Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Titania and hybrid titania - silica sol-gel thin films and their applications in integrated optical devices
(USC Thesis Other)
Titania and hybrid titania - silica sol-gel thin films and their applications in integrated optical devices
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
TITANIA AND HYBRID TITANIA – SILICA SOL-GEL THIN FILMS AND
THEIR APPLICATIONS IN INTEGRATED OPTICAL DEVICES
by
Hari Mahalingam
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfilment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ELECTRICAL ENGINEERING)
May 2014
Copyright 2014 Hari Mahalingam
ii
Dedication
To my family: Kamakshi, Mahalingam, Sriram, Dhanya and Tyke.
iii
Acknowledgements
I was introduced to the field of photonics by my advisor, Professor William H. Steier, when
I attended his graduate level class on Non Linear Optics. Six years later, I have come to
realize many important lessons learnt from him during the course of my Ph.D. studies. If
not for his unwavering support, constant guidance and patience I would never have
fulfilled my dream of getting a doctorate. He has a remarkable foresight of things, is
always available and has been flexible in letting me carry out my work independently. His
humbleness and magnanimity are qualities that I would like to exact by the time I grow
old. I feel honored to have been his last Ph.D. student and thank him for providing me
with an opportunity to work with him and in making me a better engineer.
For teaching me the tricks of the trade and making my experience at USC
memorable, I would like to express my gratitude to the past members of Professor Steier's
research group: Dr. Bipin Bhola, Dr. Satsuki Takahashi, Dr. Greeshma Gupta, Dr. Yoo Seung
Lee, Dr. Andrew Yick., Dr. Thanh Le, Nutthamon Suwanmonkha and Professor. Sang Shin
Lee for helping me set achievable research goals and guiding me through my first
waveguide cutback measurements.
I sincerely thank the members of my qualifying exam and dissertation committee
- Professor Jack Feinberg, Professor Armand Tanguay, Professor Stephen Cronin,
Professor Andrea Armani and Dean John O'Brien. I hope you found some of it interesting.
Special thanks to Dean O’Brien for making time for both. I would also like to thank
Professor Aluizio Prata for the knowledge imparted from all the Electromagnetics courses
iv
he taught and Professor Anupam Madhukar for the classes on Solid State Physics and
Solar Cells during my time here.
Some of the data collected for this dissertation would not have been possible
without the timely help of the following people: Simin Mehrabani for training me on the
FTIR machine, the microdisk fabrication process and testing them as well, Dr. Sarah
Conron for helping me sort out the chemistry in my sol-gel recipes, Andrew Bartynski and
Matt Greaney for XRD measurements which took hours, Zach Lingley for AFM
measurements, Maoqing Yao for etching samples in UCLA, Dr. Patrick Nasiatka for lab
supplies and printouts and Francisco Navarro for time sharing the OVPD furnace without
which I wouldn’t have any thin films to report about.
I also thank Ms. Susan Zarate and Ms. Kim Reid for excellent administrative
support and making sure that I received my paychecks on time and Donghai Zhu for
keeping the cleanroom equipment up and running when I needed them the most.
For great memories away from the lab and their continued friendship, I appreciate
my roommates Muralikrishna Rao, Vinay Setty and Utteerna.
Finally, but by no small margin importantly, I would like to express my heartfelt
gratitude towards my parents and brother for all their sacrifices in providing me with an
opportunity to pursue higher studies, my parents - in - law for their constant support and
my lovely wife, Dhanya, for taking the journey of life with me.
v
Table of Contents
Dedication ii
Acknowledgements iii
List of Tables vii
List of Figures ix
Abstract xv
Chapter 1 Introduction 1
1.1 Optical material systems exploiting refractive index control . . . . . . . . . 2
1.2 Competing material systems for integrated photonic devices . . . . . . . . 5
1.3 Sol-gels in photonics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4 Outline of contributions of this work . . . . . . . . . . . . . . . . . . . . . . 12
Chapter 2 Synthesis and preparation of hybrid sol - gel thin films 14
2.1 Control parameters in sol-gel synthesis . . . . . . . . . . . . . . . . . . . . 16
2.2 Titanium dioxide (TiO 2) solution . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3 Silicon dioxide (SiO 2) solution . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.4 Zirconium dioxide (ZrO 2) solution . . . . . . . . . . . . . . . . . . . . . . . . 21
2.5 Hybrid stock solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.6 Preparation of thin films on silicon, thermal oxide coated silicon and
quartz substrates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Chapter 3 Characterization of titania, titania - silica and titania -
zirconia thin films 28
3.1 Characterization of titania sol-gel thin films . . . . . . . . . . . . . . . . . . 29
3.1.1 UV – VIS spectrophotometry and FTIR spectroscopy . . . . . . . . . 31
3.2 Characterization of of diluted hybrid sol-gel thin films . . . . . . . . . . . . 34
3.2.1 General films from thin film data . . . . . . . . . . . . . . . . . . . . 34
3.3 Characterization of undiluted hybrid sol-gel thin films . . . . . . . . . . . . 40
3.4 Hybrid titania – silica sol-gel films with varying silica concentrations . . . . 42
3.5 Film crystallinity and surface quality . . . . . . . . . . . . . . . . . . . . . . 46
3.5.1 Film quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.5.2 XRD measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
vi
Chapter 4 Fabrication of ridge waveguides and characterization
of propagation losses from cutback measurements 55
4.1 Fabrication of straight waveguides . . . . . . . . . . . . . . . . . . . . . . . 57
4.2 Reactive ion etching results . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.3 Cutback loss measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.4 Propagation loss analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.4.1 2D Scattering loss calculations . . . . . . . . . . . . . . . . . . . . . . 71
4.4.2 3D Scattering loss calculations . . . . . . . . . . . . . . . . . . . . . . 73
Chapter 5 Demonstration of integrated microring resonators
using hybrid titania – silica sol-gel 80
5.1 Microring resonator theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.1.1 Parameters of the ring resonator . . . . . . . . . . . . . . . . . . . . 83
5.1.2 Coupling Schemes and Losses . . . . . . . . . . . . . . . . . . . . . . 85
5.2 Fabrication and testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.3 Microdisk and microtoroid resonators . . . . . . . . . . . . . . . . . . . . . . 91
5.3.1 Fabrication and characterization of microdisks . . . . . . . . . . . . . 91
Chapter 6 Conclusion 97
6.1 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
Bibliography 102
Appendix A Hybrid titania – silica and titania - zirconia sols:
synthesis details 111
A.1 Mixing sols by varying molar concentrations of compounds . . . . . . . . . 111
A.2 Mixing sols by varying volume ratios of stock solutions . . . . . . . . . . . . 112
Appendix B Spectroscopic ellipsometry of various sol-gel thin films 115
Appendix C Additional images of thin-films and testing setup 120
vii
List of Tables
1.1 Different optical material systems with a comparison of the refractive index
variation that has been engineered, applications & typical deposition/
synthesis techniques. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Competing optical material systems with a comparison of the refractive
index variation that has been engineered & typical deposition / synthesis
techniques with respect to the hybrid sol-gel system studied in this
dissertation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1 Products obtained according to the relative rates of hydrolysis and
condensation, reproduced from [23]. . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 Charge distribution in Ti (OR) 4 and Si (OR) 4 n-alkoxides. k h is the rate constant
for hydrolysis of silica precursors with different alkyl chains and all the
other columns represent partial charge distributions; Hydrolysis rates
decrease as the partial charge on the metal atom reduces, reproduced from
[23]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3 Film thicknesses obtained on coating titania sols diluted using different
solvents in equal volumes. They were all baked on a hotplate at
150
0
C for 17 minutes after spin coating at 2000 rpm for 1 minute. . . . . . . 22
3.1 Refractive index (at four different wavelengths) and thickness data of
various sols spin coated at 2000 rpm for 1 minute on silicon wafers. All
samples were baked at the temperatures mentioned above in column 2 and
re-baked in a nitrogen environment at 1010
0
C for 1 hour. . . . . . . . . . . . 35
3.2 Comparison of thickness and refractive index of three different sols spin
coated on silicon substrates and baked at 700
0
C in vacuum for 24 hours. . . . 41
4.1 Various control parameters used to determine optimal etch recipe. In all
cases the duration of etch was maintained at 4 minutes. . . . . . . . . . . . . 59
4.2 Finite difference simulation for waveguides of different widths indicating
effective indices and number of lobes parallel to film surface. 5 higher order
modes for the 5 µm waveguide are not shown in the table. . . . . . . . . . . . 65
4.3 Propagation loss values for waveguides of different widths at a
wavelength of 1550 nm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
viii
4.4 (2D) Scattering losses in dB/cm for waveguides of different widths using two
different r.m.s roughness values and a correlation length of 50 nm. . . . . . . 73
4.5 (3D) Scattering losses in dB/cm for waveguides of different widths using
two different r.m.s roughness values and a correlation length of 50 nm. . . . 76
4.6 Effective index of TM modes, power confinement and normalized intensity
at side wall for waveguides of different widths and air cladding. . . . . . . . . 77
4.7 Separating contribution to optical losses from absorption and scattering. . . 77
5.1 Measured and fitted values of race track micro ring parameters with
different circumferences. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
A.1 Summary of concentrations used to obtain hybrid titania – silica sols with
5%, 11% and 25% silica content. . . . . . . . . . . . . . . . . . . . . . . . . . . 112
A.2 Volume ratios required to vary silica concentrations in hybrid titania
– silica solutions based on the total volume of titania and silica stock
solutions prepared. Equivalent molar ratios are calculated in the last
column to compare the recipes in both the tables on a common scale. . . . . 113
ix
List of Figures
1.1 Change in refractive index (c) and absorption coefficient (f) at 633 nm
recorded in films obtained by varying ratio of ammonia and silane gas flow
rates. Circles indicate a frequency of 13.56 MHz and the squares indicate
380 kHz [14]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2 High resolution SEM images of TiO 2 films deposited by APCVD at 450
0
C
(a) and sintered for 22 hours at 1000
0
C [2]. . . . . . . . . . . . . . . . . . . . 8
1.3 Left: SEM image of 20 mole% Zr ridge waveguides obtained by exposure to
UV light for 1 min and post bake at 150
0
C for 5 hours [10]. Right: Index of
refraction of TiO 2-SiO 2 binary coating as a function of composition and
heat treatment atmosphere [11]. . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1 Reaction involving benzoyl acetone and titanium butoxide [35]. . . . . . . . . 19
2.2 Recipe used to synthesize titania sol-gels. . . . . . . . . . . . . . . . . . . . . 20
2.3 Glovebag used to mix the different ingredients mentioned in Figure2.2. . . . 20
2.4 (L) Vacuum Tube furnace setup. (R)Samples are placed inside another quartz
tube to avoid contamination of the chamber, at positions shown. The
furnace consists of 4 zones (fourth zone on the right is cut off in the picture)
with the zones at the ends having lower temperatures due to improper
insulation and leakage of heat via the ends. . . . . . . . . . . . . . . . . . . . 23
2.5 Furnace showing 3 zones. Heating element of 4th zone malfunctioned
during the experiment resulting in slightly darker quartz samples in Zone – 3
while quartz samples in Zone – 2 were clear and in Zone -1 was dark. This
issue was fixed on later runs. . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.6 FTIR spectra of titania sol-gel resist baked at different temperatures [35]. . . 25
2.7 TGA curves of gels with different titanium content.
Glyycidyloxipropyltrimethoxysilane (GLYMO), (TEOS) and titanium iso
propxide (TPOT) were mixed in different ratios to obtain the gels. GLYMO:
TEOS:TPOT ratios for the four curves are: a) 0.25:0.25:0.5, b) 0.3:0.3:0.4,
c) 0.35:0.35:0.3 and d) 0.4:0.4:0.2 [37]. . . . . . . . . . . . . . . . . . . . . . 26
3.1 Variation in thickness and refractive index of thin films for titania samples
vacuum baked on silicon substrates at different temperatures (Note: Sample
baked at 700
0
C was in air). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
x
3.2 Top left: Q2, Clear titania film after baking at 700
0
C. Edges are still dark
since there is buildup of thickness at the edges of the sample during spin
coating. Top right: Q3, Opaque titania film after baking at 800
0
C as sample
was placed in end zone. Bottom left: Q1, Semi-clear titania film after baking
at 700
0
C as sample was placed in end zone. Bottom right: Blank quartz
sample used as a calibration reference. . . . . . . . . . . . . . . . . . . . . . 32
3.3 UV VIS spectrophotometry data of different sol-gel samples shown earlier. 32
3.4 FTIR spectra of different sol-gel samples Q1, Q2, Q3 and blank quartz slide. 33
3.5 Absorbance spectra of quartz samples containing different sol-gel recipes.
vacuum baked at 700
0
C for 2 hours. . . . . . . . . . . . . . . . . . . . . . . . 37
3.6 Absorbance spectra of quartz samples containing different sol-gel recipes
vacuum baked at 800
0
C for 2 hours. . . . . . . . . . . . . . . . . . . . . . . 38
3.7 Absorbance spectra of quartz samples containing different sol-gel recipes
vacuum baked at 900
0
C for 2 hours. . . . . . . . . . . . . . . . . . . . . . . 38
3.8 Absorbance spectra of quartz samples containing different sol-gel recipes
air baked at 700
0
C for 2 hours. . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.9 Refractive index vs wavelength data for four different hybrid titania - silica
thin films containing 5%, 11%, 35% and 43% molar silica content after
initial vacuum bake at 700
0
C for 9 hours (5% silica sample 5 hours only)
and subsequent air rebake at 1020
0
C for 5 hours . . . . . . . . . . . . . . . 43
3.10 Refractive index vs wavelength data for three different hybrid titania- silica
thin films containing 10%, 20%, and 30% volume silica content. The 30%
doped samples were sintered in nitrogen and oxygen and the curves
suggest a lower change of refractive index for samples sintered in oxygen
at 1020
0
C. Film thickness is mentioned on the right. . . . . . . . . . . . . . 44
3.11 SEM images of titania films on silicon substrates baked at 700
0
C. Left -
sample baked in vacuum; Right – sample baked in air. . . . . . . . . . . . . 47
3.12 SEM images of hybrid titania – silica and titania – zirconia films on silicon
substrates baked at 700
0
C. Top left and bottom left– Titania – silica samples
baked in vacuum and air respectively; Top right and Bottom Right – Titania -
zirconia samples baked in vacuum and air respectively. The crack on the film
in top right is an artifact from cleaving the thin films. . . . . . . . . . . . . . . 48
xi
3.13 SEM images of titania films (left) and hybrid titania-silica films (right) on
silicon substrates sintered at 1020
0
C after air baking at 700
0
C initially.
Images with similar roughness profiles were observed even in hybrid -
zirconia films that were vacuum baked initially and susequently
sintered (right) but are not shown here. . . . . . . . . . . . . . . . . . . . . 48
3.14 XRD data of undoped titania samples that were baked in air at 700
0
C and
subsequently sintered in air at 1020
0
C. . . . . . . . . . . . . . . . . . . . . 51
3.15 XRD data of undoped titania samples that were baked in vacuum at 700
0
C
and subsequently sintered in air at 1020
0
C. . . . . . . . . . . . . . . . . . . 51
3.16 XRD data of titania - silica samples that were baked in air at 700
0
C and
subsequently sintered in air at 1020
0
C. . . . . . . . . . . . . . . . . . . . . 52
3.17 XRD data of titania – silica samples that were baked in vacuum at 700
0
C
and subsequently sintered in air at 1020
0
C. . . . . . . . . . . . . . . . . . . 52
3.18 XRD data of titania - zirconia samples that were baked in air at 700
0
C and
subsequently sintered in air at 1020
0
C. . . . . . . . . . . . . . . . . . . . . 53
3.19 XRD data of titania – zirconia samples that were baked in vacuum at 700
0
C
and subsequently sintered in air at 1020
0
C. . . . . . . . . . . . . . . . . . . 53
4.1 Stepwise procedure to fabricate hybrid titania-silica sol-gel waveguides
(Left to Right). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.2 Cleaved cross section of waveguide with photoresist as mask compared for
two samples; Left – Etched in CF 4 only. Right – Etched in CF 4 + O 2
environment. More photoresist is consumed in the presence of oxygen. . . 60
4.3 Cleaved cross section of etched waveguide with chrome as mask etched in
CF 4 + O 2 (35 & 3 scccm), 100 mT, 200 W. Vertical etched side walls and
undercutting of sol-gel below the chrome mask at the waveguide edges can
be seen. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.4 AFM image of etched surface before (left) and after (right) wet etching for 1
minute in diluted HF; Improvement in surface roughness is confirmed
visually. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.5 FESEM images of wet etched surface after 1 minute (left) and 2 minute
(right) etch durations; Longer etch duration results in peel off of
photoresist and sol-gel layers. . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
xii
4.6 Top view of etched waveguides after removal of chrome mask using a 50X
objective. The waveguide widths decrease from 10 – 2 µm from top to
bottom. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.7 Top view and cross sectional view of cleaved sample after forming initial
input and output facets. Subsequent cleave positions are shown. . . . . . . . 64
4.8 Mode field profiles of fundamental and first higher order quasi TE (top)
and quasi TM (bottom) modes from the simulations performed. The black
(blue in reality) borders are artifacts that appeared during image transfer. . . 66
4.9 Cutback loss measurement test setup schematic. . . . . . . . . . . . . . . . . 67
4.10 Mode profiles captured using an IR camera. Left – dominant mode observed
in waveguides with width 2, 3 and 4 μm. Right – Mode with three lobes
seen when lensed fiber is moved away from waveguide axis. . . . . . . . . . 68
4.11 Cutback loss measurement data for waveguides of 4 different widths, TM
Polarization at a wavelength of 1550 nm. . . . . . . . . . . . . . . . . . . . . . 69
4.12 Conversion of rectangular waveguide into a slab waveguide with rough
sidewalls for 2D scattering calculations. . . . . . . . . . . . . . . . . . . . . . 72
4.13 Decomposition of radiation problem into current element J rough with array
factor F rough [53]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.14 A: Rectangular waveguide depicting radiating sources. B: One layered
medium of interest for scattering calculations (Note: x c = x, y c = -z and z c = y).
C: Vertical profile of the current sources used to approximate the shape of
the first TE-like and TM like mode at the rough boundary [53]. . . . . . . . . . 74
4.15 Modelling the hybrid titania – silica thin films with silica spheres of radius
10 nm present in an ambient (porous) medium with refractive index 1.93. . . 78
5.1 Schematic of a dielectric waveguide coupled to a ring resonator. . . . . . . . 81
5.2 Transmission spectra of a microring resonator. . . . . . . . . . . . . . . . . . 83
5.3 Stepwise procedure to fabricate hybrid titania-silica sol-gel microring
resonators (Left - Right). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.4 Optical images of microring resonators post fabrication. To compare the
quality of photolithography the photomasks are shown with dark
waveguide regions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
xiii
5.5 Block diagram of characterization setup for testing microring resonators. . . . 88
5.6 Transmission spectra of microring resonator with a radius of 75 μm and
coupling length of 40 μm. The circled resonance wavelengths belong to the
quasi TM family with a measured FSR of 2.14 nm and the non-circled ones
belong to the quasi TE family with a measured FSR of 2.11 nm. . . . . . . . . 89
5.7 Image of scattering of light (top view) from the micro ring and MMI coupler.
The light is coupled into the waveguide from the left, a bright spot is visible
at the end of MMI junction. Also seen is scattering along the circumference
of the ring and a relatively bright spot on the bus waveguide caused due to
fabrication errors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.8 Formation of micropads and microdisks before and after XeF 2 etching. . . . 92
5.9 Coupling mechanisms in micro resonators. . . . . . . . . . . . . . . . . . . . 93
5.10 Top and side view of of microdisk and tapered fiber during measurement. . 93
5.11 Transmission spectra of a tapered fiber and an 80 μm (diameter) microdisk. 94
5.12 SEM images of CO 2 laser ablated microdisk. . . . . . . . . . . . . . . . . . . . 95
B.1 Generated and experimental ψ and Δ values for an air baked (700
0
C)
diluted titania film containing no water. The simple Cauchy model, MSE
and fit parameters along with the 90% confidence limits are shown above. . 117
B.2 Generated and experimental ψ and Δ values for an air sintered
(1010
0
C) undiluted titania - silica film containing 25% molar silica which
was vacuum baked at 700
0
C initially. The simple Cauchy model with an
intermix layer (made of equal amounts of air voids and sol-gel), MSE and fit
and fit parameters along with the 90% confidence limits are shown above. . 118
C.1 Optical micrographs (10X objective) of samples on silicon substrates
containing titania sol-gel vacuum baked at 100
0
C (left) and 200
0
C (right).
Cracking reduces as the baking temperature increases and samples were
crack free when baked at 300 0C or higher. . . . . . . . . . . . . . . . . . . . 120
C.2 Optical micrographs (50X objective) of samples on quartz substrates
containing undiluted titania (left) and titania – zirconia (right) sol-gel
vacuum baked at 700
0
C and subsequently sintered in air at 1020
0
C. . . . . 120
xiv
C.3 Optical micrographs (50X objective) of samples on quartz substrates
containing undiluted titania – silica sol-gel with 5% (top left), 11%
(top right) and 20% (bottom left) molar silica vacuum baked at 700
0
C and
subsequently sintered in air at 1020
0
C. As the molar percentage increases
the cracks reduce and larger islands are formed. SEM image of waveguide
on 4 µm silica coated silicon substrate obtained from 11% molar silica
doped titania – silica sol-gel after high temperature sintering (1020
0
C). . . . 121
C.4 Waveguide cutback measurement optical test setup showing the
different 3D stages. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
C.5 Optical test setup output side showing the iris, polarizer, IR camera and
detectors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
C.6 Optical micrograph (20X objective) showing top view of input facet
with lensed fiber and groups of waveguides of different widths. . . . . . . . 122
xv
Abstract
Sol-gel materials are prepared by the generation of colloidal suspensions (“sols”) which
are subsequently converted to viscous gels by polymerization and further into solid
materials upon densification. By controlled mixing of sols of different materials and
concentrations, organic/ inorganic composite materials with useful optical properties
have been synthesized for a variety of photonic applications. Inorganic titanium dioxide
(titania) and silicon dioxide (silica) thin films have been developed using the sol-gel
process and used as anti-reflection coatings for solar-cells. Nevertheless, most of the
integrated optical devices reported in literature are primarily based on the silica sol-gel
process.
By doping titania sol-gel with controlled amounts of silica sol-gel the refractive
index of the material obtained can be varied from ~2.5 to ~1.45 at the optical
communication wavelengths. Moreover, being a large bandgap material (~ 3.5 eV for
rutile phase) titania sol-gel is optically transparent over a large spectral region spanning
the visible and near infrared wavelengths like silica. However, the titania sol-gel process
is relatively complicated compared to the established silica process mainly due to the
reactive (hygroscopic) nature of the titanium starting compound. Apart from their use in
antireflection coatings which require sub 100 nanometer film thicknesses it has been
difficult to realize their potential in integrated optical device applications due to the lack
of a scalable and reliable process.
xvi
This work involves a modified synthesis process and densification route to obtain
variable refractive index titania, titania – silica and titania - zirconia hybrid sol-gel films in
the range of 1.6 – 2.5 at the telecommunication wavelengths. We present a synthesis
process involving a chelating agent by which stable viscous solutions of titania sol were
obtained giving ~ 650 nm crack free thin films on silicon substrates with a single spin and
bake cycle. On addition of silica sol (45% by molar ratio) as the dopant, films with
thicknesses in the range of 1 µm (refractive index ~ 2.1) after a single spin and bake cycle
were achieved which is the largest combination reported to date.
