University of Southern California Dissertations and Theses

Developing improved silica materials and devices for integrated optics applications

(USC Thesis Other)

Developing improved silica materials and devices for integrated optics applications

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

DEVELOPING IMPROVED SILICA MATERIALS AND DEVICES FOR
INTEGRATED OPTICS APPLICATIONS

by

Ashley Julia Maker

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the Requirements for the Degree
DOCTOR OF PHILOSOPHY
(Chemical Engineering)

August 2014

Copyright 2014 Ashley Julia Maker
ii
Acknowledgements

During my time at USC, I have had the great privilege and opportunity to
work with many truly amazing people, who directly and/or indirectly played a
tremendous role in making this thesis possible. I would first like to thank my
advisor, Prof. Andrea Armani, for her guidance and support. I could not have asked
for a better mentor during my undergraduate and graduate research. Thank you for
believing in me and taking an unemployed undergrad under your wing back in 2008.
Thank you for all you have done to teach me how to be a good researcher, writer,
reviewer, mentor, presenter, leader, and colleague. And for all the extra things you
have done to help, including the food, envelopes, lab parties/lunches/dinners,
carpooling, the DC trip, letting me try all my random ideas (even the ones you knew
wouldn’t work!), and especially for buying me that $40 bottle of aluminum
isopropoxide that made neodymium happy and heterodynable. What a journey it’s
been! It was such a great honor to be part of the research group from the very
beginning and learn from one of the best.
Along the way, I have been truly blessed to learn from many great teachers,
professors, and mentors. I thank my committee members: Prof. Rich Roberts, Prof.
Muhammad Sahimi, Prof. Noah Malmstadt, and Prof. Alan Willner, for their
support. And all my mentors and instructors at USC, Amador, EG, and elsewhere,
especially Coach Todd, Mr. Lomas, Mrs. Pereira, Mrs. Mistry, Mrs. Ogle, and Mr.
Grantham.
iii
It is often said that happiness comes not from the work you do, but who you
do it with. Indeed, I owe countless thanks to all members of the Armani Lab,
especially: Maria, Vinh, Sahar, Eda, Mark, Rasheeda, Kelvin, Michele, Cecilia, both
Sams (McB and K-L), Simin (BS!), Akshay, Ce, Soheil, Victoria [BTS(UP)SGB!],
Xiaomin, Tara, Garrison, Andre, Yuting, Emma, Max, and Leah. And the past
members of the group, especially Prof. Heather Hunt, Dr. Jason Gamba, Dr. Hong-
Seok Choi, Matt, Nishita, Audrey, Brian, Nic, Christine, Daniel, Bradley, Shehzad,
Chai, Dania, Carol, Lindsay, Tobi, Imran, Nick, Heaven, Sam Hsu, and Samantha.
And major thanks to Deniz Armani for inventing the toroid, all you do to support the
group, and for the great lab parties! You all are an amazing group of people. It’s
been such a great pleasure to work with you all. Thank you all for your help,
support, mentoring, testing/sol-gel/cleanroom parties, “fun” times, free food
adventures, Chick-fil-a trips, and making all the craziness that is research such a
unique experience. Keep having fun (both kinds), working hard, and closing the
XeF
2
valve. May the fun never end!
A big thank you also to my many great Chem-E classmates during undergrad
and grad school, especially Debbie and Stephanie. Additional thanks go to
numerous colleagues, neighbors in VHE, and collaborators who have gone out of
their way to help, including: Prof. Nutt’s group, Hamed, Hari, Khiza, Shermin,
Rohan, Mitzuki, Meredith, Heather Alexander, Angeline, Aimee, Martin, Andy,
Karen, Tom Lee, Donghai, and Alfonso.
And even though I spent many, many hours at USC, my family and friends
have been so supportive every step of the way. I could not have done it without you
iv
all. Especially Mom – thank you so much for everything. And I mean everything.
The road trip that started it all in July 2008, the many road trips throughout
undergrad and grad school, putting up with me, worrying about me, believing in me,
sending me food, and just being there. Also, HUGE thanks go to Aunt Alison, Uncle
Jeff, Noah, Maddy, Oreo, Trixie, Bear/“T.T.”, and Obediah “Dobie” Butterscotch.
Your love, support, food, and company have been so awesome throughout my time
at USC. Uncle Jeff, your love of family, fun, food, and all things nerdy is
contagious. We miss you.
Dad and Kevin/RSP, thanks for all your support, the fun trips to
lil’Rhody/Gelson’s/elsewhere, and the iPhone (and Figures 6-5, 6-6, and 6-8!).
Additional thanks to Randi Bug, Steve, and Ellen, the Kennedys, Pastor Ledic,
Grandma Carol and Grandpa Paul Maker (Ch. 8 ref 13); Sally, and all my family in
Michigan especially Aunt Patti and Grandma Carol/JBW. And my band/flute
friends, especially Susan, Nancy, Kim, Mandy, and Squad 47 (Geek Squad!). And to
Max the cat and Spot and Lumpy, my two frogs.
Additionally, I thank Thomas Bergersen and Nick Phoenix of TSFH for the
epic music and concert, my friends and neighbors at the Terraces, the awesome
people at CFA-USC for the food and fun, my bus and swim buddies who brightened
up the early mornings, Zhai Jiwei et al for the paper that changed everything, and
John, EdZ, and Craig for the bright light at the end of the tunnel. Finally, I thank
you, the reader, for taking the time to read my thesis. I hope it is helpful and furthers
work in the field and elsewhere. Thank you all again, and Fight on!
v
Table of Contents
Acknowledgements.............................................................................................. ii
List of Tables ...................................................................................................... ix
List of Figures...................................................................................................... x
Abstract ........................................................................................................... xvii

Chapter 1 Introduction ............................................................................................. 1
1.1 Motivation................................................................................................ 1
1.2 Chapter Overview..................................................................................... 3
Chapter 1 References ........................................................................................... 5

Chapter 2 Background and Overview....................................................................... 8
2.1 Overview: Integrated Optical Devices....................................................... 8
2.1.1 Refractive Index and Total Internal Reflection .................................. 9
2.1.2 Absorption Loss.............................................................................. 11
2.1.3 Optical Waveguides........................................................................ 12
2.1.4 Optical Resonators.......................................................................... 15
2.2 Overview of Toroid Theory .................................................................... 18
2.2.1 Quality Factor (Q)........................................................................... 18
2.2.2 Circulating Power ........................................................................... 20
2.2.3 Mode Volume................................................................................. 20
2.2.4 Finesse............................................................................................ 21
2.2.5 Purcell Factor.................................................................................. 21
2.2.6 Effective Refractive Index .............................................................. 22
2.2.7 Effective Absorption Coefficient..................................................... 22
2.2.8 Free Spectral Range ........................................................................ 23
2.2.9 Total Quality Factor........................................................................ 23
2.2.10 Radiation Loss ................................................................................ 24
2.2.11 Contamination Loss ........................................................................ 25
2.2.12 Surface Scattering Loss................................................................... 25
2.2.13 Material Absorption Loss................................................................ 26
2.2.14 Coupling Loss................................................................................. 27
2.2.15 Which Losses Dominate?................................................................ 27
2.2.16 Advantages and Disadvantages of High Qs ..................................... 28
2.3 Device Fabrication.................................................................................. 29
2.3.1 Photolithography............................................................................. 30
2.3.2 BOE Etching................................................................................... 32
2.3.3 XeF
2
Etching .................................................................................. 33
2.3.4 CO
2
Laser Reflow........................................................................... 34
2.4 Testing Procedures ................................................................................. 35
2.4.1 Testing Setup.................................................................................. 35
2.4.2 Broad Scan and Fine Scan............................................................... 39
2.4.3 Intrinsic Quality Factor ................................................................... 43
2.4.4 Tracking Resonant Wavelength Shifts............................................. 44
vi
2.4.5 Observing Thermal Effects ............................................................. 47
Chapter 2 References ......................................................................................... 48

Chapter 3 Low Loss Silica on Silicon Waveguides................................................. 53
3.1 Introduction............................................................................................ 53
3.2 Background and Motivation.................................................................... 53
3.3 COMSOL Modeling of Waveguides....................................................... 55
3.4 Fabrication Process................................................................................. 59
3.4.1 Photolithography and Etching ......................................................... 59
3.4.2 Fabrication Process – CO
2
Laser Reflow......................................... 60
3.4.3 Fabrication Process – End Firing..................................................... 61
3.4.4 Fabrication Process – Two Photolithography Steps ......................... 66
3.4.5 Testing Setup.................................................................................. 70
3.5 Results and Discussion ........................................................................... 75
3.5.1 Loss vs. Length, Polarization, and Input Power............................... 75
3.5.2 Efiron Cladding .............................................................................. 78
3.6 Conclusion ............................................................................................. 79
Chapter 3 References ......................................................................................... 80

Chapter 4 High Refractive Index Sol-gel Silica and Applications ........................... 82
4.1 Introduction............................................................................................ 82
4.2 Background and Motivation.................................................................... 83
4.3 Experimental Approach .......................................................................... 85
4.4 Results: Material Characterization .......................................................... 90
4.4.1 Infrared Spectra .............................................................................. 90
4.4.2 Refractive Index.............................................................................. 91
4.4.3 Absorption Coefficient.................................................................... 92
4.4.4 Thermo-Optic Coefficient ............................................................... 96
4.5 Application: Tailoring the Behavior of Light ........................................ 100
4.5.1 Importance of Q, Mode Volume, and Circulating Power ............... 100
4.5.2 COMSOL Modeling ..................................................................... 102
4.5.3 Results and Experimental Verification .......................................... 105
4.6 Application: Titanium-Enhanced Raman Lasers ................................... 110
4.6.1 Background and Motivation.......................................................... 110
4.6.2 Experimental and Theoretical Approach ....................................... 112
4.6.3 COMSOL Results......................................................................... 114
4.6.4 Experimental Results .................................................................... 116
4.7 Conclusion ........................................................................................... 120
Chapter 4 References ....................................................................................... 121

Chapter 5 NanoWatt-Threshold Nd
3+
and Alumina-Doped Toroidal Microlaser ... 126
5.1 Introduction.......................................................................................... 126
5.2 Background and Motivation.................................................................. 129
5.2.1 Theory .......................................................................................... 129
5.2.2 Fabrication of Yb
3+
and Nd
3+
-doped Toroid Microlasers ............... 131
vii
5.2.3 Adding CaF
2
to Improve Lasing.................................................... 137
5.2.4 Adding Alumina to Improve Lasing .............................................. 140
5.3 Materials and Methods ......................................................................... 141
5.3.1 Neodymium and Alumina-doped Sol-gel Synthesis....................... 141
5.3.2 Making Doped Sol-gel Silica Toroids ........................................... 144
5.3.3 Testing Setup................................................................................ 146
5.3.4 Q and Effective Refractive Index Measurements........................... 148
5.3.5 Laser Characterization Procedures ................................................ 150
5.4 Experimental Results and Discussion.................................................... 152
5.4.1 Effective Refractive Index ............................................................ 152
5.4.2 Quality Factor............................................................................... 153
5.4.3 Lasing Wavelength ....................................................................... 155
5.4.4 Lasing Threshold .......................................................................... 156
5.5 Conclusion ........................................................................................... 159
Chapter 5 References ....................................................................................... 160

Chapter 6 Heterodyned Toroidal Microlaser Sensor ............................................. 165
6.1 Introduction.......................................................................................... 165
6.2 Background and Motivation.................................................................. 166
6.3 Experimental Approach ........................................................................ 169
6.3.1 Fabrication of Microlasers Doped with Nd
3+
and Alumina ............ 169
6.3.2 Heterodyned Testing Setup ........................................................... 170
6.4 Comparative Temperature Sensing Experiments................................... 172
6.4.1 Linewidth (Detection Limit) Measurements .................................. 174
6.4.2 Temperature Sensing Using Resonant and Lasing Wavelengths .... 175
6.4.3 Heterodyned Temperature Sensing Experiments ........................... 176
6.5 Experimental Results and Discussion.................................................... 178
6.5.1 Measured Linewidth Results ......................................................... 178
6.5.2 Comparing Heterodyned vs. Non-Heterodyned Approaches.......... 179
6.5.3 Biodetection Experiments ............................................................. 182
6.5.4 Future Work ................................................................................. 187
6.6 Conclusion ........................................................................................... 190
Chapter 6 References ....................................................................................... 191

Chapter 7 Studying Optical and Thermal Forces around Toroids.......................... 194
7.1 Introduction.......................................................................................... 194
7.2 Background and Motivation.................................................................. 195
7.3 Heating Theory and COMSOL Modeling ............................................. 201
7.4 Experimental Verification of COMSOL................................................ 209
7.4.1 Temperature-Sensitive Fluorescent Dyes ...................................... 210
7.4.2 Attaching Rhodamine B to Silica Toroids ..................................... 211
7.4.3 Rhodamine B Temperature Sensing Experiments.......................... 217
7.4.4 Fluorescent Bead Tracking Experiments ....................................... 220
7.5 Results and Discussion ......................................................................... 221
7.6 Conclusion ........................................................................................... 226
viii
Chapter 7 References ....................................................................................... 227

Chapter 8 Future Work......................................................................................... 231
8.1 Introduction.......................................................................................... 231
8.2 Sol-gel Synthesis and Soft Lithography ................................................ 231
8.3 Integrated Heterodyned Microlaser....................................................... 232
8.4 New Materials for Higher Performance Devices ................................... 234
Chapter 8 References ....................................................................................... 235

Bibliography ........................................................................................................ 237
Appendix A : Mentoring Projects......................................................................... 252
A.1 Polymer Mentoring Projects ................................................................. 252
A.2 Sol-gel Mentoring Projects ................................................................... 260

ix
List of Tables
Table 4-1: Recipes used to make TEOS, MTES R=0.1, and MTES R=0.3 sol-gels.
The amount of each component used is shown in the highlighted rows. ... 87
Table 4-2: Material absorption and thermo-optic coefficient data for high refractive
index sol-gels . ........................................................................................ 99
Table 5-1: Sample procedures used to make 0.05wt% neodymium-doped silica sol-
gels........................................................................................................ 133
Table 5-2: Recipe for CaF
2
sol-gel synthesis. ....................................................... 138
Table 5-3: Sample recipe for sol-gel silica doped with 0.1 mol% Nd
3+
and 2 mol%
alumina. ................................................................................................ 142
Table 6-1: Comparison of resonant wavelength, lasing wavelength, and heterodyned
sensing approaches. ............................................................................... 182

x
List of Figures

Figure 2-1: Building blocks of integrated optical devices. ........................................ 9
Figure 2-2: Total Internal Reflection between air and silica.................................... 11
Figure 2-3: Schematic of an optical waveguide. ..................................................... 13
Figure 2-4: Different waveguide geometries........................................................... 13
Figure 2-5: Whispering Galleries vs. Whispering Gallery Mode Resonators........... 16
Figure 2-6: Types of WGM optical resonators and approximate quality factors...... 18
Figure 2-7: SEM images of toroid fabrication after a) photolithography, b) etching,
and c) reflow......................................................................................... 30
Figure 2-8: Schematic of the photolithography process. ......................................... 31
Figure 2-9: Schematic of CO
2
laser reflow setup. ................................................... 34
Figure 2-10: Schematic of the basic toroid characterization setup.......................... 36
Figure 2-11: Taper puller setup (a) and optical microscope images of b) stripped and
c) tapered optical fibers......................................................................... 37
Figure 2-12: Representative broad scan of ~100 micron silica toroid at 1300-
1312nm. ............................................................................................... 40
Figure 2-13: Screenshot of oscilloscope during fine scan. The vertical scale is
intensity (0.5V/division) and the horizontal scale is time (1ms/division).
The purple line is the signal from the photodetector, and the white line is
the triangle wave from the function generator. The left half of the
triangle wave shows the forward scan, and the right half shows the
reverse scan. ......................................................................................... 42
Figure 2-14: In the undercoupled regime, Q increases linearly as % coupling
decreases. The intrinsic Q is the Q at 0% coupling............................... 43
Figure 2-15: Finite element model of electric field in toroid cross-section (D=20µm),
showing evanescent field leaking into surroundings (white arrow)........ 46
Figure 2-16: Schematic of biodetection experiment and wavelength shift vs. time
data....................................................................................................... 47
Figure 3-1: Absorption coefficient and refractive index vs. wavelength for silica,
silicon, and silicon nitride. .................................................................... 54
Figure 3-2: Suspended silica waveguide design. Elevated silica slabs (a) are
reflowed with a CO
2
laser, creating pairs of smooth, cylindrical
waveguides (b-d) which are air-clad and isolated from the silicon
substrate. .............................................................................................. 55
Figure 3-3: Representative COMSOL 3.4 screenshot of waveguide cross-section
used in simulations. .............................................................................. 56
Figure 3-4: COMSOL modeling of waveguide cross-sections and mode cross
sections at 658nm (a, c) and 1550nm (b, d). The mode profiles are almost
completely confined in the waveguide and are nearly Gaussian, indicating
that light is successfully confined into the waveguide and minimally
affected by the supporting membrane.................................................... 59
Figure 3-5: Schematic of the waveguide fabrication process................................... 60
Figure 3-6: Top view optical microscope image of extremely straight reflowed
waveguides........................................................................................... 61
xi
Figure 3-7: Top view SEM image of reflowed waveguide end prior to cleaving. .... 62
Figure 3-8: Scanning electron microscope image of hand-polished waveguide end.
The polishing technique broke the samples and left acrylic residue. ...... 63
Figure 3-9: Scanning electron microscope image of waveguide sample cleaved using
dicing saw. While the dicing saw breaks samples less often, the cut can
still be rough......................................................................................... 64
Figure 3-10: SEM image of as-fabricated waveguide cleaved using a diamond scribe.
............................................................................................................. 66
Figure 3-11: Thinning the waveguide's supporting membrane to smaller than the
wavelength traveling through it reduces loss, as shown schematically and
in COMSOL finite element simulations. ............................................... 67
Figure 3-12: Thinning the waveguide membrane with a second photolithography step
............................................................................................................. 67
Figure 3-13: Thin silica slabs have increased stress, as seen in these optical
microsope images. This stress causes bending (a, b) and uneven reflow
(c)......................................................................................................... 69
Figure 3-14: Schematic of waveguide characterization setup.................................. 71
Figure 3-15: Compared to coupling with a cleaved fiber end (a), the index matching
gel (b,c) is difficult to apply to the fiber end and increases coupling loss
and sample breakage............................................................................. 73
Figure 3-16: Experimentally measured loss versus length of waveguides at 658, 980,
and 1550nm wavelengths. The measured loss is [0.97±0.22, 0.82±0.27,
0.73±0.13] dB/cm at [658nm, 980nm, 1550nm]. The coupling losses are
measured to be [7.55, 6.89, 5.70] dBm at these wavelengths. ................ 76
Figure 3-17: The waveguide loss is constant at different polarization states and loss
increases linearly with increasing input powers..................................... 77
Figure 3-18: Loss data for waveguides with Efiron cladding. We measured a loss of
approximately 0.15dB/mm or 1.5 dB/cm, which is noticeably higher than
the ~0.8dB/cm loss of the uncoated waveguides.................................... 78
Figure 4-1: Hydrolysis (a) and condensation (b) reactions of TEOS, forming a silica
matrix (c) ............................................................................................. 84
Figure 4-2: Hydrolysis (a) and condensation (b) reactions of MTES. When titanium
butoxide is added, it also undergoes hydrolysis, allowing it to integrate
within the silica matrix (c). ................................................................... 85
Figure 4-3: Flowchart summarizing the sol-gel synthesis process........................... 89
Figure 4-4: FTIR spectra of thermally grown silica, the synthesized undoped silica,
and Ti-doped sol-gel silica films. When titanium butoxide is added, an
additional peak appears around 900cm
-1
which indicates formation of Si-
O-Ti bonds in the silica matrix.............................................................. 91
Figure 4-5: Spectroscopic ellipsometry data for R=0.3 sol-gel: a) ψ and b) Δ values.
............................................................................................................. 92
Figure 4-6: Refractive index vs. wavelength data for MTES and TEOS sol-gels (from
spectroscopic ellipsometry)................................................................... 92
Figure 4-7: SEM images of a) plain silica and b) TEOS-coated silica toroids. ........ 93
xii
Figure 4-8: Representative Q spectra for TEOS, MTES R=0.1, and MTES R=0.3-
coated toroids. ...................................................................................... 94
Figure 4-9: Representative COMSOL models for a) TEOS, b) MTES R=0.1, and c)
MTES R=0.3-coated toroids at the 633nm wavelength. The toroid in the
simulation is 110 microns in diameter, with a 10 micron minor diameter.
............................................................................................................. 95
Figure 4-10: The resonant wavelength shifts as temperature increases, enabling Δλ
versus ΔT to be measured. .................................................................... 97
Figure 4-11: Representative Δλ versus ΔT data for a) TEOS, b) MTES R=0.1, and c)
MTES R=0.3 coated toroids.................................................................. 98
Figure 4-12: Theoretical (lines) and experimental (points) Q vs diameter and Q/V vs.
diameter data for silica toroids at λ=1300nm. As diameter decreases, the
total quality factor is limited by radiation losses.................................. 101
Figure 4-13: Rendering (a) and SEM image (b) of sol-gel coated toroid. .............. 106
Figure 4-14: The presence of the high refractive index coating decreases the toroid’s
mode volume (a) and increases mode volume in film as a function of
diameter (b). FEM models of the optical field in the coatings at 1300nm.
As the coating's refractive index increases from c) 1.4479 to d) 1.5066
and e) 1.5895, the light shifts radially outward into the coating. .......... 108
Figure 4-15: Theoretical and experimental Q versus diameter data for toroids coated
with a) n=1.4479, b) n=1.5066, and c) n=1.5895 sol-gels. The Q/V vs
diameter is also shown for each coating in d). ..................................... 109
Figure 4-16: (a) Rendering and (b) scanning electron microscope image of silica
toroid coated with titanium-doped sol-gel. .......................................... 113
Figure 4-17: COMSOL simulations for toroid cross-sections with 177nm-thick
coatings (left inset). As the titanium concentration increases, the
refractive index increases. The amount of light in the coating also
increases slightly while the total mode volume decreases (right inset). 116
Figure 4-18: Representative Raman spectra of sol-gel silica film. Two characteristic
Si peaks are visible at 500cm
-1
and 1000cm
-1
. ..................................... 117
Figure 4-19: Degree of polymerization I
p
vs. titanium concentration. As the titanium
concentration increases, the ratio of the 500cm
-1
and 1000cm
-1
peaks
increases, indicating that titanium is changing the silica bonds and
increasing the Raman response. .......................................................... 117
Figure 4-20: Representative spectrograph data for sample with 10mol% Ti coating.
(a) When pumped near 775nm, the first Stokes Raman peak appears near
800nm, from which the emission vs. intensity plot is produced (inset). (b)
Cascaded Raman observed at higher input powers. ............................. 118
Figure 4-21: The Raman efficiency increases as titanium concentration increases,
indicating that titanium enhances Raman behavior. The normalized
Raman threshold shows an initial decrease and gradual increase (inset),
further illustrating the trade-off between titanium’s sensitizing effects,
which lower the Raman threshold, and titanium’s increased absorption
loss, which increases the threshold by lowering the quality factor. ...... 119
xiii
Figure 5-1: Absorption coefficient of water versus wavelength, as presented by Hale.
The absorption of water is low, only ~2.5m
-1
and ~12 m
-1
at 765nm and
1064nm wavelengths. ......................................................................... 132
Figure 5-2: Lasing results for neodymium-doped silica. Neodymium-doped toroids
(a) could be pumped around 765-781nm. High Q, broadening resonance
peaks (b) produced lasing in the 1080-1150nm range, depending on
coupling. The lasing threshold is also very low (d)............................. 135
Figure 5-3: CaF
2
crystals grown in silica film....................................................... 139
Figure 5-4: Synthesis of doped sol-gel silica films. The sol-gels are spin-coated onto
bare silicon wafers to produce a thin film (a). The samples are dried on a
75°C hot plate for 5 minutes to evaporate the solvent (b), and annealed in
a 900°C tube furnace to produce a silica layer (c). This process is
repeated once more to obtain two layers for a ~700nm thick film (d). . 144
Figure 5-5: Device fabrication. First, 80µm circles of photoresist are patterned on
the doped sol-gel silica using standard photolithography steps (a).
Buffered oxide etchant (BOE) etches the uncovered silica and the
photoresist is removed, leaving silica pads (b). XeF
2
etching isotropically
etches silicon, forming elevated silica microdisks (c). Finally, the
microdisks are reflowed with a CO
2
laser to produce silica toroids (d).145
Figure 5-6: Toroidal resonant cavity laser. a) PovRay rendering of an array of silica
toroidal cavities on a silicon substrate. Light from a tapered optical fiber
waveguide is coupled into the first cavity. b) Scanning electron
microscope image of as-fabricated silica toroidal cavity...................... 146
Figure 5-7: Characterization set-up. a) Light from the tunable pump laser is coupled
into and out of the toroid with a tapered optical fiber. The output is split
using a 90/10 splitter, where 90% of the signal goes to the optical
spectrum analyzer (OSA) and 10% goes to a photodetector (PD). The
OSA signal is seen on the computer using a GPIB (PCI GPIB) input. The
photodetector signal is monitor on an oscilloscope (PCI O-scope) which
is integrated in the computer. The tunable pump laser wavelength is
controlled using the function generator (PCI Func Gen). b) The signal on
the OSA includes light from the pump laser which was not coupled into
the device (non-coupled light) as well as the emitted light (emission) from
the toroid which is back-coupled into the taper. Since the same tapered
fiber is used for both input (765nm pump laser) and output (~900-940,
1050-1150nm) signals, it is not possible to couple both the pump
wavelength and the emission wavelengths at high efficiencies. ........... 147
Figure 5-8: The effective refractive index increases as the concentration of alumina
in the sol-gel increases. Inset: Top view image of an optical resonant
cavity coupled to a tapered optical fiber, as seen using the testing setup’s
machine vision system. ....................................................................... 153
Figure 5-9: The quality factor increases as the concentration of alumina increases.
This increase is directly related to the increase in refractive index and
follows the trend in effective refractive index. Inset: Representative
transmission spectrum and Lorentzian fit at ~777nm pump wavelength
xiv
for a toroid microlaser containing 0.1mol% Nd
3+
and 2 mol% alumina
(Q=1.17x10
6
)...................................................................................... 154
Figure 5-10: Lasing from the 1mol% alumina-sensitized sol-gel device. a) Lasing
spectrum from the optical spectrum analyzer (OSA). The pump
wavelength (780nm) and two cascades at approximately 920nm and
1050nm are visible. b) As the concentration of alumina is increased, the
lasing wavelength blue-shifts to shorter wavelengths. This result
indicates that the Nd
3+
clustering is reduced by alumina, and the Nd
3+

ions become more favorably coordinated in alumina’s Al-O matrix. ... 156
Figure 5-11: Lowest threshold lasing line from the 2mol% alumina-sensitized sol-gel
device. The curve shows a clear onset of lasing, with a threshold of
530nW. The inset is the lasing spectra just after the onset of lasing.... 157
Figure 5-12: Dependence of the threshold and the slope efficiency on the alumina
concentration. a) As the concentration of alumina is increased, the
threshold/Q ratio decreases. b) As the concentration of alumina is
increased, the slope efficiency increases. This direct relationship
indicates that the alumina is acting as a sensitizer for Nd
3+
lasing. ...... 158
Figure 6-1: Schematic of an optical heterodyne. The beat signal’s frequency equals
the frequency difference between the microlaser and the reference. .... 168
Figure 6-2: Fabrication of heterodyned toroid microlasers. Nd
3+
and alumina-doped
sol-gel is spin-coated onto silicon wafers (a, rendering). Then, silica
toroids are fabricated from the films (b, rendering). c) The actual finished
devices are ~40µm in diameter, as shown in the scanning electron
microscope image c) . ......................................................................... 170
Figure 6-3: Schematic of the heterodyned testing setup. ....................................... 170
Figure 6-4: Schematics of testing setups used to track a) resonant wavelength and
microlaser wavelength and b) heterodyned beat frequency.
Representative transmission spectra, as measured with these setups near
c) 778nm and d) 1064nm. ................................................................... 173
Figure 6-5: Representative screenshot of iPhone video used to simultaneously save
beat frequency and temperature versus time........................................ 176
Figure 6-6: Heterodyned beat frequency data with and without the function
generator. As can be seen, the function generator must be kept off to
avoid excessive noise which would otherwise obstruct the desired
measurement....................................................................................... 177
Figure 6-7: Representative lasing near 1067nm for Nd
3+
sample with 2 mol%
alumina............................................................................................... 178
Figure 6-8: Comparison of resonant wavelength (a), lasing wavelength (b), and
heterodyned detection approaches (c). ................................................ 179
Figure 6-9: Lasing in PBS buffer is observed near 900-920nm and at 1055-1070nm,
exactly within the desired range. This sample was unfunctionalized (bare
silica microlaser). ............................................................................... 183
Figure 6-10: Representative Q spectra of unfunctionalized microlaser in PBS buffer
near 780nm. The Q drops from ~10
6
to ~10
5
in PBS buffer due to water’s
higher absorption losses...................................................................... 184
xv
Figure 6-11: ESA screen image of splitting beat frequency observed when
heterodyning microlaser in PBS buffer. .............................................. 184
Figure 6-12: Two beat frequency peaks are observed in PBS buffer due to mode
splitting. The observed beat frequency in PBS buffer is very unstable
even in ambient buffer shown here. This is most probably due to thermal
fluctuations and the very high input powers needed to achieve
heterodynable lasing. .......................................................................... 185
Figure 6-13: Frequency distribution of beat signal’s two peaks due to mode splitting.
........................................................................................................... 186
Figure 7-1: Flow cell used in biodetection experiments. ....................................... 196
Figure 7-2: Representative resonant wavelength shift as seen in flow experiments.
........................................................................................................... 198
Figure 7-3: Preliminary data showed increases in resonant wavelength shift and k
d
as
toroid circulating power increases....................................................... 199
Figure 7-4: Time-dependent rectangular pulse used in COMSOL modeling. ........ 206
Figure 7-5: Sample rectangular pulse used in COMSOL modeling....................... 207
Figure 7-6: Representative rectangular pulsed wave used in COMSOL modeling. 208
Figure 7-7: Structure of rhodamine B. Its carboxyl group is useful for attachment
chemistry............................................................................................ 211
Figure 7-8: Schematic of the two EDC reactions which covalently bond amine and
carboxyl groups. ................................................................................. 212
Figure 7-9: EDC attaches Rhodamine B’s carboxyl group to amine groups on
toroids. ............................................................................................... 215
Figure 7-10: Fluorescence microscopy shows no emission from bare silica toroid
controls (a, b) but rhodamine B-functionalized toroids do emit light (c, d),
indicating successful attachment ......................................................... 217
Figure 7-11: Testing setup used for temperature dependent measurements. The setup
is enclosed inside a blackout curtain to block ambient light................. 218
Figure 7-12: Sample surface temperature vs. 1550nm input power calculations in
COMSOL. Depending on the quality factor at 1550nm and the input
power, the average temperature on the entire exposed silica surface can
increase significantly. ......................................................................... 219
Figure 7-13: Schematic of particle tracking photophoresis and thermophoresis
experiments. ....................................................................................... 221
Figure 7-14: The same emission peak is observed for rhodamine B in water when
using a spectrofluorimeter (a) and the spectrograph (b). ...................... 222
Figure 7-15: Rhodamine B emission vs. temperature data when the temperature is
ramped (a). The fluorescence has minimal hysteresis, as seen in (b)... 223
Figure 7-16: Theoretical and experimental data for heating in toroids. The
experimental data does not match the trend predicted by COMSOL.... 224
Figure 7-17: Strong red-orange fluorescence near 600nm is observed in a
concentrated solution of 2µm polystyrene beads, as viewed using 532nm
excitation laser and 532nm filter to block the green excitation light. ... 225
Figure 7-18: Polystyrene beads are pushed onto silicon substrate, preventing imaging
of the optical and thermal forces around toroids. The beads can be seen
xvi
clinging to the silicon substrate when the microscope is focused on the
toroid’s silicon substrate instead of the toroid. .................................... 226
Figure A-1: Thickness vs. spin speed data for Efiron PC-409AP polymer. ........... 253
Figure A-2: LD 700 Perchlorate laser dye. When pumped near 658nm, it emits near
700nm. ............................................................................................... 255
Figure A-3: Fluorescence microscope images of LD-700 dye in PMMA films. .... 256
Figure A-4: Attachment of PEG’s hydroxyl group to epoxy functionalized toroids in
an anhydrous basic environment. ........................................................ 258
Figure A-5: Representative AFM data from a cracked sol-gel silica film. ............. 262
Figure A-6: Optical microscope images of Tm
3+
and Ce
3+
films made by Mary and
Alejandro. The lowest dopant concentration had the least amount of
cracking.............................................................................................. 264

xvii
Abstract

Due to their favorable optical and material properties, silica-based materials
and devices have found many important applications throughout science and
engineering, especially in sensing, communications, lasers, and integrated optics.
Often, silica’s properties ultimately limit the performance of these applications. To
address this limitation, this thesis investigates the development of improved silica
materials and optical devices, including silica films, coatings, waveguides,
resonators, lasers, and sensors. Using sol-gel chemistry and microfabrication
procedures, custom silica materials and devices are developed to benefit many
applications.
In this thesis, it is first demonstrated how the low optical loss of silica enables
fabrication of low loss integrated waveguides and toroidal resonators with ultra-high
quality factors. Then, by adding various rare earth and metal dopants to sol-gel
silica, hybrid silica materials and devices are made with custom properties such as
high refractive index and lasing capabilities. Finally, several applications are
demonstrated, including the use of high refractive index coatings to control the
behavior of light, development of Raman and ultra-low threshold rare earth
microlasers, and a heterodyned microlaser sensor with significantly improved
sensing performance. Future applications and directions of this research are also
discussed.
1
Chapter 1 Introduction

1.1 Motivation

Recently, an increasing amount of research effort has focused on developing
integrated optical devices, which combine optical components along with electronics
and/or microfluidics to make devices with greatly improved performance. Such
technology could greatly benefit many areas of science and engineering, including
the development of faster, more sensitive sensors for improved healthcare and
defense, more rapid and efficient communications, and high performance optical
computing [1-6]. In order to achieve these goals, it is necessary to efficiently
confine light on a silicon chip alongside the other components. To confine light on a
chip, integrated optics often relies on total internal reflection between a high
refractive index core and low refractive index cladding [7].
With its high transparency, robustness, and minimal nonlinear effects, silica
is an especially important material for confining light in integrated optics
applications [8, 9]. Due to silica’s favorable properties, many integrated silica
devices have been developed. Among them, the silica toroid resonator shows great
promise, as it is fabricated directly on a silicon wafer [10]. Because silica has very
low material absorption losses and the silica toroid is extremely smooth, silica
toroids can confine light very efficiently and achieve ultra-high quality factors (Q) of
over 100 million [10, 11]. These ultra-high Q silica resonators are useful in many
applications, as light can be stored inside a resonator for longer periods of time,
enabling very high circulating powers (~100W) and circulating intensities
2
(~gigawatts/cm
2
) to be achieved. In addition, ultra-high Q optical resonators have
very sharp, narrow linewidth resonant peaks which can be observed in real-time
using a function generator, tunable laser, detector, and oscilloscope. By tracking
changes in the resonant wavelength peak, material properties can be studied and
sensing experiments can also be performed. Therefore, ultra-high Q silica toroid
resonators have found numerous applications in integrated optical devices [12],
communications [5, 9], sensing [3, 13, 14], materials characterization [15, 16], laser
development [13, 17], and fundamental physics studies [18].
Nevertheless, the performance of toroid resonators and other silica devices is
typically limited by silica’s properties, including the material absorption loss [11,
19]. To improve the performance or add additional properties, it is necessary to
develop improved optical materials. Using sol-gel chemistry and microfabrication
procedures, custom silica materials and devices can be designed for many
applications. The present work focuses on the development of hybrid sol-gel silica
materials and toroids with enhanced properties for integrated optics, lasing, and
sensing applications.
Initially, the fabrication and characterization of silica toroids and waveguides
using thermally grown silica is discussed [20]. Then, hybrid silica materials and
devices are made using a sol-gel method to dope silica with titanium, aluminum, and
rare earth ions. By customizing silica’s properties, improved materials and devices
are developed, including high refractive index silica films [16], high index coatings
to tailor light’s behavior in resonators [21], titanium-enhanced Raman lasers [22],
ultra-low threshold alumina-Nd
3+
lasers [23], and a heterodyned alumina-Nd
3+

3
microlaser sensor with improved performance [13]. Finally, numerous related
projects are described in the appendix, which were completed while mentoring and
collaborating with undergraduate and high school students on additional research
projects.
1.2 Chapter Overview

This thesis is organized as follows:

In Chapter 2, a detailed background on integrated optical devices is
presented, including information on both optical waveguides and toroid resonators.
In subsequent chapters, numerous completed research projects are summarized
which build upon these fundamentals.
Chapter 3 provides a detailed overview on developing a new kind of silica on
silicon waveguide [20]. By optimizing the fabrication process, these low-loss silica
waveguides can be fabricated directly on silicon wafers. Also included is
information on troubleshooting and optimizing fabrication, building the waveguide
testing setup, and characterizing the waveguides’ optical properties. These
waveguide devices have since been used in applications such as optical splitters [6]
and in delivering light to and from optical toroid-based sensors [12].
Chapter 4 outlines the development of high refractive index materials and
devices made from titanium-doped sol-gel silica. This includes details from three
published projects: two in collaboration with Brian Rose [16, 21], and an additional
collaboration with Nishita Deka [22]. First, in collaboration with Brian, high
refractive index films are fabricated using titanium-doped sol-gel silica and
characterized using optical toroid resonators [16]. Then, by coating silica toroid
4
resonators with these high index materials, we experimentally and theoretically show
how the behavior of light in the resonator can be tuned [21]. Finally, the work with
Nishita is presented, in which Raman lasers are made from toroids coated with the
titanium-doped sol-gel silica films. Because the titanium enhances silica’s Raman
behavior, an improved Raman laser is demonstrated [22].
In Chapter 5, sol-gel silica materials are implemented once again to develop
improved toroid microlasers for biosensing applications. Biodetection applications
typically require aqueous, buffered environments. Therefore, it is necessary to
develop a toroid microlaser which works well in water and emits light in the range of
an available tunable laser (for heterodyned sensing applications). However, no such
laser existed. This chapter outlines several attempts (using neodymium, ytterbium,
alumina, and CaF
2
dopants) to develop an improved microlaser for biosensing
applications. Ultimately, a custom alumina and neodymium doped silica microlaser
is developed and characterized [23]. This improved laser not only emits light in the
desired 1055-1070nm range, it also does so extremely well, with an ultra-low 530nW
lasing threshold and nearly 30-fold improved efficiency [23].
In Chapter 6, the newly developed alumina-neodymium toroid microlasers
are heterodyned to develop an ultra-sensitive optical sensor [13]. By combining the
toroid microlaser’s ~1064nm lasing output with a 1064nm reference laser, a low
frequency beat signal is produced. By tracking this beat signal, significantly
improved sensing performance can be achieved. Approaches to minimize noise and
improve the detection results are also outlined.
5
Chapter 7 presents further investigation of the performance of optical sensors,
specifically, the effects of the high intensity circulating light. Since optical toroid
resonators confine light at high circulating powers, this light can have significant
thermal and optical effects on the toroid and surroundings. These thermal and
optical effects are modeled in COMSOL Multiphysics. To verify the models,
experiments are also performed by attaching the temperature-sensitive rhodamine B
dye to toroids, and by tracking fluorescent polystyrene beads.
Finally, in Chapter 8, several future directions of all this work are examined,
including the fabrication of sol-gel devices using soft lithography, development of
improved lasers for the heterodyned microlaser sensor, and studying higher
performance optical materials. In addition, I had the great opportunity to mentor
many undergraduate and high school students on various research projects. While
these projects did not directly result in publications, their findings are still important
and useful in future work. Therefore, in Appendix A, I briefly summarize the main
goals and findings of numerous mentoring projects.
As seen in this work, sol-gel silica materials and optical toroid resonators
show tremendous potential to benefit many applications. Nevertheless, these optical
materials and devices still need to overcome many challenges. This work provides
another step forward, helping to make these materials and integrated optical devices
available to scientists, engineers, and researchers in many fields.
Chapter 1 References

1. E. J. Smith, W. Xi, D. Makarov, I. Monch, S. Harazim, V. A. B. Quinones, C.
K. Schmidt, Y. F. Mei, S. Sanchez, and O. G. Schmidt, "Lab-in-a-tube: ultracompact
6
components for on-chip capture and detection of individual micro-/nanoorganisms,"
Lab on a Chip 12, 1917-1931 (2012).

2. T. Yoshie, L. Tang, and S.-Y. Su, "Optical Microcavity: Sensing down to
Single Molecules and Atoms," Sensors 11, 1972-1991 (2011).

3. A. M. Armani, R. P. Kulkarni, S. E. Fraser, R. C. Flagan, and K. J. Vahala,
"Label-free, single-molecule detection with optical microcavities," Science 317, 783-
787 (2007).

4. D. H. Kim, Y. S. Kim, J. Amsden, B. Panilaitis, D. L. Kaplan, F. G.
Omenetto, M. R. Zakin, and J. A. Rogers, "Silicon electronics on silk as a path to
bioresorbable, implantable devices," Applied Physics Letters 95 (2009).

5. H. S. Hsu, C. Cai, and A. M. Armani, "Ultra-low-threshold Er:Yb sol-gel
microlaser on silicon," Optics Express 17, 23265-23271 (2009).

6. X. M. Zhang, and A. M. Armani, "Suspended bridge-like silica 2 x 2 beam
splitter on silicon," Optics Letters 36, 3012-3014 (2011).

7. B. Saleh, and M. Teich, Fundamentals of Photonics (Wiley-Interscience,
2007).

8. R. Kitamura, L. Pilon, and M. Jonasz, "Optical constants of silica glass from
extreme ultraviolet to far infrared at near room temperature," Applied Optics 46,
8118-8133 (2007).

9. M. R. Poulsen, P. I. Borel, J. Fage-Pedersen, J. Hubner, M. Kristensen, J. H.
Povlsen, K. Rottwitt, M. Svalgaard, and W. Svendsen, "Advances in silica-based
integrated optics," Opt. Eng. 42, 2821-2834 (2003).

10. D. K. Armani, T. J. Kippenberg, S. M. Spillane, and K. J. Vahala, "Ultra-
high-Q toroid microcavity on a chip," Nature 421, 925-928 (2003).

11. X. M. Zhang, H. S. Choi, and A. M. Armani, "Ultimate quality factor of
silica microtoroid resonant cavities," Applied Physics Letters 96 (2010).

12. X. M. Zhang, and A. M. Armani, "Silica microtoroid resonator sensor with
monolithically integrated waveguides," Optics Express 21, 23592-23603 (2013).

13. A. J. Maker, and A. M. Armani, "Heterodyned Toroidal Microlaser Sensor,"
Appl. Phys. Lett. 103, 123302 (2013).

14. S. Mehrabani, P. Kwong, M. Gupta, and A. M. Armani, "Hybrid microcavity
humidity sensor," Applied Physics Letters 102 (2013).
7

15. H. S. Choi, S. Ismail, and A. M. Armani, "Studying polymer thin films with
hybrid optical microcavities," Optics Letters 36, 2152-2154 (2011).

16. B. A. Rose, A. J. Maker, and A. M. Armani, "Characterization of thermo-
optic coefficient and material loss of high refractive index silica sol-gel films in the
visible and near-IR," in Optical Materials Express(OSA, 2012), pp. 671-681.

17. L. Yang, D. K. Armani, and K. J. Vahala, "Fiber-coupled erbium microlasers
on a chip," Applied Physics Letters 83, 825-826 (2003).

18. S. M. Spillane, T. J. Kippenberg, K. J. Vahala, K. W. Goh, E. Wilcut, and H.
J. Kimble, "Ultrahigh-Q toroidal microresonators for cavity quantum
electrodynamics," Physical Review A 71 (2005).

19. M. L. Gorodetsky, A. A. Savchenkov, and V. S. Ilchenko, "Ultimate Q of
optical microsphere resonators," Optics Letters 21, 453-455 (1996).

20. A. J. Maker, and A. M. Armani, "Low-loss silica-on-silicon waveguides,"
Optics Letters 36, 3729-3731 (2011).

21. A. J. Maker, B. A. Rose, and A. M. Armani, "Tailoring the behavior of
optical microcavities with high refractive index sol-gel coatings," Optics Letters 37,
2844-2846 (2012).

22. N. Deka, A. J. Maker, and A. M. Armani, "Titanium enhanced Raman
microcavity laser," Optics Letters 39 (2014).

23. A. J. Maker, and A. M. Armani, "Nanowatt threshold, alumina sensitized
neodymium laser integrated on silicon," Optics Express 21, 27238-27245 (2013).

8
Chapter 2 Background and Overview

2.1 Overview: Integrated Optical Devices

As science, engineering, and technology continue to advance, there is an
increasing need to make existing computers, instruments, sensors, and other devices
faster, more efficient, and more compact. In particular, there is a strong demand for
more powerful computers and instruments, faster and more portable
communications, and more sensitive and efficient sensors. For all of these
applications, integrated optical devices have already shown great promise in
improving performance and meeting this need [1-6]. Put simply, integrated optical
devices have all their optical components (such as lasers, waveguides, and
resonators) fabricated directly on a silicon chip. Placing these optical components
together on a single chip has several advantages [7]. First, combining all the optical
components together on silicon makes devices more compact, simplifying the
design. Also, many optical devices are fragile and require careful, precise alignment.
Integrating the optical components together on a chip helps overcome these
limitations and is an important step in making optical devices usable outside a
laboratory setting.
Another advantage of optical components which are integrated on silicon
chips is that they can be easily combined with electrical components and
microfluidics to make faster, more powerful sensors, computers and instruments.
This is especially important in communications applications, in which optical and
electrical signals must be converted to efficiently and rapidly transfer data.
9
A final important advantage of integrated optics is increased stability,
simplicity, and robustness. Integrating an optical device and all the necessary optics,
light coupling, and other components can greatly simplify the device and make it
more robust. Integrated devices can also be easier to fabricate in bulk and are more
readily used in environments outside the lab.
One important goal of the work presented in this thesis, as well as in the
entire Armani Research Group, is to develop the building blocks of integrated optical
devices (Figure 2-1), including on-chip waveguides and resonators [1, 7-10]. Optical
waveguides are needed to confine and deliver light to other components, while
resonators are commonly used as sensors, lasers, and even as optical memory. The
primary focus of the present work is on optical resonators, with the development of
integrated optical waveguides also discussed in Chapter 3.

Figure 2-1: Building blocks of integrated optical devices.

2.1.1 Refractive Index and Total Internal Reflection

To confine light inside an integrated optical device, it is important to consider
the device’s optical properties, especially the refractive index. The refractive index
is a property inherent to a material which describes the phase velocity of light in the
material, relative to the phase velocity of light traveling through a vacuum [11]. The
10
refractive index of materials can vary widely depending on the material, temperature,
and the wavelength of light. For example, while air has a refractive index of about 1,
water’s refractive index is about 1.33, the index of silica glass is near 1.45, and
silicon’s refractive index is near 3.4.
Interesting things can happen when light passes between materials of
different refractive indices. At the boundary between materials of different refractive
index, light can be reflected and/or transmitted, and light’s angle of propagation can
change. Snell’s law (Equation 2-1) provides a simple way to predict how the light
will be affected, based on the incident angle and refractive index of each material
[11]:

1 1 2 2
sin sin n n    (2-1)
Here, n
1
and n
2
are the refractive indices of the first and second medium, and the θ
values are the incident or refracted angles of the light in the first or second medium.
If a high refractive index material is surrounded by a material of lower refractive
index, it is possible for all the light to be reflected back inside. Total internal
reflection is the phenomenon in which light is trapped in a high refractive index
material surrounded by a medium of lower refractive index.
Many integrated optical devices, including the toroids and waveguides in the
present work, rely on total internal reflection to confine light. For example, in the
silica devices fabricated in this work, light is frequently confined between an air-
silica interface. Since the refractive index of silica glass (~1.44) is larger than that of
air (~1), light becomes trapped in the silica.
11
For total internal reflection to occur, the incident light must approach the
interface at an angle greater than the critical angle [11]. The critical angle θ
c
is
defined as the angle at which the refracted angle is 90°, given by Equation 2-2:

2
1
n
c
n
arcsi
n

 

 
 
(2-2)
If the incident light in the silica approaches the interface at an angle which is larger
than the critical angle, all of the light will be reflected back inside the silica.
Otherwise, if the incident light approaches at a smaller angle than the critical angle,
some of the light may leak out of the silica (Figure 2-2). Therefore, it is extremely
important to consider the geometry and refractive index of silica devices to ensure
proper confinement of light and the critical angle. For example, radiation or bending
losses occur when light approaches at an angle smaller than the critical angle.

