URL http://dx.doi.org/10.1111/j.1151-2916.
1974.tb10903.x.
[104] D. E. Carlson, K. W. Hang, and G. F. Stockdale. Ion depletion of glass at a block-

ing anode: II, properties of ion-depleted glasses. Journal of the American Ceramic
Society, 57(7):295–300, 1974. URL http://dx.doi.org/10.1111/j.1151-2916.
1974.tb10904.x.
[105] Y. Quiquempois, N. Godbout, and S. Lacroix. Model of charge migration during

thermal poling in silica glasses: Evidence of a voltage threshold for the onset of a
second-order nonlinearity. PHYSICAL REVIEW A, 65:1–14, 2002.
[106] F. C. Garcia, I. C. S. Carvalho, E. Hering, W. Margulis, and B. Lesche. Inducing

a large second-order optical nonlinearity in soft glasses by poling. Applied Physics
Letters, 72(25):3252–3254, 1998. doi: 10.1063/1.121614. URL http://link.aip.
org/link/?APL/72/3252/1.
[107] D. W. Shin and M. Tomozawa. Electrical resistivity of silica glasses. Journal of

Non-Crystalline Solids, 163(2):203 – 210, 1993. ISSN 0022-3093. doi: DOI:10.
1016/0022-3093(93)90769-T. URL http://www.sciencedirect.com/science/
article/B6TXM-48CXRB3-1X2/2/d69566d9f4cdba18da02e9d1103c30e1.
228 Bibliography
[108] W. Margulis, O. Tarasenko, and N. Myrén. Who needs a cathode?? creating a

second-order nonlinearity by charging glass fiber with two anodes. Opt. Express, 17
(18):15534–15540, 2009. URL http://www.opticsexpress.org/abstract.cfm?
URI=oe-17-18-15534.
[109] I. Shoji, T. Kondo, A. Kitamoto, M. Shirane, and R. Ito. Absolute scale of second-
order nonlinear-optical coefficients. J. Opt. Soc. Am. B, 14(9):2268–2294, 1997.
URL http://josab.osa.org/abstract.cfm?URI=josab-14-9-2268.
[110] H. An, S. Fleming, and G. Cox. Visualization of second-order nonlinear layer in

thermally poled fused silica glass. Applied Physics Letters, 85(24):5819–5821, 2004.
doi: 10.1063/1.1835554. URL http://link.aip.org/link/?APL/85/5819/1.
[111] P. D. Maker, R. W. Terhune, M. Nisenoff, and C. M. Savage. Effects of dispersion

and focusing on the production of optical harmonics. Physical Review Letters, 8
(1):21 – 23, January 1962.
[112] J. Jerphagnon and S. K. Kurtz. Maker fringes: A detailed comparison of theory

and experiment for isotropic and uniaxial crystals. Journal of Applied Physics, 41
(4):1667–1681, 1970.
[113] W. N. Herman and L. M. Hayden. Maker fringes revisited: second-harmonic

generation from birefringent or absorbing materials. Journal of the Optical Society
of America B, 12(3):416 –427, March 1995.
[114] J. Jerphagnon and S. K. Kurtz. Optical nonlinear susceptibilities: Accurate relta-

tive values for quartz, ammonium dihydrogen phosphate, and potassium dihydro-
gen phosphate. Physical Review B, 1(4):1739–1744, 1970.
[115] Bahaa E. A. Saleh and Malvin Carl Teich. Fundamentals of Photonics. Wiley-
Interscience, 1991.
[116] K. Messa and M. Lybanon. Curve fitting using genetic algorithms. Final report,
Naval Oceanographic and Atmospheric Research Laboratory, 1991.
[117] Karr C.L., Weck B., Massart D.L., and Vankeerberghen P. Least me-
dian squares curve fitting using a genetic algorithm. Engineering Appli-
cations of Artificial Intelligence, 8:177–189, April 1995. doi: doi:10.1016/
0952-1976(94)00064-T. URL http://www.ingentaconnect.com/content/els/
09521976/1995/00000008/00000002/art00064.
[118] I. A. Heisler, L. Misoguti, S. C. Zlio, E. V. Rodriguez, and C. B. de Arajo. Spec-