We investigate the properties of thin films densified at high temperatures (> 700
0
C) both under vacuum (< 100 mTorr) and nitrogen atmospheres. The vacuum based
baking process provides a reliable technique to obtain crack free hybrid sol-gel films.
Titania, titania – silica and titania - zirconia hybrid sol-gel films were characterized by
various techniques including variable angle spectroscopic ellipsometry, ultra violet –
visible – near infrared spectrophotometry, Fourier transform infrared spectroscopy,
scanning electron microscopy and X-ray diffraction to confirm their thickness and
refractive index, optical transparency, morphology and crystallinity respectively.
In order to truly determine its optical quality we fabricated ridge waveguides using
hybrid titania-silica sol-gel and performed waveguide cut-back measurements on 2.5 cm
long waveguides. Reactive ion etch (RIE) recipes using CF 4 and O 2 gases were developed
for the various sol-gel materials and RIE induced roughness was measured using AFM
measurements and alleviated using wet etching. The optical propagation losses recorded
for the fundamental TM like mode were between 8 – 13 dB/cm at 1550 nm for
xvii
waveguides of widths 4 µm – 2 µm. From 2D and 3D scattering loss calculations the
estimated absorption losses are in the range of 6 – 8 dB/cm and waveguide width
dependent scattering losses are between 2 – 7 dB/cm.
We also demonstrate integrated micro ring resonator passive filters made of
hybrid titania - silica sol-gel coated on 4 µm silica/ silicon substrates with a group index of
1.91, FSR of 12 GHz, Q factor ~ 16,000 and extract an optical loss of ~18 dB/cm which is
scattering dominated. The straight waveguide and microring resonator characterization
are the first demonstration of its kind using hybrid titania – silica sol-gel containing
relatively large titania content (~ 90% molar).
Another possible application of these materials in fabricating microdisk /
microtoroidal resonators was explored. Using tapered fiber coupling a microdisk with Q
factor of ~ 5500 was recorded. CO 2 laser ablation experiments to melt the microdisk
preforms into microtoroids were unsuccessful.
The experimental results presented in this dissertation suggest that the proposed
synthesis process and densification route are promising but need improved fabrication
processes (RIE induced roughness) to fully realize this hybrid material system for photonic
applications.
1
Chapter 1
Introduction
Optical communications gained importance primarily due to the development of low loss
(< 1 dB/km) optical fibers as a light conduit along with the invention of lasers and light
emitting diodes. To realize an optical communication network photonic circuits
equivalent to their electronic counterparts like modulators, demodulators, amplifiers,
multiplexers, filters, detectors etc. were built to transmit and receive data. One of the
reasons this technology has revolutionized communications in the 21
st
century is because
of the availability of a wide range of materials (from semiconductors to polymers and
other inorganic oxides) with interesting optical properties which have been harnessed in
manufacturing different optical components.
Two spectral bands centered at near infra-red wavelengths of 1310 nm and 1550
nm are predominantly used in optical communications as the attenuation coefficients of
silica-glass optical fibers are the lowest in those regions with values of ~ 0.3 dB/km and ~
0.16 dB/km respectively [1]. Similar to miniaturizing the transistor, the photonics
community is striving to realize micro and nano photonic devices with smaller device
footprints and higher device densities. This brings about a unique challenge as there is a
need for optical materials with relatively high refractive indices (n >2 in the infra-red
spectrum) which is required to confine light into smaller areas and low absorption over
the spectral window ranging from visible to near infra-red wavelengths. The said material
2
should also be easy to deposit (spin-coating or vapor phase deposition) on various
substrates, processable with current photo lithography and etching techniques and
scalable (obtain sufficient optical thicknesses over large substrate areas) with minimal
costs. Furthermore, if the refractive index of the material can be controlled it offers an
additional degree of freedom in designing photonic devices. By varying the refractive
index there is an added cost benefit in that the same photo mask can be used to obtain
photonic devices whose functionality can be tuned over a wider spectral range (for e.g.
filtered wavelengths of a micro ring resonator, transmission bands of a waveguide Bragg
grating).
Inorganic titanium dioxide (titania) and silicon dioxide (silica) sols (solutions)
obtained by a sol-gel process are good candidates for a hybrid material system whose
refractive index can be well controlled and they have a large transparency window
encompassing the visible and near infra-red wavelengths. By controlling the liquid
precursor concentrations of the two materials during solution synthesis the refractive
index can be varied. This hybrid material system will be investigated in this dissertation.
1.1 Optical material systems exploiting refractive index control
Some of the important materials systems where refractive index variation has been
achieved are listed below in Table 1.1. This is by no means an extensive list covering all
materials. These material systems were chosen to compare the variety of refractive index
variation, Δn, obtained and also to compare the processing techniques involved in their
synthesis and fabrication.
3
Titania films deposited using APCVD (Atmospheric pressure chemical vapor
deposition) process by the pyrolysis of a liquid precursor, tetraisopropyl titanate, with
controlled substrate temperatures (silicon) and relative humidity gave an index variation
of ~0.7 on annealing the films in nitrogen atmosphere at high temperatures (1050
0
C) [2].
Excluding the APCVD titania films, chalcogenide glasses (As 2S 3 – Arsenic trisulphide, As 2Se 3
– Arsenic triselenide etc) and semiconductor material systems exhibit the largest Δn
among materials which use a vapor deposition process (in vacuum) and are bandgap
engineered. For example, in case of the Al xGa 1-xAs (GaAs – AlAs) system as the percentage
of aluminum (x) is increased the bandgap increases (GaAs: 1.424 eV – AlAs: 3.018 eV) and
the index of refraction decreases (n GaAs = 3. 45 – n AlAs = 2. 93 at 1300 nm) [3]. The
chalcogenide glasses, having larger bandgaps (> 2 eV) are transparent in the visible
spectrum [4]. Similarly, the Titanium (Ti) deposition step in case of Ti diffused Lithium
Niobate (LiNbO 3) is done by electron beam evaporation which typically needs high
vacuum (~10
-7
Torr) [5]. In case of manufacturing optical fibers silicon tetrachloride and
germanium tetrachloride gases are flown with oxygen into a hollow silica tube and heated
to oxidize (SiO 2 & GeO 2) and form the core of the optical fiber with a higher refractive
index [6].
The similarity between all the vapor phase deposition processes is that the
material to be deposited is introduced in the form of a molecular/gas precursor and
allowed to react on or near the substrate surface. The gas flow or precursors have to be
controlled continuously to obtain required material compositions. Hence, these
processes are relatively more expensive due to the number of variables (e.g. temperature,
4
gas flow, pressure) that need to be controlled, especially if it involves high vacuum
(Molecular Beam Epitaxy, MBE also goes by the acronym Million Buck Evaporator).
Table 1.1: Different optical material systems with a comparison of the refractive index
variation that has been engineered, applications & typical deposition/synthesis
techniques.
Material
Refractive
index
variation
Δn
Applications
Typical processing
techniques
Glasses:
Optical Fibers (Single
mode)
Chalcogenide glasses:
As 2S 3 – As 24S 38Se 38
Titania films (TiO 2)
~ 0.004 [6]
~ 0.4 [4]
~ 0.7 [2]
Optical fiber
communication
Integrated
optical
waveguides
AR coatings
(Outside) Vapor Deposition
process
Thermal evaporation, RF
sputtering
APCVD(Atmospheric
pressure)
Semiconductors:
GaAs – AlAs alloy system
In 1-xGa xAs yP 1-y alloy
system
In 0.49Ga 0.51-xAl xP alloy
system
~ 0.52 [3]
~ 0.26 [7]
~ 0.2 [8]
Diode Lasers
LED’s
Diode Lasers
Molecular beam Epitaxy,
Metal Organic Chemical
Vapor
Deposition, Metal Organic
Vapor Phase Epitaxy
Oxides
Ti diffused LiNbO 3
~ 0.005 [5]
Waveguides,
modulators
Titanium deposition
followed by baking at ~
1050
0
C
Polymers
Exguide – LFR/ZPU series
Exguide – WIR/FOWG
series
~ 0.11 [9]
~ 0.05 [9]
Single mode
Multimode
waveguides
Varying acrylate
concentrations in solutions
(unverified)
Sol-gels
Silica – Zirconia (SiO 2 –
ZrO 2)
Silica – Alumina(SiO 2 –
Al 2O 3)
Silica – Titania (SiO 2 –
TiO 2)
~ 0.05 [10]
~ 0.01 [10]
~ 1 [11]
waveguides,
resonators,
lasers, AR
coatings
Different sols mixed in
various ratios as liquid
precursors; Spin or dip
coating
5
Moreover, chamber pump down times are long (exceeding 12 hours in the case of MBE)
and the molecular/gas precursors (Trimethylgallium, Arsine etc.) are extremely toxic
requiring rigorous handling and safety measures.
In contrast, for the polymer and sol-gel systems the refractive index variation is
achieved by incorporating controlled concentrations of acrylates and of various liquid
precursors during solution synthesis respectively. The viscosities of the polymer and sol-
gel solutions can be reduced by addition of suitable solvents (cyclopentanone, propanol,
etc.) or increased in case of sol-gels by aging the solutions for longer durations [12]. Thin
films of varying thickness and refractive index can be obtained by controlling the baking
temperature and dwell time for sol-gels. Typically, they incur lower processing costs
relative to the vacuum based technologies mentioned above. Furthermore, in case of sol-
gels the Δn obtained is ~ 1 which is twice as much as the semiconductor or chalcogenide
material systems. Yet, the choice of material system is mainly dependent on the
application requirements.
1.2 Competing material systems for integrated photonic devices
For passive integrated photonic devices like waveguides and ring resonators there are
several competing material systems which have a wide optical transparency window
extending from visible to near and mid infrared wavelengths (chalcogenide glasses) with
refractive index variation, Δn, exceeding 0.5 and refractive index values of 2 or higher at
visible and infra-red wavelengths as listed in Table 1.2 below.
6
Silicon nitride films in [13] were prepared on silicon substrates by radio frequency
(RF) reactive sputtering at a frequency of 13.6 MHz. Nitrogen and argon were introduced
into the chamber and the refractive index control was achieved by controlling the flow
rates of nitrogen and argon. The refractive index was measured at 6328 A
0
using a
Rudolph ellipsometer with a maximum value of 2.8 measured for films deposited with
equal gas flow ratios and a reduced value of 1.96 for a N 2/Ar ratio of 13. However there
is no data on the absorption coefficients in the films. In case of plasma enhanced chemical
vapor deposition (PECVD), silane (SiH 4) and ammonia (NH 3) gases were flown at fixed
ratios into a parallel plate PECVD chamber resulting in silicon nitride films on glass
substrates. The silane to ammonia ratio controlled the refractive index of the film and
was measured at three wavelengths in the visible spectrum (473 nm, 532 nm, 633 nm) as
shown in Figure 1.1 below. Increase in ammonia / silane gas ratio gave rise to films with
reduced refractive index and absorption coefficient. Thus, films with better optical quality
exhibit a lower refractive index which is a crucial trade-off that cannot be mitigated [14].
Material System
Refractive index
variation, Δn
Deposition technique
Silicon Nitride (Si 3N 4)
Silicon oxynitride (SiON)
~ 0.8 [13], ~ 0.35 [14]
~0.8 [15]
RF sputtering, PECVD
Laser ablation of target
Chalcogenide glasses
As 2S 3 – As 24S 38Se 38
~ 0.4 [4]
High vacuum thermal
evaporation
Titania films (TiO 2) ~ 0.7 - 0.9 [2] APCVD and annealing
Hybrid sol-gel:TiO 2 - SiO 2 ~ 1 [11] Spin coating and annealing
Table 1.2: Competing optical material systems with a comparison of the refractive index
variation that has been engineered & typical deposition/synthesis techniques with respect
to the hybrid sol-gel system studied in this dissertation.
7
Silicon oxynitride films on silicon substrates were obtained by laser ablation (excimer
laser: 248 nm) of a silicon nitride target in a chamber with varying oxygen flow wherein
the refractive index of the film reduced from 2.3 to 1.48 (at ~ 500 nm) with increasing
oxygen gas flow. Although, the index was measured using ellipsometry there is no data
on the absorption coefficients [15].
Figure 1.1: Change in refractive index (c) and absorption coefficient (f) at 633 nm recorded
in films obtained by varying ratio of ammonia and silane gas flow rates. Circles indicate a
frequency of 13.56 MHz and the squares indicate 380 kHz [14].
Chalcogenide glass thin films were prepared by thermal evaporation of the bulk
glasses at a pressure of 10
-7
Torr using Molybdenum crucibles. To obtain bulk glasses with
different refractive indices, parameters in the glass formation process have to be
controlled carefully (furnace temperature, glass melting dwell time and quenching
temperature etc.). Melting was carried out in silica ampoules which can explode if they
develop cracks at temperatures of 1000
0
C due to high pressures exerted by sulphur gas
[16]. Once deposited, the films need to be annealed again to remove stresses and avoid
growth of crystal pits in the films [4].
8
In case of titania films deposited by APCVD [2], for deposition temperatures above
300
0
C the resultant films crystallized into the anatase phase. The grain size (see Figure
1.2 below) and refractive index increased on annealing films at 1000
0
C which was
accompanied by a reduction in film thickness and porosity implying an increase in density.
In general the porosity, 𝜑 of a film can be estimated using the following relation:
𝜑 = 1−
𝑛 𝑓 2
−1
𝑛 𝑏 2
−1
(1.1)
where n f and n b are the refractive indices of the TiO 2 film and bulk single crystal material.
Films with an index of ~ 2.57 at 400 nm and k (imaginary part of refractive index) of 0.066
were obtained by sintering them at a temperature of 1050
0
C for 6 hours. The grain size
is important as it decides the optical bulk scattering losses incurred by the film [2].
Figure 1.2: High resolution SEM images of TiO 2 films deposited by APCVD at 450
0
C (a) and
sintered for 22 hours at 1000
0
C [2].
Reactive RF sputtered amorphous and polycrystalline (anatase) titania films on
oxidized (3 µm) silicon substrates with a refractive index of ~ 2.3 at 1550 nm have been
reported in [17]. The main drawback of this technique was that deposition of a 250 nm
9
thick amorphous film required 6 hours and close to 23 hours for anatase films. (APCVD: ~
15 minutes to deposit ~ 100 nm film).
Although there are difficulties and tradeoffs involved in processing the above
mentioned materials, one of the main advantages is that, due to a vapor phase process
they can be easily deposited on different substrates (silicon, glass etc.). High aspect ratio
silicon nitride (n = 1.99) waveguides and microring resonators with very low propagation
losses of 0.1 dB/m at 1550 nm have been demonstrated in [18]. Arsenic trisulphide planar
waveguides with propagation losses of ~ 1 dB/cm at 1300 nm [4] and race track microring
resonators integrated with Ti diffused lithium niobate waveguides and Q factors in excess
of 5 X 10
4
have been reported in [19]. Likewise RF sputtered amorphous titania
waveguides with propagation losses of ~ 4 dB/cm at 1550 nm [17] and microring
resonators with Q factors of 2.2 X 10
4
at 633 nm have been recorded [20].
In contrast, the synthesis process of both silica and titania sol-gel materials are
versatile due to the availability of various high purity precursors, ability to seamlessly
intermix different sol-gels in the liquid phase to control the properties (varying refractive
index) of thin films, and control over the polymerization process in obtaining materials
with different particle sizes, viscosities and porosity at room temperatures. The materials
can be deposited cost effectively using dip-coating or spin coating techniques to obtain
thick films (> 10 µm in case of silica sol-gels). Extensive efforts in studying silica and other
metal-oxide sol-gel systems [21 - 23] have resulted in obtaining sol-gel materials with
interesting optical properties that have found their way into many main stream
10
applications in photonics from waveguides and Bragg gratings to optical resonators, lasers
and optical sensors.
1.3 Sol-gels in photonics
Thin films of silica made through the sol-gel process have been demonstrated to have
very low propagation losses (< 1 dB/cm at 514.5 nm) [24]. By doping a titania precursor
(tetra propyl orthotitanate) into a silica precursor (Δn = 0.01) ridge waveguides using
photolithography and reactive ion etching were obtained with losses of ~ 2.4 dB/cm at a
wavelength of 633nm have been reported using the hybrid silica - titania sol-gel [12].
Methyl acryloxy propyl tri methoxy silane (MAPTMS), a silica precursor, with an inclusion
of a photoinitiator (IRGACURE 184, CIBA) and zirconia precursor (to boost the refractive
index) gave hybrid silica films which were ultra violet (UV) patterned to define ridge
waveguides with smooth side walls as shown in Figure 1.3 [10] resulting in lower
propagation losses. Likewise, by varying the amount of silica content in a titania – silica
hybrid sol-gel, thin films with variable refractive indices from 1.4 – 2.4 at 633 nm (see
Figure 1.3) and transparency in the visible and near infrared regions have been obtained
with the intention of using them as anti-reflection coatings for silicon solar cells [11], [25].
Thus, based on the precursors used photonic devices (waveguides, gratings) can
be patterned by three routes: wet-etching, dry-etching and using sol-gel as a resist which
makes the material system all the more important. Generally, low temperature
processing (< 200
0
C) gives sol-gel films which are porous. These porous films absorb
water molecules resulting in an increase in the refractive index of the sol-gel. This change
11
in refractive index can be sensed which relates to the change in moisture/humidity in the
ambience and hence have been used as a humidity sensor [26]. Likewise, fabricating
transparent sol-gel films with larger titania content gave thin films with refractive index >
2 (in the visible and NIR regions). While the examples mentioned so far are all passive
photonic devices, active devices like rare earth doped (Erbium and Ytterbium) silica micro
toroid lasers with low threshold powers have also been reported [27].
Figure 1.3: Left: SEM image of 20 mole% Zr ridge waveguides obtained by exposure to UV
light for 1 min and post bake at 150
0
C for 5 hours [10]. Right: Index of refraction of TiO 2-
SiO 2 binary coating as a function of composition and heat treatment atmosphere [11].
Most of the integrated optical device examples provided above are based on the
silica sol-gel process with the exception of the hybrid titania – silica thin films used as anti-
reflection coatings. Due to the need for anti-reflection coatings that mainly encompass
the visible spectrum on various substrates having different refractive indices (silicon, glass
etc.), most of the published literature report details on variable index hybrid titania – silica
sol-gel films with sub 150 nm thicknesses and refractive index values between 1.7 – 2.2
12
in the visible spectrum [25, 28 - 30]. The few cases where film thicknesses exceeded 150
nm were obtained by repeated spin and bake cycles (upto 10 cycles to obtain 300 nm
thick films [31]) of the sol-gel solution. Using multiple spin cycles the refractive index and
extinction coefficient of anatase titania films (~ 380 nm) characterized from 400 nm –
1600 nm have been reported in [32]. The titania sol-gel process cannot be scaled easily
(by increasing viscosity of solution) to obtain thicker films due to the hygroscopic nature
of the titania precursor. Moreover, there is very little data available on the
characterization of optical losses and refractive index of hybrid sol-gel films containing
larger titania concentrations at near infra-red wavelengths, films with refractive index
values larger than 2 at those wavelengths and demonstration of functional photonic
devices using these materials which will be undertaken in this work.
1.4 Outline of contributions of this work
This work introduces a modified synthesis process and densification route to obtain
variable refractive index titania, titania – silica and titania - zirconia hybrid sol-gel films in
the range of 1.6 – 2.5 at the telecommunication wavelengths. Chapter 2 presents the
steps involved in the synthesis process using a traditional sol-gel route (for e.g. silica and
zirconia based) and that involving a chelating agent to obtain stable viscous solutions of
titania sol. It also elaborates on the thin film preparation cycle, the different molar ratios
used for the hybrid sol-gels and the choice of baking temperatures and environments for
the densification process.
13
Chapter 3, analyses the results from characterization of the different diluted and
undiluted thin film sol-gels using various techniques including: variable angle
spectroscopic ellipsometry, ultra violet – visible – near infrared spectrophotometry,
Fourier transform infrared spectroscopy, scanning electron microscopy and X-ray
diffraction to confirm their thickness and refractive index, optical transparency and
crystallinity respectively. Some of the novel outcomes and milestones from the thin film
research and general trends observed with respect to the hybrid sol-gel material systems
are discussed.
In order to truly determine the optical quality of the hybrid material system, we
fabricated ridge waveguides using hybrid titania-silica sol-gel and performed waveguide
cut-back measurements at 1550 nm. Chapter 4, describes the fabrication details including
the development of RIE etch recipes, surface roughness analysis using AFM
measurements and wet etching and the experimental setup used in measuring the
propagation losses. In an attempt to separate the material and scattering losses, 2D and
3D loss analysis calculations of the fabricated waveguide structure and bulk scattering
losses based on XRD data and SEM images are presented.
In Chapter 5, we demonstrate integrated micro ring resonator passive filters made
of hybrid titania silica sol-gel coated on 4 µm silica/ silicon substrates with Q factors of ~
16,000. Other possible applications of these materials in fabricating microdisk resonators
are discussed and a hybrid microdisk resonator is characterized. Finally, in Chapter 6 we
summarize the significant accomplishments from this work and touch upon future
research and areas still to be investigated using this hybrid material system.
14
Chapter 2
Synthesis and preparation of hybrid sol-gel thin
films
A sol is a colloidal suspension of solid particles in a liquid. In the sol-gel process, the
precursors (starting compounds) for preparation of a colloid consist of a metal or
metalloid element surrounded by various ligands. Metal alkoxides are organic compounds
which have an organic ligand attached to a metal or metalloid atom. For example, silicon
tetraethoxide Si(OC 2H 5) 4 is a silicon alkoxide with four -(OC 2H 5), i.e. ethoxy groups,
attached to it. Typically a two-step process is involved in the synthesis of sol-gels:
Hydrolysis and Condensation/Polymerization. Metal alkoxides are popular precursors
because they react readily with water. Hydrolysis is the process by which a hydroxyl ion
(-OH) attaches itself to the metal atom, as in the following reaction:
where, R represents a proton (H
+
) or other ligands and ROH is an alcohol. Depending on
the amount of water and catalyst present, hydrolysis may go to completion:
Si(OR) 4 + 4 H 2O Si(OH) 4 + 4 ROH
15
or this reaction can stop while the metal is only partially hydrolyzed Si(OR) 4-n(OH) n. Two
partially hydrolyzed molecules can link together in a condensation reaction which can be
represented as:
This condensation reaction generally liberates a small molecule, such as water or alcohol.
This type of reaction can continue to build larger silicon containing molecules by the
process of polymerization by which a Si-O-Si oxide matrix in three dimensions can be
formed. This is the basic principle behind the formation of a sol-gel network [21]. Brinker
C.J. and Scherer G.W. [21] have written a very useful book which explains the physics and
chemistry of sol-gel processing in great detail. Sol-gel technology began to gain
importance in the 1920’s. Wright J.D. and Sommerdijk N.A.J.M. [22] have compiled a good
reference covering the important milestones upto 1975.