Figure 2-2: Total Internal Reflection between air and silica.

2.1.2 Absorption Loss

The refractive index is actually a complex number with both real and
imaginary parts, given by Equation 2-3 [7, 11]:
12
ik n n  
~
(2-3)
The value n is the real part of the refractive index, and describes the phase velocity
of light as explained in the previous section. The imaginary part, k, is also known as
the extinction coefficient of the material and represents how much light is absorbed
by the material [11]. Even highly transparent materials, such as silica glass, have
nonzero k values and absorb some of the light passing through [12]. In fact, the
performance of optical devices can be limited by the absorption of light. Therefore,
it is important to consider the k values.
From the k value, the optical absorption coefficient α can be determined at a
wavelength λ
0
using Equation 2-4:

0
4



k
 (2-4)
In this equation, k is the imaginary refractive index at the wavelength λ
0
. The higher
the optical absorption coefficient, the more light is absorbed and the higher the loss.
Therefore, it is important to use materials which have small k and α values in optical
devices. No device is perfect, so loss of light is inevitable. Yet lowering the optical
loss of devices can vastly improve performance.
2.1.3 Optical Waveguides

To make optical toroids and other devices more robust, efficient, and truly
integrated on silicon, it is desirable to fabricate on-chip components for delivering
light to and from devices. This can be accomplished using optical waveguides.
Optical waveguides confine light using total internal reflection between a dielectric
core and a lower refractive index cladding [7, 11]. To effectively confine light, the
13
refractive index of the waveguide core must be higher than the cladding; otherwise,
the light will not be confined by total internal reflection (Figure 2-3).

Figure 2-3: Schematic of an optical waveguide.

Many different waveguide geometries have been developed using many
different materials, as seen in Figure 2-4 [7, 13-17]. While fiber geometries are
useful for communications, they cannot be easily integrated on a silicon chip. On-
chip waveguides are often fabricated in embedded, strip loaded, or buried
geometries. Waveguides fabricated in ring or circular geometries have also been
fabricated; however, these act as optical resonators [18, 19].

Figure 2-4: Different waveguide geometries [7].

14
One of the most important metrics of waveguide performance is the loss, or
how much light is lost while traveling through the waveguide. Ideally, a waveguide
would confine light very well and have minimal optical loss. However, various
sources of loss, such as defects, surface roughness, material absorption, and
contamination can affect waveguide performance. Therefore, waveguides which
have low loss across a wider range of wavelengths, input powers, and polarizations
are especially desirable in integrated optics applications.
The loss α of a waveguide can be calculated based on its input and output
power using Equations 2-5 and 2-6:

L
out initial
P P e
 
 (2-5)

system waveguide coupling
       (2-6)
Here, L is the waveguide length, P
initial
is the input power into the waveguide, P
out
is
the measured output power of the waveguide, and α is the waveguide’s total loss [8,
20]. Using Equations 2-5 and 2-6, the value of the loss α can be experimentally
determined by measuring the output power P
out
of a waveguide of known length L
and input power P
in
. By plotting the alpha values as a function of length, the
waveguide loss can be isolated from the system and coupling losses. Specifically,
the length-dependent α
waveguide
is determined from the slope of the line, and the value
of α
system
+ α
coupling
is given by the y-intercept of the line. Also note that the resulting
value of alpha is often given in logarithmic dB units, where
15

0
( ) 10log
m
i
P
Loss dB
P
 
   
 
 
(2-7)
For the dB
m
units, P
0
is the measured output power and P
i
is a reference
power, taken to be 1mW. In Chapter 3, the development and characterization of an
integrated silica waveguide is described in greater detail.
2.1.4 Optical Resonators

Optical resonators (also known as optical cavities or resonant cavities)
confine certain wavelengths of light inside a volume or cavity [7, 21]. Just as a
tuning fork vibrates at a specific note or acoustic frequency, optical resonators trap
light of specific optical frequencies or wavelengths. The wavelengths of light which
are confined in a resonator are known as the resonant wavelengths. These resonant
wavelengths can vary depending on the geometry and material properties of the
resonator.
While many kinds of optical resonators have been developed, they fall into
two major categories: Fabry-Perot resonators and Whispering Gallery Mode (WGM)
optical resonators [7, 21]. Fabry-Perot resonators use two or more mirrors to confine
a standing wave of light. Whispering gallery mode resonators confine light in
circular orbits, and are named after the bouncing of sound along the walls of a
whispering gallery, such as the one in St. Paul’s Cathedral (Figure 2-5).
16

Figure 2-5: Whispering Galleries vs. Whispering Gallery Mode Resonators [22].

Fabry-Perot and whispering gallery mode optical resonators can be made in
many different geometries. Fabry-Perot resonators can be made in planar-mirror,
spherical-mirror, ring-mirror, rectangular cavity, and fiber-ring geometries, among
others. There are various kinds of WGM resonators, including the micropillar,
microdisk, photonic crystal, microring, microsphere, and microtoroid [21, 23-28].
Photonic crystals can be made into either Fabry-Perot or WGM resonators depending
on how the defects or holes are arranged in the crystal.
One of the most important parameters of an optical resonator is the loss, or
how well the resonator confines light. This loss can be quantitatively described by
the quality factor, or Q [7]. The quality factor is directly related to the photon
lifetime, or how long light stays confined in the resonator. The lower the resonator’s
loss, the higher the Q of the cavity is. Therefore, using ultra-high Q resonators is
crucial to improve optical devices and minimize optical loss.
Minimizing optical loss is one of the main reasons this work focuses on
optical toroid resonators and silica materials. Silica has very low optical absorption
loss (high transparency) across a wide range of wavelengths and can be etched using
17
known microfabrication procedures, enabling many different optical resonator
geometries to be produced. While silica microsphere resonators can achieve
extremely high quality factors of over 1 billion, these devices are not easily
integrated on silicon. On the other hand, silica microdisks and microrings are more
easily integrated on silicon, but suffer from lower quality factors due to surface
roughness from fabrication. The silica toroid resonator is a unique compromise
between the high Q microsphere resonator and the lower Q but on-chip microdisk
(Figure 2-6) [28]. The silica toroid is a microdisk which has been reflowed with a
CO
2
laser, producing an extremely smooth doughnut-shaped surface with negligible
surface roughness and material-loss limited quality factors of over 10
8
(100 million)
[28, 29]. Silica toroids can also be fabricated from hybrid sol-gel silica films spin-
coated onto silicon wafers. Therefore, in pursuit of improved integrated optical
devices, we focus primarily on studying the optical toroid resonator, as it offers the
best compromise between achieving ultra-high Q factors and being integrated
directly on a silicon chip.
Since the silica toroid’s invention in 2003, it has already been used in many
integrated optics applications [28, 30-34]. Silica toroids doped with rare earth metals
can be used to make ultra-low threshold lasers [5, 33, 35-37]. In addition, silica
toroids have been used in numerous sensing applications and are sensitive enough to
detect single molecules [34, 38]. Silica toroids are also used as important tools in
material characterization [39, 40], fundamental physics studies [32], and
communications applications [5]. The key properties of silica toroids and other
resonators are summarized in the next sections.
18

Figure 2-6: Types of WGM optical resonators and approximate quality factors [21, 23, 24, 28, 41,
42].

2.2 Overview of Toroid Theory

2.2.1 Quality Factor (Q)

An ideal optical resonator could confine light perfectly for an infinite period
of time and therefore have no optical loss. In reality, optical resonators can only
confine light for finite periods of time because there are numerous sources of optical
loss which affect performance. The quality factor, or Q, is a dimensionless number
which quantitatively describes the loss of an optical resonator and is given by
Equation 2-8:
 




 Q (2-8)
Here, λ is the resonant wavelength of the resonator, Δλ is the full-width at half
maximum value of the resonant peak, ω is the resonant frequency, and τ is the
19
photon lifetime in the resonant cavity. As seen in the equations, the Q of a resonator
is directly proportional to the photon lifetime in the cavity, and inversely
proportional to the linewidth. It is also worth noting that these expressions for Q in
terms of photon lifetime and linewidth are equivalent, and related by a Fourier
transform. As a result, either expression can be used.
Experimentally demonstrated optical resonators have Q factors ranging from
roughly a thousand to over 1 billion. Because the Q is directly proportional to the
intensity of the optical field inside the cavity, many applications of optical resonators
require the highest possible Q.
It is possible to theoretically calculate the quality factor using computer
modeling such as COMSOL [43, 44] (discussed in Chapter 4). To experimentally
measure the quality factor, there are two main methods [28]:
1) Linewidth measurement: Measure the FWHM (full width half maximum)
of the Lorentzian-shaped resonance in the undercoupled regime. This is
done by scanning a single-mode laser through a resonance (for high Q
devices) or using a broadband light source and measuring the linewidth
on a spectrograph (for low Q devices).
2) Cavity ringdown: Measure the photon lifetime τ experimentally at a
known resonant frequency. This is done by repeatedly scanning a laser
into resonance using a mode that is critically coupled to the taper.
Generally, it is best to use cavity ringdown measurements with only extremely high
Q (Q > 5x10
8
) resonators. Since cavity ringdown requires direct measurement of
photon lifetimes, which are very short, it may not be possible to accurately and
20
precisely measure extremely short photon lifetimes of lower Q resonators. Also, for
very high Q resonators, the resonant peak may be narrower than the laser’s
linewidth, preventing measurement by the linewidth method. For devices with lower
quality factors (Q < 5x10
8
), the linewidth method is preferred. Since the present
work involves resonators with Q values below 5x10
8
, the linewidth method is used.
2.2.2 Circulating Power

The amount of power P
circ
circulating inside a whispering gallery mode
optical resonant cavity is given by Equation 2-9:

R n
QP
P
eff
in
circ
2


 (2-9)
Here, λ is the wavelength, Q is the resonator’s quality factor, n
eff
is the effective
refractive index of the resonator, P
in
is the input power of the resonator, and R is the
resonator’s radius. As seen in this relation, resonators with high quality factors and
small diameters can achieve especially high circulating powers. Therefore, for
applications requiring high circulating power (such as lasers), the radius R can be
optimized. For example, the radius should be made very small to enable high
circulating powers while not so small that the radiation losses due to bending reduce
the quality factor.
2.2.3 Mode Volume

The mode volume V
m
for an optical toroid resonator can be determined using
Equation 2-10:
21
(2-10)
Here, E

is the electric field strength, ( ) r 

is the value of refractive index at r


squared, and
max
E is the maximum electric field strength. V
Q
is a quantization
volume, or the volume over which the expression is integrated [32]. The mode
volume can be calculated by simulating the light in the optical resonator, and
integrating the resulting electric field intensity profile (see section 4.5.2).
2.2.4 Finesse

The finesse of an optical resonator is an alternative approach for expressing
the quality factor. Unlike the quality factor, the finesse does not account for
propagation effects within the resonator [21]. The finesse can be calculated using
Equation 2-11:

) (
2
)
2
(
2
2
2
L
F Where
d
F
c
d
t Where
c
d c
L L
t c
L
t c
Q








 
  
   



  
(2-11)
2.2.5 Purcell Factor

The Purcell factor is another measurement of how suitable an optical
resonator is for laser applications. It is given in Equation 2-12:
22

3
2
3
4
P
m eff
Q
F
V n


 

 
 
 
(2-12)
where n
eff
is the effective refractive index, V
m
is the mode volume, and λ is the
resonant wavelength of the cavity [45]. Devices with high Purcell factors are good
candidates for laser applications, because the high Q and small mode volume allow
high circulating powers to be achieved.
2.2.6 Effective Refractive Index

The effective refractive index describes the net refractive index “felt” by the
light confined in the resonator. It is therefore an average of the refractive indices of
the resonator material and the surroundings, weighted based on the amount of light
contained in each region, as shown in Equation 2-13:

eff resonator coating air
n n n n       (2-13)
In the above equation, χ is the fraction of light confined in the resonator, γ is the
fraction of light confined in the resonator’s coating (if applicable – this term is
omitted if there is no coating), and δ is the fraction of light in the surroundings [39,
40]. The coefficients χ, γ, and δ are determined using COMSOL finite element
modeling or other numerical methods [43, 46]. The refractive index values for each
component can be determined from literature or experimentally using ellipsometry.
2.2.7 Effective Absorption Coefficient

The effective absorption gives the net amount of light absorbed by the
resonator and its surroundings. Similar to the effective refractive index, it is a
23
weighted average of the absorption coefficients of the resonator and its surroundings,
as given by Equation 2-14:

eff resonator coating air
          (2-14)
In the above equation, χ is the fraction of light confined in the resonator, γ is the
fraction of light confined in the resonator’s coating (if applicable), and δ is the
fraction of light in the surroundings [39, 40]. The coefficients χ, γ, and δ are
determined using COMSOL finite element modeling or other numerical methods
[46]. The alpha values are determined from the literature or experimentally [12, 40].
The absorption coefficient of air is generally assumed to be zero.
2.2.8 Free Spectral Range

The free spectral range of a resonator describes the distance between
sequential fundamental mode peaks [28]. The free spectral range of WGM
resonators is given by Equation 2-15:

R n
FSR
eff


2
2
 (2-15)
Here, λ is the wavelength of light in the resonator, R is the radius, and n
eff
is the
effective refractive index.
2.2.9 Total Quality Factor

The quality factor (Q) of whispering gallery mode resonators, including the
optical toroid resonator, is given by Equation 2-16 [47, 48]:
24

1 1 1 1
. .
1 1      
    
coup mat cont s s rad
Q Q Q Q Q Q (2-16)
Here, Q
rad
represents radiation or bending losses caused by curvature of the
resonator, Q
s.s.
denotes surface scattering losses due to defects or roughness, Q
cont
is
losses caused by contamination on the device, Q
mat
represents intrinsic absorption of
light by the resonator’s material, and Q
coup
is the coupling loss. Each of these terms
is described in further detail in the following sections.
It is also important to distinguish between the loaded and intrinsic quality
factors of the resonator. The general quality factor equation can be rewritten in
terms of the intrinsic and loaded Q, as in Equation 2-17:

1 1 1
intrinsic loaded coupling
Q Q Q
  
  (2-17)
The loaded Q is the total loss of the resonator, including coupling losses. Therefore,
the loaded quality factor includes all the aforementioned terms (radiation, surface
scattering, contamination, material, and coupling losses). The intrinsic Q includes
only losses which are inherent to the resonator and not dependent on external factors
such as coupling and testing conditions. Therefore, the intrinsic quality factor
includes the radiation, surface scattering, material, and contamination losses, but
does not consider coupling. When performing a linewidth experiment, one
measures the loaded Q of the cavity. By varying the amount of power coupled into
the cavity or the coupling condition, the intrinsic Q can be determined [39, 49].
2.2.10 Radiation Loss

The radiation loss is the loss caused by bending of the resonator. In this
work, it is calculated using Equation 2-18:
25

) ( 2
) (
r
r
rad
f
f
Q


 (2-18)
where ( )
r
f  and ( )
r
f  are the resonant frequency’s real and imaginary components,
respectively [43]. The radiation losses can also be found by solving the analytic
characteristic equation for the real and imaginary wavenumbers [32, 50].
Radiation loss arises due to total internal reflection and the critical angle. As
the size of the optical resonator decreases, the angle at which the light reflects inside
the resonator decreases and becomes smaller than the critical angle predicted by
Snell’s law. As a result, when the toroid gets small enough, radiation losses become
important. Generally, the radiation loss is negligible for resonators with very large
diameters [44, 47, 48]. However, the loss can become significant as diameter
decreases below approximately 20µm in air (or ~80-90µm in water, due to the
reduced refractive index contrast between water and silica [50]). Radiation losses
are considered in more detail in Chapter 4.
2.2.11 Contamination Loss

The contamination losses represent losses caused by dust or other
contaminants on the resonator and can be minimized by keeping the samples clean
[47, 48]. There is no distinct mathematical expression for the contamination losses,
although it is possible to control and minimize them.
2.2.12 Surface Scattering Loss

Surface scattering loss in whispering gallery mode resonators arises due to
roughness or nanoscale non-uniformities on the surface of the resonator. It is given
by Equation 2-19:
26

2
2 2
2
ss
D
Q
B

 
 (2-19)
where λ is the wavelength of light in the resonator, D is the resonator’s diameter, and
σ and B are the root-mean squared values of roughness size and correlation length,
respectively [23]. The surface scattering losses can be minimized by fabricating very
smooth samples with minimal surface roughness [47, 48]. For example, optical
toroid resonators are assumed to have negligible surface roughness due to the reflow
fabrication step [28].
2.2.13 Material Absorption Loss

The material absorption loss of a resonator accounts for absorption of light by
the resonator itself. Dopants or surface chemistry can also increase material loss.
The material loss is given by Equation 2-20:

2
eff
mat
eff
n
Q

 

(2-20)
Here, λ is the wavelength in the resonator. The
eff
n and
eff
 terms are the effective
refractive index and effective material absorption, respectively [47]. To calculate the
eff
n

and
eff
 terms, we take the weighted average of the material properties
experienced by light in the resonator and in the surroundings, as described in the
previous sections.
The material loss is an inherent property of the material from which the
resonator is made. While the material loss can be minimized by fabricating
resonators with low loss materials (such as silica), the material loss is typically a key
limiting factor in the overall quality factor of a resonator [47, 48]. For example,
27
while the material-limited Qs of silica toroid resonators are on the order of ~500
million [48], CaF
2
resonators can achieve Qs of well over 1 billion due to the lower
absorption losses of CaF
2
[51].
2.2.14 Coupling Loss

The coupling loss of optical resonators measures how efficiently light is
transferred into and back out of the optical resonator. Depending on the coupling
methods and input powers used, the coupling losses can vary greatly. It is possible
to minimize coupling losses by using efficient coupling methods, such as tapered
optical fibers, and also by testing the devices in the under-coupled regime where the
resonator’s intrinsic losses are dominant [52].
2.2.15 Which Losses Dominate?

As mentioned in the previous sections, many different factors can influence a
WGM resonator’s quality factor. However, many of these quality factor terms can
be neglected under certain conditions. If the resonator is assumed to be clean (free
of contamination), larger than ~30µm in diameter in air or ~100µm in water (not
limited by radiation losses), tested with light efficiently coupled inside (low coupling
loss), and smooth (no scattering loss), then the resonator’s Q depends predominantly
on Q
mat
, the material absorption loss [47, 48].
Therefore, considering the aforementioned loss sources, each of the common
WGM resonator geometries can achieve different quality factors. Depending on the
material, geometry, and fabrication process, the Q of a WGM resonator can be
limited by material losses, radiation losses, surface scattering losses, or a
28
combination of them. Microsphere resonators can achieve material loss-limited
ultra-high quality factors from 100 million to over 1 billion (10
9
) [47]. Microtoroid
resonators can also have material loss-limited Qs of over 100 million [48]. (Since
the material absorption loss of thermally grown silica is slightly higher than that of
fused silica in optical fiber [12, 48], microtoroid resonators have a lower material
absorption-limited Q compared to spheres). Microrings and microdisks have
increased surface roughness, and typically have surface scattering-limited Qs
between 10
3
-10
6
, but can achieve Qs as high as 10
8
depending on the etching process
used and resulting roughness [21, 23].
The difference in Q between various resonator types arises primarily due to
surface roughness/defects which cause scattering loss. Microspheres form due to the
surface tension of melted silica, and therefore have an atomically smooth surface
enabling very high Q values. Microtoroids are also formed using surface tension
effects, making them very smooth [28, 29]. Microdisks are formed based only on
lithography, and can have significantly more roughness. By optimizing the etch
processes, it is possible to reduce this roughness. Microrings have an inner and outer
lithographically defined surface, making them even rougher than microdisks and
giving them lower Q values. The Q can also be affected by the presence of surface
chemistry, coatings, or dopants in the resonator which increase material and surface
scattering losses [30, 53].
2.2.16 Advantages and Disadvantages of High Qs

There are many important advantages of high Q optical resonators, as well as
some disadvantages [21]. The main advantages are 1) the ability to confine photons
29
for long periods of time and achieve significant circulating power in the cavity, and
2) the narrow linewidth of the resonant peak. High circulating intensities are
especially important to achieve low threshold lasers and study nonlinear optical
phenomena. A narrow linewidth is especially useful for applications in sensing as it
allows high resolution tracking of resonant wavelength shifts.
There are also some important disadvantages. First, depending on the
material and geometry used, high Q resonant cavities can be difficult or impossible
to make on-chip. For example, silica glasses can be patterned relatively easily using
standard photolithography and etching procedures. However, some materials such as
polymers and fluoride glasses require different fabrication procedures like hand
polishing or replica molding, and may not be able to withstand the high temperatures
and harsh chemicals used in silica processing [42, 51]. Also, because of the high
circulating intensities of light, high Q resonators are much more sensitive to thermo-
optic effects and radiation pressure based oscillations [21, 54, 55]. In addition, it can
be difficult to couple light into the resonator, as the coupled light must be phase-
matched with the cavity (this can be difficult if the resonator linewidth is extremely
narrow). Finally, high Q resonators are very sensitive to dust and contamination
which increase contamination losses and easily degrade the Q.
2.3 Device Fabrication

One of the most important advantages of the silica toroid is the ability to
achieve quality factors of over 100 million in a resonator which is integrated directly
on a silicon chip. These ultra high quality factors are a direct result of the toroid’s
fabrication process [28, 29]. Silica toroids are fabricated using three main steps
30
(Figure 2-7). First, photolithography is used to define silica circles on silicon wafers
(Figure 2-7a). Then, XeF
2
etching isotropically removes silicon, elevating the silica
circles to make silica microdisks (Figure 2-7b). Finally, the silica microdisks are
reflowed using a CO
2
laser to produce smooth silica toroids (Figure 2-7c). These
three steps are described in further detail in the following sections.

Figure 2-7: SEM images of toroid fabrication after a) photolithography, b) etching, and c) reflow
[29].

2.3.1 Photolithography

The photolithography is done as follows and as illustrated in Figure 2-8 [29].
Intrinsic (dopant-free) silicon wafers with a 2µm thick layer of thermally grown
silica (WRS Materials) are cleaned using acetone, methanol, isopropanol, and
deionized water, and then dried on a hot plate at 120°C for 2 minutes. Then, HMDS
is applied to the wafers to improve adhesion between the photoresist and the silica.
Using a spinner, a thin, uniform layer of S1813 photoresist is applied to the wafers at
500 rpm for 5 seconds, then 3000rpm for 45 seconds. After spin coating, the
photoresist is soft baked on a hot plate for 2 minutes at 95°C to adhere it to the
wafers and evaporate the photoresist’s solvent.
31

Figure 2-8: Schematic of the photolithography process.
After soft baking, the photoresist is exposed to UV light through a
photolithography mask. The S1813 photoresist is a positive photoresist, so any UV-
exposed photoresist will be dissolved in the subsequent developer step, while all the
covered photoresist which was not hit by UV light remains intact.
To develop the S1813 photoresist, the MF-321 developer is used. This
developer dissolves the UV-exposed photoresist, but not the unexposed photoresist.
The MF-321 developer is poured into a glass beaker and the samples are immersed
in the developer one at a time. Almost immediately, the patterned photoresist can be
visibly seen as the exposed resist is removed from the sample. During developing,
the beaker is swished in order to gently stir the solution. This ensures the developing
process occurs uniformly and is not limited by diffusion. It is important to watch the
32
developing process carefully. When no more photoresist can be seen exiting the
sample, the sample should be immediately removed, rinsed under running water, and
inspected with an optical microscope. All the UV exposed photoresist should
dissolved, leaving behind the desired pattern from the mask (usually arrays of
circles). Overdeveloping samples will cause some of the desired photoresist to be
etched and damage the desired patterns.
Once the developing step is complete, the samples are then rinsed with
deionized water and hard baked on a hot plate at 120°C. This hard baking step heats
the photoresist above its glass transition temperature, which helps repair any
roughness or defects caused by the developing process.
2.3.2 BOE Etching

To etch the silica not covered by the photoresist, the samples are immersed in
buffered oxide etchant (BOE). For safety, HF-resistant gloves and Teflon containers
are used, and calcium gluconate gel is kept readily available in case of a spill. The
improved buffered oxide etchant from Transene is used to improve etch results.
BOE contains HF, which etches silica to form circular silica pads on the silicon
wafer. For a 2 micron layer of thermal oxide silica, this etching process takes
approximately 18-22 minutes. However, the exact etch rate can vary greatly
depending on the ambient temperature, amount of silica to be etched, and porosity of
the sample. For example, some sol-gel silica samples can etch completely in only 3-
5 minutes. Therefore, it is important to watch the samples carefully during BOE
etching. Once the sample wafers appear silver in color and are hydrophobic (water
beads up), then the silica layer has been etched completely and the silicon is
33
exposed. Once BOE etching is complete, the samples are thoroughly rinsed under
running deionized water. The final step of the photolithography process is to remove
the photoresist, leaving behind the silica etched in the desired pattern. Prior to
removing the photoresist, it is good practice to inspect the samples using an optical
microscope to ensure no further etching is needed.
2.3.3 XeF
2
Etching

Once the photolithography steps are complete, the silicon underneath the
silica circles is etched using XeF
2
gas, producing silica microdisks which are
elevated above the silicon substrate (Figure 2-7b). It is important to note that XeF
2

etches silicon isotropically, or equally in all directions. This property allows the
XeF
2
to etch underneath the silica to form pillars. (Note that the buffered oxide
etchant used in the photolithography step is a semi-isotropic etchant, since it etches
primarily in a downward direction, only partially undercutting the photoresist mask).
It is also worth noting that the etch rate of the XeF
2
etcher exhibits a loading
effect, meaning the etch rate is highly dependent on how much silicon is present in
the etch chamber. While it is tempting to etch only a few samples at a time to finish
quickly, a much smoother and more uniform etch can be achieved if more samples,
and more silicon, are etched at a time. When etching sets of 16 ~5mm
2
samples, it
usually takes 30-35 80 second etch cycles (at 2800mTorr XeF
2
pressure) to etch 80
micron disks. For 160 micron disks, 60 or more cycles may be needed. Etching the
samples more slowly and carefully produces a very smooth and uniform etch which
can help improve the final devices. Another interesting effect of the XeF
2
etch is
that it fluorinates the silicon substrate [56]. This fluorination can prevent the silicon
34
substrate from being modified by surface functionalization steps. This fluorination is
very convenient when it is desired to attach molecules to the silica toroid’s surface
and not on the silicon substrate [30].
2.3.4 CO
2
Laser Reflow

It is possible to use the thermal oxide silica microdisks as optical resonators;
however, the quality factors are generally lower (~10
6
-10
7
) due to edge roughness
from the etching and photolithography steps. Therefore, to achieve the highest
possible quality factors, it is necessary to reflow the silica microdisks using a CO
2

laser to form microtoroids [28, 29].

Figure 2-9: Schematic of CO
2
laser reflow setup [29].

Thus, in the final fabrication step, the silica microdisks are reflowed using a
CO
2
laser, creating smooth, ultra-high Q toroids (Figure 2-9). Since silica absorbs
the CO
2
laser light very strongly while the silicon substrate does not, the CO
2
laser is
very effective for melting the silica into toroids. This melting process creates a
35
nearly atomically smooth surface, eliminating surface roughness. Since the melting
temperature of silicon is lower than that of silica, it would be difficult to achieve this
reflow using other methods. During the CO
2
laser reflow step, it is especially
important to keep the CO
2
laser beam properly aligned in order to achieve a smooth,
uniform reflow. For the 75W Synrad laser setup in the Armani lab, laser intensities
of 15-25% are usually optimal for reflowing toroids. However, the exact intensity
settings vary, depending on the sample size and amount of XeF
2
undercut.
2.4 Testing Procedures

2.4.1 Testing Setup

The quality factor and other important parameters of the toroid are
characterized using the linewidth measurement method [28] and the testing setup
shown schematically in Figure 2-10. First, the resonator to be tested is placed on a
piezoelectric nanopositioning stage in the center of the setup. From there, using top
and side view cameras and the nanopositioning stage, the toroid sample is aligned
with the tapered optical fiber. Light from a tunable laser is then coupled into the
toroid by bringing the tapered optical fiber very close to the toroid. The output light
from the tapered fiber is then sent to a detector. The detector output is sent to a high
speed digitizer and oscilloscope, both of which are part of a National Instruments
PCI card on the computer. A resonant wavelength can then be found by looking for
a drop in the transmitted light, which corresponds to light entering the resonator. To
more closely examine these resonant wavelength peaks, a function generator sends a
triangle wave to modulate the wavelength of the tunable laser. By quickly scanning
the tunable laser’s wavelength across a resonant wavelength, the transmission
36
spectra, including the resonant wavelength peak, can be measured and analyzed in
real-time.

Figure 2-10: Schematic of the basic toroid characterization setup.

While we use tapered optical fibers to introduce light into the toroids, other
methods of coupling, such as using prisms or lenses, could be used as well.
However, tapered optical fibers have many important advantages [52, 57]. In
particular, the tapered optical fiber can be placed alongside a WGM resonator,
allows simple focusing and alignment of the resonator beam, can be moved to adjust
and optimize coupling (unlike stagnant on-chip waveguides), and allows collection
of the output beam. Additionally, it filters out all the waveguide modes, except the
fundamental mode, at the input and output, making it an essentially single mode,
ideally matched coupler. Also, by coupling light with tapered optical fibers, it is
possible to achieve low coupling loss and phase matching. For taper-coupled WGM
37
resonators, the overall coupling efficiency depends on 1) the phase-matching
between the resonator and tapered fiber, 2) the overlap between the taper and the
resonator’s optical modes, and 3) the propagation constant [52, 57].

Figure 2-11: Taper puller setup (a) and optical microscope images of b) stripped and c) tapered
optical fibers.

Tapered optical fibers are made by slowly stretching an optical fiber while it
is heated with a hydrogen torch (Figure 2-11). When the diameter of the optical
fiber becomes thinner than the wavelength of light, the light traveling through the
tapered fiber extends slightly outside the fiber, forming an evanescent field.
Bringing the tapered optical fiber very close to the toroid causes light from the
tapered optical fiber to enter the toroid. It is important to note that the taper is not
touching the toroid during this process. If the tapered optical fiber is touching the
a)
b)
c)
38
toroid, very different modes could be observed because direct coupling is different
than the evanescent, free-space coupling.
The coupling efficiency and mode overlap can be changed by adjusting the
alignment and distance between the taper and the resonator. Using a piezoelectric
nanopositioning stage, the resonator can be moved closer and farther from the
tapered optical fiber. In doing so, the resonator can therefore be tested in the
undercoupled, overcoupled, and critically coupled regimes. These three coupling
regimes are each important and have distinct properties.
In the undercoupled regime, the tapered optical fiber is farther from the
resonator. As a result, only some light is coupled into the resonator. Since less
power is coupled into the resonator, the coupling loss is low compared to the
resonator loss. Therefore, since the resonator’s intrinsic loss predominates, the
intrinsic quality factor is most easily measured in the undercoupled regime [57].
As the tapered optical fiber and resonator are moved closer together from the
undercoupled regime, they approach the critically coupled regime [57]. Critical
coupling occurs when the tapered optical fiber’s light is completely transferred to the
resonator, and the taper’s transmission signal vanishes. When critically coupled, the
coupling and resonator losses are equal. Therefore, the extrinsic and intrinsic Q
factors are also equal.
When the resonator is overcoupled, the fiber is closer to the resonator than
the critical coupling position, and the coupling loss is higher than the resonator loss.
Since the mode overlap increases even further past critical coupling, the extrinsic Q
decreases further in the overcoupled regime [57]. It is generally not desirable to
39
measure the quality factor in the overcoupled regime since the coupling losses are
high, making the Q value even lower compared to the other coupling regimes.
As can be seen in this section, while many optical devices such as the toroid
resonator are fabricated on silicon, they are not necessarily truly integrated on a chip.
Tapered optical fibers are effective at coupling light into optical resonators, but are
also very fragile and extremely sensitive to changes in alignment. As a result,
optical toroid resonators are often difficult to use outside an optics lab because of the
delicate tapered fibers, expensive and bulky equipment, and precision alignment
needed for testing. Eventually, replacing tapered fibers with waveguides [10],
embedding them in polymers [58], or using other coupling approaches may enable
toroids to be used outside the lab.
2.4.2 Broad Scan and Fine Scan

The first step when using the resonator testing setup is to find the toroid’s
resonant wavelengths. Depending on the sample geometry and coupling conditions,
there can be few or many resonant wavelengths. Once those are found, the quality
factor can be measured, or additional experiments can be performed using the
resonant wavelength peak. There are two main approaches used to find the resonant
wavelengths: broad scan and fine scan.
In broad scan, the transmission vs. wavelength is measured directly. The
oscilloscope is set to “immediate” mode, the time/div setting is increased to several
seconds per division, and the oscope is set to save the maximum possible number of
data points. Then, light is coupled into the toroid resonator while the tunable laser’s
wavelength is continuously changed, usually at a rate of 0.1nm/s. By monitoring the
40
transmission as a function of wavelength over a wide range (~5-10 nanometers or at
least 2-3 times the resonator’s free spectral range) one can view all the resonant
wavelengths of the resonator in the range. In addition, by taking broad scans, the
free spectral range and fundamental mode resonant wavelengths can be determined
(Figure 2-12). Note that the laser wavelength is not modulated using the function
generator for this measurement.

Figure 2-12: Representative broad scan of ~100 micron silica toroid at 1300-1312nm.

A representative broad scan is shown in Figure 2-12. It is also possible to use
the broad scan to determine the quality factor of the resonant peaks. This method is
often used to measure peaks with low quality factors (<10
4
) since the large linewidth
of low-Q peaks makes them difficult or impossible to view using a the fine scan
method. However, if the resonator has high quality factors of over 10
7
, it is
generally necessary to decrease the scan rate from 0.1nm/s to 0.05 or even 0.01nm/s.
41
Since high quality factor peaks have very narrow linewidths, scanning the
wavelength too quickly can skip over the peaks altogether. Once a resonant
wavelength is found, the quality factor can be determined by fitting a Lorentzian
curve to the transmission peak and dividing the center wavelength by the peak’s full
width at half maximum value.
The other method commonly used to find and characterize resonant peaks is
fine scan. Unlike in broad scan, a function generator is used to modulate the laser’s
wavelength with a triangle wave. This allows continuous forward and backward
scanning of the laser across a resonant peak, enabling the transmission spectra and
resonant peak to be continuously viewed in real-time. To perform a fine scan, the
resonant wavelengths are approximately determined using a broad scan or by quickly
scanning the laser across a wide wavelength range and watching for peaks. Then,
while maintaining light coupling between the toroid and tapered optical fiber, the
wavelength of the tunable laser is set to a value near where the resonant wavelengths
were observed. The laser’s wavelength is slowly increased in 0.001nm steps until
the resonant peak is seen. Once the resonant peak is found, it can be saved and the Q
can be calculated.
42

Figure 2-13: Screenshot of oscilloscope during fine scan. The vertical scale is intensity
(0.5V/division) and the horizontal scale is time (1ms/division). The purple line is the signal from the
photodetector, and the white line is the triangle wave from the function generator. The left half of the
triangle wave shows the forward scan, and the right half shows the reverse scan.

One especially important advantage of fine scan is that it allows the resonant
peak to be viewed in real-time. This allows the resonant wavelength peak’s
behavior, position, and quality factor to be monitored in response to many changes,
such as changes in coupling, input power, and temperature. Also, fine scan allows
the resonant peak to be viewed as wavelength changes in both the forward and
reverse directions. When the resonant peak’s forward and reverse scans differ,
nonlinear effects such as thermal broadening can be observed (Figure 2-13) [54, 59].
Therefore, fine scan allows several important measurements to be made:
measurement of intrinsic quality factor, observation of thermal broadening/nonlinear
43
effects, and measurement of resonant wavelength shifts. These are discussed in the
next sections.
2.4.3 Intrinsic Quality Factor

As mentioned previously, the measured quality factor of a toroid can include
both losses which are intrinsic to the toroid, as well as extrinsic losses like coupling
loss. Using a fine scan, it is possible to measure Q as a function of coupling. In the
undercoupled regime, the quality factor increases linearly with decreasing coupling
(Figure 2-14). Therefore, the quality factor at zero coupling, or the intrinsic quality
factor, can be determined by fitting a line to Q versus coupling data and determining
the y-intercept.

Figure 2-14: In the undercoupled regime, Q increases linearly as % coupling decreases. The intrinsic
Q is the Q at 0% coupling.

44
It is generally desirable to know the intrinsic quality factor because it
represents the losses which are inherent to the resonator and does not include
extrinsic losses like coupling which can change depending on the testing conditions.
In addition, the intrinsic quality factor can provide other important information,
including the value of the material loss quality factor Q
mat
and the material
absorption coefficient α. The surface scattering, contamination, and radiation losses
are generally assumed negligible for toroid resonators with larger diameters in air.
Therefore, the total quality factor becomes dependent on only the material and
coupling losses as in Equation 2-21:

1 1 1 1 1 1
. rad s s cont m at coup
Q Q Q Q Q Q
     
     
1 1 1
mat coup
Q Q Q
  
  (2-21)
Since the intrinsic quality factor is determined at zero coupling of light into
the toroid, we obtain Equation 2-22:

1 1
mat
Q Q
 
 
intrinsic
2
eff
mat
eff
n
Q Q

 
  (2-22)
Therefore, assuming negligible radiation, surface scattering, and
contamination losses, measuring the intrinsic quality factor allows Q
mat
to be
determined. Since Q
mat
is related to material absorption, the effective material
absorption loss α
eff
can also be found. Knowing the α
eff
term allows additional
optical properties, such as the thermo-optic coefficient, to be studied as well [40].
More details on these material characterization experiments are in Chapter 4.
2.4.4 Tracking Resonant Wavelength Shifts

Interesting experiments can also be done by monitoring the resonant
wavelength peak. Having a high quality factor produces a sharper, narrower
45
linewidth resonant peak. These narrow peaks give higher resolution, which is crucial
in resonant wavelength tracking experiments such as biodetection [31, 34].
Using a fine scan, it is possible to view the resonant peak and monitor
resonant wavelength changes in real-time. This is extremely useful, as many
interesting biodetection and material characterization experiments can be performed
by tracking the resonant wavelength. For example, many optical properties such as
refractive index are dependent on temperature. By heating the toroid resonator
during fine scan, the change in resonant wavelength can be measured as a function of
temperature. By spin coating a thin layer of a material onto a toroid and measuring
Δλ/ ΔT, the thermo-optic coefficient dn/dT of the material can be determined at a
given wavelength [39, 40]. More information on these experiments is in Chapter 4.
Another interesting property of optical toroid resonators is the fact that light
is not completely confined inside the resonator. Some of the circulating light leaks
into the surroundings, forming an exponentially decaying evanescent field (Figure 2-
15). This evanescent field allows optical toroid resonators to detect changes in the
surrounding environment [34].

46

Figure 2-15: Finite element model of electric field in toroid cross-section (D=20µm), showing
evanescent field leaking into surroundings (white arrow) [34].

In biodetection experiments, binding of analyte to a toroid resonator causes
the resonant wavelength to shift. By tracking the resonant wavelength peak of the
toroids, one can monitor the binding and dissociation of biological molecules onto
toroids, as seen in Figure 2-16. Plotting the resonant wavelength shift as a function
of time, analyte concentration, and other parameters can give useful insight into
kinetic systems. More information is provided in Chapters 6 and 7.
47

Figure 2-16: Schematic of biodetection experiment and wavelength shift vs. time data.

2.4.5 Observing Thermal Effects

Testing toroids using fine scan also enables viewing of interesting thermal
effects. In high Q devices, some of the circulating optical power is absorbed and
converted to heat. This thermo bistability effect is described by Equation 2-23 [40,
54]:
(2-23)
Here, v and v
0
are the final and initial frequencies, n is the refractive index, dn/dT is
the thermo-optic coefficient of the cavity material, and ΔT is the temperature change.
It depends on the refractive index, wavelength/frequency of light, and the
temperature change. For many measurements, these thermal broadening effects can
48
distort or mask data, causing problems. Therefore, most quality factor measurement
experiments are done at low input powers to minimize these thermal effects. These
thermal effects can be useful in some applications, however. In Chapter 7, some
applications of this heating effect are discussed.

Chapter 2 References

1. M. R. Poulsen, P. I. Borel, J. Fage-Pedersen, J. Hubner, M. Kristensen, J. H.
Povlsen, K. Rottwitt, M. Svalgaard, and W. Svendsen, "Advances in silica-based
integrated optics," Opt. Eng. 42, 2821-2834 (2003).

2. A. L. Washburn, and R. C. Bailey, "Photonics-on-a-chip: recent advances in
integrated waveguides as enabling detection elements for real-world, lab-on-a-chip
biosensing applications," Analyst 136, 227-236 (2011).

3. C. Kopp, S. Bernabe, B. Ben Bakir, J. M. Fedeli, R. Orobtchouk, F. Schrank,
H. Porte, L. Zimmermann, and T. Tekin, "Silicon Photonic Circuits: On-CMOS
Integration, Fiber Optical Coupling, and Packaging," IEEE J. Sel. Top. Quantum
Electron. 17, 498-509 (2011).

4. M. S. Luchansky, and R. C. Bailey, "High-Q Optical Sensors for Chemical
and Biological Analysis," Anal. Chem. 84, 793-821 (2012).

5. H. S. Hsu, C. Cai, and A. M. Armani, "Ultra-low-threshold Er:Yb sol-gel
microlaser on silicon," Optics Express 17, 23265-23271 (2009).

6. H. Rokhsari, and K. J. Vahala, "Ultralow loss, high Q, four port resonant
couplers for quantum optics and photonics," Physical Review Letters 92, 253905
(2004).

7. B. Saleh, and M. Teich, Fundamentals of Photonics (Wiley-Interscience,
2007).

8. A. J. Maker, and A. M. Armani, "Low-loss silica-on-silicon waveguides,"
Optics Letters 36, 3729-3731 (2011).

9. X. M. Zhang, and A. M. Armani, "Suspended bridge-like silica 2 x 2 beam
splitter on silicon," Optics Letters 36, 3012-3014 (2011).

49
10. X. M. Zhang, and A. M. Armani, "Silica microtoroid resonator sensor with
monolithically integrated waveguides," Optics Express 21, 23592-23603 (2013).

11. E. Hecht, Optics (Addison Wesley, 2002).

12. R. Kitamura, L. Pilon, and M. Jonasz, "Optical constants of silica glass from
extreme ultraviolet to far infrared at near room temperature," Applied Optics 46,
8118-8133 (2007).

13. R. Germann, H. W. M. Salemink, R. Beyeler, G. L. Bona, F. Horst, I.
Massarek, and B. J. Offrein, "Silicon oxynitride layers for optical waveguide
applications," J. Electrochem. Soc. 147, 2237-2241 (2000).

14. A. Jain, A. H. J. Yang, and D. Erickson, "Gel-based optical waveguides with
live cell encapsulation and integrated microfluidics," Optics Letters 37, 1472-1474
(2012).

15. H. Ma, A. K. Y. Jen, and L. R. Dalton, "Polymer-based optical waveguides:
Materials, processing, and devices," Advanced Materials 14, 1339-1365 (2002).

16. A. K. Manocchi, P. Domachuk, F. G. Omenetto, and H. M. Yi, "Facile
Fabrication of Gelatin-Based Biopolymeric Optical Waveguides," Biotechnology
and Bioengineering 103, 725-732 (2009).

17. H. B. Xin, Y. Y. Li, X. S. Liu, and B. J. Li, "Escherichia coli-Based
Biophotonic Waveguides," Nano Letters 13, 3408-3413 (2013).

18. R. Adar, M. R. Serbin, and V. Mizrahi, "Less-than-1DB per meter
propagation loss of silica wave-guides measured using a ring-resonator," Journal of
Lightwave Technology 12, 1369-1372 (1994).

19. R. Adar, Y. Shani, C. H. Henry, R. C. Kistler, G. E. Blonder, and N. A.
Olsson, "Measurement of very low-loss silica on silicon wave-guides with a ring
resonator," Applied Physics Letters 58, 444-445 (1991).

20. A. J. Maker, and A. M. Armani, "Low-loss silica on silicon integrated
waveguides," in Conference on High Contrast Metastructures(San Francisco, CA,
2012).

21. K. J. Vahala, "Optical microcavities," Nature 424, 839-846 (2003).

22. COMSOL, "Heat Transfer Module User’s Guide," (COMSOL Multiphysics
4.3a, 2012).

50
23. M. Borselli, T. J. Johnson, and O. Painter, "Beyond the Rayleigh scattering
limit in high-Q silicon microdisks: theory and experiment," Optics Express 13, 1515-
1530 (2005).

24. D. W. Vernooy, V. S. Ilchenko, H. Mabuchi, E. W. Streed, and H. J. Kimble,
"High-Q measurements of fused-silica microspheres in the near infrared," Optics
Letters 23, 247-249 (1998).

25. V. M. N. Passaro, C. de Tullio, B. Troia, M. La Notte, G. Giannoccaro, and
F. De Leonardis, "Recent Advances in Integrated Photonic Sensors," Sensors 12,
15558-15598 (2012).

26. R. Colombelli, K. Srinivasan, M. Troccoli, O. Painter, C. F. Gmachl, D. M.
Tennant, A. M. Sergent, D. L. Sivco, A. Y. Cho, and F. Capasso, "Quantum cascade
surface-emitting photonic crystal laser," Science 302, 1374-1377 (2003).

27. S. Varoutsis, S. Laurent, I. Sagnes, A. Lemaitre, L. Ferlazzo, C. Meriadec, G.
Patriarche, I. Robert-Philip, and I. Abram, "Reactive-ion etching of high-Q and
submicron-diameter GaAs/AlAs micropillar cavities," Journal of Vacuum Science &
Technology B 23, 2499-2503 (2005).

28. D. K. Armani, T. J. Kippenberg, S. M. Spillane, and K. J. Vahala, "Ultra-
high-Q toroid microcavity on a chip," Nature 421, 925-928 (2003).

29. A. J. Maker, and A. M. Armani, "Fabrication of Silica Ultra High Quality
Factor Microresonators," in Journal of Visualized Experiments(2012).

30. H. K. Hunt, C. Soteropulos, and A. M. Armani, "Bioconjugation Strategies
for Microtoroidal Optical Resonators," Sensors 10, 9317-9336 (2010).

31. A. J. Maker, and A. M. Armani, "Heterodyned Toroidal Microlaser Sensor,"
Appl. Phys. Lett. 103, 123302 (2013).

32. S. M. Spillane, T. J. Kippenberg, K. J. Vahala, K. W. Goh, E. Wilcut, and H.
J. Kimble, "Ultrahigh-Q toroidal microresonators for cavity quantum
electrodynamics," Physical Review A 71 (2005).

33. L. Yang, D. K. Armani, and K. J. Vahala, "Fiber-coupled erbium microlasers
on a chip," Applied Physics Letters 83, 825-826 (2003).

34. A. M. Armani, R. P. Kulkarni, S. E. Fraser, R. C. Flagan, and K. J. Vahala,
"Label-free, single-molecule detection with optical microcavities," Science 317, 783-
787 (2007).

51
35. E. P. Ostby, L. Yang, and K. J. Vahalal, "Ultralow-threshold Yb3+: SiO2
glass laser fabricated by the solgel process," Optics Letters 32, 2650-2652 (2007).

36. A. J. Maker, and A. M. Armani, "Nanowatt threshold, alumina sensitized
neodymium laser integrated on silicon," Optics Express 21, 27238-27245 (2013).

37. S. Mehrabani, and A. M. Armani, "Blue upconversion laser based on
thulium-doped silica microcavity," Optics Letters 38, 4346-4349 (2013).

38. L. N. He, K. Ozdemir, J. G. Zhu, W. Kim, and L. Yang, "Detecting single
viruses and nanoparticles using whispering gallery microlasers," Nature
Nanotechnology 6, 428-432 (2011).

39. H. S. Choi, S. Ismail, and A. M. Armani, "Studying polymer thin films with
hybrid optical microcavities," Optics Letters 36, 2152-2154 (2011).

40. B. A. Rose, A. J. Maker, and A. M. Armani, "Characterization of thermo-
optic coefficient and material loss of high refractive index silica sol-gel films in the
visible and near-IR," in Optical Materials Express(OSA, 2012), pp. 671-681.

41. V. S. Ilchenko, P. S. Volikov, V. L. Velichansky, F. Treussart, V. Lefevre-
Seguin, J. M. Raimond, and S. Haroche, "Strain-tunable high-Q optical microsphere
resonator," Optics Communications 145, 86-90 (1998).

42. A. L. Martin, D. K. Armani, L. Yang, and K. J. Vahala, "Replica-molded
high-Q polymer microresonators," Optics Letters 29, 533-535 (2004).

43. M. I. Cheema, and A. G. Kirk, "Implementation of the perfectly matched
layer to determine the quality factor of axisymmetric resonators in COMSOL," in
COMSOL Conference(Boston, 2010).

44. A. J. Maker, B. A. Rose, and A. M. Armani, "Tailoring the behavior of
optical microcavities with high refractive index sol-gel coatings," Optics Letters 37,
2844-2846 (2012).