trally resolved femtosecond maker fringes technique. Applied Physics Letters, 92:
1–3, March 2008. doi: 10.1063/1.2890487.
Bibliography 229
[119] Gorachand Ghosh. Dispersion-equation coefficients for the refractive index and
birefringence of calcite and quartz crystals. Optics Communications, 163(1-3):
95 – 102, 1999. ISSN 0030-4018. doi: DOI:10.1016/S0030-4018(99)00091-7.
URL http://www.sciencedirect.com/science/article/B6TVF-3WFNPTY-J/2/
b8085a791766c48a0605005481fe719f.
[120] W. Margulis and F. Laurell. Interferometric study of poled glass under etching.
Opt. Lett., 21(21):1786–1788, 1996. URL http://ol.osa.org/abstract.cfm?
URI=ol-21-21-1786.
[121] G. A. C. M. Spierings. Wet chemical etching of silicate glasses in hydrofluoric acid

based solutions. JOURNAL OF MATERIALS SCIENCE, 28:6261–6273, 1993.
[122] P. Blazkiewicz, W. Xu, and S. Fleming. Optimum parameters for CO2 laser-
assisted poling of optical fibers. J. Lightwave Technol., 20(6):965, 2002. URL
http://jlt.osa.org/abstract.cfm?URI=JLT-20-6-965.
[123] K. Lee, P. Hu, J. L. Blows, D. Thorncraft, and J. Baxter. 200-m optical fiber with
an integrated electrode and its poling. Opt. Lett., 29(18):2124–2126, 2004. URL
http://ol.osa.org/abstract.cfm?URI=ol-29-18-2124.
[124] M. Fokine, L. E. Nilsson, Ãâ. Claesson, D. Berlemont, L. Kjellberg, L. Krum-

menacher, and W. Margulis. Integrated fiber Mach-Zehnder interferometer for
electro-optic switching. Opt. Lett., 27(18):1643–1645, 2002. URL http://ol.
osa.org/abstract.cfm?URI=ol-27-18-1643.
[125] M. Fokine, L. E. Nilsson, A. Claesson, L. Krummenacher, W. Margulis, and

L. Kjellberg. Method for forming a longitudinal solid body within an optical
fiber, 2009-05-26.
[126] N. Myrén, H. Olsson, L. Norin, N. Sjödin, P. Helander, J. Svennebrink, and

W. Margulis. Wide wedge-shaped depletion region in thermally poled fiber
with alloy electrodes. Opt. Express, 12(25):6093–6099, 2004. URL http://www.
opticsexpress.org/abstract.cfm?URI=oe-12-25-6093.
[127] Paul A. Tipler. College Physics. Worth Publishers Inc.,U.S., 1987.
[128] D. Faccio, A. Busacca, W. Belardi, V. Pruneri, P.G. Kazansky, T.M. Monro, D.J.
Richardson, B. Grappe, M. Cooper, and C.N. Pannell. Demonstration of thermal
poling in holey fibres. Electronics Letters, 37(2):107 –108, 18 2001. ISSN 0013-
5194. doi: 10.1049/el:20010089.
[129] F. Brechet, J. Marcou, D. Pagnoux, and P. Roy. Complete analysis of the

characteristics of propagation into photonic crystal fibers, by the finite element
230 Bibliography
method. Optical Fiber Technology, 6(2):181 – 191, 2000. ISSN 1068-5200. doi:
DOI:10.1006/ofte.1999.0320. URL http://www.sciencedirect.com/science/
article/B6WP0-45BCN8G-F/2/9228e5f54b5ffbf7a4949e1bdf6fd9da.
[130] D. Marcuse. Gaussian approximation of the fundamental modes of graded-

index fibers. J. Opt. Soc. Am., 68(1):103–109, 1978. URL http://www.
opticsinfobase.org/abstract.cfm?URI=josa-68-1-103.
[131] Werner Vogel. Glass Chemistry. SPRINGER VERLAG INC, 2nd edition, 1994.
[132] Q. Coulombier, L. Brilland, P. Houizot, T. Chartier, T. N. N’Guyen, F. Smektala,