The term metal alkoxide is inappropriate for silicon based precursors (molecular
precursor is more correct), as silicon is a semiconductor but the reason why the starting
compounds are so called is mainly because many transition metal (e.g. Cr, V, Ti, Zr, Fe
etc.) oxides can also be synthesized by the sol-gel route. In fact, in the early stages of
development of this technology the focus was more on transition metal oxide synthesis.
Livage J. et al have provided an excellent summary of the chemistry of transition metal
oxides [23].
16
2.1 Control parameters in sol-gel synthesis
On synthesizing a sol-gel the end product can be a colloid/sol, polymeric gel or a
precipitate. This is decided primarily by the relative hydrolysis and condensation rates as
shown in Table 2.1 below. This is a good rule of thumb for a sol-gel chemist. One can see
that in order to obtain spinnable solutions the synthesis of sol-gels have to be controlled
with relatively slow hydrolysis and condensation rates (highlighted first row).
Hydrolysis Rate Condensation Rate Result
SLOW SLOW COLLOIDS / SOLS
FAST SLOW POLYMERIC GELS
FAST FAST
COLLOIDAL GEL OR
GELATINOUS PRECIPITATE
SLOW FAST CONTROLLED PRECIPITATION
Table 2.1: Products obtained according to the relative rates of hydrolysis and
condensation, reproduced from [23].
While the reaction rates have been well documented in the case of silica based sol-gel
processes (as it was a new process in the preparation of glasses and ceramics) there have
been fewer reports on determination of rate constants for transition metal oxide based
sol-gel processes [21-23]. Some of the important factors that influence the hydrolysis and
condensation rates are length of alkyl chain in the precursor, type of solvent, chelating
agents in case of metal alkoxides, water to alkoxide ratio, reaction temperature and type
of catalyst [21-23, 33, 34]. In general, the hydrolysis rates reduce as the length of the alkyl
chain increases or on the addition of chelating agents. Different catalysts produce
significant variations in the properties of dried and fired gels. Base catalyzed reactions
17
give rise to highly cross-linked sol particles which link to form gels with large pores
between interconnected particles while acid catalyzed reactions result in an open
network structure [22]. Acid catalysts like HCl, HF, HNO 3 etc. reduce the gelation time by
a factor of 10 or more (from 1000 hours in case of no catalyst to < 100 hours). The gelation
rates and the properties of the gel formed are determined by the combination of alkoxide
and solvents used in the synthesis and the water to alkoxide ratio [33-34].
There is no unique route to obtain a sol containing silica or other metal oxide
matrix. One has to choose the type of alkoxide precursor, solvent, catalyst and their
respective concentrations judiciously based on the end application. Titania (TiO 2), silica
(SiO 2) and zirconia (ZrO 2) were the three different sols synthesized for the experimental
work in this dissertation. Each synthesis process is elaborated in the following sections.
2.2 Titanium dioxide (TiO
2
) solution
Transition metal alkoxides are far more reactive to water than their silicon counterparts
as the metal atoms are more electropositive and are prone to nucleophilic attack (by –
OH
-
groups) [23]. For e.g., the hydrolysis rate constants at pH 7 are 5 X 10
-9
M
-1
s
-1
for
Si(OEt) 4 and 10
-3
M
-1
s
-1
for Ti(OEt) 4 which is five orders higher [22]. The sensitivity of a
metal alkoxide to hydrolysis decreases (as the partial charge accumulated decreases) as
the length of the alkyl chain increases (-OCH 3 is more sensitive than – OC 4H 9) as seen in
Table 2.2 below.
Other ways of reducing hydrolysis rates is by using chelating agents (like acetyl
acetone or benzoyl acetone) which bond themselves to the metal precursor reducing its
18
reactivity or by organic modification of the precursor with long chain organic
polymers(like methyl cellulose or poly ethylene glycol) which reduce reactivity because of
steric hinderance effects [23]. Hence, hydrolysis and subsequent condensation of metal
alkoxides to form sols rather than precipitates can be quite challenging.
Table 2.2: Charge distribution in Ti (OR) 4 and Si (OR) 4 n-alkoxides. k h is the rate constant
for hydrolysis of silica precursors with different alkyl chains and all the other columns
represent partial charge distributions; Hydrolysis rates decrease as the partial charge on
the metal atom reduces, reproduced from [23].
As mentioned earlier, titania sols were synthesized mainly to obtain thin films (sub
100 nm) used as anti-reflection coatings for silicon solar cells [11], [25]. Brinker and
Harrington [25] synthesized diluted titania solutions using titanium tetra ethoxide,
ethanol, nitric acid and water in the volume ratio 1: 10:0.1: 0.086 respectively. The excess
ethanol content prevents the coagulation of the titania precursor due to hydrolysis. This
is a common theme observed in preparation of titania sols and thin films reported by
many other groups using different alkoxy precursors [28] – [32]. Since the alcohol to
alkoxide precursor ratios were large the sols obtained were stable and were spinnable.
However, to obtain final thicknesses of ~ 100 nm it required between 5 – 20 spin and bake
cycles. The primary trade-off in following these traditional sol-gel routes was obtaining
19
clear spinnable sols over viscous sols as reducing the solvent or increasing the water
content in the sol-gel process resulted in precipitation of the inorganic oxide.
Liu and Ho [35] reported a modified sol-gel route using equimolar titanium
butoxide, ethanol and benzoyl acetone as a chelating agent. The chelating agent (see
Figure 2.1) reduced the reactivity of the hygroscopic titanium butoxide thus eliminating
the trade-off mentioned above. The stable solutions obtained were diluted in 1-pentanol
to prepare thin films of titania which were used as an electron beam resist mask in etching
Indium phosphide. In our case, the titania stock solution prepared was using this recipe
sans the dilution step. A summary of the steps involved in the synthesis of the titania sol-
gel is shown below in Figure 2.2. The titania precursor was first chelated using benzoyl
acetone in the presence of ethanol in a vial with a magnetic stir bar for 3 hours under a
nitrogen atmosphere using a glove bag (see Figure 2.3). This final solution can be further
diluted in solvents to give films of submicron thicknesses upon spin coating. Undiluted
titania solgel coated as is gave film thicknesses ~ 1 µm even when baked at temperatures
above 500
0
C.
Figure 2.1: Reaction involving benzoyl acetone and titanium butoxide [35].
20
Figure 2.2: Recipe used to synthesize titania sol-gels
Figure 2.3: Glovebag used to mix the different ingredients mentioned in Figure2.2.
2.3 Silicon dioxide (SiO
2
) solution
Silica stock solutions were prepared using the standard sol-gel route mentioned at the
beginning of this chapter. Two different precursors were used. In one case methacryloxy
propyl trimethoxy silane (MAPTMS), a photo patternable silica precursor, was mixed with
0.01 molar hydrochloric acid as the catalyst and water in the molar ratio 1: 0.75: 1.5 based
on a recipe reported in [10]. The second recipe involved using tetra ethyl ortho silicate
(TEOS), ethanol and 0.01 molar hydrochloric acid in the molar ratio 1: 3.91: 2 respectively
21
based on [36]. Both solutions were stirred for an hour at room temperature. The former
recipe involved a methoxy precursor with an acryl substitution which could be used as a
negative photo-resist on the addition of a photo initator while the latter was a standard
quadrivalent silica precursor.
2.4 Zirconium dioxide (ZrO
2
) solution
Zirconia stock solutions were prepared using zirconium propoxide, n – propanol and
methacrylic acid in molar ratios 1: 4: 1 and were stirred for an hour at room temperature
[10]. The zirconia precursors are also hygroscopic compared to their silica counterparts
hence they were treated with methacrylic acid to avoid precipitation.
2.5 Hybrid stock solutions
The molar ratio of the silica and zirconia precursors were fixed with respect to titania
precursor. The different molar ratios prepared were: 89%:11%, 75%:25% and 57%:43%
TiO 2:SiO 2 for the titania – silica hybrid solutions and a 89%:11% :: TiO 2:ZrO 2 for the titania
– zirconia hybrid solutions. Thickness of titania films coated on silicon wafers which were
diluted in different solvents in the same ratio by volume are listed in Table 2.3 below. In
all cases, the solutions were spin coated at 2000 rpm for 1 minute on clean silicon wafers
and baked on a hot plate for ~ 17 minutes at 150
0
C. We can see that as the length of the
carbon chain in the alcohol increased the final thickness of spin coated films reduced. Di
methyl formamide (DMF) was added as Drying control chemical additive (DCCA) to
alleviate stresses during film drying and obtain crack free films. Based on the trend
22
observed in using different solvents, undoped titania stock solutions and hybrid solutions
containing 11% silica or zirconia were further diluted using 1-propanol (1:1 by weight) and
used in thin film studies to determine the refractive indices and film thickness obtained
as a function of baking temperature and environment (air or vacuum).
Solvent used for dilution
Thickness measured in nm
using a profilometer
2 – propanol ~ 910 nm
1 – butanol ~ 866 nm
1 – pentanol ~ 800 nm
1 – pentanol + Di methyl
formamide
~ 590 nm
Table 2.3: Film thicknesses obtained on coating titania sols diluted using different solvents
in equal volumes. They were all baked on a hotplate at 150
0
C for 17 minutes after spin
coating at 2000 rpm for 1 minute.
Moreover, X-ray diffraction (XRD) measurements were conducted on films coated on
silicon substrates prepared using these solutions to ascertain the nature of crystallinity
exhibited as a function of baking temperature.
Silica doped hybrid titania – silica stock solutions were prepared to study the
variability in refractive index and film thickness obtained as a function of baking
temperature. The stock solutions were not diluted in this case as the final thickness of the
film had to be such that it supported at least a single optical mode in the vertical direction
when coated on 4 µm thermally oxidized silica coated silicon wafers. Appendix A contains
calculations used in arriving at these molar ratios. The thin film preparation process is
detailed in the next few sections.
23
2.6 Preparation of thin films on silicon, thermal oxide coated
silicon and quartz substrates
The hybrid solutions were spin coated on pre cleaned substrates (silicon, quartz and
thermally grown silica coated silicon wafers) for various thin film characterization
experiments. The spin speed was fixed at 2000 rpm with a dwell time of 1 minute in most
cases. The annealing/baking process was undertaken in two different environments: air
(N 2) and vacuum (~100 mTorr). The ramp up time varied between 20 – 30 min to ramp up
the furnace to baking temperatures of 700
0
C – 1050
0
C while the ramp down time was
typically 3 hours or more for the furnace to cool down to room temperature. Yoldas [11]
reported the advantage of baking titania sol-gel films under vacuum which gave rise to
denser films (see Figure 1.3). Thus, both methods were investigated.
Figure 2.4: (L) Vacuum Tube furnace setup. (R)Samples are placed inside another quartz
tube to avoid contamination of the chamber, at positions shown. The furnace consists of
4 zones (fourth zone on the right is cut off in the picture) with the zones at the ends having
lower temperatures due to improper insulation and leakage of heat via the ends.
The quartz tube furnace setup used under vacuum is shown below in Figure 2.4.
There are 4 heating zones present in the furnace. Due to leakage of heat through the ends
24
there exists a temperature gradient when moving away from the central zones (2 & 3) in
either direction. To confirm this initially diluted titania sol (1:1 by weight in 1 - propanol)
were spin coated on multiple quartz and silicon substrates at a speed of 2000 rpm for 1
minute and placed in different heating zones inside the furnace as shown in Figure 2.5
below.
Figure 2.5: Furnace showing 3 zones. Heating element of 4
th
zone malfunctioned during
the experiment resulting in slightly darker quartz samples in Zone – 3 while quartz samples
in Zone – 2 were clear and in Zone – 1 was dark. This issue was fixed on later runs.
The temperature of all the zones (central 2 zones controlled by the same
temperature controller) were set to 700
0
C and the samples were baked for 2 hours. It
was found that the quartz slides placed in the central zones were transparent (indicating
near complete removal of organics, see Chapter 3 for details) while the ones placed in the
end zones were still dark (During this experiment the heating element corresponding to
zone 4 malfunctioned, hence, the quartz sample placed in zone 3 is opaque in Figure 2.5.
25
This issue was fixed on later runs). Hence, for all future experiments the samples were
place in the central zones 2 and 3 while setting all four zones to the same temperature
during baking.
The choice of temperature was chosen to be 700
0
C or higher because the
chelating agent and other carbon content were removed from the film only when sintered
at temperatures greater than 500
0
C as shown by the Fourier Transform Infra Red spectra
(FTIR) of the titania sol-gel thin film (see Figure 2.6). The removal of organics from the film
can be confirmed by the absence of characteristic infrared absorption peaks due to
vibration of C – C, C = C, C – O and C – H bonds in the spectral range of 1450 cm
-1
to 1600
cm
-1
, when the sample was baked at 500
0
C or higher [35].
Figure 2.6: FTIR spectra of titania sol-gel resist baked at different temperatures [35]
Further, the structural changes during heating are generally identified by thermo
gravimetric analysis (TGA) or differential thermal analysis (DTA) in conjunction with FTIR
[21, 22]. Based on the weight loss incurred by the amorphous films the shrinkage curve
26
can typically be divided into three regions. Figure 2.7 below, reproduced from [37], shows
TGA curves of four different samples containing varying amounts of titanium propoxide
(50%, 40%, 30% & 20% respectively). In each case the sols were poured into petri dishes
and dried at room temperatures for about 10 days after which they were powdered and
placed in a gold crucible and heated at 2
0
C/min from room temperature to 800
0
C. We
can see that with increasing temperature and decreasing titanium content (a>b>c>d) the
weight loss of the powder increases. There were three loss stages observed: below 200
0
C, between 200 – 310
0
C and from 310 – 480
0
C.
Figure 2.7: TGA curves of gels with different titanium content.
Glyycidyloxipropyltrimethoxysilane (GLYMO), (TEOS) and titanium iso propxide (TPOT)
were mixed in different ratios to obtain the gels. GLYMO:TEOS:TPOT ratios for the four
curves are: a) 0.25:0.25:0.5, b) 0.3:0.3:0.4, c) 0.35:0.35:0.3 and d) 0.4:0.4:0.2 [37].
The weight loss in the first region occurs with little shrinkage and is mainly due to
desorption of water and residual solvent from the films. In the second region both
shrinkage and weight loss occur attributed mainly to removal of organics (principal weight
loss), polymerization (shrinkage proportional to weight loss) and structural relaxation
27
(shrinkage only due to bond restructuring). The condensation and structural relaxation
continues in the third region with further bond restructuring and condensation
contributing to the weight loss in that region.
While TGA or DTA experiments were not conducted with the various sols used in
this work (required months to obtain powders from dried sols), we can still predict the
outcome using the FTIR data from Figure 2.6 and TGA data from Figure 2.7. In our case
we expect a similar three region TGA curve but with a weight reduction from 100 % to ~
60 % or more as the titania content in the various solutions used are at least 55% or more
(see Appendix A). Moreover, at temperatures greater than 800
0
C the titania transforms
to rutile phase (which exhibits the highest refractive index among the three phases of
titania: brookite, anatase and rutile). In some cases the samples were further sintered in
air at 1000
0
C to reduce the porosity and increase the refractive index (or density) of the
thin films obtained, as done in [2], which resulted in further shrinkage (reduction in
thickness but no weight loss). The details of the bake cycle dwell times, baking
temperatures and environments are presented in conjunction with the characterization
results in the next chapter.
28
Chapter 3
Characterization of titania, titania – silica and
titania – zirconia thin films
One of the issues involved in using a chelating agent (benzoyl acetone) in preparing the
titania films is that the samples have to be baked at temperatures above 600
0
C to get rid
of the organics and form a TiO 2 inorganic matrix. Secondly, these organic molecules
diffuse out leaving behind a porous film. In order to increase the refractive index of
vacuum baked films, they were sintered in a regular furnace at 1000
0
C for an hour or
more in a nitrogen atmosphere to further reduce the porosity. For APCVD deposited TiO 2
films this sintering process increased the refractive index from 2.08 – 2.6 measured at 600
nm [2].
Initial experiments involved baking thin films of titania sols on silicon substrates
using a tube furnace under vacuum (few µTorr) similar to the one shown in Figure 2.4, as
films baked in ambient atmosphere (or nitrogen rich) on hot plates or ovens developed
severe cracks covering the entire area of the sample. However, since that tube furnace
was made of glass tubes (not quartz) the films could not be baked at temperatures higher
than 600
0
C. Thus for baking temperatures of 700
0
C or more the furnace shown in Figure
2.4 was used, the only trade-off being lower vacuum levels (~ 50 mTorr, lowest possible).
The effect of different vacuum levels on the characteristics of thin films was not
investigated. Also, when baked in air at temperatures of 700
0
C or more the titania films
29
were crack free. Lastly, it was also observed that crack free multiple coatings could be
obtained when the samples were vacuum baked while air baked films turned out
completely cracked after the second spin and bake cycle irrespective of whether the
second bake cycle was in vacuum or air.
3.1 Characterization of titania sol-gel thin films
Titania sol was diluted with equal amounts of 1-propanol by weight. After stirring the
solution for an hour and filtering it was spin coated at 2000 rpm for ~ 1 minute on pre-
cleaned silicon wafers. This was used to calibrate the film thickness and refractive index
using an ellipsometer and a profilometer (Dektak IIA). The main purpose of diluting the
sol-gel was to obtain thin films from which the organics would burn off quickly and thus
help establish a recipe to obtain clear films with high optical quality.
Titania films coated on silicon wafers (easier to fit ellipsometric data) were
characterized using a variable angle spectroscopic ellipsometer from J.A. Woollam & Co.,
to obtain the thickness and refractive index of the films as shown below in Figure 3.1.
Preliminary data confirmed that films with refractive indices > 2 at near infrared
wavelengths can be obtained by this processing route. A Cauchy model was used to fit
the ellipsometric data and model the thin film, details of which are provided in Appendix
B. It can be seen that thickness of the films obtained are ~ 200 nm when baked at 700
0
C.
Thicker films can be obtained by multiple spin and bake cycles while baking under vacuum
or by using non – diluted titania solutions. The thickness values measured using a
profilometer gave similar results. However, on observing the surface of the films under a
30
microscope, tiny cracks could be seen which covered the entire area of the sample for
samples which were baked at 300
0
C or less (see Appendix C). Further, the films baked at
450
0
C, 500
0
C (data point missing) and 600
0
C did not give reasonable fits from the
ellipsometry data unless some absorption was included in the model (suggesting
presence of organics). Finally, since the films were coated on silicon wafers it was hard to
confirm by visual inspection if the films were clear (of organics) as the underlying
substrate absorbs very well at visible wavelengths. In order to obtain a reliable recipe for
100 200 300 400 500 600 700
100
200
300
400
500
600
700
800
900
1000
Thickness
Refractive index
Baking temperature in
0
C
Thickness in nm
1.6
1.7
1.8
1.9
2.0
2.1
2.2
2.3
2.4
Refractive index,n
Refractive index at 1550 nm, thickness vs baking temperature
Figure 3.1: Variation in thickness and refractive index of thin films for titania samples
vacuum baked on silicon substrates at different temperatures (Note: Sample baked at 700
0
C was in air).
the baking procedure, transparent (free of organics having low absorption) and crack free
films the same sol-gel was spun on clear quartz substrates and baked at 700
0
C in vacuum.
31
These samples were used to measure the transmission of the films at visible, near infrared
and infrared wavelengths by UV-VIS (ultra violet and visible) spectrophotometry (HP 4853
diode array spectrometer with 1 nm resolution) and FTIR (Fourier Transform Infra Red)
spectroscopy (Bruker Alpha – P attenuated total reflection setup with a diamond prism).
Moreover, since quartz is transparent at visible wavelengths it was a good marker in
indicating (by visual inspection) if the titania films were transparent in the visible
spectrum.
3.1.1 UV-VIS spectrophotometry and FTIR spectroscopy
Figure 3.2 shows three samples with varying degrees of clarity on visible inspection. Since
titania is transparent at visible wavelengths any opaqueness can be tied to the presence
of organics in the film.
FTIR was conducted to check for the absence of chelating agent bond vibrations
(C-C bonds etc. in the spectral range of 1450 - 1600 cm
-1
) and the Ti-O and Si-O-Ti
vibrational spectra. The presence of organics show up as a broad absorbance
encompassing a large spectral range in both UV-VIS spectrophotometry and FTIR spectra
(see Q3 data in Figures 3.3 and 3.4). The absorbance peaks at wavelengths less than 400
nm correspond to excitations above the titania band gap (Figure 3.3) and the absorbance
peaks in the FTIR spectra at around 900 cm
-1
and 650 cm
-1
correspond to the vibrational
spectra of Ti-O-Si and Ti-O-Ti bonds respectively as seen in Figure 3.3 and 3.4 below. The
semi clear sample (Q1) exhibited a relatively smaller but broad absorbance compared to
32
Figure 3.2: Top left: Q2, Clear titania film after baking at 700
0
C. Edges are still dark since
there is buildup of thickness at the edges of the sample during spin coating. Top right: Q3,
Opaque titania film after baking at 800
0
C as sample was placed in end zone. Bottom left:
Q1, Semi-clear titania film after baking at 700
0
C as sample was placed in end zone.
Bottom right: Blank quartz sample used as a calibration reference.
Figure 3.3: UV VIS spectrophotometry data of different sol-gel samples shown earlier.
200 400 600 800 1000 1200
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
Absorbance
Wavelength in nm
Quartz blank_as_sample
Q1_700
0
C (semi_clear)
Q2_700
0
C (clear)
Q3_800
0
C (opaque)
33
the opaque sample (Q3) in both UV-Vis and FTIR spectra. For the clear sample, Q2, the
small absorbance peak centered at 500 nm is due to thin film effects (see Figure 3.3).
Based on these characterizations the dwell time to obtain clear (organics free)
samples of titania sol-gel diluted 1:1 by weight in 1 – propanol while baking at 700
0
C
under vacuum was found to be 2 hours. If the baking temperature or the amount of
dilution of sol-gel by solvent was reduced the dwell time increased. From the FTIR spectra
of titania sol-gel baked at different temperatures shown in Figure 2.6 we can see that the
samples had to be baked at 600
0
C or higher to obtain clear films which pins the refractive
index variation, Δn, of diluted titania films free of organics at ~ 0.25 (2.35 – 2.1) based on
Figure 3.1.
5000 4500 4000 3500 3000 2500 2000 1500 1000 500 0
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Absorbance a.u.
Wave numbers cm
-1
Quartz blank
Q1_700
0
C (semi_clear)
Q2_700
0
C (clear)
Q3_800
0
C (opaque)
Figure 3.4: FTIR spectra of different sol-gel samples Q1, Q2, Q3 and blank quartz slide.
34
3.2 Characterization of diluted hybrid sol-gel thin films
To obtain a wider variation of refractive index the titania sol had to be doped with silica
or zirconia sols [10, 25]. The ratio of titania to silica and titania to zirconia was fixed at
89%-11% molar ratio as mentioned in Appendix A. On addition of these sols the dwell
time to clear out organics from samples containing silica or zirconia had to be recalibrated
using quartz substrates.