45. F. Pisanello, A. Qualtieri, T. Stomeo, L. Martiradonna, R. Cingolani, A.
Bramati, and M. De Vittorio, "High-Purcell-factor dipolelike modes at visible
wavelengths in H1 photonic crystal cavity," Optics Letters 35, 1509-1511 (2010).

46. M. Oxborrow, "How to simulate the whispering-gallery-modes of dielectric
microresonators in FEMLAB/COMSOL - art. no. 64520J," in Conference on Laser
Resonators and Beam Control IX(San Jose, CA, 2007), pp. J4520-J4520.

47. M. L. Gorodetsky, A. A. Savchenkov, and V. S. Ilchenko, "Ultimate Q of
optical microsphere resonators," Optics Letters 21, 453-455 (1996).
52

48. X. M. Zhang, H. S. Choi, and A. M. Armani, "Ultimate quality factor of
silica microtoroid resonant cavities," Applied Physics Letters 96 (2010).

49. H. S. Choi, X. M. Zhang, and A. M. Armani, "Hybrid silica-polymer ultra-
high-Q microresonators," Optics Letters 35, 459-461 (2010).

50. A. M. Armani, D. K. Armani, B. Min, K. J. Vahala, and S. M. Spillane,
"Ultra-high-Q microcavity operation in H2O and D2O," Applied Physics Letters 87
(2005).

51. I. S. Grudinin, N. Yu, and L. Maleki, "Generation of optical frequency combs
with a CaF2 resonator," Optics Letters 34, 878-880 (2009).

52. S. M. Spillane, T. J. Kippenberg, O. J. Painter, and K. J. Vahala, "Ideality in
a fiber-taper-coupled microresonator system for application to cavity quantum
electrodynamics," Physical Review Letters 91 (2003).

53. H. K. Hunt, and A. M. Armani, "Recycling microcavity optical biosensors,"
Optics Letters 36, 1092-1094 (2011).

54. H. S. Choi, and A. M. Armani, "Thermally stable hybrid organic/inorganic
resonant cavities," in Conference on Linear and Nonlinear Optics of Organic
Materials XI(San Diego, CA, 2011).

55. H. S. Choi, D. Neiroukh, H. K. Hunt, and A. M. Armani, "Thermo-optic
Coefficient of Polyisobutylene Ultrathin Films Measured with Integrated Photonic
Devices," Langmuir 28, 849-854 (2012).

56. B. W. Biggs, H. K. Hunt, and A. M. Armani, "Selective patterning of Si-
based biosensor surfaces using isotropic silicon etchants," Journal of Colloid and
Interface Science 369, 477-481 (2012).

57. M. Cai, O. Painter, K. J. Vahala, and P. C. Sercel, "Fiber-coupled
microsphere laser," Optics Letters 25, 1430-1432 (2000).

58. F. Monifi, S. K. Ozdemir, J. Friedlein, and L. Yang, "Encapsulation of a
Fiber Taper Coupled Microtoroid Resonator in a Polymer Matrix," IEEE Photonics
Technology Letters 25, 1458-1461 (2013).

59. H. S. Choi, and A. M. Armani, "Thermal nonlinear effects in hybrid optical
microresonators," Applied Physics Letters 97 (2010).

53
Chapter 3 Low Loss Silica on Silicon Waveguides

3.1 Introduction

As mentioned in the previous sections, integrated optical devices have many
important applications in areas such as communications, biodetection, and
fundamental physics studies [1-9]. Optical waveguides are a crucial component of
these integrated optical devices, acting as optical wires which confine and direct light
on a chip. Developing efficient and low loss waveguides is a crucial step to make
truly integrated optics available outside a research lab. For example, a waveguide
could replace the fragile tapered fiber used to couple light into toroid resonators [10].
3.2 Background and Motivation

While many different waveguides have been developed using various
geometries and materials, these devices often have low-loss, nonlinear optical
behavior across only a limited range of operating conditions, therefore limiting their
applications [7, 11]. Optical materials such as polymers and silicon can also behave
differently under different testing conditions, so loss may also vary with wavelength,
polarization, and input power. In contrast, silica is an excellent waveguiding
material because it has low optical absorption and highly linear optical behavior
across a broad range of wavelengths from the visible and near-IR (Figure 3-1) [12,
13].
54

Figure 3-1: Absorption coefficient and refractive index vs. wavelength for silica, silicon, and silicon
nitride [12].

Thus, silica materials are commonly used in communications applications
such as optical fiber, enabling signals to be transferred across long distances with
minimal losses. These low losses make silica an excellent material for developing
integrated optical waveguides as well [2, 6, 14].
However, for integrated optics applications, it is necessary to fabricate silica
waveguides directly on silicon wafers. This can be challenging for silica waveguides
because the refractive index of silicon is higher than that of silica. Unless the silica
waveguide is isolated from the silicon substrate, light will leak from the silica
directly into the silicon, causing very high optical losses. While silica waveguides
can be isolated by sandwiching or embedding them between other materials [15, 16],
these approaches increase the complexity of fabrication and limit the materials which
can be used.
To address this challenge, we have developed a silica on silicon waveguide in
a unique suspended cylinder geometry (Figure 3-2) [5, 17]. Since the silica
waveguide is elevated above the silicon substrate, the resulting waveguides are
55
surrounded by air, preventing leakage of light into the silicon. In the present work,
the fabrication, COMSOL modeling, and characterization of these waveguides is
described in detail. Because these devices have favorable optical properties and are
fabricated directly on silicon, they are useful in a variety of integrated optics
applications [9, 10].

Figure 3-2: Suspended silica waveguide design. Elevated silica slabs (a) are reflowed with a CO
2

laser, creating pairs of smooth, cylindrical waveguides (b-d) which are air-clad and isolated from the
silicon substrate [5].

3.3 COMSOL Modeling of Waveguides

One concern about this waveguide design is the possibility that light may
leak out of the cylindrical waveguide, through the supporting membrane, and into the
silicon substrate. In collaboration with Xiaomin Zhang, the confinement of light in
silica waveguide cross-sections was modeled using COMSOL Multiphysics 3.4 [5,
17]. Specifically, the RF Module (perpendicular waves) is used. This module solves
for the various TE and TM modes which could be present in the waveguide structure.
56
First, waveguide cross-sections of the desired diameters and supporting
membrane thicknesses are drawn (Figure 3-3). Note that the x and y length scale
units should be in microns (10
-6
meters). One large rectangle is drawn around the
waveguide cross-section, and four smaller rectangles are drawn along each side of
the large rectangle. These enable the boundary conditions to be properly defined
along eight border regions (four square corners and upper, lower, right, and left
boundary rectangles). These border regions act as light-absorbing PMLs (perfectly
matched layers) and are important when defining the physics and boundary
conditions later on.

Figure 3-3: Representative COMSOL 3.4 screenshot of waveguide cross-section used in simulations.

Once the main geometry is drawn, the settings are adjusted under physics 
subdomain settings. In the waveguide and air cross-section subdomains, the
refractive index is specified (~1.43 and 1 respectively). No PML is used (“none” is
checked) and all the remaining settings are kept as default. In the left and right
57
boundary rectangle regions, the values of ε, μ, and σ are specified as 1, 0, and 1
respectively, and the regions are defined as Cartesian PMLs absorbing in the x-
direction. Similarly, the upper and lower boundary rectangle regions have the values
of ε, μ, and σ are specified as 1, 0, and 1 respectively, and the regions are defined as
Cartesian PMLs absorbing in the y-direction. Finally, the four corner boundary
domains have the values of ε, μ, and σ are specified as 1, 0, and 1 respectively, and
the regions are defined as Cartesian PMLs absorbing in the x and y directions. All
other settings for the boundary domains are kept as default.
The boundary conditions are adjusted in physics  boundary settings as
follows. All the interior boundaries use the continuity boundary condition, allowing
light to pass through. The outer boundaries use the scattering boundary condition for
a plane wave. Following this procedure allows the waveguide geometry to be
defined, while surrounded by an absorbing PML on all four sides. This PML ensures
that the boundary of the problem is finite, not infinite.
Under physics  scalar variables, the default scalar variables are used,
except the value of lambda is changed to the wavelength of interest in meters (for
example, 1.55e-6m for 1550nm). A mesh is also defined with the maximum element
size set to ~200-300nm, or at least several times smaller than the wavelength of
interest.
Finally, to solve, the solver parameters are adjusted. The mode analysis
solver is used to solve for eigenvalues. As a starting point, the desired number of
effective mode indices can be set to 10 (more may be needed to find the fundamental
mode). Also, the solver should be told to search around the refractive index value
58
close to the waveguide material (~1.43). The Direct SPOOLES solver is used. Once
these settings are entered, the model is run and the fundamental mode solution is
determined by looking at the electric field surface plots.
After making and running the models, the electric field solutions are found at
658nm (Figure 3-4a) and 1550nm (Figure 3-4b) wavelengths [5]. The horizontal and
vertical mode cross-sections of the confined modes are also plotted (Figure 3-4c, d).
For simplicity, only the fundamental mode solution is plotted in Figure 3-4. These
simulations confirm that the majority of the electric field is confined inside the
cylindrical waveguide structure, and the supporting membrane has minimal impact
on the confinement of light in the waveguides (at least at the fundamental mode).
However, it is also important to note that only the fundamental mode was
considered. Since these waveguide devices are multi-mode, it is possible that higher
order modes may have greater amounts of loss and leakage into the supporting
silicon pillar.
59

Figure 3-4: COMSOL modeling of waveguide cross-sections and mode cross sections at 658nm (a, c)
and 1550nm (b, d). The mode profiles are almost completely confined in the waveguide and are
nearly Gaussian, indicating that light is successfully confined into the waveguide and minimally
affected by the supporting membrane [5].

3.4 Fabrication Process

3.4.1 Photolithography and Etching

The suspended cylinder waveguide geometry in the present work had not
been previously developed. Therefore, one significant challenge was to optimize the
fabrication procedures.
The silica on silicon waveguides are fabricated from intrinsic silicon wafers
with a 2µm thick layer of silica (Figure 3-5a). The fabrication process is modeled
after the steps used to make silica toroids [18, 19], using photolithography, buffered
oxide etching, XeF
2
etching, and CO
2
laser reflow, as described presently [5, 17].
60

Figure 3-5: Schematic of the waveguide fabrication process [17].

In the photolithography steps (Figure 3-5a-c), rectangular slabs of photoresist
are patterned on the wafers. Next, the silica not covered by the photoresist is etched
away using buffered oxide etchant (BOE), and the photoresist is removed to produce
silica slabs (Figure 3-5c-e). Afterward, the silicon underneath the silica slabs is
selectively and isotropically etched using XeF
2
gas. This creates silica slabs which
are elevated above the silicon substrate (Figure 3-5f). These silica slabs are reflowed
with a CO
2
laser to produce smooth cylindrical waveguides (Figure 3-5g).
3.4.2 Fabrication Process – CO
2
Laser Reflow

In order to produce smooth, low-loss waveguides, proper reflow using the
CO
2
laser (here, a 75W Synrad Series 48) is extremely important. Reflowing an
extremely smooth, straight waveguide sample is extremely difficult and requires
optimization of several parameters. Of thousands of waveguide samples produced
during this project, only a handful had near-perfect reflow. To optimize the
waveguide reflow process, a reflow study was performed in which the CO
2
laser
61
intensity, amount of XeF
2
undercut, and the speed at which the waveguide is moved
under the laser beam were independently varied. Reflowing 2 micron-thick
waveguides at higher intensities (~27-30% of 75W laser beam) and slow speeds
(0.05mm/s) produced the best results (Figure 3-6). Even so, the results vary
considerably from sample to sample, making the loss of the waveguide devices
ultimately limited by how well they can be reflowed without introducing bends or
other defects.
As a result, this reflow step became the most important and challenging
aspect of the project. In order to achieve consistent loss data, the waveguides needed
to reflow consistently well. However, the CO
2
laser reflow varies tremendously
from sample to sample. Finding ways to improve this reflow step could drastically
improve the fabrication and overall performance of these waveguides.

Figure 3-6: Top view optical microscope image of extremely straight reflowed waveguides.

3.4.3 Fabrication Process – End Firing

Finally, to produce flat ends for coupling light, the ends of the waveguides
are cleaved with a diamond scribe. This cleaving step is absolutely crucial as a flat,
62
cleaved waveguide end is needed in order to efficiently couple light into the
waveguide devices. After the reflow process, the ends of the waveguides are
rounded (as in Figure 3-7).

Figure 3-7: Top view SEM image of reflowed waveguide end prior to cleaving.

In order to couple light into the waveguides, the ends are cleaved to create a
smooth, flat edge. This process is known as end-firing. Four end-firing methods
were tried: polishing, cutting with dicing saw, cleaving by hand using the diamond
scribe, and using a focused ion beam (FIB) or ion milling.
Polishing by hand was the first approach used in an attempt to produce
smooth waveguide ends. The waveguide samples were super-glued to a mount and
then embedded in acrylic polymer. Once the acrylic solidified, the edges of the
waveguide samples were mechanically polished using equipment in Prof. Andrea
Hodge’s labs. Unfortunately, as shown in Figure 3-8, this polishing approach did
not work well at all. It had many issues and limitations, including the following:
63
- The entire process is very time consuming, requiring the acrylic polymer to
harden for several hours to overnight, and several additional hours to polish only
a few samples.
- Since one end of the sample needed to be super-glued to the sample holder, only
one end of the waveguide sample could be polished, while the other end is
harmed.
- The acrylic polymer was extremely difficult to remove from the waveguide
samples, and leaves behind residue even after soaking in acetone or methanol for
days.
- The waveguide ends were completely destroyed in nearly all the samples used.
Given these issues, hand polishing was not a good option for end-firing the silica
waveguides.

Figure 3-8: Scanning electron microscope image of hand-polished waveguide end. The polishing
technique broke the samples and left acrylic residue.

64
Since hand polishing did not produce smooth waveguide ends, the next
method tried was dicing using the cleanroom’s dicing saw. After reflowing the
waveguide samples with the CO
2
laser, a protective layer of photoresist was spin-
coated onto the samples at ~1500rpm for 30 seconds and softbaked at 95°C for 2-3
minutes. This protective layer helped keep the waveguides from being damaged by
the dicing saw’s cooling water during the cutting process. Then, the photoresist-
covered waveguide ends were cleaved using the dicing saw.

Figure 3-9: Scanning electron microscope image of waveguide sample cleaved using dicing saw.
While the dicing saw breaks samples less often, the cut can still be rough.

Compared to hand polishing, the dicing saw is an improvement and could
produce usable samples about 10% of the time. The results tended to be best when a
new dicing saw blade was used, as the cut would be smoother. However, the dicing
process is still time-consuming (requiring several hours at least to coat samples with
photoresist and cleave) and the extra handling steps (applying, removing photoresist)
can greatly increase the risk of sample breakage. Also, the dicing process could still
65
leave roughness on the waveguide ends, as seen in Figure 3-9. Therefore, the dicing
approach still wasn’t ideal for effectively and consistently cleaving the silica
waveguides.
The third approach used for cleaving the silica waveguides was simply
cutting the waveguides’ silicon wafer using a diamond scribe. A small scratch was
made on the very edge of the wafer where the desired cut was to be made. Then, the
samples and any waveguides on them could be manually cleaved in a straight line
propagating along the initial crack. As the silicon underneath the waveguide samples
broke, the waveguides would also be cleaved.
Compared to dicing and hand polishing, the diamond scribe is significantly
faster, taking just seconds to make a scratch and then cleave. Initially, samples were
cleaved using a large-tip diamond scribe. Since the large tip diamond scribe
produced a large crack, the waveguide ends were still destroyed about as often as
those cleaved with the dicing saw. Also, cleaving with the large tip diamond scribe
could produce silicon dust and leave a mess. Spin-coating a protective layer of
photoresist sometimes helped improve the cleaving and enabled removal of the dust,
although the additional handling steps to apply and remove photoresist could still
increase the risk of breakage.
Later, one of my labmates, Maria Chistiakova, ordered more diamond scribes
and the company sent fine-tip diamond scribes. The fine tip diamond scribes
actually cleaved the waveguides significantly better with much less dust and sample
breakage. Therefore, cleaving the waveguides with a fine-tip diamond scribe proved
to be the fastest, highest yield, and most efficient option.
66
A fourth approach for waveguide end firing involved using a focused ion
beam (FIB) or ion milling to cut the waveguide ends. While this method is very
precise and effective, it is extremely slow (taking 4-8 hours per waveguide end) and
required going to other labs to reserve and use the equipment. Therefore, the
waveguides in the present work were cleaved with the fine tip diamond scribe in
order to produce testable waveguide ends faster and more reliably (Figure 3-10).

Figure 3-10: SEM image of as-fabricated waveguide cleaved using a diamond scribe.

3.4.4 Fabrication Process – Two Photolithography Steps

While the primary focus of the waveguide project was developing
waveguides made from a two micron thick silica slab, additional experiments were
done using waveguides with thinner supporting membranes. By thinning the
supporting membrane between the silica waveguide and silicon pillar, losses into the
silicon substrate could be further reduced (Figure 3-11).
67

Figure 3-11: Thinning the waveguide's supporting membrane to smaller than the wavelength
traveling through it reduces loss, as shown schematically and in COMSOL finite element simulations.

In order to fabricate these “thin-membrane” waveguides, a second
photolithography step was performed prior to XeF
2
etching. This step thins the silica
supporting membrane, as shown in Figure 3-12.

Figure 3-12: Thinning the waveguide membrane with a second photolithography step

One may notice in Figure 3-12 that two photolithography steps are used in
order to keep a thick layer of silica along the waveguide regions. While it is possible
to simply use a thinner silica film at the beginning and omit this second
68
photolithography step to thin the membrane, there were several advantages of this
second photolithography step. The thick region of silica helped the waveguides
reflow and helped reduce breakage during fabrication. And later, it was found that
the thicker silica regions formed by the second photolithography step could be used
as trapezoidal waveguides as well [20].
For several months during development of the waveguides, these thin
membrane devices were fabricated. In addition to the increased complexity of
fabrication, there are several other challenges. First, the thin membrane waveguides
are extremely fragile. It is very difficult to handle, cleave the ends, and test such
fragile devices without breaking them. Second, the thin membrane waveguides
begin to bend due to increased stress in the silica material. This causes the devices to
gradually crack and break, and also could increase losses due to bending (Figure 3-
13a, b). Finally, the thin membrane waveguides are extremely difficult to reflow.
When exposed to the CO
2
laser beam, the thin silica melts very unevenly, often
crumbling like a piece of paper rather than forming smooth, cylindrical waveguides
(Figure 3-13c).
69

Figure 3-13: Thin silica slabs have increased stress, as seen in these optical microsope images. This
stress causes bending (a, b) and uneven reflow (c).

Another attempt to reduce the waveguide membrane thickness while avoiding
the reflow issues was to immerse finished waveguide samples in BOE again, to
further etch the silica. This could thin the waveguide membrane further, and also
reduce the diameter of the waveguides to make them single mode (or at least less
multi-mode). While the BOE immersion did indeed help reduce the diameter and
supporting membrane thickness of the waveguide devices, it was very easy to break
the waveguide samples when immersing and removing them from the BOE. Also, as
the supporting membrane became thinner (which it did very quickly,
~100+nm/minute), it could crack and break very easily. The supporting membrane
also etched much faster than the reflowed waveguide. As a result, the supporting
membrane would break before the waveguide diameters were etched enough to
become single mode.
Because of these issues, we focused primarily on development of silica
waveguides made from 2µm-thick oxide and did not use these thin membrane
devices in the final publication. If these issues can be overcome, it may be
a) b) c)
70
interesting to investigate the thin membrane waveguides in the future to further
reduce the loss of the waveguide devices. For example, immersing a finished
waveguide device in buffered oxide etchant and then reflowing the device with the
CO
2
laser at low power to repair roughness may enable fabrication of thinner
membrane, smaller waveguides, while eliminating the reflow problem. Also worth
noting is that the non-reflowed thin membrane waveguides could be used as
trapezoidal waveguides with low losses and additional interesting properties, as
studied by Xiaomin Zhang, Mark Harrison, and Audrey Harker [20].

3.4.5 Testing Setup

In addition to designing and fabricating the waveguides themselves, one of
the main challenges of this waveguide project was building and optimizing the
waveguide testing setup. Since the labs were new and no prior waveguide work had
been done in the group, optimizing the waveguide testing setup took a considerable
amount of time and effort in collaboration with Xiaomin Zhang and Heather Hunt.
The final optimized waveguide testing setup is shown schematically in Figure
3-14 [17]. The waveguide testing setup has several key components, including the
laser, lensed fiber, waveguide alignment stages, cameras, aspheric lens, beam
profiler, and a computer. Each of these components is described in the following
section.

71

Figure 3-14: Schematic of waveguide characterization setup [17].

The laser used to couple light is a Thorlabs MCLS1 4 channel laser source,
containing diode lasers at 406nm, 658nm, 980nm, and 1550nm. These wavelengths
were chosen so that the waveguide behavior could be studied at both visible and
near-IR wavelengths. The setup also includes a 1520-1630nm tunable laser from
Agilent. Both the wavelength and output power of this laser can be tuned, enabling
loss vs. input power and loss vs. wavelength measurements [5, 9, 17, 20]. All of
these lasers are fiber-coupled to FC/APC connectors, so that they can be easily
plugged into the setup and interchanged during testing.
The lensed fiber is one of the most important components of the waveguide
testing setup, as it is used to efficiently couple light from the diode or tunable lasers
into the waveguide device. The free space and efficient lensed fiber coupling is
extremely important. Initially, light was coupled into the waveguides using a
stripped and cleaved optical fiber end (Figure 3-15). This approach had many issues.
72
First, the cleaved optical fiber needed to be in contact with the waveguide end
(Figure 3-15a). Often, this approach would destroy the waveguide samples (which
are difficult enough to make already, causing added frustration). Second, the cleaved
optical fiber confined light in a much larger diameter beam on the order of several
hundred microns in diameter. As a result, not all of the light could couple into the
small (~8-10 micron diameter) waveguide. Much of the uncoupled light would
scatter along the waveguide and significantly distort the loss measurements. Finally,
index matching fluid and UV-curing adhesives were tried to improve this coupling.
However, it was nearly impossible to apply a small drop of the index matching fluid
to cover only the very tip of the optical fiber and waveguide end. Usually, the index
matching fluid or adhesive would form a messy blob on the fiber and waveguide, not
only worsening coupling but also destroying the sample (Figure 3-15b,c).
Xiaomin Zhang found that the lensed fibers could solve these light coupling
problems. The lensed fibers (Oz Optics part no. TSMJ-3A-1550-9/125-0.25-7-2-12-
1 for single mode and TPMJ-3A-1550-8/125-0.25-18-2-12-1 for polarization
maintaining) contain an optical fiber with a lens attached to the end. The lens is
optimized for focusing 1550nm light to a 2 micron-diameter spot, and has a 10
micron focal length. This allows efficient free-space coupling of the diode laser light
into the waveguides, with no index matching fluid or contact necessary. Lensed
fibers designed for other wavelengths were not currently available, so the 1550nm
lensed fibers were used for all wavelengths.

73

Figure 3-15: Compared to coupling with a cleaved fiber end (a), the index matching gel (b,c) is
difficult to apply to the fiber end and increases coupling loss and sample breakage.

While the lensed fibers work much better for efficiently coupling light into
the waveguides, they do have some important limitations. First, the lensed fibers are
very fragile and expensive (~$100 each), so they must be handled with extreme care.
The lensed fiber’s position must also be very carefully monitored on the testing setup
cameras, so that the lensed fiber does not accidentally bump or rub against the
waveguide sample, stage, or wafer. Second, the lensed fibers must be carefully and
securely mounted in a fiber holder (Newport 561-FC). It is important to only have
the very end of the lensed fiber protruding from the fiber holder, so that the lensed
fiber doesn’t sway in the ambient air and cause slight variations in coupling. (Later,
Mark Harrison and Xiaomin Zhang added a clear plastic enclosure around the
74
waveguide testing setup to further protect the lensed fiber from drifting and swaying
in ambient air).
The waveguide alignment stages in the testing setup are also very important,
as they allow precise alignment of the waveguide sample and lensed optical fiber.
The lensed fiber is placed in a holder (Newport 561-FC) on top of another XYZ
stage (Newport 561-TILT) so that the lensed fiber can be positioned close to the
sample using a motorized controller (Newport NanoPZ Series PZ5B). Another stage
(Newport 462 XYZ) is used to position the aspheric lens near the waveguide’s
output. The aspheric lens is held using Thorlabs posts and a LMR1 holder. The
waveguide sample is placed on carbon tape on a 561-GR mount which is on top of a
Newport 561D XZ stage. This allows the sample to be positioned within the travel
range of the lensed fiber stages (if placed directly on top of the stages without the
Newport 561-GR, the sample is too far from the lensed fiber and can’t be tested).
To aid in aligning the waveguide samples, Thorlabs DCC cameras, Navitar
optical columns, and Mitutoyo objectives are used to observe both the top and side
alignment of the waveguide and lensed fiber. The camera images are observed on
the computer, allowing both the vertical and horizontal alignment of the waveguide
sample and lensed fiber to be optimized.
The aspheric lens plays a critically important role in waveguide testing, as it
focuses the waveguide’s output light onto the face of a beam profiler or power meter
for analysis. Since the waveguide output is highly divergent, it was necessary to use
a lens to ensure the light can be read by the detector or beam profiler. After trying
several lenses, we found that the Thorlabs (A240TM) aspheric lens worked best, as
75
others could not sufficiently focus the waveguides’ highly divergent light. Another
challenge in focusing the waveguide output is placing the aspheric lens close enough
to the waveguide sample. To aid in focusing the light, the aspheric lens is mounted
in a small holder (Thorlabs LMR1) and placed on an XYZ stage with small enough
footprint to fit near the sample (Newport 460XYZ). It also helps to place the
waveguide sample on the stage such that the output end protrudes past the end of the
holder. This allows the aspheric lens to be brought very close (within its ~1cm focal
length) to the waveguide end for coupling.
Finally, after the waveguide alignment and input/output coupling is
optimized, the waveguide’s output can be analyzed on a power meter or beam
profiler (Thorlabs BP104-IR). In addition to measuring the output power, the beam
profiler can measure the shape of the beam and its changes in shape and phase over a
fixed distance.
To measure the loss of the fabricated waveguide devices, a lensed fiber is
used to couple light from a diode (Thorlabs MCLS) or tunable (Agilent) laser into
the waveguide ends. The waveguide’s output light is then focused with an aspheric
lens onto the face of a beam profiler (Thorlabs BP-104IR) or power meter (Thorlabs
PM-100B), as shown in Figure 3-14. Using this setup, the waveguides loss was
measured as a function of wavelength, input power, and input polarization.
3.5 Results and Discussion

3.5.1 Loss vs. Length, Polarization, and Input Power

Using the described testing setup in Figure 3-14, the loss of the waveguides
was determined at different input wavelengths, polarizations, and input powers.
76
First, waveguides of lengths ranging from 0.5-3mm were fabricated and their optical
loss values are measured at 658, 980, and 1550nm [5]. To measure the waveguide
loss, the waveguide alignment and coupling was first optimized at the 658nm visible
wavelength. Then, after measuring the loss at 658nm, the measurement was repeated
by carefully switching the laser to 980nm and 1550nm, without disturbing coupling.
By plotting the waveguides’ loss as a function of length (Figure 3-16), the
loss can be determined from the slope of the line and the coupling loss from the y-
intercept [5, 17]. Following these procedures, the waveguides’ loss was found to be
nearly constant with respect to wavelength, with measured loss values of [0.97±0.22,
0.82±0.27, 0.73±0.13] dB/cm for [658nm, 980nm, 1550nm]. The coupling losses
are measured to be [7.55, 6.89, 5.70] dBm at these wavelengths. To facilitate
comparison of the losses at different wavelengths, the coupling losses are not plotted
in Figure 3-16 as they are intrinsic the device.

Figure 3-16: Experimentally measured loss versus length of waveguides at 658, 980, and 1550nm
wavelengths. The measured loss is [0.97±0.22, 0.82±0.27, 0.73±0.13] dB/cm at [658nm, 980nm,
1550nm]. The coupling losses are measured to be [7.55, 6.89, 5.70] dBm at these wavelengths [5].

77
To characterize the dependence of loss on input power and polarization, a
similar set of experiments is performed. Using the aforementioned waveguide
testing setup, the output power of a waveguide sample was measured while changing
the power or polarization of the input light [5]. The waveguide’s output power
increased linearly with input power, indicating that the waveguides are free of
power-dependent nonlinear effects over the 5mW-15mW range studied (Figure 3-
17b). Light at various input polarizations was coupled into a single waveguide
sample using the same testing setup and a 3-paddle polarization controller. Since the
waveguide loss remained constant with changing polarization, we conclude that the
waveguide loss is also polarization-independent (Figure 3-17a).

Figure 3-17: The waveguide loss is constant at different polarization states and loss increases linearly
with increasing input powers [5].

This constant, low loss behavior is expected, since the waveguides are made
of silica and in a linear geometry. However, these favorable optical properties will
make the waveguides useful in a variety of applications, including biosensing and
integrated optics.

78
3.5.2 Efiron Cladding

To further reduce the loss of the waveguides, we also experimented with
using Efiron instead of air as a waveguide cladding. Efiron is a polymer with
refractive index of ~1.39, which is closer to silica’s refractive index of ~1.45. By
coating the silica waveguides with Efiron, the refractive index contrast between the
silica core and the cladding is decreased, which reduces the number of optical modes
in the waveguide and can potentially reduce the loss as well. Since very little data
was available for Efiron, we needed to experiment with applying, curing, and baking
the Efiron polymer to find the best parameters for coating waveguides with Efiron.
Nick Benzoni, then a high school student, assisted with the initial characterization of
Efiron films (see Appendix A). We then applied Efiron to some finished
waveguides, and obtained the loss data shown in Figure 3-18.
2 4 6 8
-2
-1
0

Loss data
Fit
Loss (db)
Waveguide length (mm)
Y=-0.12172-0.15677X

Figure 3-18: Loss data for waveguides with Efiron cladding. We measured a loss of approximately
0.15dB/mm or 1.5 dB/cm, which is noticeably higher than the ~0.8dB/cm loss of the uncoated
waveguides.

79
While we observed a higher loss for Efiron cladding waveguides (Figure 3-
18), the Efiron coating may also have higher optical absorption compared to silica
and could also increase scattering losses. The coating also seemed to increase
scattering and coupling losses at the ends of the waveguides. In the future,
optimizing the waveguide coating processes may enable development of few to
single mode waveguides with lower losses and better performance.
3.6 Conclusion

In summary, a low loss silica waveguide was successfully modeled and
fabricated directly on silicon wafers. By following a fabrication process similar to
that of the toroid, low-loss, smooth cylindrical waveguides can be made. As
anticipated, due to the low material loss and highly linear optical properties of silica,
the waveguides demonstrated low and constant optical loss at different wavelengths,
polarization states, and input powers. Ultimately, the loss of the waveguides is
limited by how smooth and straight they can be reflowed. If the reflow process can
be made more reliable and efficient, the loss of the waveguides could be significantly
improved. For example, replica molding or other soft lithography processes could
make the waveguide fabrication process faster, more reliable, and more reproducible.
This work also helped lead to development of other types of waveguides, such as
Xiaomin Zhang’s trapezoidal waveguides and splitters [9, 20], with improved
properties and behavior.

80

Chapter 3 References

1. J. F. Bauters, M. J. R. Heck, D. John, D. X. Dai, M. C. Tien, J. S. Barton, A.
Leinse, R. G. Heideman, D. J. Blumenthal, and J. E. Bowers, "Ultra-low-loss high-
aspect-ratio Si3N4 waveguides," Optics Express 19, 3163-3174 (2011).

2. M. Ferrera, L. Razzari, D. Duchesne, R. Morandotti, Z. Yang, M. Liscidini, J.
E. Sipe, S. Chu, B. E. Little, and D. J. Moss, "Low-power continuous-wave
nonlinear optics in doped silica glass integrated waveguide structures," Nature
Photonics 2, 737-740 (2008).

3. P. Kozma, A. Hamori, S. Kurunczi, K. Cottier, and R. Horvath, "Grating
coupled optical waveguide interferometer for label-free biosensing," Sensors and
Actuators B-Chemical 155, 446-450 (2011).

4. K. K. Lee, D. R. Lim, H. C. Luan, A. Agarwal, J. Foresi, and L. C.
Kimerling, "Effect of size and roughness on light transmission in a Si/SiO2
waveguide: Experiments and model," Applied Physics Letters 77, 1617-1619 (2000).

5. A. J. Maker, and A. M. Armani, "Low-loss silica-on-silicon waveguides,"
Optics Letters 36, 3729-3731 (2011).

6. T. Miya, "Silica-based planar lightwave circuits: Passive and thermally active
devices," IEEE J. Sel. Top. Quantum Electron. 6, 38-45 (2000).

7. M. R. Poulsen, P. I. Borel, J. Fage-Pedersen, J. Hubner, M. Kristensen, J. H.
Povlsen, K. Rottwitt, M. Svalgaard, and W. Svendsen, "Advances in silica-based
integrated optics," Opt. Eng. 42, 2821-2834 (2003).

8. A. L. Washburn, and R. C. Bailey, "Photonics-on-a-chip: recent advances in
integrated waveguides as enabling detection elements for real-world, lab-on-a-chip
biosensing applications," Analyst 136, 227-236 (2011).

9. X. M. Zhang, and A. M. Armani, "Suspended bridge-like silica 2 x 2 beam
splitter on silicon," Optics Letters 36, 3012-3014 (2011).

10. X. M. Zhang, and A. M. Armani, "Silica microtoroid resonator sensor with
monolithically integrated waveguides," Optics Express 21, 23592-23603 (2013).

11. Q. Lin, O. J. Painter, and G. P. Agrawal, "Nonlinear optical phenomena in
silicon waveguides: Modeling and applications," Optics Express 15, 16604-16644
(2007).

12. E. D. Palik, Handbook of Optical Constants of Solids (Elsevier, 1985).
81

13. R. Kitamura, L. Pilon, and M. Jonasz, "Optical constants of silica glass from
extreme ultraviolet to far infrared at near room temperature," Applied Optics 46,
8118-8133 (2007).

14. D. Duchesne, M. Ferrera, L. Razzari, R. Morandotti, B. Little, S. Chu, and D.
Moss, "Efficient self-phase modulation in low loss, high index doped silica glass
integrated waveguides," in Optics Express(Optical Society of America, 2009), pp.
1865-1870.

15. B. Saleh, and M. Teich, Fundamentals of Photonics (Wiley-Interscience,
2007).

16. R. Germann, H. W. M. Salemink, R. Beyeler, G. L. Bona, F. Horst, I.
Massarek, and B. J. Offrein, "Silicon oxynitride layers for optical waveguide
applications," J. Electrochem. Soc. 147, 2237-2241 (2000).

17. A. J. Maker, and A. M. Armani, "Low-loss silica on silicon integrated
waveguides," in Conference on High Contrast Metastructures(San Francisco, CA,
2012).

18. A. J. Maker, and A. M. Armani, "Fabrication of Silica Ultra High Quality
Factor Microresonators," in Journal of Visualized Experiments(2012).

19. D. K. Armani, T. J. Kippenberg, S. M. Spillane, and K. J. Vahala, "Ultra-
high-Q toroid microcavity on a chip," Nature 421, 925-928 (2003).

20. X. M. Zhang, M. Harrison, A. Harker, and A. M. Armani, "Serpentine low
loss trapezoidal silica waveguides on silicon," Optics Express 20, 22298-22307
(2012).

82

Chapter 4 High Refractive Index Sol-gel Silica and Applications

4.1 Introduction

Silica is an excellent material in many photonics applications because of its
low optical loss and minimal non-linear behavior [1-3]. However, for these same
reasons, silica-based optical devices are often limited to passive components [4, 5].
Therefore, one active area of research is developing new kinds of optical materials
with active characteristics including laser behavior, unique properties, and improved
performance over plain silica. While silica materials are commonly made by vapor
deposition or by thermal oxidation of silicon in a furnace, these methods offer only
limited tuning of silica’s optical properties [3, 6, 7]. To truly create a custom silica
matrix, it is necessary to intercalate a dopant or gain medium, such as rare earth
elements or nanoparticles. These types of dopants can enable applications such as
lasers and modulators [8-18].
Compared to other doping approaches, synthesis by a sol-gel method allows
significantly more control and flexibility in silica’s resulting material properties [19].
Sol-gel silica is made by reacting and annealing liquid silica precursors to form a
silica matrix [16, 19-21]. Dopants can be easily and uniformly added to liquid sol-
gel silica as it forms. Sol-gel silica materials can therefore provide an interesting
and useful method to develop new kinds of silica films with unique properties.
Devices fabricated from sol-gels containing active materials have found
numerous applications as low threshold lasers [10, 22-26], thin films and coatings
83
[12, 27-30], and hybrid devices [14, 15, 17, 31-34]. One application which is
especially important for communications and integrated optics is the development of
high refractive index materials [29, 35-37]. Devices made from silica materials are
difficult to directly integrate on silicon because the refractive index of silica is lower
than that of silicon. High refractive index silica materials provide a means to avoid
this problem. By sandwiching higher refractive index silica layers between layers of
lower refractive index, silica materials can be integrated on silicon wafers without
excessive loss of light to the silicon substrate [38]. In addition, high refractive index
coatings made from silica are more robust than other materials such as polymers, and
retain many of silica’s favorable optical properties, making them useful in a wider
range of applications.
4.2 Background and Motivation

In the present work, silica films with controllable refractive index are
synthesized by adding small amounts of titanium (via the titanium butoxide
precursor) to sol-gel silica [29, 37, 39]. The silica sol-gels are first prepared by acid-
catalyzed hydrolysis and condensation of tetraethoxysilane (TEOS), methyl
triethoxysilane (MTES), and titanium butoxide precursors, forming a silica matrix
(Figures 4-1 and 4-2). In addition to developing and characterizing these high index
materials, two applications are demonstrated. First, by spin-coating these sol-gel
silica films onto silica toroids, the behavior of light inside the toroid resonator can be
tuned [29]. Second, the titanium in the high index coatings is shown to enhance
Raman lasing, allowing improved Raman lasers to be developed from titanium-
doped sol-gel silica devices [40].
84
Silica films with increased refractive index are synthesized by adding various
amounts of titanium butoxide to sol-gel silica made with the MTES precursor. Like
the silica precursors, titanium butoxide can also undergo hydrolysis and
condensation reactions, enabling it to enter the silica matrix.

Figure 4-1: Hydrolysis (a) and condensation (b) reactions of TEOS, forming a silica matrix (c) [39].

85

Figure 4-2: Hydrolysis (a) and condensation (b) reactions of MTES. When titanium butoxide is
added, it also undergoes hydrolysis, allowing it to integrate within the silica matrix (c) [39].

Adding titanium to the MTES sol-gels increases the refractive index due to
the high polarizability of the metal ions and the presence of the empty Ti
4+
d orbital
which causes increased optical absorption [35]. Therefore, by adding varying
amounts of titanium to sol-gel silica films, silica coatings with a tunable range of
refractive indices can be developed [37].
4.3 Experimental Approach

The silica sol-gels are fabricated as follows and as outlined in the following
tables. First, the silica precursor (MTES or TEOS), ethanol solvent, water, and
hydrochloric acid are combined following the molar ratios in Table 4-1. It is
86
necessary to use MTES as the precursor instead of TEOS for the titanium-doped
silica films because MTES contains a pendant methyl group (Figure 4-2). Pendant
groups, such as methyls and phenyls, allow a more open silica matrix to be produced
in the resulting silica film [15]. Since the titanium atoms are significantly larger than
silicon atoms, the titanium dopant fits more easily inside the MTES matrix. As a
result, the MTES-based Ti-doped films have minimal cracking. In contrast, when
titanium butoxide was added to TEOS, significant cracking was observed. The
synthesis procedures for both films are outlined in the following section.
The exact masses of each component are listed in the highlighted rows of
Table 4-1. For this work, a total of three silica sol-gel films were fabricated and
characterized [37, 39]:
1) a silica film made from TEOS (tetraethoxysilane) precursor containing no
titanium (“TEOS”),
2) a silica film made from MTES (methyltriethoxysilane) precursor
containing a 0.1:1 ratio of titanium to MTES (“MTES R=0.1”); and
3) a silica film made from MTES containing a 0.3:1 ratio of titanium to
MTES (“MTES R=0.3”).
87
Sol-gel silica recipes
TEOS recipe
Chemicals TEOS Ethanol HCl H
2
O
Vendor
Alfa Aesar,
99.999%
EMD,
200 proof EMD Clean room
Molecular Weight 208.33 46.07 36.46 18.02
Purity (mass) 100% 100% 36% 100%
Molar ratio 1.00 4.000 0.10 2.00
Desired Chemical Mass 5.000 4.423 0.088 0.709
Desired Solution Mass (g) 5.000 4.423 0.243 0.709

MTES R=0.1 recipe
Chemicals MTES Ethanol HCl Ti(OBu)
4

Vendor
Alfa Aesar,
98%
EMD,
200 proof Sigma Aldrich Sigma Aldrich
Molecular Weight 178.30 46.07 36.46 340.36
Purity (mass) 100% 100% 36% 98%
Molar ratio 1.00 16.000 0.07 0.10
Desired Chemical Mass 1.000 4.134 0.037 0.191
Desired Solution Mass (g) 1.000 4.134 0.103 0.195

MTES R=0.3 recipe
Chemicals MTES Ethanol HCl Ti(OBu)
4

Vendor
Alfa Aesar,
98%
EMD,
200 proof Sigma Aldrich Sigma Aldrich
Molecular Weight 178.30 46.07 36.46 340.36
Purity (mass) 100% 100% 36% 98%
Molar ratio 1.00 16.000 0.07 0.30
Desired Chemical Mass 1.000 4.134 0.037 0.573
Desired Solution Mass (g) 1.000 4.134 0.103 0.584
Table 4-1: Recipes used to make TEOS, MTES R=0.1, and MTES R=0.3 sol-gels. The amount of
each component used is shown in the highlighted rows.

The ingredients are combined in the following order: add precursor (MTES
or TEOS) to ethanol; stir 5 minutes, then add water (to TEOS sol-gels only), stir 5
minutes; then add HCl. Once HCl is added, the hydrolysis reaction begins and the
sol-gels are allowed to react for 2 hours at room temperature, with continuous
~300rpm stirring by a magnetic stirrer. The titanium butoxide is added to the MTES
88
sol-gels after letting the hydrolysis run for 15-20 minutes. Hydrolysis of titanium
butoxide occurs more rapidly than that of MTES; therefore, adding the titanium
butoxide later prevents formation of a precipitate. The sol-gel is stirred for a total of
2 hours at room temperature while hydrolysis runs. After hydrolysis is complete, the
sol-gel is allowed to age at room temperature for 24 hours, filtered through a 0.45-
micron syringe filter, and then stored in a refrigerator until used (up to ~2 weeks, but
it is recommended to use sol-gels immediately or after a consistent amount of age
time to keep properties consistent).
After aging, the sol-gels are allowed to reach room temperature (if stored in
the refrigerator) and then spin coated onto bare silicon wafers or onto silica toroids.
The sol-gel is applied directly to silica toroids without any additional cleaning steps,
while the bare silicon wafers are first cleaned with acetone, methanol, and
isopropanol and dried with an air gun. After cleaning, one sample at a time is placed
on the spinner’s vacuum chuck and sol-gel is applied to the sample using a Pasteur
pipet.
Before running the spin cycle, it helps significantly to allow the sol-gel
covered wafer to stand undisturbed on the spinner chuck for 2 minutes. Brian Rose
found that this 2 minute waiting period (waiting 2 minutes between putting the sol-
gel on the sample and running the spinner) tends to produce much more uniform
films with significantly fewer cracks. This waiting step may help because the sol-
gels are highly non-Newtonian fluids and contain colloidal sol with a lot of ethanol
solvent. Giving the colloidal sol extra time to settle onto the sample may help the
remaining sol-gel stick and spin on better. After waiting 2 minutes, the spinner is
89
started, and the sol-gel is spun onto the wafers or toroids at 7000rpm for 30 seconds.
After spin coating, the samples are dried on a 75°C hot plate for 5 minutes to
evaporate the ethanol solvent. Finally, the samples are placed in a tube furnace and
annealed at 1000°C for 1 hour, with a ramp rate of 5°C/minute. This annealing or
densification step is crucial to the sol-gel synthesis process, as it allows the
condensation reaction to finish and the silica matrix to form. When annealing is
complete, additional layers of sol-gel can be added by repeating the procedure, or the
samples can be tested or etched in the cleanroom. The entire process is summarized
in Figure 4-3.

Figure 4-3: Flowchart summarizing the sol-gel synthesis process.

Once the sol-gels are made, various methods can be used to characterize the
properties. Infrared spectroscopy (FTIR) is used to verify formation of Si-O-Si and
Ti bonds, and the film thickness and refractive index can be measured by
90
ellipsometry. More complex properties such as the absorption coefficient and
thermo-optic coefficient are also measured by spin-coating the films onto toroid
resonators and studying the changes in Q and resonant wavelength. All of these
characterization results are described in the next section.
4.4 Results: Material Characterization

4.4.1 Infrared Spectra

After synthesis of the sol-gels, Fourier transform infrared (FTIR) spectra can
be measured to verify that the organic and solvent molecules have been removed and
the silica matrix has formed [41]. A Bruker-Optick ALPHA-P FTIR spectrometer is
used to perform measurements using attenuated total reflection (ATR). In the ATR
measurements, an anvil is used to place a sample in very tight contact with a
diamond. Infrared light is then passed through the diamond, and the light is
absorbed by the characteristic chemical bonds of the material. Analysis of the
resulting intensity spectrum allows the absorption peaks to be determined. Using
FTIR, the presence of Si-O-Si and Si-O-Ti bonds can be verified.
FTIR spectra of the thermally grown silica and annealed sol-gel silica films
(Figure 4-4) show distinct characteristic peaks of silica materials between 1080 and
1110 cm
-1
. All the silica films also have broad peaks near 1130 cm
-1
which
correspond to Si-O-Si bond vibrations [36, 41]. In addition, in the titanium-
containing MTES films, Si-O-Ti bonds are also present as indicated by the peak near
905 cm
-1
, as marked by an arrow in Figure 4-4.
91

Figure 4-4: FTIR spectra of thermally grown silica, the synthesized undoped silica, and Ti-doped sol-
gel silica films [37]. When titanium butoxide is added, an additional peak appears around 900cm
-1

which indicates formation of Si-O-Ti bonds in the silica matrix.

4.4.2 Refractive Index

For sol-gel silica films, it is straightforward to obtain refractive index vs.
wavelength data using an ellipsometer (Figures 4-5 and 4-6). As seen in the figures,
by increasing the titanium concentration, the refractive index of the films can be
tuned from 1.45 to over 1.6. By adding even more titanium to the films, it is possible
to achieve even higher refractive indices, although the optical absorption losses of
the film will also increase with increasing Ti concentration. Therefore, the present
work focused on films with lower Ti concentrations to minimize absorption losses.
92

Figure 4-5: Spectroscopic ellipsometry data for R=0.3 sol-gel: a) ψ and b) Δ values [37].

Figure 4-6: Refractive index vs. wavelength data for MTES and TEOS sol-gels (from spectroscopic
ellipsometry) [39].

4.4.3 Absorption Coefficient

Unlike the refractive index values, some properties such as the absorption
coefficient (absorption loss) and the thermo-optic coefficient are more difficult to
obtain. Recent work in the Armani lab demonstrated that optical toroid resonators
can be used to study and measure more complex optical properties such as material
93
loss and thermo-optic coefficient [42-44]. Since some of the light confined in optical
toroid resonators evanesces into the surroundings, spin coating materials onto toroid
resonators provides a useful means to study material properties. Therefore, to
measure the absorption loss and thermo-optic coefficient, the high refractive index
sol-gels are spin coated onto toroids (Figure 4-7) and tested as described in the
following sections.

Figure 4-7: SEM images of a) plain silica and b) TEOS-coated silica toroids [37].

Following procedures developed by Hong-Seok Choi to study polymer
optical properties [42-44], the absorption loss of the sol-gel coated toroids can be
determined by measuring the intrinsic quality factor. Since the surface scattering,
contamination, and radiation losses are generally assumed negligible for toroid
resonators with diameters larger than ~30 microns, the total quality factor becomes
dependent on only the material and coupling losses as explained in Chapter 2. The
intrinsic quality factor is independent of coupling loss and determined at zero
coupling. Thus, the intrinsic quality factor is assumed dependent only on the
material absorption losses, as described in Chapter 2 and calculated using Equation
2-22:
94

1 1
mat
Q Q
 
 
intrinsic
2
eff
mat
eff
n
Q Q

 
  (2-22)
Here, n
eff
is the effective refractive index of the toroid, α
eff
is the effective absorption
coefficient, and λ is the resonant wavelength.
Therefore, assuming negligible radiation, surface scattering, and
contamination losses, measuring the intrinsic quality experimentally factor allows
Q
mat
to be determined. The intrinsic Q is straightforward to measure by measuring Q
at different amounts of coupling, fitting a line to the data, and determining the y-
intercept, or Q at 0% coupling. Representative Q spectra are shown in Figure 4-8.