G. Renversez, A. Monteville, D. Méchin, T. Pain, H. Orain, J. C. Sangleboeuf, and
J. Trolès. Casting method for producing low-loss chalcogenide microstructured op-
tical fibers. Opt. Express, 18(9):9107–9112, Apr 2010. doi: 10.1364/OE.18.009107.
URL http://www.opticsexpress.org/abstract.cfm?URI=oe-18-9-9107.
[133] Xian Feng, A.K. Mairaj, D.W. Hewak, and T.M. Monro. Nonsilica glasses for
holey fibers. Lightwave Technology, Journal of, 23(6):2046 – 2054, june 2005.
ISSN 0733-8724. doi: 10.1109/JLT.2005.849945.
[134] H. Ebendorff-Heidepriem and T. M. Monro. Extrusion of complex preforms for

microstructured optical fibers. Opt. Express, 15(23):15086–15092, 2007. URL
http://www.opticsexpress.org/abstract.cfm?URI=oe-15-23-15086.
[135] K. J. Rowland. Guiding Light in Low-Index Media via Multilayer Waveguides.

PhD thesis, Institute for Photonics & Advanced Sensing, 2010.
[136] J. Lousteau, H. Bookey, Xin Jiang, C. Hill, A. Kar, and A. Jha. Fabrication
of multicore tellurite glass optical fibres. In Transparent Optical Networks, 2007.
ICTON ’07. 9th International Conference on, volume 2, pages 305 –308, july 2007.
doi: 10.1109/ICTON.2007.4296211.
[137] D. Crowdy, 2011. Private Communication. (Professor in Applied Mathematics,

Imperial College London).

02whole PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

02whole PDF

Uploaded by

Copyright:

Available Formats

A Study of Tellurite Glasses for

Electro-optic Optical Fibre Devices

A thesis submitted in fulﬁlment of the

1. Manning, Sean; Ebendorﬀ-Heidepriem, Heike; Heike Monro, Tanya Mary.

3. Manning, Sean; Ebendorﬀ-Heidepriem, Heike; Monro, Tanya Mary; Munch, Jes-

We investigated thermal poling as a potential post processing technique for inducing