Four different sols were prepared for these studies, each diluted in 1-propanol by
weight in the ratio of 1:1 (see Appendix - A). They are pure titania sol, titania sol with
equimolar water (intended to reduce porosity), titania-silica hybrid sol (89% - 11% molar
ratio) and titania-zirconia hybrid sol (89% - 11% molar ratio). They were all spin coated at
2000 rpm for 1 minute on pre-cleaned silicon and quartz substrates. Different samples
were baked at 700
0
C, 800
0
C and 900
0
C in vacuum for 2 hours respectively. All the
samples coated on silicon substrates were characterized using an ellipsometer and
subsequently rebaked again at 1010
0
C in a nitrogen environment to record changes in
thickness and refractive index. To compare the effect of the baking environment the sols
were also baked in air at 700
0
C for two hours and rebaked at 1010
0
C. The results are
tabulated in Table 3.1 below. Refractive index is mentioned at four different wavelengths
of 630 nm, 980 nm, 1310 nm and 1550 nm respectively.
3.2.1 General trends from thin film data
The refractive index data indicates normal dispersion in all cases. Sintering at 1010
0
C was
necessary for all vacuum baked samples to increase the refractive index further but did
35
Type of
sample
Baking
temperatur
e in
0
C (air
rebake at
1010
0
C)
Baking
duratio
n in
hours
Thicknes
s after
baking
at 1010
0
C in nm
Refractiv
e Index
at 𝝀 =
630 nm
Refractiv
e Index
at 𝝀 =
980 nm
Refractiv
e Index
at 𝝀 =
1310 nm
Refractiv
e Index
at 𝝀 =
1550 nm
Titania
– no
water
700
0
C
(VAC)
800
0
C
(VAC)
900
0
C
(VAC)
700
0
C (AIR)
2 hours
(air
rebake:
1 hour)
116 nm
165 nm
237 nm
120 nm
2.46
2.32
1.79
2.84
2.33
2.02
1.63
2.58
2.29
1.94
1.58
2.49
2.27
1.91
1.56
2.46
Titania
– 1
mole
water
700
0
C
(VAC)
800
0
C
(VAC)
900
0
C
(VAC)
700
0
C (AIR)
2 hours
(air
rebake:
1 hour)
135 nm
163 nm
272 nm
135 nm
2.5
2.6
1.61
2.77
2.36
2.27
1.59
2.49
2.32
2.16
1.58
2.41
2.31
2.12
1.58
2.37
Titania
– Silica
700
0
C
(VAC)
800
0
C
(VAC)
900
0
C
(VAC)
700
0
C (AIR)
2 hours
(air
rebake:
1 hour)
129 nm
145 nm
189 nm
125 nm
2.28
2.18
1.91
2.47
2.2
2.02
1.78
2.34
2.18
2
1.74
2.3
2.17
1.96
1.72
2.29
Titania
–
Zirconi
a
700
0
C
(VAC)
800
0
C
(VAC)
900
0
C
(VAC)
700
0
C (AIR)
2 hours
(air
rebake:
1 hour)
132 nm
181 nm
189 nm
125 nm
2.12
1.71
1.69
2.47
2.07
1.65
1.61
2.38
2.05
1.63
1.59
2.35
2.05
1.63
1.58
2.34
Table 3.1: Refractive index (at four different wavelengths) and thickness data of various
sols spin coated at 2000 rpm for 1 minute on silicon wafers. All samples were baked at
the temperatures mentioned above in column 2 and re-baked in a nitrogen environment
at 1010
0
C for 1 hour.
36
not have any effect on the 700
0
C air baked samples. Baking in air (nitrogen flowing
through the furnace) at 700
0
C gave films with the highest index, 2.5 (at 1310 nm and
1550 nm) and thickness of ~ 115 nm. Titania sol with 1 mole water spun on silicon
andquartz wafers gave films which had macro cracks post baking when observed under a
microscope. As the vacuum bake temperature increased from 700
0
C to 900
0
C the
resultant films were relatively thicker and of lower refractive index. In case of silica sol-
gel films, the addition of water or an increase in the baking temperature (in air) resulted
in a decrease in porosity and an increase in refractive index [21], [38-39]. However, among
the titania sol-gel recipes investigated here that trend is not consistent.
Addition of silica or zirconia to the titania sol resulted in films having a relatively
lower index with a larger drop in refractive index observed for zirconia doped films when
vacuum baked. Other than varying the bake temperature and baking environment,
varying the molar concentration and type of dopant (silica and zirconia) are four different
ways of controlling the refractive index of titania films. The refractive index variation, Δn,
obtained by using the various control parameters is ~ 0.92.
To achieve a refractive index variation between 2 – 2.5 at the communication
wavelengths of 1310 nm and 1550 nm it is ideal to vacuum (followed by sintering at 1010
0
C) or air bake the films at 700
0
C as seen in Table 3.1. With a decrease in refractive index
thicker films are required to confine an optical mode in the vertical direction (normal to
the substrate surface). On increasing the vacuum bake temperature the film thickness
increases accompanied by a drop in refractive index but the thickness and refractive index
values obtained together cannot support an optical mode. One approach was to use
37
multiple spin and bake cycles as done regularly in case of titania sols [31 - 32]. In our case,
it was found that this procedure was viable only if the samples were vacuum baked at 700
0
C for each spin and bake cycle. Samples air baked at 700
0
C did not give crack free films
after the second spin and bake cycle. Thus, there was a need to scale up the thickness (by
avoiding dilution) while obtaining similar refractive indices variations as before.
The quartz samples which were baked under same conditions as listed in Table 3.1
were used to record absorbances using a UV-VIS spectrophotometer. The results are
presented in Figures 3.5, 3.6, 3.7 and 3.8 for samples baked in vacuum at 700
0
C, 800
0
C,
900
0
C and air baked at 700
0
C respectively.
200 400 600 800 1000 1200
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
Absorbance A.U.
Wavelength (nm)
Q13 Titania_no water
Q14 Titania_1 mole water
Q15 Titania_silica
Q16 Titania_zirconia
All samples vacuum baked at 700
0
C for 2 hours
Figure 3.5: Absorbance spectra of quartz samples containing different sol-gel recipes
vacuum baked at 700
0
C for 2 hours.
38
200 400 600 800 1000 1200
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
All samples vacuum baked at 800
0
C for 2 hours
Absorbance A.U
Wavelength (nm)
Q32 Titania_no water
Q33 Titania_1 mole water
Q34 Titania_silica
Q35 Titania_zirconia
Figure 3.6: Absorbance spectra of quartz samples containing different sol-gel recipes
vacuum baked at 800
0
C for 2 hours.
200 400 600 800 1000 1200
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
All samples vacuum baked at 900
0
C for 2 hours
Absorbance A.U
Wavelength (nm)
Q36 Titania_no water
Q37 Titania_1 mole water
Q38 Titania_silica
Q39 Titania_zirconia
Figure 3.7: Absorbance spectra of quartz samples containing different sol-gel recipes
vacuum baked at 900
0
C for 2 hours.
39
200 400 600 800 1000 1200
0
1
2
3
4
All samples air baked at 700
0
C for 2 hours
Absorbance A.U.
Wavelength (nm)
Q21 Titania_no water
Q22 Titania_1 mole water
Q23 Titania_silica
Q24 Titania_zirconia
Figure 3.8: Absorbance spectra of quartz samples containing different sol-gel recipes air
baked at 700
0
C for 2 hours.
The absorbance tail of undoped titania sols without any water extend from 400
nm for films baked at 700
0
C to 600 nm when baked at 900
0
C under vacuum. In all cases
the absorbance of zirconia doped titania sol-gels are slightly less at wavelengths shorter
than 400 nm whereas silica doped titania sol-gels have relatively more absorption
whenthe samples were baked under vacuum at 800
0
C and 900
0
C respectively in
comparison to undoped titania films containing no water.
All samples air baked at 700
0
C had some residual absorbance throughout the
visible spectrum which can be associated with scattering from rough samples surfaces
(and micro cracks – see Appendix C) that were observed when inspected under an optical
40
microscope. Moreover the titania samples containing water had a lot of micro cracks
upon baking in air for 2 hours at 700
0
C.
On rebaking all these samples in air at 1010
0
C it was found that the film surface
degraded slightly with a hazy appearance (to the naked eye) resulting in an absorbance
tail extending throughout the visible wavelengths. This haziness was an artifact of the
high temperature sintering and it was more pronounced on quartz substrates than on
silicon substrates due to the larger thermal mismatch between fused quartz and titania
films
1
(factor of 20 for quartz compared to 5 for silicon). Hybrid titania – silica and titania
– zirconia thin films were relatively less hazy. Since, the quartz samples were mainly used
as an indicator to check for removal of organics from thin films this issue was not
investigated further.
3.3 Characterization of undiluted hybrid sol-gel thin films
The recipes used were same as before with the exception being that the sols were not
diluted using 1-propanol prior to spin coating. Since, the sols were more viscous the dwell
time to remove all organics increased. The sols were spin coated on silicon and quartz
substrates at 2000 rpm for 1 minute and baked at 700
0
C in vacuum for 24 hours. By
opening the furnace flap regularly, the quartz samples were inspected visually to obtain
the minimum dwell time required to drive out the organics. It was found that undiluted
titania sols needed around 5 hours while the ones containing 11% molar silica or zirconia
needed at least 7 hours for all the organics to be removed. Moreover, titania sols
1
The thermal expansion coefficients of silicon, quartz and titania (rutile phase) are ~2.6 X 10
-6
/
0
C, ~5.5 X
10
-7
/
0
C and ~10 X 10
-6
/
0
C respectively.
41
containing water gave cracked films which could not be characterized further. Table 3.2
compares three undiluted sol-gel recipes, one containing only titania sol and the others
containing 11% silica and zirconia respectively. Similar to results obtained in case of
diluted films we see that addition of silica or zirconia gave thicker films with relatively
lower refractive indices wherein the drop in refractive index is more for zirconia doped
films.
On visual inspections of the films under a microscope after baking, the titania films
had a hazy appearance suggesting a rough top layer which scattered light while the hybrid
titania-silica and titania-zirconia films were relatively smooth. Furthermore on sintering
these films in air at 1010
0
C all films developed cracks although the silica doped films
exhibited relatively fewer cracks and no haziness. A reasonable optical model could not
be constructed for titania and zirconia doped films (mean square error was greater than
20) although the change in refractive index after the sintering step were similar to that
obtained for the diluted films. The cracking is a result of the thermal mismatch between
Type of sample
Thickness of
film obtained
in nm
Refractive index
at 1310 nm
Refractive index
at 1550 nm
Titania sol-gel 568 nm 2.06 2.04
Titania – silica
(89% - 11%)
689 nm 1.83 1.83
Titania – zirconia
(89% - 11%)
609 nm 1.70 1.70
Table 3.2: Comparison of thickness and refractive index of three different sols spin coated
on silicon substrates and baked at 700
0
C in vacuum for 24 hours.
42
the underlying substrate and titania sol-gel and also due to the formation of crystallites
of titania in the rutile phase.
To investigate the refractive index and thickness change obtained titania – silica
(89% - 11%) sol-gel was spin coated on silicon wafers at 2000 rpm for 1 minute and baked
at 700
0
C in vacuum for 8 hours giving 557 nm thick films with an index of 1.86 at the
communication wavelengths. On sintering them in air at 1010
0
C the thickness reduced
to 406 nm and the index increased to 2.1 at those wavelengths which was sufficient to
confine an optical mode with silica as the lower cladding and air as the upper cladding.
Thus, undiluted hybrid titania – silica sol-gel is a good choice for a material system to
obtain thin films with sufficient thickness (optically) and relatively high refractive
indices (>2) at communication wavelengths. The refractive index variation that can be
achieved by varying the silica concentration is discussed in the next section.
3.4 Hybrid titania – silica sol-gel films with varying silica
concentrations
Based on the mixing procedure used in [25] different volumes of silica stock solutions
were mixed with titania stock solutions of fixed volume. 10 %, 20 % and 30 % silica sol by
volume were added to titania sol. Converting these volume percentages into molar ratios
as used previously for diluted hybrid solutions gave 11%, 25% and 43% silica by molar
ratio (for calculations see Appendix A) in the hybrid titania – silica solutions. These hybrid
sols were filtered using a 0.2 µm PTFE filter and spin coated on cleaned silicon samples at
2000 rpm for 1 min. They were baked in vacuum at 700
0
C for 9 hours and subsequently
43
sintered in a nitrogen atmosphere at 1020
0
C for 5 hours. Another hybrid titania – silica
sol containing 5% molar silica content was also spun on under similar conditions and
baked in vacuum for 5 hours at 700
0
C followed by sintering at 1020
0
C for 5 hours.
Ellipsometry data depicting refractive index in the visible and infrared
wavelengths for the four different hybrid thin films are plotted in Figure 3.9. The thickness
for the 5%, 11%, 25% and 43% (molar) silica doped thin films recorded were 395 nm, 500
nm, 554 nm and 682 nm respectively. Also films containing 5%, 11% and 25% molar silica
needed an intermix layer, 30 – 10 nm in thickness (consisting of a mixture of hybrid thin
600 800 1000 1200 1400 1600
2.0
2.1
2.2
2.3
2.4
2.5
2.6
2.7
Refractive index, n
Wavelength in nm
5% SiO
2
: TiO
2
11% SiO
2
: TiO
2
25% SiO
2
: TiO
2
43% SiO
2
: TiO
2
Figure 3.9: Refractive index vs wavelength data for four different hybrid titania-silica thin
films containing 5%, 11%, 35% and 43% molar silica content after initial vacuum bake at
700
0
C for 9 hours (5% silica sample 5 hours only) and subsequent air rebake at 1020
0
C
for 5 hours.
44
film and air voids representing micro cracks seen under an optical microscope) to be
included in the Cauchy model to obtain good fits (see Appendix B and C for details). As
expected, with increased silica concentrations the refractive index of the film reduced for
a given wavelength. At 1550 nm the refractive index variation achieved is ~ 0.30 (2.32 –
2.01). With increased silica content the film surface had fewer micro cracks when coated
on silicon or silica wafers.
The 30% by volume silica (~ 43 molar %) doped hybrid sol-gel was also spin coated
at 1000 rpm for 1 min instead to obtain thicker films. Baking the sample at 700
0
C for 7
600 800 1000 1200 1400 1600
1.9
2.0
2.1
2.2
2.3
2.4
t ~ 1050 nm
t ~ 651 nm
t ~ 791 nm
Refractive Index
Wavelength in nm
S78_ 90% TiO
2
-10% SiO
2
- N
2
S80_ 80% TiO
2
-20% SiO
2
- N
2
S79_ 70% TiO
2
-30% SiO
2
- N
2
S79_ 70% TiO
2
-30% SiO
2
- O
2
Different curves for N
2
and O
2
baked
t ~ 954 nm
Figure 3.10: Refractive index vs wavelength data for three different hybrid titania-silica
thin films containing 10%, 20%, and 30% volume silica content. The 30% doped samples
were sintered in nitrogen and oxygen and the curves suggest a lower change of refractive
index for samples sintered in oxygen at 1020
0
C. Film thickness is mentioned on the right.
45
hours gave 1400 nm thick films which reduced to ~ 960 nm upon sintering at 1020
0
C in a
nitrogen atmosphere with refractive index values at 1550 nm increasing from 1.81 to ~2
after the sintering process. Thus, with increasing silica content the recipe automatically
adjusts for the drop in refractive index with an increase in film thickness making it easy to
confine an optical mode at the communication wavelengths.
To compare films sintered in nitrogen and oxygen at 1020
0
C, hybrid titania – silica
solutions containing 10%, 20% and 30% silica by volume were spun on pre cleaned silicon
and 4 µm thermal oxide coated silicon wafers at 1000 rpm for 1 minute. They were baked
for 12 hours under vacuum at 700
0
C. The sample containing 30% silica by volume was
broken into two with one sample sintered at 1020
0
C with oxygen flow and the other
sintered along with the rest of the samples with nitrogen flowing through the tubes. From
ellipsometry data shown in Figure 3.10 we can see that sintering in oxygen rich
atmosphere gave films which had relatively lower index change.
Hybrid titania silica sols containing 10 % silica doped by volume but prepared from
a stock solution using TEOS instead of MAPTMS were also processed under similar
conditions to those used to obtain Figure 3.9. The refractive index and thickness values
were similar to those obtained from sols using MAPTMS. However, samples containing
larger silica percentages by volume (20% or higher) prepared using TEOS gave films that
had large macro cracks and were unfit for further characterization.
Based on the various recipes investigated titania stock solutions doped with silica
stock solutions prepared using MAPTMS as the starting pre cursor was used for device
fabrication reported in the next two chapters. Although formation of microcracks could
46
not be mitigated, sintering in a nitrogen atmosphere at 1020
0
C is the suggested
processing path to obtain films with higher refractive indices.
3.5 Film crystallinity and surface quality
The increase in refractive index of the various titania sol-gel films on sintering at 1020
0
C
is due to the transformation of amorphous titania to the crystalline rutile phase. This
increase in density is always accompanied by a drop in thickness. The degree of
crystallization can be ascertained by X-ray diffraction of the thin film samples coated on
silicon. By cleaving the samples and observing the cleaved edge using a Field emission
scanning electron microscope (FESEM – Hitachi S 4800) the top surface quality and the
bulk film quality were determined.
3.5.1 Film quality
The film morphology and surface quality of thin films obtained from three different sols
were analyzed. Firstly, titania films prepared using a modified sol-gel recipe (see section
2.2) and baked in air and vacuum at 700
0
C were analyzed. Furthermore, hybrid titania –
silica and titania – zirconia sol - gel films baked at the same temperature and
environments were also analyzed to observe any changes due to the addition of dopants.
In all cases the films were spin coated on cleaned silicon substrates and subsequently on
silica (thermal oxide, 4 µm) coated silicon substrates as the latter was the substrate used
for fabricating waveguides and micro-ring resonators using this material system. The
acceleration voltage and beam current was maintained at 1kV and 10 µA, respectively,
for all FE-SEM images captured.
47
Titania films baked at 700
0
C in air have a bumpy top surface and cross section due
to the formation of micro crystallites of titania in the rutile phase (confirmed by X-ray
diffraction measurements). The crystallites vary in size as seen in Figure 3.11 below. When
baked in vacuum the top surface is much smoother and the cleaved cross section reveals
an amorphous titania layer. From Table 3.1, it is clear that air baked films have the
Figure 3.11: SEM images of titania films on silicon substrates baked at 700
0
C. Left –
sample baked in vacuum; Right – sample baked in air.
highest refractive indices. Thus, there is a trade-off between higher refractive index and
film surface quality as titania films with higher refractive index tend to have rougher top
surface and cross section which will contribute to larger optical bulk losses in the material
due to light scattering.
On doping the titania sol-gel with silica or zirconia pre-cursors the resultant films
had a smooth top surface and a cross section depicting that they were amorphous. This
was the case irrespective of the baking environment when baked at 700
0
C as shown in
Figure 3.12 below. Hence, by doping the titania sol-gel with silica or zirconia precursors
the resultant thin films are largely amorphous in nature with smooth top surfaces, albeit
48
Figure 3.12: SEM images of hybrid titania – silica and titania – zirconia films on silicon
substrates baked at 700
0
C. Top left and bottom left– Titania – silica samples baked in
vacuum and air respectively; Top right and Bottom Right –Titania - zirconia samples baked
in vacuum and air respectively. The crack on the film in top right is an artifact from cleaving
the thin films.
Figure 3.13: SEM images of titania films (left) and hybrid titania-silica films (right) on
silicon substrates sintered at 1020
0
C after air baking at 700
0
C initially. Images with similar
roughness profiles were observed even in hybrid – zirconia films that were vacuum baked
initially and subsequently sintered (right) but are not shown here.
49
with a relatively lower refractive index compared to pure undoped titania films. The
doping also should help in alleviating the bulk optical losses due to scattering of light.
In case of hybrid sol-gels, to obtain the maximum refractive index of the film
possible for a give doping level a final sintering step at 1020
0
C was necessary, as seen
earlier. The downside of this process is that the surface smoothness of the thin films
degrades as shown in Figure 3.13 below. Therefore, there is an increase in optical bulk
losses due to scattering on sintering films at 1020
0
C or higher which cannot be mitigated
with the current set of processing steps. However, the size of the particles are much
smaller in case of the hybrid sol-gels and so relative to undoped titania films the optical
bulk losses due to scattering are expected to be comparatively lower.
Undoped titania and hybrid titania – silica solutions spun on 4 µm silica coated
silicon wafers and baked under similar conditions as before also had similar features when
looked at using an FESEM (as seen in Figure 3.13). In conclusion, doping titania sols with
silica or zirconia sols gave amorphous thin films with smooth top surface irrespective of
whether they were baked in vacuum or air. Undoped titania films when baked under
vacuum appeared amorphous. However, on sintering the films at a temperature of 1020
0
C the roughness of film surface increased and while the hybrid sol-gel films showed a
similar morphology as before the undoped titania films formed nano crystallites of
varying sizes. The primary trade-off of the sintering process is the increase in the
refractive index of the film at the cost of the surface quality (more optical bulk scattering
losses).
50
3.5.2 XRD measurements
X-ray diffraction measurements were conducted on thin films made using same recipes
as discussed in the previous section: undoped titania, hybrid titania-silica and hybrid
titania-zirconia sol-gel. <100> Silicon was chosen as the substrate because the XRD peak
for the silicon did not coincide (diffraction angle, 2α, greater than 70
0
C) with peaks
pertaining to various phases of titania. Thin films baked at 700
0
C in air and vacuum were
characterized using a Rigaku Ultima IV thin film diffractometer and a Cu – Kα X-ray source
(λ = 0.154056 nm) with a 10 mm slit. Subsequent to their characterization the films were
sintered at 1020
0
C in a nitrogen atmosphere and re-characterized to observe changes in
the crystal phases. In all cases the scan speed, step size and the range of scattering angles
over which the data was collected was kept fixed for consistency.
Undoped titania films when baked in air (or vacuum) at 700
0
C and sintered at
1020
0
C showed a distinct peak at 27.6
0
corresponding to rutile phase (110) as seen in
Figures 3.14 and 3.15 below. The presence of another peak at 44.6
0
corresponding to
(210) orientation was observed in both cases (although not seen here in Figure 3.14). This
second peak was found only in XRD analysis of rutile powders with particle size between
300 – 700 nm [40]. This suggests that the sintering process gives thin films with particles
sizes in that range which is visually confirmed in Figures 3.11 and 3.13.
Hybrid titania – silica and titania – zirconia films when baked under vacuum at
700
0
C registered a weak characteristic peak at 44.7
0
corresponding to rutile (210) phase
suggesting that they are mostly amorphous with a low degree of crystallization as seen in
51
10 20 30 40 50 60 70
0.0
0.5
1.0
1.5
2.0
2.5
3.0
27.6
o
Normalized counts A.U
Diffraction angle, 2
S21_undoped_titania_airbaked_700
0
C
S21_undoped_titania_sintered_1020
0
C
27.6
o
Figure 3.14: XRD data of undoped titania samples that were baked in air at 700
0
C and
subsequently sintered in air at 1020
0
C.
10 20 30 40 50 60 70
0.0
0.5
1.0
1.5
2.0
2.5
44.6
o
27.6
o
27.6
o
Normalized counts A.U
Diffraction angle, 2
S133_undoped_titania_vacuum baked_700
0
C
S133_undoped_titania_sintered_1020
0
C
Figure 3.15: XRD data of undoped titania samples that were baked in vacuum at 700
0
C
and subsequently sintered in air at 1020
0
C.