Figure 4-8: Representative Q spectra for TEOS, MTES R=0.1, and MTES R=0.3-coated toroids [37].

Knowing the intrinsic Q or Q
mat
, the values of n
eff
and α
eff
must be determined
to find the absorption loss of the coating. These are determined using Equations 2-
13 and 2-14 from Chapter 2:

eff resonator coating air
n n n n       (2-13)
and

eff resonator coating air
          (2-14)
95
In these equations, γ, χ, and δ are the fractions of light confined in the resonator,
coating, and air regions of the model, as determined from simulations in COMSOL
Multiphysics [45]. Representative COMSOL models for sol-gel coated toroids are
shown in Figure 4-9. As can be seen in the models, the circulating light shifts
radially into the coating as the coating’s refractive index increases. This change in
light’s behavior due to the coating is investigated further later in the chapter.

Figure 4-9: Representative COMSOL models for a) TEOS, b) MTES R=0.1, and c) MTES R=0.3-
coated toroids at the 633nm wavelength. The toroid in the simulation is 110 microns in diameter,
with a 10 micron minor diameter [37].

The values of γ, χ, and δ are easily calculated by integrating the electric field
intensity for each component and dividing by the total electric field intensity. The
absorption and refractive index values are taken from literature or measured
experimentally at 633 or 1300nm:
n
resonator
= 1.45067,
n
coating
= 1.4479, 1.5066, or 1.5895,
n
air
= 1;
α
resonator
= 0.01m
-1

α
air
= 0m
-1

96
After substituting the experimentally measured intrinsic Q and the above
values, it is straightforward to determine the remaining unknown, the material loss of
the coating. The calculated values are summarized in Table 4-2 [37]. As can be
seen, increasing the titanium concentration increases the absorption loss due to the
increased polarizability of the titanium in the silica matrix [35].
4.4.4 Thermo-Optic Coefficient

As with many material properties, the refractive index of the silica sol-gels is
temperature dependent. The thermo-optic coefficient dn/dT measures the change in
refractive index per unit temperature, caused by the competition between a material’s
thermal expansion and polarizability. Since the silica sol-gels studied are dielectric
materials, the polarizability term typically dominates, causing refractive index to
increase with increasing temperature and therefore resulting in a positive thermo-
optic coefficient [46]. (The opposite may be true for other materials, such as
polymers.) By tracking the resonant wavelength as a function of temperature, it is
possible to calculate the value of Δλ/ΔT [37, 43, 44]. Measuring Δλ/ΔT is a
straightforward experiment done by tracking the position of the resonant peak while
slowly increasing the temperature in ~0.5-1°C steps as shown in Figure 4-10. For
this investigation, the Δλ/ΔT was measured from approximately 20-60°C at 633nm
and 1300nm. To minimize hysteretic effects, the highest observed value of Δλ was
recorded for each temperature.
97

Figure 4-10: The resonant wavelength shifts as temperature increases, enabling Δλ versus ΔT to be
measured.

From the oscilloscope data, the resonant wavelength can be determined at
each temperature by fitting a Lorentzian to the resonant peak at each temperature
(Figure 4-10). Then, from the wavelength and temperature data, the values of Δλ
versus ΔT can be found and plotted. As shown in Figure 4-11, the resonant
wavelength of each coating varies linearly with temperature at both 633nm and
1300nm wavelengths, indicating that no significant nonlinear or other optical effects
are present [37, 39].
98

Figure 4-11: Representative Δλ versus ΔT data for a) TEOS, b) MTES R=0.1, and c) MTES R=0.3
coated toroids [37].

Using the measured value of Δλ/ΔT, the thermo-optic coefficient of the film,
dn
film
/dT can be calculated using Equation 4-1 [37, 42-44]:

eff
eff
dn
T n dT
  


(4-1)
Where λ is the resonant wavelength and n
eff
is the effective refractive index. From
the calculated value of dn
eff
/dT, the thermo-optic coefficient dn
coating
/dT is
determined with Equation 4-2:

eff coating
silica air
dn dn
dn dn
dT dT dT dT
      (4-2)
Here, χ, γ, and δ are the fractions of optical field confined in the silica toroid, sol-gel
coating, and surrounding air, respectively. These values are the same χ, γ, and δ
calculated in the COMSOL simulations and are also used to determine the absorption
and effective refractive index. Following the procedures just described, the thermo-
optic coefficients of each sol-gel material (TEOS sol-gel, MTES R=0.1, and MTES
R=0.3) were determined at 633nm and 1300nm wavelengths.
The calculated thermo-optic coefficients, absorption coefficients, and
refractive index are all summarized in Table 4-2 [37].
99
Material 
nm 
Thickness
nm
n
sol-gel

eff

m
-1


film
m
-1

 / T
nm/
o
C
dn
eff
/dT
o
C
-1
x10
-5

dn
film
/
dT
o
C
-1
x10
-5

633 350.17 ±
0.284
1.4547 0.181 0.978 0.0069 1.58 4.43 TEOS
1300 350.17 ±
0.284
1.4479 0.085 0.969 0.0144 1.59 7.13
633 373.31 ±
0.303
1.5181 15.81 28.41 0.0069 1.62 2.01 MTES R=0.1
1300 373.31 ±
0.303
1.5066 8.27 48.06 0.0142 1.60 3.96
633 360.96 ±
0.365
1.6183 63.51 79.12 0.0069 1.71 1.89 MTES R=0.3
1300 360.95 ±
0.365
1.5895 34.41 105.61 0.0146 1.66 2.83
Silicon

630 Bulk 3.8770* 32,700
#
31.2*
578 Bulk 1.4503* 1.19* Corning Fused
Silica

1367 Bulk 1.4477* 0.1
#
1.14*
* Reference [47]
# Reference [48]; reference values given for 1000 nm wavelength for fused silica, reference value for silicon given for 630 nm.

Table 4-2: Material absorption and thermo-optic coefficient data for high refractive index sol-gels
[37].

Each of the three sol-gel materials has a distinct and different thermo-optic
coefficient. The TEOS sol-gel has a thermo-optic coefficient which is approximately
twice as large as that of both the MTES films. This result is somewhat unexpected,
given the similar Δλ/ΔT values obtained. It is important to remember, however, that
the thermo-optic coefficient dn
coating
/dT is heavily influenced by how much light is
confined in the coating, toroid, and surrounding air. Since the amount of light in the
coating increases as the MTES film’s refractive index increases, the decreased
thermo-optic coefficient occurs due to the increased overlap between the toroid and
the MTES sol-gel coating. Knowing the properties of the films enables them to be
100
used in various applications, such as controlling the behavior of light, described
presently.
4.5 Application: Tailoring the Behavior of Light

4.5.1 Importance of Q, Mode Volume, and Circulating Power

Whispering gallery mode resonators integrated on silicon, such as silica
toroids, have become increasingly important in many applications because of their
high quality factors. For some applications, especially the development of lasers, it
is important to achieve very small mode volumes as well. Having a high ratio of
quality factor to mode volume, or Q/V ratio, enables high circulating powers and
Purcell factors to be achieved for development of efficient and low threshold lasers
[22, 39, 49-51]. The circulating power P
circ
and mode volume V
m
can be found using
Equations 2-9 and 2-10 from Chapter 2. The Purcell factor F
P
is given by Equation
2-12 [29, 39, 50, 51]:
P
circ
=
in
eff
P
R n
Q
*
2


(2-9)

2
3 2
2
max
( )
Q
V
m
r E d r
V
E




 
(2-10)

3
2
3
4
P
m eff
Q
F
V n


 

 
 
 
(2-12)
Based on these equations, there are several ways to improve the circulating power
and Purcell factor, including [39]:
- decreasing the size (diameter) of the devices
101
- decreasing the effective refractive index
- increasing the wavelength
Unfortunately, these approaches are not always effective or feasible. For example,
many applications require testing at specific wavelengths using devices made of
certain materials. Therefore, it may not be possible to significantly change
wavelength or effective refractive index. Decreasing the size of the devices can help,
but once the size becomes too small, high radiation losses quickly decrease the Q
factor and can negate the benefit of the decreased mode volume on the Q/V ratio
(Figure 4-12). It is therefore desirable to find other ways of achieving a high Q/V
ratio.

Figure 4-12: Theoretical (lines) and experimental (points) Q vs diameter and Q/V vs. diameter data
for silica toroids at λ=1300nm. As diameter decreases, the total quality factor is limited by radiation
losses [39].

One alternative approach to reduce the mode volume and improve the Q/V
ratio is to apply high refractive index coatings to the resonators. If the coating
material is thinner than the mode volume, the light circulating within the cavity will
become compressed in the high refractive index coating, therefore reducing mode
102
volume [29, 37, 39]. Recent work has shown that high refractive index polymer
coatings can improve the refractive index contrast and reduce mode volume, but
polymer coatings are not compatible with all fabrication processes and not as robust
as silica and silicon.
To address these concerns, the high refractive index silica sol-gels developed
in the Armani lab were spin-coated onto silica toroids [29, 37, 39]. By changing the
refractive index of the coating, the properties of the light circulating inside the toroid,
(such as the optical mode’s volume, position, and the quality factor) can be tuned.
To investigate these effects of the high refractive index coatings, both finite element
modeling and experimental measurements were performed.
4.5.2 COMSOL Modeling

In order to determine the coatings’ effects on the mode volume and quality
factor of the silica toroids, the electric field confinement and resonant frequencies
(both real and imaginary components) were simulated numerically. COMSOL
Multiphysics finite element method simulations were used, following methods
developed by Mark Oxborrow and Imran Cheema [45, 52]. These models calculate
the distribution of the light inside the toroid, as well as the real and imaginary
components of the resonant frequency. Models developed by Imran Cheema [52]
were modified by drawing in a 350nm thick high refractive index coating on the
toroid cross-section, in order to simulate the coating’s effects.
In particular, models of plain silica toroids (n=1.45067 at 1300nm) and
toroids coated with 350nm-thick layers of sol-gel (n=1.4479, n=1.5066, and
n=1.5895) were produced. When preparing the simulations, the models were drawn
103
to match experimentally fabricated toroids. Therefore, the toroid major diameter was
varied from 10-140µm, with minor diameters set to 4µm (D<20 µm) or 8µm (D>20
µm). In each model, a mesh size of 0.021 µm
2
was used and the fundamental mode
solution corresponding closest to 1300nm was used in the following calculations and
analysis [29].
4.5.2.1 Mode Volume Calculations

Using the COMSOL models, the mode volume and total quality factor were
calculated as follows. The mode volume is found using Equation 2-10 [50]:

2
3 2
2
max
( )
Q
V
m
r E d r
V
E




 
(2-10)
Both the numerator and denominator are determined directly from COMSOL. The
integral
2
3 2
( )
Q
V
r E d r 


 
is determined by integrating the electric field intensity
squared in the toroid, coating, and surrounding air regions of the model. The value
of ( ) r 

is the refractive index squared in each region. These terms are evaluated
using COMSOL’s subdomain integration tool for each region. Similarly, the value
of
2
max
E is determined using the max/min marker for the electric field intensity
squared (found in COMSOL’s plot parameters menu). Dividing the two quantities
gives the mode volume (one should make sure the result is in the desired units).
4.5.2.2 Q Calculations

The radiation quality factor Q
rad
is determined using Equation 2-18 from
Chapter 2 [52]:
104

) ( 2
) (
r
r
rad
f
f
Q


 (2-18)
Here, ( )
r
f  and ( )
r
f  are the real and imaginary components of the resonant
frequency. Both values are directly given in the COMSOL model solution. One
should be careful to ensure that the resonant frequency values used are taken from
the fundamental mode solution of the model and correspond to the desired resonant
wavelength (here, ~1300nm).
Finally, the material quality factor Q
mat
is found using the procedures in
Chapter 2 and Equation 2-20 [4, 43]

2
eff
mat
eff
n
Q

 

(2-20)
where n
eff
and α
eff
are the effective refractive index and effective absorption
coefficient at the wavelength λ. As described previously [37], these are determined
using Equations 2-13 and 2-14:

eff resonator coating air
n n n n       (2-13)

eff resonator coating air
          (2-14)
where γ, χ, and δ are the fractions of light confined in the resonator, coating, and air
regions of the model. The values of γ, χ, and δ are easily calculated by integrating
the electric field intensity for each component and dividing by the total electric field
intensity. The absorption and refractive index values are taken from literature or
measured experimentally at 1300nm:
n
resonator
= 1.45067,
n
coating
= 1.4479, 1.5066, or 1.5895,
105
n
air
= 1;
α
resonator
= 0.01m
-1

α
coating
= 0.969, 48.06, or 105.61m
-1
(for n=1.4479, 1.5066, and 1.5895 coatings,
respectively
α
air
= 0m
-1

By using these values and the γ, χ, and δ from COMSOL, both the effective
refractive index and effective absorption coefficient can be determined, allowing
Q
mat
to be calculated for each material and device.
Finally, the total quality factor is calculated assuming material loss and
radiation loss are dominant, while surface scattering, contamination loss, and
coupling loss are negligible. Therefore, the total quality factor is calculated using the
previously determined values of Q
rad
and Q
mat
and Equation 2-17 [29]:

1 1 1
total mat rad
Q Q Q
  
  (2-17)
It is important to note that since the Q
total
here does not account for other loss
sources, like coupling loss and scattering loss. Also, in the present work, it is
assumed that the toroid’s sol-gel coating is uniform on the entire toroid; however, it
is possible that the toroid’s coating is thinner or thicker in some regions due to flow
effects while spin-coating. Therefore, given these assumptions, it is possible that
the experimental and theoretical data may differ slightly.
4.5.3 Results and Experimental Verification

While the quality factor and mode volume can be theoretically determined, it
is important to also experimentally verify the theoretical values. Therefore, silica
toroids ranging from 10-140µm were fabricated using the standard procedures.
106
Then, the toroids were spin-coated with one of the sol-gel coatings developed
previously: the first synthesized with TEOS precursor with refractive index 1.4479;
the second and third using MTES precursor and titanium butoxide added to make a
refractive index of 1.5066 (R=0.1) or 1.5895 (R=0.3), respectively. Only the
titanium concentrations are varied; the molar ratios of precursor to solvent, acid
catalyst, and water are kept constant. After spin-coating the sol-gels onto the toroids
at 7000rpm, the films are annealed in a tube furnace at 1000°C following the
procedures described previously. The resulting toroids have a smooth, uniform sol-
gel coating which is approximately 350nm thick (Figure 4-13). More details on the
sol-gel fabrication can be found in the previous section.

Figure 4-13: Rendering (a) and SEM image (b) of sol-gel coated toroid [29].

The intrinsic quality factor for each toroid was determined by measuring Q
versus coupling and finding the y-intercept (Q at zero percent coupling). For each
sample, the fundamental mode peak closest to 1300nm was measured to ensure
agreement with the simulations. In addition, the quality factors were measured at
very low input powers (10µW) to minimize thermal nonlinear effects in the data.
Using the previously described procedures, the experimental intrinsic Q versus
diameter data and the theoretically calculated mode volume, Q
mat
, Q
rad
, and total
107
quality factor can be plotted for each coating and geometry. From these plots, the
effects of the coating are studied for a range of toroid geometries.
First, the mode volume is studied using a plot of mode volume as a function
of toroid diameter for the various coatings (Figure 4-14). When the refractive index
of the coating increases, the mode volume in the toroid significantly decreases, by up
to 30%. The greater the coating’s refractive index, the greater the decrease in mode
volume. This decrease in mode volume results from the light shifting radially
outward from the silica toroid into the thin, higher refractive index coating. Since
the coating is very thin, the light is compressed as it shifts into the coating. The
increasing amount of light confined in the coating is easily seen in Figure 4-14a-b: as
the coating’s refractive index increased, the volume of the light confined in the
coating increases significantly. This shift and compression of the light circulating in
the toroid is also noticeable in the mode cross-sections, as the coating’s refractive
index increases from 1.4479 (Figure 4-14c) to 1.5066(d) and 1.5895 (e). By
controlling the refractive index of the coating, both the size and position of the
confined optical mode can be tuned.
108

Figure 4-14: The presence of the high refractive index coating decreases the toroid’s mode volume
(a) and increases mode volume in film as a function of diameter (b). FEM models of the optical field
in the coatings at 1300nm. As the coating's refractive index increases from c) 1.4479 to d) 1.5066 and
e) 1.5895, the light shifts radially outward into the coating [29].

In addition to lowering the mode volume, the presence of the high refractive
index sol-gel coatings also impacts the quality factors of the resulting toroids. Both
the theoretical quality factors and the experimental data are plotted as a function of
diameter in Figure 4-15. As seen in the figure, there is very good agreement between
the theoretical and experimental results. While the experimental quality factors are
consistently a bit lower than the theoretical data, this is most likely a result of
additional losses such as scattering loss, which were assumed negligible relative to
the material and radiation losses. It is possible that slight roughness,
109
nonuniformities, or defects on the spin-coated silica coatings may contribute to the
overall quality factor as well.

Figure 4-15: Theoretical and experimental Q versus diameter data for toroids coated with a)
n=1.4479, b) n=1.5066, and c) n=1.5895 sol-gels. The Q/V vs diameter is also shown for each
coating in d) [29].

From Figure 4-15, it is clear that the presence of titanium in the high
refractive index coatings causes increased material loss and lowers the resulting
quality factors. However, quality factors remained high, at least 100,000-1 million,
for the majority of devices tested. It is also important to note that as the refractive
index of the toroid coating increased, the radiation losses decreased due to the
presence of the high index coatings. As a result, at very small device diameters, the
110
decreased radiation loss counteracts the increased material loss of the coating and
enables fabrication of smaller devices while maintaining high quality factors.

4.6 Application: Titanium-Enhanced Raman Lasers

4.6.1 Background and Motivation

After completing the initial characterization of the titanium-doped sol-gel
silica films, the logical next step was to add rare earth dopants to the titanium-doped
film as well. In doing so, the intensity of light circulating in a rare earth-doped
coating could be increased, enabling development of potentially improved lasers.
Therefore, in collaboration with Nishita Deka, titanium and neodymium-
doped sol-gel coatings were synthesized. Neodymium was chosen for several
reasons: 1) because its possible 1064nm lasing emission could potentially benefit the
heterodyned work (see Chapter 6), 2) the high refractive index coatings have even
greater compressing effects at shorter wavelengths due to higher radiation losses and
higher refractive index values, and 3) neodymium can be pumped with our high
power 765nm laser, which can input 10-15 mW of power into the coated devices.
Using previous rare earth doped silica-coated toroid work as a starting point
[26], rare earth doped sol-gel silica coatings containing various amounts of titanium
and high concentrations (10-20mol%) of neodymium were synthesized. A high rare
earth dopant concentration is needed in a coated microlaser device because only a
small portion of the toroid’s circulating light can interact with the rare earth dopants
in the coating. It was anticipated that adding the titanium and increasing the light in
111
the coating would further improve the performance of these rare earth doped silica-
coated toroid lasers.
However, upon testing the neodymium and titanium-doped sol-gel coated
devices, no emission from neodymium was observed, even at very high 765nm input
powers. Instead, there was very strong Raman lasing. This suggested that titanium
could act as a sensitizer for Raman lasing. Therefore, we stopped doping the
titanium films with neodymium and instead focused on studying titanium’s effects
on Raman behavior.
First, an extensive literature search was performed to determine whether
titanium’s effects on Raman lasing in silica had been observed previously.
Surprisingly, very little work had been done on the effects of titanium on Raman
behavior in silica glasses. However, other groups had observed interesting effects in
other glasses like tellurites [53-55]. Based on the literature, adding heavy metals to
glasses increases the polarizability of the bonds in the glass networks. This
polarizability increase can also improve the glass’s nonlinear behavior, such as
Raman lasing [37, 40]. It is possible that these effects are occurring in the titanium-
doped sol-gel silica as well, since the increase in polarizability can also explain the
higher refractive index and absorption coefficient values observed earlier in the
titanium-doped sol-gel silica coatings [37].
The potential effects of titanium on silica’s Raman lasing behavior can be
measured by monitoring the Raman lasing threshold and Raman lasing efficiency. In
optical resonators, the threshold power, or minimum power needed for Raman lasing
is given by the relation P
threshold
= n
eff
2
V
m
/(gQ
2
) where n
eff
is the effective refractive
112
index, V
m
is the mode volume, g is the Raman gain coefficient, and Q is the quality
factor [40]. Therefore, Raman lasing is a complex process which depends on many
parameters. In silica, the Raman gain coefficient is very small [54, 56]. Therefore,
silica is not usually a favorable material for Raman. However, silica optical
resonators can achieve very high quality factors, which compensate for the low
Raman gain and enable low threshold Raman lasers to be developed.
It is interesting, then, to consider the effects of a high refractive index
titanium-doped silica coating on Raman lasing. While a high index titanium coating
may lower the Raman lasing threshold by improving the Raman gain and mode
volume, the titanium coatings also decrease the quality factor [29, 37, 40].
Therefore, in the present work, a trade-off is investigated: while adding titanium
could improve Raman gain, titanium also reduces the quality factor which would
worsen lasing behavior. By varying the titanium concentration in the high refractive
index coatings, this effect can be studied and optimized for Raman lasing
applications [40].
4.6.2 Experimental and Theoretical Approach

To study the effects of Raman lasing in the coated toroids, sol-gel silica films
doped with 0-10 mol% titanium in the starting solution (R=0 to R=0.1) were
prepared following the same procedures as described previously [37]. To help
ensure consistent film thicknesses, all films were aged 24 hours and then
immediately spin-coated and annealed onto wafers (aging longer or for inconsistent
amounts of time could affect the resulting coating thickness and refractive index
values).
113
After synthesis and aging, the films are spin-coated onto bare silica toroids
approximately 30 microns in diameter (Figure 4-16). As before, the films are
allowed to let stand on the samples for two minutes prior to starting the spin coating
cycle, which significantly improves the uniformity of the resulting film. The 30
micron toroid geometry was chosen for several reasons: 1) at 30 microns, at least
50% of the circulating 765nm light becomes confined in the high index coating
(according to COMSOL), 2) at 30 microns in diameter, very high circulating powers
can be achieved, which encourages lasing behavior and 3) it becomes difficult to
consistently reflow high Q coated toroids below 30 microns in diameter due to the
increased curvature of the device.

Figure 4-16: (a) Rendering and (b) scanning electron microscope image of silica toroid coated with
titanium-doped sol-gel.

The films are also spin-coated onto bare silicon wafers so that the thicknesses
and refractive indices can be measured. With help from Hari Mahalingam and Prof.
Steier, the exact refractive index and thickness values at 765nm could be measured
using a spectroscopic ellipsometer. Additionally, to check for sensitizing effects, the
Raman spectra of the control wafers were measured with the help of Rohan Dhall
and Prof. Cronin’s Raman spectrometer (using a 200mW 532nm excitation laser).
114
To determine the mode volume and effective refractive index of the coated
devices, COMSOL simulations were also performed using the same models and
procedures developed in the previous high index coatings work [29, 37].
To experimentally characterize the coated toroids and their Raman behavior,
the standard testing setup was used [40]. Light from the 765-781nm tunable laser is
coupled into the toroids using a tapered fiber. The tapered fiber’s output is observed
on a detector and oscilloscope. Resonant wavelengths can be found by observing a
drop in the tapered fiber’s transmission spectrum, as seen on the oscilloscope, and
the quality factor at a given wavelength can be found by fitting a Lorentzian to the
full width at half maximum (FWHM) of the resonant peak using the formula Q =
λ/Δλ. By scanning the tunable laser across the resonant wavelengths, the
fundamental mode peak is found and used in subsequent Raman experiments.
Raman lasing from the fundamental mode resonant peak is observed using a
spectrograph [40, 56]. The spectrograph’s fiber tip is placed as close as possible to
the toroid sample. To characterize the Raman behavior, the transmission spectra and
spectrograph traces are saved at a range of input powers. By plotting emission
intensity as a function of input power, threshold graphs can be obtained. Fitting a
line to the decreasing part of the data gives the laser’s efficiency from the slope and
threshold from the x-intercept.
4.6.3 COMSOL Results

Based on the spectroscopic ellipsometry data of the control wafers, the
presence of titanium increased the refractive index (at 765nm) from 1.458 in
titanium-free coatings to 1.514 in coatings containing 10 mol% titanium. The
115
coating thicknesses were approximately 177nm. Using this data, toroid cross-
sections were modeled in COMSOL [40]. To closely match the experimentally
fabricated devices, toroid cross-sections with major (minor) diameters of 30 (8)
microns and 177nm thick coatings were simulated. Following the ellipsometry data,
refractive indices of 1, 1.45, 1.458, 1.45, 1.494, 1.51, and 1.514 were used for the air,
thermal silica toroid, and the 0 mol%, 2 mol%, 5 mol%, 8 mol%, and 10 mol%
titanium coatings, respectively. The fundamental mode solution corresponding
closest to the 765m pump wavelength is used in the following analysis. Using the
same procedures described previously [29], the effective refractive indices are
calculated to increase from 1.34 to 1.39 as titanium is added, and the mode volume
decreases from 80.35µm
3
to 77.46µm
3
(Figure 4-17). The experimentally measured
quality factors also gradually decrease from ~2.4x10
7
to ~1.1x10
7
as titanium is
added [40].
Since Raman lasing is a complex phenomenon, it is important to first
investigate how the high refractive index coatings would affect Raman lasing,
assuming titanium does not have any sensitizing effects. Using the relation P
threshold

= n
eff
2
V
m
/(gQ
2
), the experimentally measured Q factors, and the n
eff
and V
m
values
from COMSOL, the relative effects of the titanium coatings on lasing threshold can
be determined. Assuming that the titanium has no sensitizing effect, it is expected
that the Raman threshold power would increase up to 5-fold [40]. This increase
happens because the decrease in Q factor (due to titanium’s increased material
absorption) has a much greater effect on the Raman threshold compared to the small
decrease in mode volume. However, this analysis does not consider the possible
116
sensitizing effects of titanium, which could improve Raman gain and also impact the
lasing threshold.

Figure 4-17: COMSOL simulations for toroid cross-sections with 177nm-thick coatings (left inset).
As the titanium concentration increases, the refractive index increases. The amount of light in the
coating also increases slightly while the total mode volume decreases (right inset) [40].

4.6.4 Experimental Results

To check for titanium’s possible sensitizing effects, both the Ti-doped sol-gel
films and coated toroids are studied. Initial Raman spectroscopy measurements on
the control wafers showed two strong Raman peaks near 500cm
-1
and 1000cm
-1

corresponding to Si (Figure 4-18). By comparing the ratio of the 500cm
-1
and
1000cm
-1
peak areas for each sample, the degree of polymerization I
p
can be
calculated (Figure 4-19) [57]. As the titanium concentration increases from 0 to
10mol%, I
p
is observed to increase from 3.55 to 5.26. This increase indicates that the
titanium is changing the silica bonds and increasing the Raman response of the thin
films [57]. Therefore, Ti could enhance Raman lasing behavior in Ti-coated toroids
also.
117

Figure 4-18: Representative Raman spectra of sol-gel silica film. Two characteristic Si peaks are
visible at 500cm
-1
and 1000cm
-1
.

Figure 4-19: Degree of polymerization I
p
vs. titanium concentration. As the titanium concentration
increases, the ratio of the 500cm
-1
and 1000cm
-1
peaks increases, indicating that titanium is changing
the silica bonds and increasing the Raman response.

118
To experimentally characterize the effects of titanium on the Raman
behavior, the Raman lasing threshold and efficiency of the coated toroids were
characterized. The first Raman lasing peak in the coated toroids was observed
~20nm away from the pump wavelength, compared to ~30nm away from the pump
wavelength in plain silica toroids (Figure 4-20a). As the input power was increased,
cascaded Raman lasing could also be observed (Figure 4-20b) [40].

Figure 4-20: Representative spectrograph data for sample with 10mol% Ti coating. (a) When
pumped near 775nm, the first Stokes Raman peak appears near 800nm, from which the emission vs.
intensity plot is produced (inset). (b) Cascaded Raman observed at higher input powers [40].

By measuring the Raman intensity as a function of coupled power, the
Raman thresholds and efficiencies are calculated. As seen in Figure 4-21, adding 10
mol% titanium causes the Raman efficiency to increase over 3-fold compared to a
coated device without titanium [40]. Additionally, unlike the 5-fold threshold
increase predicted by COMSOL, the measured Raman lasing thresholds decrease as
titanium is added. The measured thresholds for the 2, 5, 8, and 10 mol% titanium
coatings are all sub-mW (52.6±5.4µW, 113±9.87 W, 159.9±11.6 W, and
117±14.2 W, respectively). In contrast, a silica toroid with a titanium-free sol-gel
coating had a lasing threshold of 170±21.4 W [40].
119

Figure 4-21: The Raman efficiency increases as titanium concentration increases, indicating that
titanium enhances Raman behavior. The normalized Raman threshold shows an initial decrease and
gradual increase (inset), further illustrating the trade-off between titanium’s sensitizing effects, which
lower the Raman threshold, and titanium’s increased absorption loss, which increases the threshold by
lowering the quality factor [40].

By looking at the threshold data alone, it is difficult to determine what effect,
if any, the titanium has on the Raman threshold. Since the lasing threshold depends
heavily on mode volume, Q, and effective refractive index, any slight change in
device size, coating thickness, or Q can have a significant effect on the lasing
threshold. To account for these variations, the threshold is divided by (n
eff
2
V
m
/Q
2
).
When this normalization is done, a trend becomes apparent. As can be seen in the
Figure 4-21 inset, the lasing threshold initially decreases due to titanium’s sensitizing
effects. However, further increasing titanium concentration lowers the quality factor
and gradually causes the Raman threshold to increase. This trend therefore
illustrates the trade-off between titanium’s sensitizing effects which improve Raman
lasing, and titanium’s increased material absorption loss, which lowers quality factor
and worsens the lasing performance [40].
120
It is also important to note that the experimentally observed improvements to
Raman lasing performance (reduced lasing threshold and increased lasing efficiency)
are significantly different than the 5-fold increase in lasing threshold predicted by
COMSOL. Since the COMSOL models did not account for titanium’s sensitizing
effects, these results illustrate that titanium can not only increase refractive index and
reduce mode volume, it can also act as a sensitizer for studying Raman lasing and
other interesting nonlinear effects [40].

4.7 Conclusion

In summary, high refractive index coatings are characterized and studied [37].
The high refractive index coatings can decrease the mode volume by as much as
30% at 1300nm, reduce the effect of radiation losses, and enable tuning of the mode
position in the resonator. At the same time, high quality factors can still be achieved
[29]. Therefore, these high refractive index sol-gel coatings can also improve lasing
behavior, such as the titanium-enhanced Raman effects. By adding up to 10 mol%
titanium to the sol-gel silica, the Raman lasing threshold can be decreased while
increasing the lasing efficiency over 3-fold [40].
Therefore, these high refractive index sol-gel coatings will prove useful for
many applications such as lasing, in which it is desirable to have high quality factors,
small mode volumes, and a strong overlap between the optical mode and gain
medium. Considering that dopants can be easily added to sol-gel coatings, these
high refractive index films may prove useful for the development of sol-gel lasers
and in integrated optics applications.
121
Chapter 4 References

1. A. J. Maker, and A. M. Armani, "Low-loss silica-on-silicon waveguides,"
Optics Letters 36, 3729-3731 (2011).

2. M. R. Poulsen, P. I. Borel, J. Fage-Pedersen, J. Hubner, M. Kristensen, J. H.
Povlsen, K. Rottwitt, M. Svalgaard, and W. Svendsen, "Advances in silica-based
integrated optics," Opt. Eng. 42, 2821-2834 (2003).

3. R. Kitamura, L. Pilon, and M. Jonasz, "Optical constants of silica glass from
extreme ultraviolet to far infrared at near room temperature," Applied Optics 46,
8118-8133 (2007).

4. M. L. Gorodetsky, A. A. Savchenkov, and V. S. Ilchenko, "Ultimate Q of
optical microsphere resonators," Optics Letters 21, 453-455 (1996).

5. D. K. Armani, T. J. Kippenberg, S. M. Spillane, and K. J. Vahala, "Ultra-
high-Q toroid microcavity on a chip," Nature 421, 925-928 (2003).

6. C. K. Kao, H. Chang, W. Y. Lim, C. H. Tsai, C. C. Chi, N. H. Tai, and I. N.
Lin, "Optical properties of PECVD TEOS-SiO2 films," (Taylor & Francis Ltd2001),
pp. 1949-1954.

7. T. Hattori, S. Semura, and N. Akasaka, "Inductively coupled plasma-
enhanced chemical vapor deposition of SiO2 and GeO2-SiO2 films for optical
waveguides using tetraethylorthosilicate and tetramethylgermanium," Japanese
Journal of Applied Physics Part 1-Regular Papers Brief Communications & Review
Papers 38, 2775-2778 (1999).

8. B. J. Ainslie, S. P. Craig, and S. T. Davey, "The absorption and fluorescence-
spectra of rare-earth ions in silica-based monomode fiber," Journal of Lightwave
Technology 6, 287-293 (1988).

9. A. Beganskiene, S. Sakirzanovas, I. Kazadojev, A. Melninkaitis, V.
Sirutkaitis, and A. Kareiva, "Sol-gel derived antireflective coating with controlled
thickness and reflective index," Mater. Sci. 25, 817-824 (2007).

10. A. J. Berry, and T. A. King, "Characterization of doped sol-gel derived silica
hosts for use in tunable glass lasers," J. Phys. D: Appl. Phys 22, 1419-1422 (1989).

11. H. S. Choi, X. M. Zhang, and A. M. Armani, "Hybrid silica-polymer ultra-
high-Q microresonators," Optics Letters 35, 459-461 (2010).

122
12. M. A. Fardad, E. M. Yeatman, E. J. C. Dawnay, M. Green, and F. Horowitz,
"Effects of H2O on structure of acid-catalyzed SiO2 sol-gel films," J. Non-Cryst.
Solids 183, 260-267 (1995).

13. L. A. Farrow, and E. M. Vogel, "Raman-spectra of phosphate and silicate
glasses doped with the cations Ti, Nb, and Bi," J. Non-Cryst. Solids 143, 59-64
(1992).

14. M. Laczka, K. Cholewa-Kowalska, and M. Kogut, "Organic-inorganic hybrid
glasses of selective optical transmission," (Elsevier Science Bv2001), pp. 10-14.

15. G. Li, M. Kanezashi, and T. Tsuru, "Preparation of organic-inorganic hybrid
silica membranes using organoalkoxysilanes: The effect of pendant groups," J.
Membr. Sci. 379, 287-295 (2011).

16. M. R. N. Monton, E. M. Forsberg, and J. D. Brennan, "Tailoring Sol-Gel-
Derived Silica Materials for Optical Biosensing," Chemistry of Materials 24, 796-
811 (2012).

17. S. Nenkova, L. Radev, N. Rangelova, B. Aleksiev, and B. Samuneva, "New
sol-gel silica hybrids containing pectin and some metal ions," (2007), pp. 164-167.

18. P. B. Wagh, A. V. Rao, and D. Haranath, "Influence of molar ratios of
precursor, solvent and water on physical properties of citric acid catalyzed TEOS
silica aerogels," Mater. Chem. Phys. 53, 41-47 (1998).

19. J. Livage, and C. Sanchez, "Sol-gel chemistry," J. Non-Cryst. Solids 145, 11-
19 (1992).

20. R. L. Dumas, I. Tejedor-Tejedor, and M. A. Anderson, "Dependence of SiO2
gel structure on gelation conditions and sol reaction temperature as followed by
FTIR and nitrogen adsorption measurements," Journal of Porous Materials 5, 95-101
(1998).

21. M. A. Fardad, "Catalysts and the structure of SiO2 sol-gel films," J. Mater.
Sci. 35, 1835-1841 (2000).

22. L. N. He, S. K. Ozdemir, and L. Yang, "Whispering gallery microcavity
lasers," Laser & Photonics Reviews 7, 60-82 (2013).

23. H. S. Hsu, C. Cai, and A. M. Armani, "Ultra-low-threshold Er:Yb sol-gel
microlaser on silicon," Optics Express 17, 23265-23271 (2009).

24. A. J. Maker, and A. M. Armani, "Nanowatt threshold, alumina sensitized
neodymium laser integrated on silicon," Optics Express 21, 27238-27245 (2013).
123

25. S. Mehrabani, and A. M. Armani, "Blue upconversion laser based on
thulium-doped silica microcavity," Optics Letters 38, 4346-4349 (2013).

26. L. Yang, D. K. Armani, and K. J. Vahala, "Fiber-coupled erbium microlasers
on a chip," Applied Physics Letters 83, 825-826 (2003).

27. R. Gupta, S. Mozumdar, and N. K. Chaudhury, "Effect of ethanol variation
on the internal environment of sol-gel bulk and thin films with aging," Biosensors &
Bioelectronics 21, 549-556 (2005).

28. S. S. Latthe, H. Imai, V. Ganesan, and A. V. Rao, "Superhydrophobic silica
films by sol-gel co-precursor method," Appl. Surf. Sci. 256, 217-222 (2009).

29. A. J. Maker, B. A. Rose, and A. M. Armani, "Tailoring the behavior of
optical microcavities with high refractive index sol-gel coatings," Optics Letters 37,
2844-2846 (2012).

30. X. Zhang, Y. Y. Wu, S. Y. He, and D. Z. Yang, "Investigation on the atomic
oxygen erosion resistance of sol-gel alumina-silica composite films on Kapton,"
Mater. Chem. Phys. 114, 179-184 (2009).

31. E. Herrero, N. Carmona, J. Llopis, and M. A. Villegas, "Sensitive glasslike
sol-gel materials suitable for environmental light sensors," Journal of the European
Ceramic Society 27, 4589-4594 (2007).

32. A. Jitianu, M. Gartner, M. Zaharescu, D. Cristea, and E. Manea,
"Experiments for inorganic-organic hybrid sol-gel films for micro- and nano-
photonics," (Elsevier Science Bv2003), pp. 301-306.

33. H. K. Kim, S. J. Kang, S. K. Choi, Y. H. Min, and C. S. Yoon, "Highly
efficient organic/inorganic hybrid nonlinear optic materials via sol-gel process:
Synthesis, optical properties, and photobleaching for channel waveguides,"
Chemistry of Materials 11, 779-788 (1999).

34. J. Ren, L. H. Wang, X. Y. Han, J. F. Cheng, H. L. Lv, J. Y. Wang, X. G. Jian,
M. S. Zhao, and L. Y. Jia, "Organic Silicone Sol-Gel Polymer as a Noncovalent
Carrier of Receptor Proteins for Label-Free Optical Biosensor Application," Acs
Applied Materials & Interfaces 5, 386-394 (2013).

35. M. Abdel-Baki, F. A. A. Wahab, and F. El-Diasty, "Optical characterization
of xTiO(2)-(60-x)SiO2-40Na(2)O glasses I. Linear and nonlinear dispersion
properties," Mater. Chem. Phys. 96, 201-210 (2006).

124
36. G. Brusatin, M. Guglielmi, P. Innocenzi, A. Martucci, C. Battaglin, S. Pelli,
and G. Righini, "Microstructural and optical properties of sol-gel silica-titania
waveguides," J. Non-Cryst. Solids 220, 202-209 (1997).

37. B. A. Rose, A. J. Maker, and A. M. Armani, "Characterization of thermo-
optic coefficient and material loss of high refractive index silica sol-gel films in the
visible and near-IR," in Optical Materials Express(OSA, 2012), pp. 671-681.

38. R. Germann, H. W. M. Salemink, R. Beyeler, G. L. Bona, F. Horst, I.
Massarek, and B. J. Offrein, "Silicon oxynitride layers for optical waveguide
applications," J. Electrochem. Soc. 147, 2237-2241 (2000).

39. A. J. Maker, B. A. Rose, and A. M. Armani, "Controlling the mode volume
in high-Q microcavities with high refractive index coatings," Integrated Optics:
Devices, Materials, and Technologies Xvii 8627 (2013).

40. N. Deka, A. J. Maker, and A. M. Armani, "Titanium enhanced Raman
microcavity laser," Optics Letters 39 (2014).

41. P. Innocenzi, "Infrared spectroscopy of sol-gel derived silica-based films: a
spectra-microstructure overview," J. Non-Cryst. Solids 316, 309-319 (2003).

42. H. S. Choi, and A. M. Armani, "Thermal nonlinear effects in hybrid optical
microresonators," Applied Physics Letters 97 (2010).

43. H. S. Choi, S. Ismail, and A. M. Armani, "Studying polymer thin films with
hybrid optical microcavities," Optics Letters 36, 2152-2154 (2011).

44. H. S. Choi, D. Neiroukh, H. K. Hunt, and A. M. Armani, "Thermo-optic
Coefficient of Polyisobutylene Ultrathin Films Measured with Integrated Photonic
Devices," Langmuir 28, 849-854 (2012).

45. M. Oxborrow, "How to simulate the whispering-gallery-modes of dielectric
microresonators in FEMLAB/COMSOL - art. no. 64520J," in Conference on Laser
Resonators and Beam Control IX(San Jose, CA, 2007), pp. J4520-J4520.

46. M. Pokrass, Z. Burshtein, and R. Gvishi, "Thermo-optic coefficient in some
hybrid organic/inorganic fast sol-gel glasses," Optical Materials 32, 975-981 (2010).

47. G. Ghosh, Handbook of Thermo-Optic Coefficients of Optical Materials with
Applications (Academic Press, 1998).

48. E. D. Palik, Handbook of Optical Constants of Solids (Elsevier, 1985).

125
49. A. J. Maker, B. A. Rose, and A. M. Armani, "Tailoring the behavior of
optical microcavities with high refractive index sol-gel coatings," Optics Letters 37,
2844-2846 (2012).

50. S. M. Spillane, T. J. Kippenberg, K. J. Vahala, K. W. Goh, E. Wilcut, and H.
J. Kimble, "Ultrahigh-Q toroidal microresonators for cavity quantum
electrodynamics," Physical Review A 71 (2005).

51. F. Pisanello, A. Qualtieri, T. Stomeo, L. Martiradonna, R. Cingolani, A.
Bramati, and M. De Vittorio, "High-Purcell-factor dipolelike modes at visible
wavelengths in H1 photonic crystal cavity," Optics Letters 35, 1509-1511 (2010).

52. M. I. Cheema, and A. G. Kirk, "Implementation of the perfectly matched
layer to determine the quality factor of axisymmetric resonators in COMSOL," in
COMSOL Conference(Boston, 2010).

53. G. Dai, F. Tassone, A. L. Bassi, V. Russo, C. E. Bottani, and F. D'Amore,
"TeO2-based glasses containing Nb2(O5), TiO2, and WO3 for discrete Raman fiber
amplification," IEEE Photonics Technology Letters 16, 1011-1013 (2004).

54. R. Stegeman, C. Rivero, G. Stegeman, P. Delfyett, K. Richardson, L.
Jankovic, and H. Kim, "Raman gain measurements in bulk glass samples," Journal of
the Optical Society of America B-Optical Physics 22, 1861-1867 (2005).

55. L. A. Farrow, and E. M. Vogel, "Raman-spetra of phosphate and silicate-
glasses doped with the cations Ti, Nb, and Bi," J. Non-Cryst. Solids 143, 59-64
(1992).

56. M. Chistiakova, and A. M. Armani, "Cascaded Raman microlaser in air and
buffer," Optics Letters 37, 4068-4070 (2012).

57. P. Colomban, and A. Slodczyk, "Raman Intensity: An Important Tool in the
Study of Nanomaterials and Nanostructures," Acta Physica Polonica A 116, 7-12
(2009).

126
Chapter 5 NanoWatt-Threshold Nd
3+
and Alumina-Doped Toroidal
Microlaser

5.1 Introduction

Efficient and low-threshold lasers made from rare earth metal-doped silica
glasses have found many applications in industry, defense, and biological
investigations [1-7]. High efficiency and low threshold lasers are desirable as they
require less pump power to emit light, and they can efficiently convert pump power
to laser output. To make these lasers, rare earth ions (such as neodymium, cerium,
erbium, thulium, and ytterbium) are frequently added to silica because they contain f
orbital electrons with known excitation and emission wavelengths [8, 9]. Often it is
possible to achieve a wide range of emission wavelengths when exciting rare earth
ions with a single pump source. For example, pumping neodymium near 800nm
produces lasing near 900-940nm, 1050-1150nm, and ~1300nm [10, 11].
Upconversion lasing can also be observed, such as thulium’s blue emission when
pumped near 1064nm [12]. Another advantage of rare earth ions is that they have
relatively large absorption cross-sections, enabling them to be efficiently excited. As
a result, rare earth doped silica glasses have been developed into high efficiency and
low threshold optical fiber lasers.
While fiber lasers can still benefit many applications, developing integrated
optical devices including on-chip lasers is becoming increasingly desirable. Ideally,
all the individual components of an optical device would be fabricated directly on
silicon. To achieve this goal, it is necessary to transition from optical fiber-based
127
lasers to lasers which are easily fabricated on-chip. One approach is to fabricate
lasers from rare earth doped silica waveguides which are integrated on silicon [13-
15]. However, due to fabrication limitations and increased material losses, the loss
of these integrated silica waveguides is considerably higher, which limits lasing
performance.
To compensate for decreased performance, one can also increase the
concentration of the gain media (rare earth dopants) and/or increase the intensity of
light inside the waveguide. Yet these approaches can only do so much. Rare earth
dopants are not very soluble in silica and will form clusters that quench lasing
emission. Additionally, only a limited amount of dopant can fit in the silica matrix
without causing defects and cracking. It is also possible to increase the optical field
intensity to improve lasing performance, often by changing the refractive index of
the optical fiber or waveguide. Unfortunately, changing the refractive index is
generally unable to significantly improve light intensity by the orders of magnitude
needed to achieve sub-milliwatt or sub-microwatt lasing thresholds.
Another more promising approach to improve laser performance is
transitioning from a waveguide or fiber platform to an optical resonator platform [3,
6, 12, 16-20]. In waveguide and fiber lasers, photons from the excitation source
have just one opportunity to interact with and excite the rare earth dopants or other
gain media. On the other hand, optical resonators trap light in circulating orbits,
significantly increasing the amount of interactions between light and the gain media.
This increase is related to the photon lifetime or quality factor (Q) of the optical
resonator. The higher the Q, the longer the photons will be confined inside, and the
128
greater the intensity of the circulating light [21, 22]. For example, in a resonator
with a Q of 100 million, the increase will be ~100,000x. As a result, rare earth-
doped optical resonators are excellent platforms for highly efficient lasers with low
threshold powers for lasing.
As mentioned in the previous sections, silica toroid resonators have both high
quality factors and are integrated directly on silicon, enabling them to benefit many
applications, including the development of high efficiency and low threshold lasers
[1, 12, 18-20]. Since toroid lasers have ultra-high quality factors, they can also
achieve extremely narrow (sub-picometer or even sub-femtometer) lasing linewidths.
By tracking the narrow linewidth emission of a toroid laser, ultra-sensitive
microlaser sensors can also be developed, as described in Chapter 6 [23, 24].
Therefore, optimizing rare earth doped silica toroid lasers can benefit many fields.
In particular, a toroid microlaser is desired which
1) has high efficiency and an ultra-low lasing threshold,
2) emits light at wavelengths where water’s absorption is low (so that it can
be used in biosensing applications), and
3) emits light within the range of an available tunable reference laser, so that
it can be used in heterodyned sensing experiments.
While some toroid microlasers have been developed which satisfy one or two of the
above criteria, none have yet been developed which meet all three. By synthesizing
new hybrid silica materials, we have developed a neodymium and alumina-doped
toroid microlaser which satisfies all three criteria [19]. This significantly improved
toroid microlaser operates at 765nm pump and ~1064nm emission wavelengths
129
where water’s absorption is low, enabling operation in both air and aqueous
environments [25]. It has improved efficiency and measured lasing thresholds as
low as 530 nanowatts [19]. In addition, this improved toroid microlaser can be
heterodyned with an available 1055-1070nm tunable reference laser and used in
ultra-sensitive heterodyned detection experiments (see Chapter 6) [24].
5.2 Background and Motivation

5.2.1 Theory

To optimize the performance of rare earth-doped optical resonator lasers, it is
important to look at the theory [26]. For optical resonator-based lasers, the threshold
power (P
thres
) is related to the normalized upper state population (N
2
/N
T
), where N
T
is
the average rare earth dopant concentration and N
2
is the upper state population. In
high Q resonators, the normalized upper state population can be expressed using
Equation 5-1 [27]:

2
*
passive
s s
T s s
N
N g
 




(5-1)
where 
s
and
*
s
g are the Giles parameters, and
passive
s
 is the passive cavity signal
loss [27]. hese parameters are defined by Equations 5-2, 5-3, and 5-4 [27]:
130

a
s s T s
N     (5-2)

* e
s s T s
g N    (5-3)

,
2
eff passive
s passive
s T s
n
Q



 (5-4)
where 
s
is the overlap factor (the interaction strength between the optical intensity
and the rare earth distribution in the material), and
 
a e
s s
  describes the optical
absorption (emission) cross sections of the rare earth dopant. 
s
is the signal
wavelength, n
eff
is the effective refractive index, and Q
T,s
is the loaded quality factor
[27].
As can be seen from these relations, optimal lasing performance can be
obtained using high Q resonators containing dopants which have large absorption
cross sections. Following this approach, researchers have achieved lasing thresholds
as low as tens to hundreds of nanowatts in silica microspheres doped with rare earth
metals [6, 27-29]. While these results are promising, silica microspheres are not
easily integrated on silicon wafers and have similar limitations as optical fiber lasers.
A promising high Q alternative to the microsphere is the silica toroid
resonator. Silica toroids have ultra-high Q factors of over 100 million, and can be
fabricated directly on silicon substrates [21, 30]. Recently, rare earth doped silica
toroids have achieved high efficiency and low threshold lasing. An ytterbium and
erbium co-doped silica toroid had an ultra-low 4.2 microwatt lasing threshold [18],
and a neodymium-doped toroid had a 69µW threshold [31].