Declaration of Authorship iii

List of Figures xiii

List of Tables xix

2 Nonlinear Optical Theory 15

3 Glass Theory, Design & Fabrication 39

3.3.2 Relationships Between the Structure and Properties . . . . . . . . 44

4 Analysis of the Microscopic Structure of Glass 55

5 Measurements of Physical and Thermal Properties 75

6 Measurements of the Optical Properties 97

7 Thermal Poling 127

8 Measurements of Second Order Nonlinearities 147

9 Fibre Preliminaries 177

9.5 Preliminary Investigations of Optical Fibre Fabrication . . . . . . . . . . . 206

10 Concluding Remarks 213

1.1 An assortment of microstructured optical ﬁbres . . . . . . . . . . . . . . . 2

2.1 Second harmonic generation energy level diagram . . . . . . . . . . . . . . 18

3.1 Atomic arrangements in the crystalline and glassy state . . . . . . . . . . 40

5.1 Experimental conﬁguration for Archimedes Density Measurement . . . . . 77

6.1 Proposed energy band diagram for tellurite glass . . . . . . . . . . . . . . 101

6.13 Schematic of Z-scan experimental setup . . . . . . . . . . . . . . . . . . . 116

9.29 Cut back measurement data showing optical attenuation of a tellurite

3.1 Deitzel’s Field Strength Parameters for Glass Formation . . . . . . . . . . 45

4.1 Comparison of certain properties of Quartz and Fused Silica . . . . . . . . 56

5.1 Molar masses, Densities and molar volumes of tellurite glasses. . . . . . . 78

7.1 Brief summary of literature reported tellurite thermal poling . . . . . . . 131

8.1 Optical parameters for quartz reference sample . . . . . . . . . . . . . . . 164

1.1 Background and Motivation

Figure 1.1: An assortment scanning electron microscope images of various microstruc-

1.2 Review of Literature

1.2.1 Tellurite Glass

1.2.2 Electro-optics in Optical Fibres

Having a technique for creating a second order nonlinearity in glass is a signiﬁcant

Signiﬁcantly the GLAMOROUS work resulted in the fabrication of an optical ﬁbre

1.3 Thesis Aims and Methodology

2. To identify suitable tellurite glasses for nonlinear optical ﬁbre devices.

3. To generate permanent second order nonlinearities in some of these glasses via

4. To design and fabricate prototype electro-optical devices from in house fabricated

1.4 Thesis Outline

1.5 Statement of Original Work and Author Contribution

1.6 List of Publications

1. Manning, Sean; Ebendorﬀ-Heidepriem, Heike; Heike Monro, Tanya Mary.

3. Manning, Sean; Ebendorﬀ-Heidepriem, Heike; Monro, Tanya Mary; Munch, Jes-

Nonlinear Optical Theory

2.2 Nonlinear Optics

P̃(t) = 0 χ(1) Ẽ(t), (2.1)

χ(1) = n20 − 1 (2.2)

and therefore to the relative permitivity r via

Therefore χ(1) determines much of the optics we experience on a day-to-day basis. We

optical electric ﬁeld can be represented as a discrete sum of frequency components by

E(−ωn ) = E(ωn )∗ . (2.6)

We are now in a position to analyse speciﬁc nonlinear processes.

2.3 Second Order Nonlinearities

E(ωn )e−iωn t E(ωm )e−iωm t = E(ωn )E(ωm )e−iωn t e−iωm t

becomes the expression for the second order nonlinear polarisation.

jk: 11 22 33 23,32 31,13 12,21

P̃(2) (t) = 0 χ(2) Ẽ2 (t) (2.13)

where the optical ﬁeld is give by

P̃(t) = 0 χ(1) Ẽ(t), (2.1)

and therefore to the relative permitivity r via

P̃(2) (t) = 0 χ(2) Ẽ2 (t) (2.13)

− P̃(2) (t) = 0 χ(2) |−Ẽ(t)|2 . (2.15)

−P̃(3) (t) = 0 χ(3) |−Ẽ(t)|2 (−Ẽ)

D̃(1) = 0 Ẽ + P̃(1) (2.26)

D̃(1) = (1) · Ẽ (2.27)

dAωi (z) ω (2) 2ω ω∗

PN L =0 χ(3) E3T

Rotation about Z by 180◦ ,

Reﬂection symmetry in the XZ plane,.

Inversion symmetry in the X and Y axes which is equivalent to a reﬂection.

sin θω = nω sin θω . (2.77)

êω = (cos θω , 0, sin θω ). (2.78)

deﬀ = 2d31 sin θω cos θω cos θ2ω

P̃(3) (t) = 0 χ(3) Ẽ(t)3 , (2.84)

P̃(3) (t) = 0 χ(3) E30 cos3 ωt. (2.86)

I = 2n0 o c|E(ω)|2 (2.90)

P(3) (ω) = 30 χ(3) (ω = ω + ω − ω)|E(ω)|2 E(ω) (2.91)

P(ω) = 0 χeﬀ E(ω). (2.94)

They will have zero-valued even-ordered nonlinear coeﬃcients.