52
10 20 30 40 50 60 70
0.0
0.5
1.0
1.5
2.0
2.5
25.6
o
27.8
o
Normalized counts A.U
Diffraction angle, 2
S23_titania_silica_airbaked_700
0
C
S23_titania_silica_sintered_1020
0
C
Figure 3.16: XRD data of titania - silica samples that were baked in air at 700
0
C and
subsequently sintered in air at 1020
0
C.
10 20 30 40 50 60 70
0.0
0.5
1.0
1.5
2.0
2.5
44.7
o
Normalized counts A.U
Diffraction angle, 2
S135_titania_silica_vacuum baked_700
0
C
S135_titania_silica_sintered_1020
0
C
44.7
o
27.7
o
Figure 3.17: XRD data of titania – silica samples that were baked in vacuum at 700
0
C and
subsequently sintered in air at 1020
0
C.
53
10 20 30 40 50 60 70
0.0
0.5
1.0
1.5
2.0
Normalized counts A.U
Diffraction angle, 2
S24_titania_zirconia_airbaked_700
0
C
S24_titania_zirconia_sintered_1020
0
C
25.5
o
27.5
o
Figure 3.18: XRD data of titania - zirconia samples that were baked in air at 700
0
C and
subsequently sintered in air at 1020
0
C.
10 20 30 40 50 60 70
0.0
0.5
1.0
1.5
2.0
44.6
o
Normalized counts A.U
Diffraction angle, 2
S136_titania_zirconia_vacuum baked_700
0
C
S136_titania_zirconia_sintered_1020
0
C
44.7
o
44.7
o
27.6
o
Figure 3.19: XRD data of titania – zirconia samples that were baked in vacuum at 700
0
C
and subsequently sintered in air at 1020
0
C.
54
Figures 3.17 and 3.19. On the contrary, when these hybrid films were baked in air at 700
0
C the XRD measurement depicted peaks at ~ 25.6
0
corresponding to the anatase (101)
phase.
On sintering the air baked hybrid titania – silica and titania – zirconia sol-gel films
at 1020
0
C in air, the peak corresponding to anatase (101) phase disappeared and the
peak due to rutile (110) phase registered as seen in Figures 3.16 and 3.18. The vacuum
baked samples depict peaks of rutile (110) phase dominantly and a reduced peak of (210)
rutile phase suggesting the presence of crystallite particles larger than 100 nm. However,
from FESEM images in Figures 3.12 and 3.13 we see that the particle size is smaller than
100 nm.
Therefore, we can conclude that irrespective of doping or prior baking
environment all sol-gel samples transitioned to the rutile phase when sintered at 1020
0
C
in air. The size of the crystallites can be used to obtain a rough estimate of the bulk
scattering losses. In general, the particle size can be extrapolated from XRD data by
looking at peak broadening of the characteristic peaks and applying Scherer’s formula
[41]. However, SEM images from the previous section suggest that in case of both
undoped and doped titania thin films a variety of particle sizes were observed. The XRD
reference broadening (using silicon dust) was much more than that observed from the
characteristic peaks of each sample and hence the particle sizes could not be deduced
from the XRD data.
55
Chapter 4
Fabrication of ridge waveguides and
characterization of propagation losses from
cutback measurements
Thin films coated on silicon or quartz substrates using different chemical compositions of
titania sols with silica and zirconia sols were characterized elaborately in the previous
chapter. The choice of substrates were based on the different characterization techniques
used to ascertain their optical properties. In order to realize integrated optical devices
using these materials the first condition to be satisfied is that the substrate should have
a relatively lower index than the sol-gel material under use so that light at communication
wavelengths can be guided through the material by the principle of total internal
reflection.
Although quartz substrates satisfy this condition, it was found that titania sol-gel
and hybrid titania sol-gel when coated on quartz substrates and sintered at high
temperatures resulted in thin films with micro cracks spread throughout the film surface.
The main reason for this cracking was due to the thermal mismatch between quartz and
the sol-gel film which was polycrystalline (when sintered at 1020
0
C) in nature. Glass
substrates (excluding fused silica) cannot be used either as they will not be able to
withstand sintering temperatures of 700
0
C or more. The best alternative is to use
thermally grown silica on silicon wafers as the substrate which contain an amorphous
56
thermal oxide (silica) coating on silicon. In order to avoid leakage of light from the core
layer (sol-gel material system) into the silicon substrate, 4 µm thick thermal oxide coated
silicon wafers were used. The orientation of the underlying silicon wafer was <100> as
this made it easier to cleave the wafer post fabrication to prepare the sample for cutback
measurements.
While earlier characterization techniques were used to determine various optical
and structural properties of the sol-gels, in order to gauge their ability to guide light over
large distances optical loss measurements are necessary. Optical losses have to be
minimal (typically a few dB/cm or lower) for this material system to be a viable choice to
transmit information in dielectric waveguides or other integrated optical devices
fabricated using these materials. While UV-VIS spectrophotometry helped determine
absorption in thin films when the light propagated in a direction normal to the film
surface, in case of waveguides the light propagates in a plane parallel to the thin film
surface.
Optical loss characterization can be done using many methods like cutback, Fabry-
Perot fringes and scattering measurements [42 - 44]. Scattering measurements are handy
when the losses are much higher than 10 dB/cm. To observe Fabry-Perot fringes the
waveguide facets need to have high reflectivities which cannot be obtained with sol-gel
material system unless they are coated with a dielectric mirror stack consisting of
alternate layers of a pair of materials having relatively high and low refractive indices
which is expensive. The losses were not expected to be more than 15 dB/cm based on
reports from waveguide losses in titania films deposited by sputtering [17]. Thus, cutback
57
measurements on straight rectangular dielectric waveguides were recorded to determine
the optical quality of these hybrid sol-gel films.
4.1 Fabrication of straight waveguides
Hybrid titania – silica sol-gel with 11% (molar) silica was the material system that was used
for the cutback measurements. The various steps involved in fabricating a hybrid titania-
silica / SiO 2 /Si dielectric waveguide are as follows:
a) On a pre-cleaned 4 µm silica coated silicon wafer (3 cm X 3 cm) hybrid titania-silica
sol-gel was spin coated at 2000 rpm and vacuum baked (~ 100 mT chamber
pressure) for 10 hours at 700
0
C. This dwell time was chosen to remove the
organics completely from the thin film.
b) The films were characterized for their thickness and refractive index using the
ellipsometer and profilometer.
c) A 5 cm long straight waveguide pattern having different widths defined on a
positive photo mask were fabricated using standard photolithography image
reversal technique on the MJB-3 Karl Suss contact aligner.
d) Samples were post-baked at 120
0
C for 90 seconds and blank exposed (dosage >
200 mJ/cm
2
) prior to developing in AZ400K (1:4 developer) to achieve image
reversal.
e) 40 nm chrome was deposited using electron beam evaporation forming the hard
mask after lift-off (AZ5214E photoresist) using acetone.
58
f) To transfer the pattern onto the hybrid sol-gel layer the samples were dry etched
using reactive ion etching (RIE) in a CF 4 and O 2 plasma. This etch process was
calibrated and optimized for different hybrid sol-gel compositions.
g) Post RIE the remaining chrome was stripped using chrome etchant (CR-7) and the
samples were sonicated.
h) The samples were cleaved to form near perpendicular input and output facets
with respect to the waveguides.
The procedure above has been shown schematically in Figure 4.1 below. Initially the
pattern was defined using S1813 as the photoresist and etch mask. It was found that using
a hard mask like chrome as the etch mask required less thickness resulting in more
anisotropic etch profiles and gave relatively smoother surface quality of etched side wall.
To prepare the samples for loss characterization they were cleaved since the quality of
the end facet was found to be much better than when the samples were diced.
Figure 4.1: Stepwise procedure to fabricate hybrid titania-silica sol-gel waveguides (Left
to Right).
59
4.2 Reactive ion etching results
Dry etching typically introduces roughness on the etched surface (waveguide side walls)
which can be detrimental in guiding light as it introduces scattering losses which increase
the total propagation loss of a dielectric waveguide [42]. Thus, it is necessary to develop
an etch recipe that gives reasonably smooth side walls with good etch selectivity with
respect to the hard mask. Based on etch recipes developed to etch sputtered titania and
silica films [17], [20], [45 - 47] and silica and titania sol-gel films [48 - 49], reactive ion
etching using a combination of carbon tetra fluoride (CF 4) and oxygen (O 2) gases were a
good starting point.
RF power
in Watts
Chamber
Pressure in
mTorr
Gases used
Etch rate in
nm / min
40 100 CF 4 ~ 25
70 100 CF 4 ~ 60
100 100 CF 4 ~ 155
100 50 CF 4 ~ 135
100 25 CF 4 ~ 115
70 100 CF 4 + O 2 ~ 100
Table 4.1: Various control parameters used to determine optimal etch recipe. In all cases
the duration of etch was maintained at 4 minutes.
Patterns were defined on hybrid sol-gel films that were coated on silicon wafers
and baked in vacuum for 7 hours at 700
0
C using the steps mentioned in the previous
section to obtain etch masks. Samples with photo resist as a mask was etched under
various conditions listed in Table 4.1 above. With CF 4 as the etch gas we can see that the
60
etch rate dropped as the RF power reduced or if the chamber pressure was reduced. The
etched samples were observed under an FESEM to determine film surface quality. On
addition of oxygen, which behaves as a good carrier of etched products, the etch rate
increased while keeping other control parameters fixed as seen in Figure 4.2 below.
Figure 4.2: Cleaved cross section of waveguide with photoresist as mask compared for two
samples; Left – Etched in CF 4 only. Right – Etched in CF 4 + O 2 environment. More
photoresist is consumed in the presence of oxygen.
Similar etch rates of ~ 100 nm / min were obtained on samples etched with
chrome as mask and CF 4 + O 2 as etch gases (equal flow rates). With increasing silica
concentration in the samples and same etch conditions the etch rate marginally increased
from ~ 80 nm / min to more than 120 nm / min.
To control the ratio of oxygen and CF 4 flow rates (CF 4: O 2 :: 35 sccm: 3 sccm) some
samples were etched in an ICP RIE dielectric etch system under varying chamber
pressures and RF power. It was found that the etched surface was much smoother but
the etch rate dropped to ~ 50 nm / min even with an RF power of 200 watts. However,
longer etch duration due to reduced etch rate resulted in removal of hybrid sol-gel below
the chrome layer by undercutting as seen in Figure 4.3 below.
61
Figure 4.3: Cleaved cross section of etched waveguide with chrome as mask etched in CF 4
+ O 2 (35 & 3 scccm), 100 mT, 200 W. Vertical etched side walls and undercutting of sol-gel
below the chrome mask at the waveguide edges can be seen.
Thus, control of flow rates of the etch gases is very important to obtain relatively
smooth etched side walls and with further optimization (to avoid undercutting) this recipe
can be used. Due to difficulty in availability of ICP etch system on a regular basis, the old
etch recipe listed in Table 4.1 (70 W RF power, CF 4 + O 2) was employed with chrome as
the etch mask to fabricate dielectric waveguides. The side walls of etched waveguides
were not vertical like the ones showed in Figure 4.3, the waveguide cross section
appeared more trapezoidal.
Quality of surface roughness was measured using an Atomic Force Microscope
(AFM) from Digital Instruments (Multimode scanning probe microscope used as an AFM)
62
in tapped mode. A 5 µm X 5 µm square area was scanned. The root mean square (r.m.s)
roughness was found to be around 30 nm. To improve the surface quality the same
sample was wet etched in 1:4 diluted buffered hydrofluoric acid solution for 1 minute. On
re-measuring the sample r.m.s roughness dropped by half to ~ 15 nm as shown in Figure
4.4 below.
Figure 4.4: AFM image of etched surface before (left) and after (right) wet etching for 1
minute in diluted HF; Improvement in surface roughness is confirmed visually.
Increasing the wet etch duration or concentration of hydrofluoric acid (with
respect to amount of deionized water) resulted in damaged waveguides with plenty of
cracks visible under SEM as seen in Figure 4.5. Although, employing a wet etch process
for hybrid sol-gel waveguides spin coated on silica coated silicon wafers can be
detrimental as HF consumes the silica quite easily, the above results suggest that by
carefully controlling the dilution and etch duration we can alleviate RIE induced roughness
on waveguide side walls without compromising structural integrity of the sol-gel and
underlying silica layer. For the waveguides fabricated for loss measurements the wet
63
etching process was not employed, therefore, the expected r.m.s roughness is around 30
nm based on AFM measurements made on thin films.
Figure 4.5: FESEM images of wet etched surface after 1 minute (left) and 2 minute (right)
etch durations; Longer etch duration results in peel off of photoresist and sol-gel layers.
4.3 Cutback loss measurements
Straight waveguides were patterned with chrome as the etch mask on hybrid titania silica
sol-gel samples as described in section 4.1. A sample area of 2.5 cm X 2.5 cm contained ~
Figure 4.6: Top view of etched waveguides after removal of chrome mask using a 50X
objective. The waveguide widths decrease from 10 – 2 µm from top to bottom.
64
80 groups of waveguides, each group consisting of 6 waveguides of different widths (10
µm, 6 µm, 5 µm, 4 µm, 3 µm and 2 µm respectively) as shown in Figure 4.6. Four
waveguides of widths 5, 4, 3 and 2 microns were used for the measurements, although
the 5 µm waveguide supported multiple – modes in the horizontal direction. The sample
was cleaved using a scriber to obtain cleaved input and output facets. The input facet was
kept fixed throughout the measurement and data from 20 groups of waveguides were
collected for three different device lengths by repeated cleaving after recording data for
one length (see Figure 4.7).
Figure 4.7: Top view and cross sectional view of cleaved sample after forming initial input
and output facets. Subsequent cleave positions are shown.
For the waveguide cross section shown quasi TE and TM modes supported for
different waveguide widths were obtained using a complex full vectorial finite difference
generic mode solver using Olympios-Temp Selene software package from C2V [50]. The
simulation grid consisted of 128 X 128 points and a simulation window of 10 µm X 6 µm
65
centered at the core encompassed the core, upper and lower claddings. The wavelength
was maintained at λ = 1550 nm for all simulations and the results are shown in Table 4.2
Waveguide
width in µm
Effective
index of
mode, n eff
Type of mode:
quasi TE or quasi
TM
Number of
lobes in
horizontal
direction
2 µm
1.6254
1.5471
1.5033
TE
TM
TE
1
1
2
3 µm
1.647
1.5908
1.5634
1.5142
1.4948
TE
TE
TM
TM
TE
1
2
1
2
3
4 µm
1.6549
1.6228
1.57
1.5685
1.5404
1.4926
1.4869
TE
TE
TM
TE
TM
TM
TE
1
2
1
3
2
3
4
5 µm
1.6586
1.5733
5 more higher
order modes
TE
TM
1
1
Table 4.2: Finite difference simulation for waveguides of different widths indicating
effective indices and number of lobes parallel to film surface. 5 higher order modes for
the 5 µm waveguide are not shown in the table.
66
above. The effective index of the fundamental quasi TE and TM modes increases as the
width of the waveguide increases as expected. Likewise the number of higher order
modes increases with larger waveguide widths. In all cases, the index contrast and
waveguide height used allow confinement of only one mode vertically. The mode profiles
for the fundamental and first higher order quasi TE and TM modes of a 3 µm wide
Figure 4.8: Mode field profiles of fundamental and first higher order quasi TE (top) and
quasi TM (bottom) modes from the simulations performed. The black (blue in reality)
borders are artifacts that appeared during image transfer.
67
waveguide are shown in Figure 4.8 above. Data from these simulations were used in
calculating 3D scattering losses from waveguides with rough side walls which is discussed
in section 4.4 below.
The cutback measurement setup is shown schematically in Figure 4.9 below. Light
from a single mode laser (1550 nm) was coupled into a lensed fiber which was positioned
at the input facet of the waveguide. Light from the output facet was collected using a 10X
lens (0.25 N.A) and power levels recorded using a photo detector. The lensed fiber,
sample, objective lens and detector were all placed on motion controlled (3D) stages. An
iris and polarizer (positioned perpendicular to substrate surface, TM polarization) were
placed in between the collection lens and detector to minimize stray light and extract
polarization sensitive data respectively. By placing a mirror (see Appendix C) in between
the iris and detector the mode profile was captured using an IR camera (Find IR scope).
On moving the input lensed fiber along the waveguide width higher order modes could
be excited which was less bright compared to the fundamental mode excited in the
waveguides for a fixed laser power. The fiber had to be positioned skewed towards one
of the waveguide edges away from the center. Figure 4.10 shows the mode profile
recorded using an IR camera of the fundamental mode and a higher order mode
Figure 4.9: Cutback loss measurement test setup schematic.
68
containing 3 lobes. For the 5 µm waveguide the dominant mode had two lobes along the
width of the guide while the rest of the guides showed modes with a single lobe when the
lensed fiber was positioned at the center aligned symmetrically to the waveguide width.
Figure 4.10: Mode profiles captured using an IR camera. Left – dominant mode observed
in waveguides with width 2, 3 and 4 µm. Right – Mode with three lobes seen when lensed
fiber is moved away from waveguide axis.
A tunable laser from New Focus set to 1550 nm was used as the source and the input
power levels were recorded prior to mounting the device. The lensed fiber was positioned
~ 3 µm away from the input waveguide facet to maximize coupling into the waveguide at
1550 nm. The coupling was varied by repeatedly moving the lensed fiber stage in all three
directions until a consistent repeatable power value was observed for each data point.
Prior to each measurement the mode profile was confirmed with the help of a mirror
placed in between the iris and detector and IR camera. From power measured for two
different waveguide lengths the propagation loss in dB/cm can be obtained by:
𝛼 =
10log(
𝑃 2
𝑃 1
)
(𝐿 2
− 𝐿 1
)
(4.1)
69
where P 2 (P 1) is the power measured for waveguide length L 2 (L 1) cm. By obtaining data
for three different lengths the loss can be extracted from the slope of the line. Cutback
measurement data is shown in Figure 4.11 below. Each data point for a given width and
6 8 10 12 14 16 18
-35
-30
-25
-20
-15
-10
Normalized Output power in dB
Device length in mm
5 m width
4 m width
3 m width
2 m width
Figure 4.11: Cutback loss measurement data for waveguides of 4 different widths, TM
polarization at a wavelength of 1550 nm.
Waveguide width in µm Propagation loss in dB/cm
5 ~ 8.4
4 ~ 8.5
3 ~ 11.4
2 ~ 12.6
Table 4.3: Propagation loss values for waveguides of different widths at a wavelength of
1550 nm.
70
length is an average of data recorded for 9 - 16 different waveguides. The propagation
losses extracted from the figure using equation (4.1) are presented in Table 4.3 below.
Firstly, as the waveguide width reduces the amount of power coupled into the
waveguide reduces resulting in lower power levels for smaller widths. Although the data
for the 5 µm width does not follow this trend as the dominant mode had two lobes.
Secondly, as the width reduces below 3 µm the propagation losses increase from ~8.5
dB/cm to ~ 12 dB/cm for TM polarization suggesting that scattering losses increase as the
width reduces.
The primary source of error in this measurement is the fluctuation of input laser
power. Although the laser was left on for an hour prior to beginning the measurement,
over time there was a drift in the optical power output. By averaging many data points
this error was minimized and the entire measurement was completed in one sitting. While
the detector was placed at least 50 cm away from the collection lens, stray light above
the waveguide and that coupled into the lower cladding at the input end reached the
detector past the iris. For shorter device lengths this can affect the power values recorded
and skew the loss values towards the higher side. This can be mitigated by using a higher
numerical aperture collection lens (> 0.4) in conjunction with the iris.
4.4 Propagation loss analysis
In case of straight waveguides the net propagation loss, α, is a sum of the material loss
(absorption, α abs) and scattering (α sc) losses (see equation 4.2). Material inhomogeneity,
crystallinity, thin film micro structure (scattering from grain boundaries) contribute to
71
absorption losses while scattering from surface roughness induced during waveguide
fabrication contribute to the scattering losses. In general
𝛼 = 𝛼 𝑎𝑏𝑠 + 𝛼 𝑠𝑐
(4.2)
It would be useful to separate the contributions due to absorption and scattering so that
the material synthesis and waveguide fabrication processes can be optimized further to
reduce the net propagation losses to the lowest values possible. 2D [51-52], [42] and 3D
[53], [18] scattering loss estimates have been reported for dielectric waveguides. The
following sections evaluate scattering losses using 2D and 3D loss calculations.
4.4.1 2D Scattering loss calculations
To calculate the scattering loss for the waveguide cross section shown in Figure
4.12 the rectangular waveguide was approximated as a dielectric slab of varying widths,
infinite in the vertical direction (x) and air cladding with rough sidewalls [42]. N
I
eff, N
II
eff
and N
III
eff are obtained by using effective index calculations using slab TE and TM
equations from [54]. For a symmetric slab of width d, with core refractive index n 1 and
cladding refractive index n 2 from the boundary conditions of the fields evaluated at the
side wall, the characteristic equations for TE and TM waves can be written as:
tan(𝑘𝑑 ) =
2𝑘𝛿
𝑘 2
− 𝛿 2
for TE waves and (4.3)
tan(𝑘𝑑 ) =
𝑛 1
2
𝑛 2
2
2𝑘𝛿
𝑛 2
4
𝑘 2
−𝑛 1
4
𝛿 2
for TM waves (4.4)
where 𝑘 = √𝑛 1
2
𝑘 0
2
− 𝛽 2
and 𝛿 = √(𝑛 1
2
− 𝑛 2
2
) 𝑘 0
2
− 𝑘 2
(4.5)
72
and k 0 and β are the free space wave number and propagation constant of TE or TM slab
mode. Using the effective index, n eff, calculated from equations (4.3) – (4.5) the scattering
losses α sc in dB/cm can be obtained by using the expression derived in [42, 51, 52]:
𝛼 𝑠𝑐
= 4.343 ∗ 10
−2
𝜎 2
√2𝑘 0
𝑑 4
𝑛 𝑒𝑓𝑓 𝑔 𝑓 𝑒 (4.6)
where σ is the r.m.s roughness of the side wall, g and f e are functions of the effective
core/cladding indices, wavelength and correlation length L c [51-52].
Figure 4.12: Conversion of rectangular waveguide into a slab waveguide with rough
sidewalls for 2D scattering calculations.
For TE modes equation (4.3) is solved first followed by (4.4) and vice versa for TM modes.
Table 4.4 shows calculated scattering losses for the fundamental TE and TM mode
supported by the waveguide in Figure 4.12 with widths 2, 3, 4 and 5 µm and a correlation
length of 50 nm for two different side wall roughness values of 30 nm and 15 nm. The
roughness values chosen were based on AFM data reported in Chapter 3. We see that as
the width of the waveguide reduces below 3 µm the scattering losses increase
exponentially. TE modes have larger scattering losses compared to TM modes. The
effective index values recorded are very similar to those obtained in Table 4.2. With an
increase in correlation length the scattering losses will increase.
73
Waveguide width
in µm
Effective index
of slab mode,
n eff
Scattering losses in
dB/cm for σ = 30
nm
Scattering losses in
dB/cm for σ = 15
nm
2
TE, 1.6251
TM, 1.5469
13.6
9.5
3.4
2.4
3
TE, 1.6466
TM, 1.5634
4.6
3.5
1.1
0.9
4
TE, 1.6546
TM, 1.5701
2
1.6
0.5
0.4
5
TE, 1.6583
TM, 1.5734
1.1
0.9
0.3
0.2
Table 4.4: (2D) Scattering losses in dB/cm for waveguides of different widths using two
different r.m.s roughness values and a correlation length of 50 nm.