131
5.2.2 Fabrication of Yb
3+
and Nd
3+
-doped Toroid Microlasers

As mentioned in the previous section, the goal of this work is to develop a
toroid microlaser which can be heterodyned and used in biosensing applications [19,
24]. Specifically, the toroid microlaser needs to 1) have a low lasing threshold and
high efficiency, 2) emit light in the range of an available tunable reference laser (in
order to produce a heterodyne in Chapter 6), and 3) operate at pump and emission
wavelengths where the absorption of water is low.
To develop a microlaser which meets these criteria, a sol-gel method was
used similar to the ones described in Chapter 4. By adding rare earth dopants to sol-
gel silica, the resulting silica film can gain lasing properties at various wavelengths
[8]. Initially, this work focused on developing rare earth lasers made from
ytterbium, but previous work by Kelvin Kuo and others showed that the lasing was
near 1020-1050nm, outside the desired 1055-1070nm range [9].
Then, Simin Mehrabani suggested trying neodymium-doped silica.
Neodymium can potentially lase at 1064nm, exactly in the 1055-1070nm range of
the tunable reference laser [10]. In addition, the absorption of water is low at
neodymium’s ~800nm pump wavelength (Figure 5-1). Therefore, silica films and
toroids were doped with neodymium to study the lasing properties.
132

Figure 5-1: Absorption coefficient of water versus wavelength, as presented by Hale [25]. The
absorption of water is low, only ~2.5m
-1
and ~12 m
-1
at 765nm and 1064nm wavelengths.

Following the basic sol-gel procedures developed and optimized by
numerous people in the Armani Lab, including Heather Hunt, Simin Mehrabani,
Kelvin Kuo, and several undergraduate researchers, neodymium-doped sol-gel
microlasers were fabricated [18, 19, 32, 33]. In the argon glove box, small amounts
of neodymium nitrate were weighed in glass bottles (corresponding to sol-gels
ranging from 0.005-0.1wt% Nd). Meanwhile, the remaining sol-gel components
were prepared in separate glass bottles following the usual procedures and 1:4:0.1:2
TEOS:ethanol:HCl:water molar ratios (Table 5-1).
Briefly, specific amounts of TEOS, then ethanol, then water, and finally HCl
were added to a glass bottle, using a mass balance in a fumehood. The mixture was
stirred for 5 minutes at 300rpm between each addition. After the HCl was added, the
mixture was stirred for 5 minutes. Then, the entire TEOS+ethanol+water+HCl+stir
133
bar mixture was poured into the glass bottle containing the rare earth salt, and
allowed to stir for 2 hours at room temperature. Once stirring completed, the sol-gel
was removed from the stir plate, filtered with a 0.45µm syringe filter, and allowed to
age for at least 24 hours. After aging, the Nd
3+
sol-gels were placed in the fridge and
could be used for up to ~2 weeks.
Nd
3+
doped Sol-gel Silica Recipe
Chemicals TEOS Ethanol HCl H
2
O
Nd(NO
3
)
3

*xH
2
O
Vendor
Alfa Aesar,
99.999%
EMD,
200
proof EMD
Clean
room
Alfa Aesar,
99.99%
MW 208.33 46.07 36.46 18.02 330.25
Purity (mass) 100% 100% 36% 100% 99.99%
Molar ratio 1.00 4.000 0.10 2.00 0.0500
Desired
Chemical Mass 2.500 2.211 0.044 0.355 0.0100
Desired Solution
Mass (g) 2.500 2.211 0.122 0.355 0.0100
Table 5-1: Sample procedures used to make 0.05wt% neodymium-doped silica sol-gels.

To prepare Nd
3+
-doped silica films from the sol-gels, the following
procedures were used. The aged sol-gels were allowed to reach room temperature,
and then spin-coated onto bare silicon wafers at 7000rpm for 30 seconds. After
spinning, the sol-gel-coated wafers were baked on a 75 °C hot plate for 5 minutes to
evaporate solvents. Finally, the samples were placed in the tube furnace and
annealed at 1000 °C for 1 hour (with the usual 5 °C/min ramp rate). After the first
layer of sol-gel was annealed, this process was repeated at least one more time so
that two or more layers of sol-gel were added. Usually, only two layers of Nd
3+
were
spin coated, as significantly more cracking appears when more layers are added (in
134
the future, annealing the sol-gels under vacuum may enable thicker sol-gel layers to
be produced with less cracking).
Once the neodymium-doped silica films were made, silica toroids were
fabricated from the films following the usual photolithography, reflow, and etching
procedures [21, 34]. One must be careful during the BOE etch step, as the thin Nd
3+

sol-gel tends to etch within 5-7 minutes, but can be done in only 3-4 minutes when
the silica has more cracks present. Additionally, the thin layer of Nd
3+
sol-gel is very
fragile. After the Nd
3+
-doped microdisks have been made, it is best to reflow them
as soon as possible, otherwise they will gradually crack and break. When reflowing
these thin sol-gel disks, it is also best to monitor the laser power closely and ensure
the CO
2
laser is well-aligned, as the thin sol-gel disks are more difficult to reflow
uniformly than the thicker oxide disks.
Following the above procedures, neodymium-doped sol-gel silica toroids
were made with a variety of dopant concentrations. To measure lasing, the standard
resonator testing setup was used, except the toroid output was connected to a 90-10
splitter. The 90% end is sent to the OSA, with the remaining 10% sent to the
detector. Using this setup, resonant wavelengths can be found and viewed on the
oscilloscope, and lasing can be observed on the OSA. The first sample tested, a
~100µm diameter toroid with ~0.02wt% Nd (Figure 5-2a) showed strong lasing
peaks near 1100nm, and had a low lasing threshold of ~50µW (Figure 5-2c, d). The
quality factor of the resonant peak was only ~3.8x10
5
but had significant thermal
broadening (Figure 5-2b). It is therefore very likely that lower lasing thresholds can
be achieved by optimizing the Nd concentration and geometry.
135

Figure 5-2: Lasing results for neodymium-doped silica. Neodymium-doped toroids (a) could be
pumped around 765-781nm. High Q, broadening resonance peaks (b) produced lasing in the 1080-
1150nm range, depending on coupling. The lasing threshold is also very low (d).

While the Nd
3+
-doped sol-gel silica emitted light very well, the emission was
unfortunately still outside the desired 1055-1070nm range of the reference laser.
Several papers in the literature describe Nd
3+
-doped sol-gel films for optical fibers
and as coatings. In bulk silica, it is known that the absorption near 800nm and lasing
near 1064nm is caused by electrons being excited to the
4
F
3/2
energy level, and
returning to the
4
I
9/2
and
4
I
11/2
levels.
Since the neodymium-doped silica did not lase at the expected 1064nm
wavelength, the next logical step was to investigate why. Upon searching the
136
literature, it became clear that the neodymium was lasing at lower energy (higher
wavelength) regions due to its amorphous sol-gel silica environment. In amorphous
silica, neodymium’s lasing red shifts from 1064nm to 1080-1090nm because the
structural properties of amorphous sol-gel silica are different than those of fused
silica or quartz [10, 11]. According to the literature, amorphous sol-gel silica not
only has a higher phonon energy, it also can have unreacted OH groups in the silica
matrix [10, 14, 19]. Both of these factors can quench rare earth lasing, increase
nonradiative decay, and cause the laser’s emission to red-shift to the lower energy
wavelengths. Additionally, rare earth dopants have limited solubility in silica and
form clusters which can further quench lasing [10, 11, 15, 19, 35-38].
Based on the literature, the amorphous silica material does not appear
particularly promising for the Nd
3+
lasers. However, sol gel silica can be further
tuned by adding other materials. By changing the neodymium’s surrounding
environment, the lasing behavior of the neodymium could be changed and
potentially shifted back into the desired 1055-1070nm range. One approach is to use
different glasses instead of silica. For example, many chalcogenide glasses such as
CaF
2
have very favorable optical properties as well as lower material absorption and
phonon energies than silica. It is even possible to prepare some of these glasses
using sol-gel approaches [39-48]. However, the glass needs to be compatible with
our current fabrication procedures for silica toroids. To overcome this issue, hybrid
glasses can be made which contain both silica and other materials which enhance the
lasing behavior of rare earth metals. Two promising approaches I tried were adding
CaF
2
crystals or alumina. These are discussed in the next sections.
137

5.2.3 Adding CaF
2
to Improve Lasing

Numerous interesting presentations at the 2012 Photonics West conference
described how rare earth dopants could behave very favorably in non-silica
environments, such as CaF
2
and chalcogenide glasses [39, 40, 48]. I found two
especially interesting hybrid materials in the literature, involving CaF
2
and alumina
[33, 48, 49]. Several groups have successfully grown CaF
2
crystals in sol-gel silica.
In fact, they even show that rare earth dopants will preferentially accumulate in the
CaF
2
instead of the silica, making the rare earth elements behave as if they were in
actual CaF
2
glass instead of a hybrid silica material [48, 49]. As detailed presently, I
successfully implemented similar procedures to grow CaF
2
crystals in sol-gel silica.
By fabricating silica toroids from these films, improved rare earth lasers as well as
frequency combs could potentially be developed in the future.
By modifying procedures in the literature, CaF
2
crystals can be successfully
grown in silica following a similar approach outlined in Table 5-2 [48]. First, the
desired volume of calcium acetate powder is weighed in a small glass vial. Since the
calcium acetate is hygroscopic, this is done in an argon glove box. Next, the calcium
acetate is combined with a small amount of water (~0.2-0.3g, enough to barely cover
the powder). Afterward, the trifluoroacetic acid is very carefully added to the
mixture (in a fumehood) and the calcium acetate is dissolved. Note that the
trifluoroacetic acid bottle is filled with argon and sealed tightly with parafilm after
use. Once the calcium acetate is dissolved, the mixture is set aside while the sol-gel
silica mixture is prepared in a separate container.
138
To prepare the sol-gel silica solution, tetraethoxysilane (TEOS), ethanol, and
deionized water are combined, stirring 5 minutes between additions. Once this
mixture is prepared, the acetic acid catalyst is added, and the entire mixture is
allowed to stir on a stir plate at room temperature for 15 minutes at 300rpm. Then,
the TEOS mixture is added to the calcium acetate bottle, stirred for 4 hours at
300rpm, and subsequently aged at room temperature. The amount of age time
necessary may vary, as aging longer causes the crystals to grow larger. However,
aging sol-gels for 2 weeks or longer causes the crystals to become quite large (50-
100+ microns) and may prevent the sol-gels from spin-coating uniformly onto the
wafers. Therefore, aging the sol-gels for 3 days to 1 week is recommended.
Chemicals TEOS Ethanol
Acetic
Acid H
2
O
Trifluoro
-acetic
acid
Calcium
Acetate
Vendor
Alfa
Aesar,
99.999%
EMD,
200 proof EMD
Clean
room
Alfa
Aesar
Alfa
Aesar
MW 208.33 46.07 60.05 18.02 114.02 176.18
Purity (mass) 100% 100% 100% 100% 99.99% 100%
Molar ratio 19.00 equal vol 3.00 90.00 3.0000 1.00
Desired
Chemical Mass 2.500 2.114 0.114 1.024 0.2160 0.111
Desired
Solution Mass
(g) 2.500 2.114 0.114 1.024 0.2161 0.111
Table 5-2: Recipe for CaF
2
sol-gel synthesis.

Once the aging step is complete, the sol-gels can be spin-coated onto silicon
wafers (usually at 7000 rpm for 30 seconds) and annealed in a tube furnace. The
samples were not dried on the hot plate prior to being placed in the tube furnace, as
rapid drying may cause the crystals to precipitate out too quickly. The suggested
anneal parameters from the literature are heating to 1000°C with a 75°C/hour ramp
rate and no soak time (not holding the temperature at 1000°C). After the anneal step,
139
additional layers of sol-gel can be spun onto the wafers to produce a thicker film.
Once at least two layers of sol-gel have been spun onto the wafers, silica toroids
could potentially be fabricated from the films.
While the initial annealed films showed significant cracking, crystals could
be observed (Figure 5-3), indicating successful formation of CaF
2
in the silica.
Smoother, more uniform films could be obtained by aging the sol-gels no longer than
1 week. After a week, the CaF
2
crystals become quite large and prevent the sol-gel
from coating the samples uniformly.

Figure 5-3: CaF
2
crystals grown in silica film.

These initial results appeared promising, and similar procedures could be
used to develop CaF
2
-doped silica toroids as well as frequency combs and greatly
improved lasers [39, 40, 48].
140
5.2.4 Adding Alumina to Improve Lasing

While the CaF
2
approach is interesting, during that same time, I found a
second approach using hybrid silica and alumina films that appeared especially
promising. One group reported that adding alumina to Nd
3+
-doped silica sol-gels
will shift the emission peak of neodymium from ~1080nm to ~1060nm – exactly
solving the red-shifting problem observed previously [33]. Therefore, adding
alumina could shift the emission of neodymium into the desired 1055-1070nm range.
In addition, adding alumina presented a second advantage. Since rare earth
ions have limited solubility in silica, they tend to form clusters which quench
emission and reduce lasing performance. Researchers have shown that adding metal
ions to silica can reduce rare earth clustering and significantly improve lasing
performance. For example, aluminum ions selectively coordinate around
neodymium ions in silica, which prevents clustering and enhances neodymium’s
lasing performance [9, 35, 38, 50]. The Nd
3+
ions coordinate very favorably with the
Al-O bonds in the silica matrix, which stabilizes the Nd
3+
ions and distributes them
uniformly in the silica. As a result, the fluorescence behavior of Nd
3+
in alumina-
doped silica has been observed to significantly improve [10, 19, 33, 35-37].
Considering these enhancements, I modified procedures in the literature to
synthesize Nd
3+
and alumina doped silica films, as outlined in the next section [19].
To study the effects of alumina, the silica films are also doped with 0 to 2 mol%
alumina. By fabricating toroids from the various doped sol-gel silica films, the
effects of alumina on lasing performance can be studied and characterized. Since the
141
alumina’s Al-O bonds create a more favorable environment for the Nd
3+
dopant and
reduce clustering, the lasing performance is significantly enhanced.
In the present work, we show how addition of 2 mol% alumina greatly
enhances Nd
3+
toroid microlasers. The alumina improves the lasing efficiency nearly
30-fold and enables ultra-low lasing thresholds of 530 nanowatts to be observed in
integrated silica devices at room temperature [19]. The alumina also shifts
neodymium’s emission into the heterodynable 1055-1070nm range, making these
significantly improved toroid microlasers useful in heterodyned sensing applications
as well [24].
5.3 Materials and Methods

5.3.1 Neodymium and Alumina-doped Sol-gel Synthesis

To fabricate the neodymium and alumina-doped silica, a modified sol-gel
method is used, as shown in Table 5-3 [19]. Tetraethoxysilane (TEOS, Alfa Aesar,
99.99%) is used as the liquid silica precursor, and solid aluminum isopropoxide
(AIP, Sigma Aldrich, 99.99%) is used as the alumina precursor. Neodymium is
added as neodymium nitrate salt. Since the neodymium and AIP are moisture
sensitive, they are handled and weighed separately in an argon glove box and kept in
separate, sealed glass bottles until use. To prepare the sol-gels, the TEOS and AIP
are combined in a small glass bottle with ethanol, water, and nitric acid (EMD,
68%), stirring 5 minutes between each addition. The molar ratio of
TEOS:ethanol:water:HNO
3
:Nd is set to (1:4:4:0.05:0.001). While the Nd
3+

concentration is fixed at 0.1mol%, the molar ratio of AIP:TEOS is set to 0:1,
0.005:1, 0.01:1, and 0.02:1 to produce silica films with varying alumina
142
concentrations. Once the nitric acid is added, the sol-gel mixture is stirred for 1 hour
at 70 °C to allow acid-catalyzed hydrolysis reactions to occur. In these acid
catalyzed hydrolysis reactions, the oxygen atoms on TEOS and AIP are attacked by
water, resulting in formation of Si-OH and Al-OH groups and the release of ethanol
and isopropanol.
Chemicals TEOS EtOH HNO
3
H
2
O
Nd(NO
3
)
3

*xH
2
O AlP
Vendor
Alfa
Aesar,
99.999%
EMD, 200
proof EMD Cleanroom
Alfa Aesar,
99.99%
Sigma
Aldrich
MW 208.33 46.07 63.00 18.02 330.25 204.24
Purity (mass) 100% 100% 68% 100% 99.99% 99.99%
Molar ratio 0.979 4.000 0.05 4.00 0.0010 0.0200
Mole fraction 0.11 0.442 0.01 0.44 0.0001 0.0022
Desired Chemical
Mass (g) 2.500 2.259 0.039 0.865 0.004 0.0490
Desired Solution
Mass (g) 2.500 2.259 0.057 0.865 0.0040 0.0490
Actual Mass (g)
Table 5-3: Sample recipe for sol-gel silica doped with 0.1 mol% Nd
3+
and 2 mol% alumina.

It is worth noting that before nitric acid is added, the sol-gel solutions appear
slightly cloudy (the solid aluminum isopropoxide has limited solubility in ethanol
and TEOS); however, after hydrolysis, the aluminum isopropoxide dissolves and the
solutions become transparent. If the solutions still appear cloudy after hydrolysis,
more acid should be added and the solutions can be stirred longer. Also, more acid
and ethanol solvent should be used if alumina concentrations greater than 2 mol%
are desired.
Meanwhile, the neodymium nitrate is dissolved in water in a 1:10 molar ratio.
Once the hydrolysis reactions are complete, the neodymium nitrate solution is
combined with the sol-gels and stirred for an additional 30 minutes at room
143
temperature. After stirring, the solutions are allowed to age for 60 hours at room
temperature. During this time, the Si-OH and Al-OH groups formed during
hydrolysis will begin to combine in condensation reactions, producing a silica and
alumina matrix. The neodymium ions will also coordinate preferentially with the Al-
O groups of alumina.
Once the aging process is complete, the sol-gels are spin-coated onto bare
silicon wafers using a spinner set to 7000 rpm for 30 seconds (Figure 5-4a) [19].
Immediately after spinning, the sol-gels are dried on a 75°C hot plate for 5 minutes
to evaporate residual solvent, leaving behind the colloidal silica sol (Figure 5-4b)
[51]. Finally, to complete the densification of the sol-gel into an amorphous silica-
alumina matrix, the sol-gels are annealed in ambient conditions using a
programmable tube furnace [19]. During annealing, the samples are heated to 900°C
for 2 hours, with a 1°C/min ramp rate. This process results in a ~350nm thick silica-
alumina-Nd film, as measured by ellipsometry (Figure 5-4c) [51]. The spin-coating
and annealing process is repeated once more to produce a ~700nm thick film (Figure
5-4d) [51]. While it is possible to produce even thicker films with additional spin
and anneal cycles, thicker films tended to form cracks and defects which could
interfere with the subsequent fabrication steps.
144

Figure 5-4: Synthesis of doped sol-gel silica films. The sol-gels are spin-coated onto bare silicon
wafers to produce a thin film (a). The samples are dried on a 75°C hot plate for 5 minutes to
evaporate the solvent (b), and annealed in a 900°C tube furnace to produce a silica layer (c). This
process is repeated once more to obtain two layers for a ~700nm thick film (d) [51].

5.3.2 Making Doped Sol-gel Silica Toroids

After fabricating the sol-gel silica films, optical toroid resonators can be
fabricated from the films using standard photolithography and etching procedures in
a Class 10/100 cleanroom [21, 34]. First, the sol-gel films are cleaned with acetone,
methanol, and isopropanol, and subsequently dried using a nitrogen gun and heating
on a 120°C hot plate for several minutes. The samples are allowed to cool, and
HMDS is applied to the silica surface by vapor deposition for 2 minutes. This
HMDS treatment enhances adhesion between the silica surface and photoresist.
Once the HMDS is applied, a layer of positive photoresist (Shipley S1813,
Microchem) is spin-coated onto the silica surface using a spinner set to 3000 rpm for
30 seconds. After spinning, the samples are baked on a 95°C hot plate for 2 minutes
to evaporate the photoresist solvent. Then, using a contact aligner, the samples are
exposed to UV light through a glass/chrome patterned mask with 80µm circular
arrays. After UV exposure, the UV-exposed photoresist is dissolved in MF-321
developer (Microchem) and rinsed thoroughly in deionized water. The samples are
145
inspected using an optical microscope to ensure the 80µm circular patterns of
photoresist are intact (Figure 5-5a). Afterward, the developed samples are heated on
a 110°C hot plate for 2 minutes. During this heating process, the photoresist
polymer is heated above its glass transition temperature, which melts the photoresist
and repairs small defects incurred during the developing.

Figure 5-5: Device fabrication. First, 80µm circles of photoresist are patterned on the doped sol-gel
silica using standard photolithography steps (a). Buffered oxide etchant (BOE) etches the uncovered
silica and the photoresist is removed, leaving silica pads (b). XeF
2
etching isotropically etches silicon,
forming elevated silica microdisks (c). Finally, the microdisks are reflowed with a CO
2
laser to
produce silica toroids (d) [51].

After the patterned wafers have cooled, they are immersed in buffered oxide
etchant (Improved Buffered Oxide Etch, Transene) containing HF, which
isotropically etches the silica not covered by photoresist. Proper safety and handling
procedures are followed, including using appropriate gloves and Teflon
tweezers/containers. Since the sol-gel samples are slightly porous, this etching
process takes only 3-5 minutes. The etching is complete when the samples appear
silver in color and become hydrophobic. Then, the samples are carefully removed
and rinsed in flowing deionized water. Finally, the photoresist is removed by rinsing
the samples thoroughly with acetone, methanol, and isopropanol, and drying them
using the nitrogen gun and 120°C hot plate as before. This process results in 80µm
146
diameter silica+alumina+Nd sol-gel pads patterned on top of a silicon wafer (Figure
5-5b).
These 80µm diameter silica sol-gel circles are subsequently etched using
XeF
2
which isotropically and selectively etches silicon, producing 80µm diameter
silica disks supported by a silicon pillar (Figure 5-5c). The samples are etched for
approximately 45-60 minutes until the silicon pillar has shrunk to ~20-25 µm in
diameter. Finally, to produce silica toroids, the silica microdisks are reflowed using
a focused laser beam from a 75W CO
2
laser (Synrad) at 20-25% intensity. Since
silica strongly absorbs the 10.6 µm CO
2
laser light while silicon does not, the silica
melts and forms a smooth toroidal structure (Figure 5-5d, Figure 5-6). These ~25-
30µm silica+alumina+Nd toroids are now ready to test, or are stored in a desiccator
until use [19].

Figure 5-6: Toroidal resonant cavity laser. a) PovRay rendering of an array of silica toroidal cavities
on a silicon substrate. Light from a tapered optical fiber waveguide is coupled into the first cavity. b)
Scanning electron microscope image of as-fabricated silica toroidal cavity [19].

5.3.3 Testing Setup

To characterize the lasing behavior and other properties of the silica toroid
microlasers, the testing setup shown in Figure 5-7 is used [19]. A tapered optical
fiber couples light from a 765-781nm tunable laser (New Focus) into the toroid
microlasers. The 765-781nm pump laser wavelength is controlled by a computer,
147
and the wavelength can be modulated using a computer-controlled function
generator (National Instruments). Then, the tapered fiber’s output is split, with 90%
sent to an optical spectrum analyzer (OSA, Agilent 86142B) to measure the emission
intensity at the lasing wavelengths. The remaining 10% is sent to a detector, which
is connected to a high speed digitizer and oscilloscope (National Instruments). This
setup enables the microlaser’s emitted light to be monitored on the OSA while
simultaneously observing the transmitted light through the taper on the oscilloscope.

Figure 5-7: Characterization set-up. a) Light from the tunable pump laser is coupled into and out of
the toroid with a tapered optical fiber. The output is split using a 90/10 splitter, where 90% of the
signal goes to the optical spectrum analyzer (OSA) and 10% goes to a photodetector (PD). The OSA
signal is seen on the computer using a GPIB (PCI GPIB) input. The photodetector signal is monitor
on an oscilloscope (PCI O-scope) which is integrated in the computer. The tunable pump laser
wavelength is controlled using the function generator (PCI Func Gen). b) The signal on the OSA
includes light from the pump laser which was not coupled into the device (non-coupled light) as well
as the emitted light (emission) from the toroid which is back-coupled into the taper. Since the same
tapered fiber is used for both input (765nm pump laser) and output (~900-940, 1050-1150nm) signals,
it is not possible to couple both the pump wavelength and the emission wavelengths at high
efficiencies [19].

Since the tapered optical fiber is responsible for effective coupling of light
both into and out of the toroid microlaser, it must be optimized for these
148
experiments. These tapered fibers are made on a custom-built taper-pulling setup
by heating silica optical fiber (F-SE, Newport) while it is stretched by computer-
controlled motorized stages (Optosigma). To optimize coupling of the 765-781nm
pump light, the optical fiber is heated and stretched until it forms a tapered region
approximately 780nm in diameter. To optimize placement of the toroid microlaser
near the tapered fiber, the toroid microlaser is precisely aligned with the tapered fiber
using a piezoelectric nanopositioning stage and top/side view cameras.
5.3.4 Q and Effective Refractive Index Measurements

Using the aforementioned testing setup, resonant peaks are found by
monitoring the transmission through the tapered optical fiber. The quality factor is
then determined by fitting a Lorentzian to peaks in the transmission spectra. For the
optical toroid microlasers in the present work [19], we assume that the quality factor
depends primarily on material absorption losses, given by the relation Q
mat
=2 n/  ,
where n and  are the effective refractive index and effective absorption or material
loss, which incorporate the refractive index and material loss of the silica [52]. Since
Nd
3+
absorbs some of the 765-781nm pump light in order to lase, higher material
absorption losses and lower Qs are inevitable. Changes in the effective refractive
index and absorption of the silica microlaser due to alumina can also affect the
quality factor and resulting lasing performance.
Experimentally, the effective refractive index can be measured by
determining the free spectral range, or FSR, of the toroid resonators [19]. Since
optical resonators confine certain resonant wavelengths by total internal reflection,
the resonant and lasing wavelengths which are confined in the silica toroid are highly
149
dependent on the toroid’s geometry, refractive index, and the wavelength of confined
light. Therefore, the distance in nanometers between the resonant and lasing
wavelengths is not random; it is a repeating pattern which can be predicted based on
the aforementioned resonator properties. The spacing (in nanometers) between
repeated resonant wavelength modes is called the free spectral range (FSR), and, to
first order, it can be approximated by the relation FSR=λ
2
/(πn
eff
D) where λ is the
wavelength, n
eff
is the effective refractive index, and D is the toroid’s major
diameter. Given a known lasing wavelength, an experimentally measured FSR at the
lasing wavelength, and a known toroid diameter, the effective refractive index of a
toroid microlaser can be easily calculated using the FSR relation.
The FSR can be measured experimentally either by finding the distance
between fundamental mode resonant wavelengths, or by measuring the distance
between adjacent TE or TM lasing emission peaks on the OSA. In the present work,
the FSR was determined based on the spacing between lasing peaks, as this data is
easily observed and saved in the process of characterizing the laser.
There are several advantages to calculating the effective refractive index in
this way. First, the FSR measurement allows the effective refractive index to be
determined in a finished, reflowed toroid. Since the reflow process melts the silica
sol-gel films, the silica toroid may have slightly different properties compared to the
silica sol-gel film from which it was made. Therefore, the effective refractive index
allows the true refractive index felt by the light in the final toroid to be measured.
Second, the effective refractive index is easily found at the exact lasing wavelengths
of interest. And finally, using COMSOL or finite element simulations, it is also
150
possible to extract the refractive index of the material alone, as demonstrated in
previous works.[53, 54]
5.3.5 Laser Characterization Procedures

The best lasing behavior will be achieved when the circulating power is very
high, so it is necessary to test at the resonant wavelength where the quality factor is
highest, corresponding to the highest possible circulating power. It is therefore
necessary to find the toroid microlaser’s fundamental mode – the wavelength at
which the toroid microlaser has the highest quality factor and confines light most
efficiently. This is accomplished as detailed in section 5.3.3.
If the toroid microlaser’s resonant peak has a high Q, lasing may be visible
on the optical spectrum analyzer. If lasing is observed, laser characterization
experiments can be performed. First, it is necessary to ensure the pump light is
efficiently coupled into and out of the microlaser. By optimizing the both the
thickness of the tapered fiber and the distance between the tapered fiber and silica
toroid microlaser, light can be coupled into and out of the resonator with very high
efficiency.
It is particularly important to optimize the coupling to ensure efficient
coupling not only of the pump light into the toroid microlaser, but also efficient
coupling of the neodymium’s emitted light (~900-940nm, 1050-1150nm) from the
resonator back into the tapered fiber [19]. If the emitted light does not successfully
couple into the tapered fiber, it will not be visible on the optical spectrum analyzer
which reads the taper output. However, because the pump and emission wavelengths
are several hundred nanometers apart, it is not possible to critically couple both the
151
780nm pump light into the device and critically couple the emitted light out of the
device. Therefore, some light is lost.
Once lasing is observed, characterization experiments are performed as
follows. First, to find the lasing threshold and efficiency, the lasing intensity is
measured as a function of power coupled into the toroid microlaser. To easily
correlate the transmission on the oscilloscope (in volts) with an input power (in
microwatts), a calibration curve is saved by measuring the transmission at multiple
known input powers.
While the neodymium-doped microlasers are capable of lasing at multiple
wavelengths, the lasing threshold and efficiency are characterized only at the
wavelengths which have the lowest lasing threshold. To do so, the emission
intensity at the given wavelengths is measured at different input powers by
attenuating the pump laser’s input power into the toroid microlaser [19]. At each
input power, both the OSA scan showing lasing intensity and the oscilloscope
showing the resonant peak are saved. Multiple sets of OSA and oscilloscope scans
are saved at each input power, and the input power is varied so that data points are
measured both above and below the lasing threshold. Other scans, such as
representative lasing peaks, full scans showing the pump and multiple lasing
wavelengths from 765-1200nm, and the oscilloscope transmission spectra of the
device at low input power, are also saved. Note that lasing near 1300nm is possible
but was not observed, probably due to the tapered fiber’s inability to efficiently
couple both the 765nm pump light and 1300nm emitted light.
152
To produce the lasing threshold curves, the coupled and emitted power are
calculated and plotted. The coupled power is determined from the oscilloscope data.
By looking at the depth of the transmission peak, the fraction of power coupled into
the toroid can be determined at each input power. Multiplying the amount of power
in the taper by the fraction of power coupled into the toroid gives the amount of
power coupled into the toroid. Similarly, the emitted power can be directly
determined by looking at the OSA data. The OSA’s emission (in dBm) is converted
to microwatts using the relation P(µW)=10
(P(dBm)/10)
*1000. By plotting the emitted
power versus coupled power, the lasing threshold curves are produced [19]. The
lasing threshold and efficiency can be determined from the x-intercept and slope,
respectively, of a line fitted to the increasing, linear region of the plot.
5.4 Experimental Results and Discussion

5.4.1 Effective Refractive Index

When testing the devices, we first measure the effective refractive index (n
eff
)
to verify the presence of alumina in the films. From the FSR values, the effective
refractive index is found to slightly increase with increasing alumina concentration
(Figure 5-8) [19]. This increase is anticipated, since alumina’s refractive index
(~1.77) is larger than silica’s (~1.44) [55].
153

Figure 5-8: The effective refractive index increases as the concentration of alumina in the sol-gel
increases. Inset: Top view image of an optical resonant cavity coupled to a tapered optical fiber, as
seen using the testing setup’s machine vision system [19].

This increase in n
eff
plays an interesting role in the threshold power. At first
glance, increasing n
eff
may appear to increase threshold power (Equations 5-1 to 5-
4). But n
eff
also affects the overlap factor ( ) and the quality factor. Therefore,
increasing n
eff
could increase or decrease threshold power.
5.4.2 Quality Factor

Another interesting effect of increasing the alumina concentration is that the
measured quality factors increase slightly (Figure 5-9) [19]. The quality factor
typically depends on material absorption losses as given by the equation
Q=2πn
eff
/λα
eff
where λ is the wavelength of light and α
eff
is the effective material
absorption coefficient. One concern about adding metal ions such as aluminum to
silica is that material absorption losses (α
eff
) could potentially increase and cause the
quality factor and lasing performance to worsen. For example, as seen in Chapter 4,
the absorption losses increased when titanium is added to silica. However, adding
alumina actually increases the quality factor, as shown in Figure 5-9.
154

Figure 5-9: The quality factor increases as the concentration of alumina increases. This increase is
directly related to the increase in refractive index and follows the trend in effective refractive index.
Inset: Representative transmission spectrum and Lorentzian fit at ~777nm pump wavelength for a
toroid microlaser containing 0.1mol% Nd
3+
and 2 mol% alumina (Q=1.17x10
6
) [19].

This increase happens for two main reasons. First, the refractive index
increases with increasing alumina concentration, making the quality factor increase
also. Second, when alumina is added to Nd
3+
-doped silica, the absorption coefficient
has been observed to decrease since Nd
3+
coordinates more favorably with alumina
in the silica. Also, the alumina concentration is relatively low – only up to 2 mol%,
so any increase in absorption loss from alumina could have negligible effects relative
to the changes in refractive index and light confinement. Therefore, high quality
factors of over 1 million can still be achieved even in the alumina-containing
devices. These effects of the alumina on the fundamental material properties further
indicate that the neodymium ions have successfully intercalated into the silica-
alumina matrix. This more uniform distribution can enable higher concentrations of
Nd
3+
to be used without causing clustering and quenching of emission.

155
5.4.3 Lasing Wavelength

The favorable coordination between alumina and the Nd
3+
also has several
significant effects on laser performance [19]. When pumped with the ~780nm laser,
all of the microlasers showed emission peaks in the 900-940nm range and 1050-
1150nm range (Figure 5-10a). These emitted wavelengths correspond to emission
from the
4
F
5/2
energy level to the
4
I
9/2
and
4
I
11/2
levels [10, 11, 15, 19, 38, 56].
The presence of alumina also noticeably affects the lasing wavelength
(Figure 5-10b). As alumina concentration increases, the lasing wavelength range
shifts from ~940nm and ~1080-1160nm in alumina-free microlasers to ~900nm and
~1050-1130nm in silica containing 2 mol% alumina. At greater alumina
concentrations, more neodymium ions are coordinated with Al-O bonds instead of
Si-O bonds, so there is less clustering, quenching and nonradiative decay compared
to plain silica [19]. The Al-O matrix of alumina also has a lower phonon energy
compared to silica, which further reduces quenching of Nd
3+
by its surroundings.
Therefore, the observed lasing at slightly higher energy wavelengths further indicates
that the neodymium ions become surrounded and enhanced by alumina.
156

Figure 5-10: Lasing from the 1mol% alumina-sensitized sol-gel device. a) Lasing spectrum from the
optical spectrum analyzer (OSA). The pump wavelength (780nm) and two cascades at approximately
920nm and 1050nm are visible. b) As the concentration of alumina is increased, the lasing wavelength
blue-shifts to shorter wavelengths. This result indicates that the Nd
3+
clustering is reduced by
alumina, and the Nd
3+
ions become more favorably coordinated in alumina’s Al-O matrix [19].

5.4.4 Lasing Threshold

The improved coordination between alumina and Nd
3+
significantly reduces
the lasing threshold. Samples with 2 mol% alumina experimentally achieved ultra-
low thresholds as small as 530 nanowatts at room temperature. A representative
lasing spectrum and the threshold curve are shown in Figure 5-11. Previous work
with similar Nd
3+
-doped cavities achieved 69µW thresholds, so this work represents
a 130-fold improvement in performance [31].

157

Figure 5-11: Lowest threshold lasing line from the 2mol% alumina-sensitized sol-gel device. The
curve shows a clear onset of lasing, with a threshold of 530nW. The inset is the lasing spectra just
after the onset of lasing [19].

Because of the lasing threshold’s significant dependence on Q, directly
plotting threshold versus the alumina concentration can lead to a misinterpretation of
the results. Also, even slight variations in device size and coupling can change Q,
distorting the results. To help account for these effects, we normalize the lasing
threshold by dividing it by the quality factor of each device. Doing so produces a
clear trend: the lasing threshold decreases linearly with increasing alumina
concentration over the 0-2 mol% alumina concentration range (Figure 5-12a).
Additionally, the lasing threshold/Q ratio decreases over 5-fold, further improving
the device performance [19].
158

Figure 5-12: Dependence of the threshold and the slope efficiency on the alumina concentration. a)
As the concentration of alumina is increased, the threshold/Q ratio decreases. b) As the concentration
of alumina is increased, the slope efficiency increases. This direct relationship indicates that the
alumina is acting as a sensitizer for Nd
3+
lasing [19].

Increasing the alumina concentration also greatly improves the slope
efficiency of the toroid microlasers. As the alumina concentration increases from 0
to 2 mol%, the slope efficiency increases up to 29-fold [19]. This efficiency increase
is linear with alumina concentration over this alumina concentration range (Figure 5-
12b). Simultaneously improving lasing threshold and slope efficiency is notable, as
typically improving one metric worsens the other.
It is also worth noting that the slope efficiency values in Figure 5-12b could
be further improved by optimizing the coupling conditions with the tapered optical
fiber, or using a different coupling approach. The tapered fiber in the present work
159
needed to simultaneously couple the ~780nm pump light into the toroid and the
~900-1150nm Nd
3+
lasing out of the toroid. As a result, the coupling over the entire
wavelength range is not ideal, and the slope efficiency values are lower [19]. Other
approaches which may improve the efficiency include using an add-drop
configuration. In this method, two waveguides are aligned with the toroid [57]. One
is optimized to couple the pump wavelength and the other is optimized to couple the
emission wavelength. However, this approach can be significantly more difficult to
achieve experimentally, especially with tapered optical fibers.
Finally, it is worth emphasizing that these improvements to laser performance
directly result from the presence of the alumina in the silica. The alumina reduces
Nd
3+
clustering, and allows the Nd
3+
ions also coordinate preferentially with the Al-
O matrix. As a result, there is less clustering and the neodymium ions and are more
uniformly distributed throughout the silica. Given the role that alumina plays in the
excitation pathway of Nd
3+
, a linear dependence on concentration for both
parameters can be theoretically anticipated from Equations 5-1 to 5-4 [19].
5.5 Conclusion

In conclusion, we have successfully fabricated toroid microlasers from
neodymium and alumina doped silica [19]. Adding alumina to the Nd
3+
-doped silica
reduces clustering of Nd
3+
dopant, causes emission at slightly shorter wavelengths,
gives up to a 29-fold increase in slope efficiency, and enables sub-microwatt lasing
thresholds to be achieved at room temperature in a lasing device which is integrated
on silicon. Adding alumina could be used to reduce clustering in lasers with other
160
rare earth dopants as well. Therefore, these efficient, ultra-low threshold lasers will
benefit applications in integrated optics and communications.

Chapter 5 References

1. R. Adar, M. R. Serbin, and V. Mizrahi, "Less-than-1DB per meter
propagation loss of silica wave-guides measured using a ring-resonator," Journal of
Lightwave Technology 12, 1369-1372 (1994).

2. C. Ge, M. Lu, S. George, T. A. Flood, C. Wagner, J. Zheng, A. Pokhriyal, J.
G. Eden, P. J. Hergenrother, and B. T. Cunningham, "External cavity laser
biosensor," Lab on a Chip 13, 1247-1256 (2013).

3. L. N. He, S. K. Ozdemir, and L. Yang, "Whispering gallery microcavity
lasers," Laser & Photonics Reviews 7, 60-82 (2013).

4. S. D. Jackson, and A. Lauto, "Diode-pumped fiber lasers: A new clinical
tool?," Laser Surg Med 30, 184-190 (2002).

5. M. R. Poulsen, P. I. Borel, J. Fage-Pedersen, J. Hubner, M. Kristensen, J. H.
Povlsen, K. Rottwitt, M. Svalgaard, and W. Svendsen, "Advances in silica-based
integrated optics," Opt. Eng. 42, 2821-2834 (2003).

6. V. Sandoghdar, F. Treussart, J. Hare, V. LefevreSeguin, J. M. Raimond, and
S. Haroche, "Very low threshold whispering-gallery-mode microsphere laser,"
Physical Review A 54, R1777-R1780 (1996).

7. O. S. Wolfbeis, "Fiber-optic chemical sensors and biosensors," Anal. Chem.
78, 3859-3873 (2006).

8. B. J. Ainslie, S. P. Craig, and S. T. Davey, "The absorption and fluorescence-
spectra of rare-earth ions in silica-based monomode fiber," Journal of Lightwave
Technology 6, 287-293 (1988).

9. A. J. Berry, and T. A. King, "Characterization of doped sol-gel derived silica
hosts for use in tunable glass lasers," J. Phys. D: Appl. Phys 22, 1419-1422 (1989).

10. W. Strek, E. Pawlik, P. Deren, A. Bednarkiewicz, J. Wojcik, V. E. Gaishun,
and G. I. Malashkevich, "Optical properties of Nd3+-doped silica fibers obtained by
sol-gel method," Journal of Alloys and Compounds 300, 459-463 (2000).

161
11. F. Q. Wu, D. Machewirth, E. Snitzer, and G. H. Sigel, "An efficient single-
mode Nd3+ fiber laser prepared by the sol-gel method," Journal of Materials
Research 9, 2703-2705 (1994).

12. S. Mehrabani, and A. M. Armani, "Blue upconversion laser based on
thulium-doped silica microcavity," Optics Letters 38, 4346-4349 (2013).

13. J. Bonar, J. A. Bebbington, J. S. Aitchison, G. D. Maxwell, and B. J. Ainslie,
"Aerosol doped Nd planar silica wave-guide laser," Electron. Lett. 31, 99-100
(1995).

14. T. Kitagawa, K. Hattori, Y. Hibino, and Y. Ohmori, "Neodymium-doped
silica-based planar wave-guide lasers," Journal of Lightwave Technology 12, 436-
442 (1994).

15. J. Yang, K. van Dalfsen, K. Worhoff, F. Ay, and M. Pollnau, "High-gain
Al2O3:Nd3+ channel waveguide amplifiers at 880 nm, 1060 nm, and 1330 nm,"
Appl. Phys. B 101, 119-127 (2010).

16. M. Cai, O. Painter, K. J. Vahala, and P. C. Sercel, "Fiber-coupled
microsphere laser," Optics Letters 25, 1430-1432 (2000).

17. M. Chistiakova, and A. M. Armani, "Cascaded Raman microlaser in air and
buffer," Optics Letters 37, 4068-4070 (2012).

18. H. S. Hsu, C. Cai, and A. M. Armani, "Ultra-low-threshold Er:Yb sol-gel
microlaser on silicon," Optics Express 17, 23265-23271 (2009).

19. A. J. Maker, and A. M. Armani, "Nanowatt threshold, alumina sensitized
neodymium laser integrated on silicon," Optics Express 21, 27238-27245 (2013).

20. L. Yang, D. K. Armani, and K. J. Vahala, "Fiber-coupled erbium microlasers
on a chip," Applied Physics Letters 83, 825-826 (2003).

21. D. K. Armani, T. J. Kippenberg, S. M. Spillane, and K. J. Vahala, "Ultra-
high-Q toroid microcavity on a chip," Nature 421, 925-928 (2003).

22. K. J. Vahala, "Optical microcavities," Nature 424, 839-846 (2003).

23. L. N. He, K. Ozdemir, J. G. Zhu, W. Kim, and L. Yang, "Detecting single
viruses and nanoparticles using whispering gallery microlasers," Nature
Nanotechnology 6, 428-432 (2011).

24. A. J. Maker, and A. M. Armani, "Heterodyned Toroidal Microlaser Sensor,"
Appl. Phys. Lett. 103, 123302 (2013).
162

25. G. M. Hale, and M. R. Querry, "Optical-Constants of Water in 200-nm to
200-mum Wavelength Region," Appl. Opt. 12, 555-563 (1973).

26. Y. Li, G. Vienne, X. Jiang, X. Pan, X. Liu, P. Gu, and L. Tong, "Modeling
rare-earth doped microfiber ring lasers," Optics Express 14, 7073-7086 (2006).

27. E. R. Giles, and E. Desurvire, "Modeling erbium-doped fiber amplifiers,"
Journal of Lightwave Technology 9, 271-283 (1991).

28. V. Lefevre-Seguin, "Whispering-gallery mode lasers with doped silica
microspheres," Opt. Mater. 11, 153-165 (1999).

29. W. von Klitzing, E. Jahier, R. Long, F. Lissillour, V. Lefevre-Seguin, J. Hare,
J. M. Raimond, and S. Haroche, "Very low threshold green lasing in microspheres by
up-conversion of IR photons," J. Opt. B-Quantum Semicl. Opt. 2, 204-206 (2000).

30. A. J. Maker, and A. M. Armani, "Fabrication of silica ultra high quality
factor microresonators," Journal of Visualized Experiment 65, e4164 (2012).

31. J. T. Lin, Y. X. Xu, J. X. Song, B. Zeng, F. He, H. L. Xu, K. Sugioka, W.
Fang, and Y. Cheng, "Low-threshold whispering-gallery-mode microlasers
fabricated in a Nd:glass substrate by three-dimensional femtosecond laser
micromachining," Optics Letters 38, 1458-1460 (2013).

32. A. J. Maker, B. A. Rose, and A. M. Armani, "Tailoring the behavior of
optical microcavities with high refractive index sol-gel coatings," Optics Letters 37,
2844-2846 (2012).

33. J. W. Zhai, B. Shen, X. Yao, and L. Y. Zhang, "Preparation and spectral
properties of Nd2O3-doped silica-based glasses prepared by the sol-gel process,"
Ceramics International 28, 737-740 (2002).

34. A. J. Maker, and A. M. Armani, "Fabrication of Silica Ultra High Quality
Factor Microresonators," in Journal of Visualized Experiments(2012).

35. C. Bartolacci, M. Laroche, T. Robin, B. Cadier, S. Girard, and H. Gilles,
"Effects of ions clustering in Nd3+/Al3+-codoped double-clad fiber laser operating
near 930 nm," Appl. Phys. B 98, 317-322 (2010).

36. T. Fujiyama, T. Yokoyama, M. Hori, and M. Sasaki, "Silica glass doped with
Nd and Al prepared by the sol-gel method - change in the state of aluminum in the
formation process," Journal of Non-Crystalline Solids 135, 198-203 (1991).

163
37. S. Mathur, M. Veith, H. Shen, S. Hufner, and M. H. Jilavi, "Structural and
optical properties of NdAlO3 nanocrystals embedded in an Al2O3 matrix," Chem.
Mat. 14, 568-582 (2002).

38. B. Wilhelm, V. Romano, and H. P. Weber, "Fluorescence lifetime
enhancement of Nd3+-doped sol-gel glasses by Al-codoping and CO2-laser
processing," J. Non-Cryst. Solids 328, 192-198 (2003).

39. I. S. Grudinin, and L. Maleki, "Ultralow-threshold Raman lasing with CaF
2
resonators," Optics Letters 32, 166-168 (2007).

40. I. S. Grudinin, N. Yu, and L. Maleki, "Generation of optical frequency combs
with a CaF2 resonator," Optics Letters 34, 878-880 (2009).

41. T. Gacoin, L. Malier, and J. P. Boilot, "New transparent chalcogenide
materials using a sol-gel process," Chemistry of Materials 9, 1502-& (1997).

42. A. Jitianu, M. Gartner, M. Zaharescu, D. Cristea, and E. Manea,
"Experiments for inorganic-organic hybrid sol-gel films for micro- and nano-
photonics," (Elsevier Science Bv2003), pp. 301-306.

43. H. K. Kim, S. J. Kang, S. K. Choi, Y. H. Min, and C. S. Yoon, "Highly
efficient organic/inorganic hybrid nonlinear optic materials via sol-gel process:
Synthesis, optical properties, and photobleaching for channel waveguides,"
Chemistry of Materials 11, 779-788 (1999).

44. M. Laczka, K. Cholewa-Kowalska, and M. Kogut, "Organic-inorganic hybrid
glasses of selective optical transmission," (Elsevier Science Bv2001), pp. 10-14.

45. B. Mashford, J. Baldauf, T. L. Nguyen, A. M. Funston, and P. Mulvaney,
"Synthesis of quantum dot doped chalcogenide glasses via sol-gel processing,"
Journal of Applied Physics 109 (2011).

46. D. L. Su, G. D. Qian, Z. Y. Wang, Z. L. Hong, and M. Q. Wang, "The
preparation of silica-based SiO2-ZrO2 films as waveguides by sol-gel process," Rare
Metal Materials and Engineering 33, 281-283 (2004).

47. C. M. Whang, C. S. Yeo, and Y. H. Kim, "Preparation and characterization of
sol-get derived SiO2-TiO2-PDMS composite films," Bulletin of the Korean
Chemical Society 22, 1366-1370 (2001).

48. L. H. Zhou, D. Q. Chen, W. Q. Luo, Y. S. Wang, Y. L. Yu, and F. Liu,
"Transparent glass ceramic containing Er3+: CaF2 nano-crystals prepared by sol-gel
method," Materials Letters 61, 3988-3990 (2007).

164
49. A. Biswas, G. S. Maciel, C. S. Friend, and P. N. Prasad, "Upconversion
properties of a transparent Er3+-Yb3+ co-doped LaF3-SiO2 glass-ceramics prepared
by sol-gel method," J. Non-Cryst. Solids 316, 393-397 (2003).

50. I. M. Thomas, S. A. Payne, and G. D. Wilke, "Optical-properties and laser
demonstration of Nd-doped sol-gel silica glasses," J. Non-Cryst. Solids 151, 183-194
(1992).

51. A. J. Maker, and A. M. Armani, "Heterodyning cavity-based microlasers to
improve sensing performance," in SPIE Photonics West, A. V. Kudryashov, A. H.
Paxton, V. S. Ilchenko, L. Aschke, and K. Washio, eds. (SPIE, San Francisco, CA,
2014).

52. M. L. Gorodetsky, A. A. Savchenkov, and V. S. Ilchenko, "Ultimate Q of
optical microsphere resonators," Optics Letters 21, 453-455 (1996).

53. H. S. Choi, S. Ismail, and A. M. Armani, "Studying polymer thin films with
hybrid optical microcavities," Optics Letters 36, 2152-2154 (2011).

54. B. A. Rose, A. J. Maker, and A. M. Armani, "Characterization of thermo-
optic coefficient and material loss of high refractive index silica sol-gel films in the
visible and near-IR," in Optical Materials Express(OSA, 2012), pp. 671-681.

55. E. D. Palik, Handbook of Optical Constants of Solids (Elsevier, 1985).

56. W. V. Moreshead, J. L. R. Nogues, and R. H. Krabill, "Preparation,
processing, and fluorescence characteristics of neodymium-doped silica glass
prepared by the sol-gel process," Journal of Non-Crystalline Solids 121, 267-272
(1990).

57. H. Rokhsari, and K. J. Vahala, "Ultralow loss, high Q, four port resonant
couplers for quantum optics and photonics," Physical Review Letters 92, 253905
(2004).

165

Chapter 6 Heterodyned Toroidal Microlaser Sensor

6.1 Introduction

Due to their high quality factors, whispering gallery mode optical resonators
have found many applications in sensing, and have demonstrated success in
detecting temperature, humidity, heavy water, and biological analytes [1-11]. In all
these applications, the sensitivity is limited by the smallest measurable wavelength
shift, which is given by the linewidth or full width at half maximum of the resonant
wavelength peak.
Since the linewidth is inversely related to quality factor, one approach to
decrease linewidth and increase sensitivity is to maximize the quality factor. Silica
toroid resonators can have quality factors of over 100 million and are especially
promising candidates for ultra-sensitive sensors. For example, a toroid with a quality
factor of 200 million at 633nm has a linewidth of about 3 femtometers (3x10
-15
m)
which is small enough to detect binding of single molecules [1].
Many applications would benefit greatly if we could not only detect but also
distinguish single molecules. For example, in the growing field of epigenetics,
researchers have found that the presence of methyl groups on DNA can significantly
affect gene expression, growth, cancer development, and mental disorders [12].
Similarly, modifying proteins with methyl, acetyl, and other groups has been found
to affect function and regulation. Developing sensors capable of detecting and
166
distinguishing modified and unmodified molecules in real-time could significantly
advance these research areas.
6.2 Background and Motivation

As mentioned before, the sensitivity of resonant wavelength-based
biodetection is ultimately limited by the linewidth (or full width at half maximum) of
the optical resonator’s resonant peak. By decreasing the toroid’s linewidth, we can
measure even smaller shifts in resonant wavelength [1, 13]. One approach to
decrease the linewidth is to increase the quality factor (Q). The challenge is that the
quality factor of silica toroids is limited to ~2x10
8
due to material absorption losses
[14]. It is therefore necessary to use other approaches to improve linewidth and
increase sensitivity.
Recently, another promising approach to improve linewidth was
demonstrated by using a toroid microlaser instead of a plain toroid. The linewidth of
the lasing emission peaks is even narrower than the linewidth of resonant wavelength
peaks. Therefore, by tracking a toroid laser’s emission wavelength, the linewidth
can be reduced and sensitivity further improved [8, 15]. However, tracking the
microlaser’s emission wavelength introduces a new challenge: It is not currently
possible to measure extremely small wavelength shifts using optical methods (such
as optical spectrum analyzers and wavelength meters), because these optical methods
have limited resolution (several picometers) and can only acquire a few
measurements per second, greatly limiting the speed and accuracy of detection.
Some groups have overcome this problem by attaching nanoparticles to the
toroid surface, which induce scattering and mode splitting in the toroid microlaser.
167
Mode splitting occurs when some of the toroid microlaser’s circulating light reflects
off a surface defect (such as a nanoparticle). This causes some of the circulating
light to propagate in the opposite direction. Thus, when mode splitting occurs, the
toroid microlaser will have light propagating in both clockwise and counterclockwise
orbits (hence, the toroid’s single circulating mode is split and becomes two modes).
Binding of additional analytes to the toroid will cause the wavelength difference
between the two circulating modes to change. By monitoring the change in mode
splitting, individual nanoparticles and viruses can be detected [4, 16]. In addition to
improving sensitivity with the narrow linewidth lasing peaks, this mode splitting
approach also reduces noise and allows lower Q devices to be used. At the same
time, mode splitting can only detect a limited number of particles and is very
sensitive to the location of the particles on the toroid surface.
Clearly, faster and more accurate wavelength analysis is needed in order for
microlaser-based sensors to achieve optimal performance and potentially
differentiate between single molecules. This chapter presents an alternative method
for sensing using an optical heterodyne, which overcomes the limitations of both the
resolution-limited optical methods and the mode-splitting methods [13]. In this
heterodyning approach, ~1064nm emission from the Nd
3+
microlaser (developed in
Chapter 5) is combined with a 1055-1070nm tunable reference laser (Figure 6-1).
When the microlaser’s and reference laser’s emission are close in wavelength, a low
frequency beat signal is produced which can be measured on an electrical spectrum
analyzer (ESA) with high frequency resolution (Δν~190Hz) and as low as 1ms
resolution. Additionally, the frequency of the resulting beat signal equals the
168
frequency difference between the microlaser’s emission wavelength and the
reference laser’s wavelength. Therefore, tracking the beat frequency allows changes
in lasing wavelength to be measured with high speed (~1-10ms) and high resolution
(sub-picometer), without relying on mode splitting and light scattering off
nanoparticles [17].

Figure 6-1: Schematic of an optical heterodyne. The beat signal’s frequency equals the frequency
difference between the microlaser and the reference.

In order to develop a heterodyned toroidal microlaser which is usable in
applications such as biodetection, two key criteria must be met. First, a microlaser
must be fabricated which emits light efficiently and within the range of an available
tunable reference laser. Second, the microlaser must also operate at wavelengths at
which the absorption of water is low [18]. This will enable the microlaser to
function efficiently in both air and aqueous environments. While many microlasers
have been developed which emit efficiently in the near-IR [19-21], water absorbs
very strongly at these wavelengths, reducing laser performance. Therefore, it was
necessary to use an entirely new toroid microlaser for the heterodyne.
After experimenting with numerous materials [22-24], a heterodynable
neodymium and alumina-doped toroid microlaser was developed, as described in
Chapter 5 [25]. Not only does this microlaser have ultra-low lasing thresholds and
169
good efficiency, it also operates at the 765nm pump and 1064nm emission
wavelengths where water’s absorption is low. Additionally, the microlaser’s
1064nm emission is easily heterodyned with a commercially available 1055-1070nm
reference laser [13, 25].
6.3 Experimental Approach

6.3.1 Fabrication of Microlasers Doped with Nd
3+
and Alumina

In order to perform heterodyned detection experiments, toroid microlasers are
fabricated from silica doped with neodymium and alumina [13] using the same
procedures described in Chapter 5 [25]. Briefly, custom Nd
3+
and alumina-doped
silica films are fabricated using a sol-gel method. The dopant concentration is set to
0.1 mol% Nd
3+
and 2 mol% alumina. As mentioned in Chapter 5, addition of 2
mol% alumina is crucial, as it enhances the neodymium’s lasing behavior and
enables lasing to occur near 1064nm. The neodymium will not lase within the
heterodynable 1055-1070nm range of the tunable laser without the alumina.
After aging for 60 hours, the Nd
3+
and alumina-doped sol-gel is spin-coated
onto bare silicon wafers (Figure 6-2a). After two spin-coating and annealing steps, a
~700nm thick film of sol-gel silica on silicon is produced. Then, silica toroids
approximately 40µm in diameter are fabricated from the films using the standard
photolithography, buffered oxide etching, XeF
2
etching, and CO
2
laser reflow steps
(Figure 6-2b, c).
170

Figure 6-2: Fabrication of heterodyned toroid microlasers. Nd
3+
and alumina-doped sol-gel is spin-
coated onto silicon wafers (a, rendering). Then, silica toroids are fabricated from the films (b,
rendering). c) The actual finished devices are ~40µm in diameter, as shown in the scanning electron
microscope image c) [13].

6.3.2 Heterodyned Testing Setup

To perform heterodyned detection, the testing setup shown schematically in
Figure 6-3 was built.

Figure 6-3: Schematic of the heterodyned testing setup.

The toroid laser is heterodyned as follows [13]. First, the fundamental mode
resonant peak in the toroid microlaser is found by scanning 765-781nm light from a
tunable laser (Newport Velocity Series). Once the fundamental resonant peak is
171
found, the tunable laser is set to the resonant wavelength to pump the neodymium
and cause emission of light near 1064nm. The high intensity 765nm light circulating
inside the toroid microlaser causes the Nd
3+
dopant to emit light near 1064nm. Both
the 765nm pump light and 1064nm emitted light are coupled from the toroid
microlaser back into the tapered optical fiber. The tapered fiber’s output containing
both the 765nm pump and 1064nm emitted light is sent to a 90:10 2x2 splitter. This
splitter combines the tapered fiber’s output with light from a 1055-1070nm tunable
reference laser (Newport Velocity Series). The combined signals are then split so
that 90% goes to an optical spectrum analyzer and 10% goes to a 12GHz detector.
The optical spectrum analyzer (OSA) is used to view and monitor the 765nm
pump light as well as the exact wavelength emitted from the toroid microlaser
(~1064nm). Then, the reference laser is tuned to closely match the microlaser’s
~1064nm emission (within ~40pm) in order to produce a beat frequency.
The remaining 10% of the combined 765nm pump, 1064nm emitted, and
1064nm reference laser light is sent to the 12GHz photoreceiver (Newport 1554-A)
and electrical spectrum analyzer (ESA, Agilent N9010A EXA with 13.6GHz range
and precision frequency reference). If the two 1064nm signals from the toroid laser
and reference laser are very close in wavelength (as determined using the OSA),
interference between them will produce a low frequency beat signal which can be
measured by the 12GHz photoreceiver. This beat signal is then sent from the
photoreceiver to the ESA for further observation and analysis. Note that a DC
blocker is attached to the photoreceiver’s output to prevent DC current from
damaging the ESA. Additionally, the cable connecting the photoreceiver to the ESA
172
is double-shielded to help prevent distortion of the beat frequency by noise and
interference from the surroundings.
Any change in wavelength from the toroid microlaser’s light will cause a
change in the beat frequency of the heterodyned signal on the ESA. Therefore, by
measuring the beat frequency on the ESA, extremely small changes in the microlaser
sensor’s light can be monitored. This ultra-sensitive heterodyned testing setup will
potentially allow detection, as well as differentiation, of similar biomolecules in
solution.
6.4 Comparative Temperature Sensing Experiments

To verify that this heterodyned setup is working and quantify the
improvements to sensor performance, a series of comparative temperature sensing
experiments are performed in air as a proof of concept. These temperature sensing
experiments are straightforward and done by heating the microlaser using a custom-
built thermocouple stage and the testing setups in Figure 6-4 [13, 26]. These testing
setups enable the microlaser’s resonant wavelength, lasing wavelength, and
heterodyned beat frequency to be tracked as the microlaser is heated. Since the
thermo-optic coefficient of silica is well-known (1.19x10
-5
/°C) [27], the expected
resonant wavelength and lasing wavelength shifts can be theoretically predicted and
compared to the experimental results.
173

Figure 6-4: Schematics of testing setups used to track a) resonant wavelength and microlaser
wavelength and b) heterodyned beat frequency. Representative transmission spectra, as measured
with these setups near c) 778nm and d) 1064nm [13].

After obtaining the shift versus temperature data, the sensitivity, detection
limit, signal to noise, and time resolution are calculated and compared for all three
cases. These four metrics are defined as follows [2, 8, 13]:
The sensitivity is the sensor’s output per unit change. In the present work,
the sensitivity is defined as the sensor’s response per °C temperature change, shown
in Equation 6-1:

C
Response Sensor
y Sensitivit

 (6-1)
This sensitivity can be converted from a wavelength change per °C to a refractive
index change per °C (using the relation n/n=      and compared to the thermo-
optic coefficient of silica for theoretical verification. The detection limit is closely
related to sensitivity, and defined as the smallest measurable temperature change
using a given sensing method.
174
The signal to noise is defined as the ratio of the signal’s strength to the noise
level. With P
signal
as the signal amplitude and P
background
as the background noise
amplitude, the signal to noise (SNR) is mathematically defined in Equation 6-2:

background
signal
P
P
SNR  (6-2)
The value of the signal to noise ratio is absolutely critical to sensing. Having a high
signal to noise ratio ensures that the sensor’s output is reliable and not obscured by
noise. Finally, the time resolution is the sensor’s sampling rate (in Hz).
Generally, it is desirable to use sensors with a low detection limit and high
values of sensitivity, signal to noise, and time resolution. By comparing these
parameters for the resonant wavelength-based, lasing wavelength-based, and beat
frequency tracking experiments, the enhancements provided by heterodyning can be
quantified.
6.4.1 Linewidth (Detection Limit) Measurements

As mentioned previously, the detection limit of optical toroid-based sensors
is defined by the linewidth of the microlaser’s resonant wavelength and lasing
wavelength peaks. For resonant wavelength-based detection, the linewidth can be
determined using the testing setup in Figure 6-4a. To perform this measurement, the
quality factor of the toroid microlaser was measured near the 765nm pump and
1064nm emission wavelengths [13]. During these measurements, the microlaser is
tested in the undercoupled regime and at low input powers to avoid nonlinear optical
effects such as thermal broadening which may distort the measurements.
175
The linewidth is given by the full width half maximum of the resonant
wavelength peak when the transmission spectrum is fitted to a Lorentzian. The
quality factor (Q) can also be found at each wavelength by dividing the resonant
wavelength by the linewidth. Since the linewidth of toroid microlasers is too narrow
to measure experimentally on the optical spectrum analyzer, it is instead estimated
using the Schawlow-Townes equation and the experimentally measured Q and
intensity values [13, 28]. Both the lasing spectra and transmission spectra are
recorded automatically using custom LabView programs, developed by Matt
Reddick and Kelvin Kuo.
6.4.2 Temperature Sensing Using Resonant and Lasing Wavelengths

In the temperature sensing experiments, the microlaser’s temperature
response is studied by placing the device on a custom-built temperature stage. The
stage’s temperature is directly monitored using a thermocouple with a feedback loop
to a controller. As the toroid microlaser is heated, its refractive index changes
slightly based on the material’s thermo-optic coefficient. This refractive index
change causes the toroid microlaser’s resonant wavelength, lasing wavelength, and
beat frequency to shift. The resonant wavelength and lasing wavelength are
monitored and tracked with time using the custom LabView programs. The
temperature versus time data is saved and synchronized with the wavelength versus
time data by recording video of the temperature controller’s display (30
frames/second) on an iPhone 5. By analyzing the video frame by frame, the
temperature versus time data is extracted and can be plotted with the resonant and
lasing wavelength data [13].
176
6.4.3 Heterodyned Temperature Sensing Experiments

In the heterodyned temperature sensing experiments, the testing setup in
Figure 6-4b is used. Here, the ~1064nm output from the toroid microlaser is
combined with a tunable reference laser. When the microlaser and reference are
sufficiently close in wavelength, a beat frequency is produced which is detected by a
12GHz photoreceiver and observed on the electrical spectrum analyzer. Since a
Labview program had not yet been developed for saving the beat frequency data, all
the initial frequency vs. time and temperature versus time data in the heterodyned
experiments were saved by recording video on an iPhone (Figure 6-5). Again, the
beat frequency and temperature versus time data could be obtained by analyzing the
video frame by frame [13]. In future measurements, a LabView program will be
used to save the beat frequency data, enabling significantly faster analysis.

Figure 6-5: Representative screenshot of iPhone video used to simultaneously save beat frequency
and temperature versus time.

177
It is also important to note that the function generator was used only in the
resonant wavelength versus temperature experiments. In the lasing wavelength
versus temperature and beat frequency versus temperature experiments, the function
generator caused the lasing wavelength to oscillate, significantly enough to
completely obscure the desired signal (Figure 6-6). Therefore, the microlaser was
pumped exactly at its resonant wavelength, without any frequency modulation from
the function generator. The resonant wavelength, lasing wavelength, and beat
frequency are also sensitive to the light from the testing setup cameras; therefore,
that light should be kept constant or turned off to reduce noise.

Figure 6-6: Heterodyned beat frequency data with and without the function generator. As can be
seen, the function generator must be kept off to avoid excessive noise which would otherwise obstruct
the desired measurement.

Based on the temperature versus time measurements, the sensitivity can be
determined by calculating the wavelength shift per °C for each measurement
technique [13]. The noise level is found by measuring the base signal for 1-3
178
minutes using the same testing conditions but without changing the temperature.
Then, the background noise can be fitted to a Gaussian and the noise level set to 3σ
from the center. All the calculated values are described in the next section and
summarized in Table 6-1.
6.5 Experimental Results and Discussion

6.5.1 Measured Linewidth Results

Using the linewidth measurement procedures described in the previous
section, the measured resonant wavelength linewidths are 2.6pm near 765nm and
1.0pm at 1064nm, corresponding to quality factors of 3.1x10
5
and 1.1x10
6
,
respectively (Fig. 6-4c, d). The Q is slightly lower at the 765nm pump wavelength
due to the absorption of light by the neodymium, which causes lasing but increases
material loss [13].

Figure 6-7: Representative lasing near 1067nm for Nd
3+
sample with 2 mol% alumina [13].

When the toroid microlasers are pumped with a 765-781nm tunable laser, the
alumina sensitizes the Nd
3+
ions and enhances their emission near 1064nm (Figure 6-
7) [13, 25]. Since the OSA is unable to measure the sub-picometer linewidth of the
179
microlaser’s ~1064nm lasing output, the linewidth is approximated using the
Schawlow-Townes equation. For a microlaser with a measured -50dBm emission
intensity at 1064nm, and Q=1.1x10
6
at 1064nm, the microlaser emission linewidth is
estimated to be 7.1x10
-6
nm [28].
6.5.2 Comparing Heterodyned vs. Non-Heterodyned Approaches

The sensor response versus temperature for each sensing approach is shown
in Figure 6-8 [13]. The data is summarized in Table 6-1 and further explained in the
following section.

Figure 6-8: Comparison of resonant wavelength (a), lasing wavelength (b), and heterodyned
detection approaches (c) [13].

The resonant wavelength versus temperature data, as measured using the
described procedures and testing setup, is plotted in Figure 6-8a [13]. The resonant
wavelength increases by only a few picometers with increasing temperature. The
measured sensitivity is 7.3 pm/°C. This corresponds to a refractive index change of
1.3x10
-5
/°C, in close agreement with the published values of 1.19x10
-5
/°C for silica.
The measured linewidth of the resonant wavelength peak is 2.6pm, corresponding to
a detection limit of 2.6pm. Given a noise level of 2.7nm and a signal of 7.3pm for a
1°C temperature increase, the signal to noise ratio is 2.7. A 100Hz scan rate is used,
enabling a time resolution of 100Hz to be achieved [13].
180
Given that the microlaser’s 1064nm emission has a much narrower linewidth
than the 2.6pm resonant wavelength, one would expect the microlaser’s detection
limit to be greatly improved. However, when the OSA is used to measure the lasing
wavelength shifts, the detection capabilities are actually much worse, as seen in
Figure 6-8b [13]. A lasing wavelength shift of 8.6 pm/°C was measured for the
lasing peak at 1064nm. This corresponds to a 1.2x10
-5
/°C refractive index change.
While this result is in excellent agreement with the thermo-optic coefficient of silica,
the desired signal is nearly obscured by noise. The measured noise level is 6.8pm,
which gives a signal to noise ratio of only 1.3. Meanwhile, the time resolution is
3.3Hz, significantly lower than the resonant wavelength-based approach. Therefore,
even though the lasing peak’s linewidth is narrower than the resonant wavelength,
the optical spectrum analyzer ultimately limits the sensor performance, negating any
benefit from the improved linewidth [13].
By heterodyning the microlaser’s narrow linewidth emission with a reference
laser, the microlaser’s wavelength can be analyzed with much greater speed and
accuracy, as seen in Figure 6-8c [13]. As a result, heterodyning enables sensing with
significantly improved performance. The measured 2.1GHz/°C sensitivity can be
converted to a wavelength shift using the relation Δν=c*Δλ/λ
2
. Therefore, the
2.1GHz/°C change corresponds to a wavelength shift of 7.9pm and a dn/dT value of
1.1x10
-5
/°C, which is in good agreement with previously published results. The ESA
used in the current work can measure changes as small as 190Hz, enabling
wavelength shifts as small as 7.1x10
-7
pm to be measured – well below the
microlaser’s theoretical linewidth (here, 7.1x10
-3
pm). The experimentally measured
181
noise level for the heterodyne beat signal is 0.030GHz, allowing wavelength shifts as
small as 0.11 pm to be detected.
Therefore, the heterodyned microlaser sensors demonstrate up to a 60-fold
improvement in resolution over the non-heterodyned microlaser and resonant
wavelength-based sensing methods. It is important to note that the resonant
wavelength and the microlaser-based detection methods nearly achieved the
theoretical detection limit value. However, the heterodyned method did not. The
experimental detection limit is higher than the microlaser’s linewidth due to noise.
By decreasing the noise in the system and optimizing the testing setup, it could be
possible to further improve the heterodyned microlaser’s detection limit by two
orders of magnitude [13].
Heterodyning with the ESA also eliminates some sources of noise, producing
a more stable signal with increased time resolution. The measured signal to noise
ratio for a 1°C temperature increase is 70, significantly higher than other approaches.
The scan rate is also faster - for these measurements, a 7.8ms (130Hz) scan rate was
used, although scan rates as fast as 1ms can be achieved.
For easy comparison, the key sensing parameters for all three sensing
methods are summarized in Table 6-1 [13]. Compared to the resonant cavity or
microlaser alone, the heterodyned microlaser can decrease detection limits by over
two orders of magnitude and enable nearly 60-fold improvements to both time
resolution and signal to noise. Therefore, heterodyning improves every sensing
metric. While the improvement to sensing properties can vary depending on the
182
testing conditions, lasers, and other equipment used, this heterodyning approach will
enable significantly improved sensing performance in many applications [13].
Parameter/Method Resonant Wavelength
Shift
Lasing Wavelength
Shift
Beat Frequency Shift
Detection Limit (signal
generated)
a,b)

0.40 °C (2.6pm)

0.77 °C (6.8 pm)

0.00080 °C (7.1x10
-3
pm)
Limiting Factor resonant peak
linewidth
OSA resolution microlaser linewidth
Demonstrated SNR for a 1
o
C change
- Measured noise level 0.0027nm 0.0068nm 0.030GHz (1.1x10
-4
nm)
- Measured signal 0.0073nm 0.0086nm 2.1GHz (0.0079nm)
- Measured SNR 2.7 1.3 70
Time Resolution 100Hz 3.3Hz 130Hz
a)
Assuming Q=3.1x10
5
at the 765nm pump wavelength and Q=1.1x10
6
at 1064nm lasing emission.
b)
The detection limit and experimental resolution limits are calculated based on a fit to the
experimental data and the equipment and devices used in the present work.

Table 6-1: Comparison of resonant wavelength, lasing wavelength, and heterodyned sensing
approaches [13].

6.5.3 Biodetection Experiments

In order to perform biodetection experiments, it is first necessary to have a
working microlaser in PBS buffer which emits a heterodynable signal. As
anticipated, the Al
2
O
3
+Nd
3+
microlasers also work in water and PBS buffer (Figure
6-9). Heterodynable lasing in the 1055-1070nm range can be measured when toroid
microlasers with a ~90-100µm major diameter are pumped near 765nm. A minimum
of -50dBm lasing intensity is needed to produce a strong enough beat signal on the
photodetector. Therefore, the microlasers generally need to be pumped with at least
5-10mW of power at 765nm to achieve strong enough lasing to heterodyne.
183

Figure 6-9: Lasing in PBS buffer is observed near 900-920nm and at 1055-1070nm, exactly within
the desired range. This sample was unfunctionalized (bare silica microlaser).

However, one concern with this heterodyned setup is that the measured
quality factors in bare (unfunctionalized) Nd
3+
toroid microlasers decrease to ~10
5
in
PBS buffer (Figure 6-10). The Qs are normally around 1 million in air, but the
increased absorption loss of water also contributes to the reduced Q [18]. In order to
achieve strong lasing, these low Q resonant peaks require very large amounts of
input power. Since the absorption loss of 1064nm light is even higher that that of
765nm, the Q near 1064nm is also expected to decrease to 10
5
or below. Lasing has
not yet been observed in a microlaser with surface chemistry attached, as the Q
factors drop even further.
184

Figure 6-10: Representative Q spectra of unfunctionalized microlaser in PBS buffer near 780nm.
The Q drops from ~10
6
to ~10
5
in PBS buffer due to water’s higher absorption losses.

Despite these issues, a heterodynable beat frequency could be tracked in PBS
buffer. However, the beat signal was very sensitive to coupling, and could only be
tracked in ambient conditions, not while flowing solutions. Upon zooming in, two
beat signals can be seen, due to mode splitting occurring in the toroid microlaser.

Figure 6-11: ESA screen image of splitting beat frequency observed when heterodyning microlaser in
PBS buffer.
185

No mode splitting was observed before when testing in air. But in the
temperature sensing experiments done in air, the temperature shift data was taken
with the ESA much more zoomed out (scanning across 3-4GHz) in order to see the
whole temperature shift. When scanning across a broader range (over ~0.5GHz), the
ESA cannot distinguish the two mode splitting peaks. When tracking a split beat
frequency, two parallel curves are seen (Figure 6-12). These parallel curves are
formed as the ESA’s peak marker jumps between the two beat peaks (the beat signal
was lost after ~20 seconds).

Figure 6-12: Two beat frequency peaks are observed in PBS buffer due to mode splitting. The
observed beat frequency in PBS buffer is very unstable even in ambient buffer shown here. This is
most probably due to thermal fluctuations and the very high input powers needed to achieve
heterodynable lasing.

If we look at the beat frequency vs. time for the marker which jumps between
both peaks, we can plot the counts and fit to a Gaussian, shown in Figure 6-13.
186
(Here, the first 15 seconds of the data in Figure 6-12 is studied, before the peaks
vanish).

Figure 6-13: Frequency distribution of beat signal’s two peaks due to mode splitting.

If we look at the left and right peaks separately, they have σ=0.0073 and 0.00875
GHz. This gives a noise level (3σ) of roughly 0.021-0.025 GHz (similar to the
0.030GHz noise in the APL paper). Also, based on the noise level, the peak
linewidths are 0.08-0.1pm.
For a toroid with a Q of 3-5x10
5
at 1064nm, emitting at -50dBm intensity, the
linewidth should be about 0.07-0.1pm as calculated by the Schawlow-Townes
equation [28]. Therefore, these peaks seem to be roughly equal to the linewidths of
the toroid microlaser in water, which ultimately limits the detection resolution.
187

6.5.4 Future Work

Unfortunately, due to linewidth and stability issues, the heterodyned
microlaser sensor did not work well in the aqueous PBS buffer environment. There
are several important limitations which must be addressed in order to use
heterodyned microlasers in biodetection applications. These include:
1. Thermal instability. At least 5-10mW of 765nm power must be coupled into
the microlaser to make it emit strongly enough for the heterodyne. However,
changing the temperature by 1 degree C causes a large beat frequency shift
(2.2GHz). Temperature fluctuations by a fraction of a degree can cause a significant
beat frequency change and add a lot of noise.
2. Noise. Due to thermal instability, optical forces, taper jitter, etc., even the
baseline signal in buffer is unstable. We need a stable baseline in order to do
detection experiments.
3. Mode splitting. Since the lasing peak gives a split signal, we need to zoom in
a lot to distinguish both peaks and get the highest resolution. In doing so, the
measurable beat frequency shift decreases to ~1GHz or ~5pm. Zooming out allows
larger shifts to be observed, but with 2-3x lower resolution (2-3x higher linewidth)
because the two peaks are seen by the ESA as a single peak.
4. Lower Qs due to increased absorption of water. In air, we can get Qs of ~10
6

at 765nm. In buffer, the Qs drop to ~10
5
. The lasing threshold also increases to 5-
10mW, which is close to the maximum possible output power through a tapered fiber
in water with the 765nm high power laser (~14mW).
188
5. Surface chemistry reduces Q further at 765nm. As seen in Heather Hunt’s
paper, the Q decreases when the NHS Ester surface chemistry is added. Any
decrease in Q will raise the lasing threshold and require even more input power to
get lasing. We are already operating near the maximum possible input power with a
bare toroid, so any further decrease in Q could prevent lasing.
6. Higher linewidth/less detection sensitivity. Using the Schawlow-Townes
equation, we can estimate the linewidth of the lasers in water. Best case scenario: if
Q doesn’t drop in water, a Q of 10
6
at 1064nm gives a laser linewidth of
approximately 0.0086pm or 8.6fm. More realistic scenario: Q drops to 100,000-
500,000 and gives linewidth of 0.034-0.08pm. Lower linewidths (3-5fm) were used
previously in single molecule detection.
7. So far, the best results are still achieved with plain silica toroids and the
resonant wavelength tracking method. A toroid with a Q of 2x10
8
at 633nm gives
10-100x lower linewidth of ~3fm [1]. These resonant wavelength methods also
don’t require very high input powers or the complex heterodyned setup.
8. The split ratio between the split microlaser peaks could be tracked instead,
although it has been done already by Prof. Lan Yang’s group and others. And, the
splitting method has its own set of limitations as mentioned in the beginning of this
chapter.
9. Maintaining coupling can be difficult. All the heterodyne work must be done
with the taper in contact with the toroid, and with the function generator off, or else
the beat signal will be too unstable to track. Even with the taper touching the toroid,
189
flowing liquid around the sample can disrupt the coupling and make the beat
unstable.
10. To make the heterodyne work, we would need to reduce the linewidth of the
laser. This may be very challenging or impossible given our existing setup and
limitations such as tunable laser availability.
One possibility is using Simin Mehrabani’s Tm-doped toroids [29]. They
could emit near 765nm, enabling a narrower lasing linewidth to be achieved since Qs
are higher in water at 765nm. However, we would need to pump with a lot of
1064nm power, and the absorption of water at 1064nm is 5-6x higher than 765nm so
the Qs are even lower. Our 1064nm laser also outputs only ~6-8mW so it would be
difficult to get enough power through the taper in water for lasing. So the thermal
instability and Q issues could still be present at the pump wavelength.
Another option is using praseodymium or samarium doped toroids, which
might emit near 633-637 nm, but they would likely require pumping in the blue.
Preferably, they could be pumped with a blue tunable laser, but could possibly be
pumped with a Tm co-dopant (although this would have similar heating/instability
issues as the Tm dopant alone, and the Q and input power at 1064nm may be too low
to get lasing). If the Pr or Sm laser gives heterodynable lasing near 633nm, it might
improve the linewidth and thermal instability issue since the Qs could be higher in
water and less input power may be needed. The difficulty here is that the blue/near
UV light might be harmful for some biomolecules, and a blue tunable laser is not
commercially available yet.
190
Another option is putting Nd or other rare earth dopants in a low phonon
energy glass. For example, if hybrid silica-CaF
2
toroids can be developed and doped
with rare earth metals, the rare earth dopants could preferentially enter the CaF
2

crystals and have a very low phonon energy environment [22]. This may enable
heterodynable, narrower linewidth emission or upconversion to be achieved and
further improve the heterodyned sensing approach.
6.6 Conclusion

In conclusion, we have successfully designed and fabricated an integrated
microlaser which can be heterodyned with a commercially available 1055-1070nm
tunable laser [13]. Since the microlaser operates at the 765nm and 1064nm
wavelengths where water’s absorption is low, this microlaser could be used in both
air and aqueous environments. Using comparative temperature sensing experiments,
we have shown that heterodyning significantly improves the microlaser sensor’s
sensitivity, signal to noise, and time resolution. These improvements are particularly
useful for sensing applications which require rapid, ultra-sensitive detection of trace
analytes. However, in order to achieve improved performance in biodetection
applications, it is necessary to significantly reduce the noise and improve the
microlaser’s linewidth in aqueous environments. If improved materials and lasers
can be developed, it may be possible to overcome these issues and achieve
significant enhancements to heterodyned detection.

191
Chapter 6 References

1. A. M. Armani, R. P. Kulkarni, S. E. Fraser, R. C. Flagan, and K. J. Vahala,
"Label-free, single-molecule detection with optical microcavities," Science 317, 783-
787 (2007).

2. D. Erickson, S. Mandal, A. Yang, and B. Cordovez, "Nanobiosensors:
optofluidic, electrical and mechanical approaches to biomolecular detection at the
nanoscale," in Microfluid Nanofluid(Springer, 2007), pp. 33-52.

3. H. K. Hunt, and A. M. Armani, "Label-free biological and chemical sensors,"
Nanoscale 2, 1544-1559 (2010).

4. J. Knittel, T. G. McRae, K. H. Lee, and W. P. Bowen, "Interferometric
detection of mode splitting for whispering gallery mode biosensors," Applied
Physics Letters 97 (2010).

5. S. Mehrabani, P. Kwong, M. Gupta, and A. M. Armani, "Hybrid microcavity
humidity sensor," Applied Physics Letters 102 (2013).

6. V. M. N. Passaro, C. de Tullio, B. Troia, M. La Notte, G. Giannoccaro, and
F. De Leonardis, "Recent Advances in Integrated Photonic Sensors," Sensors 12,
15558-15598 (2012).

7. F. Vollmer, and S. Arnold, "Whispering-gallery-mode biosensing: label-free
detection down to single molecules," Nature Methods 5, 591-596 (2008).

8. J. Yang, and L. J. Guo, "Optical sensors based on active microcavities," IEEE
J. Sel. Top. Quantum Electron. 12, 143-147 (2006).

9. T. Yoshie, L. Tang, and S.-Y. Su, "Optical Microcavity: Sensing down to
Single Molecules and Atoms," Sensors 11, 1972-1991 (2011).

10. X. M. Zhang, and A. M. Armani, "Silica microtoroid resonator sensor with
monolithically integrated waveguides," Optics Express 21, 23592-23603 (2013).

11. A. M. Armani, and K. J. Vahala, "Heavy water detection using ultra-high-Q
microcavities," Optics Letters 31, 1896-1898 (2006).

12. P. W. Laird, "The power and the promise of DNA methylation markers,"
Nature Reviews Cancer 3, 253-266 (2003).

13. A. J. Maker, and A. M. Armani, "Heterodyned Toroidal Microlaser Sensor,"
Appl. Phys. Lett. 103, 123302 (2013).

192
14. X. Zhang, H.-S. Choi, and A. M. Armani, "Ultimate quality factor of silica
microtoroid resonant cavities," Applied Physics Letters 96, 153304 (2010).

15. C. Ge, M. Lu, S. George, T. A. Flood, C. Wagner, J. Zheng, A. Pokhriyal, J.
G. Eden, P. J. Hergenrother, and B. T. Cunningham, "External cavity laser
biosensor," Lab on a Chip 13, 1247-1256 (2013).

16. L. N. He, K. Ozdemir, J. G. Zhu, W. Kim, and L. Yang, "Detecting single
viruses and nanoparticles using whispering gallery microlasers," Nature
Nanotechnology 6, 428-432 (2011).

17. D. Derickson, Fiber optic test and measurement (Prentice Hall, 1998).

18. G. M. Hale, and M. R. Querry, "Optical-Constants of Water in 200-nm to
200-mum Wavelength Region," Appl. Opt. 12, 555-563 (1973).

19. H. S. Hsu, C. Cai, and A. M. Armani, "Ultra-low-threshold Er:Yb sol-gel
microlaser on silicon," Optics Express 17, 23265-23271 (2009).

20. E. P. Ostby, L. Yang, and K. J. Vahalal, "Ultralow-threshold Yb3+: SiO2
glass laser fabricated by the solgel process," Optics Letters 32, 2650-2652 (2007).

21. L. Yang, D. K. Armani, and K. J. Vahala, "Fiber-coupled erbium microlasers
on a chip," Applied Physics Letters 83, 825-826 (2003).

22. L. H. Zhou, D. Q. Chen, W. Q. Luo, Y. S. Wang, Y. L. Yu, and F. Liu,
"Transparent glass ceramic containing Er3+: CaF2 nano-crystals prepared by sol-gel
method," Materials Letters 61, 3988-3990 (2007).

23. F. Q. Wu, D. Machewirth, E. Snitzer, and G. H. Sigel, "An efficient single-
mode Nd3+ fiber laser prepared by the sol-gel method," Journal of Materials
Research 9, 2703-2705 (1994).

24. J. W. Zhai, B. Shen, X. Yao, and L. Y. Zhang, "Preparation and spectral
properties of Nd2O3-doped silica-based glasses prepared by the sol-gel process,"
Ceramics International 28, 737-740 (2002).

25. A. J. Maker, and A. M. Armani, "Nanowatt threshold, alumina sensitized
neodymium laser integrated on silicon," Optics Express 21, 27238-27245 (2013).

26. H. S. Choi, S. Ismail, and A. M. Armani, "Studying polymer thin films with
hybrid optical microcavities," Optics Letters 36, 2152-2154 (2011).

193
27. M. Han, and A. Wang, "Temperature compensation of optical
microresonators using a surface layer with negative thermo-optic coefficient," Optics
Letters 32, 1800-1802 (2007).

28. A. L. Schawlow, and C. H. Townes, "Infrared and Optical Masers," Physical
Review 112, 1940-1949 (1958).

29. S. Mehrabani, and A. M. Armani, "Blue upconversion laser based on
thulium-doped silica microcavity," Optics Letters 38, 4346-4349 (2013).

194
Chapter 7 Studying Optical and Thermal Forces around Toroids

7.1 Introduction

The ultra-high quality factors of toroid and sphere resonators enable light to
circulate inside at extremely high intensities. Some of this high intensity circulating
light is absorbed by the resonator or its surroundings, causing heating, thermal
broadening of the resonant peak, and potentially significant changes in temperature
[1-4]. Temperature gradients from heating may also cause thermophoresis, or
temperature gradient-induced forces, to be present [5-8]. These thermal forces have
been observed to exert attractive and repulsive forces around optical resonators and
other devices. In addition to heating effects, the circulating light may be intense
enough for photophoresis, or light-induced forces, to occur, similar to those observed
in conventional optical traps [4, 9-13]. For example, light in a waveguide or
resonator can produce an attractive gradient force (from the evanescent field) as well
as a repulsive scattering force in the direction of the propagating light [14, 15].
Together, these optical and thermal forces have already been observed in
many photonic systems [10] including tapered optical fibers [16], waveguides [14],
and optical resonators [4, 17]. Because these optical and thermal forces are complex
and still not fully understood, experiments are often performed at low input powers
so that these effects (hopefully) become negligible.
Although these optical and thermal effects are at times undesirable and
unwelcome, studying these complex forces in toroids has many interesting
applications in kinetics and biological investigations. Therefore, the goal of this work
195
is to theoretically and experimentally study the various forces in and around toroid
resonators, especially at high circulating powers. Using COMSOL modeling,
temperature sensitive fluorescent dye studies, and particle tracking experiments, an
attempt is made to quantify the kinetic, optical, and heating effects around silica
toroids. Better understanding these forces and when/why they occur would enable
researchers to avoid or utilize them in many biological and optical applications [4,
10, 12, 17-26].
7.2 Background and Motivation

Initial work on this project focused purely on studying the effects of high
input power on biodetection experiments with the biotin-streptavidin system [27,
28]. In particular, we aimed to determine:
1) How changing the input power into a biotin-coated toroid will affect the
dissociation constant k
d
of streptavidin and the observed resonant
wavelength shift.
2) How closely the experimentally measured k
d
values match values in the
literature.
3) Whether temperature-dependent binding energetics and structural
changes can be observed between biotin and streptavidin by studying the
measured k
d
as a function of toroid circulating power and intensity.
If the three above items are feasible to measure, this would provide a solid proof of
concept and allow high input power toroid kinetics to be used to study additional,
more complex kinetic systems. Thus, using procedures developed in the Armani Lab
[29], biotin was covalently attached to the surface of silica toroid resonators (~100
196
µm diameter) using APTMS and NHS-biotin attachment procedures developed
previously [29]. After the biotinylated toroids are fabricated, solutions of
streptavidin in 1X PBS buffer are prepared, in concentrations ranging from 10fM to
10nM.
The testing procedures and equipment used were similar to those used for
other experiments in the Armani Lab [27, 28]. Biotinylated silica toroids are super-
glued onto a sample holder. Then, glass slides and a coverslip are placed around the
sample, forming a flow cell (Figure 7-1). A 765nm tunable laser is chosen
specifically because water can slightly absorb 765nm light, and also because we have
a tunable laser at 765nm which outputs light at high powers, up to ~20mW. While
the measured quality factors may be lower (~high 10
6
to low 10
7
) due to the
absorption of light by water, absorption of high intensity light by water could cause
heating effects. By controlling how much 765nm power is coupled into the samples,
we can control and tune how much the water surrounding the toroid is heated. The
heating may also affect the structure of the proteins attached to the toroid; therefore,
it may also be possible to tune the strength of the optical forces, thermal forces, and
binding kinetics.

Figure 7-1: Flow cell used in biodetection experiments.

197
Once the tapered fiber is prepared, the biotinylated toroid sample is super-
glued onto a sample holder and placed in the testing setup near the tapered optical
fiber. Using clean glass slides, a coverslip, and small amounts of super glue, a small
chamber or “flow cell” is built around the toroid and tapered optical fiber (Figure 7-
1). The flow cell is then filled with buffer and a syringe pump is used to purge the
cell with buffer for 10 minutes. This purging step helps remove dust or other
contaminants from the flow cell. Once purging is complete, the taper and toroid are
aligned and a resonant peak is found.
When a suitable resonant peak is found, injection of PBS buffer and
experimental solutions begins, using a syringe pump. Throughout each injection of
solutions into the toroid’s flow cell, the position of the resonant wavelength is
monitored. This data is taken using a custom LabView program which tracks the
position of the resonant peak. The program saves the x and y positions of the
minimum point on the oscilloscope. Since the minimum point on the oscilloscope
corresponds to the tip of the resonant wavelength peak, changes in the peak’s x-
position correspond to changes in resonant wavelength. Saving the y coordinate of
the resonant peak position also allows monitoring of the coupling and input power of
the resonant peak, as it is necessary to keep coupling and input power constant
throughout biodetection experiments.
198

Figure 7-2: Representative resonant wavelength shift as seen in flow experiments.

Once data collection is finished, plots of resonant wavelength versus time can
be made to study the kinetics of the experiments. A sample plot for injection is
shown in Figure 7-2. This plot corresponds to injection of a 100fM streptavidin
solution onto a toroid which was functionalized with biotin. As seen in the plot,
there is a distinct increase in resonant wavelength as the streptavidin binds the biotin,
and a noticeable decrease once the injection stops.
According to previous work in the Armani lab [27, 28], the magnitude of the
resonant wavelength shift corresponds to the amount of analyte which bound to the
toroid resonator, while the slopes of the increasing and decreasing regions can be
used to determine the binding and dissociation rate. Various kinetic models can then
be applied to the data to study the resulting kinetics.
For example, it has been shown that the model
0
ln( ( ) / )
d
t k t    
can be used to estimate the dissociation constant k
d
. In this model, λ(t) corresponds
to the linear fit of wavelength versus time over the dissociation region [28]. The
199
value of λ
0
is the dissociation curve’s initial wavelength value, and t is the time
(number of seconds) over which the dissociation occurs. Using the aforementioned
procedures, some preliminary data was measured for the binding of streptavidin to
toroids functionalized with biotin (Figure 7-3).

Figure 7-3: Preliminary data showed increases in resonant wavelength shift and k
d
as toroid
circulating power increases.