However, in applying 2-D scattering loss theory to a rectangular waveguide the
radiation profile assumed is incorrect resulting in erroneous calculation of power loss
[53]. This results in the scattering losses being overestimated from 2D calculations.
4.4.2 3D Scattering loss calculations
3-D scattering loss calculations were performed using the procedure mentioned in [53].
Since the index contrast is relatively high (1.85 – 1.44 = 0.41) the scattering loss analysis
for high index contrast waveguides were used which uses the volume current method.
The rough waveguide core is replaced by a smooth waveguide which does not radiate and
a rough side wall that radiates which is further broken into rods of the height of the
waveguide and infinitesimal cross section times a roughness array factor that relates the
far field of a single rod to that of the roughness as shown in Fig 4.13.
74
The E-field generated by the current element due to roughness is given by:
𝐸 𝑟𝑜𝑢𝑔 ℎ
⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗
= 𝐸 𝑒𝑙𝑒𝑚𝑒𝑛𝑡 ⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗
𝐹 𝑟𝑜𝑢𝑔 ℎ
(4.7)
Figure 4.13: Decomposition of radiation problem into current element J rough with array
factor F rough [53].
In calculating E element (or E pol) for high index contrast waveguides the dyadic
Green’s function for a one layer medium was used where for the fundamental TE like and
TM like mode polarizations the current profiles at the rough sidewall were assumed to be
a cosine or sine function as shown in Figure 4.14 below.
Figure 4.14: A: Rectangular waveguide depicting radiating sources. B: One layered
medium of interest for scattering calculations (Note: x c = x, y c = -z and z c = y). C: Vertical
profile of the current sources used to approximate the shape of the first TE-like and TM
like mode at the rough boundary [53].
Using the Green’s function computed in [53] the electric field,𝐸 𝑝𝑜𝑙 ⃗⃗⃗⃗⃗⃗⃗
, due to different current
sources at the rough boundary can be calculated by
𝐸 𝑝𝑜𝑙 ⃗⃗⃗⃗⃗⃗⃗⃗
(𝑟 𝑐 ⃗⃗ ) = 𝑖𝜔𝜇 𝐺 ̿ (𝑟 𝑐 ⃗⃗ ,0
⃗
) .𝐽 𝑝𝑜𝑙 ⃗⃗⃗⃗⃗⃗
(4.8)
75
where 𝐺 ̿ (𝑟 𝑐 ⃗⃗ ,0
⃗
)is the dyadic Green’s function for one layered media, ω is frequency used
and µ is the permeability of free space. For e.g. the current source of the z component of
the electric field at the rough boundary is given by
𝐽 𝑝𝑜𝑙 ,𝑧 ⃗⃗⃗⃗⃗⃗⃗⃗⃗
= 𝐽 𝑝𝑜𝑙 ,−𝑦 𝑐 ⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗
= −𝑖𝜔 𝜖 0
(𝑛 𝑐𝑜𝑟𝑒 2
− 𝑛 𝑐𝑙𝑎𝑑𝑑𝑖𝑛𝑔 2
)𝜐 𝑦 𝑐 ̂ (4.9)
where 𝜐 2
is the normalized square of the E-field, 𝜑⃗ (x’,0), at the side wall integrated along
the height of the waveguide given by
𝑣 = √∫ |𝜑⃗ (𝑥 ′
,0)|
2
𝑑𝑥 ′
𝑎 2
−
𝑎 2
(4.10)
using 4.7 – 4.10 the ensemble average of the roughness far- field Poynting vector is given
by
〈𝑆
𝑟𝑜𝑢𝑔 ℎ
〉 = |𝐹 𝑠 ℎ𝑎𝑝𝑒 |
2
〈|𝐹 𝑟𝑜𝑢𝑔 ℎ
|〉
2
𝑆
𝑝𝑜𝑙 (4.11)
where 𝑆
𝑝𝑜𝑙 is the Poynting vector calculated using equation 4.8 and F shape and F rough are the
array factors for different vertical shapes of field (sine or cosine typically) and ensemble
average of the roughness power array factor using an exponential auto correlation
function respectively [53]. The scattering loss per unit length of waveguide can be
calculated by integrating equation (4.11) in spherical coordinates which was done in
Matlab. The mode propagation constant β, parameter υ and the power ratio of z to y and
z to x components ϒ zy and ϒ zx were obtained using the mode solver used earlier. For a
correlation length, L c of 50 nm and r.m.s roughness values of 30 nm and 15 nm the 3D
76
scattering losses were computed for the waveguide in Figure 4.12 with widths 2, 3, 4 and
5 µm to compare with 2D scattering calculations as shown in Table 4.5 below.
We can see that the scattering losses computed by 2D calculations are
approximately four times those obtained from 3D calculations. TE modes suffer more
scattering compared to TM modes like before. The scattering losses for waveguide widths
of 3 µm or more are negligible for both polarizations.
Waveguide
width in µm
υ
2
for TE
modes
ϒ zy
υ
2
for TM
modes
ϒ zx
α sc in dB/cm
for TE modes
σ = 30nm (σ =
15 nm)
α sc in dB/cm
for TM modes
σ = 30nm (σ =
15 nm)
2 µm 1.96 X 10
7
3.594 1.87 X 10
7
0.189 3.6 (0.9) 2.4 (0.6)
3 µm 6.23 X 10
6
3.527 6.90 X 10
6
0.252 1.1 (0.3) 0.8 (0.2)
4 µm 2.74 X 10
6
3.375 3.21 X 10
6
0.225 0.5 (0.12) 0.4 (0.1)
5 µm 1.44 X 10
6
3.281 1.76 X 10
6
0.209 0.26 (0.07) 0.22 (0.06)
Table 4.5: (3D) Scattering losses in dB/cm for waveguides of different widths using two
different r.m.s roughness values and a correlation length of 50 nm.
For TM like modes that were measured using cutback method based on 2D and
3D scattering loss analysis the scattering losses are below the error bar observed for all
waveguide widths suggesting that losses measured are predominantly material losses.
However, loss values recorded in Table 4.3 imply that there is scattering as the values
recorded for 2 and 4 µm widths are different.
The prime reason for low scattering losses calculated using 3D calculations is that
the correlation length of 50 nm was an assumed value (typical for RIE process used) and
could not be extracted explicitly from the AFM measurements. To resolve this issue, the
effective index of the fundamental TM slab mode, the power confinement factor in the
77
core and the normalized intensity at the side wall surface were calculated for the
waveguides measured as shown in Table 4.6. We can see that as the waveguide width
increases, the effective index and power confinement increase while the intensity at the
side wall reduces. Lower power confinement will result in lower absorption losses for
waveguides of smaller widths while lower field intensity at waveguide surface results in
reduced scattering losses for waveguides with larger widths.
Waveguide width in
µm
Effective index of
slab mode (TM),
n eff
Γ , power confinement
in slab
Normalized intensity
at side wall
2 1.5469 0.988 6.787 X 10
-2
3 1.5634 0.9961 3.252 X 10
-2
4 1.5701 0.9982 1.908 X 10
-2
5 1.5734 0.9991 1.168 X 10
-2
Table 4.6: Effective index of TM modes, power confinement and normalized intensity at
side wall for waveguides of different widths and air cladding.
Comparing data for 4 and 2 µm widths from Tables 4.2 and 4.6 we can write the following
relation between scattering, absorption and total losses for the two widths respectively:
8.5 = 𝛼 𝑎𝑏𝑠 + 𝛼 𝑠𝑐
(4.12)
12.6 = 0.99𝛼 𝑎𝑏𝑠 + 3.56𝛼 𝑠𝑐
(4.13)
Waveguide
width in µm
Total loss in
dB/cm
𝜶 𝒂𝒃𝒔 in
dB / cm
𝜶 𝒔𝒄
in dB/
cm
𝜶 𝒔𝒄
in dB/ cm
from 2D
calculations
𝜶 𝒔𝒄
in dB/ cm
from 3D
calcultions
2 12.6 ~ 6.8 ~ 5.8 ~ 9.5 ~ 2.4
4 8.5 ~ 6.9 ~ 1.6 ~ 1.6 ~ 0.4
Table 4.7: Separating contribution to optical losses from absorption and scattering.
78
Solving the above system of simultaneous equations we can obtain approximate values
for scattering and absorption as listed in Table 4.7.
Based on the extensive scattering analysis presented hybrid sol-gel straight
waveguides characterized using cutback measurements have material absorption losses
in the range of 6 - 8 dB/cm and width dependent scattering losses in the range of 2-7
dB/cm. By increasing the correlation length to values of 200 nm or more the scattering
loss values from 3D calculations are close (~ 4.6 dB/cm for a 2 µm width waveguide) to
that obtained from the analysis presented in Table 4.7. The correlation length needs to
be extracted carefully from AFM measurements to pin down the scattering losses from
3D scattering loss calculations.
From FESEM images and XRD data of hybrid titania – silica sol-gel thin films
presented earlier (see Figure 3.12 and 3.17) we can conclude that the films are mostly
amorphous and the particle sizes are sub 50 nm. Using the estimated bulk scattering
losses of 6 – 8 dB/cm the thin film layer can be modelled to account for these losses.
Figure 4.15: Modelling the hybrid titania – silica thin films with silica spheres of radius 10
nm present in an ambient (porous) medium with refractive index 1.93.
79
Undoped titania films baked in vacuum for 9 hours at 700
0
C on silicon substrates
gave refractive index values of ~ 1.93 at 1550 nm. For a hybrid film containing 11 % silica
the maximum refractive index recorded when sintered at 1020
0
C was 2.2 at 1550 nm.
Thus, a film with same silica % and refractive index 1.85 (after vacuum bake prior to
sintering) is expected to have ~ 37% porosity using equation 1.1. From Mie theory,
expressions for absorption and scattering by a sphere have been derived in [55]. Using
Matlab codes written in [56] silica particles with sphere diameter assumed to be ~ 10 nm
and volume fraction of ~ 5 – 7 % gave bulk scattering losses in the range estimated
between 6 – 8 dB/cm.
While there is still some ambiguity on the separation of absorption and scattering
losses in the dielectric waveguides due to unknown correlation lengths, to the best of our
knowledge, this is the first demonstration of optical characterization of hybrid titania –
silica sol-gel rib waveguides containing relatively large titania content (~ 90% molar). From
scattering loss analysis and AFM data we can see that by improving the etch process and
reducing the r.m.s roughness scattering losses of < 1 dB/cm can be obtained for 2 µm
wide waveguides. Although, material losses have been explained and modelled in terms
of bulk scattering from silica spheres further research and experiments are required to
isolate the contributions from the two loss mechanisms and the synthesis process needs
to be optimized to minimize the material absorption.
80
Chapter 5
Demonstration of integrated micro ring
resonators using hybrid titania-silica sol-gel
This chapter demonstrates a laterally coupled microring resonator using hybrid titania
silica sol-gel with 11% molar silica. Microring resonator theory, fabrication and device
characterization details are presented below. Micro disk resonators that were fabricated
and tested using the same material system are discussed in the end.
5.1 Microring resonator theory
Microring and microdisk resonators are basically wavelength selective devices which
work on the principle of interference between the optical modes in a straight waveguide
and a ring waveguide. They have been fabricated with various material systems (e.g.: Si-
SiO 2, Si 3N 4-SiO 2, polymers) and serve as building blocks in wavelength division
multiplexing (WDM) systems as filters, modulators etc. They have also been used in
sensing applications like sensing gases, humidity, displacement, proteins, DNA etc [18 -
20], [26], [57 - 62].
The analysis of a ring resonator coupled to a single waveguide is very similar to
the one described in [58, 63-64]. A schematic diagram of a ring resonator coupled to a
single waveguide is shown above in Figure 5.1. In this scheme, we assume that a single
81
Figure 5.1: Schematic of a dielectric waveguide coupled to a ring resonator.
mode is excited in the resonator and the coupling between the ring and the waveguide is
loss-less. The interaction between the ring and the waveguide can be described by
coupled mode theory with a matrix relation given below:
|
𝑏 1
𝑏 2
| = |
𝑡 𝑖𝜅
𝑖𝜅
𝑡 ||
𝑎 1
𝑎 2
| (5.1)
where, κ and t are the electric field coupling coefficient and transmission coefficients
respectively. The complex mode field amplitudes a i and b i (i =1, 2) are normalized such
that their squared magnitude corresponds to the modal power. The coupling matrix is
unitary (assuming lossless coupling), therefore:
|𝜅 2
| + |𝑡 2
| = 1 (5.2)
In the following we will choose the input wave a 1 = 1 so that all field amplitudes are
normalized to a 1.The transmission around the ring is given by:
82
𝑎 2
= 𝑎 𝑒 𝑖𝜃
𝑏 2
; 𝑎 = 𝑒 −𝛼𝐿
2
; 𝜃 = 𝛽𝐿 ; 𝛽 =
2𝜋 𝑛 𝑒𝑓𝑓 𝜆 (5.3)
where, β, α, a, λ, n eff and L are the mode propagation constant, intensity loss coefficient,
E-field transmission for one roundtrip in the ring, wavelength of light, effective index of
refraction of the ring mode and the path length of the ring respectively. L = 2πR in case of
rings with radius R and L = 2(πR+ L s) in case of race track rings that contain two straight
sections of length L s. Using the relations from (5.1), and (5.3) we get
𝑏 1
𝑎 1
=
𝑡 −𝑎 𝑒 𝑖𝜃
1−𝑎𝑡 𝑒 𝑖𝜃
𝑎𝑛𝑑
𝑎 2
𝑎 1
=
𝑖𝜅𝑎 𝑒 𝑖𝜃
1−𝑎𝑡 𝑒 𝑖𝜃
(5.4)
The transmission past the resonator in the input waveguide is:
|𝑏 1
|
2
|𝑎 1
|
2
=
𝑎 2
+ |𝑡 |
2
−2𝑎 |𝑡 |cos (𝜃 )
1+ 𝑎 2
|𝑡 |
2
−2𝑎 |𝑡 |cos (𝜃 )
(5.5)
while the total circulating power is:
|𝑎 2
|
2
|𝑎 1
|
2
=
𝑎 2
(1− |𝑡 |
2
)
1+ 𝑎 2
|𝑡 |
2
−2𝑎 |𝑡 |cos (𝜃 )
(5.6)
At resonance, the phase shift experienced by the mode after one round trip, (θ) = m2π,
where m is some integer. Thus, at resonance, (5.5) and (5.6) reduce to:
|𝑏 1
|
2
|𝑎 1
|
2
=
(𝑎 − |𝑡 |)
2
(1− 𝑎 |𝑡 |)
2
&
|𝑎 2
|
2
|𝑎 1
|
2
=
𝑎 2
(1− |𝑡 |)
2
(1− 𝑎 |𝑡 |)
2
(5.7)
From the first part of equation (5.7) we find that when a = |t|, i.e. when the
internal losses in the ring are equal to the coupling losses, |t|, the transmitted power
83
drops down to zero, i.e., |b 1|
2
= 0. This condition is known as critical coupling, and is due
to perfect destructive interference in the outgoing waveguide between the transmitted
field ta 1 and the internal field of the resonator κa 2 coupled into the output waveguide.
On the other hand, if a < |t|, the condition is called as under-coupling and if a > |t|, it is
called over-coupling. Please note that it is assumed that the straight and curved
waveguides are phase matched in the above derivation, otherwise, κ and t are not real
and this will shift the resonance to some other value, say, ϕ’ where ϕ’ = ϕ plus the phase
mismatch. A typical transmission spectrum of a ring resonator coupled to a single
waveguide is shown below in Figure 5.2.
Figure 5.2: Transmission spectra of a microring resonator.
5.1.1 Parameters of the ring resonator
The Full Width at Half Maximum (FWHM), Δλ, of the transfer function has been
derived in [58], [64] given by:
∆𝜆 =
𝐹𝑆𝑅 [sin
−1
(
1−𝑎 .𝑡 2√𝑎 .𝑡
) ]
𝜋 (5.8)
84
It is also called the bandwidth at resonance (Δλ). The separation between consecutive
resonance dips is known as the Free Spectral Range (FSR) and it can be calculated from
the expression:
𝐹𝑆𝑅 =
𝜆 2
𝑛 𝑔 𝐿 ; 𝑛 𝑔 =
𝜕𝛽
𝜕𝑘
= 𝑛 − 𝜆 𝜕𝑛
𝜕𝜆
(5.9)
where n g is the group index (or effective index if the dispersion is negligible over a
reasonable spectral range). The figure of merit for a resonator is the quality factor Q
(loaded), defined by:
𝑄 =
𝜆 𝛥𝜆
=
𝐹𝑛
𝑔 𝐿 𝜆 (5.10)
where, F is called the Finesse of the ring. The Finesse represents the number of round
trips made by light in the ring. The Quality Factor determines the number of oscillations
of the field before the circulating energy is depleted to 1/e of the initial energy. The
fundamental measurands that characterize a ring resonator are its bandwidth Δλ, free
spectral range FSR, and normalized transmittance at resonance, T min given by equation
(5.11) below.
𝑇 𝑚𝑖𝑛
= [
(𝑎 −𝑡 )(1+𝑎 .𝑡 )
(𝑎 +𝑡 )(1−𝑎 .𝑡 )
]
2
(5.11)
From these measured values the unknowns a and t can be determined by
iteratively solving equations (5.8) and (5.11) from which κ and α can be estimated. One
important point to be noted is that cos (θ) = m2π (where m is an integer) is the condition
85
for resonance as mentioned earlier. From (5.3) we can see that it depends on the effective
index of the mode n eff, in the ring. Thus a change in the effective index of the ring will shift
the resonance of the micro ring resonator and multiple modes (say TE and TM) with
different effective indices hit resonance at different wavelengths.
5.1.2 Coupling Schemes and Losses
Figure 5.1 shows the top view of a ring resonator. If the straight waveguide and the ring
lie on the same plane, the coupling scheme is called lateral coupling. Whereas, if the ring
was on a different plane below or above the straight waveguide then the scheme is called
vertical coupling. The coupling gap decides the extent of power coupled into the ring
which has to be taken into account while designing the ring. In general the intensity loss
coefficient (see equation 5.3), α = α m (material absorption loss) + α b (radiation loss due to
bending) + α w (scattering loss due to sidewall roughness). These losses affect the line
width of the resonance. Bending losses determine the minimum ring radius before which
light in the ring is coupled into the radiation modes, larger index contrast between the
core and cladding provides ability to design smaller rings. Material loss is the amount of
light absorbed by the material at a particular wavelength. Scattering losses occur due to
imperfect or rough walls of the waveguide which is primarily introduced during
photolithography. Reference [58] covers the above topics in great detail using polymer
micro ring resonators.
86
5.2 Fabrication and testing
The fabrication procedure to obtain planar laterally coupled microring resonators are
similar to that mentioned in section 4.1 previously for straight waveguides, other than a
different mask (negative photo mask) with micro ring patterns of varying coupling lengths
and ring radii. Due to a negative photomask, to obtain chrome metal patterns by lift-off
process, a double layer photoresist consisting of S3612 (second layer) and LOL 2000 (first
layer) photo resists were used as shown in Figure 5.3. After standard UV lithography the
Figure 5.3: Stepwise procedure to fabricate hybrid titania-silica sol-gel microring
resonators (Left - Right).
samples were developed in MF-26 A where the first layer develops much faster than the
second layer leaving an undercut of the first layer that aids in the lift-off process. Figure
5.4 shows an optical image of microring resonator coupling junctions fabricated with
different coupling lengths. The images with dark waveguide regions are that of the mask
itself. We can see that the ends of the coupling regions are more rounded (or abrupt) in
case of the actual device compared to the masks where they are gradual.
87
The rounded regions act as an abrupt change in the waveguide width seen by the
optical mode incident in the bus or ring waveguides resulting in scattering of light in
addition to that caused by the side wall roughness introduced during photolithography.
The waveguide width is around 2.5 µm post fabrication everywhere except the coupling
regions where it is ~ 4 µm.
Figure 5.4: Optical images of microring resonators post fabrication. To compare the
quality of photolithography the photomasks are shown with dark waveguide regions.
To characterize the ring resonators the test setup was similar to that used for the
cutback measurements using straight waveguides. Due to the larger index contrast (~ 0.4
with respect to the silica lower cladding and ~ 0.85 with respect to the air upper cladding)
the effective index and free spectral range of a race track ring resonator with radius 75
µm and coupling length of 40 µm for the fundamental (quasi) TE and TM modes with a
waveguide width of 2 µm and height of ~ 680 nm (see Figure 4.6) using full vectorial finite
difference simulations (Olympios, Tempselene) was found to be:
88
𝑛 𝑒𝑓𝑓 𝑇𝐸
= 1.66, 𝑛 𝑒𝑓𝑓 𝑇𝑀
= 1.61, 𝐹𝑆𝑅 𝑇𝐸
= 2 𝑛 𝑚 , 𝐹𝑆𝑅 𝑇𝑀
= 2.07 𝑛𝑚
The coupling losses were high due to the small waveguide cross section (0.7 µm X 2.5
µm).Thus, instead of a tunable laser a broadband source (1520 – 1580 nm) feeding an
Erbium doped fiber amplifier (EDFA) was used as the input and an Ando optical spectrum
analyzer was used to obtain the transmission spectra. It was found that to record the
resonance spectra having a lensed fiber on the output end as well was necessary. Figure
5.5 shows a block diagram of the characterization setup used to test the ring resonators.
The lensed fibers and the device were all placed on 3D motion controlled stages to obtain
optimal coupling.
Figure 5.5: Block diagram of characterization setup for testing microring resonators.
Transmission spectra recorded for a microring of radius 75 µm and coupling length of 40
µm is shown below in Figure 5.6. There are two family of resonances observed as
expected from simulations, with a free spectral range of ~ 2.14 nm (quasi TM) and 2.11
nm (quasi TE) and measured group index of 2.38 (quasi TM) and 2.08 (quasi TE)
respectively. For the resonance dips at 1546.31 nm and 1553.69 nm the extinction ratio
is > 3dB from which the quality factor was extracted to be ~ 16000.
For the resonance at 1553.69 nm the normalized transmittance T min and 3dB
bandwidth Δλ was found to be 0.423 and 0.097 nm respectively. Using equations 5.8 and
5.11, a and t values obtained were 0.89 and 0.98 which suggests that the ring is slightly
89
under coupled (a < t) with ~ 4% power coupled into the ring and a round trip loss value of
~ 18.4 dB/cm.
Figure 5.6: Transmission spectra of microring resonator with a radius of 75 µm and
coupling length of 40µm. The circled resonance wavelengths belong to the quasi TM family
with a measured FSR of 2.14 nm and the non-circled ones belong to the quasi TE family
with a measured FSR of 2.11 nm.
This is much more that the propagation losses obtained for a straight waveguide with a
2.5 µm width and similar fabrication procedure mainly due to the abrupt coupling
junctions. This was further confirmed by observing the scattering of light from the top of
the MMI junction. Due to fabrication errors as depicted in Figure 5.4 there appears a
bright spot at the right end of the MMI junction as seen in Figure 5.7.
90
Figure 5.7: Image of scattering of light (top view) from the micro ring and MMI coupler.