200
Based on this data, it appeared that heating around the toroid becomes
significant at high circulating powers and causes a noticeable increase in resonant
shift and dissociation constant. This may indicate that more binding and faster
dissociation is occurring since the temperature at the surface may be higher.
However, we found some issues with these initial results. First, the k
d
values do not
agree with literature values (they should be at least several orders of magnitude
smaller). Also, streptavidin and biotin have an extremely high affinity for each other
and they should be extremely difficult to break apart once they bind to each other. It
is therefore unlikely that simply stopping flow or purging the flow cell with buffer
would cause streptavidin and biotin to dissociate as shown previously [27, 28].
Therefore, the measured changes in wavelength shift and k
d
may not accurately
reflect specific attachment and dissociation of streptavidin to biotin functionalized
toroids. Some other effect is probably occurring, such as a flow-induced temperature
change, or physical (non-specific) binding and dissociation to the toroid.
To better understand what is happening during these biodetection
experiments, it is necessary to take a closer look at what forces are present.
Depending on the testing conditions, several forces can impact molecular
interactions with the toroid surface. These forces include photophoresis (light-
induced motion), thermophoresis (temperature gradient-induced motion), Brownian
motion (concentration gradient-induced motion), and convection (mass and heat
transport due to flow). Quantifying and understanding these various forces will help
improve biodetection experiments and may be useful for other applications as well.
201
Therefore, a series of theoretical modeling and experiments were used to study these
forces.
7.3 Heating Theory and COMSOL Modeling

To theoretically model the effects of thermal, optical, mass transport, and
flow forces on toroids, COMSOL Multiphysics simulations (ver. 4.2) are
implemented. Current COMSOL models developed by Mark Oxborrow and Imran
Cheema can be used to model the light and intensity distribution in axisymmetric
resonators such as toroids. However, no additional thermal, optical, flow, or
transport effects are considered in these models [30, 31]. Therefore, these previously
developed models are modified to study the additional thermal, optical, mass
transport, and flow effects which may be present during toroid biodetection
experiments.
As optical, mass transport, and flow effects all have some temperature-
dependence, initial modeling focused solely on determining the light-induced heating
in toroid resonators. Once the temperature model is developed, the other forces can
be more easily investigated. First, heating effects were added to existing 2D models
using the transient heat transfer module in COMSOL. The key equations and
assumptions for the 2D heating models are described presently.
The COMSOL heat transfer module solves the fundamental heat transfer
equation (Equation 7-1) in terms of temperature T [32-34]:
Q p
T
p
T
p T
T
t
T
C
p
p







  




     






  


) ( : ) ( ) ( u S q u

  (7-1)
Where the terms are described as follows:
202
 is density (in kg/m
3
);
p
C is the specific heat capacity at constant pressure (in J/(kg*K));
T is the absolute temperature (in K);
u is the velocity vector (in m/s);
q is the heat flux due to conduction (in W/m
2
), given by Fourier’s law of
conduction as:
i
i
x
T
k


  q with the thermal conductivity tensor











zz zy zx
yz yy yx
xz xy xx
k k k
k k k
k k k
k in W/(m*K);
p is the pressure (in Pa);
 is the viscous stress tensor (in Pa);
S is the strain-rate tensor (in 1/s), given by ) ) ( (
2
1
T
S u u     ;
and
Q represents heat sources (other than viscous heating forces). Units: W/m
3

Conservation of mass is assumed, so the continuity relation 0 ) (    


v 

t
must
also be satisfied.
To simplify the heat equation, several assumptions are made [33]. First, the
viscous heating and pressure work term






  




 p
T
p
T
p T
p
) ( : u S

 is assumed
negligible (since any flow rates considered would be slow).
203
Substituting Fourier’s law of conduction for the q term gives Equation 7-2:
Q T k T C
t
T
C
p p
       


) ( u   (7-2)
For the initial modeling, it was assumed that there is no flow in the system. This
assumption further simplifies Equation 7-2 to Equation 7-3, the case of purely
conductive heat transfer:
Q T k
t
T
C
p
     


) (  (7-3)
Equation 7-3 is solved in all the heat transfer models developed thus far [32, 33]. In
future work, convective heat transfer may be considered as well using Equation 7-2
instead.
Having determined the heat transfer equation for the system, it is now
necessary to define the boundary conditions, constants (  ,
p
C , and k ), and the
source term Q. For the boundary conditions and constants, the following assumptions
are made:
- Continuity is assumed at all internal boundaries (such as the interface
between the toroid and surroundings).
- Far from the optical toroid, on the outer boundaries of the model, a
constant temperature of 293.15K is assumed.
- Initially (at time t=0), the toroid and surroundings are at room
temperature, so the initial temperature T = 293.15K.
- The constants are assumed temperature-independent. The initial values at
room temperature are used, taken either from the literature or in
COMSOL’s material library. (This is done to simplify the initial models.
204
Depending on the theoretical results and their agreement with
experiments, temperature dependence may be added in the future.)
The source term Q is more difficult to define as it is related to many different
parameters. When defining Q, the following properties of the toroid and laser
system must be considered:
- The magnitude of Q depends not only on intensity distribution, but also
the circulating power of the toroid. Therefore, the heating effect must be
related to the toroid’s circulating power, the laser’s input power and
wavelength, and the toroid’s quality factor, mode volume, refractive
index, and diameter.
- The laser’s light is not uniformly distributed. An intensity distribution
I(x,y) must be defined which accurately describes light intensity in and
around the toroid.
- Heating does not occur all the time, but only when the laser is scanning
across the toroid’s resonant wavelength. Therefore, the laser-induced
heating depends on time and the scan rate and scan speed of the laser (as
controlled by the frequency and amplitude of the function generator).
- Not all materials are heated equally by the laser. The model must account
for differences in heating intensity in materials with different absorption
coefficients.
Having considered these important parameters, the source term Q is defined in W/m
3

by Equation 7-4:
205
) ( * ) , ( * *
0
t f y x I P Q   (7-4)
Each term in Equation 7-4 is described presently. First,
0
P is the circulating
intensity, in W/m
2
, found by dividing the toroid’s circulating power by the mode
area:

D
V
P
R n
Q
P
m
in
eff
*
2
0


 (7-5)
In Equation 7-5, λ is the resonant wavelength, n
eff
is the effective refractive index of
the toroid, typically assumed to be close to silica’s value of 1.44, R is the radius of
the toroid, D is the toroid’s diameter, Q is the toroid’s quality factor, P
in
is the
amount of power coupled into the toroid, and V
m
is the mode volume.
Next,  in Equation 7-4 is the absorption coefficient of the material (in m
-1
,
taken from the literature). Depending on the location, this absorption coefficient
could be that of the silica toroid or that of the surroundings. All absorbed light is
assumed to be converted to heat, while the rest is assumed to be transmitted without
causing heating.
The ) , ( y x I term in Equation 7-4 describes the intensity profile of the light in
the toroid. The intensity profile is normalized so that 1 ) , ( 0   y x I , and is
calculated using Oxborrow’s models [31]. The fundamental mode solution closest to
the wavelength of study is used, with the maximum intensity value normalized to 1.
206
Finally, ) (t f is a periodic, time-dependent rectangular or Lorentzian-shaped
pulse, as shown in Figure 7-4. These pulses mimic the scanning of the function
generator across the resonant wavelength.

Figure 7-4: Time-dependent rectangular pulse used in COMSOL modeling.

The value of f(t) is set to either 0, 1, or a value in between, depending on whether the
laser is scanning on the toroid’s resonance at time t. The width (duration) of the
pulse is set equal to the linewidth of the toroid; which estimates the amount of time
the laser is scanning through the resonant peak and heating the toroid (Figure 7-4).
The pulse linewidth, or duration, of the rectangular pulse is calculated based
on the toroid’s linewidth (λ/Q) which approximates the time the laser is scanning on
resonance. Since the linewidth is in length units, it is converted to time units based
on the scan rate using the relation pulse duration (in seconds) = [Linewidth (in
nm)]/[scan rate (in nm/s)]. The scan rate (nm/s) is found using scan rate = [laser
frequency modulation (nm/V)]*[function generator amplitude (V)]*[198 s
-1
] (For a
1V function generator triangle wave at 100Hz, the slope is about 198 V/s. At
different function generator frequencies, this number may differ).
For example, for a peak with Q=10
4
tested with the 1550nm laser (frequency
modulation = 0.061nm/s), with the function generator set to 1V amplitude, the scan
rate and linewidth are:
207
scan rate (nm/s) = (0.061nm/s)*1V*198s
-1
= 11.96 nm/s
pulse duration (in seconds) = (1550nm)/(11.96 nm/s) = 0.003241s
Using the linewidth, a rectangular pulse or Lorentzian can be defined in
COMSOL, with a width equal to the calculated pulse duration, as shown below in
Figure 7-5.

Figure 7-5: Sample rectangular pulse used in COMSOL modeling.

Once the pulse’s duration is determined in seconds, it is necessary to find the
period of the pulses. This is calculated based on the frequency of the function
generator triangle wave: Pulse period = 1/(2*function generator frequency). The
factor of 2 is added because the triangle wave scans the resonance twice in each
cycle – once in the forward direction and again in the reverse direction. Following
the aforementioned procedures, the periodic function f(t) can be defined as an
analytic function an1(t) in COMSOL, as shown in Figure 7-6. Note that these
208
procedures can also be used to define the linewidth of a Lorentzian pulse instead of a
square wave pulse.

Figure 7-6: Representative rectangular pulsed wave used in COMSOL modeling.

Now that all the parameters and terms of the heat equation have been defined,
the heat equation can be solved for temperature. Representative results are described
in the next section. It is also worth noting that while these modeling procedures have
been developed for the case of optical toroid resonators, these assumptions and
equations could also be extended to other resonators, such as microsphere resonators.
Also, additional simulations can be run to determine the effects of other forces,
including thermophoresis and photophoresis.
As mentioned previously, two additional forces which can be present in
toroid sensors are thermophoresis and photophoresis. Thermophoresis refers to
motion of particles caused by a temperature gradient, while photophoresis is motion
caused by light. The steady-state speed of particles moved by thermophoresis is the
thermophoretic velocity, while the photophoretic velocity is the speed at which
209
photophoresis moves particles. These thermophoretic and photophoretic velocities
can be calculated using the temperature and electric field intensity profiles for the
toroid in 3D, as well as some additional equations. In this preliminary model, the
thermophoretic and photophoretic velocities were calculated for polystyrene beads,
following equations and data for polystyrene beads in the literature [5, 7, 35-38]:
In the present work, the thermophoretic velocity is given by Equation 7-6:
T D
T T
    (7-6)
Here, the temperature gradient T  is taken from COMSOL, and D
T
is the
thermophoretic coefficient [5, 36].
The photophoretic velocity can be calculated using Equation 7-7:

  

2
3
Pr
c
Q n
P
 (7-7)
where n is the refractive index of the medium (water = 1.33), P is the beam power, r
is the particle radius, Q is the photophoretic efficiency [38], c is the speed of light, ω
is the beam radius (taken from COMSOL mode profile), and η is the viscosity of the
medium (here, assumed to be between 2-8.6x10
-4
kg/m*s, depending on the
temperature of the water [39]). These thermophoresis and photophoresis equations
can also be investigated and solved using the COMSOL results.
7.4 Experimental Verification of COMSOL

Two approaches were used to verify the COMSOL models. To verify the
heating effects present, the temperature on the toroid surface could be measured
using temperature sensitive fluorescent dyes [40]. Then, to study the thermophoretic
210
and photophoretic forces, polystyrene beads could be tracked as they interact with
the toroid sensors [5, 10, 15, 41].
7.4.1 Temperature-Sensitive Fluorescent Dyes

While theoretically modeling the heating effects in toroids is fairly
straightforward, measuring the temperature at the surface of the toroid can be very
difficult. One approach is to use a temperature dependent fluorescent dye. Many
temperature dependent dyes exist, yet most have temperature sensitivity over a
limited range (few to tens of degrees). According to the COMSOL simulations, it is
possible that significant heating from 30°C to over 100°C may occur. Therefore, to
verify the COMSOL models, it is necessary to use a dye with temperature sensitivity
from room temperature to near boiling, and one which can also be covalently
attached to silica toroids. Rhodamine B is a fluorescent dye which fits these criteria
(Figure 7-7) [40]. It is known from the literature that rhodamine B’s fluorescence
intensity decreases with increasing temperature, from ~20°C to ~100°C. In addition
to being used in temperature sensing applications, rhodamine B is commonly used in
biological applications and has well-studied chemistry. Therefore, by attaching
rhodamine B to the surface of toroids and monitoring its fluorescence, the
temperature at the toroid surface can be quantitatively determined.
211

Figure 7-7: Structure of rhodamine B. Its carboxyl group is useful for attachment chemistry.

7.4.2 Attaching Rhodamine B to Silica Toroids

Having selected the rhodamine B dye, the next step is to attach it to toroids.
Several attachment options were considered: 1) spin coating rhodamine B directly
onto toroids; 2) purchasing a rhodamine B-strepavidin conjugate and attaching it to
biotinylated toroids; 3) purchasing rhodamine B isothiocyanate and attaching it to
aminated toroids; and 4) attaching rhodamine B to aminated toroids using EDC
chemistry.
The first approach, spin coating, did not work well for several reasons. First,
the rhodamine B covers the entire sample, not just the toroid, so the fluorescence on
the toroid alone cannot be easily monitored. Also, rhodamine B does not excite or
emit very strongly in air (it does so much more strongly in water and alcohols).
Finally, since the rhodamine B is physically attached, it would dissolve if the toroid
sample is placed in water. Thus, covalent attachment was considered instead.
One approach to attach rhodamine B to toroids is to use the existing surface
chemistry procedures developed in the Armani Lab to attach a rhodamine B-
streptavidin to a biotinylated toroid or a rhodamine B-isothiocyanate to a
212
hydroxylated toroid. Since both the rhodamine B streptavidin and rhodamine B
isothiocyanate had been discontinued indefinitely and could not be obtained from
companies, EDC chemistry was used. While EDC chemistry is slightly more
complex, the reactants are less expensive and easily obtained through VWR. The
EDC methods are explained in detail in the following sections.
EDC (1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride) is a
popular crosslinker commonly used to attach carboxyl and amine groups. It can
therefore covalently attach rhodamine B’s carboxyl group to amine groups on
aminated toroids [42]. These bonds form in a two step process (Figure 7-8).

Figure 7-8: Schematic of the two EDC reactions which covalently bond amine and carboxyl groups.
213

First, EDC reacts with the carboxyl group to form an activated ester
intermediate. Then, in a second reaction, this ester intermediate reacts with
nucleophilic amine groups, forming a covalent bond between the carboxyl and amine
and releasing an isourea by-product. The isourea by-product is water soluble and
can be removed by rinsing, gel filtration, or dialysis. Some side reactions can also
occur – EDC can polymerize proteins which contain both amine and carboxyl
groups, and can also form stable complexes with sulfhydryl groups, tyrosine, and
histidine.
The following protocols used to attach rhodamine B were developed based on
sample protocols in the book Bioconjugate Techniques [42] and shown in Figure 7-9:
1. Prepare aminated toroid samples using the standard oxygen plasma (120W,
30sccm, 5min), and APTMS vapor deposition treatment (15 minutes in vacuum
dessicator) previously developed in the Armani lab [29].
2. Meanwhile, pH 6 PBS buffer is prepared by diluting 10X PBS stock to 1X and
adding dilute HCl until pH 6 is reached (as verified on a pH meter).
3. The EDC reactions are performed by combining the post-APTMS aminated
toroids with two solutions: ~0.05M EDC in pH 6 PBS, and ~5mM rhodamine B in
pH 6 PBS. These solutions are prepared as follows.
a. For the rhodamine B solutions: Solid rhodamine B is weighed on a mass
balance in a dimly lit environment, and placed in a foil-covered glass bottle
to protect from light. To the foil-covered bottle, enough pH 6 PBS is added
214
to make the 5mM rhodamine B solution, and the bottle is gently swished to
completely dissolve the rhodamine B.
b. For the EDC solutions: EDC is removed from the -20°C freezer and
allowed to reach room temperature. Since EDC is water-sensitive, all
handling is done inside an argon glovebox. In the argon environment, EDC
is weighed and placed in a separate glass bottle. Upon removal from the
glovebox, the bottle is kept tightly closed. Afterward, enough pH 6 PBS
buffer is added to produce the 0.05M EDC concentration, and the bottle is
stirred until the EDC completely dissolves.
4. Once the solutions are prepared, label some 1.5ml Eppendorf tubes and protect
them from light (cover with foil and/or a box).
5. Into each Eppendorf tube, pipet 125 microliters of the EDC solution and 125
microliters of the Rhodamine B solution.
6. Then, place an aminated toroid sample inside each of the tubes.
7. Make sure the tubes are properly covered to protect from light, and allow them to
react for at least 2 hours on a gently rocking tilt tray.
8. Finally, when the reaction is complete, the samples are placed in Eppendorf tubes
filled with deionized water and put back on the tilt tray to rinse for 5 minutes.
Repeat the rinsing step two more times with fresh water tubes each time, so that the
samples have been rinsed for at least 15 minutes total. Sufficient rinsing is needed to
ensure removal of the water soluble isourea by-product which forms during the EDC
reaction.
215
9. Once the rinsing steps are complete, the samples are allowed to air dry and are
stored in the dark.

Figure 7-9: EDC attaches Rhodamine B’s carboxyl group to amine groups on toroids.

While experimenting with the surface chemistry, some other observations were
noted:
- Initially, higher concentrations (0.1M EDC, 0.01M rhodamine B) were tried
as suggested in Bioconjugate Techniques, but the EDC-rhodamine B
intermediate formed a precipitate. To avoid this, more dilute EDC and dye
concentrations are used, as recommended.
- It is important to note that EDC is water-sensitive; therefore the solutions
should be prepared fresh each time.
216
- The EDC should be stored in the -20°C freezer and protected from
moisture.
- The 1.5 to 2ml Eppendorf sample tubes work the best for this procedure.
The non-Eppendorf tubes (without the cone-shaped bottom) are not good as
the samples frequently get stuck in the bottom and then are damaged when
removed.
- The overall EDC reactions are most efficient between pH4.5-7.5. So far,
pH 6 appears to produce the strongest attachment (based on fluorescence
intensity); however, by varying the pH, reaction time, and reactant
concentrations, the attachment results could potentially be improved.

To verify attachment of rhodamine B, fluorescent microscopy is used.
Rhodamine B is known [40] to excite over the 500-570nm wavelength range, with a
peak near 550nm. Its fluorescent emission peak is around 580-630nm depending on
the environment. Using the Nikon fluorescence microscope and Texas Red filters
(which cover rhodamine B’s emission/excitation wavelengths), plain silica toroids
and rhodamine B functionalized toroids were imaged (Figure 7-10). While plain
silica toroids did not exhibit any fluorescence (Figure 7-10a,b), the rhodamine B
functionalized samples emitted light (Figure 7-10c, d), indicating that the rhodamine
B was successfully attached.
217

Figure 7-10: Fluorescence microscopy shows no emission from bare silica toroid controls (a, b) but
rhodamine B-functionalized toroids do emit light (c, d), indicating successful attachment

7.4.3 Rhodamine B Temperature Sensing Experiments

Once the rhodamine B functionalized samples have been prepared, the
fluorescence emission can be studied using the spectrograph, a 532nm laser, heated
stage, and the modified testing setup shown in Figure 7-11.
First, a rhodamine B functionalized toroid is glued onto the heated stage.
Using glass slides and coverslips, a small chamber is constructed around the toroid
and filled with water. Control experiments can then be performed by heating the
stage with a custom-built temperature controller, exciting the rhodamine B with a
532nm laser, and measuring rhodamine B’s emission on the spectrograph. Plotting
emission versus temperature produces a calibration curve which can be used later to
relate a measured fluorescence intensity to a temperature.
218

Figure 7-11: Testing setup used for temperature dependent measurements. The setup is enclosed
inside a blackout curtain to block ambient light.

During the entire process, the setup is protected from ambient light using a
blackout curtain and the sample is exposed only very briefly to the 532nm excitation
light to minimize photobleaching. Using a light guide for the 532nm light allows the
samples to be excited with the blackout curtain closed and prevents accidental
bumping/misalignment of the laser and setup. As water evaporates, the sample
chamber must be refilled with water using a syringe or slow flow from a syringe
pump.
Once the control experiments are finished, resonant peaks are found near
1300nm or 1550nm using a tunable laser, standard tapered optical fiber, function
generator, oscilloscope, cameras, and detector. As high intensity light circulates in
the toroid, heating will occur in the toroid and in the surrounding water, which
219
strongly absorbs at the 1300nm and 1550nm wavelengths. Depending on the Q and
input power of the toroid, COMSOL predicts that the average temperature on the
exposed silica will increase (Figure 7-12). Any experimentally observed
fluorescence change in Rhodamine B can be correlated with a temperature change on
the toroid’s exposed silica surface, allowing the temperature to be estimated.

Figure 7-12: Sample surface temperature vs. 1550nm input power calculations in COMSOL.
Depending on the quality factor at 1550nm and the input power, the average temperature on the entire
exposed silica surface can increase significantly.

Once the control experiments are completed, the temperature on actual toroid
samples can be measured. To determine the temperature of the toroid, the rhodamine
B dye is excited with a 532nm laser and the dye’s emission intensity as a function of
wavelength is recorded on the spectrograph. By varying the input power to the
toroid, the fluorescence intensity can be determined as a function of circulating
power. Then, by comparing the fluorescence intensity with the emission vs.
temperature control experiments, the temperature at the toroid surface can be
220
determined. Ultimately, this setup should enable experimental verification of the
COMSOL simulations, and allow the light-induced heating to be further
investigated.
7.4.4 Fluorescent Bead Tracking Experiments

Once the heating experiments are completed, the photophoresis and
thermophoresis forces can also be studied by tracking fluorescent polystyrene beads
around silica toroids. By tuning the bead diameter, input power, toroid diameter, and
other properties, the forces around the toroids could potentially be quantified based
on the bead velocity and behavior. To do these experiments, a similar testing setup
is used as for the rhodamine B temperature sensing experiments, including the
532nm laser and filter (Figure 7-13). Fluorescent polystyrene beads are excited by
the 532nm laser, and their position is monitored on a microscope camera. Videos of
the beads could then be analyzed using particle tracking software which Yuting Liu,
an undergraduate student, had helped prepare.
221

Figure 7-13: Schematic of particle tracking photophoresis and thermophoresis experiments.

7.5 Results and Discussion

As a proof of concept, fluorescence intensity vs. temperature results were
first obtained for a rhodamine B solution in water. To ensure that the correct
emission peak is studied for the emission vs. temperature experiments, the
fluorescence of rhodamine B dye in water was measured using a spectrofluorimeter
(PTI Quantamaster 4, Figure 7-14a) and the spectrograph (Figure 7-14b). Both
instruments detected the same broad emission peak centered around 635nm. This
confirms that the spectrograph is able to monitor rhodamine B’s fluorescence in
water.
222

Figure 7-14: The same emission peak is observed for rhodamine B in water when using a
spectrofluorimeter (a) and the spectrograph (b).

Since increased temperature causes a decrease in the rhodamine B
fluorescence intensity, one concern about the temperature dependent fluorescent
measurements is that photobleaching and hysteresis could distort the results. To
check this, a solution of rhodamine B in water was placed on the testing setup’s
heated stage and the emission intensity was measured at ramped temperatures (from
30-90°C). Emission was recorded during two cycles of increasing and decreasing
temperature, while allowing the temperature to stabilize before each measurement.
Then, the area under each emission peak (from 610-680nm) was integrated and
plotted as a function of temperature. As seen in Figure 7-15, the change in
fluorescence remains highly consistent throughout both temperature ramp cycles.
There is some slight variation in the intensity data due to power fluctuations from the
532nm laser pointer used (later, a more stable 532nm diode laser was used). Based
on these results and data in the literature, any fluorescence intensity change for
Rhodamine B could be correlated to a change in temperature with minimal
hysteresis, although photobleaching is still a possible concern for long experiments.
223

Figure 7-15: Rhodamine B emission vs. temperature data when the temperature is ramped (a). The
fluorescence has minimal hysteresis, as seen in (b).

Having verified that rhodamine B behaves as expected, the emission intensity
versus circulating power measurements were performed on rhodamine B-
functionalized toroids with the 1300nm laser. The 1300nm laser was chosen due to
the laser’s moderately high output moderate powers (~5-8mW), the moderate
absorption coefficient of 1300nm in water (~130m
-1
), and the COMSOL results
predicting both heating and photophoretic effects. Emission intensity data was then
recorded both with the 1300nm light coupled into the toroid and without coupling (as
a background).
Unfortunately, these experiments did not work well. Since only a monolayer
of rhodamine B dye is attached to the toroids, any emitted fluorescence is extremely
weak (and only becomes weaker as the sample is heated). As a result, the data
becomes incredibly noisy due to the weak signals present. In addition, space
limitations made setting up and running the experiments challenging. The
224
spectrograph tip, 532nm filter, tapered fiber, excitation light from the 532nm laser,
and top view camera all needed to be very close to the toroid sample, making the
setup extremely crowded. Needing to place the 532nm filter between the toroid
sample and spectrograph tip also increased the distance between the sample and
spectrograph tip, further weakening any fluorescent signal. Also, the fluorescence
intensity values seemed to fluctuate greatly, regardless of whether light was coupled
into the toroid. Finally, the rhodamine B on the sample likely photobleached
quickly, as the samples did not show a strong emission peak for long. Because of all
these issues, the limited experimental data obtained has poor agreement with the
predicted COMSOL results.

Figure 7-16: Theoretical and experimental data for heating in toroids. The experimental data does
not match the trend predicted by COMSOL.

225
A final attempt to study the optical and thermal forces around the toroids
involved tracking fluorescent polystyrene beads near toroids in water. Using a 5W
532nm diode laser at ~15-20% intensity, 2 micron fluorescent polystyrene beads
could be easily seen on the testing setup’s top view camera (Figure 7-17). Videos of
the beads could be recorded and analyzed using particle tracking software which
Yuting Liu helped develop. By tracking the position and velocity of the beads under
different conditions, it may be possible to quantify the thermal and optical forces
present.

Figure 7-17: Strong red-orange fluorescence near 600nm is observed in a concentrated solution of
2µm polystyrene beads, as viewed using 532nm excitation laser and 532nm filter to block the green
excitation light.

However, once again, some issues arose making these measurements
unfeasible. First, exciting the polystyrene beads required extremely very high laser
powers at 532nm (several hundred milliwatts to 1 watt). When the beads are
exposed to the high laser power, they absorb the 532nm light strongly and are pushed
226
onto the substrate by the 532nm laser (Figure 7-18). Also, beads smaller than ~1
micron were very difficult to see on the testing setup cameras. In order to visualize
smaller beads (which were more likely to experience the optical and thermal forces
of interest), higher magnification lenses and modified optics would be needed.
These higher magnification optics would prevent the entire toroid sample from being
visible, and make it difficult to experimentally study the whole toroid system.
Therefore, since experimental verification of the COMSOL was not possible after
months of trying unsuccessfully, this project was ultimately discontinued.

Figure 7-18: Polystyrene beads are pushed onto silicon substrate, preventing imaging of the optical
and thermal forces around toroids. The beads can be seen clinging to the silicon substrate when the
microscope is focused on the toroid’s silicon substrate instead of the toroid.

7.6 Conclusion

From the limited results of this project, it is clear that the optical and thermal
forces can play a significant role in optical toroid resonators. Based on the
COMSOL models, significant heating and optical forces can occur around toroid
sensors. Using rhodamine B dye and fluorescent polystyrene beads, an attempt was
made to study these complex forces experimentally. However, noise and limitations
in the testing setup ultimately prevented clear results from being obtained. While
227
these forces could not be experimentally quantified, the limited results can provide
useful insight to build upon in future experiments.

Chapter 7 References

1. D. K. Armani, T. J. Kippenberg, S. M. Spillane, and K. J. Vahala, "Ultra-
high-Q toroid microcavity on a chip," Nature 421, 925-928 (2003).

2. H. S. Choi, and A. M. Armani, "Thermally stable hybrid organic/inorganic
resonant cavities," in Conference on Linear and Nonlinear Optics of Organic
Materials XI(San Diego, CA, 2011).

3. H. S. Choi, and A. M. Armani, "Thermal nonlinear effects in hybrid optical
microresonators," Applied Physics Letters 97 (2010).

4. S. Y. Lin, and K. B. Crozier, "Trapping-Assisted Sensing of Particles and
Proteins Using On-Chip Optical Microcavities," ACS Nano 7, 1725-1730 (2013).

5. S. Duhr, and D. Braun, "Why molecules move along a temperature gradient,"
Proceedings of the National Academy of Sciences of the United States of America
103, 19678-19682 (2006).

6. R. Piazza, and A. Parola, "Thermophoresis in colloidal suspensions," Journal
of Physics-Condensed Matter 20 (2008).

7. S. A. Putnam, and D. G. Cahill, "Transport of nanoscale latex spheres in a
temperature gradient," Langmuir 21, 5317-5323 (2005).

8. X. Serey, S. Mandal, Y. F. Chen, and D. Erickson, "DNA Transport and
Delivery in Thermal Gradients near Optofluidic Resonators," Physical Review
Letters 108 (2012).

9. A. Ashkin, "Forces of a single-beam gradient laser trap on a dielectric sphere
in the ray optics regime," Biophysical Journal 61, 569-582 (1992).

10. D. Erickson, X. Serey, Y. F. Chen, and S. Mandal, "Nanomanipulation using
near field photonics," Lab on a Chip 11, 995-1009 (2011).

11. D. G. Grier, "A revolution in optical manipulation," Nature 424, 810-816
(2003).

12. C. L. Kuyper, and D. T. Chiu, "Optical trapping: A versatile technique for
biomanipulation," Appl. Spectrosc. 56, 300A-312A (2002).
228

13. E. J. G. Peterman, F. Gittes, and C. F. Schmidt, "Laser-induced heating in
optical traps," Biophysical Journal 84, 1308-1316 (2003).

14. A. H. J. Yang, S. D. Moore, B. S. Schmidt, M. Klug, M. Lipson, and D.
Erickson, "Optical manipulation of nanoparticles and biomolecules in sub-
wavelength slot waveguides," Nature 457, 71-75 (2009).

15. T. Funatsu, Y. Harada, H. Higuchi, M. Tokunaga, K. Saito, Y. Ishii, R. D.
Vale, and T. Yanagida, "Imaging and nano-manipulation of single biomolecules,"
Biophys. Chem. 68, 63-72 (1997).

16. H. B. Xin, X. M. Li, and B. J. Li, "Massive photothermal trapping and
migration of particles by a tapered optical fiber," Optics Express 19, 17065-17074
(2011).

17. S. Arnold, D. Keng, S. I. Shopova, S. Holler, W. Zurawsky, and F. Vollmer,
"Whispering gallery mode carousel - a photonic mechanism for enhanced
nanoparticle detection in biosensing," Optics Express 17, 6230-6238 (2009).

18. G. Balgi, D. E. Leckband, and J. M. Nitsche, "Transport Effects on the
Kinetics of Protein-Surface Binding," Biophysical Journal 68, 2251-2260 (1995).

19. J. R. Bielenberg, and H. Brenner, "A hydrodynamic/Brownian motion model
of thermal diffusion in liquids," Physica A-Statistical Mechanics and Its Applications
356, 279-293 (2005).

20. A. P. Fields, and A. E. Cohen, "Anti-Brownian traps for studies on single
molecules," Methods in Enzymology, Vol 475: Single Molecule Tools, Pt B 474,
149-174 (2010).

21. J. W. J. Kerssemakers, M. E. Janson, A. van der Horst, and M. Dogterom,
"Optical trap setup for measuring microtubule pushing forces," Applied Physics
Letters 83, 4441-4443 (2003).

22. S. C. Kuo, and M. P. Sheetz, "Force of single kinesin molecules measured
with optical tweezers," Science 260, 232-234 (1993).

23. G. V. Shivashankar, and A. Libchaber, "Single DNA molecule grafting and
manipulation using a combined atomic force microscope and an optical tweezer,"
Applied Physics Letters 71, 3727-3729 (1997).

24. E. J. Smith, W. Xi, D. Makarov, I. Monch, S. Harazim, V. A. B. Quinones, C.
K. Schmidt, Y. F. Mei, S. Sanchez, and O. G. Schmidt, "Lab-in-a-tube: ultracompact
229
components for on-chip capture and detection of individual micro-/nanoorganisms,"
Lab on a Chip 12, 1917-1931 (2012).

25. H. Yin, M. D. Wang, K. Svoboda, R. Landick, S. M. Block, and J. Gelles,
"Transcription against an applied force," Science 270, 1653-1657 (1995).

26. Y. Zhang, H. X. Lei, Y. Z. Li, and B. J. Li, "Microbe removal using a
micrometre-sized optical fiber," Lab on a Chip 12, 1302-1308 (2012).

27. C. Shi, S. Mehrabani, and A. M. Armani, "Leveraging bimodal kinetics to
improve detection specificity," Optics Letters 37, 1643-1645 (2012).

28. C. E. Soteropulos, H. K. Hunt, and A. M. Armani, "Determination of binding
kinetics using whispering gallery mode microcavities," Applied Physics Letters 99
(2011).

29. H. K. Hunt, C. Soteropulos, and A. M. Armani, "Bioconjugation Strategies
for Microtoroidal Optical Resonators," Sensors 10, 9317-9336 (2010).

30. M. I. Cheema, and A. G. Kirk, "Implementation of the perfectly matched
layer to determine the quality factor of axisymmetric resonators in COMSOL," in
COMSOL Conference(Boston, 2010).

31. M. Oxborrow, "How to simulate the whispering-gallery-modes of dielectric
microresonators in FEMLAB/COMSOL - art. no. 64520J," in Conference on Laser
Resonators and Beam Control IX(San Jose, CA, 2007), pp. J4520-J4520.

32. COMSOL, "Introduction to the Heat Transfer Module," (COMSOL
Multiphysics 4.3a, 2001).

33. COMSOL, "Heat Transfer Module User’s Guide," (COMSOL Multiphysics
4.3a, 2012).

34. R. B. Bird, W. E. Stewart, and E. N. Lightfoot, Transport Phenomena,
Second Edition (Wiley, 2007).

35. A. E. Cohen, and W. E. Moerner, "Method for trapping and manipulating
nanoscale objects in solution," Applied Physics Letters 86 (2005).

36. S. Duhr, and D. Braun, "Thermophoretic depletion follows Boltzmann
distribution," Physical Review Letters 96 (2006).

37. X. Y. Ma, J. Q. Lu, R. S. Brock, K. M. Jacobs, P. Yang, and X. H. Hu,
"Determination of complex refractive index of polystyrene microspheres from 370 to
1610 nm," Physics in Medicine and Biology 48, 4165-4172 (2003).
230

38. H. Monjushiro, A. Hirai, and H. Watarai, "Size dependence of laser-
photophoretic efficiency of polystyrene microparticles in water," Langmuir 16,
8539-8542 (2000).

39. L. Korson, Drosthan.W, and F. J. Millero, "Viscosity of water at various
temperatures," Journal of Physical Chemistry 73, 34-& (1969).

40. H. D. Duong, and J. I. Rhee, "Exploitation of thermo-effect of rhodamine B
entrapped in sol-gel matrix and silica gel for temperature detection," Sensors and
Actuators B-Chemical 124, 18-23 (2007).

41. T. Lu, H. Lee, T. Chen, S. Herchak, J. H. Kim, S. E. Fraser, R. C. Flagan, and
K. Vahala, "High sensitivity nanoparticle detection using optical microcavities,"
Proceedings of the National Academy of Sciences of the United States of America
108, 5976-5979 (2011).

42. G. T. Hermanson, Bioconjugate Techniques (Second Edition) (Elsevier
Science, 2008).

231
Chapter 8 Future Work

8.1 Introduction

Having completed the various projects outlined in Chapters 3-7, it is
important to consider future directions and applications for these research areas. In
each of the following sections, some interesting further projects are described in
greater detail.
8.2 Sol-gel Synthesis and Soft Lithography

Sol-gel silica has been implemented throughout this work in order to develop
custom optical materials. The silica films doped with titanium found numerous
applications in the development of high index coatings [1-3], and the alumina and
neodymium-doped sol-gel silica enabled a heterodynable toroid microlaser and
sensor to be created [4, 5]. One of the most important advantages of sol-gel silica in
all these cases is its liquid precursor, which enables dopants to be easily added and
allows the liquid sol-gels to be easily spin-coated onto substrates.
Yet sol-gel silica’s liquid state has another tremendous advantage which we
have not investigated. With enough age time (~2+ weeks), sol-gel silica can become
a viscous liquid, and could be used in replica molding/soft lithography [6-10]. For
example, silica toroid-shaped molds could be produced using PDMS and an array of
toroids fabricated using the standard procedures. Then, the sol-gel silica could be
poured into the PDMS molds and aged/annealed further. Then, the molded silica
could be placed on silicon wafers and the PDMS could be peeled off, producing sol-
gel silica devices. Compared to the spin coating, annealing, and standard lithography
232
procedures used in this work, soft lithography offers a significantly simpler, cheaper,
and faster alternative which also eliminates the need for large amounts of XeF
2
, HF,
and other toxic chemicals. Soft lithography could also enable these sol-gel silica
devices to be more easily and reliably mass-produced. While replica molding has
already been used to make polymer toroids [6], potential challenges of replica
molding sol-gel devices include shrinkage of the sol-gel and cracking or destruction
of the elastomeric mold during the sol-gel anneal process. If these issues could be
addressed, replica molded devices could have greatly simplified fabrication, a wider
range of usable materials, and potentially better performance in more applications.
8.3 Integrated Heterodyned Microlaser

One of the main accomplishments of this work is the successful development
of a heterodyned toroidal microlaser sensor which works in both air and aqueous
environments. In order to make this microlaser sensor truly integrated and useful in
biodetection applications, it is necessary to 1) improve the linewidth of the toroid
microlaser to increase sensitivity and 2) integrate the toroid microlaser with on-chip
waveguides. These are important to maximize sensitivity and signal to noise while
allowing the laser to become a more integrated sensing platform.
Another challenge in developing sensors such as the heterodyned microlaser
is minimizing signal to noise. While the heterodyne approach helped improve noise
to some extent, a significant amount of noise remains in the system. These noise
sources can include:
- taper jitter (even though the taper is in contact with the toroid, it could
move around, especially in water and during flow)
233
- thermal fluctuations (due to high circulating power in toroid, as well as
from flow)
- laser drift from the tunable 765nm pump and 1064nm reference lasers
- vibrations in the setup which cannot be completely canceled out by the
testing setup
- reduced linewidth of the toroid microlaser (due to the increased
absorption loss of neodymium, water, and surface chemistry)
Improving the microlaser sensor’s linewidth and integrating the microlaser sensor
with waveguides for coupling can help solve or reduce many of these noise sources.
Narrowing the microlaser sensor linewidth can significantly improve the sensitivity
of the heterodyned device by allowing smaller wavelength shifts to be measured. As
seen in the previous sections, the smallest measurable wavelength shift was
approximately 25-30MHz with the heterodyned setup, which corresponds to
wavelength shifts of about 0.1 picometers (100 femtometers). In order to detect
single molecules as in Prof. Armani’s Science paper, wavelength shifts below 5-10
femtometers need to be measurable. Ideally, the heterodyned microlaser would be
an improvement and be capable of measuring even smaller wavelength shifts.
To improve the microlaser’s linewidth, it is necessary to improve its quality
factor. If the toroid microlaser’s quality factor could be increased from 10
5
to 10
6
or
10
7
, the microlaser’s linewidth could be improved. Otherwise, the heterodyned
approach will not offer significant improvement over the resonant wavelength
tracking-based sensing approach developed by Prof. Armani and others.
234
There are several possible approaches to improve the heterodyned
microlaser’s quality factor. First, different glass materials could be tried, such as the
hybrid CaF
2
-silica sol-gels. These may provide a better lasing environment for the
neodymium and enable higher quality factors to be obtained. If that does not work, it
may be possible to produce a toroid microlaser which emits at heterodynable
wavelengths in the visible. For example, since the optical absorption coefficient of
water is much lower at 633nm, developing a heterodynable laser which emits at
633nm could vastly improve the quality factor. One challenge, however, is that rare
earth dopants or other materials must be found which both absorb and emit light
within the range of a commercially available tunable lasers. This can be a significant
challenge.
Also important to reduce noise and improve sensor performance is
integrating the microlasers with on-chip waveguides. Integrating the toroid lasers
with a waveguide [11] could eliminate many sources of noise, including taper jitter.
However, fabricating these devices can be a significant challenge.
8.4 New Materials for Higher Performance Devices

Silica’s low optical loss, minimal nonlinearity, and compatibility with
microfabrication techniques make it a good material for integrated optics. However,
the performance of many optical devices is limited by silica’s properties. If new and
improved materials can be developed, this material limitation can be overcome and
even greater enhancements to device performance could be possible. For example,
some non-silica materials such as anisotropic crystals can exhibit fascinating and
interesting nonlinear optical effects, including second harmonic generation [12, 13].
235
Additionally, materials with lower phonon energies, such as CaF
2
, could enable
observation of enhanced Raman lasing, frequency combs, and upconversion effects
[14-16]. Finding ways to easily and reliably make ultra high Q resonators and other
integrated optical devices out of these interesting materials could enable higher
performance optical devices with even greater capabilities.
As seen in this thesis, one way to develop improved optical devices is to
develop custom materials using the sol-gel approach [1, 3, 5]. By continuing to
explore the sol gel approach and new optical materials, the capabilities of integrated
optical devices can expand even further.

Chapter 8 References

1. N. Deka, A. J. Maker, and A. M. Armani, "Titanium enhanced Raman
microcavity laser," Optics Letters 39 (2014).

2. A. J. Maker, B. A. Rose, and A. M. Armani, "Tailoring the behavior of
optical microcavities with high refractive index sol-gel coatings," Optics Letters 37,
2844-2846 (2012).

3. B. A. Rose, A. J. Maker, and A. M. Armani, "Characterization of thermo-
optic coefficient and material loss of high refractive index silica sol-gel films in the
visible and near-IR," in Optical Materials Express(OSA, 2012), pp. 671-681.

4. A. J. Maker, and A. M. Armani, "Heterodyned Toroidal Microlaser Sensor,"
Appl. Phys. Lett. 103, 123302 (2013).

5. A. J. Maker, and A. M. Armani, "Nanowatt threshold, alumina sensitized
neodymium laser integrated on silicon," Optics Express 21, 27238-27245 (2013).

6. A. L. Martin, D. K. Armani, L. Yang, and K. J. Vahala, "Replica-molded
high-Q polymer microresonators," Optics Letters 29, 533-535 (2004).

7. Y. N. Xia, and G. M. Whitesides, "Soft lithography," Annual Review of
Materials Science 28, 153-184 (1998).

236
8. F. Back, M. Bockmeyer, E. Rudigier-Voigt, and P. Lobmann, "Hybrid
polymer sol-gel material for UV-nanoimprint: microstructure and thermal
densification," Journal of Sol-Gel Science and Technology 66, 73-83 (2013).

9. X. M. Han, J. Lin, R. B. Xing, J. Fu, and S. B. Wang, "Patterning and optical
properties rhodamine B-doped organic-inorganic silica films fabricated by sol-gel
soft lithography," Materials Letters 57, 1355-1360 (2003).

10. S. Sarkar, P. K. Biswas, and S. Jana, "Nano silver coated patterned silica thin
film by sol-gel based soft lithography technique," Journal of Sol-Gel Science and
Technology 61, 577-584 (2012).

11. X. M. Zhang, and A. M. Armani, "Silica microtoroid resonator sensor with
monolithically integrated waveguides," Optics Express 21, 23592-23603 (2013).

12. J. Jerphagnon, and S. K. Kurtz, "Maker fringes - A detailed comparison of
theory and experiment for isotropic and uniaxial crystals," Journal of Applied
Physics 41, 1667-& (1970).

13. P. D. Maker, C. M. Savage, R. W. Terhune, and M. Nisenoff, "Effects of
dispersion and focusing on production of optical harmonics," Physical Review
Letters 8, 21-& (1962).

14. I. S. Grudinin, and L. Maleki, "Ultralow-threshold Raman lasing with CaF
2
resonators," Optics Letters 32, 166-168 (2007).

15. I. S. Grudinin, N. Yu, and L. Maleki, "Generation of optical frequency combs
with a CaF2 resonator," Optics Letters 34, 878-880 (2009).

16. L. H. Zhou, D. Q. Chen, W. Q. Luo, Y. S. Wang, Y. L. Yu, and F. Liu,
"Transparent glass ceramic containing Er3+: CaF2 nano-crystals prepared by sol-gel
method," Materials Letters 61, 3988-3990 (2007).

237
Bibliography

M. Abdel-Baki, F. A. A. Wahab, and F. El-Diasty, "Optical characterization of
xTiO(2)-(60-x)SiO2-40Na(2)O glasses I. Linear and nonlinear dispersion
properties," Mater. Chem. Phys. 96, 201-210 (2006).

R. Adar, M. R. Serbin, and V. Mizrahi, "Less-than-1 DB per meter propagation loss
of silica wave-guides measured using a ring-resonator," Journal of Lightwave
Technology 12, 1369-1372 (1994).

R. Adar, Y. Shani, C. H. Henry, R. C. Kistler, G. E. Blonder, and N. A. Olsson,
"Measurement of very low-loss silica on silicon wave-guides with a ring resonator,"
Applied Physics Letters 58, 444-445 (1991).

B. J. Ainslie, S. P. Craig, and S. T. Davey, "The absorption and fluorescence-spectra
of rare-earth ions in silica-based monomode fiber," Journal of Lightwave
Technology 6, 287-293 (1988).

A. M. Armani, D. K. Armani, B. Min, K. J. Vahala, and S. M. Spillane, "Ultra-high-
Q microcavity operation in H2O and D2O," Applied Physics Letters 87 (2005).

A. M. Armani, R. P. Kulkarni, S. E. Fraser, R. C. Flagan, and K. J. Vahala, "Label-
free, single-molecule detection with optical microcavities," Science 317, 783-787
(2007).

A. M. Armani, and K. J. Vahala, "Heavy water detection using ultra-high-Q
microcavities," Optics Letters 31, 1896-1898 (2006).

D. K. Armani, T. J. Kippenberg, S. M. Spillane, and K. J. Vahala, "Ultra-high-Q
toroid microcavity on a chip," Nature 421, 925-928 (2003).

S. Arnold, D. Keng, S. I. Shopova, S. Holler, W. Zurawsky, and F. Vollmer,
"Whispering gallery mode carousel - a photonic mechanism for enhanced
nanoparticle detection in biosensing," Optics Express 17, 6230-6238 (2009).

A. Ashkin, "Forces of a single-beam gradient laser trap on a dielectric sphere in the
ray optics regime," Biophysical Journal 61, 569-582 (1992).

F. Back, M. Bockmeyer, E. Rudigier-Voigt, and P. Lobmann, "Hybrid polymer sol-
gel material for UV-nanoimprint: microstructure and thermal densification," Journal
of Sol-Gel Science and Technology 66, 73-83 (2013).

G. Balgi, D. E. Leckband, and J. M. Nitsche, "Transport Effects on the Kinetics of
Protein-Surface Binding," Biophysical Journal 68, 2251-2260 (1995).

238
C. Bartolacci, M. Laroche, T. Robin, B. Cadier, S. Girard, and H. Gilles, "Effects of
ions clustering in Nd3+/Al3+-codoped double-clad fiber laser operating near 930
nm," Appl. Phys. B 98, 317-322 (2010).

J. F. Bauters, M. J. R. Heck, D. John, D. X. Dai, M. C. Tien, J. S. Barton, A. Leinse,
R. G. Heideman, D. J. Blumenthal, and J. E. Bowers, "Ultra-low-loss high-aspect-
ratio Si3N4 waveguides," Optics Express 19, 3163-3174 (2011).

A. Beganskiene, S. Sakirzanovas, I. Kazadojev, A. Melninkaitis, V. Sirutkaitis, and
A. Kareiva, "Sol-gel derived antireflective coating with controlled thickness and
reflective index," Mater. Sci. 25, 817-824 (2007).

A. J. Berry, and T. A. King, "Characterization of doped sol-gel derived silica hosts
for use in tunable glass lasers," J. Phys. D: Appl. Phys 22, 1419-1422 (1989).

J. R. Bielenberg, and H. Brenner, "A hydrodynamic/Brownian motion model of
thermal diffusion in liquids," Physica a-Statistical Mechanics and Its Applications
356, 279-293 (2005).

B. W. Biggs, H. K. Hunt, and A. M. Armani, "Selective patterning of Si-based
biosensor surfaces using isotropic silicon etchants," Journal of Colloid and Interface
Science 369, 477-481 (2012).

R. B. Bird, W. E. Stewart, and E. N. Lightfoot, Transport Phenomena, Second
Edition (Wiley, 2007).

A. Biswas, G. S. Maciel, C. S. Friend, and P. N. Prasad, "Upconversion properties of
a transparent Er3+-Yb3+ co-doped LaF3-SiO2 glass-ceramics prepared by sol-gel
method," J. Non-Cryst. Solids 316, 393-397 (2003).

J. Bonar, J. A. Bebbington, J. S. Aitchison, G. D. Maxwell, and B. J. Ainslie,
"Aerosol doped Nd planar silica wave-guide laser," Electron. Lett. 31, 99-100
(1995).

M. Borselli, T. J. Johnson, and O. Painter, "Beyond the Rayleigh scattering limit in
high-Q silicon microdisks: theory and experiment," Optics Express 13, 1515-1530
(2005).

G. Brusatin, M. Guglielmi, P. Innocenzi, A. Martucci, C. Battaglin, S. Pelli, and G.
Righini, "Microstructural and optical properties of sol-gel silica-titania waveguides,"
J. Non-Cryst. Solids 220, 202-209 (1997).

M. Cai, O. Painter, K. J. Vahala, and P. C. Sercel, "Fiber-coupled microsphere laser,"
Optics Letters 25, 1430-1432 (2000).

239
M. I. Cheema, and A. G. Kirk, "Implementation of the perfectly matched layer to
determine the quality factor of axisymmetric resonators in COMSOL," in COMSOL
Conference(Boston, 2010).

M. Chistiakova, and A. M. Armani, "Cascaded Raman microlaser in air and buffer,"
Optics Letters 37, 4068-4070 (2012).

H. S. Choi, and A. M. Armani, "Thermal nonlinear effects in hybrid optical
microresonators," Applied Physics Letters 97 (2010).

H. S. Choi, and A. M. Armani, "Thermally stable hybrid organic/inorganic resonant
cavities," in Conference on Linear and Nonlinear Optics of Organic Materials
XI(San Diego, CA, 2011).

H. S. Choi, S. Ismail, and A. M. Armani, "Studying polymer thin films with hybrid
optical microcavities," Optics Letters 36, 2152-2154 (2011).

H. S. Choi, D. Neiroukh, H. K. Hunt, and A. M. Armani, "Thermo-optic Coefficient
of Polyisobutylene Ultrathin Films Measured with Integrated Photonic Devices,"
Langmuir 28, 849-854 (2012).

H. S. Choi, X. M. Zhang, and A. M. Armani, "Hybrid silica-polymer ultra-high-Q
microresonators," Optics Letters 35, 459-461 (2010).

A. E. Cohen, and W. E. Moerner, "Method for trapping and manipulating nanoscale
objects in solution," Applied Physics Letters 86 (2005).

P. Colomban, and A. Slodczyk, "Raman Intensity: An Important Tool in the Study of
Nanomaterials and Nanostructures," Acta Physica Polonica A 116, 7-12 (2009).

R. Colombelli, K. Srinivasan, M. Troccoli, O. Painter, C. F. Gmachl, D. M. Tennant,
A. M. Sergent, D. L. Sivco, A. Y. Cho, and F. Capasso, "Quantum cascade surface-
emitting photonic crystal laser," Science 302, 1374-1377 (2003).

COMSOL, "Heat Transfer Module User’s Guide," (COMSOL Multiphysics 4.3a,
2012).

COMSOL, "Introduction to the Heat Transfer Module," (COMSOL Multiphysics
4.3a, 2001).