The light is coupled into the waveguide from the left, a bright spot is visible at the end of
MMI junction. Also seen is scattering along the circumference of the ring and a relatively
bright spot on the bus waveguide caused due to fabrication errors.
Ring
circumference
in µm
Extinction
Ratio (dB)
Band
Width
(nm)
FSR
(nm)
Q
factor
a t Power
coupling
k
2
603 -4 dB 0.161 2.27 9629 0.84 0.96 8%
723 -3.7 dB 0.097 2.11 16000 0.89 0.98 4%
1351 -3.6 dB 0.13 1.31 11920 0.78 0.95 10 %
1477 -4.8 dB 0.39 1.16 3970 0.52 0.83 31%
Table 5.1: Measured and fitted values of race track micro ring parameters with different
circumferences.
Thus, the loss values obtained are higher due to these scattering spots seen from
the top view. With the available combinations of radii and coupling lengths on the photo
mask, it was not possible to observe critical coupling. For other devices tested with same
ring radius but increasing coupling lengths the extinction ratio improved to ~ - 5 dB but
they had lower Q factors due to similar fabrication errors. Some of the other micro ring
91
devices that were characterized are listed in Table 5.1. In all cases the microrings were
under coupled as a < t.
5.3 Microdisk and microtoroid resonators
Thermally grown silica and silica sol-gels have been used to fabricate ultra-high Q
resonators with Q factors ~ 10
8
[65]. There have also been demonstrations of sol-gels
doped with Erbium ions to attain lasing in these high Q factor resonators [27]. These are
based on forming microtoroids from microdisks. There are two main reasons why this
technology has succeeded. First, after fabrication of silica microdisks on silicon the silicon
layer beneath the silica disk has to be under etched using xenon di-fluoride etch
chemistry. The selectivity of XeF 2 to Silicon over Silica is > 1000:1 and thus the under etch
is achieved easily without harming the silica microdisk providing an air cladding for the
silica disks. Secondly, silica has a higher absorption coefficient at 10. 6 µm (CO 2 laser) than
silicon and thus by controlled exposure the silica in the overhanging region can be melted
and reforms due to surface tension to form microtoroids [66]. TiO 2 has a higher
absorption coefficient than silica at 10.6 µm while the selectivity with respect to silicon in
the XeF 2 etch system is unknown. Thus, micro disks and subsequently microtoroid
fabrication using titania or hybrid titania – silica sol-gel were investigated.
5.3.1 Fabrication and characterization of microdisks
The fabrication procedure to obtain microdisk resonators using titania – silica hybrid thin
films with 11% molar silica are similar to that mentioned in section 4.1 previously for
92
straight waveguides with one important distinction, other than a different mask (photo
mask) with micro disk pad patterns of varying radii the sol-gel thin films are coated on
silicon wafers. Following the steps in Figure 4.1 there is one additional processing step
where the silicon substrate is etched isotropically in XeF 2 gas system to obtain silicon
pillars with microdisks supported at the center. Typical etch durations were ~ 40 pulses
Figure 5.8: Formation of micropads and microdisks before and after XeF 2 etching
long with each pulse lasting 80 seconds during which the chamber containing the samples
was filled with XeF 2 gas. Figure 5.8 above shows etched hybrid silica-titania micro pads
and micro disks before and after the XeF 2 etching process.
There have been predominantly two schemes to couple light in and out of these
resonators: prism coupling [70 - 71] and tapered fiber coupling [72 - 73] as shown below
93
in Figure 5.9. Both the techniques rely on the coupling of the evanescent wave from the
prism/fiber into the microresonator as the fundamental whispering gallery mode
propagates close to the outer surface of the resonator.
Figure 5.9: Coupling mechanisms in micro resonators.
Figure 5.10: Top and side view of of microdisk and tapered fiber during measurement.
Using a tapered fiber coupler one of the micro disks were characterized. Figure 5.10
shows the top and side view of the taper and micro disk positions during measurements.
The best Q factor obtained was ~ 5500 for the family of resonances with measured FSR of
2.57 nm. The second family had a larger FSR of 3.97 nm but relatively broad resonances
with Q factors of ~ 2000 or lower as seen in Figure 5.11 below. The larger FSR corresponds
to higher order radial modes which are situated further away from the periphery. The low
94
Q factors are mainly due to rough sidewalls of the microdisk. As the fundamental mode
exists near the periphery any sidewall roughness affects the scattering losses.
Figure 5.11: Transmission spectra of a tapered fiber and an 80 µm (diameter) microdisk.
To test if the hybrid titania sol-gel based microdisks can be reflowed using CO 2 laser to
form microtoroids, some of the microdisks were placed in the path of the laser beam such
that the beam was perpendicular to substrate plane, with increasing laser energy and
fixed time. While the microdisks do melt due to heating it was found that in all cases the
melting process was not self-quenched (like in the case of silica microdisks ) and instead
of forming toroids due to surface tension the material flowed downwards towards the
substrate forming a blanket like structure as shown in Figure 5.12 below.
95
Figure 5.12: SEM images of CO 2 laser ablated microdisk.
From the microdisk measurements and laser reflow experiments it is evident that
the work on microdisks is still largely unexplored. There is a need to obtain microdisks
with smoother sidewalls and preferably low material losses and study the laser reflow
process on films of varying thickness and doping levels (of silica) to develop a reliable
technique to obtain microtoroids using this new material system.
96
By improving the fabrication process involved in fabricating microrings the loss
values obtained can be further reduced. By coating suitable upper cladding materials (like
cytop, n ~ 1.34) before and after characterization of the micro-rings the scattering losses
and absorption losses can be extrapolated. Just like the straight waveguide
characterizations to the best of our knowledge, this is the first demonstration of hybrid
titania – silica sol-gel microring resonators containing relatively large titania content (~
90% molar).
97
Chapter 6
Conclusion
The experimental work undertaken as a part of this dissertation was mainly focused on
developing a scalable and reliable process to obtain optically thick sol-gel thin films using
titania and hybrid titania and silica sols. Using benzoyl acetone as a chelating agent to
reduce the reactivity of hygroscopic titania precursors stable viscous solutions of titania
sol were prepared. By spin coating these solutions and subsequently baking them under
vacuum at temperatures of 700
0
C or more crack-free films on silicon and silica coated
silicon wafers with thicknesses of upto 650 nm in a single spin and bake cycle were
obtained.
Based on UV-Vis spectrophotometry and FTIR spectroscopy a reliable baking
process was established to obtain transparent films from different sols studied. Using
silica and zirconia sols in controlled molar ratios hybrid stable titania – silica and titania –
zirconia sols were synthesized and thin films with refractive index variation in the range
of 1.6 – 2.5 at 1550 nm were reported. Varying the silica molar percentage from 5% - 43%
in hybrid titania-silica sols gave thin films with refractive index between 2.37 – 2 at 1550
nm. Thin films with refractive index in the range of 1.6 – 2 at 1550 nm can be realized by
either using zirconia sols as dopants or increasing the vacuum bake temperatures above
700
0
C for undoped titania films. Undiluted solutions gave thin films with sufficient optical
98
thickness and refractive index to support an optical mode in a direction normal to film
surface.
For vacuum baked films a high temperature (1000
0
C) sintering step was necessary
in achieving refractive index values > 2 at near infrared wavelengths. Firstly, irrespective
of doping, this sintering process resulted in all films transforming from amorphous to
rutile phase which were confirmed from XRD measurements. Secondly, the trade-off in
attaining films with refractive index values >2 is that the film surface quality deteriorated
and micro cracks appeared. FESEM images showed the presence of large particles (100
nm or more) and bumpy surfaces in case of undoped titania films but relatively lower
particle sizes (sub 50 nm) for hybrid titania – silica and titania - zirconia thin films. Thus,
the trade-off using the current processing technique is that films with high refractive
index invariably had poorer surface quality, large particle sizes contributing to increased
bulk-scattering losses. In case of undiluted hybrid titania – silica thin films, with increasing
silica content the microcracks introduced from the sintering process reduced and films
with a refractive index of 2.1 (45% molar silica) and thickness of ~ 1 µm were achieved
which is the largest combination reported to date.
To determine its optical quality, using standard photolithography processes ridge
waveguides were fabricated using hybrid titania-silica sol-gel. For the RIE etch recipe
developed using CF 4 and O 2 gases, AFM data suggests that the r.m.s roughness of etched
waveguide sidewall was ~ 30 nm. By wet-etching in diluted HF solutions this roughness
was reduced to 15 nm (giving calculated scattering losses < 1dB/cm). From waveguide
cut-back measurements on ~ 2.5 cm long waveguides, the optical propagation losses
99
recorded for the fundamental TM-like mode were between 8 – 13 dB/cm at 1550 nm for
waveguides of widths 4 µm – 2 µm. From 2D and 3D scattering loss calculations the
estimated absorption losses are in the range of 6 – 8 dB/cm and waveguide width
dependent scattering losses between 2 – 7 dB/cm.
An integrated micro ring resonator passive filter was demonstrated using hybrid
titania - silica (11%) sol-gel coated on 4 µm silica/ silicon substrates. The highest Q factors
extracted were ~ 16,000 with a group index of 1.91, FSR of 12 GHz and optical loss of ~18
dB/cm which was scattering dominated mainly due to fabrication errors. The straight
waveguide and microring resonator characterization are the first demonstration of its
kind using hybrid titania – silica sol-gel containing relatively large titania content (~ 90%
molar).
Another possible application of these materials in fabricating microdisk /
microtoroidal resonators was explored. Using tapered fiber coupling a microdisk with Q
factor of ~ 5500 was recorded. CO 2 laser ablation experiments to melt the microdisk
preforms into microtoroid were unsuccessful.
Titania, hybrid titania – silica and titania – zirconia sol-gel based material system
is a promising candidate for optical materials with relatively high refractive indices (n>2)
and large refractive index variations (Δn ~ 0.9). The synthesis and deposition process
developed is scalable and reasonably reliable and the material is processable by standard
lithography and etching techniques. From preliminary (optical loss) measurements we
can conclude that the material system needs improved fabrication processes (alleviate
reactive ion etching induced roughness) and a better densification route (sintered films
100
with high refractive index but improve optical quality) to fully realize its potential for
integrated photonic device applications.
6.1 Future work
As mentioned earlier the first issues to be resolved are the reactive ion etch recipes to
obtain the lowest possible etching induced roughness during fabrication and better
lithography (pattern transfer). Alternately, the structures can be photo patterned
eliminating the need to use RIE. The primary reason for using MAPTMS as the precursor
for obtaining silica sols was that the resulting sol could be photo patterned using UV light
by adding a photoinitator (IRGA CURE 184) [11]. Using the hybrid titania – silica sol doped
with photoinitiator molecules UV patterning was found possible experimentally.
However, the samples have to be developed in a solvent and the baking temperatures
prior to patterning were required to be less than 200
0
C to avoid decomposition of the
photoinitiator. Since, the chelating agent was still present in the film after baking, the
patterned region was attacked during development. By developing a scaled recipe using
glycidylmethacrylate and propylene oxide reported in [28] obtaining photo patterned
structures need to be investigated.
Analysis of the composition of films will be conducted using Auger electron
spectroscopy (AES) and X-ray photon spectroscopy (XPS) [32] and porosity studies using
TGA or DTA experiments [37]. The effect of different vacuum levels on thin film optical
properties also need to be recorded. Using these results better optical models can be
constructed to obtain improved fits of the ellipsometric data.
101
Due to the use of chelating agents the thin films need high temperature processing
which give rise to additional problems due to crystallization. The processing route (or the
use of alternate recipes) needs to be explored further to obtain relatively high refractive
index amorphous films similar to that obtained by sputtering. The baking and sintering
process needs to be studied in greater detail to validate the results obtained in this
dissertation and ascertain any fundamental limitations of the material system as a whole
as some of the results presented are purely empirical.
Measurement of waveguide propagation losses at other visible and near infrared
wavelengths for waveguides containing various percentages and types of dopants (silica
and zirconia) will be undertaken to evaluate the merits of the hybrid sol-gel material
system as an optical grade material. With an increase in dopant levels as the refractive
index decreases the scattering losses at a given wavelength will reduce due to lower index
contrasts but for a fixed dopant level shorter wavelengths are expected to exhibit larger
scattering losses due to Rayleigh scattering. Microtoroidal resonators using this material
system needs to be evaluated further as the optical properties give rise to more flexibility
in using them as sensors etc. Waveguide Bragg gratings will make an interesting
demonstration as the wide refractive index variation can be used to make filters with large
pass band variations.
Titania thin films have interesting electrical properties in that they can be switched
between different conductivity states by heating [74, 75]. It would be worthwhile to see
if titania films and hybrid titania – silica sol-gel films exhibit similar switching properties.
102
Bibliography
[1] B.E.A Saleh and M.C. Teich, “Fundamentals of Photonics”, John Wiley & Sons,
(2007)
[2] Richards B.S., “Single-material TiO 2 double layer antireflection coatings”, Solar
Energy Materials and Solar Cells, Vol. 79, pp. 369 – 390, (2003)
[3] “Refractive index n of Al xGa 1-xAs alloys.” Batop. Batop Optoelectronics, n.d. Web.
5 Apr 2012.
[4] Viens J-F, Meneghini C., Villeneuve A., Galstian T.V., Knystautas E.J., Duguay M.A.,
Richardson K.A. and Cardinal T., “Fabrication and characterization of Integrated
optical waveguides in sulfide chalcogenide glasses”, Journal of Lightwave
Technology, Vol. 17, No. 7, pp. 1184 – 1191 (1999)
[5] Fouchet S., Carenco A., Daguet C., Guglielmi R., Riviere L., "Wavelength dispersion
of Ti induced refractive index change in LiNbO 3 as a function of diffusion
parameters" Journal of Lightwave Technology, Vol. 5, No. 5, pp. 700 ‐ 708 (1987)
[6] “Corning SMF-28 ULL optical fiber.” Corning. Corning Incorporated, n.d. Web. 5
Apr 2012.
[7] Jensen B. and Torabi A., “Refractive index of quaternary In 1−xGa xAs yP 1−y lattice
matched to InP”, Journal of Applied Physics, Vol. 54, No. 6, pp. 3623- 3625 (1983)
[8] Tanaka H., Kawamura Y. and Asahi H., “Refractive indices of In 0.49Ga 0.51−xAl xP
lattice matched to GaAs”, Journal of Applied Physics, Vol. 59, No. 3, pp. 985- 986
(1986)
[9] “Single mode waveguide resins.”Chemoptics. Chemoptics Inc, n.d. Web. 5 Apr 2012.
103
[10] Du X.M., Touam T., Degachi L., Guilbault J.L., Andrews M.P. and Najafi S.I., “Sol-gel
waveguide fabrication parameters: an experimental investigation” Optical
Engineering, Vol. 37, No. 4, pp. 1101 – 1104 (1998)
[11] Yoldas B.E and O’Keeffe T.W., “Antireflective coatings applied from metal-organic
derived liquid precursors”, Applied Optics, Vol. 18, No. 18, pp. 3133 – 3138 (1979)
[12] Holmes A.S., Syms R.R.A., Li M. and Green M., “Fabrication of buried channel
waveguides on silicon substrates using spin-on glass”, Applied Optics, Vol. 32, No.
25, pp. 4916 - 4921 (1993)
[13] Nayar P.S., “Refractive index control of silicon nitride films prepared by radio-
frequency reactive sputtering”, Journal of Vacuum Science and Technology, A, Vol.
20, No.6, pp. 2137 – 2139 (2002)
[14] Gorin A., Jaouad A., Grondin E., Aimez V. and Charette P., “Fabrication of silicon
nitride waveguides for visible-light using PECVD: a study of the effect of plasma
frequency on optical properties”, Optics Express, Vol. 16, No. 18, pp. 13509 –
13516 (2008)
[15] Machorro R., Samano E.C., Soto G., Villa F. and Cota-Araiza L., “Modification of
refractive index in silicon oxynitride films during deposition”, Materials Letters,
Vol. 45, pp. 47 – 50, (2000)
[16] Seddon A.B., “Chalcogenide glasses: a review of their preparation, properties and
applications”, Journal of Non-Crystalline Solids, Vol. 184, pp. 44 – 50 (1995)
[17] Bradley J.D.B, Evans C. C., Choy J.T., Reshef O., Deotare P. B., Parsy F., Philips K. C.,
Loncar M. and Mazur E., ‘Submicrometer-wide amorphous and polycrystalline
TiO 2 waveguides for microphotonic devices’, Optics Express, Vol. 20, No. 21, pp.
23821 – 23831 (2012)
104
[18] Bauters J. F., Heck M.J.R., John D., Dai D., Tien M-C., Barton J.S., Leinse A.,
Heideman R.G., Blumenthal D.J. and Bowers J.E., “Ultra-low-loss high-aspect ratio
Si 3N 4 waveguides”, Optics Express, Vol. 19, No.4, pp. 3163 – 3174 (2011)
[19] Solmaz M.E., Adams D.B., Tan W.C., Snider W.T. and Madsen C.K., “Vertically
integrated As 2S 3 ring resonator on LiNbO 3”, Optics Letters, Vol. 34, Issue 11, pp.
1735 – 1737 (2009)
[20] Choy J.T., Bradley J.D.B., Deotare P.B., Burgess I.B., Evans C.C., Mazur E. and Loncar
M., “Integrated TiO 2 resonators for visible photonics”, Optics Letters, Vol. 37, Issue
4, pp. 539-541 (2012)
[21] Brinker C.J. and Scherer G.W., “Sol-gel Science: The Physics and Chemistry of Sol-
Gel Processing”, Academic Press, Inc (1990)
[22] Wright J.D. and Sommerdijk N.A.J.M., “Sol-Gel materials: Chemistry and
Applications”, Gordon and Breacher Science Publishers (2001)
[23] Livage J., Henry M. and Sanchez C., “Sol-gel chemistry of transition metal oxides”,
Progress in Solid State Chemistry, Vol. 18, pp. 259 – 341 (1988)
[24] Yang L., Saavedra S.S., Armstrong N.R. and Hayes J., “Fabrication and
characterization of low-loss, sol-gel planar waveguides”, Analytical Chemistry, Vol.
66, No. 8 pp. 1254-63 (1994)
[25] Brinker C.J. and Harrington M.S., “Sol-gel derived antireflective coatings for
Silicon”, Solar Energy Materials, Vol. 5, pp. 159 – 172 (1981)
[26] Bhola B., Nosovitskiy P., Mahalingam H. and Steier W.H., “Sol-gel based Integrated
Optical Microring Resonator Humidity Sensor”, IEEE Sensors Journal, Vol.9, No.
7,pp. 740 – 747 (2009)
[27] H.-S. Hsu, C. Cai, A. M. Armani, “Ultra-low-threshold Er:Yb sol-gel microlaser on
silicon,” Optics Express, Vol. 17, No. 25, pp. 23265 (2009).
105
[28] Himmelhuber R., Gangopadhyay P., Norwood R.A., Loy D.A. and Peyghambarian
N., “Titanium oxide sol-gel films with tunable refractive index”, Optical Materials
Express, Vol.1, No.2, pp. 252 – 258 (2011)
[29] Hinczewski D.S., Hinczewski M., Tepehan F.Z., and Tepehan G.G., “Optical filters
from SiO 2 and TiO 2 multi-layers using sol-gel spin coating method”, Solar Energy
Materials & Solar Cells, Vol. 87, pp. 181- 196 (2005)
[30] Chrysicopoulou P., Davazoglou D., Trapalis C. and Kordas G., “Optical properties
of SiO 2- TiO 2 sol-gel thin films”, Journal of Materials Science, Vol. 39, pp. 2835 –
2839 (2004)
[31] Mechiakh R., Meriche F., Kremer R., Bensaha R., Boudine B. and Boudrioua A.,
“TiO 2 thin films prepared by sol-gel method for waveguiding applications:
Correlation between the structural and optical properties”, Optical Materials, Vol.
30, pp. 645 – 651 (2007)
[32] Wang Z., Helmersson U. and Kall P-O., “Optical properties of anatase TiO 2 thin
films prepared by aqueous sol-gel process at low temperature”, Thin Solid Films,
Vol. 405, pp. 50 – 54 (2002)
[33] Chen K.C., Tsuchiya T. and Mackenzie J.D., “Sol-gel Processing of Silica: I. The role
of the starting compounds”, Journal of Non-Crystalline Solids, Vol. 81, pp. 227 –
237 (1986)
[34] Pope E.J.A. and Mackenzie J.D., “Sol-gel Processing of Silica: II. The role of the
catalyst”, Journal of Non-Crystalline Solids, Vol. 87, pp. 185 – 198 (1986)
[35] Liu B., Ho S.T., “Sub-100 nm nanolithography and pattern transfer on compound
semiconductor using sol-gel derived TiO 2 resist”, J. of Electrochem. Soc., Vol. 155,
Issue 5, pp. 57 - 60, (2008)
106
[36] Fardad M.A., Yeatman E.M., Dawnay E.J.C., Green M. and Horowitz F., “Effects of
H 2O on structure of acid-catalysed SiO 2 sol-gel films”, Journal of Non-Crystalline
Solids, Vol. 183, Issue 3, pp. 260-267 (1995)
[37] Que W., Sun Z., Lam Y. L., Chan Y.C. and Kam C.H., “Effects of titanium content on
properties of sol-gel silica-titania films via organically modified silane precursors”,
Journal of Physics D: Applied Physics, Vol. 34, No.4, pp. 471 – 476 (2001)
[38] Matsuda A., Matsuno Y., Tatsumisago M. and Minami T., “Changes in Porosity and
Amounts of Adsorbed Water in Sol-Gel Derived Silica films with Heat Treatment”,
Journal of Sol-Gel Science and Technology, Vol. 20,pp. 129-134, (2001)
[39] Yeatman E. M., Green M., Dawnay E. J. C., Fardad M. A. and Horowitz F.,
“Characterization of Micro-porous Sol-gel Films for Optical Device Applications”,
Journal of Sol-Gel Science and Technology, Vol. 2, pp. 711-715, (1994)
[40] Thamaphat K., Limsuwan P. and Ngotawornchai B., “Phase Characterization of
TiO 2 Powder by XRD and TEM”, Kasetsart Journal of Natural Science, Vol. 42, No.
5, pp. 357 – 361 (2008)
[41] Kasai N., Kakudo M., “X-Ray Diffraction by Macromolecules”, Kodansha and
Springer Series (Chemical Physics) (2005)
[42] Lee K.K., Lim D.R., Kimerling L.C., Shin J. and Cerrina F., “Fabrication of ultralow-
loss Si/SiO 2 waveguides by roughness reduction”, Optics Letters, Vol. 26, No. 23,
pp. 1888 – 1890 (2001)
[43] Feuchter T. and Thirstrup C., “High precision planar waveguide propagation loss
measurement technique using a Fabry–Perot cavity,” IEEE Photonics Technology
Letters, Vol. 6, pp. 1244–1247 (1994)
[44] S. Taebi, M. Khorasaninejad and S. S. Saini, “Modified Fabry-Perot interferometric
method for waveguide loss measurement”, Applied Optics, Vol. 47, No. 35, pp.