G. Dai, F. Tassone, A. L. Bassi, V. Russo, C. E. Bottani, and F. D'Amore, "TeO2-
based glasses containing Nb2(O5), TiO2, and WO3 for discrete Raman fiber
amplification," IEEE Photonics Technology Letters 16, 1011-1013 (2004).

240
N. Deka, A. J. Maker, and A. M. Armani, "Titanium enhanced Raman microcavity
laser," Optics Letters 39 (2014).

D. Derickson, Fiber optic test and measurement (Prentice Hall, 1998).

D. Duchesne, M. Ferrera, L. Razzari, R. Morandotti, B. Little, S. Chu, and D. Moss,
"Efficient self-phase modulation in low loss, high index doped silica glass integrated
waveguides," in Optics Express(Optical Society of America, 2009), pp. 1865-1870.

S. Duhr, and D. Braun, "Thermophoretic depletion follows Boltzmann distribution,"
Physical Review Letters 96 (2006).

S. Duhr, and D. Braun, "Why molecules move along a temperature gradient,"
Proceedings of the National Academy of Sciences of the United States of America
103, 19678-19682 (2006).

R. L. Dumas, I. Tejedor-Tejedor, and M. A. Anderson, "Dependence of SiO2 gel
structure on gelation conditions and sol reaction temperature as followed by FTIR
and nitrogen adsorption measurements," Journal of Porous Materials 5, 95-101
(1998).

H. D. Duong, and J. I. Rhee, "Exploitation of thermo-effect of rhodamine B
entrapped in sol-gel matrix and silica gel for temperature detection," Sensors and
Actuators B-Chemical 124, 18-23 (2007).

D. Erickson, S. Mandal, A. Yang, and B. Cordovez, "Nanobiosensors: optofluidic,
electrical and mechanical approaches to biomolecular detection at the nanoscale," in
Microfluid Nanofluid(Springer, 2007), pp. 33-52.

D. Erickson, X. Serey, Y. F. Chen, and S. Mandal, "Nanomanipulation using near
field photonics," Lab on a Chip 11, 995-1009 (2011).

M. A. Fardad, "Catalysts and the structure of SiO2 sol-gel films," J. Mater. Sci. 35,
1835-1841 (2000).

M. A. Fardad, E. M. Yeatman, E. J. C. Dawnay, M. Green, and F. Horowitz, "Effects
of H2O on structure of acid-catalyzed SiO2 sol-gel films," J. Non-Cryst. Solids 183,
260-267 (1995).

L. A. Farrow, and E. M. Vogel, "Raman-spectra of phosphate and silicate glasses
doped with the cations Ti, Nb, and Bi," J. Non-Cryst. Solids 143, 59-64 (1992).

M. Ferrera, L. Razzari, D. Duchesne, R. Morandotti, Z. Yang, M. Liscidini, J. E.
Sipe, S. Chu, B. E. Little, and D. J. Moss, "Low-power continuous-wave nonlinear
241
optics in doped silica glass integrated waveguide structures," Nature Photonics 2,
737-740 (2008).

A. P. Fields, and A. E. Cohen, "Anti-Brownian traps for studies on single
molecules," Methods in Enzymology, Vol 475: Single Molecule Tools, Pt B 474,
149-174 (2010).

T. Fujiyama, T. Yokoyama, M. Hori, and M. Sasaki, "Silica glass doped with Nd and
Al prepared by the sol-gel method - change in the state of aluminum in the formation
process," J. Non-Cryst. Solids 135, 198-203 (1991).

T. Funatsu, Y. Harada, H. Higuchi, M. Tokunaga, K. Saito, Y. Ishii, R. D. Vale, and
T. Yanagida, "Imaging and nano-manipulation of single biomolecules," Biophys.
Chem. 68, 63-72 (1997).

T. Gacoin, L. Malier, and J. P. Boilot, "New transparent chalcogenide materials
using a sol-gel process," Chemistry of Materials 9, 1502-& (1997).

C. Ge, M. Lu, S. George, T. A. Flood, C. Wagner, J. Zheng, A. Pokhriyal, J. G.
Eden, P. J. Hergenrother, and B. T. Cunningham, "External cavity laser biosensor,"
Lab on a Chip 13, 1247-1256 (2013).

R. Germann, H. W. M. Salemink, R. Beyeler, G. L. Bona, F. Horst, I. Massarek, and
B. J. Offrein, "Silicon oxynitride layers for optical waveguide applications," J.
Electrochem. Soc. 147, 2237-2241 (2000).

G. Ghosh, Handbook of Thermo-Optic Coefficients of Optical Materials with
Applications (Academic Press, 1998).

E. R. Giles, and E. Desurvire, "Modeling erbium-doped fiber amplifiers," J.
Lighwave Technol. 9, 271-283 (1991).

M. L. Gorodetsky, A. A. Savchenkov, and V. S. Ilchenko, "Ultimate Q of optical
microsphere resonators," Optics Letters 21, 453-455 (1996).

D. G. Grier, "A revolution in optical manipulation," Nature 424, 810-816 (2003).

I. S. Grudinin, and L. Maleki, "Ultralow-threshold Raman lasing with CaF
2
resonators," Optics Letters 32, 166-168 (2007).

I. S. Grudinin, N. Yu, and L. Maleki, "Generation of optical frequency combs with a
CaF2 resonator," Optics Letters 34, 878-880 (2009).

242
R. Gupta, S. Mozumdar, and N. K. Chaudhury, "Effect of ethanol variation on the
internal environment of sol-gel bulk and thin films with aging," Biosensors &
Bioelectronics 21, 549-556 (2005).

G. M. Hale, and M. R. Querry, "Optical-Constants of Water in 200-nm to 200-mum
Wavelength Region," Appl. Opt. 12, 555-563 (1973).

M. Han, and A. Wang, "Temperature compensation of optical microresonators using
a surface layer with negative thermo-optic coefficient," Optics Letters 32, 1800-1802
(2007).

X. M. Han, J. Lin, R. B. Xing, J. Fu, and S. B. Wang, "Patterning and optical
properties rhodamine B-doped organic-inorganic silica films fabricated by sol-gel
soft lithography," Materials Letters 57, 1355-1360 (2003).

T. Hattori, S. Semura, and N. Akasaka, "Inductively coupled plasma-enhanced
chemical vapor deposition of SiO2 and GeO2-SiO2 films for optical waveguides
using tetraethylorthosilicate and tetramethylgermanium," Japanese Journal of
Applied Physics Part 1-Regular Papers Brief Communications & Review Papers 38,
2775-2778 (1999).

L. N. He, K. Ozdemir, J. G. Zhu, W. Kim, and L. Yang, "Detecting single viruses
and nanoparticles using whispering gallery microlasers," Nature Nanotechnology 6,
428-432 (2011).

L. N. He, S. K. Ozdemir, and L. Yang, "Whispering gallery microcavity lasers,"
Laser & Photonics Reviews 7, 60-82 (2013).

E. Hecht, Optics (Addison Wesley, 2002).

G. T. Hermanson, Bioconjugate Techniques (Second Edition) (Elsevier Science,
2008).

E. Herrero, N. Carmona, J. Llopis, and M. A. Villegas, "Sensitive glasslike sol-gel
materials suitable for environmental light sensors," Journal of the European Ceramic
Society 27, 4589-4594 (2007).

H. S. Hsu, C. Cai, and A. M. Armani, "Ultra-low-threshold Er:Yb sol-gel microlaser
on silicon," Optics Express 17, 23265-23271 (2009).

H. K. Hunt, and A. M. Armani, "Label-free biological and chemical sensors,"
Nanoscale 2, 1544-1559 (2010).

H. K. Hunt, and A. M. Armani, "Recycling microcavity optical biosensors," Optics
Letters 36, 1092-1094 (2011).
243

H. K. Hunt, C. Soteropulos, and A. M. Armani, "Bioconjugation Strategies for
Microtoroidal Optical Resonators," Sensors 10, 9317-9336 (2010).

V. S. Ilchenko, P. S. Volikov, V. L. Velichansky, F. Treussart, V. Lefevre-Seguin, J.
M. Raimond, and S. Haroche, "Strain-tunable high-Q optical microsphere resonator,"
Optics Communications 145, 86-90 (1998).

P. Innocenzi, "Infrared spectroscopy of sol-gel derived silica-based films: a spectra-
microstructure overview," J. Non-Cryst. Solids 316, 309-319 (2003).

S. D. Jackson, and A. Lauto, "Diode-pumped fiber lasers: A new clinical tool?,"
Laser Surg Med 30, 184-190 (2002).

A. Jain, A. H. J. Yang, and D. Erickson, "Gel-based optical waveguides with live cell
encapsulation and integrated microfluidics," Optics Letters 37, 1472-1474 (2012).

J. Jerphagnon, and S. K. Kurtz, "Maker fringes - a detailed comparison of theory and
experiment for isotropic and uniaxial crystals," Journal of Applied Physics 41, 1667-
& (1970).

A. Jitianu, M. Gartner, M. Zaharescu, D. Cristea, and E. Manea, "Experiments for
inorganic-organic hybrid sol-gel films for micro- and nano-photonics," (Elsevier
Science Bv2003), pp. 301-306.

C. K. Kao, H. Chang, W. Y. Lim, C. H. Tsai, C. C. Chi, N. H. Tai, and I. N. Lin,
"Optical properties of PECVD TEOS-SiO2 films," (Taylor & Francis Ltd2001), pp.
1949-1954.

J. W. J. Kerssemakers, M. E. Janson, A. van der Horst, and M. Dogterom, "Optical
trap setup for measuring microtubule pushing forces," Applied Physics Letters 83,
4441-4443 (2003).

D. H. Kim, Y. S. Kim, J. Amsden, B. Panilaitis, D. L. Kaplan, F. G. Omenetto, M. R.
Zakin, and J. A. Rogers, "Silicon electronics on silk as a path to bioresorbable,
implantable devices," Applied Physics Letters 95 (2009).

H. K. Kim, S. J. Kang, S. K. Choi, Y. H. Min, and C. S. Yoon, "Highly efficient
organic/inorganic hybrid nonlinear optic materials via sol-gel process: Synthesis,
optical properties, and photobleaching for channel waveguides," Chemistry of
Materials 11, 779-788 (1999).

T. Kitagawa, K. Hattori, Y. Hibino, and Y. Ohmori, "Neodymium-doped silica-based
planar wave-guide lasers," J. Lightwave Technol. 12, 436-442 (1994).

244
R. Kitamura, L. Pilon, and M. Jonasz, "Optical constants of silica glass from extreme
ultraviolet to far infrared at near room temperature," Applied Optics 46, 8118-8133
(2007).

J. Knittel, T. G. McRae, K. H. Lee, and W. P. Bowen, "Interferometric detection of
mode splitting for whispering gallery mode biosensors," Applied Physics Letters 97
(2010).

C. Kopp, S. Bernabe, B. Ben Bakir, J. M. Fedeli, R. Orobtchouk, F. Schrank, H.
Porte, L. Zimmermann, and T. Tekin, "Silicon Photonic Circuits: On-CMOS
Integration, Fiber Optical Coupling, and Packaging," IEEE J. Sel. Top. Quantum
Electron. 17, 498-509 (2011).

L. Korson, Drosthan.W, and F. J. Millero, "Viscosity of water at various
temperatures," Journal of Physical Chemistry 73, 34-& (1969).

P. Kozma, A. Hamori, S. Kurunczi, K. Cottier, and R. Horvath, "Grating coupled
optical waveguide interferometer for label-free biosensing," Sensors and Actuators
B-Chemical 155, 446-450 (2011).

S. C. Kuo, and M. P. Sheetz, "Force of single kinesin molecules measured with
optical tweezers," Science 260, 232-234 (1993).

C. L. Kuyper, and D. T. Chiu, "Optical trapping: A versatile technique for
biomanipulation," Appl. Spectrosc. 56, 300A-312A (2002).

M. Laczka, K. Cholewa-Kowalska, and M. Kogut, "Organic-inorganic hybrid glasses
of selective optical transmission," (Elsevier Science Bv2001), pp. 10-14.

P. W. Laird, "The power and the promise of DNA methylation markers," Nature
Reviews Cancer 3, 253-266 (2003).

S. S. Latthe, H. Imai, V. Ganesan, and A. V. Rao, "Superhydrophobic silica films by
sol-gel co-precursor method," Appl. Surf. Sci. 256, 217-222 (2009).

K. K. Lee, D. R. Lim, H. C. Luan, A. Agarwal, J. Foresi, and L. C. Kimerling,
"Effect of size and roughness on light transmission in a Si/SiO2 waveguide:
Experiments and model," Applied Physics Letters 77, 1617-1619 (2000).

V. Lefevre-Seguin, "Whispering-gallery mode lasers with doped silica
microspheres," Opt. Mater. 11, 153-165 (1999).

G. Li, M. Kanezashi, and T. Tsuru, "Preparation of organic-inorganic hybrid silica
membranes using organoalkoxysilanes: The effect of pendant groups," J. Membr.
Sci. 379, 287-295 (2011).
245

Y. Li, G. Vienne, X. Jiang, X. Pan, X. Liu, P. Gu, and L. Tong, "Modeling rare-earth
doped microfiber ring lasers," Opt. Express 14, 7073-7086 (2006).

J. T. Lin, Y. X. Xu, J. X. Song, B. Zeng, F. He, H. L. Xu, K. Sugioka, W. Fang, and
Y. Cheng, "Low-threshold whispering-gallery-mode microlasers fabricated in a
Nd:glass substrate by three-dimensional femtosecond laser micromachining," Opt.
Lett. 38, 1458-1460 (2013).

Q. Lin, O. J. Painter, and G. P. Agrawal, "Nonlinear optical phenomena in silicon
waveguides: Modeling and applications," Optics Express 15, 16604-16644 (2007).

S. Y. Lin, and K. B. Crozier, "Trapping-Assisted Sensing of Particles and Proteins
Using On-Chip Optical Microcavities," Acs Nano 7, 1725-1730 (2013).

J. Livage, and C. Sanchez, "Sol-gel chemistry," J. Non-Cryst. Solids 145, 11-19
(1992).

T. Lu, H. Lee, T. Chen, S. Herchak, J. H. Kim, S. E. Fraser, R. C. Flagan, and K.
Vahala, "High sensitivity nanoparticle detection using optical microcavities,"
Proceedings of the National Academy of Sciences of the United States of America
108, 5976-5979 (2011).

M. S. Luchansky, and R. C. Bailey, "High-Q Optical Sensors for Chemical and
Biological Analysis," Anal. Chem. 84, 793-821 (2012).

H. Ma, A. K. Y. Jen, and L. R. Dalton, "Polymer-based optical waveguides:
Materials, processing, and devices," Advanced Materials 14, 1339-1365 (2002).

X. Y. Ma, J. Q. Lu, R. S. Brock, K. M. Jacobs, P. Yang, and X. H. Hu,
"Determination of complex refractive index of polystyrene microspheres from 370 to
1610 nm," Physics in Medicine and Biology 48, 4165-4172 (2003).

A. J. Maker, and A. M. Armani, "Fabrication of silica ultra high quality factor
microresonators," Journal of Visualized Experiment 65, e4164 (2012).

A. J. Maker, and A. M. Armani, "Heterodyned Toroidal Microlaser Sensor," Appl.
Phys. Lett. 103, 123302 (2013).

A. J. Maker, and A. M. Armani, "Heterodyning cavity-based microlasers to improve
sensing performance," in SPIE Photonics West, A. V. Kudryashov, A. H. Paxton, V.
S. Ilchenko, L. Aschke, and K. Washio, eds. (SPIE, San Francisco, CA, 2014).

A. J. Maker, and A. M. Armani, "Low-loss silica-on-silicon waveguides," Optics
Letters 36, 3729-3731 (2011).
246

A. J. Maker, and A. M. Armani, "Low-loss silica on silicon integrated waveguides,"
in Conference on High Contrast Metastructures (San Francisco, CA, 2012).

A. J. Maker, and A. M. Armani, "Nanowatt threshold, alumina sensitized
neodymium laser integrated on silicon," Optics Express 21, 27238-27245 (2013).

A. J. Maker, B. A. Rose, and A. M. Armani, "Controlling the mode volume in high-
Q microcavities with high refractive index coatings," Integrated Optics: Devices,
Materials, and Technologies Xvii 8627 (2013).

A. J. Maker, B. A. Rose, and A. M. Armani, "Tailoring the behavior of optical
microcavities with high refractive index sol-gel coatings," Optics Letters 37, 2844-
2846 (2012).

P. D. Maker, C. M. Savage, R. W. Terhune, and M. Nisenoff, "Effects of dispersion
and focusing on production of optical harmonics," Physical Review Letters 8, 21-&
(1962).

A. K. Manocchi, P. Domachuk, F. G. Omenetto, and H. M. Yi, "Facile Fabrication of
Gelatin-Based Biopolymeric Optical Waveguides," Biotechnology and
Bioengineering 103, 725-732 (2009).

A. L. Martin, D. K. Armani, L. Yang, and K. J. Vahala, "Replica-molded high-Q
polymer microresonators," Optics Letters 29, 533-535 (2004).

B. Mashford, J. Baldauf, T. L. Nguyen, A. M. Funston, and P. Mulvaney, "Synthesis
of quantum dot doped chalcogenide glasses via sol-gel processing," Journal of
Applied Physics 109 (2011).

S. Mathur, M. Veith, H. Shen, S. Hufner, and M. H. Jilavi, "Structural and optical
properties of NdAlO3 nanocrystals embedded in an Al2O3 matrix," Chem. Mat. 14,
568-582 (2002).

S. Mehrabani, and A. M. Armani, "Blue upconversion laser based on thulium-doped
silica microcavity," Optics Letters 38, 4346-4349 (2013).

S. Mehrabani, P. Kwong, M. Gupta, and A. M. Armani, "Hybrid microcavity
humidity sensor," Applied Physics Letters 102 (2013).

B. Min, T. J. Kippenberg, and K. J. Vahala, "Compact, fiber-compatible, cascaded
Raman laser," Optics Letters 28, 1507-1509 (2003).

T. Miya, "Silica-based planar lightwave circuits: Passive and thermally active
devices," IEEE J. Sel. Top. Quantum Electron. 6, 38-45 (2000).
247

F. Monifi, S. K. Ozdemir, J. Friedlein, and L. Yang, "Encapsulation of a Fiber Taper
Coupled Microtoroid Resonator in a Polymer Matrix," IEEE Photonics Technology
Letters 25, 1458-1461 (2013).

H. Monjushiro, A. Hirai, and H. Watarai, "Size dependence of laser-photophoretic
efficiency of polystyrene microparticles in water," Langmuir 16, 8539-8542 (2000).

M. R. N. Monton, E. M. Forsberg, and J. D. Brennan, "Tailoring Sol-Gel-Derived
Silica Materials for Optical Biosensing," Chemistry of Materials 24, 796-811 (2012).

W. V. Moreshead, J. L. R. Nogues, and R. H. Krabill, "Preparation, processing, and
fluorescence characteristics of neodymium-doped silica glass prepared by the sol-gel
process," J. Non-Cryst. Solids 121, 267-272 (1990).

S. Nenkova, L. Radev, N. Rangelova, B. Aleksiev, and B. Samuneva, "New sol-gel
silica hybrids containing pectin and some metal ions," (2007), pp. 164-167.

E. P. Ostby, L. Yang, and K. J. Vahalal, "Ultralow-threshold Yb3+: SiO2 glass laser
fabricated by the sol-gel process," Optics Letters 32, 2650-2652 (2007).

M. Oxborrow, "How to simulate the whispering-gallery-modes of dielectric
microresonators in FEMLAB/COMSOL - art. no. 64520J," in Conference on Laser
Resonators and Beam Control IX(San Jose, CA, 2007), pp. J4520-J4520.

E. D. Palik, Handbook of Optical Constants of Solids (Elsevier, 1985).

V. M. N. Passaro, C. de Tullio, B. Troia, M. La Notte, G. Giannoccaro, and F. De
Leonardis, "Recent Advances in Integrated Photonic Sensors," Sensors 12, 15558-
15598 (2012).

E. J. G. Peterman, F. Gittes, and C. F. Schmidt, "Laser-induced heating in optical
traps," Biophysical Journal 84, 1308-1316 (2003).

R. Piazza, and A. Parola, "Thermophoresis in colloidal suspensions," Journal of
Physics-Condensed Matter 20 (2008).

F. Pisanello, A. Qualtieri, T. Stomeo, L. Martiradonna, R. Cingolani, A. Bramati,
and M. De Vittorio, "High-Purcell-factor dipolelike modes at visible wavelengths in
H1 photonic crystal cavity," Optics Letters 35, 1509-1511 (2010).

M. Pokrass, Z. Burshtein, and R. Gvishi, "Thermo-optic coefficient in some hybrid
organic/inorganic fast sol-gel glasses," Optical Materials 32, 975-981 (2010).

248
M. R. Poulsen, P. I. Borel, J. Fage-Pedersen, J. Hubner, M. Kristensen, J. H. Povlsen,
K. Rottwitt, M. Svalgaard, and W. Svendsen, "Advances in silica-based integrated
optics," Opt. Eng. 42, 2821-2834 (2003).

S. A. Putnam, and D. G. Cahill, "Transport of nanoscale latex spheres in a
temperature gradient," Langmuir 21, 5317-5323 (2005).

J. Ren, L. H. Wang, X. Y. Han, J. F. Cheng, H. L. Lv, J. Y. Wang, X. G. Jian, M. S.
Zhao, and L. Y. Jia, "Organic Silicone Sol-Gel Polymer as a Noncovalent Carrier of
Receptor Proteins for Label-Free Optical Biosensor Application," ACS Applied
Materials & Interfaces 5, 386-394 (2013).

H. Rokhsari, and K. J. Vahala, "Ultralow loss, high Q, four port resonant couplers
for quantum optics and photonics," Physical Review Letters 92, 253905 (2004).

B. A. Rose, A. J. Maker, and A. M. Armani, "Characterization of thermo-optic
coefficient and material loss of high refractive index silica sol-gel films in the visible
and near-IR," in Optical Materials Express (OSA, 2012), pp. 671-681.

B. Saleh, and M. Teich, Fundamentals of Photonics (Wiley-Interscience, 2007).

V. Sandoghdar, F. Treussart, J. Hare, V. LefevreSeguin, J. M. Raimond, and S.
Haroche, "Very low threshold whispering-gallery-mode microsphere laser," Physical
Review A 54, R1777-R1780 (1996).

S. Sarkar, P. K. Biswas, and S. Jana, "Nano silver coated patterned silica thin film by
sol-gel based soft lithography technique," Journal of Sol-Gel Science and
Technology 61, 577-584 (2012).

A. L. Schawlow, and C. H. Townes, "Infrared and Optical Masers," Physical Review
112, 1940-1949 (1958).

X. Serey, S. Mandal, Y. F. Chen, and D. Erickson, "DNA Transport and Delivery in
Thermal Gradients near Optofluidic Resonators," Physical Review Letters 108
(2012).

C. Shi, S. Mehrabani, and A. M. Armani, "Leveraging bimodal kinetics to improve
detection specificity," Optics Letters 37, 1643-1645 (2012).

G. V. Shivashankar, and A. Libchaber, "Single DNA molecule grafting and
manipulation using a combined atomic force microscope and an optical tweezer,"
Applied Physics Letters 71, 3727-3729 (1997).

E. J. Smith, W. Xi, D. Makarov, I. Monch, S. Harazim, V. A. B. Quinones, C. K.
Schmidt, Y. F. Mei, S. Sanchez, and O. G. Schmidt, "Lab-in-a-tube: ultracompact
249
components for on-chip capture and detection of individual micro-/nanoorganisms,"
Lab on a Chip 12, 1917-1931 (2012).

C. E. Soteropulos, H. K. Hunt, and A. M. Armani, "Determination of binding
kinetics using whispering gallery mode microcavities," Applied Physics Letters 99
(2011).

S. M. Spillane, T. J. Kippenberg, O. J. Painter, and K. J. Vahala, "Ideality in a fiber-
taper-coupled microresonator system for application to cavity quantum
electrodynamics," Physical Review Letters 91 (2003).

S. M. Spillane, T. J. Kippenberg, K. J. Vahala, K. W. Goh, E. Wilcut, and H. J.
Kimble, "Ultrahigh-Q toroidal microresonators for cavity quantum electrodynamics,"
Physical Review A 71 (2005).

R. Stegeman, C. Rivero, G. Stegeman, P. Delfyett, K. Richardson, L. Jankovic, and
H. Kim, "Raman gain measurements in bulk glass samples," Journal of the Optical
Society of America B-Optical Physics 22, 1861-1867 (2005).

W. Strek, E. Pawlik, P. Deren, A. Bednarkiewicz, J. Wojcik, V. E. Gaishun, and G. I.
Malashkevich, "Optical properties of Nd3+-doped silica fibers obtained by sol-gel
method," Journal of Alloys and Compounds 300, 459-463 (2000).

D. L. Su, G. D. Qian, Z. Y. Wang, Z. L. Hong, and M. Q. Wang, "The preparation of
silica-based SiO2-ZrO2 films as waveguides by sol-gel process," Rare Metal
Materials and Engineering 33, 281-283 (2004).

I. M. Thomas, S. A. Payne, and G. D. Wilke, "Optical-properties and laser
demonstration of Nd-doped sol-gel silica glasses," J. Non-Cryst. Solids 151, 183-194
(1992).

K. J. Vahala, "Optical microcavities," Nature 424, 839-846 (2003).

S. Varoutsis, S. Laurent, I. Sagnes, A. Lemaitre, L. Ferlazzo, C. Meriadec, G.
Patriarche, I. Robert-Philip, and I. Abram, "Reactive-ion etching of high-Q and
submicron-diameter GaAs/AlAs micropillar cavities," Journal of Vacuum Science &
Technology B 23, 2499-2503 (2005).

D. W. Vernooy, V. S. Ilchenko, H. Mabuchi, E. W. Streed, and H. J. Kimble, "High-
Q measurements of fused-silica microspheres in the near infrared," Optics Letters 23,
247-249 (1998).

F. Vollmer, and S. Arnold, "Whispering-gallery-mode biosensing: label-free
detection down to single molecules," Nature Methods 5, 591-596 (2008).

250
W. von Klitzing, E. Jahier, R. Long, F. Lissillour, V. Lefevre-Seguin, J. Hare, J. M.
Raimond, and S. Haroche, "Very low threshold green lasing in microspheres by up-
conversion of IR photons," J. Opt. B-Quantum Semicl. Opt. 2, 204-206 (2000).

P. B. Wagh, A. V. Rao, and D. Haranath, "Influence of molar ratios of precursor,
solvent and water on physical properties of citric acid catalyzed TEOS silica
aerogels," Mater. Chem. Phys. 53, 41-47 (1998).

A. L. Washburn, and R. C. Bailey, "Photonics-on-a-chip: recent advances in
integrated waveguides as enabling detection elements for real-world, lab-on-a-chip
biosensing applications," Analyst 136, 227-236 (2011).

C. M. Whang, C. S. Yeo, and Y. H. Kim, "Preparation and characterization of sol-get
derived SiO2-TiO2-PDMS composite films," Bulletin of the Korean Chemical
Society 22, 1366-1370 (2001).

"Whispering Gallery at St. Paul's Cathedral,"
http://en.wikipedia.org/wiki/File:Dome_of_st_pauls.jpg.

B. Wilhelm, V. Romano, and H. P. Weber, "Fluorescence lifetime enhancement of
Nd3+-doped sol-gel glasses by Al-codoping and CO2-laser processing," J. Non-
Cryst. Solids 328, 192-198 (2003).

O. S. Wolfbeis, "Fiber-optic chemical sensors and biosensors," Anal. Chem. 78,
3859-3873 (2006).

F. Q. Wu, D. Machewirth, E. Snitzer, and G. H. Sigel, "An efficient single-mode
Nd3+ fiber laser prepared by the sol-gel method," Journal of Materials Research 9,
2703-2705 (1994).

Y. N. Xia, and G. M. Whitesides, "Soft lithography," Annual Review of Materials
Science 28, 153-184 (1998).

H. B. Xin, X. M. Li, and B. J. Li, "Massive photothermal trapping and migration of
particles by a tapered optical fiber," Optics Express 19, 17065-17074 (2011).

H. B. Xin, Y. Y. Li, X. S. Liu, and B. J. Li, "Escherichia coli-Based Biophotonic
Waveguides," Nano Letters 13, 3408-3413 (2013).

A. H. J. Yang, S. D. Moore, B. S. Schmidt, M. Klug, M. Lipson, and D. Erickson,
"Optical manipulation of nanoparticles and biomolecules in sub-wavelength slot
waveguides," Nature 457, 71-75 (2009).

J. Yang, and L. J. Guo, "Optical sensors based on active microcavities," IEEE J. Sel.
Top. Quantum Electron. 12, 143-147 (2006).
251

J. Yang, K. van Dalfsen, K. Worhoff, F. Ay, and M. Pollnau, "High-gain
Al2O3:Nd3+ channel waveguide amplifiers at 880 nm, 1060 nm, and 1330 nm,"
Appl. Phys. B 101, 119-127 (2010).

L. Yang, D. K. Armani, and K. J. Vahala, "Fiber-coupled erbium microlasers on a
chip," Applied Physics Letters 83, 825-826 (2003).

H. Yin, M. D. Wang, K. Svoboda, R. Landick, S. M. Block, and J. Gelles,
"Transcription against an applied force," Science 270, 1653-1657 (1995).

T. Yoshie, L. Tang, and S.-Y. Su, "Optical Microcavity: Sensing down to Single
Molecules and Atoms," Sensors 11, 1972-1991 (2011).

J. W. Zhai, B. Shen, X. Yao, and L. Y. Zhang, "Preparation and spectral properties
of Nd2O3-doped silica-based glasses prepared by the sol-gel process," Ceramics
International 28, 737-740 (2002).

X. M. Zhang, and A. M. Armani, "Silica microtoroid resonator sensor with
monolithically integrated waveguides," Optics Express 21, 23592-23603 (2013).

X. M. Zhang, and A. M. Armani, "Suspended bridge-like silica 2 x 2 beam splitter
on silicon," Optics Letters 36, 3012-3014 (2011).

X. M. Zhang, H. S. Choi, and A. M. Armani, "Ultimate quality factor of silica
microtoroid resonant cavities," Applied Physics Letters 96 (2010).

X. M. Zhang, M. Harrison, A. Harker, and A. M. Armani, "Serpentine low loss
trapezoidal silica waveguides on silicon," Optics Express 20, 22298-22307 (2012).

X. Zhang, Y. Y. Wu, S. Y. He, and D. Z. Yang, "Investigation on the atomic oxygen
erosion resistance of sol-gel alumina-silica composite films on Kapton," Mater.
Chem. Phys. 114, 179-184 (2009).

Y. Zhang, H. X. Lei, Y. Z. Li, and B. J. Li, "Microbe removal using a micrometre-
sized optical fiber," Lab on a Chip 12, 1302-1308 (2012).

L. H. Zhou, D. Q. Chen, W. Q. Luo, Y. S. Wang, Y. L. Yu, and F. Liu, "Transparent
glass ceramic containing Er3+: CaF2 nano-crystals prepared by sol-gel method,"
Materials Letters 61, 3988-3990 (2007).

252
Appendix A : Mentoring Projects

In this section, I describe some of the undergraduate and high school student
projects in which I served as a research mentor. While these projects did not directly
result in publications, their results are interesting and have been/could be built upon
in future work.
A.1 Polymer Mentoring Projects

Nick’s Project: Efiron Coatings

Nick Benzoni, then a high school student, visited the lab in Summer 2009,
and helped study waveguide coatings made from the Efiron polymer. Normally, the
silica waveguides we fabricate are air-clad, meaning the surrounding air serves as the
waveguide’s cladding layer. However, due to the high refractive index contrast
between the waveguide and air, the waveguide does not have single mode operation.
By spin-coating and curing the Efiron coatings (which have a refractive index around
1.4) onto waveguides, the number of modes in the silica waveguides can be reduced,
improving the loss (higher order modes can be a parasitic loss mechanism) and
helping us develop a waveguide device with few or single mode operation.
Using ellipsometry, Nick measured the thickness vs. spin speed of the Efiron
polymer (Figure A-1), so that we could apply a thin layer to the waveguides by spin-
coating. Then, I measured the loss vs. length of the waveguides when coated with
the Efiron, as shown in Chapter 3.
253
Efiron Thickness vs. Spin Speed
0
5000
10000
15000
20000
25000
30000
35000
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 1000
0
Spin Speed (RPM)
Efiron Thickness (Angstroms)
Thickness Data
Efiron Thickness

Figure A-1: Thickness vs. spin speed data for Efiron PC-409AP polymer.

Audrey: Reflowing Disks On-chip

Audrey’s first project upon joining the Armani lab was to determine whether
circular silica pads could be reflowed and smoothed with a CO
2
laser prior to XeF
2

etching. If the silica circles’ roughness (from BOE etching) could be reduced in this
way, it would provide many benefits.
The most significant benefit of this project, if successful, would be the
development of a high Q silica disk resonator with smooth edges. At the time, the
only way to improve the Q of the silica microdisk was to reflow it with a CO
2
laser
after XeF
2
etching. This reflow process eliminates roughness and vastly improves Q,
but the resonator also shrinks significantly during the process, making integration of
the toroids with waveguides extremely difficult. In contrast, if we developed a high
254
Q, reflowed resonator which does not shrink, the resonator could be easily integrated
with on-chip waveguides for efficient coupling of light.
To investigate whether silica pads could be reflowed on-chip after BOE
etching but prior to XeF
2
etching, Audrey reflowed on-chip silica pads of various
sizes and thickness with the CO
2
laser set at various intensities. Then, using AFM,
SEM, profilometry, and optical microscopy, we inspected the samples to determine
whether the surface roughness from the BOE etching was reduced. Unfortunately,
the on-chip reflow seemed to have no measurable effect. The most probable reason
for this is the fact that the silicon chip underneath the silica pad acts as a heat sink,
reducing the heating effects of the laser on the silica. In contrast, the undercut silica
pads become elevated silica microdisks, and the suspended silica loses this efficient
heat sink. Therefore, when the silica microdisks are exposed to the CO
2
laser, they
reflow and shrink to form toroids, while the on-chip silica pads undergo negligible
change when reflowed.
Although this particular project did not work, it is important to note that
Xiaomin Zhang was able to overcome the shrinkage problem of toroids and
successfully integrate waveguides and toroids together on-chip (although with a
much more complex fabrication process). Also, work in the Vahala group at Caltech
showed how optimizing the BOE etch step can produce smoother microdisks with
high quality factors.
Daniel + Chelsea: Organic Laser Dyes

Daniel Guevara and Chelsea Beirne worked in the lab during Summer 2010
as part of the USC Center for Engineering Diversity’s Summer Institute program.
255
Since they were only in the lab for 10 days, their project was very brief and involved
optimization of laser dye concentrations in PMMA films.
While most of the work in this thesis focuses on developing lasers by doping
silica, another approach to make a laser is to use materials containing fluorescent
laser dyes. For example, by coating a silica toroid with a film containing laser dyes,
it may be possible to cause the laser dye to emit light at a desired wavelength. One
advantage of laser dyes is that many pump wavelengths are available with broad
absorption bands. Therefore, lasers can be developed with many different pump and
emission wavelengths, unlike rare earth doped lasers which can have very narrow
optical cross sections at well-defined wavelengths. Two dyes were ordered from
Exciton: LD 700 Perchlorate and Pyrromethene 597-8C9. Since Daniel and Chelsea
were only available for a short time, they focused on characterizing LD-700
perchlorate, a dye which absorbs near 658nm and emits near 700nm.

Figure A-2: LD 700 Perchlorate laser dye. When pumped near 658nm, it emits near 700nm.

For their project, Chelsea and Daniel weighed various amounts of laser dye
and PMMA polymer, and dissolved them in toluene to produce PMMA+dye
solutions with different concentrations. The resulting solutions were stirred for ~30
minutes using a magnetic stirrer. Then, the PMMA+toluene+laser dye solutions
256
were spin-coated onto silicon wafers and characterized using fluorescence
microscopy.
Ideally, the films would have bright, uniform fluorescence without clumps.
We observed fluorescence near 700nm, however, the films still contained noticeable
clumps of undissolved dye. By examining the strength and uniformity of the
resulting films, we could determine the best dye-polymer concentrations to use on
optical devices. Based on fluorescence microscopy images, dye concentrations of
1x10
-4
M to 5x10
-4
M and PMMA concentrations of 2 to 4% produced the most
uniform film with less clumping (Figure A-3).

Figure A-3: Fluorescence microscope images of LD-700 dye in PMMA films.

Therefore, in future experiments, the dye may need to be stirred longer or used with
different PMMA or toluene concentrations. Based on what we learned in this
project, we may be able to coat silica toroids with laser dyes and develop lasers in
the future.
257
John: PEG and Chitosan Hydrogels

John Casabar joined the lab during Summer 2012-Spring 2013 as part of his
research class at Bravo Medical Magnet High School. The goal of John’s project
was to attach a hydrogel material to silica toroids, so that cells could be grown and
studied using toroids with a more biocompatible medium.
Initially, we focused on studying hydrogels made from various
concentrations of PEG, or poly ethylene glycol dissolved in water. The molecular
weight of the PEG chain was also varied between 400, 1000, and 4000. First,
various PEG+water solutions were spin-coated onto (O
2
plasma-treated) silica on
silicon wafers. Then, we attempted to characterize the film thicknesses and
refractive indices using ellipsometry and FTIR. During this process, we encountered
a few issues: 1) the PEG solutions did not spin-coat uniformly, even if a slow spinner
ramp rate was used, and 2) Even if the PEG solutions did spin-coat uniformly,
considering that these solutions are hydrogels, they would quickly dissolve while
testing a PEG-coated toroid in water.
Given these issues, we decided to covalently bond PEG instead. The
challenge here is that there are not many efficient and safe ways to bond PEG’s
hydroxyl groups with a toroid. Using procedures developed by Rasheeda, we
attached epoxy groups to hydroxylated toroids using GPTMS vapor deposition for 30
minutes. Then, PEG was dissolved in an anhydrous, basic environment
(THF+triethanolamine base to pH 11-12) and incubated overnight to make PEG’s
OH groups bind the epoxy rings. However, after this process, we could not detect
any PEG attachment by fluorescence microscopy, TGA, ellipsometry, or FTIR.
258

Figure A-4: Attachment of PEG’s hydroxyl group to epoxy functionalized toroids in an anhydrous
basic environment.

Two other approaches could be tried to attach PEG. First, we could attempt
vapor deposition of an isocyanate or isothiocyanate group onto OH-functionalized
toroids, and then PEG’s OH groups would readily attach (along with any other OH
containing molecule, such as water). Another alternative is to purchase PEG with a
carboxyl or amine on one end of the chain. Then, the PEG could be attached to
aminated toroids by EDC or hydroxylated toroids by epoxy chemistry. However,
both of these approaches required use of dangerous chemicals and/or the argon
glovebox, and were a bit too complex for John to be able to work on during the 2-3
hour sessions when he came to lab.
Since the PEG approach was therefore not working as well, we shifted our
focus to chitosan hydrogels. Chitosan is a biocompatible polysaccharide commonly
259
found in crab and shrimp shells. It contains amine groups which can easily attach to
an epoxy functionalized toroid. Therefore, John attached epoxy groups to silica
toroids using GPTMS and the same procedures as before. He prepared slightly
acidic, aqueous solutions of ~1-10mM chitosan in 1% acetic acid (to improve
solubility). Then, John incubated the epoxy-functionalized toroids overnight in the
chitosan solutions. Successful attachment was verified by attaching rhodamine B to
chitosan’s amine groups using EDC (see section 7.4.2).
The next step in the project was to verify that the chitosan-functionalized
toroids have minimal impact on Q factor and are biocompatible with cells.
Unfortunately, we were not able to fully investigate these areas as John’s school year
was ending. However, we were able to measure Q factors on chitosan-functionalized
toroids, and it seems that the attachment of chitosan to epoxy functionalized toroids
is effective based on our surface chemistry and fluorescence measurements.

Leah: Hydrogels

Leah Tsui, a biomedical engineering undergraduate, took over part of John’s
project after he left the group. Leah’s goal was to grow cells successfully on silicon
chips and on toroids. To do so, Leah investigated various hydrogels and scaffolds,
including the chitosan-functionalized silica John worked on. By measuring cell
viability on substrates with various surface functionalization, an optimal
environment can be developed for studying cells in vivo with toroids. Before Leah
started working with biologists in the group, we worked on literature searches and
discussed possible materials which could be used in her project.
260

A.2 Sol-gel Mentoring Projects

Christine: Optimizing Sol-gel Synthesis

Christine Zimmerman first joined the group as a high school student during
summer 2010. The goal of her project was to study and optimize our sol-gel
synthesis parameters. At the time, we made sol-gel silica using tetraethoxysilane
(TEOS), ethanol solvent, HCl catalyst, and water in a set 1:4:0.1:2 molar ratio.
However, several issues had come up during sol-gel synthesis: 1) the sol-gel films
often formed cracks, preventing the formation of smooth uniform films; and 2) we
could not spin coat thicker films (at spin speeds below ~7000 rpm) without excessive
cracking. Christine prepared many sol-gels using various ratios of precursors,
spin/age/stir times, and annealing parameters in order to figure out which procedures
produced the best films. Unfortunately, at the time we did not have our FTIR or
ellipsometer, so Christine’s data was mostly qualitative and involved comparing
optical microscope images of various films and seeing which had the fewest cracks.
Based on her analysis, our current sol-gel parameters (same molar ratio, 2 hour stir
time, 24-48 hour age time, 1000°C anneal for 2 hours) seemed to work well. Since
her data was highly qualitative, it is difficult to compare the various films which
turned out well and had minimal cracks.
Christine: Porous Sol-gels

Christine also rejoined the lab after starting her undergraduate studies at
USC. Her second project focused on developing porous sol-gel silica materials. By
adding surfactants such as CTAB (cetyltrimethylammonium bromide) to sol-gel
261
silica, spherical micelles could be produced which form pores upon annealing. She
closely followed the procedures developed by Brian in his Master’s thesis in order to
make porous silica films. Then, by tuning the age time, amount of surfactant, and
other parameters, she attempted to tune the pore size. Unfortunately, she was unable
to complete this project and another undergraduate, Tara, took over.
Nic: Reflowing Cracked Sol-gels

Nic Murillo joined the lab in 2009 as an undergraduate researcher. While he
was working on sol-gel projects he encountered a lot of cracked films. Nic, Heather
Hunt, and I were curious to see if reflowing a cracked sol-gel silica film with our
CO
2
laser could help eliminate cracking. To do so, Nic prepared sol-gel films, and
we exposed them to the CO
2
laser at various intensities. Then, we inspected the sol-
gel films using optical microscopy and atomic force microscopy (AFM) to see if
cracking could be removed.
262

Figure A-5: Representative AFM data from a cracked sol-gel silica film.

While the CO
2
laser appeared to help reduce very large cracks, it would not
repair smaller cracks. It is possible that the silicon substrate underneath the cracked
silica layer could act as a heat sink (similar to what Audrey observed in her project)
and prevent the cracks from being repaired. It was difficult to image the silica films
using AFM since the topography of the extremely cracked film could vary greatly.
While the CO
2
laser has been shown effective at reducing cracking in the literature, it
was not able to repair the cracked sol-gels into smooth enough films from which we
could make devices. Therefore, Nic moved on to other projects.

Heaven: Er-doped Sol-gels

Heaven Saunders joined the lab in summer 2011 as a high school researcher
and worked on erbium-doped sol-gel silica films. Her project had two goals: first, to
263
determine the optimal concentration of erbium dopant to add to sol-gels for minimal
cracking, and second, to see if we could make a toroid microlaser by spin-coating
erbium directly onto toroids.
Following the standard sol-gel procedures (see section 5.2.2), Heaven
prepared sol-gels containing various amounts of erbium and spin-coated them onto
bare silicon wafers and finished silica toroids. Unfortunately, that summer we were
having problems with our spinner and sol-gel processes, and it was very difficult to
get thin, uniform films. So even when Heaven varied the synthesis parameters such
as dopant concentration, stir/age time, anneal parameters, and spin speed, her films
had a lot of cracking. Therefore, it was difficult to determine the optimal dopant
concentration and synthesis parameters for her project.

Alejandro + Mary : Tm
3+
and Ce
3+
Sol-gels

Alejandro Sanchez and Mary Williams joined the lab in summer 2011 as part
of the Center for Engineering Diversity’s Summer Institute. Their goal was to help
Simin determine the optimal concentration of thulium (Tm
3+
) and cerium (Ce
3+
)
dopants in silica films. Using standard sol-gel synthesis procedures, they
synthesized sol-gel silica films containing the dopants at concentrations ranging from
0.05-0.2 atomic percent. Then, they could inspect the films using the optical
microscope to determine which had the least cracking and defects. After Alejandro
and Mary spin-coated and annealed at least 2-3 layers of sol-gel, silica toroids could
be fabricated from the films.
264
Based on the amount of cracking present in the films, the lowest dopant
concentration produced the best results. This is somewhat expected. As seen in the
alumina-sensitized nanolaser project, rare earth dopants have limited solubility in
silica and form clusters. The lowest concentration of rare earth dopants likely
disrupts the silica matrix least. Other issues, such as contamination and the spinner
not working well that summer, could also contribute to the amount of cracking seen
in the final films. Despite the cracking issues, Mary and Alejandro were able to
fabricate thick enough and uniform enough films for fabricating toroids. While no
lasing was observed, the toroids did have measurable Q factors (~10
5
).

Figure A-6: Optical microscope images of Tm
3+
and Ce
3+
films made by Mary and Alejandro. The
lowest dopant concentration had the least amount of cracking.
265

Abstract (if available)

Abstract Due to their favorable optical and material properties, silica-based materials and devices have found many important applications throughout science and engineering, especially in sensing, communications, lasers, and integrated optics. Often, silica’s properties ultimately limit the performance of these applications. To address this limitation, this thesis investigates the development of improved silica materials and optical devices, including silica films, coatings, waveguides, resonators, lasers, and sensors. Using sol-gel chemistry and microfabrication procedures, custom silica materials and devices are developed to benefit many applications. ❧ In this thesis, it is first demonstrated how the low optical loss of silica enables fabrication of low loss integrated waveguides and toroidal resonators with ultra-high quality factors. Then, by adding various rare earth and metal dopants to sol-gel silica, hybrid silica materials and devices are made with custom properties such as high refractive index and lasing capabilities. Finally, several applications are demonstrated, including the use of high refractive index coatings to control the behavior of light, development of Raman and ultra-low threshold rare earth microlasers, and a heterodyned microlaser sensor with significantly improved sensing performance. Future applications and directions of this research are also discussed.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Titania and hybrid titania - silica sol-gel thin films and their applications in integrated optical devices

PDF

Silica sol-gel thin film coatings for integrated photonic devices

PDF

Development of integrated waveguide biosensors and portable optical biomaterial analysis systems

PDF

Development of novel optical materials for whispering gallery mode resonator for nonlinear optics

PDF

Development of optical devices for applications in photonic integrated circuit and sensing

PDF

Development of hybrid microsensors for environmental monitoring and biodetection

PDF

Development of hybrid optical microcavities for Plasmonic laser and improving biodetection

PDF

Silicon integrated devices for optical system applications

PDF

Biological and chemical detection using optical resonant cavities

PDF

Integration of practical high sensitivity whispering gallery mode resonator sensors

PDF

Surface-enhanced on-chip whispering gallery mode resonators for improved nonlinear optics

PDF

Nonlinear optical nanomaterials in integrated photonic devices

PDF

Fabrication and characterization of yoroidal resonators for optical process improvement

PDF

Application of optical forces in microphotonic systems

PDF

Active integrated photonic devices in single crystal LiNbO₃ micro-platelets and a hybrid Si-LiNbO₃ platform

PDF

Optical simulation and development of novel whispering gallery mode microresonators

PDF

Development of organic-inorganic optical microcavities for studying polymer thin films

PDF

Integrated photonics assisted electron emission devices

PDF

Resonant light-matter interactions in nanophotonic structures: for manipulating optical forces and thermal emission

PDF

Integrated silicon waveguides and specialty optical fibers for optical communications system applications

Asset Metadata

Creator Maker, Ashley Julia (author)

Core Title Developing improved silica materials and devices for integrated optics applications

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Chemical Engineering

Publication Date 06/13/2014

Defense Date 05/21/2014

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag high index films and coatings,integrated optics,OAI-PMH Harvest,optical resonators,optical sensors,rare earth lasers,sol-gel silica materials,toroid microcavities,waveguides

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Armani, Andrea M. (committee chair), Malmstadt, Noah (committee member), Roberts, Richard W. (committee member), Willner, Alan E. (committee member)

Creator Email ashleyjmaker@gmail.com,maker@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-419619

Unique identifier UC11285890

Identifier etd-MakerAshle-2548.pdf (filename),usctheses-c3-419619 (legacy record id)

Legacy Identifier etd-MakerAshle-2548.pdf

Dmrecord 419619

Document Type Dissertation

Format application/pdf (imt)

Rights Maker, Ashley Julia

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