6625 – 6630, (2008)
107
[45] Kupsta M.R., Taschuk M.T., Brett M. J. and Sit J. C., “Reactive Ion Etching of
Columnar Nanostructured TiO 2 Thin Films for Modified Relative Humidity Sensor
Response Time”, IEEE Sensors Journal, Vol. 9, No. 12, pp. 1979 – 1986 (2009)
[46] Matsutani A., Koyama F. and Iga K., “Microfabrication of dielectric multilayer
reflector by reactive ion etching and characterization of induced wafer damage”,
Japanese Journal of Applied Physics, Vol. 30, No.2, pp. 428 – 429 (1991)
[47] Furuhashi M., Fujiwara M., Ohshiro T., Tsutsui M., Matsubara K., Taniguchi M.,
Takeuchi S. and Kawai T., “Development of microfabricated TiO 2 channel
waveguides”, AIP Advances, Vol.1, pp. 032102 - (1 -5) (2011)
[48] Wee T.C.L., Ooi B.S., Zhou Y., Chan Y.C. and Lam Y.L., “Characterization of Reactive
Ion Etching of Sol-Gel SiO 2 Using Taguchi Optimization Method”, SPIE Proceedings,
Vol. 3896, pp. 438 – 444 (1999)
[49] Abe K., Teraoka E.Y.M., Kita T. and Yamada H., “Nonlinear optical waveguides with
rutile TiO 2”, Proceedings of SPIE, Vol. 7940, pp. 79401G1 – G7 (2011)
[50] “C2V – 200 micro GC.”ThermoScientific. Thermo Fisher Scientific Inc, n.d. Web. 5 Apr
2012
[51] J. P. R. Lacey and F. P. Payne, “Radiation loss from planar wave-guides with
random wall imperfections,” Proceedings of Institute of Elect. Eng., Vol. 137, No.
4, pp. 282–288, (1990)
[52] F. P. Payne and J. P. R. Lacey, “A theoretical analysis of scattering loss from planar
optical waveguides,” Optical and Quantum Electronics, Vol. 26, No. 10, pp. 977–
986, (1994)
[53] Barwicz T. and Haus H.A., “Three-Dimensional Analysis of Scattering Losses Due to
Sidewall Roughness in Microphotonic Waveguides”, Journal of Lightwave
Technology, Vol. 23, No. 9, pp. 2719 – 2732 (2005)
108
[54] D .Marcuse, “Theory of Dielectric Optical Waveguides”, 2
nd
edition, Academic
Press Inc, 1994.
[55] Bohren C.F. and Huffman D.R., “Absorption and Scattering of light by small
particles”, John Wiley & Sons, 1983.
[56] Prahl S., “Maetzler's MATLAB code for Mie theory.” Oregon Medical Laser Center.
n.d .Web. 5 Oct 2013.
[57] Bhola B., “Applications of optical microring and micro disk resonators as physical,
chemical and biological sensors”, Ph.D Dissertation, University of Southern
California, Los Angeles (2007)
[58] Rabiei P. and Steier W.H., “Polymer Microring Resonators”, in “Optical
Microcavities” by Vahala K.J., World Scientific, (2004)
[59] B. E. Little, S. T. Chu, W. Pan, and Y. Kokubun., “An eight-channel add-drop filter
using vertically coupled microring resonators over a cross grid”, IEEE Photonics
Technology Letters, Vol. 11, pp. 691–693, (1999)
[60] B. E. Little, J. S. Foresi, G. Steinmeyer, E. R. Thoen, S. T. Chu, H. A. Haus, E. P. Ippen,
L. Kimerling, and W. Greene, “Ultra-compact Si-Sio 2 microring resonator optical
channel dropping Filters”, IEEE Photonics Technology Letters, Vol. 10, pp. 529–
551, (1998)
[61] Rafizadeh D., Zhang J.P., Hagness S.C. Taflove A., Stair K.A., Ho S.T., “Waveguide
coupled AlGaAs/ GaAs microcavity ring and disk resonators with high finesse and
21.6 nm free spectral range”, Optics Letters, Vol. 22, No. 16, 1244, (1997)
[62] G. Gupta, “Microring resonator based filters and modulators: optical coupling
control and applications to digital communications”, Ph.D Dissertation, University
of Southern California, Los Angeles (2008)
109
[63] Yariv A., “Universal relations for coupling of optical power between
microresonators and dielectric waveguides”, Electronics Letters, Vol. 36, No. 4, pp.
321 – 322 (2000)
[64] Rabus D.G., “Integrated Ring Resonators- The Compendium”, Springer Series
(2007)
[65] Armani D.K., Kippenberg T.J., Spillane S.M., Vahala K.J. “Ultra-high-Q toroid
microcavity on a chip”, Nature, Vol. 421, 925, (2003)
[66] Vahala K.J., “Optical Microcavities”, World Scientific, (2004)
[67] K. D. Djordjev, S. J. Choi, S. J. Choi, and P. D. Dapkus, “Active semiconductor
microdisk device”, IEEE Journal of Lightwave Technology, Vol. 20, No. 1,pp. 105–
113, (2002)
[68] Maleki L., Ilchenko V.S., Savchenkov A.A., Matsko A.B., “Crystalline whispering
gallery mode resonators in optics and photonics” in Practical Applications of
Microresonators in Optics and Photonics, CRC Press, (2009)
[69] Heebner J., Grover R., Ibrahim T. A., “Optical Microresonators”, Springer Series in
Optical Sciences, (2008)
[70] Tien P.K. and Ulrich R., “Theory of Prism Film Coupler and Thin Film Light Guides”,
Journal of Optical Society of America, Vol. 60, No. 10, 1325, (1970)
[71] Cohen D.A. and Levi A.F.J., “Microphotonic millimeter-wave reciever
architecture”, Electronics Letters, Vol. 37, No. 1, 37, (2001)
[72] Knight J.C., Cheung G., Jacques F., Birks T.A., “Phase matched excitation of
whispering gallery model resonances by a fiber taper”, Opt. Lett., Vol. 22, 1129-
1131, (1997)
[73] “Model2010/M Overview.” Metricon Corporation. Metricon Corporation, n.d.
Web. 5 Apr 2012
[74] Argall F., “Switching Phenomena in titanium oxide thin films”, Solid State
Electronics, Vol. 11, pp. 535- 541(1967)
110
[75] Earle M.D., “The Electrical Conductivity of Titanium Dioxide”, Physical Review,
Vol. 61, pp. 56- 62 (1942)
111
Appendix A
Hybrid titania - silica and titania - zirconia
sols: synthesis details
There were two different doping procedures used while synthesizing the hybrid sols used
for the experimental work in this thesis. One was the conventional mixing scheme by
varying molar concentrations of the different compounds which is typically followed in
most cases. The other was a procedure similar to that reported in [25] where the sols are
intermixed in different volume ratios to obtain hybrid solutions containing different molar
ratios of the constituents.
A.1 Mixing sols by varying molar concentrations of compounds
Three vials, each containing 3 ml (8.8 X 10
-3
moles) of titanium butoxide was mixed with
equal moles of benzoyl acetone and ethanol. The total volume of this titania stock
solution was ~4. 822 ml and contained 1.825 moles/litre of Titanium. Three different silica
stock solutions were prepared with the molar percentage of methyl acryloxy propyl
trimethoxy silane (MAPTMS, silica precursor) fixed at 5%, 11% and 25% with respect to
titanium butoxide. In preparing each of the three silica stock solutions 0.01 M HCl
(hydrochloric acid) and water were added in the molar ratio of 1 : 0.75 and 1 : 1.5 with
respect to the silica precursor respectively. To obtain larger volumes of total hybrid titania
– silica solutions with different silica concentrations the volume of titania precursor can
112
be increased and all other quantities adjusted to reflect the molar ratios decided. The
above procedure is summarized in Table A.1 below.
Compounds Molar ratios # of moles Volume of total
solution
mol / litre of Ti
and Si in final
solution
Titanium butoxide
Benzoyl acetone
Ethanol
1
1
1
8.8 X 10
-3
8.8 X 10
-3
8.8 X 10
-3
4.822 ml
1.825 mol/l
MAPTMS
0.01 M HCl
DI water
0.05
0.05 X 0.75
0.05 X 1.5
4.4 X 10
-4
3.3 X 10
-4
6.6 X 10
-4
0.152 ml
Ti : 1.769 mol / l
Si : 0.089 mol / l
(Ti : Si :: 1 : 0.05)
MAPTMS
0.01 M HCl
DI water
0.11
0.11 X 0.75
0.11 X 1.5
9.68 X 10
-4
7.26 X 10
-4
1.452 X 10
-3
0.269 ml
Ti : 1.729 mol / l
Si : 0.19 mol / l
(Ti : Si :: 1 : 0.11)
MAPTMS
0.01 M HCl
DI water
0.25
0.25 X 0.75
0.25 X 1.5
1.76 X 10
-3
1.32 X 10
-3
2.64 X 10
-3
0.492 ml
Ti : 1.656 mol / l
Si : 0.337 mol / l
(Ti : Si :: 1 : 0.25)
Table A.1: Summary of concentrations used to obtain hybrid titania – silica sols with 5%, 11% and
25% silica content.
A.2 Mixing sols by varying volume ratios of stock solutions
An alternate way to obtain hybrid titania – silica stock solutions with varying silica content
is by preparing large quantities of titania and silica stock solutions and mixing them in
different volume ratios to obtain the necessary molar concentrations.
15 ml (4.4 X 10
-2
moles) of titanium butoxide was mixed with equal moles of
benzoyl acetone and ethanol. The total volume of this titania stock solution was 24. 113
113
ml and contained 1.825 moles/litre of Titanium. Silica stock solution was prepared with
equi molar (4.4 X 10
-2
moles) methyl acryloxy propyl trimethoxy silane (MAPTMS, silica
precursor) (with respect to titanium butoxide). In preparing the silica stock solutions 0.01
M HCl (hydrochloric acid) and water were added in the molar ratio of 1 : 0.75 and 1 : 1.5
with respect to the silica precursor respectively. The total volume of silica stock solution
was 11. 846 ml and contained 3.714 moles/litre of Silicon. 5ml of titania stock solution
was pipetted into a vial and different volumes of silica stock solution were added based
on the volume ratios mentioned in Table A.2 below. For e.g. to obtain 90%:10% hybrid
solution, 1ml of silica stock solution and (0.9*24.113/(0.1*11.846)) =18.32 ml of titania
stock solution is required. The volume ratios change when the total volume of the titania
and silica stock solutions change.
TiO 2 % – SiO 2 %
Volume ratio
of TiO 2 : SiO 2
Volume of
total solution
mol / litre of Ti and Si in final solution
(equivalent molar ratio)
90% - 10 % 18.32 : 1 5.273 ml
Ti : 1.730 mol / l; Si : 0.192 mol / l
(Ti : Si :: 1 : 0.11)
80% - 20 % 8.142 : 1 5.614 ml
Ti : 1.625 mol / l; Si : 0.406 mol / l
(Ti : Si :: 1 : 0.25)
70% - 30 % 4.75 : 1 6.053 ml
Ti : 1.507 mol / l; Si : 0.646 mol / l
(Ti : Si :: 1 : 0.43)
Table A.2: Volume ratios required to vary silica concentrations in hybrid titania – silica solutions
based on the total volume of titania and silica stock solutions prepared. Equivalent molar ratios
are calculated in the last column to compare the recipes in both the tables on a common scale.
Zirconia stock solutions used to prepare hybrid titania – zirconia sols were
prepared based on the recipe listed in Table A.1. 11% (4.84 X 10
-4
moles) molar zirconium
propoxide (with respect to titanium butoxide - 8.8 X 10
-4
moles) was mixed with equal
114
volume of n – propanol and equal moles (4.84 X 10
-4
moles) of methacrylic acid to obtain
zirconia stock solution. Titania stock solution used was as mentioned in Table A.1.
Silica stock solutions using 4.4 X 10
-2
moles of tetra ethyl ortho silicate (standard
quadrivalent silica precursor) were prepared similar to the procedure mentioned in Table
A.2. The total volume of silica stock solution was 21. 44 ml and contained 2.052
moles/litre of Silicon. The volume ratios and equivalent molar ratios were determined
using the procedure mentioned above.
115
Appendix B
Spectroscopic ellipsometry of various sol -
gel thin films
All ellipsometric measurements were carried out at room temperature at the incident
angles of 65
0
, 70
0
and 75
0
in the wavelength range of 400 nm – 1600 nm with a 10 nm
step size using a computer controlled rotating analyzer type (VB-200) variable angle
spectroscopic ellipsometer (VASE) from J.A. Woollam Co.. The beam diameter was around
3 mm and the ‘zone average polarizer’ and ‘dynamic averaging’ options were selected in
all cases to obtain more accurate data points.
The VASE measures the complex reflectance ratio, ρ, which is related to the
measured ellipsometric parameters ψ and Δ by the parallel and perpendicular
polarization (with respect to the plane of incidence) reflection coefficients r p and r s given
by
𝜌 =
𝑟 𝑝 𝑟 𝑠 = tan (𝜓 )𝑒 𝑖 ∆
(B.1)
Once an optical model containing the different optical layers present in the sample was
constructed ellipsometric data was analyzed using the VASE software package (J.A.
Woollam Co.) which uses least – square regression analysis to obtain the unknown fitting
parameters and their 90% confidence limits. In general for a sol-gel layer on a silicon
substrate the primary unknown parameters are thickness and refractive index. Since the
titania sol-gel and hybrid sol-gels have a large bandgap (> 3eV) the refractive index can be
116
represented using a Cauchy model given by
𝑛 (𝜆 ) = 𝐴 +
𝐵 𝜆 2
+
𝐶 𝜆 4
(B.2)
where A, B and C are the unknown parameters. The imaginary part of the refractive index
is negligible in this spectral range. By varying the fitting parameters (thickness, A, B and
C) the difference between the measured ψ exp, Δ exp and the calculated ψ cal, Δ cal are
minimized. This difference is the mean squared error (MSE) which is defined as
𝑀𝑆𝐸 = √
1
𝑀 −𝑃 +1
∑ [(
𝜓 𝑖 𝑐 − 𝜓 𝑖𝑒
𝜎 𝑖𝜓
)
2
+ (
Δ
𝑖𝑐
− Δ
𝑖𝑒
𝜎 𝑖 Δ
)
2
]
𝑖 (B.3)
where M is the total number of experimental observations (for the spectral range of 400
nm – 1600 nm in 10 nm intervals it mounts to 121 data points per angle * 3 = 363 data
points), P is the number of fit parameters (3 – 5 generally), σ iψ and σ iΔ are the standard
deviations in ψ and Δ, respectively and i Є [1,M] is the summation index. A good optical
model and fit corresponds to a low positive value of MSE. MSE values below 20 are
considered a successful fit as long as the 90% confidence limits are logical.
In most cases inclusion of parameter C did not improve the quality of fits
drastically. For films that were sintered at a temperature of 1020
0
C or those obtained
using undiluted sols an intermix layer above the Cauchy layer or thickness non-uniformity
as a result of spin coating had to be incorporated as a fitting parameter to improve the
quality of the fit. Figures B.1 and B.2 below show the fits along with measured and
calculated ψ, Δ curves for two different samples one which followed a simple Cauchy
model and another that required an intermix layer to be added to the model.
117
Figure B.1: Generated and experimental ψ and Δ values for an air baked (700
0
C) diluted titania
film containing no water. The simple Cauchy model, MSE and fit parameters along with the 90%
confidence limits are shown above.
118
Figure B.2: Generated and experimental ψ and Δ values for an air sintered (1010
0
C) undiluted
titania - silica film containing 25% molar silica which was vacuum baked at 700
0
C initially. The
simple Cauchy model with an intermix layer (made of equal amounts of air voids and sol-gel), MSE
and fit parameters along with the 90% confidence limits are shown above.
119
MSE values for data points shown in Figure 3.1 were between 2 and 6. For the
various vacuum bake temperatures and subsequent sintering at 1020
0
C used to obtain
data in Table 3.1 the MSE values were in the range of 0.1 – 3 after the vacuum bake step
and between 3 – 10 after the sintering step. Similar trends were observed for the other
refractive index data reported in Chapter 3 (i.e. Table 3.2, Figures 3.9 and 3.10).
120
Appendix C
Additional images of thin films and testing
setup
Figure C.1: Optical micrographs (10X objective) of samples on silicon substrates containing titania
sol-gel vacuum baked at 100
0
C (left) and 200
0
C (right). Cracking reduces as the baking
temperature increases and samples were crack free when baked at 300
0
C or higher.
Figure C.2: Optical micrographs (50X objective) of samples on quartz substrates containing
undiluted titania (left) and titania – zirconia (right) sol-gel vacuum baked at 700
0
C and
subsequently sintered in air at 1020
0
C.
121
Figure C.3: Optical micrographs (50X objective) of samples on quartz substrates containing
undiluted titania – silica sol-gel with 5% (top left), 11% (top right) and 20% (bottom left) molar
silica vacuum baked at 700
0
C and subsequently sintered in air at 1020
0
C. As the molar percentage
increases the cracks reduce and larger islands are formed. SEM image of waveguide on 4 µm silica
coated silicon substrate obtained from 11% molar silica doped titania – silica sol-gel after high
temperature sintering (1020
0
C).
Figure C.4: Waveguide cutback measurement optical test setup showing the different 3D stages.
122
Figure C.5: Optical test setup output side showing the iris, polarizer, IR camera and detectors.
Figure C.6: Optical micrograph (20X objective) showing top view of input facet with lensed fiber
and groups of waveguides of different widths.
Abstract (if available)
Abstract
Sol‐gel materials are prepared by the generation of colloidal suspensions ("sols") which are subsequently converted to viscous gels by polymerization and further into solid materials upon densification. By controlled mixing of sols of different materials and concentrations, organic/ inorganic composite materials with useful optical properties have been synthesized for a variety of photonic applications. Inorganic titanium dioxide (titania) and silicon dioxide (silica) thin films have been developed using the sol‐gel process and used as anti‐reflection coatings for solar‐cells. Nevertheless, most of the integrated optical devices reported in literature are primarily based on the silica sol‐gel process. ❧ By doping titania sol‐gel with controlled amounts of silica sol‐gel the refractive index of the material obtained can be varied from ~2.5 to ~1.45 at the optical communication wavelengths. Moreover, being a large bandgap material (~ 3.5 eV for rutile phase) titania sol-gel is optically transparent over a large spectral region spanning the visible and near infrared wavelengths like silica. However, the titania sol‐gel process is relatively complicated compared to the established silica process mainly due to the reactive (hygroscopic) nature of the titanium starting compound. Apart from their use in antireflection coatings which require sub 100 nanometer film thicknesses it has been difficult to realize their potential in integrated optical device applications due to the lack of a scalable and reliable process. ❧ This work involves a modified synthesis process and densification route to obtain variable refractive index titania, titania – silica and titania - zirconia hybrid sol‐gel films in the range of 1.6 – 2.5 at the telecommunication wavelengths. We present a synthesis process involving a chelating agent by which stable viscous solutions of titania sol were obtained giving ~ 650 nm crack free thin films on silicon substrates with a single spin and bake cycle. On addition of silica sol (45% by molar ratio) as the dopant, films with thicknesses in the range of 1 μm (refractive index ~ 2.1) after a single spin and bake cycle were achieved which is the largest combination reported to date. ❧ We investigate the properties of thin films densified at high temperatures (> 700 °C) both under vacuum (< 100 mTorr) and nitrogen atmospheres. The vacuum based baking process provides a reliable technique to obtain crack free hybrid sol‐gel films. Titania, titania – silica and titania - zirconia hybrid sol‐gel films were characterized by various techniques including variable angle spectroscopic ellipsometry, ultra violet – visible – near infrared spectrophotometry, Fourier transform infrared spectroscopy, scanning electron microscopy and X‐ray diffraction to confirm their thickness and refractive index, optical transparency, morphology and crystallinity respectively. ❧ In order to truly determine its optical quality we fabricated ridge waveguides using hybrid titania‐silica sol‐gel and performed waveguide cut‐back measurements on 2.5 cm long waveguides. Reactive ion etch (RIE) recipes using CF₄ and O₂ gases were developed for the various sol‐gel materials and RIE induced roughness was measured using AFM measurements and alleviated using wet etching. The optical propagation losses recorded for the fundamental TM like mode were between 8 – 13 dB/cm at 1550 nm for waveguides of widths 4 μm – 2 μm. From 2D and 3D scattering loss calculations the estimated absorption losses are in the range of 6 – 8 dB/cm and waveguide width dependent scattering losses are between 2 – 7 dB/cm. ❧ We also demonstrate integrated micro ring resonator passive filters made of hybrid titania - silica sol‐gel coated on 4 μm silica/ silicon substrates with a group index of 1.91, FSR of 12 GHz, Q factor ~ 16,000 and extract an optical loss of ~18 dB/cm which is scattering dominated. The straight waveguide and microring resonator characterization are the first demonstration of its kind using hybrid titania – silica sol‐gel containing relatively large titania content (~ 90% molar). ❧ Another possible application of these materials in fabricating microdisk / microtoroidal resonators was explored. Using tapered fiber coupling a microdisk with Q factor of ~ 5500 was recorded. CO₂ laser ablation experiments to melt the microdisk preforms into microtoroids were unsuccessful. ❧ The experimental results presented in this dissertation suggest that the proposed synthesis process and densification route are promising but need improved fabrication processes (RIE induced roughness) to fully realize this hybrid material system for photonic applications.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Silica sol-gel thin film coatings for integrated photonic devices
PDF
Developing improved silica materials and devices for integrated optics applications
PDF
Active integrated photonic devices in single crystal LiNbO₃ micro-platelets and a hybrid Si-LiNbO₃ platform
PDF
Development of optical devices for applications in photonic integrated circuit and sensing
PDF
Application of optical forces in microphotonic systems
PDF
Development of integrated waveguide biosensors and portable optical biomaterial analysis systems
PDF
Development of hybrid microsensors for environmental monitoring and biodetection
PDF
Development of organic-inorganic optical microcavities for studying polymer thin films
PDF
Nonlinear optical nanomaterials in integrated photonic devices
PDF
Optical studies in photonics: terahertz detection and propagation in slot waveguide
PDF
Development of hybrid optical microcavities for Plasmonic laser and improving biodetection
PDF
Biological and chemical detection using optical resonant cavities
Asset Metadata
Creator
Mahalingam, Hari
(author)
Core Title
Titania and hybrid titania - silica sol-gel thin films and their applications in integrated optical devices
School
Viterbi School of Engineering
Degree
Doctor of Philosophy
Degree Program
Electrical Engineering
Publication Date
08/26/2014
Defense Date
12/19/2013
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
cut back measurements,doping sol‐gels,high refractive index,micro ring resonators,OAI-PMH Harvest,propagation losses,rectangular waveguides,scattering losses,sol‐gel thin films,titania - silica sol‐gel,titania sol‐gel
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Steier, William Henry (
committee chair
), Armani, Andrea M. (
committee member
), O'Brien, John D. (
committee member
)
Creator Email
dhanyahiremath@gmail.com,hari.m1783@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-365650
Unique identifier
UC11288138
Identifier
etd-Mahalingam-2273.pdf (filename),usctheses-c3-365650 (legacy record id)
Legacy Identifier
etd-Mahalingam-2273.pdf
Dmrecord
365650
Document Type
Dissertation
Rights
Mahalingam, Hari
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
cut back measurements
doping sol‐gels
high refractive index
micro ring resonators
propagation losses
rectangular waveguides
scattering losses
sol‐gel thin films
titania - silica sol‐gel
titania sol‐gel